Review

Epithelial Plasticity: A Common Theme in Embryonic and Cancer Cells

See allHide authors and affiliations

Science  08 Nov 2013:
Vol. 342, Issue 6159, 1234850
DOI: 10.1126/science.1234850

Structured Abstract

Background

During embryonic development, cells often travel long distances to form tissues and organs. To be able to migrate, embryonic cells undergo a process known as epithelial-to-mesenchymal transition (EMT). Once migratory embryonic cells reach their destination, they undergo the reverse process, mesenchymal-to-epithelial transition (MET), to later differentiate into multiple cell types. This reveals a high degree of cell plasticity, referring to the ability of cells to reversibly change phenotype, a common feature of embryonic cells. Research indicates that the EMT program is reactivated in cancer cells in the delamination from a primary tumor, the first step toward the colonization of distant organs to form secondary tumors (metastasis). The dissemination of cancer cells and the subsequent formation of metastasis are responsible for the vast majority of cancer-associated deaths. As in EMT, recent advances show that cancer cells rely on the reactivation of developmental programs through MET for the localization and proliferation of disseminating cells. The embryo provides clues to understanding the complex cell biology of EMT and MET in cancer and moving toward improved therapeutic strategies.

Embedded Image

Epithelial plasticity in three-dimensional (3D) cultures. Epithelial cells [Madin-Darby canine kidney (MDCK) cells] form ducts when grown on 3D matrices resembling the in vivo microenvironment (left). When grown under identical conditions, MDCK cells expressing Prrx1 (an EMT inducer) form networks of mesenchymal cells (right). Note the dramatic phenotypic change that is accompanied by the acquisition of motility and invasive properties. Blue, nuclei revealed by 4′,6-diamidino-2-phenylindole staining; red, actin filaments as seen after phalloidin binding; green, E-cadherin (epithelial marker) (left) and vimentin (mesenchymal marker) (right).

Advances

Due to the importance of the EMT and MET programs in normal development for the generation of tissues and organs, as well as their role in cancer, stringent regulatory mechanisms are needed. Multiple extracellular signals converge in the activation of transcription factors that can trigger the full EMT program. In addition, epigenetic and splicing programs, as well as microRNA regulatory networks, control epithelial plasticity toward EMT or MET. As differentiated normal and cancer cells can reenter an undifferentiated stemlike state, another level of cell plasticity has become apparent, helping to elucidate complex cell behaviors and interactions.

Outlook

Technical advances in noninvasive in vivo imaging of embryos will help define cell behavior and plasticity in normal development, fundamental to the understanding of congenital malformations. This knowledge will undoubtedly facilitate the study of tumor progression in animal models of cancer. Cancer cells can also be directly interrogated about their plastic states in molecular terms after the purification and analysis of circulating or disseminated single cells from animal models and also from patients, aiding in the design of improved therapies. With respect to antimetastatic therapies, inhibiting EMT may be counterproductive in tumors that disseminate early, as rather than preventing metastasis, it could favor the formation of secondary tumors from already disseminated cells. Strategies aimed at targeting cancer stem cells are very promising, but it is important to consider that new cancer stem cells can be produced from differentiated nonstem bulk tumor cells.

From Here to There

To form different tissues and organs, embryonic cells must migrate to new locations. Specific transcription factors, epigenetic and splicing programs, and microRNA regulatory networks regulate this process, which is known as the epithelial-to-mesenchymal transition (EMT). During EMT, considerable cellular plasticity is observed, and once activated at their new location, cells must again change into their new differentiated form. This “reverse” event is called the mesenchymal-to-epithelial transition (MET). Nieto (p. 10.1126/science.1234850) reviews EMT and MET as observed during normal development and in the generation of cancer when cells leave the primary tumor and travel to other parts of the body forming metastases and secondary tumors.

Abstract

During embryonic development, many cells are born far from their final destination and must travel long distances. To become motile and invasive, embryonic epithelial cells undergo a process of mesenchymal conversion known as epithelial-to-mesenchymal transition (EMT). Likewise, EMT can be seen in cancer cells as they leave the primary tumor and disseminate to other parts of the body to colonize distant organs and form metastases. In addition, through the reverse process (mesenchymal-to-epithelial transition), both normal and carcinoma cells revert to the epithelial phenotype to, respectively, differentiate into organs or form secondary tumors. The parallels in phenotypic plasticity in normal morphogenesis and cancer highlight the importance of studying the embryo to understand tumor progression and to aid in the design of improved therapeutic strategies.

Cellular plasticity refers to the ability of cells to reversibly change their phenotype. An example is seen in early metazoan embryogenesis, where epithelial cells take on mesenchymal characteristics, a process termed epithelial-to-mesenchymal transition (EMT). The EMT implies the conversion of an epithelial cell into a mesenchymal cell with the ability to migrate and invade adjacent tissues. Through the associated changes in adhesion and behavior, cells can move into the interior of the embryo, travel long distances, and participate in the formation of internal organs (1). Importantly, the latter implies the reversibility of the EMT, as when cells arrive to their destination they usually revert to the epithelial phenotype, undergoing a mesenchymal-to-epithelial transition (MET) to settle, proliferate, and differentiate into different organs (2) (Fig. 1).

Fig. 1 The control of epithelial plasticity in development and disease.

Embryonic epithelial cells undergo EMT to migrate and then revert to the epithelial phenotype at their destination to give rise to different tissues and organs. After injury, epithelial cells undergo a partial EMT to heal the wound. Later, these cells undergo a MET to maintain epithelial homeostasis. Likewise, disseminated cancer cells need to return to the epithelial state during metastatic colonization, accompanied by the recovery of a high proliferation potential for the establishment of a secondary tumor. A MET is also required during reprogramming of fibroblasts to iPSCs. The development of organ fibrosis also involves EMT, which does not revert and leads to organ degeneration and failure. Both developmental and pathological EMTs not only induce morphological changes but also provide motility and invasive properties together with the ability to overcome safe-guard mechanisms, leading to cell survival and chemotherapy resistance, which is particularly relevant in cancer patients. Global cellular programs—including epithelial-specific splicing, epigenetic mechanisms, and miRNA regulatory networks—are in place to protect epithelial homeostasis.

The hallmarks of the EMT can be summarized as follows: loss of cell-cell junctions, loss of apico-basal polarity, and acquisition of migratory and invasive properties (3) (Fig. 1). This phenotypic change is accompanied by a dramatic change in cell behavior (see supplementary movies) and is triggered in response to extracellular signals by the activation of one or several transcription factors (TFs) belonging to different families, including Snail, Twist, Zeb, and others (4), which are generally termed EMT-TFs (5). These EMT-TFs elicit the EMT program by repressing the epithelial phenotype, enhancing mesenchymal traits including motility, and inducing the ability to degrade the basement membrane and the extracellular matrix in general. In addition, EMT-TFs also impinge into other basic cellular processes that help maintain the mesenchymal phenotype, the efficacy of migration, and survival in adverse environments. As such, EMT-TFs attenuate proliferation and protect from cell death–activating survival signaling cascades (2) (Fig. 1). Due to the pivotal role of E-cadherin loss, EMT-TFs are sometimes referred to as E-cadherin repressors; also, a decrease in functional E-cadherin is often associated with the activation of EMT. However, it is important to bear in mind that the EMT program and EMT-TFs are much more than E-cadherin transcriptional repressors, that E-cadherin down-regulation (or endocytosis) is not necessarily associated with the EMT program, and that E-cadherin reexpression is not sufficient to revert the fibroblastic phenotype (6).

Cells that escape from carcinomas and are converted into mesenchymal cells with migratory and invasive properties have undergone an EMT (2, 3), and many EMT-TFs have been characterized that operate during tumor progression (7). However, the relevance of the EMT in cancer biology has been a matter of debate (8). The application of sophisticated imaging techniques to animal models and the characterization of circulating tumor cells (CTCs) from patients as cells with an EMT signature have provided evidence that the EMT also occurs during the dissemination of cells from a primary tumor (912).

Because the EMT can also endow cells with stem cell properties, and given that cancer can be initiated, maintained, and propagated by stationary and motile cancer stem cells (CSCs) (5, 13), a new field of study has emerged that requires the concerted effort of developmental biologists and cancer researchers. However, even though developmental and pathological EMTs have many common features, they also have obvious differences, leading to the classification of three different EMT types (14). Type 1 refers to developmental EMTs; type 2 denotes those related to wound healing, tissue regeneration, and organ fibrosis; and type 3 indicates the EMT associated with cancer. Developmental EMTs imply the conversion of epithelial into mesenchymal cells, but embryos lack inflammatory responses, typical of type 2 and 3 EMTs that occur in the adult (14, 15). Type 2 EMTs are also characterized by the appearance of myofibroblasts with the ability to secrete an excess of extracellular matrix molecules in response to inflammation (14). In contrast to the situation in embryos, in cancer cells and during regeneration, the acquisition of a mesenchymal state in fibrosis can be considered an end stage, which leads to organ degeneration and failure (Fig. 1). During cancer progression, tumor cells undergo type 3 EMT, which, in addition to invasion and motility, involves intravasation of delaminated cells into lymphatic and blood vessels and their subsequent extravasation. The intravasation and extravasation steps do not occur in fibrosis and do not have obvious parallels with the migratory processes undergone by embryonic cells. In this Review, I will focus on the similarities—in particular, between embryonic and cancer cells—and how the latter rely on developmental programs, not only to delaminate from the initial tumor but also to later colonize distant sites. The common theme of epithelial cell plasticity unifies both events.

Global Cellular Programs Regulate Epithelial Plasticity

The EMT is triggered by many extracellular signals and agents in both embryos and cancer cells. The most potent inducers are members of the transforming growth factor–β (TGFβ)/bone morphogenetic protein (BMP) family, but also Wnt, Notch, epidermal growth factor (EGF), fibroblast growth factor, hypoxia, obesity, alcohol, nicotine, ultraviolet light, and others [reviewed in (2)]. Such pathways and agents may act alone or in combination, and the response is dependent on the cell context (2). The response to such signals converges on the activation of the EMT-TFs, which can cooperate to induce or maintain the mesenchymal phenotype (4, 16). The existence of many EMT-TFs provides robustness to the system, ensuring the implementation of the EMT program during embryonic development (1, 17). Analyzing regions where EMT is induced in embryos indicates that Snail expression normally precedes the expression of other factors, which are more relevant for the maintenance of the mesenchymal phenotype (2, 4). This temporal hierarchy of EMT-TF activation and cooperation also seems to operate in tumors, as observed for Snail factors and Twist in breast epithelial cells and breast cancer progression (16, 1820). Once activated, EMT-TFs can, in turn, activate the same signaling pathways to generate positive regulatory autocrine loops that help maintain the mesenchymal phenotype both in normal and transformed breast stem cells (21).

Because EMT may be extremely deleterious when aberrantly activated in the adult (3), it needs to be very tightly regulated to ensure epithelial homeostasis. Besides the transcriptional level, there is also posttranscriptional, posttranslational, and epigenetic control, including a splicing program specifically associated with the epithelial phenotype and regulation by noncoding RNAs [recently reviewed in (2224)] (Fig. 1). The epigenetic, splicing, and microRNA (miRNA) networks operating during EMT are now being defined at the whole-genome level (2528), identifying global regulatory programs that control the EMT. Such programs provide specificity and robustness through their phenotypic impact on each aspect of the EMT in both embryos and cancer cells [reviewed in (17, 24)] (Fig. 1).

EMT in Embryonic Development and the Delamination of Cancer Cells from Primary Tumors

The internalization of cells from the surface of the embryo to form organs often involves an EMT. Cells disrupt cell-cell adhesion contacts within the primitive epithelial layer and ingress. The best-studied examples are the ingression of mesendodermal precursors at the primitive streak in amniotes and the delamination of the neural crest (Fig. 2). However, it is worth noting that many additional EMT processes exist in the embryo, as with the exception of the anterior central nervous system and the epidermis, all vertebrate adult tissues originate from cells that underwent at least one EMT. The best evidence of EMT in cancer comes from the observation of single cells delaminating from primary tumors (9) and from the finding that CTCs and disseminated tumor cells (DTCs) from patients show EMT features and a high degree of epithelial plasticity (10, 11) (Fig. 2).

Fig. 2 Cell migration in embryos and tumors.

Individual migration occurs after epithelial cells undergo EMT, as for neural crest migration during normal embryonic development and the delamination of breast cancer cells from the primary tumor. Partial EMT can also occur when cells activate the EMT program but maintain some contacts with their migrating neighbors. Integrated tissue migration implies a decrease in cell-cell adhesion forces but the maintenance of epithelial traits, including cadherin expression and apico-basal polarity. This type of migration is observed in the developing zebrafish lateral line and in the invasive front of some carcinomas. Cells at the invasive front can also activate the EMT program, in which case they can be seen delaminating as individual or loosely organized groups of cells. Blue, epithelial cells; yellow, mesenchymal cells (after EMT).

In development and disease, EMT is not necessarily an “all-or-nothing” transition. Partial EMTs, in which cells share epithelial and mesenchymal traits, can occur. Partial EMTs occur during Drosophila gastrulation and during the early migration of groups of neural crest cells in anamniotes (amphibian and fish). These cell movements are sometimes referred to as collective migration to indicate that cells move in a coordinated manner while maintaining some cell-cell contacts (29). However, in terms of molecular mechanisms, cells have engaged into the EMT program during these processes, down-regulating E-cadherin transcription and losing apico-basal polarity upon activation of EMT-TFs (3). Partial EMTs are also observed in the epithelial component of carcinosarcomas (30) and at intermediate states during the progression of organ fibrosis, when parenchymal (epithelial) cells have activated the EMT program and show both epithelial and mesenchymal markers (31).

Cell movements can also occur in the absence of an EMT, when cells move while maintaining epithelial integrity. Examples of this type of movement include the migration of the lateral line primordium in the zebrafish embryo (Fig. 2) and that of border cells in the Drosphila egg chamber (32). These cells do not express EMT-TFs, maintain cell junctions (including E-cadherin expression), and normally also maintain apico-basal polarity. These movements can also occur in cancer, when tumor cells invade without segregating from the invasive front (Fig. 2), and are able to colonize lymph nodes. However, in contrast to cancer cells that have undergone an EMT, cells that have not undergone an EMT are not able to intravasate into blood vessels to later develop blood-borne distant metastasis (33).

Reversal of Developmental EMTs for Organ Formation

As noted above, developmental EMTs are reversible (34). As such, the cells that go through so called primary EMTs in the embryo later undergo the reverse process, MET, to differentiate into multiple cell types. Early mesodermal cells, generated by EMT and located at different mediolateral levels of the embryo (axial, paraxial, intermediate, and lateral plate mesoderms), condense into transient epithelial structures through MET to give rise to the notochord, the somites, the primordia of the urogenital system, and the splanchopleura and somatopleura, respectively (Fig. 3). In the majority of these structures, a second round of EMT is completed. For example, the dorsal somite converts into the dermal and myotomal mesenchyme, which will give rise to components of the dermis and the muscle satellite cells. The ventral part will become the mesenchymal sclerotome that later generates the vertebrae, the ribs, and the tendons (Fig. 3). In turn, precursor cells of endodermal organs, such as the pancreatic endocrine cells, migrate upon undergoing EMT and revert this migratory phenotype through a MET to form the Langherhans islets. A notable example of epithelial plasticity is the development of the cardiac valves, for which the initial mesenchymal precursors are specified at gastrulation stages and must undergo three consecutive rounds of EMT and MET (2).

Fig. 3 Reversibility of EMT during embryonic development.

Physiological EMT are reversible. With the exception of the epidermis and part of the central nervous system, adult vertebrate tissues are the result of one or several rounds of EMT and MET. This figure follows a temporal developmental sequence: (1) The precursors of the endoderm and the mesoderm ingress at the primitive streak, a prototypical example of primary EMT. (2) The mesodermal precursors migrate to occupy different positions along the medio-lateral axis of the embryo to give rise to axial, paraxial, intermediate, and lateral mesoderm, which upon undergoing MET will generate notochord, somites, the urogenital system, and the splanchno- and somatopleura, respectively. (3) The majority of these epithelial derivatives undergo a second round of EMT. The figure exemplifies secondary EMTs, showing those occurring in the somites (see text). Blue, epithelial cells; yellow, mesenchymal cells after EMT.

Although many signaling pathways have been identified for EMT (2, 24), the signals involved in the induction of MET have not been well characterized in embryonic development. The most studied and amenable processes are those associated with the epithelialization of the paraxial and intermediate mesoderm to form the somites and the precursors of the renal system, respectively. An increasing gradient of BMP signaling along the mediolateral axis patterns the different mesoderms, with lower concentration specifying the paraxial mesoderm. BMP7 counteracts TGFβ signaling and is the most prominent known epithelializing agent, with examples seen during kidney development (35), during reprogramming of fibroblasts to induced pluripotent stem cells (iPSCs) (36), in mouse models of renal fibrosis (37), and in cancer cells (38). Eph and Ephrin signaling is also involved in the epithelialization of somites (39), and Wnt signaling is required for MET during nephron development (40).

For nearly all MET processes, there is a concomitant down-regulation of the corresponding EMT-TFs [i.e., (41, 42)]. This down-regulation is accompanied by the appearance of specific epithelial markers that are associated with differentiation into the distinct tissues and organs. In the case of the epithelialization of the somitic and pronephric tissues, MET is associated with the induction of Paraxis and Pax2, respectively.

Reversal of EMTs for Metastatic Colonization

Despite the observations above, it is not universally accepted that tumor cell dissemination is related to a change in cell identity [a requirement for EMT, see (8)]; rather, this process might rely on mutations that weaken the cell-cell adhesion of cancer cells. Nevertheless, it is clear that carcinoma metastases usually adopt a well-differentiated epithelial phenotype. Ironically, this is one of the arguments used to dismiss the importance of the EMT in cancer progression. However, the differentiated phenotype of metastases reflects the epithelial plasticity implicated in tumor progression—in other words, the reversion of EMT first suggested by Brabletz et al. (43) and again resembling the situations described during embryonic development (34).

As in the embryo, MET concurs with the down-regulation of EMT-TFs in cancer cells. MET not only implies a reversion to the epithelial phenotype but also an increase in cell proliferation, important for the growth of the secondary tumor. Several transcriptional repressors that can repress EMT-TFs have recently been identified, including Kruppel-like factor 17 (KLF17) (44), E74-like factor 5 (ELF5), (45) and Single-minded 2s (SIM2s) (46). Both ELF5 and SIM2s repress Snail2 expression. Conversely, Snail represses Single-minded during Drosophila gastrulation (47), likely revealing an additional regulatory loop in the control of epithelial plasticity.

Complex regulatory loops also control epithelial plasticity posttranscriptionally, as numerous miRNAs are integrated into EMT and MET regulatory networks (22, 24). Double-negative feedback loops have been described between EMT-TFs (Snail, Zeb, and Gata3 factors) and several miRs, including members of the miRNA-200 (miR-200) family and miR-34a (4850), which promote MET and protect the epithelial phenotype (Fig. 4). Recently, an additional miRNA has been added to this network, as miR-22 promotes EMT and cancer progression by indirectly repressing miR-200 (51), and miR-506 has been identified as a node that promotes MET (28) (Fig. 4).

Fig. 4 Epithelial plasticity and miRNA regulatory networks.

(A to D) Signaling pathways mediated by TGFβ or the Notch ligand Jagged induce the expression of EMT-TFs (Zeb, Snail, Prrx, or Gata3 factors), which are down-regulated by a series of miRNAs (miR-200, miR-130, miR-34a, and miR-506) to protect epithelial integrity. Double-negative feedback loops are established in which miRNAs and the TFs behave as reciprocal repressors. The tumor suppressor p53 prevents EMT by activating miR-200 and miR-34a. By contrast, mutant p53 and miR-22 activate EMT by repressing miR-130 and miR-200, respectively. Epigenetic regulation of the Zeb1 and miR-200 promoters impinges into their feedback loop. The bivalent state of the Zeb promoter allows its rapid activation, EMT, and reentry into the CSC state by increasing the expression of the surface marker CD44. By contrast, the TET complex demethylates the miR-200 promoter, thereby favoring MET. Sirtuin 1 (Sirt1) favors EMT by cooperating with EMT-TFs in the repression of E-cadherin expression. miR-506 is a newly described EMT regulator able to repress several EMT-TFs, including Snail2 and Prrx1. Blue, MET related molecules; yellow, EMT related molecules; pink, CSC markers.

Although TGFβ is a potent inducer of EMT-TFs, the tumor suppressor p53 activates miR-34a and miR-200 (50, 52) (Fig. 4). The latter can be antagonized by the opposing role of mutant p53, which activates Zeb1 trough the repression of miR-130 (53). Another p53 family member, p73, and its truncated antagonistic form DNp73, also play opposing roles in tumor progression: Whereas p73 behaves as a tumor suppressor, DNp73 induces EMT (54). Going back to the aforementioned miRNAs, though much less is known about their function during embryonic development, it is clear that miRs help the establishment of embryonic territories (55) and that a double-negative regulatory loop is established between miR-200 and Sox2 transcription factor during neuronal differentiation (56). The expression pattern of miR-200 in developing embryos is compatible with its role in preserving the epithelial phenotype and promoting MET.

Induction of MET by miR-200 was shown to promote breast cancer metastasis (57), in agreement with earlier findings indicating that MET facilitates bladder cancer metastasis (58) and the observation that myeloid cells induce a MET in the metastatic niche (59). However, until very recently, a requirement for epithelial plasticity and the reversion of EMT had yet to be directly assessed in vivo.

In their 2012 study, Tsai et al. used a spontaneous squamous cell carcinoma model in mice carrying skin-specific inducible Twist to show that Twist-mediated EMT is sufficient to promote the dissemination of cancer cells into the bloodstream. Yet more importantly, Tsai et al. used this model to show that Twist needs to be down-regulated for metastasis to occur (60). Further in vivo evidence for epithelial plasticity and the requirement of MET has been found in the progression of breast cancer to the metastatic state. EMT mediated by an EMT-TF, the paired-related homeobox 1 (Prrx1 factor) needs to be reverted, and Prrx1 loss is required for metastatic colonization (16). Therefore, a reversible EMT seems to be necessary for the metastasis of primary carcinomas, revealing a developmental program that is reactivated by cancer cells for them to successfully complete the final step of the metastatic cascade, the formation of macrometastasis.

Uncoupling EMT and Stemness

EMT can confer stemlike properties on cells (61, 62). This is consistent with the concept of “migratory cancer stem cells” (63) and provides a link between the EMT program and the characteristics associated with CSCs; that is, self-renewal and the capacity to seed secondary tumors and produce differentiated nonstem cells (5). However, the relation between EMT and stemness is another controversial issue in tumorigenesis, as it was later shown that fibroblasts must undergo a MET to complete the initial reprogramming of fibroblasts to iPSCs (36, 64). Again, help may come from development, although CSCs differ from canonical embryonic stem cells (ESCs) in that they revert to the phenotypes of only the primary tumor and, therefore, have a more restricted potency than both ESCs and iPSCs (65).

The need for a MET during reprogramming is consistent with the epithelial nature of ESCs. In fact, the Yamanaka factors (Sox2, Klf4, Oct4, and Myc) act together with BMP7 to repress TGFβ signaling and EMT-TFs to revert fibroblasts to the epithelial phenotype (64). In agreement with this, both Snail1 and Prrx1 are down-regulated during reprogramming before cells acquire pluripotency genes (66). In line with the incompatibility of a stable EMT phenotype and metastatic colonization, Celià-Terrassa et al. (67) have shown that EMT can suppress tumor-initiating capacities (TICs) in epithelial cells. Together, these results link the MET with stemlike properties, rather than EMT with stemness as proposed previously. A plausible explanation for this apparent contradictory data may be that epithelial plasticity and stemness are regulated independently. The MET undertaken during reprogramming of fibroblasts requires continuous presence of the Yamanaka factors and does not initially involve the acquisition of stemness, which represents a later event (36). In addition, one of the EMT-TFs, Snail2, cooperates with Sox9 in the acquisition of the mesenchymal mammary stem cell state, although it seems that Snail2 is mainly involved in the induction of EMT, and Sox9 is responsible for the entry into the stem cell state (68). Further evidence of such independent regulation comes from the analysis of Prrx1, which, unlike classical EMT-TFs (Snail, Twist, and Zeb factors), induces EMT without concomitantly inducing CSC properties. By contrast, the loss of Prrx1 facilitates self-renewal and mammosphere-forming ability in breast cancer cells (16), simultaneously inducing MET and TIC, both favoring metastatic colonization.

How Prrx1 down-regulation is achieved at the metastatic niche and how TIC is maintained upon the down-regulation of EMT-TFs linked to stemness (Snail-type) is a matter of future investigation (Fig. 5). Another pending issue is whether the two EMT types are parallel or exclusive processes in a particular tumor. On the one hand, Prrx and Snail do not show any correlation in data sets from patients. However, Prrx1 and Twist are correlated, and in these cases, Prrx1 prevails (16), as PRRX1 down-regulation is sufficient to induce MET and CSC properties. It is possible that both types (Snail and Prrx) could coexist, as depicted in Fig. 5, but information is still lacking. On the other hand, isolated CTCs depict both EMT and CSC traits (10, 11), suggesting that they might not include the Prrx1 type. This has not been specifically addressed, but it is likely that in purified CTCs, Prrx1-positive cells are underrepresented, as they are fully mesenchymal and might have escaped isolation with current purification protocols, which use epithelial markers to identify CTCs (10, 11).

Fig. 5 Reversibility of EMT during metastatic colonization.

Cancer cells delaminate from the primary tumor-expressing EMT-TFs, which endow them with invasive properties to migrate through the extracellular matrix and enter the lymphatic and blood vessels. Those migratory cancer cells expressing classical EMT-TFs (Snail, Twist, and Zeb factors) present CSC properties, whereas those expressing Prrx1 do not. These two classes of CTCs are depicted, although whether the two types of EMT occur in parallel or are exclusive for individual tumors has not been shown. After extravasation, DTCs can colonize distant organs, undergoing a MET upon down-regulation of the EMT-TFs, which reverts them to the epithelial state and increases proliferation. Further work is necessary to characterize the signals that down-regulate the classical EMT-TFs while maintaining CSC properties. CTGF has been shown to induce MET concomitantly with CSC traits in cancer cells, a global effect that promotes metastatic colonization and is similar to that produced after Prrx1 down-regulation. Stationary cancer stem cells can be converted into differentiated tumor cells, and dedifferentiation of non-CSCs can generate cells with TIC. This bidirectional conversion (represented by white double arrows) adds a high degree of phenotypic plasticity in tumors that requires the design of improved therapeutic strategies.

Data accumulated over the past few years indicate that EMT-TFs are developmental factors that, when aberrantly activated in tumors, initiate the invasion-metastatic cascade. This finding also favors the cancer stem cell model over the clonal evolution model by which successive mutations accumulate in a tumor, providing a defined phenotype (69). However, two additional issues need to be discussed here: On the one hand, stemness is also a plastic state during tumor progression, whereby both stationary and migratory CSCs can coexist (13). On the other hand, there are different subsets of EMT-TFs that, though all able to induce the mesenchymal phenotype, decrease proliferation, and promote invasion, can either provide or repress CSC properties (classical EMT-TFs versus Prrx1, respectively); in other words, these EMT-TFs can either promote or repress TICs. The bottom line is that (i) EMT-TFs need to be down-regulated for metastasis to form, which implies reversion to the epithelial phenotype and increased proliferation (Fig. 5), and (ii) the regulation of stem cell properties is independent from that of epithelial plasticity. Malignancy by cancer cells requires the reversion to the epithelial phenotype and the maintenance of a “stemness” state (70), which may be achieved when the Prrx1 type of EMT-TFs is repressed (16) and/or when connective tissue growth factor (CTGF), recently described as both a MET inducer and a stemlike properties enhancer, is activated in cancer cells (71) (Fig. 5).

EMT and MET in Cancer: Conflicting Data or Just Plasticity?

The high degree of epithelial plasticity observed during cancer progression, together with the independent regulation of EMT and stemness, has several implications. EMT-related invasion plus dissemination is required but is not sufficient for completing the metastatic cascade, and therefore invasive and metastatic may not be equivalent terms.

In agreement with the requirement for invasion and dissemination for metastasis formation, the presence of CTCs has prognostic and predictive value in breast cancer patients (72). Accordingly, abrogating EMT-TFs, such as Snail factors, in cancer cells makes these cells less prone to invade and later generate metastasis when injected in mice (18). However, constitutive expression of Snail, pRRX1, or Twist also suppresses metastasis formation (60, 67), as EMT-TFs must be down-regulated for metastasis to form (16, 60). Rather than representing a paradox, these data simply reflect the plasticity of epithelia and the dynamics of the whole metastatic process, resembling that of embryonic cell migration to populate distant territories. Thus, whereas the EMT is associated with the early steps in metastasis (such as delamination, invasion, and intravasation), it is not associated with metastatic colonization. There is no association between the expression of Snail or Twist in primary tumors and relapse-free survival in patients with breast or lung squamous carcinomas (16). These findings are important when discussing the role of different elements in tumor progression, and the phrase “invasive and metastatic” should be avoided when referring to factors involved in the acquisition of invasive properties by cancer cells. A similar situation applies when we contemplate embryonic development. The EMT-TFs are expressed in the embryo wherever EMT processes take place. However, EMT-TFs are not expressed in the derivatives of these migratory populations, because once embryonic cells reach their destination to differentiate into distinct tissues, the EMT inducers are repressed. Thus, EMT-TFs are related to cell behavior rather than to cell fate, and hence, their expression is only transient. Accordingly, EMT-TFs cannot be used as markers of differentiated cell populations, the equivalent of differentiated distant metastases.

The description of signals that down-regulate EMT-TFs at the metastatic sites awaits further investigation, but it is clear that interactions between CSCs and the tumor microenvironment dictate (i) EMT and invasion of the primary tumor (73, 74) and (ii) colonization at the metastatic niche (57, 58, 74). The EMT program also seems to help recently extravasated cells to extend filopodium-like protrusions and interact productively with the niche, promoting homing and proliferation (75). Cytokines are important to modify both microenvironments, as they are fundamental to recruit mesenchymal stem cells and inflammatory cells. The stroma influences metastasis formation, and recent data point to the regulation of TGFβ/BMP signaling in both the primary tumor and the metastatic niche in colorectal and breast carcinomas (33, 76, 77).

Toward Improved Metastatic Therapeutic Strategies

Cell plasticity and heterogeneity in tumor biology also has an impact on the design of therapeutic strategies (69). Classical cytotoxic chemotherapy has proven beneficial in patients with carcinoma. However, one of the main caveats is the appearance of tumor recurrence, associated with acquired multidrug resistance and concomitant with the appearance of signs of EMT and stemness in the residual or relapsed tumor after chemotherapy. Resistance to cell death is a property of embryonic migratory cells upon undergoing EMT, implemented to promote their arrival to their destination (78). A similar strategy is used by cancer stem cells, and therefore, efforts are being devoted to fighting not only proliferation but also EMT and stem cell–like properties to prevent the metastatic disease in cancer patients.

To target the EMT, the strategy has been to inhibit some of its inducing signaling pathways, including those of EGF receptor, platelet-derived growth factor receptor, and TGFβ. Recent data indicate that inhibiting the TGFβ pathway might be beneficial in colorectal cancer metastatic disease (77). But even if the bulk of the tumor can be considerably decreased, chemoresistant stem cells will support the subsequent growth of the tumor. Therefore, different strategies have been used to directly target CSCs (including inhibitors of additional pathways such as those triggered by the developmental factors Notch, Wnt, and Shh) using cytoprotective agents and/or small molecules identified in high-throughput screenings.

Recent data indicate that stem cells can generate differentiated cells and also that adult normal and (nonstem) cancer cells can reenter the stem state, pointing to a bidirectional conversion between stem and nonstem populations (79, 80) (Fig. 5). This additional plasticity seems to be dependent on the chromatin state of the Zeb1 promoter, which is in a bivalent state in non-CSCs so that it can be rapidly activated in response to EMT-inducing signals to help the conversion of non-CSCs into CSCs (81) (Fig. 4). Thus, therapeutic strategies should be adjusted to this superimposed level of cellular plasticity, as combating CSCs is unlikely to be sufficient because new stem cells can be generated from the remaining differentiated tumor cells. The idea is to combine classical chemotherapy with anti-CSC drugs to simultaneously hit nonstem cells and CSC in a tumor. Given the importance of the tumor microenvironment in cancer progression, therapeutic strategies should also consider targeting the stroma.

Finally, it is clear that fully preventing delamination from the primary tumor will impede metastasis, which is the principle that has inspired efforts by academia and pharmaceutical companies to block the EMT. A recent report shows that this seems to be a valid strategy, as observed in a mouse model of ovarian carcinoma (28). Nonetheless, in the light of recent data on epithelial plasticity during metastatic colonization and the need for cancer cells to revert to the epithelial phenotype through a MET (16, 60), rather than preventing metastasis, inhibiting EMT could favor the formation of secondary tumors from already disseminated cells. This is particularly important in carcinomas in which EMT is a very early event in tumorigenesis, where cancer cells have already disseminated before the tumor is diagnosed, as in pancreatic and breast carcinomas (82, 83). By contrast, as EMT reversion does not happen during the development of the fibrotic disease, therapeutic intervention inhibiting EMT is a promising strategy to ameliorate fibrosis and organ degeneration (37).

Recent advances reveal an unanticipated degree of cell plasticity during the progression of carcinomas to the metastatic state and in the generation of CSCs, reflecting the complex biology of cancer. This concept will undoubtedly influence the strategies for therapeutic intervention. Further complexity may arise when considering other tumor types. An example is found in melanomas, where recent data show that ZEB1 and TWIST1 promote dedifferentiation and are associated with late-stage melanoma and poor prognosis. By contrast, SNAIL2 and ZEB2 do not behave as classical EMT inducers, as they are expressed in normal melanocytes and work as tumor suppressors (84). Fortunately, the process of morphogenesis provides clues as to how cell plasticity may occur in cancer, including the responses elicited by the different EMT-TFs and the mechanisms used in the maintenance and differentiation of CSCs. Together with embryos, cancer animal models such as “avatar” mice (mice that carry cancer cells from individual human patients) will help define the best therapeutic strategy, even in a personalized manner. In general terms, although inhibiting EMT may be counterproductive, strategies aimed at targeting cancer stem cells are very promising, keeping in mind that new cancer stem cells can be produced from differentiated nonstem bulk tumor cells.

Supplementary Materials

References and Notes

  1. Acknowledgments: I thank members of my lab for helpful discussions, in particular R. Corcoles for contributing the Review Summary figure and movies S1 and S2. I would also like to thank S. Ingham (Instituto de Neurociencias de Alicante–CSIC–UMH) for his help with the figures. Work in my lab is supported by grants from the Spanish Ministry of Science and Innovation (BFU2008-01042 and CONSOLIDER-INGENIO 2010 CSD2007-00017 and 00023), the Generalitat Valenciana (Prometeo II/2013/002 and ISIC 2012/010), and the European Research Council (ERC AdG322694).
View Abstract

Stay Connected to Science

Navigate This Article