Acquired features of cells under epithelial-mesenchymal transition (EMT) have not yet been fully identified. The current study was conducted to assess alterations in both the proliferative potential and the responsiveness to extracellular matrices (ECMs) in EMT.
MDCK cells and SLUG-transfected MDCK clones (SLUG-MDCK) were used in this study. The cell cycle was analyzed by using flow cytometry and Western blotting. ECM-stimulated cell proliferation was examined by using the following ECMs, type I collagen, type IV collagen, fibronectin, and laminin. Protein phosphorylation was detected by immunoprecipitation-Western by using the 4G10 antibody.
Both G1 and G2/M arrest were found in the SLUG-MDCK cells, and the responsible molecules for the cell-cycle arrests were, at least in part, p21WAF1/Cip1 and Wee1. Once in contact with type I collagen, the SLUG-MDCK cells, showing the Wee1 degradation, dramatically started to proliferate up to 6-fold in cell number at Day 5, in contrast to only a 2-fold increase in the control. The analysis of the collagen receptors in the SLUG-MDCK cells disclosed a striking increase in the discoid domain receptor (DDR) 2 expression and a clear decrease in the DDR1 expression. The immunoprecipitated DDR2 protein extracted from SLUG-MDCK cells, which were cultured on collagen for 30 minutes, was tyrosine-phosphorylated, indicating valid functionality of the up-regulated receptor. The altered expression from DDR1 to DDR2 was also found in the naturally dedifferentiated sister cell lines of human liver cancer.
Epithelial-mesenchymal transition (EMT) is a biological program that induces a change in cell fate characterized by a loss in the epithelial phenotype and acquisition of a mesenschymal state.1 EMT is essential for several developmental processes, including mesoderm formation and neural-tube formation, and appears to be critical for fibrosis and cancer cell progression associated with invasion and metastasis. Several signal transduction pathways, including TGF-β, insulin-like growth factors, and Wnt, and transcription factors of the Snail family have been shown to be involved in the regulation of EMT.2–5 Although much attention has focused on the involvement of EMT in cancer invasion and metastasis, less attention has been paid to the mechanism regulating the cell cycle during EMT.
In eukaryotes, the cell cycle is tightly regulated by several protein kinases composed of cyclin-dependent kinase (CDK) subunit(s), and corresponding regulatory cyclin subunit(s), and CDK inhibitors (CKIs).6–8 CKIs are grouped into 2 distinct families based on sequence homology and targets of inhibition9; one is the INK4 family, including p15INK4b, p16INK4a, p18INK4c, and p19INK4d, and the other is the CIP/KIP family, including p21WAF1/Cip1, p27Kip1, and p57Kip2. The protein product of the retinoblastoma tumor-suppressor gene (pRb) is a negative regulator of cell proliferation and a potential substrate for cyclin E/CDK2 complex at the G1-to-S phase transition of the cell cycle.10 Both p21WAF1/Cip1 and p27Kip1 inhibit the activity of the cyclin D/CDK4, cyclin E/CDK2, and cyclin A/CDK2 complexes, whereby the phosphorylation of pRb is blocked.11 In addition, p21WAF1/Cip1 also blocks DNA replication depending on proliferation cell nuclear antigen, resulting in G1 arrest. In cells entering into mitosis, cdc2 kinase plays an important role. The critical regulatory step in activating cdc2 during progression into mitosis appears to be dephosphorylation of tyrosine (Tyr) 15 and threonine (Thr) 14.12, 13 Phosphorylation at Tyr15 and Thr14 and inhibition of cdc2 is carried out by Wee1 and the myelin transcription factor 1 protein kinases, whereas Tyr15 dephosphorylation and activation of cdc2 is carried out by the cdc25 phosphatase,12, 14, 15 leading to the activation of the cdc2/cyclin B complex to allow entry into mitosis. The amount of Wee1 protein is regulated by ubiquitin-dependent degradation through the proteasomal pathway, and the accumulation in the Wee1 protein causes G2/M cell cycle arrest.16
In the current study, we show that Wee1 is involved in the cell-cycle arrest induced by SLUG and that contact by the SLUG-MDCK cells with type I collagen triggers the decrease in the Wee1 expression. We further demonstrate that SLUG strikingly alters the expression profile of the nonintegrin-type collagen receptors discoid domain receptors (DDRs).
MATERIALS AND METHODS
Cell Lines and Cultures
Madin-Darby canine kidney (MDCK) cells and the human osteosarcoma Saos-2 cells obtained from RIKEN BioResource Center (Tsukuba, Japan), the human fetus-derived immortalized hepatocyte cell line OUMS-29,17, 18 and the human hepatoma cell lines HAK-1A, HAK-1B,19 and HLF were used in this study. The OUMS-29 cell line has cell polarity and several hepatocyte-specific gene expressions, such as albumin, α-fetoprotein (AFP), glutathione S-transferase, and asialoglycoprotein receptor.17, 18 HAK-1A and HAK-1B were 2 clonally related HCC cell lines established from a single HCC nodule that shows a 3-layered structure with a different histologic grade in each layer.19 HAK-1A resembles well differentiated HCC cells in the outer layer of the original tumor, and HAK-1B resembles poorly differentiated cells in the inner layer. HAK-1B is presumed to be derived from HAK-1A through its clonal dedifferentiation, judging from p53 mutation analysis.19 HLF cells, which have fibroblastic morphology, were obtained from the Japanese Cancer Research Resources Bank (Tokyo, Japan). Each cell line was grown in Dulbecco modified Eagle medium (Sigma-Aldrich Japan, Tokyo, Japan) supplemented with 10% heat-inactivated (56°C, 30 minutes) fetal bovine serum (BioWest, Nuaill, France), 100 U/mL penicillin, and 100 μg/mL streptomycin (Invitrogen, Carlsbad, Calif) in a humidified atmosphere of 5% CO2 at 37°C.
Transfection of cDNAs and Establishment of SLUG-overexpressing Clones
The human SLUG clone (FLAG-tagged SLUG in pPGS-CMV-CITE-Neo [-] vector) was a kind gift from Professor Eric Fearon (University of Michigan). Transient transfection of the cDNAs was performed by using TransIT-LT1 Transfection Reagent (Mirus Bio Corporation, Madison, Wis), according to the manufacturer's protocol. To create stable transfectants, MDCK cells were transfected with the SLUG cDNA or the empty vector by using the TransIT-LT1 Reagent. SLUG-FLAG-overexpressing clones were selected by 400 μg/mL G418 after transfection. Mock-transfected cells were similarly selected by G418 as the control clones.
Immunocytochemistry was performed as previously reported.11 The cells, grown on Lab-Tek Chamber Slides (Nalge Nunc Int, Naperville, Ill), were fixed with 4% paraformaldehyde for 10 minutes at room temperature, and then washed in phosphate-buffered saline (PBS) that contained 0.05% Tween 20 (PBS-T). Nonspecific reactions were blocked with Protein Block Serum-Free (Dako Japan, Kyoto, Japan) and then incubated with each 1 of the following primary antibodies (Abs) at 4°C overnight; anti-SLUG Ab (Santa Cruz Biotechnology, Santa Cruz, Calif), anti-E-cadherin Ab (clone HECD-1; Takara Bio, Otsu, Japan), anti-7H6 Ab (IBL, Takasaki, Japan), anti-ZO-1 Ab (Zymed, South San Francisco, Calif), and anti-Occludin Ab (Santa Cruz). After being washed in PBS-T, the specimens were treated with Alexa Fluor goat antimouse or antirabbit IgG Ab (Molecular Probes, Eugene, Ore) for 30 minutes at room temperature. Then, after their RNA was digested by RNase (Nippon Gene, Tokyo, Japan), the specimens were counterstained by propium iodide (PI). A confocal laser scanning microscope (FLUOVIEW FV300; Olympus, Tokyo, Japan) equipped with an argon/krypton laser capable of dual excitation and detection was used to observe the immunostaining for proteins and the nuclear localization of PI. A negative-stain control was performed by applying the mouse or rabbit IgG instead of the primary Ab in the above experiment.
Scanning Electron Microscopy (SEM)
Cultured cells were washed with culture medium and fixed with 2% glutaraldehyde in 0.1 M cacodylate buffer for 2 hours at 4°C. Samples were further fixed with 2% OsO4 in 0.1 M cacodylate buffer for 1 hour and dehydrated with ethanol. For scanning electron microscopy, dehydrated samples were critical-point dried, spatter coated with gold, and viewed under a scanning electron microscope (S-800; Hitachi, Tokyo, Japan).
Western Blot Analysis and Immunoprecipitation
Western blot analyses were performed as previously described.20 The Western blot analysis used 50 μg of protein from each sample. We used the primary antibodies E-cadherin (Takara Bio, Otsu, Japan), FLAG (Sigma-Aldrich Japan, Tokyo, Japan), p21WAF1/Cip1 and p27Kip1 (BD Biosciences, San Jose, Calif), Skp1, cyclin B1, cdc2, Wee1, Cdk2, DDR1, DDR2, osteopontin, β1 integrin, and αVβ3 integrins (Santa Cruz Biotechnology, Santa Cruz, Calif). For immunoprecipitation, cells were lysed in lysis buffer (50 mmol/L Tris-HCl [pH 8.0], 150 mmol/L NaCl, 0.5% Nonidet P-40, 1 mmol/L EDTA [pH 8.0], 1 mmol/L EGTA [pH 8.0], 0.1 mmol/L sodium fluoride [NaF], 0.1 mmol/L sodium orthovanadate [Na3VO4], 1 mmol/L dithiothreitol, 2 g/mL aprotinin, and 2 g/mL leupeptin). Cell lysates were centrifuged at 12,000 g for 20 minutes at 4°C, and the supernatants were separated. The protein concentration was measured by using a Bio-Rad protein assay kit (Bio-Rad, Hercules, Calif). The cell lysates (500 μg of protein) were subjected to immunoprecipitation with the use of anti-DDR2 antibody (Santa Cruz) and Recombinant Protein G Agarose (Invitrogen, Carlsbad, Calif). Immunoblots were probed with antibodies to DDR2 and to phosphotyrosine (4G10) (Upstate, Lake Placid, NY). The bound antibodies were detected with horseradish peroxidase-labeled sheep antimouse IgG or horseradish peroxidase-labeled donkey antirabbit IgG (Amersham Pharmacia Biotech, Buckinghamshire, UK) by using the enhanced chemiluminescence detection system (ECL Advanced kit, Amersham). A positive signal from the target proteins was observed by using an image analyzer LAS-1000 plus (Fujifilm, Tokyo, Japan), and the signal intensity was semiquantitatively determined by using an Image Gauge version 3.45 (Fujifilm).
The DNA content was assessed by staining ethanol-fixed cells with PI and monitoring by FACSCaliber (Becton Dickinson, Franklin Lakes, NJ). The percentage of cells in the G0/G1 phase, the S phase, and the G2/M phase was determined by using ModFIT software (Verity Software House, Topsham, Me).
Wee1 Protein Stability on Type I Collagen
The SLUG-MDCK cells cultured in poly-D-lysine (PDL)-coated dishes were detached, and the cells were seeded onto new PDL-coated dishes and type I collagen-coated dishes for incubation for 10, 20, and 30 minutes. The cells harvested at these times were immediately subjected to Western blot analysis to determine the expression level of Wee1 protein. To assess the stability of the Wee1 protein on the type I-collagen coat, the proteasome inhibitor MG132 (Calbiochem, San Diego, Calif) was used.
Cell Proliferation on Extracellular Matrices (ECMs)
From both MDCK cells and SLUG-MDCK cells, 1 × 104 cells each were seeded onto 3 wells in the 6-well plate coated with 1 ECM, type I collagen, type IV collagen, fibronectin, or laminin (BD Biosciences, San Jose, Calif). After these cells were cultured for 5 days, the cell numbers in 6 wells for each ECM were counted in duplicate by using a CDA-500 automated cell counter (Sysmex, Kobe, Japan). To compare the proliferative response to type I collagen among MDCK cells, mock-transfected MDCK cells, and SLUG-MDCK cells, 3 × 104 cells each were cultured in 6 dishes (60 mm in diameter), and the cell numbers were counted in duplicate on Days 1, 2, 3, and 5. An increased proliferative response to type I collagen was judged by the ratio of the cell number on the type I collagen-coated dish to the cell number on the PDL-coated dish (the collagen:PDL ratio).
Statistical significances were assessed by using the Mann-Whitney U test. P < .05 was considered statistically significant.
EMT Caused by SLUG
We initially confirmed a strong nuclear staining of SLUG in the SLUG-MDCK cells (Fig. 1A). These cells either did not show any cell surface localization of E-cadherin (Fig. 1A) or the expression level of E-cadherin was undetectable by Western blot analysis (Fig. 1B), suggesting disruption of the cell-to-cell adhesion. Indeed, no clear tight junction-associated protein, such as 7H6, ZO-1, or occludin, was detected in the SLUG-MDCK cells (Fig. 1C). A fibroblastic transformation was found in SLUG-MDCK cells by scanning electron microscopy, in contrast to polygonal MDCK cells, which showed a sheet-like epithelial growth pattern and abundant surface villi (Fig. 1A). The mesenchymally transformed SLUG-MDCK cells showed a cleaved (active) form of MMP-2, and clear increases in the expression levels of MMP-9, osteopontin, and vimentin. These findings confirmed establishment of EMT in the SLUG-MDCK cells, demonstrating mesenchymal features in cellular enzymes, secreted protein,21 and in structural protein (Fig. 1D).
Cell-Cycle Arrest in SLUG-MDCK Cells
Because we found lower proliferative potential in the SLUG-MDCK cells than in either the parental MDCK cells or the mock-transfected MDCK cells, we performed cell-cycle analysis using these 3 kinds of cells. In the SLUG-MDCK cells, a significant increase was found in the cell population at both the G0/G1 and the G2/M phases, accompanied by a significant decrease at the S phase (Fig. 2A). The percentage of the cell population at the G0/G1 phase in the mock-transfected MDCK cells versus the SLUG-MDCK cells was 33.85 ± 0.80 versus 43.42 ± 1.69 (%, mean ± standard deviation [SD]; P < .01), and 22.65 ± 0.29 versus 28.33 ± 0.62 at the G2/M phase (%, mean ± SD; P < .01), respectively. Then, we examined the possible molecules responsible for the cell-cycle arrest at both the G0/G1 and the G2/M phases by Western blot analysis and found that p21WAF1/Cip1 and Wee1 were each related somehow to these arrests (Figs. 2 B and C).
Type I Collagen Promoted Proliferation of SLUG-MDCK Cells
To investigate any role of ECMs in stimulating cell proliferation of the arrested SLUG-MDCK cells, we seeded the cells onto dishes coated with an ECM, type I collagen, type IV collagen, fibronectin, or laminin, and counted the cell number after incubation for 5 days. In the SLUG-MDCK cells, the highest proliferative activities were found with type I collagen, and with fibronectin, in contrast to no significant difference in the ECM-induced cell proliferation in MDCK cells (Fig. 3A). Because we found a profound expression of Wee1 protein, a fragile protein, in the SLUG-MDCK cells cultured on PDL-coated dishes (Fig. 2B), we focused on its possible rapid decrease in contact with type I collagen. Indeed, the expression level of Wee1 was clearly decreased, immediately after the cells were seeded onto the type I collagen-coated dish (Fig. 3B), suggesting a rapid degradation of this protein. When SLUG-MDCK cells keeping contact with type I collagen were treated with the proteasome inhibitor MG-132, the expression level of Wee1 was clearly increased (Fig. 3C), suggesting that the type I collagen triggered the degradation of this protein through the ubiquitin-proteasome pathway. To semiquantitatively assess the increased responsiveness to type I collagen in SLUG-MDCK cells, the number of the cells on type I collagen-coated dishes was compared with that on PDL-coated (control) dishes on Day 1, 2, 3, and 5. The actual cell numbers of the MDCK cells on control dishes and type I collagen-coated dishes were 5.11 ± 0.07 and 5.46 ± 0.12 on Day 1, 11.10 ± 0.14 and 14.24 ± 0.24 on Day 2, 19.98 ± 0.16 and 25.66 ± 0.22 on Day 3, and 77.82 ± 0.91 and 158.40 ± 1.14 on Day 5 (mean ± SD). The respective cell numbers in the mock-transfected MDCK cells were 5.06 ± 0.10 and 5.35 ± 0.13 on Day 1, 12.61 ± 0.16 and 15.28 ± 0.53 on Day 2, 20.02 ± 0.51 and 25.19 ± 0.78 on Day 3, and 80.22 ± 0.98 and 156.10 ± 1.44 on Day 5 (mean ± SD). In case of the SLUG-MDCK cells, the cell numbers were 2.44 ± 0.05 and 3.67 ± 0.11 on Day 1, 3.53 ± 0.06 and 8.57 ± 0.15 on Day 2, 4.99 ± 0.09 and 14.74 ± 0.15 on Day 3, and 9.08 ± 0.09 and 53.72 ± 0.57 on Day 5 (mean ± SD). The calculated collagen:PDL ratios in cell numbers for the MDCK cells (and for the mock-transfected MDCK cells) versus the SLUG-MDCK cells were 1.069 ± 0.037 (1.057 ± 0.045) versus 1.504 ± 0.060 on Day 1 (P < .01), 1.283 ± 0.015 (1.212 ± 0.021) versus 2.427 ± 0.063 on Day 2 (P < .01), 1.284 ± 0.011 (1.258 ± 0.029) versus 2.955 ± 0.074 on Day 3(P < .01), and 2.036 ± 0.032 (1.946 ± 0.039) versus 5.920 ± 0.112 on Day 5 (P < .01), respectively (Fig. 3D). The significantly increased collagen:PDL ratios in the SLUG-MDCK cells indicated that these cells were highly sensitive to type I collagen, driving their cell cycle.
DDR Switching in SLUG-MDCK Cells
On the basis of the higher sensitivity of SLUG-MDCK cells to type I collagen, shown in this study, we focused on a possible alteration in the expression profile of collagen receptors, including DDR1 and DDR2. In SLUG-MDCK cells, a striking increase in the DDR2 expression was found, in contrast to a clear decrease in the DDR1 expression (Fig. 4A). In MDCK cells, the predominant DDR was DDR1, but not DDR2, suggesting involvement of switching from DDR1 to DDR2 in the SLUG-induced EMT. The integrin β1 expression level was slightly increased in the SLUG-MDCK cells. The integrin αVβ3 expression level was not changed (data not shown). To assess whether the DDR2-abandunt SLUG-MDCK cells respond well to type I collagen, we examined the tyrosine phosphorylation status of the DDR2 from the collagen-contacted SLUG-MDCK cells by using the immunoprecipitation-Western blotting assay. The DDR2 derived from the SLUG-MDCK cells seeded onto type I collagen-coated dish was highly tyrosine-phosphorylated, whereas the DDR2 from the control cells was, in contrast, barely phosphorylated (Fig. 4B).
Expression of DDR2 in HCC Dedifferentiation
Because HAK-1B hepatoma cells were the clonally aggressive type derived from HAK-1A hepatoma cells, we investigated whether EMT-associated DDR switching was caused (or not) in these naturally dedifferentiated sister cell lines. As shown in Figure 5, the expression level of DDR2 was clearly predominant in HAK-1B cells compared with that in HAK-1A cells. In contrast, the expression level of DDR1 was reciprocally decreased in HAK-1B cells, although the decrease was not so striking in comparison with that in SLUG-MDCK cells. The polarized human hepatocyte cell line OUMS-29 was similar to HAK-1A in DDR expression profile, as both showed a clear expression of the epithelial marker pan-cytokeratin (Fig. 5). The representative malignant mesenchymal cell line Saos-2 demonstrated a remarkably higher expression level in DDR2 than in DDR1 (Fig. 5).
In the current study, we have demonstrated the following new findings in the SLUG-induced EMT model, 1) Wee1-associated cell-cycle arrest, 2) a significantly accelerated cellular proliferation in response to type I collagen, and 3) a clear switching from DDR1 to DDR2.
It has recently been demonstrated that the transcription factor Snail attenuated the cell cycle in the MDCK cells and the human hepatoma HepG2 cells.22, 23 The Snail-induced G0/G1 cell-cycle arrest shown in previous studies was due to the repression of Cyclin D2 transcription, the increase22 in p21WAF1/Cip1, and the possible direct up-regulation23 in p15INK4b. In the current study, up-regulation of p21WAF1/Cip1 was also found in the SLUG-MDCK cells, showing cell-cycle arrest at both the G0/G1 and the G2/M phases. The expression level of p27Kip1, another CKI, was not altered by SLUG in the current study, and this was consistent with the previous finding in the Snail study.22 Both Snail and SLUG are categorized as zinc-finger-type transcription factors. Thus, this type of EMT-inducing factor was thought to up-regulate CKIs, thereby affecting the cell-cycle progression during the EMT process. In addition to the G0/G1 cell-cycle arrest by Snail,22, 23 the G2/M cell-cycle arrest was first demonstrated in the SLUG-overexpressing cells used in our study. It is well known that the G2/M check point in the cell cycle is tightly regulated by the Wee1 protein. Indeed, in the current study, the Wee1 expression was strikingly increased in the SLUG-MDCK cells, suggesting some involvement in the G2/M cell-cycle arrest. Of interest, the Wee1 expression was immediately diminished, when the SLUG-MDCK cells were in contact with type I collagen. This strongly suggested that the Wee1 protein was rapidly degraded, at least in part, by the proteasome pathway, allowing the cells to escape from the Wee1-associated G2/M cell-cycle arrest, thereby leading to a boost in cellular proliferation. Besides the cell-cycle-blocking effect of Snail, the previous study also emphasized the Snail-induced resistance to cell death.22 Thus, the findings on the cell-cycle arrest induced by Snail and SLUG together may suggest that these factors do not always confer deregulation or an increase in proliferation to the transformed cells, which is crucial for tumor formation and growth.22
Although there was no significant difference in the cell number of the MDCK cells cultured on each distinct ECM tested for 5 days, a significant increase in the cell number was found in the SLUG-MDCK cells with type I collagen and with fibronectin. The SLUG-MDCK cells proliferated approximately 6-fold when they were in contact with type I collagen on Day 5, whereas the parental MDCK cells proliferated approximately 2-fold. The highly elevated proliferation ratio in the type I collagen-contacted SLUG-MDCK cells was coupled with a rapid degradation in Wee1 protein, suggesting that the release from the G2/M cell-cycle arrest was, at least in part, a trigger for their accelerated proliferation.
The type I collagen-induced accelerated proliferation prompted us to investigate alteration in the expression profile of collagen receptors in the SLUG-MDCK cells, with our focus on the expression levels in integrin β1 and DDRs. It is known that both DDR1 and DDR2 are widely expressed in human tissues, but their distribution is mutually exclusive: DDR1 is expressed primarily in epithelial tissues, and DDR2 is found in mesenchymal cells.24 Indeed, in our study, a striking increase in the DDR2 expression and a reciprocal decrease in the DDR1 expression were found in the mesenchymally transformed SLUG-MDCK cells. In addition, the newly expressed DDR2 in the SLUG-MDCK cells was phosphorylated in contact with type I collagen, suggesting that the induced receptor was physiologically active. Despite this clear switching from DDR1 to DDR2, there was relatively less alteration in the expression level of integrin β1. The underlying mechanism for the preferential impact of SLUG on the DDR expressions was not identified in this study; however, it was noteworthy that the DDR signaling was independent from the integrin β1 signaling.25 Therefore, it was speculated that DDRs were more flexible in their expression levels than integrin β1 in response to type I collagen during the SLUG-induced EMT program, regulating the distinct downstream intracellular signaling from that of integrin β1. Although the increase in the expression level of integrin β1, a major component of the fibronectin receptor, was slight in the SLUG-MDCK cells, it may contribute to the high proliferative response to fibronectin.
Recent evidence has suggested that DDRs can regulate cell proliferation and extracellular matrix remodeling mediated by MMP activities during normal development or pathological conditions26, 27; however, their relevance to cancer progression has not yet been assessed fully. In human nasopharyngeal carcinoma, the up-regulation of DDR2, but not of DDR1, has been demonstrated.28 Indeed, in our study, the DDR2 expression was more profound in the aggressive HAK-1B hepatoma cells than in the clonally sister cell line HAK-1A, which resembled well-differentiated hepatoma cells.19 These findings suggested that the predominant DDR2 expression reflected a result of the dedifferentiation and/or EMT process toward more malignant cells. However, this concept may not always be applied to every cancer type, because DDR1, and not DDR2, showed increased expression level in human lung cancer tissues, demonstrating its predictive potential for patient survival.29 Further investigation is needed to determine the clinical implication of the switching from DDR1 to DDR2 in the context of EMT-associated cancer invasion and metastasis.
The authors thank Masako Hayakawa for technical assistance.