While normal melanocytes are growth-inhibited by TGF-β, melanoma cells exhibit a variable degree of resistance to its antiproliferative activity. The mechanisms underlying evasion from a cytostatic response to TGF-β in these cells remain somewhat elusive. Alterations in the components of the TGF-β signaling in melanoma, including loss of expression or mutation of TβR and SMADs, are very rare12 and therefore do not explain the mechanisms of resistance of melanoma cells to TGF-β-mediated growth arrest and apoptosis.

Various mechanisms have been suggested to explain a possible attenuation or inhibition of the TGF-β signaling cascade in melanoma cells. For example, the type III TGF-β receptor CD105 (also called endoglin) was reported to bind TGF-β1 in melanoma cells and to contribute to the regulation of the anti-proliferative effects of TGF-β30,31. In melanoma cells deficient in filamin, a cytoskeletal actinbinding protein, TGF-β/SMAD signaling is defective, suggesting that filamin, which was identified as a SMAD-interacting protein, may serve as an anchor protein that allows localization of SMAD proteins near cell surface receptors, thus facilitating their phosphorylation and subsequent activation32. More recently, distal-less homeobox (DLX2), identified as a novel TGF-β-inducible transcription factor in B16 melanoma cells, was shown to exert protective function against the growth suppressive effects of TGF-β33. Also, forced expression of H11/HSPB8, an atypical heat shock protein with kinase activity that inhibits the growth of melanoma cells through activation of TGF-β activated kinase 1 (TAK1)-dependent death pathways, is silenced through aberrant promoter methylation in melanoma cells, thus contributing to resistance to TGF-β-mediated cytostasis34.

High levels of the oncoproteins SKI and SNON are detected in melanoma35,36. These proteins have been described as potent inhibitors of TGF-β signaling, acting both in the cell nucleus where they recruit transcriptional co-repressors to active SMAD complexes, and in the cytoplasm where they sequester SMAD proteins11. While it may be easily interpreted that such high expression of TGF-β pathway inhibitors is an obvious mechanism for melanoma cells to escape the growth inhibitory activity of TGF-β, reports from various laboratories indicate a complex regulatory network. In some melanoma cell lines, SKI or SNON knockdown inhibit cell and xenograft tumor growth36-38. Melanoma inhibitory activity (MIA), a protein expressed in melanoma but not in melanocytes39, has been shown to upregulate SKI and SNO expression via the MAPK/ERK cascade. This, in turn, was suggested to disrupt TGF-β signaling in melanoma cells40. It needs to be emphasized, however, that TGF-β signaling is functional in melanoma and that autocrine activation of the pathway in melanoma cell has been well documented12,17,18,41, underlying the complexity of TGF-β signaling in melanoma progression. Regarding SKI and SNON, TGF-β is a potent and rapid inducer of SKI and SNON degradation by the proteasome in a variety of cancer cell lines, including melanoma35,42. In the latter study, active and efficient TGF-β signaling and responsiveness was verified in melanoma cell lines, associated with rapid degradation of SKI and SNON proteins, within minutes after TGF-β stimulation. This event takes place not only in melanoma cells but also in various breast carcinoma cell lines42. Also, SKI knockdown prior to exposure to TGF-β did not significantly affect the extent of TGF-β responsiveness. Together, the latter studies indicate that TGF-β efficiently counteracts the ability of SKI and SNON to inhibit metastasis by inducing their degradation by the proteasome.

A number of other possible mechanisms of resistance to TGF-β may exist in melanoma, yet their functional relevance in disease is still ambiguous given the efficacy of TGF-β inhibitors to antagonize experimental melanoma progression24. For example, phosphorylation of the SMAD2 and SMAD3 linker regions was found to be high in both melanoma cell lines and melanoma samples43. Such event has previously been identified as important in the control of SMAD2/3 activity, leading to downregulation of TGF-β-responsive genes such as p21, p15 and c-MYC, and subsequent escape from TGF-β-induced cell growth arrest19. Yet, activation of other TGF-β target genes, such as PAI1, was not modified, suggesting additional, currently unknown, levels of activity control. The mechanisms responsible for SMAD2/3 linker domain phosphorylation may include MAPKs (p38, JNK and ERK)44-46, as well as GSK3β47 and CDKs48,49. This is compatible with the fact that melanomas are characterized by a high incidence of activating mutations of the BRAF and NRAS oncogenes, which lead to constitutive activation of MEK/ERK signaling, and thus could participate in SMAD2/3 linker phosphorylation. Also, stable overexpression of NRAS in melanoma cells was found to interfere with TGF-β-induced growth inhibition via alteration of Rb protein phosphorylation50.

ID proteins have been shown to enhance cell cycle progression as they interfere with bHLH, ETS and PAX transcription factors (that negatively regulate CDK inhibitors such as p15^Ink4b) and bind tumor suppressor proteins like Rb51. Interestingly, ID2 overexpression was shown to protect melanoma cells from TGF-β suppression of proliferation. ID2 is differently expressed in melanoma cells according to their proliferative vs. invasive phenotype. ID2 may be downregulated by TGF-β in proliferative cells that respond to its growth inhibitory effect, while TGF-β cannot repress ID2 expression in invasive cells52. PAX3, which is repressed by TGF-β in primary melanocytes, also participates to melanoma cell survival53. In murine B16 melanoma cells, PAX3 overexpression was found to desensitize cells to TGF-β-induced growth inhibition15, thus representing another potential mechanism of resistance to TGF-β anti-proliferative effects.

Melanoma cells secrete high levels of TGF-β that acts on tumor cells and their microenvironment by both autocrine and paracrine mechanisms, to initiate or amplify various processes, either cell-autonomous or non cell-autonomous, that contribute to tumor progression (Fig. 2).

In epithelial tumor cells, TGF-β pathway promotes an EMT, a critical multistep process that converts epithelial cells into motile and invasive mesenchymal-like cells. This mechanism is believed to allow cancer cells to escape from the primary tumor and to become metastatic54. During the transition from radial growth phase to vertical growth phase, melanoma cells undergo a process that may be considered a pseudo-EMT whereby melanoma cells acquire a fibroblastoid morphology55, via mechanisms that recapitulate many of the molecular characteristics of EMT56. In this context, TGF-β has been shown to promote melanoma cell pseudo-EMT through up-regulation of matrix metalloproteinase MMP9, β1 and β3 integrin and down-regulation of the epithelial marker E-cadherin.

The role of autocrine TGF-β/SMAD signaling in controlling the invasiveness and metastatic potential of melanoma cells has been described in several studies from our laboratory. In melanoma cells, we showed that overexpression of SMAD7, an inhibitory SMAD that interferes with TGF-β receptor function, decreases their capacity to form colonies in an anchorage-independent manner, reduces MMP2 and MMP9 secretion and Matrigel-invading capacity, and delays xenografted tumor growth in immuno-compromized mice57. Moreover, melanoma cells overexpressing SMAD7 were less prone than their wild-type counterparts to form experimental bone metastases following intraventricular inoculation in mice, highlighting the role of TGF-β signaling in metastatic progression23. Mechanistically, SMAD7 was found to induce β-catenin degradation and promote homotypic interactions with N-cadherin between melanoma cells, and to stabilize their association with neighboring dermal fibroblasts, an additional mechanism that likely attenuates invasion58. Another approach consisted in using the small molecule TβRI inhibitor SD-208 to interfere with TGF-β signaling. This pharmacologic inhibitor efficiently blocked TGF-β-induced SMAD3 phosphorylation and SMAD3/4-specific transcription, as well as Matrigel invasion by melanoma cells. Furthermore, SD-208 inhibited the expression of the TGF-β target genes parathyroid hormone-related peptde (PTHrP), IL-11, connective tissue growth factor (CTGF), and RUNX2, all involved in the development of osteolytic bone metastases. Systemic administration of SD-208 to nude mice prior to melanoma cell inoculation dramatically inhibited the establishment of osteolytic bone metastases. Also, SD-208 treatment significantly reduced the growth of established osteolytic lesions in the same mouse model of bone metastases24. Most recently, we found that halofuginone, an alkaloid that interferes with TGF-β signaling through the induction of SMAD7 and repression of TβRII expression59, inhibits melanoma experimental metastasis to bone and brain60.

Taken together, these results demonstrate that therapeutic targeting of TGF-β may prevent the development of melanoma bone metastases and decrease the progression of established osteolytic lesions.

We also demonstrated that GLI2, a transcription factor involved in the hedgehog (HH) signaling cascade, is a direct transcriptional target of TGF-β/SMAD signaling in various cell lines, including melanoma61,62. Accordingly with numerous studies that have established the pro-oncogenic role of GLI2 in various malignancies, GLI2 is highly expressed in aggressive melanoma cells where it regulates their invasive capacity by upregulating MMP2 and MMP9 expression. In melanoma cells, GLI2 expression is driven, at least in part, by TGF-β, and is associated with a loss of E-cadherin expression63. In addition, GLI2 is a potent inducer of WNT5A and PTHrP expression, two proteins involved in the metastatic process64-66.

TGF-β also induces the expression of integrins, proteases such as cathepsin B and matrix remodeling proteins like SPARC that may be involved in the acquisition of mesenchymal traits that favor metastatic dissemination. TGF-β1 contribution to cathepsin B-mediated melanoma invasiveness is highly Secreted Protein Acidic and Rich in Cystein-dependent67. SPARC also cooperates with interferon (IFN)-γ to repress the expression of a disintegrin and metalloproteinase (ADAM)1568. Yet, ADAM15 may have tumor-suppressive functions in melanoma as it was shown to reduce the formation of lung metastases in mice and to reduce migration and invasive capacity of melanoma cells in vitro69. In another study, genome-wide microarray analyses of invasive human melanomas and benign nevi revealed up-regulation of cathepsins B and L, MMP-1 and MMP-9, and urokinase- and tissue-type plasminogen activators. The mRNA levels of cathepsins B/L and plasminogen activators, but not MMPs, correlated with metastasis. Primary melanomas and benign nevi revealed that cathepsin B is predominantly expressed by melanoma cells while cathepsin L is preferentially expressed by tumor-associated fibroblasts surrounding the invading melanoma cells. Cathepsin B was shown to be important for both the production and secretion of TGF-β signaling, which in turn activates fibroblasts and subsequent invasive capacity of melanoma cells70.

In spindle tumor cells including melanoma cells, TGF-β was shown to induce the expression of fascin, an actin-bundling protein critical for metastasis. By stabilizing the actin cytoskeleton within invadopodia and filopodia, fascin coordinates the invasion and metastasis promoted by TGF-β71.

TGF β induced protein (TGF-βI) is a TGF-βinducible, secreted extracellular matrix (ECM) protein shown to impair the adhesive contacts of melanoma cells with the ECM molecules fibronectin, type I collagen and laminin. TGF-βI was found to localize with the fibrillar fibronectin/tenascin-C/periostin structures that characteristically surround the invasive fronts of melanomas72.

Several cytoplasmic proteins have been identified that interact with the TGF-β receptors. Aberrant expression and/or regulation of these receptor interactors could interfere with TGF-β signaling and promote tumor development. In Xenopus, the pseudoreceptor BAMBI (bone morphogenetic protein and activin membrane-bound inhibitor) is a naturally occurring truncated type I receptor that lacks a kinase domain. It can interact with TGF-β in the presence of TβRII, binding TβRI and TβRII and disrupting ligand-induced receptor heterodimerization73. Interestingly, NMA, a mammalian orthologue of BAMBI, was identified by differential display analysis as decreased in highly metastatic melanoma74, suggesting that derepression of TGF-β signaling following a decrease of NMA expression could account for the increased metastatic potential of these cells.

Paracrine secretion of TGF-β may modulate the tumor microenvironment for the benefit of melanoma growth, invasion and metastasis, especially through its capability to activate stromal fibroblasts, its immune-suppressive effects and pro-angiogenic properties.

TGF-β-activated stromal fibroblasts can stimulate ECM protein production, including collagen, fibronectin, tenascin and α2-integrin, thus providing a scaffold that promotes melanoma cell survival, migration and metastatic progression75. Interestingly, autocrine as well as paracrine TGF-β may also promote endothelial-to-mesenchymal transition (EndMT), a process whereby endothelial cells acquire mesenchymal features and contribute to the accumulation of fibroblasts in the tumor microenvironment that facilitates tumor progression in a B16F10 melanoma model76.

During melanoma progression, tumors acquire a rich vascular network due, at least in part, to paracrine secretion of TGF-β that allows the recruitment and proliferation of endothelial cells for the development of new blood vessels. In melanoma cells, TGF-β-induced angiogenesis involves up-regulation of several pro-angiogenic factors, including IL-8 and vascular endothelial growth factor (VEGF)77.

Periostin is a TGF-β-induced ECM protein involved in cell invasion, metastasis and angiogenesis. In a murine model of human melanoma, the use of an antibody directed against a periostin epitope that corresponds to a binding site for integrins significantly reduced tumor growth and angiogenesis78.

Moreover, treatment of a B16 melanoma model with a chimeric protein derived from the fusion of IL-2 and the soluble extracellular domain of TβRII resulted in reduced angiogenesis and tumor growth79.

TGF-β negatively regulates the activity of immune cells and plays an important role in the control of T-cell functions and proliferation. For example, TGF-β blocks the production of IL-2 (a lymphokine that activates T cells, natural killer (NK) cells and other immune cells) and cell cycle regulators in T cells. TGF-β also controls the expression of effector molecules produced by cytotoxic T lymphocytes, and is an important differentiation factor for regulatory T lymphocytes (Treg), which are present in high quantities in tumors including melanoma and contribute to TGF-β-mediated immunosuppression11.

Melanoma cells, but not normal melanocytes, express monocyte chemoattractant protein-1 (MCP-1), a potent chemokine that recruits macrophages80 that in turn infiltrate melanomas and produce factors that help tumor invasion, angiogenesis and immunosuppression81. IL-10 expression has been shown in vitro in many melanoma cell lines82, and in vivo, in primary and in metastatic melanoma lesions83. Also, elevated levels of IL-10 have been observed in patients with advanced melanoma and correlate with tumor progression84. TGF-β was shown to increase the expression of MCP-1 and IL-10 in melanoma cells, an effect mediated by the crosstalk between the SMAD, PI3K/AKT and BRAF/MAPK signaling pathways85.