Ras proteins act as molecular switches that cycle between active GTP-bound and inactive GDP-bound forms and function as essential components of signal transduction pathways regulating cell growth. Activating mutations of the Ras family members are among the most common genetic events in human tumorigenesis.1 For example, K-ras is mutated in nearly 50% of colorectal tumors at a relatively early stage of the carcinogenic process2 and despite extensive research, the primary reason for this high frequency remains unclear.
The most studied downstream effector pathway of K-ras is the mitogenic serine/threonine kinase cascade called Raf/MEK/mitogen-activated protein kinase (MAPK). Indeed, upon activation by growth factor-stimulated receptors, activated Ras complexes with and promotes Raf kinases, which in turn activate MAPK kinases (MEK1 and MEK2), resulting in activation of extracellular signal-regulated kinases (ERK1 and ERK2).3 Activated ERKs then translocate into the nucleus where they phosphorylate and activate nuclear transcription factors, such as Elk-1, ATF-2 and ETS1/2 resulting in immediate-early gene induction.4 Studies on cultured intestinal epithelial cells and many other cell types have revealed a close correlation between ERK activation and DNA synthesis, while pharmacological or molecular inhibition of cellular ERK activity has been shown to block cell cycle progression.5–9 Notably, the phosphorylated and activated forms of ERK1/2 have mostly been detected in the nucleus of undifferentiated proliferative crypt cells in human fetal small intestine,10 hence supporting the role of these kinases in the control of cell proliferation in intestinal crypts.
The critical involvement of these Ras downstream effectors in intestinal tumorigenesis is also supported by a number of experimental observations. First, mutations of BRAF, a member of the Raf family, are associated with increased kinase activity and have been found in 9–11% of colorectal cancers; many of those with such mutations are at early Dukes' stage (A and B).11, 12 Furthermore, BRAF mutations are frequently identified in sporadic colorectal cancer with microsatellite instability.13 Second, it has been demonstrated that MEK is phosphorylated and activated in 30–40% of adenomas14, 15 and in 76% of colorectal tumors. Third, colorectal cancers exhibit particularly high frequencies of ERK activation16 and some studies have reported that ERK1/2 activities are elevated in intestinal tumors.17, 18 Fourth, blockade of the MEK/ERK cascade suppresses growth of colon tumors in vivo, suggesting that ERK activation is indeed involved in intestinal tumor proliferation.19 Fifth, we20 and others21 have recently shown that expression of constitutively active MEK1 in nontransformed rat intestinal epithelial crypt cell lines is sufficient to induce growth factor relaxation for DNA synthesis and to promote morphological transformation and growth in soft agar.
The mechanisms by which constitutive activation of MEK1 induces epithelial transformation are however still under investigation. Loss of cell–cell adhesion followed by the dissociation of epithelial structures is a prerequisite for increased cell motility and tumor invasion.22, 23 This phenotypic switch, designated epithelial-to-mesenchymal transition (EMT), also plays a fundamental role in morphogenetic tissue remodeling during embryogenesis.24 However, inappropriate reactivation of the EMT program is commonly observed during transition of benign adenomas to metastatic carcinomas.23 Elucidating the molecular mechanisms that govern EMT is therefore essential in understanding the process of tumor progression. A signaling pathway possibly involved in the control of EMT is the ERK MAPK cascade. Indeed, there is some evidence suggesting a link between ERK pathway signaling and cell adhesion, invasion and metastasis in colorectal cancer.25 For example, interactions between the cell-surface urokinase-type plasminogen activator (uPA) receptor and integrins are crucial for tumor invasion and metastasis, and uPA increases basal ERK activation in colon cancer cells.26 Furthermore, protein kinase CβII can activate colorectal cancer cell invasion by signals transmitted through the Ras/MEK pathway.27 However, despite accumulating evidences showing a positive link between high MEK/ERK MAPK activities and intestinal transformation and cancer, the specific involvement of MEK/ERK signaling in EMT and metastasis of intestinal epithelial cells remains elusive. Hence, our study was aimed at characterizing whether constitutive activation of MEK1 is sufficient to induce EMT and in vivo tumor invasion as well as metastasis and to analyze the potential involvement of some of the downstream molecular mechanisms in these processes.
Material and methods
The antibodies against total ERK, HA epitope, occludin, N-cadherin, GST, Fra-1, β-tubulin, c-fos and c-jun were from Santa Cruz Biotechnologies (Santa Cruz, CA). Antibodies recognizing phosphorylated ERK1/2 and total Egr-1 were obtained from Cell Signaling (Danvers, MA). Antibodies against E-cadherin were from BD Pharmingen (Mississauga, ON, Canada). The ZO-1 antibody was from Zymed Laboratories (Invitrogen, Burlington, ON, Canada). Antibodies recognizing β-actin and vimentin were purchased from Chemicon International (Billerica, MA). Osteopontin antibody was purchased from DSHB (Developmental Studies Hybridoma Bank, Iowa City, IA). MMP-2/MMP-9 inhibitor and U0126 were from Calbiochem-Novabiochem Corp. (San Diego, CA). The recombinant ERK2 protein was purchased from Upstate Biotechnology (Lake Placid, NY). For immunofluorescence, the goat anti-rabbit, goat anti-mouse AlexaFluoro488 FITC-labeled and the rabbit anti-rat IgG-FITC-labeled secondary antibodies were from Molecular Probes (Invitrogen). Other materials were obtained from Sigma-Aldrich unless stated otherwise.
The expression vectors for HA-tagged wild-type MEK1 (wtMEK) and constitutively active MEK1 mutant (caMEK, in which the Raf1-dependent regulatory phosphorylation sites, S218 and S222, were substituted by aspartic residues)28 were kindly provided by Dr. J. Pouysségur (Université de Nice, France). wtMEK and caMEK were subcloned into the retroviral expression vector pLXIN (Clontech) to produce viruses in HEK293T cells in cotransfection with helper amphotropic DNA vector as previously described.20
The rat intestinal epithelial crypt cell line IEC-6 and IEC-6 cells stably overexpressing pLXIN empty vector (EV), wtMEK or caMEK were cultured as previously described.6, 20 These cell populations were generated after viral infection of wtMEK and caMEK cloned in the retroviral vector pLXIN. The percentage of retrovirally transduced cells ranged between 70 and 80%, as estimated by parallel infections using viruses expressing the Green Fluorescent Protein gene product.20 The pLXIN retroviral vector coexpressed a G418 resistance gene that allowed selection of pure populations of transduced cells within 10 days. The phenotype of these 3 cell lines was previously described.20 More specifically, the caMEK-expressing cells formed foci at postconfluence, in contrast to pLXIN- and wtMEK-expressing epithelioid cells, which formed a monolayer of contact-inhibited cells.20 Foci from postconfluent caMEK-expressing cells were picked by aspiration with a pipet and pooled as 1 caMEK-expressing cell population. All the experiments were performed with this caMEK-expressing cell population in comparison to pLXIN- and wtMEK-expressing cell populations. This strategy was repeated independently 3 times with other IEC-6 cell cultures, and similar results were obtained with all caMEK-expressing cell populations. HT-1080 fibrosarcoma cells, obtained from ATCC (Manassas, VA) and used as positive control in invasion assays, were cultured in DMEM containing 10% FCS. The colon carcinoma cell line HCT116 was obtained from ATCC (CCL-247) and cultured in McCoy's medium containing 10% FCS. The colon adenocarcinoma cell line Lovo (CCL 229, ATCC) was cultured in Ham's F12 medium containing 10% FCS, and the SW480 (CCL 228, ATCC) was cultured in DMEM containing 10% FBS. The colon adenocarcinoma cell line DLD-1 was obtained from Dr. F. Boudreau (University of Sherbrooke, QC, Canada) and cultured in RPMI medium containing 10% FCS.
Protein expression and immunoblotting
Cells were lysed in SDS sample solution (62.5 mM Tris-HCl pH 6.8, 2.5% SDS, 10% glycerol, 5% β-mercaptoethanol, 0.02% bromophenol blue). Protein concentrations were measured using a modified Lowry procedure with bovine serum albumin as standard.29 Western blot analyses were performed as previously described.20
MEK kinase assay
Cells were washed twice with ice-cold PBS, lysed in chilled lysis buffer (100 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, pH 7.5, 1% Triton X-100, 1 mM PMSF, 0.2 mM orthovanadate and 40 mM β-glycerophosphate) and lysates were cleared of cellular debris by centrifugation. HA antibodies were added to 200 μg of each cell lysate and incubated overnight at 4°C under agitation. Four milligrams of protein A-Sepharose (GE Healthcare Life Science) was subsequently added for 1 hr (4°C under agitation). Immunocomplexes were harvested by centrifugation and washed 3 times with ice-cold lysis buffer. Proteins were solubilized in Laemmli's buffer and separated by SDS-PAGE. In other experiments, the beads were washed 3 times with lysis buffer followed by 2 rinses in ice-cold kinase buffer (40 mM Hepes pH 7.4, 20 mM MgCl2, 1 mM dithiothreitol) before performing the MEK kinase assay with 1 μg of recombinant ERK2 as substrate. Phosphorylation of ERK2 was revealed by Western blot with antibodies recognizing phosphorylated ERK1/2.
Immunofluorescence microscopy of cultured cells
Immunofluorescence experiments on cells grown on sterile glass coverslips were performed exactly as previously described.20 Negative controls (no primary antibody) were included in all experiments.
RNA extraction and quantitative RT-PCR analysis
Total RNA was isolated using the totally RNA extraction kit (Ambion). RT-PCR analysis was performed using AMV-RT (Roche Diagnostics) according to the manufacturer's instructions. Snail1 and Snail2 expression: qPCR was performed using a LightCycler apparatus (Roche Diagnostics). Experiments were run and analyzed with the LightCycler software 4.0 according to the manufacturer's recommendations. Synthesis of double-stranded DNA during the PCR cycles was monitored with SYBR Green I (QuantiTect SYBR Green PCR Kit; Qiagen). All samples were run in triplicate. Target expression was quantified relatively to TATA-binding protein (TBP) expression. A standard calibration curve was prepared for each gene using serial dilutions of the calibrator sample, and crossing point values were plotted vs. the log of the relative concentration of each dilution. This standard curve was used to correct for differences in PCR efficiencies. Primers: rat Snail1 forward: ATGAGGACAGTGGCAAAAGC; rSnail1 backward: TCGGATGTGCATCTTCAGAG. Primers: rSnail2 forward: GCACTGTGATGCCCAGTCTA; rSnail2 backward: CAGT GAGGGCAAGAGAAAGG. Primers: human Snail1 forward: CATCCTTCTCACTGCCATG; hSnail1 backward: GTCTTCAT CAAAGTCCTGTGG; hSnail2 forward: ATGAGGAATCTGGCT GCTGT; hSnail2 backward: CAGGAGAAAATGCCTTTGGA.
Generation of shRNAs against Egr-1 and Fra-1
The lentiviral shRNA expression vector (pLenti6-U6) was constructed by cloning the U6 promoter from pSilencer 2.0-U6 (Ambion) into pLenti6/V5-D-TOPO (Invitrogen). Briefly, the U6 promoter was amplified by PCR from pSilencer 2.0-U6 using the forward primer: CCA TCG ATC ATG ATT ACG AAT TGC AAC G (insertion of a ClaI site) and the reverse primer: GGC CAG TGC CAA GCT TG. The PCR product was digested by BamHI and ClaI and cloned into recirculated pLenti6/V5-D-TOPO between BamHI and ClaI sites that replaced the cytomegalovirus promoter. shRNA oligonucleotides were designed according to Ambion guidelines (technical bulletin no. 506) using the siRNA sequences GCA TCT GTT CCC CCT TGA TAT (shEgr-1) or GAA GTT CCA CCT TGT GGC AAG (shFra-1) and TTCAAGAGA as the loop sequence. The oligonucleotide-annealed product was subcloned into the pLenti6-U6 between BamHI and XhoI sites, giving rise to pLenti6-shEgr-1 and pLenti6-shFra-1. The control shRNA targeted the GFP sequence was as following: GCC ACA ACG TCT ATA TCA TGG. Lentiviruses were produced and used for cell infection according to Invitrogen recommendations (ViraPower Lentiviral Expression System, instructions manual).
Aliquots of the diluted cDNA preparations were used as templates for PCR reactions with the primers. Reactions were performed using 1 unit of Taq DNA polymerase (Qiagen). Parameters for DNA amplification were 94°C for 45 sec, annealing temperature (59°C) for 30 sec and 72°C for 45 sec. Oligonucleotide primers used for DNA amplification were synthesized by Invitrogen. DNA amplification products were analyzed by gel electrophoresis on a 2% agarose gel stained with ethidium bromide. Primer sequences are available upon request.
The invasion assay was performed using 6-well BD Biocoat Matrigel invasion chambers with 8-μm polycarbonated filters (Becton Dickinson, Bedford, MA). A total of 2.5 × 105 HT-1080 (positive control), wtMEK- and caMEK-expressing cells were seeded on 6-well Matrigel invasion chamber plates in the presence of 20 mM hydroxyurea, a pharmacological inhibitor of cellular ribonucleoside reductase to arrest the cell cycle in G1/S phase30 and as we have previously done31; the cells were cultured in routine medium in the presence or absence of specific inhibitors. Nonmigratory cells on the upper surface of the filter were removed by wiping with a cotton swab. Invasive cells that penetrated through pores and migrated to the underside of the membrane were stained with crystal violet (1%) solution after fixation with methanol.
MMP activation and production by zymography
Zymography was done according to the procedure described in Ref.32. Cells were plated in 6-well plates in complete medium. Cells were serum-deprived for 24 hr, after which medium samples were harvested and centrifuged at 2,000g for 10 min. After normalization for cell protein content from cell lysates, aliquots from medium samples were electrophoresed in 10% sodium dodecyl sulfate (SDS)-polyacrylamide gels containing gelatin (0.1%) or β-casein (0.1%). After electrophoresis, the gels were washed twice for 15 min at room temperature in a buffer containing 2.5% Triton X-100, 50 mM Tris and 10 mM CaCl2 pH 7.4 and incubated for 18 hr in 50 mM Tris, 50 mM NaCl, 10 mM CaCl2, pH 7.6, at 37°C. Gels were then stained with Coomassie brilliant blue. Clear zones against the blue background indicated the presence of gelatinase or caseinase activity.
Chromatin Immunoprecipitation assays
For chromatin immunoprecipitation (ChIP) assays, 107 cells were crosslinked with 1% formaldehyde at room temperature for 15 min, and the reaction stopped by the addition of glycine to a final concentration of 0.125 mM. After washing twice with ice-cold PBS, cells were scraped off the plate, centrifuged and resuspended in 1 ml of nucleus/chromatin preparation buffer [Triton X-100 0.25%, Hepes 10 mM (pH 6.5), EDTA 10 mM, EGTA 0.5 mM, 0.5 μg/ml aprotinin, pepstatin (1 μg/ml) and 1 mM phenylmethylsulfonyl fluoride (PMSF)]. After incubation on ice for 10 min, cells were dounced to isolate nuclei. The nuclei were then collected, resuspended in 500 μl of nuclei lysis buffer [50 mM Tris-HCl (pH 8.1), 10 mM EDTA, 1% SDS, 0.5% NP40, 0.5 μg/ml aprotinin, pepstatin (1 μg/ml) and 1 mM PMSF] and sonicated (5 pulses of 10 sec) (sonicator: Branson Sonifier 250). After centrifugation, the supernatant was diluted 1:10 with immunoprecipitation dilution buffer [0.1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.1), 150 mM NaCl, 0.5 μg/ml aprotinin, pepstatin (1ug/ml) and 1 mM PMSF]. The resulting cell lysates were precleared with protein A-agarose beads and subjected to immunoprecipitation with 2 μg of anti-Egr-1 antibody, anti-Fra-1 antibody or anti-GST at 4°C overnight. Immunoprecipitated complexes were collected by adding salmon sperm DNA/protein A-agarose for 1 hr at 4°C. Immunoprecipitates were initially washed twice with low-salt buffer [2 mM EDTA, 20 mM Tris-HCl (pH 8.1), SDS 0.1%, Triton X-100 1% and 150 mM NaCl], then twice with high-salt buffer [2 mM EDTA, 20 mM Tris-HCl (pH 8.1), SDS 0.1%, Triton X-100 1% and 500 mM NaCl] and twice again with LiCl buffer [1 mM EDTA, 10 mM Tris-HCl (pH 8.1), NP-40 1%, LiCl 250 mM and Deoxycholate 1%]; the DNA/protein complex was eluted twice using elution buffer (0.1 M NaHCO3, 1% SDS) and treated with RNase A for 30 min at 37°C. The protein/DNA crosslinks were reversed by heating at 65°C for 16 hr, and the DNA subsequently extracted using the QIAquick PCR purification kit (Qiagen). For PCR, 3 μl from a 50-μl DNA preparation was used for amplifications. Specific sequences of the rat Snail1 promoter in the immunoprecipitates were detected by PCR with primers forward 5′-CCTTGGGCTCCATTTACCTT-3′ and reverse 5′-TGCCTGACTCCTGACCTGTT-3′. The resulting PCR product spans the −1380 to −1200 region that includes the putative Egr-1-binding sites. Specific sequences of the rat Snail2 promoter in the immunoprecipitates were detected by PCR with primers forward 5′-CACCGTAGCTTGGGAAGAAA-3′ and reverse 5′-GGGGATGAGAAAATGGGAGT-3′. The resulting PCR product spans the region −450 to −650 that includes the 2 putative AP-1-binding sites.
Tumor growth in nude mice
A total of 2 × 106 cells suspended in 0.1 ml DMEM were injected into the dorsal subcutaneous tissue of 5-week-old female nude mice CD1 nu/nu (Charles River, Canada). Tumor volume was determined by external measurement according to published methods (d2 × D)/2.33
The tail vein of CD1 nu/nu mice (Charles River, Canada) was injected with 106 cells suspended in 100 μl. Animals were sacrificed at any sign of respiratory distress or weight loss, or after 21 days and examined for the presence of lung metastases. Lungs were stained with Bouin's fluid.
Assays were performed in either duplicate or triplicate. Typical results shown are representative of 3 independent experiments. Densitometric analysis was performed using Image J software. Results were analyzed by the Student's t-test and were considered statistically significant at p < 0.001.
Constitutive active MEK1 induces loss of epithelial morphology and inhibition of epithelial junctional proteins
Cancer cells of epithelial origin undergo EMT as an early event that leads to local invasion and metastasis to distant sites.22 In a previous report, we had shown that expression of a constitutive active mutant of MEK1 (caMEK) in the intestinal epithelial cell line IEC-6 induced morphological transformation and growth in soft agar.20 In marked contrast, wtMEK expression had no effect on IEC-6 phenotype when compared to EV expression.20 In addition, at higher cell densities, the caMEK-expressing cells formed foci in contrast to EV- and wtMEK-expressing epithelioid cells.20 The multilayered morphology in caMEK-expressing cells became apparent at cell densities in which EV- and wtMEK-expressing cells were forming a monolayer of contact-inhibited cells.20 The essential features of EMT are the disruption of intercellular contacts and the enhancement of cell motility, thereby leading to the release of cells from the parent epithelial tissue.22–24 Therefore, to determine whether expression of caMEK is sufficient to induce EMT, foci from postconfluent caMEK-expressing cells were picked by aspiration and pooled in 1 caMEK-expressing cell population. The kinase activity of exogenous MEK in this cell population in comparison to the kinase activity of exogenous MEK present in control cell populations (EV and wtMEK-expressing cells) was first analyzed. Exogenous MEK activities were measured after immunoprecipitation with antibodies against HA tag and using recombinant ERK2 protein as substrate. Figure 1a (left panel) shows that caMEK-expressing IEC-6 cells exhibited strong HA-MEK activity when compared to wtMEK-expressing IEC-6 cells after normalization of expression levels. The molecular manifestation of wild-type and activated MEK expression on endogenous ERK expression and activity was also analyzed. As shown in Figure 1a (right panel), caMEK-expressing IEC-6 cells consistently showed a very slight increase in ERK1/2 phosphorylation levels in comparison to EV- and wtMEK-expressing cells. Expression of ERK proteins was similar in all cell populations. However, an additional species with low electrophoretic mobility was detected with the antibody recognizing the biphosphorylated and activated forms of ERK1/2 in caMEK-expressing cell population (Fig. 1a, right panel, see arrowhead). Interestingly, treatment with the MEK inhibitor, U0126, completely abrogated MEK activity as well as ERK phosphorylation.
We next verified whether activation of MEK1 was sufficient to induce EMT phenotype. As visualized in Figure 1b, phase-contrast microscopy revealed that caMEK-expressing cells underwent a morphological change from an epithelial morphology to an elongated morphology. Indeed, cells clearly exhibited loose cell-to-cell contacts and a fibroblast-like appearance when compared to wtMEK-expressing control cells (Fig. 1b). After caMEK expression, the epithelial markers E-cadherin and ZO-1 disappeared from cell–cell junctions (Fig. 1c, panels 3 and 6). F-actin pattern was rearranged from the typical epithelial circumferential pattern to a pattern typical of fibroblastic/mesenchymal cells (Fig. 1c, panel 9). E-cadherin, ZO-1, occludin and Vimentin expression in these cells was also examined by Western blotting, and in agreement with their phenotypic change, E-cadherin, ZO-1 and occludin expression was clearly decreased (Fig. 1d, left panel). Interestingly, expression of Vimentin, a mesenchymal marker, was markedly upregulated in caMEK-expressing cells compared to wtMEK-expressing cells (Fig. 1d, right panel). In addition, caMEK-expressing cells exhibited N-cadherin expression, whereas the wtMEK-expressing cells clearly did not express this molecule (Fig. 1d, right panel). Upregulation of Vimentin and N-cadherin is a characteristic phenomenon during EMT.22 Finally, treatment with U0126 during 72 hr efficiently rescued the epithelial phenotype of caMEK-expressing IEC-6 cells (Fig. 1b, panel 4), markedly attenuated Vimentin and N-cadherin expression (Fig. 1d, right panel) and partially restored E-cadherin and ZO-1 expression, while occludin protein levels were returned to control levels (Fig. 1d, left panel). Treatment of our cell populations with U0126 had no effect on cell survival as verified by Trypan blue exclusion assay (data not shown). Indeed, as shown in Figure 1b (panel 4), cells appear healthy and neither nuclei fragmentation nor detached cells were observed after the use of U0126. These data indicate that induction of EMT by caMEK was a direct consequence of enhanced MEK activity in IEC-6 cells.
Constitutive active MEK1 downregulates E-cadherin expression and enhances Snail1 and Snail2 expression through Egr-1 and Fra-1
In recent years, several direct transcriptional repressors of E-cadherin (Snail1, Snail2, deltaEF1, SIP1 and E47) have been identified.34–39 These proteins act downstream in EMT-inducing signal transduction pathways activated by transforming growth factor-β, fibroblast growth factor (FGF), epidermal growth factor (EGF), integrin engagement and hypoxia.40–44 Snail family members directly interact with the E-box response elements in the proximal E-cadherin gene promoter and could actively repress transcription by recruiting transcriptional corepressors such as mSin3A.43 In addition, Snail1/Snail2 and deltaEF1/SIP1 proteins mediate upregulation of genes implicated in cell invasion and motility (e.g., Vimentin, members of the MMP family of proteases, fibronectin). Herein, RT-PCR analysis confirmed that E-cadherin gene expression was clearly downregulated in caMEK-expressing cells in comparison to wtMEK-expressing cells (Fig. 2a). Therefore, we next determined whether caMEK directly affected expression of Snail1, Snail2, SIP1, E47 and Twist by quantitative RT-PCR analysis. As shown in Figure 2b, Snail1 and Snail2 mRNA levels were significantly upregulated by 4.4- and 5-fold (p < 0.001), respectively, in caMEK-expressing cells when compared to cells expressing wtMEK. However, Twist, SIP1 and E47 expressions were not significantly altered in caMEK-expressing cells in comparison to wtMEK-expressing cells (data not shown), indicating that Snail1 and Snail2 are the most prominent E-cadherin transcriptional repressors induced by caMEK in intestinal epithelial cells.
We further analyzed the molecular mechanisms by which caMEK signaling induced Snail1 and Snail2 expression and investigated the potential involvement of candidate transcription factors. Recently, hepatocyte growth factor (HGF) has been shown to induce scattering of lens epithelial cells through ERK/Egr-1-mediated upregulation of Snail1.45 Egr-1 is a zinc-finger transcription factor that binds DNA sequences containing the consensus binding site GCG(G/T)GGGCG.46 Herein, a survey of the Snail1 rat promoter sequence using Transcription Element Search System (TESS) software identified 1 strong putative Egr-1 binding site between nucleotides −1349 and −1341 upstream of the Snail1 gene transcriptional start site. We thus verified whether Egr-1 was induced by caMEK in intestinal epithelial cells. Western blot analysis shown in Figure 3a demonstrates that Egr-1 protein was markedly induced by 3-fold in caMEK-expressing cells when compared to control cells. Of note, treatment of cells with the MEK inhibitor, U0126, completely abolished Egr-1 expression. To determine whether Egr-1 physically associates with the endogenous promoter of Snail1, ChIP assays were performed with anti-Egr-1 antibodies in wtMEK- and caMEK-expressing cells. As shown in Figure 3b, increased in vivo binding of Egr-1 to Snail1 promoter was found in caMEK-expressing cells compared to wtMEK-expressing cells. The survey of the Snail1 promoter sequence also identified 4 putative AP-1 binding sites upstream of the transcriptional start site. Because AP-1 transcription factor complexes are downstream effectors of ERK MAPKs,4 expression of the main proteins forming AP-1 complexes in colon cancer cells,47 namely c-jun, c-fos and Fra-1, was analyzed. Western blot analysis shown in Figure 3a illustrates that Fra-1 protein was strongly induced by 15-fold in caMEK-expressing cells comparatively to control cells, whereas no detectable change was observed in c-fos and c-jun expression. It is noteworthy that treatment of cells with the MEK inhibitor, U0126, completely abolished Fra-1 expression. However, ChIP assays did not reveal any difference in the binding of Fra-1 to the Snail1 promoter in cells expressing the activated MEK1 mutant (data not shown). The analysis of the Snail2 promoter sequence identified 2 strong putative AP-1 binding sites between nucleotides −334 and −327 and −264 and −257 upstream of the Snail2 gene transcriptional start site. ChIP assays revealed increased in vivo binding of Fra-1 to Snail2 promoter in caMEK-expressing cells compared to wtMEK-expressing cells (Fig. 3c). Hence, our findings suggest that activated MEK1 induces the expression of Snail1 and Snail2 through Egr-1 and AP-1 binding sites, respectively.
To further investigate the importance of Egr-1 and Fra-1 in caMEK-induced Snail1 and Snail2, recombinant lentiviruses encoding anti-Egr-1 and anti-Fra-1 short hairpin RNAs (shRNAs) were developed to stably suppress Egr-1 and Fra-1 mRNA levels. Several lentiviral constructs were generated and tested for their ability to knock down these transcription factors. Two of these viral shRNAs were selected and designated as shEgr-1 and shFra-1. caMEK-expressing cells were henceforth infected with shEgr-1 or shFra-1 lentiviruses. The pLentiV5-U6 lentiviral vector, which coexpresses a blasticidin S resistance gene, allowed the selection of pure populations of transduced cells within 6 days. In addition, a shRNA against GFP RNA sequence was also generated leading to the production of a caMEK cell population stably expressing this shRNA as a control. Cells infected with shEgr-1 exhibited strongly reduced Egr-1 protein synthesis, in contrast to noninfected cells (NI) and to cells infected with shRNA targeting the GFP sequence (Fig. 3d). Q-PCR analysis shown in Figure 3e demonstrates that the downregulation of Egr-1 expression in caMEK-expressing cells significantly decreased Snail1 expression, hence confirming the contribution of Egr-1 in MEK1-induced Snail1. Cells infected with shFra-1 exhibited strongly reduced Fra-1 protein synthesis, in contrast to NI and to cells infected with shRNA targeting the GFP sequence (Fig. 3d). Interestingly, downregulation of Fra-1 expression in caMEK-expressing cells significantly decreased Snail2 expression (Fig. 3f), hence confirming the contribution of Fra-1 in MEK1-induced Snail2.
Constitutive active MEK1 induces invasion capacity through MMP2 and MMP9 protease expression and activities
Because the suppression of E-cadherin is related to cancer cell migration during EMT,22 the net effect of caMEK expression was determined on invasion using BD Biocoat Matrigel invasion chambers, in the presence of 20 mM hydroxyurea, a pharmacological inhibitor of the cellular ribonucleoside reductase, to arrest cell cycle in G1/S phase,30 as we previously done.31 As shown in Figure 4a, caMEK-expressing cells clearly acquired an invasive capacity when compared to wtMEK-expressing cells, which never invaded Matrigel; this effect was completely abolished by treatment of caMEK-expressing cells with U0126. Because the migration of cells through extracellular matrices requires the activity of MMPs,23 β-casein and gelatin zymographies were therefore used to determine whether caMEK-expressing cells secrete β-casein- and gelatin-degrading MMPs. Gelatin works well for MMP-2 and MMP-9, whereas MMP-1, MMP-3, MMP-7, MMP-8 and MMP-10 are better identified in casein-containing gels.48 Using casein zymography, no significant difference was detected between conditioned media of caMEK-expressing cells compared to conditioned media of wtMEK-expressing cells (data not shown). By contrast, in gelatin zymography, the Mr 72,000 MMP-2 proenzyme (progelatinase A) was revealed in conditioned media of caMEK-expressing cells, in addition to the Mr 64,000 and Mr 62,000 activated MMP-2 species (Fig. 4b). MMP-9 (gelatinase B) activity was also detected in conditioned media of caMEK-expressing cells (Fig. 4b). Addition of the specific inhibitor of MMP-2/9 to the medium of caMEK-expressing cells resulted in the complete disappearance of the molecular species revealed in gelatin zymography (data not shown) and significantly reduced invasion capacity of caMEK-expressing cells (Fig. 4c).
To exert its enzymatic activity, MMP-2 requires cleavage and activation by MT1-MMP or MMP14. As shown in Figure 4d, MT1-MMP as well as MMP-2 and MMP-9 expressions were barely detectable in wtMEK-expressing cells, whereas high levels of expression of each MMP were observed in caMEK-expressing cells. Conversely, their expressions by caMEK-expressing cells were significantly downregulated after a 24-hr incubation with the MEK inhibitor U0126. Taken together, these results indicate that MMP-2 and MMP-9 secreted by caMEK-expressing cells are the most important MMPs involved in invasion through Matrigel.
Intestinal cells expressing activated MEK1 induce tumor formation and metastasis in nude mice
The tumorigenicity of IEC-6 cell lines was next assessed after their s.c. injection into the flank of nude mice. Cells expressing wtMEK failed to develop tumors 90 days after inoculation. In contrast, IEC-6 transformed with caMEK induced palpable tumors with a short latency of 7 days after their injection (Fig. 5a). We next investigated whether caMEK could induce tumor metastasis in vivo in an experimental metastasis assay. Nude mice injected into the tail vein with cells expressing the caMEK showed extensive lung metastasis within 21 days, whereas cells expressing wtMEK failed to show any significant lung colonization (Fig. 5b). To further address this question, caMEK/IEC-6 cells expressing GFP protein were used.49 Cells were injected s.c. into nude mice and segments of lungs were excised and processed for fluorescence analysis 29 days after implantation of the tumor cells. As shown in Figure 5c, lungs from mice bearing caMEK tumors exhibited GFP fluorescence. Accordingly, RT-PCR analysis revealed the presence of HA-tagged MEK1 in lungs (Fig. 5d). Finally, the expression of factors reported to be involved in colon cancer metastasis50–54 was also analyzed in cultured cells and in tumors resulting from the subcutaneous injection of caMEK-expressing cells. As shown in Figures 5e and 5f, expression of cyclooxygenase-2 (COX-2), osteopontin (OPN) and α6 integrin subunit was upregulated in caMEK-expressing cells in culture and in tumors, whereas uPA expression was well detected in tumors. Taken together, these data show that activated MEK1 is sufficient to induce metastatic spread of intestinal epithelial cells probably through the increased expression of factors known to promote invasion and metastasis.
MEK/ERK signaling pathway controls Snail1 and Snail2 gene expression in human colorectal cancer cells
To evaluate the contribution of endogenous MEK/ERK activities in EMT in human cell models, we analyzed the impact of U0126 treatment on Snail1 and Snail2 expression levels. As shown in Figure 6a, a 48-hr treatment of 4 colorectal cancer cell lines with U0126 efficiently blocked endogenous MEK activity as visualized by the marked inhibition of ERK1/2 phosphorylation. Interestingly, treatment of these cell lines with U0126 also resulted in the strong downregulation of Egr-1 and Fra-1 protein levels. We have also analyzed expression levels of Snail1 and Snail2 in these cell lines by Q-PCR and found that inhibition of MEK/ERK activity significantly decreased Snail1 and Snail2 expression (Fig. 6b), indicating that expression of these transcription factors is likely dependent on ERK activity in these cell lines.
EMT is a type of cell plasticity during which epithelial cells lose many of their epithelial characteristics and acquire properties that are typical of mesenchymal cells. EMT promotes tumor invasion and metastasis, and compelling evidence shows that it plays an important role in determining the dissemination of tumors.55 Our findings demonstrate that expression of activated MEK1 (not wtMEK1) is sufficient for EMT of intestinal epithelial crypt cells. The expression of caMEK resulted in reproducible loss of epithelial characteristics and a gain in mesenchymal characteristics associated with the ability to invade Matrigel (Figs. 1, 2 and 4). More importantly, cells expressing activated MEK1 gained the ability to grow as tumors and had metastatic properties when injected in vivo in nude mice (Fig. 5).
Various mechanisms leading to specific gene repression and activation, transduction signaling pathways and a multitude of mediator molecules seem to cooperate in controlling EMT.56 The key molecular change occurring in EMT is E-cadherin downregulation that represents the determinant step in reducing cell-to-cell adhesion, thereby leading to destabilization of the epithelial architecture.56, 57 A number of transcription factors negatively affecting E-cadherin expression are now known. The zinc-finger factor Snail1 is a strong repressor of E-cadherin transcription and a well-known inducer of EMT. Snail1 has been found overexpressed in the tumor cells located at the invasive front, where it is responsible for the disruption of E-cadherin-mediated cell–cell contacts and invasion.57 Other E-cadherin repressors that have been implicated in EMT are Snail2, E47, Twist and the 2-handed zinc factors deltaEF1 and SIPl.34–44, 57 In our experimental approach, we first investigated the regulatory mechanisms by which activated MEK1 stimulates the induction of Snail1 and Snail2 expression, because no modulation of Twist, SIP1 and E47 expression was observed. In addition, the importance of Snail1 in HGF-mediated cell scattering through ERK activation was previously demonstrated.45 The survey of the Snail1 rat promoter sequence identified only 1 strong putative Egr-1 binding site, and it is located between nucleotides −1349 and −1341 upstream of the ratSnail1 gene transcriptional start site. We demonstrated that this is the putative regulatory element for Egr-1 binding in the rat Snail1 promoter in caMEK-expressing cells. This binding site that we have identified herein seems identical in sequence to the one previously identified by Grotegut et al. at position −417 in the Snail1 human promoter. Previous studies have shown that increased activation of the ERK pathway enhances expression of Egr-1 and the binding of Egr-1 to a target sequence.45 Our data demonstrate that activated MEK1 increased the expression and in vivo binding of Egr-1 with the endogenous Snail1 promoter. A similar correlation has also been recently reported in human hepatocellular carcinomas, where Egr-1 expression, Snail1 expression, E-cadherin downregulation and Erk activation are all associated with cell scattering.45 In human colonic epithelial cells, it was recently reported that neurotensin stimulates expression of Egr-1 and the EGF receptor through MAPK activation.59 In addition, curcumin, the active ingredient of the rhizome of the turmeric plant (Curcuma longa), inhibits human colon cancer cell growth by suppressing gene expression of the EGF receptor through reduction of Egr-1 activity.60 Thus, the MEK/ERK/Egr-1 axis may be a valuable target in colorectal tumors.
The AP-1 family of transcription factors is known to be secondary transcriptional targets of ERK signaling.61 In fibroblasts, transformation by Ras induces constitutive expression of c-Jun and Fra-1, which by themselves can promote a transformed phenotype.62 The signaling pathway from Ras to AP-1 appears to be through the ERK MAPK cascade, as signaling through MEK1 modifies AP-1 activity and composition.63 Significantly, Fra-1 rather than c-Fos is the predominant protein that contributes to AP-1 activity in Ras- and MEK1-transformed fibroblasts62, 63 as well as in colon cancer cells.47 The strong expression of Fra-1 in intestinal cells expressing activated MEK1 suggests that it may also contribute to some other aspects of the phenotype induced by activated MEK1. Recently, it has been demonstrated that in colorectal carcinoma cells with high levels of ERK signaling, Fra-1 acts to provide survival signals.64 Moreover, K-ras uses a MEK-ERK-Fra-1 pathway to deregulate Rho signaling, thus maintaining a disrupted actin cytoskeleton in colon cancer cells.65 In addition, Fra-1 has also been shown to be very prominent in the induction of Vimentin expression during Ha-RAS-induced EMT, in association with colon cell migration.47 Vimentin is not just a simple EMT marker; there is also evidence that Vimentin is increased in tumors from patients with colorectal cancer.66, 67 Herein, we showed that Fra-1 expression was also important in the induction of Snail2 by activated MEK1. To our knowledge, this is the first report describing the binding of this AP-1 protein to the Snail2 promoter. We could not detect, however, any difference in the binding of Fra-1 to the corresponding rat proximal AP-1 binding site described previously in mouse Snail1 promoter.68 This does not mean that this site was not important, because other AP-1 proteins might be involved. Finally, we cannot totally exclude the potential implication of other transcription factors such as NF-κB and Sp1, which have been previously reported to induce Snail1 human promoter in response to oncogenic stimulation.69
Our results obtained in human colorectal cancer cell lines in which MEK activity was blocked by U0126 highlight the close correlation between ERK activation and expression of Egr-1, Fra-1, Snail1 and Snail2 (Fig. 6). Current experiments are in progress in the laboratory to determine whether the inhibition of MEK restores some aspects of the epithelial phenotype in colorectal cancer cell lines. A modest increase in E-cadherin mRNA expression was detected in some cell lines after 48 hr of treatment with U0126 (preliminary data), suggesting that prolonged MEK inhibition might be necessary. Alternatively, other molecular signaling mechanisms, known to mediate EMT, could also be implicated. Such signaling mechanisms include PI3K/Akt,70, 71 TGFβ/JNK/Smad3,72 Src73 and NFκB.74 In this respect, some of these signaling proteins are mutated and/or hyperactivated in colorectal cancer and therefore should contribute to the maintenance of EMT-like phenotype in cancer cells. Nevertheless, our data argue that a key role of sustained MEK activity resulting from the constitutive activation of KRAS or BRAF in colorectal carcinoma cells may be to provide signals inducing not only proliferation19, 21, 25, 26, 75 but also EMT.
Other studies have also demonstrated that expression of activated MEK1 induces EMT in other cell systems, particularly in MDCK cells.76–78 For example, expression of caMEK in MDCK results in the destabilization of the 3D architecture and the conversion of polarized epithelial cells into migrating mesenchymal-like cells.77 Furthermore, expression of activated MEK1 has been previously shown to also induce such morphological changes associated with EMT in intestinal epithelial cells,21, 79 although the authors have not provided any experimental evidences and/or mechanistic insights to prove that this was the case. Our study provides experimental proof that constitutive activation of MEK1 in intestinal epithelial cells is sufficient to cause EMT very likely through the induction of Egr-1 and Fra-1 transcription factors, which bind to Snail1 and Snail2 promoters, enhancing their expression. The central role of Snail1 and Snail2 in inducing EMT has been described extensively.40
In addition to cellular transformation and anchorage-independent growth, metastatic tumor cells need to gain numerous biological properties including cell motility, cell survival and secretion of extracellular matrix-degrading enzymes to escape the primary tumor mass, and subsequently invade and colonize distant organs.80 Previous studies have suggested a role of MAPKs in invasion and metastasis of tumor cells.81, 82 Indeed, ERK activity was shown to be higher in metastatic cells when compared to nonmetastatic human breast cancer cells.83–85 Herein, we demonstrate by several evidences (shown in Figs. 4 and 5) that activated MEK1 plays a major role in intestinal tumor progression and metastasis. First, constitutive activation of MEK1 in normal intestinal epithelial cells induced migration through Matrigel as well as induced MT1-MMP, MMP-2, MMP-9 and uPA that promote degradation of the basement membrane, which is a key histological marker of a tumor's transition to invasive carcinoma.80 Second, 29 days after subcutaneous injection, the presence of caMEK-expressing cells was detected in lungs of nude mice. Third, expression of activated MEK1 was sufficient to induce metastatic spread of intestinal transformed cells in the lungs of nude mice. Fourth, increased expression of factors known to promote colorectal cancer cell metastasis was found in caMEK-expressing tumors. Indeed, it has been reported that COX-2 and OPN overexpressions are associated with metastatic potential of colorectal cancer cells,50–52 and clinical studies have identified uPA as an indicator of poor overall survival in patients with colorectal cancer.53 Furthermore, Enns et al.54, 86 showed that α6β1 and α6β4 integrins are crucial for cancer cell adhesion in liver sinusoids. Finally, in support of our findings, microarray analysis in intestinal epithelial cells aimed at identifying MEK1 regulated genes showed significant OPN, COX2 and uPAR upregulation.87 Hence, in addition to demonstrating a role for MEK1 in inducing metastasis, our study also provides new insights in the nature of the downstream effectors possibly involved in this process.
The most common sites of colorectal metastasis are regional lymph nodes, liver, lungs and peritoneum and the first site of dissemination is usually the liver as a result of portal drainage.88 With the experimental approaches chosen to analyze if our intestinal cells expressing activated MEK1 can metastasize, we did not detect macroscopically liver metastases, although our cells express several markers of colorectal cancer liver metastasis including integrin α6.54 We speculate that within the time-frame of the experimental metastasis assay (21 days because after this time, mice demonstrated signs of respiratory distress), liver metastases did not have enough time to significantly develop and that caMEK-expressing cells were rapidly trapped in the well-vascularized lungs after their injection into tail vein or subcutaneously.
Finally, as we previously reported,20 little increase in ERK1/2 phosphorylation was observed in cells expressing activated MEK1 (Fig. 1a). One plausible explanation is that cells permanently stimulated by autoactive MEK1 are desensitized via multiple mechanisms including the previously reported ERK-mediated feedback inhibition of MEK89 and possible increased basal levels of MAPK phosphatases,90–92 2 phenomena previously observed in rodent fibroblasts. However, an additional species with low electrophoretic mobility was detected with the antibody recognizing the biphosphorylated and activated forms of ERK1/2 in all 3 IEC-6 cell populations (Fig. 1a). Of note, this higher molecular mass form was prominent in caMEK-expressing cells. We have previously demonstrated that this high molecular form (46 kDa) correlates nicely with IEC-6 cell proliferation and transformation induced by activated MEK1.20 As proposed previously, we believe that this 46-kDa band is in fact ERK1b, an alternatively spliced isoform of ERK1, which has recently been cloned and characterized by the group of Seger and coworkers.93, 94 Current experiments are in progress in the laboratory to understand the contribution of ERK1b variant in caMEK-induced intestinal epithelial cell transformation and EMT.
In summary, we have elucidated the consequences of expressing a constitutive activated mutant of MEK1 in normal intestinal epithelial cells. We have to mention that although this manuscript was in revision, Voisin et al. have also reporting that activation of MEK1 and MEK2 is sufficient to fully transform intestinal epithelial cells and induce invasive and metastatic tumors.75 Taken together, all these results confirm the importance of MEK signaling in intestinal tumorigenesis. Thus, MEK-ERK signaling may therefore represent an important target for developing new therapeutic approaches for treatment of colorectal cancer.
The authors thank Ms. Anne Vézina for technical assistance, Mr. Pierre Pothier for the critical reading of the manuscript and Dr. J. Carrier for her help with nude mice. The osteopontin antibody developed by Dr. M. Solursh and Dr. A. Franzen was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. This research was supported by a grant from the Canadian Institutes of Health Research Grant to Dr. Nathalie Rivard. Dr. Sébastien Bergeron is a recipient of a postdoctoral fellowship from the Canadian Association of Gastroenterology/CIHR/CCFC. Dr. Nathalie Rivard is a recipient of a Canadian Research Chair in Signaling and Digestive Physiopathology.