gene expression profiling reveals cross-talk between melanoma … · 36618_g_at x77956 491.3 161.2...

Gene Expression Profiling Reveals Cross-talk between Melanoma

and Fibroblasts: Implications for Host-Tumor

Interactions in Metastasis

Paul G. Gallagher,1Yongde Bao,

1Alyson Prorock,

1Paola Zigrino,

2Roswitha Nischt,

2

Vincenzo Politi,3Cornelia Mauch,

2Bojan Dragulev,

1and Jay William Fox

1

1Department of Microbiology, University of Virginia, Charlottesville, Virginia; 2Department of Dermatology, University of Cologne,Cologne, Germany; and 3Polifarma SpA, Rome, Italy

Abstract

Host-tumor interaction is considered critical in carcinogen-esis, tumor invasion, and metastasis. To explore the recipro-cal effects of host-tumor interaction, we developed a systemto assess the gene expression patterns of A2058 humanmelanoma cells cocultured in fibrillar collagen with HS-68primary human fibroblasts. The gene expression pattern ofthe cocultured A2058 cells was only modestly affected,whereas the HS-68 fibroblast gene expression pattern wassignificantly altered. Interleukin-11 and inhibitor of DNA-binding domain-1 gene expression in the cocultured A2058cells was down-regulated, indicative of a proinflammatoryresponse and resistance to apoptosis, respectively. The overallpattern of up-regulated genes indicated triggering of theproinflammatory process. In addition, the melanoma growthand migration stimulatory chemokines CXCL1 and CXCL2were significantly up-regulated in the cocultured fibroblasts.These results were corroborated by additional cocultureexperiments with the melanoma cell lines WM-164, BLM,and SK-Mel-28 and immunohistochemistry on invasive hu-man melanoma sections. Taken together, these resultsindicate that tumor cells cause a proinflammatory andmelanoma growth-promoting response in stromal fibroblasts.The role of inflammation in carcinogenesis, tumor promo-tion, invasion, and metastasis is viewed as being increasinglyimportant and the results of these studies underscore this aswell as identify certain key proteins that are expressed as aresult of the complex interactive processes in the host-tumormicroenvironment. (Cancer Res 2005; 65(10): 4134-46)

Introduction

Increasing attention is being focused on the interaction of tumorcells with the host in terms of exploring how that interaction isconducive to the induction, selection, and expansion of the tumorultimately leading to the malignant progression of the tumor (1–5).For example, studies on human breast carcinomas have shown thatvascular stroma formation occurs before invasion by tumor cells.In situ hybridization of breast carcinomas at various stagesdetected expression of vascular permeability factor and vascularendothelial growth factor, whereas the stromal cells were shown to

express thrombospondin-1, collagen type I, fibronectin, versican,and decorin, suggesting that the tumor induces vascularization inregions where they invade and host stroma promotes thisvascularization (6).The microenvironment in which host-tumor interaction occurs

has also been implicated in malignancy. The extracellular matrix(ECM) of the microenvironment is well known to regulate a varietyof cellular phenomena. Perturbation of the matrix by eitherproteolysis or alteration of its architecture due to changes inmolecular composition and/or stoichiometry resulting from host-tumor interaction can disrupt the homeostasis of the microenvi-ronment (7–9). For example, Thomasset et al. (10) have shown thatlow-level expression of the transgene matrix metalloproteinase(MMP) stromelysin-1 in mouse mammary epithelia causes an up-regulation of endogenous stromelysin-1 in fibroblasts. Thesechanges resulted in the development of preneoplastic andneoplastic lesions in mice that were associated with the up-regulation of other MMPs, ECM components, and mammary glandvascularization.The concept that inflammation plays a central role in tumor

progression has received increasing attention, particularly in thecontext of host-tumor interaction (1, 11). For example, macrophageinfiltrates of malignant melanoma have been associated withtumor stage and angiogenesis. Production of transforming growthfactor-h, tumor necrosis factor-a (TNF-a), interleukin (IL)-1a,arachidonate metabolites, and extracellular proteinases by themacrophages elicited the expression of IL-8 and vascularendothelial growth factor-A in melanocytes, thereby promotinginflammation and angiogenesis (12). Several studies have shownthat host-tumor interaction results in the production of proin-flammatory cytokines and chemokines, thereby promoting therecruitment of host leukocytes in the microenvironment of thetumor (13). However, in these investigations and most others, thereis always some question as to the cell source of the chemokinesand cytokines involved in the development of the proinflammatorypathway within the tumor microenvironment.Unlike melanocytes, melanoma cells constitutively produce a

large number of growth factors and cytokines and their respectivereceptors that enable the cells to progress to a more aggressivephenotype. Autocrine growth factors (e.g., basic fibroblast growthfactor, IL-8, and hepatocyte growth factor) stimulate proliferationand migration of melanoma cells themselves, whereas paracrinefactors (e.g., platelet-derived growth factor, transforming growthfactor-h, basic fibroblast growth factor, vascular endothelial growthfactor, and monocyte chemoattractant protein-1) are believed tomodulate the microenvironment to the benefit of melanomagrowth, invasion, and metastasis (14). On the other hand,fibroblasts are also a rich source of growth factors but only after

Note: Supplementary data for this article are available at Cancer Research Online(http://cancerres.aacrjournals.org/).

Requests for reprints: Jay William Fox, Department of Microbiology, University ofVirginia, P.O. Box 800734, Charlottesville, VA 22908-0734. Phone: 434-924-0050; Fax:434-924-2514; E-mail: [email protected].

I2005 American Association for Cancer Research.

Cancer Res 2005; 65: (10). May 15, 2005 4134 www.aacrjournals.org

Research Article

Research. on September 6, 2020. © 2005 American Association for Cancercancerres.aacrjournals.org Downloaded from

http://cancerres.aacrjournals.org/

activation. When stimulated by melanoma cell–released factors,they can produce growth factors (e.g., insulin-like growth factor-I,hepatocyte growth factor, and ET-3) that in turn contribute to theorchestra of virtual modulation of cellular activities (15).Several investigators have shown that host-tumor interactions

also play a crucial role in the regulation of connective tissuebreakdown in different tumors. Tumor cell–derived factors, such asextracellular MMP inducer, have been shown to be expressed byvarious tumor cells, including melanoma, and to induce productionof MMP-1, MMP-2, and MMP-3 in normal fibroblasts (16). Recently,we could show in human primary melanoma and in lymph nodemetastases that the activity of MMPs (e.g., MMP-2 and MMP-9) waslocated primarily in those areas where tumor cells lay adjacent tothe ECM or within connective tissue septa among the aggregates ofmelanoma cells (17). This observation stresses the importance ofhost-tumor interactions with structural and cellular componentsof the surrounding ECM.

In this investigation, we report the results of assays designed toassess the reciprocal effect of soluble cofactors resulting from thecoculture of human melanoma cell lines A2058, BLM, WM-164, andSK-Mel-28 with human HS-68 fibroblasts on their respective geneexpression profiles using oligonucleotide microarrays. Interestingly,the gene expression profile of the fibroblasts was dramaticallyaltered under coculture conditions, whereas the effect on themelanoma A2058 cell gene expression profile was more modest.Overall, the fibroblasts responded to coculture by up-regulation ofa variety of genes associated with the proinflammatory pathwayand cellular proliferation and this up-regulation was confirmed byquantitative real-time PCR (qRT-PCR). Several transcripts of ECMcomponents were observed to be down-regulated in coculturedfibroblasts suggesting a potential alteration of the extracellularenvironment. Furthermore, the presence of several of theseproteins was detected by immunohistochemistry of invasivehuman melanoma.

Table 1. Genes with altered expression levels in A2058 melanoma cells when grown in coculture with HS-68 fibroblasts

Feature no. Accession no. Monoculturemean intensity

Coculturemean intensity

Foldchange

P Gene description

1916_s_at V01512 43.0 101.5 2.4 0.004398 v-fos FBJ murine osteosarcomaviral oncogene homologue

41182_at D17547 191.0 446.6 2.3 0.000313 Dopachrome tautomerase

(dopachrome y-isomerase,

tyrosine-related protein 2)1052_s_at M83667 128.3 260.4 2.0 0.012085 CCAAT/enhancer-binding protein y37251_s_at AF016004 80.6 145.2 1.8 0.005486 Glycoprotein M6B

235_at M59488 460.5 746.6 1.6 0.001581 M59488/feature = mRNA/

definition = HUMS100B3human S100 protein

h-subunit gene, exon 3

1364_at M93426 122.9 198.9 1.6 0.000999 Protein tyrosine phosphatase,receptor type, Z polypeptide 1

36159_s_at U29185 2023.6 1215.8 �1.7 0.000908 Prion protein (p27-30)

(Creutzfeld-Jakob disease,

Gerstmann-Strausler-Scheinkersyndrome, fatal familial insomnia)

37762_at Y07909 222.1 131.2 �1.7 0.00145 Epithelial membrane protein 1

1278_at 848.9 481.6 �1.8 0.003165 Tyrosine kinase, receptor Axl,

Alt. splice 238433_at M76125 920.0 512.5 �1.8 0.005982 AXL receptor tyrosine kinase

160025_at X70340 452.6 240.5 �1.9 0.001444 Transforming growth factor-a

38261_at AF085692 123.2 63.8 �1.9 0.006146 ATP-binding cassette, subfamily C(CFTR/MRP), member 3

37157_at X56667 206.0 93.9 �2.2 0.00684 Calbindin 2 (29 kDa, calretinin)

37279_at U10550 221.2 97.7 �2.3 0.007357 GTP-binding protein overexpressed

in skeletal muscle35726_at AI539439 436.7 188.3 �2.3 0.001857 S100 calcium-binding protein A2

2027_at M87068 885.9 381.4 �2.3 0.001572 M87068/definition = HUMCAN

Homo sapiens CaN19 mRNA

sequence36671_at M27396 582.3 246.5 �2.4 0.000562 Asparagine synthetase

37043_at AL021154 1206.5 505.9 �2.4 0.002939 E2F transcription factor 2

36618_g_at X77956 491.3 161.2 �3.0 0.001036 Inhibitor of DNA-binding 1,

dominant-negativehelix-loop-helix protein

35464_at X58377 381.5 107.1 �3.6 0.00129 IL-11

NOTE: Affymetrix feature numbers, accession numbers, and fold changes in gene expression compared with cells grown in collagen fibrils alone.

Melanoma-Fibroblast Cross-talk

www.aacrjournals.org 4135 Cancer Res 2005; 65: (10). May 15, 2005



In summary, we observed a profound effect on the geneexpression profile of fibroblasts when in coculture with melanomacells. Analysis of the gene expression profiles indicates theinitiation of the proinflammatory pathway and an altered ECM,conditions that have been implicated in tumor progression,invasion, and metastasis. However, the response to coculture withfibroblasts by the melanoma cell line is modest. Therefore, weconclude based on this study that the stroma, due to signalsoriginating from the tumor, is primarily responsible for thegeneration of a microenvironment that is proinflammatory,proproliferation, proinvasion, and prometastatic.

Materials and Methods

HS-68 fibroblast and melanoma cell culture and RNA isolation.Human primary foreskin fibroblasts (HS-68, American Type Culture

Collection, Manassas, VA) were grown in DMEM/10% FCS, and the

melanoma cell lines A2058, BLM, WM-164, and SK-Mel-28 were grown inDMEM/10% FCS and 1 Amol/L sodium pyruvate. Fibroblasts and melanoma

cells were seeded individually during polymerization of the collagen type Ilattice and cultured on opposite sides of a 1 Am pore, six-well cell culture

insert (Becton Dickinson, Sparks, MD). Collagen lattices containing cells

were composed of 1� DMEM, 0.9 mg/mL bovine skin collagen type I (IBFB,

Leipzig, Germany), 14.8 mmol/L NaOH, 10% FCS, and 3 � 105 cells/mL.Collagen/cell mixture (540 AL) was added to each side of the insert and

allowed to polymerize to form the cell-embedded lattice. The fibroblast/

collagen lattice was situated above the filter and melanoma cell/collagen

lattice below. After collagen fibers had formed, DMEM/10% FCS (1 mL) wasinjected below each insert and DMEM/10% FCS (250 AL) was overlaid on

top of the fibroblast/collagen lattice. Monoculture control samples were

generated by culturing only melanoma or fibroblast cells in collagen latticeson the same side of the filter as in the coculture design. Cells were

incubated for 2 or 72 hours at 37jC/5% CO2, and total RNA was isolated

from two inserts for each cell type using TRIzol reagent (Invitrogen,

Carlsbad, CA). All experiments were done in duplicate.Microarray hybridization. Biotin-labeled cRNA was generated from

total RNA samples according to standard Affymetrix GeneChip protocols

(Santa Clara, CA). Briefly, a poly-T oligo primer containing a 5VT7 RNA po-

lymerase promoter was used to generate double-stranded cDNA followed

Table 2. Pathways and biological process classes associated with differentially expressed genes in cocultured A2058melanoma cells

No. genes

changed

No. genes

measured

No. genes

on MAPP

% Changed % Present Z score

Up-regulated genesPathways

Cholesterol biosynthesis 3 13 15 23.1 86.7 5.365Acetylcholine synthesis 1 6 7 16.7 85.7 2.534

G13 signaling pathway 2 27 31 7.4 87.1 1.973

Inflammatory response pathway 2 28 31 7.1 90.3 1.910

Small ligand GPCRs 1 11 16 9.1 68.8 1.654GPCRs, class C metabotropic

glutamate, pheromone

1 12 18 8.3 66.7 1.542

Biological process classes

Melanin biosynthesis from tyrosine 3 5 6 60 83.3 8.483Glutamate transport 2 4 4 50 100 6.271

Isoprenoid biosynthesis 3 7 13 42.9 53.8 4.994

Dicarboxylic acid transport 2 9 10 22.2 90 4.215Response to oxidative stress 2 17 22 11.8 77.3 3.855

Defense response 4 56 70 7.1 80 3.303

Antigen presentation,

exogenous antigen

4 18 35 22.2 51.4 3.263

Antigen processing, exogenous

antigen via MHC class

4 18 35 22.2 51.4 3.263

Neurogenesis 9 158 215 5.7 73.5 3.146

Down-regulated genesPathways

Cysteine metabolism 1 5 5 20 100 2.944

Mitogen-activated protein kinase cascade 2 21 26 9.5 80.8 2.537

Methionine metabolism 1 7 7 14.3 100 2.382Orphan GPCRs 1 11 16 9.1 68.8 1.731

Small ligand GPCRs 1 11 16 9.1 68.8 1.731

Complement activation, classic 1 12 16 8.3 75 1.617Eicosanoid synthesis 1 14 16 7.1 87.5 1.423

Nucleotide metabolism 1 15 16 6.7 93.8 1.338


Angiogenesis 3 22 26 13.6 84.6 3.997Phosphate transport 2 4 10 50 40 3.954

Glutamine metabolism 2 7 7 28.6 100 3.920

Positive regulation of cell proliferation 6 91 107 6.6 85 3.221

Ubiquitin cycle 5 62 105 8.1 59 2.983

Cancer Research




Table 3. Genes with altered expression levels in HS-68 fibroblasts grown in coculture with A2058 melanoma cells

Feature

no.

Accession

no.

Monoculture

mean intensity

Coculture

mean intensity

Fold

change

P Gene description

35372_r_at M17017 62.5 424.3 6.8 0.000016 IL-8

37187_at M36820 33.6 221.5 6.6 0.000893 GRO2 oncogene

35692_at AL080235 514.0 2,015.4 3.9 0.000026 DKFZP586E1621 protein39310_at X86163 266.9 895.9 3.4 0.000091 Bradykinin receptor B2

408_at X54489 140.7 447.9 3.2 0.004722 GRO1 oncogene (melanoma

growth stimulating activity, a)

37310_at X02419 798.2 2,518.3 3.2 0.000014 Plasminogen activator, urokinase34476_r_at D30783 23.4 67.2 2.9 0.002951 Epiregulin

33534_at X89426 31.9 90.1 2.8 0.002845 Endothelial cell–specific

molecule 138326_at M69199 466.1 1,238.9 2.7 0.010686 Putative lymphocyte G0-G1

switch gene

38428_at M13509 1,193.8 3,064.2 2.6 0.000464 MMP-1

32805_at U05861 514.1 1,271.7 2.5 0.000874 Cluster Incl. U05861:humanhepatic dihydrodiol

dehydrogenase gene

1520_s_at X04500 250.5 612.8 2.4 0.005425 IL-1h40812_at AF027153 70.5 167.8 2.4 0.016044 Solute carrier family 5

(inositol transporters), member 3

1237_at S81914 915.4 2,148.3 2.3 0.00083 Immediate-early response 3

36589_at X15414 3,202.6 7,367.9 2.3 0.000071 Aldo-keto reductase family 1,member B1 (aldose reductase)

32521_at AF056087 624.7 1,401.2 2.2 0.000251 Secreted frizzled-related protein 1

1491_at M31166 333.4 746.6 2.2 0.000095 Pentaxin-related gene,

rapidly induced by IL-1h41354_at U25997 559.9 1,215.0 2.2 0.00463 Stanniocalcin 1

875_g_at M26683 1,778.5 3,533.1 2.0 0.002517 Small inducible cytokine A2

(monocyte chemoattractant

protein-1, homologous tomouse Sig-je)

34375_at M28225 1,874.3 3,671.2 2.0 0.000511 Cluster Incl. M28225:human JE

gene encoding a monocyte

secretory protein35396_at U54804 79.3 155.2 2.0 0.002128 Hyaluronan synthase 2

37156_at AF070641 44.9 86.4 1.9 0.018795 Cluster Incl. AF070641:H. sapiens

clone 24421 mRNA sequence31824_at AL049699 58.1 105.9 1.8 0.004246 Malic enzyme 1, NADP(+)-dependent,

cytosolic

32796_f_at U66061 264.6 471.7 1.8 0.002439 Protease, serine, 1 (trypsin 1)

40043_at X71345 211.1 365.1 1.7 0.001157 Protease, serine, 4 (trypsin 4, brain)33452_at M15518 728.9 1,259.7 1.7 0.001882 Tissue plasminogen activator

1372_at M31165 74.7 129.1 1.7 0.001901 TNF-a-induced protein 6

38183_at U13219 176.5 301.5 1.7 0.001388 Forkhead box F1

37279_at U10550 326.1 552.8 1.7 0.000507 GTP-binding protein overexpressedin skeletal muscle

34942_at AF070524 182.6 308.7 1.7 0.012919 Cluster Incl. AF070524:H. sapiens

clone 24453 mRNA sequence34666_at X07834 89.0 149.6 1.7 0.020578 Superoxide dismutase 2,

mitochondrial

37603_at X52015 181.4 297.0 1.6 0.004994 IL-1RA

36130_f_at R92331 4,168.3 6,767.7 1.6 0.000264 Metallothionein 1E ( functional)39594_f_at R93527 4,721.7 7,647.9 1.6 0.000549 Metallothionein 1H

41048_at D90070 92.3 148.4 1.6 0.005756 Phorbol-12-myristate-13-acetate-

induced protein 1

1461_at M69043 986.2 1,579.5 1.6 0.002294 Nuclear factor of n light polypeptidegene enhancer in B cells inhibitor, a

837_s_at U43944 1,064.4 1,693.0 1.6 0.000207 Malic enzyme 1, NADP(+)-dependent,

cytosolic

(Continued on the following page)





Table 3. Genes with altered expression levels in HS-68 fibroblasts grown in coculture with A2058 melanoma cells (Cont’d)

Feature

no.

Accession

no.

Monoculture

mean intensity

Coculture

mean intensity

Fold

change

P Gene description

39081_at AI547258 232.4 368.6 1.6 0.011862 Metallothionein 2A

1173_g_at M77693 672.7 1,062.5 1.6 0.000637 Spermidine/spermineN1-acetyltransferase

546_at S76965 89.7 141.4 1.6 0.012067 Protein kinase (cyclic AMP-dependent,

catalytic) inhibitor a33466_at AF038182 592.9 921.2 1.6 0.005106 Cluster Incl. AF038182:H. sapiens

clone 23860 mRNA sequence

39385_at M22324 1,485.1 2,296.2 1.5 0.001294 Alanyl (membrane) aminopeptidase

(aminopeptidase N, aminopeptidase M)437_at X05232 1,184.1 1,826.7 1.5 0.029174 MMP-3

31622_f_at M10943 5,107.1 7,814.7 1.5 0.000624 Metallothionein 1F ( functional)

41635_at D14661 551.6 840.5 1.5 0.000124 Wilms’ tumor 1-associating protein

36941_at U16954 494.5 747.7 1.5 0.001928 ALL1-fused gene from chromosome 1q609_f_at M13485 4,585.9 6,922.1 1.5 0.000761 M13485/feature = cds/definition =

HUMMT1B2 human metallothionein

I-B gene, exon 333700_at AF039843 207.0 311.7 1.5 0.005918 Sprouty (Drosophila) homologue 2

980_at AF002020 395.7 593.5 1.5 0.005965 Niemann-Pick disease, type C1

1788_s_at U48807 93.9 140.8 1.5 0.00659 Dual specificity phosphatase 4

34304_s_at AL050290 169.8 254.6 1.5 0.010397 Spermidine/spermineN1-acetyltransferase

41446_f_at H68340 5,059.9 7,575.5 1.5 0.00176 RNA helicase-related protein

32305_at J03464 1,846.5 1,235.1 �1.5 0.01459 Collagen, type I, a2

38754_at AI557295 1,297.1 867.6 �1.5 0.00039 p8 protein (candidate of metastasis 1)37283_at X82209 135.3 90.1 �1.5 0.003498 Meningioma (disrupted in balanced

translocation) 1

37929_at AB017563 766.7 509.8 �1.5 0.009205 Cluster Incl. AB017563:H. sapiensIGSF4 gene

160043_at X66087 262.8 172.5 �1.5 0.002992 v-myb avian myeloblastosis viral

oncogene homologue-like 1

39333_at M26576 637.9 413.7 �1.5 0.013582 Cluster Incl. M26576:human a1collagen type IV gene

35829_at AL080181 387.2 250.2 �1.5 0.041304 Immunoglobulin superfamily, member 4

36513_at U37283 187.2 119.7 �1.6 0.004712 Microfibril-associated glycoprotein-2

32551_at U03877 394.1 251.8 �1.6 0.031589 Epidermal growth factor–containingfibulin-like ECM protein 1

1088_at M61176 153.5 97.0 �1.6 0.035594 Brain-derived neurotrophic factor

32755_at X13839 2,655.0 1,672.3 �1.6 0.000792 Actin, a2, smooth muscle, aorta

37765_at X54162 600.9 376.1 �1.6 0.002885 Leiomodin 1 (smooth muscle)40621_at U63809 222.3 138.9 �1.6 0.009002 PRKC, apoptosis, WT1, regulator

39710_at U30521 2,321.6 1,447.4 �1.6 0.011166 P311 protein

36917_at Z26653 565.6 345.7 �1.6 0.010019 Laminin, a2 (merosin, congenitalmuscular dystrophy)

41401_at U57646 658.9 387.0 �1.7 0.001607 Cysteine and glycine–rich protein 2

40060_r_at AF061258 733.9 423.7 �1.7 0.009087 LIM protein (similar to rat protein

kinase C-binding enigma)2053_at M34064 2,862.2 1,645.7 �1.7 0.004121 N-cadherin

37678_at U23070 229.6 130.4 �1.8 0.005819 Putative transmembrane protein

39098_at X52896 433.5 236.4 �1.8 0.000298 Elastin (supravalvular aortic stenosis,

Williams-Beuren syndrome)34012_at Y16790 424.2 229.6 �1.8 0.001367 Keratin, hair, acidic, 4

32542_at AF063002 2,615.5 1,394.8 �1.9 0.000937 Four-and-a-half LIM domains 1

36671_at M27396 197.2 103.6 �1.9 0.005675 Asparagine synthetase37892_at J04177 877.1 445.6 �2.0 0.004028 Collagen, type XI, a1

1451_s_at D13666 1,804.9 889.8 �2.0 0.000968 Osteoblast-specific factor 2

( fasciclin I-like)

33878_at W27472 774.4 346.5 �2.2 0.001289 Cluster Incl. W27472:31d4H. sapiens cDNA

39063_at J00073 535.7 138.7 �3.9 0.000553 Cluster Incl. J00073:human

a-cardiac actin gene

Cancer Research




by biotin-labeled cRNA synthesis using BioArray High-Yield RNA Transcript

Labeling kit (Enzo, New York, NY). cRNA was fragmented according toAffymetrix standard protocol, and fragmented cRNA (10 Ag) was hybridizedto Affymetrix Hg-U95A probe arrays for 16 hours. The arrays were washed

and stained in an Affymetrix Automated Fluidics Station 400 and scannedwith a HP GeneArray scanner.

Gene expression analysis. Scanned gene array images were first

examined for visible defects and then checked for the fitness of the

gridding. When passed, the image file was analyzed to generate compositedata files (‘‘cell files’’). From this point on, a coordination of two paths of

analysis was carried on using the Affymetrix Microarray Analysis Suite

version 5.0 and the Dchip software version 1.1 (18). The detection of a

particular gene, called ‘‘present’’, ‘‘absent’’, or ‘‘marginal’’, was made usingthe nonparametric Wilcoxon ranked score algorithm in Microarray Analysis

Suite version 5.0. Those detection calls were later imported into and used

by the Dchip program. Scatter plots were also generated using this software

to inspect the reproducibility of the replicates as well as the degree ofchanges of the samples under comparison. Quantification of gene

expression was obtained using Dchip, which applied a model-based

approach to derive the probe sensitivity index and expression index. Thetwo indices were used in a linear regression to quantify a particular gene.

When certain probes or transcripts deviated from the model to a set extent,

they were excluded from the quantification process. Normalization of the

arrays was done using the invariant set approach. Comparative analysis ofthe samples was done based on Dchip-generated fold changes and unpaired

sample t test. Typically, we considered a P of less than 0.05, a fold change

greater than 1.5 or less than �1.5, and a signal intensity difference of greater

than 100 or less than �100 as indications of significant change in geneexpression.

To aid in discovery of the potential biological processes represented by

the differentially expressed genes identified from the microarray data, theMAPPFinder program (19) was used in conjunction with GenMAPP

program (20). These programs were developed to reveal global gene

expression profiles across all areas of biology by integrating the annotation

of the Gene Ontology Project (The Gene Ontology Consortium, 2000).Briefly, the GeneChip data were reformatted to use Genbank accession

numbers as the gene identification to query the GenMAPP database. Search

terms included the possible pathways and the gene ontology terms. A Z

score is generated with each hit to indicate the strength of the associationof the cluster of genes to the gene ontology terms or pathways discovered.

For our data, we reported the genes with of a Z score equal to or greater

than 1.

Quantitative real-time PCR. Total RNA samples used in qRT-PCR werefrom the same preparations as described for the Affymetrix GeneChip

experiments. Reverse transcription was done with MultiScribe reverse

transcriptase (Applied Biosystems, Foster City, CA) and random hexamers

as per the manufacturer’s instruction. The resulting cDNA was thensubjected to qRT-PCR. For each of the transcripts of interest identified from

the GeneChip, primers were designed from the target sequences retrieved

from the Affymetrix Probe Sequence Database using the Primer Express 2.0software (Applied Biosystems). Quantitative PCRs were carried out in

triplicates using equal amounts of each cDNA sample equivalent to 50 ng of

starting total RNA. Each reaction contained the fluorescent indicator

SYBR Green I dye and 6 AL of each respective forward and reverse primer(5 Amol/L) in a total volume of 50 AL. Amplification PCR and monitoring of

the fluorescent emission in real-time were done in an ABI Prism 7900HT

Sequence Detection System (Applied Biosystems) as recommended by the

manufacturer (ABI SYBR Green Protocol). The data collected from thesequantitative PCRs defined a threshold cycle (Ct) of detection for the target

or the housekeeping genes in each cDNA sample.

To convert the Ct value into a relative abundance of target and

housekeeping gene per sample, a standard curve was generated for thehousekeeping gene using serial dilutions of cDNA sample: an arbitrary value

of template was first assigned to the highest standard and then

corresponding values were assigned to the subsequent dilutions, and theserelative values were plotted against the Ct value determined for each

dilution, resulting in the generation of the standard curve. The relative

amount of target and housekeeping genes in each sample was then

determined using the comparative Ct method (Applied Biosystems). Therelative quantity of target, normalized to an endogenous reference (usually a

housekeeping gene) and relative to a calibrator (the Rox reference dye), is

given by: relative quantity = 2� DDCt, where DDCt represents the difference

in Ct between the transcript and the housekeeping gene for the same RNAsample. The ratio of the relative quantities for the treated sample and the

experiment sample was used to derive the fold change. ANOVA was then

used to determine the mean and SE for each comparison.Immunohistochemistry. Cryosections (8 Am thick) were fixed with cold

acetone for 5 minutes and rinsed for 10 minutes in TBS. Paraffin sections

used for CD68 detection were first deparaffinized by xylol and ethanol

incubations and washed in TBS. Sections were blocked for 1 hour with 10%FCS in TBS before applying the primary antibodies diluted in TBS-FCS for

16 hours at 4jC. The sections incubated with a primary goat antibody were

also incubated with a bridging mouse anti-goat antibody (1:100, DAKO

Envision, Hamburg, Germany) for 30 minutes. After 3 � 15–minute washes,bound antibodies were detected with alkaline phosphatase–labeled anti-

mouse/anti-rabbit polymer (DAKO Envision) and neofuchsin as substrate.

Nuclei were counterstained with hematoxylin solution for 1 minute

(Shandon, Pittsburgh, PA).The following antibodies were used: monoclonal antibody directed

against human fibroblasts (anti-Thy-1, 1:50, Dianova, Hamburg, Germany),

Table 4. Comparisons of expression levels measured for selected genes from HS-68 fibroblasts grown in coculture withA2058, BLM, WM-164, and SK-Mel-28 melanoma cells

Gene name Accession

no.

GeneChip measurement for HS-68

cells cocultured with A2058 cells

qRT-PCR measurement for HS-68

cocultured with melanoma cells

A2058, 95%

confidence interval

BLM, 95%

confidence interval

Fold

change

P

Fold

change

Range P Fold

change

IL-8 M17017 6.79 0.00002 7.14 3.95-12.91 0. 00463 35.71

GRO2 (CXCL2) M36820 6.59 0.00089 6.31 3.49-11.42 0.00004 18.55

GRO1 (CXCL1) X54489 3.18 0.00472 3.11 1.72-5.62 0.00001 28.10MMP-1 M13509 2.57 0.00046 1.30 1.18-1.44 0.01083 2.06

IL-1h X04500 2.45 0.00543 2.14 1.84-2.48 0.00047 9.59

MMP-3 X05232 1.54 0.02917 1.37 0.81-2.31 0.00023 3.27





monoclonal antibody to human IL-1h (1:100, BioSource/Laboserv, Giessen,

Germany), goat anti-human GROa (1:50, R&D Systems, Wiesbaden,Germany), IL-8 (1:50, Santa Cruz Biotechnology, Inc., Heidelberg, Germany),

and mouse anti-human CD68 (1:25, DAKO Envision).

Results

Gene expression analysis. Within the 16 HGU95AV2 arraysused for the study, the percentage of genes determined to bepresent ranged from 49.6% to 55.4% with a median of 54.3%. Theaverage correlation coefficient between biological duplicates was>0.99. In a global view, the two different cell lines display quitedifferent expression profiles as evidenced by the correlationcoefficients extracted from the individual expression values forthe f12,000 genes for each sample under various conditions (seeSupplementary Data for details). The correlation coefficients withinthe A2058 cells have a mean of 0.985152 and a SD of 0.007936, andthose within the human fibroblast HS-68 cells have a mean of0.9763 and a SD of 0.01493. The correlation coefficients between thetwo cell lines have a mean of 0.876462 and a SD of 0.008629.Effect of coculture on the gene expression profiles of

melanoma A2058 cells. The change in the gene expression profileof the melanoma A2058 cells grown in coculture with the humanHS-68 fibroblasts compared with growth as a monoculture alonewas relativity modest. About 21 genes were determined to havesignificant fold changes compared with the gene expression profileof the melanoma A2058 cells cultured alone (Table 1).Dopachrome tautomerase , an enzyme involved in melanin

synthesis and melanocyte cell proliferation (21), was observed tobe up-regulated in A2058 cells cocultured with fibroblasts,indicating a growth stimulatory effect of fibroblast coculture onthe A2058 cells.The gene showing the greatest down-regulation in A2058

melanoma cells when cocultured with fibroblasts was IL-11 . Thiscytokine has been shown to be anti-inflammatory, functioning byinhibiting the secretion of proinflammatory cytokines by macro-phages (22, 23). Another gene that was down-regulated in thecocultured A2058 cells is the inhibitor of DNA-binding domain-1 .In mammary epithelial cell culture, inhibitor of DNA-bindingdomain-1 expression induces apoptosis (24); hence, in coculturewith fibroblasts, the A2058 cells with a decreased level of inhibitorof DNA-binding domain-1 may gain some resistance to apoptosis.

The results of analysis of the gene expression data by theMAPPFinder and GenMAPP programs to detect expression patternsassociated with biological ontology and pathways, respectively, areshown in Table 2. These data identified nine ontology classes and sixpathways as being significantly populated with up-regulated genesin the cocultured melanoma cells. Interestingly, two pathwaysassociated with chemokine signal transduction, the G13 signalingpathway (guanine nucleotide-binding protein) and small ligand Gprotein-coupled receptors (GPCR), were identified. The inflamma-tory response pathway was also identified as being active. Thissuggests involvement of the proinflammatory process as a result ofcoculture of A2058 melanoma cells with fibroblasts. In the up-regulated biological process class, melanin biosynthesis had thegreatest Z score. Up-regulation of this process class suggests anenhanced growth of the melanoma cells grown in coculture with thefibroblasts possibly due to a response to GROa (CXCL1) productionby cocultured HS-68 fibroblasts (see below). In addition, the defenseresponse biological process class was up-regulated.As seen in Table 2, eight pathway classifications and five

biological process classes were identified to be populated by asignificant number of down-regulated genes in the cocultured cells.Effect of coculture on the gene expression profiles of human

HS-68 fibroblasts. Quantitatively and qualitatively, the change inthe gene expression profile of the HS-68 fibroblast cells grown incoculture with the human A2058 melanoma cells was in sharp con-trast to that observed for the cocultured melanoma cells. From thegene expression data of cocultured HS-68 fibroblasts, 85 genes weredetermined to have significant fold changes compared with the geneexpression profile of the HS-68 fibroblasts cultured alone (Table 3).From the literature, 23 of the 85 genes that were altered have beenimplicated in processes associated with proinflammatory response,cell growth, proteolysis of ECM, and tumor invasion and metastasis.The GenMAPP and MAPPFinder programs identified 8 biological

pathways and 11 biological process classes up-regulated. In the up-regulated pathways, the blood clotting cascade had the greatestZ score (5.770). Eighteen genes were identified in that pathwayfrom the data set representing 90% of members of the pathway.The inflammatory response biological process class had the

greatest Z score (9.996) in the up-regulated group, with 132 genesin the class being identified from the data set (82% of the classmembers).

Table 4. Comparisons of expression levels measured for selected genes from HS-68 fibroblasts grown in coculture withA2058, BLM, WM-164, and SK-Mel-28 melanoma cells (Cont’d)

WM-164, 95%

confidence interval

SK-Mel-28, 95%

confidence interval

Range P Fold

change

Range P Fold

change

Range P

33.08-38.54 9.60278E�08 1.41 1.31-1.53 0.00042 1.35 1.23-1.47 0.23046

15.36-22.41 1.59039E�05 1.31 1.08-1.58 0.11886 1.39 1.15-1.67 0.46001

24.90-31.71 1.32284E�06 0.88 0.63-1.23 0.57749 2.33 2.06-2.62 0.002271.86-2.28 0.000291108 0.71 0.64-0.78 0.01319 0.80 0.65-0.97 0.13903

8.41-10.94 1.15074E�05 0.93 0.86-1.00 0.14278 1.28 1.05-1.57 0.00401

2.95-3.62 4.2211E�05 1.80 1.66-1.94 1.2E�05 2.24 2.02-2.49 0.00012

Cancer Research




Nine biological pathways and 25 biological process ontologyclasses were associated with the down-regulated gene expressiondata. Interestingly, the inflammatory response pathway wasidentified from the down-regulated gene set as having the greatestZ score (3.975) populated by 29 genes representing 94% of thepathway members, whereas the humoral immune responsebiological process class had the greatest Z score (5.075) of thiscategory with 23 genes (85% of the members) identified in this class.Based on the data, it seems that there is a significant response in

HS-68 fibroblasts to coculture with A2058 melanoma cells, muchmore so than that observed with the A2058 cells cocultured withfibroblasts. As noted from the list of genes whose expression wassignificantly altered in the HS-68 fibroblasts and the results fromthe pathway and biological process class surveys, the fibroblastsrespond to coculture with a proinflammatory response accompa-nied by other changes that have typically been associated withpromotion of tumor invasion and metastasis.Quantitative real-time PCR. Table 4 shows the results of qRT-

PCR on six genes based on identification from the gene analysisdata of cocultured HS-68 fibroblasts, which we determined to be ofparticular interest due to their role in the proinflammatorypathways and matrix degradation. These results, in terms of foldchanges, show relatively good concordance with the microarraydata in terms of fold change of gene expression.qRT-PCR results are also shown in Table 4 for HS-68 fibroblasts

cocultured with three additional human melanoma cell lines: BLM,WM-164, and SK-Mel-28. The human melanoma cell line BLM is ahighly invasive melanoma that readily forms metastases in nudemice (25). The human melanoma lines WM-164 and SK-Mel-28 areonly modestly invasive (26, 27). The BLM melanoma cells exhibitedsignificant fold changes for the six genes analyzed, whereas the foldchanges for the WM-164 and SK-Mel-28 cell lines were modestcompared with the BLM and A2058 cells.Immunohistochemistry. To corroborate the in vitro gene

expression results, immunohistochemical studies on invasive humanmelanoma specimens were done. Nevus and melanoma biopsies ofprimary tumors were analyzed for the presence of selected genes(e.g., GROa, IL-1h, and IL-8) that were surprisingly up-regulated infibroblasts in cocultures. Antibodies against CD68 and fibroblast-specific antigen (Thy-1) were used to identify macrophages andfibroblasts in the tissue, respectively. Figure 1 is a representative ofseveral biopsies that were examined. In a nodular melanoma with amaximal tumor thickness of 2.3 mm, CD68-positive macrophageswere detected between the tumor cells, whereas Thy-1-positive,spindle-shaped fibroblasts embedded in a spare stroma surroundedthe melanoma cell nests. GROa staining was mainly associatedwith the stromal fibroblasts located close to the tumor cells. IL-8 andIL-1h staining was detected in the stromal fibroblasts adjacent to thetumor cells but also in somemelanoma cells. In the congenital nevus,some macrophages were depicted in the vicinity to the nevus cellnests at the junctional zone. Thy-1-positive fibroblasts, however,were not stained for IL-8, IL-1h, and GROa.Table 5 presents a summary of an immunohistochemical survey

for the expression of GROa, IL-1h, and IL-8 in biopsies from ninehuman malignant melanomas. In all sections, strong positivestaining associated with fibroblasts was observed for IL-1h and IL-8,with modest staining observed for GROa. These data corroboratethe representative immunohistochemistry shown in Fig. 1.Interestingly, there did not seem to be any significant correlationof staining intensity to whether the biopsies were from nodularmalignant melanomas or superficial spreading melanomas.

Discussion

The relationship between host and tumor has recently beenappreciated to be a dynamic one, whereby the environment of hostas represented by the stroma, tissues, organs, and ECM affects thebehavior of the tumor and similarly the presence of the tumor has areciprocal effect on the host. This interplay between the host andthe tumor may in fact be a critical factor in terms of whether thehost eliminates the tumor or the tumor continues to progress andmetastasize (28). As early as 1863, cancer was noted to often beassociated with chronic inflammation and subsequent cell prolif-eration, and over the years, anti-inflammatory therapies have beensuggested as an approach for cancer prevention and treatment(13, 29). Thus, the host-tumor interaction can result in a variety ofproinflammatory processes that may lead to the successfulimmunologic abatement of the tumor by the host (13). However,if the proinflammatory response resulting from host-tumorinteraction is blunted, such that a full, effective immune responsedoes not occur, then an environment may occur that could promotetumorigenesis, tumor progression, and metastasis (1).Effect of coculture of A2058 melanoma cells and HS-68

fibroblasts. The design of the experimental system for thecoculture of melanoma cells with fibroblasts in the context offibrillar collagen allowed for the exchange of soluble mediators inthe culture medium yet segregated the cells. The pore size of themembrane separating the two cell lines was such that physicalinteraction of invadopods from the cells could occur but not cellularpassage through the membrane. As seen from Table 1, there isrelatively little change in the gene expression profile of A2058 cellsgrown in coculture with fibroblasts compared with growth alone infibrillar collagen. However, there seems to be a significant effect ofcoculture on the gene expression profile of HS-68 fibroblasts in thepresence of A2058 cells (Table 3). This suggests the possibility thatthe fibroblasts are more responsive to the pool of soluble mediatorsproduced by the two cell lines than the melanoma cells. It could alsobe interpreted that the melanoma cells are constitutively activatedand not responsive to factors released by fibroblasts.Proinflammatory environment in cocultured A2058 cells

and HS-68 fibroblasts. From the gene expression profile of thecocultured fibroblasts, there are a variety of changes in geneexpression that could be considered to be of a proinflammatorynature (Tables 3 and 6). For example, the proinflammatorychemokine IL-8 (CXCL8) and the inflammatory cytokine IL-1bwere observed to increase 6.8- and 2.4-fold, respectively. The up-regulation of these genes is indicative of the development of aproinflammatory/inflammatory environment resulting from thecoculture of melanoma cells and fibroblasts. IL-8 is produced bymany cell types, and in addition to being a neutrophil chemo-attractant, it has been shown to have other activities that can beconsidered prometastatic (30, 31). However, in a recent report, Liet al. (32) showed a direct role of IL-8 in angiogenesis by inducingMMP-2 synthesis, inhibiting endothelial cell apoptosis, andenhancing antiapoptotic gene expression. It is well establishedthat melanoma cells constitutively express increased amounts ofIL-8. Melanoma cell–induced expression of this cytokine byadjacent stromal fibroblast therefore would amplify enhancedneovascularization of the tumor.IL-1 receptor antagonist (IL-1RA) was observed to be modestly

up-regulated (1.6-fold) in cocultured fibroblasts (Table 3). IL-1RAbinds to the IL-1 receptor, thereby inhibiting interaction with IL-1.IL-1 is a key, pleiotropic inflammatory cytokine that promotes





angiogenesis, tumor growth, and metastasis (33). IL-1RA is oftencoexpressed with IL-1, and depending on the ratio betweencytokine and its antagonist, there is diminution of the biologicaleffects of IL-1. For example, IL-1 is constitutively expressed at highlevels in the metastatic melanoma cell line SMEL (34). When thiscell line is transduced with IL-1RA, tumor growth resulting fromthe transduced cells was significantly reduced in athymic nudemice, as was metastasis to the lung, clearly indicating a role for IL-

1RA in modulating IL-1 activities. In addition, IL-1 is required forangiogenesis by stimulating the proliferation of endothelial cellsand the synthesis of adhesion molecules and proteolytic enzymeson endothelial cells (35). Using IL-1a and IL-1h knockout mice,these authors were able to show that local tumor or lungmetastases of B16 melanoma cells were not observed comparedwith wild-type animals. In addition, vascularization of melanomacell–populated Matrigel plugs by endothelial cells was observed in

Figure 1. Immunohistologic staining of Thy-1(fibroblast-specific antigen), CD68 (macrophageantigen), IL-8, GROa, and IL-1h in specimens of arepresentative sample of human congenital nevusand a primary nodular melanoma. Serial sections(8 Am) of paraffin-embedded tissue were stainedusing monoclonal or polyclonal antibodies raisedagainst Thy-1, CD68, IL-8, GROa, and IL-1h.Bound antibodies were detected by alkalinephosphatase and neofuchsin as a substrate andcounterstained with hematoxylin. Magnification,�20. Bar , 50 Am.

Cancer Research




wild-type and inhibited by the addition of IL-1RA, whereas in-growth of endothelial cells was absent in IL-1h knockout mice,suggesting that host-derived IL-1 has an important impact ontumor growth and metastasis.In our studies of the cocultured fibroblasts, we have not examined

the protein levels of either IL-1h or IL-1RA and hence have no realunderstanding of the relative amounts of these proteins; however,from the perspective of the gene expression data, both the inten-sities and the fold changes were greater for IL-1b than IL-1RA .Therefore, we would expect that the net result is that the IL-1RAexpression in this system may serve to attenuate the inflammatoryresponse due to IL-1h, but it is not likely to completely ablate it.The down-regulation of IL-11 represents another interesting

change in the gene expression profile of the cocultured A2058 cellsthat could play a role in an inflammatory environment. As seen inTable 1, gene expression of IL-11 was down-regulated. IL-11 isinvolved in the regulation of type I cytokine proinflammatorypathways and has been shown to decrease keratinocyte prolifer-ation and cutaneous inflammation by attenuating the expressionof a variety of critical genes associated with inflammation anddisease. Therefore, it is reasonable that the down-regulationobserved for IL-11 in cocultured A2058 cells could contribute toa proinflammatory state.From the MAPPFinder and GenMAPP analyses of the gene

expression data for pathways and biological process classes thatare altered due to the coculture of melanoma cells and fibroblasts,it was seen that the inflammatory response pathway was identifiedas being up-regulated in the cocultured melanoma cells (Table 2).However, closer consideration of this suggests that this pathwaymay not be fully engaged. Of the 31 genes associated with thepathway, 28 were detected as ‘‘present,’’ although only 2 of thesegenes were actually changed due to the coculture conditions. TheZ score for this pathway was 1.9, and although significant, it isrelatively low; hence, it is probable that the up-regulation of thispathway under these conditions is rather modest and does notrepresent a fully effective inflammatory response.

One interesting consideration whether the conditions for acomplete inflammatory response could be developed under theseconditions stem from the MAPPFinder and GenMAPP analyses ofthe gene expression data of cocultured HS-68 fibroblasts. Fromthese data, the inflammatory response biological process class wasidentified as being up-regulated (Table 6) with a highly significantZ score, yet the inflammatory response pathway was identified asbeing down-regulated. We interpret these superficially conflictingdata to be due to the fact that a sufficient variety of genesassociated with proinflammation and inflammation are presentand up-regulated and hence identified as an up-regulated biologicalprocess class. However, they are not sufficient in number or foldchange to be considered as an up-regulated pathway. This, inconjunction with the down-regulation of IL-11 in the A2058 cells,leads us to conclude that it is likely that only a partial, incompleteproinflammatory/inflammatory response is being developed due tothe coculture of the A2058 cells with the fibroblasts.Effect of coculture on cell growth, proliferation, matrix

degradation, and angiogenesis. There is a close relationshipbetween cell growth, proliferation, matrix degradation, andangiogenesis and the proinflammatory pathway is considered tobe central to these processes in melanoma (36). Although we didnot assay for cellular growth and proliferation in these studies, it islikely that these events were occurring in the coculture systembased on the results from the gene expression profiles. Epiregulinwas observed to be up-regulated in the cocultured fibroblasts(Table 3). Epiregulin is a member of the epidermal growth factorfamily, and in addition to its abilities to stimulate cell growth, it hasbeen implicated in the pathobiology of pancreatic ductaladenocarcinoma (37). The gene for immediate-early response wasalso seen to be up-regulated in the cocultured fibroblasts. This genehas been shown to function to protect cells from Fas ligand orTNF-a-induced apoptosis and its up-regulation in fibroblaststherefore may serve to protect these cells from apoptosis underthe conditions of inflammation during tumor invasion (38).Matrix degradation has been shown to play a crucial role during

tumor invasion and to involve various classes of proteases,including MMPs. Genes of MMP-1 and MMP-3 were found to beup-regulated in fibroblasts cocultured with melanoma cells. Bothproteases play a key role in the degradation of fibrillar collagensand have been immunolocalized at the host-tumor junction at adermal invasion zone (39).The primary driver of cell growth, proliferation, matrix

degradation, and angiogenesis in invasive melanoma is likely IL-1h (40). Stimulation of melanoma cells by IL-1h causes theexpression of the chemokine IL-8, which can result in a widevariety of biological responses, including proliferation of keratino-cytes and melanoma cells, haptotatic migration of the melanomacells, and induction of angiogenesis (41–44). Interestingly, theseactivities seem to have some dependence on the environment ofthe tumor cells. IL-8 production in A375 melanoma cells is highwhen these cells are cocultured with human keratinocytes thatproduce IL-1, whereas when A375 cells are cocultured withhepatocytes that do not produce IL-1 their IL-8 level of expressionwas decreased (45). These studies indicate the role of the tumorenvironment in terms of promoting processes associated withtumor progression and metastasis. In our studies of the cocultureof A2058 human melanoma cells with HS-68 primary humanfibroblasts in fibrillar collagen, no up-regulation of chemokine orcytokine expression was observed in the A2058 cells; however, inthe cocultured fibroblasts, IL-1b , IL-8 , GROb , GROa , and CCL2

Table 5. Expression of GROa, IL-1h, and IL-8 in humanmalignant melanomas

Specimen Depth

(mm)

Melanoma

type

GROa IL-1h IL-8

1 1.61 SSM ++ ++ ++

2 4.5 NMM ++ +++ +++

3 2.25 NMM +++ +++ ++4 4.05 SSM + +++ +++

5 8 SSM + +++ +++

6 3.51 SSM ++ +++ +++

7 2.8 SSM + +++ +++8 3.6 NMM + +++ +++

9 3.15 SSM + +++ +++

NOTE: Expression of cytokines and chemokines was analyzed by

immunohistochemistry on cryosections of melanomas from ninedifferent patients. Intensities of specific stainings were arbitrarily set

as the following: +, modest expression; ++, moderate expression; +++,

strong expression. SSM, superficial spreading melanoma; NMM,

nodular malignant melanoma.





(monocyte chemoattractant protein-1) were observed to be up-regulated (Table 3). Overexpression of CXCL1-3 by stimulatedmelanoma cells has been shown to function in an autocrine fashionto promote melanoma cell growth and proliferation (46, 47). Basedon our data, it is quite possible that host stroma, in this caserepresented by melanoma-stimulated fibroblasts, can also expressthese chemokines to promote melanoma growth and proliferation.As noted above, the chemokine CCL2 was also up-regulated in

fibroblasts in coculture with melanoma cells. This chemokine has

been shown to be important in attracting macrophage infiltrates totumors, a process that has been linked to tumor progression,invasion, and angiogenesis (48, 49). As observed in our experi-ments, the stimulation of cocultured fibroblasts by the A2058 cellspromotes an environment rich in chemokines known to stimulatemelanoma proliferation, synthesis of proteolytic enzymes, tumorcell invasion, and angiogenesis associated with macrophageinvasion, underscoring the potential for a central role of hoststroma in melanoma metastasis.

Table 6. Pathways and biological process classes associated with differentially expressed genes in cocultured HS-68fibroblasts

No. genes

changed

No. genes

measured

No. genes

on MAPP

% Changed % Present Z score

Up-regulated genesPathways

Blood clotting cascade 3 18 20 16.7 90 5.770

Apoptosis 2 42 43 4.8 97.7 1.988

Eicosanoid synthesis 1 14 16 7.1 87.5 1.936Wnt signaling 2 45 62 4.4 72.6 1.869

Nucleotide metabolism 1 16 16 6.2 100 1.754


MMPs 1 22 31 4.5 71 1.349Calcium channels 1 22 28 4.5 78.6 1.349


Inflammatory response 11 132 161 8.3 82 9.996

Chemotaxis 7 82 97 8.5 84.5 7.455Calcium ion homeostasis 3 19 19 15.8 100 7.342

Epidermal growth factor receptor signaling pathway 2 9 9 22.2 100 5.891

Response to pathogenic bacteria 2 9 10 22.2 90 5.613Xenobiotic metabolism 3 35 52 8.6 67.3 4.739

Blood coagulation 4 63 68 6.3 92.6 4.340

Angiogenesis 2 23 26 8.7 88.5 3.832

Antiapoptosis 4 52 66 7.7 78.8 3.591Mitogen-activated protein kinase kinase kinase cascade 2 7 7 28.6 100 3.161

Superoxide metabolism 3 11 11 27.3 100 3.142

Down-regulated genesPathways

Inflammatory response pathway 9 29 31 31 93.5 3.975

Methionine metabolism 3 7 7 42.9 100 3.023

GPCRs, class B secretin-like 5 19 29 26.3 65.5 2.515

Ovarian infertility genes 6 27 31 22.2 87.1 2.270Nucleotide metabolism 4 16 16 25 100 2.128

Alanine-aspartate metabolism 4 17 17 23.5 100 1.986

GPCRs, other 4 19 178 21.1 10.7 1.731Wnt signaling 7 45 62 15.6 72.6 1.402


G protein signaling 9 63 80 14.3 78.8 1.314

Biological process classesHumoral immune response 7 23 27 30.4 85.2 5.075

Development 57 346 479 16.5 72.2 4.962

RAS protein signal transduction 10 24 24 41.7 100 4.697

Signal transduction 68 699 895 9.7 78.1 4.494Actin filament organization 3 4 6 75 66.7 4.358

Methionyl-tRNA aminoacylation 2 2 3 100 66.7 4.265

Proline biosynthesis 4 4 11 100 36.4 4.265Cell surface receptor linked signal transduction 17 110 131 15.5 84 4.173

Muscle contraction 14 59 72 23.7 81.9 3.792

Physiologic processes 2 12 18 16.7 66.7 3.769

Cell adhesion 44 302 441 14.6 68.5 3.725Skeletal development 15 73 90 20.5 81.1 3.633

Intracellular signaling cascade 59 375 596 15.7 62.9 3.487

Cancer Research




Taken together, there appears from our gene expression data tobe an activation of a pathway(s) initiated with IL-1h. IL-1 has beenshown to activate nuclear factor-kB, which in turn can up-regulatethe expression of MMPs, urokinase-type plasminogen activator,CXCL1-3, chemokines, and TNF-a (49). TNF-a has also been shownto up-regulate IL-8 and CCL2 (50). In the cocultured fibroblasts, weobserved the up-regulation of IL-1b , nuclear factor-kB , MMP-1 ,MMP-3 , tissue plasminogen activator, CXCL1 , and CXCL2 . Further-more, we also saw IL-8 and CCL2 up-regulated, but not TNF-a .Nevertheless, it could be that TNF-a protein levels, but not geneexpression, was increased in the coculture system because the genefor TNF-a-induced protein 6 was up-regulated, suggesting an effectof TNF-a on the system.Relevance of gene expression data to invasive human

melanoma. Over the years, many experimental and clinical effortshave aimed to improve the criteria for melanoma diagnosis andtreatment. To improve the knowledge in molecular pathology ofmelanoma, particularly in discriminating between very earlymelanoma lesions and benign nevi, cDNA/oligonucleotide arraytechnology is increasingly being used to identify new biologicalmarkers for malignancy or for invasive potential. However, analysisof tumor tissue alone may not identify small changes in geneexpression occurring in the microenvironment of tumor-stroma.Identification of regulated genes resulting from cross-talk andactivation due to coculture of melanoma cells and fibroblasts fromour studies have provided insight into the molecular mechanismsinvolved in tumor-stroma interactions.Our studies have shown that coculture of melanoma cells and

fibroblasts has a profound reciprocal effect on gene expressionprofiles, with dramatic alterations in fibroblasts compared with amore modest effect on the melanoma cells. The fibroblasts werefound to be more easily activated and responded to coculture byup-regulation of a large number of genes associated with theproinflammatory pathway, matrix proteolysis, and cellular prolif-eration. Interestingly, cytokines and chemokines (e.g., IL-1a,

CXCL1/GROa, and CXCL8/IL-8) that were thought to be producedby inflammatory cells or by melanoma cells were identified to beup-regulated in fibroblasts, suggesting that these cells are likely toalso contribute to the inflammatory reaction.In this study, we extended the coculture experiments of A2058

cells, a modestly invasive human melanoma cell line with HS-68fibroblasts to other human melanoma cell lines. The humanmelanoma BLM cell line has been shown to be highly metastaticwith early and frequent formation of metastasis in nude mice afters.c. inoculation (25). SK-Mel-28 and WM-164 melanoma cells are alow invasive cell lines (26) and WM-164 melanoma cells show noorgan metastasis unless injected i.v. in mice (29). Interestingly, weobserved that the fibroblasts grown in coculture with the highlymetastatic BLM melanoma cell lines showed the significantlygreater fold changes in the six relevant gene transcripts comparedwith the low invasive A2058, SK-Mel-28, and WM-164 melanomacell lines. It is interesting to speculate that melanoma with agreater invasive potential may elicit a greater ‘‘proinvasion’’ effecton the gene expression profiles of stromal cells.Therefore, we conclude that stroma, due to signals originating

from the tumor, by responding to the presence of transformed cellsis primarily responsible for the generation of a microenvironmentthat is proinflammatory, proproliferation, proinvasion, and prom-etastatic. Studies directed at disrupting the cross-talk between hostand tumor may define new strategies for therapeutic intervention.

Acknowledgments

Received 2/6/2004; revised 1/6/2005; accepted 3/1/2005.Grant support: Polifarma S.p.A (J.W. Fox), University of Virginia Cancer Center

(J.W. Fox), Alexander von Humboldt Stiftung (J.W. Fox), Wilhelm Sander Foundationgrant 1999.093.2 (R. Nischt and C. Mauch), Koeln Fortune Program of the Faculty ofMedicine grant 10/2004 (P. Zigrino), Deutsche Forschungsgemeinschaft through theSFB 587 at the University of Cologne (C. Mauch), and German Research Foundationgrant Ni 304/9-1 (R. Nischt).

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

References1. Coussens LM, Werb Z. Inflammation and cancer.Nature 2002;240:860–7.

2. Wernert N. The multiple roles of tumour stroma.Virchows Arch 1997;430:433–43.

3. Fidler I. The pathogenesis of cancer metastasis: the‘‘seed and soil’’ hypothesis revisited. Nat Rev Cancer2002;3:1–6.

4. Tlsty TD. Stromal cells can contribute oncogenicsignals. Cancer Biol 2001;11:97–104.

5. Mueller MM, Fusenig NE. Tumor-stroma interactionsdirecting phenotype and progression of epithelial tumorcells. Differentiation 2002;70:486–97.

6. Brown LF, Guidi AJ, Schnitt SJ, et al. Vascular stromaformation in carcinoma in situ , invasive carcinoma, andmetastatic carcinoma of the breast. Clin Cancer Res1999;5:1041–56.

7. Radisky D, Muschler J, Bissell MJ. Order and disorder:the role of extracellular matrix in epithelial cancer.Cancer Invest 2002;20:139–53.

8. Bissell MJ, Radisky D. Putting tumours in context. NatRev Cancer 2001;1:46–54.

9. Lukashev ME, Werb Z. ECM signaling: orchestratingcell behaviour and misbehaviour. Trends Cell Biol1998;8:437–41.

10. Thomasset N, Lochter A, Sympson CJ, et al.Expression of autoactivated stromelysin-1 in mam-mary glands of transgenic mice leads to a reactivestroma during early development. Am J Pathol1998;153:457–67.

11. Kuper H, Adami HO, Trichopoulos D. Infections as amajor preventable cause of human cancer. J Intern Med2000;248:171–83.

12. Torisu H, Ono M, Kiryu H, et al. Macrophageinfiltration correlates with tumor stage and angio-genesis in human malignant melanoma: possibleinvolvement of TNFa and IL-1a. Int J Cancer 2000;85:182–8.

13. Brigati C, Noonan DM, Albini A, Benelli R. Tumorsand inflammatory infiltrates: friends or foes? Clin ExpMetastasis 2002;19:247–58.

14. Shih IM, Elder DE, Hsu MY, Herlyn M. Regulation ofMel-CAM/MUC18 expression on melanocytes of differ-ent stages of tumor progression by normal keratino-cytes. Am J Pathol 1994;45:837–45.

15. Halaban R. Growth factors and melanomas. SeminOncol 1996;23:673–81.

16. Van den Oord JJ, Paemen L, Opdenakker G, de Wolf-Peeters C. Expression of gelatinase B and extracellularmatrix metalloproteinase inducer EMMPRIN in benignand malignant pigment cell lesions of the skin. Am JPathol 1997;151:665–70.

17. Kurschat P, Wickenhauser C, Groth W, et al.Identification of activated matrix metalloproteinase-2(MMP-2) as the main gelatinolytic enzyme in malignantmelanoma by in situ zymography. J Pathol 2002;197:179–87.

18. Li C, Wong WH. Model-based analysis of oligonu-cleotide arrays: expression index computation andoutlier detection. Proc Natl Acad Sci U S A 2001;98:31–6.

19. Donige SW, Salomonis N, Dahlquist KD, et al.MAPPFinder: using gene ontology and GenMAPP tocreate a global gene-expression profile from microarraydata. Genome Biol 2003;4:R7. Epub.

20. Dahlquist KD, Salomonis N, Vranizan K, et al.GenMAPP, a new tool for viewing and analyzingmicroarray data on biological pathways. Nat Genet2002;31:19–20.

21. Sonesson B, Rosengren E, Hansson AS, HanssonC. UVB-induced inflammation gives increased d-dopachrome tautomerase activity in blister fluidwhich correlates with macrophage migration inhib-itory factor. Exp Dermatol 2003;12:278–82.

22. Trepicchio WL, Wang L, Bozza M, Dorner AJ. IL-11regulates macrophage effector function through theinhibition of nuclear factor-nB. J Immunol 1997;159:5661–70.

23. Trepicchio W, Ozawa M, Walters IB, et al. Interleukin-11 selectively downregulates type I cytokine proinflam-matory pathways in psoriasis lesions. J Clin Invest 1999;104:1527–37.

24. Parrinello S, Lin CQ, Murata K, et al. Id-1, ITF-2, andId-2 comprise a network of helix-loop-helix proteinsthat regulate mammary epithelial cell proliferation,differentiation and apoptosis. J Biol Chem 2001;276:39213–9.

25. van Mujen GNP, Cornelissen IMHA, Jansen KFJ,et al. Antigen expression of metastasizing and non-metastasizing human melanoma cell lines xeno-grafted into nude mice. Clin Exp Metastasis 1991;9:259–72.





26. Luca M, Hunt B, Bucana CD, et al. Direct correlationbetween MUC18 expression and metastatic potential ofhuman melanoma cells. Melanoma Res 1993;3:35–41.

27. Herlyn D, Adachi K, Koprowski H, Herlyn M.Experimental model of human melanoma metastases.Cancer Treat Res 1991;54:105–18.

28. Hsu M-Y, Meier F, Herlyn M. Melanoma developmentand progression: a conspiracy between tumor and host.Differentiation 2002;70:522–36.

29. Balkwill F, Mantovani A. Inflammation and cancer:back to Virchow? Lancet 2001;357:539–45.

30. Herbert CA, Baker JB. Interleukin-8: a review. CancerInvest 1993;11:743–50.

31. Kim SJ, Uehara H, Karashima T, et al. Expression ofinterleukin-8 correlates with angiogenesis, tumorige-nicity, and metastasis of human prostate cancer cellsimplanted orthotopically in nude mice. Neoplasia 2001;3:33–42.

32. Li A, Dubey S, Varney ML, et al. IL-8 directlyenhanced endothelial cell survival, proliferation, andmatrix metalloproteinases production and regulatedangiogenesis. J Immunol 2003;170:3369–76.

33. Pantschenko AG, Pushkar I, Anderson KH, et al. Theinterleukin-1 family of cytokines and receptors inhuman breast cancer: implications for tumor progres-sion. Int J Oncol 2003;23:269–84.

34. Weinreich DM, Elaraj DM, Puhlmann M, et al. Effectof interleukin 1 receptor antagonist gene transductionon human melanoma xenografts in nude mice. CancerRes 2003;63:5957–61.

35. Voronov E, Shouval DS, Krelin Y, et al. IL-1 is requiredfor tumor invasiveness and angiogenesis. Proc NatlAcad Sci U S A 2003;100:2645–50.

36. Mattie S, Colombo MP, Melani C, et al. Expression ofcytokine/growth factors and their receptors in humanmelanoma and melanocytes. Int J Cancer 1995;56:853–7.

37. Zhu H, Kleeff J, Friess H, et al. Epiregulin is up-regulated in pancreatic cancer and stimulates pancre-atic cancer cell growth. Biochem Biophys Res Commun2000;273:1019–24.

38. Kumar K, Kobayashi T, Warner GM, et al. A novelimmediate early response gene, IEX-1, is induced byultraviolet radiation in human keratinocytes. BiochemBiophys Res Commun 1998;253:336–41.

39. Walker RA, Wooley DE. Immunolocalization studiesof matrix metalloproteinases-1, -2 and -3 in humanmelanoma. Virchows Arch 1999;435:574–9.

40. Rennekampff HO, Hansbrough JF, Kiessig V, et al.Bioactive interleukin-8 is expressed in wounds andenhances wound healing. J Surg Res 2000;93:41–54.

41. Schadendorf D, Moller A, Algermissen B, et al. IL-8 produced by human malignant melanoma cells in vitrois an essential autocrine growth factor. J Immunol1993;151:2667–75.

42. Singh RK, Gutman M, Radinsky R, et al. Expression ofinterleukin-8 correlates with the metastatic potential ofhuman melanoma cells in nude mice. Cancer Res 1994;54:3242–7.

43. Wang JM, Taraboletti G, Matsushima K, et al.Induction of haptotactic migration of melanoma cells

by neutrophil activating protein/IL-8. Biochem BiophysRes Commun 1990;169:165–70.

44. Yoshida S, Ono M, Shono T, et al. Involvement ofinterleukin-8, vascular endothelial growth factor, andbasic fibroblast growth factor in tumor necrosisfactor a-dependent angiogenesis. Mol Cell Biol 1997;17:4015–23.

45. Gutman M, Singh RK, Xie K, et al. Regulation ofinterleukin-8 expression in human melanoma cells bythe organ environment. Cancer Res 1995;55:2470–5.

46. Luan J, Shattuck-Brandt R, Haghnegahdar H, et al.Mechanism and biological significance on the consti-tutive expression of MGSA/GRO chemokines in malig-nant melanoma tumor progression. J Leukoc Biol 1997;62:588–97.

47. Takamori H, Oades ZG, Hoch OC, et al. Autocrinegrowth effect of IL-8 and GROa on a human pancreaticcancer cell line, Capan-1. Pancreas 2000;21:52–6.

48. Leek RD, Lewis CE, Whitehouse R, et al. Associationof macrophage infiltration with angiogenesis andprognosis in invasive breast carcinoma. Cancer Res1996;56:4625–9.

49. Pantschenko AG, Pushkar I, Anderson KH, et al. Theinterleukin-1 family of cytokines and receptors inhuman breast cancer: implications for tumor progres-sion. Int J Oncol 2003;23:269–84.

50. Nozaki S, Sledge GW Jr, Nakshatri H. Cancer cell-derived interleukin 1a contributes to autocrine andparacrine induction of pro-metastatic genes in breastcancer. Biochem Biophys Res Commun 2000;275:60–2.

Cancer Research




2005;65:4134-4146. Cancer Res Paul G. Gallagher, Yongde Bao, Alyson Prorock, et al. Interactions in MetastasisMelanoma and Fibroblasts: Implications for Host-Tumor Gene Expression Profiling Reveals Cross-talk between

Updated version

http://cancerres.aacrjournals.org/content/65/10/4134

Access the most recent version of this article at:

Cited articles

http://cancerres.aacrjournals.org/content/65/10/4134.full#ref-list-1

This article cites 47 articles, 12 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/65/10/4134.full#related-urls

This article has been cited by 5 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. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://cancerres.aacrjournals.org/content/65/10/4134To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/content/65/10/4134.full#ref-list-1

http://cancerres.aacrjournals.org/content/65/10/4134.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]



gene expression profiling reveals cross-talk between melanoma … · 36618_g_at x77956 491.3 161.2...

Documents