EMT has a role in several disease states including pathologic organ degradation (ie fibrosis) and cancer. How EMT fits into these processes will be reviewed, including the initiating signal pathways, downstream targets, and mechanisms for regulating these pathologic EMTs.
EMT and cancer
The pathologic potential of EMT is seen in organ fibrosis and similar mechanisms are seen in the development of cancer, evolution of disease, and metastatic progression. In cancer, EMTs rely on: 1) exogenous induction agents acting through transcriptional control 2) Non-coding RNAs and 3) creating and maintaining cancer stem cells [
29].
1. Transcriptional control mechanisms
The discovery of the transcription factor SNAIL1 interacting with the CDH1 promoter to repress the production of E-cadherin began our understanding of how EMT is regulated [
30],[
31]. Numerous other transcription factors have been found to play a role in the progression of epithelial cells to a mesenchymal phenotype including SNAIL2, ZEB1, ZEB2, E47, KLF8, TWIST1, and FOXC2, all which have the ability to promote EMT in various cell cancer lines [
32],[
33]. Of these, the nuclear factors which seem to be most important are SNAIL, ZEB, and TWIST proteins and these can be modulated to initiate EMT by exogenous stimuli including hypoxia, bile acids, and nicotine [
34]-[
36].
The association between disease progression and solid tumor oxygenation is seen in various malignancies. It is apparent that tumor hypoxia contributes to several features of aggressive cancers by enhancing invasive growth, local spread, and distant deposits [
37]. The mechanism for hypoxic influence of these features is likely multifactorial, however there is clear interplay with EMT. In pancreatic tumor cells, hypoxia (induced by culturing cells in a hypoxic chamber) resulted in expression of EMT makers when compared to normal conditions [
34]. This observation is seen in other solid tumors including lung and breast [
38].
Cholestatic disorders seem to promote hepatic fibrosis and to the progression of carcinogenesis in cholangiocytes and hepatocytes. Studies have shown that chendeocycholic acid (CDCA), a bile acid, can activate EGFR to mediate gene expression. In a cDNA microarray, CDCA increased the expression of SNAIL to repress E-cadherin both in vitro and in vivo in liver cancer models, initiating an EMT program [
36]. The ability for bile acids to play an activating role is just one of many examples where exogenous substrates can induce EMT.
Another substrate which activates EMT through a similar mechanism is nicotine. Specifically, in breast and lung cancer cell lines (A549, MDA-MB-468, MCF-7), treatment with nicotine increased the proliferation, invasion, and migration of both lung and breast cancer cells in a dose-dependent fashion. Notably, expression of E-cadherin was repressed while the expression of fibronectin and vimentin, both mesenchymal markers, increased. The mechanism for these changes occurred through signaling by nicotinic acetylcholine receptors (nAChRs) [
35].
Recent data demonstrates a pathway for cigarette smoke to stimulate EMT in small bronchi of patients demonstrating a mechanism contributing to the progression of disease. In primary human bronchial epithelial cells (HBEC), exposure to cigarette smoke led to expression of EMT markers by increased production of intracellular ROS, release of TGFβ, and TGFβ activation of the transcription factors Smad3 and ERK1/2 [
39]. This kind of interaction between an exogenous substrate and the cellular machinery activating EMT has significant clinical implication as treatments can be designed to potentially halt or reverse the progression of disease.
2. Non-coding RNAs
Non-coding RNAs and, in particular, microRNAs (miRNAs) have gained significant attention for their ability to modify gene expression allowing for quick cellular adaptation to new environments and conditions [
29],[
40]. The regulatory association of miRNAs is seen in many settings including development, maintaining cell homeostasis, and cell-cell interaction [
41]. Li et al., in 2009, showed a role for miRNAs in reversing EMT in pancreatic tumor cells that were resistant to common chemotherapeutics [
42].
Since that time, numerous miRNAs have been found to interact either directly or indirectly with EMT machinery in several cell models [
43]. Interestingly, these miRNAs can act as either tumor suppressors or tumor activators based on the cell type and environment they interact with. Additionally, their adaptive properties play a role in chemotherapy resistance.
Two major classes of miRNAs have been studied extensively in the setting of EMT: the miR-200 family and the miR-34 family. Other families include the miR-205, miR-9, and miR-10 families, respectively. The miR-200 family has five named members located on human chromosome 1 (mir-200a, mir-200b, mir-429) and 12 (mir-200c, mir-141). Both the miR-200 and miR-34 families are typically considered to be “tumor suppressive” as their members all stabilize the epithelial phenotype [
44].
The interplay between miR-200 and the ZEB family of transcription factors to maintain E-cadherin expression has been validated in several systems. In a TGFβ-induced model of EMT in a murine mammary cell line, miR-200 family member expression was repressed. This same cell line was transfected with synthetic pre-miRNAs to stimulate overexpression of corresponding miR-200 members in vitro. The result was inhibiting EMT in the transfected cell lines, when compared to control, by increasing the expression of E-cadherin. Furthermore, using an elegantly designed luciferase assay, miR-200 family members were noted to interact directly with the transcription factors ZEB1 and ZEB2 [
45]. ZEB1 and ZEB2 typically act as EMT inducers by suppressing E-cadherin expression, however in the presence of miR-200, ZEB1 and ZEB2 could not interact with E-cadherin and the epithelial phenotype was maintained [
45],[
46].
The clinical relevance of miR-200 family members can be seen in malignant breast, prostate, and colon cancer cells. Silencing of the miR-200 family induces early transformation of epithelial cells leading to EMT, reinforcing the role of these miRNAs in tumor suppression [
47].
An interesting link between the p53 tumor suppressor gene, SNAIL, and the miR-34 was described by Siemans et al. in 2011. In colon cancer cells, activation of p53 notably induced the expression of miR-34 family genes, resulting in the inhibition of SNAIL activation. As would be predicted, a correlate decrease in EMT markers was seen in the absence of SNAIL activation in this cell line. Treatment of cells with miR-34 halted TGFβ dependent EMT. It was noted, however, that the action of miR-34 family members was not unidirectional. The miR-34 family promoter sequences contain E-boxes which can be bound by SNAIL, resulting in decreased expression of miR-34. This double negative feedback loop between miR-34 and SNAIL offers an interesting regulatory mechanism for EMT creating unique challenges in drug development targeting either of these factors [
48].
In contrast to the tumor suppressive functions of miR-200 and miR-34, the miR-9 family appears to be tumor activating through a direct interaction with the CDH1 promoter which subsequently initiates EMT [
49]. miR-9 expression is increased in primary breast, gastric, and brain cancers [
29]. Additionally, increased expression of miR-9 is seen in metastatic disease further supporting a tumorigenic role for this family of miRNAs.
It must be emphasized that work with non-coding RNAs is far from complete and their relevance to cancer (and cancer therapy) remains imprecise. While the miR-200 family appears to have compelling data for a tumor suppressive function, as discussed above, there is are also data indicating a pro-metastatic niche for miR-200 family members.
In samples from colon cancer patients, there is lowered expression of miR-200 within the primary tumor which fits the impression that miR-200 is tumor suppressive. However, in metastatic lesions, miR-200 expression is high [
47],[
50]. This allows for argument that miR-200 plays a pro-metastatic role in cancer by stabilizing the mesenchymal phenotype after deposition to distant sites (i.e. an important regulatory element in the mesenchymal to epithelial transition). Other evidence for a pro-metastatic role for miR-200 is seen in pancreatic, breast, and gastric cancer.
Another important miRNA family is the miR-10 group including miR10b. In the study of metastatic breast cancer cells, miR-10b was noted to be highly expressed and positively regulate cell invasion. It was also able to initiate tumor formation and played a role in distant metastases. Taken together, these properties are highly suggestive that miR-10b is implicated in EMT [
51].
3. Cancer stem cells
The concept of the cancer stem cell is based on the hypothesis that a single cell type or subset of cells can act as tumor-initiators in various cancers ranging from hematopoietic to solid organ. By definition, these cancer stem cells (CSC) have features which give them “stem-like” properties allowing for self-renewal and differentiation leading to the histologic heterogeneity displayed by tumors [
52].
Seminal work done in the mid-1990s provided compelling evidence supporting the role of CSCs in tumor initiation and progression in acute myeloid leukemia (AML). The AML-initiating cells, defined by their cell surface marker expression (CD34
+,CD38
-), were able to proliferate, disseminate, and maintain an AML phenotype similar to the patients they were derived from after transplantation into severe combined immune-deficient (SCID) mice [
53].
Furthermore, the AML-initiating cells appeared to be derived from normal hematopoietic stem cells directly linking stem cells to tumorigenic cells. In other words, normal primitive cells underwent leukemic transformation resulting in the first experimentally described CSC population [
54]. This pattern of tumor-initiation has subsequently been described in various solid tumors including brain, breast, pancreatic, prostate, and others [
55]-[
59].
Ultimately, CSCs exert their pathophysiologic impact in several ways. Their ability to self-renew and differentiate is at least partly responsible for the heterogeneity seen within tumors. One potential consequence of this is chemotherapy and radiotherapy resistance within populations of tumors. Additionally, CSCs can regenerate a tumor when transplanted into an appropriate host in experimental settings. This offers a proposed mechanism for metastatic disease and cancer recurrence. These properties have made CSCs, irrespective of their tumor of origin, extremely important for cancer therapy and an important new cancer target.
The machinery of EMT appears to connect directly to the formation and maintenance of CSCs. Mani et al. induced EMT in human, non-tumorigenic mammary epithelial cells using either TWIST or SNAIL transcription factors. As seen in prior studies, exposed cells acquired a mesenchymal phenotype [
30]. Studying these cells using flow cytometry revealed a pattern of cell surface expression that was seen in breast cancer stem cells (CD44
High/CD24
Low). EMT made non-tumorigenic epithelial cells, at least in breast tissue, look like cancer stem cells.
While it was interesting that EMT made epithelial cells look like stem cells, more important was the discovery that these cells acted like stem cells. CD44
High/CD24
Low cells were able to self-renew and give rise to multiple cell lineages. Furthermore, it was noted that CD44
High/CD24
Low cells expressed a mesenchymal phenotype and up-regulated the expression of EMT transcription factors [
60]. Together, these findings connected two important cancer-related processes - CSCs and EMT.
Other important EMT signaling pathways are also implicated in the transformation and maintenance of CSCs. For example, Wnt signaling is a well-studied pathway necessary for a variety of EMTs, both developmental and pathologic [
61]. Wnt signaling is classically described as being canonical or non-canonical based on the involvement of β-catenin [
62]. Hyperactivity of canonical and non-canonical Wnt pathways can trigger EMT programs and, more recently, play key roles in CSC biology [
61].
In fact, Wnt activity may be used as a marker for defining CSCs in some malignancies. Using colon CSCs derived from patient samples, Wnt signaling was assessed using a GFP reporter system. Cells with high GFP expression notably had increased clonogenicity compared to those cells with low GFP expression. Stated another way, colon CSCs with high Wnt expression were highly potent at inducing tumors following injection into mice. The extent of Wnt activation was positively correlated with expression of well-studied cell surface makers for CSCs [
63].
Notch signaling is another important pathway required for the conversion and maintenance of CSCs using EMT. Multiples studies show the importance of Notch signaling in multiple breast cancer subtypes and its role in invasive disease. Notch signaling, specifically Notch4 receptor activation, is increased in breast CSCs. In addition, Notch signal pathway inhibition decreases the number of breast CSCs both in vitro and in vivo [
64].
The importance of Notch signaling is also seen in pancreatic CSCs. The β-secretase inhibitor IX can be used to inactivate Notch signaling. After treating pancreatic CSCs with this secretase inhibitor, cell proliferation, migration, and invasion are all decreased. The mechanism for this appears to be through inhibiting the mesenchymal transition of pancreatic cells lines demonstrating the importance of Notch signaling to EMT [
65]. Treatments directed at Wnt and Notch signal pathways may allow for regulation of EMT in tumors and modulation of CSCs.
A property of CSCs with important clinical relevance is the observation that they demonstrate resistance to both chemotherapy and radiotherapy [
59]. Imatinib mesylate is a tyrosine kinase inhibitor used for the treatment of chronic myeloid leukemia (CML) with excellent clinical result. Early studies, following its approval for use by the Food and Drug Administration in 2001, showed a complete response to therapy in 76% of patients at a median follow-up of 19 months [
66].
However, these results have since been somewhat tempered by the finding that patients with an early response to imatinib show evidence of disease recurrence [
67]. The mechanism for this appears to be supported by a significant role for CSCs in resisting initial treatment and leading the progression to cancer recurrence as CSCs isolated from patients with CML show insensitivity to imatinib [
68]. Some authors describe this as the “weed” or “dandelion” hypothesis as it may appear the problem is solved, but until the root is removed the weed can and will come back [
69].
The “dandelion” hypothesis was tested by Li et al. in a breast cancer model. Cells were taken from patients with HER2+ breast cancer before and after treatment with lapatinib. Following chemotherapy, the percentage of CD44
High/CD24
Low cells increased. These cells, with a molecular signature consistent with breast CSCs, were able to self-renew and had a propensity for tumor formation [
70]. CD44
High/CD24
Low breast CSCs appear to also show resistance to conventional radiation therapies [
71]. The implication of these data are two-fold. First, chemotherapy and radiotherapy appear to select for the survival CSCs. Second, chemotherapy and radiotherapy do not impair the ability of CSCs to grow and replicate.
The focus of this review to this point has been on EMT, however an important process that must be discussed is its “opposite” pathway, mesenchymal to epithelial transition (MET). Just as EMT plays a critical role in development, MET also is a process first recognized in development. One specific example is the formation of epithelial nephrons which require the transcription factor Pax-2 to initiate MET. This illustrates that while MET functions to transition cells to an epithelial phenotype, the regulatory factors required are not necessarily the same for both processes [
72]. MET is an expansive topic and the migrating cancer stem cell (MCSC) model will be discussed to briefly emphasize the role MET plays in metastatic disease and distant seeding of tumor cells.
The (MCSC) model attempts to integrate the concepts of CSC, EMT, MET, and metastatic disease. CSCs are generated via an EMT allowing them the ability to migrate and disseminate. Upon reaching a distant site, the CSC is exposed to a new environment which can either promote dormancy or undergo a MET to recapture epithelial characteristics. The next step would be the formation of a metastatic lesion with conserved features from the primary tumor [
73].
The relevance to this model on a clinical level cannot be understated as therapies targeting EMT could inadvertently promote MET and progression of disease. Overcoming these challenges will likely define the next steps in cancer drug development directed at EMTs.