Journal of Dermatological Science 58 (2010) 97–104

TGF-beta1 causes epithelial-mesenchymal transition in HaCaT derivatives,but induces expression of COX-2 and migration only in benign, not inmalignant keratinocytes

Kati Rasanen *, Antti Vaheri

Haartman Institute, POB 21, FI-00014 University of Helsinki, Finland

A R T I C L E I N F O

Article history:

Received 27 November 2009

Received in revised form 4 February 2010

Accepted 5 March 2010

Keywords:

Transforming growth factor

Keratinocyte

Carcinogenesis

Epithelial-mesenchymal transition

Cyclooxygenase 2

Migration

A B S T R A C T

Background: Transforming growth factor b (TGF-b) acts as a tumor promoter by inducing epithelial-

mesenchymal transition (EMT), which leads to a motile phenotype, enabling invasion and metastasis of

cancer cells. Cancer-related inflammation, mediated by prostaglandins, has been proposed as a critical

mechanism in conversion of benign cells to malignant.

Objective: Induction of cyclooxygenase 2 (COX-2), producer of prostaglandins, is thought to be a

prerequisite for TGF-b-induced EMT in benign cells. We used HaCaT derivatives, representative of skin

cancer progression, to investigate TGF-b1 mediated EMT response, and the role of COX-2 in it.

Methods: Effect of TGF-b1 was investigated by analyzing cell proliferation, morphology and protein

expression. Chemotaxis and scratch-wound assays were used to study migration.

Results: TGF-b1 caused proliferation arrest of benign and malignant HaCaT cells, and changed the

epithelial morphology of benign and low-grade malignant cells, but not metastatic cells, to

mesenchymal spindle-shape. Epithelial junction proteins ZO-1 and E-cadherin were downregulated

in all cell lines in response to TGF-b1, but mesenchymal markers were not induced, suggesting a partial

EMT response. COX-2 and migration were induced only in benign HaCaT derivatives. Malignant

derivatives did not induce COX-2 in response to TGF-b 1 treatment, thus emphasizing the role of

inflammation in EMT response of benign cells.

Conclusions: TGF-b1 operates via distinct mechanisms in inducing EMT and metastasis, and supporting

this we show that TGF-b1 induces COX-2 and promotes the migration of benign cells, but does not further

augment the migration of malignant cells, indicating their resistance to TGF-b1 in the context of motility.

� 2010 Japanese Society for Investigative Dermatology. Published by Elsevier Ireland Ltd. All rights

reserved.

Contents lists available at ScienceDirect

Journal of Dermatological Science

journa l homepage: www.e lsev ier .com/ jds

1. Introduction

Epithelial-mesenchymal transition (EMT) is a biological processthat allows tightly organized epithelial cells to acquire a motilemesenchymal phenotype. Recently it has been suggested that EMTshould be classified into three different subtypes, as it occurs invarious biological and pathological settings. Type 1 EMT isassociated with embryonal and organ development, type 2 withwound healing, tissue regeneration and fibrosis, and type 3 withcancer progression and metastasis [1]. In the multi-step process ofcancer progression, invasion through the basement membrane is acrucial event, and EMT has been proposed as the criticalmechanism [2]. A key event in breaking the polarity is the lossof epithelial junctions, mediated by downregulation of tight andadherens junction proteins, zonula occludens 1 (ZO-1) and E-

* Corresponding author. Tel.: +358 919126608; fax: +358 919126491.

E-mail address: kati.rasanen@helsinki.fi (K. Rasanen).

0923-1811/$36.00 � 2010 Japanese Society for Investigative Dermatology. Published b

doi:10.1016/j.jdermsci.2010.03.002

cadherin, respectively. At the same time mesenchymal proteins,such as vimentin and fibronectin, are upregulated, facilitating themigratory phenotype [3]. Cytoskeletal rearrangement is critical formigration of epithelial cells, often observed as formation offilamentous actin stress fibers [4]. The HaCaT keratinocyte cellpanel represents different stages of skin carcinogenesis fromspontaneously immortalized non-malignant cells (HC) to H-ras

transformed benign A5 cells, to low-grade malignant HaCaT II-4cells forming locally invasive highly differentiated squamous cellcarcinomas (SCC), and to high-grade malignant, metastasizing RT3cells [5,6]. Transforming growth factor b (TGF-b) is a regulatorycytokine that has paradoxical roles in cancer. It suppresses tumorprogression of normal and premalignant cells, but when cancercells lose their response to TGF-b, it induces differentiation into aninvasive phenotype and is thus a potent inducer of EMT [7]. TGF-b1has been shown to induce EMT in immortal and malignant HaCaTcells through mitogen-activated protein kinase (MAPK) and Smad(transcriptional regulator of TGF-b) signaling pathways. However,the immortal HC cells initiated EMT only when the Ras pathway

y Elsevier Ireland Ltd. All rights reserved.

K. Rasanen, A. Vaheri / Journal of Dermatological Science 58 (2010) 97–10498

was activated in these cells by epidermal growth factor (EGF) [8].In a study by Wilkins-Port et al. [9], TGF-b1 together with EGFpotentiated the invasion of II-4 cells through collagen type I. Thecollagenolytic phenotype was mediated by increased expression ofmatrix metalloproteinase 10 (MMP-10).

Studies using non-tumorigenic intestinal [10] and mammary[11] epithelial cells have shown that TGF-b1 induces expression ofcyclooxygenase 2 (COX-2) through the p38 MAPK dependentpathway and causes inactivation of Smad signaling. COX-2 isexpressed aberrantly in a variety of carcinomas, leading toincreased conversion of its metabolic product prostaglandin E2

[12]. Prostaglandins are executors of inflammation, but are alsolinked to cancer progression, and PGE2 has been shown to promoteproliferation, survival, angiogenesis, migration and invasion [13].EMT mediated through upregulation of COX-2 has been shown tobe PGE2 dependent [11].

We have previously shown that co-culturing HaCaT derivativeswith nemotic fibroblasts (for review, see [14]) does not causeinduction of COX-2 or EMT phenotype in HaCaT cells [15]. Instead,as we recently reported, nemosis increases proliferation andaugments motility of the HaCaT derivatives [16]. Therefore thepurpose of the present study was to investigate the EMT responseof the HaCaT cell panel induced by TGF-b1 and the role of COX-2 inthis process. The effect of TGF-b1 on HaCaT derivatives wasinvestigated by analyzing changes in cell proliferation, expressionof known EMT markers and migratory response.

2. Materials and methods

2.1. Cell lines and cell culture

The HaCaT cell line panel, representing keratinocyte tumor-igenesis, was kindly provided by Prof. Petra Boukamp and Prof.Norbert E. Fusenig (DKFZ, Heidelberg, Germany). All cell lines werecultured at +37 8C in 5% CO2 atmosphere in Dulbecco’s modifiedEagle’s medium (DMEM) (Invitrogen, Carlsbad, CA) supplementedwith 5% fetal calf serum (FCS) (Invitrogen), 0.3 mg/ml glutamine,100 mg/ml streptomycin, and 100 U/ml penicillin. For experi-ments, cells were seeded and grown for 48 h, after which they wereserum-starved in medium containing no FCS for 24 h. TGF-b1-containing or control medium was added and cells were culturedfurther 48 h. HaCaT cells were used until passage number 50.

For morphological analysis 0.75 � 105 cells/well (6-well plate)were seeded and after 48 h of incubation with or without TGF-b1the cells were imaged with Olympus CXK41 phase contrastmicroscope and photographed with Olympus DP12 digital camera.

For chemotaxis assay HaCaT cell lines were transduced withpLV-PGK/GFP (kind gift from Professor Seppo Yla-Herttuala, AIVInstitute, Kuopio, Finland) green fluorescent protein-lentivirus.Cells seeded (3 � 105 cells/well) into 6-well plates were trans-duced with pLV-PGK/GFP virus (titer of 2.8 � 106) in the presenceof polybrene (6 mg/ml, Sigma–Aldrich, St. Louis, MO). After over-night incubation, virus-containing medium was removed, and cellswere washed and resuspended in normal growth medium forexperimentation. The transduction efficiency was over 95% in allcell lines.

2.2. Reagents and antibodies

Recombinant human TGF-b1 was from R&D Systems (Minnea-polis, MN) and was used at 2 ng/ml concentration throughout. Forimmunohistochemistry, rabbit monoclonal anti-Ki-67 antibodywas used (Labvision, Fremont, CA). The following primaryantibodies were used for immunoblotting and immunofluores-cence: mouse monoclonal anti-E-cadherin, mouse monoclonalanti-ZO-1 (both from BD Biosciences, Bedford, MA), mouse

monoclonal anti-PCNA, rabbit polyclonal anti-COX-2 (both fromLabvision), rabbit polyclonal anti-GAPDH (Santa Cruz Biotechnol-ogy, Santa Cruz, CA), mouse monoclonal anti-vimentin [17] andmouse monoclonal anti-fibronectin [18]. Secondary antibodieswere: horseradish peroxidase coupled anti-mouse IgG and anti-rabbit IgG (both from Jackson Immunoresearch, Cambridgeshire,UK) for immunoblotting and goat anti-mouse IgG Alexa Fluor 488(Invitrogen) for immunofluorescence. Rhodamine-labeled phalloi-din (Molecular Probes, Eugene, OR) was used to stain for F-actinand Hoechst 33342 (Molecular Probes) for nuclei.

2.3. Proliferation assay

The MTT cell proliferation assay (Cayman Chemical Company,Ann Arbor, MI) was used to confirm that TGF-b1 causesproliferation arrest of HaCaT cells. In the assay, performedaccording to manufacturer’s protocol, HaCaT cells were seededto 96-well plates (20,000 cells/well) and let to plate for 2 h. Culturemedium with unattached cells was aspirated and replaced with100 ml of either control medium or TGF-b1. Cells were incubatedfor 48 h, after which 10 ml MTT reagent (5 mg/ml) was added andincubated for 4 h at +37 8C in 5% CO2 incubator. Culture mediumwas aspirated and 100 ml Crystal Dissolving Solution was added toeach well. The absorbance of the dissolved formazan crystals wasmeasured at 540 nm using Multiscan EX microplate reader(Thermo Labsystems, Vantaa, Finland).

2.4. Immunohistochemistry

HaCaT derivatives (5 � 104 cells inside a silicon ring) weregrown on SuperFrost Plus adhesion slides (Menzel-Glazer, ThermoFisher Scientific, Cheshire, UK) for 24 h, after which medium wasreplaced with experimental media (ctrl or TGF-b1) and incubatedfor further 48 h. Ice-cold methanol was used for fixation, afterwhich cells were washed with PBS. Ventana Discovery immuno-histochemistry Slide Stainer (Ventana Medical Systems, Tucson,AZ) was used for immunohistochemical staining. Slides wereincubated with primary antibody (Ki-67) for 30 min. The stainingwas performed with the Ventana 3,30-diaminobenzidine tetra-hydrochloride (DAB) biotin avidin detection kit. Images werecaptured with Olympus BX50 microscope using Olympus DP12digital camera.

2.5. Immunoblotting

Cells (3 � 105 in 6-well plates) were washed twice with PBS andcollected in 2� sample buffer (125-mM Tris (pH 6.8), 4% sodiumdodecyl sulfate (SDS), 0.01% bromophenol blue, 10% b-mercap-toethanol, 10% glycerol) and boiled for 10 min. Equal amounts ofprotein from each sample were resolved by 10% SDS-PAGE,transferred to nitrocellulose membrane (Schleicher & Schuell,Dassel, Germany) and blocked with 2.5% non-fat powdered milk inTBS (20 mM Tris–HCl pH 7.5, 150 mM NaCl and 0.1% Tween-20).Immunoreactive proteins, incubated with appropriate primary andsecondary antibodies, were visualized using ECL detection (Pierce,Rockford, IL).

2.6. Immunofluorescence

For immunofluorescence, the HaCaT derivatives (0.75 � 105 in6-well plates) treated with our without TGF-b1 were fixed with acytoskeleton-preserving fixative (4% paraformaldehyde, 0.32 mol/lsucrose, 10 mmol/l MES, 138 mmol/l KCl, 3 mmol/l MgCl2, 2 mmol/l EGTA). To diminish background, fixed cells were incubated withNH4Cl, permeabilized with 0.5% Triton-X-100/PBS and blockedwith 0.2% BSA/PBS containing calcium and magnesium. Cells were

Fig. 1. Proliferation arrest caused by TGF-b1 treatment. (A) HaCaT derivatives were

grown either in control or TGF-b1-containing medium and cell proliferation was

measured using MTT assay. P < 0.05 in all cells lines in control vs. TGF-b1. (B) The

expression of proliferation marker Ki-67 was analyzed immunohistochemically

from HaCaT cell panel; TGF-b1 reduced significantly the expression of Ki-67. The

percentage of Ki-67 positive nuclei (brown) from representative images is

indicated.

K. Rasanen, A. Vaheri / Journal of Dermatological Science 58 (2010) 97–104 99

incubated with appropriate primary and secondary antibodiesdiluted in 0.2% BSA/PBS with Ca and Mg, and mounted. Phalloidinor Hoechst staining were performed after antibody staining.Fluorescent images were produced using Zeiss Axioplan 2microscope and Hamamatsu digital camera.

2.7. Migration assays

GFP-labeled HaCaT derivatives were used for chemotaxis assay.To test the migration response towards TGF-b1 modified Boydenchambers with 3-mm pore polycarbonate filters (Millicell, Milli-pore, Billerica, MA) were inserted in a 24-well plate. The lowerchamber was filled with control medium or TGF-b1 containingmedium and GFP-labeled keratinocytes (50,000/200 ml) wereadded to the upper chamber. Cells were let to migrate for 24 h,after which the inserts were washed with PBS and non-migratedcells from top of the filter were removed. Migrated cells werephotographed with Olympus DP12 camera using Olympus IX71inverted immunofluorescence microscope. The intensity of theGFP, relating to the amount of migrated cells, was measured usingImage J software (NIH Image).

Scratch-wound assay was used to confirm the migratoryresponse of HaCaT derivatives. HaCaT cells were grown toconfluence, serum-starved and scratched with pipette tip to createa wound. After wounding, cells were washed twice with serum-free medium to remove cell debris and grown for 24 h in control orTGF-b1 containing medium. Photographs were taken withOlympus DP12 camera and the percentage of the open woundarea at 24 h was analyzed using TScratch software [19] (http://www.cse-lab.ethz.ch).

2.8. Statistical analysis

All experiments were done in duplicate and repeated threetimes. The mean and SEM of all three experiments are shown.Statistical significance, calculated using GraphPad software, wasdetermined by unpaired t-test. P-Value of less than 0.05 wasconsidered to be statistically significant.

3. Results

3.1. TGF-b1 inhibits proliferation of benign and malignant HaCaT

cells

In order to go through EMT, cells must switch from proliferationto a differentiation program. Therefore we used the MTT cellproliferation assay to measure the proliferation rate of HaCaT cells.As expected, TGF-b1 inhibited cell proliferation of benign HC andA5 HaCaT clones. It also inhibited the proliferation of malignant II-4 and RT3, which contradicts previous results [20] (P < 0.05compared to control in all HaCaT derivatives) (Fig. 1A). Theproliferation arrest was further confirmed by immunohistochem-ical staining of widely used proliferation marker Ki-67 (Fig. 1B),that is present during all active phases of the cell cycle but absentin resting cells. Its expression was significantly (P < 0.05)decreased in all cell lines in response to TGF-b1 treatment,corroborating the MTT result.

3.2. Change in cell morphology caused by TGF-b1

A classical feature of TGF-b induced EMT is the morphologicalchange of epithelial cells to mesenchymal spindle-shape that isaccompanied with cytoskeletal reorganization through formationof stress fibers [21]. Phase contrast images (Fig. 2A) revealed thatwithin 48 h, TGF-b1 had caused a clear change in morphology ofHC, A5 and II-4 cells, but in the metastatic RT3 cells no such change

to elongated spindle-shape morphology was seen. Presence offilamentous actin (F-actin) detected by phalloidin staining (Fig. 2B)confirmed the change in cell morphology; in control cells no stressfibers were observed, whereas in HC, A5 and II-4 cells TGF-b1induced formation of actin stress fibers crossing the cell body.Interestingly, in RT3 cells phalloidin staining was diffuse both incontrol and TGF-b1-treated cells.

3.3. Loss of epithelial cell junction proteins in response to TGF-b1

Epithelial cell junctions are critical in maintaining the cellpolarization, and expression of proteins involved in these

Fig. 2. TGF-b1-induced morphological change and F-actin stress fiber formation (A) 48-h treatment with TGF-b1 changed the morphology of serum-starved HC, A5 and II-4

cells from cobblestone-like to elongated bipolar spindle-shape; this effect was not observed with metastatic RT3 cells. Images were captured with phase contrast microscope.

(B) Immunofluorescence images of phalloidin staining (red) revealed stress fiber formation in HC, A5 and II-4 cells in response to TGF-b1, which was not seen in RT3 cells.

Nuclei (blue) were stained with Hoechst.

K. Rasanen, A. Vaheri / Journal of Dermatological Science 58 (2010) 97–104100

structures is repressed during EMT. Therefore we investigated theexpression of the tight junction protein ZO-1 and adherensjunction protein E-cadherin in cells treated with TGF-b1 for48 h. Immunoblotting (Fig. 3A) revealed that expression of ZO-1was completely abolished in response to TGF-b1 in all cell lines. E-cadherin levels were downregulated to some extent, more so inbenign HC and A5 cells, correlating with study by Herfs et al. [22] inwhich TGF-b1 downregulated expression of E-cadherin time-dependently in HC cells. Immunostaining revealed that TGF-b1diminished expression of ZO-1 (Fig. 3B) totally in HC and A5 cells,and a weak diffuse staining was observed in II-4 and RT3 cells. E-cadherin expression (Fig. 3C) was not completely lost in any celllines in response to TGF-b1, correlating with immunoblottingresult. However, the staining pattern was clearly more diffuse in allHaCaT derivatives treated with TGF-b1 compared to the clearcortical staining of cell junctions in control cells. Particularly in RT3cells treated with TGF-b1 E-cadherin staining appeared sharp-pointed towards cytoplasm.

3.4. TGF-b1 causes induction of COX-2 in benign HaCaT cell lines

We used immunoblotting to investigate the COX-2 expressionin response to TGF-b1 (Fig. 4). The proliferation arrest in thesesamples was confirmed by PCNA (proliferating cell nuclearantigen) levels. As expected, PCNA expression was downregulatedin all HaCaT derivatives treated with TGF-b1. The effect was morepronounced in the benign HC and A5 cells, than in the malignant II-4 and RT3 cell lines. In line with our previously published results[15], COX-2 was not expressed in any of the HaCaT cell lines undernormal culture conditions. COX-2 was induced in benign HC andA5 cells in response to TGF-b1 treatment. We did not detect COX-2in II-4 cells, and a slight increase was observed in RT3 cells treated

with TGF-b1. We also investigated the expression of mesenchymalmarkers vimentin and fibronectin from these samples byimmunoblotting, but did not detect any expression of theseproteins (data not shown).

3.5. TGF-b1 promotes migration of benign HaCaT cells

The release from tight and adherens junctions causes the loss ofepithelial cell polarity and thus enables cell migration. Weinvestigated the migratory response of HaCaT derivatives inresponse to TGF-b1 treatment. In chemotaxis assay (Fig. 5A) TGF-b1 significantly enhanced the migration of benign HC and A5 cells(P < 0.05). As expected, malignant II-4 and RT3 cells grown undernormal culture conditions migrated more than benign HC and A5cells, but interestingly TGF-b1 slightly reduced the migratoryresponse of these cells (P = 0.06). The effect of TGF-b1 on HaCaTderivative migration was further confirmed using scratch-woundassay (Fig. 5B). Wound closure in TGF-b1 treated cells wassignificantly augmented in benign HC and A5 cells (P < 0.05)compared to control cells at 24 h after wounding; this was not seenwith malignant II-4 and RT3 cells.

4. Discussion

The objective of this study was to investigate the epithelial-mesenchymal transition of HaCaT derivatives in response to TGF-btreatment. EMT is a highly conserved, fundamental process ofnormal morphogenesis and members of the transforming growthfactor b family control many aspects of normal development andhomeostasis [23]. TGF-b is synthesized in an inactive form andsequestered to the ECM. Latent TGF-b is activated by proteases,such as MMPs, cleaving the ECM. Binding of activated TGF-b leads

Fig. 3. Downregulation of epithelial markers ZO-1 and E-cadherin in response to TGF-b1. (A) Immunoblotting showed complete loss of tight junction protein ZO-1 in all HaCaT

derivatives treated with TGF-b1, expression of adherens junction protein E-cadherin was downregulated to a lesser extent. GAPDH served as a loading control.

Immunofluorescence of (B) ZO-1 (green) confirmed immunoblotting results: in HC and A5 cells ZO-1 was completely lost in response to TGF-b1, in II-4 and RT3 cells weak

diffuse staining was observed. (C) E-cadherin (green) staining revealed a diffuse pattern of expression in response to TGF-b1 in all HaCaT cell lines compared to cortical

staining in control cells. Nuclei (blue) were stained with Hoechst.

K. Rasanen, A. Vaheri / Journal of Dermatological Science 58 (2010) 97–104 101

to formation and activation of receptor kinase complex andphosphorylation of Smad transcription factors. Activated Smadcomplex then translocates to the nucleus and thereby inducestarget gene transcription. TGF-b signals mainly through activationof Smads, but it also leads to activation of several MAP kinasepathways [24]. Smad-independent signaling can be activated byRas and EGF [25]. In some epithelial cells cooperative signalingbetween TGF-b and Ras is required for maintenance of EMT [23],and Oft et al. [26] have shown that non-malignant HaCaT cellsrequire the activation of Ras for TGF-b1 induced EMT. Ras theninduces both autocrine production of TGF-b1 and nuclearaccumulation of phosphorylated Smad and thus upregulatestranscription of target genes.

Often in cancer, malignant cells have increased production ofTGF-b1 leading to increased invasion via EMT-dependent mechan-isms. Of the TGF-b isoforms, TGF-b1 is most frequentlyupregulated in cancers. TGF-b1 exerts its growth-inhibitoryresponse via inducing expression of cell cycle inhibitors [27]and downregulating the expression of c-Myc [28,29]. High levels ofc-Myc in cancer cells render them resistant to TGF-b growthinhibition [30]. Interestingly, in the case of squamous cellcarcinomas TGF-b expression is often lost, correlating withreduced sensitivity to TGF-b1-induced growth arrest and highrisk to malignant, metastatic conversion [31]. In the panel ofmurine carcinomas SCC cells did not secrete TGF-b, only the highlyinvasive spindle-shaped cancer cell secreted significant amounts

Fig. 4. TGF-b1 induces expression of COX-2. Immunoblot revealed induction of

inflammation-related COX-2 only in benign HC and A5 cells, but not in malignant

derivatives treated with TGF-b1. PCNA was used as a marker for proliferation arrest

and GAPDH served as a loading control.

Fig. 5. Migration of benign HaCaT cells promoted by TGF-b1. (A) Chemotaxis

towards TGF-b1 was tested using a modified Boyden chamber assay. TGF-b1

significantly increased the migration of benign HC and A5 cells (P < 0.05), but

decreased slightly the migration of malignant II-4 and RT3 cells. (B) Scratch-wound

assay was used to confirm the migration-promoting effect of TGF-b1 on benign

HaCaT clones. The percentage of open wound area at 24 h after wounding was

calculated using TScratch software.

K. Rasanen, A. Vaheri / Journal of Dermatological Science 58 (2010) 97–104102

of TGF-b [26]. TGF-b isoforms are differentially expressed inHaCaT derivatives; TGF-b1 was shown to be expressed mainly innon-malignant cells, whereas the highly aggressive cells had lowerlevels of TGF-b1 [32]. Another study showed that tumorprogression of these derivatives was associated with progressiveabrogation of TGF-b1 and EGF growth control [33]. Contradictingthese results and the results by Davies et al. [20], we show here thatTGF-b1 is able to cause proliferation arrest also in metastatic RT3cells. These cells seem to hyperproliferate, as shown by the cellproliferation assay, but given their metastatic potential themorphology is not fibroblastoid spindle-shape that is normallyassociated with invasive cells, and curiously TGF-b1 did not causemorphological change in RT3 cells, but did change the morphologyof HC, A5 and II-4 cells.

Morphological change induced by TGF-b1 is accompanied withreorganization of actin cytoskeleton, and correlating with theresults from Xie et al. [34] where benign epithelial cells were used,we observed F-actin stress fiber formation in benign HC and A5cells and in low-grade malignant II-4 cells. Cytoskeletal rearrange-ment was not found in metastatic RT3 cells treated with TGF-b1,corroborating with our morphological data. Notably the phalloidinstaining was more diffuse in these cells than in the HC, A5 and II-4cells even when they were grown in normal culture conditions, andthe staining pattern did not change in response to TGF-b1treatment, implying an intrinsic change in actin cytoskeleton in theRT3 cells.

Next we investigated the levels of epithelial junction proteinsZO-1 and E-cadherin in response to TGF-b1 treatment. Surpris-ingly, also the malignant II-4 and RT3 cells had high basal levels ofthese epithelial junction proteins, as detected both by immuno-blotting and immunofluorescence. Expression of ZO-1 decreasedconsiderably in all cell lines in response to TGF-b1 within 48 h. E-cadherin levels were downregulated to a lesser extent; thestaining pattern changed from junctional localization to morediffuse in cells treated with TGF-b1, correlating with results fromHerfs et al. [22]. The sharp-pointed localization of E-cadherintowards cytoplasm in TGF-b1-treated RT3 cells could be theexplanation for the lack of morphological restructuring that wasobserved in phase contrast and phalloidin images. However, it isworth noting that in the case of this HaCaT keratinocyte panel, lossof differentiation is not a prerequisite for malignant growthbehavior [5]. Also supporting this notion, a study by Han et al. [35]showed that EMT and metastasis mediated by TGF-b1 are distinctprocesses and cells that overexpress TGF-b1 but have lost thecognate receptor (TGF-bRII) do metastasize without losing theexpression of E-cadherin, thus corroborating our observations.

Study by Neil et al. [11] suggested that induction of COX-2 isrequired for TGF-b1-induced EMT of non-malignant cells. In thatstudy, COX-2 was shown to be induced in normal mammaryepithelial cells in response to TGF-b1 treatment, leading to EMT,invasion and anchorage-independent growth, and that inhibitionof COX-2 inhibited PGE2 production and thus the EMT phenotype.In non-tumorigenic rat intestinal epithelial cells TGF-b1 enhancedCOX-2 mRNA stability via the p38 MAPK-signaling pathway [10].Concurring with these results, we show here that TGF-b1 inducesexpression of COX-2 in non-malignant HC and A5 cells, but not inmalignant II-4 and RT3 cells.

Contradicting a previous report [36], we did not detect de novo

expression of mesenchymal markers vimentin and fibronectin incells treated with TGF-b1. However, in that report the incubationperiod was much longer than in our experiments (8 days vs. 2days), suggesting that a longer time is needed for the expression ofthese proteins. Supporting this, Tomakidi et al. [37] showed thatvimentin was not expressed in benign or malignant HaCaT cells,and synthesis occurred only in the malignant cells after 2 weeksculturing as cell transplants in nude mice. In a study by Szell et al.

K. Rasanen, A. Vaheri / Journal of Dermatological Science 58 (2010) 97–104 103

[38] only proliferating HaCaT cells released from quiescencestarted to express fibronectin, in particular the oncofetal EDAisoform. Here we show that TGF-b1 causes a proliferation arrest,and this is the probable cause for lack of fibronectin expression,even though the cells become spindle-shaped and lose epithelialjunctions. This also suggests that the cells have gone through apartial, not complete EMT.

Epithelial plasticity is an absolute requirement for tumorprogression, and EMT is the mechanism for local disseminationleading to motility and invasion [39]. Therefore we investigatedthe migration response of HaCaT clones. As expected, the degree ofchemotaxis correlated with malignancy of the cell lines, increasingfrom benign to malignant. Bachmeier et al. [40] have shown thatmalignant cells secrete more MMPs, and MMP-10 is expressed onlyby malignant HaCaT derivatives, which could lead to activation oflatent TGF-b1 and thus explain the increased migratory responseof II-4 and RT3 cells. The migratory response of benign HC and A5cells increased significantly in response to TGF-b1. Several studieshave shown that expression of COX-2 promotes cell migration [41–44], and that actin stress fiber formation is required for epithelialcell migration [40], thus corroborating the increased migration ofbenign HC and A5 in response to TGF-b1. Also supporting this,Johansson et al. [45] showed that TGF-b1 causes activation of p38MAPK and ERK2 (extracellular signal-regulated kinase) in A5 cellsand this leads to expression of MMP-13 and MMP-1 and henceincreased invasion and furthermore, Xie et al. [34] reported thatactivation of the ERK pathway is required for TGF-b1-inducedEMT. Surprisingly, TGF-b1 seemed to inhibit the chemotaxis ofmalignant II-4 and RT3 cells, thus suggesting the reducedsensitivity of these cells to TGF-b1 in context of motility andmetastatic potential.

In conclusion, we report here that TGF-b1 induces epithelial-mesenchymal transition both in benign and malignant HaCaTclones by causing proliferation arrest, cytoskeletal reorganizationand downregulation of epithelial cell junction proteins. The EMTphenotype is accompanied by induction of COX-2 in benignkeratinocyte derivatives, whereas COX-2 induction is not observedin malignant HaCaT cells, indicating that COX-2 production isrequired for EMT response of non-malignant cells. Induction ofCOX-2 in benign HaCaT cell clones, but not in malignant clones, isaccompanied by increased migration in response to TGF-b1. Theprostaglandin synthase pathway is known to be a majorcontributor to inflammation, and cancer-related inflammationhas been proposed as a key event in progression of tumors [46], andour results here further strengthen that notion.

Acknowledgements

Prof. Petra Boukamp and Prof. Norbert E. Fusenig (DKFZ,Heidelberg) are thanked for providing the HaCaT cell panel. IrinaSuomalainen is thanked for her expert technical assistance andadvice, and Pertteli Salmenpera for providing the GFP-labeled cells.This work was supported by grants from Magnus EhrnroothFoundation, Finnish Cancer Organizations and Academy of Finland.

References

[1] Kalluri R, Weinberg RA. The basics of epithelial-mesenchymal transition. JClin Invest 2009;119:1420–8.

[2] Guarino M, Rubino B, Ballabio G. The role of epithelial-mesenchymal transitionin cancer pathology. Pathology 2007;39:305–18.

[3] Zeisberg M, Neilson EG. Biomarkers for epithelial-mesenchymal transitions. JClin Invest 2009;119:1429–37.

[4] Zhang L, Deng M, Parthasarathy R, Wang L, Mongan M, Molkentin JD, et al.MEKK1 transduces activin signals in keratinocytes to induce actin stress fiberformation and migration. Mol Cell Biol 2005;25:60–5.

[5] Boukamp P, Stanbridge EJ, Foo DY, Cerutti PA, Fusenig NE. c-ha-ras oncogeneexpression in immortalized human keratinocytes (HaCaT) alters growth

potential in vivo but lacks correlation with malignancy. Cancer Res1990;50:2840–7.

[6] Fusenig NE, Boukamp P. Multiple stages and genetic alterations in immorta-lization, malignant transformation, and tumor progression of human skinkeratinocytes. Mol Carcinog 1998;23:144–58.

[7] Massague J. TGFbeta in cancer. Cell 2008;134:215–30.[8] Davies M, Robinson M, Smith E, Huntley S, Prime S, Paterson I. Induction of an

epithelial to mesenchymal transition in human immortal and malignantkeratinocytes by TGF-beta1 involves MAPK, smad and AP-1 signalling path-ways. J Cell Biochem 2005;95:918–31.

[9] Wilkins-Port CE, Ye Q, Mazurkiewicz JE, Higgins PJ. TGF-beta1 + EGF-initiatedinvasive potential in transformed human keratinocytes is coupled to a plas-min/MMP-10/MMP-1-dependent collagen remodeling axis: role for PAI-1.Cancer Res 2009;69:4081–91.

[10] Guo YS, Chen Z, Wen XD, Ko TC, Townsend Jr CM, Hellmich MR. Synergisticregulation of COX-2 expression by bombesin and transforming growth factor-beta. Dig Dis Sci 2008;53:2045–52.

[11] Neil JR, Johnson KM, Nemenoff RA, Schiemann WP. Cox-2 inactivates smadsignaling and enhances EMT stimulated by TGF-beta through a PGE2-depen-dent mechanisms. Carcinogenesis 2008;29:2227–35.

[12] Wang MT, Honn KV, Nie D. Cyclooxygenases, prostanoids, and tumor progres-sion. Cancer Metastasis Rev 2007;26:525–34.

[13] Greenhough A, Smartt HJ, Moore AE, Roberts HR, Williams AC, Paraskeva C,et al. The COX-2/PGE2 pathway: key roles in the hallmarks of cancer andadaptation to the tumour microenvironment. Carcinogenesis 2009;30:377–86.

[14] Vaheri A, Enzerink A, Rasanen K, Salmenpera P, Nemosis. A novel way offibroblast activation, in inflammation and cancer. Exp Cell Res 2009;315:1633–8.

[15] Rasanen K, Salmenpera P, Baumann M, Virtanen I, Vaheri A. Nemosis offibroblasts is inhibited by benign HaCaT keratinocytes but promoted bymalignant HaCaT cells. Mol Oncol 2008;2:340–8.

[16] Rasanen K, Vaheri A. Proliferation and motility of HaCaT keratinocyte deri-vatives is enhanced by fibroblast nemosis. Exp Cell Res, in press, doi:10.1016/j.yexcr.2010.01.020.

[17] Virtanen I, Miettinen M, Lehto VP, Kariniemi AL, Paasivuo R. Diagnosticapplication of monoclonal antibodies to intermediate filaments. Ann NY AcadSci 1985;455:635–48.

[18] Vartio T, Laitinen L, Narvanen O, Cutolo M, Thornell LE, Zardi L, et al. Differ-ential expression of the ED sequence-containing form of cellular fibronectin inembryonic and adult human tissues. J Cell Sci 1987;88:419–30.

[19] Geback T, Schulz MM, Koumoutsakos P, Detmar M. TScratch: a novel andsimple software tool for automated analysis of monolayer wound healingassays. BioTechniques 2009;46:265–74.

[20] Davies M, Paterson IC, Ganapathy A, Prime SS. Cell death induced by N-(4-hydroxyphenyl) retinamide in human epidermal keratinocytes is modulatedby TGF-beta and diminishes during the progression of squamous cell carci-noma. Int J Cancer 2006;119:2803–11.

[21] Funderburgh JL, Funderburgh ML, Mann MM, Corpuz L, Roth MR. Proteoglycanexpression during transforming growth factor b-induced keratocyte-myofi-broblast transdifferentiation. J Biol Chem 2001;276:44173–8.

[22] Herfs M, Hubert P, Kholod N, Caberg JH, Gilles C, Berx G, et al. Transforminggrowth factor-beta1-mediated Slug and Snail transcription factor up-regula-tion reduces the density of Langerhans cells in epithelial metaplasia byaffecting E-cadherin expression. Am J Pathol 2008;17:1391–402.

[23] Zavadil J, Bottinger EP. TGF-beta and epithelial-to-mesenchymal transitions.Oncogene 2005;24:5764–74.

[24] Massague J. TGF-beta signal transduction. Annu Rev Biochem 1998;67:753–91.

[25] Li AG, Lu SL, Han G, Hoot KE, Wang XJ. Role of TGFbeta in skin inflammationand carcinogenesis. Mol Carcinog 2006;45:389–96.

[26] Oft M, Akhurst RJ, Balmain A. Metastasis is driven by sequential elevation of H-ras and Smad2 levels. Nat Cell Biol 2002;4:487–94.

[27] Datto MB, Li Y, Panus JF, Howe DJ, Xiong Y, Wang XF. Transforming growthfactor beta induces the cyclin-dependent kinase inhibitor p21 through a p53-independent mechanism. Proc Natl Acad Sci USA 1995;92:5545–9.

[28] Pietenpol JA, Holt JT, Stein RW, Moses HL. Transforming growth factor beta 1suppression of c-myc gene transcription: role in inhibition of keratinocyteproliferation. Proc Natl Acad Sci USA 1990;87:3758–62.

[29] Cerezo A, Kalthoff H, Schuermann M, Schafer B, Boukamp P. Dual regulation oftelomerase activity through c-myc-dependent inhibition and alternative spli-cing of hTERT. J Cell Sci 2002;115:1305–12.

[30] Derynck R, Akhurst RJ, Balmain A. TGF-beta signaling in tumor suppressionand cancer progression. Nat Genet 2001;29:117–29.

[31] Missero C, Ramon y Cajal S, Dotto GP. Escape from transforming growth factorbeta control and oncogene cooperation in skin tumor development. Proc NatlAcad Sci USA 1991;88:9613–7.

[32] Gold LI, Jussila T, Fusenig NE, Stenback F. TGF-beta isoforms are differentiallyexpressed in increasing malignant grades of HaCaT keratinocytes, suggestingseparate roles in skin carcinogenesis. J Pathol 2000;190:579–88.

[33] Game SM, Huelsen A, Patel V, Donnelly M, Yeudall WA, Stone A, et al.Progressive abrogation of TGF-beta 1 and EGF growth control is associatedwith tumour progression in ras-transfected human keratinocytes. Int J Cancer1992;52:461–70.

[34] Xie L, Law BK, Chytil AM, Brown KA, Aakre ME, Moses HL. Activation of the Erkpathway is required for TGF-b1-induced EMT in vitro. Neoplasia 2004;6:603–10.

K. Rasanen, A. Vaheri / Journal of Dermatological Science 58 (2010) 97–104104

[35] Han G, Lu SL, Li AG, He W, Corless CL, Kulesz-Martin M, et al. Distinctmechanisms of TGF-beta1-mediated epithelial-to-mesenchymal transitionand metastasis during skin carcinogenesis. J Clin Invest 2005;115:1714–23.

[36] Ikuta T, Kawajiri K. Zinc finger transcription factor slug is a novel target gene ofaryl hydrocarbon receptor. Exp Cell Res 2006;312:3585–94.

[37] Tomakidi P, Stark HJ, Herold-Mende C, Bosch FX, Steinbauer H, Fusenig NE,et al. Discriminating expression of differentiation markers evolves in trans-plants of benign and malignant human skin keratinocytes through stromalinteractions. J Pathol 2003;200:298–307.

[38] Szell M, Bata-Csorgo Z, Koreck A, Pivarcsi A, Polyanka H, Szeg C, et al. Pro-liferating keratinocytes are putative sources of the psoriasis susceptibility-related EDA+ (extra domain A of fibronectin) oncofetal fibronectin. J InvestDermatol 2004;123:537–46.

[39] Thiery JP. Epithelial-mesenchymal transitions in tumour progression. Nat RevCancer 2002;2:442–54.

[40] Bachmeier BE, Boukamp P, Lichtinghagen R, Fusenig NE, Fink E. Matrixmetalloproteinases-2,-3,-7,-9 and-10, but not MMP-11, are differentiallyexpressed in normal, benign tumorigenic and malignant human keratinocytecell lines. Biol Chem 2000;381:497–507.

[41] Broom OJ, Massoumi R, Sjolander A. Alpha2beta1 integrin signalling enhancescyclooxygenase-2 expression in intestinal epithelial cells. J Cell Physiol2006;209:950–8.

[42] Takaoka K, Kishimoto H, Segawa E, Hashitani S, Zushi Y, Noguchi K, et al.Elevated cell migration, invasion and tumorigenicity in human KB carcinomacells transfected with COX-2 cDNA. Int J Oncol 2006;29:1095–101.

[43] Lee EJ, Choi EM, Kim SR, Park JH, Kim H, Ha KS, et al. Cyclooxygenase-2promotes cell proliferation, migration and invasion in U2OS human osteo-sarcoma cells. Exp Mol Med 2007;39:469–76.

[44] Banu SK, Lee J, Speights Jr VO, Starzinski-Powitz A, Arosh JA. Cyclooxygenase-2regulates survival, migration, and invasion of human endometriotic cellsthrough multiple mechanisms. Endocrinology 2008;149:1180–9.

[45] Johansson N, Ala-aho R, Uitto V, Grenman R, Fusenig NE, Lopez-Otin C, et al.Expression of collagenase-3 (MMP-13) and collagenase-1 (MMP-1) by trans-formed keratinocytes is dependent on the activity of p38 mitogen-activatedprotein kinase. J Cell Sci 2000;113(Pt 2):227–35.

[46] Colotta F, Allavena P, Sica A, Garlanda C, Mantovani A. Cancer-related inflam-mation, the seventh hallmark of cancer: links to genetic instability. Carcino-genesis 2009;30:1073–81.