ZEB/miR-200 feedback loop: At the crossroads of signal transduction in cancer



Embryonic differentiation programs of epithelial–mesenchymal and mesenchymal–epithelial transition (EMT and MET) represent a mechanistic basis for epithelial cell plasticity implicated in cancer. Transcription factors of the ZEB protein family (ZEB1 and ZEB2) and several microRNA species (predominantly miR-200 family members) form a double negative feedback loop, which controls EMT and MET programs in both development and tumorigenesis. In this article, we review crosstalk between the ZEB/miR-200 axis and several signal transduction pathways activated at different stages of tumor development. The close association of ZEB proteins with these pathways is indirect evidence for the involvement of a ZEB/miR-200 loop in tumor initiation, progression and spread. Additionally, the configuration of signaling pathways involving ZEB/miR-200 loop suggests that ZEB1 and ZEB2 may have different, possibly even opposing, roles in some forms of human cancer.

Epithelial–mesenchymal transitions (EMT) are embryonic morphogenetic programs whereby epithelial cells are converted into mesenchymal cells. Diverse molecular and phenotypic changes occur during this transition; cells lose their epithelial characteristics and acquire mesenchymal traits. The molecular hallmarks of EMT programs are loss of epithelial cadherin and other junction proteins and de novo expression of mesenchymal markers, such as vimentin, N-cadherin and S100A4.1, 2 EMT programs operate at early (gastrulation, neural crest delamination) or late (organ formation) stages of embryonic development downstream of several signaling pathways, such as Wnt, TGFβ, BMP, Notch, FGF, etc.3 The reverse process, a mesenchymal–epithelial transition (MET), has also been shown to play an important role in development by generating epithelial tissues from mesenchymal progenitors. In pathological conditions, aberrant activation of various EMT/MET programs contributes to fibrosis and cancer metastases. Several pleiotropically activated transcription factors acting downstream of EMT pathways are categorized as master regulators of EMT (MR-EMT). MR-EMT include Zn finger transcription factors of the SNAIL (SNAI1 and SNAI2) and ZEB (ZEB1 and ZEB2) families, the basic helix-loop-helix (bHLH) proteins E47, TWIST1, TWIST2, a forkhead transcription factor FOXC2 and a few other less comprehensively studied candidates.4 These factors interact with promoters of target genes and recruit transcriptional corepressors, coactivators and chromatin remodeling complexes to regulate transcription.3, 5 Upon ectopic expression in epithelial cells, these factors induce complete redifferentiation and generate individually migrating and highly invasive cells. In recent years, MR-EMT have been shown to control different key cancer-related features such as cell cycle progression,6–8 cell survival,9, 10 escape from cellular senescence,11 drug resistance,10, 12 DNA damage response13 and stemness.14, 15 This multifunctionality may be important at different stages of tumorigenesis, including oncogenic transformation, metastatic dissemination, escape from oncogene addiction and drug resistance. In particular, these activities of MR-EMT have been discussed in the context of the parallel model of tumor progression.9, 16 This model envisages epithelial cell plasticity to be a driving force for the formation of dormant micrometastases, which then escape dormancy and evolve into secondary tumors.17, 18 Cells in growing primary tumors accumulate genetic defects and produce late metastases, which are therefore genetically dissimilar to those generated during the early waves of dissemination. MR-EMT have been suggested to regulate virtually every step in tumor development including early and late metastatic dissemination.9, 16

In recent years, the cancer-related pathways controlling the expression and function of particular MR-EMT have been identified in an overwhelming number of reports. These studies demonstrate crosstalk between EMT and signaling pathways which are considered to be active at both early and late tumorigenic stages. Therefore, these data indirectly support the view of MR-EMT as factors which are active throughout the entire process of tumorigenesis. As an exploration of novel data on all MR-EMT is beyond the scope of this review we will focus on two proteins, ZEB1 and ZEB2 and discuss their involvement in several signal transduction pathways fundamental for different cancer stages, initiation, progression and spread.

ZEB Proteins and miRNA

The ZEB family of transcription factors includes ZEB1 (also known as δEF1, TCF8, AREB6 and Zfhx1a) and ZEB2 (also known as SIP1 and Zfhx1b). Both proteins have roles in development, and mutations in the ZEB2 allele cause Mowat–Wilson syndrome, a form of Hirschprung's disease manifested by cranio-facial abnormalities, mental retardation, pigmentation defects, etc.19 ZEB1 mutations are associated with polymorphous corneal dystrophy resulting from endothelial cell migration defects.20 ZEB1 and ZEB2 are large transcription factors (124 and 136 kDa, respectively) containing two zinc-finger domains, separated by a homeodomain. The zinc-finger structures are highly conserved and allow DNA binding at E-boxes present within the promoter regions of target genes, such as E-cadherin. ZEB family members are able to interact with multiple transcription factors and cofactors, including Smads, p300/pCAF, BRG1, the NuRD complex and CtBP. These interactions determine whether ZEB1 and ZEB2 function as transcriptional repressors or activators and have been discussed in detail in recently published excellent reviews.5, 21–23

An important regulatory link between ZEB proteins and several species of microRNA (miRNA) has been established by several groups. miRNA are small noncoding RNA that silence cognate target genes by binding to the 3′UTRs in corresponding mRNAs resulting in inhibition of translation or mRNA degradation. A group of miRNAs highly implicated in the regulation of EMT is the miR-200 family consisting of five members, which form two clusters. MiR-200b, miR-200a and miR-429 are clustered on human chromosome 1, whereas miR-200c and miR-141 are grouped on chromosome 12, with each cluster expressed as a polycistronic transcript. Binding specificities differ within the miR-200 family, with seed sequences differing between miR-200a-141 (subgroup I) and miR-200b-200c-429 (subgroup II).24

In initial studies, inverse correlations between miR-200c and ZEB125 and miR-200b and ZEB226 were established in carcinoma cells in culture. Repression of ZEB1 and ZEB2 by the miR-200 family resulted in enhanced expression of the key epithelial marker, E-cadherin, and acquisition of an epithelial phenotype.26, 27 Moreover, from the analyses of NCL-60 cell lines, expression of the miR-200 family was established as a marker for the epithelial phenotype.24 miRNA-200 family members have subsequently been studied in a number of EMT-related in vitro model systems. During induction of EMT in MDCK cells with either TGF-β or ectopic expression of the protein tyrosine phosphatase Pez, the miR-200 family and E-cadherin were repressed in parallel with an increase in ZEB1 and ZEB2 expression. The ability to induce an EMT was dependent upon repression of the miR-200 family and induction of ZEB1 and ZEB2 expression. Conversely, a MET could be induced by expression of the miR-200 family in cells that were originally mesenchymal in nature. These results confirm that the miR-200 family represses ZEB1 and ZEB2 expression and consequently inhibits the progression of an EMT by establishing and maintaining an epithelial phenotype.28 The repression of ZEB expression by miRNA-200 family is direct, and occurs as a result of the miRNA binding to eight and nine sites in the 3′UTRs of ZEB1 and ZEB2 mRNA.24, 28, 29

An additional level of miR200-ZEB protein regulation was identified in breast and colon cancer cell lines in which ZEB1 was constitutively downregulated by shRNA. Cells underwent MET, with corresponding upregulation of the miR-200 family, most notably the miR-141/200c transcript (Fig. 1a). The miR-141/200c promoter contains multiple highly conserved E-boxes which are occupied by ZEB1 in mesenchymal cells leading to the transcriptional repression. This finding was complemented by data showing that ZEB1-depleted cells retained epithelial phenotype upon miRNA-200 inhibition. Taken together these data established the presence of a double-negative self-enforcing feedback loop, whereby miR-200 and ZEB1 negatively control the expression of each other.21, 30 The presence of this regulatory loop was confirmed and further expanded upon to include other members of the miR-200 family (Fig. 1a).31 Data suggest that the majority, if not all, epithelial cells express high levels of the miR-200 family, which directly repress ZEB1 and ZEB2 and so enable the expression of E-cadherin. However, if an extracellular signal stimulates the expression of ZEB1, the miR-200 family is suppressed allowing EMT to proceed.

Figure 1.

Hypothetical schemes depicting crosstalk between ZEB/miR loop (detailed in a) and several cancer-associated pathways including TGFβ (b), stem cell (c), NOTCH (d), WNT (e), RAS (f), Rb (g) and p53 (h). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

The relevance of the miR-200/ZEB loop to EMT has been addressed in vivo. In the p53-/KRASG12D lung adenocarcinoma model the miR-200 family had the most prominent differential expression in metastasis-prone relative to metastasis-incompetent tumors. Moreover, forced expression of the miR-200b cluster in metastasis-prone tumor cells abrogated their capacity to undergo EMT and metastasize in syngeneic mice.32 Likewise, in a mouse model of pancreatic neuroendocrine tumors, low expression levels of miR-200 family members were part of a metastatic gene expression signature.33 However in clinical samples, enhanced levels of miR-200a and miR-200b and hypermethylation of the ZEB2 promoter was associated with different forms of human pancreatic cancer and chronic pancreatitis.34 A recent study in breast cancer revealed a dual role for the miRNA-200/ZEB loop in tumor progression.35 As in pancreatic cancer specimens, expression of the miR-200 family members positively correlated with metastatic colonization in breast cancer samples, and their expression was elevated in cells with a higher rate of metastatic spread. Enhanced expression of these miRNA species was required for lung colonization in an orthotopic mammary fat pad model. This contrasted with expression of miRNA-200 family members reducing invasion in vitro and decreasing the number of tumor cells circulating in the blood in vivo. This suggests a novel Janus-like role for the miRNA-200 family in cancer, hindering entry of tumor cells into circulation but promoting colonization of distant organs for those cells that do intravasate.35 Mechanistically, the prometastatic function of miRNA-200 was largely dependent on Sec23a, a new miRNA-200 target that functions in the establishment of a metastases-suppressive secretome.

miRNA that target ZEB proteins are not limited to the miR-200 family. ZEB1 and ZEB2 3′UTRs contain one and two sites, respectively that are targeted by miRNA-205, which was shown to act with miRNA-200 in an additive manner to downregulate expression of ZEB proteins leading to a MET.28 miRNA-205 is downregulated in breast cancer compared with normal tissue (Fig. 1a),36 but the association of this miRNA with the differentiation status of NCI60 cell lines is not as significant as in the case of the miRNA-200 family.28 miR-205 has been suggested to have a specific function in maintaining epithelial differentiation in cells of the mammary gland.37 A direct link between ZEB2 and miRNA-192, although established in the diabetic kidney,38 has never been fully addressed in cancer. However, given that miRNA-192 expression is regulated by both the p53 and TGFβ pathways,39, 40 the role of a miRNA-192/ZEB2 axis in tumor biology is likely to be the target of future investigation.

TGFβ Pathway and miR-200/ZEB Loop

TGFβ/SMAD signaling is a prototypical EMT-inducing pathway, which inhibits early tumor development, but promotes tumor progression in the late stage. Homozygous deletions of different TGFβ ligands and receptors results in embryonic lethality due to disruption of different morphogenetic EMT programs implicated in palate, lung and heart development.41 Type I and II TGFβ receptors form complexes in response to the binding of a ligand with subsequent phosphorylation of the R-SMADs, SMAD2 and SMAD3. Activated R-SMADs bind cytoplasmic SMAD4 and enter nuclei to control transcription of target genes.41, 42 The interplay between the TGFβ pathway and MR-EMT occurs at several levels. SMAD complexes interact with a number of transcription factors implicated in EMT including both ZEB proteins43, 44 and SNAI1.45 The domain that interacts with R-SMADs in ZEB2 encompasses 51 amino acids (437–487). This area is minimally conserved between ZEB2 and ZEB1 proteins (only 27% amino acid identity) suggesting that their affinity for SMADs may differ. Indeed, ZEB1 is a poor SMAD binder and is likely to interact only indirectly with SMAD3 via p300.44 Binding of either ZEB protein to p300 increases the stability of SMAD-p300 complexes, possibly contributing to transcriptional activation by SMADs.44 This hypothesis is, however, not supported by observations made in a classical model where EMT was induced by treating NMuMG cells with TGFβ. Here, upregulation of mesenchymal markers by TGFβ was independent of either ZEB1 or ZEB2, whereas both proteins equally downregulated E-cadherin.46 Other levels of interaction between ZEB proteins and TGFβ involve complex mutual transcriptional regulation of feedback ZEB/miR-200 loops and TGFβ (Fig. 1b). TGFβ ligands are either direct (TGFβ2 is directly repressed by miR-141/200c30) or indirect targets of miR-200.47 Prolonged treatment of MDCK cells with TGFβ1, as well as manipulation of ZEB protein or miR-200 levels, activates the autocrine TGFβ pathway with subsequent irreversible EMT. The mechanism by which the ZEB/miR-200 loop activates autocrine TGFβ signaling is presently not well understood. However, at least two important observations have been made. First, SMADs are directly associated with the ZEB2 gene promoter in vivo and SMAD4 is required for the induction of ZEB2 transcription by TGFβ1.47, 48 Second, the repression of both miR-200 loci occurs through reversible DNA methylation via a mechanism that possibly involves the recruitment of histone-modifying complexes by ZEB proteins. The finding that two other components of TGFβ pathway, SMAD2 and TGFβ-RI are miR-200 targets49 adds further complexity to the ZEB/miR-200/TGFβ regulatory network (Fig. 1b).

The importance of the ZEB/miR-200/TGFβ regulatory loop has been confirmed in nonsmall cell lung carcinoma cells. Argast et al.50 carried out gene expression profiling of lung cancer cells H358 undergoing EMT in response to ectopic expression of ZEB1, SNAI1 or constitutively active TGFβ. Almost all ZEB1-regulated genes represented a subset of the TGFβ-dependent transcriptome. In contrast, TGFβ and SNAI1 regulated overlapping but different gene sets.50 Significant progress has been made in the identification of mechanistic pathways that execute EMT downstream of the ZEB/miR-200/TGFβ loop. An important mechanism involving regulation of splicing programs during EMT was first observed years ago for the FGFR2 transcript,51 and was later documented in a number of different experimental models. Recent studies identified two splicing factors, epithelial splicing regulatory proteins (ESRP) 1 and 2, which orchestrate epithelial splicing regulatory networks and seem to be important players in EMT programs.52, 53 Ablation of ESRPs is sufficient to abrogate the epithelial splicing pattern leading to the development of many mesenchymal features and activation of signaling pathways contributing to EMT, such as Akt, FGF or ILK.53, 54 In a model of TGFβ-induced EMT, global alterations in splicing pattern have been described.55, 56 The underlying mechanism was through transcriptional repression of the ESPR genes by ZEB1 and ZEB2, which for ESRP2 occurred via direct promoter binding by ZEB1/2. In particular, the ZEB-ESRP module appears to be a critical element of EMT in TGFβ signaling. Indeed, ectopic expression of ESRP1+2 attenuated EMT in TGFβ-treated NMuMG cells and induced a partial MET (including an increase in E-cadherin expression) in mesenchymal MDA-231 breast carcinoma cells.56

These findings accord well with the hypothesis, first formulated by Frisch in1997,57 that the epithelial phenotype is the default state of cancer cells, and that intrinsic or extrinsic signals are required for the transition to mesenchymal state.57 The above data exemplify this idea in relation to TGFβ-induced EMT, where the default epithelial state is established by the presence of ESRPs, members of miR-200 family, lack of ZEB expression and the absence of autocrine TGFβ signaling. Prolonged treatment with TGFβ switches this equilibrium towards a mesenchymal pattern where autocrine TGFβ signaling represses miR-200 and ESRPs, activates ZEB1 and ZEB2 and leads to the global reprogramming of gene expression and splicing patterns.

The functional significance of ZEB proteins in TGFβ-induced EMT has been highlighted in esophageal cell lines. Forced expression of EGFR strongly increased the number of immortalized esophageal cells which were competent to undergo EMT in response to TGFβ. This observation was explained by the finding that upon TGFβ treatment ZEB proteins were induced exclusively in cells overexpressing EGFR. Depletion of ZEB proteins by RNAi induced cellular senescence via TGFβ-mediated upregulation of the tumor suppressor proteins p16INK4A, p15INK4B and p21.58 Notably, TWIST proteins have been shown to override oncogene-induced senescence and cooperate with activated Ras in oncogenic transformation. Aberrant activation of RAS results in premature senescence via the induction of the tumor suppressors p16INK4a and p21. This pathway, however, was abrogated by TWIST1 and 2, implying that under the conditions used in this study TWIST proteins acted as bona fide oncogenes.11 ZEB proteins may have similar functions in cells exposed to TGFβ suggesting a widespread role for MR-EMT in oncogenic transformation.

ZEB Proteins and Stem Cell Pathways

Cancer heterogeneity has been recognized for more than a hundred years. However, it has only recently been experimentally demonstrated that different populations of cells in solid tumors exhibit different tumorigenic potential. In xenograft assays, breast cancer cells expressing high CD44 and low CD24 cell surface markers appeared to be >100-fold more tumorigenic than tumor cells with a high CD24 expression signature.59 These highly tumorigenic cells (termed cancer stem cells, CSC) are usually underrepresented in primary tumors, but may represent the origin of metastases. The link between CSC and EMT was theoretically proposed by Brabletz et al. in 2005,60 and later experimentally proven in two independent studies.14, 15 Immortalized HMEC cells undergoing EMT in response to ectopic expression of TWIST, SNAI1 or TGFβ1 treatment acquired a CD44high/CD24low antigen phenotype and enhanced tumorigenicity. In agreement with this finding, CD44high/CD24low cells isolated from different sources (including immortalized HMEC cultures or normal mammary glands) exhibited mesenchymal characteristics. Of relevance to the theme of this review, these stem-like cells contained very high levels of ZEB2 mRNA.14

Experimental evidence linking ZEB proteins with stemness in cancer arose from the observation that miR-200c targets the polycomb group member BMI1, an essential regulator of stem cell self-renewal. miR-200c suppressed the tumorigenicity of breast CSC in immunodeficient mice by downregulating BMI1, suggesting that EMT and stem cell properties are jointly controlled by this species of miR.61 In pancreatic cancer cells, ZEB1 maintains stem-like properties via downregulation of miR-200c, miR-203 and miR-183, miRNAs which inhibit stemness by suppressing the stem cell factors BMI1, KLF4, SOX-2 and possibly p63. Expression of ZEB1 was detected at the invasive front of pancreatic tumors, and these areas were proposed to represent a reservoir in which CSC are generated (Fig. 1c).62 Further evidence for the important role of ZEB1 in promoting a CSC phenotype in pancreatic cancer was provided in a recent study where ZEB1, EMT and consequent dissemination of circulating cancer cells correlated with increased tumor initiating capacity and early metastases.63

In contrast to ZEB1, the role of ZEB2 has not been systematically studied in CSC. However, recent reports have established its role in stem cell biology in normal embryonic development. Tissue-specific conditional deletion of ZEB2 gene in hematopoetic stem cells resulted in severe defects in their differentiation and mobilization (Fig. 1c).64 In human embryonic stem cells ZEB2 regulates cell-fate decisions by promoting the neuroectodermal state and repressing mesoendoderm development. ZEB2 transcription is repressed by two stem cell factors responsible for the maintenance of pluripotency, NANOG and OCT4. NANOG is in turn repressed by ZEB2 indicating that ZEB2 expression shifts embryonic stem cells from self-renewal towards differentiation (Fig. 1c). The TGFβ-related factors Activin-NODAL play key role in stem cell biology. Via interactions with other pathways (BMP or FGF) they promote either pluripotency or mesendoderm differentiation. NODAL signaling cooperates with NANOG and OCT4 to repress ZEB2 and the resultant release of NANOG repression reverses ZEB2-induced neuroectodermal differentiation.65 The downregulation of ZEB2 by NODAL occurs via direct binding of the NODAL effectors SMAD2/3 to a regulatory element ∼4 kb upstream of the main transcriptional start site in the ZEB2 gene. In spite of the emerging role of NODAL pathway in cancer66 the potential link between NODAL and ZEB proteins remains to be investigated in CSC.

Regulatory Loops Associated with Early Stages of Cancer: NOTCH and WNT

Notch signaling is an embryonic pathway that maintains the self-renewal ability of embryonic stem cells. Abnormal activation of Notch has been associated with different cancer types including pancreatic, lung, breast carcinoma, malignant melanoma, etc.67, 68 In the Notch pathway, a cell fate signal is generated upon interaction of two adjacent cells expressing a Notch ligand (DLL1, 2, 4, JAG1 and JAG2) and a cognate receptor (NOTCH1-4). The binding of the ligand results in the proteolytic cleavage of the receptor by γ-secretase, followed by nuclear translocation of the receptor intracellular domain. In nuclei, this fragment interacts with transcription factors and coactivators to stimulate transcription of target genes. Notch cooperates with several pathways activated in cancer and has been implicated in the induction of an EMT program via SNAI1 and SNAI2.69, 70 Recently, two important findings established a mechanistic link between the ZEB/miR-200 axis and the Notch pathway in pancreatic and prostate cancer cells.71–73 First, the Notch ligand JAG1 and coactivators MAML2 and MAML3 are targets of miR-200 family members. Second, activation of Notch signaling increases ZEB1 expression (Fig. 1d). The underlying mechanism may involve JAG2-stimulated upregulation of the transcription factor GATA3, which acts as a repressor for both miR-200 loci. This pathway has been shown to contribute to EMT and tumor progression in a KrasLA1/+p53R172HΔG/+ mouse model of lung adenocarcinoma.74 These observations suggest a mechanism for cooperation between Notch signaling and other EMT inducing stimuli, such as TGFβ or hypoxia. Activation of ZEB1, e.g., by TGFβ, would promote EMT and a CSC phenotype via repression of stemness-inhibiting miRNAs, including miR-200 family members. This would release the repression of JAG1, MAML2 and MAML3, stabilizing Notch signaling and further reinforcing the EMT switch and generation of CSCs mediated by ZEB1 (Fig. 1d).72

Wnt signaling represents another example of a developmental pathway that is functionally linked with the ZEB/miR-200 network in cancer. In more than 80% of cases, colorectal cancer is driven either by loss-of-function mutation of APC or (less frequently) by stabilizing mutations in β-catenin. These mutations lead to the constitutive activation of the canonical Wnt pathway, which would be expected to lead to the nuclear accumulation of β-catenin. However, while its nuclear localization is often detected at the invasive front, cells within the bulk of a tumor usually retain normal membranous β-catenin expression (so called β-catenin paradox).75 A combination of intrinsic (additional mutations) and paracrine (microenvironmental) factors facilitates β-catenin signaling and promotes localized tumor cell invasion. As nuclear accumulation of β-catenin at the invasive front of colorectal tumors is associated with EMT, the interplay between the Wnt pathway and EMT has been investigated. Despite ZEB1 immunopositivity being detected at the invasive front of colon tumors and in tumor-associated fibroblasts76, 77 manipulation of ZEB1 expression levels in colon carcinoma cell lines had no effect on β-catenin/TCF4 transcriptional activity.76 Accordingly, an effect of ZEB1 on the upstream components of canonical or noncanonical Wnt signaling has not, to our knowledge, been reported. However, a functional link between the canonical Wnt pathway and ZEB1 has recently been established in work demonstrating that ZEB1 is a direct transcriptional target of β-catenin in colon cancer cells.77 Moreover, ZEB1 mediates the effects of Wnt signaling on the expression of other key proteins implicated in invasive growth of colorectal tumors, MT1-MMP and LANC2 (Fig. 1e).77 Interestingly, expression of ZEB2 has also recently been detected in the invasive front of colorectal cancer at a significantly higher level than in the tumor center.78 However, the ZEB2/Wnt relationship in colon cancer remains unclear. The expression pattern of ZEB2 in colon carcinoma may indicate an interaction, perhaps similar to that of ZEB1, with the Wnt pathway.

A possible link between ZEB2 and the Wnt pathway in cancer is suggested by results obtained by Tarabykin and coworkers where ZEB2 ablation in the precursors of cortical neurons results in reduced cell proliferation and increased cell death.79 This is caused by compromised activity of the noncanonical Wnt/JNK pathway due to overproduction of the Wnt antagonist, secreted Frizzled-related protein 1 (SFRP1). In normal hippocampal development the transcription of SFRP1 is directly repressed by ZEB2. SFRP1 is often downregulated in breast cancer leading to activation of the Wnt pathway and a subsequent reduction in CD24 levels, a feature associated with highly aggressive basal like carcinomas. This suggests a possible mechanistic link between ZEB2 and the Wnt pathway in this cancer type.80 Experiments carried out in our laboratory have indicated that ectopic expression of ZEB2 in squamous carcinoma cells leads to the induction of the Wnt receptor Fzd4 and the repression of the Wnt pathway inhibitors DKK1 and WISP2. The latter has been shown to suppress TGFβ-induced EMT by acting as transcriptional corepressor of the TGFBR2 gene.81 Interestingly, CCN6 (WISP3) another Wnt-inducible factor attenuates ZEB1-induced EMT by inhibiting secretion of insulin-like growth factor-1 and reducing phosphorylation of the cognate receptor.82, 83

The RAS-MEK-ERK-Fra-1 Pathway

Gain-of-function mutations in RAS or BRAF oncogenes are thought to be tumor-initiating events in several forms of cancer. Initial studies investigating the cooperation between the RAS and TGFβ pathways demonstrated that RAS activation predisposes epithelial cells to TGFβ-induced EMT by overriding TGFβ-induced cell cycle arrest.84 In mammary epithelial cells, RAS-induced TGFβ1 secretion resulted in an autocrine mechanism of mesenchymal state maintenance.85 Increased cell motility is an important characteristic of tumor cells and is cooperatively regulated by the TGFβ and RAS pathways.86 TGFβ-induced EMT in NMuMG cells was associated with increased stress fiber and focal adhesion formation, resulting in strong adhesion. RAS activation alleviated this effect by blocking the induction of tropomyosins (TM1) and counteracting integrin signaling. Interestingly, Fos-related antigen-1 (Fra-1), an important effector of the RAS pathway, also regulates cytoskeletal dynamics and cell adhesion by uncoupling active Rho from stress fiber formation and by repressing integrin β1 signaling.87, 88 Additionally, Fra-1 has the potential to override the inhibitory effect of TGFβ on cell cycle progression.89

This view of Fra-1, as a key molecule mediating the collaboration between the RAS and TGFβ pathways, was supported by a study performed in the Blenis' laboratory. Here, Fra-1 was shown to contribute to an EMT program downstream of ERK2 or RAS-G12V via the upregulation of ZEB1 and ZEB2 (Fig. 1f).90 The mechanism of ZEB protein upregulation was not addressed and it remains unclear if they are activated directly or if miR-200 family members are involved. Another novel and complex Fra-1-mediated EMT regulatory pathway involving ZEB2 may function in breast cancer. Transcription of miR-221/222 is directly activated by the RAS-MEK pathway via occupation of a distant AP-1 binding site by Fra-1. miR-221/222 then targets the 3′-UTR of TRPS1 (tricho-rhino-phalangeal syndrome type 1), a member of the GATA family of transcription factors (Fig. 1f). TRPS1 is in turn a transcriptional repressor, directly binding at two sites within the ZEB2 promoter.91 This novel pathway was shown to operate in basal rather than in luminal subtypes of breast cancer. Both ZEB1 and ZEB2 genes are transcribed in the majority of mesenchymal breast cancer cell lines,28 and they both represent a part of the EMT signature in the claudin-low subtype of breast tumors.92 It remains to be elucidated whether ZEB1 is also a downstream effector of RAS/Fra-1/miR-221;222/TRPS1 signaling and whether TRPS1 upregulation is a common event in different EMT programs. The recent finding that the interaction between the corepressor CtBP and ZEB1 is MEK-dependent adds an additional layer of complexity to the interplay between the RAS/MEK and TGFβ pathways (Fig. 1f).46

The ability to override oncogenic addiction to mutant K-Ras in pancreatic and lung cancer cell lines is known to be dependent on the expression of ZEB1.93 Accordingly, K-Ras pathway inhibition in A549 and PANC-1 cancer cells resulted in apoptosis in ZEB1-depleted cells. These data suggest that mutational activation of Ras signaling may activate a regulatory loop to drive development of cells independent of this initial mutation, via the upregulation of ZEB1 and induction of EMT.

Recent work described a novel link between ZEB2/miR-200 loop and the PI3K/AKT branch of the RAS pathway (Fig. 1f).94 In a mouse model of malignant melanoma ZEB2 has been shown to increase expression of the PTEN (Phosphatase and TENsin homologue) tumor suppressor protein, loss of which accelerates melanomagenesis. PTEN activation was independent of ZEB2 protein function, but dependent on several microRNA species including miR-200b, −181, −25 and −92a. The underlying mechanism was direct competition between PTEN and ZEB2 mRNAs for microRNA binding and attenuated ZEB2 expression resulted in PTEN downregulation and PI3K/AKT pathway activation. The study demonstrated a tumor-suppressive function of ZEB2 in malignant melanoma.94

Context-Dependent Interactions between ZEB Proteins and the Retinoblastoma Pathway

ZEB1 null embryos exhibit multiple defects, at least some of which are caused by the reduced proliferation rate of mesenchymal progenitors.95 Consequently, mouse embryonic fibroblasts obtained from knockout animals display reduced replicative capacity in culture and undergo early senescence. Mechanistically, ZEB1 was found to attenuate senescence via the transcriptional repression of p15Ink4B, a tumor suppressor protein that prevents activation of CDK4 and CDK6 by type D cyclins and impedes G1-S cell cycle progression.95 Importantly, p15Ink4B is a key effector of TGFβ-induced cell cycle arrest in different genetic contexts.96 Thus, both ZEB1 and ZEB2 proteins are required for normal proliferation of different mesenchymal progenitor cells in embryonic development. ZEB2 stimulates prosurvival Wnt signals in cortical precursors,79 whereas ZEB1 uncouples TGFβ signaling from cell cycle arrest and senescence in some mesenchymal precursors (perichordium) and in the ventricular zone.95 In contrast with the role of ZEB2 in hippocampal development, its ectopic expression in squamous or bladder carcinoma cell lines results in G1 cell cycle arrest.7, 13 This occurs via transcriptional repression of cyclin D1 and hypophosphorylation of Rb protein (Fig. 1g). Another link between the Rb and ZEB pathways has been established in a study addressing the way in which mutations in E2F1 or Rb family members affect ZEB expression. In the absence of Rb proteins ZEB1 was activated by E2F1 via an E2F-binding DNA element in its distal gene promoter. However, it was actively repressed in the presence of hypophosphorylated Rb1.97, 98 In contrast, transcription of ZEB2 was strongly repressed in mouse embryo fibroblasts in which the G1/S checkpoint was lost as a result of disruption of all three Rb family members (Fig. 1g).97 The mechanism of ZEB2 gene regulation was not addressed in this study, and it remains unclear whether transcriptional regulation of ZEB2 by Rb-E2F is direct. These data, however, indicate that there is an Rb-E2F-dependent mechanism ensuring high expression of ZEB1 and low expression of ZEB2 in actively proliferating fibroblasts. In resting cells, the ZEB protein expression pattern is converted to a low ZEB1/high ZEB2 mode. Low expression of ZEB2 in actively proliferating cells is consistent with the observation that ZEB2 induces replicative senescence in breast and hepatic cancer cell lines via the negative regulation of hTERT expression.99, 100 Overall, ZEB2 exhibits some context-dependent tumor suppressor features, which can possibly be circumvented by oncogenic signaling (such as RAS or Wnt).

The ZEB/miR-200 Network and p53 Family Members

p53 and its homologues p63 and p73 control cell cycle progression, apoptosis, senescence, differentiation and stemness.101 p63 and p73 exist as full-sized proteins (TAp63 and TAp73) or truncated versions (DNp63 and DNp73), which lack transcriptional activation domains. TA isoforms are akin to p53 in activating proapoptotic programmes in response to different types of stresses. In contrast, the DNp63 and DNp73 isoforms may have anti-p53 and, therefore, prosurvival functions and are overexpressed in several forms of human cancer. A link between ZEB proteins and the p53 family has been established in early studies demonstrating a direct repression of DNp73, and to a lesser extent TAp73 isoforms during myoblast differentiation and in MEFs (Fig. 1h).102 Likewise, DNp63 (but not TAp63 or p53) are repressed by ZEB1 in MEFs and differentiating keratinocytes.103 A cell survival effect of ZEB1 via repression of the proapoptotic TAp73 isoform has been described in cortical neurons exposed to ischemic conditions.104

Recent studies have revealed a second arm in the regulatory network linking p53 family with the ZEB/miR-200 circuit. p53 was shown to directly activate transcription of the miR-200c/141 cluster in breast epithelial cells,105 and both miR-200c/141 and miR-200a/200b/429 units are under the positive control of p53 in hepatocellular carcinoma cell lines (Fig. 1h).39 Consistent with these findings, knockdown of p53 in MCF12A epithelial cells upregulated ZEB1, induced EMT and influenced EMT-associated stem cell properties. Overexpression of p53 in the same cells reverted the TGFβ-induced EMT and stem cell characteristics.105 These data demonstrate a novel function of p53 in which it acts not only as a guardian of genome integrity but also to safeguard the differentiated epithelial phenotype. The p53 gene is mutated in the majority of cancers, but different p63 and p73 isoforms are often overexpressed. Therefore, in the absence of p53 these proteins may be involved in the control of the ZEB/miR-200 equilibrium and EMT-MET plasticity. Indeed, both proteins were identified as positive direct regulators of miR-200 in ovarian carcinoma cells.106 In addition to ZEB proteins, other MR-EMT are involved in p53 signaling. For instance, miR-34a/34b/34c are upregulated by wild type p53 but downregulated by SNAI1; and SNAI1 mRNA is a target of these microRNA species.107 These data suggest that miR-34 and SNAI1 form a p53-controlled double negative feedback loop analogous to the p53/ZEB/miR-200 module. The functional interaction of EMT and p53 pathways appears to be a common theme in tumor progression and is critical for the activation of prosurvival or antiapoptotic responses and generation of EMT-permissive conditions.


  • The “linear model” of tumor progression suggests that cells within a tumor pass through successive rounds of mutation and selection for the most aggressive genetic backgrounds.108 At the final stages of this process there is an expansion of metastatic clones giving rise to the formation of secondary tumors. According to this model, activation of different signaling pathways does not occur simultaneously, but in an ordered manner. Indeed, in gastric cancer, for example, even small and benign lesions contain mutations leading to the activation of canonical Wnt pathway, but only large and aggressive primary tumors harbor inactivating mutations in p53. The alternative “parallel progression model” of tumor development, whilst retaining the idea of clonal selection in tumor growth, suggests that metastatic spread and formation of dormant micrometastases is an early event in the disease rather than the endpoint of cancer evolution. Accumulating evidence regarding the role of CSC in cancer indicates that highly tumorigenic cell populations are generated from differentiated somatic cells at different stages of tumor progression. Data reviewed in this article illustrate that the ZEB/miR-200 loop is an integral part of signaling pathways which are mutationally activated at early (WNT, RAS, NOTCH) or late (TGFβ, TP53, RB) stages of tumor growth. These data support the idea that factors controlling EMT act at different stages of tumor progression and contribute to major hallmarks of cancer acquired during tumor initiation, progression and metastatic dissemination.

  • The literature reviewed here suggests that many complexities remain regarding the precise, and sometimes contradictory, role of each ZEB family member in tumor biology. For example whilst we see ZEB1 cooperating in the acquisition of a Notch dependent embryonic stem cell-like phenotype in pancreatic and prostate cancer, in embryonic development the induction of ZEB2 triggers differentiation away from the embryonic pluripotent state and towards a neuroectodermal fate. Likewise, interactions with the Rb pathway suggest a tumor-suppressive role for ZEB2, but not ZEB1. In contrast, enhanced expression of ZEB2 in several forms of malignancies indicates a tumor-promoting role. Further questions arise on the role of the miR-200/ZEB feedback inhibition loop in the different stages of tumor metastasis, where temporal and spatial regulation of the balance within the loop may be crucial to allow both successful metastatic migration and colonization. Clearly, a significant amount of further research is needed before our understanding of molecular pathways regulating the ZEB protein family can be converted into novel treatment approaches.


The authors apologize to those colleagues whose relevant studies were not cited here.