Epithelial-to-Mesenchymal and Mesenchymal-to-Epithelial Transition in Mesenchymal Tumors: A Paradox in Sarcomas?

The epithelial-to-mesenchymal transition (EMT) is a multi-facetted process, which was initially described during embryogenesis (1). The EMT is defined as the reprograming of an epithelial cell into a full mesenchymal phenotype and is comprised of various subprograms that promote the loss of cell–cell interaction and a lack of polarity, as well as increases in extracellular matrix interfaces (2). The first clues regarding the involvement of the EMT in cancer were provided by Michael and colleagues, who discovered that yolk sac tumors with a mesenchymal phenotype originate from the metaplasia of the yolk sac epithelium and that yolk sac tumors are chemoresistant and pluripotent (3). The EMT allows individual cells or groups of cohesive cancer cells to escape the rigid epithelial organization, acquire motility, and gain the ability to invade other loci, which explains why the EMT is implicated in cancer progression, metastasis, and drug resistance in several carcinomas (1). In fact, it has been shown that cancer invasiveness very rarely initiates as a single-cell process but is rather a collective event. During this collective cell migration, a subpopulation of cells undergoes partial EMT, changing their morphology toward a spindle-like phenotype and reorganizing the cell cytoskeleton and cell–cell contacts (4, 5), but maintaining the expression of some epithelial markers, e.g., E-cadherin, which are necessary to mediate cohesiveness during migration, and allows resistance to the mechanical stress that cells experience during migration (6, 7).

Importantly, the EMT is a reversible process, and tumor cells undergoing the mesenchymal-to-epithelial transition (MET) can re-acquire an epithelial phenotype, which is a more sessile and proliferative phenotype (1). Accordingly, the MET is defined as the complete reprograming of a mesenchymal cell into an epithelial cell (1).

While the term “carcinoma” refers to malignant tumors of epithelial (i.e., ectodermal or endodermal) origin, the term “sarcoma” refers to malignant tumors arising from cells that descend from the mesenchyme (8). To date, at least 70 sarcoma subtypes have been identified, though their precise etiologies are often unknown (8). Mesenchyme-derived sarcoma cells are already equipped with the phenotypic features typically induced by the EMT in epithelial cells and do not need to undergo the EMT to acquire them. However, because the EMT is a dynamic and gradual process, the existence of intermediate EMT states or incomplete EMT has become widely accepted (2).

Many sarcoma subtypes exhibit an aggressive clinical phenotype characterized by early metastasis and frequent relapse, which is associated with unfavorable clinical outcomes (8). Although the underlying biological mechanisms remain largely elusive, several reports indicate that sarcomas can undergo phenotypic changes reminiscent to the EMT/MET (9–11), which could result in such highly aggressive clinical behaviors. However, because sarcomas are locked, by definition, in the mesenchymal stage ab initio, it is unlikely that sarcoma cells can undergo full EMT/MET. However, it is reasonable that, depending on their histiotype, they can shuttle in their phenotype through EMT/MET-related processes, which can contribute to their progression and aggressiveness. Indeed, the expression of some epithelial markers in sarcoma is not a rare event but is quite variable across different histiotypes (12), and accumulating data suggest that many sarcomas can undergo EMT- and MET-related processes. These processes are especially evident in certain sarcoma subtypes, such as biphasic synovial sarcomas and uterine carcinosarcomas, which display a biphenotypic morphology with the characteristics of both mesenchymal and epithelial tumors. In accordance, a recent study from The Cancer Genome Atlas provided evidence that uterine carcinosarcomas exhibit a prominent EMT-related transcriptomic signature, which was associated with cell cycle and survival pathways (13). In line with this observation, previous studies showed that the stimulation of uterine carcinosarcoma cell lines with TGFβ—a major EMT trigger in carcinomas—induces an EMT-like process, which was associated with increased cellular migration (11).

In this review, we summarize current knowledge on this topic in the literature and propose that sarcomas can reside in an intermediate, i.e., “metastable” phenotype between the epithelial and mesenchymal stages enabling them to undergo EMT- or MET-related processes under specific conditions, which likely contributes to their aggressive clinical behavior (Fig. 1).

Although full EMT has mainly been observed in carcinomas, EMT-like processes have also been reported in nonepithelial cancers. This is particularly demonstrated by malignant melanomas originating from neuroectodermal cells, which spread in a mesenchymal state throughout the body during embryogenesis and finally settle in the skin. It appears that a subpopulation of melanoma cells transiently acquires a mesenchymal-like state, which is characterized by fluctuations in gene expression programs that are typically observed during the EMT/MET (14, 15). These fluctuations may allow individual or groups of cohesive melanoma cells to transiently acquire mesenchymal or epithelial features, equipping them with either a more motile/invasive or sessile/proliferative phenotype (14, 15). Indeed, the concomitant induction of TWIST and ZEB1—two master transcription factors (TF) contributing to the induction of the EMT—can be observed during melanoma progression (14, 15).

We propose that analogously to melanoma, certain sarcoma subtypes can also undergo EMT/MET-related processes and that transient fluctuations in the activation of EMT/MET-related programs can lead to reversible phenotypical switches given specific stimuli. In fact, the EMT is a complex program in which cells can adopt and reside in multiple intermediate phenotypic states of the epithelial–mesenchymal spectrum (16). Interestingly, although most mesenchymal cancers to do not primarily form lymph node metastases (17), epithelioid sarcoma predominantly metastasizes, like carcinomas, via the lymphatic system (18), suggesting that the specific intermediate phenotype and the subsequent activation of EMT- or MET-related programs may influence the preferred routes of metastasis. In fact, in rare cases, Ewing sarcoma can also metastasize to the lymph nodes (19), and it is tempting to speculate that these tumors activate an EMT-related program and thus acquire carcinoma-like metastatic behavior. Lymph node metastasis is similarly rare in osteosarcoma (2%–4% of cases) but tends to be more frequent in more differentiated osteoblastic osteosarcomas (20), which often show an epithelioid phenotype with strong cytokeratin expression (21).

A specific intermediate EMT state, termed a “metastable” phenotype, was identified in trophoblast stem cells (22). Such a metastable phenotype is defined as a cellular state, which is characterized by the presence of both epithelial and mesenchymal features, as well as stem cell properties, which can be associated with more aggressive behavior in cancer cells (22, 23). This is mediated, at least in part, by the fact that a metastable phenotype enables the collective cell migration of clusters of cohesive cells, which protects metastasizing cells from apoptosis and the harsh microenvironment (24). During collective cell migration, cancer cells typically express P-cadherin, which mediates cell polarization and collective cell migration via activation of CDC42 (25). Interestingly, in normal myoblasts, high levels of P-cadherin promote malignant transformation, migration, and invasion, whereas its knockdown in rhabdomyosarcoma cells has the opposite effect (26). In breast cancer, P-cadherin is involved in cell adhesion, invasion, and collective cell migration, and its expression is associated with the expression of stem cell markers such as CD44, CD49f, and ALDH1 (27).

In fact, few studies have suggested that breast cancer cells displaying a metastable phenotype have a higher capacity to form mammospheres with the concomitant expression of stemness markers such as CD44 and CD24, which confer poor patient outcomes (24, 28). Based on these findings, Ribeiro and colleagues suggest that P-cadherin may serve as a marker for the metastable phenotype (27).

In sarcoma tumors, such a metastable phenotype may allow individual tumor cells to acquire the features of more either epithelial or mesenchymal differentiated cells. Such molecular intratumor heterogeneity could lead to highly aggressive clinical behavior because the entire tumor will take advantage of both the EMT- and MET-related biological features of its cells.

The concept of a metastable phenotype is further supported by the variable degree of epithelial/mesenchymal differentiation observed in various sarcoma histiotypes, which can be either more epithelial-like (Ewing sarcoma, synovial sarcomas, and epithelioid sarcomas) or more mesenchymal-like (osteosarcoma, chondrosarcoma), as well as by the existence of sarcoma subtypes displaying both extreme phenotypes within one tumor.

For instance, biphasic synovial sarcomas, which are composed of spindle-shaped and epithelioid cells, express the epithelial marker E-cadherin in their epithelioid cells, whereas monophasic fibrous synovial sarcomas lack an epithelioid component and show strongly express the EMT TFs SNAIL, SLUG, and TWIST1, as well as the EMT-inducer TGFβ (29, 30). These phenotypic switches are fully reversible, as demonstrated by TWIST1 knockdown experiments, which reduced migration and invasiveness of synovial sarcoma cell lines (30), and it appears that they are at least partially caused by the interference of the driving-fusion oncogenes SYT-SSX1/2 with SNAIL and SLUG (31). Another example is provided by epithelioid sarcoma, which characteristically exhibits both mesenchymal and epithelial markers. In fact, virtually all cases are positive for cytokeratin and epithelial membrane antigen, but most cases coexpress the mesenchymal marker vimentin (32).

Collectively, these findings imply that certain sarcomas may reside in an intermediate EMT state, termed “metastable” state, which is clinically relevant (Fig. 1).

In Ewing sarcoma—the second most common sarcoma in children and adolescents, for which a neuroectodermal or mesenchymal origin is proposed—several research groups have shown that individual tumor cells can switch back and forth between more epithelial and more mesenchymal phenotypes (33–36). This is most likely mediated by stochastic fluctuations in the expression of the pathognomonic master oncogene EWSR1–FLI1 in individual Ewing sarcoma cells (36). It has been demonstrated that EWSR1–FLI1 induces a highly proliferative and more epithelial phenotype through the upregulation of tight junction proteins such as claudin-1, ZO1, and occludin, which is accompanied by a more disorganized actin cytoskeleton (33, 34, 36). Recently, Franzetti and colleagues proved that some Ewing sarcoma cells transiently show low levels of EWSR1-FLI1 expression, which induces an EMT-related process, with dramatic gains in motility and invasive capacity, accompanied by a clear shift from the expression of cell–cell proteins to the expression of cell-matrix proteins (36). In fact, low-EWSR1–FLI1 cells were capable of detaching from the entire tumor and invading Matrigel as either single cells or in more collective fashion depending on the cell line (36).

Although only a minority of cells within a given Ewing sarcoma tumor are thought to undergo such EMT/MET-related switches, this may equip the entire tumor with functional compartments that are either more epithelial-like and thus more proliferative or more mesenchymal-like and thus more motile/invasive. This phenomenon may also explain the long-lasting observation that Ewing sarcoma is a highly proliferative but also metastatic type of tumor—features that indicate its highly aggressive clinical nature. The basis for this is, in our opinion, the metastable state of Ewing sarcoma cells. On the molecular level, this could be mediated by the fact that EWSR1–FLI1 transcriptionally represses the expression of TGFβ receptors (37), which may make high-EWSR1–FLI1 cells insensitive and low-EWSR1–FLI1 cells susceptible to the induction of EMT-related programs through TGFβ.

In carcinomas, the EMT and MET are driven by a variety of EMT-TFs, such as SNAIL, SLUG, TWIST1, and ZEB1/2, which induce significant changes in cell–cell interaction, cell-matrix adhesion, and the cytoskeleton. Important downstream effectors, such as epithelial (E-cadherin, occludin, and ZO1) and mesenchymal markers (vimentin, N-cadherin, and fibronectin), can be used to determine the degree to which a given cell has undergone the EMT, MET, or EMT/MET-related processes.

Although little is known about how sarcoma cells with a metastable phenotype can undergo EMT- and MET-related processes, several in vitro studies have provided first clues. For instance, osteosarcoma genes such as TIM3 (10), ST6GAL1 (38), TRIM66 (9), UHRF1 (39), and CYR61 (40) contribute to the positive regulation of EMT-TFs. However, MET-related processes also occur in osteosarcoma because it has been shown that TSSC3 acts as an EMT repressor via the upregulation of GSK3β and the subsequent suppression of the WNT pathway promoting cell migration and invasion (41).

In addition, miRNAs are implicated in the suppression and induction of EMT-related processes in osteosarcoma (42–45). For instance, miR-370 and miR-132 suppress EMT-related processes inhibiting tumor growth and metastasis via targeting forkhead box protein M1 (FOXM1) and SRY-box 4 (SOX4), respectively (43, 44). In contrast, miR-23a and miR-130a promote EMT-related processes by targeting the tumor-suppressor phosphatase and tensin homolog (PTEN; refs. 42, 45). In addition, the PI3K/AKT/mTOR and MEK/ERK pathways were found to induce EMT-related processes, leading to a more aggressive rhabdomyosarcoma and sacral chondrosarcoma (46–48).

Taken together, these data suggest that certain sarcomas can undergo to an EMT- and MET-related process through pathways classically involved in the EMT/MET in carcinomas. The activation of one or another pathway appears to be crucial for the phenotypic switching of sarcomas toward either a more epithelial or mesenchymal phenotype. For instance, the activation of the TGFβ, WNT, AKT/PI3K/mTOR, and MEK/ERK pathways can lead to a more aggressive and mesenchymal phenotype. In contrast, the upregulation of tumor suppressors, such as PTEN or GSK3β, can promote epithelial differentiation.

Recently, a research group generated a three-dimensional model of mammary epithelial cells to study TGFβ–induced EMT (23). Interestingly, after the interruption of TGFβ administration, the cells underwent a partial reversion of the EMT, coming to resemble the “metastable” phenotype, via the re-expression of E-cadherin at the plasma membrane, along with the concomitant expression of fibronectin and vimentin (23). In addition, the mRNA levels of the epithelial marker occludin were increased, together with those of the mesenchymal marker N-cadherin and the EMT-TF ZEB2 (23). This three-dimensional model may represent a helpful tool with which to investigate the EMT-related process in sarcomas.

Sarcomas are mesenchymal cancers that often display an aggressive clinical behavior (8). Thus far, few studies have considered the possibility that these cancers can undergo EMT/MET-related processes. However, the findings summarized in this review suggest that certain sarcomas reside in an intermediate, i.e., metastable phenotype, one characterized by the combined presence of epithelial and mesenchymal features. This metastable phenotype, which is a cellular state, allows tumor cells to switch between epithelial and mesenchymal differentiation, which likely depends on the spatio-temporal expression and/or activity of classical EMT-TFs due to specific stimuli. As a result, many sarcomas may contain only a few individual or groups of tumor cells, which equip these tumors with additional biological and clinical features related to the epithelial- or mesenchymal phenotype. The combined presence of epithelial and mesenchymal features likely contributes to the aggressiveness of such sarcomas, which is why we propose that a metastable phenotype is more common in aggressive sarcomas. Moreover, studies of carcinomas and sarcomas suggest that P-cadherin promotes collective migration, whereas the stemness markers CD44 and CD24 were also found to be positive in the metastable phenotype. Thus, we suggest that the combinatorial expression of P-cadherin, CD44, and CD24 may represent a promising signature for the metastable phenotype.

The first in vitro studies of sarcomas indicated that EMT- and MET-related processes and the expression of EMT/MET-related markers are regulated via several pathways, such as TGFβ, WNT, MEK/ERK, PI3K/AKT, NF-κb, PTEN, and GSK3β, which are also known to regulate EMT in carcinomas and to enhance motility and invasiveness. However, further studies are necessary to precisely define the triggers and effectors of EMT/MET-related processes in metastable sarcomas. Three-dimensional models, such as that established by Bidarra and colleagues, may help to accelerate research on metastable phenotypes and explore whether EMT/MET-related processes, such as those in carcinomas, contribute to metastasis and drug resistance in sarcomas and how these processes can be therapeutically blocked.

We propose that certain sarcoma subtypes reside in a peculiar metastable state that enables individual tumor cells to undergo EMT/MET-related processes due to specific cues, combining both epithelial and mesenchymal biological features in a single tumor, which makes metastable sarcomas highly aggressive.

No potential conflicts of interest were disclosed.

We thank Dr. Isidora Romani for her help in drafting the figure, and Drs. Gabriel Leprivier and Sarah Maria Fendt for critical reading of the article. We apologize to those colleagues whose work could not be cited.

T.G.P. Grünewald is supported by grants from the “Verein zur Förderung von Wissenschaft und Forschung an der Medizinischen Fakultät der LMU München (WiFoMed),” the Daimler and Benz Foundation in cooperation with the Reinhard Frank Foundation, by LMU Munich's Institutional Strategy LMUexcellent within the framework of the German Excellence Initiative, the “Mehr LEBEN für krebskranke Kinder – Bettina-Bräu-Stiftung,” the Walter Schulz Foundation, the Wilhelm Sander-Foundation (2016.167.1), and by the German Cancer Aid (DKH-111886 and DKH-70112257). G. Sannino is supported by the Fritz-Thyssen Foundation (FTF-2015-01046).