Skip to main content
NCBI home page
Proceedings of the National Academy of Sciences of the United States of America logoLink to Proceedings of the National Academy of Sciences of the United States of America
. 2001 Jun 5;98(12):6686–6691. doi: 10.1073/pnas.111614398

Genetic programs of epithelial cell plasticity directed by transforming growth factor-β

Jiri Zavadil *, Markus Bitzer *, Dan Liang *, Yaw-Ching Yang *, Aldo Massimi †, Susanne Kneitz †, Ester Piek ‡, Erwin P Böttinger *,†,§
PMCID: PMC34413  PMID: 11390996

Abstract

Epithelial–mesenchymal transitions (EMTs) are an essential manifestation of epithelial cell plasticity during morphogenesis, wound healing, and tumor progression. Transforming growth factor-β (TGF-β) modulates epithelial plasticity in these physiological contexts by inducing EMT. Here we report a transcriptome screen of genetic programs of TGF-β-induced EMT in human keratinocytes and propose functional roles for extracellular response kinase (ERK) mitogen-activated protein kinase signaling in cell motility and disruption of adherens junctions. We used DNA arrays of 16,580 human cDNAs to identify 728 known genes regulated by TGF-β within 4 hours after treatment. TGF-β-stimulated ERK signaling mediated regulation of 80 target genes not previously associated with this pathway. This subset is enriched for genes with defined roles in cell–matrix interactions, cell motility, and endocytosis. ERK-independent genetic programs underlying the onset of EMT involve key pathways and regulators of epithelial dedifferentiation, undifferentiated transitional and mesenchymal progenitor phenotypes, and mediators of cytoskeletal reorganization. The gene expression profiling approach delineates complex context-dependent signaling pathways and transcriptional events that determine epithelial cell plasticity controlled by TGF-β. Investigation of the identified pathways and genes will advance the understanding of molecular mechanisms that underlie tumor invasiveness and metastasis.

Transforming growth factor-β (TGF-β) elicits distinct cellular response patterns of epithelial cells in a highly context-dependent manner (1). Although TGF-β family members play pivotal roles at the interphase of epithelial and mesenchymal cell fates during embryonic development and organogenesis (2–4), they are essential for maintenance of epithelial homeostasis in several mouse models, presumably by exerting potent antiproliferative effects and tumor suppressor activities (5–7). TGF-β may inhibit cell cycle progression through regulation of cyclin-dependent kinase inhibitors p15INK4B, p21CIP1, and p27KIP1 (8, 9) and induces apoptosis via activation of caspases (10).

Most carcinomas are characterized by loss of normal growth-inhibitory and apoptotic responses to TGF-β (11). In a subset of colon and pancreatic tumors, this is caused by inactivating defects of TGF-β receptors and Smad4/DPC4, but the majority of tumors with loss of growth inhibition do not acquire genetic defects in TGF-β signaling pathways. Hence, epigenetic mechanisms of inactivation of TGF-β's growth inhibitory and proapoptotic activities have been proposed (for review, see ref. 12), leaving the possibility that interference with TGF-β signaling mechanisms in these cases blocks only selected responses. For example, partial inhibition of TGF-β type II receptor function results in loss of growth inhibition but not regulation of matrix genes by TGF-β (13).

In contrast with epithelial homeostasis, TGF-β signals appear to promote tumor progression in advanced carcinomas. In this context, TGF-β is often synthesized in excess and has been associated with epithelial–mesenchymal transitions (EMT) of tumor cells in skin and mammary carcinogenesis models (14, 15). Most invasive and/or metastatic carcinomas are characterized by EMT, in which the epithelial phenotype, exhibiting strong cell–cell junctions and polarity across the epithelial layer, is replaced by a mesenchymal phenotype with reduced cell–cell interactions and increased motility (16). Activation of TGF-β signaling is sufficient to induce EMT in cultured epithelial cells (17–19). Although both phosphatidylinositol-3 kinase and extracellular response kinase (ERK) mitogen-activated protein (MAP) kinase signaling pathways have been implicated in TGF-β-induced EMT (20, 21), their specific roles and proximal target gene programs in EMT remain unclear. Similarly, persistent activation of Ras and Raf proteins that activate ERK is sufficient to promote EMT and contributes to invasiveness and metastasis formation in carcinomas with activated ras oncogenes (22). However, the underlying genetic programs have not been identified.

Although Smad proteins are considered important mediators (12, 23, 24), Smad-independent TGF-β signaling has been demonstrated, involving predominantly MAP kinase pathways. MAP kinase pathways have been implicated in mediating induction of fibronectin, p21Cip1 and TGF-β1 gene transcription by TGF-β (25–27). However, physiological roles and the identities of transcriptional targets of MAP kinase activation by TGF-β have not been determined.

To explore molecular mechanisms underlying the roles of TGF-β as a tumor promoter and to define the functional relevance and transcriptional targets of ERK MAP kinase in the context of human malignancies, we established a model of EMT induced by TGF-β in human keratinocytes (HaCaT). Using a gene expression profiling approach, we identified discrete gene expression patterns that are directed by TGF-β at the onset of EMT. We show that engagement of ERK MAP kinase signaling by TGF-β in this context regulates predominantly genes with functions in cell–matrix adhesion and endocytosis, consistent with a requirement of ERK signaling in cell motility and disassembly of adherens junctions in TGF-β-induced EMT.

Methods

Cell Culture.

HaCaT keratinocytes (28) were expanded to ≈90% confluency and serum deprived by culturing in 0.2% FBS for 48 h before timed stimulation with 10 pM rhTGF-β1 (R & D Systems) in the absence or presence of either 15 μM MAP kinase/ERK kinase (MEK)/ERK inhibitor U0126 (Promega) or 10 μM cycloheximide (Sigma). Experiments were performed in three independent repeats.

Immunofluorescence and Immunoblotting.

Cells were fixed and permeabilized for immunofluorescence staining as described (29). Antigen detection: F-actin (FITC-phalloidin; Sigma) and E-cadherin (mouse anti-E-cadherin; Zymed) followed by Cy3-conjugated donkey anti-mouse IgG antibody (Jackson ImmunoResearch). Cell nuclei were counterstained with 4′,6-diamidino-2-phenylindole (Sigma). Fluorescent images were generated by using a cooled charge-coupled device camera (AECOM Analytical Imaging Facility, Albert Einstein College of Medicine, Bronx, NY). For immunoblotting, we used polyclonal antibody anti-pYTEpY of ERK1/2 (Anti-ACTIVE MAP kinase pAb; Promega) and total anti-ERK1/2 antibody (Anti-ERK1/2 pAb, Promega).

Motility Assay.

Confluent serum-deprived monolayer cultures of HaCaT cells were “wounded” by a scratch in the presence or absence of TGF-β and/or MEK/ERK inhibitor U0126 and monitored by phase contrast microscope photography as described (30). Motility of the cells was quantitated by 24 measurements in each experimental category and expressed as a percentage of distance coverage by cells migrating into the scratch wound area 48 h after wounding.

Microarray Procedures.

Unique human expressed sequence tags (ESTs) (16,580), available from Research Genetics (Huntsville, AL), were used as cDNA probes, spotted on two separate glass slides: 5H arrays contained 7,873 probes (non-sequence verified), and H1 arrays contained 8,707 sequence-verified probes. Detailed descriptions of microarray hardware and procedures are available from http://www.aecom.yu.edu/home/molgen/facilities.html. For each hybridization, cDNA targets were prepared from the RNA sample (Cy5-labeled) obtained from TGF-β stimulated cells and the reference RNA sample (Cy3-labeled) representing a common time point zero (T0), to provide baseline expression measurements. Common baseline reference cDNA target allows comparisons of the relative expression of each gene across the entire series of timed TGF-β stimulations (0, 0.3, 1, 2, and 4 h, respectively). Microarray analysis of time series experiments was performed in three independent repeats. Independent images were obtained for Cy3 and Cy5 fluorescence emitted from hybridized microarrays by using a custom-built dual channel laser scanning microscope (see http://www.aecom.yu.edu/home/molgen/facilities.html for specifications). scanalyze Verison 2.44 software (M. Eisen, Stanford University, Palo Alto, CA; http://www.microarrays.org/software.html) was used as described in the scanalyze manual to generate raw data files containing measurements of signal and background fluorescence emissions of Cy3 and Cy5, respectively, for each element.

Quality Control, Data Analysis, and Statistics.

Signal and background intensities of each spotted cDNA element in Cy5 images were calibrated by a correction factor defined as e(mean lnCy3)/e(mean lnCy5), typically between 0.9 and 1.1, to normalize for effects of global variations between Cy3 and Cy5 fluorescence intensities in each hybridized array. Net signal intensity (NSI) in each channel (Cy3 and Cy5) was determined by subtracting the local background from signal intensity values. Negative NSI values were set to the smallest intensity unit equal to 1. A Cy5/Cy3 ratio represents the relative abundance of a target transcript in stimulated and reference samples, respectively. The median value of Cy5/C3 ratios of three experimental repeats was taken as a representative value. Signal-to-noise (S/N) threshold criteria for genes to pass as “expressed” were set independently for both Cy3 and Cy5 channels. Elements with S/N >1 + ρ (ρ = one standard deviation of the mean of background values in each hybridized array divided by the mean of background values; typically 0.15–0.30) in either Cy3, Cy5, or both channels, respectively, were considered well measured and expressed. Elements with S/N <1 + ρ in both channels were flagged (marked as nonexpressed). The value ρ was determined for each hybridized array to account for variations in global signal and background intensities between individual experiments.

Time course gene expression profiles were required to fulfill “expressed” S/N criteria in (i) at least two of three repeats at each time point, and (ii) at least three of the five time points. Genes were considered regulated by TGF-β stimulation if their normalized Cy5/Cy3 ratios deviated more than 1.75-fold from baseline (T0) Cy5/Cy3 ratios. These thresholds reflected the three SD intervals around the mean Cy5/Cy3 of expressed transcripts at T0 (1 SD = 0.2467 used as measure of random variability of ratios at T0). Hierarchical clustering was performed using Pearson correlation metric on Cy5/Cy3 ratios for each transcript with the tree program of GENESPRING Version 3.2.4 (Silicon Genetics, Redwood City, CA). Primary data tables can be obtained at http://www.aecom.yu.edu/BottingerLab/microarray.htm.

Quantitative Real-Time PCR (rtPCR).

Quantitative rtPCR analysis of TGF-β controlled gene expression of SLUG, BNC, SMAD7, SOX9, CTSD, and c-FOS was performed by using the iCycler apparatus (Bio-Rad) with sequence-specific primer pairs for all genes tested. The SYBR Green PCR Core Reagents system (Perkin–Elmer Applied Biosystems) was used for real-time monitoring of amplification. Results were evaluated by icycler iq real time detection system software (Bio-Rad).

Results and Discussion

ERK MAP Kinase Is Required for Disassembly of Cell Adherens Junctions and Induction of Cell Motility by TGF-β.

Quiescent serum-deprived human HaCaT keratinocytes were treated with TGF-β and manifested phenotypic features of EMT. Within 48 h, cortical actin filaments were replaced by actin stress fibers, intercellular junctions were disassembled, and cell motility increased, resulting in complete cell–cell separation and fibroblast-like cell morphology (Fig. 1a). As functional roles for MEK/ERK in TGF-β signaling and in EMT have been proposed (17, 22, 31), we treated cells with TGF-β in the presence of MEK/ERK inhibitor U0126. Disassembly of tight junctions and desmosomal junctions was demonstrated by delocalization of ZO-1 and desmoplakin (data not shown), and de novo formation of actin stress fibers was observed (Fig. 1b). In contrast, E-cadherin-mediated adherens junctions did not dissolve, and cells failed to separate and did not acquire a fibroblastoid morphology (Fig. 1b). TGF-β treatment induced cell motility (assessed by scratch wound assay) when compared with untreated cells (46% wound distance coverage). This effect was inhibited in the presence of U0126 (16% wound distance coverage) (Fig. 1c). In contrast, potent antimitogenic effects of TGF-β that are typically observed in subconfluent proliferating HaCaT cells were not inhibited by U0126 (Fig. 1d). To examine activation profiles (phosphorylation) of ERK1/2, c-Jun N-terminal kinase (JNK), and p38 MAP kinases by TGF-β, we used phospho-MAP kinase-specific antibodies. Immunoblotting for phospho-ERK1/2 revealed transient activation of ERK2 at 1 and 2 h after addition of TGF-β to quiescent cells. The activation was completely inhibited by U0126 and by cycloheximide, an inhibitor of protein synthesis (Fig. 1e). TGF-β had no effect on phospho-JNKs or phospho-p38 levels under identical conditions (data not shown). The kinetic and functional profiles of ERK activation by TGF-β were distinct from those observed after treatment with epidermal growth factor (EGF). EGF activated ERK within minutes (Fig. 1e) and increased DNA synthesis as assessed by [3H]thymidine incorporation, but had no detectable effect on epithelial cell morphology even after 48 h (data not shown). Our findings suggest that TGF-β indirectly stimulates a distinct activation profile of MEK/ERK that exerts a specific functional role in disassembly of E-cadherin-mediated adherens junctions and induction of cell motility in the context of the described EMT model.

Figure 1.

Figure 1

Open in a new tab

Activation of MEK/ERK signaling is required for disassembly of adherens junctions and cell–cell separation during EMT induced by TGF-β1. (a) Quiescent HaCaT cells treated with TGF-β1 and labeled against F-actin and E-cadherin, and (b) addition of MEK/ERK inhibitor U0126 [arrows: cell–cell junctions; arrowheads: actin stress fibers; cell nuclei stained by 4′,6-diamidino-2-phenylindole (DAP1)]. (c) Inhibition of MEK/ERK pathway by U0126 impedes TGF-β-induced motility of cells as percentage of migration distance (measured across scratch wound width) covered by cells 48 h after induction. (d) MEK/ERK activity is not required for growth-inhibitory effects of TGF-β on proliferating HaCaT cultures as assayed by [3H]thymidine incorporation. (e) Immunoblots show Thr185/Tyr187-phosphorylated Erk2 (P-Erk2) and total Erk1/2 proteins (Erk1/2) in quiescent HaCaT treated, as indicated.

Dynamic Changes in Expression of up to 4,000 Genes May Be Associated with Initiation of EMT by TGF-β in HaCaT.

Gene expression profiling was used to comprehensively delineate early genetic programs associated with EMT in response to TGF-β. Fluorescent cDNAs derived from quiescent TGF-β-treated cells (Cy5) and from a common reference RNA sample T0 (Cy3) were cohybridized to cDNA microarrays produced at our institution. These arrays represent 16,580 unique human ESTs. Relative transcript abundance was expressed as Cy5/Cy3 ratios of signal intensities after background subtraction in each channel. Data analysis and quality control procedures are described in detail in Methods.

We identified 1,716 TGF-β-regulated transcripts, including 728 known genes plus 88 duplicates (Fig. 2a), and 900 anonymous ESTs. Nonhierarchical clustering algorithms [self-organizing maps (32)] were used to determine 33 kinetically defined classes of gene regulation by TGF-β (Fig. 5, which is published as supplemental data on the PNAS web site, www.pnas.org). Normalized expression data for the 728 known genes are available in Table 1 (which is published as supplemental data) and on our web site (see Methods). Fifty-two percent of the regulated genes with known functions have roles in signaling, gene, and protein expression, cell–matrix adhesion, and cytoskeletal remodeling (Fig. 2d). Interestingly, gene repression is an important part of the cellular response to TGF-β, as it affected one-third of TGF-β target genes (Fig. 2d).

Figure 2.

Figure 2

Open in a new tab

Identification and functional classification of ERK-independent and -dependent TGF-β-regulated transcripts. (a) Hierarchical cluster alignment: the rows show 1,716 expression profiles of TGF-β-regulated transcripts in three experimental repeats (R1, R2, R3) of timed (0, 0.3, 1, 2, and 4 h) TGF-β stimulation. Colors indicate ratio of transcript abundance according to color bar scale. (b) Cluster diagrams show expression profiles (based on median ratios normalized against baseline) of ERK-dependent TGF-β target genes (see Methods for criteria) in different functional categories. ERK-dependent target genes fulfill criteria in both experimental designs “TGF-β ” and “TGF-β/ERK,” as indicated. ERK-independent target genes fail criteria in experimental design “TGF-β/ERK” (not shown). (c) Comparison of expression profiles (lines show normalized median ratios of three repeats) for ERK-independent and ERK-dependent TGF-β target genes, as determined by microarray or quantitative rtPCR assays. (d) Functional classification of 728 TGF-β target genes. Red (induced) and green (repressed) bars indicate proportions of 648 ERK-independent genes, and shaded inserts/italicized numbers indicate distribution of 80 ERK-dependent genes in functional categories.

Recent estimates predict less than 40,000 genes in the human genome. On the basis of a ≈10% gene redundancy among the 16,580 unique ESTs on our DNA arrays, we anticipate that the presented transcriptome screens cover ≈14,922 genes, representing ≈38% of all human genes. About 10% of genes covered in our screens were modulated by TGF-β within 4 h. We therefore predict dynamic changes in expression of up to 4,000 genes associated with the onset of TGF-β-initiated EMT in HaCaT. Our results allow the first genome-wide estimate of the extent of rapid genetic reprogramming induced by TGF-β in a defined physiological context.

A Subset of ERK-Dependent Target Genes of TGF-β Is Enriched for Genes with Functions in Integrin-Mediated Cell–Matrix Adhesion.

We designed a microarray hybridization strategy to specifically identify ERK-dependent target genes of TGF-β by replacing the common reference RNA T0 with RNA of cells treated with TGF-β in the presence of U0126. Eighty of 525 TGF-β-regulated known genes (15%) represented on the H1 arrays (5H arrays were not used in these experiments) were ERK-dependent (Fig. 2b) and generally required protein synthesis (data not shown). ERK dependence was notably more common in the cell–matrix adhesion category (20%), compared with all other functional categories (3–13%). Interestingly, although increased levels of phospho-ERK1/2 were not detectable at time points earlier than 1 h after TGF-β stimulation, regulation of few ERK-dependent target genes, including c-FOS (Fig. 2c), showed peak induction at 1 h and did not require protein synthesis (data not shown). This finding indicates that TGF-β may induce two distinct activation profiles of ERK in the described EMT model: direct, perhaps more rapid activation is required for some immediate-early genes (c-FOS), whereas indirect delayed activation is required for most ERK-dependent TGF-β target genes.

Reproducibility and validity of ERK-dependent and ERK-independent target genes were verified by comparing microarray and quantitative rtPCR measurements (Fig. 2c). Microarray assay reliability was further confirmed by concordance of multiple expression profiles for single genes represented by multiple ESTs (i.e., SLUG, etc.; see Fig. 6, which is published as supplemental data) and by confirmation of previously reported expression patterns for TGF-β-responsive genes (i.e., PAI1, etc.; see Fig. 7, which is published as supplemental data).

TGF-β Directs a Functionally Defined Genetic Program of Epithelial Cell Plasticity.

We identified a tightly coordinated program of genes and signaling pathways with putative roles in EMT, a manifestation of epithelial cell plasticity (Fig. 3a). It consists of regulators of epithelial dedifferentiation and maintenance of epithelial/transitional cell progenitors, including Notch and Wnt signaling pathways (33, 34), as suggested by a synchronized regulation of multiple positive and negative mediators in both pathways (Figs. 3a and 4). Genes in the Wnt pathway were consistently induced by TGF-β at 4 h and ERK-independent (Fig. 3a). In contrast, Notch pathway activation by TGF-β may ensue earlier (1–2 h) and may in part require ERK activation, as indicated by ERK-dependency of JAG1 and TLE3 (Figs. 2b and 3a). Key regulators of differentiation of early mesenchymal progenitor cells into chondrogenic (SOX9), hematopoietic and angiogenic (CBFA2, GATA2, KDR, VEGF), and cardiac lineages (SOX4) were induced as early as 1–4 h (Fig. 3a). TGF-β thus rapidly orchestrates an epithelial plasticity gene expression program that may specify a pluripotent mesenchymal progenitor cell phenotype (see Fig. 4).

Figure 3.

Figure 3

Open in a new tab

Cluster diagrams of gene expression profiles (normalized median ratios) in TGF-β-activated genetic programs for epithelial plasticity, cytoskeletal reorganization, and cell–matrix adhesion/cell motility. (a) Epithelial plasticity program: genes that promote mesenchymal progenitor cells (“mesenchyma”); genes and pathways (Notch, Wnt) that promote epithelial dedifferentiation and/or undifferentiated transitional progenitor cells (“transitional progenitor”); epithelial cell marker genes (“epithelial”). (b) Cytoskeletal remodeling program. (c) Cell–matrix adhesion and motility program. Column “ERK” indicates ERK-dependent (“yes”) and ERK-independent (“no”) gene regulation, respectively; ND, not determined. *, ERK dependence of c-FOS determined by quantitative rtPCR. Gene symbols follow Unigene nomenclature.

Figure 4.

Figure 4

Open in a new tab

Signaling modules and genetic programs underlying EMTs induced by TGF-β. Established (Smads and MEK/ERK) and unknown (Others) signaling mediators of TGF-β are shown in shaded boxes. Induced genes are colored in red, repressed genes in green. Genes and pathways are aligned according to time of peak activation. Solid line arrows indicate pathways, programs, and cellular responses suggested in this report. Broken-line arrows show mediators and dependent responses demonstrated previously. Dotted-line arrows show putative pathways proposed in this report. Genes in Notch and Wnt pathways are grouped together, as are mesenchymal progenitor factors.

TGF-β Regulates a Coordinated Program of Genes with Roles in Remodeling of Actin Cytoskeleton and Actin Stress Fiber Formation.

The cytoskeletal remodeling program includes largely ERK-independent (70%) genes with putative roles in F-actin turnover in response to TGF-β (degradation and de novo synthesis) (Fig. 3b). It involves mediators of F-actin degradation (RhoE, GSN), actin nucleation (ZYX, VASP), actin polymerization (ACTR3, ARPC5), actin capping (CAPZA1, GSN), and actin monomers (ACTA2, ACTG2). TGF-β also induced expression of phosphatidyl-4-phosphatase 5-kinase (PIP5K1A). PIP5K1A regulates generation of the phosphoinositide PIP2, which provides a potential link between Rho proteins and F-actin assembly. The ERK independence of the cytoskeletal remodeling program is consistent with our immunohistochemical observations of ERK-independent reorganization of cortical actin bundles to actin stress fibers induced by TGF-β (Fig. 1b).

In contrast, a group of genes involved in cell adhesion, extracellular matrix (ECM) remodeling, and cell motility [cell–matrix adhesion/motility program (Fig. 3c)], was strikingly enriched for ERK-dependent genes (60% ERK-dependent), compared with epithelial plasticity (25%) and cytoskeletal reorganization programs (30%). In this program, ERK-dependent fibronectin 1 (FN1), thrombospondin 1 (THBS1), laminin-5 (LAMC2, LAMA3) and tenascin-C (HXB), β4-integrin (ITGB4), and α2-integrin (ITGA2) constitute a core of ECM components and matrix receptors that are known to promote cell motility (35). Matrix metalloproteinases 1 (MMP1) and 12 (MMP12) modulate adhesive sites and clear ECM to break down physical barriers of cell motility (36). Analysis of the expression profiles of ERK-dependent genes in this group strongly suggests a specific functional role of ERK MAP kinase activation in mediating induction of cell motility during TGF-β-induced EMT. This finding is directly supported by the observation that U0126 inhibited TGF-β induced cell motility and cell–cell separation in our model (see Fig. 1 b and c).

Disassembly of Adherens Junctions in TGF-β-Induced EMT May Depend on ERK-Dependent Regulation of Genes with Roles in Cell Surface Membrane Turnover.

E-cadherin turnover and assembly or disassembly of adherens junctions are dynamically regulated by a balance of Rho and Rab GTPase-mediated endocytosis and exocytotic lateral vesicle targeting (37). MEK/ERK-dependent disassembly of adherens junctions observed in our system may be thus mediated by ERK-dependent target genes with roles in endocytosis (RHOB, NET1, RAB5EP), vesicle transport (RANs and RAB2), and/or membrane vesicle trafficking [LLGL2, a homologue of Drosophila lethal giant larvae (lgl)] (38) (see Figs. 2b and 4). Although U0126 inhibited TGF-β-initiated disassembly of adherens junctions, cell spreading and loosening of cell–cell contacts occurred (Fig. 1b). This phenotype appeared similar to that of cells overexpressing the zinc-finger protein Slug, which mediates disassembly of desmosomes but not adherens or tight junctions (39). Indeed, we observed ERK-independent induction of SLUG (Fig. 3a), and TGF-β-induced disassembly of desmosomes and tight junctions was not affected by U0126. Our results suggest that ERK activation by TGF-β may control a set of genes with functional roles in endocytosis that may be critical for disassembly of adherens junctions but not of desmosomes and tight junctions.

Conclusions

Our report provides a comprehensive view of coordinated regulation of genetic programs induced by TGF-β during the initial phase of EMT. A remarkably high number of genes (by extrapolation 4,000 or ≈10% of all genes, according to current estimates) undergo rapid changes in expression levels within only 4 hours of TGF-β treatment. This observation further underscores a broad and multifunctional impact of TGF-β signaling in cell biology.

Future design of disease process-specific interventions, such as prevention of tumor progression, will to a large extent depend on knowledge of mechanisms that determine biological specificity of TGF-β actions. We demonstrate that functional genomic approaches provide a new tool to define roles of critical nodes in TGF-β's signaling circuitry, based on inspection of their proximal signaling targets in the context of well-defined physiological responses. For example, we show that in the context of TGF-β-induced EMT, a subset of target genes, activated strongly and specifically via the ERK MAP kinase, functions in remodeling of integrin-based cell–matrix adhesion and promotes cell motility. The role of ERK, as predicted from temporal pattern classification of its proximal signaling targets, is consistent with alterations of phenotypic responses at the cellular level that are associated with ERK inhibition. We conclude that the control of cell–matrix adhesion and cell motility in TGF-β-induced EMT is, to our knowledge, a novel physiological role of ERK MAP kinase, independent of cell growth control. Contrary to a recent report (26), we did not observe ERK activation or an effect of ERK inhibition on growth inhibition by TGF-β in proliferating cells. This discrepancy may be explained by different concentrations of the MAP kinase inhibitor used in both studies.

Results of genome-wide expression profiling in cellular signaling systems (40) and clinical settings (41–44) are being reported and deposited in publicly accessible databases at an increasing pace. This report presents a database of hundreds of genes that are regulated by TGF-β and substantially expands the list of its known targets. By adapting DNA array methodology to a defined experimental context (EMT) and by using specific inhibition of MEK/ERK signaling, we were able to correlate ERK-dependent and -independent transcriptional programs of TGF-β signaling with complex manifestations of epithelial plasticity (Fig. 4).

Many of the ERK-dependent TGF-β target genes identified here (FN1, HXB, etc.) have recently been described in screens for genes that mediate tumor invasiveness and metastasis (41, 43, 44). Our results thus suggest a mechanism of promotion of tumor invasion and metastasis by TGF-β (19) that involves ERK-dependent activation of a described set of target genes strongly associated with aggressive malignancies.

The observations presented in this report provide insights into the complex molecular spectrum of epithelial plasticity controlled by TGF-β, an important mechanism of morphogenesis and tumor progression.

Supplementary Material

Supplemental Data
pnas_98_12_6686__index.html (1.3KB, html)

Acknowledgments

We thank Anita Roberts, Lalage Wakefield, Aris Moustakas, and Raju Kucherlapati for their critical review of the manuscript. We are grateful to Raju Kucherlapati for critical discussions and advice, and to Carl Baker for his support with quantitative rtPCR. We thank the National Institute of Diabetes and Digestive and Kidney Diseases, The Speaker's Fund for Biomedical Research: Toward the Science of Patient Care, and the American Cancer Society, who provided support for this research study. M.B. has been funded by the National Kidney Foundation of New York/New Jersey and Deutsche Forschungsgemeinschaft, Germany.

Abbreviations

TGF-β

transforming growth factor-β

EMT

epithelial–mesenchymal transition

ERK

extracellular response kinase

MAP

mitogen-activated protein

HaCaT

human keratinocytes

rtPCR

real-time PCR

MEK

MAP kinase/ERK kinase

EST

expressed sequence tag

Footnotes

This paper was submitted directly (Track II) to the PNAS office.

References

  • 1.Sporn M B. Cytokine Growth Factor Rev. 2000;11:3. doi: 10.1016/s1359-6101(99)00023-4. [DOI] [PubMed] [Google Scholar]
  • 2.Kimelman D, Kirschner M. Cell. 1987;51:869–877. doi: 10.1016/0092-8674(87)90110-3. [DOI] [PubMed] [Google Scholar]
  • 3.Potts J D, Runyan R B. Dev Biol. 1989;134:392–401. doi: 10.1016/0012-1606(89)90111-5. [DOI] [PubMed] [Google Scholar]
  • 4.Whitman M. Genes Dev. 1998;12:2445–2462. doi: 10.1101/gad.12.16.2445. [DOI] [PubMed] [Google Scholar]
  • 5.Böttinger E P, Jakubczak J L, Roberts I S, Mumy M, Hemmati P, Bagnall K, Merlino G, Wakefield L M. EMBO J. 1997;16:2621–2633. doi: 10.1093/emboj/16.10.2621. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Tang B, Böttinger E P, Jakowlew S B, Bagnall K M, Mariano J, Anver M R, Letterio J J, Wakefield L M. Nat Med. 1998;4:802–807. doi: 10.1038/nm0798-802. [DOI] [PubMed] [Google Scholar]
  • 7.Tang B, de Castro K, Barnes H E, Parks W T, Stewart L, Böttinger E P, Danielpour D, Wakefield L M. Cancer Res. 1999;59:4834–4842. [PubMed] [Google Scholar]
  • 8.Reynisdottir I, Polyak K, Iavarone A, Massague J. Genes Dev. 1995;9:1831–1845. doi: 10.1101/gad.9.15.1831. [DOI] [PubMed] [Google Scholar]
  • 9.Hannon G J, Beach D. Nature (London) 1994;371:257–261. doi: 10.1038/371257a0. [DOI] [PubMed] [Google Scholar]
  • 10.Oberhammer, F., Fritsch, G., Pavelka, M., Froschl, G., Tiefenbacher, R., Purchio, T. & Schulte-Hermann, R. (1992) Toxicol. Lett.64–65 Spec. No., 701–704. [DOI] [PubMed]
  • 11.Reiss M. Oncol Res. 1997;9:447–457. [PubMed] [Google Scholar]
  • 12.Massague J, Chen Y G. Genes Dev. 2000;14:627–644. [PubMed] [Google Scholar]
  • 13.Chen R H, Ebner R, Derynck R. Science. 1993;260:1335–1338. doi: 10.1126/science.8388126. [DOI] [PubMed] [Google Scholar]
  • 14.Cui W, Fowlis D J, Bryson S, Duffie E, Ireland H, Balmain A, Akhurst R J. Cell. 1996;86:531–542. doi: 10.1016/s0092-8674(00)80127-0. [DOI] [PubMed] [Google Scholar]
  • 15.Oft M, Peli J, Rudaz C, Schwarz H, Beug H, Reichmann E. Genes Dev. 1996;10:2462–2477. doi: 10.1101/gad.10.19.2462. [DOI] [PubMed] [Google Scholar]
  • 16.Birchmeier W, Behrens J. Biochim Biophys Acta. 1994;1198:11–26. doi: 10.1016/0304-419x(94)90003-5. [DOI] [PubMed] [Google Scholar]
  • 17.Miettinen P J, Ebner R, Lopez A R, Derynck R. J Cell Biol. 1994;127:2021–2036. doi: 10.1083/jcb.127.6.2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Piek E, Moustakas A, Kurisaki A, Heldin C H, ten Dijke P. J Cell Sci. 1999;112:4557–4568. doi: 10.1242/jcs.112.24.4557. [DOI] [PubMed] [Google Scholar]
  • 19.Portella G, Cumming S A, Liddell J, Cui W, Ireland H, Akhurst R J, Balmain A. Cell Growth Differ. 1998;9:393–404. [PubMed] [Google Scholar]
  • 20.Bakin A V, Tomlinson A K, Bhowmick N A, Moses H L, Arteaga C L. J Biol Chem. 2000;275:36803–36810. doi: 10.1074/jbc.M005912200. [DOI] [PubMed] [Google Scholar]
  • 21.Santibanez J F, Iglesias M, Frontelo P, Martinez J, Quintanilla M. Biochem Biophys Res Commun. 2000;273:521–527. doi: 10.1006/bbrc.2000.2946. [DOI] [PubMed] [Google Scholar]
  • 22.Lehmann K, Janda E, Pierreux C E, Rytomaa M, Schulze A, McMahon M, Hill C S, Beug H, Downward J. Genes Dev. 2000;14:2610–2622. doi: 10.1101/gad.181700. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Heldin C H, Miyazono K, ten Dijke P. Nature (London) 1997;390:465–471. doi: 10.1038/37284. [DOI] [PubMed] [Google Scholar]
  • 24.Attisano L, Wrana J L. Curr Opin Cell Biol. 2000;12:235–243. doi: 10.1016/s0955-0674(99)00081-2. [DOI] [PubMed] [Google Scholar]
  • 25.Hocevar B A, Brown T L, Howe P H. EMBO J. 1999;18:1345–1356. doi: 10.1093/emboj/18.5.1345. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Hu P P, Shen X, Huang D, Liu Y, Counter C, Wang X F. J Biol Chem. 1999;274:35381–35387. doi: 10.1074/jbc.274.50.35381. [DOI] [PubMed] [Google Scholar]
  • 27.Yue J, Mulder K M. J Biol Chem. 2000;275:30765–30773. doi: 10.1074/jbc.M000039200. [DOI] [PubMed] [Google Scholar]
  • 28.Boukamp P, Petrussevska R T, Breitkreutz D, Hornung J, Markham A, Fusenig N E. J Cell Biol. 1988;106:761–771. doi: 10.1083/jcb.106.3.761. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Bitzer M, von Gersdorff G, Liang D, Dominguez-Rosales A, Beg A A, Rojkind M, Böttinger E P. Genes Dev. 2000;14:187–197. [PMC free article] [PubMed] [Google Scholar]
  • 30.Cano A, Perez-Moreno M A, Rodrigo I, Locascio A, Blanco M J, del Barrio M G, Portillo F, Nieto M A. Nat Cell Biol. 2000;2:76–83. doi: 10.1038/35000025. [DOI] [PubMed] [Google Scholar]
  • 31.Hartsough M T, Mulder K M. J Biol Chem. 1995;270:7117–7124. doi: 10.1074/jbc.270.13.7117. [DOI] [PubMed] [Google Scholar]
  • 32.Tamayo P, Slonim D, Mesirov J, Zhu Q, Kitareewan S, Dmitrovsky E, Lander E S, Golub T R. Proc Natl Acad Sci USA. 1999;96:2907–2912. doi: 10.1073/pnas.96.6.2907. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Artavanis-Tsakonas S, Rand M D, Lake R J. Science. 1999;284:770–776. doi: 10.1126/science.284.5415.770. [DOI] [PubMed] [Google Scholar]
  • 34.Polakis P. Genes Dev. 2000;14:1837–1851. [PubMed] [Google Scholar]
  • 35.Ridley A. J Cell Biol. 2000;150:F107–F109. doi: 10.1083/jcb.150.4.f107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Vu T H, Werb Z. Genes Dev. 2000;14:2123–2133. doi: 10.1101/gad.815400. [DOI] [PubMed] [Google Scholar]
  • 37.Kamei T, Matozaki T, Sakisaka T, Kodama A, Yokoyama S, Peng Y F, Nakano K, Takaishi K, Takai Y. Oncogene. 1999;18:6776–6784. doi: 10.1038/sj.onc.1203114. [DOI] [PubMed] [Google Scholar]
  • 38.Bilder D, Li M, Perrimon N. Science. 2000;289:113–116. doi: 10.1126/science.289.5476.113. [DOI] [PubMed] [Google Scholar]
  • 39.Savagner P, Yamada K M, Thiery J P. J Cell Biol. 1997;137:1403–1419. doi: 10.1083/jcb.137.6.1403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Iyer V R, Eisen M B, Ross D T, Schuler G, Moore T, Lee J F, Trent J M, Staudt L M, Hudson J J, Boguski M S, et al. Science. 1999;283:83–87. doi: 10.1126/science.283.5398.83. [DOI] [PubMed] [Google Scholar]
  • 41.Perou C M, Sorlie T, Eisen M B, van de Rijn M, Jeffrey S S, Rees C A, Pollack J R, Ross D T, Johnsen H, Akslen L A, et al. Nature (London) 2000;406:747–752. doi: 10.1038/35021093. [DOI] [PubMed] [Google Scholar]
  • 42.Alizadeh A A, Eisen M B, Davis R E, Ma C, Lossos I S, Rosenwald A, Boldrick J C, Sabet H, Tran T, Yu X, et al. Nature (London) 2000;403:503–511. doi: 10.1038/35000501. [DOI] [PubMed] [Google Scholar]
  • 43.Bittner M, Meltzer P, Chen Y, Jiang Y, Seftor E, Hendrix M, Radmacher M, Simon R, Yakhini Z, Ben Dor A, et al. Nature (London) 2000;406:536–540. doi: 10.1038/35020115. [DOI] [PubMed] [Google Scholar]
  • 44.Clark E A, Golub T R, Lander E S, Hynes R O. Nature (London) 2000;406:532–535. doi: 10.1038/35020106. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Data
pnas_98_12_6686__index.html (1.3KB, html)
pnas_98_12_6686__1.html (1.7KB, html)
pnas_98_12_6686__2.pdf (977.4KB, pdf)
pnas_98_12_6686__6143Table1.xls (217.5KB, xls)
pnas_98_12_6686__3.html (1.1KB, html)
pnas_98_12_6686__4.pdf (19.1KB, pdf)
pnas_98_12_6686__5.html (1KB, html)
pnas_98_12_6686__6.pdf (1.6MB, pdf)

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences

RESOURCES