Identification of Molecules Relevant for the Invasiveness of Fibrosarcomas and Melanomas

Helsinki University Biomedical Dissertations No. 148

IDENTIFICATION OF MOLECULES RELEVANT FOR THE INVASIVENESS OF FIBROSARCOMAS AND MELANOMAS

Pirjo Nummela

Department of Pathology Haartman Institute Faculty of Medicine

and

Division of Biochemistry Department of Biosciences Faculty of Biological and Environmental Sciences

University of Helsinki Finland

Academic dissertation

To be presented for public examination with the permission of the Faculty of Biological and Environmental Sciences of the University of Helsinki in the Lecture Hall 2 of Haartman Institute (Haartmaninkatu 3, Helsinki), on 13.5.2011 at 12 o’clock noon.

Helsinki 2011 Supervisor Docent Erkki Hölttä, M.D., Ph.D. Department of Pathology Haartman Institute University of Helsinki

Thesis committee Docent Jouko Lohi, M.D., Ph.D. Department of Pathology Haartman Institute University of Helsinki and

Pirjo Nikula-Ijäs, Ph.D. Division of Biochemistry Department of Biosciences University of Helsinki

Reviewers Professor Veli-Matti Kähäri, M.D., Ph.D. Department of Dermatology University of Turku and Turku University Hospital and

Docent Jouko Lohi, M.D., Ph.D.

Opponent Professor Jyrki Heino, M.D., Ph.D. Department of Biochemistry and Food Chemistry University of Turku

Custos Professor Kari Keinänen, Ph.D. Division of Biochemistry Department of Biosciences University of Helsinki

ISBN 978-952-92-8821-2 (paperback) ISBN 978-952-10-6924-6 (PDF) ISSN 1457-8433 http://ethesis.helsinki.fi

Helsinki University Print Helsinki 2011

To Juha, Joona, and Joel

TABLE OF CONTENTS

LIST OF ORIGINAL PUBLICATIONS ...... 7 ABSTRACT ...... 8 ABBREVIATIONS ...... 10 INTRODUCTION ...... 12 REVIEW OF THE LITERATURE ...... 14 1. CANCER DEVELOPMENT ...... 14 2. TUMOR MICROENVIRONMENT ...... 16 2.1 Stromal cells and surrounding epithelia ...... 16 2.1.1 Keratinocytes ...... 17 2.1.2 Fibroblasts ...... 18 2.1.3 Endothelial cells ...... 19 2.1.4 Inflammatory and immune cells ...... 19 2.2 Extracellular matrix molecules ...... 20 2.2.1 Collagens...... 22 2.2.2 Elastin ...... 23 2.2.3 Laminins ...... 24 2.2.4 Fibronectin ...... 24 2.2.5 Tenascins ...... 25 2.2.6 Transforming growth factor beta-induced ...... 26 3. LOCAL INVASION ...... 27 3.1 Cell adhesion and adhesion receptors ...... 27 3.1.1 Cell-cell interactions ...... 27 3.1.2 Cell-matrix interactions ...... 28 3.2 Cell migration ...... 29 3.2.1 Actin polymerization ...... 30 3.2.2 Acto-myosin contraction ...... 31 3.3 Proteolysis of extracellular matrix ...... 31 3.3.1 Matrix metalloproteinases ...... 32 3.3.2 Serine proteases ...... 34 3.3.3 Cysteine proteases ...... 34 AIMS OF THE STUDY ...... 37 MATERIALS AND METHODS ...... 38 1. CELL LINES AND TISSUE SPECIMENS (I-IV) ...... 38 2. ANTIBODIES (I-IV) ...... 40 3. RNA ANALYSES (I-IV) ...... 41 3.1 Microarray analyses (I-IV) ...... 41 3.2 Northern blot analysis (I) ...... 41 3.3 RT-PCR analysis (I-IV) ...... 41 4. PROTEIN ANALYSES (I-IV) ...... 43 4.1 Western blotting (I-IV) ...... 43 4.2 Immunofluorescence staining (I-II, IV) ...... 43 4.3 Flow cytometry (II) ...... 43 4.4 Immunoprecipitation (II) ...... 43 4.5 Assay of cathepsin L activity (II) ...... 44 5. TRANSFECTION EXPERIMENTS (I-II, IV) ...... 44 5.1 Tetracycline-inducible antisense-RNA expression vector (I) ...... 44

5.2 siRNA oligonucleotides (I-II) ...... 44 5.3 shRNA expression vectors (II, IV) ...... 45 6. FUNCTIONAL ASSAYS (I-IV) ...... 45 6.1 Adhesion assays (II, IV) ...... 45 6.2 Analysis of cell growth (I-II, IV) ...... 46 6.3 Migration assays (I, III-IV) ...... 46 6.4 Invasion assays (I-II, IV) ...... 46 7. IMMUNOHISTOCHEMISTRY (II-IV) ...... 47 RESULTS AND DISCUSSION ...... 48 1. MICROARRAY ANALYSES (I-IV) ...... 48 1.1 Incyte Genomics versus Affymetrix arrays (I-IV) ...... 48 1.2 Mouse fibrosarcoma cells (I-II) ...... 51 1.3 Human melanoma cells (III-IV) ...... 52 1.4 Fibrosarcoma versus melanoma (I-IV) ...... 54 2. FUNCTIONAL CHARACTERIZATION (I-IV) ...... 55 2.1 Thymosin β4 (I) ...... 55 2.1.1 Tβ4 is overexpressed in Amdc-s, Amdc-as, and E4 fibrosarcoma cells ...... 56 2.1.2 Tβ4 overexpression is related to the transformed phenotype ...... 56 2.1.3 LAT-A, an inhibitor of Tβ4, affects the morphology and growth of Amdc-s and -as cells ...... 57 2.1.4 LAT-A impairs the migration of Amdc-s, Amdc-as, and E4 cells ...... 58 2.1.5 LAT-A blocks the invasion of Amdc-s and -as cells in Matrigel ...... 58 2.1.6 Tβ4 shows strong immunostaining in Amdc-s-induced mouse fibrosarcomas .. 59 2.2 Integrins α6 and β7 (II) ...... 60 2.2.1 Itgα6 and Itgβ7 are up-regulated in Amdc-s, Amdc-as, and E4 fibrosarcoma cells ...... 61 2.2.2 Itgα6β1 mediates the binding of Amdc-s cells to laminin ...... 61 2.2.3 Itgα6β1 is relevant for the invasion of Amdc-s cells and human HT-1080 fibrosarcoma cells in Matrigel ...... 62 2.2.4 ITGα6 localizes into the invasion fronts of human high-grade fibrosarcomas ... 63 2.3 Tenascin-C and fibronectin (III) ...... 63 2.3.1 TN-C and FN are overexpressed in WM793 and WM239 melanoma cells and primary melanomas ...... 64 2.3.2 TN-C and FN show intense immunostaining in fibrillar structures surrounding invasive melanoma cells in tumor tissues...... 64 2.3.3 TN-C and FN stimulate the migration of melanoma cells in 3D-collagen-I ...... 66 2.4 Transforming growth factor beta-induced (IV) ...... 67 2.4.1 TGFBI is up-regulated in WM793 and WM239 melanoma cells ...... 67 2.4.2 TGFBI is an anti-adhesive protein for melanoma cells and skin fibroblasts ...... 68 2.4.3 TGFBI knockdown stimulates the migration and invasion of melanoma cells in vitro ...... 68 2.4.4 TGFBI-KD cells show highly concentrated actin filaments co-localizing with activated ITGβ1 in pseudopodia-like structures ...... 69 2.4.5 TGFBI-KD cells show remarkably impaired tumor growth in nude mice ...... 69 CONCLUDING REMARKS AND FUTURE PERSPECTIVES ...... 71 ACKNOWLEDGEMENTS ...... 73 REFERENCES ...... 74

LIST OF ORIGINAL PUBLICATIONS

This thesis is based on the following original publications, which are referred to in the text by Roman numerals I-IV.

I Nummela P, Yin M, Kielosto M, Leaner V, Birrer MJ, Hölttä E. Thymosin β4 is a determinant of the transformed phenotype and invasiveness of S-adenosylmethionine decarboxylase-transfected fibroblasts. Cancer Res 66: 701-12, 2006

II Kielosto M*, Nummela P*, Järvinen K, Yin M, Hölttä E. Identification of integrins α6 and β7 as c-Jun- and transformation-relevant genes in highly invasive fibrosarcoma cells. Int J Cancer 125: 1065-73, 2009

III Kääriäinen E*, Nummela P*, Soikkeli J*, Yin M, Lukk M, Jahkola T, Virolainen S, Ora A, Ukkonen E, Saksela O, Hölttä E. Switch to an invasive growth phase in melanoma is associated with tenascin-C, fibronectin, and procollagen-I forming specific channel structures for invasion. J Pathol 210: 181-91, 2006

IV Nummela P, Lammi J, Soikkeli J, Saksela O, Laakkonen P, Hölttä E. Transforming growth factor beta-induced (TGFBI) is an anti-adhesive protein regulating the progression of melanoma cells. Manuscript

*equal contribution

The original publications have been reprinted with the permission of the respective copyright holders. In addition, some unpublished data are presented.

ABSTRACT

Cancer is becoming the leading cause of deaths in the world. As 90% of all deaths from cancer are caused by metastasis, discovery of the mechanisms behind cancer cell invasion and metastasis is of utmost importance. Only new effective therapies targeting cancer progression can reduce cancer mortality rates.

The aim of this study was to identify molecules that are relevant for tumor cell invasion and spreading in fibrosarcomas and melanomas, and to analyze their potential for cancer biomarkers or therapeutic targets. First, the gene expression changes of normal cells and transformed cells showing high invasiveness, S-adenosylmethionine decarboxylase (AdoMetDC)-transfected murine fibroblasts and human melanoma cells, were studied by microarray analyses. The function of the identified candidate molecules were then studied in detail in these cell lines. Finally, the physiological relevance of the identified changes was studied by immunohistochemical analyses of human sarcoma and melanoma specimens or by a mouse xenograft model.

In fibrosarcoma cells, the most remarkable change detected was a dramatic up-regulation of the actin-sequestering molecule thymosin β4 (Tβ4), which was shown to be important for the transformed phenotype of the AdoMetDC-transfected cells (Amdc-s and -as). A sponge toxin latrunculin A, inhibiting the binding of Tβ4 to actin, was found to selectively inhibit the migration and invasion of these cells. Further, Amdc-s-induced mouse tumors and human high-grade sarcomas were found to show intense Tβ4 immunostaining. In addition to Tβ4, integrin subunits α6 and β7 (Itgα6 and Itgβ7) were found to be up-regulated in Amdc-s and - as cells. Itgα6 was shown to dimerize mainly with Itgβ1 in Amdc-s. Inhibition of Itgα6 or Itgβ1 function with neutralizing antibodies fully blocked the invasiveness of Amdc-s cells, and importantly also human HT-1080 fibrosarcoma cells, in three-dimensional (3D)-Matrigel mimicking tumor extracellular matrix (ECM). By immunohistochemical analyses, strong staining for ITGα6 was detected in human high-grade fibrosarcomas and other sarcomas, especially at the invasion fronts of the tumors.

In the studied melanoma cell lines, the expression levels of the adhesion-related ECM proteins tenascin-C (TN-C), fibronectin (FN), and transforming growth factor beta-induced

(TGFBI) were found to be highly up-regulated. By immunohistochemistry, intense TN-C and FN staining was detected in invasive and metastatic melanoma tumors, showing co- localization (together with procollagen-I) in tubular meshworks and channels around the invading melanoma cells. In vitro, TN-C and FN were further found to directly stimulate the migration of melanoma cells in 3D-collagen-I matrix. The third candidate protein, TGFBI, was found to be an anti-adhesive molecule for melanoma cells, and knockdown of its expression in metastatic melanoma cells (TGFBI-KD cells) led to dramatically impaired tumor growth in immunocompromized mice. Interestingly, the control tumors showed intense TGFBI immunostaining in the invasion fronts, showing partial co-localization with the fibrillar FN staining, whereas the small TGFBI-KD cell-induced tumors displayed amorphous, non-fibrillar FN staining. These data suggest an important role for TGFBI in FN fibrillogenesis and melanoma progression.

In conclusion, we have identified several invasion-related molecules, which show potential for cancer diagnostic or prognostic markers, or therapeutic targets. Based on our previous and present fibrosarcoma studies, we propose the possibility of using ITGα6 antagonists (affecting tumor cell adhesion) in combination with Tβ4 inhibitors (affecting tumor cell migration) and cathepsin L inhibitors (affecting the degradation of basement membrane and ECM proteins) for the treatment of fibrosarcomas and other tumors overexpressing these molecules. With melanoma cells, in turn, we point to the importance of three secreted ECM proteins, TN-C, FN, and TGFBI, in melanoma progression. Of these, especially the potential of TN-C as a prognostic melanoma biomarker and TGFBI as a promising therapeutic target molecule are clearly worth additional studies.

ABBREVIATIONS

AdoMetDC S-adenosylmethionine decarboxylase BM basement membrane BSA bovine serum albumin CAF cancer-associated fibroblast cFN cellular fibronectin CM conditioned media COL collagen CSF-1 colony stimulating factor-1 DAPI 4’,6-diamidino-2-phenylindole ECM extracellular matrix EGF epidermal growth factor EMT epithelial to mesenchymal transition FAK focal adhesion kinase FAS fasciclin FBS fetal bovine serum FGF fibroblast growth factor FITC fluorescein isothiocyanate FN fibronectin Gapdh glyceraldehyde-3-phosphate dehydrogenase HGF hepatocyte growth factor IHC immunohistochemistry IL interleukin Itg/ITG integrin (mouse/human) KD knockdown LAP latency-associated peptide LAT-A latrunculin A LN laminin LTBP latent TGF-β-binding protein MMP matrix metalloproteinase ODC ornithine decarboxylase pFN plasma fibronectin PDGF platelet derived growth factor POSTN periostin RGP radial growth phase RT-PCR reverse transcription polymerase chain reaction shRNA short hairpin RNA siRNA small interfering RNA Tβ4 thymosin β4 Tβ10 thymosin β10 TIMP tissue inhibitor of metalloproteinases TGF-β transforming growth factor β TGFBI transforming growth factor beta-induced TLN talin TN tenascin TNF-α tumor necrosis factor α TRITC tetramethyl rhodamine isothiocyanate uPA urokinase-type plasminogen activator

UV ultraviolet VEGF vascular endothelial growth factor VGP vertical growth phase VN vitronectin

INTRODUCTION

Worldwide cancer incidence and the number of cancer-related deaths are rising all the time. In Finland, there are currently more than 205,000 persons with cancer diagnosed at some point in their life (some of them are cured), and over 10,000 persons die of cancer every year (Finnish Cancer Registry 2009). The corresponding worldwide numbers are as high as 25 million and over 7 million, respectively (Ferlay et al. 2004). Regardless of the improved cancer diagnostics and treatments leading to decreasing age-adjusted cancer mortality rates, the highly increasing incidence rate, especially in the developing countries, results in a steady rise in the number of cancer victims. In Finland, it has been estimated that by 2020, there will be annually about 25-30% more new cancer cases and 10-20% more cancer deaths (Finnish Cancer Registry 2009). As the main cause (90%) for cancer-related deaths is the dissemination of the cancer, metastasis (Gupta and Massagué 2006), discovery of the molecules and mechanisms relevant for cancer cell invasion and metastasis is highly important. New, more effective therapies targeting cancer progression are needed to stop the rise in worldwide cancer mortality.

In this study, we investigated mouse fibroblasts transformed by the polyamine biosynthetic enzyme S-adenosylmethionine decarboxylase (AdoMetDC), along with human fibrosarcoma specimens, human melanoma cells, and human melanoma specimens. The polyamines spermidine and spermine, and their diamine precursor putrescine, are small polycationic molecules essential for normal cell growth (for a review see Pegg 1988; Pegg 2009). Polyamines function by binding to various anions in the cells, including DNA, RNA, and phospholipids, thereby stabilizing these molecules. Further, polyamines are known to influence the transcription of many genes, to regulate ion channels, and to affect the activity of many enzymes. AdoMetDC and ornithine decarboxylase (ODC) are the two key regulatory enzymes of the tightly controlled biosynthetic pathway of polyamines. ODC catalyzes the first step, synthesis of putrescine, and AdoMetDC catalyses the formation of decarboxylated S-adenosylmethionine, which is needed in the conversion of putrescine to spermidine and spermine. As an indication of the need for tightly controlled polyamine biosynthesis to control cell growth, aberrant polyamine metabolism is known to be related to carcinogenesis (Pegg 1988 and the references therein). Interestingly, overexpression of both ODC (Auvinen et al. 1992; Moshier et al. 1993; Shantz and Pegg 1994) and AdoMetDC (Paasinen-Sohns et

al. 2000), by themselves, has been shown to induce tumorigenic cell transformation of rodent cells. In addition, also blockage of AdoMetDC expression has been found to induce cell transformation (Paasinen-Sohns et al. 2000). The AdoMetDC-transformed mouse fibroblasts, transfected in either sense (Amdc-s) or antisense orientation (Amdc-as), form highly invasive fibrosarcomas in nude mice, rapidly penetrating from the subcutaneous injection site in the flanks into the peritoneal cavity. In addition, these cells are in vitro even more invasive than the c-Ha-rasVal12-oncogene-transformed murine fibroblasts (E4), human HT-1080 fibrosarcoma cells, or human MDA-MB-231 breast cancer cells (Ravanko et al. 2004; Paasinen-Sohns et al. 2011). The aggressive Amdc-s and Amdc-as cells thus offer a good model to study the molecules relevant for the invasion and dissemination of fibrosarcomas.

Melanoma is a malignant tumor originating from the melanocytes, the cells responsible for the production of the dark pigment melanin (reviewed in Villanueva and Herlyn 2009). Melanoma is one of the less common types of skin cancer, but due to its high tendency to invade and metastasize, it causes the great majority of the skin cancer-related deaths (Kumar et al. 2007). Further, melanoma affects relatively young people. As the worldwide incidence of melanoma is rising and no effective treatment currently exists for the treatment of disseminated melanoma (Villanueva and Herlyn 2009; Garbe et al. 2011), it is extremely important to solve the mechanisms of melanoma cell invasion and dissemination.

REVIEW OF THE LITERATURE

1. CANCER DEVELOPMENT

Malignant tumors arise from normal cells, any cells in the body, which escape the control of normal growth. According to the common view, six hallmark capabilities are required for malignant growth: self-sufficiency for external growth signals, insensitivity to negative growth signals, resistance to programmed cell death (apoptosis), limitless replicative potential, sustained formation of new blood vessels (angiogenesis), and acquisition of tissue invasiveness (Hanahan and Weinberg 2000). However, clinically the most important of these capabilities is tissue invasiveness, which, in contrast to the other capabilities, is only found in malignant tumors and not in benign ones (Lazebnik 2010). Cancer development thus is a multistep process that usually takes a time span of decades. Normal tissues may first develop hyperplasia, just by uncontrolled proliferation, and form then benign tumors, which may gradually develop dysplasia, showing also cytological abnormality and architectural distortion. Finally these benign tumors may develop into malignant tumors able to invade the adjacent tissues (for a review see Weinberg 2007). The initiating events in the pre-malignant cells are mainly gene mutations and deletions leading to activation of oncogenes and inactivation of tumor suppressor genes, but at the later stages of cancer development, the changes are mostly seen in gene expression (Sager 1997). In this thesis I will concentrate on these gene expression changes in two highly aggressive malignancies.

Fibrosarcoma, the first target disease, is a non-epithelial cancer type arising from fibroblasts, a cell type originating from the mesoderm of the embryo. As compared to many more common fibroblastic proliferations, fibrosarcomas are highly malignant neoplasms often harboring the ability to recur locally or to invade and spread to distant organs (metastasize) (Kumar et al. 2007). Malignant melanoma, the second target, is a non-epithelial cancer type as well, evolving from melanocytes, the pigmented cells of the skin and the retina of the eye. Melanocytes, in turn, arise from an ectodermal embryonic structure termed the neural crest and are distributed as single cells within the basal layer of epidermis (Satyamoorthy and Herlyn 2002). The role of melanocytes in the skin is to produce melanin-pigment upon UV- light induction, which protects the skin from UV-light-induced DNA damage. Melanoma is less common skin cancer than basal and squamous cell carcinomas, nevertheless it is far more

deadly than those two main skin cancer types (Kumar et al. 2007). Indeed, melanomas often have a high and unpredictable tendency to metastasize. The phases of melanoma development are relatively well known (nicely reviewed in Villanueva and Herlyn 2009) and are depicted in Figure 1. The way towards melanoma starts with initial proliferation of melanocytes, leading to benign nevi. Some of the nevus cells may then continue anomalous growth and become dysplastic, but still stay in the epidermis. Then, again some of the nevus cells may acquire the ability to invade the surrounding epidermis and are said to be at the radial growth phase (RGP). Later in melanoma progression, some of the tumor cells further acquire the ability to invade vertically deep into the dermis (vertical growth phase, VGP), and, finally, to metastasize into the regional lymph nodes and distal organs (metastatic phase), and kill the patient. Recently, cancer stem-like cells, already defined in various human and mouse tumors, have also been identified in melanoma tumors (Zabierowski and Herlyn 2008). Thus, in addition to the classical view of melanoma development, melanoma tumors may also arise from these interesting cells, possessing self-renewal capacity, differentiation potential, and, most importantly, high tumorigenicity.

Figure 1. A schematic representation of the classical stages of melanoma development and progression. Environmental factors such as UV-light and the tumor microenvironment alterations are influencing these processes. RGP and VGP stand for radial and vertical growth phases, respectively. Modified from Villanueva and Herlyn 2009.

A locally growing tumor is usually not a life-threatening condition. The process of metastasis makes it fatal. Indeed, metastasis accounts for 90% of human cancer deaths (Gupta and Massagué 2006). The process of metastasis has been viewed as a series of events including dissociation of cancer cells from the primary tumor, local invasion, intravasation into blood or lymphatic vessels, survival in these channels, extravasation from the vessels, and colonization at a distant site (for a review see Gupta and Massagué 2006; Weinberg 2007; Nguyen et al. 2009). To achieve these successes, cancer cells have to become able to move and to degrade the surrounding tissue matrix (unless exploiting amoeboid migration). To obtain these abilities, cancer cells reactivate gene expression programs related to embryonic development and tissue remodeling like wound healing. Indeed, tumors have been depicted as “wounds that do not heal” (Dvorak 1986). In this thesis, I will concentrate on the process of local invasion, a prerequisite for metastasis.

2. TUMOR MICROENVIRONMENT

Malignant tumor cells do not exist alone, but are surrounded by multiple types of stromal cells, as well as various proteins of the extracellular matrix (ECM) (see Fig. 2). Growing number of studies emphasize the importance of the tumor microenvironment in cancer progression. Under normal conditions, the cellular microenvironment is able to suppress malignant cell growth, but the tumor-modified microenvironment, developing as a product of continuing crosstalk between tumor cells and the surrounding stromal cells (paracrine signaling), has been shown to permit or even promote tumor progression. The modifications taking place in the tumor microenvironment include for instance modification of ECM composition (degradation of the existing molecules by proteases, and synthesis and deposition of new ECM components), activation of stromal fibroblasts, and the recruitment of immune and inflammatory cells into the area. Instead of cancer cells alone, cancer research has, therefore, started to increasingly concentrate on the tumor tissue as a whole.

2.1 Stromal cells and surrounding epithelia

The tumor may contain, in addition to the tumor cells, various types of stromal cells, including fibroblasts, endothelial cells, pericytes, smooth muscle cells, and various

leukocytes (Satyamoorthy and Herlyn 2002; Coussens and Werb 2002; Lorusso and Rüegg 2008). In addition, surrounding normal tissue components such as epidermal keratinocytes in melanoma interact with tumor cells. In many tumors the number of these ancillary cells greatly exceeds that of the neoplastic cells.

Figure 2. The tumor microenvironment contains multiple cell types, which crosstalk by soluble mediators, growth factors, cytokines, and chemokines, embedded in the extracellular matrix.

2.1.1 Keratinocytes

Epidermal keratinocytes are important for the homeostasis of normal skin (Satyamoorthy and Herlyn 2002). Keratinocytes keep the proliferation of melanocytes under control via homotypic cell-adhesion mediated by E-cadherin (see the section 3.1.1) expressed at the surface of both the cell types. Thus, in normal skin, melanocytes do not divide. Malignant melanoma cells, in turn, often show down-regulated expression of E-cadherin (Hoek 2007) and hence escape the control of keratinocytes. In addition to E-cadherin-mediated interactions, epidermal keratinocytes regulate the phenotype of neighboring melanocytes via paracrine signaling utilizing growths factors and cytokines such as interleukin-1β (IL-1β), tumor necrosis factor α (TNF-α), and epidermal growth factor (EGF) (Fukunaga-Kalabis et al. 2008).

2.1.2 Fibroblasts

Cancer-associated fibroblasts (CAFs) (reviewed in Kalluri and Zeisberg 2006; Räsänen and Vaheri 2010) appear to be highly important for tumor growth and progression. Fibroblasts, the most abundant cell type in connective tissues, are in normal conditions in an inactive, quiescent state, but become activated to myofibroblasts (which start to express α-smooth muscle actin, hence the name) in wound healing and fibrosis, and start to synthesize ECM components such as collagens. Similarly, in the stroma of tumors, “wounds that do not heal”, some of the fibroblasts become “activated CAFs” that start to secrete ECM components and ECM modifying proteases, including matrix metalloproteinases (MMPs) and urokinase-type plasminogen activator (uPA). In contrast to wound myofibroblasts, which are removed by apoptosis after the wound is healed, activated CAFs remain in the tumor-modified microenvironment (Räsänen and Vaheri 2010). In addition to ECM remodeling, the CAFs crosstalk with tumor cells by producing various growth factor molecules such as hepatocyte growth factor (HGF) and transforming growth factor β (TGF-β), and regulate the inflammatory response by secreting cytokines and chemokines. Further, CAFs stimulate angiogenesis by secreting factors like vascular endothelial growth factor (VEGF) that activate endothelial cells and pericytes. Importantly, CAFs may also stimulate the process of epithelial to mesenchymal transition (EMT; see 3.1.1), an important determinant of epithelial cancer progression (Kalluri and Zeisberg 2006).

As an example of the complex crosstalk between tumor cells and fibroblasts, melanoma cells are known to stimulate the proliferation of stromal fibroblasts by producing basic fibroblast growth factor (bFGF), which is also an important autocrine growth factor for melanoma cells themselves, and platelet derived growth factor (PDGF) targeting only stromal cells (Satyamoorthy and Herlyn 2001). PDGF is known to induce the proliferation of fibroblasts, but not their activation (Kalluri and Zeisberg 2006). Further, melanoma cells produce TGF- β1, which is able to activate stromal fibroblasts to synthesize ECM components, such as collagens, fibronectin, and tenascins, leading to increased melanoma cell survival and metastasis formation (Berking et al. 2001). Notably, TGF-β is also known to mediate fibroblast activation during wound healing and fibrosis (Kalluri and Zeisberg 2006). Activated fibroblasts, in turn, reciprocally stimulate melanoma cells by producing e.g. insulin-like growth factor-1, HGF, bFGF, and TGF-β (Satyamoorthy and Herlyn 2001; Lee and Herlyn 2007; Melnikova and Bar-Eli 2009). However, the involvement of fibroblasts in

melanoma progression seems to be stage-dependent, as co-cultured fibroblasts have been shown to enhance the growth of metastatic melanoma cells, but to repress that of RGP melanoma (Lee and Herlyn 2007).

2.1.3 Endothelial cells

To be able to grow beyond 1 mm in diameter, the developing tumor needs a blood supply to deliver oxygen and nutrients. In order to stimulate the process of angiogenesis, normally occurring during embryonic development, wound healing, and tissue regeneration, tumor cells and surrounding stromal cells (like CAFs and inflammatory cells, see below) produce soluble pro-inflammatory and pro-angiogenic cytokines, including VEGF, TNF-α, and IL-8, which stimulate the endothelial cells of a nearby blood vessel to proliferate and migrate towards the tumor (Brooks et al. 2010). These migrated endothelial cells then form a vascular network, which is less organized and worse functioning than normal vasculature, but still nourishes the growing tumor and provides a route for the hematogeneous spreading of the tumor. In addition, tumor cells use similar process of lymphangiogenesis (exploiting VEGF- C) to attract endothelial cells of the lymphatic vessels (Alitalo et al. 2005). Besides stimulation of endothelial cell proliferation and migration, tumor and stromal cell-secreted VEGF induces separation of endothelial cells in venules, thereby causing normal blood vessels to leak plasma proteins, like fibrinogen and plasminogen (Dvorak 1986). Some evidence also exists for reciprocal signaling, where lymphatic endothelial cells actively attract tumor cells by secreting chemokines like CCL21 (Alitalo et al. 2005).

2.1.4 Inflammatory and immune cells

Tumor tissues are often infiltrated by inflammatory and immune cells. Traditionally these cells have been related to host defense against the tumor, but nowadays the tumor cells and the infiltrating inflammatory and immune cells are also known to have reciprocal interactions between each other, promoting tumor progression (reviewed in Coussens and Werb 2002; Lorusso and Rüegg 2008; Qian and Pollard 2010). Initially, tumor cells and the surrounding stromal cells release growth factors and cytokines that are chemoattractive for leukocytes, including colony stimulating factor-1 (CSF-1) and monocyte chemotactic protein, thereby attracting leukocytes into the area and stimulating them to differentiate. In addition, for

instance melanoma cells use these chemokines for autocrine stimulation of their own proliferation (Coussens and Werb 2002). The recruited leukocytes include e.g. neutrophils, eosinophils, basophils, dendritic cells, monocytes/macrophages, natural killer cells, and lymphocytes (Coussens and Werb 2002), but macrophages are thought to be the most important in respect to tumor progression (Qian and Pollard 2010). Indeed, over 80% of studies report a correlation between macrophage density and poor patient prognosis (Qian and Pollard 2010).

As the signaling is bidirectional, the activated leukocytes, in turn, secrete cytokines (like TGF-β and TNF-α) that influence tumor cells and the surrounding stromal cells such as fibroblasts, and recruit additional inflammatory cells (Coussens and Werb 2002). Importantly, TNF-α and IL-1α secreted by the activated macrophages may stimulate melanoma cells to produce angiogenic IL-8 and VEGF (Melnikova and Bar-Eli 2009). Further, macrophages stimulated by the tumor-secreted CSF-1 in turn produce EGF, which reciprocally activates migration of the tumor cells (Qian and Pollard 2010). The tumor- associated macrophages produce many proteases, including MMPs and serine proteases such as uPA, facilitating tumor invasion. These proteases make angiogenic growth factors (like VEGF) bioavailable by releasing ECM-bound molecules, but more importantly, macrophages themselves produce VEGF. Macrophages thus play an important role in cancer progression and, consistently, macrophage-deficient mice susceptible to mammary cancer have shown delayed tumor progression and metastasis (Lin et al. 2001).

2.2 Extracellular matrix molecules

The extracellular matrix (ECM) is a dynamic structure found around cells in animal tissues, in the intercellular space, and is composed of specialized proteins, proteoglycans, and glycosaminoglycans secreted by the surrounding cells, mainly fibroblasts. To mention some, the specialized proteins include collagens, elastin, laminins, fibronectins, and tenascins, the proteoglycans include versican, brevican, perlecan, decorin, and biglycan, and the glycosaminoglycans include heparin and hyaluronan. The exact composition of the ECM depends on the tissue type and it is subject to constant remodeling. In addition to the interstitial matrix between various cells, a specialized form of the ECM, the basement membrane (BM, also known as basal lamina), is found as thin sheets underlying epithelial

and endothelial cells, and surrounding nerve, muscle, and fat cells, thereby separating them from stromal components. The BM also prevents the migration of normal epidermal melanocytes into the dermis (Satyamoorthy and Herlyn 2002). The main components of the BM are collagen-IV (COL-IV), entactin, laminins (LNs), and heparan sulfate proteoglycans like perlecan (Finger and Giaccia 2010).

The most obvious function of the ECM is to provide support and anchorage for the resident cells (mainly through the integrin adhesion receptors, see the section 3.1.2), but the interactions of ECM molecules are also important in regulation of many cellular processes, including cell survival, proliferation, differentiation, cell migration, and cancer cell invasion. As the ECM proteins are often large, multi-domain molecules, they are able to interact with several other molecules through different domains (nicely reviewed in Hynes 2009). Indeed, several families of ECM proteins have domains in common, favoring their mutual interactions. In addition, ECM sequesters water and growth factors/angiogenic factors that are important for tumor growth. Many of these growth factors, like FGFs and VEGFs, are known to bind to the glycosaminoglycan parts of ECM proteoglycans, but increasing evidence shows that growth factors can also directly bind to ECM proteins, such as HGF binding to fibronectin (FN) and vitronectin (VN), and VEGF binding to FN and tenascin-C (TN-C) (Hynes 2009). Further, ECM does not only function as a reservoir for soluble growth factors (released by degradation of ECM), but it plays an important role in the presentation of growth factor signals. One of the best examples of this role is the regulation of TGF-β signaling (reviewed in Hynes 2009). First, the TGF-β precursor is cleaved to the mature TGF- β molecule (still inactive) and its propeptide (latency-associated peptide, LAP), which still remain non-covalently associated and then further bind to latent TGF-β-binding proteins (LTBPs) to form large latent complexes. These complexes are bound to the ECM via interactions of LTBPs and ECM proteins (including fibrillins and FNs). This LTBP-mediated binding of latent TGF-β molecules to the ECM is necessary for their subsequent effective activation, which can be triggered by degradation of the ECM proteins, LTBPs, or LAPs, or directly by conformational change in LAP, leading to exposed TGF-β able to bind to and activate its receptors. Thus, the activated TGF-β growth factor triggers its downstream signaling while still remaining bound to the ECM proteins.

2.2.1 Collagens

The main characteristic of the ECM is a collagen (COL) scaffold into which the cells are anchored, indirectly via adhesive glycoproteins and proteoglycans or directly via their surface collagen receptors (see below). COL is best known as the tensile element of tissues such as tendon, cartilage, bone, and skin (elongated COL fibrils, the most abundant structural elements in adult tissues) and the main component of BM (COL-IV meshwork) (Kadler et al. 2007). Collagens (reviewed in Kadler et al. 2007; Heino 2007; Gordon and Hahn 2010), 28 named to date, represent a large family of triple helical proteins composed of three α chains held together by interchain hydrogen bonds. Collagen α chains, in turn, typically contain triple helical domains (COL domains), composed of repeated glycine-X-Y motifs (where X and Y mean any amino acids, often, however, occupied by proline and 4-hydroxyproline), flanked by non-collagenous domains (NC domains). Fibril-forming collagens (the most important being COL-I and -III) are synthesized as procollagen molecules containing N- and C-terminal propeptides, which are then cleaved by proteinases before fibrillogenesis (Kadler et al. 2007). Collagen fibrils are then further stabilized by non-reducible covalent crosslinks. Collagens are mostly synthesized by fibroblasts, myofibroblasts, osteoblasts, and chondrocytes (Bosman and Stamenkovic 2003).

In addition to the collagen fibrils, collagens form various other supramolecular aggregates, including chicken wire-like networks formed by the BM collagen COL-IV, hexagonal networks formed by collagens like COL-X, beaded filaments formed by COL-VI, and anchoring fibrils formed by COL-VII (Heino 2007; Gordon and Hahn 2010). Further, fibril- associated collagens with interrupted triple helices (FACITs, COL-IX as an example) can be found at the surfaces of collagen fibrils mediating their interactions with other matrix proteins. Transmembrane collagens (COL-XIII, -XVII, -XXIII, and -XXV), in turn, are inserted into the plasma membrane and multiplexins (COL-XV and -XVIII) participate in the stabilization of BM structures (Heino 2007; Gordon and Hahn 2010). The most abundant collagens, COL-I and -II are able to spontaneously form fibrils in vitro, but in vivo they need non-collagenous molecules like FN and FN- and COL-binding integrin receptors as organizers (COL-I), and COL-V (COL-I) or COL-XI (COL-II) to initiate the process (Kadler et al. 2008). In addition to these interaction partners, numerous other molecules have been shown to influence collagen fibrillogenesis, including tenascin-X (Kadler et al. 2008). Indeed, COL-I is known to have nearly 50 binding partners (Di Lullo et al. 2002).

The main cellular collagen receptors are integrins (integrins α1β1, α2β1, α10β1, and α11β1 being the most important), but also receptor tyrosine kinases (discoidin domain receptors 1 and 2) and immunoglobulin-like receptor family members (glycoprotein VI and leukocyte- associated immunoglobulin-like receptor-1) are known to bind collagens, as well as proteoglycans (CD44, syndecans, and glypicans) (Heino 2007). Collagen fibrils play a central role in various physiological processes like embryogenesis, tissue repair, fibrosis, and tumor invasion (Kadler et al. 2008).

2.2.2 Elastin

Another key structural component of the ECM is elastin (reviewed in Mithieux and Weiss 2005; Muiznieks et al. 2010). Elastin is a self-assembling protein providing elasticity to vertebrate tissues dependent on repeated distension and relaxation, such as large arteries, lung, skin, bladder, ligaments, tendons, and elastic cartilage. Elastin self-assembly involves self-association of the soluble tropoelastin precursor molecules through hydrophobic domains (a process known as coacervation) followed by massive cross-linking accomplished by lysyl oxidases. This cross-linking is indeed essential for the function of elastin. Tropoelastin monomers are typically secreted by fibroblasts, vascular smooth muscle cells, endothelial cells, and chondrocytes (Mithieux and Weiss 2005; Rodgers and Weiss 2005). In the ECM, elastin is found as the main component (90%) of elastic fibers, forming the internal core of the fibers. This core is interspersed with and surrounded by unbranched microfibrils (the main components being fibrillin-1 and -2, and microfibril-associated glycoprotein-1), which, in fact, precede the self-assembly of tropoelastin molecules and act as a scaffold for their deposition (Mithieux and Weiss 2005; Muiznieks et al. 2010). Other proteins present in the elastic fibers include VN, LTBPs (see above), microfibrillar associated proteins, and proteoglycans such as versican, biglycan, and decorin (Mithieux and Weiss 2005). The elastin receptors studied include elastin binding protein, a G protein-coupled receptor, and integrin αVβ3 (Rodgers and Weiss 2005). Interestingly, elastin has a half-life of about 70 years, and thus very little turnover is seen during the adult life (Mithieux and Weiss 2005). In addition to barely providing elasticity to the tissues, elastin, tropoelastin, and elastin degradation products are nowadays known to influence cellular responses, including chemotaxis, proliferation, and cell adhesion (Rodgers and Weiss 2005).

2.2.3 Laminins

LNs (reviewed in Miner 2008; Durbeej 2010), the major constituents of BM, are cross- , Y-, or rod-shaped heterotrimeric proteins forming polymeric networks able to bind other matrix proteins (also interacting with each other), thus organizing the formation of a highly cross- linked structure of BM. LNs, synthesized by numerous cell types, are composed of one α-, one β-, and one γ-chain, all sharing a common domain structure consisting of a number of globular and rod-like domains. In vertebrates, there exist five α, three β, and three γ-laminin chains, and their heterotrimers show tissue-specific distribution. The LN heterotrimers are assembled inside the cell, but further proteolytic processing occurs outside the cell to produce the mature protein and, finally, the LNs form the BM networks. In addition to other ECM molecules, LNs interact with cell surface receptors, linking them to intracellular signaling pathways. The major LN receptors are integrins (including α6β1, α6β4, and αVβ3) and non- integrin receptors (including dystroglycan, syndecans, and Lutheran blood group glycoprotein). LNs are essential for early embryonic development and organogenesis (Durbeej 2010), but they seem to play a role in malignancy as well (Patarroyo et al. 2002). Notably, LN (LN-111) is also one of the main components of Matrigel, a basement membrane-like matrix extracted from murine sarcoma tumor, which has been extensively utilized in cancer research (Benton et al. 2011).

2.2.4 Fibronectin

FN molecules (reviewed in Pankov and Yamada 2002; Kaspar et al. 2006) are ubiquitous high-molecular-weight ECM glycoproteins forming covalently-linked dimers, which assemble into a fibrillar matrix (Singh et al. 2010). This matrix is relevant for several cellular processes, including cell adhesion, migration, growth, and differentiation, and, consistently, FN is essential for early embryonic development. Further, FN plays an important role in wound healing, where FN is included in the initial clot, together with fibrin and platelets, and further in the FN-rich granulation tissue, the young connective tissue later displaced by the collagenous matrix of the scar (Dvorak 1986). FN is also a component of tumor stroma (Dvorak 1986).

FN monomers consist of three types of repeating units, FN repeats type I, II, and III, which account for approximately 90% of the FN sequence. These units fold into a series of protein

domains, often displaying functionally distinctive properties (Kaspar et al. 2006). Indeed, FN molecules are known to possess a collagen-binding domain and heparin-binding domains, as well as fibrin- and integrin-binding sites (including the RGD sequence), through which they interact with other ECM macromolecules and cell surface integrins (the most important being integrins α5β1 and αVβ3). Indeed, even the FN polymerization, the FN fibrillogenesis, requires direct interactions with the integrin receptor α5β1 (Kadler et al. 2008; Singh et al. 2010). Further, FN molecules bind to each other, which is essential for fibril formation (Singh et al. 2010). As mentioned above, FN also functions as an organizer in collagen fibrillogenesis, and the assemblies of FN and COL fibrils seem to be dependent on each other (Kadler et al. 2008; Singh et al. 2010). Further, FN matrix is required for the assembly of microfibrils, which, in turn, have an important role in elastic fiber assembly, and for the incorporation of growing number of other ECM proteins (including LTBPs and TN-C) into the matrix (Singh et al. 2010). In addition to the insoluble ECM FN, liver hepatocytes secrete soluble FN found in plasma and other body fluids (Pankov and Yamada 2002). The single FN gene gives rise to multiple variant molecules by alternative splicing and these variants have different cell-adhesive, ligand-binding, and solubility properties (Pankov and Yamada 2002). As this spicing is cell type-specific, cells can utilize FN variants in a developmental and tissue-specific manner.

2.2.5 Tenascins

Tenascin (TN) family (reviewed in Jones and Jones 2000; Bosman and Stamenkovic 2003; Midwood and Orend 2009) contains the high-molecular-weight glycoproteins tenascin-C, -R, -W, -X, and -Y. Of these, TN-C, which was discovered first and is the most extensively studied, forms hexamers (hence also called hexabrachion) and functions to modulate cell migration, proliferation and cellular signaling. The other members of TN family, in turn, most often form trimers. In addition to the oligomerization domain, TNs have varying numbers of EGF-like repeats, various FN type III domains, and a fibrinogen-like globular domain, making them able to interact with numerous ECM proteins and proteoglycans, including FN (Chiquet-Ehrismann et al. 1988; Chung et al. 1995) and COL-I (Di Lullo et al. 2002), and multiple cell surface receptors, such as integrins α2β1, α7β1, α9β1, and αvβ3, and EGF receptor (Midwood and Orend 2009). In contrast to FN, being an adhesive ECM protein, TN-C is mainly anti-adhesive and can interfere with cell attachment to FN and LN

(Chiquet-Ehrismann et al. 1988; Herlyn et al. 1991). Like FN, also TNs have several alternatively spliced isoforms, having different adhesive properties and showing tissue and cell type-specificity. TN-C is highly expressed during embryogenesis, but it is largely absent in normal adult tissues. Its expression is, however, specifically induced at sites of inflammation, wound healing, and neovascularization, and persistently induced in fibrotic diseases and in solid tumors (Midwood and Orend 2009). TN-C is known to be produced by stromal fibroblasts, both in healing wounds and in tumors, and by osteoblasts, perichondrial cells, and muscle cells (Chiquet-Ehrismann and Tucker 2004), but some tumor cells also produce TN-C, including melanoma cells (Herlyn et al. 1991). Interestingly, TN-C has been found to stimulate MMP expression, including MMP-1 (Fukunaga-Kalabis et al. 2008).

2.2.6 Transforming growth factor beta-induced

Of the minor ECM proteins, transforming growth factor beta-induced (TGFBI, originally known as βig-h3, also keratoepithelin) is an interesting protein expressed ubiquitously in various tissues, except the brain (Skonier et al. 1994; Ivanov et al. 2008). It is a TGF-β- inducible, 68 kDa secreted protein containing an N-terminal secretory signal peptide, a cysteine-rich domain (EMI), four consecutive fasciclin 1 (FAS) repeats, and a C-terminal Arg-Gly-Asp (RGD) motif (Skonier et al. 1992). TGFBI has been found to interact with several other ECM components, including COL-I, -II, -IV and -VI, FN, LN, periostin, and proteoglycans like biglycan and decorin (Hashimoto et al. 1997; Billings et al. 2002; Kim et al. 2002; Reinboth et al. 2006; Kim et al. 2009). Further, TGFBI has been shown to directly (independent of other macromolecules) bind to collagen fibers of the triple helical conformation (not binding to denatured COL-I, gelatin), whereas FN binds both collagens and gelatin (Hashimoto et al. 1997). In addition, TGFBI has been shown to interact with various cell surface integrins (Thapa et al. 2007), and because of this feature and the RGD motif, has been thought to function as a linker between ECM and the cell surface integrins. TGFBI function is related to cell adhesion, but opposite results have been achieved in adhesion experiments: TGFBI has been shown to promote the adhesion of some cell types, including human dermal, lung, and bladder fibroblasts (LeBaron et al. 1995; Ohno et al. 1999; Billings et al. 2002), corneal epithelial cells (Kim et al. 2000), keratinocytes (Bae et al. 2002), and endothelial cells (Nam et al. 2003), but to inhibit the adhesion of human lung adenocarcinoma cells, HeLa cells, lung fibroblasts (Skonier et al. 1994), and scleral

fibroblasts (Shelton et al. 2008; Shelton and Summers-Rada 2009). Thus, TGFBI’s role in cell adhesion may depend on the cell type and the molecules of the microenvironment. In line with its multiple interaction partners, TGFBI has been shown to participate in multiple physiological as well as pathological processes, including morphogenesis, adhesion/migration, corneal dystrophy, wound healing, inflammation, and tumorigenesis (reviewed in Thapa et al. 2007).

3. LOCAL INVASION

Disaggregation of tumor cells and their movement into the surrounding tissues are the first steps in metastatic spread. For this, modification of cell-cell and cell-ECM interactions is essential, as well as increased proteolysis to remove physical obstruction created by the ECM.

3.1 Cell adhesion and adhesion receptors

To maintain normal tissue architecture the cells need to maintain proper contacts with their neighboring cells and the ECM. Invading cancer cells, in turn, need to break the existing contacts and to make new contacts that are beneficial to them. In addition to cancer cell invasion, a similar ability to change adhesion partners is needed during normal embryonic development and the healing of tissue injuries.

3.1.1 Cell-cell interactions

The family of classical cadherins, of which the type I cadherins E-cadherin (epithelial), N- cadherin (neural), and P-cadherin (first identified in placenta) are the best characterized, consists of transmembrane adhesion molecules mediating calcium-dependent cell-cell interactions (reviewed in Behrens 1999; Angst et al. 2001; Rickelt et al. 2008). With their cytoplasmic regions, classical cadherins interact with the actin cytoskeleton via the cytoplasmic proteins α-, β-, and γ-catenin, and this interaction, directly connecting the cytoskeletons of the adjacent cells, is crucial for efficient cell-cell adhesion. During

melanoma development, melanoma cells switch their cadherin expression profile, usually during the transition from RGP to VGP (reviewed in Satyamoorthy and Herlyn 2002; Rickelt et al. 2008; Villanueva and Herlyn 2009). In the normal skin environment, epidermal keratinocytes and melanocytes both express E-cadherin (and to lesser extent also P-cadherin), that mediates their homophilic interaction controlling melanocyte proliferation. The dermal fibroblasts and vascular endothelial cells, in turn, express N-cadherin at their surfaces. During melanoma development, melanoma cells down-regulate E-cadherin and P-cadherin expression, and up-regulate N-cadherin expression, thereby reducing their interaction with the division-controlling keratinocytes, and promoting interactions with fibroblasts and endothelial cells. This may facilitate the migration/invasion of melanoma cells into the dermis and intravasation into the vasculature. A similar cadherin profile switch occurs in the progression of epithelial cancers, in EMT. In this process, the non-motile, polarized epithelial cells are transformed into motile, non-polarized mesenchymal cells able to disseminate (Finger and Giaccia 2010).

3.1.2 Cell-matrix interactions

Integrins are the main cellular receptors used for cell-ECM adhesive interactions (reviewed in Giancotti and Ruoslahti 1999). Additionally, some leukocytic integrins are involved in cell- cell interactions. Integrins are obligate heterodimers, consisting of transmembrane α and β subunits, which are non-covalently linked to each other. In mammals, there exist 18 α and 8 β subunits, forming at least 24 integrin dimers. The subunit composition of integrin dimers confers them specificity to different ligands, like COL, FN, LN, TN, and VN. Most integrins have several ligand molecules and multiple integrins can recognize the same ligand molecule. While integrins bind their ECM ligands through their large extracellular domains, their small cytoplasmic tails are linked to the actin cytoskeleton via a variety of linker proteins, such as talin, α-actinin, vinculin, and paxillin. Following ligand binding, integrins trigger reorganization of the cytoskeleton. In addition to anchoring of the cells into the surrounding ECM, integrins mediate bidirectional signals between the cells and ECM. First, integrins can be activated by cytoplasmic molecules, like talin, leading to “inside-out” signaling. When binding to their ECM ligands, integrins undergo conformational changes and activate multiple cytoplasmic signaling pathways (“outside-in” signaling). Integrins become clustered upon ligand binding. When integrins bind their ECM ligands and associate with the

cytoskeletal and signaling complex, they form protein aggregates on both sides of the membrane. This property of integrins to serve as integrators of the ECM and cytoskeleton has given them their name. In cell cytoplasm, integrins, not having any enzymatic activity of their own, activate various protein tyrosine kinases, including focal adhesion kinase (FAK) and Src-family kinases, which, in turn, activate various signaling pathways. Further, integrins are also necessary for optimal activation of growth factor receptors. Numerous studies have reported altered integrin expression in different cancers and integrins have been shown to clearly contribute to cancer progression (reviewed in Desgrosellier and Cheresh 2010; Rathinam and Alahari 2010).

In addition to the large family of integrins, also non-integrin receptors to ECM molecules exist on the cell surface. These include for instance the FN-binding syndecans, the LN- binding 67 kDa LN receptor and α-dystroglycan, the COL-binding discoidin domain receptors, glycoprotein VI, and leukocyte-associated immunoglobulin-like receptor-1, and the hyaluronan binding hyaladherins (like RHAMM and CD44) (Heino and Käpylä 2009). Nevertheless, these non-integrin receptors often show more limited tissue distribution than the integrins and function in collaboration with the integrin receptors.

3.2 Cell migration

In addition to cancer cell invasion and metastasis, cell motility is essential for many physiological processes, including embryonic development, wound healing and inflammation. Cell migration occurs due to cycles of five independent steps (Wolf and Friedl 2006). First, membrane protrusion formation (lamellipodia, filopodia, or pseudopodia, or invadopodia in invading cancer cells in 3D-matrix) and polarization are initiated by actin polymerization (in an adhesion-independent manner), and after that, the cell protrusions adhere to the ECM molecules (integrins being also the major migration-mediating receptors) forming focal adhesion sites. At this stage, also the cell surface proteases may become recruited to the attachment sites, to cleave the matrix barriers. Next, actomyosin contraction of the cell body creates bipolar tension in the cell body, and, finally, this tension leads to the cell body translocation and tail retraction.

3.2.1 Actin polymerization

The driving force in eukaryotic cell motility is polymerization of actin monomers (G-actin) into polarized filaments (F-actin) (Pollard and Borisy 2003). New monomers are constantly added at the “barbed” or “plus” ends of the filaments and, simultaneously, the “pointed” or “minus” ends are depolymerized. The predominant site of actin polymerization in motile cells is proximal to the plasma membrane, and the actin filaments are oriented in the direction of the movement, barbed ends directed towards the plasma membrane (Olson and Sahai 2009). Actin polymerization in motile cells can be enhanced in many ways: formin homology (FH) proteins increase the rate of polymerization at the barbed ends (generating linear F-actin arrays found in filopodia), the Arp2/3 complex binds to the actin filament and nucleates a branching point, thereby increasing the number of barbed ends (generating sheets of F-actin found in lamellipodia), and cofilin creates new barbed ends by cutting existing actin filaments (reviewed in Olson and Sahai 2009). In addition to polymerization enhancing molecules, capping proteins are used to sterically hinder monomer addition. Further, the availability of actin monomers is regulated by maintaining them complexed with profilin or thymosin. Of these actin sequestering molecules, thymosins β4 and β10 (Tβ4 and Tβ10) play a major role (reviewed in Mannherz and Hannappel 2009). Tβ4 and Tβ10 bind G-actin at 1:1 stoichiometric ratio and maintain a dynamic equilibrium between G- and F-actin. From this pool of G-actin, the actin monomers can easily be used for actin polymerization when needed, but inappropriate polymerization is prevented.

As actin cytoskeleton and actin polymerization are enormously important for the cell, the activity of the complex polymerization machinery must be tightly regulated (reviewed in Olson and Sahai 2009). To mention some regulatory molecules, the Arp2/3 complex is regulated by the WAVE and WASP family proteins, binding both the Arp2/3 complex and actin monomers, and thereby increasing the rate of actin polymerization. The WAVE and WASP family proteins, as well as FH proteins, are, in turn, also tightly regulated themselves through a conformational switch. Their inactive “closed” conformation can only become active “open” conformation after interactions with GTP-binding proteins like Cdc42 (WASP and FH), RhoA, RhoB, and Rif (FH), and various adaptor molecules. In addition, many regulators of actin polymerization are themselves regulated by phosphorylation. The goal of these complex regulatory mechanisms is to ensure that inappropriate actin polymerization does not occur, and, on the other hand, to enable increased actin polymerization in response

to appropriate stimuli. Accordingly, many of these multiple regulatory pathways have become deregulated in invasive cancer cells (coordinated up-regulation) (Olson and Sahai 2009).

3.2.2 Acto-myosin contraction

Cell movement needs, in addition to actin polymerization, force generation accomplished by the myosin family of motor proteins (reviewed in Vale and Milligan 2000; Olson and Sahai 2009). For that, the polymerized actin filaments assemble with myosin II filaments, composed of heavy and regulatory light chains, and the associated myosin molecules then use energy from ATP hydrolysis to move towards the barbed end of the actin filament. When multimeric myosin is associated with multiple actin filaments, this movement causes the actin filaments to move relative each other, generating the contractile force driving morphological reorganization. Not surprisingly, this step of cell movement is tightly regulated as well. The key mechanism for regulation of actin-myosin contractibility is phosphorylation of the myosin II light chain, accomplished by numerous kinases, promoting the formation of myosin filaments and facilitating the association with F-actin.

The processes of actin polymerization and contraction are both essential steps of cell motility and for effective movement they must be coordinated. Polymerization needs to occur prior to contraction and more proximal to the plasma membrane (Olson and Sahai 2009). As actin polymers are oriented barbed ends directed towards the plasma membrane, the more recently polymerized actin exists adjacent to plasma membrane and the older F-actin is further away. Indeed, newly polymerized actin is not associated with the contractile machinery. Anyhow, it is not clear yet how this spatial and temporal separation is achieved.

3.3 Proteolysis of extracellular matrix

Invading cancer cells use proteolytic enzymes to break down the surrounding ECM, thereby clearing a path for the migrating tumor cells and liberating/activating beneficial growth factors and angiogenic factors anchored in the ECM. As mentioned above, also activated fibroblasts and macrophages residing in the tumor microenvironment play a significant role in ECM degradation by producing ECM modifying proteases. In addition to being the major

migration-mediating receptors (see above), the integrins also assist in targeted ECM remodeling by influencing protease expression and localizing them near their substrates (Ivaska and Heino 2000).

3.3.1 Matrix metalloproteinases

Matrix metalloproteinases (MMPs) are a family of zinc-dependent endopeptidases involved in various physiological processes including tissue remodeling and organ development, as well as regulation of inflammatory processes. Most importantly, MMPs have been thought to be the main proteases behind cancer cell invasiveness (reviewed in Ala-aho and Kähäri 2005; Hofmann et al. 2005; Kessenbrock et al. 2010; Bourboulia and Stetler-Stevenson 2010). The family of human MMPs, comprising 23 members (both secreted and membrane bound), is divided into five main groups according to their structure and substrate specificity: collagenases, gelatinases, membrane type MMPs, stromelysins, and matrilysins (Bourboulia and Stetler-Stevenson 2010). To mention some, MMP-2 and MMP-9 degrade COL-IV, the major component of BM, and MMP-1 degrades type I, II, and III collagens (Finger and Giaccia 2010). Collectively MMPs are able to degrade all the components of ECM. In normal intact tissues, MMPs are expressed at low levels, but when tissue remodeling is needed, their expression is induced by signals such as growth factors, cytokines, and physical stress (Ala- aho and Kähäri 2005). Most MMPs are secreted as inactive proproteins, which are active only after proteolytic cleavage accomplished by other MMPs or serine proteases like plasmin (see below), or intracellularly by furin (Kessenbrock et al. 2010). Further, MMPs are additionally kept in check by tissue inhibitors of metalloproteinases (TIMPs) and inhibitors secreted by the liver (such as α2-macroglobulin). In invasive tumors, the levels of active MMPs (as well as TIMPs) are usually increased, originating from secretion by both tumor cells and stromal cells (the major source) such as tumor-associated inflammatory cells (especially macrophages), activated fibroblasts, and endothelial cells (Kessenbrock et al. 2010). Melanoma cells have been reported to express various MMPs, including MMP-1, -2, -9, -13, and MT1-MMP (Hofmann et al. 2005 and the references therein).

In addition to making space for the invading cancer cells by ECM degradation, MMPs also promote tumor progression in multiple other ways. MMPs for instance influence the bioavailability or functionality of various important factors, such as ECM sequestered growth

factors (an important representative being TGF-β) and angiogenic/lymphangiogenic factors (like VEGF), as well as chemokines and proinflammatory cytokines recruiting inflammatory cells (reviewed in Kessenbrock et al. 2010). MMPs also degrade cell surface E-cadherin molecules, thereby disrupting cell-cell interactions, and interfere with the induction of apoptosis. Importantly, MMPs are now known to interact with cell surface receptors, such as integrins, (which also influence MMP expression), providing a way to localize them near their substrate molecules (a process called integrin-guided proteolysis; Ivaska and Heino 2000). Interestingly, MMP-3 has also been related to tumor initiation (Sternlicht et al. 1999). In contrast to their invasion-promoting effects, MMP catalyzed degradation of collagens and plasminogen has also been shown to generate anti-angiogenic fragments, and the catalytic function of some MMPs, such as MMP-8 in melanoma, have been shown to suppress the metastatic behavior (Gutiérrez-Fernández et al. 2008; Palavalli et al. 2009). Further, in addition to the traditional proteolytic-dependent roles, MMPs have more recently been reported also to have non-catalytic functions, where the hemopexin domain plays an important role. These functions include enhanced cell migration (MMP-2, -9, and -14), antimicrobial function (macrophage elastase MMP-12), and inhibition of the complement system (MMP-14) (Kessenbrock et al. 2010).

Besides MMPs, also closely related families of ADAMs (a disintegrin and metalloproteinases) and ADAMTSs (ADAMs with thrombospondin type 1 motifs) that share the metalloproteinase domain with MMPs, have recently been related to cancer development and progression (reviewed in Rocks et al. 2008; Klein and Bischoff 2011). Indeed, dysregulated expression of both types of molecules has been reported in various cancers. ADAMs (nicely reviewed in Toriseva and Kähäri 2009; Klein and Bischoff 2011) are transmembrane proteins possessing both adhesive and proteinase activities (12 of the 22 known human ADAMs display proteolytic activity; Klein and Bischoff 2011) and mainly functioning as sheddases, cleaving and releasing membrane bound proteins including cytokines and growth factors, and their receptors, as well as adhesion proteins and proteases. By liberating these biologically active proteins, ADAMs are known to regulate cell proliferation. ADAMTSs, in turn, are secreted proteases that mainly cleave ECM proteins.

The proteolytic activities of ADAMs are known to be inhibited by certain integrins, α2- macroglobulin and, to some extent, also by TIMPs, and those of the ADAMTSs by TIMPs. Similarly to some MMPs showing a tumor progression-inhibiting function (such as MMP-8),

also some ADAMs and ADAMTSs have been recently shown to possess anti-angiogenic activities (Rocks et al. 2008).

3.3.2 Serine proteases

The serine protease family contains e.g. the digestive enzymes chymotrypsin, trypsin, and elastase, as well as mast cell tryptase and chymotryptase, and neutrophil cathepsin G. It also includes the urokinase plasminogen activator system, which plays an important role in tumor cell invasion and metastasis (reviewed in Dass et al. 2008). This system in turn consists of urokinase-type plasminogen activator (uPA), tissue-type plasminogen activator (tPA), the uPA receptor (uPAR), and plasminogen activator inhibitors (PAIs). Several malignant tumors have been shown to overexpress uPA and uPAR, including breast and ovarian cancers. uPA catalyzes the cleavage of plasminogen into plasmin, a broad-spectrum serine protease directly degrading ECM components and activating MMPs. uPA itself is secreted as an inactive zymogen, which, in turn, is cleaved to its active form by plasmin (a feed-back-type activation). Before its activation uPA binds to its receptor uPAR, which is attached to the plasma membrane via a glycosyl phosphatidyl inositol anchor. tPA serves the same roles as uPA, but it is mainly involved in intravascular thrombolysis, whereas uPA is involved in pericellular proteolysis during migration, wound healing, and tissue remodeling. The plasminogen activator inhibitors, PAI-1 and PAI-2, function to keep the activities of uPA and tPA in control, but, interestingly, PAI-1 has been reported to also facilitate tumor growth and angiogenesis.

3.3.3 Cysteine proteases

Cysteine cathepsins, belonging to the papain subfamily of cysteine proteases, are lysosomal proteases related to tumor dissemination, as well as angiogenesis (reviewed in Mohamed and Sloane 2006; Gocheva and Joyce 2007; Vasiljeva and Turk 2008). The family of human cysteine cathepsins comprises 11 members, cathepsins B, C, F, H, L, K, O, S, V, W, and X/Z, of which especially cathepsins B and L often show up-regulation in human malignancies, including melanoma (Gocheva and Joyce 2007; Vasiljeva and Turk 2008). Cathepsins are synthesized as inactive precursors, zymogens, which are then activated by proteolytic removal of the N- terminal propeptide (Vasiljeva and Turk 2008). In normal cells, the main

function of cathepsins is terminal protein degradation in the lysosomes, but they also participate in MHC class II-associated antigen processing and presentation, and bone remodeling (Mohamed and Sloane 2006; Vasiljeva and Turk 2008). In addition, cysteine cathepsins participate in pathological conditions, such as rheumatoid arthritis and atherosclerosis (Vasiljeva and Turk 2008). During malignant progression, cysteine cathepsins promote tumor cell invasion through several mechanisms: by directly cleaving ECM components (including LN, FN, TN-C, and COL-IV), by activating other proteases (such as pro-MMPs and pro-uPA) leading to activation of proteolytic cascades degrading ECM molecules, and by cleaving cell surface E-cadherin molecules resulting in disrupted cell-cell interactions (Gocheva and Joyce 2007). In addition to promoting tumor cell invasion, cysteine cathepsins have, interestingly, been shown to mediate cell apoptosis (Vasiljeva and Turk 2008). A key feature separating these two opposing functions appears, however, to be differential distribution: secreted or plasma membrane-associated cathepsins participate in tumor cell invasion, whereas cytosol translocated cathepsins trigger apoptosis (Vasiljeva and Turk 2008). In addition to tumor cells themselves, also other cell types of the tumor microenvironment have been shown to express cathepsins, including tumor-associated fibroblasts and macrophages (Mohamed and Sloane 2006). Interestingly, the AdoMetDC- transformed mouse fibroblasts used in this study (Amdc-s cells) have been found to show highly increased secretion of pro-cathepsin L as compared to normal fibroblasts (Ravanko et al. 2004). Indeed, cathepsin L was found to account for the major protease activity in these highly invasive cells, and cathepsin L inhibitors were able to totally inhibit the invasion of these cells in Matrigel (Ravanko et al. 2004).

Altogether, in addition to heterotypic signaling and cell adhesion, proteolytic activity is one of the most important aspects of cancer progression. As proteases are collectively able to cleave all the ECM molecules, they make it possible for the cancer cells to invade the surrounding tissues and to intravasate or extravasate the blood and lymphatic vessels. The most important effector proteases behind these events may be MMPs, but serine and cysteine proteases also play an important role, especially in the activation of proteolytic cascades. In addition to making space for the invading cancer cells, the proteolytic networks also influence the effects of various growth factors and cytokines, thereby influencing heterotypic signaling, and disrupt cell adhesions. An opposite role, the suppression of cancer progression, has, however, been reported for some of the MMPs, and cysteine cathepsins have been found

to mediate cell apoptosis. These more recent findings make following the story of proteases in cancer very interesting.

AIMS OF THE STUDY

The aim of this study was to search for molecules that are relevant for the high invasiveness of fibrosarcoma and melanoma cells by comparing the gene expression patterns of representative malignant cells to those of corresponding normal cells. Further, an attempt was made to functionally characterize the most interesting molecules and to analyze their potential for cancer biomarkers or therapeutic targets.

MATERIALS AND METHODS

1. CELL LINES AND TISSUE SPECIMENS (I-IV)

The cell lines used in this study are detailed in Tables 1 and 2. Mouse cell lines Amdc-s, Amdc-as, and E4 are transformed fibroblasts (fibrosarcoma cells), and 4N and N1 are corresponding normal cells. Amdc-s-pLRT-Tβ4-as and Amdc-as-pLRT-Tβ4-as are transformed fibroblasts with inducible expression of thymosin β4 antisense-RNA, whereas Amdc-pLRT-TAM67 cells show inducible expression of TAM67, a dominant-negative mutant of c-Jun. HT-1080 is a human fibrosarcoma cell line originating from a clinical fibrosarcoma tumor.

The human cell lines WM793 and WM239 are invasive melanoma cells isolated from primary melanoma and melanoma metastasis, respectively, whereas 42V and Mela3 are normal melanocytes. PMEL29 cells are invasive melanoma cells isolated from a primary melanoma tumor, primary fibroblasts are isolated from normal adult skin, and TGFBI-KD cells are melanoma cells expressing TGFBI shRNA.

Table 1. Mouse cell lines. Name Parental cell line Vector Vector insert Publication 4N NIH3T3 pLTRpoly - I-II Amdc-s NIH3T3 pLTRpoly AdoMetDC sense I-II Amdc-as* NIH3T3 pLTRpoly AdoMetDC antisense I N1 NIH3T3 pSV2neo - I E4 NIH3T3 pGEJ6.6 c-Ha-rasVal12 I Amdc-s-pLRT-Tβ4-as§ Amdc-s pLRT Tβ4 antisense-RNA I Amdc-as-pLRT-Tβ4-as§ Amdc-as pLRT Tβ4 antisense-RNA I Amdc-pLRT-TAM67§ Amdc-s pLRT TAM67 I-II *To prevent the counter selection of Amdc-as cells expressing AdoMetDC antisense-mRNA at levels sufficient to block the synthesis of essential polyamines, a low concentration (1 μM) of spermidine was added to the growth medium. §For induction of Tβ4 antisense-RNA or TAM67 expression, 1 μg/ml doxycycline was added the day after plating.

Table 2. Human cells. Name Cell type Publication HT-1080 Fibrosarcoma cell line II 42V Primary melanocytes I, III-IV Mela3 Primary melanocytes IV WM793 Melanoma cell line (VGP)* I, III-IV WM239 Melanoma cell line (metastatic) I, III-IV PMEL29 Primary melanoma cells III Fibroblast Primary culture of adult skin fibroblasts IV TGFBI-KD WM239 cells expressing TGFBI shRNAs IV *VGP = vertical growth phase

Tissue specimens of benign nevi (n = 32; from healthy volunteers), primary melanomas (n = 101), clinically detected melanoma metastases (n = 19), and soft-tissue sarcomas (n = 24) were obtained from surgical excisions at the Helsinki University Central Hospital. The protocols for taking the specimens have been approved by the local ethical committee. Further, informed consent was given by the patients before operations. The primary melanomas were mostly from patients with cutaneous nodular melanomas with a thickness of ≥1 mm or Clark’s level IV-V. For RNA isolation, the tissue specimens were frozen in liquid nitrogen, and for immunohistochemistry (IHC), the paraffin-embedded tissue samples were obtained from the Department of Pathology, Helsinki University Central Hospital or other hospitals of the district. In addition to human patient specimens, paraffin-embedded tissue samples of nude mice bearing tumors induced by Amdc-s cells or TGFBI short hairpin RNA (shRNA) expressing WM239 cells (or their control, non-target shRNA expressing cells) were used.

2. ANTIBODIES (I-IV)

The antibodies used in this work are presented in Table 3. In function blocking experiments, the neutralizing antibody was preincubated for 30 min with the cells and was also included in the growth medium/matrix.

Table 3. Antibodies. Target Clone/Code Source Specificity* Publication Method** Actin JLA20 mouse br II WB c-Jun/AP-1 Ab-1 rabbit h, m, r II WB c-Jun L70B11 mouse h, m, r, mk II WB ab31419 rabbit h, m, r, mk II WB E254 rabbit h, m, r II IHC p-c-Jun 9164 rabbit h, m, r II IHC COL-I ab6308 mouse h, r, b, mk IV IF FN NCL-FIB mouse h III, IV IHC IST-4 mouse h III IHC BP8036 rabbit h IV IF Itgα6 MA6 rat m II IF, IP GoH3 rat h, b, m II FC, IP, FB H-87 rabbit m, r, h II IHC 3750 rabbit h, m, r, mk II WB HPA012696 rabbit h II WB ItgαV AV1 mouse h IV IF Itgβ1 AB1952 rabbit h, m, r II WB HA2/5 hamster r, m II FB 6S6 mouse h II FB P5D2 mouse h IV IF Itgβ4 346-11A rat m II WB M126 mouse h, m, r II WB Itgβ7 FIB27 rat m II FB M293 rat m II IF PCOL-I M-58 mouse h III IHC P-FAK (Y397) ab4803 rabbit h, m IV IF Tβ4 1-43 rabbit h I, IV IF 1-14 rabbit h UP§ IHC TGFBI 10188-1-AP rabbit h, m IV WB, IHC 2719 rabbit h IV IF TLN TD77 mouse h IV IF TN-C BC-24 mouse h III IHC DB7 mouse h III IHC α-tubulin 600-401-880 rabbit br II WB DM1A mouse br IV IF β-tubulin ab21057 goat h IV IF Vinculin VIIF9 mouse b, h, m, mk, rb IV IF VN 8E6(LJ8) mouse h IV IF *br=broad range, h=human, m=mouse, r=rat, mk=monkey, b=bovine, rb=rabbit §UP=unpublished **WB=Western blotting, IHC=immunohistochemistry, IF=immunofluorescence staining, IP=immunoprecipitation, FC=flow cytometry, FB=function blocking

3. RNA ANALYSES (I-IV)

3.1 Microarray analyses (I-IV)

PolyA RNAs or total RNAs, extracted from cell pellets or frozen tissue specimens, were analyzed using Incyte Genomics' cDNA microarrays (Mouse GEM 2/Unigene 1 LifeArrays and Human Unigene1 LifeArray) and Affymetrix oligonucleotide microarrays (MGU74A, MOE430 Set, HG U133 Set, or HG 1.0 ST). Mouse GEM 2/Unigene 1 arrays contained 9596 elements corresponding to 9307 total unique genes/clusters, Human Unigene1 contained 9182 elements representing 8472 unique genes or expressed sequence tags, and the MGU74A, MOE430 Set, HG U133 Set, and HG 1.0 ST arrays contained 12 654, >30 000, >39 000, and 28 869 genes/transcript variants, respectively. The validity of the Affymetrix probe sets of the genes selected for further studies were checked by BLAST searches of non- redundant databases at the National Center for Biotechnology Information. Further, the probe accuracies of the arrays have been reported to be quite high (the number of sequence-verified probes with perfect matches ≥ 85% in the older and ≥ 95% in the newer arrays; Alberts et al. 2007).

3.2 Northern blot analysis (I)

The Tβ4 mRNA levels of normal and transformed mouse fibroblasts, human melanocytes and melanoma cell lines, and a clinical melanoma metastasis were confirmed by Northern blot analysis. Hybridization with human β-actin cDNA control probe was used to check the uniform loading of the samples.

3.3 RT-PCR analysis (I-IV)

The expression levels of the identified candidate genes, various related molecules, and the Tβ4 antisense-RNA were confirmed by reverse transcription-PCR (RT-PCR). The primers and PCR conditions used are listed in Tables 4 and 5. For Tβ4 antisense-RNA, the forward primer was specific for the pLRT-vector and the reverse one for the Tβ4 antisense-RNA. The primers and PCR conditions for human β-actin, used as a loading control, were the same as previously described (Soikkeli et al. 2007).

Table 4. Primer sequences and PCR conditions used for mouse mRNAs. Gene name (Gene ID) Polarity Sequence 5' to 3' Annealing (ºC) Cycles* Ctsl (NM_009984) F ATGAGGAATTCAGGCAGGTGGTGA 56 18 R ACTCAGTGAGATCAGTTTGCCGGT Gapdh (NM_008084) F GGGTGTGAACCACGAGAAAT 59 30 R GGTCCTCAGTGTAGCCCAAG Itgα4 (AF109136) F AGGGTCCTTGTCGGAAGTGACAAA 58 35 R AGCCTGGGACCTCTTTGCCTTTAT Itgα6 (BC058095) F TGGAGGTACAGTTGTTGGTGAGCA 55 20 R AAACACCGTCACTCGAACCTGAGT ItgαE (DQ386608) F AACGGAGGGTCCTTTCTGTGTGAA 58 30 R GCATGTCTGAATTGAAGGCTGCGT Itgβ7 (BC011184) F AGTGTGCGACTGTAACTGTGGTGA 53 23 R ACTCTGCACAATCCCTGTACTGCT Itgβ1 (BC050906) F TCGCTGATTGGCTGGAGGAATGTA 57 20 R TCCTGCAGTAAGCGTCCATGTCTT Itgβ4 (L04678) F AGGGAGGCTGGCTTTCAATGTAGT 55 25 R TTCACCAGGTGCTCAGTGTCATCA Tβ4 (NM_021278) F CCTCATCCTCCTCGTCCTTA 56 30 R TGATCCAACCTCTTTGCATC Tβ4 antisense-RNA F CCTACAGGTGGGGTCTTTCA 62 23 R TCGCCAGCTTGCTTCTCTTG Tβ10 (NM_025284) F GTAAGAAAATGGCAGACAAGCC 50 30 R AGTCCGATTAGTGGAGGG *The numbers of cycles were optimized to be in the linear range. F = forward, R = reverse

Table 5. Primer sequences and PCR conditions used for human mRNAs. Gene name (Gene ID/reference) Polarity Sequence 5' to 3' Annealing (ºC) Cycles* FN (Tavian et al. 1994) F GGAGAGAGTCAGCCTCTGGTTCAG 56 25 R TGTCCACTGGGCGCTCAGGCTTGTG ITGβ1 (NM_002211) F TGCGAGTGTGGTGTCTGTAAGTGT 56 30 R TGACCACAGTTGTTACGGCACTCT Tβ4 (BT007090) F TGTCTGACAAACCCGATATGGCT 57 35 R GCCTGCTTGCTTCTCCTGTTCAAT TGFBI (NM_000358) F TGTGTGCTGAAGCCATCGTT 61 25 R ACATGGACCACGCCATTTGT TGF-β1 (BC022242) F ACAATTCCTGGCGATACCTCAGCA 61 25 R TCTTCTCCGTGGAGCTGAAGCAAT TGF-β2 (BC096235) F AAACAAGAGCAGAAGGCGAATGGC 61 30 R CTCCAGCACAGAAGTTGGCATTGT TGF-β3 (BC018503) F TGGACTTCGGCCACATCAAGAAGA 61 40 R TCTCCATTGGGCTGAAAGGTGTGA TLN1 (AF177198) F AGGAAGCATGTGGACCTTTGGAGA 61 30 R TGGCAGCCAGAAGAAGTTTGCTTG TN-C (Katenkamp et al. 2004) F GAGATTTAGCCGTGTCTGAGGTTG 57 25 R AGGAGAGATTGAAGCTCTCG TN-C (Latijnhouwers et al. 2000) F ACCGCTACCGCCTCAATTAC 57 25 R GGTTCCGTCCACAGTTACCA *The numbers of cycles were optimized to be in the linear range. F = forward, R = reverse. The used FN primers amplify both EDA+ and EDA- FN mRNAs. The TN-C primers published by Katenkamp et al. (TnA4F and TnBR) amplify the alternatively spiced domains A4-B containing mRNAs of large TN-C isoforms and those published by Latijnhouwers et al. (TEN5 and TEN6) amplify TN-C mRNAs containing or not containing the A1-D domains.

4. PROTEIN ANALYSES (I-IV)

4.1 Western blotting (I-IV)

Whole cell lysates, serum-free conditioned media (CM), or immunoprecipitated proteins in 1xLSB were resolved by SDS-PAGE and transferred onto nitrocellulose membrane. The membranes were then immunoblotted using primary antibodies detailed in Table 3 and horseradish peroxidase-conjugated secondary antibodies.

4.2 Immunofluorescence staining (I-II, IV)

Cells grown on glass coverslips were fixed with 3.5% paraformaldehyde, permeabilized with 0.1% Triton X-100 (except for TGFBI), and immunostained with the antibodies detailed in Table 3. The secondary antibodies used were conjugated with TRITC, FITC, or Alexa Fluor 488 dyes, and actin filaments were detected using Alexa Fluor 594 -conjugated phalloidin (Molecular Probes). Mounting was done with the Prolong Antifade Kit (Molecular Probes) or Vectashield H-1200 mounting medium containing DAPI (Vector Laboratories).

4.3 Flow cytometry (II)

The cell surface level of Itgα6 in 4N and Amdc-s cells were analyzed by flow cytometry. The primary antibody used was the clone GoH3 and secondary antibody was biotinylated goat anti-rat antibody. Finally, r-phycoerythrin-streptavidin (Vector Laboratories, Inc) was used for detection with the flow cytometer (FACScan; Becton Dickinson).

4.4 Immunoprecipitation (II)

For analysis of integrin heterodimers, Amdc-s cells were immunoprecipitated with Itgα6 antibody (GoH3 or MA6), and the immunoprecipitates were analyzed for the existence of the candidate partners Itgβ1 and Itgβ4 by Western blotting. Epithelial carcinoma cell line A431 was used as a positive control for Itgβ4.

4.5 Assay of cathepsin L activity (II)

The effect of neutralizing Itgα6 antibody on cathepsin L activity was tested on Amdc-s cells cultured both on plastic plates and on BD BioCoat Thin Layer Matrigel dishes. The cells were preincubated with neutralizing Itgα6 antibody before plating and grown for 2 days in the presence of the antibody. The CM were then collected and concentrated (the added immunoglobulins were first excluded using 100 kDa filter devices), and the activity of secreted cathepsin L was assayed as previously reported (Ravanko et al. 2004) in the presence of heparin.

5. TRANSFECTION EXPERIMENTS (I-II, IV)

5.1 Tetracycline-inducible antisense-RNA expression vector (I)

For tetracycline-inducible Tβ4 antisense-RNA expression, the rat Tβ4 cDNA (coding region) was inserted in reversed orientation into the inducible pLRT expression vector (Watsuji et al. 1997), and the resulting pLRT-Tβ4-antisense vector was then transfected into Amdc-s and - as cells using Lipofectamine Plus reagent (Invitrogen). To obtain stable cell lines (Amdc-s- pLRT-Tβ4-as and Amdc-as-pLRT-Tβ4-as), blasticidin S selection was used. To study the effect of decreased Tβ4 protein levels, the expression of Tβ4 antisense-RNA was induced by the addition of doxycycline (1 μg/ml; Sigma-Aldrich).

5.2 siRNA oligonucleotides (I-II)

In addition to Tβ4 antisense-RNA expression, small interfering RNAs (siRNAs) were used to diminish Tβ4 expression. For that, two siRNAs targeted towards the mouse Tβ4 mRNA coding region, as well as a negative control siRNA targeted towards a scrambled sequence were transfected into Amdc-s and -as cells using Oligofectamine, Lipofectamine 2000 (both from Invitrogen), or Fugene 6 (Roche Diagnostics GmbH) reagents. In addition, to target c- Jun expression, siRNA oligonucleotide pool to mouse c-Jun (sc-29224) and control siRNA-A

(sc-37007; both from Santa Cruz Biotechnology) were transfected into Amdc-s cells using the same transfection reagents.

5.3 shRNA expression vectors (II, IV)

To target c-Jun expression with short hairpin RNA (shRNA) expression, empty pSuper.retro.puro vector and the c-Jun shRNA construct, kindly provided by Dr. Kazushi Inoue (Wake Forest University Health Sciences, Winston-Salem, NC), were transfected into Amdc-s cells using the above mentioned reagents. To study the effect of TGFBI knockdown on melanoma cells, empty pLKO.1 vector and shRNA target set for NM_000358 (TGFBI; both from Open Biosystems), and MISSION Non-target shRNA control vector (Sigma- Aldrich) were transfected into WM239 cells using Lipofectamine 2000. To obtain stable cell lines, puromycin selection was used and the cells were cloned using the cloning cylinder technique. Two selected TGFBI shRNA clones, 239 G5B and 239 G6C (bearing TGFBI shRNA constructs G5 and G6, respectively), as well as one non-target shRNA clone 239 nt3A and one empty vector clone 239 pLKO.1 were used in the following experiments (except only 239 nt3A and 239 G6C were tested in vivo).

6. FUNCTIONAL ASSAYS (I-IV)

6.1 Adhesion assays (II, IV)

To study cell attachment onto coated surfaces, two assay formats were used: the 12-well plate format (II), and the 96-well plate format (IV). In both formats, the wells were first coated with the studied proteins (LN and pFN used for 12-wells; TGFBI, cFN, pFN, COL-I, COL- IV, LN, and VN used for 96-wells), after which the cells were allowed to attach. Fatty acid free, low-endotoxin BSA served as a control in both the assay formats. The quantification of cell attachment differed between the assay formats. In the 12-well format, the wells were washed once, the attached cells were trypsinized, and the cell numbers were counted with a Coulter particle counter (Beckman Coulter). In the 96-well format, the wells were

photographed after 1h 30 min (± 15 min) adhesion and the percentage of attached and spread cells was then calculated manually from printed images (2 replicas, 3 images from each well).

6.2 Analysis of cell growth (I-II, IV)

To study the proliferation rates of the cells, a fixed number of cells were added onto 12-well plates in triplicates, and the total number of trypsinized cells were counted at selected time points using a Coulter particle counter. To test the effect of latrunculin A (LAT-A) on cell growth (I), varying concentrations of the toxin were added to the cells 2 days after subculture. To test the effect of Itgα6, Itgβ1, and Itgβ7 antibodies, the antibodies were preincubated with the cells and included in the growth medium.

6.3 Migration assays (I, III-IV)

Cell migration was tested using the traditional scratch wound assay (I), three-dimensional (3D) collagen gel assay (I, III-IV), and Boyden chamber assay (IV). To study the effect of LAT-A using scratch wound assay, 80-90% confluent cell cultures were wounded using sterile pipette tips and cell migration into the wound area was then followed by microscopy for 24 h in the presence or absence of LAT-A. In 3D-collagen gel assays, the cells were cast between two COL-I gel layers, covered with growth medium (LAT-A, when studied, was included in this medium; TN-C and FN were added into the upper gel layer and the growth medium) and followed for several days by microscopy. In Boyden chamber assays, the chambers were first coated with COL-I on both sides of the membrane, and migration of the cells from the upper compartment (serum-free medium) into the lower compartment (10% serum) was quantitated by crystal violet staining, photography, and manual counting from printed images.

6.4 Invasion assays (I-II, IV)

Cell invasiveness in vitro was analyzed using thick gels of Growth Factor Reduced Matrigel (BD Biosciences), where the cells were cast between two Matrigel layers. To see the effect of LAT-A on cell invasion, it was included in the Matrigel layers and the uppermost growth

medium. Neutralizing antibodies were, in turn, preincubated with the cells and included into the Matrigel layers. The invasive growth of the cells in the matrix was then followed by microscopy for one week. To see the effect of TGFBI knockdown on melanoma cell invasion in vivo, 3 x 106 TGFBI-KD cells (239 G6C) or non-target shRNA control cells were injected subcutaneously (diluted 1:1 with 75 µl Growth Factor Reduced Matrigel) into Balb/c nu/nu mice (6 animals per group, one tumor per animal). Animal protocols were approved by the Experimental Animal Committee of the Provincial Government of Southern Finland. Tumor growth was monitored thrice a week using calipers, and the mice were euthanized after 30 days. The tumors, lungs, livers and lymph nodes (one per animal) were removed and embedded in paraffin or frozen in liquid nitrogen (two tumors per group).

7. IMMUNOHISTOCHEMISTRY (II-IV)

Tissue sections were cut at 5 μm thickness and deparaffinised, and antigen retrieval was accomplished by heating in microwave oven, except for FN, TN-C, and PCOL-I staining, where the samples were treated with trypsin. Endogenous peroxidase activity was then blocked and the sections were incubated with the primary antibodies detailed in Table 3 at 4°C overnight. The immunoreactivity was detected by processing with the StreptABComplex/HRP Duet kit, the chromogen being either 3,3'-diaminobenzidine (DAB) or 3-amino-9-ethylcarbazole (AEC). Finally, the sections were gently counterstained with Harris or Mayer’s Hematoxylin (Polysciences Inc) and mounted. In immunohistochemical analysis of melanoma tumors, the tumors were semi-quantitatively scored as −, +, ++, or +++ corresponding to absent, weak, moderate and strong/very strong immunoreactivity, respectively. This quantification was made double blindly. Correlative analyses of the data were done by chi-square tests.

RESULTS AND DISCUSSION

1. MICROARRAY ANALYSES (I-IV)

1.1 Incyte Genomics versus Affymetrix arrays (I-IV)

To identify molecules potentially related to the invasiveness of fibrosarcomas and melanomas, we first compared the gene expression profiles of the studied transformed/malignant cells and the corresponding normal cells using two different microarray platforms: cDNA (Incyte Genomics' LifeArrays) and oligonucleotide (Affymetrix GeneChips) arrays. Two platforms were used to make the results more confident, an idea that has been validated in a previous study (Lee et al. 2003). Incyte Genomics’ analyses comprised competitive hybrization of two differently fluorescently labeled RNA samples to cDNA fragments (average length 1000 bp) printed onto glass slides, whereas Affymetrix analyses are based on two independent hybridizations of biotinylated cRNA samples to glass- immobilized oligonucleotides (25mers) synthesized in situ, and fluorescent detection using antibody amplification.

With both the platforms used, the highest changes were quite similar (I: Tables 1 and 2; Table 6 and IV: Supplemental Table 1). As expected, the differing gene contents of the arrays resulted in some differences in the lists of the topmost changes, one example being transforming growth factor beta-induced (further studied in IV), which was only present in Affymetrix’ arrays. Also, the exact fold change values expectedly varied between the two analysis methods (see II), but the topmost changes detected with both of the systems usually nicely confirmed each other in terms of being markedly changed in both. As the fold changes in repeated analyses using the same method also showed variation (one reason being biological variation), it is advisable not to concentrate too much on the exact fold changes. The Incyte Genomics' competitive system was still applicable with this small number of samples, but for more extensive studies, the Affymetrix non-competitive platform is more useful.

Table 6. The most marked gene expression changes in the studied melanoma cell lines as compared to normal melanocytes (42V) by Incyte Genomics Human Unigene1 LifeArrays.

Incyte Genomics, increased expression Genbank Gene Title Signal Fold Signal Fold 42V WM793 WM793 42V WM239 WM239 BF795929 MHC, class II, DR alpha (HLA-DRA) 934 29466 31.5 1133 9544 8.4 AW385690 Fibronectin 1 1715 46491 27.1 1783 22024 12.4 BF339409 Serpin peptidase inhibitor, clade A, member 3 972 25952 26.7 961 6330 6.6 BG679767 Tenascin C 586 14978 25.6 739 9945 13.5 AW609791 MHC, class II, DR beta 3 (HLA-DRB3) 674 14544 21.6 793 4113 5.2 BG333618 CD74 molecule, MHC, class II invariant chain 1151 24341 21.1 2194 14871 6.8 NM_002421 Matrix metallopeptidase 1 557 10903 19.6 1263 20123 15.9 NM_006953 Uroplakin 3A 1065 19970 18.8 562 456 <2.5 NM_000885 Integrin alpha 4 118 2177 18.4 583 2293 3.9 BE439689 MHC, class II, DP alpha 1 (HLA-DPA1) 1083 18419 17.0 1173 6590 5.6 M20503 MHC, class II, DR beta 1 (HLA-DRB1) 1188 19944 16.8 1016 4542 4.5 BG685150 MHC, class II, DR beta 1 (HLA-DRB1) 1430 23375 16.3 1767 6917 3.9 AF085224 Solute carrier organic anion transporter family, member 1A2 570 9066 15.9 1057 3590 3.4 AF053356 Procollagen C-endopeptidase enhancer 1614 24267 15.0 3381 19029 5.6 M62424 Coagulation factor II (thrombin) receptor 118 1727 14.6 112 538 4.8 NM_006533 Melanoma inhibitory activity 637 8993 14.1 1041 3659 3.5 AI588970 Scrapie responsive protein 1 141 1988 14.1 121 641 5.3 AL549837 Cysteine-rich, angiogenic inducer, 61 1018 11342 11.1 766 3656 4.8 NM_001673 Asparagine synthetase 568 5940 10.5 639 2192 3.4 AB002351 Synemin 513 5340 10.4 1546 3884 2.5 NM_001901 Connective tissue growth factor 703 7137 10.2 2631 5838 <2.5 AU137979 Lumican 188 1924 10.2 395 281 <2.5 AV714998 UDP-glucose pyrophosphorylase 2 1058 10840 10.2 2602 2921 <2.5 NM_004911 Protein disulfide isomerase family A, member 4 2043 20484 10.0 2973 20588 6.9 AL581504 Solute carrier family 20, member 1 472 4598 9.7 754 2055 2.7 AV655628 Heat shock protein 90 kDa beta 2924 27982 9.6 1641 10538 6.4

Table 6 continues

Incyte Genomics, decreased expression Genbank Gene Title Signal Fold Signal Fold 42V WM793 WM793 42V WM239 WM239 BG765374 Silver homolog (mouse) 56917 519 100.0 47875 1020 46.9 X51420 Tyrosinase-related protein 1 23817 574 41.5 13442 591 22.7 AF035307 Chromosome 12 genomic cintig 4051 334 12.1 2781 230 12.1 AI097486 Oculocutaneous albinism II 3499 292 12.0 2379 504 4.7 AV717752 Microphthalmia-associated transcription factor 6263 534 11.7 3086 235 13.1 AL533154 Guanosine monophosphate reductase 7400 646 11.5 6658 959 6.9 BF800512 Tetratricopeptide repeat domain 3 8689 775 11.2 5153 709 7.3 AB002360 MCF.2 cell line derived transforming sequence-like 2897 261 11.1 1823 705 2.6 NM_001793 P-cadherin (placental) 7268 718 10.1 6017 1119 5.4 NM_001935 Dipeptidyl-peptidase 4 (CD26) 8013 848 9.4 4050 1349 3.0 BE302598 Calpain 3 (p94) 4228 500 8.5 3230 1115 2.9 AI826630 RAB32, member RAS oncogene family 5492 651 8.4 4097 818 5.0 BF245562 Microphthalmia-associated transcription factor 3174 389 8.2 2660 458 5.8 AW162151 Chromosome 4 (unique) 234 expressed sequence 2117 268 7.9 1463 289 5.1 AI127103 Sorting nexin 19 4904 642 7.6 3383 1443 <2.5 BG704431 Apolipoprotein E 3900 517 7.5 1674 465 3.6 BF573292 Spermidine/spermine N1-acetyltransferase 1 15322 2221 6.9 9124 1226 7.4 BG760324 Potassium voltage-gated channel, shaker-related subfamily, beta member 2 6168 972 6.3 6772 2117 3.2 NM_000222 V-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog 4036 688 5.9 3201 763 4.2 BG763127 Vesicle amine transport protein 1 homolog 21020 3617 5.8 15538 6543 <2.5 AL120794 RAB27A, member RAS oncogene family 7660 1321 5.8 3659 1228 3.0 AB002438 Chromosome 5 genomic contig 1939 343 5.7 1523 321 4.7 NM_001266 Carboxylesterase 1 4170 751 5.6 4759 1641 2.9 NM_003272 G protein-coupled receptor 137B 4330 784 5.5 379 126 3.0 BE837728 Endothelin receptor type B 5522 1011 5.5 3646 791 4.6 Z13009 E-cadherin (epithelial) 2920 561 5.2 2870 571 5.0 The gene annotations have been updated by BLAST searches of non-redundant databases at the National Center for Biotechnology Information. The genes showing parallel changes by Affymetrix HG U133 Set’s A-arrays are shown in bold. Note that TGFBI was not included in the Human Unigene1 arrays.

1.2 Mouse fibrosarcoma cells (I-II)

To study the gene expression changes behind the high invasiveness of mouse Amdc-s and -as fibrosarcoma cells, we compared their gene expression patterns with that of the empty vector- bearing 4N cells. The results (I: Tables 1 and 2) revealed the gene expression patterns of Amdc-s and -as cells to be surprisingly similar even if the former was transformed by overexpression and the latter by blockage of the expression of the same gene, S- adenosylmethionine decarboxylase (AdoMetDC). This result was, however, in concert with their similar, transformed phenotype. As we also analyzed an exceptionally aggressive clone of Amdc-as cells, we could study the effect of the degree of transformation on the gene expression levels. Thymosin β4, as an example, showed 14.0-fold up-regulation in the pool of Amdc-as cells (26.4 and 20.6 in pools of Amdc-s cells), whereas it showed 62.2-fold up- regulation in an extremely aggressive clone (I: Table 1). Similarly, integrin β7 showed 6.5- fold up-regulation in the pool of Amdc-as cells (8.4 and 8.2 in pools of Amdc-s cells), and 22.4-fold up-regulation in the aggressive clone (I: Table 1). Thus, the fold change values obtained greatly correlates with cell aggressiveness.

The highest gene expression changes potentially relevant for the high invasiveness of Amdc-s and -as cells included thymosin β4, encoding an actin-sequestering protein related to cell motility (the highest up-regulation with both the platforms), cell surface adhesion molecules encoding integrin α6, integrin β7, and protocadherin 7 (all increased), ECM components/modifiers encoding cathepsin L, chondroitin sulfate proteoglycan 2, decorin, matrix metallopeptidase 2 (type IV collagenase), hyaluronan synthase 2, ECM-associated protein Sc1, osteopontin (all increased), and procollagen type I α1, microfibrillar-associated protein 5, and thrombospondin 1 (all decreased). Of these, the cysteine proteinase cathepsin L has been previously shown to be the major secreted protein of Amdc-s cells and to play a major proteolytic role in the invasiveness of these cells (Ravanko et al. 2004). In subsequent studies, we concentrated on studying the function of the actin cytoskeleton-related molecule thymosin β4 (Tβ4), and the adhesion molecules integrin α6 (Itgα6) and integrin β7 (Itgβ7) in Amdc-s and -as cell invasion. Further, we studied the presence and localization of Tβ4, ITGα6, and cathepsin L in human sarcoma specimens by immunohistochemistry (see below).

1.3 Human melanoma cells (III-IV)

To study the invasion/progression-related gene expression changes in melanoma cells, we analyzed the gene expression patterns of two melanoma cell lines, vertical growth phase (VGP) WM793 cells and metastatic WM239 cells, as compared to those of two primary cultures of melanocytes, 42V and Mela3 (Table 6 and IV: Supplemental Table 1). The gene expression patterns of the studied melanoma cells verified many of the known melanoma- related changes, reported for instance in an analysis of multiple melanoma cell line microarray studies utilizing the same Affymetrix HG-U133A arrays (Hoek 2007). However, some changes not reported in that study were also identified. The reason for this discrepancy may be the varying characteristics (e.g. invasiveness) of the studied cell lines, as in Hoek’s analysis only genes showing up/down-regulation in all the studied cell lines were included in the results. First, both the studied melanoma cell lines showed decreased expression of the gene encoding the cell-cell adhesion molecule E-cadherin, a change that allows melanoma cells to lose adhesive interactions with keratinocytes and become more migratory and invasive (Satyamoorthy and Herlyn 2002; Lee and Herlyn 2007). Consistently, both cell lines also showed up-regulation of N-cadherin (2.2- and 3.3-fold changes, respectively, with Affymetrix arrays) helping them to interact with dermal fibroblasts and endothelial cells (Satyamoorthy and Herlyn 2002). Similarly to E-cadherin, the melanocyte-keratinocyte interaction molecule encoding P-cadherin (Rickelt et al. 2008), was found to be down- regulated in both the cell lines. Further, up-regulated expression of the desmosomal cadherin encoding desmoglein 2, previously reported in some melanomas (Rickelt et al. 2008), was detected in both the cell lines (8.0 and 7.3, respectively, with Affymetrix arrays).

Besides E- and P-cadherins, the shared down-regulated genes in the studied amelanotic WM793 and WM239 cells (ceased to produce the pigment melanin) included melanin synthesis related molecules encoding silver homolog (mouse), melan-A, tyrosinase, microphthalmia-associated transcription factor, and oculocutaneous albinism II. Further, both cell lines showed down-regulated expression of the genes encoding serine protease dipeptidyl-peptidase 4 [whose down-regulation has previously been found to be an important early event in melanoma pathogenesis (Van den Oord 1998; Wesley et al. 1999)], a nucleotide modifying enzyme guanosine monophosphate reductase, an actin-regulatory molecule capping protein (actin filament), gelsolin-like, and a cation channel melastatin 1

[10.4- and 9.8-fold changes with Affymetrix arrays; whose loss of expression has previously been found to correlate with melanoma aggressiveness (Duncan et al. 1998)].

In addition to the up-regulated expression of the cell-cell adhesion molecules encoding N- cadherin and desmoglein 2, WM793 and WM239 cells showed up-regulated expression of genes encoding the ECM modification related molecules matrix metallopeptidase 1 [also known as collagenase-1; the highest increase in both cell lines using Affymetrix arrays; expression previously reported to correlate with melanoma invasion and rapid progression (Airola et al. 1999; Nikkola et al. 2002)], procollagen C-endopeptidase enhancer, cathepsin C (6.3- and 6.1-fold changes with Incyte Genomics’ arrays), and tissue inhibitor of metalloproteinase 3 (11.8- and 5.7-fold changes with Affymetrix arrays). Further, both the cell lines showed up-regulated expression of the cell adhesion molecules encoding thrombospondin 1, integrin α4, integrin α6 (4.8- and 3.8-fold changes with Incyte Genomics’ arrays), and integrin β3 (10.5- and 14.5-fold changes with Affymetrix arrays), as well as integrin activity modulating protein melanoma inhibitory activity. Further, the cell lines were found to show increased expression of genes encoding various mitogens (and their function modulating proteins), chemokines, and cytokines, such as connective tissue growth factor, platelet-derived growth factor α (11.5- and 15.4-fold changes with Affymetrix arrays), transforming growth factor α, interleukin 8, monocyte chemotactic protein (9.6- and 3.6-fold changes with Affymetrix arrays), and insulin-like growth factor binding proteins 2 and 3. Interestingly, the cell lines also showed overexpression of the angiogenesis-related molecules encoding cysteine-rich angiogenic inducer 61, vascular endothelial growth factor A, neuropilin-1 (6.4- and 5.4-fold changes with Affymetrix arrays), and angiopoietin 1 (2.3- and 2.1-fold changes with Affymetrix arrays). One more, interesting gene, encoding an inhibitor of retinoic acid signaling, preferentially expressed antigen in melanoma (7.7- and 24.3-fold changes with Affymetrix arrays) showed up-regulation. In line with previous studies (Hoek et al. 2004; Hoek 2007; Degenhardt et al. 2010), major histocompatibility complex (MHC) class II molecules (especially DR α), normally expressed in antigen presenting cells, were also found to be up-regulated in both the cell lines. As reported by Degenhardt et al. (Degenhardt et al. 2010), the metastatic cells (WM239) most often showed lower expression of MHC class II molecules than the VGP cells (WM793).

Moreover, besides the above mentioned shared changes, the VGP WM793 cells also showed increased expression of the proteoglycan lumican encoding gene. Metastatic WM239 cells, in 53

turn, additionally showed increased expression of the genes encoding ECM modifying proteases matrix metallopeptidase 15 (also known as membrane-type 2 MMP; 4.1-fold change with Affymetrix arrays) and matrix metallopeptidase 8 [6.6-fold change with Affymetrix arrays; high serum levels found to correlate with melanoma ulceration and vascular invasion (Vihinen et al. 2008); wild type protein shown to inhibit melanoma progression, but inactivating mutations often found in melanoma tumors (Palavalli et al. 2009)], ECM component collagen type XV, α1 (20.4-fold change with Affymetrix arrays), angiogenic mitogen pleiotrophin (3.5-fold change with Affymetrix arrays), and cytokine macrophage-specific colony-stimulating factor (3.8-fold change with Affymetrix arrays). As most of these “cell line specific” changes (apart from matrix metallopeptidase 8) were not included in the Hoek’s list of changes typical to melanoma (Hoek 2007), they appear not to represent universal melanoma-related changes, but may be related to metastatic alterations.

In addition to the above mentioned cell adhesion and ECM modification-related molecules, WM793 and WM239 cell lines interestingly showed increased expression of various ECM components, including fibronectin 1 (FN) and tenascin-C (TN-C) (being the most highly up- regulated ECM proteins with both microarray platforms), as well as collagen type IX, α3 (9.0- and 16.4-fold changes with Affymetrix arrays), consistent with a previous study (Hoek et al. 2004). One additional interesting ECM protein, transforming growth factor beta-induced (TGFBI), was also among the highest up-regulated genes using Affymetrix arrays (2nd highest in WM239 cells; 108th highest in Hoek’s analysis). In subsequent studies, we concentrated on studying the role of the cell adhesion-related molecules TN-C, FN, and TGFBI in melanoma cell invasion and melanoma progression.

1.4 Fibrosarcoma versus melanoma (I-IV)

As predicted from the different origins of fibrosarcoma and melanoma cells, the gene expression patterns of AdoMetDC-transformed cells and melanoma cells were quite different, especially the topmost changes. Among the changes in common were, however, up-regulated expression of integrin α6 and TN-C (2.5- and 2.0-fold changes in Amdc-s and -as cells as analyzed with Incyte Genomics’ arrays). Interestingly, TGFBI was found to show up- regulated expression only in the extremely aggressive clone of Amdc-as fibrosarcoma cells (6.8-fold change with Incyte Genomics’ array), which is consistent with the previous studies

reporting elevated TGFBI expression to be related to the aggressiveness of the tumor (Zajchowski et al. 2001; Sasaki et al. 2002; Ma et al. 2008).

In addition to the shared gene expression changes between fibrosarcoma and melanoma cells, a few genes showing contrasting behavior were identified. They included e.g. cysteine-rich protein 61 (Cyr61), thrombospondin 1, and protease, serine, 23, which were down-regulated in Amdc-s and -as cells, but up-regulated in the studied melanoma cells (14.6- and 4.4-fold changes in WM793 and WM239 cells with Affymetrix arrays). Of these, Cyr61 has previously been shown to induce apoptosis of fibroblasts, but to promote the survival of endothelial cells (Todorovicc et al. 2005), so the difference in its expression levels is understandable and can indeed be cell type-specific. Also the angiogenesis inhibitor thrombospondin-1 has been reported to show varying expression levels in different cancers. Protease, serine, 23 (also known as SPUVE), in turn, is still a poorly studied ovarian protease which has been found to be overexpressed in papillary thyroid carcinomas (Huang et al. 2001).

2. FUNCTIONAL CHARACTERIZATION (I-IV)

2.1 Thymosin β4 (I)

Tβ4 is a strongly polar 5 kDa polypeptide that functions as a major actin sequestering molecule (see the section “Actin polymerization”). Tβ4 sequesters a large pool of monomeric actin (G-actin) by forming a 1:1 complex with it, and from this pool, the actin monomers can then be easily released for the rapid polymerization of actin filaments (F-actin). Interestingly, previous reports have shown up-regulated expression of Tβ4 in various cancer cell lines and human tumors (Hall 1991; Yamamoto et al. 1993; Santelli et al. 1999; Xie et al. 2002), and in functional studies, increased Tβ4 expression has been related to increased tumorigenicity and metastatic potential (Clark et al. 2000; Kobayashi et al. 2002; Cha et al. 2003; Ji et al. 2003; Wang et al. 2003; Wang et al. 2004;). Based on these data and the highly up-regulated expression, we decided to study the possible role of Tβ4 in the high invasiveness of AdoMetDC-transformed cells.

2.1.1 Tβ4 is overexpressed in Amdc-s, Amdc-as, and E4 fibrosarcoma cells

The highly up-regulated expression of Tβ4 in Amdc-s and -as cells was first verified with Northern blotting (I: Fig. 1A) and RT-PCR (data not shown). In addition to these cells, a marked up-regulation of Tβ4 was detected in ras-tranformed mouse fibroblasts (E4; I: Fig. 1A) and the metastatic WM239 melanoma cells (I: Fig. 1C). Additionally, the change at the protein level was confirmed in fibrosarcoma cells by immunofluorescence staining (I: Fig. 1D).

2.1.2 Tβ4 overexpression is related to the transformed phenotype

In line with the role of Tβ4 as a G-actin reservoir, the Tβ4-overexpressing cells were found to show distinctive actin staining pattern as compared to 4N cells (Fig. 3). Previously, our group has shown that activation of the proto-oncogene c-jun, a member of the dimeric activator protein 1 (AP-1) transcription factor complex, is required for AdoMetDC-induced transformation and expression of a dominant-negative mutant of c-Jun (TAM67), inhibiting the function of c-Jun, is able to reverse the transformed phenotype (Paasinen-Sohns et al. 2000; II). Here, we also analyzed the Tβ4 expression levels of Amdc-pLRT-TAM67 cells before and after TAM67-induced reversal of transformation, and found them to be clearly regulated by c-jun (I: Fig. 1B). Tβ4 expression level did not, however, decrease back to the level of control 4N cells upon TAM67 induction, indicating the existence of other regulatory molecules as well.

Figure 3. The staining patterns of actin filaments in 4N (a), Amdc-s (b), and ras-tranformed E4 cells (c), as detected with Alexa Fluor 594 -conjugated phalloidin. Note the areas of highly concentrated actin filaments (containing pseudopodia-like structures) at the edges of the migratory Amdc-s and E4 cells (b and c). Exposure time 4000ms in a and 2000ms in b and c.

To study the effects of Tβ4 overexpression in AdoMetDC-transformed cells, we blocked its expression by inducible antisense-RNA expression or addition of siRNAs (inducible antisense-RNA shown in I: Fig. 2A). This expression knockdown led to reversion of the transformed morphology and impaired dispersion/migration of the cells in three-dimensional (3D)-collagen gel (I: Fig. 2B-C). The degree of morphological reversion, however, varied, and the effect tended to weaken over an extended time period, indicating that some compensatory mechanism may be operational.

2.1.3 LAT-A, an inhibitor of Tβ4, affects the morphology and growth of Amdc-s and -as cells

Latrunculin A (LAT-A; structure shown in Fig. 4) is a sponge toxin shown to inhibit the binding of Tβ4 to actin (Yarmola et al. 2000). LAT-A binds G-actin at 1:1 stoichiometry, similarly to that of Tβ4, and thereby inhibits actin polymerization (Spector et al. 1983; Coue et al. 1987). When Amdc-s and -as cells were exposed to this toxin (0-350 ng/ml), their morphology changed in a concentration- and time-dependent manner into ball-and-stick model-looking cells with a round center and thin protrusions (I: Fig. 3C-F). With high concentrations or prolonged periods of exposure, the cells finally detached and become swollen, probably due to impaired cytokinesis. No activation/cleavage of caspase-3, indicating apoptosis, was detected in these detached cells by immunoblotting. The morphological effect of LAT-A on these cells was reversible after a short exposure, but after a longer exposure of high concentration (350 ng/ml for 2 days or longer) the cells did not recover fully, and many of the cells had multiple nuclei (implying impaired cytokinesis as well) (I: Fig. 3G-H). In contrast to Tβ4-overexpressing cells, normal fibroblasts were quite resistant to LAT-A’s morphological effect (I: Fig. 3A-B).

In addition to the morphological effect, LAT-A selectively inhibited the growth of Amdc-s and -as cells as compared to that of 4N cells (I: Fig. 4). Even with only 50 ng/ml LAT-A (half of the concentration used for morphological effects), the growth of Tβ4-overexpressing cells was seriously inhibited. As the untreated 4N and Amdc-s cells showed similar growth rates, Tβ4 overexpression itself does not confer any growth advantage to these transformed cells.

Figure 4. The structure of LAT-A, a macrolide isolated from the Red Sea sponge Negombata magnifica (formerly Latrunculia magnifica). The sponge uses this toxin to defend itself from hungry fishes. Modified from Morton et al. 2000.

2.1.4 LAT-A impairs the migration of Amdc-s, Amdc-as, and E4 cells

As reorganization of the actin cytoskeleton is important for cell migration, we also tested the effect of LAT-A on cell motility. In routine cell culturing we had noticed Amdc-s and -as cells to be more migratory than control 4N cells. As anticipated, the toxin clearly affected the migration of both Amdc-s and -as cells, as well as that of Tβ4-overexpressing E4 cells (ras- transformed fibroblasts), when analyzed by scratch wound closure proliferation/migration assay (Amdc-s shown in Fig. 5), commonly used to assess the effects of drugs, or 3D- collagen gel assay (data not shown). Again, the toxin had a milder effect on control 4N cells showing lower Tβ4 expression.

2.1.5 LAT-A blocks the invasion of Amdc-s and -as cells in Matrigel

Matrigel matrix is a basement membrane extract from mouse sarcoma tumor, consisting mainly of LN, COL-IV, entactin, and heparan sulfate proteoglycan. Matrigel allows the reconstitution of 3D-ECM around tumor cells in vitro, mimicking the situation in vivo. Amdc-s and -as cells normally grow and invade well in this ECM matrix, being more invasive than human HT-1080 fibrosarcoma cells, MDA-MB-231 breast cancer cells (Ravanko et al. 2004), or ras-oncogene-transformed fibroblasts (E4; Paasinen-Sohns et al. 2011). When Amdc-s and -as cells were grown in 3D-Matrigel matrix in the presence of LAT-A, their invasion was dramatically inhibited (I: Fig. 5A-C). Even with a concentration of 50 ng/ml, the invasion was effectively impaired, and with a concentration of 100 ng/ml it was totally inhibited. In addition, LAT-A was also able to block Amdc-s and -as cell invasion when added to already formed invasive colonies, and even seemed to shrink the existing

colonies (I: Fig. 5F-H). The growth of the control, noninvasive 4N cells was, as expected, not affected by LAT-A (I: Fig. 5D-E).

Figure 5. Scratch wound assay comparing the migration of control 4N cells and Tβ4-overexpressing Amdc-s cells in the absence and presence of LAT-A (25 ng/ml).

2.1.6 Tβ4 shows strong immunostaining in Amdc-s-induced mouse fibrosarcomas

To study the distribution of Tβ4 in tissue specimens, Amdc-s-induced murine fibrosarcomas and human sarcoma specimens (n = 10) were analyzed by immunohistochemistry (IHC). In normal mouse tissues, high Tβ4 immunostaining was detected especially in hair follicles (Fig. 6a), whereas in the induced fibrosarcomas, the immunoreactivity was found scattered around the tumor (Fig. 6b). Further, high-grade human sarcomas were found to show intense

Tβ4 immunostaining (Fig. 6c and 7b). These data thus support the idea of Tβ4 as a potential target for cancer therapy.

Figure 6. Amdc-s cell-induced mouse fibrosarcomas and human sarcomas show intense Tβ4 immunostaining. a) Normal mouse tissue, b) Amdc-s cell-induced tumor tissue, and c) undifferentiated pleomorphic human sarcoma. Magnification 400x.

Our study was the first to show that a small-molecular toxin interfering with Tβ4 binding to actin can selectively inhibit cancer cell invasion. Two years later, El Sayed et al. (Sayed et al. 2008) reported that LAT-A shows anti-invasive activity against highly metastatic human prostate cancer cells in vitro, and, interestingly, suppress hypoxia-induced HIF-1 activation in breast tumor cells. Three years after our work, Konishi et al. (Konishi et al. 2009) further demonstrated the ability of LAT-A to inhibit the dissemination of cancer cells in vivo (peritoneal dissemination of human gastric cancer in mice). Notably, no major side-effects were observed in the mice after intraperitoneal injection of LAT-A. Still more recently, Khanfar et al. (Khanfar et al. 2010) have shown LAT-A derivatives to be even more potent than the parental molecule in inhibiting breast cancer cell invasion. Taken together, these data strongly encourage testing LAT-A-like molecules for the treatment of Tβ4-overexpressing tumors showing high invasive and metastatic potentials.

2.2 Integrins α6 and β7 (II)

Integrins (ITGs) are the major cell surface receptors mediating adhesion to ECM molecules, being, at the same time, the major migration-mediating receptors (see the section “Cell- matrix interactions”). Integrins are heterodimeric molecules, consisting of one α and one β subunit, and the subunit composition specifies their ligand specificity (e.g. FN, LN, COL, and 60

VN). Integrin α6 subunit (ITGα6) can dimerize with subunits β1 and β4, and both the heterodimers function as laminin receptors. Of these partners, integrin β1 is universally expressed, whereas the β4 subunit is mainly expressed in epithelial cells. The leukocyte- specific integrin, integrin β7, in turn, can pair with subunits α4 and αE. The resulting heterodimers show specificity to FN, the vascular addressins MAdCAM-1 and VCAM-1 (α4), and E-cadherin and LEEP-CAM (αE). High ITGα6 expression has previously been reported in several epithelial cancers, including high-grade breast carcinomas (Friedrichs et al. 1995) and invasive oral squamous cell carcinomas (Shinohara et al. 1999). ITGβ7 up- regulation, in turn, has been reported in multiple myelomas and T cell lymphomas (Hurt et al. 2004; Morito et al. 2006). As cell adhesion is one of the processes essential to cancer cell invasion, we further concentrated on studying the importance of the up-regulated integrins α6 and β7 in the invasiveness of Amdc-s cells.

2.2.1 Itgα6 and Itgβ7 are up-regulated in Amdc-s, Amdc-as, and E4 fibrosarcoma cells

The up-regulated expression of Itgα6 and Itgβ7 was verified in Amdc-s cells by RT-PCR analysis (II: Fig. 2a’) and confirmed at protein level by immunofluorescence staining (II: Fig. 2b’), flow cytometric analysis (II: Supplementary Fig. 2A), and Western blotting (II: Supplementary Fig. 2B). In addition, increased expression of both genes was found in ras- transformed E4 cells, and that of ITGα6 was found in human WM793 and WM239 melanoma cells, and HT-1080 fibrosarcoma cells (data not shown). Both Itgα6 and Itgβ7 expression and immunostaining were decreased in Amdc-pLRT-TAM67 cells upon TAM67 induction (II: Fig. 2a’ and c’), showing that c-Jun is also able to regulate the expression of Itgα6 and Itgβ7, in addition to Tβ4.

2.2.2 Itgα6β1 mediates the binding of Amdc-s cells to laminin

To find out the main dimerization partner of Itgα6 in Amdc-s cells, we did immunoprecipitation experiments. Of the two known dimerization partners, β1 and β4 subunits, we found Itgβ1 to be the principal partner of Itgα6 in these cells (II: Fig. 4a’). In further studies, function blocking antibodies specific to either Itgα6 or Itgβ1 inhibited the

binding of Amdc-s cells to laminin and Matrigel (consisting mainly of laminin) (II: Fig. 4b’- d’). With the antibody specific to Itgβ1 the inhibition was total and longer lasting, probably due to the numerous dimerization partners of β1 (more than ten).

We could not solve out the main dimerization partner of Itgβ7 by immunoprecipitation. However, by RT-PCR we found Amdc-s cells to express ItgαE, but not Itgα4 (data not shown). Further, we did not find Itgβ7 antibody to inhibit Amdc-s cell binding to fibronectin (ligand of Itgα4β7). Despite our failure in identification of the partner and its function, to our knowledge, this is the first study to associate increased Itgβ7 level with the transformed phenotype of non-leukocytic cells. Similarly to those of Tβ4 and Itgα6, the expression of Itgβ7 was found to be regulated by c-Jun in Amdc-s cells (II: Fig. 2a’ and c’).

2.2.3 Itgα6β1 is relevant for the invasion of Amdc-s cells and human HT-1080 fibrosarcoma cells in Matrigel

The possible role of Itgα6β1 in the high invasiveness of Amdc-s cells was tested in 3D- Matrigel assays, mimicking the tumor-surrounding ECM. Importantly, both the Itgα6 and Itgβ1 antibodies effectively inhibited the invasion of Amdc-s cells, as well as human HT- 1080 fibrosarcoma cells, in this laminin-rich matrix (II: Fig. 5a’-b’). Both the used antibodies thus show potential for the treatment of ITGα6β1-utilizing, invasive tumors. As ITGα6 has many fewer dimerization partners than ITGβ1, ITGα6 specific antibodies (or their derivatives) could show less unwanted side-effects than the more studied ITGβ1 antibodies. As cancer cells are known to develop resistance to various therapies, ITGα6 antibodies could be used in combination with Tβ4 inhibitors and cathepsin L inhibitors [previously shown to be the major protease related to Amdc-s cell invasiveness (Ravanko et al. 2004)], in the treatment of fibrosarcomas/sarcomas overexpressing these molecules. The neutralizing ITGα6 antibody was not found to affect the expression or activity of cathepsin L, but the two molecules seem to function independently (II: Supplementary Fig. 4).

2.2.4 ITGα6 localizes into the invasion fronts of human high-grade fibrosarcomas

To survey the clinical relevance of ITGα6 overexpression in human tumors, we analyzed by IHC a preliminary set (n = 24) of human fibrosarcomas and other soft-tissue sarcomas. Interestingly, ITGα6 showed faint to moderate staining in low-grade tumors, whereas a strong staining was detectable in high-grade tumors, especially in the invasion fronts (II: Fig. 5c’ and Supplementary Fig. 5). This is in line with a previous IHC study on soft tissue sarcomas (not involving fibrosarcomas; Benassi et al. 1998). Interestingly, this broader study of Benassi et al. further reported ITGα6 expression levels to show positive correlation with the occurrence of metastases in the sarcoma types studied, and thus to show prognostic potential. Similarly, in invasive breast carcinomas, high ITGα6 expression has been reported to correlate with reduced survival (Friedrichs et al. 1995). When ITGα6 immunostaining was compared to those of Tβ4 and cathepsin L, somewhat similar patterns were detected in the tumors, especially with ITGα6 and Tβ4 (Fig. 7). The idea of combinatory therapy might thus be rational. When compared to c-Jun immunostaining, ITGα6 immunoreactivity was found to correlate best with the activated, phosphorylated form of c-Jun (II: Supplementary Fig. 6).

Figure 7. Immunostaining of ITGα6 (a; published in II: Fig. 5c’ c), Tβ4 (b), and cathepsin L (c) in a high-grade (G4) fibrosarcoma. Note that cathepsin L, a secreted molecule, shows a somewhat different pattern of staining. Magnification 400x.

2.3 Tenascin-C and fibronectin (III)

TN-C (originally hexabrachion) and FN are large disulfide-linked ECM glycoproteins functioning in cell-matrix adhesion (see the section “Extracellular matrix molecules”). Of these, FN is an adhesive protein, whereas TN-C is able to function anti-adhesively for

melanoma cells, and even to affect their binding to FN (Herlyn et al. 1991). In addition to interacting with each other, TN-C and FN are able to interact with other ECM molecules and cell surface integrins. Previously, both increased TN-C and FN expression (Clark et al. 2000; Hoek et al. 2004) and secretion (Herlyn et al. 1991) have been reported in melanoma cells. We were, however, interested to study the localization of these highly up-regulated ECM molecules in melanoma tissue specimens and to clarify their functional role in melanoma cell invasion.

2.3.1 TN-C and FN are overexpressed in WM793 and WM239 melanoma cells and primary melanomas

The highly up-regulated expression of TN-C and FN in the VGP WM793 and metastatic WM239 melanoma cells was first verified using RT-PCR analyses (Fig. 8A), and an increase in secreted protein levels was confirmed by Western blotting (Fig. 8B). Adult skin fibroblasts were also found to express FN, but not TN-C (Fig. 8A). Further, microarray analyses of primary melanoma tissue specimens, compared to benign melanocytic nevi, showed increased expression of TN-C and FN (III: Fig. 1B).

2.3.2 TN-C and FN show intense immunostaining in fibrillar structures surrounding invasive melanoma cells in tumor tissues

To see the distribution and clinical relevance of TN-C and FN overexpression in melanoma, these molecules were immunohistochemically analyzed from benign nevi (n = 32), primary melanomas (n = 101), and clinically detected melanoma metastases (macrometastases; n = 19). In normal skin and benign nevi, TN-C immunostaining was detected as a thin rim at the basement membrane (III: Fig. 2A) and surrounding blood vessels (III: Fig. 2B). Horizontally growing primary melanomas additionally showed very faint TN-C immunostaining in the immediate vicinity of the basal lamina (III: Fig. 2B). However, of the primary melanomas and melanoma metastases studied, strong TN-C staining was detected in 59% and 58% of the specimens, respectively (III: Table 1). Most importantly, in invasive VGP primary nodular melanomas and melanoma metastases, TN-C staining revealed strong, elongated strands and

ring structures surrounding melanoma cell nests and rows, particularly at the sites of invasion (III: Fig. 2C-D, Fig. 3A-E, Supplementary Fig. 1A and E, Supplementary Fig. 2).

Figure 8. Confirmation of FN and TN-C up-regulation in WM793 and WM239 melanoma cells. A) RT- PCR and B) Western blotting of secreted proteins (10 µg/lane). The antibody NCL-FIB detects FN, whereas BC-24 is specific for TN-C. Specificity of the used antibodies (later used in IHC) is shown in C), using Western blotting of purified TN-C, FN, and COL-I proteins (2 µg/lane).

Further, in analysis of several serial sections, these invasion-associated fibrillar networks seemed to form tubular structures around the melanoma cells. Occasionally also red blood

cells were detected inside these tubules (III: Supplementary Fig. 1E). In addition to promoting invasion, these structures might thus aid, purposely or not on purpose, the survival and haematogenous spreading of the tumor cells. Interestingly, FN immunoreactivity was invariably found to co-localize with TN-C staining in these structures (III: Fig. 5A-B), suggesting that the interaction of these two molecules is important in melanoma progression. In benign nevi, FN was only detectable in the basal laminae of the epidermis and blood vessels (III: Fig. 5E).

In addition to TN-C and FN, highly up-regulated expression of the ECM proteins collagen type I, α1 and collagen type I, α2 was detected in microarray analyses of melanoma lymph node metastases compared to normal lymph nodes (III: Fig. 1C). The studied WM793 and WM239 cell lines, in turn, were not found to express high levels of COL-I, which is expected, as COL-I is known to be mainly synthesized by fibroblasts and myofibroblasts. When studied by IHC, immunoreactivity to procollagen-I (PCOL-I), the precursor molecule of COL-I and an indicator of newly synthesized collagen, was found to co-localize with TN-C and FN staining in the fibrillar structures surrounding the invading melanoma cells (III: Fig. 6B-D). PCOL-I staining of these structures was, however, more variable than that of TN-C and FN, and PCOL-I was also detected intracellularly in the neighboring stromal fibroblasts. In benign nevi, PCOL-I immunoreactivity was restricted to the basal laminae (III: Fig. 6A).

2.3.3 TN-C and FN stimulate the migration of melanoma cells in 3D-collagen-I

To clarify the functional role of secreted TN-C and FN in melanoma cell invasion, the effect of purified TN-C and FN proteins on the behavior of primary melanoma cells (PMEL29) was then tested in 3D-COL-I cultures, mimicking skin ECM in vivo. As compared to pure COL-I, both TN-C and FN, especially when combined, stimulated the infiltrative growth of these melanoma cells in COL-I matrix (III: Supplementary Fig. 3), suggesting that they have a functional role in melanoma cell migration.

Taken together, our study on TN-C and FN shows these ECM molecules, secreted by melanoma cells, to form fibrillar/tubular structures, together with newly synthesized COL-I (secreted by nearby fibroblasts), around the invading melanoma cells and to induce the migration/invasion of melanoma cells. By secreting these adhesion-related molecules (FN

acting as an adhesive protein and TN-C as an anti-adhesive one), melanoma cells presumably are able to modify the microenvironment in a direction more permissive for migration and invasion promoting contacts (strong enough, but not too strong). The co-localization of PCOL-I, secreted by stromal fibroblasts, in the invasion-associated structures formed by tumor produced TN-C and FN, further emphasizes the role of tumor microenvironment in cancer spreading. According to our observations, both TN-C and FN, as well as some other, yet unidentified components of the invasion-related structures, could be potential target molecules in cancer therapy. In addition, TN-C further shows prognostic potential, as its staining intensity in primary melanoma tumors was found to correlate with metastasis to sentinel lymph nodes, better than the Breslow score of tumor thickness routinely used for melanoma grading and assessment of metastasis risk.

2.4 Transforming growth factor beta-induced (IV)

TGFBI is a transforming growth factor β (TGF-β)-inducible ECM protein which has been shown to interact with several other ECM components and various cell surface integrins, and thought to function as a linker between these ECM molecules and the integrin receptors (see the section “Extracellular matrix molecules”). Interestingly, overexpression of TGFBI has been reported in several malignant human tumors, including colorectal, lung, pancreatic, and kidney cancers, glioblastoma, neuroblastoma, and aggressive sarcomatoid cholangiocarcinoma (Zhang et al. 1997; Kitahara et al. 2001; Buckhaults et al. 2001; Sasaki et al. 2002; Schneider et al. 2002; Becker et al. 2006; Tso et al. 2006; Ivanov et al. 2008; Ma et al. 2008; Yoo et al. 2009). In functional studies, increased expression of TGFBI has been shown to promote invasion and metastatic spread of colon cancer and hepatoma cells (Tang et al. 2007; Ma et al. 2008). As we found TGFBI expression to show the second highest up- regulation in metastatic WM239 cells, we decided to study the role and mechanism of this interesting, less studied molecule in melanoma progression.

2.4.1 TGFBI is up-regulated in WM793 and WM239 melanoma cells

The overexpression of TGFBI in WM793 and WM239 melanoma cells was confirmed by RT-PCR (IV: Fig. 1A). Further, the metastatic WM239 cells were shown to secrete high

levels of TGFBI protein (IV: Fig. 1B). In addition, both the studied melanoma cells were found to express the TGFBI’s inducers TGF-β1 and TGF-β2 (IV: Fig. 1A), and also showed up-regulated expression of an additional inducer, interleukin 1β (Nam et al. 2006; 5.8- and 10.9-fold changes, respectively, with Affymetrix arrays).

2.4.2 TGFBI is an anti-adhesive protein for melanoma cells and skin fibroblasts

Due to its FAS domains and RGD motif, TGFBI has been thought to function as a cell adhesion molecule. However, TGFBI has been shown to both promote and inhibit cell adhesion, probably depending on cell type and the surface molecules expressed. To test the effect of TGFBI on melanoma cell adhesion, we coated cell culture wells with different known cell adhesive proteins: pFN, cFN, COL-I, LN, VN, as well as recombinant human TGFBI, and counted the percentage of bound and spread cells. In these adhesion assays, TGFBI was found to be anti-adhesive for melanoma cells, and also adult skin fibroblasts, inducing even less cell spreading than the negative control BSA (IV: Fig. 2 and Supplemental Fig. 1). Further, when TGFBI was combined 1:1 with the indicated adhesive proteins, it substantially weakened melanoma and fibroblast cell adhesion to these molecules (except for VN, where only a minor effect was seen) (IV: Fig. 3). To our knowledge, there exist no report on interaction of TGFBI and VN, whereas all the other tested molecules have been reported to bind to TGFBI, which could explain the different result obtained with VN.

2.4.3 TGFBI knockdown stimulates the migration and invasion of melanoma cells in vitro

To study the function of overexpressed TGFBI in melanoma cells, we knocked down its expression in WM239 cells using TGFBI short hairpin RNA (shRNA) expression (IV: Fig. 4A and Supplemental Fig. 2). The resulting WM239 TGFBI knockdown (TGFBI-KD) clones showed increased migration in COL-I-coated Boyden chambers (IV: Supplementary Fig. 3), as well as in 3D-COL-I gels (IV: Fig. 4B). Similarly, when cultured in 3D-Matrigel matrix, the TGFBI-KD cells formed larger invasive colonies than the control cells (IV: Fig. 4C).

2.4.4 TGFBI-KD cells show highly concentrated actin filaments co-localizing with activated ITGβ1 in pseudopodia-like structures

As the cytoskeleton plays an important role in cell adhesion and migration, we immunostained TGFBI-KD cells for α- and β-tubulins, and F-actin. Interestingly, the TGFBI- KD cells were found to show highly concentrated actin filaments in pseudopodia-like structures, not detected in control cells (IV: Fig. 5). Previously, we had seen similar pseudopodia-like structures at the edges of the highly migratory Amdc-s, Amdc-as, and E4 murine fibrosarcoma cells (Amdc-s and E4 cells shown in Fig. 3). In further studies, ITGβ1 and talin immunostainings were found to co-localize in these areas, indicating integrin activation (IV: Fig. 5 e-f and g-h). TGFBI might thus achieve its anti-adhesive function by preventing ITGβ1 activation. By RT-PCR, ITGβ1 and talin mRNA levels were not found to be changed in TGFBI-KD cells, but the cells were found to show markedly reduced expression of the actin monomer sequestering protein Tβ4 (IV: Fig. 6).

2.4.5 TGFBI-KD cells show remarkably impaired tumor growth in nude mice

To study the effect of TGFBI knockdown on melanoma cell tumorigenicity in vivo, we injected TGFBI-KD cells (3 x 106 cells suspended in Matrigel) subcutaneously into nude mice. In this experiment, the TGFBI-KD cells were clearly less tumorigenic than the non- target shRNA expressing control cells (IV: Fig. 7A and Supplemental Fig. 4), implying an important role for TGFBI in the growth of these melanoma cells in the complex environment in vivo. Bearing in mind the multiple interaction partners of TGFBI, the opposite results between the simple and shorter in vitro assays and the complex environment in vivo are conceivable. No such good model system exists that could in vitro fully mimic the versatile microenvironment the tumors encounter in vivo.

Interestingly, the TGFBI-overexpressing control tumors showed intense TGFBI immunostaining at the edges of the infiltrating tumors, invasion fronts (IV: Fig. 7B a and c), forming similar fibrillar structures around the melanoma cells that we found previously with TN-C, FN, and PCOL-I immunostainings of clinical specimens. Indeed, TGFBI was found to show partial co-localization with FN, its interaction partner, in these structures (IV: Fig. 7B a-d). In TGFBI-KD cell-induced tumors, only faint TGFBI staining was detected at some

areas of the tumor center (IV: Fig. 7B e and data not shown) and the tumors were not able to grow. Of special interest, the TGFBI-KD cell-induced tumors showed amorphous, non- fibrillar pattern of FN staining (IV: Fig. 7B f), implicating that TGFBI may be important for the formation of FN fibrils (an integrin-dependent process), which, in turn, may be important for the assembly of the invasion-associated structures.

Our studies on the effects of TGFBI on melanoma cell growth in mice identify this ECM protein, secreted by metastatic melanoma cells, to clearly contribute to the progression of melanoma tumors and to show potential for cancer therapy. Interestingly, periostin (POSTN), an anti-adhesive protein showing a highly similar structure to TGFBI, has been recently found to co-localize with FN and COL-I in the fibrillar networks in melanoma metastases (Soikkeli et al. 2010). As FN, COL-I, TGFBI, and POSTN all are TGF-β-inducible proteins, TGF-β receptors might also be effectively exploited as therapy target to control melanoma progression.

CONCLUDING REMARKS AND FUTURE PERSPECTIVES

In our fibrosarcoma study, we found two new, interesting up-regulated molecules relevant for fibrosarcoma cell invasion, an actin sequestering protein thymosin β4 (Tβ4) and a cell adhesion receptor integrin α6 (ITGα6) that was complexed with integrin β1 (ITGβ1). Based on the presented and previous data, we propose the possibility of using a combination therapy consisting of ITGα6β1 antagonists (affecting tumor cell adhesion), Tβ4 inhibitors (affecting tumor cell migration), and cathepsin L inhibitors (affecting the degradation of ECM/BM proteins) for the treatment of fibrosarcomas and other sarcomas/tumors overexpressing these molecules. This kind of combinatory therapy could be more effective than therapies exploiting only one feature of cancer cells, as the development of resistance should be slower.

In the other part of this work, melanoma study, we found the extracellular matrix proteins tenascin-C (TN-C), fibronectin (FN), and transforming growth factor beta-induced (TGFBI) to be highly expressed and secreted by invasive melanoma cells. As all these molecules function in cell adhesion and showed co-localization in melanoma tissues, we propose that melanoma cells secrete these molecules to modify their microenvironment to become more supportive for their adhesive interactions and, in that way, their migration and spreading. In addition, these molecules may be important for the regulation of growth and apoptosis of melanoma and other cells. Our results on TN-C, FN, and TGFBI, as well as the stromal fibroblast secreted protein collagen-I, thus strengthen the view of the importance of tumor microenvironment on the aggressiveness of human tumors and identify these molecules as potential melanoma biomarkers or therapeutic targets. In the future, the promising potentials of TN-C, especially the large splice variants, as a melanoma prognostic marker and the knockdown or inhibition of TGFBI as a melanoma treatment strategy should be studied further.

In this study, we managed to find several candidate molecules and characterized in more detail six molecules showing most promising for cancer diagnostics or prognostics, or therapeutic targets. The fact that the identified molecules represent both tumor cell-specific and ECM-related molecules (produced also by stromal cells) emphasizes the importance of not only tumor cells but also the tumor microenvironment in tumor dissemination. Most

potent therapeutic approaches are, therefore, likely to target multiple features of this complex process.

ACKNOWLEDGEMENTS

This study was carried out at the Department of Pathology, Haartman Institute, University of Helsinki during the years 1999-2010. I would like to thank the former and present heads of the department, Professors Eero Saksela and Veli-Pekka Lehto, for providing us with excellent research facilities.

Above all, I want to express my deepest gratitude to my supervisor Docent Erkki Hölttä for giving me the opportunity to acquire a strong know-how of laboratory techniques and to mature into a real scientist. I honestly admire your expertise, patience, and unbelievable memory.

Professor Veli-Matti Kähäri and Docent Jouko Lohi, the official reviewers of my thesis, are gratefully acknowledged for their constructive criticism.

My thesis committee members Pirjo Nikula-Ijäs and Jouko Lohi are profoundly thanked for helpful comments and encouragement during the way. Of my collaborators, Tiina Jahkola is appreciated for providing us with melanoma tissue specimens, Olli Saksela for his dermatological expertise and contribution to financial support, Susanna Virolainen for pathological expertise (as well as kindly letting us use her superior microscope), and Pirjo Laakkonen and Johanna Lammi for enabling the in vivo tumorigenicity assays.

I owe my warmest thanks to Miao, Kirsi, Aino, Riikka, and Essi, the former and present members of our research group, for creating such a nice team. Merja, Lennu, Taina, and Ulla are specially thanked for their technical expertise. Professor Leif C. Andersson and his lab are thanked for “the past golden years” of co-work. I sincerely admire Leif’s enormous knowledge of pathology and life in general. The third floor coffee room gang is thanked for a warm and relaxing atmosphere.

My lab members and friends, Mari, Johku, and Krisse, deserve huge hugs for sharing scientific, as well as mothers’ worries (whether talking about babies or puppies).

My parents-in-law, Mirja and Aarre, are warmly appreciated for support and especially taking care of sick children. My parents, Ulpu and Pentti, are dearly thanked for their continuing love and support. As my father wasn’t allowed to use his talent for studying, I dedicate part of this work to him.

I owe unlimited thanks to my dear husband Juha for understanding and patience. Now you have got your own company and I have got my thesis, it’s time to just enjoy life with our two lovely sons! Joona and Joel, the absolutely most important parts of my life! Thank you for magically turning my thoughts away from the scientific matters. You are the most wonderful wizards outside Hogwarts!

This study was supported by the Helsinki Biomedical Graduate School, the Maud Kuistila Memorial Foundation, the Ida Montin Foundation, the Finnish Cancer Organizations, the Paulo Foundation, the University of Helsinki, the Biomedicum Helsinki Foundation, the Academy of Finland, and the Helsinki University Central Hospital Research Funds.

Helsinki, April 2011

REFERENCES

Airola K, Karonen T, Vaalamo M, Lehti K, Lohi J, Kariniemi AL, Keski-Oja J, and Saarialho-Kere UK (1999). Expression of collagenases-1 and -3 and their inhibitors TIMP-1 and -3 correlates with the level of invasion in malignant melanomas. Br J Cancer 80:733-43.

Ala-aho R and Kähäri VM (2005). Collagenases in cancer. Biochimie 87:273-86.

Alberts R, Terpstra P, Hardonk M, Bystrykh LV, de Haan G, Breitling R, Nap JP, and Jansen RC (2007). A verification protocol for the probe sequences of Affymetrix genome arrays reveals high probe accuracy for studies in mouse, human and rat. BMC Bioinformatics 8:132.

Alitalo K, Tammela T, and Petrova TV (2005). Lymphangiogenesis in development and human disease. Nature 438:946-53.

Angst BD, Marcozzi C, and Magee AI (2001). The cadherin superfamily: diversity in form and function. J Cell Sci 114:629-41.

Auvinen M, Paasinen A, Andersson LC, and Hölttä E (1992). Ornithine decarboxylase activity is critical for cell transformation. Nature 360:355-8.

Bae JS, Lee SH, Kim JE, Choi JY, Park RW, Yong Park J, Park HS, Sohn YS, Lee DS, Bae Lee E, and Kim IS (2002). Betaig-h3 supports keratinocyte adhesion, migration, and proliferation through alpha3beta1 integrin. Biochem Biophys Res Commun 294:940-8.

Becker J, Erdlenbruch B, Noskova I, Schramm A, Aumailley M, Schorderet DF, and Schweigerer L (2006). Keratoepithelin suppresses the progression of experimental human neuroblastomas. Cancer Res 66:5314-21.

Behrens J (1999). Cadherins and catenins: role in signal transduction and tumor progression. Cancer Metastasis Rev 18:15-30.

Benassi MS, Ragazzini P, Gamberi G, Sollazzo MR, Molendini L, Ferrari C, Merli M, Böhling T, and Picci P (1998). Adhesion molecules in high-grade soft tissue sarcomas: correlation to clinical outcome. Eur J Cancer 34:496-502.

Benton G, Kleinman HK, George J, and Arnaoutova I (2011). Multiple uses of basement membrane- like matrix (BME/Matrigel) in vitro and in vivo with cancer cells. Int J Cancer 128:1751-7.

Berking C, Takemoto R, Schaider H, Showe L, Satyamoorthy K, Robbins P, and Herlyn M (2001). Transforming growth factor-beta1 increases survival of human melanoma through stroma remodeling. Cancer Res 61:8306-16.

Billings PC, Whitbeck JC, Adams CS, Abrams WR, Cohen AJ, Engelsberg BN, Howard PS, and Rosenbloom J (2002). The transforming growth factor-beta-inducible matrix protein (beta)ig-h3 interacts with fibronectin. J Biol Chem 277:28003-9.

Bosman FT and Stamenkovic I (2003). Functional structure and composition of the extracellular matrix. J Pathol 200:423-8.

Bourboulia D and Stetler-Stevenson WG (2010). Matrix metalloproteinases (MMPs) and tissue inhibitors of metalloproteinases (TIMPs): positive and negative regulators in tumor cell adhesion. Semin Cancer Biol 20:161-8.

Brooks SA, Lomax-Browne HJ, Carter TM, Kinch CE, and Hall DM (2010). Molecular interactions in cancer cell metastasis. Acta Histochem 112:3-25.

Buckhaults P, Rago C, St Croix B, Romans KE, Saha S, Zhang L, Vogelstein B, and Kinzler KW (2001). Secreted and cell surface genes expressed in benign and malignant colorectal tumors. Cancer Res 61:6996-7001.

Cha HJ, Jeong MJ, and Kleinman HK (2003). Role of thymosin beta4 in tumor metastasis and angiogenesis. J Natl Cancer Inst 95:1674-80.

Chiquet-Ehrismann R, Kalla P, Pearson CA, Beck K, and Chiquet M (1988). Tenascin interferes with fibronectin action. Cell 53:383-90.

Chiquet-Ehrismann R and Tucker RP (2004). Connective tissues: signalling by tenascins. Int J Biochem Cell Biol 36:1085-9.

Chung CY, Zardi L, and Erickson HP (1995). Binding of tenascin-C to soluble fibronectin and matrix fibrils. J Biol Chem 270:29012-7.

Clark EA, Golub TR, Lander ES, and Hynes RO (2000). Genomic analysis of metastasis reveals an essential role for RhoC. Nature 406:532-5.

Coue M, Brenner SL, Spector I, and Korn ED (1987). Inhibition of actin polymerization by latrunculin A. FEBS Lett 213:316-8.

Coussens LM and Werb Z (2002). Inflammation and cancer. Nature 420:860-7.

Dass K, Ahmad A, Azmi AS, Sarkar SH, and Sarkar FH (2008). Evolving role of uPA/uPAR system in human cancers. Cancer Treat Rev 34:122-36.

Degenhardt Y, Huang J, Greshock J, Horiates G, Nathanson K, Yang X, Herlyn M, and Weber B (2010). Distinct MHC gene expression patterns during progression of melanoma. Genes Chromosomes Cancer 49:144-54.

Desgrosellier JS and Cheresh DA (2010). Integrins in cancer: biological implications and therapeutic opportunities. Nat Rev Cancer 10:9-22.

Di Lullo GA, Sweeney SM, Körkkö J, Ala-Kokko L, and San Antonio JD (2002). Mapping the ligand-binding sites and disease-associated mutations on the most abundant protein in the human, type I collagen. J Biol Chem 277:4223-31.

Duncan LM, Deeds J, Hunter J, Shao J, Holmgren LM, Woolf EA, Tepper RI, and Shyjan AW (1998). Down-regulation of the novel gene melastatin correlates with potential for melanoma metastasis. Cancer Res 58:1515-20.

Durbeej M (2010). Laminins. Cell Tissue Res 339:259-68.

Dvorak HF (1986). Tumors: wounds that do not heal. Similarities between tumor stroma generation and wound healing. N Engl J Med 315:1650-9.

Ferlay J, Bray F, Pisani P, and Parkin DM (2004). GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide, IARC CancerBase No. 5., version 2.0., IARC Press, Lyon.

Finger EC and Giaccia AJ (2010). Hypoxia, inflammation, and the tumor microenvironment in metastatic disease. Cancer Metastasis Rev 29:285-93.

Finnish Cancer Registry (2009). Cancer in Finland 2006 and 2007. Cancer Society of Finland Publication No. 76, Helsinki.

Friedrichs K, Ruiz P, Franke F, Gille I, Terpe HJ, and Imhof BA (1995). High expression level of alpha 6 integrin in human breast carcinoma is correlated with reduced survival. Cancer Res 55:901-6.

Fukunaga-Kalabis M, Santiago-Walker A, and Herlyn M (2008). Matricellular proteins produced by melanocytes and melanomas: in search for functions. Cancer Microenviron 1:93-102.

Garbe C, Eigentler TK, Keilholz U, Hauschild A, and Kirkwood JM (2011). Systematic review of medical treatment in melanoma: current status and future prospects. Oncologist 16:5-24.

Giancotti FG and Ruoslahti E (1999). Integrin signaling. Science 285:1028-32.

Gocheva V and Joyce JA (2007). Cysteine cathepsins and the cutting edge of cancer invasion. Cell Cycle 6:60-4.

Gordon MK and Hahn RA (2010). Collagens. Cell Tissue Res 339:247-57.

Gupta GP and Massagué J (2006). Cancer metastasis: building a framework. Cell 127:679-95.

Gutiérrez-Fernández A, Fueyo A, Folgueras AR, Garabaya C, Pennington CJ, Pilgrim S, Edwards DR, Holliday DL, Jones JL, Span PN, Sweep FC, Puente XS, and López-Otín C (2008). Matrix metalloproteinase-8 functions as a metastasis suppressor through modulation of tumor cell adhesion and invasion. Cancer Res 68:2755-63.

Hall AK (1991). Differential expression of thymosin genes in human tumors and in the developing human kidney. Int J Cancer 48:672-7.

Hanahan D and Weinberg RA (2000). The hallmarks of cancer. Cell 100:57-70.

Hashimoto K, Noshiro M, Ohno S, Kawamoto T, Satakeda H, Akagawa Y, Nakashima K, Okimura A, Ishida H, Okamoto T, Pan H, Shen M, Yan W, and Kato Y (1997). Characterization of a cartilage- derived 66-kDa protein (RGD-CAP/beta ig-h3) that binds to collagen. Biochim Biophys Acta 1355:303-14.

Heino J (2007). The collagen family members as cell adhesion proteins. Bioessays 29:1001-10.

Heino J and Käpylä J (2009). Cellular receptors of extracellular matrix molecules. Curr Pharm Des 15:1309-17.

Herlyn M, Graeven U, Speicher D, Sela BA, Bennicelli JL, Kath R, and Guerry D,4th (1991). Characterization of tenascin secreted by human melanoma cells. Cancer Res 51:4853-8.

Hoek K, Rimm DL, Williams KR, Zhao H, Ariyan S, Lin A, Kluger HM, Berger AJ, Cheng E, Trombetta ES, Wu T, Niinobe M, Yoshikawa K, Hannigan GE, and Halaban R (2004). Expression

profiling reveals novel pathways in the transformation of melanocytes to melanomas. Cancer Res 64:5270-82.

Hoek KS (2007). DNA microarray analyses of melanoma gene expression: a decade in the mines. Pigment Cell Res 20:466-84.

Hofmann UB, Houben R, Bröcker EB, and Becker JC (2005). Role of matrix metalloproteinases in melanoma cell invasion. Biochimie 87:307-14.

Huang Y, Prasad M, Lemon WJ, Hampel H, Wright FA, Kornacker K, LiVolsi V, Frankel W, Kloos RT, Eng C, Pellegata NS, and de la Chapelle A (2001). Gene expression in papillary thyroid carcinoma reveals highly consistent profiles. Proc Natl Acad Sci U S A 98:15044-9.

Hurt EM, Wiestner A, Rosenwald A, Shaffer AL, Campo E, Grogan T, Bergsagel PL, Kuehl WM, and Staudt LM (2004). Overexpression of c-maf is a frequent oncogenic event in multiple myeloma that promotes proliferation and pathological interactions with bone marrow stroma. Cancer Cell 5:191-9.

Hynes RO (2009). The extracellular matrix: not just pretty fibrils. Science 326:1216-9.

Ivanov SV, Ivanova AV, Salnikow K, Timofeeva O, Subramaniam M, and Lerman MI (2008). Two novel VHL targets, TGFBI (BIGH3) and its transactivator KLF10, are up-regulated in renal clear cell carcinoma and other tumors. Biochem Biophys Res Commun 370:536-40.

Ivaska J and Heino J (2000). Adhesion receptors and cell invasion: mechanisms of integrin-guided degradation of extracellular matrix. Cell Mol Life Sci 57:16-24.

Ji P, Diederichs S, Wang W, Boing S, Metzger R, Schneider PM, Tidow N, Brandt B, Buerger H, Bulk E, Thomas M, Berdel WE, Serve H, and Muller-Tidow C (2003). MALAT-1, a novel noncoding RNA, and thymosin beta4 predict metastasis and survival in early-stage non-small cell lung cancer. Oncogene 22:8031-41.

Jones FS and Jones PL (2000). The tenascin family of ECM glycoproteins: structure, function, and regulation during embryonic development and tissue remodeling. Dev Dyn 218:235-59.

Kadler KE, Baldock C, Bella J, and Boot-Handford RP (2007). Collagens at a glance. J Cell Sci 120:1955-8.

Kadler KE, Hill A, and Canty-Laird EG (2008). Collagen fibrillogenesis: fibronectin, integrins, and minor collagens as organizers and nucleators. Curr Opin Cell Biol 20:495-501.

Kalluri R and Zeisberg M (2006). Fibroblasts in cancer. Nat Rev Cancer 6:392-401.

Kaspar M, Zardi L, and Neri D (2006). Fibronectin as target for tumor therapy. Int J Cancer 118:1331-9.

Katenkamp K, Berndt A, Hindermann W, Wunderlich H, Haas KM, Borsi L, Zardi L, and Kosmehl H (2004). mRNA expression and protein distribution of the unspliced tenascin-C isoform in prostatic adenocarcinoma. J Pathol 203:771-9.

Kessenbrock K, Plaks V, and Werb Z (2010). Matrix metalloproteinases: regulators of the tumor microenvironment. Cell 141:52-67.

Khanfar MA, Youssef DT, and El Sayed KA (2010). Semisynthetic latrunculin derivatives as inhibitors of metastatic breast cancer: biological evaluations, preliminary structure-activity relationship and molecular modeling studies. ChemMedChem 5:274-85.

Kim BY, Olzmann JA, Choi SI, Ahn SY, Kim TI, Cho HS, Suh H, and Kim EK (2009). Corneal dystrophy-associated R124H mutation disrupts TGFBI interaction with periostin and causes mislocalization to the lysosome. J Biol Chem 284:19580-91.

Kim JE, Kim SJ, Lee BH, Park RW, Kim KS, and Kim IS (2000). Identification of motifs for cell adhesion within the repeated domains of transforming growth factor-beta-induced gene, betaig-h3. J Biol Chem 275:30907-15.

Kim JE, Park RW, Choi JY, Bae YC, Kim KS, Joo CK, and Kim IS (2002). Molecular properties of wild-type and mutant betaIG-H3 proteins. Invest Ophthalmol Vis Sci 43:656-61.

Kitahara O, Furukawa Y, Tanaka T, Kihara C, Ono K, Yanagawa R, Nita ME, Takagi T, Nakamura Y, and Tsunoda T (2001). Alterations of gene expression during colorectal carcinogenesis revealed by cDNA microarrays after laser-capture microdissection of tumor tissues and normal epithelia. Cancer Res 61:3544-9.

Klein T and Bischoff R (2011). Active metalloproteases of the a disintegrin and metalloprotease (ADAM) family: biological function and structure. J Proteome Res 10:17-33.

Kobayashi T, Okada F, Fujii N, Tomita N, Ito S, Tazawa H, Aoyama T, Choi SK, Shibata T, Fujita H, and Hosokawa M (2002). Thymosin-beta4 regulates motility and metastasis of malignant mouse fibrosarcoma cells. Am J Pathol 160:869-82.

Konishi H, Kikuchi S, Ochiai T, Ikoma H, Kubota T, Ichikawa D, Fujiwara H, Okamoto K, Sakakura C, Sonoyama T, Kokuba Y, Sasaki H, Matsui T, and Otsuji E (2009). Latrunculin a has a strong anticancer effect in a peritoneal dissemination model of human gastric cancer in mice. Anticancer Res 29:2091-7.

Kumar V, Abbas AK, Fausto N, and Mitchell RN (2007). Robbins basic pathology, 8th edition. Saunders, Elsevier Inc., Philadelphia.

Latijnhouwers MA, de Jongh GJ, Bergers M, de Rooij MJ, and Schalkwijk J (2000). Expression of tenascin-C splice variants by human skin cells. Arch Dermatol Res 292:446-54.

Lazebnik Y (2010). What are the hallmarks of cancer? Nat Rev Cancer 10:232-3.

LeBaron RG, Bezverkov KI, Zimber MP, Pavelec R, Skonier J, and Purchio AF (1995). Beta IG-H3, a novel secretory protein inducible by transforming growth factor-beta, is present in normal skin and promotes the adhesion and spreading of dermal fibroblasts in vitro. J Invest Dermatol 104:844-9.

Lee JK, Bussey KJ, Gwadry FG, Reinhold W, Riddick G, Pelletier SL, Nishizuka S, Szakacs G, Annereau JP, Shankavaram U, Lababidi S, Smith LH, Gottesman MM, and Weinstein JN (2003). Comparing cDNA and oligonucleotide array data: concordance of gene expression across platforms for the NCI-60 cancer cells. Genome Biol 4:R82.

Lee JT and Herlyn M (2007). Microenvironmental influences in melanoma progression. J Cell Biochem 101:862-72.

Lin EY, Nguyen AV, Russell RG, and Pollard JW (2001). Colony-stimulating factor 1 promotes progression of mammary tumors to malignancy. J Exp Med 193:727-40.

Lorusso G and Rüegg C (2008). The tumor microenvironment and its contribution to tumor evolution toward metastasis. Histochem Cell Biol 130:1091-103.

Ma C, Rong Y, Radiloff DR, Datto MB, Centeno B, Bao S, Cheng AW, Lin F, Jiang S, Yeatman TJ, and Wang XF (2008). Extracellular matrix protein betaig-h3/TGFBI promotes metastasis of colon cancer by enhancing cell extravasation. Genes Dev 22:308-21.

Mannherz HG and Hannappel E (2009). The beta-thymosins: intracellular and extracellular activities of a versatile actin binding protein family. Cell Motil Cytoskeleton 66:839-51.

Melnikova VO and Bar-Eli M (2009). Inflammation and melanoma metastasis. Pigment Cell Melanoma Res 22:257-67.

Midwood KS and Orend G (2009). The role of tenascin-C in tissue injury and tumorigenesis. J Cell Commun Signal 3:287-310.

Miner JH (2008). Laminins and their roles in mammals. Microsc Res Tech 71:349-56.

Mithieux SM and Weiss AS (2005). Elastin. Adv Protein Chem 70:437-61.

Mohamed MM and Sloane BF (2006). Cysteine cathepsins: multifunctional enzymes in cancer. Nat Rev Cancer 6:764-75.

Morito N, Yoh K, Fujioka Y, Nakano T, Shimohata H, Hashimoto Y, Yamada A, Maeda A, Matsuno F, Hata H, Suzuki A, Imagawa S, Mitsuya H, Esumi H, Koyama A, Yamamoto M, Mori N, and Takahashi S (2006). Overexpression of c-Maf contributes to T-cell lymphoma in both mice and human. Cancer Res 66:812-9.

Morton WM, Ayscough KR, and McLaughlin PJ (2000). Latrunculin alters the actin-monomer subunit interface to prevent polymerization. Nat Cell Biol 2:376-8.

Moshier JA, Dosescu J, Skunca M, and Luk GD (1993). Transformation of NIH/3T3 cells by ornithine decarboxylase overexpression. Cancer Res 53:2618-22.

Muiznieks LD, Weiss AS, and Keeley FW (2010). Structural disorder and dynamics of elastin. Biochem Cell Biol 88:239-50.

Nam EJ, Sa KH, You DW, Cho JH, Seo JS, Han SW, Park JY, Kim SI, Kyung HS, Kim IS, and Kang YM (2006). Up-regulated transforming growth factor beta-inducible gene h3 in rheumatoid arthritis mediates adhesion and migration of synoviocytes through alphavbeta3 integrin: regulation by cytokines. Arthritis Rheum 54:2734-44.

Nam JO, Kim JE, Jeong HW, Lee SJ, Lee BH, Choi JY, Park RW, Park JY, and Kim IS (2003). Identification of the alphavbeta3 integrin-interacting motif of betaig-h3 and its anti-angiogenic effect. J Biol Chem 278:25902-9.

Nguyen DX, Bos PD, and Massagué J (2009). Metastasis: from dissemination to organ-specific colonization. Nat Rev Cancer 9:274-84.

Nikkola J, Vihinen P, Vlaykova T, Hahka-Kemppinen M, Kähäri VM, and Pyrhönen S (2002). High expression levels of collagenase-1 and stromelysin-1 correlate with shorter disease-free survival in human metastatic melanoma. Int J Cancer 97:432-8.

Ohno S, Noshiro M, Makihira S, Kawamoto T, Shen M, Yan W, Kawashima-Ohya Y, Fujimoto K, Tanne K, and Kato Y (1999). RGD-CAP ((beta)ig-h3) enhances the spreading of chondrocytes and fibroblasts via integrin alpha(1)beta(1). Biochim Biophys Acta 1451:196-205.

Olson MF and Sahai E (2009). The actin cytoskeleton in cancer cell motility. Clin Exp Metastasis 26:273-87.

Paasinen-Sohns A, Kielosto M, Kääriäinen E, Eloranta T, Laine A, Jänne OA, Birrer MJ, and Hölttä E (2000). c-Jun activation-dependent tumorigenic transformation induced paradoxically by overexpression or block of S-adenosylmethionine decarboxylase. J Cell Biol 151:801-10.

Paasinen-Sohns A, Kääriäinen E, Yin M, Järvinen K, Nummela P, and Hölttä E (2011). Chaotic neovascularization induced by aggressive fibrosarcoma cells overexpressing S-adenosylmethionine decarboxylase. Int J Biochem Cell Biol 43:441-54.

Palavalli LH, Prickett TD, Wunderlich JR, Wei X, Burrell AS, Porter-Gill P, Davis S, Wang C, Cronin JC, Agrawal NS, Lin JC, Westbroek W, Hoogstraten-Miller S, Molinolo AA, Fetsch P, Filie AC, O'Connell MP, Banister CE, Howard JD, Buckhaults P, Weeraratna AT, Brody LC, Rosenberg SA, and Samuels Y (2009). Analysis of the matrix metalloproteinase family reveals that MMP8 is often mutated in melanoma. Nat Genet 41:518-20.

Pankov R and Yamada KM (2002). Fibronectin at a glance. J Cell Sci 115:3861-3.

Patarroyo M, Tryggvason K, and Virtanen I (2002). Laminin isoforms in tumor invasion, angiogenesis and metastasis. Semin Cancer Biol 12:197-207.

Pegg AE (1988). Polyamine metabolism and its importance in neoplastic growth and a target for chemotherapy. Cancer Res 48:759-74.

Pegg AE (2009). Mammalian polyamine metabolism and function. IUBMB Life 61:880-94.

Pollard TD and Borisy GG (2003). Cellular motility driven by assembly and disassembly of actin filaments. Cell 112:453-65.

Qian BZ and Pollard JW (2010). Macrophage diversity enhances tumor progression and metastasis. Cell 141:39-51.

Rathinam R and Alahari SK (2010). Important role of integrins in the cancer biology. Cancer Metastasis Rev 29:223-37.

Ravanko K, Järvinen K, Helin J, Kalkkinen N, and Hölttä E (2004). Cysteine cathepsins are central contributors of invasion by cultured adenosylmethionine decarboxylase-transformed rodent fibroblasts. Cancer Res 64:8831-8.

Reinboth B, Thomas J, Hanssen E, and Gibson MA (2006). Beta ig-h3 interacts directly with biglycan and decorin, promotes collagen VI aggregation, and participates in ternary complexing with these macromolecules. J Biol Chem 281:7816-24.

Rickelt S, Franke WW, Doerflinger Y, Goerdt S, Brandner JM, and Peitsch WK (2008). Subtypes of melanocytes and melanoma cells distinguished by their intercellular contacts: heterotypic adherens junctions, adhesive associations, and dispersed desmoglein 2 glycoproteins. Cell Tissue Res 334:401- 22.

Rocks N, Paulissen G, El Hour M, Quesada F, Crahay C, Gueders M, Foidart JM, Noel A, and Cataldo D (2008). Emerging roles of ADAM and ADAMTS metalloproteinases in cancer. Biochimie 90:369-79.

Rodgers UR and Weiss AS (2005). Cellular interactions with elastin. Pathol Biol (Paris) 53:390-8.

Räsänen K and Vaheri A (2010). Activation of fibroblasts in cancer stroma. Exp Cell Res 316:2713- 22

Sager R (1997). Expression genetics in cancer: shifting the focus from DNA to RNA. Proc Natl Acad Sci U S A 94:952-5.

Santelli G, Califano D, Chiappetta G, Vento MT, Bartoli PC, Zullo F, Trapasso F, Viglietto G, and Fusco A (1999). Thymosin beta-10 gene overexpression is a general event in human carcinogenesis. Am J Pathol 155:799-804.

Sasaki H, Kobayashi Y, Nakashima Y, Moriyama S, Yukiue H, Kaji M, Kiriyama M, Fukai I, Yamakawa Y, and Fujii Y (2002). Beta IGH3, a TGF-beta inducible gene, is overexpressed in lung cancer. Jpn J Clin Oncol 32:85-9.

Satyamoorthy K and Herlyn M (2001). Melanoma. In: Encyclopedia of Life Sciences (ELS). John Wiley & Sons, Ltd, Chichester.

Satyamoorthy K and Herlyn M (2002). Cellular and molecular biology of human melanoma. Cancer Biol Ther 1:14-7.

Sayed KA, Khanfar MA, Shallal HM, Muralidharan A, Awate B, Youssef DT, Liu Y, Zhou YD, Nagle DG, and Shah G (2008). Latrunculin A and its C-17-O-carbamates inhibit prostate tumor cell invasion and HIF-1 activation in breast tumor cells. J Nat Prod 71:396-402.

Schneider D, Kleeff J, Berberat PO, Zhu Z, Korc M, Friess H, and Büchler MW (2002). Induction and expression of betaig-h3 in pancreatic cancer cells. Biochim Biophys Acta 1588:1-6.

Shantz LM and Pegg AE (1994). Overproduction of ornithine decarboxylase caused by relief of translational repression is associated with neoplastic transformation. Cancer Res 54:2313-6.

Shelton L, Troilo D, Lerner MR, Gusev Y, Brackett DJ, and Summers Rada J (2008). Microarray analysis of choroid/RPE gene expression in marmoset eyes undergoing changes in ocular growth and refraction. Mol Vis 14:1465-79.

Shelton L and Summers-Rada J (2009). TGFBIp inhibits the attachment of human scleral fibroblasts to collagen type I. Invest Ophthalmol Vis Sci 50:3542-52.

Shinohara M, Nakamura S, Sasaki M, Kurahara S, Ikebe T, Harada T, and Shirasuna K (1999). Expression of integrins in squamous cell carcinoma of the oral cavity. Correlations with tumor invasion and metastasis. Am J Clin Pathol 111:75-88.

Singh P, Carraher C, and Schwarzbauer JE (2010). Assembly of fibronectin extracellular matrix. Annu Rev Cell Dev Biol 26:397-419.

Skonier J, Neubauer M, Madisen L, Bennett K, Plowman GD, and Purchio AF (1992). cDNA cloning and sequence analysis of beta ig-h3, a novel gene induced in a human adenocarcinoma cell line after treatment with transforming growth factor-beta. DNA Cell Biol 11:511-22.

Skonier J, Bennett K, Rothwell V, Kosowski S, Plowman G, Wallace P, Edelhoff S, Disteche C, Neubauer M, and Marquardt H (1994). Beta ig-h3: a transforming growth factor-beta-responsive gene encoding a secreted protein that inhibits cell attachment in vitro and suppresses the growth of CHO cells in nude mice. DNA Cell Biol 13:571-84.

Soikkeli J, Lukk M, Nummela P, Virolainen S, Jahkola T, Katainen R, Harju L, Ukkonen E, Saksela O, and Hölttä E (2007). Systematic search for the best gene expression markers for melanoma micrometastasis detection. J Pathol 213:180-9.

Soikkeli J, Podlasz P, Yin M, Nummela P, Jahkola T, Virolainen S, Krogerus L, Heikkilä P, von Smitten K, Saksela O, and Hölttä E (2010). Metastatic outgrowth encompasses COL-I, FN1, and POSTN up-regulation and assembly to fibrillar networks regulating cell adhesion, migration, and growth. Am J Pathol 177:387-403.

Spector I, Shochet NR, Kashman Y, and Groweiss A (1983). Latrunculins: novel marine toxins that disrupt microfilament organization in cultured cells. Science 219:493-5.

Sternlicht MD, Lochter A, Sympson CJ, Huey B, Rougier JP, Gray JW, Pinkel D, Bissell MJ, and Werb Z (1999). The stromal proteinase MMP3/stromelysin-1 promotes mammary carcinogenesis. Cell 98:137-46.

Tang J, Zhou HW, Jiang JL, Yang XM, Li Y, Zhang HX, Chen ZN, and Guo WP (2007). BetaIg-h3 is involved in the HAb18G/CD147-mediated metastasis process in human hepatoma cells. Exp Biol Med (Maywood) 232:344-52.

Tavian D, De Petro G, Colombi M, Portolani N, Giulini SM, Gardella R, and Barlati S (1994). RT- PCR detection of fibronectin EDA+ and EDB+ mRNA isoforms: molecular markers for hepatocellular carcinoma. Int J Cancer 56:820-5.

Thapa N, Lee BH, and Kim IS (2007). TGFBIp/betaig-h3 protein: a versatile matrix molecule induced by TGF-beta. Int J Biochem Cell Biol 39:2183-94.

Todorovicc V, Chen CC, Hay N, and Lau LF (2005). The matrix protein CCN1 (CYR61) induces apoptosis in fibroblasts. J Cell Biol 171:559-68.

Toriseva M and Kähäri VM (2009). Proteinases in cutaneous wound healing. Cell Mol Life Sci 66:203-24.

Tso CL, Shintaku P, Chen J, Liu Q, Liu J, Chen Z, Yoshimoto K, Mischel PS, Cloughesy TF, Liau LM, and Nelson SF (2006). Primary glioblastomas express mesenchymal stem-like properties. Mol Cancer Res 4:607-19.

Vale RD and Milligan RA (2000). The way things move: looking under the hood of molecular motor proteins. Science 288:88-95.

Van den Oord JJ (1998). Expression of CD26/dipeptidyl-peptidase IV in benign and malignant pigment-cell lesions of the skin. Br J Dermatol 138:615-21.

Vasiljeva O and Turk B (2008). Dual contrasting roles of cysteine cathepsins in cancer progression: apoptosis versus tumour invasion. Biochimie 90:380-6.

Vihinen P, Koskivuo I, Syrjänen K, Tervahartiala T, Sorsa T, and Pyrhönen S (2008). Serum matrix metalloproteinase-8 is associated with ulceration and vascular invasion of malignant melanoma. Melanoma Res 18:268-73.

Villanueva J and Herlyn M (2009). Melanoma. In: Encyclopedia of Life Sciences (ELS). John Wiley & Sons, Ltd, Chichester.

Wang WS, Chen PM, Hsiao HL, Ju SY, and Su Y (2003). Overexpression of the thymosin beta-4 gene is associated with malignant progression of SW480 colon cancer cells. Oncogene 22:3297-306.

Wang WS, Chen PM, Hsiao HL, Wang HS, Liang WY, and Su Y (2004). Overexpression of the thymosin beta-4 gene is associated with increased invasion of SW480 colon carcinoma cells and the distant metastasis of human colorectal carcinoma. Oncogene 23:6666-71.

Watsuji T, Okamoto Y, Emi N, Katsuoka Y, and Hagiwara M (1997). Controlled gene expression with a reverse tetracycline-regulated retroviral vector (RTRV) system. Biochem Biophys Res Commun 234:769-73.

Weinberg RA (2007). The biology of cancer, 1st edition. Garland Science, New York.

Wesley UV, Albino AP, Tiwari S, and Houghton AN (1999). A role for dipeptidyl peptidase IV in suppressing the malignant phenotype of melanocytic cells. J Exp Med 190:311-22.

Wolf K and Friedl P (2006). Molecular mechanisms of cancer cell invasion and plasticity. Br J Dermatol 154 Suppl 1:11-5.

Xie D, Jauch A, Miller CW, Bartram CR, and Koeffler HP (2002). Discovery of over-expressed genes and genetic alterations in breast cancer cells using a combination of suppression subtractive hybridization, multiplex FISH and comparative genomic hybridization. Int J Oncol 21:499-507.

Yamamoto T, Gotoh M, Kitajima M, and Hirohashi S (1993). Thymosin beta-4 expression is correlated with metastatic capacity of colorectal carcinomas. Biochem Biophys Res Commun 193:706- 10.

Yarmola EG, Somasundaram T, Boring TA, Spector I, and Bubb MR (2000). Actin-latrunculin A structure and function. Differential modulation of actin-binding protein function by latrunculin A. J Biol Chem 275:28120-7.

Yoo HJ, Yun BR, Kwon JH, Ahn HS, Seol MA, Lee MJ, Yu GR, Yu HC, Hong B, Choi K, and Kim DG (2009). Genetic and expression alterations in association with the sarcomatous change of cholangiocarcinoma cells. Exp Mol Med 41:102-15.

Zabierowski SE and Herlyn M (2008). Melanoma stem cells: the dark seed of melanoma. J Clin Oncol 26:2890-4.

Zajchowski DA, Bartholdi MF, Gong Y, Webster L, Liu HL, Munishkin A, Beauheim C, Harvey S, Ethier SP, and Johnson PH (2001). Identification of gene expression profiles that predict the aggressive behavior of breast cancer cells. Cancer Res 61:5168-78.

Zhang L, Zhou W, Velculescu VE, Kern SE, Hruban RH, Hamilton SR, Vogelstein B, and Kinzler KW (1997). Gene expression profiles in normal and cancer cells. Science 276:1268-72.