The epithelial to mesenchymal transition (EMT) is a shift from one cell type to a dramatically different one. Whereas epithelial cells are comparatively static, with a group mentality, mesenchymal cells are solo wanderers. The transformation from a rigid to a fluid cell is central to much of the tissue remodeling that takes place during the highly dynamic process of embryonic development.
Epithelia are comprised of single or stratified cell layers that form a protective barrier at a tissue surface or around organs. Accordingly, epithelial cells are replete with cell–cell adhesions such as adherens junctions, desmosomes, tight junctions, and gap junctions that keep them “joined at the hip.” Thus, any epithelial cell migration occurs en masse to preserve barrier function. Also integral to cell function is an apical-basal polarity that structures distribution of cell–cell adhesions, basally secreted lamina, and other features unique to distinct epithelial types such as apical microvilli, or apically localized components of Golgi apparatus found in secretory epithelia (Fig. 1A).
In contrast to the relative rigidity of epithelia, mesenchymal cells are designed to move. They have a front-to-back polarity that positions leading edge filopodia that pull cells forward. Yet overall, the cells are relatively unstructured and irregularly shaped, characteristics that also benefit motility (Fig. 1B).
EMT is central to the creation of numerous organs and complex tissues in vertebrates, and in the adult underlies certain diseases such as carcinoma progression and organ fibrosis. This primer introduces the reader to these concepts as well as to cellular and molecular processes underlying the cell's metamorphosis. For more information on topics covered in this primer, the reader is referred to other comprehensive reviews and references therein (Hay,1968; Barrallo-Gimeno and Nieto,2005; Hay,2005; Liu,2006; Boutet et al.,2006; Thiery and Sleeman,2006; Baum et al.,2008; Chua et al., in press).
EMT SHAPES THE EMBRYO
EMT arose nearly 600 million years ago as a way to establish the middle primary germ layer, mesoderm. The advent of this process allowed for great flexibility in morphogenesis, and creation of structurally complex tissues. In vertebrates, EMT is a source of mesenchyme in diverse places, especially in mesodermally derived tissues. Below are three examples of morphogenetic events that are dependent on EMT.
A crucial morphogenic event in embryonic development is gastrulation, the process by which the three germ layers are established. In mammals, birds, and reptiles, epiblast cells —the upper layer of cells in the blastula—carry out EMT, allowing them to ingress into the primitive streak. Once interior, some mesenchymal cells differentiate as mesoderm, while others displace the underlying hypoblast and differentiate as definitive endoderm.
Neural Crest Development
Neural crest is a transient population that forms between the dorsal neural tube and epidermis in vertebrates. Transition to a mesenchymal state is an essential step, allowing them to migrate long distances to various sites in the body. Neural crest give rise to several derivatives, including neurons and glia of the peripheral nervous system, connective tissue, pigment, and facial cartilage and bone.
Nascent somites are epithelial balls of cells that bud from presomitic mesoderm on either side of the neural tube. As somitogenesis progresses, subpopulations of cells become mesenchymal at different times, giving rise to distinct differentiated cell types. The first population to undergo EMT is the ventral somite, or sclerotome, enabling cells to surround the neural tube and notochord, where they differentiate as axial bones and cartilage. Next, cells from the dorsal somite, or dermamyotome, move ventrally to populate the myotome, which eventually gives rise to skeletal muscle. At the same time, cells from the medial dermamyotome migrate to the ectoderm and differentiate as dermis of the back. Finally, lateral dermamyotomal cells disperse to form muscle of the limbs, tongue, and diaphragm.
BREAKING FROM THE PACK
EMT directs an extensive restructuring of the cell—a multistep process that culminates in alienation of an epithelial cell from its tight-knit community. The inductive cue varies among cell types, suggesting that activation is context-dependent. For example, TGFβ2 induces EMT in the atrioventricular canal, and scatter factor/hepatocyte growth factor (SF/HGF) does so in the lateral dermamyotome during somitogenesis; other inducers include EGF, FGF, BMPs, Wnts, Notch, and contact with extracellular matrix (ECM) components like collagen.
The EMT process consists of several stereotyped events. Cells extinguish expression of epithelial cell adhesion and structural components, like E-cadherin, cytokeratins, and laminins, in favor of components that favor mobility, vimentin, vitronectin, fibronectin, and N-cadherin. In addition, contacts with its epithelial community, mediated by desmosomes, adherens and tight junctions, are severed. Combined, these alterations promote changes in cell shape and polarity that support migration.
It is widely believed that each of the various EMT triggers sets forth a conserved cascade of downstream events. Recent evidence provides new insight into this matter. Sheng and colleagues show that during chick gastrulation, EMT consists of several temporally separable steps: BM breakdown, followed by loss of tight junctions, and finally a shift in cadherin expression where N-cadherin replaces E-cadherin (Nakaya et al.,2008). Moreover, breakdown of cell–basement membrane (BM) integrity is independently regulated by RhoA down-regulation and microtubule instability. The findings suggest that EMT may consist of a sequence of separately regulated steps.
The product of EMT, the mesenchymal cell, is built to move. In a developmental context, the cell will often follow molecular cues, even penetrating through surrounding tissues to reach its final position. At its destination, it can undergo the reverse process, mesenchymal to epithelial transition (MET) to form a new structure or to integrate with existing epithelia. In response to local cues, the cell then differentiates into a specified cell type found within the target tissue domain.
THE TIPPING POINT
Up- or down-regulation of E-cadherin is a nodal point in the transition between epithelial and mesenchymal states. The calcium dependent cell adhesion molecule is a core component of adherens junctions, the most prevalent point of contact between epithelial cells. Upon transfection of E-cadherin into mesenchymal cells, they acquire epithelial characteristics. The converse is also true. In the embryo, E-cadherin is down-regulated at sites of EMT.
E-cadherin expression is subject to regulation by several factors that promote EMT transition. These include the transcription factors Snail/Snai1, Slug/Snai2, Zeb1, Zeb2/Sip1, Twist, and E12/E47, each of which bind the E-cadherin promoter and repress its transcription. More recently, it has been shown that E-cadherin expression is indirectly regulated by the miR-200 microRNA (miRNA) family, which directly represses Zeb1 and/or Zeb2/Sip1 (Cano and Nieto,2008). In response to TGFβ-induced EMT in vitro, the miRNAs are down-regulated, allowing for E-cadherin repression.
E-cadherin may also exert its influence over EMT by means of its regulation of β-catenin localization. E-cadherin sequesters β-catenin to the cytoplasm, and upon down regulation of E-cadherin, β-catenin localizes to the nucleus where it activates LEF/TCF transcription factors. These factors then regulate genes associated with the mesenchymal fate, including repression of E-cadherin itself.
Another family of genes often considered as master regulators of EMT are the Snail genes, which consist of Snai1 and Snai2 in mice and humans. Snail genes are zinc-finger transcription factors that have been found in every embryonic EMT event in which they have been examined. The Snail family orchestrates changes in expression of genes required for EMT; they repress epithelial markers, including E-cadherin, and activate mesenchymal markers and genes involved in cell shape change and motility. It is important to note that this is not the only function of the Snail family. They also inhibit proliferation, enhance cell survival, and regulate cell adhesion and migration independent of EMT. The latter two functions may represent ancestral roles for Snail genes that precede the evolution of mesoderm.
EMT CRIPPLES THE ADULT
In the adult, EMT regulates the development of at least two diseases: cancer and fibrosis. During carcinoma progression and metastasis, epithelial cells destroy the basement membrane, sever cell–cell contacts, and acquire invasive motility. Sound familiar? Indeed, involvement of key EMT regulators in cancer suggest that tumorigenic cells deploy this process. For example, mutations in E-cadherin cause inherited gastric cancer (OMIM 192090). In addition, expression of E-cadherin repressors, Zeb, Twist, and Snail, accompanied by a loss of E-cadherin expression have been found in metastatic tumors. Moreover, the miR-200 family, which down-regulate Zeb1 and Zeb2/Sip1, are markers for epithelial and well-differentiated cancers.
The idea that EMT is an integral part of tumor progression also has its skeptics, mostly because functional in vivo evidence has been hard to come by. Because tumor cells are a heterogeneous, rapidly changing population, it is difficult to find and follow developing tumors and cancer progression in the body. To secure the role of EMT in the malignancy pathway, it will be imperative to do so.
EMT also plays a central role in kidney fibrosis, a cause of chronic kidney disease (CKD). During fibrosis, tubular epithelial cells undergo EMT to become myofibroblasts. These and resident fibroblasts secrete excess extracellular matrix, which leads to scarring and eventual kidney failure. The disease process can be viewed as wound healing gone awry, where normal deposition of ECM for repair fails to be terminated. TGFβ has been shown to cause renal fibrosis in transgenic mice, stimulate EMT, and is up-regulated in CKD. These data suggest that aberrant TGFβ signaling may uncontrollably stimulate tubule epithelial cells to become ECM secreting myofibroblasts. Indeed, aberrant activation of Snail, a well known TGFβ target, is sufficient to induce renal fibrosis leading to renal failure.
WHAT DO THE EXPERTS SAY?
Featured below is a discussion with EMT experts Angela Nieto, Ph.D., and Jean Paul Thiery, Ph.D. (Fig. 2) about current topics in the field.
Dev Dyn: What is your lab's research focus?
JPT: My laboratory is focusing on identifying signaling pathways underlying EMT in cancer cell lines and human carcinoma. We are also addressing the mechanisms controlling EMT in the ontogeny of murine neural crest.
MAN: We are interested in the study of cell movements both in physiology and pathology. Thus, the EMT is a central focus in my lab. We use the mouse, chick, and zebrafish embryos as experimental models together with the analysis of samples from patients.
Dev Dyn: What initially provoked your interest in this field?
JPT: I first became interested in the mechanisms driving epithelial cell plasticity in 1978 when I started on a project to investigate avian neural crest cell migration. I had only a very basic knowledge of developmental biology and among other things, I was not aware that cells could move from one place to another in embryos. I became quickly puzzled by the fact that avian neural crest cells could delaminate from the dorsal border of the neural tube and migrate. I filmed the delaminating neural crest cells both in vivo and in vitro and observed the dynamic nature of these cells during delamination. I then realized, that neural crest cell EMT and migration could represent an ideal model to study cancer dissemination. However, it would have been very difficult to study the molecular mechanisms driving EMT in these very few migratory neural crest cells and it became apparent that I would need to use another in vitro model to study EMT.
Greenburg and Hay in 1982 published a landmark paper showing epithelial cell plasticity that prompted me to search for a simpler model than crest cells. Thus I searched for cell lines that could exhibit such plasticity. My final choice came from a lucky encounter with Ruy Tchao, a pathologist in Philadelphia. He convinced me with a remarkable movie showing small cell clusters of NBT-II rat bladder carcinoma responded to type 1 collagen substrate by strong activation of lamellipodial activity. Although he did not show clear scattering in his movie, it reminded me of the behavior of embryonic and adult epithelial cells embedded in 3D collagen gels shown by Hay and colleagues. I immediately decided to work on this cell line. I subsequently found that the cell line was also able to undergo an EMT when stimulated with a crude extract from bovine brain, later identified as FGF1.
The seminal paper of Sir Michael Stoker in 1985 suggested that scatter factor could be produced by fibroblastic cells. However, it wasn't until 1990 that Michael Stoker, Ermanno Gherhardi and other teams headed by Walter Birchmeier and Toshikazu Nakamura identified that the scatter factor was identical to hepatocyte growth factor. EMT then became amenable to further molecular analyses. An important step in understanding EMT was the discovery that Snail was acting as a master gene and was subsequently shown to be a critical transcriptional repressor of E-cadherin, the prototypic epithelial cell adhesion molecule.
MAN: I isolated the first Snail family member from mouse and chick 15 years ago, when I was about to go back to Spain after finishing a postdoc in David Wilkinson's lab at the National Institute for Medical Research in London. I had been working on hindbrain segmentation with David, and we were very close to Robb Krumlauf′s lab, where there was already a very high interest in the neural crest. I was fascinated by the described migratory behavior and by the evolutionary and developmental potential of the neural crest cells. Therefore, when I saw the expression patterns of Snail1 in the mouse and Snail2 in the chick (at that time called Snail and Slug, respectively) in migratory cells within the embryo, I decided that I wanted to study how they worked.
When back in Madrid, we first described its role in the induction of EMT during embryonic development. Migratory neural crest cells were very difficult to distinguish from neighboring cells in the embryo although fundamental work by Nicole Le Douarin using chick–quail chimaeras and the NC1-HNK1 antibody characterized by Jean Paul Thiery had been crucial in defining migratory pathways. The description of Snail genes as the first neural crest markers that were expressed in the premigratory crest cells and could simultaneously identify the migratory population very much contributed to the impact of our work. Ironically, their use as markers also contributed to specifically associate these genes with the neural crest, an issue that we always tried to clarify. In the original work, the migration of the early mesodermal cells from the primitive streak was also analyzed and the message was that these transcription factors act by controlling cell behavior and the induction of the EMT rather than defining a particular cell fate. In this initial study, we proposed that Snail pathological activation could be involved in the acquisition of invasive properties during the malignization of epithelial tumors. We then started a very fruitful collaboration with Amparo Cano, an expert of cadherin regulation in cancer progression. Although it took us years to demonstrate it, now it is well-established that this is the case.
Dev Dyn: Which papers have most impacted your research?
JPT: The three following references were of great interest to me during my earlier studies on EMT. Greenburg and Hay were able to show that EMT can be studied in vitro. Michael Stoker and colleagues were the first to show that EMT can be induced by a soluble factor later identified as scatter/hepatocyte factor, while the seminal paper of Nieto et al. (1994) provided us for the first time with a fundamental mechanism for the execution of the EMT program.
1Greenburg G, Hay ED. 1982. Epithelial suspended in collagen gels can lose polarity and express characteristics of migrating mesenchymal cells. J Cell Biol 95:333–339.
2Stoker M, Perryman M. 1985. An epithelial scatter factor released by embryo fibroblast. J Cell Sci 77:209–223.
3Nieto MA, Sargent MG, Wilkinson DG, Cooke J. 1994. Control of cell behavior during vertebrate development by Slug, a zinc finger gene. Science 264:835–839.
MAN: 1. Bellairs R. 1987. The primitive streak and the neural crest: comparable regions of cell migration? In: Maderson P, editor. Developmental and evolutionary aspects of the neural crest. New York: John Wiley and Sons. p 123–145.
This paper gave me all the inspiration and background to better understand the phenotype of embryos treated with antisense oligos to Slug in 1994.
2. Hay ED. 1995. An overview of epithelio-mesenchymal transformation. Acta Anat (Basel) 154:8–20.
Betty Hay has been a reference for all of us interested in cell adhesion and motility. She coined the term “epithelial to mesenchymal transformation” and described the transient nature of the process and the reversion to the epithelial character (MET). This is why the currently preferred term is “epithelial to mesenchymal transition”.
3. I have always been impressed by the notion that by the 1980s, Jean Paul Thiery had already pioneered the view that tumor progression and embryonic development have some mechanisms in common.
Dev Dyn: There are several different signaling molecules that activate EMT. How do you explain that? Are they activating equivalent or distinct processes?
MAN: If we define EMT as the transition from an epithelial to a mesenchymal cell and confirm that this is the case in terms of cell biology, then the outcome is the same. However, the signal transduction pathways used by the different signals are obviously different although they usually converge in the activation of the so-called EMT inducers. Very importantly, the cell context very much influences the response. The capabilities of the cells that have undergone EMT vary depending on whether they are embryonic or adult cells and in this case, whether they are normal or transformed. It is also interesting to note that the classic signals for EMT and the described EMT inducers are unable to complete the process in some cell types, which takes us to the next question.
JPT: Although EMT is an evolutionary conserved mechanism, the molecular details differ between species during the formation of the body plan and subsequently in organogenesis. Angela Nieto correctly states that these different signaling pathways converge to induce master genes, which in turn control the appropriate expression of genes that execute the EMT program.
Dev Dyn: Is EMT an all or none process, or can cells arrest at intermediate steps in vivo? If so, what is the function of such cells?
MAN: Indeed, cases of partial EMT have been described in different systems. As such, particular Drosophila snail mutant alleles show an intermediate phenotype in the cells that normally express the gene. These cells exhibit both mesodermal and ectodermal markers, in agreement with the finding that Snail regulates genes involved in the loss of the epithelial phenotype and the gain of the mesenchymal character, and probably reflects that these genes can be independently regulated in different mutants. The partial EMT phenotype is referred to as the metastable phenotype by Pierre Savagner and is characterized by the maintenance of cell–cell adhesion structures that do not preclude active cell migration. This process seems to be at work during wound healing. In relation to this, we have recently shown that the function of Snail genes in regulating adhesion may or may not be associated with the induction of a full EMT, depending on the cellular context. As such, while Snail2 induces a full EMT in the mesoderm of the early chick embryo, two Snail1 proteins cooperate in the migration of the axial mesendoderm in the zebrafish embryo by decreasing adhesion but without inducing a full EMT.
JPT: There are several instances where EMT may be incomplete, such as during gastrulation in Xenopus laevis where convergence extension dominates. In contrast, complete EMT occurs in Pleurodeles waltlii, a well-characterized Urodele. Recent findings from Mina Bissell's laboratory show that there is incomplete EMT during mammary gland branching morphogenesis. Angela Nieto also emphasizes that a metastable state can contribute to general epithelial cell plasticity. Cells can reorganize in a monolayer without necessarily losing all the epithelial characteristics.
Dev Dyn: Is MET simply a reversal of EMT or can it be viewed as a completely different process?
MAN: As a developmental biologist, I tend to think that MET is the reverse of the EMT and is related to the competence of the cells to respond to signals that modulate their response and unveil their potential for phenotypic plasticity. Sequential EMTs and METs are necessary for the formation of different tissues and organs, reflecting the high degree of cell plasticity in embryos. Obviously, plasticity is much more restricted in the adult. However, in pathological contexts such as transformed tumor cells or even during normal ageing, epithelial cells can regain the ability to undergo EMT, contributing to the progression of tumors or to organ fibrosis, respectively. Interestingly, the induction of MET can be regarded as a therapeutic strategy to palliate organ fibrosis and hopefully, tumor progression.
JPT: MET is in most cases a reversal of EMT. For instance, heart morphogenesis involves three cycles of EMT-MET. Angela Nieto suggests that induction of MET could abrogate tumor progression or fibrosis. This is indeed a very good strategy. However, further evidence is required to validate activation of an MET program in dedifferentiated tumors and in fibrotic lesions.
Dev Dyn: Targeted repression of Snail genes has been discussed as a means to prevent or inhibit tumor progression or fibrosis. What do you think of this strategy? Do you favor other approaches?
JPT: This may be strategy to explore, however Snail has other important functions in cells including cell survival. It may difficult to attribute the inhibitory function of Snail solely to EMT. It would be also important to determine its tissue distribution in normal cells to avoid important side effects. Thus far, it has been difficult to identify small molecules that specifically interfere with transcription factors. The big Pharmas still prefer to target enzymes and a downstream target of Snail may turn out to be a more effective target. There are already several small molecules that interfere with EMT, although they were originally discovered as anti mitotic agents. A small review on this topic has been recently written (Chua et al., in press). Several pharmaceutical companies are currently focusing on drug screening using EMT in vitro models. Similar approaches to identify new drugs targeting EMT signaling nodes are also being carried out by several academic laboratories.
MAN: Indeed, Snail has other functions, including regulation of cell proliferation and cell death. These functions can work concomitantly with the changes in cell shape and motility and as a matter of fact, be part of the EMT process or they can be independent in different cell contexts. By blocking Snail function in invasive tumor cells, perhaps one could not only modulate invasion but also make cells more susceptible to destruction. With respect to tumor progression, studies from Amparo Cano′s lab in Madrid have recently shown that inhibiting Snail expression has an impact on the metastatic ability of tumors induced in mice. Regarding side effects due to Snail silencing in the adult as a result of a putative therapeutic strategy, we have been always concerned about them. However, it is fair mentioning that although it has not yet been systematically analyzed, the more we know about Snail the more convinced we are that they are (and must be) kept silenced in the adult, precluding at least in theory side effects problems due to the targeted repression of its pathological activation. Having said this, I completely agree with Jean Paul that targeting transcription factors is a very difficult task, and thus, the Pharma Companies prefer to target other molecules.
Dev Dyn: Is EMT in tumor progression a repeat of EMT in development or are there distinct differences?
JPT: The mechanisms governing EMT in embryos have not been fully deciphered because EMT is often embedded in mechanisms that also specify cell lineages. Our current understanding is that the mechanisms driving epithelial-mesenchymal transitions in development share common features with those inducing the disappearance of epithelial features in carcinoma cells during tumor progression. Carcinoma-derived epithelial cell lines can undergo EMT following activation of tyrosine kinase receptors; a mechanism that also operates during gastrulation in Drosophila, chick, and mouse. TGFβ induces EMT in heart endocardium, and is also found to promote EMT in several normal and malignant epithelial cells. However, not all steps involved in EMT that are described in embryos can be observed in all carcinoma cell lines and it is not possible to follow the EMT process in solid tumors. Pathologists can observe single cells at the invasive front of carcinoma; these cells are well documented in the case of colon carcinoma. Often single carcinoma cells can also be detected in the blood and bone marrow of cancer patients. Some murine models have served to address these issues using live cell imaging. There is evidence for EMT in these models but not all steps may be amenable in the near future for advanced molecular analysis.
However, one must remember that even though similar EMT inducing signaling pathways may be activated, their contribution to the phenotypic conversion may not be complete. Other activated pathways in cancer cells could alter significantly the EMT pathways. In summary, it remains to be discovered how carcinoma through genetic alteration and epigenetic changes may activate successive steps in EMT to disseminate cells. It is far more complex than what we can describe by short-term EMT in cell lines or in embryos.
MAN: I am convinced of the many similarities between these two processes. However, and as mentioned by Jean Paul, cancer cells carry many genetic alterations and are highly responsive to epigenetic cues, very much influencing their responses.
Dev Dyn: What are some important questions that remain to be answered?
JPT: There are many unsolved issues as EMT during development is embedded into multiple developmental programs. EMT during gastrulation occurs in cell collectives that are responding to numerous inductive signals from the node. Neural crest specification programs operate at the time of EMT from the neural epithelium. Understanding how these different signaling pathways progressively orchestrate EMT is a major challenge. The situation is even more complex in tumors where the loss of the epithelial phenotype may take several years. Clonal evolution in tumors creates enormous phenotypic heterogeneity. For this reason, some pathologists still argue that EMT does not play a role in tumor progression.
MAN: Undoubtedly, the understanding of the individual contribution and the integration of the different signaling pathways that trigger EMT is one of the challenges. Similarly, understanding the biological nature of the different cell responses is another complex issue that requires a global analysis and the knowledge of the process from a systems biology perspective.
Dev Dyn: What exciting ideas are emerging in the field?
JPT: The concept that tumor initiating cells in carcinoma have a mesenchymal-like phenotype is opening a new avenue of exciting research in the EMT field. Stemness may be better preserved in the mesenchymal state. It is already well established that mesenchymal stem cells can give rise to different lineages of the three primary germ layers. In contrast, it has recently been shown that crest cells can contribute to mesenchymal stem cells. Robert Weinberg's laboratory has provided the first evidence of this concept showing that normal breast progenitor cells and clonogenic cancer cells have a mesenchymal phenotype. These findings vindicate the proposal of Brabletz and colleagues suggesting that pioneer colon cancer cells have a mesenchymal stem cell like phenotype, whereby single tumor cells can invade and continuously give rise to organized cell clusters that resemble normal colon architecture.
MAN: I agree that the new proposal of tumor initiating cells is important and exciting. Its relationship with stemness and the data coming out from Weinberg′s lab are also very stimulating. It is interesting to note that the concept of stemness associated with the neural crest has been put forward by Nicole Le Douarin for many years, and work from Lukas Sommer′s lab in collaboration with Jean Paul Thiery has also recently shown that neural crest-derived cells have stem cell properties.
On the other hand, the emergence of new regulatory mechanisms driven by noncoding RNAs is expanding the field of EMTs research while providing more complexity to the process. State-of-the-art imaging techniques are helping us to understand the cell biology of the process in vivo, such as the work carried out in John Condeelis' lab in New York showing individual cells emigrating from the primary tumor.
Many thanks to Angela Nieto for critically evaluating the manuscript, and to both she and Jean Paul Thiery for sharing their insightful dialog.