c-Myc expression and MEK1-induced Erk2 nuclear localization are required for TGF-beta induced epithelial–mesenchymal transition and invasion in prostate cancer

Abstract

Understanding the initial mechanisms by which epithelial cells transform to an invasive phenotype is critical to the development of diagnostics that can identify the metastatic potential of cancers as well as therapeutic agents that can prevent metastases. Changes in cellular response to the transforming growth factor-beta (TGF-β) cytokine are known to promote epithelial cell invasion and metastasis in part through induction of epithelial–mesenchymal transitions (EMTs). In this report, we demonstrate that non-metastatic human prostate cancer cell lines of increasing Gleason score can be induced to undergo EMT when treated with TGF-β in combination with epidermal growth factor. Mechanistic studies revealed that in cells stably transfected with activated Ras, TGF-β alone induced EMT and that a Ras-Raf-MEK1, but not MEK2, signaling cascade is necessary and sufficient for Erk2 nuclear localization that works in concert with TGF-β to promote EMT. Furthermore, we show for the first time that expression of the transcription factor c-myc, which is phosphorlyated by Erk2, is required for EMT. Characteristically, EMT involved adoption of a spindle-shaped morphology, loss of E-cadherin and increased expression of Vimentin, Fibronectin and Fibroblast Specific Protein-1 (S100A4). Prostate cells undergoing EMT became invasive and expressed several genes associated with metastasis, including MT-MMP1, MMP-2/9, the MMP-9 homodimer, Slug and Twist2. In sum, we demonstrate a novel mechanism by which non-invasive primary prostate tumor cells transition to an invasive phenotype characteristic of malignant tumor cells in response to TGF-β signaling.

Introduction

Epithelial–mesenchymal transition (EMT) is mostly described as part of germ layer reorganization and tissue remodeling during embryonic development. However, it has become increasingly clear that a reactivation of the EMT developmental program primes malignant epithelial cells for the dissemination and invasion required for metastatic spread of solid tumors, the foremost cause of mortality in prostate cancer patients (1). During EMT, tumor cells lose cell–cell contacts and the cobblestone networks characteristic of epithelial tissues and adopt a spindle-shaped morphology and migratory phenotype typical of fibroblasts (2). Additionally, E-cadherin and β-catenin expression at cell–cell junctions is lost as cells express mesenchymal-associated genes such as Vimentin, Fibronectin and Fibroblast Specific Protein-1 (FSP-1, also known as S100A4) (3). Importantly, these changes in gene expression are correlated with an increasingly invasive and aggressive tumor cell phenotype that is associated with a poorer patient prognosis (4–6). Silencing of Vimentin or re-expression of E-cadherin in invasive cells also decreases their invasive phenotype, emphasizing that these genes play a major role in controlling the metastatic behavior of tumor cells (7–9). Likewise, transcription factors that serve as master regulators of EMT, including those of the Snail, Zeb and Twist families, have repeatedly been shown to be associated with increased malignancy and to regulate carcinoma cell movement and metastasis (10–17). Therefore, understanding the initial molecular mechanisms regulating the EMT phenotype in prostate cancer will aid in identification of new tumor biomarkers or therapeutics to target cells with a higher metastatic potential. Currently little is known on what the key regulators of metastatic potential are in prostate cancer.

EMT is induced by various growth factors; specifically, transforming growth factor-beta (TGF-β) appears to be the most ubiquitous instigator of EMT during development and cancer (3,18,19). In canonical TGF-β signaling, TGF-β ligands activate TGF-β transmembrane receptors that phosphorylate latent Smad proteins that form transcription factor complexes, which regulate the expression of TGF-β-responsive genes (20,21). In addition, TGF-β activates a variety of non-canonical pathways, including the AKT, mitogen-activated protein kinase (MAPK), c-Jun N-terminal kinase and NF-kappaB pathways (21–25). MAPK activation by TGF-β also represents an important mechanism for Smad signaling by phosphorylating various transcription factors in the nucleus of cells that physically interact with Smads and regulate TGF-β responses (21,26,27). Interestingly, both TGF-β-induced Smad signaling and non-canonical Ras-MAPK activation are required for EMT; however, many cancer cell lines exhibiting proficient TGF-β signal transduction do not undergo TGF-β-mediated EMT (28–31). These findings suggest that TGF-β may require significant crosstalk with other pathways to coordinate EMT. In some instances, TGF-β-induced EMT and metastasis is dependent on sustained elevated levels of active Ras-MAPK signaling resulting from Ras overexpression or hyperactivity (32–34). Thus, although the importance of Ras signaling in promoting EMT is well documented, why non-canonical TGF-β activation of the Ras-MAPK pathway is not sufficient to induce EMT alone in these models remains unresolved.

In studies of the prostate cancer, ArCAP model using transformed cells, simultaneous treatment with epidermal growth factor (EGF) and TGF-β induces both EMT and increased metastatic potential (33). One plausible explanation is that EGF activates signaling events controlling Ras signaling dynamics that work in concert with TGF-β to help induce EMT in earlier stages of cancer. Using non-transformed and hTERT immortalized primary prostate cells isolated from human prostates of increased Gleason score (GS 6 to 8), we report that TGF-β combined with EGF or Ras overexpression drives EMT and invasion in earlier cancer stages. Specifically, we found that MEK1 signaling downstream of Ras was necessary and sufficient for TGF-β-induced EMT and that EGF and MEK1 signaling was sufficient to induce nuclear accumulation of the MEK1/2 effector molecule, Erk2, which correlated with EMT. Notably, TGF-β treatment alone was unable to induce Erk2 nuclear accumulation despite inducing its phosphorylation. Furthermore, we demonstrate that a mutant Erk2 construct that accumulates in the nucleus is sufficient to drive TGF-β-induced EMT in early-grade prostate cancer cells, and that this relies on expression of the c-myc transcription factor. In sum, we demonstrate a novel mechanism by which MEK1 signaling promotes the transition of primary non-invasive tumor cells to an invasive phenotype characteristic of malignant tumor cells in response to TGF-β.

Materials and methods

Cells

IBC-10a , PCa-20a and PCa-30a cells were isolated from the right peripheral zone of a GS 6, 7 and 8 prostate tumors respectively, as described previously (35). IBC-10a cells were immortalized by stable transfection with the pLXSN-hTERT retroviral plasmid (courtesy of John Rhim, USUHS, Bethesda, MD) and identified as intermediate basal cells. They possess minimal gross chromosomal abnormalities and express CK5, CK18, p63, PSA and PTEN (35). PCa-20a and PCa-30a cells expressed CK18, PTEN and PSA but not CK7 or p63. Cells were maintained in serum-free complete keratinocyte media (Km) containing EGF, bovine pituitary extract (Invitrogen Inc., Carlsbad, CA) and 50 ug/ml penicillin/streptomycin (Mediatech, Manassas, VA). PC3-ML cells were isolated from PC3 prostate cancer cells (ATCC, Bethesda, MD) based on their ability to metastasize to the lumbar vertebrae (36). PC3-ML cells were maintained in DMEM with 10% fetal bovine serum (Atlanta Biologicals, Lawrenceville, GA) and 50 ug/ml penicillin/streptomycin.

RasV12, Ras V12S35, RasV12C40 and RasV12G37 were stably overexpressed in both IBC-10a and PCa-20a cells using the pBABE-puro retroviral vector (a kind gift from Dr Christian Sell, Drexel University College of Medicine). MEK1-DD and MEK2-DD were also overexpressed in cells using the pBABE-puro retroviral vector (a kind gift of Dr Mauricio Reginato, Drexel University College of Medicine). HA-Erk2 WT and HA-Erk2 D319N were expressed in cells using the pLNCX retroviral vector (a kind gift from Dr Claudio Torres, Drexel University College of Medicine). Scrambled shRNA constructs and shRNA constructs targeting c-myc (TRCN0000039641) were purchased from Sigma (St Louis, MO). shRNA constructs targeting Erk2 were a kind gift from the lab of Dr John Blenis (Harvard Medical School) and Dr Peter Lelkes (Drexel University) (37). Retroviral and lentiviral production and maintenance of transfected cells was carried out according to methods described previously (38).

Antibodies

Western blot and immunoflourescence was carried out according to methods described previously (39). For western analysis, primary antibodies targeting Vimentin and Fibronectin were purchased from Sigma–Aldrich (St Louis, MO) (V6389 and F3648, respectively); E-cadherin, Tubulin, phosphorylated-Erk1/2, phosphorylated-Smad3, phosphorylated-Akt, c-myc and Slug were purchased from Cell Signaling Technology (Beverly, MA) (2148, 4065, 9106, 9520, 4060, 9402 and 9585, respectively); FSP-1 and Twist2 were purchased from Abcam (Cambridge, MA) (ab27597 and ab57997, respectively); and phosphorylated-c-myc was purchased from Millipore (Bellerica, MA) (04-217). For immunoflourescence, primary antibodies targeting Vimentin were purchased from Sigma–Aldrich (V6389); β-catenin was purchased from Cell Signaling Technology (9582); and Erk2 was purchased from Santa Cruz biotechnology (Santa Cruz, CA) (sc-1647).

EMT induction

Unless otherwise stated, for in vitro induction of EMT, cells were trypsinized and plated in growth media at a low density (2 × 10⁴ cells/ml). The next day (day 0), cells were washed once in minimal media (Km) without supplements (i.e. pituitary extract and EGF), and media was replaced with Km supplemented with TGF-β1 (10ng/ml; Peprotech, Rocky Hill, NJ) and/or EGF (10ng/ml; Invitrogen). Media in all experiments were changed on days 3, 6 and 9, and cells were analyzed on day 10.

Quantitative real-time PCR

Total RNA was isolated using Qiagen’s RNAeasy isolation kit per manufacturer instructions. Target genes were amplified using the one-step Brilliant SYBR Green qRT–PCR kit (Stratagene, La Jolla, CA) and the MX3000P thermocycler (Stratagene) per manufacturer instructions. Primers used were as follows: Cyclophilin A: Fwd-GTGACTTCACACGCCTATATG, Rev-ACAAGATGCCAGGACCGTA;

 Snail: Fwd-GCTGCCAATGCTCATCTGGGACT,

Rev-CAGGGAGGTCAGCTCTGCCA;

 Slug: Fwd-TCAGCTCAGGAGCATACAGC,

Rev-GACTCCACTCGCCCCAAAGA;

 Twist1: Fwd-GTCCGCAGTCTTACGAGGAG,

Rev-CCAGCTTGAGGGTCTGAATC;

 Twist2: Fwd-AGCAAGAAGTCGAGCGAAGA,

Rev-CAGCTTGAGCGTCTGGATCT;

 Zeb1: Fwd-TATGAATGCCCAAACTGCAA,

Rev-TGGTGATGCTGAAAGAGACG;

 Zeb2: Fwd-CGCTTGACATCAATGAAGGA,

Rev-CTTGCCACACTCTGTGCATT;

 Vimentin: Fwd-CCCTCACCTGTGAAGTGGAT,

Rev-TCCAGCAGCTTCCTGTAGGT;

 MT-MMP1: Fwd-ACATTGGAGGAGACACCCAC,

Rev-TAGGCAGTGTTGATGGACGC;

 MMP-2: Fwd-CAAAAACAAGAAGACATACATCTT.

Rev-GCTTCCAACTTCACGCT;

 MMP-9- Fwd: CCCTGGAGACCTGAGAACCA, Rev:

CCCGAGTGTAACCATAGCGG.

Using the 2^-ddCt method, empty vector or parent cells grown in minimal media were used to normalize gene expression across treatments. Relative internal mRNA expression of target genes was normalized to Cyclophilin-A expression in each sample. Each sample for each experiment was run in duplicate and averages are representative of three independent experiments. Statistical significance was determined using Welch’s unpaired t-test.

Nuclear isolation

Cells were initially lysed with a Triton-X cytoplasmic extraction buffer for 10min at 4°C (10mM HEPES, pH 7.9, 50mM NaCl, 0.5M sucrose, 0.1mM EDTA, 0.5% Triton-X and protease inhibitor cocktail). Cells were scraped and nuclei collected by centrifugation at 1000rpm in a swing bucket rotor at 4^oC for 10min. Pellet was washed in wash buffer (10mM HEPES, pH 7.9, 10mM KCl, 0.1mM EDTA and 0.1mM EGTA) and centrifuged at 1000rpm for 5min. Nuclei were lysed with lysis buffer (10mM HEPES, pH 7.9, 500mM NaCl, 0.1mM EDTA, 0.1mM EGTA, 0.1% NP40, 0.05% SDS and protease inhibitor cocktail), vortexed and extract was cleared by centrifugation at 14 000rpm at 4°C for 10min.

Zymography

Evaluation of enzymatic activity of matrix metalloproteinases (MMPs) was assessed using gelatin zymography as described previously (40). Media conditioned for 24h on day 10 was collected and protein in conditioned media was concentrated 10-fold using Amicon Ultra-15 centrifugation filter devices according to manufacturer instructions (Millipore). Between 0.1 and 1 ug of protein depending on cell type was loaded and run on a 10% polyacrylamide gel containing 2mg/ml of gelatin A.

Invasion assays

Following experimental treatments, cells were trypsinized and seeded onto Matrigel-coated invasion inserts with 0.8 um porous membranes (BD Biosciences, Bedford, MA) at a density of 5 × 10⁴ cells per well in growth media and allowed to attach for 2h. Medium on the top chamber was then changed to experimental condition and bottom chamber was filled with growth medium containing 5% fetal bovine serum (Atlanta Biologicals). Transwells were placed at 37°C for 48h. Cells in top compartment were scraped off and cells that migrated to bottom were either fixed with 4% paraformaldehyde and stained with 0.1% crystal violet or trypsinized and counted using a hemocytometer. Data were averaged from three independent experiments. Prostashperes were produced as described previously and topped with minimal media (Km) containing experimental condition, 0.2% fetal bovine serum (Atlanta Biologicals) and 5% Matrigel (41). Medium was changed every 3 days with experimental condition and 5% Matrigel. Prostasphere acini were analyzed after 12 days of culture.

Results

EGF and TGF-β function synergistically to induce EMT in primary non-invasive epithelial cells isolated from prostate cancer.We previously isolated three different human prostate epithelial cell lines (termed IBC-10a, PCa-20a and PCa-30a) from tumors of increasing GS (6, 7 and 8, respectively) (35,42). Previous studies have shown that TGF-β alone or in conjunction with other growth factors can induce EMT in transformed cells, but whether these ligands might normally induce EMT in non-immortalized primary cells has yet to be shown. Therefore, we treated each cell line with either minimal media (Km) as a control, EGF, TGF-β1 (TGF) or both EGF and TGF-β1 in combination (E + T) (as described in Materials and methods) and analyzed the expression of mesenchymal and epithelial-associated proteins. Treatment of all three cell lines with Km or EGF failed to induce expression of several EMT-associated genes, including Fibronectin and Vimentin (Figure 1A). In all cell lines, TGF-β alone was sufficient to induce Fibronectin; however, a significant loss in E-cadherin expression and induction of Vimentin and FSP-1 only occurred in more malignant PCa-30a cells (Figure 1A). In contrast, cotreatments of all three cell lines with E + T induced a robust EMT response as characterized by expression of Vimentin and FSP-1, loss of E-cadherin, disruption of epithelial cell–cell contacts, cytoplasmic accumulation of β-catenin and adoption of a spindle-shaped morphology (Figure 1A and 1B, and Supplementary Figure 1, available at Carcinogenesis Online). Expression of these EMT markers may be associated with the metastatic phenotype in prostate cancer; therefore, we sought to understand if these markers were expressed in the highly metastatic PC3-ML cell line or if they were regulated by TGF-β and EGF. We found that PC-3ML cells constitutively expressed Fibronectin, Vimentin and FSP-1 and lacked E-cadherin expression (Figure 1A, and Supplementary Figure 1, available at Carcinogenesis Online).

Fig. 1.

EGF and TGF-β act synergistically to induce EMT in primary human prostate epithelial cells. (A) Western blots of crude cell extracts from cells treated with minimal media (Km), or Km containing 10ng/ml EGF (EGF), 10ng/ml TGF-β (TGF) or 10ng/ml of EGF + TGF-β (E + T) as described in Materials and methods. IBC-10a cells, PCa-20a cells and PCa-30a cells were isolated from GS 6, 7 and 8 prostate cancers, respectively. Blots were probed with antibodies specific for Fibronectin, E-cadherin, Vimentin, FSP-1 and Tubulin (loading control). (B) Images of IBC-10a cells cultured in Km, EGF, TGF-β or E + T. Phase contrast images show a reduction in cell–cell contacts in E + T-treated IBC-10a and acquisition of a spindle-shaped morphology (see red arrows; top panel). Immunofluorescent images of cells labeled with antibodies specific for β-catenin (middle panel; green) and Vimentin (bottom panel; green). Cell nuclei (red) were labeled with propidium iodide (top/middle panels:×400 magnification; bottom panel: ×200 magnification). (C) Western blot analysis of E-cadherin, Vimentin and Actin (loading control) expression in IBC-10a cells treated with E + T at day 0 through 9 and 4 days post-treatment where cells were cultured in Km. (D) Induction of EMT in prostate organ cultures. Phase contrast and immunofluorescent imaging shows prostate epithelial cells derived from GS6 prostate cancer tissue, emigrating out of the tissue and strongly expressing Vimentin (red) in response to concomitant E + T treatment, but not TGF-β alone; nuclei labeled with DAPI (×200 magnification).

Notably, a stable EMT phenotype was maintained as indicated by continued expression of Vimentin in cells cultured for an additional 4 days following discontinuation of the EMT-inducing treatments (Figure 1C). To ensure that E + T-induced EMT was not an artifact associated with cell lines, cell passage or continued growth in EGF-containing media, we treated freshly established organ cultures from a GS 6 prostate cancer specimen with the different ligands. These organ cultures developed outgrowths of prostate epithelial cells and we observed that E + T, but not TGF-β alone, induced significant morphological changes reminiscent of EMT and promoted Vimentin expression after ~6 days of treatment (Figure 1D). Taken together, these results suggest that signaling pathways activated by both EGF and TGF-β function synergistically to induce EMT in epithelial cells derived from low-grade prostate tumors. Furthermore, they imply that induction of EMT by TGF-β does not require transformation of primary cell lines; rather TGF-β induction of EMT may be a characteristic of epithelial cells isolated from higher grade tumors.

EGF signaling modulates cellular responses to TGF-β to induce the upregulation of pro-metastatic genes and an invasive phenotype. Several transcription factors, including those of the Snail (Snail and Slug), Twist (Twist1 and Twist2) and Zeb (Zeb1 and Zeb2) families, have been identified as important regulators of EMT and are required for cell movement and metastatic spread in a variety of cancers (10–16,43). We observed that E + T treatment induced expression of Slug and Twist2 in IBC-10a cells (4- and 9-fold increase, respectively) and PCa-20a cells (3- and 8-fold increase, respectively) (Figure 2A and 2B). Treatment of these cells with EGF or TGF-β alone failed to elicit significant changes in the expression of Slug. EGF alone induced Twist2 expression in both IBC-10a and PCa-20a cells but less than that observed by E + T treatment (Figure 2A and 2B). In PC3-ML cells, TGF-β alone was sufficient to upregulate Slug and Twist2 mRNA 2.5- and 3-fold, respectively (Figure 2A). EGF alone had no effect on the expression of these genes, and E + T treatment was as efficacious as TGF-β treatment alone (Figure 2A). In contrast, the expression of Snail, Twist1 and Zeb1/2 was not induced by these ligands in any of our primary cell lines (Supplementary Figure 2A, available at Carcinogenesis Online). However, PC3-ML cells expressed a high basal level of Zeb1 and Twist2 (Supplementary Figure 2B, available at Carcinogenesis Online). As expected, PC3-ML cells constitutively expressed high levels of Vimentin in minimal media regardless of treatment (Figure 2A, and Supplementary Figure 2B, available at Carcinogenesis Online).

Fig. 2.

EGF signaling modulates TGF-β signaling to upregulate genes associated with an invasive phenotype in primary prostate epithelial cells. (A) Expression of Vimentin, Slug, Twist2, MT-MMP-1, MMP-2 and MMP-9 genes in IBC-10a (left), PCa-20a (middle) and PC3-ML cells (right). Effects of EGF (10ng/ml) (light gray), TGF-β (10ng/ml) (dark gray), and EGF + TGF-β (10ng/ml each) (E + T) (black) treatments on gene expression was determined by qRT–PCR. Values were normalized to cyclophilin-A and represent the fold increase in expression relative to cells treated with Km (white). Data represent the mean ± SD from three independent experiments. Statistical significance comparing TGF-β treatment alone with E + T was determined using Welch’s unpaired t-test, *P < 0.05, **P < 0.01; ***P < 0.001; #, no significant difference (P > 0.1). (B) Western blot analysis of Slug and Twist2 in IBC-10a cells treated 9 days with Km, EGF, TGF-β or E + T. Tubulin served as the loading control. (C) Representative gelatin zymographs of conditioned media from IBC-10a, PCa-20a and PC3-ML cells treated with EGF, TGF-β or E + T. Equal amounts of protein from conditioned medium from each treatment was resolved by 10% PAGE containing gelatin-A (2mg/ml). Coomassie Blue staining reveals relative levels of enzyamtically active MMP-9 homodimer (210kDa), MMP-9 (92kDa) and MMP-2 (72kDa) in media. (D) Modified Boyden chamber invasion assays showing cell migration following treatment with Km, EGF, TGF-β or E + T for 48h. Representative image of cells on bottom of chamber stained with brilliant blue dye. (E) Prostasphere culture showing three-dimensional acinar structures formed by IBC-10a cells grown as single-cell suspensions in Matrigel for 12 days in the presence of different growth factors Top panel: phase contrast images show cells grew as rounded spheres in the presence of Km, EGF and TGF-β. In contrast, cells grown in the presence of E + T were irregularly shaped and cells invaded the surrounding Matrigel (see red arrows). Lower panel: immunofluorescent labeling of prostaspheres with Vimentin antibodies (green); nuclei labeled with propidium iodide (×200 magnification).

The upregulation of MMPs, including MMP-2, MMP-9 and MT-MMP1, is also associated with acquisition of an EMT phenotype and is important to break down stromal barriers during invasion and metastasis (44). In IBC-10a and PC-20a cells, treatment with E + T induced a robust increase in MMP-2, MMP-9 and MT-MMP-1 gene expression and accumulation of catalytically active MMP-2, MMP-9 and MMP-9 homodimer in conditioned media (Figure 2A and 2C). In contrast, treatment of PC3-ML cells with TGF-β alone was sufficient to promote the enzymatic activity of MMP-2, MMP-9 and the MMP-9 homodimer in conditioned media, and EGF had no additive effect when combined with TGF-β (Figure 2C).

To functionally demonstrate the invasive capacity of cells undergoing EMT, we tested the affect of EGF, TGF-β and E + T on IBC-10a cells’ ability to migrate through a Matrigel-coated modified Boyden chamber. Although minimal media (Km), EGF and TGF-β alone induced little to no invasion, IBC-10a cells treated with E + T exhibited significant increases in cell invasion and migration (Figure 2D). Furthermore, using a three-dimensional Matrigel model that recapitulates in vivo glandular organization (41), we observed that IBC-10a cells formed tight acinar-like structures (termed prostaspheres) in the presence of Km, EGF or TGF-β alone; however, in the presence of E + T, prostaspheres were disrupted, and treatment promoted cell to emigration from the acini and their invasion through the surrounding Matrigel (Figure 2E). Notably, the invading IBC-10a cells were spindle-shaped and expressed Vimentin, suggestive of EMT (Figure 2E).

Ras activation of Raf promotes TGF-β-induced EMT. Ras is a major effector molecule of EGF signaling and has previously been implicated in promoting TGF-β-mediated EMT (34). To determine the role of Ras in modulating TGF-β responses in IBC-10a and PCa-20a cells, we stably transfected these cells with either a constitutively active Ras construct (pBABE:RasV12) or empty vector control (pBABE) and treated with minimal media (Km), EGF, TGF-β or E + T. In response to TGF-β or E + T treatments, Ras-transfected cells showed a reduction in both cell–cell junctions and E-cadherin expression, along with concomitant upregulation of Vimentin (Figure 3A, and Supplementary Figure 3A, available at Carcinogenesis Online). Activated Ras is known to mediate its signaling through several downstream pathways; we, therefore, transfected IBC-10a and PCa-30a cells with specific Ras effector mutants including RasV12-C40, which binds PI3-kinase to activate AKT signaling; RasV12-G37, which binds RalGDS to activate phospholipase D signaling; and RasV12-S35, which binds c-Raf to activate MAPK signaling (45). Although all cells increased expression of Vimentin and FSP-1 in response to treatment with E + T, only cells transfected with RasV12-S35 also did so in the presence of TGF-β alone (Figure 3B and 3C, and Supplementary Supplementary Figure 3B and E, available at Carcinogenesis Online). In response to TGF-β treatment, RasV12-S35-transfected cells also expressed increased activity of MMP-2, MMP-9 and the MMP-9 homodimer and demonstrated enhanced cell motility and invasion exhibiting a >3-fold increase in migration and invasion in modified Boyden chamber assays when compared with controls (Figure 3D and 3E). Moreover, TGF-β treatment of IBC-10a or PC-20a cells transfected with either RasV12 or RasV12-S35 significantly increased expression of Vimentin, Slug, Twist2, MMP-2 and MMP-9 mRNA (Figure 3F, and Supplementary Figure 3F and G, available at Carcinogenesis Online). In contrast, IBC-10a and PCa-20a cells transfected with empty vector, RasV12-C40 or RasV12-G37 failed to elicit any increase in expression of these genes in response to TGF-β (Figure 3F, and Supplementary Figure 3G, available at Carcinogenesis Online). Taken together, these results indicate that EGF signaling through the Ras-Raf-MAPK cascade potentiates TGF-β induction of EMT in non-invasive prostate epithelial cells.

Fig. 3.

Ras activation of Raf induces TGF-β’s pro-invasive response. (A) Western blot analysis for E-cadherin, Vimentin and Tubulin (loading control) expression in IBC-10a (left) and PCa-20a (right) cells stably transfected with pBABE:RasV12 and treated with minimal media (Km), EGF (10ng/ml), TGF-β (10ng/ml) or EGF + TGF-β in combination (E + T). (B) Western blot analysis for E-cadherin, Vimentin and Tubulin (loading control) expression in IBC-10a and PCa-20a cells stably transfected with pBABE:RasV12-C40, pBABE:RasV12-G37 and pBABE:RasV12-S35 vectors. Cells were treated with Km, EGF, TGF-β or E + T. (C) Western blot analyses of FSP-1 and Tubulin (loading control) expression in IBC-10a cells stably transfected with empty vector (Empty), pBABE:RasV12-C40, pBABE:RasV12-G37 or pBABE:RasV12-S35 and treated with TGF-β. (D) Quantification of modified Boyden chamber invasion assays showing the invasive activity of IBC-10a cells stably transfected with empty vector (Empty), pBABE:RasV12-C40, pBABE:RasV12-G37 and pBABE:RasV12-S35 variants and treated with TGF-β. Values are representative of three independent replicates. Mean ± SD, **P < 0.01 compared with empty vector control. (E) Gelatin zymographs of conditioned media 24h after last treatment from pBABE:RasV12-C40, pBABE:RasV12-G37 and pBABE:RasV12-S35 and pBABE:Empty variants of IBC-10a cells treated with TGF-β. (F) Relative expression of Slug, Twist-2, MMP-2 and MMP-9 mRNA in IBC-10a cells transfected with pBABE:RasV12-C40 (light gray bars); pBABE:RasV12-G37 (dark gray bars); pBABE:RasV12-S35 (black bars); and pBABE.Empty (white bars) variants treated with TGF-β by qRT–PCR. Values normalized against cyclophilin-A and represent the fold increase in expression compared with pBABE:Empty cells. Mean ± 1 SD from three independent experiments. Statistical significance was derived using Welsh’s unpaired t-test, *P < 0.05, **P < 0.01; ***P < 0.001 when comparing RasV12 and RasV12-S35 cells with pBABE:Empty cells.

MEK1, but not MEK2, activity is necessary and sufficient for TGF-β-induced EMT. MEK1/2 activation of Erk1/2 is the most well-characterized downstream effect of Ras/Raf signaling and is critical for Ras-induced transformation (46). To better understand the signaling dynamics regulating EMT, IBC-10a cells were treated with increasing concentrations of either a MEK 1/2 inhibitor (PD098059), a PI3K inhibitor (LY290042) or a SMAD3 inhibitor (SIS3). As indicated by Vimentin and FSP-1 expression, we observed that the EMT response was dramatically inhibited in a dose-dependent manner by both PD098059 and SIS3 in IBC-10a cells (Figure 4A, and Supplementary Figure 4A, available at Carcinogenesis Online) suggesting that signaling through MAPK and Smad3 is both necessary for E + T-induced EMT.

Fig. 4.

MEK1, but not MEK2, activity is necessary and sufficient for TGF-β-induced EMT. (A) Western blot analysis of phosphorylated-Erk 1/2, Vimentin, FSP-1 and Tubulin (loading control) expression in IBC-10a cells cultured in either Km with 0.5% DMSO (Km + DMSO), Km with 10ng/ml of EGF + TGF-β with 0.5% DMSO (E + T + DMSO), or Km with 10ng/ml of EGF + TGF-β with increasing concentrations of the MEK1/2 inhibitor PD09859 (+PD098). (B) Western blot analysis of E-cadherin, Vimentin and Tubulin (loading control) expression in IBC-10a cells stably transfected with constitutively active MEK1 (pBABE:MEK1-DD), MEK2 (pBABE:MEK2-DD) and empty vector (pBABE:Empty) and treated with TGF-β (10ng/ml). (C) IBC-10a cells from (B) were cultured in minimal media and the expression of phosphorylated-Erk1/2, total-Erk 1/2 and Tubulin (loading control) examined by western blot. (D) Western blot analysis of phosphorylated-Erk1/2, total-Erk1/2, Ras and Tubulin (loading control) expression in IBC-10a, PCa-20a, PCa-30a and PC3-ML cells following treatment with TGF-β. (E) Western blot analysis of phosphorylated-Erk1/2, total-Erk1/2 and Tubulin (loading control) expression in IBC-10a cells treated with minimal media (Km), EGF (10ng/ml), TGF-β (10ng/ml) or EGF + TGF-β (10ng/ml) (E + T). (F) Western blot analysis of E-cadherin, Vimentin, FSP-1, total Erk1/2 and Tubulin (loading control) expression in PCa-20a and PCa-30a cells stably transfected with a scrambled shRNA construct or a shRNA construct targeting Erk2 after treatment with TGF-β, EGF or E + T.

We stably transfected IBC-10a cells with a constitutively active MEK1 or MEK2 construct (pBABE: MEK1-DD, pBABE:MEK2-DD) and empty vector (pBABE) as a control. MEK1-DD- and MEK2-DD-transfected IBC-10a overexpressed MEK-1 and MEK-2, respectively, with no change in expression to the other MEK protein (Supplementary Figure 4B, available at Carcinogenesis Online). In response to TGF-β, MEK1-DD-transfected cells demonstrated a decrease in E-cadherin expression and induction of Vimentin (Figure 4B). In contrast, MEK2-DD-transfected cells showed a partial reduction in E-cadherin expression but showed no induction of Vimentin (Figure 4B). Immunofluorescence imaging further demonstrated that Vimentin expression was ubiquitously induced by TGF-β in MEK1-DD but not in MEK2-DD-transfected IBC-10a cells (Supplementary Figure 4C, available at Carcinogenesis Online). MEK1-DD- and MEK2-DD-transfected cells also both significantly increased phosphorylation of Erk 1/2 compared with the empty vector cells (Figure 4C). We also observed that phosphorylation of Erk1/2 was elevated in IBC-10a, PCa-20a and PCa-30a cells when treated with TGF-β alone, and levels of activated Erk 1/2 were equal in IBC-10a cells treated with either EGF, TGF-β or E + T (Figure 4D and 4E). Surprisingly, metastatic PC3-ML cells exhibited decreased levels of Erk1/2 phosphorylation when compared with IBC-10a, PCa-20a and PCa-30a cells despite expressing significantly more Ras (Figure 4D). Functional Erk2, but not Erk1, is previously shown to be essential for EMT, and given the conflicting results above, we wanted to determine if Erk2 expression was required for EMT in our model (47). We transfected PCa-20a and PCa-30a cells with a scrambled shRNA or shRNA vector targeting Erk2 and observed that treatment with E + T or TGF-β in PCa-20a and PCa-30a cells with Erk2 knockdown failed to induce Vimetin and FSP-1 or downregulate E-cadherin (Figure 4F). Taken together, these findings suggest that although MEK1 signaling specifically regulates EMT and Erk2 expression is required for EMT, differential levels of Erk2 phosphorylation are not regulating EMT.

Erk2 nuclear accumulation promotes and c-myc expression is required for TGF-β-induced EMT.MEK1 and MEK2 are often considered to be redundant in function, although MEK1 and MEK2 are shown to have differential effects on cellular localization of Erk2 (48). Consistent with this observation, Erk2 accumulated in the nucleus of MEK1-transfected IBC-10a cells but not in MEK2- or empty vector-transfected IBC-10a cells cultured in minimal media (Km) (Figure 5A). Additionally, we observed by immunofluorescence that TGF-β alone was insufficient to induce nuclear accumulation of Erk2 in PCa-20a cells, whereas E + T induced a dramatic increase in Erk2 nuclear staining (Figure 5B). Significantly, both TGF-β and E + T treatments induced sustained Erk2 accumulation in the nucleus of PCa-30a cells that undergo EMT with TGF-β treatment alone (Figure 5B). These observations were confirmed by western blot of PCa-20a and PCa-30a nuclear fractionations for Erk2 in cells treated with minimal media, EGF, TGF-β, and EGF and TGF-β in combination (Figure 5C, and Supplementary Figure 4D, available at Carcinogenesis Online). To further investigate the role of Erk2 nuclear accumulation, PCa-20a cells were transfected with a phosphatase-resistant Erk2 mutant (D319N) that accumulates in the nucleus of cells and WT Erk2 as a control (Supplementary Figure 4E, available at Carcinogenesis Online) (49). TGF-β treatment alone was sufficient to induce Vimentin and FSP-1 expression and promote EMT in cells transfected with mutant Erk2 but not WT Erk2 (Figure 5D). It is well established that nuclear Erk2 induces c-myc phosphorylation as a functional consequence of Erk2 nuclear accumulation, and we also observed an increase in phosphorylation of c-myc at serine 62 (Figure 5C) (50,51). Moreover, transfection with MEK1 induced c-myc phosphorylation, whereas knockdown of Erk2 decreased c-myc phosphorylation in response to E + T treatments in PCa-20a cells and treatment of TGF-β alone in PCa-30a cells further indicating that Erk2 nuclear accumulation is phosphorylating c-myc during EMT(Supplementary Figure 4F and G, available at Carcinogenesis Online). These observations prompted us to discern the role of c-myc in promoting TGF-β-induced EMT. We transfected IBC-10a cells with a c-myc overexpression construct and a c-myc-targeting shRNA and treated them with TGF-β and E + T. We observed that c-myc overexpression was insufficient to promote TGF-β-induced EMT, however, c-myc expression was required for induction of EMT in both IBC-10a and PCa-20a cells in response to E + T (Figure 5E and 5F). Knockdown of c-myc also significantly inhibited the invasive potential of IBC-10a cells in response to E + T (Figure 5G). Furthermore, knockdown of c-myc or Erk2 in PC3-ML cells decreased expression of Vimentin and FSP-1 (Figure 5H).

Fig. 5.

Erk2 nuclear localization and c-myc expression promote TGF-β-induced EMT. (A) Immunofluorescent imaging of Erk2 localization (red) in IBC-10a cells transfected with pBABE:Empty, pBABE:MEK1-DD or pBABE:MEK2-DD. Cells were grown in minimal media and probed for Erk2 localization (×400 magnification). (B) Immunofluorescent images of PCa-20a and PCa-30a cells treated with TGF-β (10ng/ml) (TGF) or EGF + TGF-β (10ng/ml each) (E + T) and probed for Erk2 localization (×400 magnification). (C) Western blot analysis of Erk2 expression in nuclear fractions of PCa-20a and PCa-30a cells treated with minimal media (Km), EGF (10ng/ml), TGF-β (10ng/ml) or EGF + TGF-β (10ng/ml) in combination (E + T). Lamin A/C was used as a loading control. (D) Western blot analysis of E-cadherin, Vimentin, serine 62 phosphorylated c-myc, FSP-1 and Tubulin (loading control) expression in PCa-20a cells stably transfected with Erk2 WT construct (pLNCX:HA-Erk2), a phosphatase resistant mutant Erk2 (pLNCX:HA-Erk2 D319N) or empty vector (pBABE:Empty) as control and treated with TGF-β. (E) Western blot analysis of E-cadherin, Vimentin, c-myc, FSP-1 and Tubulin (loading control) expression in IBC-10a cells stably transfected with an empty vector, c-myc overexpression, scrambled shRNA or a shRNA targeting c-myc construct. Cells were treated with TGF-β alone or concomitant EGF + TGF-β (E + T). (F) Western blot analysis of E-cadherin, Vimentin and Tubulin (loading control) expression in PCa-20a cells stably transfected with a scrambled shRNA or a shRNA targeting c-myc construct and treated with concomitant EGF + TGF-β (E + T). (G) Quantification of modified Boyden chamber invasion assays showing the invasive activity of IBC-10a cells stably transfected with a scrambled shRNA or a shRNA targeting c-myc and treated with E + T. Values are representative of three independent replicates. Mean ± SD, **P < 0.01 compared with empty vector control. (H) Western blot analysis of Vimentin, c-myc, total-Erk 1/2, FSP-1 and Tubulin (loading control) expression in PC3-ML cells stably transfected with a scrambled shRNA, a shRNA targeting Erk2 or a shRNA targeting c-myc construct.

To test the enhanced metastatic potential associated with EMT, PC3-ML cells containing either Erk2 or c-myc shRNA constructs were injected intercardiacally into male NOD-SCID mice. Previous studies have demonstrated that PC3-ML cells readily metastasize in mice to distant organ sites by 4 weeks post-injection (52). We found that at 5 weeks post-injection, 2/3 of mice injected with PC3-ML cells carrying a control-scrambled shRNA construct exhibited liver and adrenal metastasis, and 1/3 of these mice exhibited a brain metastasis (n = 3) (Supplementary Figure 5A and B, available at Carcinogenesis Online, and Table I). In contrast, shRNA-mediated knockdown of c-myc failed to produce distant metastasis in mice (n = 6), and shRNA-mediated knockdown Erk2 produced only one distant metastasis (n = 6) (Supplementary Figure 5A and B, available at Carcinogenesis Online, and Table I). Knockdown of c-myc and Erk2 also inhibited the invasive phenotype typically observed in PC3-ML cells (Supplementary Figure 5C , available at Carcinogenesis Online). Taken together, these results suggest that nuclear accumulation of Erk2, which is stimulated by MEK1, but not MEK2, is a key regulator of TGF-β-induced EMT and invasion. Moreover, these results indicate that c-myc expression, a target of activated Erk2 in the nucleus, is required for EMT and that inhibition of this pathway results in an overall decreased metastatic potential of highly invasive prostate cancer cells.

Table I.

Summary of PC3-ML tumor development in NOD-SCID

Cell line	Thoracic cavity	Tumor mass (mg)	Abdominal cavity metastasis	Adrenal metastasis	Liver metastasis	Brain metastasis
PC3-ML sh Scram	3/3	16.5±11	2/3	2/3	2/3	1/3
PC3-MLsh Erk2	6/6	7.2±3.0	1/6	1/6	1/6	0/6
PC3-MLsh c-myc	5/6	6.3±3.1	0/6	0/6	0/6	0/6

Male NOD-SCID mice were injected intercardiacally with 5 × 10⁴ cells of PC3-ML cells expressing GFP and transfected with a scrambled shRNA vector (n = 3), an Erk2-targeted shRNA construct (n = 6) or a c-myc-targeted shRNA construct (n = 6). Tumor development in thoracic cavity, abdominal cavity, adrenal gland, liver and brain was assessed 5 weeks after injection. Totals represent number of tumors identified at given location per total number of mice injected. Tumors from the thoracic cavity were harvested, weighed and averaged showing standard deviation (n = 9, 12 and 12 for tumors harvested from sh Scram, sh Erk2 and sh c-myc, respectively).

Discussion

To our knowledge, this is the first report to show that downstream of EGF, Ras and Raf signaling, active MEK1, but not MEK2, is necessary and sufficient for TGF-β-induced EMT in a variety of normally non-invasive primary cells. These findings imply that activation of MEK1 and MEK2 has differential effects on TGF-β signaling and that their role in growth factor signaling is not interchangeable. Although MEK1 and MEK2 share extensive homology, it is shown that MEK1-activated Erk2 preferentially accumulates in the nucleus (48). In agreement with a previous report (53), our findings indicate that overexpression of a mutant of Erk2 that accumulates in the nucleus, given its resistance to MAPK phosphatases, is sufficient for TGF-β alone to induce an EMT phenotype. These data strongly indicate that EGF signaling plays an important role in modulating TGF-β responses in prostate epithelial cells by inducing differential Erk2 shuttling to the nucleus, which is critical for EMT. These data also suggest that there may be a role for MAPK phosphatases, which reside in the nucleus, in regulating EMT and TGF-β responses.

MEK1-induced Erk2 nuclear accumulation is in part accomplished through a proline-rich domain in MEK1 that is absent in MEK2, which interacts with proteins associated with adhesion structure signaling, such as PAK1, which phosphorylates MEK1 at serine 298 in response to cellular adhesion to fibronectin (48,54). Interestingly, previous studies have shown that functionally blocking the association between fibronectin and its receptor inhibits EMT induction (55). Interestingly, EMT in our model was achieved after 9 days of treatment with growth factors; therefore, it is possible that EMT induction requires such a time frame to allow for sufficient deposition of extracellular matrix proteins for cells to interact with to promote MEK1-induced Erk2 accumulation. This hypothesis is partially supported by the observation that although the EMT phenotype was stable after withdrawal of EMT-inducing growth factors (Figure 1C), trypsinization and replating of cells resulted in reversion to an epithelial phenotype (data not shown).

One of the functions of nuclear Erk2 is phosphorylation and stabilization of the transcription factor c-myc (50). Although in vivo breast cancer modeling suggests that overexpression of c-myc can elicit an EMT phenotype and that overexpression of c-myc alone can induce EMT in mammary epithelial cells, there is a lack of studies directly indicating whether c-myc expression is required for EMT in regard to TGF-β-induced invasion (56–58). In this report, we demonstrate that expression of c-myc is critical for the EMT program and for TGF-β-induced invasion. Interestingly, in normal epithelia, TGF-β acts as a tumor suppressor in part by repressing c-myc; therefore, it is conceivable that inhibition of c-myc downregulation by TGF-β through the Ras-MAPK pathway is critical for the tumor-promoting activities of TGF-β (59). Furthermore, our findings suggest that overexpression of c-myc is not sufficient for EMT, suggesting that post-translational phosphorylation of c-myc may have a larger functional role in tumor progression than simply stabilization of the c-myc protein. This finding is in agreement with a recent report that in mammary epithelial cells, expressing a mutant myc protein possessing elevated levels of phosphorylated serine 62 results in invasive mammary carcinoma (60). In addition, c-myc is a driver of the pluripotent phenotype, regulating stem cell self-renewal and differentiation and is shown to be required for growth of tumor-initiating prostate cancer cells (35,61,62). Interestingly, EMT in human mammary epithelial cells also includes induction of classical stem cell markers, and cells undergoing EMT exhibit some level of cellular plasticity (63,64). Therefore, c-myc activity might play a critical role in regulating EMT, the cellular plasticity associated with EMT and the tumor-initiating characteristics of cells undergoing EMT.

Reportedly, Ras and Raf mutations, and/or amplification, are a rare event during the prostate and breast cancer progression and has led pathological studies to doubt the clinical contribution of Ras alone to cancer metastasis and EMT (65,66). However, alternative molecular processes may transiently upregulate Ras and Raf activity, including increased expression of Ras GEFs and reduced expression of Ras GAPs. For example, enhancer of zeste homolog 2, a member of the Polycomb Repressive Complex 2, is shown to silence disabled homolog 2-interacting protein, a Ras GAP, thereby inducing hyperactive Ras and promoting increased prostate cancer metastasis (67). Since enhancer of zeste homolog 2 expression is greatly increased in metastatic prostate cancer cells compared with localized prostate cancers, it is possible that a transient upregulation of Ras activity may contribute to EMT invasion and metastatic progression of human prostate cancer (68).

Non-canonical MAPK activation by TGF-β is known to be an important mechanism for Smad signaling by phosphorylating various transcription factors in the nucleus of cells that physically interact with Smads and regulate TGF-β responses (21,26,27). While MAPK activation by TGF-β appears to be required for TGF-β-mediated EMT, it is also apparent that constitutive activation of Ras along with TGF-β can act cooperatively to promote EMT when TGF-β alone cannot (69,70). Our findings suggest that the ability for EGF and MEK1 to differentially direct Erk2 cellular localization may serve as a functional mechanism for the synergistic signaling between Ras and TGF-β to induce EMT. From our findings, we propose a model by which Erk2 must be activated and shuttled to the nucleus where it can phosphorylate c-myc and, in cooperation with TGF-β signaling, induce EMT (Supplementary Figure 6, available at Carcinogenesis Online). Therefore, in circumstances where TGF-β alone cannot induce EMT, Erk2 may not have sufficiently accumulated in the nucleus, or c-myc may not be adequately expressed. In this case, auxiliary pathways, such as EGF activation of Ras, may be required for TGF-β-mediated EMT. In agreement with this hypothesis, other studies have shown that sustained MAPK signaling directed by Ras, Raf, EGF or Erb2 overexpression is often necessary to promote robust and sustainable EMT in response to TGF-β treatment (32–34).

Recent studies have suggested that EMT and metastatic dissemination may be an early event in tumorigenesis (71,72). Our results support this concept and suggest that early-stage prostate cancer cells possess the genetic repertoire necessary for EMT and invasion. In early-stage tumors, it is feasible that increased TGF-β and EGF levels may arise from chronic inflammation or the reactive stroma associated with early tumors to induce EMT and invasion (73,74). Future studies examining the nuclear localization of Erk2 in cancer cells at the leading edges of tumors may aid identification of early-stage cancers that are poised to metastasize and identify patients with poorer prognosis and who may require more aggressive therapeutic intervention.

Supplementary material

Supplementary Figures 1–5 can be found at http://carcin.oxfordjournals.org/

Funding

National Institutes of Health National Cancer Institute (grant CA-577083 to M.E.S.) and DOD-PCRP Pre-Doctoral Fellowship (W81XWH-10-1-1044 to M.D.A.).

Glossary

Abbreviations:

EGF

epidermal growth factor

EMT

epithelial–mesenchymal transition

FSP-1

Fibroblast Specific Protein-1

GS

Gleason score

Km

keratinocyte media

MAPK

mitogen-activated protein kinase

MMP

matrix metalloproteinase

TGF-β

transforming growth factor-beta