transforming growth factor- 1 increases survival of human ......effects on the immune system may...

[CANCER RESEARCH 61, 8306–8316, November 15, 2001]

Transforming Growth Factor-�1 Increases Survival of Human Melanoma throughStroma Remodeling1

Carola Berking, Richelle Takemoto, Helmut Schaider, Louise Showe, Kapaettu Satyamoorthy, Paul Robbins, andMeenhard Herlyn2

The Wistar Institute, Philadelphia, Pennsylvania 19104 [C. B., R. T., H. S., L. S., K. S., M. H.], and the University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania15261 [P. R.]

ABSTRACT

Transforming growth factor (TGF)-� is growth inhibitory for normalepithelial cells and melanocytes but can stimulate mesenchymal cells.Resistance to its inhibitory effects is characteristic of human melanoma,the growth of which may instead be promoted by TGF-�, because itsproduction is increased with melanoma progression. Whether TGF-� hasan autocrine function for melanoma cells or is important for paracrinestimulation of the tumor stroma is not known. In this study, TGF-�1 wasexpressed in melanoma cells via adenoviral gene transfer, and tumorgrowth was analyzed in vitro, in human skin grafts, and in mixtures withfibroblasts that were injected s.c. into immunodeficient mice. The TGF-�1produced by the melanoma cells activated the fibroblasts to producematrix within and around the tumor mass, whereas control tumorsshowed less stroma and more cell death. High expression of collagen,fibronectin, tenascin, and �2 integrin was detected in the TGF-�1-expressing tumors by immunohistochemistry. Number and size of lungmetastases were significantly increased. cDNA expression array analysisof TGF-�1-transduced fibroblasts embedded in type I collagen and ofTGF-�1-transduced melanoma cells demonstrated induction of types XV,XVIII, and VI collagens, tenascin, plasminogen activator inhibitor-I, vas-cular endothelial growth factor, cysteine-rich fibroblast growth factorreceptor-1, and platelet-derived growth factor receptor-�, which could belinked to promotion of growth and survival in melanoma. These datasuggest that remodeling of the neighboring stroma, which provides asupporting scaffolding and a positive feedback stimulation of tumorgrowth, is an important function of TGF-�1 in melanoma.

INTRODUCTION

TGF3-� is an almost ubiquitously expressed protein with diversefunctions in embryogenesis and adult tissue homeostasis. There existthree isoforms of TGF-� (TGF-�1, TGF-�2, and TGF-�3) in mam-mals with 75–80% homology, which arise from proteolytic cleavageof longer precursors (1, 2). Mature biologically active TGF-� resultsfrom dissociation of the latent inactive TGF-� complex, which can bestored in the ECM. TGF-� antagonizes the mitogenic activities ofmany other growth factors by interfering with cell cycle progressionand is the most potent growth inhibitor known for epithelial cells andcells of the immune system. On the other hand, TGF-� can stimulatemesenchymal cells, such as fibroblasts, smooth muscle cells, andchondrocytes, and induce the synthesis of proteins found in the ECM

including collagens, fibronectin, tenascin, thrombospondins, os-teopontin, osteonectin, and elastin (1, 3, 4). Simultaneously, TGF-�can reduce the synthesis of proteases, such as collagenases, andincrease the synthesis of protease inhibitors, such as TIMP-1 andPAI-I (5).

Carcinoma cells of the breast, prostate, lung, and colon produceTGF-� and are resistant to its growth-inhibitory effects, in contrastwith their benign precursor cells (6, 7). It has been hypothesized thatat these advanced stages of transformation, TGF-� can act as a tumorpromoter. Likewise in melanoma, expression of all three isoforms ofTGF-� has been found in malignant cells in culture (8) and in situ(9–11), and an association with progression has been proposed. Bothhigh-affinity receptors for TGF-�, which form a heteromeric complexupon activation and transmit signals to the cytoplasmic SMAD pro-teins (2), are expressed in melanoma (12, 13). Upon exogenousTGF-� stimulation, melanoma cells display various degrees of resist-ance to TGF-�-induced inhibition of DNA synthesis, whereas mela-nocytes are highly sensitive (8, 14). However, in contrast with whathas been found in some other TGF-�-resistant carcinomas, no inac-tivating mutations in the TGF-� receptor system or of the SMADsignaling cascade have been detected in melanoma, suggesting addi-tional mechanisms for resistance to the growth-inhibitory functions ofTGF-� in this tumor type (15).

The biological benefits for melanoma cells to constitutively pro-duce TGF-� remain unclear. An autocrine TGF-�-mediated up-regu-lation of integrins and matrix metalloproteinase-9 as well as a down-regulation of E-cadherin has been described and may facilitatemelanoma cell migration (16) and adhesion to the endothelium (13).Paracrine effects of TGF-� on host cells in the tumor microenviron-ment may also be advantageous for melanoma cells. Suppressiveeffects on the immune system may allow tumor cells to escape fromimmune surveillance (17, 18), and angiogenic properties of TGF-�could support nutrition of the tumor and facilitate metastasis (3, 19).In addition, stimulation of stromal cells by TGF-� could lead toincreased production of reciprocally paracrine-acting growth factorsand to ECM production, which could, in turn, provide a scaffoldingfor melanoma cells to adhere and migrate. Finally, the modulation ofproteases and their inhibitors by TGF-� could facilitate remodeling ofthe stroma and invasion (20–22).

In this report, we provide evidence that melanoma cells can mod-ulate their surrounding stroma for their own benefits through theparacrine activity of TGF-�1. Stimulation of production of ECMproteins by stromal fibroblasts provided a scaffolding for the mela-noma cells, which showed increased survival and metastasis forma-tion compared with the controls. Global gene expression analyses ofTGF-�1-expressing melanoma cells and fibroblasts in organotypicculture indicated a complex interplay between matrix proteins, adhe-sion molecules, and growth factors, providing an optimal environmentfor tumor growth and progression.

MATERIALS AND METHODS

Cell Culture. Normal human keratinocytes and melanocytes were isolatedfrom the epidermis, and fibroblasts were isolated from the dermis of neonatal

Received 6/8/01; accepted 9/19/01.The costs of publication of this article were defrayed in part by the payment of page

charges. This article must therefore be hereby marked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.

1 Supported by NIH Grants CA80999, CA25874, and CA10815 (to M. H.) and apostdoctoral research fellowship BE2189/1-1 from the Deutsche Forschungsgemeinschaft(to C. B.).

2 To whom requests for reprints should be addressed, at The Wistar Institute, 3601Spruce Street, Philadelphia, PA 19104. Phone: (215) 898-3950; Fax: (215) 898-0980;E-mail: [email protected].

3 The abbreviations used are: TGF, transforming growth factor; ECM, extracellularmatrix; PAI, plasminogen activator inhibitor; TIMP, tissue inhibitor of metalloproteinase;VEGF, vascular endothelial growth factor; CFR, cysteine-rich fibroblast growth factorreceptor; PDGF, platelet-derived growth factor; PDGFR, PDGF receptor; SFM, serum-free medium; FBS, fetal bovine serum; SCID, severe combined immunodeficient;TUNEL, terminal deoxynucleotidyltransferase-mediated nick end labeling; Ad5, adeno-virus serotype 5; pfu, plaque-forming unit; VGP, vertical growth phase; RGP, radialgrowth phase.

8306

on June 7, 2021. © 2001 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

http://cancerres.aacrjournals.org/

human foreskins. Keratinocytes were cultured in SFM (Life Technologies,Inc., Rockville, MD) supplemented with human recombinant epidermal growthfactor and bovine pituitary extract. Melanocytes were cultured in MCDB153(Sigma Chemical Co., St. Louis, MO) supplemented with 2% FBS, 10%chelated FBS, 2 mM glutamine (Mediatech, Herndon, VA), 20 pM choleratoxin(Sigma Chemical Co.), 150 pM recombinant human basic fibroblast growthfactor, 100 nM endothelin-3 peptide (Peninsula, Belmont, CA), and 10 ng/mlrecombinant human stem cell factor (R&D Systems, Minneapolis, MN). Fi-broblasts and human embryonic kidney-derived 293 cells used for adenovirusreplication were cultured in DMEM with glutamine (Life Technologies, Inc.)and 10% FBS (Hyclone, Logan, UT). Human primary and metastatic mela-noma cells were isolated from clinically and histologically defined lesions andcultured as described (23, 24). They were maintained in MCDB153 with 20%Leibovitz’s L-15 medium (Life Technologies, Inc.), 2% FBS, and 5 �g/mlinsulin (Sigma Chemical Co.).

Growth Factor Detection. Melanoma cells, melanocytes, and fibroblastswere plated in six-well plates at 1.5 � 105 cells/well in their culture mediumfor 24 h. They were transduced with TGF-�1 or LacZ via adenoviral vectors,washed with SFM after 24 h, and cultured in SFM for another 48 h. Cellsupernatants were then analyzed for the presence of TGF-�1 using a humanTGF-�1 immunoassay (R&D Systems) that uses TGF-� soluble receptor typeII, which binds TGF-�1, as the coating reagent and an enzyme-linked poly-clonal antibody to TGF-�1 as the second reagent. The procedure followed themanufacturer’s instructions. The measurements represent the total of bothactive and latent forms of TGF-�1 and are given in pg/105 cells. For analysisof VEGF protein, an ELISA was used according to the manufacturer’s instruc-tions (R&D Systems). Melanoma cells and fibroblasts were transduced withTGF-�1 or LacZ control, respectively, and conditioned medium was collected72 h later. Samples were frozen at �70°C and analyzed within the following1–2 days. Results are expressed as mean � SD ng/ml per 106 cells. Allexperiments were performed in duplicates.

[3H]Thymidine Incorporation Assay. Melanoma cells, fibroblasts, andmelanocytes were seeded at 2–4 � 104 cells/well in 96-well plates andinfected with adenoviral vectors. After 2 days, 1 �Ci of [3H]thymidine wasadded per well, and 18 h later, cells were harvested, and the activity wascounted with a beta counter. Experiments were performed twice and intriplicates. Results are expressed as average difference (in percentages) � SDcompared with the respective LacZ-transduced control cells.

Adenoviral Vectors for TGF-�1 and LacZ. The adenoviral vector TGF-�1/Ad5 carrying the gene for the TGF-�1 protein has been described (25). Thecontrol adenoviral vector LacZ/Ad5 (Vector Core; University of Pennsylvania,Philadelphia, PA) induces expression of the reporter gene �-galactosidase fromEscherichia coli. The vectors were prepared, purified, and titered to1–5 � 1010 pfu/ml.

Melanoma cells were infected with 2 or 20 pfu/cell, and fibroblasts wereinfected with 40 pfu/cell in serum-free base medium for 3–4 h. Medium wasthen changed to growth medium, and cells were used the next day forexperiments.

Human Skin Grafting. Human foreskins from newborns were kept insterile transport media (HBSS supplemented with antibiotics) and graftedwithin 48 h of excision as described (26). Female and male CB-17 SCID micewere bred at the Animal Facility of the Wistar Institute and housed underpathogen-free conditions in groups of up to five animals/isolator cage. Graftswere well healed after 4–6 weeks, and mice were then used for the experi-ments. Mice received injections intradermally with the adenoviral vectorsusing a 26-gauge needle at a concentration of 0.5–5 � 108 pfu in a total volumeof 100 �l of sterile PBS. Melanoma cells (2–5 � 106 cells) were injectedintradermally with a 23-gauge needle in 100 �l of cell culture medium. TheWistar Institutional Animal Care and Use Committee approved all protocols.

Tumor Growth in Vivo. For in vivo growth studies, 451Lu melanoma cellswere mixed with normal human fibroblasts in 100 �l of medium and 100 �lof Matrigel matrix (Collaborative Biomedical Products, Bedford, MA) andinjected s.c. into SCID mice with a 23-gauge needle. One day before injection,melanoma cells were transduced with TGF-�1 or LacZ using adenoviralvectors, respectively, at an infection dose of 20 pfu/cell. The total injected cellnumber per mouse was 2.2 � 106 to 4 � 106 cells after mixing melanoma cellswith fibroblasts at a ratio of 1:10 (2 � 105 melanoma cells and 2 � 106

fibroblasts), 1:1 (2 � 106 and 2 � 106 cells), and 10:1 (2 � 106 and 2 � 105

cells). Six mice/group received injections and were sacrificed after 2 weeks,

and 10 additional mice that received injections of TGF-�1- or LacZ-transducedmelanoma cells mixed with fibroblasts at a ratio of 1:1, respectively, weresacrificed after 5.5 weeks. Tumor growth was monitored twice weekly.

Western Blot. 451Lu tumors were harvested from the mouse and mincedwith RIPA buffer (TGF-�1) containing protease inhibitors (1 mM phenylmeth-ylsulfonyl fluoride, 10 �g/ml aprotinin, 10 �g/ml leupeptin, and 1.8 mg/mliodoacetamide). After centrifugation at 4°C for 20 min at 12000 � g, proteinin the supernatants was quantified using the BCA kit (Pierce, Rockford, IL).Equal amounts of total protein from each sample were resolved in a 15%SDS-polyacrylamide gel, electroblotted onto a polyvinylidene difluoride mem-brane (Bio-Rad Laboratories, Richmond, CA), and blocked with a 5% solutionof dry milk and 0.05% Tween 20 in PBS at room temperature for 1 h. Themembrane was incubated with rabbit polyclonal antihuman-TGF-�1 antibody(Santa Cruz Biotechnology, Santa Cruz, CA) followed by peroxidase-labeledsecondary antibody (Jackson ImmunoResearch Laboratories, Inc., WestGrove, PA). Immunoreactive bands were developed using the ECL detectionsystem (Amersham, Arlington Heights, IL) and exposed to Kodak Biomax film(Eastman Kodak, Rochester, NY).

Histology and Immunohistochemistry. At the end of each experiment,mice were sacrificed by CO2 inhalation, and skin grafts or s.c. tumors wereexcised. Half of the samples were fixed in 10% neutral-buffered formalin(Fisher Scientific, Pittsburgh, PA) for 6–12 h at room temperature and em-bedded in paraffin. The other half was dehydrated by increasing concentrationsof sucrose solutions (5, 10, and 20%) at 4°C overnight, embedded in OCTmedium (Miles, Elkhart, IN), snap-frozen, and stored at �70°C until cryosec-tioning at 6–8 �m. Formalin-fixed sections were stained with H&E forhistopathological evaluation. Masson’s trichrome stain was used for estimationof the amount and distribution of collagen in the tissues. The DNA-bindingfluorochrome Hoechst 33258 (Sigma Chemical Co.) was used to distinguishhuman from murine cells.

Immunohistochemistry was performed on serial sections using an avidin-biotin-peroxidase system kit (Vector Laboratories, Burlingame, CA) and 3,3�-diaminobenzidine tetrahydrochloride (Sigma Chemical Co.) or 3-amino-9-ethylcarbazole (Vector) as chromogens. Antigens in the formalin-fixed tissueswere retrieved by trypsin digestion at 37°C or microwave heat treatment incitrate buffer. Cryostat sections of 6–8 �m were air-dried and fixed in ice-coldacetone for 10 min. Prior to incubation with the primary antibodies in ahumidified chamber at 4°C overnight or at room temperature for 1–2 h,nonspecific binding was blocked with 10% normal horse or 10% normal goatserum. Primary mouse monoclonal antibodies against the following humanantigens were used: Ki-67 (Immunotech, Westbrook, ME); HMB45 (Bio-genex, San Ramon, CA); type IV collagen (hybridoma from ATCC, Manassas,VA); fibronectin (American Type Culture Collection); tenascin (27); �2 inte-grin (Chemicon, Temecula, CA); �3 integrin (28); smooth muscle actin(Zymed, South San Francisco, CA); aminopeptidase N (29); and CD31 (PE-CAM; Dako Carpinteria, CA). Primary rabbit polyclonal antibodies used inthis study were rabbit anticow S100 (Dako) and rabbit antihuman TGF-�1(Santa Cruz Biotechnology). Mouse IgG1 isotype antibody (P3) was used asnegative control for each staining with mouse monoclonal antibodies and arabbit antihuman involucrin antibody (Biomedical Technologies, Stoughton,MA) as negative control for each staining with rabbit polyclonal antibodies.Between each incubation step, slides were rinsed twice in PBS for 3–5 min.Endogenous peroxidase was quenched with 3% H2O2 in methanol for 20–30min at room temperature. A biotin-labeled antimouse secondary antibody wasapplied for 30 min at room temperature, followed by incubation with apreformed avidin-biotinylated enzyme complex for 30 min. After color devel-opment by addition of the chromogen and counterstaining with Mayer’shematoxylin (Sigma Chemical Co.), sections were mounted and evaluatedunder a light microscope.

TUNEL. Detection of apoptosis was done on formalin-fixed, paraffin-embedded tumor sections with a commercially available TUNEL kit (Boeh-ringer Mannheim, Indianapolis, IN) according to the manufacturer’s directionswith modifications. Briefly, after deparaffinization, rehydration, and quench-ing of endogenous peroxidase, sections were treated with 0.1% Triton X-100(Sigma Chemical Co.) and 0.1% citrate buffer for cell permeabilization. Afterincubation with terminal deoxynucleotidyl transferase and fluorescein-labelednucleotides for 60 min at 37°C and incubation with peroxidase-conjugatedanti-fluorescein antibody Fab fragments for 40 min at 37°C, 3-amino-9-ethylcarbazole substrate and H2O2 were added for color development. Coun-

8307

TGF-�1 IN MELANOMA



terstain was done with Mayer’s hematoxylin. Positive labeled cells were scoredin a blinded manner in randomly chosen fields at �200.

Skin Reconstruction. Skin reconstructs were prepared essentially as de-scribed with modifications (30). Human fibroblasts (FF2441) were added toneutralized bovine type I collagen (Organogenesis, Canton, MA) to a finalconcentration of 0.8–1 mg/ml of collagen in MEM (BioWhittaker, Walkers-ville, MA), 1.66 mM L-glutamine (Life Technologies, Inc.), 10% FBS, and0.21% sodium bicarbonate (BioWhittaker). Three ml of fibroblast-containingcollagen (2.5 � 104 cells/ml) were added to each insert of a six-well tissue-culture tray (Organogenesis) after precoating with 1 ml of acellular collagen.Mixtures were allowed to constrict in DMEM with 10% FBS for 5–7 days. Theday before seeding, melanoma cells were infected with TGF-�1/Ad5, andcontrols were infected with LacZ/Ad5 at 20 pfu/cell for 4 h in protein-freeSFM and then incubated overnight in complete SFM. Keratinocytes weremixed with melanoma cells at a ratio of 5:1 to 10:1 in low-calcium epidermalgrowth medium containing DMEM, F-12 Ham’s (Life Technologies, Inc.), 1%newborn calf serum (Hyclone), 4 mM glutamine, 1.48 � 10�6 M hydrocorti-sone, 4 pM progesterone, 20 pM triiodothyronine, 0.1 mM O-phosphoryleth-anolamine, 0.18 mM adenine (Sigma Chemical Co.), 5 mg/ml insulin, 5 mg/mltransferrin, 5 mM ethanolamine, 5 g/ml selenium (BioWhittaker), and 50 �g/mlgentamicin (Mediatech, Herndon, VA). A total of 5–6 � 105 cells was seededon each contracted collagen gel. Cultures were maintained submerged in lowcalcium growth medium for 2 days and in normal calcium (1.88 mM) growthmedium for another 2 days and then raised to the air-liquid interface for 10–12days with feeding from below with normal calcium and high-serum (20%)medium.

RNA Preparation and Labeling. Total RNA was isolated with TrizolReagent (Life Technologies, Inc.) according to the manufacturer’s protocol.After DNase I treatment (Roche Diagnostics, Mannheim, Germany) for 15 minat 37°C and ethanol precipitation, the RNA quality and quantity was visualizedon a 1% agarose gel with ethidium bromide. One to 2 �g of total RNA werereverse transcribed with 300 units of Superscript II reverse transcriptase (LifeTechnologies, Inc.) in the presence of 2 �g of oligo(dT)15 primer (PromegaCorp., Madison, WI), 1 �l of 10 � decamers (Ambion, Austin, TX), 1 mMdATP, dGTP, and dTTP (Amersham Pharmacia, Piscataway, NJ), respectively,0.1 mCi of [33P]dCTP (ICN Biomedicals, Costa Mesa, CA), and 3.5 mM DTTat 39°C for 90 min. The labeled cDNA targets were separated from theunincorporated [33P]dCTP through Sephadex G-50 Quick Spin columns(Roche Diagnostics, Indianapolis, IN) and denatured at 100°C before hybrid-ization to the filter arrays. [33P]CTP incorporation was quantitated by scintil-lation counting.

cDNA Expression Array. Human arrays from the Genomics Core of theWistar Institute were used. Each array consisted of a nylon membrane(2.5 � 7.5 cm) spotted with 200–600-bp cDNA fragments from sequence-validated human clones (Research Genetics, Huntsville, AL) representing 2280different genes, 9 housekeeping genes, and negative controls. Arrays wereproduced with a GM417 array station (Genetic MicroSystems, Bedford, MA)using 300-�m pins with 750-�m center-to-center spacing.

The labeled cDNA targets were hybridized to the arrays in Church bufferwith boiled and chilled Cot-1 DNA and sheared salmon sperm DNA inplastic hybridization bags at 65°C for 18 h. The membranes were washedat 65°C in 2� SSC/1% lauryl sulfate sodium salt (SDS) for 30 min 3 timesin 0.1� SSC/0.5% SDS for 30 min once and then exposed to a Phosphorscreen (Molecular Dynamics, Sunnyvale, CA) for 24 h to 5 days. Duplicatehybridizations were performed with RNA from two independent experi-ments. Sequential filters from the same printing were used for the analyses.Phosphorscreens were scanned with a Storm Phosphorimager (MolecularDynamics) at a resolution of 50 �m. Imagequant files from scans wereimported into ArrayVision (Imaging Research, St. Catharine, Ontario,Canada) for quantification. Spot intensities were background subtracted,globally normalized, and reported as median pixel densities by the WistarGenomics Core.

Statistics. For ELISA, TUNEL assay, proliferation, and tumor growthexperiments, the arithmetic mean and SDs were calculated. Statistical differ-ences to LacZ-treated controls were validated by the two-sided Student’s t test.P � 0.05 was considered significant.

RESULTS

Constitutive TGF-�1 Production in Melanoma and Sensitivityto Induced TGF-�1 Expression. Constitutive TGF-�1 production in21 human melanoma cell lines from different progression stagesranged between 0 and 248 pg/105 cells with an average of 49 � 70pg/105 cells as measured by ELISA (Fig. 1A). There was a tendencyof higher TGF-�1 levels in late progression stages but not exclusively.More than 40 pg of TGF-�1 per 105 cells were found in 3 of 7 (43%)metastatic melanoma lines, 2 of 9 (22%) primary VGP, and 1 of 5(20%) primary RGP melanoma lines, whereas �5 pg/105 cells werefound in 1 of 7 (14%) metastatic lines, 3 of 9 (33%) primary VGP, and3 of 5 (60%) primary RGP melanoma lines. Metastatic lines WM373,WM1617, and 1205Lu secreted higher levels of TGF-�1 than theirprimary melanoma counterparts WM75, WM278, and WM793 estab-lished from the same patient, respectively. The metastatic lineWM239A, on the contrary, showed little or no TGF-�1 production(2 � 2 pg/105 cells), whereas the respective primary tumor lineWM115 produced 36 � 35 pg/105 cells. Normal human melanocytesproduced negligible levels of TGF-�1 (2 pg/105 cells), whereas neo-natal foreskin fibroblasts produced 64 pg/105 cells.

The sensitivity of 21 melanoma cell lines to TGF-�1 was analyzedafter transduction with an adenoviral vector expressing TGF-�1 (Fig.1B). Cell proliferation was measured by [3H]thymidine incorporation

Fig. 1. Constitutive production of TGF-�1 and proliferative response to transductionwith TGF-�1. A, TGF-�1 production in 21 human melanoma cell lines, in foreskinmelanocytes (FM), and in foreskin fibroblasts (FF) analyzed by ELISA of cell culturesupernatants 72 h after seeding. Progression stage decreases from left to right. No., normalcells. Bars, SD. B, growth of the same melanoma cell lines as above after transductionwith TGF-�1. Given is the percentage relative to the [3H]thymidine uptake after trans-duction of each respective line with LacZ control. Bars, SD.

8308




and compared with LacZ control vector-transduced cells. Resistanceor a �30% inhibition upon TGF-�1 transduction was found in 12 of21 lines (57%). Inhibition defined as a �30% reduction comparedwith the controls was observed in 7 of 21 lines (33%), and stimulationwas found in 2 of 21 lines (10%). Melanocytes were inhibited, andfibroblasts were resistant. In general, cell lines with high constitutiveTGF-�1 levels were resistant to TGF-�1 transduction, and cell lineswith low or no endogenous production of TGF-�1 showed eitherstimulation, resistance, or inhibition. The strongest inhibition wasobserved for WM239A, which was found to be the lowest TGF-�1producer among the metastatic lines analyzed.

The dose-dependent protein production of TGF-�1 after adenoviraltransduction was analyzed in six selected melanoma cell lines byELISA (Fig. 2A). An infection dose of 2 pfu/cell led to a 1.2–4-foldincrease in TGF-�1 protein production in five of six cell lines and a153-fold increase in cell line WM239A. An infection dose of 20 pfuled to a 9–23-fold increase in TGF-�1 protein production in four ofsix cell lines and an 115-fold and 333-fold increase in the cell lines1205Lu and WM239A, respectively. This dose was therefore used inthe following in vivo studies.

The effect of TGF-�1 on cell morphology after adenoviral trans-duction was tested in 24 melanoma cell lines. Three of the 24 celllines, WM1552c, WM793, and 1205Lu, which had an epitheloidmorphology, displayed a more spindle-shaped, fibroblastoid morphol-ogy starting 3 days after transduction (Fig. 2B). It should be noted thatthe VGP primary melanoma cell line WM793 and the metastaticmelanoma cell line 1205Lu were derived from the same patient (31).

Stroma Activation and Decreased Tumor Cell Death throughTGF-�1 in Melanoma. To analyze melanoma-stroma interactionsmediated by TGF-�1, we used an in vivo model of close proximity of

human melanoma cells with fibroblasts. Melanoma cell line 451Lu(32) was chosen, because the cells showed low constitutive levels ofTGF-�1, were resistant to TGF-�1 transduction (see Fig. 1), produced23-fold higher levels of TGF-�1 than controls after transduction (seeFig. 2), and were highly tumorigenic in immunodeficient mice (33).After transduction with TGF-�1 in vitro, 451Lu melanoma cells weremixed with normal human skin fibroblasts in ratios of 1:10, 1:1, and10:1 and injected together with Matrigel matrix s.c. into SCID mice.After 2 weeks, solid tumors were palpable in all groups with nosignificant difference in tumor volume from controls (data notshown). Increased protein expression of TGF-�1 in tumors of TGF-�1-transduced melanoma cells was confirmed by Western blot anal-ysis (Fig. 3). Histologically, thick stroma septae were found through-out and around the TGF-�1-expressing tumors (Fig. 4A), whereasthere was only little stroma in the control tumors (Fig. 4B). Controltumors were characterized by a higher proportion of necrosis withblood extravasation (Fig. 4B), and TUNEL assay demonstrated asignificantly higher number of apoptotic cells (Fig. 4D) comparedwith the TGF-�1-expressing tumors (Fig. 4C). Decreased cell death inmelanoma by TGF-�1 was also observed in organotypic cultures.Melanoma cell line WM793, which showed low constitutive levels ofTGF-�1 and growth resistance as well as spindle-shaped morphologyafter TGF-�1 transduction, was incorporated into human skin recon-structs after TGF-�1 and LacZ transduction, respectively (Fig. 4, Eand F). In this organotypic culture model, fibroblasts embedded intype I collagen form the dermis, and keratinocytes seeded on top forma stratified epithelium, i.e., epidermis. After 2 weeks, melanoma cellclusters were found in the epidermis and upper dermis in both groups;however, much less cell death was observed in the tumors formed byTGF-�1-transduced melanoma cells (Fig. 4E) compared with thecontrols (Fig. 4F).

Paracrine stroma activation by TGF-�1 expression in melanomacells was also detected in an orthotopic melanoma model. Primarymelanoma cell line WM3248, which was stimulated by TGF-� trans-duction in vitro, was transduced with TGF-�1 or LacZ in vitro andthen injected intradermally into human foreskins grafted to SCIDmice. After 2 weeks, solid dermal tumors were palpable in both theTGF-�1 and LacZ groups without a significant difference in volumeuntil sacrifice 3–4 weeks after injection. A capsule-like stromal ma-trix had formed around the TGF-�1-expressing melanoma with in-creased numbers of fibroblasts (Fig. 4G) when compared with thecontrols (Fig. 4H). The stromal cells produced high levels of collagenas demonstrated by Masson’s trichrome stain (not shown), and stromacapsules directly surrounding tumor cell clusters contained type IVcollagen, which was not seen in the controls (not shown). Numerous

Fig. 3. Detection of TGF-�1 precursor forms by Western blot analysis of 2-week-old451Lu melanomas after transduction of melanoma cells with TGF-�1 or LacZ in vitro andinjection s.c. into SCID mice together with human fibroblasts in a ratio 1:1. Equal amountsof protein were loaded.

Fig. 2. Induction of TGF-�1 production by adenoviral gene transfer and morphologicaleffect on melanoma. A, TGF-�1 production in six different melanoma cell lines aftertransduction with TGF-�1 at 2 and 20 pfu, LacZ control vector (20 pfu), or withouttransduction. Bars, SD. B, morphology of 1205Lu melanoma cells 3 days after transduc-tion with LacZ control (A) or TGF-�1 (B). Note the TGF-�1-induced, elongated spindleshape of the cells.

8309




vessels were detected in this stroma; however, there seemed to be nosignificant difference in vascularity compared with the LacZ controls.

Induction of ECM Proteins by TGF-�1 in Melanoma. Thestroma activation by TGF-�1 produced by melanoma cells was furthercharacterized in 2-week-old tumors of TGF-�1-transduced 451Lumelanoma cells, which had been mixed with fibroblasts 1:1 and s.c.injected into SCID mice. Masson’s trichrome stain revealed an in-crease in collagens (Fig. 5A) when compared with LacZ controls (Fig.5B). Immunohistochemical analyses demonstrated an induction offibronectin (Fig. 5C) and �2 integrin (Fig. 5E) expression by TGF-�1,

whereas they were only weak or undetectable in the controls (Fig. 5,D and F). Tenascin expression was not stronger in intensity but morewidely distributed throughout the TGF-�1-expressing tumors (notshown). Detection of CD13 (aminopeptidase N) as a human fibroblastmarker revealed high positivity in the interstitium of the TGF-�1-expressing tumors (Fig. 5G) in contrast with the controls (Fig. 5H),suggesting the close proximity of human fibroblasts to the melanomacells.

The described remodeling of stroma was obviously mediated byparacrine effects of TGF-�1 produced by melanoma cells. Therefore,

Fig. 4. Stroma activation and decreased tumor celldeath through TGF-�1 in melanoma. A–D, 451Lumelanoma 19 days after s.c. injection into SCID mice.Before injection, melanoma cells were transduced withTGF-�1 (A and C) or LacZ (B and D) and mixedtogether with normal human fibroblasts in Matrigel ata ratio 1:1. A and B, increase in stroma around theTGF-�1-transduced (A) versus LacZ-transduced (B)melanoma cells (H&E, �100). The control tumors (B)show more necrosis and blood extravasation. C and D,TUNEL assay for evaluation of apoptosis (red) showsonly few positive cells in the TGF-�1-transduced mel-anoma cells (C) but high positivity in the controltumors (D; �100). E and F, histological section of16-day-old human skin reconstructs with normal hu-man fibroblasts, keratinocytes, and WM793 primarymelanoma cells (arrowheads) transduced withTGF-�1 (E) or LacZ control (F). Note the higherdegree of cell death in the control tumor (F; H&E,�200). G and H, WM3248 melanoma after transduc-tion with TGF-�1 (G) or LacZ control (H) and intra-dermal injection into human skin grafts. Shown is anH&E-stained histological section of a sample 3 weeksafter injection (�200). Note the capsule-formingstroma reaction around the tumor WM3248 transducedwith TGF-�1 (G).

8310




we tested direct expression of TGF-�1 in human skin by intradermalinjection of the adenoviral vectors into human skin xenografts, whichresults in a highly efficient gene transduction of fibroblasts in thedermis.4 As early as 1 week, a strong thickening of the skin wasvisible and palpable, whereas there was no change in LacZ-injected

skins. TGF-�1 production in the dermis led to the formation of astrong collagen fiber network, as determined by H&E (not shown) andMasson’s trichrome stain (Fig. 5I), which was not seen in the controls(Fig. 5J).

The TGF-�1/Ad5-treated animals suffered from systemic effects inthe second week after injection. They became weak and apathic andeventually died. Autopsy showed normal lung and liver tissue but4 C. J. Gruss, K. Satyamoorthy, C. Berking, J. Lininger, M. Nesbit, H. Schaider, Z-J.

Liu, M. Oka, M-Y. Hsu, T. Shirakawa, G. Li, P. Carmeliet, W. El-Deiry, S. L. Eck, J. S.Rao, A. H. Baker, J. Bennett, T. Crombleholme, J. Karmacharya, D. J. Margolis, J. M.Wilson, S. Werner, M. Detmar, M. Skobe, P. D. Robbins, C. Johnson, D. Carbone, C.Buck, and M. Herlyn. Re-modeling of the human skin architecture in vivo by adenovirus-

mediated gene transfer of growth factors, adhesion molecules, proteolytic enzymes,oncogenes and tumor suppressor genes, submitted for publication.

Fig. 5. Induction of ECM proteins by TGF-�1 inmelanoma. A–H, 451Lu melanoma 19 days after s.c.injection into SCID mice. Before injection, melanomacells were transduced with TGF-�1 (A, C, E, and G) orLacZ (B, D, F, and H) and mixed together with normalhuman fibroblasts in Matrigel at a ratio 1:1. A and B,Masson’s trichrome stain illustrates an increase incollagen (blue) induced by TGF-�1-expressing mela-noma cells (A) compared with the controls (B; �100).C and D, immunohistochemical detection of fibronec-tin (red) reveals an increase in the stroma septaearound the TGF-�1-transduced (C) compared with theLacZ-transduced (D) melanoma cells (�50). E and F,�2 integrin expression in the interstitium of TGF-�1-transduced melanoma cells (E) is not seen in the LacZ-transduced controls (F). G and H, immunohistochem-ical detection of human fibroblasts by CD13(aminopeptidase N; red) in the TGF-�1-transduced(G) compared with the LacZ-transduced (H) controlgroup (�100). I and J, human foreskin graft 10 daysafter intradermal injection of 5 � 108 pfu of TGF-�1/Ad5 (I) or LacZ/Ad5 control vector (J) in 100 �l ofsterile PBS. Masson’s trichrome stain reveals an in-crease in collagen in blue (I) compared with the con-trol (J; �100).

8311




myeloid metaplasia in the spleen and acute tubular necrosis in thekidney. Lethal effects of the TGF-�1/Ad5 treatment were also ob-served with 10-fold lower injection doses. The systemic serum levelsof circulating TGF-�1 were at ng/ml levels, as analyzed by ELISA.

Increased Metastasis Development by TGF-�1-transducedMelanoma Cells. For metastasis studies, SCID mice were injecteds.c. with TGF-�1- or LacZ-transduced 451Lu melanoma cells mixedwith normal human skin fibroblasts in Matrigel matrix in a ratio of 1:1and observed for 39–41 days. Tumor volume was not significantlydifferent between groups during the first 32 days. At day 39, TGF-

�1-transduced tumors were 1.7-fold larger than LacZ controls(1.46 � 0.95 cm3) versus 0.8 � 0.55 cm3). S100-positive microme-tastases in the lungs were found in 8 of 10 mice with TGF-�1-transduced melanomas (Fig. 6A) and in 7 of 10 mice with LacZ-transduced control melanomas (Fig. 6B). Metastases in 10 randomlychosen fields at �100 were counted in each lung. The average numberof micrometastases/microscopic lung field was significantly(P � 0.03) higher in the TGF-�1 group (4.1 � 4.7) compared with thecontrols (0.4 � 0.7), and the average size of each metastasis wasbigger as well. The average diameter of each metastasis was in the

Fig. 6. Increased metastasis formation by TGF-�1-transduced melanoma cells. A and B, S100-positivemicrometastases in lungs from SCID mice 39 daysafter s.c. injection of TGF-�1-transduced (A) or LacZ-transduced (B) 451Lu melanoma cells mixed in a ratio1:1 with normal human fibroblasts in Matrigel matrix.Number and size of metastases were significantly in-creased in the TGF-� group. Metastases stained posi-tive for TGF-�1 (C), whereas controls showed no oronly weak stain for TGF-�1 (D; �100). E and F,smooth muscle actin expression (red) in the lung invessel walls (arrowheads) and around melanoma me-tastases (arrows) after TGF-�1 (E) and LacZ (F) trans-duction of s.c.-injected 451Lu melanoma cells (�200).G and H, smooth muscle actin expression (immuno-histochemical detection in brown) is induced by injec-tion of adenoviral vectors for TGF-�1 (G) in humanskin grafts but not by injection of LacZ control vectors(H; �50).

8312




TGF-�1 group (85 � 44 �m), 1.8-fold larger than in the LacZ group(47 � 29 �m; Table 1). The metastases were found to expressTGF-�1 (Fig. 6C), whereas the controls showed no or only weakstaining for TGF-�1 (Fig. 6D). Smooth muscle actin, which wasdetected in the vessel walls of the murine lungs, was commonlyexpressed around the micrometastases in the TGF-�1 group (Fig. 6E)but only rarely in the control group (Fig. 6F).

Induction of smooth muscle actin expression was also observed infibroblasts in human skin grafts when injected with adenoviral vectorsfor TGF-�1 (Fig. 6G). This indicated that TGF-�1 can induce trans-differentiation of fibroblasts in the dermis into myofibroblasts, sug-gesting that they are also induced by TGF-�1-derived from melanomacells.

Gene Expression Profiling of TGF-�1-transduced Fibroblastsand Melanoma Cells. The mRNA expression of 2280 different genesin melanoma cells and fibroblasts after transduction with TGF-�1 wasanalyzed by cDNA microarray. Melanoma lines 1205Lu, 451Lu, andWM793 were tested 3 days after adenoviral infection with 20 pfu ofTGF-�1/Ad5 or LacZ/Ad5 per cell. At this harvesting time point,1205Lu and WM793 displayed in 40–50% of all cells a TGF-�1-induced fibroblastoid phenotype, which was not seen in the LacZ-transduced or nontransduced controls. Normal human foreskinfibroblasts were analyzed after adenoviral infection with 40 pfu ofTGF-�1/Ad5 or LacZ/Ad5 per cell and organotypic culture in type Icollagen for 4 days. Table 2 summarizes the expression results ofselected genes encoding for ECM proteins, adhesion receptors,growth factors, their binding proteins and receptors, as well as pro-tease inhibitors.5

Among the different collagen subtypes, expression of the �1 chainof type XVIII and type XV was increased up to 12-fold in bothmelanoma cells and fibroblasts. Tenascin and osteonectin expressionincreased 5- and 6-fold, respectively, in TGF-�1-transduced fibro-blasts, whereas it was unchanged or 2-fold reduced in melanoma cells.Integrins �1, �5, �V, �3, and �6 showed a 2–6-fold reduction inTGF-�1-transduced fibroblasts. In at least 2 melanoma cell lines, a

2–4-fold increase in �v and �3 integrin after TGF-�1 transductionwas detected, whereas other subtypes were not changed in more thanone of the three analyzed lines. VEGF-A increased after TGF-�1transduction up to 35-fold in all analyzed cell lines and was thereforethe growth factor that exhibited the strongest induction. The inductionof VEGF was confirmed at the protein level by ELISA (Fig. 7).Although WM793 melanoma cells and dermal fibroblasts werestrongly induced to secrete VEGF by TGF-�1 transduction, VEGFlevels in 451Lu melanoma cells were already relatively high in thecontrols (1.53 ng/ml/106 cells) and stayed at similar levels afterTGF-�1 transduction.

PDGF receptor-� expression was induced by TGF-�1 transduction5- and 21-fold in two of three analyzed melanoma lines, and CFR-1was increased up to 6-fold in TGF-�1-transduced melanoma cells and4-fold in fibroblasts. PA1-I was, with 14- to 43-fold increase inexpression in all analyzed TGF-�1-transduced cell lines, the highestinduced protease inhibitor.

DISCUSSION

Interactions of tumor cells with their microenvironment and theinfluence of stroma on tumor and vice versa have been increasinglyrecognized to be essential for tumor survival and progression and havebeen primarily studied in carcinomas of the breast (33), pancreas (34),prostate (35), skin (36), and cervix (37). In melanoma, a wide varietyof different cytokines and growth factors are expressed (38), whichoften act in an autocrine way, but also may influence the tumorenvironment via paracrine loop (39). Most studies have hereby fo-cused on the induction of angiogenesis by VEGF, basic fibroblastgrowth factor, PDGFs, and IL-8 (40). In this study, it is shown thatstroma can be remodeled by the paracrine effects of TGF-�1 producedby melanoma cells, which results in an increased deposition of ECMproteins in the interstitium of the tumor. The previously describedinduction of collagen, fibronectin, tenascin, and �2 integrin by TGF-�(41–43) could be demonstrated in the human fibroblast-containingstroma surrounding TGF-�-producing melanoma cells in an in vivomodel. Microarray studies of TGF-�1-transduced fibroblasts in orga-notypic culture, in which tenascin and types VI, XV, and XVIIIcollagen were up to 12-fold increased, partly mirrored the in vivo data.However, other collagen subtypes, �2 integrin, or fibronectin precur-sor were not increased, which might be because of limitations of thein vitro culture system, degradation of RNA, missed transient timepoints of induction, or the fact that fibroblasts were transduced withTGF-� and not stimulated by exogenous TGF-�.

Concomitant with the stroma reaction, the absence of larger ne-crotic areas and the fewer number of apoptotic cells in the TGF-�1-transduced tumors were the most striking differences to the controls.The fact that this did not result in a greater volume of the ECM-richmelanomas in the first 4 weeks was most likely attributable to in-creased edema and blood content in the more necrotic control tumors,which, however, at later time points were found to be smaller than theTGF-�1-transduced tumors, possibly as a consequence of resorptionof the edema and necrotic cells. Decreased death attributable toTGF-�1 transduction was also seen in human melanoma skin recon-structs. TGF-� obviously conferred a selective survival advantage tothe melanoma cells, which culminated in the significant increase innumber and size of lung metastases in mice injected with TGF-�1-transduced 451Lu melanoma cells. No growth induction by TGF-�1was seen in 451Lu melanoma cells in vitro, indicating that directautocrine effects of TGF-�1 were not responsible for this phenome-non, but that neighboring stroma cells in vivo were a prerequisite forthese beneficial effects of TGF-�1 on the tumor. The stroma forma-tion might have provided the melanoma cells a scaffolding, to which

5 The detailed results of all 2280 genes can be found on our Web site, http://www.wistar.upenn.edu/herlyn.

Table 1 Lung metastases of TGF-�1-transduced melanoma in SCID mice

451Lu melanoma cells were injected s.c. into SCID mice after transduction withTGF-�1 or LacZ and after mixing with normal human fibroblasts in a ratio 1:1. After39–41 days, lungs were harvested and analyzed microscopically. Micrometastases werecounted in 10 randomly chosen fields at �100, and the largest diameter of each metastasiswas measured.

Experimental group Mouse

Metastases

Number(mean � SD)

Diameter (mm)(mean � SD)

TGF-�1 1 13.1 � 2.47 0.11 � 0.182 10.7 � 2.58 0.11 � 0.093 6.5 � 3.89 0.15 � 0.184 4.6 � 3.81 0.12 � 0.025 4.08 � 3.8 0.06 � 0.066 0.73 � 1.22 0.05 � 0.057 0.5 � 0.53 0.01 � 0.018 0.3 � 0.67 0.08 � 0.059 0 0

10 0 0LacZ 1 2.3 � 2.19 0.04 � 0.03

2 0.9 � 0.99 0.05 � 0.063 0.4 � 0.7 0.04 � 0.044 0.23 � 0.44 0.03 � 0.025 0.2 � 0.42 0.05 � 06 0.2 � 0.42 0.11 � 0.137 0.15 � 0.38 0.01 � 08 0 09 0 0

10 0 0

8313




they can adhere and along which they can migrate. The stimulation offibroblasts might have induced in turn other growth factors, whichpositively regulate survival and growth of melanoma. Microarray andELISA analyses demonstrated VEGF to be strongly induced byTGF-�1 in both fibroblasts and melanoma cells, confirming previousin vitro data (44). However, angiogenesis was not significantly in-creased in tumors from TGF-�1-transduced 451Lu melanoma cells.This may have been attributable to the already high constitutiveproduction of VEGF protein in 451Lu cells and the already highvascularization of 451Lu control tumors. On the other hand, TGF-�1-induced VEGF in fibroblasts might have contributed to the survivaladvantage of the TGF-�1-transduced tumors by acting as an antiapo-ptotic factor both for the fibroblasts themselves and for the melanomacells via reciprocal paracrine routes.

Microarray analyses further revealed an induction of PDGF recep-tor-� by TGF-�1 in melanoma cells, which normally do not expressthis receptor (45). Also an increase in CFR-1 was detected, indicatingthat melanoma cells might have become increasingly responsive togrowth factors produced by stromal cells through up-regulation ofgrowth factor receptors. Increased survival and motility were addi-tionally observed by an increase in systemic metastases. This mighthave been either attributable to an altered expression of adhesionmolecules, proteases, or plasminogen activators on the melanomacells themselves (13, 16, 21, 22) or to the altered microenvironmentenhancing the chance of surviving melanoma cells to migrate alongthe stromal septae and enter the lymphatic or blood system.

Although TGF-�1 had stimulatory effects in melanoma in vivo, itdoes not seem to be essential for tumor growth and progression ingeneral. This is reflected in the heterogeneous pattern of endogenousTGF-�1 production in melanoma, which included advanced stages ofmelanoma that showed no or low levels of TGF-�1. It cannot beexcluded that the ratio between the active and the latent form ofTGF-�1 differed among the tested cell lines, because only the total

levels of TGF-�1 were measured in the ELISA used. However, inselected cases, constitutive TGF-�1 production was associated withprogression with highest concentrations found in single metastatic andadvanced primary cell lines (WM373, WM1617, and WM902B) andhigher concentrations in advanced cell lines when compared with theirprimary counterparts derived from the same patient (WM373-WM75,WM278-WM1617, and 1205Lu-WM793). The levels of TGF-�1 pro-duction in transduced melanoma cells were within the range of con-stitutive TGF-�1 levels in selected “high producer” melanoma celllines.

Induced expression of TGF-�1 in melanoma cells had differenteffects on their phenotype and proliferation capability. Many mela-noma cell lines were resistant to the growth-inhibitory effects ofTGF-�1, which was in line with previously reported effects of exog-

Fig. 7. Production of VEGF protein by human melanoma cells (WM793 and 451Lu)and dermal fibroblasts (FF2441) 72 h after transduction with TGF-�1 (f) or LacZ control(o) via adenoviral vectors. Results are expressed in average ng/ml per 106 cells; bars, SD.

Table 2 Modulation of gene expression by TGF-�1 in fibroblasts and melanoma cells

Fibroblasts and melanoma cell lines were transduced with either TGF-�1 or LacZ using adenoviral vectors. Fibroblasts were then embedded in type 1 collagen gels for 72 h, whereasmelanoma cells were maintained in monolayers. Microarray analysis of 2280 different human clones was performed. Genes coding for matrix proteins, adhesion receptors, growthfactors, their binding proteins and receptors, as well as protease inhibitors are shown. Detailed results can be found at Internet address http:/www.wistar.upenn.edu/herlyn.

Gene group

Fibroblasts (F) and melanoma cells (M)a

F1b M1 F2 M2

Matrix proteins Collagen XV �1c Collagen XVIII �1c Collagen III �1 Thrombospondin 2Collagen XVIII �1d Collagen XV �1d Collagen IV �2 OsteonectinOsteonectind Collagen IV �2 Collagen IX �3Tenascind Syndecan1Collagen VI �3 VitronectinSyndecan 4 Thrombospondin 1

Thrombospondin 2Thrombospondin 4

Integrins �V �1d

�3 �5�V�3�6

Growth factors VEGF-Ac VEGF-Ac NGFe-�c

TGF-�1d TGF-�1d IGF-1BMP-6 BMP-6d IGF-2VEGF-B IGF-2 PDGF-B

BMP-7Growth factor receptors TGF-�RII PDGFR-�c IGF-R1

CFR-1 IGF-R2d PDGFR-�TGF-�RIICFR-1

Protease inhibitors PAI-Ic PAI-Ic TIMP-IPAI-II PAI-IId

TIMP-3 TIMP-3a Melanoma cell lines WM793, 1205Lu, and 451Lu were used. Results indicated were found in at least two of the three cell lines.b Results are expressed as at least 2-fold higher (1) or lower (2) than LacZ controls.c Expression was more than 10-fold higher or lower than in LacZ controls.d Expression was 5–10-fold higher or lower than in LacZ controls.e NGF, nerve growth factor; BMP, bone morphogenetic protein; IGF, insulin-like growth factor.

8314




enous TGF-� on melanoma cells (8, 14). For a more detailed analysisof resistance of melanoma cells to TGF-�1, both the active and thelatent form of TGF-�1 need to be determined separately before andafter TGF-�1 transduction of each cell line and different infectiondoses need to be tested to be able to determine the threshold concen-tration of sensitivity, which was, however, not the aim of this study.Resistance to TGF-� has been linked in several tumors to mutationsin genes involved in the TGF-� signaling pathway, such as TGF-�receptors I and II, Smad2, or Smad4 (6); however, none of thesemutations could be demonstrated in melanoma (15).

Three of 24 cell lines displayed a morphological transdifferentia-tion toward a spindle cell-like fibroblastoid phenotype. This has beenreported previously after addition of exogenous TGF-� and wasassociated with increased metastatic capacity (16). Metastatic capacityof the transdifferentiated cell lines was not tested in this study;however, in skin reconstructions, increased tumor cell survival wasfound. Transdifferentiation was also detected in fibroblasts, whenTGF-�1 was overexpressed in human skins. The widespread detectionof smooth muscle actin in the dermis indicated a differentiation of theskin fibroblasts into myofibroblasts, which has been described only invitro (46) and not yet in human skin in vivo. The myofibroblastphenotype has been discussed to be the main source of increased ECMdeposition in fibrosis of the kidney, liver, and lung (47), and the dataof this study suggest that the same holds true for fibrosis of the skin.A strong collagen fiber network was induced by TGF-�1, which wasstronger than for any other growth factor studied.6

In summary, we have shown that TGF-�1 expression in humanmelanoma cells can lead to stimulation of the neighboring stroma cellswith increased production and deposition of ECM proteins. Theactivation of stroma in turn leads to a survival advantage and in-creased metastasis formation of the melanoma cells. A TGF-�1-triggered complex interplay between matrix proteins, proteases, inte-grins, and growth factors is suggested by global gene expressionstudies and excludes attempts to limit the characterization of melano-ma-stroma interactions to just single genes or a single gene group.

ACKNOWLEDGMENTS

We thank Sylvia Major, Katerina Chruma, Adrien Jarvis, and Dr. Ling Lifor technical assistance with the in vitro cell and immunohistochemistrystudies; Dr. Jonathan Garlick for invaluable support for the skin reconstructionmodel; Emma DeJesus and Rena Finko for technical assistance with cellcultures and skin reconstruction; Elsa Aglow for excellent histological proc-essing of the samples; Dr. Dirk Ruiter for helpful discussions about thehistopathological sections; and Wen Hwai Horng for biostatistical analysis ofthe microarray data.

REFERENCES

1. Roberts, A. B., McCune, B. K., and Sporn, M. B. TGF-�: regulation of extracellularmatrix. Kid. Int., 41: 557–559, 1992.

2. Massagué, J. TGF-� signal transduction. Annu. Rev. Biochem., 67: 753–791, 1998.3. Roberts, A. B., Sporn, M. B., Assoian, R. K., Smith, J. M., Roche, N. S., Wakefield,

L. M., Heine, U. I., Liotta, L. A., Falanga, V., Kehrl, J. H., et al. Transforming growthfactor type �: rapid induction of fibrosis and angiogenesis in vivo and stimulation ofcollagen formation in vitro. Proc. Natl. Acad. Sci. USA, 83: 4167–4171, 1986.

4. Verrechia, F., Chu, M-L., and Mauviel, A. Identification of novel TGF-�/Smad genetargets in dermal fibroblasts using a combined cDNA microarray/promoter transac-tivation approach. J. Biol. Chem., 276: 17058–17062, 2001.

5. Lund, L. R., Riccio, A., Andreasen, P. A., Nielsen, L. S., Kristensen, P., Laiho, M.,Saksela, O., Blasi, F., and Dano, K. Transforming growth factor-� is a strong and fastacting positive regulator of the level of type-1 plasminogen activator inhibitor mRNAin WI-38 human lung fibroblasts. EMBO J., 6: 1281–1286, 1987.

6. Reiss, M. Transforming growth factor-� and cancer: a love-hate relationship? Oncol.Res., 9: 447–457, 1997.

7. Gold, L. I. The role of transforming growth factor-� (TGF-�) in human cancer. Crit.Rev. Oncog., 10: 303–360, 1999.

8. Krasagakis, K., Kruger-Krasagakes, S., Fimmel, S., Eberle, J., Tholke, D., von derOhe, M., Mansmann, U., and Orfanos, C. E. Desensitization of melanoma cells toautocrine TGF-� isoforms. J. Cell. Physiol., 178: 179–187, 1999.

9. Reed, J. A., McNutt, N. S., Prieto, V. G., and Albino, A. P. Expression of transform-ing growth factor-�2 in malignant melanoma correlates with the depth of tumorinvasion. Implications for tumor progression. Am. J. Pathol., 145: 97–104, 1994.

10. Van Belle, P., Rodeck, U., Nuamah, I., Halpern, A. C., and Elder, D. E. Melanoma-associated expression of transforming growth factor-� isoforms. Am. J. Pathol., 148:1887–1894, 1996.

11. Moretti, S., Pinzi, C., Berti, E., Spallanzani, A., Chiarugi, A., Boddi, V., Reali, U. M.,and Giannotti, B. In situ expression of transforming growth factor � is associated withmelanoma progression and correlates with Ki67, HLA-DR and �3 integrin expres-sion. Melanoma Res., 7: 313–321, 1997.

12. Schmid, P., Itin, P., and Rufli, T. In situ analysis of transforming growth factor-�s(TGF-�1, TGF-�2, TGF-�3), and TGF-� type II receptor expression in malignantmelanoma. Carcinogenesis (Lond.), 16: 1499–1503, 1995.

13. Teti, A., De Giorgi, A., Spinella, M. T., Migliaccio, S., Canipari, R., Onetti Muda, A.,and Faraggiana, T. Transforming growth factor-� enhances adhesion of melanomacells to the endothelium in vitro. Int. J. Cancer, 72: 1013–1020, 1997.

14. Rodeck, U., Bossler, A., Graeven, U., Fox, F. E., Nowell, P. C., Knabbe, C., and Kari,C. Transforming growth factor � production and responsiveness in normal humanmelanocytes and melanoma cells. Cancer Res., 54: 575–581, 1994.

15. Rodeck, U., Nishiyama, T., and Mauviel, A. Independent regulation of growth andSMAD-mediated transcription by transforming growth factor � in human melanomacells. Cancer Res., 59: 547–550, 1999.

16. Janji, B., Melchior, C., Gouon, V., Vallar, L., and Kieffer, N. Autocrine TGF-�-regulated expression of adhesion receptors and integrin-linked kinase in HT-144melanoma cells correlates with their metastatic phenotype. Int. J. Cancer, 83: 255–262, 1999.

17. Letterio, J. J., and Roberts, A. B. Regulation of immune responses by TGF-�. Annu.Rev. Immunol., 16: 137–161, 1998.

18. Conrad, C. T., Ernst, N. R., Dummer, W., Brocker, E. B., and Becker, J. C.Differential expression of transforming growth factor �1 and interleukin 10 inprogressing and regressing areas of primary melanoma. J. Exp. Clin. Cancer Res., 18:225–232, 1999.

19. Iruela-Arispe, M. L., and Sage, E. H. Endothelial cells exhibiting angiogenesis invitro proliferate in response to TGF-�1. J. Cell. Biochem., 52: 414–430, 1993.

20. Festuccia, C., Angelucci, A., Gravina, G. L., Villanova, I., Teti, A., Albini, A.,Bologna, M., and Abini, A. Osteoblast-derived TGF-�1 modulates matrix degradingprotease expression and activity in prostate cancer cells. Int. J. Cancer, 85: 407–415,2000.

21. Santibánez, J. F., Frontelo, P., Iglesias, M., Martı́nez, J., and Quintanilla, M. Uroki-nase expression and binding activity associated with transforming growth factor�1-induced migratory and invasive phenotype of mouse epidermal keratinocytes.J. Cell. Biochem., 74: 61–73, 1999.

22. Farina, A. R., Coppa, A., Tiberio, A., Tacconelli, A., Turco, A., Colletta, G., Gulino,A., and Mackay, A. R. Transforming growth factor-�1 enhances the invasiveness ofhuman MDA-MB-231 breast cancer cells by up-regulating urokinase activity. Int. J.Cancer, 75: 721–730, 1998.

23. Herlyn, M., Thurin, J., Balaban, G., Bennicelli, J. L., Herlyn, D., Elder, D. E., Bondi,E., Guerry, D., Nowell, P., Clark, W. H., et al. Characteristics of cultured humanmelanocytes isolated from different stages of tumor progression. Cancer Res., 45:5670–5676, 1985.

24. Hsu, M-Y., Elder, D. E., and Herlyn, M. The Wistar melanoma (WM) cell lines. In:J. R. W. Masters and B. Palsson (eds.), Human Cell Culture, Vol. 3, Solid Cancers,pp. 259–274. Norwell, MA: Kluwer Academic Publishers, 1999.

25. Lee, W. C., Zhong, C., Qian, S., Wan, Y., Gauldie, J., Mi, Z., Robbins, P. D.,Thomson, A. W., and Lu, L. Phenotype, function, and in vivo migration and survivalof allogeneic dendritic cell progenitors genetically engineered to express TGF-�.Transplantation, 66: 1810–1817, 1998.

26. Berking, C., and Herlyn, M. Experimental induction of atypical melanocytic lesionsand melanoma in ultraviolet-irradiated human skin grafted to immunodeficient mice.In: B. Nickoloff (ed.), Melanoma Techniques and Protocols. Molecular Diagnosis,Treatment, and Monitoring, Methods in Molecular Medicine, pp. 71–84. Totowa, NJ:Humana Press, 2001.

27. Herlyn, M., Graeven, U., Speicher, D., Sela, B. A., Bennicelli, J. L., Kath, R., andGuerry, D. Characterization of tenascin secreted by human melanoma cells. CancerRes., 51: 4853–4858, 1991.

28. Hsu, M. Y., Shih, D. T., Meier, F. E., Van Belle, P., Hsu, J. Y., Elder, D. E., Buck,C. A., and Herlyn, M. Adenoviral gene transfer of �3 integrin subunit inducesconversion from radial to vertical growth phase in primary human melanoma. Am. J.Pathol., 153: 1435–1442, 1998.

29. Menrad, A., Speicher, D., Wacker, J., and Herlyn. M. Biochemical and functionalcharacterization of aminopeptidase N expressed by human melanoma cells. CancerRes., 53: 1450–1455, 1993.

30. Meier, F., Nesbit, M., Hsu, M. Y., Martin, B., Van Belle, P., Elder, D. E.,Schaumburg-Lever, G., Garbe, C., Walz, T. M., Donatien, P., Crombleholme, T. M.,and Herlyn, M. Human melanoma progression in skin reconstructs: biological sig-nificance of bFGF. Am. J. Pathol., 156: 193–200, 2000.

31. Juhasz, I., Albelda, S. M., Elder, D. E., Murphy, G. F., Adachi, K., Herlyn, D.,Valyi-Nagy, I. T., and Herlyn, M. Growth and invasion of human melanomas inhuman skin grafted to immunodeficient mice. Am. J. Pathol., 143: 528–537, 1993.6 Unpublished data.

8315




32. Herlyn, D., Iliopoulos, D., Jensen, P. J., Parmiter, A., Baird, J., Hotta, H., Adachi, K.,Ross, A. H., Jambrosic, J., Koprowski, H., et al. In vitro properties of humanmelanoma cells metastatic in nude mice. Cancer Res., 50: 2296–2302, 1990.

33. Shekhar, M. P., Werdell, J., Santner, S. J., Pauley, R. J., and Tait, L. Breast stromaplays a dominant regulatory role in breast epithelial growth and differentiation:implications for tumor development and progression. Cancer Res., 61: 1320–1326,2001.

34. Lohr, M., Schmidt, C., Ringel, J., Kluth, M., Muller, P., Nizze, H., and Jesnowski, R.Transforming growth factor-�1 induces desmoplasia in an experimental model ofhuman pancreatic carcinoma. Cancer Res., 61: 550–555, 2001.

35. Harding, M. A., and Theodorescu, D. Prostate tumor progression and prognosis.Interplay of tumor and host factors. Urol. Oncol., 5: 258–264, 2000.

36. Lengyel, E., Gum, R., Juarez, J., Clayman, G., Seiki, M., Sato, H., and Boyd, D.Induction of Mr 92,000 type IV collagenase expression in a squamous cell carcinomacell line by fibroblasts. Cancer Res., 55: 963–967, 1995.

37. Turner, M. A., Darragh, T., and Palefsky, J. M. Epithelial-stromal interactionsmodulating penetration of Matrigel membranes by HPV 16-immortalized keratino-cytes. J. Investig. Dermatol., 109: 619–625, 1997.

38. Moretti, S., Pinzi, C., Spallanzani, A., Berti, E., Chiarugi, A., Mazzoli, S., Fabiani,M., Vallecchi, C., and Herlyn, M. Immunohistochemical evidence of cytokine net-works during progression of human melanocytic lesions. Int. J. Cancer, 84: 160–168,1999.

39. Lazar-Molnar, E., Hegyesi, H., Toth, S., and Falus, A. Autocrine and paracrineregulation by cytokines and growth factors in melanoma. Cytokine, 12: 547–554,2000.

40. Rofstad, E. K., and Halsor, E. F. Vascular endothelial growth factor, interleukin 8,platelet-derived endothelial cell growth factor, and basic fibroblast growth factor

promote angiogenesis and metastasis in human melanoma xenografts. Cancer Res.,60: 4932–4938, 2000.

41. Varga, J., Rosenbloom, J., and Jimenez, S. A. Transforming growth factor � (TGF �)causes a persistent increase in steady-state amounts of type I and type III collagen andfibronectin mRNAs in normal human dermal fibroblasts. Biochem. J., 247: 597–604,1987.

42. Ignotz, R. A., Endo, T., and Massagué, J. Regulation of fibronectin and type Icollagen mRNA levels by transforming growth factor-�. J. Biol. Chem., 262: 6443–6446, 1987.

43. Raghow, R., Postlethwaite, A. E., Keski-Oja, J., Moses, H. L., and Kang, A. H.Transforming growth factor-� increases steady state levels of type I procollagen andfibronectin messenger RNAs posttranscriptionally in cultured human dermal fibro-blasts. J. Clin. Investig., 79: 1285–1288, 1987.

44. Dolecki, G. J., and Connolly, D. T. Effects of a variety of cytokines and inducingagents on vascular permeability factor mRNA levels in U937 cells. Biochem. Bio-phys. Res. Commun., 180: 572–578, 1991.

45. Barnhill, R. L., Xiao, M., Graves, D., and Antoniades, H. N. Expression of platelet-derived growth factor (PDGF)-A, PDGF-B and the PDGF-� receptor, but not thePDGF-� receptor, in human malignant melanoma in vivo. Br. J. Dermatol., 135:898–904, 1996.

46. Vaughan, M. B., Howard, E. W., and Tomasek, J. J. Transforming growth factor-�1promotes the morphological and functional differentiation of the myofibroblast. Exp.Cell Res., 257: 180–189, 2000.

47. Fan, J-M., Ng, Y-Y., Hill, P. A., Nikolic-Paterson, D. J., Mu, W., Atkins, R., and Lan,H. Y. Transforming growth factor-� regulates tubular epithelial-myofibroblast trans-differentiation in vitro. Kidney Int., 56: 1455–1467, 1999.

8316




2001;61:8306-8316. Cancer Res Carola Berking, Richelle Takemoto, Helmut Schaider, et al. Melanoma through Stroma Remodeling

1 Increases Survival of HumanβTransforming Growth Factor-

Updated version

http://cancerres.aacrjournals.org/content/61/22/8306

Access the most recent version of this article at:

Cited articles

http://cancerres.aacrjournals.org/content/61/22/8306.full#ref-list-1

This article cites 44 articles, 14 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/61/22/8306.full#related-urls

This article has been cited by 24 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/61/22/8306To request permission to re-use all or part of this article, use this link

http://cancerres.aacrjournals.org/content/61/22/8306http://cancerres.aacrjournals.org/content/61/22/8306.full#ref-list-1http://cancerres.aacrjournals.org/content/61/22/8306.full#related-urlshttp://cancerres.aacrjournals.org/cgi/alertsmailto:[email protected]://cancerres.aacrjournals.org/content/61/22/8306http://cancerres.aacrjournals.org/

transforming growth factor- 1 increases survival of human ......effects on the immune system may...

Documents