Transforming growth factor-β in cutaneous melanoma


Address correspondence to A. Mauviel, e-mail:


Transforming growth factor-β (TGF-β) plays a complex role during carcinogenesis. It may either act as a tumor suppressor through its broad antiproliferative potential or as a tumor promoter either via direct effects on tumor cell aggressiveness or indirectly by modulating stromal responses, angiogenesis and immune surveillance. Increased production of TGF-β by cancer cells is often associated with tumor grade. Melanoma cells largely escape cell cycle arrest normally induced by TGF-β in normal melanocytes, yet produce active TGF-β and are capable of efficient transcriptional responses to the growth factor. In this review, we summarize the current knowledge about the role played by TGF-β in melanoma progression and hypothesize about the appropriateness of targeting TGF-β signaling for therapeutic intervention.

Transforming growth factor-β

The transforming growth factor-β (TGF-β) family of growth factors comprises above 40 members, including, but not restricted to, TGF-β, Activins, Bone Morphogenetic proteins (BMP), and Nodal. They all are ubiquitous multifunctional cytokines that regulate a plethora of cellular activities, such as proliferation, differentiation, migration or survival, as well as physiological processes, among which various phases of embryonic development, angiogenesis or immune surveillance (Javelaud and Mauviel, 2004). There are three mammalian isoforms of TGF-β, sharing between 75% and 80% peptide sequence similarities, encoded by distinct genes. They are synthesized as latent precursors that require maturation before being able to bind their cellular receptors and induce biological responses. Proteolytic cleavage of the latency-associated peptide leads to the release of a biologically active, mature form of TGF-β.

The expression of each isoform is distinct during embryonic development. Their spatial and temporal patterns of expression are somewhat superimposed but also identify distinct and non-redundant functionalities that have been confirmed by specific gene ablation in mice. Mice knockout for Tgfb1 exhibit a severe post-natal inflammatory reaction accompanied with tissue necrosis leading to organ failure and rapid death, identifying TGF-β1 as critical for the control of inflammation and immune responses. Tgfb2 knockout results in severe developmental defects affecting several organs including the heart. The Tgfb3 knockout phenotype is more subtle, as it is not fully penetrant and consists in severely impaired palatal closure (Bottinger and Kopp, 1998).

The TGF-β signaling machinery from cell membrane to nucleus

Signal transduction is initiated by the binding of TGF-β ligands to their cellular receptors (Figure 1). Transforming growth factor-β binds to the TGF-β type II receptor, TβRII, a receptor with constitutive kinase activity which, upon ligand binding, transphosphorylates TGF-β receptor type I (TβRI) in its Serine/Glycine-rich domain onto serine and threonine residues. This, in turn, induces TβRI kinase activity, leading to the phosphorylation of Smad2 and Smad3, proteins that constantly shuttle between the nucleus and cytoplasm (Reguly and Wrana, 2003). Upon phosphorylation in their C-terminal domain, Smad2/3 associate with Smad4 to form heterocomplexes that accumulate in the nucleus and function as transcription factors to activate the transcription of target genes, binding to their 5′ regulatory regions either directly or indirectly via association with other transcription factors (ten Dijke and Hill, 2004; Feng and Derynck, 2005; Javelaud and Mauviel, 2004; Shi and Massague, 2003).

Figure 1.

 The TGF-β signaling machinery. Upon TGF-β ligation to TβRII, the latter phosphorylates TβRI (TGF-β type I receptor), which in turn phosphorylates Smad2/3 which, in the absence of ligand, shuttle between cytoplasm and nucleus. Activated Smad2/3 bind Smad4 and accumulate in the nucleus to act as transcription factors. Inhibitory Smad7 binds activated TβRI, thereby preventing phosphorylation of Smad2/3, or recruits ubiquitine ligases (Smurf1,Smurf2,WWP1) or phosphatase PP1 through its subunit Gadd34 to induce proteasomal degradation or dephosphorylation, respectively, of the receptor complexes. c-SKI and SnoN act as transcriptional co-repressors to repress transcription induced by TGF-β. Aside from Smad signaling, activated TGF-β receptor complexes may also signal via non-canonical pathways.

Smad7, another member of the Smad family of proteins, is an inhibitor of TGF-β signaling that controls TGF-β receptor function. Smad7 may either bind the receptor complexes directly to prevent Smad2/3 phosphorylation (Moustakas et al., 2001), or recruit either the protein phosphatase PP1/GADD34 or E3 ubiquitin ligases of the Smurf family to the activated receptors. These events, in turn, result in receptor dephosphorylation or ubiquitination followed by proteasomal degradation, respectively (Komuro et al., 2004; Shi et al., 2001).

Ski and SnoN oncoproteins have been identified as inhibitors of TGF-β responses (Liu et al., 2001). Ski has been shown to interact directly with Smad2/3/4 at the level of target gene promoters, recruiting transcriptional co-repressors of the N-CoR family as well as histone deacetylases, thereby blocking Smad-driven transcription. Ski may also stabilize the binding of inactive Smad complexes to DNA and may interfere with TβRI phosphorylation of Smad2/3. Similar mechanisms of action have also been identified for the related protein SnoN, which also interferes with Smad signaling by sequestering Smad complexes in the cytoplasm (Krakowski et al., 2005).

Other signal transduction pathways may be activated by TGF-β, including all MAP kinase pathways. Activation of MAPK activity by TGF-β exerts a regulatory role on the transcriptional responses elicited by Smads (Javelaud and Mauviel, 2005). Furthermore, activation of MAPK by a variety of cytokines (TNF-α or IL-1 for example) or pharmacological agents (halofuginone or rapamycin for example) is often critical for their ability to counteract TGF-β effects (Poulalhon et al., 2006; Tacheau et al., 2007; Verrecchia and Mauviel, 2004; Verrecchia et al., 2000, 2001, 2003; Xavier et al., 2004). Remarkably, melanomas often exhibit constitutive ERK activation, due to activating mutations in the upstream kinases BRAF or N-Ras. Functional crosstalks between the ERK and Smad pathways will be discussed later in the text.

TGF-β and cancer

Transforming growth factor-β is a potent inhibitor of epithelial cell proliferation (Figure 2), and as such is considered a potent tumor suppressor during early stages of carcinogenesis (Elliott and Blobe, 2005; Wakefield and Roberts, 2002). Transforming growth factor-β operates through the induction of the cyclin-dependent kinase (CDK) inhibitors p21 and p15, as well as p27Kip1 and p57Kip2, leading to cell cycle arrest in G1 phase (Massague and Gomis, 2006). Alterations of TGF-β signaling, including loss-of-function mutations in genes encoding TGF-β receptors or Smad proteins, are often found in human tumors. They may represent mechanisms by which tumor cells escape from the antiproliferative activity of TGF-β (de Caestecker et al., 2000).

Figure 2.

 The multifaceted role of TGF-β in cancer.

Remarkably, TGF-β may also exert tumor promoter activities at later stages of carcinogenesis. Most human tumors secrete large amounts of TGF-β, which directly influences the microenvironment and promotes tumor growth, invasiveness and capacity to disseminate and form metastases (Leivonen and Kahari, 2007). Transforming growth factor-β also contributes to the development of peri-tumoral angiogenesis, and as a negative regulator of immune functions, may also favor tumor escape from immune surveillance.

TGF-β in melanoma

Melanomas are cancers originating from melanocytes, primarily in the skin but also in the eye, the brain and mucosas. Although melanomas represent only approximately 4% of all cutaneous cancers, they are responsible for >80% casualties from skin-related cancers (Houghton and Polsky, 2002). Melanoma affects patients in all age ranges and its incidence has dramatically risen over the past 50 years in industrialized countries. While very early stage melanoma (localized, stage I) is >90% curable, disseminated stage IV melanoma leads to life expectancy of less than a year (Kim et al., 2002). Metastases may establish in various organs, including skin, lungs, liver, brain, and bone.

Increased expression and secretion of the different TGF-β isoforms in melanoma cell lines when compared with normal melanocytes has been documented by several studies (Albino et al., 1991; Krasagakis et al., 1994; Rodeck et al., 1991, 1994). In situ, TGF-β1 is secreted by normal melanocytes and melanomas at various stages, while TGF-β2 and TGF-β3 are not expressed in normal melanocytes, but only heterogeneously in nevi and melanomas. TGF-β2 and TGF-β3 seem to appear early in melanoma progression and to increase with tumor progression (Van Belle et al., 1996). Also, a correlation between TGF-β2 expression and tumor thickness has been reported (Reed et al., 1994). Increased TGF-β1 and TGF-β2 plasma levels are observed at later stages of tumor development, while no significant differences have been reported between those of healthy patients and those from patients with primary or locally invasive melanoma (Krasagakis et al., 1998). Thus, despite some discrepancies, all these studies point toward an increase in TGF-β expression levels that correlates with tumor progression.

Resistance of melanoma cells to the antiproliferative activity of TGF-β

Although normal melanocytes are very sensitive to the antiproliferative effects of TGF-β, melanoma cells exhibit increased resistance, proportional to tumor progression stage. Melanoma cell proliferation is only moderately inhibited by TGF-β in contrast to the strong antiproliferative effect exerted on normal melanocytes. Also, it has been shown that metastatic melanoma cell populations exhibit a further decreased response to TGF-β-dependent growth inhibition than melanoma cells originating from primary tumors (Krasagakis et al., 1999). Yet, TGF-β is perfectly capable of inducing Smad signaling and Smad-dependent transcription in melanoma cells (Rodeck et al., 1999), suggesting that desentization of melanoma cells to the anti-proliferative activity of TGF-β is highly specific to cell cycle progression. Contrary to other tumor types, no genetic alteration of TGF-β signaling molecules has been identified in melanoma that could explain their resistance to the growth inhibitory activity of TGF-β.

One hypothesis arises from the fact that melanoma cells express high levels of both SKI and SnoN, when compared with normal melanocytes (Reed et al., 2001; Poser et al., 2005). It has been suggested that SKI and SnoN interfere with Smad-dependent transcription and prevent p21 induction by TGF-β in melanoma cells, thus aborting one of the main mechanisms by which TGF-β exerts its antiproliferative activity. Using siRNA targeting SKI expression in two melanoma cell lines, the authors were able to restore both p21 induction and growth inhibition in response to TGF-β. Although the SKI hypothesis is tempting, it is somewhat hard to reconcile with a number of reports from the literature, indicating that melanoma cells are capable of responding to TGF-β with solid transcriptional responses, not compatible with an antagonistic activity of SKI/SnoN on Smad signaling. For example, it has been clearly established that a number of TGF-β target genes are upregulated in melanoma cells exposed to TGF-β, in particular those involved in invasion and metastasis (Janji et al., 1999; Miyoshi et al., 1995; Park et al., 1996). Also, we previously demonstrated an absence of correlation between autocrine TGF-β-dependent Smad signaling and transcription and the degree of growth inhibition exerted by this growth factor on melanoma cells (Rodeck et al., 1999). Most recently, we have identified that transcriptional responses to TGF-β, measured using either a Smad3/ 4-specific reporter construct in transient cell transfections or as the rapid induction of the TGF-β immediate-early target gene GLI2 (Dennler et al., 2007), is not affected by high basal levels of c-SKI protein in melanoma cells (Mauviel et al., 2007a). We also observed that in a dozen of distinct melanoma cell lines, TGF-β induced a rapid degradation of c-SKI protein, followed by delayed induction of c-SKI gene expression. However, proteasome blockade with either MG132 or ALLN resulted in further increased c-SKI accumulation and to partial reduction in TGF-β-driven c-SKI degradation, accompanied with attenuated TGF-β-dependent gene transcription. Thus, endogenous c-SKI is not necessarily able to antagonize TGF-β-dependent transcription in melanoma cells, unless stabilized by proteasome blockade (D. Javelaud et al., unpublished data). Together, these results suggest that mechanisms leading to c-SKI accumulation in melanoma, and subsequent biological outcome, remain to be further studied.

We previously showed that basal Smad-dependent transcription in melanoma cells is inhibited by a pan-TGF-β neutralizing antibody (Rodeck et al., 1999), therefore proving that autocrine Smad signaling is largely dependent upon secretion and pericellular activation of TGF-β. A number of other studies from various laboratories have confirmed that melanoma cells efficiently respond to TGF-β. For example, it has been shown that TGF-β is a potent inducer of integrins and VEGF gene expression (Ijland et al., 1999; Janji et al., 1999; Liu et al., 2005; Schadendorf et al., 1993), implicated in metastasis and angiogenesis, respectively. A recent genome-wide transcriptome analysis in nearly a hundred human cell lines in culture was able to identify two populations with very distinct gene expression profiles: the first one was characterized by high expression of neural crest and melanocytic differentiation markers, the other one characterized by the expression of a number of genes associated with a more aggressive phenotype, whose concomitant expression is reminiscent of a TGF-β signature, indicating that TGF-β responsiveness may vary across melanomas (Hoek et al., 2006).

Autocrine TGF-β signaling contributes to melanoma aggressiveness and metastasis

The importance of constitutive TGF-β signaling through the Smad pathway in controlling the invasive capacity and metastatic potential of melanoma cells has been highlighted by several recent studies from our laboratory (Figure 3). Specifically, we found that overexpression of the inhibitory Smad, Smad7, in melanoma cells reduces their capacity to form colonies in an anchorage-independent manner, dramatically reduces MMP2 and MMP9 secretion and capacity to invade Matrigel™, and delays tumor growth in a subcutaneous injection model in nude mice (Javelaud et al., 2005). Because TGF-β promotes metastases to bone by several types of solid tumors including breast cancer (Kang et al., 2005; Yin et al., 1999), we hypothesized that TGF-β may contribute to melanoma metastases to bone. Consistent with this, overexpression of the inhibitory Smad7 blocked the capacity of 1205Lu melanoma cells to form bone metastases by preventing TGF-β induction of genes known to mediate osteolytic breast cancer bone metastases such as PTHrP, IL-11, CXCR4, and osteopontin. Furthermore, using a mouse model of bone metastasis in which tumor cells are inoculated to the left cardiac ventricle, we determined that Smad7-overexpressing melanoma cells exhibited a largely delayed bone metastasis growth when compared with mock-transfected cells, associated with reduced bone tumor burden and increased survival (Javelaud et al., 2007).

Figure 3.

 Effects of Smad7 overexpression in melanoma cells. This schematic summarizes our knowledge of the effects of Smad7 overexpression in melanoma cells. For details, see Javelaud et al., 2005, 2007.

Although it is highly likely that Smad7 effects are directly linked to its inhibitory function against TGF-β/Smad signaling, one cannot exclude that Smad7 may also function independently (Pulaski et al., 2001). Thus, our next approach to determine the exact role played by TGF-β signaling in melanoma metastasis, consisted in blocking TGF-β signaling in mice, via systemic administration of a highly specific small molecule inhibitor of TβRI, SD-208 (Uhl et al., 2004). In vitro, SD-208 blocked TGF-β induction of Smad3 phosphorylation, TGF-β-induced Smad3/4-specific transcription in several melanoma lines, as well as invasion through MatrigelTM. Since Smad7 overexpression resulted in dramatically impaired capacity of human melanoma cells to form bone metastases, we hypothesized that pharmacologic blockade of TGF-® signaling using the small-molecule TβRI inhibitor SD-208, would be an effective treatment for melanoma metastasis to bone. To this aim, nude mice were inoculated with 1205Lu melanoma cells into the left cardiac ventricle and treated with SD-208 administered via oral gavage in both prevention and treatment protocols. In the former, SD-208 started 2 days before tumor inoculation, reduced the development of osteolytic bone metastases compared with vehicle. In the latter, mice given SD-208 2 weeks after tumor cell inoculation had significantly smaller osteolytic lesions than vehicle-treated mice 4 weeks later. These data provide definitive evidence of a direct role of TGF-β in the development of osteolytic bone metastases caused by malignant melanoma, likely via mechanisms similar to those implicated in breast cancer osteolysis (Mauviel et al., 2007b; K.S. Mohammad et al., unpublished data). Whether preventive administration of the TβRI inhibitor may alter the capacity of host tissues to function as an adequate premetastatic niche for melanoma cells remains to be understood.

Crosstalks between Smad and ERK signaling

A genetic hallmark of melanoma is the frequent presence of activating mutations within the kinases BRAF (approximately 70% of melanomas; Daniotti et al., 2004; Libra et al., 2005; Maldonado et al., 2003; Uribe et al., 2003) and N-RAS (approximately 15% of melanomas; Bloethner et al., 2005; van Elsas et al., 1995; Goydos et al., 2005; Platz et al., 1994; van ‘t Veer et al., 1989), resulting in constitutive activation of the MEK/ERK pathway.

It is generally understood that integration of Smad signaling within crosstalks with other signal transduction cascades, and in particular with the Ras/MEK/ERK cascade, is likely to dramatically alter the initial Smad signals and contribute to the broadly pleiotropic activities of TGF-β.

Initial observations pointed toward an antagonistic activity of ERK activation against Smad signaling [reviewed in (Massagué and Chen, 2000; Derynck and Zhang, 2003; Javelaud and Mauviel, 2005)], and it was thought that the opposing effects of Ras signaling against Smad activity or accumulation in the nucleus may allow for reduced anti-proliferative response of epithelial cells to TGF-β (Kretzschmar et al., 1999). Also, the two pathways have been shown to exert opposite effects on cyclin-dependent kinases during the G1 phase of the cell cycle (Hannon and Beach, 1994; Reynisdottir et al., 1995), which may also contribute to Ras-dependent attenuation of TGF-β-induced tumor suppression.

Contrary to the antagonistic activities of Ras signaling with the Smad pathway, it is also possible that constitutive activation of ERK/MAPK may allow for a switch of the TGF-β response from tumor suppressor to pro-oncogenic. For example, aberrant MAPK activation by Ha-Ras or Ki-Ras in human prostate and cancer cells converts the growth inhibitory activity of TGF-β into a Smad-independent mitogenic response (Lehmann et al., 2000). Also, in breast cancer cells, oncogenic Ras not only attenuates the cytostatic activity of TGF-β but also potentiates epithelial to mesenchymal transition in response to autocrine TGF-β signaling (Oft et al., 1998, 2002). Accordingly, EGFR-dependent Ras activation enhances TGF-β-dependent capacity of breast cancer cells to metastasize to bone (Yin et al., 1999). It appears therefore that the Ras/MEK/ERK pathway is not solely capable of blocking Smad signaling, but may also reprogram the response of epithelial cells to TGF-β and participate or enhance their aggressive and metastatic potential.

Given the highly frequent constitutive activation of both ERK and TGF-β signaling cascades in melanoma, together with the specificity of the melanocytic origin of this cancer, unraveling these crosstalks in the context of melanoma progression is of utmost importance. The complexity of signaling crosstalks is further illustrated by a recent study that identified the MEK/ERK pathway as an upstream activator of the JNK pathway in melanoma cells (Lopez-Bergami et al., 2007). Given the capacity of the latter to interfere with TGF-β signaling (see above), this work suggests additional levels of complexity in the ERK/Smad crosstalks.

Paracrine roles of tumor-secreted TGF-β

Transforming growth factor-β produced by tumor cells may promote tumor growth and progression by modifying the microenvironment (Li et al., 2003). It has been demonstrated that forced overexpression of TGF-β1 by melanoma cells activates stromal fibroblasts, leading to increased collagen, fibronectin, tenascin, and α2 integrin expression, and that, in experimental tumors generated subcutaneously by co-injection of fibroblasts with melanoma cells, TGF-β-overexpressing melanoma cells exhibit lesser necrotic and apoptotic cells, and form more lung metastases than control melanoma cells (Berking et al., 2001). Thus, activation of stromal fibroblasts by tumor-derived TGF-β provides an optimal microenvironment for tumor progression and metastasis, suggesting that therapeutic targeting of tumor stroma may represent a valuable alternative for future melanoma treatments (Li et al., 2003).

Transforming growth factor-β may also contribute to peri-tumoral angiogenesis by stimulating the expression of pro-angiogenic factors such as IL-8 and VEGFA. In addition, TGF-β is an activator of endothelial cell growth and migration, thereby allowing the establishment of a neovasculature around the tumor, which accelerates its development and allows further spreading.

Transforming growth factor-β is one of the most potent immunosuppressive cytokines (Li et al., 2006; Rubtsov and Rudensky, 2007; Wahl, 2007). It controls T-lymphocyte homeostasis by directly inhibiting their proliferation and activation, and by inhibiting antigen presentation by antigen-presenting cells. It was also recently identified as an important differentiation factor, along with interleukin-2, for regulatory T cells (Treg), a subset of strongly immunosuppressive T cells often found in tumors, including melanoma. On the other hand, antitumor immunosuppression by TGF-β may also occur via reduction of the level of T-helper cells, cytotoxic T cells, natural killer cells, dendritic cells and macrophages. Collectively, the various immunosuppressive activities of TGF-β secreted by tumor cells are likely to contribute to decrease the efficacy of immunotherapies and tumor vaccines. Accordingly, experimental blockade of TGF-β signaling in bone marrow cells with a viral vector expressing a dominant negative form of TβRII in a syngeneic mouse model of melanoma was sufficient to completely abolish melanoma development (Shah et al., 2002). Likewise, mice transplanted with bone marrow cells expressing the mutant TβRII were able to generate cytotoxic T lymphocytes specific for the tumor that blocked its metastatic progression (Shah et al., 2002). Together, these data provide strong evidence for TGF-β-dependent escape of melanoma from immune surveillance. With regards to human melanoma, it should be noted that high plasma levels of TGF-β have been observed in patients with advanced disease (Krasagakis et al., 1998). It is likely that circulating TGF-β contributes to systemic immunosuppression, thereby promoting tumor development.

Other TGF-β family members may also play an important role in melanoma progression. For example, melanomas overexpress both BMP-4 and BMP-7 (Rothhammer et al., 2005). Inhibition of BMP-4 expression in melanoma cells reduced their invasive and migratory potential, and reduced their paracrine activity on neo-angiogenesis (Rothhammer et al., 2007).

Another TGF-β family member that requires further attention in the context of melanoma biology is Nodal, whose expression was until recently thought to be restricted to embryonic development. It was found that melanoma cells express Nodal and that levels of expression are correlated with the stage of tumor progression. Also, Nodal was found to be potentially involved in the control of melanoma cell aggressiveness (Topczewska et al., 2006), although the matter is still controversial (Salomon, 2006).

Anticancer treatments targeting TGF-β

The vast number of experimental situations whereby TGF-β targeting has shown efficacy in preventing or delaying cancer development has raised interest for inhibition of TGF-β signaling for cancer treatment. Various strategies have been developed over the years and are described in several specific reviews (Akhurst, 2002; Lahn et al., 2005; Mourskaia et al., 2007; Saunier and Akhurst, 2006). Some of the most promising ones are summarized thereafter (Table 1).

Table 1.   Various means to interfere with TGF-β signaling
ReagentsSubgroupsTargetType of cancer cellsReference
NucleotidesAntisense reagents (AP11014, AP12009)TGF-β1Prostatic, colorectal and non-small cell lung cancer cellsMourskaia et al., 2007
TGF-β2Glioma, pancreatic cancer, adenocarcinoma and breast cancer cellsSchlingensiepen et al., 2006
Small interfering RNATGF-β1, TGF-β2Glioblastoma cellsFriese et al., 2004
Short hairpin RNASmad4Breast cancer cellsKang et al., 2005
ProteinsPan anti-TGF-β antibody (1D11, 2G7)TGFβ1-3Breast cancer cellsArteaga et al., 1993; Nam et al., 2006
sRIII (soluble form of TβRIII)TβRIBreast cancer, prostate cancer and melanoma cellsBandyopadhyay et al., 2006, 2002, 2005
Chimeric molecules (TβRII/Fc)TGF-β1, TGF-β3Lung cancer cellsKomesli et al., 1998
Small molecule kinase inhibitorsPyrazoles (LY364947, LY566578, LY580276, A-83-01)TβRIBreast cancer cellsPeng et al., 2005; Tojo et al., 2005; Arteaga, 2006
Imidazoles (SB-431542, SB-505124)TβRISeveral cancer cell linesHalder et al., 2005
Pyridopyrimidine-based molecules (SD-093 and SD-208)TβRIGlioma, pancreatic carcinoma and melanoma cellsGe et al., 2006; Uhl et al., 2004; Gaspar et al., 2007; Mauviel et al., 2007b

Antisense strategies against TGF-β are mostly developed to target malignant brain tumors, such as glioblastoma and gliosarcomas that typically express high TGF-β2 levels, creating an immunosuppressive environment enabling aggressive tumor growth. Small interfering RNA against TGF-β1 and TGF-β2 reduced the tumorigenicity and invasiveness of glioblastoma cells in vivo and made them more vulnerable to the immune system (Friese et al., 2004).

Antisense reagents targeting TGF-β1 (AP11014) and TGF-β2 (AP12009) were also recently developed. A phase I/II study consisting of intra-tumoral administration of AP-12009 has shown remarkable efficacy in some patients with glioma (Schlingensiepen et al., 2006). Although the number of patients enrolled in these studies was very small, it should be noted that there is no current efficient treatment against these tumors. Thus, any beneficial effect even on a single patient should be considered a progress, indicating that the antisense strategy against TGF-β2 requires further attention. Antisense approaches are currently being tested in melanoma and pancreatic adenocarcinoma (Schlingensiepen et al., 2008). Preclinical antisense TGF-β therapies in combination with IFNγ have been shown to inhibit tumor outgrowth and metastasis of aggressive breast cancer cells (Wu et al., 2001). RNAi strategies and short hairpin RNA have been recently used to target components of the TGF-β pathway and inhibit breast cancer metastasis to bone (Kang et al., 2000, 2005).

Pan-TGFβ antibodies (1D11 and 2G7), neutralizing TGFβ1-3, have been shown to reduce lung metastasis of breast cancer cells in mice (Arteaga et al., 1993; Nam et al., 2006), however, with a high risk of a widespread inflammatory response. A TGF-β co-receptor, known as betaglycan or TβRIII, binds all three TGF-β isoforms showing, however, the highest affinity for TGF-β2 (Vilchis-Landeros et al., 2001). Administration of a soluble form of TβRIII, sRIII, decreased angiogenesis, and reduced lung and bone metastases by breast, prostate, and melanoma cancer cells (Bandyopadhyay et al., 2002, 2005, 2006).

A chimeric molecule containing the extracellular domain of TβRII fused to the constant region (Fc) of a human immunoglobin heavy chain (TβRII/Fc) has been developed, exhibiting higher affinity for both TGF-β1 and TGF-β3 (Komesli et al., 1998). Experiments in mice revealed that TβRII/Fc may function by reducing tumor cell viability, migration, and invasion, but it had no effect on tumor angiogenesis.

The latest technology in TGF-β-targeting cancer therapy has been the development of small molecule kinase inhibitors with high affinity to TβRI (DaCosta Byfield et al., 2004; Halder et al., 2005; Inman et al., 2002; Tojo et al., 2005). Such molecules may be subcategorized according to their chemical structure: (i), pyrazoles (LY3649467, LY566578, LY580276, A-83-01), which are ATP-binding competitors for TβRI, showing 200-fold higher selectivity for TβRI compared with other kinases and (ii), imidazoles (SB-431542, SB-505124) that are more selective for TβRI than pyrazoles. SB-431542 was shown to inhibit several TGF-β tumor-promoting effects including EMT, cell motility and VEGF secretion in several cancer cell lines (Halder et al., 2005).

Recently new highly selective pyridopyrimidine-based, orally bioavailable inhibitors of TβRI have been developed. These compounds, SD-093 and SD-208, have been shown to inhibit the motility and invasion of glioma cells in vitro and to impair their growth after intracranial injection by promoting anti-tumor immunity (Ge et al., 2006; Uhl et al., 2004). They are also capable of reducing pancreatic carcinoma cell growth and aggressiveness (Gaspar et al., 2007). In melanoma cells, SD-208 potently inhibits TGF-β signaling, accompanied with dramatically reduced invasive properties in vitro and capacity to form bone metastases in a mouse model of intracardiac tumor cell inoculation (Mauviel et al., 2007b); K.S. Mohammad et al., unpublished data).

Although the efficacy of small-molecule inhibitors has been proven in vitro and in animal models, toxicity issues have arisen that may be because of either a lack of precise specificity to TβRI or to an in vivo efficacy that goes against the beneficial protective effects of TGF-β on healthy organ physiology, such as the heart for example. Curiously, these toxic effects are less or not seen with bigger molecules such as antibodies or soluble receptors, possibly because of a certain leakiness that would allow enough TGF-β signaling to prevail within healthy organs, or to a better specificity to TGF-β ligands than small-molecule kinase inhibitors.

Concluding remarks

There is ample evidence in the literature for an important role of the TGF-β signaling pathway during the course of melanoma progression and metastasis. Yet, a number of critical issues have to be resolved before targeted therapies can be envisioned. It is fundamental that better understanding of the outcome of signaling crosstalks is achieved: several signal transduction cascades are known to play an important role in melanoma progression, including, but not restricted to, MAP kinases, Wnt/β-catenin, PI3K/AKT or NF-κB. All are critical at various steps of melanocyte transformation, for melanoma progression from radial to vertical phase, and for metastasis. Transforming growth factor-β is both capable of tumor suppressor and pro-oncogenic activities, alone or in combination with other pathways. Understanding how and when signals are integrated within a tumor cell is a prerequisite for targeted therapy, as it will allow for a reasonable assessment of the risk-benefit. Likewise, understanding tumor cell heterogeneity and identifying specific signals within the population(s) to be targeted is of utmost importance. In particular, a major leap forward will likely come from the isolation and characterization of melanoma stem cells. Understanding signaling crosstalks within this cancer-initiating cell population will provide critical cues toward melanoma treatment.