Emerging molecular targets in melanoma invasion and metastasis

Authors


CORRESPONDENCE V. Sanz-Moreno, e-mail: victoria.sanz_moreno@kcl.ac.uk

Summary

Metastatic cutaneous melanoma accounts for the majority of skin cancer deaths due to its aggressiveness and high resistance to current therapies. To efficiently metastasize, invasive melanoma cells need to change their cytoskeletal organization and alter contacts with the extracellular matrix and the surrounding stromal cells. Melanoma cells can use different migratory strategies depending on varying environments to exit the primary tumour mass and invade surrounding and later distant tissues. In this review, we have focused on tumour cell plasticity or the interconvertibility that melanoma cells have as one of the factors that contribute to melanoma metastasis. This has been an area of very intense research in the last 5 yr yielding a vast number of findings. We have therefore reviewed all the possible clinical opportunities that this new knowledge offers to both stratify and treat cutaneous malignant melanoma patients.

Introduction

Cutaneous melanoma is the most serious and deadliest type of skin cancer, despite its relatively low incidence (4% of dermatological cancers). When diagnosed early, most patients with primary melanomas can be cured by surgical resection. However, metastatic melanoma accounts for more than 80% of deaths by skin cancer, due to its aggressiveness and resistance to existing therapies (Gray-Schopfer et al., 2007a; Miller and Mihm, 2006). Therefore, new therapeutic approaches are urgently needed for this devastating disease. To efficiently metastasize, invasive melanoma cells need to change their cytoskeletal organization and alter contacts with the extracellular matrix (ECM) and the surrounding stromal cells. Furthermore, metastatic melanoma cells are highly plastic and accommodate their mode of invasion depending on the differing microenvironments that they encounter during the metastatic cascade. In the last decades, much effort has been placed on deciphering molecules and signalling pathways that are deregulated in melanoma cells and how these factors promote their invasive behaviour. In this review, we summarize our current knowledge of cutaneous melanoma invasion and focus on potential molecular targets that could be employed in the treatment of metastatic melanoma.

Cutaneous malignant melanoma and metastatic spread

Cutaneous melanoma cells arise from skin melanocytes, melanin-producing cells that reside in the basal layer of the epidermis and in the hair follicles and whose homoeostasis is tightly controlled by epidermal keratinocytes (Gray-Schopfer et al., 2007a; Slominski et al., 2004). Transformation of melanocytes into melanoma entails a number of genetic and environmental factors. Loss of adhesion receptors and mutation of growth regulatory genes enable melanocytes to escape from the regulation by keratinocytes. One key event that facilitates this escaping is the loss of E-cadherin and the gain of N-cadherin, which allows melanocytes to get away from the control of keratinocytes and start interacting with stromal cells from the dermis, such as fibroblasts and endothelial cells (Haass et al., 2004). These events are crucial to allow melanoma cells to metastasize. Metastasis is the dissemination of cancer cells from the primary tumour and the subsequent formation of new tumour masses in peripheral tissues. Metastasis is one of the main clinical parameters known to affect the prognosis of cancer patients, including melanoma, and it is associated with resistance to chemotherapeutic treatments, higher post-treatment recurrence rates and poor cancer patient survival rates (Valastyan and Weinberg, 2011). Metastasis can be divided into a number of stages. The first stage occurs in the primary tumour, where subpopulations of cancer cells develop a metastatic phenotype and, to exit the tumour mass, migrate through an environment composed of ECM, stromal cells and cancer cells. This abnormal cell migration and invasion of the surrounding tissue is an essential component of metastasis. In fact, the acquisition of invasive behaviour is the key transition in the progression of benign melanocyte hyperplasia to life-threatening melanoma. Understanding this transition and the mechanisms of invasion are the key to understanding why malignant melanoma is such a devastating disease and this will aid treatment strategies. Underlying the invasive behaviour is increased cell motility caused by changes in cytoskeletal organization and altered contacts with the ECM and the stroma. The second stage involves the dissemination of cancer cells away from the primary tumour either via lymph vessels or the vasculature in a process called intravasation (Bustelo, 2012). Circulating cancer cells then adhere to the microvasculature and move across the vascular endothelial cell layer and the ECM of the peripheral tissue in a process called extravasation. Finally, the extravasated cells colonize the new peripheral niche, a process called macrometastasis (Bustelo, 2012).

More than 20 yr ago, Clark et al. (1984) proposed a model to describe melanoma development and progression based on clinical and histopathological features. Transformed melanocytes can proliferate and spread, giving rise to a common naevus. Naevi are generally benign lesions that rarely progress to melanoma, probably due to oncogene-induced cell senescence. Further alterations and a more aberrant growth can lead to a dysplastic naevus with structural and architectural atypia, which can stay latent for long periods or progress towards a more malignant stage. Melanoma cells in the radial-growth phase (RGP) usually arise de novo and proliferate and spread laterally mostly within the epidermis, although some local microinvasion of the dermis can occur. More dangerous is the transition to the vertical-growth phase (VGP), in which cells have metastatic potential. Vertical-growth phase melanoma cells are able to invade the dermis and intravasate into the blood or lymph circulation, and eventually disseminate to distant organs. While there are no strong genetic candidates to mediate the switch from RGP to VGP, microarray and immunohistochemistry analysis have identified numerous genes whose expression level is altered as melanoma becomes invasive (Alonso et al., 2004, 2007; Bittner et al., 2000; Clark et al., 2000; Gaggioli and Sahai, 2007; Gray-Schopfer et al., 2007a; Villanueva and Herlyn, 2009). Importantly, some melanomas do not follow this progression stepwise, given that both RGP and VGP can progress directly to metastatic melanoma (Clark, 1991; Clark et al., 1984; Gray-Schopfer et al., 2007a; Miller and Mihm, 2006).

Generally, the diagnosis and prognosis of primary melanoma lesions entail their classification based on histopathological criteria, such as the thickness of the tumour. However, there are a significant number of cases with thin melanomas that rapidly metastasize (Fecher et al., 2007), while a large percentage of thick lesions do not lead to metastatic disease (Lomuto et al., 2004). Given melanoma heterogeneity and the difficulty in its classification and staging, the last decade has yielded a myriad of gene expression profile studies that intended to establish a molecular classification of melanoma. Those studies have confirmed the great heterogeneity of this disease and the key role the tumour microenvironment has in determining melanoma gene expression programmes and progression to metastasis (Bittner et al., 2000; Hoek et al., 2006; Seftor et al., 2005); interested readers are suggested to read specific reviews elsewhere (Fecher et al., 2007; Hoek, 2007).

Genetic changes and altered gene expression leading to melanoma metastasis

Cutaneous melanoma is a complex genetic disease influenced by both genetic and environmental factors. Most of the genetic alterations associated with melanoma development result from sporadic somatic mutations. Familial melanoma accounts only for 10% of melanomas, predominantly harbouring germline mutations in genes involved in regulating cell cycle progression, like CDKN2A and CDK4, or pigmentation, such as melanocortin receptor MC1R (Chin, 2003). Generally, mutations activating oncogenes and inactivating tumour suppressors lead to deregulation of critical intracellular signalling pathways (cell cycle, apoptotic machinery, motility and cytoskeleton, etc.) and the interaction of melanoma cells with the tumour microenvironment, resulting in primary tumour formation and eventually metastatic progression. To develop new and more effective therapeutic strategies, it is essential to gain a full understanding of the deregulated signalling pathways in melanoma.

During the last decades, new technologies involving comprehensive strategies such as comparative genomic hybridization and mutation analysis by gene sequencing have provided insights into crucial cell signalling pathways altered in melanoma (Gray-Schopfer et al., 2007a). In this section, we will only focus on melanoma genetic changes and altered gene expression involved in cell invasion and metastasis; for a more exhaustive description, excellent reviews can be found elsewhere (Bennett, 2008; Chin, 2003; Gray-Schopfer et al., 2007a; Miller and Mihm, 2006; Sekulic et al., 2008; Villanueva and Herlyn, 2009).

BRAF, RAS and the MAPK pathway in invasion and metastasis

Perhaps, one of the most important and exciting discoveries on the biology of melanoma development has been the finding that 50–70% of melanomas harbour mutations in BRAF (Davies et al., 2002). Braf belongs to the Raf family of serine/threonine kinases, which are effectors of the small GTPase Ras in the ERK/MAPK pathway. Traditionally, this pathway has been involved in regulation of cell growth, survival and differentiation and is activated by a lot of different membrane-bound receptors, such as receptor tyrosine kinases and G-protein-coupled receptors (Gray-Schopfer et al., 2005). Activation of membrane receptors by growth factors promotes Ras switching to its active GTP-bound state, which then binds and activates multiple effector proteins, including the three Raf members (Araf, Braf and Craf), leading to subsequent activation of a cascade of kinases (MEK1/2, ERK1/2). Activated ERK phosphorylates a number of nuclear and cytoplasmic substrates that mediate multiple cellular processes (Gray-Schopfer et al., 2005).

Common gain-of-function mutations leading to enhanced melanoma proliferation–through hyperactivation of ERK–occur in one member of the Ras family of oncogenes, NRAS, which is mutated in 15–30% of melanomas, the majority harbouring a leucine for glutamine substitution in position 61 (Q61L; Davies et al., 2002). Regarding the other RAS family members, HRAS activation has occasionally been detected in melanoma, although it is more frequently associated with Spitz naevi, while KRAS mutations have not been described in human melanocytic lesions (Chin et al., 2006). RAS signals through a number of effectors, especially those in the RAF/MEK/ERK pathway and the phosphoinositide-3-OH kinase (PI3K) pathway, thereby affecting cell proliferation, survival, migration and invasion (Gray-Schopfer et al., 2007a; Shaw and Cantley, 2006). Importantly, oncogenic RAS is able to transform melanocytes presumably involving PI3K pathway (Wellbrock et al., 2004). RAS can also induce melanoma in p16INK4a-deficient mice, and it is involved in tumour maintenance (Ackermann et al., 2005; Chin et al., 1999).

As stated previously, the most commonly mutated component of the MAPK pathway is BRAF, and the most frequent mutation is a glutamic acid for valine substitution at position 600 (V600E; Davies et al., 2002). Interestingly, both NRAS and BRAF mutations are mutually exclusive in melanoma (Gray-Schopfer et al., 2007a).

Activating non-V600EBRAF mutations (e.g. G465A, T598, K600E, A727V), although very rare, are also important in melanoma as they are able to directly phosphorylate and activate MEK and ERK signalling (Garnett and Marais, 2004; Gray-Schopfer et al., 2005; Wan et al., 2004). Other mutants have low or impaired activity (G466E/V, G469E, G596R, D594G), and even though they are not able to directly phosphorylate MEK, they appear to retain enough activity to transphosphorylate and activate CRAF in a RAS-independent manner, eventually activating the pathway indirectly through CRAF (Garnett et al., 2005; Wan et al., 2004).

Over 90% of BRAF-mutant melanomas harbour the V600E substitution. V600EBRAF triggers constitutive ERK signalling, in turn stimulating proliferation and survival (Gray-Schopfer et al., 2005). Intriguingly, BRAF is also mutated in up to 80% of benign naevi (Pollock et al., 2003), even though most naevi stay indolent for decades and rarely progress to melanoma. It is believed that mutant BRAF and NRAS would lead to melanocyte hyperproliferation and consequent p16INK4a-mediated senescence, giving rise to naevi with senescent melanocytes and, therefore, benign dormant lesions (Gray-Schopfer et al., 2007a). V600EBRAF also contributes to neoangiogenesis by stimulating autocrine vascular endothelial growth factor (VEGF). Other relevant downstream targets of mutant BRAF are the microphtalmia-associated transcription factor (MITF; Wellbrock and Marais, 2005), the POU-domain class 3 transcription factor BRN2 (Goodall et al., 2004) and the matrix metalloproteinase-1 (MMP-1; Huntington et al., 2004), which will be discussed later; for other targets of mutant BRAF, see reference (Packer et al., 2009).

Importantly, several studies have shown the implication of V600EBRAF in melanoma invasion and metastasis. The major mediators of cell–ECM interactions are integrins, which form heterodimeric complexes consisting of an α and β subunit. Numerous groups have reported deregulated expression of integrins in invasive melanomas and shown the functional importance of integrins [(Melchiori et al., 1995; Seftor et al., 1999; Van Belle et al., 1999); reviewed in (Kuphal et al., 2005a)]. The αvβ3 dimer that binds a range of ligands containing the amino acid sequence RGD (including fibronectin) appears to be important for the invasive potential of melanoma (Dang et al., 2006; Hsu et al., 1998; Van Belle et al., 1999). The main ECM components that integrins bind are collagen, laminin and fibronectin; peptides that block the ability of integrins to interact with laminin and fibronectin have been shown to reduce melanoma motility in vitro and metastatic spread in vivo (Kuratomi et al., 1999; Makabe et al., 1990). BRAF signalling is capable of promoting fibronectin expression (Gaggioli et al., 2007) and may also promote the expression of the fibronectin receptor component β3 integrin (Woods et al., 2001).

V600EBRAF promotes melanoma cell extravasation in vitro and subsequent development of lung metastasis in vivo (Liang et al., 2007). Further mechanistic studies have shown hints of how mutant BRAF enhances melanoma metastasis. V600EBRAF knockdown by RNA interference blocks melanoma cell invasion and decreases MMP-2 activity and β1 integrin expression (Sumimoto et al., 2004). V600EBRAF induces activity of MMP-1 (Huntington et al., 2004) and secretion of interleukin-8 (IL8; Liang et al., 2007), which eventually leads to enhanced melanoma cell invasion. V600EBRAF is also capable of inducing melanoma cell invasion, metastatic spread and colonization by downregulating the cGMP-selective phosphodiesterase PDE5A. V600EBRAF downregulates PDE5A via activation of the transcription factor BRN2, which binds to the promoter of PDE5A and suppresses its expression. The regulation of PDE5A by BRN2 might be one function of BRN2 that contributes to invasive behaviour in vivo (Pinner et al., 2009). As such, downregulation of PDE5A by V600EBRAF increases short- and long-term lung colonization in mice; and downregulation of PDE5A is observed in patients with melanoma metastasis compared to those with primary tumours (Arozarena et al., 2011b). Furthermore, miR-211, which normally represses BRN2 expression, is decreased in melanoma cells compared with melanocytes, resulting in increased BRN2 levels promoting the invasive potential of melanoma (Boyle et al., 2011).

Therefore, not only is V600EBRAF involved in the first steps in melanoma development, but also in the metastatic spread and distant organ colonization, providing a rationale and mechanistic basis for targeting V600EBRAF in patients both with primary and metastatic melanoma, which will be discussed later in the review.

PI3K pathway in invasion and metastasis

The phosphoinositide-3-OH kinase (PI3K) pathway is also another key signalling cascade that plays important roles in melanoma development (Madhunapantula and Robertson, 2009). PI3K is activated by receptor tyrosine kinases and RAS, in turn promoting conversion of phosphoinositide membrane lipids into second messenger phosphatidyl-inositol-3,4,5-triphosphate, which activate a number of downstream effector pathways. This signalling is terminated by the tumour suppressor phosphate and tensin homologue (PTEN; Shaw and Cantley, 2006). PI3K signalling controls cell survival, proliferation and motility, and is hyperactivated in a high percentage of melanomas. In particular, activating mutations of PI3K occur in 3% of metastatic melanomas, while mutations leading to loss of PTEN are found in 5–20% of late-stage melanomas (Wu et al., 2003). The PI3K effector protein kinase B (PKB, also known as AKT) is overexpressed in up to 60% of melanomas (Stahl et al., 2004).

Apart from controlling cell proliferation, PTEN regulates cell–cell adhesion, cell migration, and metastasis. Ectopic expression of PTEN protein can inhibit invasion of melanoma cells (Madhunapantula and Robertson, 2009; Wu et al., 2003). Interestingly, a recent study utilizing mouse models has shown that PTEN-loss and V600EBRAF cooperate to promote invasive, metastatic melanoma (Dankort et al., 2009). Upon PTEN-loss, AKT is activated and can control multiple cell processes by phosphorylating different substrates. AKT inhibits the small GTPase RhoB in melanoma, inducing tumour cell invasion and metastasis (Jiang et al., 2004). Even though there is evidence of PI3K activating Rac GTPase activity (Han et al., 1998; Nimnual et al., 1998; Rodriguez-Viciana et al., 1997), in melanoma cells AKT has been shown to phosphorylate serine 71 of the GTPase Rac1, which may inhibit GTP binding of Rac1 and eventually attenuate the Rac1-driven cell motility (Kwon et al., 2000).

Rho GTPases in invasion and metastasis

The Rho family of GTPases act as molecular switches involved in a wide variety of processes, including control of cell morphology and the actin cytoskeleton, adhesion, cell migration, cell proliferation and survival (Sahai and Marshall, 2002). Rho GTPases cycle between GTP-bound (active) and GDP-bound (inactive) states, and this cycling is regulated by activators, guanine nucleotide exchange factors (GEFs); and inactivators, GTPase accelerating proteins (GAPs). When active, Rho GTPases interact and activate downstream effector proteins to trigger a signalling cascade to direct cellular responses. Seminal studies on the three prototypical members of the Rho family of small GTPases showed that they each coordinate specific rearrangements of the actin cytoskeleton. Rac was shown to drive actin assembly to form lamellipodia, Cdc42 actin assembly for filopodia while Rho stimulates stress fibre formation both through actin assembly and the generation of actomyosin contractility (Etienne-Manneville and Hall, 2002). Some of the aforementioned pathways can activate Rho GTPases, such as MAPK and PI3K pathways (Sahai and Marshall, 2002). Later in the review there is a more detailed description on how Rho GTPases control cell movement, melanoma invasion and metastatic dissemination.

Even though Rho GTPases are related to Ras, traditionally mutations in Rho proteins have been thought to be relatively rare in tumours, whereas their expression and/or activity is indeed frequently altered, as well as that of some GEFs and GAPs (Sahai, 2005; Sahai and Marshall, 2002; Vega and Ridley, 2008). For example, RhoC was found to be overexpressed in in vivo-selected melanoma metastatic cells compared with the parental melanoma cells (Clark et al., 2000). Interestingly, two recent studies have shown that another frequently mutated gene in melanoma is Rac1 (Hodis et al., 2012; Krauthammer et al., 2012). Between 5 and 9% of melanomas harbour an oncogenic mutation changing Pro29 to serine, resulting in increased binding of Rac1 to downstream effectors and promoting melanocyte proliferation and migration (Krauthammer et al., 2012). This important finding suggests Rac1 and its downstream targets as potential new drug targets. Furthermore, mutations in Cdc42, Rac2 and RhoT1 have also been found in melanoma (Hodis et al., 2012). The Rac1 GEF – and also PTEN-interacting protein (Fine et al., 2009) – phosphatidylinositol-3,4,5-trisphosphate-dependent Rac exchange factor 2 (PREX2) has been found to be mutated in 14% of melanomas (Berger et al., 2012). These mutations generate truncated or variant PREX2 proteins with oncogenic activity in melanoma cells, promoting melanoma tumour formation. Apart from its role in melanomagenesis, PREX2 mutants could be involved in promoting melanoma cell migration and invasion via two possible mechanisms: activation of Rac1, given that PREX2 is a GEF for Rac1; and/or via down modulation of PTEN and consequently, hyperactivating PI3K pathway (Fine et al., 2009). These functions remain to be tested. Furthermore, PREX1, a related Rac GEF, has also been described to be important for melanoblast migration and melanoma metastasis in a Q61KNRAS/INK4a(−/−) mouse model (Lindsay et al., 2011).

These recent findings highlight the importance of GTPases in melanomagenesis and prove that Rho GTPases can be considered melanoma oncogenes.

JAK-STAT pathway in invasion and metastasis

Janus-activated kinases (JAKs) are membrane-bound proteins that transduce signals from a variety of receptors (growth factor and cytokine receptors) into cytoplasmic transcription factors signal transducer and activator of transcription (STAT), which translocate to the cell nucleus and directly bind DNA (Kortylewski et al., 2005). The balance between the activation of individual STAT factors determines unique gene expression profiles. Different STATs can have opposite roles: STAT1 is thought to be a tumour suppressor, even though in certain situations may play a paradoxical melanoma promoting role (Kortylewski et al., 2005), whereas STAT3 is considered an oncogene that inhibits apoptosis and promotes cell proliferation, angiogenesis, invasion and metastasis (Yu et al., 2009). STAT3 is constitutively activated in multiple tumour types, including melanoma (Yu et al., 2009) especially those of inflammatory origin. Most melanoma cell lines and clinical specimens display constitutive phosphorylation of STAT3 at the Tyr705 residue (Yu and Jove, 2004). Activated STAT3 induces expression of VEGF and hypoxia-inducible factor 1-α (HIF1α), thereby promoting tumour angiogenesis (Kortylewski et al., 2005; Niu et al., 2002). Furthermore, STAT3 stimulates invasion and metastasis by regulating the expression of matrix metalloproteinases MMP-1, MMP-2, and MMP-9 (Itoh et al., 2006; Song et al., 2003; Xie et al., 2004) and episialin/MUC1 (Gaemers et al., 2001), all of which have roles in tumour invasion. Importantly, STAT3 and JAK1 cooperate with Rho-ROCK in promoting actomyosin contractility in melanoma (Sanz-Moreno et al., 2011).

Wnt/β-catenin in invasion and metastasis

The Wnt/β-catenin pathway (or canonical Wnt pathway) is required for different cellular processes in melanocytes such as cell adhesion, proliferation, survival and migration. This is due in part to the direct transcriptional activation of MITF by β-catenin when Wnt signalling is activated (Takeda et al., 2000). β-catenin/Wnt pathway is frequently constitutively activated in melanoma, as it happens in other cancers. Mutations on both β-catenin and other Wnt pathway components can lead to dysregulation of β-catenin. β-catenin mutations were originally described in 23% of melanoma cell lines (Rubinfeld et al., 1997), and they have also been recurrently found in around 5% of primary melanomas (Demunter et al., 2002; Omholt et al., 2001; Reifenberger et al., 2002). Nuclear β-catenin, thought to reflect activation of canonical Wnt pathway, is found in approximately 33% of human melanoma patients (Rimm et al., 1999). A recent report utilizing PTEN/BRAF mouse models has unveiled that β-catenin is a key mediator of melanoma metastasis to lymph nodes and lungs, cooperating with PTEN-loss and V600EBRAF in promoting primary tumour and metastasis formation (Damsky et al., 2011). Regarding Wnt family members, Wnt5A has been implicated in melanoma progression, being downregulated in melanomas with better outcome and its overexpression correlates with more aggressive disease (Bittner et al., 2000; Weeraratna et al., 2002).

MITF in invasion and metastasis

Microphtalmia-associated transcription factor is a transcription factor considered to be the master regulator of melanocyte and melanoma cell biology. MITF drives the expression of melanogenic proteins and regulates melanoblast survival, melanocyte lineage commitment and cell invasion (Levy et al., 2006). MITF is expressed in most human melanomas and its target genes are diagnostic markers for this disease (Levy et al., 2006). MITF gene is amplified in 10% of primary melanomas and 21% of metastatic tumours (Garraway et al., 2005). MITF amplification seems to be a useful prognostic marker for metastatic melanoma, as high MITF amplification correlates with reduced disease-specific patient survival (Garraway et al., 2005; Ugurel et al., 2007). Continued MITF expression is necessary for melanoma cell proliferation and survival, but MITF protein levels and activity must be tightly controlled: very high levels promote cell cycle arrest and differentiation, whereas critically low levels lead to cell cycle arrest and apoptosis. Thus, only at intermediate levels proliferation is favoured (Levy et al., 2006). Furthermore, MITF is a key player in the control of melanoma invasion. Low MITF levels result in downregulation of the diaphanous-related formin Dia1, reorganization of the actin cytoskeleton, and increased Rho-associated protein kinase (ROCK)-dependent invasiveness. In contrast, high MITF expression decreases invasiveness in certain experimental conditions (Arozarena et al., 2011a; Carreira et al., 2006). MITF can also control melanoma invasion via direct regulation of expression of c-Met, the receptor of hepatocyte growth factor (HGF; McGill et al., 2006).

Other signalling pathways altered in melanoma metastasis

From the membrane to the nucleus, several signalling pathways have been recently involved in melanoma invasion and metastatic dissemination.

Secreted factors

A wide variety of growth factors, cytokines and their receptors are upregulated during metastatic progression of melanoma; examples are melanoma inhibitory activity (Bosserhoff, 2006), HGF and its receptor c-Met (McGill et al., 2006; Vergani et al., 2011), transforming growth factor (TGF)-β, interleukin 8, Nodal and several members of fibroblast growth factor (FGF) family (Al-Alousi et al., 1996; Albino et al., 1991; Gaggioli and Sahai, 2007; Topczewska et al., 2006). These factors have an autocrine effect on melanoma cells via upregulation of cell motility genes, and are also believed to act in a paracrine fashion on stromal cells, forcing them to produce pro-invasive molecules like HGF and tenascin C (De Wever et al., 2004; Gaggioli and Sahai, 2007). Conversely, expression of multifunctional protein pigment epithelium-derived factor (PEDF) is decreased in melanoma progression due to its anti-invasive and anti-angiogenic properties (Orgaz et al., 2009).

Some MMPs have also their expression and/or activity increased in invasive melanomas; particularly expression of MMP-1, MMP-2 and MMP-9 has been shown to correlate with an invasive phenotype (Hofmann et al., 2000; MacDougall et al., 1995; Overall and Lopez-Otin, 2002).

Membrane receptor signalling

The role of the stem cell factor tyrosine kinase receptor (RTK) c-Kit in melanoma had been uncertain until very recently. Early studies described that expression of c-Kit progressively was decreased during local tumour growth and invasion of human melanomas (Bar-Eli, 1997; Natali et al., 1992), partially due to the fact that enforced c-Kit expression significantly inhibited tumour growth and metastasis, via triggering apoptosis of these cells (Bar-Eli, 1997). However, activating mutations had been found in about 2% of metastatic melanomas (Willmore-Payne et al., 2005), and more recent genomic screen studies have reported activating mutations of c-Kit in 20–25% of melanomas (Carvajal et al., 2011). Activation of c-Kit results in activation of MAPK and PI3K/AKT pathways, and hence, increased cell migration and proliferation. Metastatic melanomas could also need to activate c-Kit to have MITF at low levels compatible with high invasion, given that c-Kit triggers dual phosphorylations, which couple activation and degradation of MITF (Wu et al., 2000).

Upregulation of integrin expression is associated with the acquisition of a more metastatic phenotype in melanoma (Marshall et al., 1998). Among the most consistent observation is the upregulation of the β3 subunit of the αvβ3 vitronectin receptor in VGP melanoma in situ [reviewed in (Kuphal et al., 2005a)]. β3 integrin overexpression in RGP melanoma cells induces conversion to VGP by enhancing the invasive growth from the epidermis into the dermis in three-dimensional skin reconstructs, preventing apoptosis of invading cells and increasing tumour growth in vivo (Hsu et al., 1998). For a more detailed description of other integrins overexpressed in melanoma, see reference (Kuphal et al., 2005a).

Mutations in neurofibromatosis type 2 (NF2) have been found in melanoma patients (Bianchi et al., 1994). The protein product of NF2 is called merlin (moesin-ezrin-radixin-like protein) and it shares significant homology to ezrin-radixin-moesin proteins. Merlin has been shown to bind numerous transmembrane and intracellular proteins, including β1 integrin (Obremski et al., 1998), paxillin (Fernandez-Valle et al., 2002), neural Wiskott–Aldrich Syndrome protein (N-WASP; Manchanda et al., 2005), microtubules (Xu and Gutmann, 1998), and calpain (Kimura et al., 1998). Merlin is believed to play an essential role in inhibiting tumour cell motility and invasion (Poulikakos et al., 2006). Loss of merlin destabilizes cadherin containing cell–cell junctions, increases Rac activity, lamellipodia formation and cell motility (Lallemand et al., 2003; Shaw et al., 2001).

Tyrosine kinase signalling

Src family kinases are pleiotropic kinases involved in cell migration, adhesion and invasion, as well as proliferation and survival (Yeatman, 2004). Upon interaction with transmembrane receptor tyrosine kinases and also integrins, it directly activates downstream effectors such as PI3K, AKT and STAT3, which eventually can promote expression of MMPs, VEGF and IL8, among other targets (Yeatman, 2004; Zhang and Yu, 2012). Activation of Src family kinases is found in melanomas and other tumour types (Summy and Gallick, 2003).

Neural precursor expressed, developmentally downregulated 9 (NEDD9) is an adaptor protein that belongs to the Cas family of signal transduction molecules. NEDD9 is amplified at the genetic level in melanoma, and elevated expression levels have been shown to correlate with melanoma progression and metastasis (Kim et al., 2006).

Somatic mutations leading to inactivation of the serine/threonine kinase STK11/LKB1 occur in 10% of cutaneous melanomas and result in highly metastatic tumours (Liu et al., 2012).

Transcription factors

The Snail family of transcription factors Snail and Slug are also upregulated in melanoma, as they decrease expression of cell adhesion receptors that favour interaction with keratinocytes, such as E-cadherin and occludin, and upregulate N-cadherin to allow melanoma cells to interact with fibroblasts (Gaggioli and Sahai, 2007; Kajita et al., 2004). In addition, Snail enhances the expression of pro-invasive factors such as MMP-2 secreted protein acidid and rich in cysteine, tissue inhibitor of metalloproteinases-1 TIMP-1 and RhoA (Kuphal et al., 2005b).

Melanoma cell plasticity in migration and invasion

It is now accepted that tumour cells have multiple forms of movement. Cancer cells can move as collective groups or as individual cells (Friedl, 2004). Collective migration is classified into subclasses, chain migration and collective invasion. In chain migration, individual cells form transient cell–cell contacts and move along a track (Teddy and Kulesa, 2004). In collective invasion, cell–cell contacts remain intact as cells invade. Collective invasion can occur in many forms including irregularly shaped masses, multicellular tubes, strands, or sheets (Friedl and Wolf, 2009). While collective cell movement permits entry into the lymphatic system, individual cell movement is necessary for tumour cells to cross basement membranes and enter blood vessels to enable dissemination to distant organs (Giampieri et al., 2009). Therefore, in this review, we will focus on individual cell migration, for more information on collective cell migration see reviews (Friedl and Gilmour, 2009; Ilina and Friedl, 2009).

Actomyosin contractility as a key regulator of melanoma invasion

The most studied modes of individual migration are the so-called elongated-mesenchymal mode and rounded-amoeboid mode. Other strategies used by tumour cells are also emerging [reviewed in (Bustelo, 2012)], but we will focus on these two main types. Both actin assembly and actomyosin contractility contribute to determining cell morphology. Actin assembly is involved in the generation of localized structures such as protrusions and adhesions between cells and to the matrix. Actomyosin filaments under the plasma membrane form a cortical structure that provides the cell with structural integrity [reviewed in (Sanz-Moreno and Marshall, 2010)]. Actomyosin contractility generates physical force that creates tension within the cell. Thus cell morphology is determined by the balance of adhesion and tension. On rigid substrates where adhesion to the substrate is high, cells can spread and have an elongated morphology but on deformable substrates, the high tension generated by actomyosin contractility leads to a rounded morphology. Actin assembly to generate protrusions and actomyosin contractility to create traction forces are key processes in individual cell migration (Ridley et al., 2003). Actomyosin contractility can be used by cells to squeeze through voids in 3D matrices or can generate blebs in the plasma membrane through hydrostatic pressure breaking the linkage between plasma membrane and cortical actin (Charras et al., 2006; Lammermann and Sixt, 2009). Therefore, in movement driven by high levels of actomyosin contractility, cells are less adhesive and have a rounded shape, so we will call them rounded moving cells (Figure 1). In contrast, mesenchymal cells have an elongated shape, move by forming integrin-mediated adhesions that provide traction (Ballestrem et al., 2001; Tamariz and Grinnell, 2002) and have a higher requirement of peri-cellular degradation via protease remodelling of the ECM (Wolf et al., 2003). We will refer to this type of motion as elongated movement (Figure 1). Elongated movement is known to require lower levels of actomyosin contractility (Sanz-Moreno et al., 2008), but a certain degree of actomyosin contractility is also needed to efficiently retract the long lived protrusions that characterize elongated moving cells (Li et al., 2011).

Figure 1.

Modes of individual cancer cell movement. Melanoma cells can move and disseminate through the body using different strategies that are interconvertible depending on the microenvironment. Pictures of A375M2 cells (left) or A375P cells (right) on top of a collagen I matrix. Upper panels: bright field images. Lower panels: confocal images after phalloidin staining.

Mechanistically, asymmetry is a prerequisite of all cell movement (Petrie et al., 2009). For elongated moving cells, lamellipodia or protrusion formation at the front of the cell determines the direction of movement. The front of the cell adheres to the substrate in focal adhesions through integrins. On the other hand, blebbing rounded melanoma cells are polarized at the cell rear, with a uropod-like structure that determines the direction of movement similar to leucocyte movement (Lorentzen et al., 2011).

In melanoma, elongated/mesenchymal-type motility, rounded/amoeboid-type motility and collective movement have all been described (Friedl and Gilmour, 2009; Sahai and Marshall, 2003; Sanz-Moreno et al., 2008). Furthermore, melanoma cells appear to be highly plastic and can convert between different modes of movement (Charras et al., 2006; Hegerfeldt et al., 2002; Sahai and Marshall, 2003; Sanz-Moreno et al., 2008). As such, histopathological studies from human samples have shown that both rounded- and spindle-shaped cells are observed in malignant melanoma (Rosai and Ackerman, 2004), corresponding to the cell shapes observed in rounded-amoeboid motility and elongated-mesenchymal motility, respectively. Furthermore, the dermal invasive front of the majority of primary melanomas consists of cells with a rounded morphology. Round cells predominate in the invasive front even when the cells in the body of the tumour have an elongated morphology (Sanz-Moreno et al., 2011), showing that plasticity in human melanoma might be regulated by the tumour microenvironment, independently of genetic background.

Seminal studies on tumour cell plasticity showed that treatment with pericellular protease inhibitors resulted in conversion of the elongated mode into the rounded mode of movement in certain environments (Wolf et al., 2003), explaining in part why therapies targeting proteases have failed in the clinic. Cell migration in vivo and efficient metastasis may benefit from tumour cell plasticity. When measured in vivo rounded tumour cell movement can be 10–100 times faster than Rac-dependent protrusive movement of around 100 μm per day. While the rounded form of movement can be much faster, movement through a rigid ECM requires extracellular proteolysis; therefore, the ability to convert to this form of movement is essential where tumour cells meet such barriers (Demou et al., 2005; Sabeh et al., 2009; Van Goethem et al., 2010). High actomyosin contractility may also provide mechanical strength to resist shear forces in the bloodstream (Calvo et al., 2011; Pinner and Sahai, 2008; Sanz-Moreno et al., 2008), and amoeboid-rounded cells have been described to express specific blood cell cytoskeletal proteins (Charras et al., 2006). Melanoma cells, as well as other tumour types, can use multiple modes of migration depending on environmental conditions, so that blockade of alternative modes of movement might be required to stop melanoma dissemination (Ahn et al., 2012; Sahai and Marshall, 2003; Sanz-Moreno et al., 2008).

Signalling pathways regulating melanoma cell plasticity

In the search for new therapeutic targets to stop invasion of melanoma cells, intense research has been carried out in the last years focused on elucidating signalling pathways involved in controlling migration and invasion. Rho-family GTPases are key regulators of cytoskeletal dynamics and actomyosin contractility. Therefore, much of the work on the mechanisms underlying different modes of cell migration has focused on this family of GTPases. Once activated by the GEFs, Rho GTPases signal through effector pathways. Rounded, amoeboid contractile forms of movement are driven by high actomyosin contractility through Rho and Cdc42 signalling, while Rac signalling is required for actin assembly in elongated-protrusive movement (Sahai and Marshall, 2003; Sanz-Moreno et al., 2008; Wilkinson et al., 2005; Yamazaki et al., 2009). Figure 2 summarizes the main signalling pathways and key molecules that regulate plasticity in melanoma cell invasion. Some of the genes controlling such plasticity have been discussed in the previous section. It remains to be elucidated if other melanoma metastasis regulators also control such tumour cell plasticity.

Figure 2.

Signalling pathways that regulate plasticity in melanoma cell invasion. Schematic diagram outlining signalling pathways that control the interconversion between modes of individual cell movement in melanoma. Two main branches regulating rounded or elongated movement can be highlighted: rounded invasion is controlled by Rho/ROCK with the cooperation of JAK1/STAT3, whereas the Rac1 branch, including also Src and NEDD9, drives elongated invasion.

Rho GTPases

RhoA and RhoC appear to be important in rounded movement, while RhoA has also been implicated in elongated movement (Clark et al., 2000; Sahai and Marshall, 2003; Stoletov et al., 2007). Downregulation of SMURF1, an ubiquitin ligase that targets RhoA for degradation at Rac-dependent protrusions results in conversion from elongated to rounded movement (Sahai et al., 2007). Low levels of the cell cycle inhibitor p27Kip1 promote the rounded form of movement (Berton et al., 2009), as p27Kip1 can bind RhoA in the cytoplasm and prevent its activation (Besson et al., 2004; Hoshino et al., 2011). Rho activates Rho-kinases, ROCK I and II, which generate actomyosin contractility through phosphorylation and inactivation of the myosin phosphatase target subunit 1 (MYPT1; Jacobelli et al., 2009; Sahai and Marshall, 2003; Wilkinson et al., 2005). Inactivation of MYPT promotes myosin light chain II (MLC2) phosphorylation and this is a major driver in melanoma rounded movement. RhoE inhibits ROCK, but this is opposed by 3-phosphoinositide-dependent kinase-1 (PDK1), which has been shown to positively regulate the rounded form of movement in melanoma both in vitro and in vivo (Pinner and Sahai, 2008). Activation of the formin-like 2 (FMNL2) that assembles cortical actin is a specific RhoC driven pathway in rounded movement (Kitzing et al., 2010). The role of formins in the rounded form of movement is further supported by the observation that Dip, an interaction partner of the formin mDia2, promotes the blebbing characteristic of rounded cell movement (Eisenmann et al., 2007).

Actomyosin contractility can also be generated through Cdc42 signalling through the kinases myotonic dystrophy kinase-related Cdc42-binding kinase (MRCK; Wilkinson et al., 2005) or p21 protein (Cdc42/Rac)-activated kinase 2 (PAK2; Gadea et al., 2008). Cdc42 is required for both elongated and rounded movement. Rounded movement in melanoma cells involves the activation of Cdc42 through the GEF dedicator of cytokinesis 10 (DOCK10) and the Cdc42 effectors N-WASP and PAK2 (Gadea et al., 2008), while the GEF and effectors for elongated movement remain to be discovered. Furthermore, RasGRF2, a Ras GEF, suppresses rounded movement by inhibiting the activation of Cdc42 independently of its capacity to activate Ras. RasGRF1/2 directly binds to Cdc42, outcompeting Cdc42 GEFs, thereby preventing Cdc42 activation in melanoma cells (Calvo et al., 2011).

Rac1 activity is key for the interconversion or plasticity between modes of movement (Sanz-Moreno et al., 2008; Yamazaki et al., 2009). In melanoma, activation of Rac through the Rac GEF DOCK3 promotes elongated movement and suppresses rounded movement. As mentioned earlier, another Rac GEF, PREX, has been involved in melanoma metastasis, but whether these two GEFs are cooperating remains to be tested. On the other hand, Rac activity is low in rounded movement through high actomyosin contractility activating a Rac-specific GAP, ARHGAP22 (Sanz-Moreno et al., 2008). Localization of active Rac to protrusions is dependent on Rab5-dependent endocytosis (Palamidessi et al., 2008). Rac signals through WAVE2 to promote actin assembly and protrusion formation (Sanz-Moreno et al., 2008; Yamazaki et al., 2009). WAVE2 also has an important role in regulating rounded movement by suppressing actomyosin contractility (Sanz-Moreno et al., 2008).

Nevertheless, therapies aiming at blocking melanoma invasion will have to stop plasticity of tumour cells by blocking both Rho and Rac-driven mechanisms (Ahn et al., 2012; Sahai and Marshall, 2003; Sanz-Moreno et al., 2008). Since Cdc42 is involved in the two alternative types of individual movement and as activating mutations in the Cdc42 gene have been recently found in melanoma (Hodis et al., 2012), Cdc42 effector kinases could be attractive therapeutic targets.

Other regulators

Progress has been made in understanding intracellular pathways regulating alternative modes of cell migration. Information about the extracellular signals and their receptors that regulate modes of migration is provided by some studies. Receptor tyrosine kinase signalling has been linked to Rac-dependent elongated movement but interestingly overexpression of EphA2 seems to be involved in the rounded form of movement (Parri et al., 2009). On the other hand, cytokine signalling through the receptor subunit GP130-IL6ST and the kinase JAK1 generates actomyosin contractility through Rho-kinase-dependent signalling (Sanz-Moreno et al., 2011). This pathway generates high levels of actomyosin contractility required for migration of melanoma cells in the rounded mode of movement (Sanz-Moreno et al., 2011).

On the other hand, the adaptor molecule NEDD9, mentioned earlier to be a melanoma metastasis gene (Kim et al., 2006), interacts with the GEF DOCK3 to promote Rac activation and the elongated movement and invasion in melanoma (Sanz-Moreno et al., 2008). Increased NEDD9 leads to elevated phosphorylation of β3 integrin which results in increased Src and focal adhesion protein (FAK) to drive elongated mode of invasion only upon vitronectin binding (Ahn et al., 2012). NEDD9 overexpression decreases ROCKII signalling through Src-dependent phosphorylation of the negative regulatory site Tyr722. In NEDD9-overexpressing melanoma cells, inhibition of Src with dasatinib results in a switch from Rac-driven elongated, mesenchymal-type invasion to ROCK-dependent rounded, amoeboid invasion. These findings suggest that a combination treatment of dasatinib and a ROCK inhibitor might be a better alternative to inhibit both elongated and rounded modes of invasion (Ahn et al., 2012).

As mentioned before, PEDF is a multifunctional secreted glycoprotein that displays broad antitumour activity based on dual targeting of the tumour microenvironment (anti-angiogenic action) and the tumour cells (direct antitumour action). Pigment epithelium-derived factor blocks tumour extravasation by suppressing rounded contractile movement and elongated-mesenchymal driven proteolysis (Ladhani et al., 2011).

As melanomas have frequent mutations in the MAPK pathway, it is starting to be clear that tumour cell plasticity may also be controlled by RAF or Ras signalling. In those lines, melanomas with BRAF mutations showed distinct morphological features such as larger, rounder, and more pigmented tumour cells (Viros et al., 2008). The same correlation was not found in melanomas with N-Ras mutations. An additional mechanism that melanoma cells may use to regulate actomyosin contractility is downregulation of PDE5A (Arozarena et al., 2011b). PDE5A ablation leads to an increase in cGMP, which augments cytosolic Ca2+. Cytosolic Ca2+ stimulates phosphorylation of MLC2, thereby increasing actomyosin contractility (Somlyo and Somlyo, 2003) and eventually enhancing cell invasion (Arozarena et al., 2011b). The kinase responsible for these effects remains to be investigated, but it is most likely that myosin light chain kinase is involved in such cGMP-mediated effects. Therefore, BRAF could be regulating actomyosin contractility through PDE5A (Arozarena et al., 2011b) and in that manner actomyosin contractility could be driving rounded movement in BRAF-mutant melanomas. On the other hand, MEK inhibition by selumetinib (AZD6244, ARRY-142886) or PD184352, suppressed actin-cortex contraction, but increased integrin-mediated adhesion in vitro (Ferguson et al., 2012). MEK inhibition in melanoma cells induced an elongated phenotype with a significant increase in MMP-2 and membrane-type 1-MMP expression (Ferguson et al., 2012). BRAF has been described to regulate MMP-2 (Sumimoto et al., 2004), so it remains to be elucidated how MEK inhibitors are mediating these effects and if BRAF inhibitors will also promote elongated movement.

Unexpectedly loss of tumour suppressor p53 function has been linked with promotion of rounded movement, suggesting unanticipated links between p53 signalling and cell movement (Gadea et al., 2007). It will be important to assess if other tumour suppressors relevant in melanoma progression are also regulating tumour cell plasticity.

To gain a better understanding of melanoma heterogeneity, it would be useful to elucidate or define genetic signatures or programs for rounded and elongated types of movement to validate them as biomarkers for metastatic dissemination. Such signatures most likely will be regulated by the Rho GTPases themselves and will help in the design of tailored therapies for patients. Since the recent discovery of mutations in Rho GTPases, patients carrying such activating mutations would be suited for treatment with drugs targeting signalling pathways or genes regulated by the Rho GTPases.

Angiogenesis and vasculogenic mimicry in melanoma metastasis

Angiogenesis is the formation of new blood vessels from pre-existing ones, and is a hallmark of cancer (Hanahan and Weinberg, 2011). Primary tumours require activating angiogenesis (the so-called tumour angiogenic switch) to have an adequate supply of nutrients and oxygen. These very same blood vessels, along with lymphatic vessels, are used by tumour cells to enter into the general circulation and reach other organs to form metastases (Carmeliet and Jain, 2000; Folkman, 1971).

Melanoma cells have been described to secrete a number of angiogenic growth factors such as VEGF, placenta growth factor (PlGF), bFGF, TGF-α and TGF-β, IL8 and platelet-derived growth factor (PDGF-B), among others (Mahabeleshwar and Byzova, 2007; Rofstad and Halsor, 2000). Several studies have shown positive correlation between melanoma neovascularization with poor patient prognosis, overall survival, ulceration and a higher rate of relapse (Mahabeleshwar and Byzova, 2007).

Apart from inducing neoangiogenesis, aggressive melanoma cells can engage in vasculogenic mimicry by expressing endothelial-associated genes, such as vascular endothelial cadherin (VE-cadherin), leading to the formation of an extracellular fluid-conducting network in aggressive tumours and which correlate with poor clinical outcome (Hendrix et al., 2003; Maniotis et al., 1999). Pigment epithelium-derived factor knockdown in poorly aggressive melanoma cell lines not only augmented migration and invasion, but also increased vasculogenic mimicry (Orgaz et al., 2009). These effects translated into an increased in vivo metastatic potential (Orgaz et al., 2009), indicating again that delivery of PDEF to tumours could be considered in the clinic. Vasculogenic mimicry can occur at the same time as angiogenesis, vessel co-option and intussusceptive microvascular growth (Hendrix et al., 2003). Therefore, vasculogenic mimicry is another example of plasticity of melanoma cells that allows them to adapt to hypoxic microenvironments and also to escape from conventional anti-angiogenic therapies, which target endothelial cell-specific neovascularization.

Clinical opportunities

Melanoma is an extremely aggressive, highly metastatic cancer with a notoriously high resistance to cytotoxic agents (Gray-Schopfer et al., 2007a). Both chemoresistance and high metastatic potential are thought to be caused due to the fact that melanocytes derive from highly motile neural crest precursors that have enhanced survival properties (Gray-Schopfer et al., 2007a; Soengas and Lowe, 2003; Thomas and Erickson, 2008). Most chemotherapeutic drugs act through induction of apoptosis in malignant cells, but melanoma cells have low spontaneous apoptosis in vivo compared with other tumour types and are relatively refractory to drug-induced apoptosis in vitro; therefore, resistance to apoptosis is probably the main cause for chemoresistance in melanoma (Soengas and Lowe, 2003). Several approved postoperative adjuvant therapies for malignant melanoma include interferon-γ, IL2, dacarbazine (DITC), carmustine, paclitaxel (Taxol), temozolomide and cisplatin, among others, even though these therapies contribute little to overall patient survival [reviewed in (Gray-Schopfer et al., 2007a; Tarhini and Agarwala, 2006)]. It will be important to assess if chemoresistance and invasive behaviour are linked by some common regulators and explore them as therapeutic targets.

Identification of signalling pathways that are essential not only for melanoma initiation but also for metastatic progression is providing the opportunity to develop targeted therapies and to design treatments specific to certain patients according to the genetic lesions that underlie their individual disease (Gray-Schopfer et al., 2007a). However, given the multiplicity and redundancy of disrupted molecular pathways in melanoma (like other tumours) and the plasticity in terms of tumour cell migration, concurrent, sequential and/or combined treatments would be needed to produce effective, actual and long-lasting benefit for patients (Sekulic et al., 2008). For further reading, interested readers are referred to several excellent recent reviews (Flaherty et al., 2012; Tsao et al., 2012).

Targeting MAPK pathway

The most appealing cancer targets are those on which the cancer cells are highly dependent for progression, the so-called concept of oncogene addiction. This dependence cancer cells have on hyperactivated pathways provides a therapeutic opportunity because cancer cells would be more sensitive to inhibition of those pathways than normal cells. Enzymes such as kinases, proteases and phosphatases are promising targets because their catalytic sites include deep pockets into which carefully tailored drugs can bind quite selectively (Gray-Schopfer et al., 2007a).

Drugs targeting the MAPK pathway and particularly BRAF, given its dual role in melanoma initiation and metastatic progression, are of considerable interest. The multi-kinase inhibitor sorafenib (BAY 43-9006) targets BRAF, CRAF, VEGF receptors 1, 2 and 3, PDGF receptor, Flt-3, p38, c-Kit and FGFR-1 (Wilhelm et al., 2004) and therefore inhibiting tumour growth, angiogenesis and metastatic spread. As a single agent, it has little or no antitumour activity in advanced melanoma patients (Eisen et al., 2006), but in combination with carboplatin and paclitaxel more encouraging results were obtained (Flaherty, 2006). It is not clear why sorafenib monotherapy is so inefficient in melanoma, it could be due to survival escape routes provided by other pathways such as tumour-necrosis factor-α when BRAF is inhibited (Gray-Schopfer et al., 2007a,b).

Although BRAF is the current focus of RAF drug development, the other RAF isoforms should not be overlooked, particularly because BRAF activates CRAF in mammalian cells (Garnett et al., 2005). In fact, drugs that selectively inhibit BRAF in the presence of oncogenic RAS drive RAS-dependent BRAF binding to CRAF, and consequently CRAF activation and MEK-ERK signalling (Heidorn et al., 2010). Elevated CRAF has also been suggested as a possible mechanism underlying acquired resistance to BRAF inhibition in melanoma (Montagut et al., 2008), and low activity G469E and D594G BRAF mutants were found to be highly reliant on CRAF activity while displaying resistance to MEK inhibition (Smalley et al., 2009). These findings highlight the importance of tailoring treatments according to the genetic lesions in the patients. For example, between 15 and 30% of patients treated with type I BRAF inhibitors such as vemurafenib and dabrafenib (GSK-2118436) develop non-melanoma skin cancers (cutaneous squamous-cell carcinomas and keratoacanthomas). This is due to the presence of RAS mutations in around 60% of cases in patients treated with vemurafenib, which accelerates the progression of pre-existing subclinical non-melanoma cancerous lesions. Therefore, caution must be taken when single-agent BRAF inhibitor treatments are administered to patients with cancers driven by RAS and with strong upstream MAPK signalling. These findings also point to the usefulness of combined BRAF-MEK inhibition to prevent this toxic effect (Su et al., 2012), particularly as mutant BRAF melanoma cell lines and in vivo models are especially dependent on MEK/ERK signalling (Solit et al., 2006). In line with this, alternative drugs targeting MEK, such as PD0325901 and AZD6244, and which inhibit not only proliferation but also invasion in vitro, have been developed and are being tested in clinical trials (Gray-Schopfer et al., 2007a). It will be important to assess if these drugs affect metastatic dissemination and tumour plasticity in vivo.

NRAS is also another therapeutic target whose potential has been validated in preclinical models with siRNAs (Kelleher and McArthur, 2012). So far there are not potent and selective pharmacological inhibitors, in part due to its GTPase activity, which has prevented the design of specific small-molecule antagonists (Flaherty et al., 2012). Ras farnesyl transferase inhibitors, such as R115777 (tibifarnib), were the first to enter the clinic; however, they are not very specific and did not yield promising results in clinical trials (Ji et al., 2012; Kelleher and McArthur, 2012). Given the dependence oncogenic RAS has on MEK/ERK and PI3K signalling, therapies utilizing combinations of inhibitors targeting those pathways are likely to result in better outcomes [reviewed in (Ji et al., 2012; Kelleher and McArthur, 2012)].

Targeting PI3K pathway

Targeting PI3K-PTEN signalling would likely to be a good approach for the treatment of melanoma, particularly when used in combination with other agents targeting RAS, RAF or MEK (Dancey, 2006). Among the agents targeting PI3K, PKB and other downstream components such as mammalian target of rapamycin (mTOR), mTOR inhibitors such as CCI-779 (temsirollimus) or RAD001 (everolimus) are the most advanced (Dancey, 2006).

Targeting receptor tyrosine kinase signalling

Activating mutations of the RTK c-Kit are relatively common in melanoma (Flaherty et al., 2012). Treatment of c-Kit-mutant melanoma cells with the receptor tyrosine kinase inhibitor imatinib (Gleevec) greatly inhibits their proliferation and induces apoptosis (Jiang et al., 2008). Even though its clinical testing has yielded only moderate responses limited to a subset of patients carrying certain c-Kit mutations [reviewed in (Flaherty et al., 2012)], other RTK inhibitors such as sunitinib, nilotinib and dasatinib are being tested and could be more effective (Tsao et al., 2012).

Several clinically applicable small-molecule Src inhibitors have been developed and are undergoing clinical testing in melanoma and other tumours, such as dasatinib (Sprycel), saracatinib and bosutinib (Zhang and Yu, 2012). However, results from Src inhibitor monotherapy have not been very promising so far (Mayer and Krop, 2010).

Given the role of cytokine signalling in melanoma invasion (Pedranzini et al., 2006; Sanz-Moreno et al., 2011), therapeutic agents such as blocking antibodies against cytokines (Rose-John et al., 2007), and small-molecule inhibitors of JAK kinase (Pedranzini et al., 2006) or STAT activity (Yue and Turkson, 2009) may be useful agents to block invasion and metastasis (Sanz-Moreno et al., 2011). Some of these agents are already in clinical trials for halting the growth of the primary tumour in other malignancies. Therefore, they seem attractive drugs to stop both tumour growth and metastatic dissemination of melanoma cells.

Blocking integrin activity

Inhibition of integrin activation may have potential in melanoma treatment, especially in preventing angiogenesis and metastasis. Available inhibitors of integrins are functionally blocking antibodies, peptide antagonists and matrix-mimicking cyclic peptides (Kuphal et al., 2005a). Integrin αvβ3 can be targeted by the humanized monoclonal antibody vitaxin, and is currently in clinical trials (Gutheil et al., 2000; Patel et al., 2001; Posey et al., 2001). CNTO 95 is another human monoclonal antibody that inhibits αv integrins with showed promising preclinical results (Trikha et al., 2004).

Delivery of metastasis suppressors

Delivery of metastasis suppressors such as PEDF (Ladhani et al., 2011; Orgaz et al., 2009) or PDE5A (Arozarena et al., 2011b) has been discussed earlier, and they could be considered for future strategies.

Targeting Rho GTPase signalling

Rho GTPases are key players in metastatic dissemination. As Rac1 and Cdc42 have been found mutated in some melanoma patients (Hodis et al., 2012; Krauthammer et al., 2012) and their effectors include various protein kinases, they offer pharmacological inhibition opportunities. Interestingly, Rac activity has been shown to be important for melanoblast cell cycle progression (Li et al., 2011) and for melanocyte proliferation (Krauthammer et al., 2012), proving that Rac signalling could be targeted to stop mesenchymal invasion and proliferation of melanoma cells. For example, inhibiting members of the PAK family of protein kinases might be of therapeutic benefit in the treatment of melanoma patients that harbour Rac or Cdc42 mutations. RhoC seems to be a potent driver of melanoma metastasis (Clark et al., 2000), so that Rho effector kinases should also be considered, as we will discuss later.

Targeting angiogenesis

As aforementioned angiogenesis has a critical role in melanoma. The VEGF receptor inhibitors SU5416 (semaxanib) and AG013736 have been described to exert antitumour activity in melanoma xenografts in mice (Gray-Schopfer et al., 2007a; Peterson et al., 2004). However, targeting angiogenesis alone would probably be an ineffective treatment for melanoma, given the poor response to sorafenib, thought to work through an anti-angiogenic mechanism in melanoma.

Targeting actomyosin contractility in combinatorial therapies

It has been noted that successful melanoma treatments would probably entail targeting several pathways together. For example, failure of therapies targeting Src could be explained by recent reports showing that the Src kinase inhibitor dasatinib blocks elongated motility, but promotes rounded motility driven by actomyosin contractility (Ahn et al., 2012), consistent with previously published findings of Src kinase inhibitor PP2 (Carragher et al., 2006). It would be interesting to combine integrin blocking antibodies, dasatinib and ROCK inhibitors. Dasatinib and ROCK inhibitors should inhibit elongated and rounded motility (Ahn et al., 2012), and integrin blocking antibodies may contribute to blocking elongated, mesenchymal-type motility and target tumour-associated vessels. Naturally, there is great interest in developing ROCK inhibitors for clinical use due to promising preclinical data in cardiovascular disease and glaucoma. The great abundance of compounds targeting ROCK may enable pairing them to Src inhibitors.

As explained earlier, actomyosin contractility is necessary for both rounded-blebbing moving cells and for the retraction of long lived protrusions needed for efficient elongated-mesenchymal moving cells. Therefore, it is becoming apparent that ROCK is a good therapeutic target that should be re-considered. Some ROCK inhibitors are fasudil (HA-1077; Davies et al., 2000), currently being tested for use in acute stroke and pulmonary artery hypertension and the only ROCK inhibitor approved for human use (Liao et al., 2007); hydroxyfasudil (HA-1100) is the major active metabolite following oral administration of fasudil and has greater selective inhibitory effect on ROCK (Higashi et al., 2003; Shimokawa et al., 1999); H1152 (H-1152P), which is the most potent inhibitor of the three (Sasaki et al., 2002); Y-27632 is a potent ROCK inhibitor that blocks smooth muscle contractility and normalizes blood pressure in a non-specific manner for the two ROCK isoforms (Ishizaki et al., 2000). However, it also inhibits members of Rho-dependent PRK family (Davies et al., 2000). A closely related compound, Y-32885, has been developed more recently. It has similar potency on ROCKs and protein kinase C-related kinase 1 (Loge et al., 2002). Two aminofurazan-based inhibitors GSK269962A and CSB772077B have been shown to inhibit both ROCKs more potently that Y27632 and fasudil, especially on ROCKI (Doe et al., 2007). Currently, more ROCK inhibitors are being developed and will need to be tested for their efficiency in blocking metastatic dissemination (Vigil et al., 2012).

Other drug combinations with dasatinib are being explored in melanoma with most combinations aimed at targeting survival and growth. These combinations include a phase I/II study using dasatinib with dacarbazine (Montero et al., 2011) or dasatinib with PI3K inhibitors, and platinum compounds like cisplatin (Homsi et al., 2009).

Matrix metalloproteinase also seemed to be attractive cancer targets, and two decades ago different drugs were developed to block their matrix-degrading activities and entered clinical trials (Overall and Kleifeld, 2006). However, trials yielded disappointing results, highlighting the need for better understanding into the mechanisms by which MMPs contribute to tumour growth and metastasis (Overall and Lopez-Otin, 2002). MMPs not only degrade ECM but they also exert essential signalling functions (Egeblad and Werb, 2002). Some MMPs have tumourigenic and prometastatic actions, whereas other MMPs display tumour-suppressive effects (Lopez-Otin and Matrisian, 2007). Therefore, the initial strategy of broadly blocking MMPs to abrogate invasion into basement membranes and tumour dissemination, has inevitably failed in patients, due partially to unexpected loss of these antitumour actions of tumour-suppressive MMPs (Overall and Kleifeld, 2006). New therapeutic approaches blocking specific MMPs or their targets should prove more effective in cancer treatment (Egeblad and Werb, 2002; Lopez-Otin and Matrisian, 2007; Overall and Kleifeld, 2006; Overall and Lopez-Otin, 2002). On the other hand, protease inhibition in mesenchymal moving cells leads to cells switching due to plasticity to a rounded movement by an as yet unknown mechanism (Sahai and Marshall, 2003; Wolf et al., 2003). This compensation mechanism has been suggested to explain in part the failure of MMP targeting therapies. Therefore, combined protease inhibitors and actomyosin contractility inhibitors, such as the above-mentioned ROCK inhibitors, have also been suggested to stop invasion (Sahai and Marshall, 2003).

The final question that arises is when we should start treating the patients to block metastatic dissemination. Early diagnosis and biomarker discovery combined with gene sequencing will help decide which patients might benefit from combined treatments to target tumour growth and dissemination of melanomas. This area of research might bring important advances for melanoma targeted therapies.

Concluding remarks

Intensive molecular research during the last decades has led to the identification of a myriad of key molecular pathways involved in the pathogenesis and progression of melanoma. This research has also highlighted the fact that melanoma is a heterogeneous group of diseases that arise from complex molecular changes. Therefore, better molecular disease classification methods would help design more efficient individually tailored targeted therapies for this devastating disease. Successful design of such treatments will involve a combination of drugs that will target the multiple signalling pathways involved in tumour growth and metastatic spread. Emphasis should be put in finding signalling modules that regulate both tumour growth and dissemination or signalling pathways that regulate multiple modes of migration at once. Such good therapeutic targets would minimize the amount of drugs administered to each patient.

Acknowledgements

This work was supported by Cancer Research UK Grant C33043/A12065 and Royal Society RG110591. VS-M is a CRUK Career Development Fellow. We apologize to those authors whose work could not be cited due to space constraints.

Ancillary