ConSEnSuS Statement Guidelines and definitions for research on epithelial–mesenchymal transition Jing Yang 1 ✉, Parker Antin2, Geert Berx3, Cédric Blanpain4, Thomas Brabletz5, Marianne Bronner 6, Kyra Campbell 7, Amparo Cano 8, Jordi Casanova 9, Gerhard Christofori10, Shoukat Dedhar 11, Rik Derynck 12, Heide L. Ford 13, Jonas Fuxe 14, Antonio García de Herreros 15, Gregory J. Goodall 16, Anna-Katerina Hadjantonakis17, Ruby J. Y. Huang 18, Chaya Kalcheim 19, Raghu Kalluri20, Yibin Kang 21, Yeesim Khew-Goodall 22, Herbert Levine 23, Jinsong Liu24, Gregory D. Longmore25, Sendurai A. Mani26, Joan Massagué27, Roberto Mayor 28, David McClay29, Keith E. Mostov 30, Donald F. Newgreen 31, M. Angela Nieto 32, Alain Puisieux 33,47, Raymond Runyan 34, Pierre Savagner35, Ben Stanger36, Marc P. Stemmler 5, Yoshiko Takahashi37, Masatoshi Takeichi 38, Eric Theveneau 39, Jean Paul Thiery 40, Erik W. Thompson 41, Robert A. Weinberg42, Elizabeth D. Williams 43, Jianhua Xing 44, Binhua P. Zhou45, Guojun Sheng46 ✉ and On behalf of the EMT International Association (TEMTIA) Abstract | Epithelial–mesenchymal transition (EMT) encompasses dynamic changes in cellular organization from epithelial to mesenchymal phenotypes, which leads to functional changes in cell migration and invasion. EMT occurs in a diverse range of physiological and pathological conditions and is driven by a conserved set of inducing signals, transcriptional regulators and downstream effectors. With over 5,700 publications indexed by Web of Science in 2019 alone, research on EMT is expanding rapidly. This growing interest warrants the need for a consensus among researchers when referring to and undertaking research on EMT. This Consensus Statement, mediated by ‘the EMT International Association’ (TEMTIA), is the outcome of a 2-year-long discussion among EMT researchers and aims to both clarify the nomenclature and provide definitions and guidelines for EMT research in future publications. We trust that these guidelines will help to reduce misunderstanding and misinterpretation of research data generated in various experimental models and to promote cross-disciplinary collaboration to identify and address key open questions in this research field. While recognizing the importance of maintaining diversity in experimental approaches and conceptual frameworks, we emphasize that lasting contributions of EMT research to increasing our understanding of developmental processes and combatting cancer and other diseases depend on the adoption of a unified terminology to describe EMT. Epithelial–mesenchymal transition (EMT) is a cellular characteristics and the acquisition of mesenchymal char­ process during which epithelial cells acquire mesen­ acteristics. Cells then display fibroblast-like morphol­ chymal phenotypes and behaviour following the down­ ogy and cytoarchitecture, as well as increased migratory regulation of epithelial features. EMT is triggered in capacity. Furthermore, these now migratory cells often response to signals that cells receive from their micro­ acquire invasive properties (Fig. 1). environment. The epithelial state of the cells in which EMT was first described by researchers studying EMT is initiated is characterized by stable epithelial early embryogenesis as a programme with well-defined ✉e-mail: jingyang@ucsd.edu; cell–cell junctions, apical–basal polarity and interac­ cellular features1,2. It is now widely accepted that EMT sheng@kumamoto-u.ac.jp tions with basement membrane. During EMT, changes occurs normally during early embryonic development, https://doi.org/10.1038/ in gene expression and post-translational regulation to enable a variety of morphogenetic events, as well as s41580-020-0237-9 mechanisms lead to the repression of these epithelial later in development and during wound healing in adults. nATuRe RevIeWS | MolECulAr CEll BioloGy volume 21 | June 2020 | 341 C o n S e n S u S S tat e m e n t Moreover, EMT is known to be activated during cancer have been published in the past 5 years alone (Fig. 2), and pathogenesis and tissue fibrosis. The reverse process, half of those have reported on studies of EMT in the known as mesenchymal–epithelial transition (MET), context of cancer biology. The growing complexity and also occurs frequently during development. A salient diversity of the EMT literature has resulted in vague characteristic of EMT occurring in vivo, whether during and often confusing definitions of EMT and associated normal development or in a pathological context, is that nomenclature. Cell biologists have traditionally focused the transition from an epithelial to a mesenchymal state is on the microscopically visible and profound changes in often incomplete, resulting in cells that reside in interme­ cell–cell interactions, cell motility, cytoskeletal organi­ diate states that retain both epithelial and mesenchymal zation, cell proliferation and resistance to various stress­ characteristics. Importantly, these intermediate states can ors that occur during EMT. Molecular biologists have be diverse, depending on the biological context3. focused on changes in the activity of EMT-associated The EMT research field has grown explosively over transcription factors (EMT-TFs) and in aspects of the past 20 years. More than half of all articles on EMT their regulation, often involving various chromatin Author addresses 1 Departments of Pharmacology and Pediatrics, 15 Programa de Recerca en Càncer, Institut 31 Murdoch Children’s Research Institute, Royal Moores Cancer Center, University of California, Hospital del Mar d’Investigacions Mèdiques Children’s Hospital, Parkville, VIC, Australia. San Diego, La Jolla, CA, USA. (IMIM) and Departament de Ciències 32 Instituto de Neurociencias (CSIC-UMH) Avda 2 Department of Cellular and Molecular Experimentals i de la Salut, Universitat Pompeu Ramon y Cajal s/n, Sant Joan d´Alacant, Spain. Medicine, University of Arizona, Tucson, Fabra, Barcelona, Spain. 33 Université Claude Bernard Lyon 1, INSERM AZ, USA. 16 Centre for Cancer Biology, An alliance of SA 1052, CNRS 5286, Centre Léon Bérard, Cancer 3 Molecular and Cellular Oncology Lab, Pathology and University of South Australia, Research Center of Lyon, Lyon, France. Department of Biomedical Molecular Biology, Adelaide, SA, Australia. 34 Department of Cellular and Molecular Medicine, Ghent University, Cancer Research Institute 17 Developmental Biology Program, Sloan University of Arizona, Tucson, AZ, USA. Ghent (CRIG), VIB Center for Inflammation Kettering Institute, Memorial Sloan Kettering 35 INSERM UMR 1186, Integrative Tumor Research, Ghent, Belgium. Cancer Center, New York, NY, USA. Immunology and Genetic Oncology, Gustave 4 Laboratory of Stem Cells and Cancer, Université 18 School of Medicine, College of Medicine, Roussy, University Paris-Saclay, Villejuif, France. Libre de Bruxelles, Bruxelles, Belgium. National Taiwan University, Taipei, Taiwan. 36 Department of Medicine, Perelman School of 5 Department of Experimental Medicine 1, 19 Department of Medical Neurobiology, Institute Medicine, University of Pennsylvania, Nikolaus-Fiebiger-Center for Molecular for medical Research Israel-Canada and the Philadelphia, PA, USA. Medicine, Friedrich-Alexander-University Safra Center for Neurosciences, Hebrew 37 Department of Zoology, Graduate School of Erlangen-Nürnberg, Erlangen, Germany. University of Jerusalem, Hadassah Medical Science, Kyoto University, Kyoto, Japan. 6 Division of Biology and Biological Engineering, School, Jerusalem, Israel. 38 RIKEN Center for Biosystems Dynamics California Institute of Technology, Pasadena, 20 Department of Cancer Biology, Metastasis Research, Kobe, Japan. CA, USA. Research Center, MD Anderson Cancer Center, 39 Centre de Biologie du Développement (CBD), 7 Department of Biomedical Science and Houston, TX, USA. Centre de Biologie Intégrative (CBI), Université Bateson Centre, University of Sheffield, 21 Department of Molecular Biology, Princeton de Toulouse, CNRS, UPS, Toulouse, France. Sheffield, UK. University, Princeton, NJ, USA. 40 Guangzhou Regenerative Medicine and Health, 8 Departamento de Bioquímica, Universidad 22 Centre for Cancer Biology, an Alliance of SA Guangdong Laboratory, Guangzhou, China. Autónoma de Madrid (UAM), Instituto de Pathology and the University of South Australia, 41 School of Biomedical Sciences and Institute of Investigaciones Biomédicas ‘Alberto Sols’ Adelaide, SA, Australia. Health and Biomedical Innovation, Queensland (CSIC-UAM), IdiPAZ & Centro de Investigación 23 Department of Physics, Northeastern University of Technology, Translational Research Biomédica en Red de Cáncer (CIBERONC), University, Boston, MA, USA. Institute, Woolloongabba, QLD, Australia. Madrid, Spain. 24 Department of Anatomic Pathology, The 42 Whitehead Institute for Biomedical Research, 9 Institute for Research in Biomedicine (IRB Division of Pathology and Laboratory Medicine, Department of Biology, MIT Ludwig Center for Barcelona), Barcelona Institute of Science and University of Texas MD Anderson Cancer Molecular Oncology, Massachusetts Institute of Technology/Institut de Biologia Molecular de Center, Houston, TX, USA. Technology, Cambridge, MA, USA. Barcelona (IBMB-CSIC), Barcelona, Spain. 25 Department of Medicine (Oncology) and 43 Australian Prostate Cancer Research 10 Department of Biomedicine, University of Department of Cell Biology and Physiology, Centre-Queensland (APCRC-Q) and Queensland Basel, Basel, Switzerland. ICCE Institute, Washington University, St. Louis, Bladder Cancer Initiative (QBCI), School of 11 Department of Biochemistry and Molecular MO, USA. Biomedical Sciences and Institute of Health and Biology, University of British Columbia and 26 Department of Translational Molecular Biomedical Innovation, Queensland University of British Columbia Cancer Research Centre, Pathology, University of Texas MD Anderson Technology, Woolloongabba, QLD, Australia. Vancouver, BC, Canada. Cancer Center, Houston, TX, USA. 44 Department of Computational and Systems 12 Departments of Cell and Tissue Biology, and 27 Cancer Biology and Genetics Program, Sloan Biology and UPMC-Hillman Cancer Center, Anatomy, University of California at San Kettering Institute, Memorial Sloan Kettering University of Pittsburgh, Pittsburgh, PA, USA. Francisco, San Francisco, CA, USA. Cancer Center, New York, NY, USA. 45 Department of Molecular and Cellular 13 Department of Pharmacology, University of 28 Department of Cell and Developmental Biology, Biochemistry and UK Markey Cancer Center, Colorado Anschutz Medical Campus, Aurora, University College London, London, UK. University of Kentucky College of Medicine, CO, USA. 29 Department of Biology, Duke University, Lexington, KY, USA. 14 Department of Laboratory Medicine Durham, NC, USA. 46 International Research Center for Medical (LABMED), Division of Pathology, Karolinska 30 Departments of Anatomy and Biochemistry/ Sciences (IRCMS), Kumamoto University, University Hospital and Department of Biophysics, University of California, San Kumamoto, Japan. Microbiology, Tumor and Cell Biology (MTC), Francisco, School of Medicine, San Francisco, 47 Present address: Institut Curie, PSL Research Karolinska Institutet, Stockholm, Sweden. CA, USA. University, Paris, France. 342 | June 2020 | volume 21 www.nature.com/nrm C o n S e n S u S S tat e m e n t Crumbs complex Par complex Tight junction Actin Adherens junction Scribble complex Desmosome Gap junction Epithelial cell Mesenchymal cell Apical–basal polarity Back–front polarity (migrating cell) Vimentin Actin-based subcortical mesh Actin stress fibres Nucleus Cytokeratin Hemi- desmosome Basement Extracellular membrane matrix Core EMT changes • Cytoskeleton remodelling Interstitial matrix • Loss of apical–basal cell polarity (e.g. fibrillar collagen I) • Cell–cell adhesion weakening • Cell–matrix adhesion remodelling • Cell individualization • Establishment of front–back cell polarity • Acquisition of cell motility • Basement membrane invasion Actin filament Focal Integrin adhesion E/M intermediate state E/M intermediate state Fig. 1 | EMT diversity represented by an epithelial–mesenchymal plasticity model. Various cellular features associated with an epithelial or a mesenchymal cell state are found in a range of combinations and to different degrees in cells in different developmental contexts. Epithelial cells are connected with each other via a variety of epithelial cell junctions, including adherens junctions, desmosomes, gap junctions and tight junctions. Adherens junctions are connected to cortical actin bundles, while desmosomes are linked with cytokeratin intermediate filaments. Tight junctions are localized at the apical-lateral contact points in order to help maintain epithelial polarity. Apical–basal polarity guides proper organization of the tight junctions, adherens junctions and desmosomes in epithelial cells. Polarity complexes, including the Par, Crumbs and Scribble complexes, define the apical versus basolateral domains of an epithelial cell. Epithelial cells are attached to the underlying basement membrane via hemidesmosomes, which contain integrin to allow binding to the basement membrane and are also linked to cytokeratins inside the cell. By contrast, mesenchymal cells do not contain functional epithelial junctions and present a back–front polarity in their actin stress fibres. Mesenchymal cells contain vimentin- based intermediate filaments and utilize integrin-containing focal adhesions to attach to the extracellular matrix. The accumulated loss or gain of epithelial/mesenchymal (E/M) characteristics pushes a cell towards various intermediate states (bottom left and right) in a fluid and reversible manner, between a complete epithelial (middle left) and a complete mesenchymal (middle right) state. EMT, epithelial–mesenchymal transition. nATuRe RevIeWS | MolECulAr CEll BioloGy volume 21 | June 2020 | 343 C o n S e n S u S S tat e m e n t 6,000 5,000 Number of publications 4,000 3,000 EMT plasticity in metastasis 2,000 miRNA regulation of EMT EMT promotes cancer stemness EMT in organ 1,000 fibrosis Experimental TGFβ First EMT-TF EMT-TF found analysis of EMT in induces found in the in metastatic embryonic tissue EMT embryo tumours 0 1980 1990 2000 2010 2019 Year Fig. 2 | Growth of the primary literature in EMT. The first experimental analysis of epithelial–mesenchymal transition (EMT) in development was published in 1979. The relationship of EMT to growth factors was found in 1989. Transcriptional regulation of EMT was identified in 1994. Subsequent growth of such research was stimulated by linkage of EMT to metastasis, organ fibrosis and stem cells. Growth in the field has been logarithmic since the first TEMTIA meeting in 2003. The graph indicates primary papers published each year, identified by a search of the Web of Science database. The total numbers of publications in 2018 and 2019 exceeded 5,000 and 5,700 articles, respectively. EMT-TF, EMT-associated transcription factors. modifications to orchestrate changes in EMT-associated associated with EMT using a limited set of molecular gene expression. Cancer biologists have often empha­ markers. However, the later identification of EMT as a sized the acquisition of various malignancy-associated crucial programme in cancer progression10,11 indicated cell phenotypes, notably invasiveness, as well as dis­ that EMT involves more than the originally identified semination and different degrees of cell responsive­ developmental EMT programmes and that, clearly, ness to various therapeutic modalities. However, these many different variations of the EMT programme exist context-specific, or even research-community-specific, that cannot be accurately defined by those limited sets criteria to define or refer to EMT do not reflect the of markers. current experimental shortcomings in the abil­ As the complexity of EMT events and EMT regula­ ity to identify EMT events. As we learn more about tors in both development and cancer becomes increas­ EMT-associated changes, it is becoming apparent that ingly appreciated, there is a need for the community of there is a great diversity of EMT phenotypic manifes­ researcher experts on EMT to agree on a number of key tations. Thus, narrow definitions have become inap­ points. These include: definitions of major EMT-related propriate or inaccurate, and a more encompassing terms, a description of EMT-associated phenomena, definition is required in order to describe this complex cell a description of the diverse versions of EMT that can biological programme. occur in different contexts, the context-dependent func­ tion and activity of EMT regulators, and the relation­ Purpose of this Consensus Statement ship between the core and non-core EMT functions of The use of the term EMT in research areas as diverse as EMT-TFs. In the context of cancer, there is also a need to developmental biology, cell biology, tissue homeostasis further consider the contributions of genetic alterations, and disease (notably cancer and fibrosis) has created the complex input of changing tumour environments discrepancies in data interpretation and persistent disa­ and the EMT-like changes that occur in non-epithelial greements about whether the process studied is EMT4–9, cancers such as melanoma, sarcoma and leukaemia. largely because the plasticity and heterogeneity of EMT Provoked by the passionate town hall discussions that programmes have been insufficiently considered. took place during the 2017 and 2019 meetings of the EMT was originally described as an important process EMT International Association (TEMTIA), TEMTIA in embryonic development during which epithelial cells proposes the following guidelines to define the EMT underwent a phenotypic transformation to mesenchy­ programme, its phenotypic plasticity and the resulting mal cells. These early analyses and subsequent valida­ multiple intermediate epithelial–mesenchymal states. tions in cell culture described the phenotypic changes By building such a consensus on EMT-related concepts, 344 | June 2020 | volume 21 www.nature.com/nrm C o n S e n S u S S tat e m e n t we aim to eliminate semantic problems in the EMT Slug expression can convert epithelial carcinoma cells debate and facilitate genuine cross-disciplinary discus­ into mesenchymal derivatives made a strong case for a sion of the roles of EMT in both normal development connection between embryonic EMT and cancer pro­ and pathological conditions. gression24. This notion was reinforced by the observation that the Snail family of transcription factors are capable A brief history of EMT of inducing EMT and invasiveness (the capacity to leave Modern EMT studies began with research aimed at the epithelial tissue and migrate into the underlying tis­ understanding tissue morphogenesis during develop­ sue) in normal epithelial cells, in part through transcrip­ ment, cell behaviour in culture and carcinoma invasive­ tional repression of the gene encoding E-cadherin25–28. ness in cancer progression. Elizabeth Hay recognized Additional EMT-TFs, notably E47, Twist1, Zeb1 and Neural crest A multipotent cell population the importance of EMT in embryogenesis1 and began to Zeb2, were identified by means of their ability to evoke formed at the interface discuss the concept of “epithelial–mesenchymal trans­ morphological and molecular changes associated with between the neuroepithelium formation” in the late 1970s. EMT was subsequently EMT29–32. It is important to note that these EMT-TFs and the epidermis. Neural crest observed in the context of neural crest formation12,13, usually cooperate with one another to orchestrate cells give rise to multiple cell types, including the neurons, heart valve formation14 and Müllerian duct regression15, EMT. A large number of studies have also revealed glial cells, pigment cells, as well as in epithelial tissue explants in vitro16. This that EMT-inducing signals can regulate the expres­ fibroblasts, smooth muscle ‘epithelial–mesenchymal transformation’ process was sion and activity of these EMT-TFs, doing so via both cells, odontoblasts and alternately referred to as ‘epithelial–mesenchymal tran­ transcriptional and post-transcriptional mechanisms. adipocytes. sition’, to distinguish it from the process of neoplastic As our understanding of the inductive signals and Müllerian duct transformation commonly used by the cancer research transcriptional control of EMT evolved, it became A coelomic epithelium-derived community. ‘Epithelial–mesenchymal transition’ became apparent that the activation and execution of EMT does tubular structure present in the term of use after the first TEMTIA meeting, which not require changes in DNA sequence and can be revers­ early reproductive tract brought the field together in 2003. ible. This made it clear, in turn, that EMT occurs as a development. Müllerian duct gives rise to the oviduct, Many observations originating more than a quar­ result of complex epigenetic regulatory programmes, uterus, cervix and part of the ter century ago described EMT as being induced by a much like those operating at different stages of devel­ vagina in the female embryo. diverse array of contextual signals. For example, cul­ opment. During development, some cell popu­lations It degenerates in the male tured amnion cells were ‘transformed’ from an epithe­ may undergo multiple rounds of EMT and MET, indi­ embryo through the action of lial phenotype into fibroblast-like cells in response to a cating substantial phenotypic plasticity. For example, anti-Müllerian hormone. leukocyte medium17. Endocardial cells underwent EMT during renal morphogenesis, the epithelial cells lining Renal vesicles in response to signals from adjacent cardiac muscle18. renal vesicles are derived from renal mesenchymal Epithelialized vesicles derived Hepatocyte growth factor (HGF) was found to induce cells via MET, while these mesenchymal cells in turn from the metanephric transformation of epithelial cells into migratory fibro­ are descendants of epithelial cells in the epiblast via mesenchyme during kidney development. Each renal blasts19. Fibroblast growth factor 1 (FGF1) induced an EMT33. During somite formation, paraxial mesenchyme vesicle gives rise to one ‘epithelial plasticity’ response in bladder carcinoma10, cells undergo MET to form epithelial somites, which then nephron and, together with connecting EMT to cancer. TGFβ, overexpressed in undergo EMT to give rise to the sclerotome34. Likewise, vascular-derived glomerulus cancers and required for cardiac EMT20, was found to during the pathogenesis of cancers and fibrosis, EMT is and ureteric bud-derived be a potent inducer of EMT in cultured cells11. These activated to various degrees (from partial to fully) and collecting duct, forms the basic unit of renal filtration. observations provided the first indications that diverse is often reversible, revealing a plasticity that can yield extracellular signals, including soluble factors and cells residing in a spectrum of states, between a fully epi­ Epiblast components of the extracellular matrix (ECM), could thelial phenotype and a fully mesenchymal phenotype3 A population of pluripotent act together to evoke EMT programmes in responding (considered as the end points of EMT). Thus, EMT does cells during early mammalian, reptilian and avian development. epithelial cells. not result in a single mesenchymal state, but rather in a The epiblast undergoes These early descriptive studies shed little light on variety of intermediate states with various degrees of epi­ epithelialization, initiates the mechanisms operating within individual cells that thelial and mesenchymal features. This finding presents gastrulation morphogenesis enable the induction of EMT. Discoveries in the field a major challenge to the EMT research community: how and gives rise to the three of Drosophila melanogaster developmental genetics led best to capture the diversity and plasticity of the EMT principal germ layers (ectoderm, mesoderm and to the identification of master regulators of EMT — for programmes operating in various biological contexts. endoderm). example, the transcription factors Snail and Twist, which act pleiotropically to orchestrate mesoderm formation EMT in development, cancer and fibrosis Somite during gastrulation21. The identification of related tran­ A driving paradigm for the growth of this research field An epithelial vesicle formed through periodic budding from scription factors in chordates revealed the high degree has been that EMT operates in normal tissues during the mesenchymal pre-somitic of conservation of these factors during metazoan evo­ development and wound healing, but is also a driver in mesoderm. The somite is located lution22 and thus highlighted the importance and rele­ the pathogenesis of cancer and fibrosis. The common between the neural tube and the vance of studying various developmental animal model starting point of diverse EMTs is the downregulation of intermediate mesoderm and systems in order to understand EMT regulation. certain features of the epithelial phenotype. Importantly, gives rise to the sclerotome (axial bones) and dermomyotome Research aimed at identifying molecular regulators of however, although it is now recognized that the EMT (dermis and skeletal muscles). EMT began on a large scale in the 1990s. For example, programmes do not operate as binary switches that shunt identification of the Snail-related transcription factor cells from fully epithelial to fully mesenchymal extremes, Sclerotome Slug (also known as Snai2) as an inducer of EMT dur­ it remains unclear whether discrete phenotypic states are A part of the epithelial somite that undergoes ing chick gastrulation and neural crest cell formation arrayed along the epithelial-to-mesenchymal (E-to-M) EMT first and gives rise illustrated that specific transcription factors can act as phenotypic spectrum or, alternatively, a continuum of to the vertebrae and ribs. key upstream regulators of EMT23. The finding that such states exist that lack distinct, definable boundaries. nATuRe RevIeWS | MolECulAr CEll BioloGy volume 21 | June 2020 | 345 C o n S e n S u S S tat e m e n t The extent to which such intermediates represent sta­ be able to perform collective migration locally without ble states in specific biological contexts is also unclear. activating EMT, possibly using collective migration A continuum of EMT intermediate states might enable mechanisms similar to those used during development. rapid interconversion between cells possessing various However, it is unclear whether primary carcinoma cells combinations of these traits, a process viewed as hav­ can complete the entire process of metastatic dissemina­ ing high phenotypic plasticity. Moreover, it is possible tion without activating, at least transiently, components that the phenotypic states between the fully epithelial of the EMT programme. The behaviour of carcinoma and fully mesenchymal end points might not be arrayed cells that transition to intermediate epithelial/mesenchy­ along a linear spectrum, and that multiple alternative mal states (E/M states) (that is, partial EMTs) echoes the paths can operate to enable an epithelial cell to advance behaviour of epithelial cells during normal development. towards a mesenchymal state. Finally, the cells activating Cancer cells proceed through a gradation of phenotypic EMT programmes in adult tissues under pathological states, each associated with combinations of epithelial conditions commonly express combinations of epithe­ and mesenchymal markers3,51,52. lial and mesenchymal markers and rarely complete the The activation of alternative EMT programmes entire EMT programme, suggesting that ‘partial EMTs’ and the progression of individual cells to different states represent the norm rather than the exception. along the E-to-M spectrum can generate extensive pheno­typic heterogeneity within tumours. Supporting Development. During animal development, cells of epi­ this notion, multiple E/M cell subpopulations with dis­ thelial origin often migrate long distances from their tinct chromatin landscapes and gene expression signa­ original position to their final destination. The exit and tures have been reported in skin and mammary primary detachment from the epithelial sheet and the following tumours, and these subpopulations are often spatially migration to distant locations depend on cells acquiring localized within specific areas of a tumour53. Moreover, a mesenchymal state through EMT. Moreover, in many hybrid E/M states are enriched in circulating tumour cells cases, the cells later undergo a permanent or temporary (CTCs) that are released by primary breast and lung can­ reversal to an epithelial state through MET (for example, cers and their metastases54,55, ostensibly reflecting the cel­ Tight junctions endoderm cells) or switch to a different state (for example, lular heterogeneity seen within the originating primary Intercellular barriers formed neural crest cells). A great degree of morphological vari­ tumours. Such phenotypic plasticity and heterogeneity between epithelial cells. Such ability is associated with epithelial cells that participate may provide cancer cells with increased adaptability junctions regulate transepithelial particle transport and prevent in developmental EMTs, ranging from cells possessing and resistance, enabling them to respond to a variety of the free diffusion of cell fully formed epithelial cell–cell junctions (including external cues and physiological stresses3,49,51,52,56,57. Thus, membrane proteins between tight junctions, adherens junctions and/or desmosomes) because tumour cells encounter diverse microenviron­ the apical and basolateral and an underlying basement membrane (to which they ments as they navigate the multiple steps of the metastatic domains. adhere through hemidesmosomes), such as the pluri­ cascade (Box 1) and migrate through and reach different Adherens junctions potent epiblast cells of amniotes35, to the primitive epithe­ tissues, various hybrid E/M phenotypes may provide a A type of cell–cell junction in lial cells giving rise to mesendoderm in Xenopus laevis survival advantage in these distinct environments, such which the plasma membranes and zebrafish, which exhibit only apical–basal polarity and as blood and lymphatic vessels and primary and second­ of apposing cells adhere to incompletely assembled cell–cell junctions36. ary tumour sites. The tissue of origin of the tumour cell, each other through cadherin- mediated homophilic In many cases, the quasi-mesenchymal state is not specific combinations of expressed EMT-TFs, and chro­ interactions. Adherens junctions reached through a complete loss of cell–cell junctions, matin modifications may also determine the phenotypic associate with condensed actin but instead by changes in the nature and dynamics heterogeneity of these various hybrid E/M states. filaments intracellularly. of junction formation and dissolution, which may The diversity of EMT-associated cancer cell pheno­ Desmosomes explain how cells with mesenchymal characteristics can types is reflected in the discrepancies in experimental A type of cell–cell junction exhibit collec­tive cell migration37 — that is, the migra­ and histopathological observations of human tumours that provides strong adhesion tion of cohorts of cells that seem to be held together that have fueled a long-standing debate regarding the between cells and resists by various types of cell–cell junctions. Such plastic, roles of EMT in cancer progression4,5,58–60. Such discrep­ mechanical stress. Desmosomes quasi-mesenchymal phenotypes are observed in cells ancies can often be attributed to the use of different interact with intermediate filaments intracellularly. that migrate collectively and are held together partially by EMT markers and the analysis of particular EMT-TFs cadherin-based cell–cell contacts, in endoderm and meso­ as markers of this programme; in addition, the EMT Hemidesmosomes derm cells of D. melanogaster, zebrafish, X. laevis and programmes operating in different tissues might differ A type of cell–matrix mouse38–43, and in neural crest cells of zebrafish, chick from one another. For instance, Snai1 and Twist1 have attachment structure mediated by integrin– and X. laevis44,45. It is important to note that not every both been shown to be important for metastasis in the matrix interactions migratory process employed by epithelial cells involves PyMT-driven breast cancer model61,62, but dispensable for extra­cellularly and EMT, as is the case in chicken epiblast morphogenesis metastasis in a pancreatic cancer model7, which instead intermediate filaments before the formation of the primitive streak46. depends on Zeb1 (ref.63). Further confusion has come intracellularly. from the observation that carcinoma cells undergoing PyMT Cancer. During the multistep progression of carcino­ a partial EMT can reduce their epithelial phenotype A breast cancer metastasis mas that are initially benign, epithelial cells acquire a through post-translational mechanisms64, making it model in which the few distinctly mesenchymal traits that confer to them challenging to interpret studies that rely solely on the polymavirus middle the ability to invade adjacent tissues, locally, and then to perturbation of transcriptomes by EMT-TFs. These and T-antigen is overexpressed in mammary epithelial cells disseminate to distant tissues. Much of this phenotypic other examples indicate that the versions of EMT pro­ as a driver of breast tumour progression towards increased invasiveness depends grammes and the functions of the involved EMT-TFs development. on the activation of EMT3,47–51. Carcinoma cells might are tissue context dependent. Moreover, as is often 346 | June 2020 | volume 21 www.nature.com/nrm C o n S e n S u S S tat e m e n t Box 1 | The metastasis cascade The metastatic process is thought to consist of the following sequential steps (see the figure). Initial escape from the primary site (invasion) requires that the epithelial tumour cells loosen their cell–cell junctions to become motile (step 1) and that they degrade the basement membrane and extracellular matrix (ECM); breakdown of these physical barriers allows cells to migrate and invade into nearby normal tissues (step 2). The next step of metastasis is termed ‘intravasation’, during which tumour cells invade across the endothelial lamina, penetrate into the vascular or lymphatic vessels and thereby enter the systemic circulation (step 3). Only a small number of the released neoplastic cells appear to be capable of surviving attack by the shear forces encountered in the circulation and by anoikis (a form of programmed cell death that occurs when cells detach from the surrounding ECM) provoked by the loss of anchorage to solid ECM. Eventually, some of the survivors may arrest in and extravasate through the capillary endothelium of distant organs into the parenchyma of these organs (extravasation) (step 4). In the new stromal environment that they encounter, an even smaller subset succeed in establishing themselves (step 5) and in proliferating from minute growths (micrometastases) into fully malignant secondary tumours that are clinically detectable and eventually life-threatening (secondary growth) (step 6). The activation of epithelial–mesenchymal transition (EMT) can provide tumour cells with the ability to migrate, invade, intravasate and extravasate. Once they reach distant organs, these mesenchymal cells revert to a more epithelial identity via mesenchymal–epithelial transition (MET) in order to regain proliferating ability, to form secondary growth in distant organs. Figure adapted from ref.93, CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Primary tumour EMT 1 2 Dissociation, Intravasation Extravasation migration, invasion 3 4 Circulating Blood vessel tumour cells 5 Tumour dormancy 6 MET and Epithelial cell state macrometastasis Mesenchymal cell state formation seen during the course of embryonic development, of fibrosis-associated myofibroblasts68. More recent cancer-associated EMT is only activated partially and studies have shown that renal epithelial cells undergo transiently, making end-stage markers of a fully mesen­ a partial EMT that is crucial for disease progression, but chymal state uninformative in cancer studies. Another that they do not directly contribute to the formation complicating factor is that EMT programmes have been of the myofibroblast population68,69. Instead, they lose linked to additional traits that are not associated with their normal tubular function, and these damaged cells canonical EMT regulation, such as stemness, cell sur­ release paracrine signals to the renal interstitium, reshap­ vival rate, metabolic changes and, in the case of cancer ing the microenvironment. The release of TGFβ converts cells, resistance to anticancer therapeutic drugs65,66. existing fibroblasts into myofibroblasts, and the secre­ tion of additional cytokines and chemokines probably Fibrosis. EMT has also been observed to occur and recruits macrophages to the stroma. Hence, damaged play a role in diverse types of fibrosis (including in renal epithelial cells promote both fibrogenesis and Interstitial fibrosis the lung, liver and kidney), with EMT-TF expression inflammation, which are hallmarks of renal fibrosis69,70. Excessive deposition of shown to be a prerequisite for fibrosis development in While the debate concerning the contribution of EMT extracellular matrix by fibroblasts in interstitial space, mouse models3. As in cancer progression, the role of to different types of fibrosis continues, the demonstrated leading to tissue hardening and EMT in organ fibrosis has been the subject of active requirement of EMT-TF expression strongly suggests scarring. debate. A central issue in this debate is the origin of that the activation of EMT is indeed required for the the myofibroblasts that accumulate in fibrotic tissues. development of several types of fibrosis. Renal interstitium These cells represent a specialized fibroblast population The renal parenchymal space that is outside the tubular, involved in collagen secretion and thus in the develop­ Definitions of EMT and its associated terms glomerular and vascular ment and progression of interstitial fibrosis, which is the To facilitate the investigation of multifaceted EMT pro­ structures. major cause of the disease in different tissues. Early cesses and discussion among diverse groups of research­ lineage-tracing studies supported the hypothesis that ers studying EMT, we propose the following definitions of Fibrogenesis The phenomenon of excessive myofibroblasts arise from EMT-driven conversion67, but EMT and its associated terms to stand as a reference. We production of extracellular subsequent lineage-tracing analyses have not provided encourage researchers to adhere to this recommended matrix by fibroblasts. compelling evidence of epithelial cells as precursors nomenclature. nATuRe RevIeWS | MolECulAr CEll BioloGy volume 21 | June 2020 | 347 C o n S e n S u S S tat e m e n t EMT. A multifaceted and often reversible change in E/M phenotypic states arrayed along the epithelial– cellular phenotypes during which epithelial cells lose mesenchymal spectrum that cannot be easily distinguished their apical–basal polarity, modulate their cytoskeleton on the basis of our current understanding. This plasticity and exhibit reduced cell–cell adhesive properties. Cells has been variably referred to as partial EMT, hybrid E/M may individually or collectively acquire mesenchy­ status, a metastable EMT state, EMT continuum and mal features and increase motility and invasive ability. EMT spectrum; in all cases, the cells express a mixture of Typically, a switch in intermediate filament usage from epithelial features (such as cytokeratins) and mesenchy­ cytokeratins to vimentin is observed after a complete mal features (such as cell migration) and markers. EMP EMT. Cortical actin filament in epithelial cells also indicates an ability to move readily between these various undergoes marked rearrangement during EMT. While states, although the stability of the various states varies the characteristics of fully epithelial cells are relatively in different biological contexts. EMP is widely observed clearly defined, our current knowledge does not allow in development, wound healing and cancer. In addition us to define the mesenchymal state with specific cellular to a mesenchymal type of migration, as observed during characteristic or molecular markers that are universal mesoderm formation, EMP can also participate in collec­ end-products of all EMT programmes. tive migration — for example, during tubulogenesis and wound healing. EMP also accounts for the reversibility MET. Reciprocal changes in cellular phenotype that of the EMT programme. Epithelial cells going through reverse EMT-induced phenotypes, during which EMT give rise to cell populations that may enter reversi­ mesenchymal-like cells may acquire apical–basal polar­ bly into states with various proportions of epithelial and ity, reorganize their cytoskeleton, and exhibit increased mesenchymal features. EMP is thought to provide cells cell–cell adhesion, resulting in an organized epithelium. with the fitness and flexibility to fulfil the diverse require­ MET occurs during embryonic development (for exam­ ments during the course of either developmental or ple, cardiac development, kidney morphogenesis and pathological processes. somite formation) and cancer. EMT-TFs. In many if not most settings, both in cell Endothelial–mesenchymal transition. As with epithelial culture and in vivo, EMP involves some degree of tran­ cells, endothelial integrity depends on cell–cell junctions, scriptional regulation. Several transcription factors apical–basal polarity and interactions with an underly­ belonging to the Snail, Twist and Zeb families have ing basement membrane. Endothelial–mesenchymal been found to control cell–cell adhesion, cell migra­ transition (EndoMT) more accurately indicates the phe­ tion and ECM degradation, and to play evolutionarily nomenon in such cell populations and resembles EMT conserved central roles in the execution of EMT in in most aspects, except for the replacement of E-cadherin various biological settings and organisms (Table 1). Of by VE-cadherin. EndoMT thereby enables endothelial note, all the developmental EMT processes described cells to attenuate or deconstruct their functional integrity to date involve at least one member of these fami­ and apical–basal polarity, to acquire motile and invasive lies of core EMT-TFs. Other transcription factors behaviour and to activate changes in gene expression that have been shown to impact EMT in certain contexts are driven by certain EMT-TFs. Similar to epithelial cells in (Supplementary Table 1). However, these transcrip­ EMT, endothelial cells that have activated EndoMT pro­ tion factors are also involved in other cellular processes grammes exhibit a variety of intermediate or partial pheno­ (for example, proliferation, apoptosis or stemness). types, as discussed above for EMT. EndoMT was described In addition, many of the EMT-TFs are also expressed in initially during embryonic heart development71 and non-epithelial cells, ranging from fibroblasts to haema­ subsequently in the context of cardiac fibrosis72. topoietic precursors, and in cancer types involving non-epithelial derivatives (melanoma, glioblastoma Epithelial–mesenchymal plasticity. We favour and rec­ and leukaemia), where they play important roles during ommend use of the term ‘epithelial–mesenchymal plas­ tumour progression, often beyond classic EMT phases. ticity’ (EMP) to describe the ability of cells to adopt mixed Although we use the term EMT-TFs to describe all tran­ E/M features and to interconvert between intermediate scription factors associated with EMT, it is important to Table 1 | Core EMT transcription factors, with key studies reporting their discovery Transcription Type Development Cancer Fibrosis factor Snai1 (Snail) Zinc finger Boulay et al., 1987 (ref.84); Batlle et al., 2000 (ref.25); Boutet et al., 2006 (ref.86) Nieto et al., 1992 (ref.85) Cano et al., 2000 (ref.26) Snai2 (Slug) Zinc finger Nieto et al., 1994 (ref.23) Savagner et al., 1997 (ref.24) — Zeb1 Zinc finger Funahashi et al., 1993 (ref. ) 87 Grooteclaes and Frisch, Oba et al., 2010 (ref.89) 2000 (ref.88) Zeb2 (SIP1) Zinc finger Verschueren et al., 1999 (ref.90) Comijn et al., 2001 (ref.31) Oba et al., 2010 (ref.89) Twist1 bHLH Thisse et al., 1988 (ref. ) 91 Yang et al., 2004 (ref. ) 30 Kida et al., 2007 (ref.92); Lovisa et al., 2015 (ref.70) EMT, epithelial–mesenchymal transition. 348 | June 2020 | volume 21 www.nature.com/nrm C o n S e n S u S S tat e m e n t keep in mind that their expression alone is not sufficient manifestations, of the epithelial phenotype. Epithelial to indicate that EMT is occurring. cells harbour complexes that mediate cell–cell interac­ tions, most notably adherens junctions, tight junctions Recommendations on the criteria to define EMT and desmosomes (Fig. 1). Apical–basal polarity guides the In the current EMT literature, both cellular and molec­ proper organization of tight junctions, adherens junctions ular descriptors have been used to define EMT in var­ and desmosomes in epithelial cells. Polarity complexes, ious biological systems. Below we provide several key including the Par, Crumbs and Scribble complexes75–77, recommendations on how to better use this informa­ define the apical-lateral and basal-lateral domains of epi­ tion to cover the enormous complexity and plasticity thelial cells (Fig. 1). During the early phase of EMT, loss of of the EMT programme in diverse developmental and apical–basal polarity is often the first event to be observed pathological settings. and can lead to the destabilization of adhesion complexes, such as the tight junctions and adherens junctions at EMT status cannot be assessed on the basis of one or a the lateral membrane78,79, as well as to the activation of small number of molecular markers. EMT constitutes EMT-TFs80. The decrease or loss of epithelial adherens changes of cell behaviour that involve the loss of certain junctions and desmo­somes occurs via transcriptional epithelial characteristics and the gain of certain mes­ repression by the core EMT-TFs of the genes encod­ enchymal traits. The complex series of cellular changes ing junctional proteins. The cytoplasmic relocalization occurring during EMT require the cooperation of a of adherens junction proteins, such as E-cadherin, via large number of molecular factors. On the basis of their post-transcriptional regulation is also an early feature of involvement in the process, these factors can be divided EMT initiation in various EMT models81. into three groups: EMT-inducing signals, EMT-TFs Another key function of the EMT programmes is to and EMT markers that define and constitute various provide stationary epithelial cells with the ability epithelial and mesenchymal cell characteristics. In the to migrate by invading through extracellular matri­ literature, diverse cellular and molecular descriptors ces secreted by both epithelial and mesenchymal cells. have been used to define EMT in different biological sys­ Thus, during EMT, epithelial cells often need to breach tems, which has been a major source of confusion. For the basement membrane in order to migrate away from example, some studies define partial loss of E-cadherin their epithelium of origin35,82. Migration of cells that have as an indication of EMT, while others argue that the undergone EMT does not necessarily require the cells to maintenance of certain levels of expression of epithelial lose all epithelial features, and a switch of intermediate markers such as cytokeratins is indicative of cells not filaments from cytokeratin to vimentin can facilitate cell having undergone EMT. Given the complex manifesta­ migration. Depending on the extent of cell–cell adhe­ tions of the EMT programme, it has become clear that sion loss, epithelial cells can migrate as single cells, in inferring the involvement of EMT in any process cannot a mesenchymal manner, or collectively, while remain­ rely solely on a few salient molecular markers, such as ing attached with one another via weakened but still E-cadherin and vimentin73. operative cell–cell interactions. This indicates that the More importantly, the use of various EMT molec­ breakdown of these tightly regulated epithelial struc­ ular markers to characterize the phenotypic state of tures, the gain of motility and the ability to degrade ECM individual tumour cells has revealed that such cells, as during EMT are inaccurately represented by the simple described earlier, can simultaneously express both epi­ expression, or lack of expression, of selected markers. thelial and mesenchymal genes. The core EMT-TFs are Furthermore, complex post-translational modifica­ often co-expressed in various combinations in order to tions of key proteins play critical roles in governing the orchestrate complex EMT programmes, and they involve complex cellular processes that occur during EMT. For various members of EMT-TF families, such as Snai these reasons, researchers should describe EMT and (Snai1 and Snai2) and Zeb (Zeb1 and Zeb2), depend­ MET as functional changes in the biological proper­ ing on the specific biological context74. Importantly, ties of cells rather than focusing largely on changes in a post-transcriptional regulation of EMT regulators at few readily monitored molecular markers. It is through both the mRNA and protein levels is crucial in con­ this perspective that our understanding of EMT could trolling EMT. Such regulation is often neglected in faithfully reflect the function of EMT during animal studies that use RNA expression exclusively to survey development and pathological events (Fig. 1). Therefore, EMT molecular markers. A focus on defining EMT whenever it is experimentally feasible, EMT should be programmes exclusively on the basis of the expression assessed through the combination of cellular properties of specific molecular markers such as these under­ and multiple molecular markers. represents the enormous complexity and plasticity of the EMT programmes in diverse developmental and EMT-TFs and other molecular markers are valuable indi- pathological settings. cators of EMT, but they should be assessed in conjunction with changes in cellular characteristics to define EMT. The primary criteria for defining EMT status should The morphological and functional changes that can be be changes in cellular properties together with a set observed in cells during EMT often result from changes of molecular markers, rather than relying solely on in gene expression. Many, but not all, EMT-associated molecular markers. One major feature that unites all changes in gene expression result directly or indirectly the variant EMT programmes is the initial attenuation from the actions of EMT-TFs, which play key roles or deconstruction, to varying degrees and with diverse in driving EMT. Indeed, most EMT programmes are nATuRe RevIeWS | MolECulAr CEll BioloGy volume 21 | June 2020 | 349 C o n S e n S u S S tat e m e n t associated with the activation of expression of one phenotypes are independent of cell differentiation or or several core EMT-TFs. Although core EMT-TFs cell dedifferentiation. Therefore, it is important to note often initiate EMT-associated changes in gene expres­ that the cell-biological definition of EMT strictly refers sion (Table 1), a large number of other EMT-TFs and to cellular features describing the E-to-M phenotypes, numerous microRNAs and long non-coding RNAs while we should keep these associated phenotypes (lncRNAs) have also been shown to contribute to or in mind when examining the role of EMT in various play critical roles in diverse EMTs (Supplementary biological settings. Table 1). Decreased association between β-catenin or p120-catenin and E-cadherin, achieved by Implications for future EMT research post-translational modifications, can also greatly Fifty years and over 38,000 publications after Betty Hay’s weaken the adhesive functions of adherens junctions. pioneering observations, the concept of EMT has now Reduced expression of junctional and polarity proteins been widely applied in biomedical research. It provides is often visible during EMT. Depending on the cell a unifying framework for developmental and cancer type and the extent to which cells advance through an studies, which is evidenced by the exponentially grow­ EMT programme, cells undergoing an EMT may begin ing number of EMT-related publications. Such a frame­ to express vimentin, to suppress cytokeratin, to shift work holds the promise of far-reaching breakthroughs expression of key integrins and so forth. These changes in cancer diagnosis and treatment, for cancer biologists, in gene expression are often seen as being indicative and of bridging the gap in understanding normal and of EMT or as markers of EMT, although, considering pathological epithelial organization and morphogene­ the extensive variations in EMT, their overall value in the sis, for developmental and cell biologists. To realize this diagnosis of EMT needs to be considered with caution. promise, it is desirable that the EMT community reach Beyond this small set of commonalities, it is difficult a consensus on the definition of EMT-related terms and to define other changes as contributing universally to on the conceptual framework for approaching EMT all the diverse manifestations of EMT programmes as a biological process with quantifiable molecular that have been described in the rapidly expanding lite­ descriptors and cellular readouts. All the authors of this rature. Furthermore, our current knowledge does not Consensus Statement article have agreed to adhere to the allow us to know whether there is a linear succession recommendations on nomenclature presented here in of cell-biological changes as cells advance progressively their future research publications and recommend that through an EMT programme, or whether a diverse series other researchers in the EMT and the larger biological of routes radiates in multiple directions from the starting research communities also follow these guidelines. Only point of attenuation or loss of epithelial junctions. by minimizing semantic misinterpretation and data mis­ communication can we begin to appreciate the diversity Finding reliable EMT markers requires a combina- of individual EMTs and uncover conserved themes in torial approach, as well as distinguishing between EMT regulation between development and disease. EMT-associated and non-EMT-associated functions. Studies using cell lines, developmental systems and Analysing the state of a cell that is engaged in EMT cancer models have revealed a diversity of EMT-induced often requires the use of markers that are specific to a phenotypes and have highlighted remarkable complex­ specific biological context. To be able to assess in which ity in the execution and regulation of EMT. Looking position on the EMT spectrum a cell resides, it might forward, to decipher the complexity and plasticity of be necessary to use a set of criteria that are standard­ the EMT programme, we propose that EMT research, ized for the specific biological context. Obtaining quan­ while remaining anchored in traditional developmen­ titative EMT marker measurements should always be tal, cell and cancer biology, should be explored within coupled with cellular and functional analyses of EMT a broader conceptual context. The EMT field has in status, as described above. Importantly, recent studies recent years attracted the interest of a diverse group of have linked EMT to various other cellular programmes researchers with expertise in systems biology, biophysics, and functions, including cancer cell stemness, resistance stem cell biology, pathology and mathematical model­ to apoptosis, genome instability, cancer drug resistance ling. This remarkable strength of interest will enable and metabolic adaptation. Many components of EMT cross-disciplinary collaborations and push this field regulatory pathways, including EMT-TFs, also affect of research forward. We expect that future EMT stud­ other important cellular functions and phenotypes and ies will apply multidisciplinary approaches in order to are themselves regulated through diverse signals that gain increased mechanistic understanding of EMT. One Cancer cell stemness may or may not involve canonical EMT. For example, open question in the cancer EMT field is the extent to The presence of subpopulations EMT-TFs such as Snai1/2 and Twist1 also regulate can­ which the stabilization of specific hybrid E/M states, or of cancer cells that exhibit stem cer cell survival and cancer drug resistance66,83, and it the dynamic switch between E/M states in response to cell-like behaviour. is currently unknown whether cell survival and drug distinct cues from the microenvironment, favours the Dedifferentiation resistance are regulated independently of EMT. Likewise, metastatic process. With many important aspects of The conversion of cells from many ECM remodeling proteins that are important EMT remaining unexplored, advancing EMT research a more differentiated state for breaching the basement membrane can be regu­ will require technological innovations to enable the to a less differentiated one. lated in both EMT-dependent and EMT-independent study of both developmental and cancer-associated Dedifferentiation is often associated with the acquisition manners. Furthermore, although cells may switch to a EMT at the single-cell level. These innovations will of progenitor or stem cell-like different cell fate upon EMT, the EMT-associated cel­ include single-cell live imaging, lineage tracing, gene properties. lular changes from epithelial to more mesenchymal expression analyses and studies of genetic and epigenetic 350 | June 2020 | volume 21 www.nature.com/nrm C o n S e n S u S S tat e m e n t modifications. Finally, a combination of mathematical An increased understanding of the EMT mechanisms modelling with carefully constructed experimental associated with these behaviours offers the poten­ analyses will be important to gaining a mechanistic tial for targeted therapy to prevent cancer metastasis. understanding of EMT plasticity. For example, while the inhibition of EMT-associated Another major challenge is the translation of the cur­ changes might reduce cancer cell dissemination in rent knowledge of EMT heterogeneity and plasticity into early-stage carcinoma, preventing MET in dissemi­ clinical practice. While we are far from understanding nated tumour cells might inhibit metastatic outgrowth the functional implications of EMT heterogeneity, sev­ in distant organs. Experimental and clinical studies eral clinical trials have already incorporated the notion have shown that the development of resistance to vari­ of EMT plasticity3, thereby opening the way for novel ous therapies, including chemotherapies and immuno­ therapies that exploit EMT heterogeneity. Single-cell therapies, is tightly associated with EMT phenotypes83. sequencing of normal tissues, primary tumours, circu­ These studies suggest that targeting EMT, or the cells lating tumour cells and metastases, combined with cellu­ capable of executing it, holds promise in overcoming lar analyses and functional validations, will capture the therapy resistance, which is a major challenge in cancer diversity and plasticity of EMT and have the potential treatment. to reveal the molecular alterations underlying tumour progression and the diverse responses to therapy. Published online 16 April 2020 1. Hay, E. D. in Epithelial–Mesenchymal Interactions: 20. Potts, J. D. & Runyan, R. B. Epithelial–mesenchymal 38. Campbell, K. & Casanova, J. A role for E-cadherin 18th Hahnemann Symposium (eds R. Fleischmajer cell transformation in the embryonic heart can be in ensuring cohesive migration of a heterogeneous & Billingham, R. E.) 31–35 (Williams and Wilkins, mediated, in part, by transforming growth factor beta. population of non-epithelial cells. Nat. Commun. 6, 1968). Dev. Biol. 134, 392–401 (1989). 7998 (2015). 2. Hay, E. D. An overview of epithelio–mesenchymal 21. Leptin, M. twist and snail as positive and negative 39. Montero, J. A. et al. Shield formation at the onset transformation. Acta Anat. 154, 8–20 (1995). regulators during Drosophila mesoderm development. of zebrafish gastrulation. Development 132, 3. Nieto, M. A., Huang, R. Y., Jackson, R. A. & Genes Dev. 5, 1568–1576 (1991). 1187–1198 (2005). Thiery, J. P. EMT: 2016. Cell 166, 21–45 (2016). 22. Nieto, M. A. The snail superfamily of zinc-finger 40. Clark, I. B., Muha, V., Klingseisen, A., Leptin, M. 4. Tarin, D., Thompson, E. W. & Newgreen, D. F. transcription factors. Nat. Rev. Mol. Cell Biol. 3, & Muller, H. A. Fibroblast growth factor signalling The fallacy of epithelial mesenchymal transition in 155–166 (2002). controls successive cell behaviours during mesoderm neoplasia. Cancer Res. 65, 5996–6000 (2005). 23. Nieto, M. A., Sargent, M. G., Wilkinson, D. G. & layer formation in Drosophila. Development 138, 5. Thompson, E. W., Newgreen, D. F. & Tarin, D. Cooke, J. Control of cell behavior during vertebrate 2705–2715 (2011). Carcinoma invasion and metastasis: a role for development by Slug, a zinc finger gene. Science 264, 41. Dumortier, J. G., Martin, S., Meyer, D., Rosa, F. M. epithelial–mesenchymal transition? Cancer Res. 65, 835–839 (1994). & David, N. B. Collective mesendoderm migration 5991–5995 (2005). 24. Savagner, P., Yamada, K. M. & Thiery, J. P. relies on an intrinsic directionality signal transmitted 6. Fischer, K. R. et al. Epithelial-to-mesenchymal The zinc-finger protein Slug causes desmosome through cell contacts. Proc. Natl Acad. Sci. USA 109, transition is not required for lung metastasis dissociation, an initial and necessary step for growth 16945–16950 (2012). but contributes to chemoresistance. Nature 527, factor-induced epithelial–mesenchymal transition. 42. Weber, G. F., Bjerke, M. A. & DeSimone, D. W. 472–476 (2015). J. Cell Biol. 137, 1403–1419 (1997). A mechanoresponsive cadherin–keratin complex 7. Zheng, X. et al. Epithelial-to-mesenchymal transition is 25. Batlle, E. et al. The transcription factor Snail is a directs polarized protrusive behavior and collective dispensable for metastasis but induces chemoresistance repressor of E-cadherin gene expression in epithelial cell migration. Dev. Cell 22, 104–115 (2012). in pancreatic cancer. Nature 527, 525–530 (2015). tumour cells. Nat. Cell Biol. 2, 84–89 (2000). 43. Viotti, M., Nowotschin, S. & Hadjantonakis, A. K. 8. Ye, X. et al. Upholding a role for EMT in breast cancer 26. Cano, A. et al. The transcription factor Snail controls SOX17 links gut endoderm morphogenesis and germ metastasis. Nature 547, E1–E3 (2017). epithelial–mesenchymal transitions by repressing layer segregation. Nat. Cell Biol. 16, 1146–1156 9. Aiello, N. M. et al. Upholding a role for EMT in E-cadherin expression. Nat. Cell Biol. 2, 76–83 (2014). pancreatic cancer metastasis. Nature 547, E7–E8 (2000). 44. Theveneau, E. & Mayor, R. Neural crest delamination (2017). 27. Hajra, K. M., Chen, D. Y. & Fearon, E. R. The SLUG and migration: from epithelium-to-mesenchyme 10. Valles, A. M. et al. Acidic fibroblast growth factor zinc-finger protein represses E-cadherin in breast transition to collective cell migration. Dev. Biol. 366, is a modulator of epithelial plasticity in a rat bladder cancer. Cancer Res. 62, 1613–1618 (2002). 34–54 (2012). carcinoma cell line. Proc. Natl Acad. Sci. USA 87, 28. Bolos, V. et al. The transcription factor Slug represses 45. Scarpa, E. et al. Cadherin switch during EMT in neural 1124–1128 (1990). E-cadherin expression and induces epithelial to crest cells leads to contact inhibition of locomotion via 11. Miettinen, P. J., Ebner, R., Lopez, A. R. & Derynck, R. mesenchymal transitions: a comparison with Snail and repolarization of forces. Dev. Cell 34, 421–434 (2015). TGF-β induced transdifferentiation of mammary E47 repressors. J. Cell Sci. 116, 499–511 (2003). 46. Rozbicki, E. et al. Myosin-II-mediated cell shape epithelial cells to mesenchymal cells: involvement of 29. Perez-Moreno, M. A. et al. A new role for E12/E47 changes and cell intercalation contribute to primitive type I receptors. J. Cell Biol. 127, 2021–2036 (1994). in the repression of E-cadherin expression and streak formation. Nat. Cell Biol. 17, 397–408 (2015). 12. Newgreen, D. F., Ritterman, M. & Peters, E. A. epithelial–mesenchymal transitions. J. Biol. Chem. 47. Ocana, O. H. et al. Metastatic colonization requires the Morphology and behaviour of neural crest cells of 276, 27424–27431 (2001). repression of the epithelial–mesenchymal transition chick embryo in vitro. Cell Tissue Res. 203, 115–140 30. Yang, J. et al. Twist, a master regulator of inducer Prrx1. Cancer Cell 22, 709–724 (2012). (1979). morphogenesis, plays an essential role in tumor 48. Tsai, J. H., Donaher, J. L., Murphy, D. A., Chau, S. 13. Thiery, J. P., Duband, J. L., Rutishauser, U. & metastasis. Cell 117, 927–939 (2004). & Yang, J. Spatiotemporal regulation of epithelial– Edelman, G. M. Cell adhesion molecules in early 31. Comijn, J. et al. The two-handed E box binding zinc mesenchymal transition is essential for squamous chicken embryogenesis. Proc. Natl Acad. Sci. USA 79, finger protein SIP1 downregulates E-cadherin and cell carcinoma metastasis. Cancer Cell 22, 725–736 6737–6741 (1982). induces invasion. Mol. Cell 7, 1267–1278 (2001). (2012). 14. Markwald, R. R., Fitzharris, T. P. & Manasek, F. J. 32. Eger, A. et al. DeltaEF1 is a transcriptional repressor 49. Tsai, J. H. & Yang, J. Epithelial–mesenchymal plasticity Structural development of endocardial cushions. of E-cadherin and regulates epithelial plasticity in in carcinoma metastasis. Genes Dev. 27, 2192–2206 Am. J. Anat. 148, 85–119 (1977). breast cancer cells. Oncogene 24, 2375–2385 (2013). 15. Trelstad, R. L., Hayashi, A., Hayashi, K. & Donahoe, P. K. (2005). 50. Brabletz, T. To differentiate or not—routes towards The epithelial–mesenchymal interface of the male rate 33. Stark, K., Vainio, S., Vassileva, G. & McMahon, A. P. metastasis. Nat. Rev. Cancer 12, 425–436 (2012). Mullerian duct: loss of basement membrane integrity Epithelial transformation of metanephric mesenchyme 51. Chaffer, C. L., San Juan, B. P., Lim, E. & Weinberg, R. A. and ductal regression. Dev. Biol. 92, 27–40 (1982). in the developing kidney regulated by Wnt-4. Nature EMT, cell plasticity and metastasis. Cancer Metastasis 16. Greenburg, G. & Hay, E. D. Epithelia suspended 372, 679–683 (1994). Rev. 35, 645–654 (2016). in collagen gels can lose polarity and express 34. Christ, B. & Ordahl, C. P. Early stages of chick somite 52. Lambert, A. W., Pattabiraman, D. R. & Weinberg, R. A. characteristics of migrating mesenchymal cells. development. Anat. Embryol. 191, 381–396 (1995). Emerging biological principles of metastasis. Cell 168, J. Cell Biol. 95, 333–339 (1982). 35. Nakaya, Y., Sukowati, E. W., Wu, Y. & Sheng, G. RhoA 670–691 (2017). 17. Nahmias, A. J., Cooperband, S. R., Green, J. A. & and microtubule dynamics control cell–basement 53. Pastushenko, I. et al. Identification of the tumour Kibrick, S. Transformation of epithelial cells in vitro. membrane interaction in EMT during gastrulation. transition states occurring during EMT. Nature 556, Nature 216, 1349–1350 (1967). Nat. Cell Biol. 10, 765–775 (2008). 463–468 (2018). 18. Runyan, R. B. & Markwald, R. R. Invasion of 36. Shook, D. & Keller, R. Mechanisms, mechanics and 54. Yu, M. et al. Circulating breast tumor cells exhibit mesenchyme into three-dimensional collagen gels: function of epithelial–mesenchymal transitions in dynamic changes in epithelial and mesenchymal a regional and temporal analysis of interaction in early development. Mech. Dev. 120, 1351–1383 composition. Science 339, 580–584 (2013). embryonic heart tissue. Dev. Biol. 95, 108–114 (1983). (2003). 55. Schliekelman, M. J. et al. Molecular portraits of 19. Stoker, M. & Perryman, M. An epithelial scatter 37. Mayor, R. & Etienne-Manneville, S. The front and rear epithelial, mesenchymal and hybrid states in lung factor released by embryo fibroblasts. J. Cell Sci. 77, of collective cell migration. Nat. Rev. Mol. Cell Biol. adenocarcinoma and their relevance to survival. 209–223 (1985). 17, 97–109 (2016). Cancer Res. 75, 1789–1800 (2015). nATuRe RevIeWS | MolECulAr CEll BioloGy volume 21 | June 2020 | 351 C o n S e n S u S S tat e m e n t 56. Diepenbruck, M. & Christofori, G. Epithelial– 74. Stemmler, M. P., Eccles, R. L., Brabletz, S. & and binds to 5′-CACCT sequences in candidate target mesenchymal transition (EMT) and metastasis: Brabletz, T. Non-redundant functions of EMT genes. J. Biol. Chem. 274, 20489–20498 (1999). yes, no, maybe? Curr. Opin. Cell Biol. 43, 7–13 transcription factors. Nat. Cell Biol. 21, 102–112 91. Thisse, B., Stoetzel, C., Gorostiza-Thisse, C. & (2016). (2019). Perrin-Schmitt, F. Sequence of the twist gene and 57. Santamaria, P. G., Moreno-Bueno, G., Portillo, F. & 75. Assemat, E., Bazellieres, E., Pallesi-Pocachard, E., nuclear localization of its protein in endomesodermal Cano, A. EMT: present and future in clinical oncology. Le Bivic, A. & Massey-Harroche, D. Polarity complex cells of early Drosophila embryos. EMBO J. 7, Mol. Oncol. 11, 718–738 (2017). proteins. Biochim. Biophys. Acta 1778, 614–630 2175–2183 (1988). 58. Ledford, H. Cancer theory faces doubts. Nature 472, (2008). 92. Kida, Y., Asahina, K., Teraoka, H., Gitelman, I. & 273 (2011). 76. Lee, M. & Vasioukhin, V. Cell polarity and cancer–cell Sato, T. Twist relates to tubular epithelial–mesenchymal 59. Maheswaran, S. & Haber, D. A. Transition loses its and tissue polarity as a non-canonical tumor transition and interstitial fibrogenesis in the obstructed invasive edge. Nature 527, 452–453 (2015). suppressor. J. Cell. Sci. 121, 1141–50 (2008). kidney. J. Histochem. Cytochem. 55, 661–673 (2007). 60. Ruben, B. & Gerhard, C. The relevance of EMT in 77. Ngok, S. P., Lin, W. H. & Anastasiadis, P. Z. 93. Yeung, K. T. & Yang, J. Epithelial–mesenchymal breast cancer metastasis: correlation or causality? Establishment of epithelial polarity–GEF who’s transition in tumor metastasis. Mol. Oncol. 11, FEBS Lett. 589, 1577–1587 (2015). minding the GAP? J. Cell Sci. 127, 3205–3215 28–39 (2017). 61. Tran, H. D. et al. Transient SNAIL1 expression is (2014). necessary for metastatic competence in breast 78. Barrios-Rodiles, M. et al. High-throughput mapping Acknowledgements cancer. Cancer Res. 74, 6330–6340 (2014). of a dynamic signaling network in mammalian cells. We sincerely apologize to the many researchers whose work we 62. Xu, Y. et al. Breast tumor cell-specific knockout of Science 307, 1621–1625 (2005). were unable to cite due to space restrictions. Twist1 inhibits cancer cell plasticity, dissemination, 79. Ozdamar, B. et al. Regulation of the polarity protein and lung metastasis in mice. Proc. Natl Acad. Sci. USA Par6 by TGFβ receptors controls epithelial cell Author contributions 114, 11494–11499 (2017). plasticity. Science 307, 1603–1609 (2005). All authors contributed to the discussion of content and to the 63. Krebs, A. M. et al. The EMT-activator Zeb1 is a key 80. Jung, H. Y. et al. Apical–basal polarity inhibits writing and/or revision of the article. factor for cell plasticity and promotes metastasis in epithelial–mesenchymal transition and tumour pancreatic cancer. Nat. Cell Biol. 19, 518–529 metastasis by PAR-complex-mediated SNAI1 Competing interests (2017). degradation. Nat. Cell Biol. 21, 359–371 (2019). The authors declare no competing interests. 64. Aiello, N. M. et al. EMT subtype influences epithelial 81. Janda, E. et al. Raf plus TGFβ-dependent EMT is plasticity and mode of cell migration. Dev. Cell 45, initiated by endocytosis and lysosomal degradation Peer review information 681–695 (2018). of E-cadherin. Oncogene 25, 7117–7130 (2006). Nature Reviews Molecular Cell Biology thanks P. Howe and 65. Goossens, S., Vandamme, N., Van Vlierberghe, P. 82. Eckert, M. A. et al. Twist1-induced invadopodia the other, anonymous, reviewer(s) for their contribution to the & Berx, G. EMT transcription factors in cancer formation promotes tumor metastasis. Cancer Cell 19, peer review of this work. development re-evaluated: beyond EMT and MET. 372–386 (2011). Biochim. Biophys. Acta Rev. Cancer 1868, 584–591 83. Dongre, A. & Weinberg, R. A. New insights into the Publisher’s note (2017). mechanisms of epithelial–mesenchymal transition and Springer Nature remains neutral with regard to jurisdictional 66. Puisieux, A., Brabletz, T. & Caramel, J. Oncogenic implications for cancer. Nat. Rev. Mol. Cell Biol. 20, claims in published maps and institutional affiliations. roles of EMT-inducing transcription factors. Nat. Cell 69–84 (2019). Biol. 16, 488–494 (2014). 84. Boulay, J. L., Dennefeld, C. & Alberga, A. The Supplementary information 67. Iwano, M. et al. Evidence that fibroblasts derive from Drosophila developmental gene snail encodes a Supplementary information is available for this paper at epithelium during tissue fibrosis. J. Clin. Invest. 110, protein with nucleic acid binding fingers. Nature 330, https://doi.org/10.1038/s41580-020-0237-9. 341–350 (2002). 395–398 (1987). 68. LeBleu, V. S. et al. Origin and function of 85. Nieto, M. A., Bennett, M. F., Sargent, M. G. & Open Access This article is licensed under myofibroblasts in kidney fibrosis. Nat. Med. 19, Wilkinson, D. G. Cloning and developmental a Creative Commons Attribution 4.0 1047–1053 (2013). expression of Sna, a murine homologue of the International License, which permits use, 69. Grande, M. T. et al. Snail1-induced partial Drosophila snail gene. Development 116, 227–237 sharing, adaptation, distribution and reproduction in any epithelial-to-mesenchymal transition drives renal (1992). medium or format, as long as you give appropriate credit to fibrosis in mice and can be targeted to reverse 86. Boutet, A. et al. Snail activation disrupts tissue the original author(s) and the source, provide a link to the established disease. Nat. Med. 21, 989–997 homeostasis and induces fibrosis in the adult kidney. Creative Commons licence, and indicate if changes were (2015). EMBO J. 25, 5603–5613 (2006). made. The images or other third party material in this article 70. Lovisa, S. et al. Epithelial-to-mesenchymal transition are included in the article’s Creative Commons licence, unless 87. Funahashi, J., Sekido, R., Murai, K., Kamachi, Y. & induces cell cycle arrest and parenchymal damage in indicated otherwise in a credit line to the material. If material Kondoh, H. Delta-crystallin enhancer binding protein is not included in the article’s Creative Commons licence and renal fibrosis. Nat. Med. 21, 998–1009 (2015). delta EF1 is a zinc finger-homeodomain protein your intended use is not permitted by statutory regulation or 71. Boyer, A. S. et al. TGFβ2 and TGFβ3 have separate and implicated in postgastrulation embryogenesis. exceeds the permitted use, you will need to obtain permission sequential activities during epithelial–mesenchymal Development 119, 433–446 (1993). directly from the copyright holder. To view a copy of this cell transformation in the embryonic heart. Dev. Biol. 88. Grooteclaes, M. L. & Frisch, S. M. Evidence for a licence, visit http://creativecommons.org/licenses/by/4.0/. 208, 530–545 (1999). function of CtBP in epithelial gene regulation and 72. Zeisberg, E. M. et al. Endothelial-to-mesenchymal anoikis. Oncogene 19, 3823–3828 (2000). transition contributes to cardiac fibrosis. Nat. Med. 89. Oba, S. et al. miR-200b precursor can ameliorate Related links 13, 952–961 (2007). renal tubulointerstitial fibrosis. PLoS One 5, e13614 temtIa: https://temtia.org 73. Savagner, P. Epithelial–mesenchymal transitions: from (2010). cell plasticity to concept elasticity. Curr. Top. Dev. Biol. 90. Verschueren, K. et al. SIP1, a novel zinc finger/ 112, 273–300 (2015). homeodomain repressor, interacts with Smad proteins © The Author(s) 2020 352 | June 2020 | volume 21 www.nature.com/nrm