Endothelial to Mesenchymal Transition: Role in Physiology and in the Pathogenesis of Human Diseases

Abstract

Numerous studies have demonstrated that endothelial cells are capable of undergoing endothelial to mesenchymal transition (EndMT), a newly recognized type of cellular transdifferentiation. EndMT is a complex biological process in which endothelial cells adopt a mesenchymal phenotype displaying typical mesenchymal cell morphology and functions, including the acquisition of cellular motility and contractile properties. Endothelial cells undergoing EndMT lose the expression of endothelial cell-specific proteins such as CD31/platelet-endothelial cell adhesion molecule, von Willebrand factor, and vascular-endothelial cadherin and initiate the expression of mesenchymal cell-specific genes and the production of their encoded proteins including α-smooth muscle actin, extra domain A fibronectin, N-cadherin, vimentin, fibroblast specific protein-1, also known as S100A4 protein, and fibrillar type I and type III collagens. Transforming growth factor-β1 is considered the main EndMT inducer. However, EndMT involves numerous molecular and signaling pathways that are triggered and modulated by multiple and often redundant mechanisms depending on the specific cellular context and on the physiological or pathological status of the cells. EndMT participates in highly important embryonic development processes, as well as in the pathogenesis of numerous genetically determined and acquired human diseases including malignant, vascular, inflammatory, and fibrotic disorders. Despite intensive investigation, many aspects of EndMT remain to be elucidated. The identification of molecules and regulatory pathways involved in EndMT and the discovery of specific EndMT inhibitors should provide novel therapeutic approaches for various human disorders mediated by EndMT.

I. INTRODUCTION

The endothelium is a thin membrane-like structure that lines the inner surface of all vessels in the body, including capillaries, arterioles, arteries, veins, and lymphatic vessels, with the primary essential function of regulating and maintaining vessel wall permeability. The endothelium also plays a crucial role in the pathogenesis of numerous vascular and nonvascular disorders (3, 37, 258). The vascular endothelium comprises a monolayer of highly differentiated cells, that display specific morphological, metabolic, structural, functional, and molecular/gene expression characteristics depending on the vascular system of which they are a cellular component (18, 62, 87, 110, 292). Although the monolayer of cells lining the posterior surface of the cornea has also been referred to as “corneal endothelium,” these cells display marked differences from the endothelial cells lining the vasculature, including distinct embryological origin, functional role, and gene expression profiles. Corneal endothelial cells are derived from the neural crest, whereas vascular endothelium is of mesoderm origin. Regarding their function, vascular endothelial cells are continually exposed to circulating biological fluids (blood and lymph) and to hemodynamic perturbations caused by circulatory flow, whereas corneal endothelial cells are not exposed to the functional consequences that continuously flowing biological fluids exert on the cells. Furthermore, there are profound differences in gene expression between these two cell types (97, 115). Given these important considerations, we have not included studies involving corneal endothelium or corneal endothelial cells in this review.

Under normal conditions, the endothelial cell phenotype is precisely maintained; however, numerous studies have demonstrated that endothelial cells display remarkable phenotypic plasticity (67, 75) including their ability to undergo endothelial to mesenchymal transition (EndMT), a newly recognized type of cellular transdifferentiation (11, 12, 14–16, 79, 177, 200, 201, 362, 363). EndMT is a complex biological process in which endothelial cells lose their specific phenotype and progressively evolve into cells with a mesenchymal phenotype that includes a spindle-shaped elongated cell morphology, loss of cell-cell junctions and polarity, and the acquisition of cellular motility and invasive and contractile properties. At the molecular level, EndMT results in the initiation of expression and production of mesenchymal cell-specific proteins including α-smooth muscle actin (α-SMA), extra domain A (EDA) fibronectin, N-cadherin, vimentin, fibroblast-specific protein-1 (FSP-1; also known as S100A4 protein), fibroblast activating protein (FAP), and fibrillar collagens type I and type III. The initiation of expression of mesenchymal cell-specific genes is accompanied by the progressive reduction and the eventual loss of endothelial cell-specific proteins including von Willebrand factor (vWF), CD31/platelet-endothelial cell adhesion molecule-1 (CD31/PECAM-1), and vascular-endothelial cadherin (VE-cadherin) (251, 252, 273, 274). Extensive studies have shown that members of the transforming growth factor-β (TGF-β) family of growth factors, and most prominently the TGF-β1 isoform, are the main inducers of EndMT (16, 117, 200, 209, 212, 311, 339). However, EndMT is an extremely complex process involving numerous TGF-β and non-TGF-β signaling pathways that are modulated by multiple and often redundant molecular mechanisms depending on the physiological or pathological status of the cells and on their specific cellular context.

EndMT has been shown to participate in highly important embryonic developmental processes and also plays a role in the pathogenesis of various genetically determined as well as acquired human diseases, including malignant, vascular, inflammatory, and fibrotic disorders (TABLES 1 and 2). Thus the precise identification of the molecules and regulatory pathways involved in the loss of the endothelial cell differentiated state and in the change of the endothelial cell phenotype to that of mesenchymal cells should lead to the development of novel therapeutic approaches for numerous EndMT-mediated human disorders.

Table 1.

Putative role of EndMT in human disease pathogenesis

Human Disease	EndMT Role	Reference Nos.
Malignant diseases	Originates cancer-associated fibroblasts	191, 256
	Stimulates angiogenesis	191, 193
	Stimulates metastasis	105, 158
Fibrotic diseases	Originates activated myofibroblasts	148, 201
	Increases production of profibrotic molecules	11, 174, 254
	Increases tissue stiffness that further stimulates fibrosis	195, 215
Pulmonary arterial hypertension	Structural and functional pulmonary vessel alterations	11, 309
Atherosclerosis	Increased vascular wall fibrosis leading to vessel narrowing	48, 91, 166, 329
	Calcification of atherosclerotic vessel walls	36, 119
Diabetes mellitus	Vascular fibrosis and vascular loss	40, 167, 362
Cavernous malformations	Cavernous vessel proliferation	38, 164, 192, 298
Fibrodysplasia ossificans progressiva	Soft tissue calcification	208, 259

Table 2.

Demonstration of EndMT in human fibrotic and vascular diseases

	Evidence of EndMT in Affected Tissues
Fibrotic Disease	Tissue source	Method(s)	Reference Nos.
Systemic sclerosis-associated pulmonary  fibrosis	Lung transplants	Immunohistochemistry	211
		Immunofluorescence
		Gene expression
Radiation-induced pulmonary fibrosis	Lung tissues (surgery)	Immunofluorescence	58
Systemic sclerosis-associated pulmonary  hypertension	Lung biopsies	Immunofluorescence	113
Idiopathic pulmonary hypertension	Lung transplants	Immunofluorescence transmission  electron microscopy	260
		Immunoelectron microscopy
Cardiac fibrosis	Heart transplants	Gene expression	345
Chronic kidney disease-associated cardiac  fibrosis	Heart tissue (autopsies and  cardiac surgery)	Immunohistochemistry	46
		Gene expression
Diabetic kidney disease-associated renal  fibrosis	Kidney biopsies	Immunohistochemistry	171
Idiopathic portal hypertension	Liver biopsies	Immunohistochemistry	152
Intestinal fibrosis	Colonic mucosa	Immunohistochemistry	266
Radiation-induced rectal fibrosis	Rectal tissues (surgery)	Immunofluorescence	216, 217

II. CELLULAR TRANSDIFFERENTIATION AND PLASTICITY OF THE CELLULAR PHENOTYPE

During early vertebrate embryonic development, totipotent cells evolve into fully differentiated cells and cell types specialized for specific functions in various tissues. The process of cellular differentiation is highly complex and precisely determined, involving numerous molecular signals and pathways that are DNA sequence independent and that do not require DNA replication. According to the Mendelian principles of inheritance, it was generally accepted that fully differentiated cells were stable and maintained their differentiated phenotype throughout the life of the organism. This notion was emphasized by Waddington who depicted differentiated cells as cellular entities destined to remain in a groove or valley and were unable to switch into other cells or cell types (317). This concept gained widespread acceptance among the scientific community; however, it had to be revised following studies showing that when nuclei from specialized amphibian intestinal cells were introduced into enucleated eggs, some of the eggs developed into normal tadpoles (121). Further studies showed that when fully differentiated keratinocytes cultured from adult Xenopus frog skin were transplanted into enucleated eggs from the same species they evolved into tadpoles containing highly specialized and fully functional cell types including muscle, nervous system, cardiac, hematologic, and ocular cells, thus providing strong evidence supporting the cellular plasticity of fully differentiated cells (122). Extensive subsequent studies with heterokaryons conclusively demonstrated that fully differentiated cells are able to acquire the phenotypic characteristics of other cells when exposed to specific stimuli or under specific cellular contexts (28–30). Most importantly, it was shown that this transdifferentiation process does not require DNA replication and must, therefore, be mediated by the combined effects of activation of previously silent genes and/or the silencing of genes that were being expressed in the cells displaying the original phenotype. This novel concept introduced by Waddington in 1942, who coined the term epigenetic landscape (318), refers to the remarkable influences exerted by DNA sequence-independent mechanisms and pathways present at various stages of the cellular differentiation process leading to the establishment of a specific cellular phenotype (33, 74). Epigenetic modifications result in profound alterations and crucial functional effects and involve multiple molecular pathways and diverse mechanisms including changes in chromatin structure, such as the creation of DNase hypersensitive sites, removal of histones, or changes in DNA methylation, as well as regulation of mRNA transcription rates or changes in RNA stability. The molecules responsible for these effects are quite heterogeneous and include various species of noncoding RNAs, histones, and other regulatory proteins. Epigenetic effects modulate protein-protein interactions capable of precisely regulating gene expression levels that are critical for the expression and maintenance of the differentiated state and for the phenotypic transition of fully differentiated cells into cells with a different and distinct phenotype (25, 34, 60).

III. EndMT: A RECENTLY IDENTIFIED CELLULAR TRANSDIFFERENTIATION PROCESS

A. Definition and Early Descriptions of EndMT

The EndMT process is characterized by profound morphological, functional, and molecular changes in the endothelial cell phenotype. The morphological changes take effect owing to alterations in cell polarity and to remarkable cytoskeleton restructuring. At the molecular level, the endothelial cells initiate the expression of mesenchymal cell-specific genes and the production of their corresponding proteins. These events are diagrammatically illustrated in FIGURE 1. The earliest observations of the occurrence of EndMT were made in studies that examined the morphological changes and the cellular components involved in the development of cardiac valves during embryonic chicken and murine development (200, 201). The results demonstrated that the cardiac cushion cells that eventually were to become the cardiac valves arose from a distinct subpopulation of endocardial endothelial cells that transitioned into mesenchymal cells (200, 201). It was subsequently shown that this transition was induced by the expression of certain extracellular matrix (ECM) macromolecules such as fibrillin and fibulin, growth factors such as TGF-β and bone morphogenetic protein (BMP), and other regulatory molecules such as SOX9 that were present within the primordial subendothelial cardiac mesenchymal tissue known as cardiac jelly (6, 16, 17, 201, 220, 249). The occurrence of a morphological conversion of endothelial cells to mesenchymal cells was also described during aortic artery development in the chick embryo (14, 15). Subsequent studies demonstrated that EndMT was not confined to embryonic development but that it could also occur in endothelial cells isolated from adult vascular endothelium (16). These pioneering studies showed that endothelial cells isolated from adult bovine aortas were able to acquire the expression of α-SMA through a transdifferentiation process initiated and driven by TGF-β1. Remarkably, the transition of endothelial cells to mesenchymal cells expressing α-SMA and other mesenchymal cell markers did not require cell division as it occurred even in the absence of cellular proliferation. These studies also identified hybrid cells expressing both an endothelial cell-specific molecular marker (factor VIII antigen) and α-SMA, most likely representing an intermediary phenotype between endothelial cells and mesenchymal/myofibroblastic cells, and demonstrated that the phenotypic change was not permanent as the transition was partially reversible following culture in TGF-β1-free conditions (16). Similar studies in isolated mature bovine aortic and pulmonary arterial endothelial cells cultured in vitro under various experimental conditions including treatment with TGF-β1 confirmed these observations (98, 244). Following these early observations, extensive investigations described the occurrence of EndMT in numerous in vitro studies, in experimental animal models of various human diseases in vivo, as well as in a growing number of genetically determined and acquired benign and malignant human pathological conditions. These studies have taken advantage of the development of procedures that allowed a precise and accurate endothelial cell lineage analysis and tracking based on the expression of microscopically or molecularly detectable markers under the control of endothelial cell-specific gene promoters (7, 8, 316).

FIGURE 1.

Diagram illustrating the endothelial to mesenchymal transition (EndMT) process. The diagram illustrates the morphological, phenotypic, and gene expression program changes occurring during the process of endothelial to mesenchymal cell transition leading to the phenotypic conversion of endothelial cells into activated myofibroblasts displaying increased production of various mesenchymal-specific macromolecules including α-smooth muscle actin (α-SMA), COL1, COL3, fibronectin (FN), FN-extra domain A (EDA)/vimentin, N-cadherin, and fibroblast-specific protein-1 (FSP-1). These events are accompanied by the loss of endothelial cell-specific gene products such as CD31/platelet-endothelial cell adhesion molecule-1 (CD31/PECAM-1), vascular-endothelial cadherin (VE-cadherin), COL4, vascular epidermal growth factor receptor (VEGFR), and von Willebrand factor.

B. Intermediate EndMT Phenotype

Although the initial descriptions of EndMT suggested that the conversion of endothelial cells into mesenchymal cells was a permanent shift between two alternative cellular differentiation phenotypes, more recent investigations have shown a greater flexibility in this transitional process, and it is now generally accepted that cells undergoing EndMT evolve through a spectrum of intermediary phases (327). This plasticity implies that cells could remain in intermediary stages of transdifferentiation and may frequently undergo a partial EndMT program (FIGURE 1). Indeed, hybrid intermediate EndMT states have been identified during angiogenesis, thus providing conclusive evidence of the occurrence of intermediary EndMT phases (339). This plasticity also implies that the transitional phenotype is metastable and that partial or total reversal of the process may occur under certain conditions (16, 327, 339).

IV. MOLECULAR MECHANISMS OF EndMT

Despite extensive investigation of the mechanisms regulating the EndMT process, the detailed molecular and regulatory alterations involved have not been fully elucidated. Although TGF-β1, TGF-β2, and TGF-β3 are the most important inducers and initiators of EndMT under most physiological and pathological conditions (117, 209, 212, 243, 311), there are substantial differences in their ability to induce EndMT. A direct comparison of their effects on EndMT in human microvascular endothelial cells showed the three TGF-β isoforms induced EndMT; however, TGF-β2 was most potent (271). These studies also showed that EndMT induced by TGF-β1 and TGF-β3 may be mediated through paracrine stimulation of TGF-β2 production and secretion since siRNA silencing of TGF-β2 blunted the expression of EndMT markers in cells treated with TGF-β1 and TGF-β3 (271). It should be emphasized, however, that although the role of the TGF-β family of growth factors in the initiation of EndMT has been conclusively demonstrated, the molecular events are extremely complex and involve numerous signaling pathways triggered by multiple and diverse molecular mechanisms. The most important of these pathways will be discussed in the following sections.

A. TGF-β Initiated Pathways

The TGF-β family of proteins comprises a large number of pleiotropic growth factors that play crucial roles in numerous physiological processes including embryogenesis, cellular development and differentiation, immunological system development, inflammatory response functions, and fibrosis and wound repair (35, 267, 313). Following the initial studies of the occurrence of EndMT, numerous studies conclusively demonstrated that the TGF-β growth factors play a crucial role in the initiation and progression of EndMT (117, 208, 311). Confirmation of the role of TGF-β in EndMT was provided by the demonstration of a reduction in the number of mesenchymal cells originating from endothelial cells when the TGF-β response was blunted in transgenic mice, as well as by the abrogation of EndMT employing a TGF-β receptor kinase inhibitor which inhibits activated TGF-β or employing small molecule inhibitors of TGF-β-mediated intracellular phosphorylation reactions (117, 168, 209, 363). It was subsequently shown that both Smad-dependent or canonical and Smad-independent or noncanonical TGF-β signaling pathways may be involved in the EndMT process and that this process is mediated by numerous transcriptional regulators such as Snail1, Snail2 (or Slug), Twist, and some members of Zeb family of proteins (80, 107, 117, 251).

Extensive recent studies have demonstrated that TGF-β signaling and the resulting molecular events and biological responses to TGF-β signaling are highly complex and involve a large number of molecular components requiring a precise regulation of TGF-β signaling to generate a specific biological response. Furthermore, the regulatory effects of TGF-β are extremely dependent on the cellular types involved and on the specific cellular context, as well as on the complex interactions with the surrounding extracellular microenvironment within specific tissues (5, 202, 371). Indeed, it has been demonstrated that under specific conditions even completely opposite effects may be elicited in different tissue environments or in different cell types (202, 371).

Owing to their remarkable pleotropic effects and their modulation of a wide spectrum of regulatory pathways, the TGF-β molecules have been implicated in the pathogenesis of numerous benign and malignant human diseases (31, 35, 114). One of the most important effects induced by the TGF-β family of proteins is a potent stimulation of fibroproliferative pathways in numerous target cell populations (267, 312, 313). Therefore, several TGF-β-mediated disorders most commonly involve exaggerated tissue fibrotic responses such as pancreatic cancer among the malignant disorders, and pulmonary and cardiac fibrosis and systemic sclerosis among the nonmalignant diseases (35, 142, 269).

1. Canonical Smad2/3-mediated pathways

The TGF-β pathway is triggered upon binding of activated TGF-β homodimers to a transmembrane heterodimeric TGF-β receptor complex in the cell surface. TGF-β molecules are secreted from the TGF-β producing cells in an inactive or latent form that is held latent by a noncovalent binding with a peptide known as the latency-associated peptide (LAP) that results from cleavage of the NH₂-terminal domain of the TGF-β molecules. Latent TGF-β requires to be activated extracellularly to be able to bind to the TGF-β-specific cell membrane receptors and exert any of the diverse and pleotropic biological effects of this potent family of growth factors. The extracellular activation process is dependent on complex interactions of the latent TGF-β with cell surface integrins for activation. Several integrins that share the promiscuous αv subunit are capable of initiating this activation process, although the integrins αvβ1 and αvβ6 are among the most potent inducers of TGF-β activation (130, 226, 227, 278). Following binding of the activated TGF-β molecules to their corresponding cell surface receptors a complex cascade of molecular interactions is initiated. In the canonical TGF-β signaling, this cascade is mediated by the Smad proteins, a large family of intracellular proteins (204) transducing the signaling initiated by TGF-β cell surface binding from the cell membrane to the nucleus as illustrated in FIGURE 2 (53, 80, 128, 129, 203, 205, 281). The TGF-β receptor initiating the canonical signaling pathway consists of two transmembrane serine-threonine kinases, one is the TGF-β receptor type I or ALK5 and the other is the TGF-β receptor type II. The binding of TGF-β to the heterodimeric receptor complex leads to the autophosphorylation of the TGF-β receptor II that once phosphorylated causes transphosphorylation of specific residues of the TGF-β receptor I leading to the formation of the active TGF-β receptor I/II complex. The active TGF-β receptor I/II complex phosphorylates specific serine residues in the cytoplasmic Smad2 and Smad3 proteins resulting in their activation. The phosphorylated Smad2/Smad3 forms a complex with the co-Smad, Smad4, that allows its translocation to the nucleus. Inside the nucleus, the Smad2/Smad3/Smad4 complex interacts with Smad binding elements (SBE) of TGF-β-responsive genes. Binding of Smad2 and Smad3 to SBE sites within the promoter regions of various TGF-β target genes stimulates their transcription. Two inhibitory Smad proteins, Smad6 and Smad7, abrogate or modulate TGF-β-induced signaling cascades and are potent negative regulators of TGF-β-induced gene expression responses (139, 219, 232).

FIGURE 2.

Canonical transforming growth factor (TGF)-β and noncanonical pathways in endothelial to mesenchymal transition (EndMT). The canonical TGF-β pathway is triggered upon binding of TGF-β homodimers to a transmembrane heterodimeric TGF-β I/II receptor complex in the cell surface. The binding of TGF-β to the heterodimeric receptor complex leads to the phosphorylation-mediated activation of the TGF-β receptor I/II complex. The active TGF-β receptor I/II complex phosphorylates cytoplasmic Smad2 and Smad3 proteins resulting in their activation. Activated Smad2 and Smad3 form a complex with Smad4 that translocates to the nucleus. Inside the nucleus, the Smad2/Smad3/Smad4 complex interacts with Smad binding elements (SBE) of TGF-β-responsive genes including those encoding profibrotic extracellular matrix (ECM) macromolecules and transcription factors such as Snail1 and 2 and Twist and stimulates their transcription. Another intracellular Smad protein, the inhibitory Smad7, abrogates TGF-β-induced signaling cascades by inhibitory effects on the activated TGF-β receptors and is a potent negative regulator of TGF-β-induced gene expression responses. Noncanonical TGF-β pathways include the mitogen-activated protein kinase (MAPK) family of serine/threonine-specific protein kinases and numerous other kinases such as phosphatidylinositol 3-kinase (PI3K), RhoA, Rac, c-Abl, and protein kinase C (PKC)-δ. TGF-β can activate all three known MAPK pathways: extracellular signal-regulated kinase (ERK), p38 MAPK, and c-Jun NH₂-terminal kinases (JNK). Signaling through these pathways mediates Smad2/3-independent TGF-β responses. These pathways result in either reduced transcription of endothelial-specific genes or increased the transcription of mesenchymal-specific genes effects mediated by EndMT-related transcription factors such as Snail1 and 2 and TWIST1.

Although the canonical Smad2/3-mediated pathway is the most potent inducer of EndMT, several other molecular Smad2/3-dependent pathways have also been shown to modulate the EndMT process employing endothelial cells in vitro and in animal models in vivo. These pathways involve molecules that are stimulated or activated by TGF-β and that utilize the canonical TGF-β signaling pathways for their effects on EndMT. Among these are certain Toll-like receptor (TLR) proteins such as TLR5, the Friend leukemia virus integration 1 (Fli-1) protein, and SIRTUIN 1 (SIRT1). Regarding the TLR5, a recent study showed that endothelial cell TLR5 expression is low under normal conditions but is increased in response to stimuli such as pressure overload (183). Evaluation of EndMT in TLR5-deficient mice indicated that EndMT was significantly less frequent and of less severity in the TLR5-deficient group compared with the control wild-type group. Gene expression analysis showed that the expression levels of several transcription factors that play an important role in EndMT, including snail1, snail2, twist1, and twist2, were decreased in the TLR5-deficient mice compared with control mice. Additional evidence that endothelial TLR5 promotes EndMT was obtained from studies using cultured human umbilical vein endothelial cells (HUVECs) with induced TLR5 overexpression. These cells displayed significantly enhanced EndMT in response to in vitro TGF-β1 stimulation, strongly suggesting that TLR5 expression in endothelial cells plays an important role in promoting EndMT in response to chronic pressure overload (183). Regarding Fli-1, a recent study showed that Fli-1 depletion in human dermal microvascular endothelial cells in vitro caused reduced VE-cadherin expression and induced typical EndMT features including remarkable morphological rearrangements of the actin filaments. Additionally, mRNA levels of the EndMT regulatory transcription factor SNAIL1 were moderately, but significantly, increased in the cells with decreased Fli-1 levels (290). A recent study of the effects of SIRT1 on TGF-β-induced EndMT in human endothelial cells showed that SIRT1 was decreased in TGF-β1-induced EndMT and that SIRT1 was a potent inhibitor of TGF-β1-induced EndMT, an effect that was mediated through de-acetylation of Smad4. These observations were validated employing specific adenovirus to overexpress or knockdown SIRT1 in the endothelial cells (175). In a related in vitro study, it was shown that another SIRT molecule, SIRT3, was also involved in the regulation of EndMT in murine glomerular endothelial cells (178). The results were confirmed in transgenic mice with either SIRT3 deficiency or with endothelial-specific SIRT3 overexpression following angiotensin II-induced hypertensive renal injury, indicating that SIRT3 deficiency was required for the development of EndMT and contributed to angiotensin II-induced hypertensive renal disease and fibrosis (178).

2. Noncanonical pathways

Besides the canonical TGF-β pathways, there are numerous noncanonical Smad2/3-independent signaling pathways activated by TGF-β that are also involved in the potent transcriptional events induced by TGF-β binding to its corresponding TGF-β cell surface receptors (80, 225, 370). The noncanonical pathways of TGF-β activation include the mitogen-activated protein kinase (MAPK) family of serine/threonine-specific protein kinases, and numerous other kinases such as phosphatidylinositol 3-kinase (PI3K), RhoA, Rac, c-Abl, protein kinase C (PKC)-δ, and others. TGF-β can activate all three known MAPK pathways: extracellular signal-regulated kinase (ERK), p38 MAPK, and c-Jun NH₂-terminal kinases (JNK). Signaling through these pathways mediates Smad2/3-independent TGF-β responses. However, it should be emphasized that these pathways are also capable of modifying the Smad pathways depending on a particular cell type or cellular context causing either increases or decreases in Smad signaling and, therefore, substantially affecting and modulating the full TGF-β-induced biological response.

Although noncanonical TGF-β signaling is involved in numerous intracellular transduction cascades and plays important roles in multiple cellular functions, there are only a few studies that examined their participation in EndMT. An analysis of EndMT induction by TGF-β2 in highly purified human cutaneous microvascular endothelial cells employing an extensive array of kinase inhibitors as well as RNA interference showed that TGF-β2 caused a very potent stimulation of Snail1 expression and protein production that was dependent on the activation of MEK, PI3K, and p38 MAPK in convergence with the canonical Smad pathway (209). Other studies have shown that, whereas ERK is capable of either increasing or decreasing the Smad pathway effects depending on the cell type studied, p38 MAPK and JNK usually potentiate TGF-β/Smad-induced responses. In addition, it has been shown that TGF-β-induced activation of Ras/Erk MAPK signaling can induce TGF-β1 expression, thereby amplifying the TGF-β response and inducing secondary TGF-β effects. Activation of MAPK pathways by TGF-β may also affect transcriptional responses through direct effects on Smad-interacting transcription factors such as the JNK substrate c-Jun or the p38 MAPK substrate ATF-2 (activating transcription factor 2), thus further increasing the interactions and convergence of TGF-β-induced Smad and MAPK pathways (80, 209, 370).

The Smad2/3-independent pathways of TGF-β signaling also include the PKC-δ isoform of the PKC family of kinases, the cellular Abl kinase (c-Abl), and the Janus kinase 2 (JAK2) that are activated by TGF-β. With regard to c-Abl, it has been found that the most upstream component currently identified in this cascade is PI3K, and these effects are independent of the canonical TGF-β-Smad pathway. Further results demonstrated that c-Abl activation by TGF-β is cell type-specific and Smad-independent and that the initiation and propagation of the regulatory signal(s) from the TGF-β receptor complex to (and from) Smad and c-Abl proteins are distinct. More recently, it was demonstrated that TGF-β induction of EndMT is mediated by c-Abl and PKC-δ, as specific inhibition of their kinase activity or specific siRNA knockdown of their corresponding transcripts abrogated the marked increase in TGF-β-induced α-SMA and Snail-1 expression and protein levels (174). Regarding JAK2, this kinase is likely to play an important role in EndMT during the development of pulmonary arterial hypertension (PAH) associated with idiopathic pulmonary fibrosis (IPF) since high JAK2 expression levels and elevated phosphorylation (JAK2p) were found in pulmonary artery endothelial cells from these patients (214).

Another important pathway involves glycogen synthase kinase-3β (GSK-3β), and it has been shown that GSK-3β kinase phosphorylation was a crucial event in the EndMT process resulting in inactivation of its kinase activity which in turn induced the nuclear accumulation of Snail1 followed by a marked increase in the expression of EndMT triggering genes (209). In contrast, in the absence of GSK-3β phosphorylation, the GSK-3β kinase is active and induces the proteosomal degradation of Snail1, thus abrogating EndMT. The most important TGF-β noncanonical pathways involved in the EndMT process are illustrated in FIGURE 2.

B. Non-TGF-β Initiated Pathways

Several important regulatory pathways besides TGF-β signaling have been also shown to participate in the induction of EndMT under certain specific cellular contexts. These molecular pathways include Notch signaling, the canonical Wnt pathway, as well as other important regulatory mechanisms mediated by cytokines and other inflammatory mediators, caveolin-1 (CAV1), endothelin-1 (ET-1), the hypoxia inducible factor (HIF)-1α hypoxia induced pathway, and the response to oxidative stress. Additional pathways involve those triggered by alterations in fatty acid oxidation (FAO), responses to hyperglycemia, and modulation resulting from changes in mechanical strain caused by the effects of hemodynamic forces and the accumulation of pericellular matrix proteins. The most relevant of these pathways are shown in FIGURE 3.

FIGURE 3.

Diagrammatic representation of other pathways involved in endothelial to mesenchymal transition (EndMT) regulation. Molecular signaling pathways beside the transforming growth factor (TGF)-β pathways that induce or participate in the EndMT process. These include endothelin (ET)-1, NOTCH, caveolin (CAV)-1, Wnt, high glucose, and hypoxia inducible factor (HIF)-1α pathways. The morphogen pathways NOTCH and Wnt are important modulators of EndMT. CAV-1 exerts an inhibitory effect owing to the internalization of TGF-β receptors and their subsequent degradation. ET-1 induces a synergistic stimulation of TGF-β-induced EndMT involving the canonical Smad2/3 pathway. Hypoxia induces EndMT through the effects of HIF-1α activation of Snail1. High glucose concentrations have been shown to induce EndMT involving extracellular signal-regulated kinase (ERK) 1/2 phosphorylation. Shear stress forces (represented by undulating arrows) induce EndMT through several different molecular mechanisms. One mechanism is initiated by the mechanical force-induced release and liberation of TGF-β from the latency associated peptide (LAP) followed by activation of the TGF-β canonical pathway. Other mechanisms include reactive oxygen species (ROS) generation and activation of NFκB followed by the activation of NOX1/4 oxidases resulting in increasing production and accumulation of ROS.

1. NOTCH pathway

Numerous studies have shown that the Notch pathway is a widely used intercellular communication mechanism that following its activation is involved in numerous and crucial regulatory functions including cellular proliferation, apoptosis, and differentiation (133, 247). Endothelial cells express ligands, receptors, and other components involved in this pathway.

The role of Notch signaling in EndMT was first described in vitro in human microvascular endothelial cells and HUVECs (239). These observations were confirmed by subsequent studies showing the crucial contribution of Notch signaling in the EndMT process during cardiac valve development (43, 44). Although the mechanisms involved in Notch-mediated EndMT induction have not been fully elucidated, extensive molecular studies have shown that Notch and TGF-β synergistically stimulate Snail expression and upregulate numerous target genes by transcriptional activation of their corresponding gene promoter regulatory elements (32, 43, 99, 237).

Activation of the Notch signaling pathway has been shown to be involved in the induction of EndMT in animal models as well as in various human pathological conditions. For example, an extensive study in a conditional transgenic mouse model demonstrated that constitutive activation of Notch1 signaling was capable of inducing EndMT in the absence of any other known EndMT stimuli (180). With regard to the role of NOTCH1 activation in the induction of EndMT in human disorders, it was recently shown that the chemokine receptor CXCR7 was a potent regulator of EndMT and that this effect was mediated by inhibition of the Jag1-Notch pathway and was markedly elevated in TGF-β1-treated endothelial cells and in animal models of pulmonary fibrosis (118). The TGF-β1 upregulation of CXCR7 expression was mediated by the Smad 2/Smad 3 canonical signaling pathway. These results suggested that CXCR7-induced Notch pathway inhibition may serve as a feedback mechanism to control TGF-β-induced EndMT and the resulting fibrotic process (118). The role of Notch-induced EndMT in the development of the fibrotic process was investigated in experimentally induced pulmonary fibrosis, and the results showed a significant elevation of expression levels of the Notch1 pathway‐related signaling molecules Notch1 and Jagged1, as well as of α‐SMA protein in pulmonary microvascular endothelial cells isolated from rats developing pulmonary fibrosis following the intratracheal administration of bleomycin. In addition, it was demonstrated that activation of Notch1 signaling was crucial in inducing the transdifferentiation of pulmonary microvascular endothelial cells into profibrotic myofibroblasts during the development of bleomycin-induced pulmonary fibrosis (356, 357).

The majority of studies that examined the role of Notch signaling in the pathogenesis of human diseases indicated that Notch activation stimulated EndMT and the resulting pathological effects and that these effects were accompanied by TGF-β pathway activation. However, it was recently demonstrated that endothelial cell-specific inhibition of canonical Notch signaling in knockout mice resulted in EndMT acceleration during certain stages of wound healing and scar formation, causing exaggerated tissue fibrosis and increased TGF-β1 expression (245). These controversial results are in agreement with previous observations that disruption/abrogation of Notch signaling in adult mice induced by Cre-lox P-mediated RBP-J knockout caused remarkable morphological and molecular changes in endothelial cells in the aorta and aortic valves, indicating the occurrence of EndMT compared with control mice (180). Thus the results described in these two later studies emphasize the complex role of Notch as a driver of EndMT and of the molecular cascades that cause and regulate EndMT in the tissues.

2. Wnt pathway

The Wnt proteins comprise a multigene family of secreted glycoproteins with highly complex receptor signaling pathways (236) that play crucial roles during embryonic development and in the pathogenesis of numerous pathological processes including malignant, cardiovascular, neurological, musculoskeletal, inflammatory, autoimmune, and fibrotic diseases (24, 96, 236, 240). The involvement of Wnt signaling in EndMT was examined recently in cardiac, aortic valve, and renal fibrosis. In cardiac fibrosis it was shown that a significant proportion of α-SMA-positive myofibroblasts present in cardiac tissues following experimentally induced myocardial infarction originated from endothelial cells undergoing EndMT and expressed β-catenin-dependent Wnt signaling (4). Numerous subsequent investigations have confirmed the role of β-catenin in EndMT (171, 366). A related investigation performed an unbiased proteomic analysis of the secretome of profibrogenic in comparison to fibrosis-resistant cultured renal microvascular endothelial cells (181). The profibrogenic cells were obtained from a genetically engineered strain of mutant mice prone to develop fibrosis owing to an endothelial cell-specific mutation in the gene encoding sirtuin 1 (Sirt1endo−/−). The fibrosis-resistant renal microvascular endothelial cells were isolated from a strain of transgenic mice with TGF-β receptor II (TGFβRIIendo+/−) endothelial cell deficiency that is protected from fibrosis. Very high levels of Dickkopf-3 (DKK3) were detected exclusively in the secretome of the profibrogenic cells derived from Sirt1endo−/− mice. Subsequently it was shown that DKK3 was a potent inducer of EndMT and that, whereas DKK1 is a well-known inhibitor of canonical Wnt/β-catenin signaling, DKK3 displays opposite effects (179). However, these results were apparently contradictory to previous observations that the Wnt inhibitor Dkk-1 enhanced EndMT in aortic endothelial cells and induced a mineralizing mesenchymal phenotype (56). These contradictory results indicate that the participation of Wnt pathways in EndMT may be highly dependent on the specific cell type and cellular context and that further investigation will be required to conclusively determine the precise role of Wnt and Wnt-related pathways on EndMT.

3. ET-1

ET-1, the most potent endogenous vasoconstrictor polypeptide identified to date, plays crucial roles in the regulation of multiple vascular functions and in the pathogenesis of various human diseases including primary and secondary PAH. Owing to its high expression in endothelial cells, there has been substantial interest in determining whether ET-1 is capable of inducing EndMT. The role of ET-1 in EndMT was first examined in experimentally induced diabetes mellitus in mice with normal ET-1 levels and in EDN1 knockout mice, and it was shown that ET-1 promoted the development of cardiac fibrosis. Further studies showed that silencing of the EDN1 gene encoding ET-1 abrogated the phenotypic transition of cultured human endothelial cells to a fibroblast phenotype and that this effect occurred through the inhibition of TGF-β signaling (330). More recently, the interactions between ET-1 and TGF-β1 in the induction of EndMT in murine lung microvascular endothelial cells were examined in vitro, and changes in expression levels of various relevant genes participating in EndMT modulation were identified (328). The most remarkable changes included a strong potentiation of TGF-β1-induced EndMT by ET-1 resulting in a synergistic stimulation of the expression of multiple mesenchymal cell-specific genes and of relevant transcriptional coactivators such as Snai1 and Twist1. Most intriguingly, it was also demonstrated that ET-1 induced a marked increase in the expression of TGF-β1 and TGF-β2 and of the three TGF-β receptors, suggesting that ET-1 may be capable of initiating an autocrine loop by inducing expression of the TGF-β ligands and, therefore, resulting in amplification of the induction of EndMT (328). These results were further validated in vivo in mice that had received continuous administration of either TGF-β1 alone, ET-1 alone, or the combination of TGF-β1 plus ET-1. Immunohistological analysis indicated that a small number of cells undergo EndMT in mice treated with either TGF-β1 or ET-1 alone but that treatment with TGF-β1 plus ET-1 greatly increased the number of cells undergoing EndMT (328). Another study of the role of ET-1 in EndMT was performed employing immunopurified CD31-expressing dermal endothelial cells isolated from systemic sclerosis (SSc) patients. The results showed that ET-1 alone was able to induce EndMT in normal and SSc dermal endothelial cells and did not require additional TGF-β treatment and that the ET-1 effects involved the Smad pathway. It was also shown that ET-1-induced EndMT was blocked by the specific ET-1 receptor antagonist macitentan, thus conclusively demonstrating that ET-1 alone or in combination with TGF-β is capable of inducing EndMT (63). Subsequently, the ability of dual ET-1 receptor antagonists to abrogate the molecular and functional effects of ET-1-induced EndMT was examined in cultured human microvascular endothelial cells. The results showed that both ET-1 and TGF-β1 induced EndMT in these cells, an effect associated with a reduced ability to initiate the process of endothelial cell damage repair in vitro. Pretreatment with ET-1 dual receptor antagonists significantly restored the in vitro functional repair capacity of the endothelial cells and antagonized the EndMT process reversing the altered cell function and the morphological and molecular changes of EndMT induced by ET-1 in cultured human microvascular endothelial cells (285).

4. CAV-1

CAV-1, the main protein component of caveolae, plays an important role in the internalization, trafficking, and degradation of TGF-β receptors and, therefore, is intimately involved in the regulation of TGF-β signaling and the associated pleotropic TGF-β cellular responses as well as in the pathogenesis of TGF-β-mediated fibrotic diseases (76, 123, 262). Given the crucial role of TGF-β in the induction of EndMT, the involvement of CAV-1 in EndMT was examined recently employing immunopurified CD31+/CD102+ endothelial cells isolated from lungs of Cav-1 knockout mice (Cav-1−/−mice), and the results demonstrated that EndMT occurred spontaneously in Cav-1−/−mice in vivo (176). In the same study, endothelial cells isolated from lung tissues from wild-type and from Cav-1−/−mice were cultured and treated in vitro with TGF-β1 before and following restoration of functional CAV-1 domains using a cell-permeable CAV-1 scaffolding domain peptide. Lung microvascular endothelial cells from Cav-1 knockout mice displayed high levels of spontaneous molecular changes indicative of EndMT, which increased even further following TGF-β treatment. The important role of CAV-1 in the development of spontaneous and TGF-β1-stimulated EndMT was confirmed by the demonstration that EndMT morphological and molecular alterations were abrogated following the restoration of functional CAV-1 domains (176). CAV-1 as a mediator of endothelial cells undergoing EndMT was also shown by the demonstration that serum from patients with sepsis induced HUVECs to differentiate into mesenchymal cells in vitro (136). A proteomic analysis identified 29 proteins that were differentially expressed in these cells including a marked increase in mesenchymal-specific proteins and a concomitant reduction in the levels of CAV-1. Furthermore, the crucial role of CAV-1 in this process was documented by the demonstration that EndMT was markedly attenuated upon restitution of CAV-1 expression levels by transfection of a CAV-1 expression vector into endothelial cells exposed to serum from patients with sepsis (136).

5. Cytokine and cytokine-mediated pathways

The earliest observations demonstrating the involvement of inflammatory mediators in the induction of EndMT described that HUVECs cultured in the presence of supernatants from activated peripheral blood mononuclear leukocytes underwent remarkable morphological and phenotypic changes that subsequently were to be known as EndMT, and demonstrated for the first time an important role of IL-1β in this process (221). Numerous subsequent observations have confirmed that IL-1β and other inflammatory cytokines are potent inducers of EndMT as discussed in recent reviews (57, 248) and diagrammatically illustrated in FIGURE 4. One of these studies showed that in vitro treatment of human dermal microvascular endothelial cells with IL-1β resulted in profound phenotypic changes of EndMT and the loss of endothelial-specific functions such as the ability to form tubule like structures indicative of intact angiogenic effects (268). Similar results were obtained in cultured human intestinal mucosal endothelial cells showing that IL-1β was an even more powerful inducer of EndMT than TGF-β, an observation that led the authors to propose that IL-1β is the main inducer of EndMT in intestinal mucosa (266). Other studies demonstrated that IL-1β acts synergistically in the presence of TGF-β2 to induce EndMT in HUVECs (197).

FIGURE 4.

Diagrammatic representation of endothelial to mesenchymal transition (EndMT) induction by cytokines and other inflammatory mediators. The cytokines tumor necrosis factor (TNF)-α, interleukin (IL)-1β, IL-6, and IL-13 have been show to induce EndMT either directly or through molecular interactions with other pathways. IL-1β, IL-6, and TNF-α induction is mediated by IRAk and is followed by activation of NFκB. IL-13-induced EndMT involves an alternate pathway mediated by STAT6 and the AKT molecular cascade.

Other cytokines from the interleukin family such as IL-6 and IL-13 have also been shown to stimulate EndMT. The involvement of IL-6 and of TNF-α in the induction of EndMT was first demonstrated in tridimensional in vitro cultures of porcine embryonic and adult valvular endothelial cells, and it was shown that these effects involved Akt/NFκβ and canonical TGF-β pathway activation in embryonic cells, whereas in the adult cells, the TGF-β pathway was not involved as the cytokine effects were observed even after TGF-β inhibition (194). A related study demonstrated that the L1 transmembrane glycoprotein also known as L1CAM, which is a potent inducer of IL-6 and of the IL-6-receptor alpha (IL-6Rα), caused marked stimulation of IL-6-mediated STAT phosphorylation resulting in a potent induction of EndMT in murine immortalized cultured endothelial cells (193). In contrast to the extensive investigation of the role of type I cytokines in EndMT, there is only one study that examined the contribution of the type II cytokine IL-13 to the EndMT process and described potent effects of IL-13 on the phenotype of human pulmonary artery endothelial cells in vitro showing that IL-13 induced EndMT in these cells, an effect mediated by a potent downregulation of the level of miR-424/503 that resulted in markedly enhanced expression of Rictor, a key molecule in the mTOR pathway (300).

Tumor necrosis factor (TNF)-α is a cytokine that has been implicated in numerous endothelial cell activation responses to inflammatory events inducing expression of adhesion molecules, such as vascular cell adhesion molecule 1 (VCAM-1) and intercellular adhesion molecule 1 (ICAM-1) was examined on human dermal microvascular endothelial cells and in rat mesenteric lymphatic endothelial cells in vitro, and it was shown that TNF‐α demonstrated a potent EndMT effect on the cells (42, 47). Another group of cytokines that have been shown to be important in the pathogenesis of numerous inflammatory and immunologic disorders are the interferons (IFNs). However, the potential role of IFNs in EndMT has not been fully explored. One study investigated changes in profibrotic gene expression induced by IFN-γ treatment on normal human dermal microvascular endothelial cells in culture (61). The results raised the possibility that IFN-γ could induce phenotypic changes consistent with EndMT. Gene expression analysis demonstrated increased expression of numerous genes associated with the EndMT process including TGF-β2 and SNAI1. Furthermore, treatment with IFN-γ resulted in morphological changes of EndMT similar to those induced by treatment with TGF-β1 that correlated with downregulation of VE-cadherin, as well as with major changes in the actin cytoskeleton. However, since these studies were performed with neonatal foreskin endothelial cells, further investigation will be required to validate and generalize these observations (61).

NF-κB has also been shown to play an important role in cytokine-induced EndMT as costimulation with IL-1β plus TGF-β2 synergistically induced EndMT in HUVECs and that these effects were mediated by NF-κB activation (197). Stimulation of primary endothelial cells with TNF-α and IL-6 has also been shown to induce NF-κB activation-mediated EndMT (248). These observations collectively support the notion that NF-κB is a central mechanism involved in EndMT development under inflammatory conditions. It is important to emphasize that NF-κB activation can be induced not only by cytokines but also by diverse other stimuli; therefore, it is likely that NF-κB plays a role in EndMT development mediated by a wider spectrum of conditions besides inflammatory events. Among other investigations of EndMT induction by inflammatory mediators, the role of STAT3 on the induction of EndMT was examined in rat mesenteric endothelial cells cultured in vitro. Although it is generally accepted that STAT3 can inhibit TGF-β1-induced effects by interacting with Smad3 and decreasing TGF-β-mediated transcriptional responses, it was recently reported that the inflammatory process-induced suppression of STAT3 expression in endothelial cells resulted in their conversion into activated fibroblasts. These observations were validated by fluorescence immunohistology analysis of intact rat aortic arteries. The underlying molecular mechanism involved modulation of TGF-β1 secretion, changes in ALK5 activation pathway, and translocation of phosphorylated Smad into the nucleus (23). These studies further suggested that any inflammatory pathway resulting in the subsequent downregulation of STAT3 would translate into an unopposed fibrotic process mediated by EndMT.

The direct role of inflammatory cells, particularly of polymorphonuclear leukocytes and of macrophages and other mononuclear cells in EndMT induction, has not been examined in detail; however, there is evidence that these cells may directly cause EndMT. Thus considerable recent interest has been focused on the direct effect of inflammatory cells on the development of EndMT, and it was demonstrated that macrophages are able to induce EndMT by upregulating C-C motif chemokine ligand 4 (CCL-4), as well as by increasing endothelial permeability and monocyte adhesion (353). Furthermore, it was shown that nitro-oleic acid, an endogenously generated nitrolipid characterized by potent inhibitory effects on activated macrophages, was capable of regulating the interactions between macrophages and endothelial cells and decreasing TGF-β activation pathways resulting in a potent reduction of macrophage and TGF-β-mediated EndMT (9). Although not directly related to the induction of EndMT by inflammatory cells, the effects of the powerful bacterial cell membrane inflammatory mediator lipopolysaccharide (LPS) on EndMT were examined employing HUVECs in culture and in perfused whole umbilical vein vessels. The results demonstrated that LPS causes potent induction of EndMT that was mediated by ALK5 activation and required LPS-mediated Smad2 phosphorylation (86), and it was further shown that these effects involve the transient receptor protein melastatin 7 (84, 85). More recently it was demonstrated that neutrophil extracellular traps or NETS, fibrous networks derived from activated neutrophil granule proteins that protrude from their cell membranes allowing them to catch and destroy bacteria, may cause deleterious effects on endothelial cells and that these effects are mediated by their ability to induce EndMT (255).

C. Induction of EndMT by Oxidative Stress, Metabolic Alterations, Hypoxia, and Shear Stress Forces

1. Oxidative stress

Oxidative stress and reactive oxygen species (ROS) have been recently recognized as important inducers of EndMT. Recently, primary HUVECs exposed to increasing concentrations of H₂O₂ in culture were shown to display a dose-related transition into mesenchymal cells, and this effect was mediated by enhanced endogenous TGF-β production and involved Smad3, p38, and NF-κB activation (222). Although there are multiple sources of intracellular ROS, extensive studies have shown that most ROS production derives from the activation of the nicotinamide adenine dinucleotide phosphate (NADPH) oxidase system (NOX). The most relevant NOX species regarding endothelial cell functions and pathology are NOX2 and NOX4 (82). Both NOX isoforms are present at high levels in the vascular endothelial compartment. Unlike other members of the NOX family, NOX4 is constitutively active and does not require any cofactors to exert its potent functions; therefore, its overall effects depend on the levels of NOX4 protein within the cell that are, in turn, determined by the levels of the corresponding NOX4 mRNA. Although the role of NOX4 in EndMT has not been directly examined, given that NOX4 is an important downstream mediator of TGF-β-induced effects, it has been recently suggested that it may also induce EndMT. This suggestion is based on recent observations that TGF-β causes a potent stimulation of NOX4 gene expression and of NOX4 protein levels and that NOX4 is involved in several downstream molecular effects of TGF-β1 (250, 253). Although NOX4 induction of EndMT was not specifically examined, it was recently demonstrated that hyperglycemia enhances NOX4-mediated ROS generation and stimulates renal hedgehog interacting protein (Hhip) gene expression and that the elevated renal Hhip protein levels were shown to activate the TGF-β1-Smad2/3 cascade in vivo and in vitro and promoted EndMT leading to endothelial cell fibrosis/apoptosis (373). Owing to the high levels of NOX2 in endothelial cells, there has been recent interest in its role in EndMT. To investigate the effects of increased endothelial cell NOX2 activation on the development of EndMT, a transgenic mouse model with endothelial-specific NOX2 overexpression was generated. The results showed that the NOX2 overexpressing mice had increased EndMT compared with wild-type mice. The role of endothelial NOX2 on EndMT was confirmed in cultured human aortic endothelial cells with adenovirally mediated overexpression of NOX2 demonstrating that the transition from endothelial- to fibroblast-specific gene expression was markedly enhanced in NOX2-overexpressing endothelial cells (229).

In relation to ROS-mediated induction of EndMT, the effects of antioxidant treatment on EndMT have been investigated, and it was demonstrated that N-acetyl-l-cysteine, a potent ROS scavenger compound, prevented endotoxin-induced EndMT in HUVECs. Furthermore, it was shown that ROS production in endotoxin-treated endothelial cells involved NOX activity since the NOX inhibitor apocynin abrogated the induction of EndMT markers, indicating that NOX activation is a crucial step in endotoxin-induced EndMT (86). Another study examining human aortic endothelial cells under H₂O₂-induced oxidative stress showed that the cells exposed to H₂O₂ displayed EndMT. It was further demonstrated that the EndMT-related changes were abolished by low-intensity pulsed ultrasound (LIPUS) and that the anti-EndMT effects of LIPUS were mediated by its ability to inhibit oxidative stress (170). Thus, collectively, these extensive investigations provide convincing evidence to support the crucial role of ROS and NOX in the initiation and/or progression of EndMT as discussed in detail in a recent review (307).

2. Fatty acid oxidation

Although the role of metabolic alterations in EndMT has not been extensively examined, it has been recently demonstrated that TGF-β signaling-induced EndMT is accompanied by inhibition of fatty acid oxidation (FAO) and by a marked reduction in the intracellular levels of acetyl-CoA and that FAO inhibition potentiates EndMT through inhibition of SMAD7 signaling. The molecular mechanisms involved a reduction in the acetylation of SMAD7 resulting in its decreased stability and, therefore, in a decrease in the inhibitory effect of TGF-β signaling normally caused by SMAD7. It was further shown that TGF-β plus IL-1β treatment of human lung microvascular endothelial cells resulted in a marked reduction in expression of carnitine palmytoil transferase II (Cpt2), an enzyme involved in FAO, and that endothelial cell-specific genetic disruption of Cpt2 causes increased in vivo EndMT mediated by TGF-β Smad pathway activation (343). These observations are an important contribution to understanding the molecular mechanisms of EndMT by identifying remarkable metabolic alterations that may play a crucial role in the induction of a mesenchymal phenotype in endothelial cells and that FAO is a metabolic process required for maintenance of the endothelial cell phenotypic identity and that regulation of FAO and modulation of endothelial cell phenotypic plasticity are intimately connected. Therefore, FAO might be considered a novel checkpoint that regulates the maintenance of endothelial cell identity and that further understanding of the role of metabolic alterations within the endothelial cells may lead to the development of metabolomic interventions capable of modulating EndMT and abrogating its deleterious effects (186).

3. Hypoxia

Extensive investigations have demonstrated that hypoxia is a potent inducer of EndMT in a variety of endothelial cells and under diverse experimental conditions, and have shown that these effects are mediated by HIF-1α (58, 81, 181, 184, 222, 229, 282, 301, 345, 347). One of these studies found that hypoxia efficiently induces human coronary artery endothelial cells to undergo EndMT through a HIF-1α-dependent pathway that involved a marked stimulation of the expression of the transcription factor Snail1 (347). Subsequently, it was shown that HIF-1α and TGF-β/SMAD signaling pathways synergistically stimulated hypoxia-induced EndMT through both DNMT3a-mediated hypermethylation of RASAL1 promoter and direct Snail1 induction and that hypoxia mediated a potent autocrine stimulation of TGF-β2, an effect documented both at the gene expression level, as well as at the level of protein production (345, 347). The in vitro observations of the effects of hypoxia on EndMT in various types of endothelial cells have been confirmed in vivo in various experimental animal models including the demonstration that administration of the experimental agent SUS416 under hypoxic conditions induced prominent EndMT phenotypic transition associated with PAH. The role of hypoxia-induced EndMT in the generation of pulmonary hypertension was confirmed in extensive endothelial cell lineage tracking in mice expressing green fluorescent protein specifically in endothelial cells (297). A similar investigation showed that hypoxia- and monocrotaline-induced PAH in rats was also associated with typical changes of EndMT and was mediated by hypoxia stimulation of platelet-derived growth factor and TGF-β expression (286). Although it is well recognized that most of the effects of hypoxia are mediated by HIF-1α, it was recently shown that HIF-2α was substantially elevated in lung endothelial cells isolated from patients with idiopathic PAH cultured in vitro under normoxic conditions and that HIF-2α induced elevated levels of SNAI1 and SNAI2 and resulted in a strong transition of these cells into a mesenchymal phenotype (301).

The mechanisms involved in the induction of EndMT by hypoxia have been only partially elucidated; however, numerous pathways have been implicated. For example, it was shown in human cardiac microvascular endothelial cells that both TGF-β and Notch were involved and that this pathway was mediated by RUNX3 transcription factor, since it was abrogated or attenuated by RUNX3 knockdown (184). The role of endothelial cells in cardiac remodeling under hypoxic conditions and the potential paracrine effects of hypoxic endothelial cells on adult rat ventricular cardiomyocytes were investigated in ventricular cardiomyocytes exposed to conditioned medium from hypoxic/reoxygenated microvascular endothelial cells. The results showed that the conditioned medium caused activation of TGF-β signaling including enhanced SMAD1/5 and SMAD2 phosphorylation and that these effects were blocked by specific ALK1 or ALK5 inhibition, indicating that hypoxia induces EndMT through TGF-β1/SMAD2 signaling (282). A related investigation examined the molecular mechanisms involved in the development of EndMT under hypoxia in pulmonary artery endothelial cells. The results showed a reduction of the expression of bone morphogenetic protein 7 (BMP-7) both in vivo and in vitro demonstrating that under normoxic conditions, BMP-7 deficiency induced spontaneous EndMT and increased endothelial cell migration. Furthermore, these alterations were markedly attenuated following pretreatment with rh-BMP-7.

In other studies an extensive analysis of the molecular alterations induced by hypoxia was performed. The results indicated that mTOR phosphorylation was involved in EndMT and that BMP-7 suppressed hypoxia-induced mTORC1 phosphorylation (368). Subsequently, it was demonstrated that RhoJ depletion attenuated hypoxia-induced EndMT and that the EndMT inhibitory effect of RhoJ depletion was mediated by blockade of binding of the transcriptional repressors TWIST and SNAIL to the promoter region of relevant endothelial cell-specific genes such as VE-cadherin. RhoJ also mediated the induction of TWIST and SNAIL expression and promoted the binding of HIF-1α to the promoters of genes involved in EndMT. The mechanisms mediating these potent effects involved RhoJ upregulation of WDR5 expression, a key component of the trimethylated histone H3K4 resulting in the accumulation of the proEndMT trimethylated histone H3K4 (181).

4. Hyperglycemia

Although the molecular mechanisms implicated in diabetes-associated vascular alterations have been extensively investigated, there has also been recent interest in examining the possibility that EndMT may play a role in the initiation, development, and progression of the frequently severe vasculopathy affecting the large vessels and the microvasculature in diabetes, particularly affecting the kidneys (167, 168). The direct role of elevated glucose levels on cultured human aortic endothelial cells in vitro was examined, and it was shown that high glucose concentrations induced EndMT, an effect mediated by angiotensin II (304). It was further demonstrated that high-glucose-induced EndMT was abrogated by treatment of the cultures with an angiotensin II receptor 1 inhibitor (Ibesartam). These observations suggested a novel mechanism for the occurrence of EndMT induced by hyperglycemia involving stimulation of angiotensin II synthesis (305). Similar results were obtained in human retinal endothelial cells in culture, and it was shown that high glucose induced EndMT in these cells. EndMT changes were also shown to occur in vivo in the retinal vessels of mice with short-term experimentally induced diabetes. It was also demonstrated that TGF-β played a key role in this process and that the EndMT changes were regulated by miR-200b. These observations were confirmed in transgenic mice with endothelial-specific miR-200b overexpression following experimentally induced diabetes (40). A related study demonstrated that the level of miR-328 was significantly upregulated in high-glucose-induced EndMT and that miR-328 displayed the capability of inducing EndMT independently of TGF-β1, and this effect was abrogated by antagomiR-328. Further investigation demonstrated a significantly higher expression of p-MEK1/2 and p-ERK1/2 in HUVECs transduced with miR-328, and a lower expression of p-MEK1/2 and p-ERK1/2 in cells transduced with antagomiR-328 (54).

The underlying molecular mechanisms and pathways responsible for high glucose-induced EndMT were examined recently in HUVECs cultured under high-glucose conditions. The results showed that high-glucose-induced EndMT was mediated by MAPK pathway-dependent activation with increased phosphorylation of ERK1/2. The results were confirmed utilizing a specific ERK signaling pathway blocker that caused a significant attenuation of the high-glucose-induced increase in phosphorylated ERK1/2 and abrogated the expression α-SMA expression. It was also found that the ROS inhibitor N-acetylcysteine decreased the high-glucose-induced TGF-β1 expression and the EndMT transition, indicating an important role of ROS/ERK1/2/MAPK-dependent mechanisms in high-glucose-induced EndMT (359). Mechanistic investigations of the molecular alterations occurring in rat glomerular endothelial cells exposed to high glucose demonstrated that high glucose medium resulted in EndMT and that this process was mediated by activation of ROCK1 (246) and that it may also involve serum response factor (372).

5. Shear stress forces

Vascular endothelial cells are constantly exposed to a cyclic pumping force caused by circulatory system blood flow throughout the entire body and respond in a highly specific manner to these hemodynamic forces. The vascular endothelium senses the frictional shear stress generated by blood flow, and in response, the vessel endothelial cells initiate multiple molecular reactions including remarkable changes in gene expression (325), that eventually alter vessel function to appropriately adapt to alterations in blood flow characteristics. Furthermore, changes in blood flow also induce endothelial cell paracrine signaling to the smooth muscle cells in the medial layer of the vessel to regulate vessel tone to appropriately respond to blood flow and volume changes (124).

The endothelial cells lining the vascular endothelium are highly sensitive to the mechanical forces generated by the linear laminar shear (LLS) from intraluminal blood flow as well to the cyclic strain (CS) resulting from the pulsatile pressure changes caused by the repetitive vessel wall mechanical deformation during the cardiac cycle. Transduction of these signals is achieved through a variety of cell receptor-triggered intracellular signaling pathways and specific flow patterns and cyclic strain regulate differentially multiple aspects of the endothelial cell phenotype and induce important modulation of endothelial cell gene expression (20, 104, 157, 264, 308). Although previous studies have demonstrated that shear stress induces EndMT in embryonic endothelial cells (88, 306), the effects of variations in shear stress profiles on EndMT induction in adult endothelial cells have not been fully elucidated. To further examine the mechanisms involved, a shear stress bioreactor was developed that allowed to expose adult cardiac valve endothelial cells embedded on a three-dimensional matrix containing relevant ECM macromolecules to varying and multiple shear stresses. The results demonstrated that certain changes in ECM composition coupled with low shear stress induced EndMT causing marked upregulation of mesenchymal cell-specific molecules and downregulation of endothelial cell-specific gene expression (195, 215). The molecular mechanisms underlying low shear stress regulation of EndMT were investigated in cultured HUVECs and porcine aortic endothelial cells exposed to low shear stress (196), and it was found that the phenotypic changes were associated with a marked induction of Snail expression and that Snail positively regulated the expression of EndMT markers (Slug, N-cadherin, α-SMA). The role of Snail on low shear stress induction of EndMT was validated in a murine animal model with experimentally induced variations in carotid artery flow levels (196). It was further demonstrated that EndMT induction mediated by variations in shear stress caused by altered ECM composition involved ERK1/2 signaling in aortic valve endothelial cells (72).

Multiple in vitro and in vivo investigations employing various types of endothelial cells in culture and several animal models have shown that normal linear blood flow ensures the maintenance of endothelial homeostasis and prevents EndMT. In contrast, oscillatory or nonuniform shear stress, also referred to as disturbed shear stress (DSS), has detrimental effects on endothelial homeostasis (157). There are numerous mechanisms involved in the induction of EndMT by alterations in the patterns and turbulence of the vascular blood flow. The most important mechanism is the stress force-mediated release and liberation of TGF-β from the LAP followed by TGF-β activation and TGF-β-induced Smad2/3 signaling. Disturbed flow shear stress also stimulates the expression of BMP4 that may cause ROS formation and activation of NFκB. Lastly, disturbed shear stress induces NOX activity resulting in the generation of ROS. All these signaling pathways are responsible for induction of EndMT as illustrated in FIGURE 3. Therefore, local hemodynamic changes are capable of profoundly affecting the endothelial cell phenotype and the production of associated proteins that are involved in the transduction of shear and mechanical stress signals into the cytosol. These include integrins, receptor and non-receptor tyrosine kinases, heterotrimeric G proteins, and endothelial cell specific adhesion molecules such as CD-31/PECAM-1 and VE-cadherin (124). The intracellular signaling cascades involved in the transduction of mechanical shear forces include Wnt pathway proteins, MAPK activation, and NF-κB signaling, resulting in a variety of endothelial cell specific responses including EndMT. Thus the extensive literature examining the role of blood flow-generated stress forces indicates that high LSS exerts a protective effect inhibiting EndMT either directly or through indirect mechanisms that interfere with TGF-β signaling. In contrast, DSS suppresses these protective mechanisms rendering endothelial cells more prone to EndMT. Furthermore, DSS can induce EndMT directly through stimulation of BMP4 or increased ROS production. Indeed, exposing aortic endothelial cells to DSS both in vitro and in vivo efficiently induces EndMT in the absence of other stimuli (223).

V. REGULATION OF EndMT

A. Regulation by Growth Factors

1. TGF-β

As described in previous sections, it is generally accepted that the TGF-β family of growth factors are the most important triggering molecules for the induction of EndMT. However, it must be emphasized that there are multiple sources of the TGF-β molecules that may initiate the EndMT process. Some of these multiple sources include TGF-β molecules released from a TGF-β storage pool tethered to the pericellular ECM, or newly secreted molecules released from epithelial and endothelial cells, fibroblasts, and immune/inflammatory cells, such as dendritic cells, eosinophils, macrophages, and various subsets of lymphocytes as illustrated in FIGURE 5.

FIGURE 5.

Potential sources of transforming growth factor (TGF)-β for initiation of endothelial to mesenchymal transition (EndMT). The most important cellular and noncellular sources of TGF-β available to endothelial cells for initiation of EndMT are shown. Among the cellular sources, the inflammatory and immunologic cells and the autogenous endothelial cells following their stimulation are the most important. The TGF-β bound to extracellular matrix (ECM) molecules is the most important noncellular source.

2. Other growth factors

The contribution of other growth factors has been examined recently and will be reviewed in the following sections.

A) FIBROBLAST GROWTH FACTORS.

There has been recent interest in examining the participation of several of the twenty-two members of the extended fibroblast growth factor (FGF) family of proteins to the development of EndMT, owing to their widespread cellular distribution, their pleotropic functions, their close functional interactions with the TGF-β growth factors, and their high relevance to the vascular system and to endothelial functions (228). The effects of FGF2 to EndMT depend on cell and tissue type, and on the particular conditions of the cellular microenvironment. It has been found that FGF2 has a protective role against EndMT in numerous types of endothelial cells including human aortic, dermal, lymphatic, and umbilical arterial and vein endothelial cells (49, 50, 69, 137). The molecular mechanisms involved in the protective effects of FGF2 towards TGF-β-induced EndMT are of substantial relevance and have been studied intensively, although they have not been fully elucidated. However, recent evidence suggests that FGF2 stimulates several distinct pathways to modulate EndMT. Recent studies have shown that FGF2 regulated gene expression in dermal lymphatic endothelial cells treated with TGF-β or isolated from mice with experimentally induced lymphedema through Smad2 in a Ras-MAPK dependent manner (137) and inhibited TGF-β-induced EndMT in HUVECs by regulating ALK5, TGFBR2, and SARA, effects mediated through miR-20a (69). One related investigation performed high throughput molecular screening of TGF-β-induced EndMT in ovine mitral valve endothelial cells and identified FGF2 as the strongest negative regulator of TGF-β-induced EndMT (321). FGFR1 was found to be essential for the protective effect of FGF preventing EndMT development in several in vitro and in vivo studies, and it was shown that these protective effects of FGFR1 were mediated by induction of the MAP4K4 pathway (169). Other recently described mechanisms have identified FGF regulation of EndMT through induction of microRNAs (miRNAs) that exert inhibitory effects on EndMT. One of the most important miRNA involved in this regulatory mechanism is miRNA let-7. These studies indicated that basal FGF signaling in endothelial cells is required to maintain let-7 miRNA expression and that high let-7 miRNA levels are in turn responsible for suppression of TGF-β signaling and inhibition of TGF-β-induced EndMT. Thus FGF control of let-7 miRNA expression levels directly regulates TGF-β induction of EndMT signaling and links these two critical growth factor signaling pathways that are required to maintain the normal homeostatic state of the vasculature (49). While many factors can and do affect this process, the occurrence of FGF resistance is clearly associated with inflammation. Indeed, this link was traced to the ability of key inflammatory mediators, including IFN-γ and TNF-α, to reduce FGFR1 expression and to markedly decrease FGF-dependent let-7 miRNA levels (49). These observations were further confirmed by the demonstration that intense vascular inflammation such as that associated with allogeneic artery transplantation induce an extensive EndMT response that is mediated by inhibition of FGF2 signaling (50). A second major signaling pathway that contributes to the antagonistic ability of FGF2 is the MEK/ERK pathway. Activation of MEK/ERK in endothelial cells by FGF2 is critical for cell proliferation, migration, survival, and angiogenesis, and extensive studies have shown that activation of the downstream MEK/ERK pathway is also capable of counteracting TGF-β-stimulated EndMT (137).

B) BONE MORPHOGENETIC PROTEINS.

There has been increasing interest in the role of BMP-7 as a potent inhibitor of EndMT following pioneering observations demonstrating that administration of recombinant human BMP-7 abrogated the development of EndMT and inhibited the progression of cardiac fibrosis in mouse models of cardiac fibrosis induced by pressure overload or chronic allograft rejection. In vitro studies with TGF-β1-induced EndMT in human pulmonary microvascular endothelial showed that TGF-β1 stimulation resulted in a reduction of BMPR2 expression and that addition of exogenous BMPs to restore the balance in BMPR2/TGF-β signaling abrogated TGF-β-induced EndMT (363). The effects of BMP-7 on EndMT were further examined in vivo in rats induced to develop PAH following exposure to chronic hypoxic conditions, and in vitro in cultured pulmonary artery endothelial cells following silencing of BMP-7 expression employing specific siRNA inhibition (368). These results showed that BMP-7 expression was markedly decreased under hypoxia both in vitro and in vivo and that hypoxia-induced EndMT and cell migration were markedly attenuated following pretreatment with recombinant human BMP-7. It was further shown that mTOR phosphorylation was involved in EndMT and that BMP-7 suppressed hypoxia-induced mTOR complex 1 (mTORC1) phosphorylation in vitro in cultured rat pulmonary artery endothelial cells. Thus collectively these results demonstrate that BMP-7 is a potent antagonist of hypoxia-induced EndMT in pulmonary artery endothelial cells and that the role of BMP-7 involves the mTORC1 signaling pathway (368).

C) VASCULAR ENDOTHELIAL GROWTH FACTOR.

Vascular endothelial growth factor (VEGF) is a potent angiogenic factor that plays an important role in the cardiovascular system and participates in numerous physiological events including bone formation, wound healing, and hematopoiesis. During embryonic development, VEGF participates in the formation of the atrioventricular canal cushion causing a potent inhibition of the conversion of endocardial cells to mesenchymal tissue. It has also been recently shown that VEGF inhibits TGF-β-induced EndMT in cultured ovine aortic valve endothelial cells and that these effects are opposite to the EndMT-inducing effects of Notch1 (350). Subsequently, it has been demonstrated that owing to its inhibitory effects on EndMT, VEGF displays a protective role in EndMT-mediated kidney and cardiac fibrosis pathogenesis (138, 279, 280).

D) PLATELET-DERIVED GROWTH FACTORS AND HEPATOCYTE GROWTH FACTOR.

The possible contribution of platelet-derived growth factors (PDGFs) to the initiation or progression of EndMT was examined in isolated bovine pulmonary artery endothelial cells in vitro, and it was shown that both PDGFAA and PDGFBB were capable of inducing EndMT, although PDGFBB was substantially a more potent inducer than PDGFAA. A remarkable observation was that PDGF inhibition with imatinib mesylate was able to prevent the development of PAH in vivo in rats with monocrotaline-induced PAH (286). The role of hepatocyte growth factor (HGF) in the regulation of EndMT was examined in HUVECs and human renal glomerular endothelial cells treated with TGF-β1 and/or varying doses of HGF. The results showed that HGF attenuated the development of TGF-β1-induced EndMT and modified other features of the EndMT transdifferentiated cells including a marked reduction of their motility and migration. Furthermore, it was demonstrated that the inhibitory effect of HGF on the EndMT process involved the TGF-β/Smad and Akt/mTOR signaling pathways (322).

B. Transcription Factors Involved in EndMT Regulation

Numerous transcription factors that have been shown to participate in the regulatory gene expression effects are involved in the induction of EndMT in humans including SNAI1, SLUG, ZEB1, ZEB2 or SIP1, and TWIST. However, given the complexity of the EndMT process, other as yet not identified transcription factors and regulatory proteins will very likely be shown to be involved.

1. SNAI, SLUG, and ZEB

SNAI1, SNAI2, SLUG, ZEB1, and ZEB2 belong to the same SNAIL family of zinc finger transcription factors that exert potent transcriptional repressor activity mediated by a shared NH₂-terminal protein-interaction domain, known as the SNAG domain, that is required for their transcriptional repressor activity. The downstream signaling pathway initiated by TGF-β results in strong upregulation of the transcriptional repressor Snail1. Extensive experimental evidence indicates that Snail1 plays a crucial role in EndMT, although the mechanisms involved have not been fully elucidated. The upregulation of Snail was shown to be dependent on the activation of the Smad pathway and the MEK, PI3K, and p38 MAPK pathways by TGF-β (209). Although overexpression of Snail was sufficient to induce EndMt, Snail expression alone was not sufficient to induce EndMT, and it was shown that inhibition of GSK-3β was required for TGF-β-mediated induction of EndMT (209). The crucial role of Snail1 in EndMT was further documented in cultured murine embryonic stem cell-derived endothelial cells induced to become mesenchymal mural cells following treatment with TGF-β showing that siRNA knockdown of Snail blocked TGF-β-induced EndMT (154). The role of Snail1 in retinal neovascularization was also examined, and it was found that high levels of SNAI1 expression by endothelial cells were involved in pathological ocular neovascularization (293). It was also shown that experimentally induced SNAI1 overexpression triggered cell morphological changes and enhanced cell motility in normal endothelial cells, whereas loss of SNAI1 attenuated their migration and invasion and inhibited vascular sprouting. Gene expression analysis revealed that SNAI1 knockdown decreased the expression of numerous genes related to EndMT including genes involved in cytoskeleton rearrangement and ECM remodeling. These results were validated by intravitreal injection of SNAI1 inhibitor siRNA that suppressed new vessel formation in a murine model of choroidal neovascularization (293).

Regarding the role of Snail1 in EndMT, it has been shown that in contrast to TGF-β-induced EndMT that requires Snail1, EndMT induction by Notch does not involve Snail1 but the related family member Slug (99, 100). Related observations demonstrated that in human microvascular endothelial cells siRNA-induced suppression of expression of the receptor tyrosine kinase Tie1 caused these cells to undergo EndMT and that this process involved SLUG since a reduction of SLUG levels abrogated the EndMT process (103). The mechanisms mediating the induction of EndMT by Slug have not been studied as extensively as those involving Snail. However, a study on human dermal microvascular endothelial cells cultured in the presence of conditioned media from breast cancer cells showed that the cells lost VE-cadherin expression, an effect induced by Slug through inhibition of the VE-cadherin gene promoter (185). Besides Snail and Slug, it is very likely that the other members of this family of transcriptional repressors such as ZEB1 and ZEB2 participate in EndMT under specific cellular contexts. However, this possibility needs further investigation.

2. TWIST

Twist1 and Twist2 are highly conserved members of a large family of helix-loop-helix (HLH) transcription factors that function as molecular switches to activate and repress the transcriptional regulation of expression of numerous genes characteristic of the mesenchymal cell lineage. Although these two proteins are highly conserved and exhibit similar functions in vitro, recent studies have demonstrated that they are not redundant and that under specific conditions and cellular contexts they may perform different functional roles and cause different effects. Although the involvement of Twist1 in the EndMT process has been demonstrated experimentally both in vitro and in vivo, the participation of Twist2 in EndMT has not been described.

Twist1 plays a crucial role as an activator of EndMT in hypoxia-induced pulmonary hypertension (198). In this report, it was demonstrated that Twist1 overexpression induced EndMT in cultured human pulmonary arterial endothelial cells through TGF-β signaling, whereas a nonfunctional mutant Twist1construct containing a mutation that prevented serine 24 phosphorylation (Twist1S42A) failed to induce EndMT. The cellular changes occurring in a fibrin gel-containing human endothelial cells when implanted in the lungs of recipient normal mice in vivo were also examined. The results showed that Twist1-overexpressing cells underwent EndMT in the gel; in contrast, cells overexpressing theTwist1S42A mutant did not show any evidence of transition to a mesenchymal phenotype. Furthermore, hypoxia-induced EndMT was inhibited in endothelial cells overexpressing the Twist1S42A mutant construct in vitro and was attenuated in endothelial cell-specific Twist1 conditional knockout mice in vivo. These findings demonstrated conclusively a crucial role of Twist1 and of the phosphorylation of serine-24 within the molecule in the induction of EndMT (198). Similar observations were made employing murine pulmonary artery endothelial cells cultured under hypoxic conditions, and it was further documented that HIF-1α activated Twist gene transcription by binding directly to its promoter region (364). Twist1 has also been shown to be an important component of various tissue fibrotic processes, and it is very likely that some of its profibrotic effects may be mediated by its stimulation of EndMT as discussed recently (238). Furthermore, mice lacking Twist1 selectively in fibroblasts were protected from experimentally induced skin fibrosis and were comparable to mice with ubiquitous inactivation of Twist1 (241). In agreement with these observations, increased Twist1 expression has been shown in bleomycin‐induced mouse models of tissue fibrosis and in cultured HUVECs undergoing EndMT. Related to these observations, it has been described that geniposide, an agent previously shown to prevent EndMT in the bleomycin‐induced fibrosis mouse model, decreased Twist1 expression, thus collectively demonstrating that Twist1 induces EndMT and that this novel mechanism may be involved in the pathogenesis of various fibrotic disorders.

3. Other transcription factors

Several other transcription factors have been recently shown to be also involved in the EndMT process, although their role in EndMT regulation is just beginning to be explored and their relevance and potential physiological and pathological roles remain to be fully elucidated. Some recent studies have shown that ERG, the most abundant ETS-related gene in endothelial cells, is a crucial determinant of endothelial cell lineage identity and maintenance, and regulates the homeostatic balance between TGF-β and BMP signaling pathways. EGR knock down in vitro as well as extensive in vivo experiments with transgenic mice expressing a constitutive endothelial cell-specific heterozygous deletion of EGR showed that EGR loss induced EndMT and spontaneous liver fibrosis through modulation of SMAD-mediated pathways (83). Other transcriptional regulators including myocardin-related transcription factor (MRTF) have been shown to be also involved in EndMT (64, 149, 212, 276, 379), but their precise role and the detailed molecular mechanisms for their involvement require further study.

A) REGULATION OF EndMT BY NONCODING RNAs.

Recent high-throughput sequencing of the human genome has revealed that only a minor proportion of transcribed genes (<2%) are actually protein-coding genes, and it has become apparent that the great majority of the human transcriptome can be referred to as noncoding RNA. According to their size, noncoding RNAs are subdivided into small noncoding RNAs (<200 nt) and long noncoding RNAs (lncRNAs; >200 nt). One important class of small noncoding RNAs are the miRNAs characterized by a nucleotide length of only 22 nucleotides. Following the discovery of noncoding RNAs, there has been an explosion in the number of studies examining their role in the regulation of normal and pathological cellular functions, and on their effects on gene expression, epigenetic pathways, and cell signaling, as well as their alterations in a large spectrum of pathological conditions.

I) miRNAs. miRNAs modulate the expression of protein coding genes at the epigenetic level targeting a large number of translated mRNAs and, therefore, resulting in the posttranscriptional modulation of numerous key molecular pathways. Of particular relevance to the regulation of TGF-β-induced EndMT are the observations that miRNAs can affect the expression of mRNA transcripts coded by genes involved in the regulation of multiple TGF-β pathways including Smad signaling proteins, and can even modulate TGF-β mRNA transcripts. Given the remarkable widespread and ubiquitous regulatory effects of miRNAs, it is not surprising that numerous recent studies have described the stimulatory or inhibitory effects of various miRNA on the EndMT process as reviewed recently (150). Indeed, the exploration of miRNA contribution to EndMT is rapidly expanding, and these studies will yield valuable information both to unravel the complex molecular mechanisms involved, as well as to allow the identification of potential therapeutic targets for numerous disorders in which EndMT plays an important pathogenetic role. The large number of studies that examined the regulatory role of various miRNAs on the induction of EndMT under various experimental conditions are listed in TABLE 3. These results allowed the identification of elevated levels of numerous miRNAs including miR-125b during the development of experimentally induced EndMT in vitro (109), the demonstration of potent inhibitory effects of miR-126 on EndMT of endothelial progenitor cells (369), the demonstration that TGF-β-induced EndMT is partially mediated through miR-21 (120, 160), and the description that miR-155 is a potent inhibitor of TGF-β induced EndMT, an effect mediated through the modulation of RhoA signaling (26). It was also recently described that miR-31 is a positive modulator of TGF-β-induced EndMT and that it was required for the expression of α-SMA and other EndMT molecular markers following TGF-β-treatment (149). The role of the miR-200 miRNA cluster comprised of miR-429, miR-141, miR-200c, miR-200b, and miR-200a on EndMT regulation was investigated recently. Although it was previously reported that this miRNA cluster was able to inhibit epithelial to mesenchymal transition, it was not known whether they also were involved in EndMT. The role of miR-200a in TGF-β1-induced EndMT its effects on EndMT was examined in human aortic endothelial cells, and it was shown that its overexpression blocked EndMT in these cells and that these effects were mediated by regulation of expression of the gene encoding growth factor receptor bound protein 2 (GRB2) through direct binding to its 3′ untranslated region. It was also shown that miR-200a expression was suppressed during the EndMT process induced by GRB2 upregulation, that EndMT was promoted by ectopic expression of GRB2, and that cotransfection of miR-200a with an expression construct for GRB2 caused partial reversal of EndMT (366). The expression of miR-18a-5p and its functional role in the regulation of EndMT was assessed in human aortic valvular endothelial cells cultured with high glucose levels, which caused a reduction of miR-18a-5p and concomitantly induced Notch2 expression, which promoted EndMT. Furthermore, Notch2 was identified as a target of miR-18a-5p and miR-18a-5p overexpression downregulated Notch2 expression and suppressed EndMT. These findings demonstrated an important role of miR-18a-5p/Notch2 signaling pathway in the regulation of high-glucose-induced EndMT (108). More recently, the role of miR-148b on EndMT was examined in HUVECs (218). The results showed that miR-148b inhibition promotes EndMT in vitro and that these effects are mediated through downregulation of TGFB2 and SMAD2 gene expression. These observations were confirmed in an in vivo model of wound healing showing that miR-148b inhibition resulted in pronounced EndMT alterations in the wound tissues associated with impaired wound closure (218). The role of other miRNAs in the modulation of EndMT in newborn rat pulmonary microvascular endothelial cells cultured under hypoxic conditions as an in vitro model of the clinical disorder known as persistent pulmonary hypertension of the newborn were investigated. The results showed that miR-126a-5p expression was increased when the cells were exposed to hypoxia and that endothelial cell specific protein expression decreased whereas α-SMA increased in the hypoxic cells. Knockdown of miR-126a-5p in the cultured cells followed by their exposure to hypoxic conditions resulted in abrogation of the EndMT molecular alterations, and it was further shown that these effects were mediated by changes in PI3K levels and downstream phosphorylation of AKT (348).

Table 3.

Regulation of EndMT by microRNA and circular RNA

RNA	Gene Target	Effect on EndMT	Reference Nos.
MicroRNAs
miR-Let7	TGFBR1	Inhibition	49, 231
miR18a-5p	Notch2	Inhibition	108
miR20a	TGFBR1,TGFBR2, and SARA	Inhibition	69
miR21	PTEN	Stimulation	120, 160, 161
miR23	Has2	Inhibition	162
miR27b	Elk1, Neutropilin2, Plexin A2, Plexin D1	Stimulation	295
miR29	DPP-4	Inhibition	145
miR31	VAV3	Stimulation	149
miR125b	p53	Stimulation	109
miR126	PIK3R2	Inhibition	369
miR126a-5p	Unknown	Stimulation	348
miR130a	BMPR2	Stimulation	172
miR145	TGFBR2/SMAD3	Inhibition	338
miR148b	TGFB	Inhibition	218
miR155	RhoA	Inhibition	26, 319
miR200a	GRB2	Inhibition	366
miR200b	P300/Smad2, Snai1	Inhibition	40, 94
miR302c	Metadherin (MTDH)	Inhibition	378
miR483	CTGF	Inhibition	127
miR532	prss23	Inhibition	22
miR630	Slug	Inhibition	294
Circular RNAs
circHECW2	miR30d/ATG5	Stimulation	351
circDLGAP4	miR143/HECTD1	Stimulation	19
circHECTD1	HECTD1	Stimulation	93

II) Circular RNAs. Circular RNAs (circRNAs) are a recently identified novel class of noncoding RNAs generated by back-splicing events within genes in which exons rather than being spliced in a linear fashion become circularized. Extensive recent studies have shown that circRNA play an important role in the regulation of gene expression through a variety of mechanisms, including their ability to bind specific miRNAs or groups of miRNAs, sequestering them and suppressing their function, a process known as competitive endogenous RNA regulation. CircRNA also may serve as sponges for RNA binding proteins and may function as transcriptional regulators. Owing to their important effects in gene expression regulation and in maintenance of normal cellular homeostasis, circRNA are now beginning to be recognized as important regulators of multiple cellular functions. Indeed, alterations of their expression have been reported in association with various human diseases including malignancies, heart disease, neurological disorders, diabetes, and atherosclerosis, although the precise mechanisms involved have not been fully elucidated. Despite the recent interest in circRNAs, their potential role in EndMT has not been fully elucidated. However, as shown in TABLE 3, it was recently demonstrated that circHECW2, one specific circRNA that is particularly abundant in brain cells, regulated EndMT by directly binding to miR-30d, a significantly downregulated miRNA that caused an increased expression of the autophagy gene ATG5 (351). These findings provide novel information regarding the regulatory role of ATG5 in the occurrence of EndMT in brain tissues. In a related study it was shown that the circular RNA DLGAP4 (circDLGAP4) functions as an endogenous inhibitor of miR-143 activity. It was further demonstrated that circDLGAP4 overexpression significantly inhibited EndMT through regulation of miR-143, and it was suggested that it might play a role in the regulation of endothelial cell functions and their recovery following ischemic brain damage (19). Other circRNAs involved in the induction of EndMT by TGF-β1 in rat coronary artery endothelial cells have been identified employing high-throughput RNA sequencing. The results identified 102 differentially expressed circRNAs in the TGF-β1 treated cells (66 upregulated and 36 downregulated) and indicated an important role of specific circRNAs in the TGF-β1-mediated development of EndMT in these cells (135). The possibility that circRNAs may play a role in the pathogenesis of pulmonary fibrosis in silicosis through induction of EndMT was recently investigated in lung tissue samples from Tie2-GFP mice exposed to SiO₂ and from patients with pulmonary fibrosis caused by silicosis (93). The results showed that in vitro and in vivo exposure to SiO₂ induced EndMT. A remarkable observation was that SiO₂ concomitantly increased circHECTD1 expression, a key regulator of the levels of the HECT domain ES ubiquitin ligase 1 (HECTD1), which, in turn, inhibited HECTD1 protein levels. SiO₂-induced EndMT phenotypic changes were reversed either by a siRNA targeting circHECTD1 or by induced overexpression of HECTD1. These results were confirmed in tissue samples from SiO₂-exposed mice and from patients with silicosis. Thus these studies conclusively demonstrated the involvement of the circHECTD1/HECTD1 pathway in the induction of EndMT by SiO₂ and suggested a novel and plausible mechanism for the development of tissue fibrosis in patients with pulmonary silicosis (93).

III) lncRNAs. lncRNAs are a recently discovered novel class of RNA transcripts longer than 200 nucleotides that lack protein coding potential. On a functional level, lncRNAs have been implicated in numerous complex regulatory processes through diverse mechanisms including gene regulation by titration of transcription factors, splicing alteration, sponging of miRNAs, and recruitment of chromatin modifying enzymes. Several recent studies have identified lncRNAs in the regulation and control of numerous endothelial cell functions including EndMT. An initial RNA deep sequencing-based screening revealed hypoxia as a powerful regulator of lncRNA expression and uncovered that besides previously annotated endothelial lncRNAs the 40-GATA6-AS lncRNA was robustly and consistently upregulated. Since hypoxia induces GATA6-AS, the role of GATA6-AS in the cellular mechanisms mediating EndMT was analyzed (233). To this end, GATA6-AS expression was silenced in HUVECs by transfecting a specific locked nucleic acid, and the effects of silencing were assessed in TGF-β2-induced EndMT assays. Whereas TGF-β2 treatment strongly induced EndMT, the silencing of GATA6-AS diminished these effects and prevented EndMT. Together, these data showed that GATA6-AS plays an important role in the regulation of EndMT (233).

VI. EndMT IN EMBRYOGENESIS AND DEVELOPMENT

The first demonstration of the occurrence of EndMT was made during a study of the process of endocardial cushion in mouse embryos (200, 201). These pioneering studies showed that during embryonic development when endocardial cells evolve to become cardiomyocytes in the atrioventricular canal, they initiate biosynthetic processes leading to the formation of cardiac cushion mesenchyme and cardiac valves. The morphological conversion of endothelial cells to mesenchymal cells was also described during aortic (15) and pulmonary (16) artery development in the chick embryo and that a similar process participated in vascular intimal thickening and in pulmonary vascular structural remodeling (12, 14). The molecular mechanisms involved in EndMT during embryonic organ development are quite complex as discussed in recent reviews (89, 111); however, it has been recently shown that insulin-like growth factor II promoted EndMT throughout the interaction with mannose-6-phosphate (M6P) and urokinase-type plasminogen activator receptor (uPAR) and its ligand uPA (13) and also implicated NFκB in this transition in embryonic vascular remodeling (10). Another mechanistic pathway involving BMPR type I-induced endocardial EndMT has shown that BMPR is required for AV valve formation (146) and that this process is controlled by several transcription factors including SNAI1 and is mediated by NOTCH signaling (335). Furthermore, it was recently shown that Hand2 is a key regulator of EndMT controlling upstream Snai1 regulation in the development of the AVC cardiac cushions (165). The EndMT process has been implicated in the embryonic development of numerous additional vascular tissues, including the abdominal aorta (79), and cardiac and semilunar valve development (155, 220, 314). It has also been shown that the angiogenic process in the early postnatal retina is mediated by EndMT, and this transition is regulated by VEGF (70).

VII. EndMT IN PATHOPHYSIOLOGY

A. In Vitro Studies of EndMT in Human Endothelial Cells

Extensive investigations have examined the occurrence of EndMT in vitro utilizing human endothelial cells of various organs. The pioneering observations of Zeisberg et al. (363) showed that human coronary artery endothelial cells when passaged in vitro and cultured with TGF-β1 became spindly shaped and lost their endothelial cell molecular markers, simultaneously acquiring the ability to produce various fibroblast-specific macromolecules including vimentin and type I procollagen. Subsequently, the occurrence of EndMT in cultured human endothelial cells of various types was demonstrated, and it was shown that TGF-β1 was a potent in vitro inducer of EndMT in cultured human coronary artery endothelial cells and that the addition of apolipoprotein A1 (ApoA1) to the TGF-β1-treated cells abrogated EndMT development associated with inhibition of Smad2/3 phosphorylation and a marked decrease in the expression of Snail and Slug transcription factors (95). Similar studies were performed in TGF-β1-treated cultured human dermal microvascular endothelial cells (152). More recently, human microvascular endothelial cells isolated from small intestine mucosal layers transitioned into mesenchymal cells in vitro following exposure to a combination of proinflammatory cytokines including TGF-β, TNF-α, and IL-1β (266). The effects of IFN-α and -γ on human dermal microvascular endothelial cells were also examined recently, and it was demonstrated that IFN-γ induced morphological and molecular/gene expression changes in these cells consistent with their transdifferentiation into profibrotic myofibroblasts (61). The role of Fli1, and of several specific TLR ligands, including Poly(I:C) and LPS, in the induction of EndMT in cultured normal human dermal microvascular endothelial cells was also investigated, and the results demonstrated that Fli1 reduction induced EndMT and that LPS and Poly(I:C) also induced EndMT in these cells, although not with the same power as that of Fli1 reduction (290).

More recently, the mechanisms of radiation-induced EndMT were examined in human pulmonary artery endothelial cells in vitro. The results showed induction of EndMT in a radiation dose and length of exposure-dependent process and that the effects were mediated by the canonical SMAD signaling cascade triggered by TGF-β receptor activation. The same study described the occurrence of EndMT changes in these cells following hypoxic exposure and that this transition resulted in increased Smad3 phosphorylation, and increased expression of TGF-β receptor I and Snail1 (58). In a related investigation it was shown that radiation treatment of HUVECs in vitro was capable of inducing EndMT by increasing the expression of Snail and vimentin and reducing the expression of CD31 in these cells. Further investigation in a coculture system indicated that HUVECs irradiation induced the ability to cause paracrine differentiation of fibroblasts into activated myofibroblasts, effects mediated through the Snail/miR-199a-5p axis and therefore induce tissue fibrosis (355).

B. EndMT in Animal Models

The occurrence of EndMT in animal models was first described during the development of experimentally induced cardiac fibrosis in transgenic mice employing endothelial and mesenchymal cell-specific lineage tracing (363). These studies conclusively demonstrated the emergence of fibroblasts originating from endothelial cells and showed that from 27 to 35% of fibroblasts present in the fibrotic myocardium of mice with aortic banding originated through EndMT. Furthermore, it was found that TGF-β1 induced endothelial cells to undergo EndMT since there was a significant reduction in the relative proportion of EndMT-derived mesenchymal cells in aortic banded Smad3+/− transgenic mice in which the TGF-β response is blunted owing to the deficiency of Smad3 (363). Extensive subsequent investigations have demonstrated the occurrence of EndMT in animal models of various human diseases and have explored the molecular mechanisms involved. In experimentally induced cardiac fibrosis it was emphasized that TGF-β plays an important role in mediating EndMT (116) and that EndMT represents the most important trigger of tissue fibrosis acting as a profibrotic switch in cardiac fibrosis as well as in other fibrotic disorders. In other studies, it was shown that endothelial cell-derived ET-1 promotes EndMT and cardiac fibrosis in the setting of experimentally induced diabetes mellitus and that elevated ET-1 levels promoted accumulation of tissue fibroblasts through EndMT, an effect that was abolished in hearts from transgenic mice with endothelial cell specific ET-1 deletion (330). A subsequent study demonstrated that EndMT played a role in the development of cardiac fibrosis in high-fat-fed apolipoprotein E knockout mice and that this process was substantially augmented following induction of a systemic inflammatory process employing parenteral administration of casein (189).

In other animal model studies, EndMT was also examined in three mouse models of chronic kidney disease: unilateral ureteral obstructive nephropathy, streptozotocin-induced diabetic nephropathy, and a model of Alport renal disease. The results showed that 30–50% of fibroblasts in the fibrotic kidneys were found to coexpress endothelial and mesenchymal/myofibroblast molecular markers indicative of EndMT (362). These results were confirmed by other investigators and indicated that EndMT was an important mechanism in the development of kidney fibrosis (153, 167) and that it also played an important role in vascular rarefaction following ischemia/reperfusion acute kidney injury (21). The development of EndMT was also examined in bleomycin-induced lung fibrosis in double-transgenic mice in which LacZ expression was used to mark endothelial cells and any cells originated from an endothelial cell lineage. Morphological evaluation of lungs showed severe pulmonary fibrosis with the areas of fibrotic involvement containing large numbers of LacZ-positive fibroblasts indicative of their endothelial origin. Quantitative analysis of endothelial cell-derived lung fibroblasts showed that ~16% of lung fibroblasts from bleomycin-treated mice were from endothelial cell origin (125). More recently, mice subjected to mechanical ventilation displayed marked pulmonary fibrotic alterations associated with the occurrence of EndMT, demonstrating that mechanical stretch of the lungs may promote EndMT and pulmonary fibrosis (187). Regarding liver fibrosis, the occurrence of EndMT during CCl₄-induced chronic liver injury in mice was investigated, and although it was conclusively demonstrated that EndMT phenotypic transition occurred employing extensive endothelial cell genetic lineage tracing, a quantitative assessment of the number of the liver endothelial cells displaying EndMT was found to be remarkably small (265).

C. EndMT in Human Diseases

Extensive recent cell-lineage analysis has indicated that EndMT may be an important mechanism in the pathogenesis of numerous human diseases including cancer, atherosclerosis, and other vascular diseases, as well as various fibrotic diseases such as SSc and pulmonary, cardiac, and kidney fibrosis. The role of EndMT in these disorders will be reviewed in the following sections.

1. Malignant diseases

It is well known that the fibroblasts present within the tumor tissues, known as cancer-associated fibroblasts (CAFs), and display unique properties and biological activities that facilitate tumor progression. CAFs deposit various extracellular matrix molecules and secrete paracrine factors including TGF-β and VEGF, capable of affecting other cells present in the tumors, as well as cells in the tissues surrounding the tumors. The origin of CAF has been extensively investigated, and it has been estimated that ~40% may originate by the EndMT process (191, 256, 361). However, the molecular mechanisms involved in EndMT in malignancies have not been fully elucidated, although it has been shown that TGF-β2 induces the mesenchymal transdifferentiation of human microvascular endothelial cells to CAF-like cells. Furthermore, TGF-β2 stimulation resulted in activation of MRTF and in its upregulation. Such interaction induced integrin-linked kinase (ILK) overexpression, and these observations suggested a novel ILK-MMP9-MRTF axis that appears to be critical for EndMT-mediated increase of CAF-like cells (64). The mechanisms involved in lymph node metastasis that were examined recently implicated Wnt signaling and EndMT molecular interactions in this process, demonstrating that Wnt5b modulates lymphatic vessel angiogenesis via induction of partial EndMT (320). It has also been demonstrated that osteopontin exerts a potent angiogenic effect in various human cancers resulting in migration and invasion of new vessels into the surrounding tissues (73) and that some of these effects involve osteopontin’s potent EndMT-inducing activity. It was subsequently shown that this process was mediated through integrin αVβ3-activated PI3K, resulting in increased HIF-1α protein levels. In turn, HIF-1α transactivates TCF12 gene expression and the increased TCF12 further represses VE-cadherin gene expression at the transcriptional level and, thus, facilitates EndMT (92). It has also been shown that different levels of TGF-β-induced EndMT in various tumors may be related to the plasticity and heterogeneity of the tumor endothelial cells with some cells expressing high-basal SMA and responding to TGF-β by generating α-SMA+ myofibroblast-like cells, whereas others have low-basal α-SMA and transition to a population of fibroblast-like cells expressing low levels of α-SMA or not expressing α-SMA at all (340).

An important mechanism in the progression of cancer is the metastatic cascade that is initiated when the primary tumor cells invade the extracellular matrix of the surrounding tissues and then detach and intravasate into the lumina of blood and lymphatic vessels. The extravasation of cancer cells from the vasculature to the surrounding tissues is a crucial previous step required for the establishment of metastasis. Recent evidence indicates that endothelial cells do not play a passive role during transendothelial migration of cancer cells, and it has been demonstrated that EndMT plays an important role in the local progression and extension of various malignancies as well as in the metastatic process. Indeed, it has been shown that EndMT is necessary for both intravasation and extravasation of metastatic tumor cells. The role of EndMT was demonstrated in extravasation of brain endothelial cells (158) and in intravasation of breast carcinoma cells into lymphatic endothelium (315). Metastatic extravasation through EndMT has been also proposed (105, 158). A recent review discussed extensive experimental evidence supporting the concept that EndMT is one of the most important mechanisms involved in cancer cell metastasis and that extravasation and, to some extent, intravasation of malignant cells require the active involvement of endothelial cells undergoing EndMT and that EndMT also may participate in the generation of signals that regulate the premetastatic niche influencing the metastatic capabilities of tumor cells (105).

The role of EndMT in the remarkable fibrotic response/desmoplasia characteristic of numerous malignancies has also been examined recently to determine whether mesenchymal cells generated through EndMT may participate in this serious complication that has been considered to play an important role in the poor response of some of these malignancies to chemotherapeutic interventions. The role of EndMT in malignancy-associated desmoplastic responses was examined employing confocal microscopy of human pancreatic tumors. The results revealed coexpression of both CD31 and various mesenchymal markers in a subpopulation of cells within the tumors, and the double-labeled cells were identified as endothelial cells in intermediate stages of EndMT (103). A related study examined an orthotropic model of pancreatic adenocarcinoma and described a crucial role of EndMT transition in tumor vasculogenesis. Furthermore, it was shown that EndMT and increased tumor vasculogenesis were mediated by the neural adhesion molecule L1, also known as CD171 (cluster of differentiation 171) and involved JAK-STAT signaling. These observations were validated employing human pancreatic cancer tissues demonstrating markedly increased expression of L1/CD171 in the microvasculature of the tumoral tissues (193). A related study showed that the angiogenic co-receptor neuropilin-1 (NRP-1) that is highly upregulated in pancreatic ductal adenocarcinoma can upregulate the profibrotic effects of TGF-β1 and be a mediator of EndMT and, at the same time, stimulate the fibrotic process that is a prominent characteristic of this type of cancer (206).

Primary myelofibrosis (PMF) is a myeloproliferative disease characterized by abnormal proliferation of myeloid cell lineages accompanied by excessive production and accumulation of fibrotic tissue macromolecules in the bone marrow stroma. The mechanisms involved in the marked fibrotic response in this disease have not been fully elucidated. However, it has been shown that up to 30% of the endothelial cells of the microcirculation of the bone marrow and spleen of PMF patients display molecular evidence of EndMT. Furthermore, it was shown that EndMT was induced by PMF megakaryocytes constitutively producing and releasing high levels of bioactive TGF-β and that direct incubation of spleen endothelial cells with PMF megakaryocytes induced EndMT in vitro, a transition that was abrogated by specific TGF-β inhibitors (90). The acquisition of these mesenchymal characteristics by endothelial cells is an early event that appears to be induced by a profibrotic and inflammatory microenvironment and that the EndMT switch of the bone marrow and spleen endothelial cells in PMF might contribute to the overall development of the fibrotic process. Confirmation of these observations was obtained in vivo employing a murine model of PMF engineered to express a mutation in the JAK 2 gene specifically in the hematopoietic compartment. These mice developed a myeloproliferative-like phenotype with bone marrow and splenic fibrosis, and immunofluorescence analysis of the tissues showed that bone marrow and spleen endothelial cells displayed coexpression of VE-cadherin with mesenchymal cell phenotype markers, indicating that they were undergoing EndMT phenotypic conversion (90).

The participation of EndMT in esophageal adenocarcinoma was investigated, and it was shown to be induced by esophageal adenocarcinoma tumor cells cocultured with primary human esophageal microvascular endothelial cells. The results showed typical EndMT changes, and mechanistic studies demonstrated that IL-1β and TGF-β2 secreted by the esophageal cancer cells were responsible for the phenotypic transition of the endothelial cells into mesenchymal cells (235).

The role of miRNAs in regulating EndMT in malignant disorders has generated intense interest. One study described a potent inhibition of EndMT by miR-302c in hepatocellular carcinoma (378) resulting in inhibition of tumor growth. Other studies showed that miR-145, a tumor suppressor that has been reported to inhibit SMAD3-mediated EMT of cancer cells, also contributes to EndMT regulation in neointimal hyperplasia following cell therapy and that metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) modulates TGF-β1-induced EndMT through regulation of TGFBR2 and SMAD3 via miR-145 (338).

As discussed in a previous section, shear stress is a potent EndMT inducer; thus mechanical forces within the tumor tissue are very likely important to metastasis and effectivity of treatment. Indeed, it has been demonstrated that EndMT induced by shear stress exerted marked effects on breast cancer cell proliferation and migration (215).

2. Human fibrotic disorders

Fibrotic diseases encompass a wide spectrum of entities characterized by the progressive and uncontrolled accumulation of exaggerated amounts of fibrotic tissue in various organs. The resulting fibrosis disrupts the normal architecture of the affected organs, leading to their progressive dysfunction and eventually to their functional failure (269, 270). The cells responsible for the production and assembly of fibrotic tissues are the activated fibroblasts or myofibroblasts (131, 270). There is substantial controversy and extensive discussion in the literature regarding the origin of the activated fibroblasts that are ultimately responsible for the exaggerated production and accumulation of fibrotic tissue macromolecules in the fibrotic diseases. Furthermore, the origin of myofibroblasts may vary depending on the organ affected and the particular fibrotic disorder, although several potential sources have been recognized. These include the proliferation and activation of quiescent tissue-resident fibroblasts, the recruitment of fibroblast precursor cells from the bone marrow (also known as fibrocytes), and the phenotypic conversion of various differentiated cell types including pericytes, adipocytes, and epithelial, mesothelial, and endothelial cells into cells with a mesenchymal phenotype that subsequently can acquire an activated myofibroblast phenotype (270). Besides the phenotypic transdifferentiation of these cells, it has been recently demonstrated that certain progenitor mesenchymal stemlike cells with preferential perivascular localization may also generate myofibroblasts in certain fibrotic diseases. However, it is currently generally accepted that tissue-resident fibroblasts are the major source of activated myofibroblasts in the fibrotic disorders, although epithelial to mesenchymal transition (EMT), endothelial to mesenchymal transition (EndMT), or pericyte to myofibroblast transition may play a role in specific diseases and under specific conditions and cellular contexts (1, 156, 188, 270).

The molecular mechanisms involved in the development of pathological tissue fibrosis are quite complex and have not been fully elucidated; however, as discussed above, it has been recently demonstrated that a potential source of activated profibrotic myofibroblasts in the fibrotic diseases may be endothelial cells that have acquired a profibrotic phenotype through EndMT. This hypothesis, first proposed by Karasek (148), has been pursued intensively. In the subsequent sections, the extensive recent literature examining the role of EndMT in the pathogenesis and development of human fibrotic diseases will be summarized.

A) SYSTEMIC SCLEROSIS.

SSc is the prototypic systemic fibrosing disorder. It is characterized by a usually progressive fibrotic process involving the skin and multiple internal organs including the lungs, heart, and gastrointestinal tract (78, 101). Although the mechanisms responsible for the exaggerated fibrotic process in SSc are quite complex and have not been fully established, recent interest has been focused on the role of EndMT in this process (143, 199, 224, 234, 252, 254). The possibility that EndMT is involved in the fibrotic process of SSc-associated pulmonary fibrosis was examined employing immunohistology of SSc lung tissues. The results showed expression of the endothelial cells marker CD31 in mesenchymal cells embedded within the neointima of small pulmonary arteries as well as in the parenchymal fibrotic areas (211). Furthermore, coexpression of CD31 or vWF with the mesenchymal markers, collagen type I, or α-SMA was demonstrated in numerous endothelial cells lining the small and medium-sized pulmonary arteries, indicating that mesenchymal cells of endothelial origin may be responsible for the vascular fibrosis and obliteration. These structural alterations were associated with marked differences in gene expression patterns between microvascular endothelial cells isolated from normal lungs compared with microvascular endothelial cells isolated from lungs from patients with SSc-associated interstitial lung disease. The expression of mesenchymal genes and EndMT-related transcription factor genes such as SNAI2 and TWIST was substantially increased in the endothelial cells isolated from the lungs of patients with SSc-associated interstitial lung disease providing very strong evidence for an important role for EndMT in the development of the fibrotic and vascular alterations in SSc (211).

Numerous observations have provided support to the concept that EndMT plays a role in the prominent vascular and fibrotic alterations characteristic of SSc. Among these, the induction of typical EndMT molecular alterations in CD31+ dermal microvascular endothelial cells isolated from skin biopsies of patients with SSc and from gender- and age-matched normal individuals following exposure in vitro to TGF-β1 plus ET-1 was demonstrated (63). Although these results support the notion that EndMT may participate in the development of SSc-associated cutaneous fibrosis, the occurrence of a phenotypical change from endothelial cells to activated myofibroblasts was not demonstrated in the affected skin of patients with SSc. However, an extensive recent immunohistological analysis of affected SSc skin demonstrated the presence of numerous endothelial cells simultaneously expressing endothelial cell-specific and myofibroblast markers in SSc dermal microvasculature. The in vitro phenotype of dermal microvascular endothelial cells isolated from SSc skin displayed an intermediate EndMT phenotypic state with coexpression of molecular markers of endothelial and mesenchymal cells (199). The presence of transitional EndMT cells in dermal vessels was validated in two murine models of SSc. Thus these results yielded direct evidence that EndMT is a process occurring in the dermal endothelium of patients with SSc and provided a crucial link between endothelial cell dysfunction, the development of EndMT, and the resulting pathological tissue fibrosis (199). Strong support for the role of EndMT in the development of tissue fibrosis was provided by studies employing an in vitro model of EndMT of relevance to SSc pathogenesis. In this model, CD31+ immunopurified microvascular endothelial cells isolated from unaffected skin from patients with SSc were cocultured with affected SSc dermal fibroblasts isolated from skin that had been previously incubated with ET-1 and TGF-β. The coculture resulted in the reduction of the ability of the endothelial cells to form tubular structures and in the increased expression of numerous genes associated with EndMT. The study further demonstrated that ET-1 inhibition with bosentan and macitentan effectively inhibited the ability of ET-1/TGF-β-treated SSc fibroblasts to induce EndMT. Thus these results provided strong evidence indicating that dermal fibroblasts from SSc skin that have been exposed in the SSc patient’s tissues to ET-1 and TGF-β may be responsible for the subsequent transdifferentiation of the endothelial cells toward the mesenchymal phenotype and, therefore, may lead to the establishment and further progression of the fibrotic process (68).

B) ORGAN-SPECIFIC FIBROTIC DISEASES.

Despite the clinical importance and the demonstration of morphological and molecular alterations indicative of EndMT during experimentally induced pulmonary fibrosis (125) and the demonstration of EndMT changes in lung tissues from patients with SSc-associated interstitial lung disease (211), there are very few reports describing EndMT in IPF or in secondary pulmonary fibrotic diseases. As reviewed recently most studies on the pathogenesis of these diseases have focused on the phenotypic conversion of epithelial cells to activated profibrotic myofibroblasts, a process known as epithelial to mesenchymal transition or EMT (277); however, the occurrence of EndMT in the pulmonary arterioles of lung tissues from patients with IPF and IPF associated with PAH was recently demonstrated as indicated by the observations that endothelial cells in the lung fibrotic tissue pulmonary arteries displayed characteristic immunohistological evidence of EndMT (214).

The occurrence of EndMT in human tissues from patients with cardiac fibrosis has been described in several publications (46, 346, 358). Immunofluorescence and gene expression analyses of myocardial tissues from patients with cardiac fibrosis associated with chronic kidney disease demonstrated that ~17% of fibroblasts/myofibroblasts present in the fibrotic myocardium from patients with stage 3–7 chronic kidney disease were EndMT derived (46). The expression of various EndMT-associated genes was assessed in left ventricular myocardial tissues from patients undergoing cardiac transplantation for end-stage cardiac failure, and a remarkable increase from 5- to 10-fold in the mRNA levels for Snail, Slug, and Twist was demonstrated (346). Regarding kidney fibrosis, the presence of glomerular cells expressing EC and mesenchymal/myofibroblast cell-specific molecules was examined in kidney biopsy tissues obtained from patients with diabetic kidney disease, and numerous glomerular cells coexpressing α-SMA and CD31 were identified in the tissues providing strong evidence for the occurrence of EndMT in vivo (171). The contribution of EndMT to the development of renal interstitial fibrosis was also examined in renal biopsy tissues from kidney transplant recipients who developed chronic allograft dysfunction and kidney fibrosis (323, 324). These and other results collectively confirm the occurrence of EndMT cellular alterations during the development and progression of various fibrotic kidney diseases (71, 77, 126, 323, 324). Regarding liver fibrosis, there are studies that examined the occurrence of EndMT in liver fibrotic diseases; however, a recent investigation of human cirrhotic livers from patients undergoing liver transplantation found conclusive evidence for the transition of endothelial to mesenchymal cells, although the proportion of cells displaying the EndMT phenotype was very small (265).

C) RADIATION-INDUCED TISSUE FIBROSIS.

Tissue fibrotic reactions represent some of the most serious complications of radiation therapy, and intense investigation is being conducted to understand the pathogenesis and molecular mechanisms involved in the devastating organ fibrosis and dysfunction induced by radiation therapy and to investigate the potential role of EndMT in this fibrotic process. The occurrence of EndMT in radiation-induced pulmonary fibrosis was examined using immunofluorescence analysis, and extensive colocalization of α-SMA and CD31 markers accompanied by exaggerated collagen accumulation in the lung parenchyma was demonstrated during the development of the fibrotic process. These observations indicated that radiation-induced EndMT was an important mechanism involved in the development of the severe and progressive fibrotic process occurring in the irradiated lungs these extensive (58). The occurrence of EndMT in the course of tissue fibrotic reactions induced by radiation therapy was further documented in a recent histopathological analysis of rectal tissues from patients who received radiation therapy for the treatment of rectal malignancies, observations that were confirmed in a murine model of radiation-induced proctitis and rectal fibrosis (217). Mechanistic studies in this animal model demonstrated that radiation-induced EndMT was mediated by the Notch pathway through the Hey 2 transcription factor (216).

3. Vascular diseases

There has been an intense interest in the role of EndMT in the pathogenesis of numerous vascular diseases including pulmonary arterial hypertension, atherosclerosis, diabetic vasculopathy, and several other cardiovascular disorders such as mitral valve prolapse (65, 112, 173, 287, 288). Numerous publications have also described the occurrence of EndMT in related pathological conditions including vascular remodeling leading to vascular occlusion and neointima formation in human vein graft tissues (66), intestinal fibrosis associated with inflammatory bowel diseases (266, 272), and idiopathic portal vein hypertension (152, 275).

A) PULMONARY ARTERIAL HYPERTENSION.

The EndMT process has been implicated in primary and secondary PAH. The earliest studies revealed the presence of intimal thickenings in the pulmonary arterial wall, and it was suggested that EndMT contributed to the pathological alterations in chronic PAH (11). It was subsequently shown employing immunohistochemical analysis of endothelial cell-specific proteins that the concentric subintimal layers in pulmonary arteries associated with plexogenic vessels in PAH were most likely produced by endothelial cells undergoing EndMT, resulting in intimal layer remodeling and a reduced vascular lumen (309). Additional recent investigations have also implicated EndMT in primary PAH and in PAH secondary to SSc. In the primary PAH study employing transmission electron microscopy and correlative light and electron microscopy, unequivocal ultrastructural-level evidence of ongoing EndMT in lung tissue samples from patients with primary PAH was obtained. In this study it was shown that typical endothelial cells, identified by the presence of Weibel-Palade bodies, acquired expression of the myofibroblast-specific marker α-SMA and displayed invaginations into the neointima of the abnormal pulmonary arterioles (260). Similar results were obtained in an extensive examination of the cellular phenotype of intimal and plexiform lesions from SSc-PAH lungs showing the unambiguous coexpression of endothelial and mesenchymal markers in up to 4% of pulmonary arterioles (113). A more recent analysis of human lung tissues from patients with PAH associated with IPF who underwent lung transplantation in comparison with lung explant samples from healthy controls confirmed the occurrence of EndMT in the PAH tissues. The results showed that pulmonary arteries from healthy controls displayed strong positive immunostaining for CD31 in the endothelial layer, whereas pulmonary arteries from PAH/IPF patients showed strong coimmunostaining of CD31 and mesenchymal cell markers in endothelial cells, as well as in the vessel intima, suggesting that EndMT was involved in pulmonary arterial remodeling. The results further indicated that these alterations were mediated by the increased levels of JAK2/p-JAK2 (214). Collectively, these novel observations provide conclusive evidence for the occurrence of EndMT in small and medium-sized arterioles of lung tissues from patients with primary PAH and in PAH associated with IPF or PAH in SSc and offer a promising therapeutic target for PAH as discussed recently (261, 291, 342).

B) ATHEROSCLEROSIS.

Although numerous pathophysiological investigations have examined the genetic, metabolic, and biochemical mechanisms involved in the complex pathogenesis of the atherosclerotic process and extensive knowledge of the molecular alterations involved has been acquired, the possible contribution of EndMT was not considered until quite recently. However, a large number of recent studies have demonstrated that EndMT may be a major mechanism in the initiation and progression of atherosclerosis and may be a crucial determinant of the formation of the calcified plaque deposits and also play a role in plaque instability leading to plaque rupture and in the subsequent release of calcified atherosclerotic nodules into the circulation (36, 48, 91, 166, 273, 329). The occurrence of EndMT has been demonstrated in various animal models of genetically determined or induced atherosclerotic lesions as well as in arterial tissues from affected atherosclerosis patients (36, 48, 91, 166, 273, 329). An extensive study examined the molecular mechanisms involving EndMT focusing on the role of FGF and demonstrated that FGFR1 is a key factor in the progression of the disease and EndMT provides the link between the functional and molecular effects of FGF, local tissue inflammation, and disturbed shear stress that leads to the activation of TGF-β signaling (48). The possibility that apolipoprotein A1 (ApoA1) may abrogate the activation of the EndMT process induced by TGF-β was examined in vitro employing human coronary artery endothelial cells, and it was shown that ApoA1 reduced the expression levels of the key transcriptional activators mediating EndMT, Snail and Slug, and caused a potent inhibition of EndMT in these cells (95). Another mechanistic investigation examined the role of the CLOCK gene in the development of EndMT in HUVECs cultured in high glucose and in carotid artery endothelium of patients with carotid artery stenosis (302). The results uncovered a unique mechanism by which CLOCK suppresses EndMT reducing atherosclerotic development, and a strong positive relationship between loss of CLOCK expression, development of EndMT, and atherosclerotic plaque vulnerability was demonstrated. These observations indicate that CLOCK has a distinct function in atherosclerotic progression and that loss of CLOCK induced by high glucose is a specific mechanism to induce EndMT and thus worsen the development of atherosclerosis (302). An additional observation was that inhibition of ROCK1 significantly attenuated EndMT induced by loss of CLOCK expression (302). These latter results are in agreement with previous evidence indicating that ROCK1 activation induced by high glucose mediates EndMT in kidney glomerular cells and aggravates albuminuria in diabetic nephropathy (246). In other studies related to the role of lipoproteins in atherosclerosis, it has been shown that radiation-induced atherosclerosis may be mediated through EndMT and that this process can be accelerated by oxidized low-density lipoprotein (151).

The role of shear stress in atherosclerotic lesion development and stability has also been examined extensively. Shear stress has been shown to modulate neointimal hyperplasia and atherogenesis, effects mediated by induction of EndMT, and the molecular mechanism involved in this modulation is dependent on the ERK5 pathway (223). In a related investigation it has also been shown that shear stress from blood flow markedly influences the pathological alterations of atherosclerosis and that some of these effects are likely mediated by induction of Snail, a regulatory transcription factor crucial for EndMT, that was found to be highly expressed in atherosclerotic plaques (196). Recent investigations showed that mechanosensitive miRNAs are involved in the transduction of flow-mediated effects and in the induction of endothelial cell dysfunction and atherosclerosis development and these miRNA have been designated as “mechanosensitive athero-miRs,” and it has been suggested that they may also exert their effects through EndMT, although experimental evidence for this novel concept is not currently available (159).

C) VASCULAR CALCIFICATIONS.

Vascular wall calcifications are a common feature of atherosclerotic vascular disease and diabetic vasculopathy, and it has recently been suggested that EndMT contributes to their development (36, 119, 134). Mechanistic studies have demonstrated that endothelial cells transition to cells capable of undergoing osteochondrogenesis through EndMT and that this sequence was associated with an increase in endothelial cell stem cell-like properties, suggesting that, rather than a direct transdifferentiation from endothelial to osteogenic cells, a certain level of endothelial cell dedifferentiation occurred before the osteochondrogenic differentiation (210). Subsequently, the transition of endothelial cells to chondrocytes was demonstrated, and it was shown that high glucose levels mediated this process, suggesting highly important mechanistic interactions that are likely to be involved in the development and/or progression of diabetic vasculopathy and vascular calcifications (303). Other mechanistic studies linking atherosclerotic vascular calcification with EndMT have shown that this process involves proteolytic activation of specific proteases including elastase and various kallikreins through a process mediated by Sox2 allowing through EndMT the emergence of multipotent endothelial cells capable of ultimately inducing vascular calcification (354).

D) PROLIFERATIVE DIABETIC RETINOPATHY.

Proliferative diabetic retinopathy (PDR) is a common and serious medical complication of diabetes mellitus. Although the detailed pathogenetic mechanisms have not been fully elucidated, recent investigations to characterize the mechanisms involved in PDR epiretinal membrane have explored the role of EndMT (1, 40, 45). It was demonstrated that endothelial cells in epiretinal fibrovascular membranes from PDR patients contribute to the emergence of activated myofibroblasts via the process of EndMT and that the endothelial cells undergoing EndMT further inhibit the spontaneous reappearance of new blood vessels in these patients (1). In another study, membranes from eyes undergoing vitrectomy for complicated PDR were examined employing confocal laser scanning microscopy for the distribution of cellular markers for fibroblasts and endothelial cells (40). The role of ET-1 and of its receptors was also examined. Diabetic membranes showed significantly higher ET-1 labeling and colocalization of ET-1 with FSP1/S100A4 and CD31. Numerous cells displaying coexpression of CD31 and FSP1/S100A4 protein were identified in PDR tissues, indicating that the vasculature within these membranes is undergoing a transition towards fibrosis, via EndMT (45). A related investigation found that SNAI1 was highly expressed by endothelial cells in pathological ocular neovascularization and demonstrated that SNAI1 overexpression triggered enhanced cell motility, whereas SNAI1 knockdown attenuated neovasculature development as well as vascular invasion and sprouting. Of remarkable importance were observations that intravitreal injection of SNAI1 small interfering RNA suppressed new vessel formation in developing retina as well as in a murine model of oxygen-induced retinopathy and choroidal neovascularization (293).

E) PORTAL VEIN FIBROSIS.

The possibility that EndMT of portal vein endothelium induced by TGF-β/Smad activation may be involved in portal vein fibrosis in idiopathic portal hypertension (IPH) was examined recently. The results showed enhanced expression of phosphorylated Smad2 in the endothelium of smaller portal veins in IPH that was associated with colocalization of endothelial cell and myofibroblast protein markers, and it was concluded that the conversion of portal vein endothelial cells into myofibroblasts may be responsible for exaggerated deposition of fibrous tissue proteins in periportal veins and may represent a crucial mechanism responsible for portal venous obliteration in IPH (152, 275).

F) CARDIAC VALVULAR DISEASE.

Calcific aortic valve disease (CAVD) is a common human cardiac valve disorder characterized by fibrotic thickening and extensive and frequently progressive calcification of the aortic valve leaflets leading to aortic valve stenosis often requiring surgical valve replacement. In this disorder, the aortic valve interstitial cells undergo complex molecular alterations resulting in their differentiation into activated myofibroblast-like and osteoblast-like osteogenic cells that are ultimately responsible for the valvular tissue calcification in CAVD. Furthermore, it has been demonstrated that the endothelial cell monolayer that covers the surface of the heart valve can undergo EndMT in adult valves (132, 242) and that mitral valve endothelial cells differentiate in vitro into mesenchymal cells including some with osteogenic potential (27, 336). Thus it has been suggested that EndMT may contribute to the replenishment of valvular mesenchymal interstitial cells and that under certain pathological conditions the endothelial cells covering the valves may also become osteogenic and capable of inducing valvular calcification (27, 336). In a related study, the possibility that EndMT may play a role in the mitral valve changes associated with mitral valve prolapse was examined. Very high levels of circulating osteoprotegerin in patients undergoing mitral valve replacement for mitral valve prolapse were identified. Subsequently, it was demonstrated that osteoprotegerin was capable of inducing EndMT in cultured endothelial cells isolated from the pathological mitral valves and, remarkably, it was shown that mitral valve endothelial cells undergoing EndMT phenotypic transition were capable of producing and secreting osteoprotegerin, thus creating an autocrine pathogenetic mechanism for EndMT (287).

G) IDIOPATHIC DILATED CARDIOMYOPATHY.

Dilated cardiomyopathy (DCM), the third most common cause of heart failure, is characterized histopathologically by moderate-to-severe myofibril and myocyte degeneration, interstitial vacuolization, and replacement of myocardial cells by tissue fibrosis. Although the pathogenesis of DCM is unknown, the possibility that EndMT contributed to the progression of cardiac fibrosis in this disease was recently examined in cardiac tissue samples from patients with DCM for colocalization of endothelial and mesenchymal markers and for the expression and protein levels of Wnt, β-catenin, and Snail. The results demonstrated that numerous cells displayed typical EndMT molecular changes and Wnt and Snail signaling were significantly increased in DCM samples, indicating that EndMT may participate in the pathogenesis of myocardial fibrosis and remodeling during the development of DCM (341).

H) CEREBRAL AND ORBITAL CAVERNOUS MALFORMATIONS AND CEREBRAL HEMANGIOBLASTOMAS.

Cerebral cavernous malformations (CCMs) are a unique congenital intracranial vascular malformation that can occur in a sporadic or autosomal dominant inherited form characterized by single or multiple clusters of enlarged capillary-like channels with a lumen covered by a single layer of flattened endothelial cells (41, 59). Although less frequent, similar malformations can occur in the eyes and are termed orbital cavernous malformations. The mechanism involved in the induction of vascular malformations by mutations in the genes involved in CCM and the role of EndMT in their development have been investigated recently. For this purpose, endothelial-specific tamoxifen-inducible loss-of-function of the Ccm1 gene (iCCM1) mice were generated (192), and it was shown that Ccm1 deletion resulted in the development of vascular lesions within the central nervous system quite similar to the spontaneous lesions of patients harboring CCM1 mutations. Extensive histopathological and molecular analysis demonstrated the occurrence of EndMT in the endothelial cells forming the vascular malformations. An important observation was that the acquisition of the mesenchymal cell phenotype occurred only in endothelial cells expressing the CCM1 loss-of-function mutation, thus confirming that endothelial cell-specific deletion of the CCM1 gene induced EndMT. It was further shown that EndMT in CCM1-ablated endothelial cells is mediated by the upregulation of TGF-β and BMP signaling pathways since inhibitors of these pathways prevented EndMT and reduced the number and size of vascular lesions in Ccm1-deficient mice (192). Recent microscopy imaging and immunohistochemistry analysis in CCM human tissues validated the observations from the animal studies and conclusively demonstrated that the endothelial cells involved in the lesions of patients with sporadic CCM had undergone EndMT (38, 164). A semiquantitative assessment of the occurrence of EndMT in human CCM and orbital CM indicated that the occurrence of EndMT was observed in ~70% of endothelial cells in CCMs and in ~35% of endothelial cells in orbital CMs (298). However, despite the demonstration that EndMT contributes to the pathogenesis of CCM, it remains unknown whether EndMT occurs in central nervous system hemangioblastomas, lesions that display substantial similarities to those of CCM. Thus recently the possibility that EndMT participates in the severe vascular abnormalities that are characteristic of cerebral hemangioblastomas was investigated in surgically removed cerebral hemangioblastomas. The results demonstrated that endothelial cells in essentially all microvessels in the lesions displayed EndMT with a mosaic-like staining pattern with positive and negative cells mixed in the endothelial layer of all microvessels. These results indicated that EndMT occurred in the microvessels of central nervous system hemangioblastomas and very likely this process contributed to the notorious vascular abnormalities that are characteristic of these tumors (299).

4. Inflammatory and infectious diseases

The first link between inflammation and EndMT was described in human microvascular endothelial cells treated with supernatants from activated peripheral blood mononuclear cells (221). Following such treatment, the endothelial cells underwent a marked phenotypic change to that of mesenchymal cells and augmented their ECM production, an effect that was shown to be induced by IL-1 present in the mononuclear cell supernatants (221). Following these groundbreaking studies, there has been intensive investigation into the pathogenetic role of EndMT in the clinical manifestations of various systemic and organ-specific inflammatory diseases (248). In contrast, the possibility that infectious agents may directly induce EndMT by infection of endothelial cells has not been explored extensively. One of the few studies that have examined this topic investigated the mechanisms of myocardial fibrosis resulting from viral myocarditis and tested the hypothesis that cardiac microvascular endothelial cells undergo EndMT following coxsackievirus-B3 (CVB3) infection (352). For this purpose, capillary microvascular endothelial cells isolated from rat cardiac ventricles were infected in vitro with CVB3 virus. The results showed that the CVB3-infected cells developed gene expression changes indicative of EndMT and that abrogation of the EndMT process could be induced pharmacologically resulting in a remarkable reduction of α-SMA expression levels (352). A related study demonstrated that Kaposi sarcoma herpes virus infection of endothelial cells can induce EndMT and that through this mechanism the invasiveness of the infected endothelial cells is enhanced, contributing to the malignant progression. This process was shown to be mediated by canonical Notch signaling and did not appear to involve the TGF-β signaling pathway (106). More recently, the role of EndMT in patients with Kawasaki disease was investigated. The results indicated that sera from these patients caused EndMT in HUVECs in vitro and that these effects were mediated by connective tissue growth factor (CTGF) and suggested that suppression of CTGF may be of benefit for the treatment of these patients (127).

5. Wound healing and tissue regeneration

Endothelial to mesenchymal transition has also been implicated in the development of numerous wound healing and repair disorders occurring in particular in response to musculoskeletal injury (2, 365). Interestingly, it has been demonstrated that endothelial cells that have undergone EndMT are capable of supporting increased transendothelial migration of immune and other inflammatory cells as well as malignant cells by regulating their adhesion to normal endothelium and promoting their migration into the subendothelial space by inducing increased endothelium permeability (256). On the other hand, there has been intense interest in the possibility that endothelial cells undergoing EndMT may also be able to become pluripotential cells with certain features of stem cells and that under specific conditions may be utilized for regenerative purposes, particularly for cartilage and bone repair (207, 208, 360).

6. Other disorders

A) CHRONIC OBSTRUCTIVE PULMONARY DISEASE.

The occurrence of EndMT has recently been demonstrated in the vasculature of human lung tissues obtained from explanted lungs of five patients with chronic obstructive pulmonary disease (COPD) (263, 283, 284). Similar findings were observed in lungs from mice with tobacco smoke-induced lung emphysema (exposure for 3 and 8 mo) showing significant upregulation of S100A4 mRNA and protein levels in intrapulmonary vessels of the COPD lung tissues and in small parenchymal vessels isolated employing laser-microdissection from mice after eight months of smoke exposure (263). Collectively, these observations indicate that substantial vascular remodeling that included the occurrence of EndMT was present in affected lung tissues but also occurred even in nonhypoxic tissues, thus suggesting a potentially important role for EndMT in the development of COPD-associated pathologic alterations (263, 283, 284).

B) FIBRODYSPLASIA OSSIFICANS PROGRESSIVE.

Fibrodysplasia ossificans progressive (FOP) is a congenital heterotopic ossification syndrome caused by gain-of-function mutations of the BMP type I receptor ACVR1. This unusual disorder manifests with progressive ossification of skeletal muscles, tendons, ligaments, and joints often triggered by injury or inflammation (147). Mice harboring an Acvr1^R206H knock-in allele recapitulate the phenotypic spectrum of FOP, including the development of injury-associated intramuscular ossification and spontaneous articular, tendon, and ligament ossification. Recent publications have suggested that this unusual heterotopic ossification process is mediated by certain populations of endothelial cells undergoing EndMT, resulting in the accumulation in affected tissues of mesenchymal cells with a chondrogenic phenotype that are responsible for the deposition of a bonelike calcified tissue matrix (208, 259).

VIII. INHIBITION OF EndMT

Given the large number of studies across multiple disciplines conclusively demonstrating the crucial participation of EndMT in the pathogenesis, development, and progression of a variety of common as well as rare human pathologies including numerous malignant, fibrotic, and vascular disorders, there has been great interest in the abrogation or modulation of EndMT as a novel therapeutic approach for these disorders (51, 140). Indeed, there has been intensive investigation focused on the identification of compounds, substances, chemicals, and pharmacological agents as inhibitors of EndMT and their use as potential therapeutic approaches for various diseases in which EndMT has been shown or suggested to play a role in their pathogenesis and/or in the resulting pathological alterations (listed in TABLE 4). The mechanisms postulated for these inhibitory effects are quite diverse and include the inhibition of various growth factors implicated in the initiation or progression of EndMT, the regulation of the activity or expression of various transcription factors and specific genes involved in EndMT, the modulation of epigenetic EndMT regulators, or the stimulation or inhibition of the numerous molecular pathways that play crucial regulatory roles in the cellular phenotypic transition of endothelial cells into mesenchymal cells.

Table 4.

Studies of pharmacological inhibitors of EndMT in vitro or in disease models

Molecular Targets	Drug	In Vitro or Disease Model	Reference Nos.
TGF-β pathways
Canonical	ApoA-1	In vitro	95
	Glycyrrhizin	Bleomycin SSc	349
mTOR	Rapamycin	Cardiac fibrosis	102, 380
	Geniposide	Bleomycin SSc	257
pERK	Losartan	Myocardial infarction	331, 336
Notch	Relaxin	Cardiac fibrosis/renal fibrosis	39, 374, 377
	Spironolactone	Cardiac fibrosis/in vitro	39, 52
	Scutellarin	Cardiac fibrosis	375
WNT	PEDF	Myocardial ischemia	367
DPP-4	Linagliptin	Diabetic kidney fibrosis	145
	Vildagliptin	Lung fibrosis	296
	Sitagliptin	Renal fibrosis	344
Endothelin-1	Bosentan/macitentan	In vitro	68
PTH	Cinacalcet	Renal fibrosis	332, 333
Endothelial protection
HMG-CoA reductase	Simvastatin	Myocardial fibrosis	310
	Lovastatin	Chronic kidney disease	190
Other pathways
PDGF	Imatinib	PAH	286
PPAR-γ	Evodiamine	Cardiac fibrosis	141, 334
	Puerarin	Cardiac fibrosis	144
	Pioglitazone	Cardiac fibrosis	326
	EP4 agonist	PAH	163
HDAC	Givinostat	Myocardial infarction	213
	Valpronic acid	In vitro	230
Nrf2	Salviaonolic acid	Pulmonary vascular remodeling	55

TGF, transforming growth factor; SSc, systemic sclerosis; PAH, pulmonary arterial hypertension; PDGF, platelet-derived growth factor; PPAR, peroxisome proliferator-activated receptor.

A. Endothelial Cell Protection

Several compounds are endowed with overall protective effects for endothelial cells and have been examined for potential inhibitors of EndMT alterations. Among these are the high-density lipoproteins (HDL), ApoA1, and the statins. The endothelial-protective effects of HDL are well established; however, only a few investigations have examined their effects on the development of EndMT. This concept was directly examined in cultured human aortic endothelial cells, and the results provided the first in vitro evidence that HDL reduced TGF-β1-induced EndMT (289). Subsequently, it was shown ApoA1 also is capable of abrogating TGF-β1-induced endothelial dysfunction, providing a protective effect for TGF-β1-induced EndMT in cultured human coronary artery endothelial cells causing elevated expression of endothelial cell-specific phenotypic markers and a marked reduction of expression levels of Slug and Snail, crucial transcriptional factors of EndMT (95). Owing to the extensive evidence describing numerous beneficial effects of statins independent from their effects of lipid metabolism, it has been of substantial interest to examine whether the statins may also have effects on the EndMT process (190, 310). One such study showed that simvastatin inhibited TGF-β1-induced human cardiac microvascular endothelial cell transition into a mesenchymal cell phenotype and prevented endothelial cell filopodia formation and cell detachment in a concentration-dependent effect, and it was suggested that inhibition of EndMT may directly contribute to the antifibrotic effects of simvastatin (310). Similar results were obtained in in vitro and in vivo investigations of the effects of lovastatin on EndMT. The results demonstrated that lovastatin caused potent inhibition of EndMT in glomerular endothelial cells and that these events were mediated by suppression of oxidative stress (190).

B. Inhibition of TGF-β and Other Pathways

Given the crucial role of TGF-β in EndMT induction, there has been intense interest in the inhibition of TGF-β and of related growth factors as an effective approach to prevent or revert the development of EndMT. Indeed, it was recently demonstrated that abrogation of endothelial cell-specific TGF-β signaling in genetically modified mice with endothelium-specific heterozygous TGF-β receptor knockout resulted in markedly reduced EndMT and kidney fibrosis in two animal models of chronic kidney disease (337). Therapeutic inhibition of TGF-β1 in human clinical trials, however, has been accompanied by numerous side effects and undesirable complications owing to the potent pleotropic effects of this family of proteins in a large number of crucial cellular functions and in a large spectrum of important regulatory pathways. Given these serious limitations, there has been extensive research examining the possible effects of inhibition of downstream pathway components or of selected TGF-β-associated molecules on EndMT as described in greater detail in previous sections and shown in TABLE 4.

Among the most relevant compounds capable of inhibiting EndMT by modifying various EndMT-inducing pathways are Notch pathway inhibitors including relaxin, spironolactone, and scutellarin (39, 52, 374–377); macitentan, a newly approved inhibitor of endothelin-1 (63); and statins, in particular simvastatin and lovastatin, that are able to ameliorate fibrosis in an effect mediated by EndMT inhibition (190, 310). Rapamycin and geniposide, autophagy enhancers, also attenuated EndMT through mTOR signaling pathway, a mechanism that was demonstrated in vitro using HUVECs and cardiac microvascular endothelial cells and in vivo in a bleomycin mouse model of tissue fibrosis, respectively (257, 380).

Protein phosphatase 2A (PP2A) is a major serine/threonine protein phosphatase in eukaryotic cells and regulates many signaling pathways. However, the participation of PP2A in EndMT is poorly understood. In a recent study, the role of PP2A in EndMT was evaluated and it was demonstrated that PP2A was activated in endothelial cells during their EndMT phenotype acquisition, and it was found that specific inhibition of PP2A activity prevented EC undergoing EndMT (77). Importantly, PP2A activation was dependent on tyrosine nitration at residue 127 in the catalytic subunit of PP2A (PP2Ac) and blockade of tyrosine127 nitration to inhibit PP2A activation by using a mimic peptide derived from PP2Ac conjugated to a cell-penetrating peptide prevented TGF-β1-induced EndMT (77). Another investigation in diabetes-associated kidney fibrosis showed that the N-acetyl-seryl-aspartyl-lysyl-proline peptide inhibited the fibrotic process through inhibition of EndMT (231). More recently, it was shown that irisin, a myokine/adipokine with important effects on hyperglycemia and obesity-associated metabolic alterations, was a potent regulator of high-glucose-induced EndMT and that these effects caused a dose-dependent bidirectional effect on the development of diabetic cardiomyopathy (182).

C. Miscellaneous EndMT Inhibitors

Numerous other compounds have been shown to mediate potent antifibrotic effects in various experimental animal models of human fibrotic diseases as well as to cause significant improvement in animal models of human vascular diseases through modulation of the EndMT process. Among these, there are several activators of PPAR-γ that have been shown to inhibit EndMT in cardiac fibrosis (144, 326, 334) and in monocrotaline-induced PAH in rats (163). Salvianolic acid and ginsenoside-Rb3, both naturally occurring compounds, reduce EndMT progression in pulmonary vascular remodeling and in myocardial infarction, respectively (55, 352). Another study showed that nitro-oleic acid was a potent inhibitor of EndMT in HUVECs (9). Also, in myocardial infarction, losartan was shown to inhibit TGF- β-mediated EndMT through p-ERK (332, 336), and more recently, givinostat, a type II histone acetylate inhibitor, has been shown to decrease EndMT in a mouse model of acute myocardial infarction (213). Dipeptidyl peptidase (DPP-4) inhibitors utilized for treatment of kidney and lung fibrosis have been shown to exert antifibrotic effects owing to the downregulation of EndMT (145, 296). The mechanisms involved in the DPP-4 inhibition effects were examined following the intraperitoneal injection of monocrotaline to rats (344). Monocrotaline resulted in EndMT in renal tissue microvasculature, and the DPP-4 inhibitor sitagliptin decreased EndMT through glucagon-like peptide-1 (GLP-1) signaling as the sit gliptin effect was blocked by the GLP-1 receptor antagonist exendin-3. Furthermore, activation of GLP-1 signaling with liraglutide also reversed monocrotaline-induced EndMT. Thus the results indicated that DPP-4 inhibition prevented normal endothelial cells from undergoing EndMT in a GLP-1 receptor-dependent manner by regulating TGF signaling (344). A series of other studies demonstrated that reduction of parathyroid hormone levels by cinacalcet during treatment of renal fibrosis was accompanied by inhibition of EndMT (332, 333), and it was suggested that some of the beneficial effects of normalization of parathyroid hormone levels such as the reduction of aortic wall calcification in uremic rats were mediated by EndMT inhibition. More recently, the effect of the histone deacetylase inhibitor givinostat was examined in a mouse model of acute myocardial infarction, and it was found that its administration induced a decrease in EndMT and in the associated inflammatory process. These effects resulted in a reduction of the extent and severity of cardiac fibrosis and in an improvement in myocardial function and performance mediated by the abrogation of EndMT, suggesting that givinostat may be an effective therapeutic intervention for various cardiovascular diseases associated with EndMT-induced cardiac fibrosis (213).

IX. CONCLUDING REMARKS AND FUTURE DIRECTIONS

The results of the extensive studies involving numerous scientific disciplines, varied areas of investigational focus including evidence from in vitro studies with human endothelial cells and in vivo investigations of experimental animal models of various benign and malignant diseases, and morphological, histopathological, and molecular studies with human pathological tissues conclusively demonstrate an important role of EndMT in the pathogenesis of numerous human disorders. Although the phenotypic transition of endothelial cells to cells with mesenchymal characteristics and functions was well accepted to occur during embryologic development, there has been substantial reluctance and controversy among the scientific community to recognize that a similar phenotypic transition could take place in the postembryonic development or in adult tissues. More importantly, there has been substantial skepticism regarding the possibility that the EndMT process may be of relevance to the pathogenesis, development, and progression of human pathological conditions and disorders. Thus we believe that this review will certainly provide a critical and unbiased assessment of the background and significance of the extensive scientific literature published about this important topic, and it is certainly expected that it may be of value to mitigate some of this controversy and skepticism. However, we are convinced that there are still a large number of unanswered questions that will need to be addressed in future investigations, including the identification of the earliest event or events that initiate the gene expression changes and the subsequent molecular pathway modifications comprising an extremely well-coordinated activation and inhibition of crucial intracellular molecular reactions. We are also aware that the relative contribution of the different mechanisms that are capable of inducing EndMT as well as the potential epigenetic regulators participating in the process require further study and quantitative assessment. Finally, we are extremely optimistic that this review will certainly encourage the development and testing of novel and potentially effective drugs and other therapeutic approaches to overcome the multiple and quite deleterious effects of EndMT in a large number and variety of human diseases.

We believe that the extensive studies reviewed here clearly indicate that elucidation of the molecular mechanisms involved in EndMT may provide novel molecular targets and therapeutic approaches for numerous diseases in which the EndMT process plays a role in their development and/or pathogenesis and, eventually, these studies may lead to a change in the therapeutic paradigm for these disorders.

GRANTS

This work was supported by National Institute of Arthritis and Musculoskeletal and Skin Diseases Grant AR 19616 (to S. A. Jimenez).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.