Abberant activation of the process of epithelial-mesenchymal transition in cancer cells is a late event in tumor progression. A key inducer of this transition is the transcription factor Snail, which represses E-cadherin. We report that conditional expression of the human transcriptional repressor Snail in colorectal cancer cells induces an epithelial dedifferentiation program that coincides with a drastic change in cell morphology. Snail target genes control the establishment of several junctional complexes, intermediate filament networks, and the actin cytoskeleton. Modulation of the expression of these genes is associated with loss of cell aggregation and induction of invasion. Chromatin immunoprecipitation experiments showed that repression of selected target genes is associated with increased binding of Snail to their promoters, which contain consensus Snail-binding sites. Thus, Snail constitutes a master switch that directly represses the epithelial phenotype, resulting in malignant carcinoma cells.
- epithelial-mesenchymal transition
- tumor progression
Snail is a transcriptional repressor that plays a central role in epithelial-mesenchymal transition, a process by which epithelial cells lose their polarity and are converted to a mesenchymal phenotype ( 1). Epithelial-mesenchymal transition is important in many developmental processes, such as gastrulation and neural crest migration, but its deregulation in cancer cells can lead to tumor progression. Multiple signaling pathways seem to converge to Snail expression during different normal developmental steps but also during tumor progression. Besides their involvement in epithelial-mesenchymal transition, Snail family members have been implicated in a variety of other processes, such as apoptosis and left-right asymmetry ( 2, 3). Up to now, the mechanism by which Snail influences these different cellular processes remains largely unresolved. Snail can repress E-cadherin through binding to E-boxes in the E-cadherin promoter ( 4, 5). Other candidate repressors for E-cadherin are Slug, E12/E47, δ-EF1 (ZEB1), and SIP1 (ZEB2; reviewed in ref. 6). Furthermore, Snail suppresses the expression of claudins and occludins ( 7), and other epithelial genes such as MUC1 and cytokeratin 18 ( 8). However, until now, few studies have focused on the global effects of Snail transcriptional activity, and information on the early response genes in cells expressing Snail is limited. Here, we show that conditional expression of human Snail (hSnail) in human colon cancer cells promotes an epithelial-mesenchymal transition–like process in which loss of intercellular adhesion coincides with induction of invasiveness. To identify transcriptional changes that are specific responses toward hSnail expression, a comparative differential gene expression analysis using cDNA microarrays was done. This molecular functional analysis revealed that Snail induction leads to a general (de)regulation of epithelial differentiation, metabolism, and signal transduction. Using chromatin immunoprecipitation, we identified several new direct cellular targets of Snail. These molecular data provide a better understanding of the functional consequences of Snail expression during tumor progression.
Materials and Methods
Cell culture and generation of stable cell lines. DLD-1TR21 cells were cultured in RPMI with 10% FCS, 100 units/mL penicillin, and 100 μg/mL streptomycin. Linearized pcDNA4/TO-hSnailMyc/His plasmid was stably transfected by electroporation. Clones were isolated after 2 weeks of selection on 500 μg/mL zeocin and 10 μg/mL blasticidin (Invitrogen, Carlsbad, CA). Expression was induced using doxycycline (1 μg/mL, Sigma, St. Louis, MO).
DLD-1TR21-hSnailMyc/His cells were retrovirally transduced with pFB-Neo-hEcad. Cells were selected on 300 μg/mL neomycin for 2 weeks.
DLD-1TR21 cells were retrovirally transduced with either the pFB-Neo-enhanced green fluorescent protein (EGFP) or the pFB-Neo-EGFP-hSnail vector. After selection for 2 weeks, cells underwent two cycles of sorting using the FACSVantage (Becton Dickinson, San Jose, CA).
All constructs used were obtained with standard cloning techniques. Sequence verification was carried out with sequence-specific primers.
Quantitative real-time PCR. Design of primers and probes, cDNA synthesis, and PCR amplification were described previously ( 9). In addition, we used TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA). Sequences of primers and probes are listed in Supplementary Table S1. The average threshold cycle of triplicate reactions was used for all subsequent calculations using the ΔCt method.
Immunocytochemistry. Immunocytochemistry was done using standard procedures ( 9). In addition, mouse monoclonal antibodies recognizing E-cadherin (HECD-1, Takara, Kyoto, Japan), p120ctn (Transduction, San Jose, CA), plakophilin-2 (Progen, Heidelberg, Germany), and claudin-4 (Zymed, San Francisco, CA) were used.
Collagen invasion and fast aggregation assay. The assays were done as described ( 9).
RNA preparation. Cells were grown to subconfluency, trypsinized, washed, and resuspended in 4 mol/L guanidine thiocyanate buffer. Lysates were homogenized on ice with a syringe and needle. RNA was pelleted by ultracentrifugation through CsCl buffer at 32,000 rpm. The pellet was resuspended in 0.3 mol/L sodium acetate (pH 6.0) and precipitated with 100% ethanol. RNA was further purified by phenol/chloroform extraction and ethanol precipitation before removal of genomic DNA by DNase treatment in the presence of RNase inhibitor (Promega, Madison, WI). Treated samples were then purified again in the same way.
cDNA microarray analysis. Construction of the four human 5K microarrays, probe labeling, hybridization, washing, and scanning were carried out at the MicroArray Facility of the Flanders Interuniversity Institute for Biotechnology (details in ref. 10). A gene was scored as down-regulated if the [normratio 3]Av + 2.33σ[normratio 3]Av < 0.75 and the [normratio 2]Av < 0.57; a gene was scored as up-regulated if the [normratio 3]Av − 2.33σ[normratio 3]Av > 1.25 and the [normratio 2]Av > 1.75. Genes were selected only if they complied with these selection criteria in the three hybridization experiments representing RNA samples harvested at three time points after Snail induction.
Northern blot analysis. Total RNA was isolated using the RNeasy kit (Qiagen, Valencia, CA). Hybridizations were done as described before ( 9). Specific sequences were amplified from MicroArray Facility clones that had been sequenced with universal M13 forward and reverse primers.
Chromatin immunoprecipitation analysis. Chromatin immunoprecipitation analysis was done as previously described ( 11). The Snail antibody was from Santa Cruz Biotechnology (Santa Cruz, CA). For EGFP, we used living colors peptide antibody (Clontech, Mountain View, CA). Fragments were analyzed by quantitative real-time PCR.
Results and Discussion
Human Snail induction in colorectal cancer cells leads to morphologic changes that coincide with loss of aggregation and induction of invasion. To elucidate the functional consequences of Snail expression on epithelial differentiation, we made use of an inducible cell system, the epithelial DLD-1TR21 cell line, which expresses high levels of the tetracycline activator ( 12). These cells were stably transfected with an expression vector harboring a Myc-tagged full-length human Snail under control of a responsive tetracycline operator element. hSnail expression was induced by adding doxycyclin to the medium, and mRNA could readily be observed 12 hours after induction, as shown by real-time quantitative PCR ( Fig. 1A ). This was confirmed by immunofluorescent analysis, which clearly showed nuclear localization of induced hSnail ( Fig. 1B). Changes in morphology from an epithelioid morphotype to a fibroblast-like type could be seen 48 hours after hSnail induction, and became more pronounced after long-term doxycyclin administration ( Fig. 1C). This transition is in agreement with previous reports of mouse Snail expression in the dog Madin-Darby canine kidney (MDCK) cell line ( 4, 5). To our knowledge, an inducible cell line for hSnail has not been reported thus far. Expression of hSnail altered the aggregation behavior of the cells. Whereas noninduced cells showed significant aggregation after 30 minutes, hSnail induction abrogated this normal cell-to-cell aggregation to an extent similar to that caused by an E-cadherin blocking antibody, DECMA-1 ( Fig. 1E). In addition, invasion into collagen type I gels was efficiently induced by hSnail ( Fig. 1D). Altogether, this colon cancer cell system for conditional hSnail expression enables the study of the very early molecular mechanisms by which hSnail executes its function as promoter of cancer invasion.
Functional analysis of induced Snail expression using gene expression profiling. To understand the molecular mechanism by which hSnail expression induces loss of epithelial differentiation, we analyzed differential gene expression upon hSnail induction. To search in the population of differentially expressed genes for those directly under the control of hSnail, microarrays were used to monitor expression profiles 12, 24, and 48 hours after hSnail induction. Of the 17,268 cDNAs examined, hSnail consistently repressed 167 genes and induced 23 others relative to the noninduced state (Supplementary Table S2). These data indicate that hSnail acts mainly as a transcriptional repressor. It may be surprising that a set of genes is up-regulated by a transcriptional repressor, but it had already been shown that Snail induces the promoter activity of ZEB1 ( 8).
Our data mining comprised the annotation of the most characteristic gene ontology term to a cluster of genes. The analysis was done with the Web-based gene ontology tools FatiGO (http://fatigo.bioinfo.cnio.es/) and AmiGO (http://www.godatabase.org). Gene ontology analysis revealed that Snail has a strong impact on genes involved in epithelial differentiation, metabolism, and signal transduction ( Table 1 ).
Snail acts on different epithelial adhesion junctions. The first cluster of down-regulated genes concerns epithelial differentiation, suggesting a profound effect on cytoskeletal organization and different intercellular junctional and adhesive complexes ( Table 1). This cluster was studied in more detail. Surprisingly, E-cadherin was not detected in the microarray analysis as a gene repressed by hSnail. E-cadherin had been described as a major target gene for mSnail ( 4, 5). However, we found only a moderate repression of E-cadherin mRNA levels upon hSnail expression in the DLD-1TR21 system ( Fig. 2A ). In accordance with this, transient transfection of luciferase reporter constructs driven by the E-cadherin core promoter in this inducible cell system showed only a minor repression of E-cadherin promoter activity (data not shown). Furthermore, a drop in E-cadherin protein expression is detectable only after 4 days of hSnail induction ( Fig. 2B). Additional retroviral transduction of E-cadherin in the inducible cell line DLD-1TR21-hSnail generated high E-cadherin expression levels independently of hSnail repression, but was unable to restore the loss of aggregation or prevent induction of invasion ( Fig. 1D and data not shown). Seemingly, high E-cadherin expression by itself is unable to counteract the epithelial dedifferentiation induced by Snail. This is in agreement with the observation that reintroduction of E-cadherin in MDCK-Snail cells does not alter the fibroblastic morphology of the cells ( 13). Claudin-4, an important component of the tight junctions, was dramatically repressed on both the mRNA and protein levels ( Fig. 2A-C). There exists in vitro and in vivo evidence that this protein acts as a tumor suppressor ( 14). Moreover, transforming growth factor-β (TGF-β), a known inducer of members of the Snail family, is able to repress claudin-4. Interestingly, it has been reported that members of the TGF-β family induce Snail family members in several different systems ( 15, 16). It was previously shown that mSnail can repress the promoter activities of the mouse claudin-3, claudin-4, and claudin-7 genes ( 7). Also, plakophilin-2 is repressed after hSnail induction ( Fig. 2C). Plakophilin-2 is a desmosomal component that can directly interact with desmoplakin, plakoglobin, desmoglein-1 and desmoglein-2, and desmocollin-1a and desmocollin-2a ( 17). Eliminating plakophilin-2 by hSnail would then likely lead to a significant perturbation of the desmosomal assembly.
Human Snail regulates genes involved in a program that remodels the cytoskeleton. A remarkably high number of genes down-regulated by hSnail belong to the cytoskeleton group ( Table 1). This group embraces microfilaments, intermediate filaments, and microtubules. The various elements of the cytoskeleton not only serve the maintenance of cell shape but also have roles in other cellular functions, including cell movement, cell division, endocytosis, and movement of organelles. hSnail modulates a large subset of seven different epithelial-specific cytokeratins whose transcriptional repression was confirmed by Northern blot analysis for four cytokeratins ( Fig. 3A ). The primary function of these proteins is to impart resistance to mechanical stress to cells, as cytokeratins are the major structural proteins in epithelial cells, forming a cytoplasmic network of intermediate filaments. Studies in knockout mice showed functional redundancy between keratins. However, in double-knockout mice, a failure of the compensation mechanism can lead to fragility of the cell ( 18). In this respect, the modulation of a whole set of keratins seems to be an excellent strategy for breaking the topological organization of the cell. Intriguingly, a cluster of cytokeratin genes was previously found to be up-regulated after expression of KLF-4, an epithelial enriched, zinc finger–containing transcription factor ( 19). We showed that KLF-4 is down-regulated after hSnail expression ( Fig. 4A ). KLF-4 controls in vivo differentiation of specific epithelial functions, as was shown by knockout studies ( 20). It is likely that some key factors, such as KLF4, are involved in the coordinated up-regulation of a cluster of cytokeratin genes, conferring epithelial characteristics to the cell, whereas other factors, such as Snail, have the opposite effect. Actin filaments represent another important part of the cytoskeletal framework. They are linked to the plasma membrane through various linker proteins and are associated with many actin-modulating proteins. Induction of hSnail leads to a dramatic change in actin cytoskeleton organization, as was shown by staining with phalloidin-FITC ( Fig. 3C). Noninduced cells show a cortical localization of actin, whereas hSnail induction leads to the organization of actin in stress fibers. Gelsolin and its homologue CAPG are capping proteins that are strongly down-regulated upon hSnail induction ( Fig. 3B). Our microarray analysis revealed down-regulation of other actin-modulating genes as well. SLC9A3R1, better known as NHE-RF, is highly expressed in the epithelia of many tissues, particularly in cells with numerous microvilli, and is often concentrated at the luminal membrane. It interacts with merlin (NF2 tumor suppressor), ezrin, radixin, and moesin (MERM proteins), which are involved in cytoskeletal reorganization and signal transduction ( 21). Another putative linker that is repressed is ABLIM-1 ( Fig. 4A), which binds F-actin via its villin headpiece and has been suggested to function in signaling pathways through its LIM domain ( 22). EPLIN would promote the formation of stable actin filament structures at the expense of more dynamic actin filament structures ( 23), and reduction in expression of EPLIN, as is observed here under the influence of hSnail, may contribute to the motility of invasive tumor cells.
Identification of human Snail target genes by chromatin immunoprecipitation analysis. Next, we wanted to evaluate whether hSnail was able to directly repress the genes identified in our microarray analysis. For this purpose, the DLD-1TR21 cell line was retrovirally transduced with an EGFP-hSnail fusion construct. Real-time quantitative reverse transcription-PCR (RT-PCR) of a subset of sequence verified cDNAs confirmed their repression in the cells expressing EGFP-hSnail ( Fig. 4A). These quantitative RT-PCRs together with the Northern experiments validated without exception >10% of the microarray data for corresponding genes with diverging levels of signal ratios. Consequently, the complete list of differentially expressed genes can be considered largely reliable. We analyzed in more detail the promoters of 11 genes identified in the microarray analysis ( Fig. 4B). The depicted promoters have E-boxes close to the transcription initiation site. E-boxes are the characterized binding sites for Snail family members, as was already showed by in vitro experiments using promoter reporter constructs for several genes, such as E-cadherin ( 4, 5), claudins, and occludins ( 7). Using EGFP- and Snail-specific antibodies, we did chromatin immunoprecipitation analysis for a subset of potential target genes which had been fully validated as Snail repressed genes. We were able to show the in vivo binding of hSnail to target genes implicated in epithelial differentiation, signaling, and metabolism ( Fig. 4C). Rab25 is a small GTP-binding protein with an epithelial distribution. It is associated with the apical recycling system, through which polarized epithelial cells maintain the polarized distribution of basolateral and apical membrane proteins ( 24). Binding to the E-cadherin promoter (CDH1) could be detected as well. Because of the low repression capacity of hSnail on this promoter, we suggest that other cell-specific corepressors are needed for complete shutdown of the E-cadherin transcription. Furthermore, binding of hSnail to the promoters of PFKP and SLC27A2 was also shown, suggesting that other pathways including lipid and carbohydrate metabolism could also be involved in the dedifferentiation of cells. The data we obtained from a genome-wide screening for target genes not only give a comprehensive view of the thorough impact of hSnail on epithelial differentiation, they also offer new candidate genes that are putatively involved in invasion and metastasis.
Grant support: Association for International Cancer Research, United Kingdom; Geconcerteerde Onderzoeksacties of Ghent University; Fund for Scientific Research, Flanders, Fortis Insurances (Belgium); and Interuniversity Attraction Poles Programme (Belgian Science Policy).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. P. Van Hummelen (Microarray Facility, VIB, Leuven, Belgium) for the microarray experiments; Dr. J. Gettemans (Department of Medical Protein Research, VIB-Ghent University, Ghent, Belgium) for the gelsolin and CAPG antibodies; and all unit members for helpful discussions, P. De Bleser and D. Vlieghe for the informatics support, D. De Wispelaere for the excellent technical assistance, and E. Parthoens for the confocal images.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
B.D. Craene is a research assistant, and C. Stove and G. Berx are postdoctoral fellows, with the Fund for Scientific Research, Flanders.
- Received October 1, 2004.
- Revision received May 4, 2005.
- Accepted May 18, 2005.
- ©2005 American Association for Cancer Research.