|Home | About | Journals | Submit | Contact Us | Français|
Epithelial to mesenchymal transition (EMT) is implicated in the progression of primary tumours towards metastasis and is likely caused by a pathological activation of transcription factors regulating EMT in embryonic development. To analyse EMT-causing pathways in tumouri-genesis, we identified transcriptional targets of the E-cadherin repressor ZEB1 in invasive human cancer cells. We show that ZEB1 repressed multiple key determinants of epithelial differentiation and cell–cell adhesion, including the cell polarity genes Crumbs3, HUGL2 and Pals1-associated tight junction protein. ZEB1 associated with their endogenous promoters in vivo, and strongly repressed promotor activities in reporter assays. ZEB1 downregulation in undifferentiated cancer cells by RNA interference was sufficient to upregulate expression of these cell polarity genes on the RNA and protein level, to re-establish epithelial features and to impair cell motility in vitro. In human colorectal cancer, ZEB1 expression was limited to the tumour–host interface and was accompanied by loss of intercellular adhesion and tumour cell invasion. In invasive ductal and lobular breast cancer, upregulation of ZEB1 was stringently coupled to cancer cell dedifferentiation. Our data show that ZEB1 represents a key player in pathologic EMTs associated with tumour progression.
Invasion and metastasis are the primary causes of morbidity and mortality in human cancer. Despite their central role in disease progression the molecular mechanisms driving tumour cell invasion and metastasis are poorly understood. The escape of cancer cells from the primary tumour is mostly associated with epithelial dedifferentiation, loss of intercellular adhesion and enhanced migratory potential, collectively termed epithelial to mesenchymal transition (EMT). While EMT is an essential process during embryogenesis, it is a major pathological event in cancer progression (Thiery, 2002; Eger and Mikulits, 2005; Huber et al., 2005).
EMT-inducing pathways, such as transforming growth factor-β (TGFβ) and Wnt signalling trigger epithelial dedifferentiation by impairing the expression/function of the epithelial adhesion protein E-cadherin (De Craene et al., 2005b; Thiery and Sleeman, 2006). Hence, loss of E-cadherin is one of the hallmarks of EMT and late-stage cancer progression, while re-expression of E-cadherin in tumour cells inhibits invasion and metastasis (Peinado et al., 2004).
Repression of E-cadherin is mediated by distinct transcription factors, including Snail1 (Snail) (Batlle et al., 2000; Cano et al., 2000), Snail2 (Slug) (Hajra et al., 2002), SIP1 (ZEB2) (Comijn et al., 2001) and E47 (Perez-Moreno et al., 2001). These repressors may regulate developmental transcriptional programmes of EMT in tumour cells predisposing them to invasion and metastasis (Gupta et al., 2005). Therefore, the comprehensive identification of their bona fide targets is of prime importance for understanding tumour cell invasion at the molecular level.
The zinc finger (C2H2-type) and homeobox pit1, Oct 1/2, Unc 86 homology region (POU domain) containing protein ZEB1 (δEF1, TCF8 and AREB6) has recently been identified as a potent transcriptional repressor of E-cadherin (Grooteclaes and Frisch, 2000; Guaita et al., 2002; Eger et al., 2005). However, in contrast to other E-cadherin repressors, comprehensive studies on ZEB1’s function in cancer cell plasticity are missing. Here, we show that ZEB1 inhibits the expression of epithelial genes central to adhesion and epithelial polarity. In human colon and breast tumours, ZEB1-expressing tumour cells show impaired adhesion and invaded host tissues. We propose that ZEB1 functions as a master regulator of epithelial plasticity in cancer cell invasion and human tumour progression in vivo.
To identify ZEB1 target genes in cancer cells we examined the genome-wide transcriptional repertoire of ZEB1 in the highly metastatic MDA-MB-231 breast cancer cell line, which has been widely used as an invasion and metastasis cell model (Lacroix and Leclercq, 2004). Knockdown of ZEB1 by small interfering RNAs (siRNAs) has previously been shown to induce E-cadherin expression and to re-establish epithelial features (Eger et al., 2005). To determine global gene expression changes in MDA-MB-231 cells following a 3 days knockdown of ZEB1, which reduced ZEB1 mRNA levels by 80–90% (Figure 1a), we performed Affymetrix GeneChip analyses (Human Genome U133 Plus 2.0). ZEB1 knockdown caused upregulation of ~200 genes (repressed by ZEB1) and downregulation of ~30 genes (activated by ZEB1) (complete microarray data will be published elsewhere). ZEB1 mRNA level was strongly reduced, confirming the efficiency of the knockdown, while other E-cadherin repressors were not affected (data not shown).
In this study, we focused on potential ZEB1 targets involved in epithelial differentiation. Besides E-cadherin, ZEB1 depletion also induced re-expression of a multitude of genes crucial for epithelial cell–cell adhesion and differentiation (Table 1). Potential ZEB1 targets include the cell polarity genes Crumbs3, Pals1-associated tight junction protein (PATJ) and human lethal giant larvae homologue 2 (LLGL2; HUGL2); members of the classical cadherin superfamily (placental (P-) and retinal (R-) cadherin); components of tight junctions (occludin, JAM1, claudin 7, tricellulin and shroom), desmosomes (desmoplakin, plakophilin 3, desmocollin 2 and desmoglein 2) and gap junctions (connexin 26 and 31); epithelial-specific adhesion molecules with Ig-like domains (epithelial V-like antigen 1); the apically localized protein Mucin 1 and various genes involved in vesicular trafficking and transcytosis in epithelial cells, including the yeast homologue CDC50p implicated in the polarized establishment of actin patches at bud sites (Table 1). Transcript levels were confirmed by reverse transcription (RT)–PCR in three independent siRNA experiments (Figure 1c, one representative experiment is shown).
We have recently demonstrated that, unlike ZEB1, the expression of Snail1 did not tightly correlate with repression of E-cadherin in a panel of 20 breast cancer cell lines (Eger et al., 2005). To compare directly the functions of ZEB1 and Snail1 in MDA-MB-231 cells, we abrogated expression of Snail1 by siRNA treatment. Although transcript levels of Snail1 were reduced by ~90% (Figure 1a), the downregulation of Snail1 was not sufficient to activate expression of E-cadherin Crumbs3 and HUGL2 (Figure 1b and d). Consequently, MDA-MB-231 cells remained fibroblastoid (data not shown). Thus, ZEB1 rather than Snail1 controls epithelial plasticity of MDA-MB-231 cells.
De-repression of Crumbs3, HUGL2 and PATJ upon ZEB1 knockdown also induced strong upregulation of respective proteins (Figure 2a and b). Upregulated Crumbs3 and HUGL2 accumulated in the cytoplasm and at the plasma membrane in MDA-MB-231 cells (Figure 2a). PATJ accumulated in the nucleus but failed to localize to the membrane (Figure 2a). As PATJ has been implicated in the biogenesis of tight junctions (Michel et al., 2005; Shin et al., 2005), we analysed tight junction formation after ZEB1 knockdown. The tight junction marker ZO1 was detected at cell–cell contact sites (Figure 2a), indicating that PATJ may be dispensable for initial stages of tight junction formation. Altogether, ZEB1 depletion caused de novo expression of cell polarity proteins, their partial translocation to the plasma membrane and formation of rudimentary peripheral tight junction complexes.
Expression of E-cadherin in undifferentiated cancer cells rescues epithelial architecture and affects several signalling pathways (Wheelock and Johnson, 2003). Since E-cadherin upregulation is a hallmark of ZEB1 downregulation in MDA-MB-231 cells (Eger et al., 2005) (Table 1, Figure 1b and d), the observed changes in gene expression upon ZEB1 depletion may be indirectly caused by increased E-cadherin levels. To address this issue, we stably expressed E-cadherin in MDA-MB-231 cells. In 90% of the cells, E-cadherin was detected in the cytoplasm (Figure 2c, E-cadherin-1) and cells were unable to develop pronounced epithelial features (Figure 2c, E-cadherin-1). Only 10% of cells formed ‘epitheloid’ cell clusters in which E-cadherin accumulated at cell–cell contact sites (Figure 2c, E-cadherin-2). RT–PCR analyses revealed that transcription of the ZEB1 target genes (Crumbs3, HUGL2 and PATJ) was unaffected upon expression of E-cadherin (Figure 2d). Crumbs3 and HUGL2 proteins were neither upregulated nor redistributed, even in the 10% of E-cadherin expressing ‘epitheloid’ cells (Figure 2c).
Next, we examined whether ZEB1 can directly repress the promoters of Crumbs3, HUGL2 and PATJ. Since ZEB1 represses transcription via binding to E-box elements (5′-CACCTG-3′) (Grooteclaes and Frisch, 2000; Eger et al., 2005), we screened the proximal promoter regions of Crumbs3 and HUGL2 for the presence of E-box consensus sites and cloned relevant fragments into reporter constructs. For human Crumbs3, we generated three reporter constructs spanning different, albeit overlapping promoter regions (CRB3-prom 1–3, Figure 3a), which cover ~3.5 kbp (−3421 to +74) and contain 20 E-box consensus sites mostly arranged in three clusters (Figure 3a). For HUGL2, a reporter construct containing a ~900 bp fragment with four E-box consensus sites was generated (−743 to +135) (Figure 3c).
Epithelial MCF7 cells expressing endogenous Crumbs3 and HUGL2 were transfected with the reporter plasmids together with a ZEB1 expression vector or empty control vector and a constitutive β-galactosidase expression plasmid for normalization of transfection efficiency. Unlike the control, ZEB1 repressed the activity of the two proximally located Crumbs3 promoter fragments CRB3-prom1 (−1402/ +74) and CRB3-prom2 (−1958/ +74) as well as the HUGL2 promoter fragment HUGL2-prom (−743/ +135) (Figure 3b and c). ZEB1 expression had no effect on the more distally located Crumbs3 promoter region (CRB3-prom3; −3421/−1402) (Figure 3b). Ectopic expression of Snail1 also repressed promoter activities of both genes (Figure 3b and c), but unlike ZEB1, Snail1 preferentially repressed the activity of the most upstream promoter fragment (CRB3-prom3; −3421/−1402).
To test whether ZEB1 can directly interact with the endogenous promoters in vivo, we performed chromatin immunoprecipitations (ChIP). In line with the reporter analyses, chromatin fragments containing the first or the second proximal E-box clusters of the Crumbs3 promoter were efficiently pulled down by ZEB1 antibodies (Abs) (Figure 3d, ChIP1 and ChIP2), whereas fragments located at the most distal E-box cluster were barely detectable above background levels (Figure 3d, ChIP3). Similarly, ZEB1 physically interacted with the proximal PATJ promoter, which contained two E-box consensus sites (Figure 3e). To demonstrate specificity, we knocked down ZEB1 via siRNA before ChIP. Reduction of ZEB1 significantly decreased coprecipitated Crumbs3 and PATJ promoter fragment levels (Figure 3d and e; lower panels, siRNA-ChIP). Thus, ZEB1 can directly repress Crumbs3 and PATJ by binding to defined proximal promoter segments.
Next, we performed transpore migration assays to determine whether knockdown of ZEB1 also affects cell motility in vitro. MDA-MB-231 cells treated with ZEB1-specific siRNA for 3 days were seeded onto Transwell filter inserts and cell migration through pores was analysed 24h later. ZEB1 depletion impaired cell motility by 80% (Figure 4a). The reduced motility is not only a result of E-cadherin, upregulation, since ectopic expression of E-cadherin in MDA-MB-231 cells barely affected motility (Figure 4b). Interestingly, also Snail1 depletion affected transmigration of MDA-MB-231 cells through filters (Figure 4a), although they retained a fibroblastoid morphology.
Our findings suggested that aberrant upregulation of ZEB1 in human tumours may induce cancer cell dissemination. To test this hypothesis, we screened paraffin-embedded human colon and breast neoplasm specimens immunohistochemically for ZEB1, E-cadherin, cytokeratin and HUGL2 expression.
In colon cancer samples, the bulk tumour area stained positively for E-cadherin, cytokeratin and HUGL2 (Figure 5a and Supplementary Figure 1, arrows). ZEB1 expression was not detected in differentiated tumour cells but in many stroma cells adjacent to the tumour areas (Figure 5a and Supplementary Figure 1, arrowheads). In normal colon tissue, ZEB1-positive stroma cells were mostly absent (Supplementary Figure 1, arrowheads). Unspecific goat IgG controls did not reveal specific signals (Supplementary Figure 2, cytokeratin/goat IgG double stainings).
As EMT-like phenotypic conversions at the tumour–host interface are prominent features of colorectal cancer progression (Brabletz et al., 2001, 2005), we screened tumour–host interfaces for ZEB1 expression. In eight out of ten colon tumours, we found strong ZEB1 expression at tumour margins exhibiting tumour cell dedifferentiation. ZEB1-positive cells expressed low levels of cytokeratin, formed loosely attached clusters or invaded the tumour stroma as single cells (Figure 5b and Supplementary Figure 3, arrows). ZEB1-positive cancer cells were clearly distinguishable from tumour-associated stroma cells by perinuclear and/or cytoplasmic cytokeratin remnants and by nuclear morphology (Figure 5b and Supplementary Figure 1, ZEB1/cytokeratin, arrows). In addition, we found co-expression of ZEB1 and β-catenin in a significant number of invasive tumour cells (Supplementary Figure 4, lower panel, arrows).
We also examined ZEB1 expression in eight ductal and five lobular breast cancer specimens. In all ductal carcinomas, ZEB1 expression inversely correlated with cancer cell differentiation. Nuclear accumulation of ZEB1 was always associated with low expression of cytokeratin (Figure 5c, Ductal-Dediff., arrows) and weakened intercellular adhesion, while differentiated tumour areas were strongly positive for cytokeratin and lacked ZEB1 (Figure 5c, Ductal-Diff., arrow). In three ductal carcinomas, ZEB1 was highly expressed in tumour cells invading host tissue as loosely associated group or single cells, reminiscent of infiltrating tumour cells in invasive lobular breast cancer (Figure 5c, Mixed, arrows). Accordingly, the most striking upregulation of ZEB1 was found in breast cancer specimens exhibiting all histological criteria of invasive lobular carcinomas (Figure 5c, Lobular).
Here, we show that, ZEB1 affects expression of several genes critically involved in epithelial polarity, supporting ZEB1’s regulatory role in EMT. Our approach for the identification of ZEB1 target genes was to selectively knockdown endogenous ZEB1, thus avoiding potential overexpression artefacts. Genes affected by ZEB1 include constituents of all junctional complexes located along the lateral membrane of epithelial cells and apical membrane proteins and epithelial Ig-domain adhesion molecules. Most interestingly, we identified ZEB1 as potent direct transcriptional repressor of the cell polarity genes Crumbs3, PATJ and HUGL2, as shown by ChIP and reporter assays. Loss of epithelial cell polarity proteins causes severe defects in epithelial polarity, which is a primary diagnostic mark for malignant carcinomas. Their functional loss also enhances tumour cell proliferation and invasion (Bilder, 2004; Brumby and Richardson, 2005). Hence, ZEB1-mediated repression of Crumbs3, PATJ and HUGL2 in cancer cells may affect multiple aspects of normal epithelial physiology favouring cancer progression, invasion and metastasis.
Like ZEB1 the E-cadherin repressors Snail1, Snail2, SIP1 and E47 also modulate epithelial architecture and induce EMT (Peinado et al., 2004; De Craene et al., 2005b). However, the specific contributions of these repressors to EMT in development and tumour progression are poorly understood. First insights have been gained from recent microarray studies, in which specific transcription programmes were determined by overexpression of these proteins in human colon cancer or Madin–Darby canine kidney cells (De Craene et al., 2005a; Vandewalle et al., 2005; Moreno-Bueno et al., 2006). Despite the different cell lines and experimental set-ups used, a few important conclusions can be drawn from the synopsis of the four data sets. First, among the E-cadherin repressors ZEB1 showed the strongest negative effect on epithelial-specific transcription. Second, none of the other repressors was capable of repressing cell polarity genes. Third, the related repressors ZEB1 and SIP1 may regulate different sets of junctional genes.
Interestingly, Snail1 is not always sufficient to repress E-cadherin, as demonstrated in DLD-1 colon cancer cells overexpressing Snail1 (De Craene et al., 2005a), and in our previous studies, revealing that Snail1 expression did not correlate with E-cadherin repression in human breast cancer cell lines (Eger et al., 2005). In this study, we show that knockdown of Snail1 was not sufficient to activate expression of E-cadherin, Crumbs3 or HUGL2 in MDA-MB-231 cells. Yet, numerous studies have shown a negative impact of Snail1 on E-cadherin expression and epithelial differentiation (Batlle et al., 2000; Cano et al., 2000; Palmer et al., 2004; Peinado et al. 2004; Pena et al., 2005, 2006). In addition, we found that overexpressed Snail1 also affected Crumbs3 and HUGL2 promoter activity, indicating Snail1’s intrinsic capability to affect these genes in breast cancer cells. Therefore, overall expression levels of the E-cadherin repressors and the availability of cofactors may determine target gene specificity and their final impact on EMT, tumour progression and metastasis (Pena et al., 2006).
To analyse the role of ZEB1 in human cancer in vivo we tested ZEB1 expression in colon and breast cancer specimens. In colon cancer, ZEB1 was upregulated at the tumour–host interface and was accompanied by epithelial dedifferentiation and tumour cell invasion. In invasive ductal carcinomas of the breast, ZEB1 upregulation correlated with epithelial dedifferentiation and undifferentiated lobular breast tumours expressed ZEB1 in a large proportion of cells. ZEB1 was also highly expressed in tumour-associated stroma cells. It is unclear whether ZEB1-positive stromal fibroblasts may be derived from epithelial cancer cells through a bona fide ZEB1-dependent EMT in vivo.
Several recent reports provide supporting evidence for a role of ZEB1 in the malignant progression of different cancer types. (1) In non-small cell lung cancer and in renal cell carcinomas ZEB1 repressed E-cadherin expression in response to cyclooxygenase-2 activation or hypoxia-inducible factor-1 mediated signalling, respectively (Dohadwala et al., 2006; Krishnamachary et al., 2006). (2) In high-grade endometrioid adenocarcinomas and other aggressive types of uterine cancers, ZEB1 was strongly expressed in E-cadherin negative carcinoma cells (Spoelstra et al., 2006). (3) In colon tumours, ZEB1 repressed the expression of specific laminin genes and this transient basement membrane loss correlated with increased distant metastasis and poor patient survival (Spaderna et al., 2006). Likewise, an inverse correlation of ZEB1 and E-cadherin levels was observed in colon tumours lacking Snail1 (Pena et al., 2005).
The molecular cues that induce ZEB1 expression in particular cancer settings or in particular tumour areas are still elusive. Candidates for ZEB1 inducers are TGFβ, tumour necrosis factor-α (TNF-α) and Wnts, which are often excessively produced by different tumour-associated stroma cells (Ohira et al., 2003; Chua et al., 2006; Nishimura et al., 2006). One attractive hypothesis is that tumour infiltrating macrophages, which are abundantly detected at the tumour–host interface, express cytokines that may induce ZEB1 expression (Condeelis and Pollard, 2006).
In summary, the data of this study provide strong evidence for a key function of ZEB1 in late-stage cancer progression. Large-scale tumour studies with well-documented clinical records will be critical to determine whether ZEB1 can be used as a prognostic parameter to predict metastatic burden and patient survival.
MDA-MB-231 and CAMA1 were cultivated as reported (Eger et al., 2005). For generation of E-cadherin-expressing cells, MDA-MB-231 cells were transfected with E-cadherin-green fluorescent protein vector and selected in 800 μg/ml Neomycin (Invitrogen, Carlsbad, CA, USA). RNA isolation, quality control, labelling and hybridization on Affymetrix GeneChip (Human Genome U133 Plus 2.0) and data acquisition were done as described in Pacher et al. (2006).
Proximal human Crumbs3 promoter fragments −1402/+74 and −1958/+74 were cloned into NheI/SmaI and SacI/SmaI sites of the pGL3-basic reporter vector (Promega, Madison, WI, USA), respectively, the human Crumbs3 promoter fragment −3421/−1402 into KpnI/NheI sites and the human −743/+135 HUGL2 fragment into MluI/XhoI sites. Transient reporter assays were done as described (Eger et al., 2000).
Preparation of RNA and cDNA, and PCR analyses were done as described (Eger et al., 2005) For PCR primers see Supplementary Table 1. ZEB1 and Snail1 transcript levels were determined by real-time PCR as described in Eger et al. (2005).
The following Abs were used: mouse monoclonal Ab to E-cadherin and ZO1 (BD Biosciences, Franklin Lakes, NJ, USA); rabbit polyclonal Ab to Crumbs3, provided by Ben Margolis (University of Michigan Medical School, Ann Arbor, MI, USA); rabbit polyclonal Ab against HUGL2; goat polyclonal Ab to ZEB1 (ZEB-E20) and PATJ (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA); anti-active-β-catenin Ab (Upstate Biotechnology, Lake Placid, NY, USA); mouse monoclonal Ab to cytokeratin 18 and rabbit polyclonal Ab against actin (Sigma, St Louis, MO, USA); secondary Abs coupled to Alexa Fluor 488 (Molecular Probes Inc., Eugene, OR, USA); Texas Red or peroxidase (Jackson Laboratories, West Grove, PA, USA).
Cells were fixed in 2.5% formaldehyde (Merck Inc., Whitehouse Station, NJ, USA) and processed for immunofluorescence microscopy (Eger et al., 2000). Immunoblotting of total cell lysates is described elsewhere (Eger et al., 2000).
Formalin-fixed, paraffin-embedded colorectal adenocarcinomas or ductal and lobular breast tumours were obtained from surgical resection specimens. Tumour samples were obtained from the archive of the Institute for Pathology, University of Erlangen-Nürnberg and from the Pathology Laboratory Obrist and Brunhuber, Tyrolpath OEG. The age of colon cancer patients ranged from 45 to 65 years and breast cancer patients from 40 to 60 years. Immunohistochemistry was performed according to the instructions of the Vectastain ABC Kit (Vector Laboratories, Burlingame, CA, USA).
Three days after ZEB1- and Snail1-siRNA treatment, 50 000 cells were transferred to 24-well Transwell filter inserts (Corning, NY, USA) (8 μm, Costar). After 24h, cells were fixed in 2.5% formaldehyde and cells in the lower chamber were stained with Hoechst33258 (Invitrogen) and quantified by fluorescence microscopy.
We thank Gabriele Stengl and Peter Steinlein, Institute of Molecular Pathology, Vienna, for Flow Cytometry (FACS) experiments. Furthermore, we thank Jürgen Pollheimer and Martin Knofler for their technical expertise in transpore migration assays and Heidemarie Huber for support in immunohistochemistry. This study was supported by funds from the Hochschuljubiläumsstiftung of the city of Vienna to AE (H-703/2005), by grants from the Austrian Science Research Fund (FWF) No. SFB 006 to RF (603) and HB (612), SFB-F28 to WM and by funds of the Austrian Ministry of Education, Science, and the Arts (Austrian Genome Research Program GEN-AU) to MS, WS, NS, AW. AS is supported by the ÖAD-Pakistan Scholarship programme.