Cancer and Metastasis Reviews

, Volume 28, Issue 1, pp 151–166

E-cadherin, β-catenin, and ZEB1 in malignant progression of cancer


  • Otto Schmalhofer
    • Department of Visceral SurgeryUniversity of Freiburg
  • Simone Brabletz
    • Department of Visceral SurgeryUniversity of Freiburg
    • Department of Visceral SurgeryUniversity of Freiburg

DOI: 10.1007/s10555-008-9179-y

Cite this article as:
Schmalhofer, O., Brabletz, S. & Brabletz, T. Cancer Metastasis Rev (2009) 28: 151. doi:10.1007/s10555-008-9179-y


The embryonic program ‘epithelial-mesenchymal transition’ (EMT) is activated during tumor invasion in disseminating cancer cells. Characteristic to these cells is a loss of E-cadherin expression, which can be mediated by EMT-inducing transcriptional repressors, e.g. ZEB1. Consequences of a loss of E-cadherin are an impairment of cell-cell adhesion, which allows detachment of cells, and nuclear localization of β-catenin. In addition to an accumulation of cancer stem cells, nuclear β-catenin induces a gene expression pattern favoring tumor invasion, and mounting evidence indicates multiple reciprocal interactions of E-cadherin and β-catenin with EMT-inducing transcriptional repressors to stabilize an invasive mesenchymal phenotype of epithelial tumor cells.


E-cadherinEMTZEB1CancerInvasionFeedback/forward loop

1 Introduction

Tumors of epithelial origin develop from benign precursor lesions to invasive carcinomas and metastases. In certain epithelial tumors, e.g. colorectal carcinomas, activation of the embryonic program ‘epithelial-mesenchymal transition’ (EMT) is crucial for the dissemination and invasion of cancer cells [1]. Loss of epithelial differentiation and acquisition of a mesenchymal phenotype allows detachment of cancer cells from the primary tumor mass and dissemination into the surrounding stroma [2]. The most important event of EMT is loss of E-cadherin, which was demonstrated to be a prerequisite for epithelial tumor cell invasion. E-cadherin, encoded by the CDH1 gene, has dual functions in epithelial cells: As cell-cell adhesion molecule and as negative regulator of the canonical WNT signaling cascade, in particular of its central mediator β-catenin. Several transcriptional repressors of the CDH1 gene have been indentified. They include the basic helix-loop-helix (bHLH) family members twist and the E2A gene product E12/47, members of the snail family, snail1 and snail2, and members of the zinc finger homeobox (ZFH) family of repressors, ZEB1 and ZEB2 [314]. The purpose of this review is to summarize the knowledge about the molecular interconnections of E-cadherin, β-catenin, and WNT signaling to these transcriptional repressors, in particular to ZEB1. Subsequently, the contribution of this interdependence to induction of EMT, tumor cell invasion, and metastasis will be discussed and exemplified by focusing on the colorectal carcinoma as a model tumor, which is one of the best studied epithelial tumors.

2 Physiology of E-cadherin

2.1 E-cadherin in development and tissue homeostasis

A hallmark of epithelial cell layers is expression of E-cadherin, which is localized at the basolateral membrane in adherens junctions, defining apico-basal polarity [15]. As the core molecule of adherens junctions E-cadherin connects neighbouring epithelial cells by calcium-dependent homotypic interactions of its extracellular tail. The cytoplasmic part of E-cadherin interacts with the other components of adherens junctions, in particular the armadillo repeat proteins p120-catenin, γ-catenin/plakoglobin, and β-catenin. By binding of β-catenin to α-catenin the adherens junction complex is linked to the cortical actin cytoskeleton, thereby mediating mechanical stability [15, 16]. Impairment of the adherens junction components, in particular E-cadherin and β-catenin, has been demonstrated to play an important role in induction of EMT in development and tumorigenesis [1719]. This raises the question of how E-cadherin and β-catenin contribute to maintain an epithelial differentiation. For further reading about adherens junctions and the contributions of p120-catenin and γ-catenin/plakoglobin to epithelial homeostasis and tumorigenesis, see fig. 1 and the excellent review of Perez-Moreno et al. and references therein [16].
Fig. 1

Adherens junctions. Adherens junctions allow homophilic cell-cell adhesion through direct, Ca++ dependent interaction of E-cadherin (E) molecules. E-cadherin is indirectly linked to the actin and microtubule cytoskeletons, through associated proteins, which is essential for cell-cell adhesion. E-cadherin’s direct interaction partner β-catenin (β) binds to α-catenin (α) and links cadherin/catenin complexes to the actin cytoskeleton. Thereby α-catenin is the central player in the linkage to F-actin, a process critical for coordinating actin dynamics in the cell. β-catenin binds to the E-cadherin cytoplasmic tail via its armadillo repeats. The affinity for this key interaction is increased by phosphorylation of several serine residues in the cadherin tail and reduced by phosphorylation (P) of β-catenin Y654, a known site of action for activated growth factor receptor tyrosine kinases (RTK). Thus, through posttranslational modifications, the strength of the adherens junction complex can be adapted to the particular needs of the epithelial cell within the context of its tissue. β-catenin and γ-catenin (γ) compete for binding to E-cadherin and a-catenin. β-catenin also interacts with microtubule-associated proteins such as IQGAP and the dynein/dynactin complex. β-catenin and IQGAP also crossreact with the Rho/rac-family of small GTPases. p120-catenin (p120) binds independently to E-cadherin and promotes E-cadherin clustering. p120-catenin has also been suggested to regulate the GTPase Rho

Much of the understanding of E-cadherin and β-catenin function was achieved by experiments with mouse model systems. In mouse embryogenesis zygotic E-cadherin expression starts between the late 4- and the 8-cell stage at E2.0, but is lost gradually in the trophectoderm and parietal endoderm after implantation of the developing embryo into the uterus [2022]. During gastrulation E-cadherin expression is temporarily suppressed in cells of the epiblast, which loose their epithelial differentiation and acquire a mesenchymal phenotype. These cells delaminate at the primitive streak to populate the intraembryonic mesoderm and the definitive endoderm. Cells of the definite endoderm ultimatively line the gut tube and in these cells loss of CHD1 gene expression is only transient, as they start to reexpress E-cadherin and differentiate into a stratified intestinal epithelium [2226]. Upregulation of E-cadherin expression is also observed during development of the kidney, brain, and melanocytes [2731]. These results indicate that E-cadherin expression is dynamically regulated in embryogenesis. E-cadherin is transiently lost in migrating cells, but reexpressed, when cells start to differentiate in epithelial tissues. The physiological importance of E-cadherin for organ development and tissue morphogenesis also becomes evident from different gene targeting experiments of the CDH1 gene in the mouse. Ectopic injection of CDH1+/+ embryonic stem cells gave rise to teratomas, which were composed of differentiated cell types originating from all three germ-layers. In contrast, tumors derived from CDH1−/− embryonic stem cells failed to do so. Of note, reconstitution of E-cadherin expression exclusively caused formation of epithelia [32]. In the developing embryo E-cadherin is a key player during morula compaction and CDH1 null embryos die early during embryogenesis as development of the blastocyst and trophectoderm is impaired [33, 34]. Conditional CDH1 gene targeting approaches revealed severe defects of epidermal as well as mammary and thyroid gland development [3339]. Loss of E-cadherin expression in mammary epithelial cells was associated with impaired organ function, as alveolar structures in the lactating gland were not properly built up [35]. Furthermore, loss of E-cadherin expression in the skin was demonstrated to impair establishment of apico-basal cell polarity and formation of tight junctions [36]. A transgenic approach, causing overexpression of E-cadherin in intestinal crypts, resulted in reduced proliferation and increased apoptosis in the crypt as well as slower migration of enterocytic cells along the crypt villus axis [40].

Taken together, in embryogenesis E-cadherin is important for proper development, organ morphogenesis and tissue formation. Thereby, E-cadherin is a determinant of epithelial cell differentiation. E-cadherin not only is an important mediator of cell-cell adhesion, but also contributes to regulate diverse cellular processes, such as proliferation, migration, apoptosis or maintenance of epithelial cell polarity. This raises the question by which means E-cadherin can influence such a broad variety of responses. E-cadherin must emit downstream signals and the main mediator of these E-cadherin dependent effects appears to be β-catenin. For the involvement of other molecules involved in E-cadherin associated signaling refer to the review of Cavallaro and Christofori [41]. Strikingly, β-catenin was demonstrated to localize to the nucleus, after E-cadherin expression is lost [4244].

2.2 E-cadherin, β-catenin, and WNT signalling

β-catenin functions in a dual manner in epithelial cells, depending on the intracellular localization (Fig. 2). At the plasma membrane β-catenin is an important component of adherens junctions, acting in cell-cell adhesion by linking E-cadherin, in conjunction with α-catenin, to the actin cytoskeleton. However, β-catenin can also act as the main effector of the canonical WNT signaling cascade in the nucleus. In brief, free cytosolic β-catenin is quickly turned over, unless the WNT signaling cascade is activated. Cytosolic β-catenin assembles with a multiprotein degradation complex, consisting of serine/threonine kinases CK1 and GSK3β, the adenomatous polyposis coli protein (APC), and the scaffold protein axin. This destruction complex targets β-catenin for proteasomal breakdown by CK1- and GSK3β-dependent phosphorylation of N-terminal serin/threonin residues [4551] (Fig. 2A). Binding of WNT molecules to the serpentine frizzled receptors activates the signaling cascade and LRP5/6 coreceptors are recruited to transduce the signal to the cytoplasm, causing inhibition of GSK3β [52, 53]. Thus, cytoplasmic β-catenin evades degradation, accumulates in the cytoplasm and finally translocates to the nucleus. There β-catenin acts as transcription factor in a complex with the HMG-box proteins of the TCF/LEF family [5456]. TCF/LEF molecules interact with DNA in promoter regions of target genes through their HMG-boxes in a sequence-specific manner by recognizing the consensus sequence motif T/A T/A CAAAG [57] (Fig. 2B). Thereby, TCFs act contrarily in regulating target gene transcription, depending on the nuclear β-catenin level. In absence of nuclear β-catenin various additional proteins with transcriptionally repressive properties, such as members of groucho/TLE protein family, are recruited and serve as intermediary molecules to recruit histone deacetylases, thus causing inhibition of target gene transcription by condensing chromatin. Numerous additional interactions with inhibitory factors have been reported, for instance with CtBP. Otherwise, accumulating nuclear β-catenin displaces groucho/TLE proteins, forming a bipartite complex with TCFs. β-catenin provides the transactivating domain by its C-terminus, but may also interact with CBP/p300 or BRG1 for chromatin relaxation to efficiently initiate transcription [5861].
Fig. 2

The canonical WNT signaling pathway. (A) In absence of WNT signals, β-catenin (β) is localized in adherens junctions at the plasma membrane, acting in cell-cell adhesion. Cytosolic β-catenin is targeted for proteasomal breakdown by a multiprotein degradation complex. (B) Binding of WNT molecules to the frizzled receptors inhibits activity of the destruction complex and allows β-catenin to accumulate and to translocate to the nucleus, where a specific transcriptional program is activated. (C) In colorectal carcinomas the signaling cascade is abberantly activated, e.g. due to mutations in the APC gene, thus impairing degradation of β-catenin, which accumulates, and translocates to the nucleus. There, a specific target gene program, independent of exogenous WNT signals, is permanently activated, thus driving malignant tumor progression. Figure is adapted from Fodde et al. [51]

During embryogenesis canonical WNT signaling, defined by nuclear accumulation of β-catenin, was demonstrated to promote invasive trophoblast differentiation, and activation of the signaling cascade was suggested to contribute to trophoblastic hyperplasia and local invasion [62]. Furthermore, β-catenin is involved in induction of EMT in physiological processes, e.g. gastrulation or heart cushion formation [6365]. This is supported by the fact that targeted disruption of β-catenin in mice leads to very early embryonic lethality by affecting development at gastrulation [66]. Canonical WNT signaling emerged also as a potent regulator of various additional developmental processes, contributing to hair follicle morphogenesis, thymocyte and neural delevopment, as well as angiogenesis or adipogenesis [6772]. WNT signaling is also essential for intestinal development. Thereby, β-catenin binds TCF4, a TCF-family member, specific for the intestinal and mammary epithelium [73]. TCF4 expression characterizes the intestinal stem cell compartment and targeted disruption of TCF4 in mice leads to a severe impairment of gut development [74]. The molecular role of β-catenin in the development of these organs and tissues can be deduced by the functions of its defined target genes (see below). Some of these target genes give the competence to act as a stem cell [75]. This is in line with the observed loss of stem cells after targeted deletion of β-catenin in intestine or skin [68, 74]. The function in characterizing the stem cell compartment also indicates the role of nuclear β-catenin in the adult organism. Thereby β-catenin maintains tissue homeostasis, in particular in strong-proliferative, self renewing tissues, like skin or gut.

In recent years it became clear that E-cadherin and the associated catenins, in particular β-catenin, not only are static components of adherens junctions, but in addition act as important mediators of downstream signaling cascades. In this respect, E-cadherin contributes to maintain epithelial differentiation in two ways. First, by establishing cell-cell adhesion and apico-basal polarization, and second by acting as negative regulator of active WNT signaling. By recruiting β-catenin to adherens junctions, nuclear localization of β-catenin and thus β-catenin activated target gene expression is inhibited. Still, loss of E-cadherin is not necessarily linked to nuclear localization of β-catenin and transcriptional activation of responsive target genes. It appears that in addition to loss of E-cadherin expression, also components of the destruction complex (see above) have to be impaired to allow β-catenin to localize to the nucleus and exert its transcriptional activity [76, 77]. Of utmost importance is the interaction of GSK3β, APC, and axin with β-catenin [78, 79]. These molecules are negative regulators of the WNT pathway and only if these molecules are functionally assembled in the destruction complex, the amount of nuclear β-catenin can be efficiently regulated. Compiling all data, it is becoming evident that the interconnection of E-cadherin, β-catenin, and WNT signaling not only controls single events, but contributes to control the complex morphogenetic process of EMT. This regulation needs spatial and temporal coordination of multiple events like proliferation, migration, cell-cell adhesion, and differentiation. Loss of E-cadherin expression or loss of function of either APC, axin or GSK3β leads to a reduced degradation and subsequent overexpression of free cytoplasmic β-catenin, which can exert its nuclear function without control (see below). Not surprisingly, E-cadherin and β-catenin have been described to contribute to EMT also in tumorigenesis [19].

3 E-cadherin in cancer

Sustained CHD1 gene expression contributes to retain an epithelial phenotype and E-cadherin was demonstrated to act as an inhibitor of invasion in numerous tumor cell lines and in in vivo tumor models. In the Rip-Tag mouse model of beta-cell tumors of the pancreas, E-cadherin loss is a causal prerequisite for progression from adenoma to invasive carcinomas, and in human an inverse correlation between expression of E-cadherin and survival of tumor patients was demonstrated [80, 81]. For aggressive carcinomas deregulation of E-cadherin and impairment of adherens junction mediated cell-cell adhesion by different means were reported [80, 8286]. These different possibilities include genetic alterations, epigenetic inactivation, and transcriptional silencing of the CHD1 gene.

3.1 Genetic and epigenetic inactivation of E-cadherin

Genomic alterations of the CDH1 gene causing loss-of-function of E-cadherin, have been identified in a variety of tumors. In breast cancer, somatic E-cadherin mutations were reported exclusively for sporadic lobular subtype, however not for breast cancers of other histopathological subtypes [87]. Furthermore, somatic mutations of the CHD1 gene were identified in endometrial and ovarial carcinomas, as well as in primary gastric cancers and gastric cancer cell lines [8890]. In addition, germline mutations of the CHD1 gene were discovered in patients, predisposed to suffer from diffuse gastric cancer [91, 92].

Another means to loose E-cadherin expression in tumor cells is CpG-island hypermethylation of the CDH1 gene promoter [93, 94]. Aberrant methylation of 5′CpG islands in the promoter region of CDH1 was reported for different types of mammary carcinomas and increased with malignant progression [95, 96]. Other reports showed methylation in thyroid, hepatocellular, and prostate cancer, as well as in laryngeal cancer and colorectal cancer [97, 98].

However, although these mechanisms have been described in numerous studies, the most frequent event appears to be another mechanism of transcriptional inhibition of E-cadherin: silencing of CHD1 gene expression by specific transcription factors. Several specific transcriptional repressors, interacting with E-boxes located in the proximal CDH1 gene promoter, have been identified so far (see below).

Taken together, loss of functional E-cadherin, and thus of adherens junction-mediated cell-cell contacts, enables the first step of metastasis: local invasion and dissemination of cancer cells from the main tumor mass.

3.2 E-cadherin and EMT

Cancer cells disseminating from the main tumor mass frequently are characterized by hallmarks of EMT [19]. EMT is a developmental process fundamental to various steps of embryogenesis and has been implicated e.g. during embryo implantation into the uterus, gastrulation, palatal fusion, heart cushion formation, development of the peripheral nervous system, and branching morphogenesis of the mammary gland [65, 99103]. Thereby, cells loose their epithelial differentiation and acquire a mesenchymal, spindle-shaped phenotype, characterized by loss of cell-cell adhesion, notably disruption of the cadherin–catenin complex, and loss of apico-basal polarization, as well as acquisition of migratory and invasive capabilities [104, 105]. Epithelial markers include cytokeratins 8 and 18, as well as E-cadherin localized at the plasma membrane. In contrast, vimentin expression, membraneous N-cadherin localization, and cytoplasmic relocation or loss of E-cadherin are indicative of a mesenchymal phenotype [106]. Inducers of EMT, such as hepatocyte growth factor, tumour necrosis factor α (TNFa), and transforming growth factor (TGF)β are produced by infiltrating cells or the tumor cells themselves, and trigger expression of a variety of transcriptional repressors. These transcriptional repressors are the intracellular mediators of EMT and were originally identified as inhibitors of E-cadherin expression, contributing to malignant progression of numerous human tumors by repression of CDH1 [107]. These factors include the basic helix-loop-helix (bHLH) family members twist and the E2A gene product E12/47, members of the snail family, snail1 and snail2, and members of the zinc finger homeobox family of repressors, ZEB1 and ZEB2 [314]. However, although their involvement in tumorigenesis is just unravelling, it goes far beyond transcriptional inhibition of E-cadherin expression. The importance of these proteins can be deduced from a steadily growing number of identified target genes, e.g. components of the basement mebrane, epithelial cell polarity factors or apoptosis pathways. In addition, most recently two studies reported induction of stem cell phenotypes in breast cancer by EMT [108, 109]. A comprehensive overview on the field was published most recently by Peinado et al. [110]. Whereas in the first years after the notion that EMT is involved in tumorigenesis, the EMT inducers snail1 and snail2 were in the limelight of attention, more recently another player entered center stage: ZEB1.

3.3 ZEB1 in cancer

The transcription factor ZEB1, encoded by the TCF8 gene, is the vertebrate homologue of the ZFH gene family of zinc finger/homeodomain proteins. Much insight on physiological functions was gained by experiments with model organisms. In Drosophila melanogaster, ZEB1 regulates cell fate determination, as well as development of the central nervous system and of mesodermally derived tissues [111, 112]. Involvement in mammalian development was suggested by a conventional gene targeting approach of the TCF8 gene in mouse. Heterozygous animals were vial and fertile, whereas homozygous mutants developed to term and died after birth. Their phenotype was characterized by a severe T-cell deficiency of the thymus, as well as craniofacial, cartilage, and skeletal defects. Mutation of ZEB1 caused a mesenchymal-epithelial-like transition characterized by ectopic expression of E-cadherin in a variety of tissues in these mice, such as nasal and palate mesenchyme or the developing cartilage, as well as embryonic fibroblasts, which clearly displayed an abnormal epithelial-like morphology [113115].

ZEB1 interacts with the regulatory regions of responsive target genes, by binding to the sequence motif CAGGTG/A, related to the consensus E-box sequence CANNTG [116119]. ZEB1-dependent repression of genes involved in the development of hematopoietic cells, as well as of extracellular matrix (ECM) components was reported [116, 117, 120124]. Although the research focused on the ability of ZEB1 to repress target gene transcription, increasing evidence demonstrates that ZEB1 may also act as transcriptional activator, thereby interacting with acetyltransferases p300/pCAF and activated SMADs. This appears to target genes such as smooth muscle actin and myosin or components or the VitD signaling pathway, which were demonstrated to contribute to mesenchymal differentiation [125130].

Numerous signaling pathways were demonstrated to regulate TCF8 gene expression. TGFβ and TNFα are important mediators of EMT and were shown to be activators of TCF8 gene expression in various cancer cell lines [131136]. Also IGF1 signaling was suggested to induce EMT in cancer and demonstrated to activate expression of ZEB1 [137140]. Accumulating evidences also indicates a role of epidermal growth factor receptor (EGFR) signaling in cancer cell metastasis, in particular through activation of the TCF8 gene and inhibition of E-cadherin expression [141145]. Furthermore estrogen and progesterone dependent activation of the TCF8 gene was reported [126, 146]. A recent report demonstrated a COX2-dependent inhibition of E-cadherin expression. The COX2 metabolite PGE2 induced expression of snail1 and ZEB1 [147]. Repressors of TCF8 are a certain isoform of ZEB1 itself and the Rb-E2F1 cell cycle repressor complex [148, 149].

In human cancers, high ZEB1 expression correlated significantly with a high Gleason score in prostate carcinomas and was suggested as biomarker for metastasis [140, 150].

Moreover, a role of ZEB1 was implicated in the tumorigenesis of renal clear cell carcinomas (RCC). In RCC, expression of ZEB1 was shown to be regulated by HIF1 and, typically, in biopsies of renal clear cell carcinomas (RCC) the expression patterns of HIF1 and E-cadherin were mutually exclusive [151]. Aberrant ZEB1 expression was also detected in tumors of uterine origin, in particular in myometrium-derived malignant leiomyosarcoma and in the stroma of low grade (type I) endometrioid adenocarcinomas. No epithelial ZEB1 expression was observed in normal endometrium and low-grade endometrial cancers. However, in more aggressive type II cancers of endometroid origin, such as FIGO grade 3 endometrioid adenocarcinomas, uterine papillary serous carcinomas, and in carcinosarcoidal tumors of the malignant mixed Mullerian tumor type, ZEB1 is aberrantly expressed in tumor cells of epithelial origin. The high expression level of ZEB1 correlated with loss of E-cadherin expression in surgical specimens and increased migratory and invasive potential of endometroid cancer cells [152, 153]. ZEB1 seems also to be involved in malignant progression of breast cancer, Screening of breast cancer cell lines revealed an inverse correlation of TCF8 and CHD1 gene transcript levels and cell lines with high ZEB1 expression displayed a spindle-like, fibroblastoid phenotype [146, 154156]. Another regulator of ZEB1 in breast cancer appears to be CCN6. Low CCN6 expression caused an upregulation of snail1 and ZEB1 and was associated with decreased expression of E-cadherin and axillary lymph node metastasis in breast cancer [157159]. A correlation of ZEB1 expression and loss of E-cadherin was demonstrated in tumor cell lines of lung adenocarcinomas of the non small cell lung cancer (NSCLC) subtype [160]. Another study suggested that E-cadherin repression is regulated in a COX2-dependent manner in NSCLC [147]. Thereby ZEB1 is induced by the COX2 metabolite prostaglandin E2 (PGE2). The inverse pattern of E-cadherin and ZEB1 expression was also demonstrated in human lung adenocarcinomas in vivo, suggesting that PGE2 acts as a paracrine or autocrine modulator of ZEB1 expression.

In summary, ZEB1 contributes to malignant progression of various epithelial tumors, Thereby, ZEB1 is a crucial mediator of EMT, exerting its effects on induction of EMT by inhibiting expression of E-cadherin and miRNAs, which induce an epithelial phenotype [135, 144, 161]. This raises the question of interconnections of ZEB1, E-cadherin and β-catenin with WNT signaling and EMT in cancerogenesis. This point will adressed using the colorectal carcinoma as model tumor in the next section.

4 Compilation: colorectal cancer as a model tumor

The initial mutational event in colorectal carcinogenesis, common to almost all hereditary and sporadic colorectal tumors, is an aberrant activation of the canonical WNT signaling pathway and its key downstream effector, β-catenin [162165]. In 80% of sporadic colorectal carcinomas and in familiar adenomatous polyposis (FAP), the founding genetic alteration is a loss-of-function mutation in the APC tumor suppressor gene, causing stabilization of the key effector of WNT signaling, β-catenin (Fig. 2C) [166]. Thus, the β-catenin mediated transcriptional program is not longer restricted to its physiological compartment at the bottom of colonic crypts. The permanent WNT signal disrupts the proliferation/differentiation switch along the crypt axis, causing perturbance of mucosal homeostasis and initiating the multistage process of malignant progression towards invasion and metastasis [167, 168]. However, although all neoplastic cells of well-to-moderately differentiated colorectal adenocarcinomas harbor mutations in the WNT signaling pathway, these tumors are characterized by a temporal and spatial heterogeneity. That is an increasing amount of nuclear β-catenin during malignization, as well as an intratumorous morphological heterogeneity [165, 169, 170]. The central tumor mass resembles normal colonic mucosa, as neoplastic cells retain a polarized, epithelial phenotype and grow in tubular structures. These cells lack nuclear β-catenin, which instead is localized at the plasma membrane, associated with E-cadherin in adherens junctions, and involved in cell-cell adhesion. The polar morphology of the epithelium is lost with malignization towards high-grade undifferentiated tumors, and is often coupled to a loss of membranous E-cadherin. Contrarily, solitary disseminated neoplastic cells in the periphery, at the invasive front, have undergone an EMT, that is they lost epithelial characteristics, such as membranous E-cadherin, and acquired a fibroblastoid phenotype [19, 165]. These budding tumor cells accumulate β-catenin in the nucleus, which is a direct factor for prognosis and thus of utmost clinical importance [171]. However, in the central areas of metastases tumor cells display again an apico-basically differentiated epithelial phenotype, re-express E-cadherin and recapitulate the morphology of the primary tumor, indicating a MET during metastasis formation [169]. β-catenin colocalizes at the plasma membrane with reexpressed E-cadherin in adherens junctions and thus again contributes to cell-cell adhesion. This relocation from the nucleus to the plasma membrane is associated with silencing of WNT target gene expression and colonic epithelial cell differentiation [19, 172174]. However, invading tumor cells in the metastasis periphery have again undergone EMT. That is they lost the differentiated phenotype and display nuclear β-catenin (Fig. 3). This indicates the involvement of an inductive tumor environment, regulating the cycling dynamics of EMT/MET during malignant progression of colorectal carcinomas [169, 175].
Fig. 3

Nuclear accumulation of β-catenin at the tumor invasion front is associated with EMT. Heterogenous expression of E-cadherin (A and B) and β-catenin (C and D) in human colorectal adenocarcinomas. Serial sections of central (A and C) and invasive (B and D) areas of the same tumor. Specific staining in brown, nuclear counterstaining in blue. Note membranous staining of β-catenin and E-cadherin in central, differentiated tumor areas, characterized by tubular structures, built up by more or less polarized epithelial tumor cells (arrows in A, C). No or little nuclear β-catenin. Dedifferentiation towards a mesenchyme-like phenotype of tumor cells at the invasive regions, with loss of membranous E-cadherin and β-catenin, but cytoplasmic E-cadherin and nuclear and cytoplasmic β-catenin (arrows in B, D)

β-catenin has been imlpicated in cancer stemness and among the genes activated by β-catenin during colorectal carcinogenesis are factors associated with proliferation, such as cyclin D1 or c-myc, and genes associated with tumor cell survival and a tumor stem cell phenotype, such as survivin and MDR [176181]. These traits are necessary to give rise to a primary tumor during early stages of tumorigenesis, but insufficient to promote invasion and metastasis. Among the genes targeted by nuclear β-catenin and selectively expressed at the tumor-host interface are many of crucial importance to tumor invasion, such as L1CAM, CD44, TNC, VEGF, PLAUR, PLAU, MMP7, MMP14, LAMC2 and JUN. To invade the host stroma, tumor cells must loose epithelial differentiation and acquire a mesenchymal, motile phenotype. The β-catenin-dependent transcription program being activated in budding tumor cells also comprises mesenchymal markers, e.g. fibronectin, as well as repressors of epithelial differentiation, such as CDX1 and ENC1 [19, 165].

However, the most important hallmark of epithelial differentiation is expression of E-cadherin. Its loss is a prerequisite for detachment, invasion and finally dissemination and metastasis of neoplastic cells and can be linked directly to an activated WNT signaling cascade (Fig. 4). Snail1 is overexpressed in human primary colorectal carcinomas, and was demonstrated to interact with β-catenin, thereby pomoting WNT target gene expression in colorectal cancer cells [182, 183]. Furthermore, overexpression of snail1 in colorectal cancer cell lines, caused upregulation of ZEB1 [7]. Nuclear localization of β-catenin was furthermore demonstrated to induce expression of snail2 in colorectal cancer cells and snail2 expression in primary colorectal carcinomas correlated significantly with metastatic spread of the tumor [184, 185]. As a consequence of snail1 and snail2 mediated repression of the CHD1 gene, loss of E-cadherin could be a trigger to induce expression of ZEB1, which was suggested to be important for maintenance of an invasive phenotpye of epithelial cancer cells [186]. A recent study using a human breast epithelial cell system demonstrated that loss of E-cadherin expression caused upregulation of EMT associated transcriptional repressors, in particular ZEB1 and twist, suggesting a feed forward loop of E-cadherin repression [44]. Both factors have been described to be selectively expressed at the invasion front of colorectal carcinomas, which can be explained by signals from the tumor environment [187, 188]. Prominent among the stimuli to induce EMT in colorectal cancer cells are TGFβ and TNFα, both of which upregulate ZEB1 expression [135, 189]. In addition, other putative links between, β-catenin and ZEB1 are provided by described WNT targets. Overexpression of the WNT target COX2 is frequently observed in colon cancer and is associated with tumor metastasis and poor prognosis [190195]. As ZEB1 is activated by the COX2 metabolite PGE2 in lung adenocarcinomas and as COX2 inhibitors upregulate E-cadherin expression in colorectal carcinoma cells, β-catenin dependent induction of COX2 might also activate expression of the TCF8 gene in colorectal carcinomas [147, 190, 195]. As COX2/PGE2 causes activation of a β-catenin dependent target gene activation in colon cancer, this suggests another feed forward loop to stabilize ZEB1 expression and to induce EMT [196]. Another link between ZEB1 and WNT signaling is provided by IGF1. IGF1 is a WNT target and IGF1 signaling, in turn, was demonstrated to activate the WNT signaling cascade, suggesting a feed forward loop [194, 197]. As IGF1 was also demonstrated to induce expression of ZEB1 in prostate cancer cell lines, ZEB1 might enhance this loop in colorectal carcinomas by stabilizing nuclear localization of β-catenin via suppression of E-cadherin, thus stabilizing its own expression.
Fig. 4

Putative functional interdependence of E-cadherin, β-catenin and ZEB1 to induce EMT in colorectal cancer. External stimuli like TGFβ and TNFα might activate expression of EMT inducers, in particular snail1, snail2 and ZEB1, at the invasion front of colorectal cancers. By repressing E-cadherin expression, β-catenin is liberated from the membrane compartment and localizes to the nucleus and activates, together with snail1, transcription of responsive target genes that support tumor cell invasion. Some WNT targets like snail2 directly stabilize a mesenchymal phenotpye, whereas others, like COX2 and IGF1, might indirectly contribute to mesenchymal differentiation by induction and maintenance of TCF8 gene expression. ZEB1 exerts its effects to induce EMT by different means, e.g. suppression of basement membrane components (BM), cell polarity factors and members of the miR200 family

ZEB1 expression has been associated with tumor cell migration, invasion, and metastasis, in particular in colorectal carcinomas [140, 144, 152, 188, 198]. Epithelial tissues are surrounded by a basement membrane, which induces epithelial differentiation and basal-apical polarization. ZEB1 interferes with these traits of an epithelial phenotype by inhibiting expression of basement membrane components, of E-cadherin and desmosomal proteins, as well as intracellular effectors of cell polarity [155, 156, 188, 198]. Overcoming the basement membrane is a prerequisite for epithelial tumor cells to invade the stroma and subsequently to disseminate and metastasize [199]. In this context, ZEB1-mediated loss of the basement membrane contributes to tumor progression and disturbance of tissue architecture in a dual fashion: by loosing a mechanical barrier and by loss of inductive signals, and thus subsequent loss of epithelial polarity. ZEB1 also inhibits basal-apical differentiation by directly targeting cell polarity factors, such as AP1M2, PATJ, and CRB3 [156, 198, 200, 201]. These data also demonstrated a direct role of ZEB1 in metastasis by suppression of LGL2. Indeed, LGL2 is classified as tumor suppressor, as loss of its Drosophila homologue lgl caused metastatic tumors [202]. ZEB1 also inhibits epithelial differentiation by transcriptional inhibition of the microRNA (miR)-200 family. miR-200 family members were demonstrated to be expressed in epithelial tissues and negatively correlated with expression of TCF8 [203, 204]. When transfected into tumor cell lines of colorectal origin, a mesenchymal to epithelial differentiation was induced by miR-200 family members, including upregulation of E-cadherin expression. The miR-200 family inhibited migration and invasion of undifferentiated cancer cell lines. Intriguingly, ZEB1 is also a direct target of the miR-200 family and ZEB1 was demonstrated to directly suppress transcription of two members of the miR-200 family, miR-141 and miR-200c [135]. As both miRNAs affect the expression of molecules acting in a proinvasive manner, such as cofilin, the leptin receptor, TGFβ2 and, in particular, ZEB1, this suggests an EMT-enhancing feed forward loop in invading cancer cells [131, 132, 135, 161, 205]. Supporting this model, knockdown of ZEB1 in colorectal cancer cells abolished liver metastasis in a nude mice xenograft approach [198]. This is in line with the observation that expression of ZEB1 and selective loss of basement membrane in invasive tumor regions of colorectal carcinomas are a strong predictor of poor patient survival and metachronous distant metastasis [188].

5 Conclusions

An enormous knowledge about the invasion and metastasis of epithelial cancer cells has been gained in the last 20 years. This process is best understood in colorectal cancers. In colorectal carcinomas disseminating cancer cells undergo EMT at the tumor-host interface, that is they loose their epithelial differentiation and acquire a mesenchymal phenotype, allowing detachment of cancer cells from the primary tumor mass. The most important hallmark of EMT is loss of E-cadherin, which is mediated by different transcriptional repressors, e.g. ZEB1 and snail1. E-cadherin has a dual function in epithelial cells as cell-cell adhesion molecule and as negative regulator of the canonical WNT pathway, in particular its central mediator β-catenin. After loss of E-cadherin expression, β-catenin localizes to the nucleus and activates a target gene expression program, linking EMT to WNT signaling. However, WNT signaling is also linked to EMT, by direct activation of snail2 or by indirect activation of ZEB1 via other WNT target genes, e.g. COX2 or IGF1. Taken together, this suggests a feed forward loop of dedifferentiation in invading cancer cells in colorectal carcinomas. This loop involves E-cadherin, β-catenin and ZEB1 and interference with these crosstalks might offer novel therapeutic opportunities in the future.


This work was supported by grants to T.B. from the EU MCSC contract no. 037297, the DFG (no. BR 1399/4-3) and the Deutsche Krebshilfe (no. 106958).

Copyright information

© Springer Science+Business Media, LLC 2009