Cellular and molecular mechanisms controlling the migration of melanocytes and melanoma cells

Authors

  • Jacky Bonaventure,

    1. Developmental Genetics of Melanocytes, Institut Curie, Centre de Recherche, Orsay, France
    2. CNRS UMR3347, Orsay, France
    3. INSERM U1021, Orsay, France
    Search for more papers by this author
  • Melanie J. Domingues,

    1. Developmental Genetics of Melanocytes, Institut Curie, Centre de Recherche, Orsay, France
    2. CNRS UMR3347, Orsay, France
    3. INSERM U1021, Orsay, France
    Search for more papers by this author
  • Lionel Larue

    Corresponding author
    1. CNRS UMR3347, Orsay, France
    2. INSERM U1021, Orsay, France
    • Developmental Genetics of Melanocytes, Institut Curie, Centre de Recherche, Orsay, France
    Search for more papers by this author

CORRESPONDENCE L. Larue, e-mail: lionel.larue@curie.fr

Summary

During embryonic development in vertebrates, the neural crest-derived melanoblasts migrate along the dorsolateral axis and cross the basal membrane separating the dermis from the epidermis to reach their final location in the interfollicular epidermis and epidermal hair follicles. Neoplastic transformation converts melanocytes into highly invasive and metastatic melanoma cells. In vitro, these cells extend various types of protrusions and adopt two interconvertible modes of migration, mesenchymal and amoeboid, driven by different signalling molecules. In this review, we describe the major contributions of natural mouse mutants, mouse models generated by genetic engineering and in vitro culture systems, to identification of the genes, signalling pathways and mechanisms regulating the migration of normal and pathological cells of the melanocyte lineage, at both the cellular and molecular levels.

Introduction

At a given time during embryonic development, cells migrate, in a complex and heterogeneous process underlying morphogenesis and tissue maintenance. Migration of most cells including stromal, epithelial and neuronal cells normally ceases with their terminal differentiation, but the process can be reactivated during tissue regeneration or tumour invasion. In this review, we focus principally on the migration of melanocytes and their precursors, melanoblasts, during normal embryonic development in vertebrates. We also describe the cellular and molecular mechanisms controlling cell migration in vitro, in two-dimensional (2D) and three-dimensional (3D) systems, considering, in particular, the extension of the plasma membrane of the cells at the leading edge for movement. Finally, we discuss the pathological aspects of migration associated with tumour invasion and metastasis, frequent events in patients with melanoma.

Migration of melanocytes during normal development

Melanocytes are melanin-containing dendritic cells involved in pigmentation in many animal species (Lamoreux et al., 2010). In humans, they determine skin, hair and eye colour. It has to be noted that in the eye, two types of melanin-producing cells are present, one is derived from the neural tube, the retinal pigment epithelial cells and the other one from the neural crest cells (NCCs), the uveal melanocytes including choroid, iris and ciliary body (Colombo et al., 2011). In the skin and hair, pigment cells are derived from NCCs. The truncal region contains two subpopulations of NCCs, which proliferate and then follow two major migration pathways during their development (Figure 1). NCC derivatives migrating along the dorsoventral pathway give rise to cells of the peripheral nervous system, including neurons, Schwann cells (from spinal sensory, sympathetic and parasympathetic ganglia) and chromaffin cells in the adrenal medulla. The dorsolateral pathway is the main source of melanocytes (Larue et al., 2003). In mouse embryos, NCC are first specified to differentiation into founder melanoblasts on embryonic days 8.5–9.5 (E8.5–E9.5), in the trunk region. These founder melanoblasts proliferate actively in a wedge-shaped area known as the migration staging area (MSA) lying between the ectoderm, the neural tube and the somites, being restricted to the somite-containing region. On E10.5, these melanoblast precursors, which are still proliferating, begin to migrate from the MSA, between the somites and the ectoderm, along a rostro-caudal temporal gradient. Some of the migrating melanoblasts begin to cross the basal lamina separating the dermis from the epidermis on E11.5. The remaining cells continue to migrate dorsolaterally in the dermis until stage E15.5 while continuing to cross the basal lamina (Luciani et al., 2011). Despite the occurrence of active melanoblast proliferation in the dermis, the number of melanoblasts in this compartment remains constant. Asymmetric melanoblast division, with one of the daughter cells migrating into the epidermis and the other remaining in the dermis, may account for this observation (Luciani et al., 2011). Alternatively, it may be hypothesized that parallel orientation of the mitotic spindle of the mother dermal melanoblast is predetermined. This would favour preferential localization in the epidermis of the daughter cell closest to the basal membrane (Larue et al., 2012). Because the numbers of melanoblasts in the dermis and epidermis during development are unaffected by the absence of the crucial Notch proteins Notch 1 and Notch 2 (Schouwey et al., 2007), the putative asymmetric cell division would be Notch independent. Colonization of the hair follicles begins on E15.5. At birth, most of the melanocytes are present in the epidermis of the follicular and interfollicular zones in the hairy parts of the mouse body. The interfollicular melanocytes disappear within a few days, resulting in a follicular distribution of melanocytes. More than 10 yr ago, an analysis of melanoblast establishment in patch and rump-white embryos suggested that a putative second source of melanoblasts may exist (Jordan and Jackson, 2000). The origin and characteristics of these cells were unknown at that time, but one may hypothesize that they emerge from the plastic NCC derivatives migrating along the dorsoventral pathway. In a recent study, some of the Schwann cell precursors (SCPs) migrating ventrally along nerves innervating the skin were reported to differentiate into melanocytes both in chick and mouse embryos (Adameyko et al., 2009). The classical melanoblasts migrating along the dorsolateral pathway may be referred to as the primary wave of melanocyte precursors (Figure 1). Bipotent cells with molecular characteristics of SCPs, acquiring a melanocytic fate, would then represent an alternative source (‘secondary wave’) of melanocyte precursors. The relative contributions of these two waves of migration to the final pool of melanocytes in adult wild-type skin is still a matter of debate, given the difficulty to estimate precisely the number or relative number, of nerve-derived melanocytes. It remains also unclear whether these cells arise from the differentiation of multipotent neural crest cells, from bipotent cells or from the conversion of SCP to melanoblasts.

Figure 1.

Melanocyte specification from neural crest cells. In the trunk region, early migrating neural crest cells (NCCs) move through the dorso-ventral pathway (blue arrow) in between the somites (S) and the neural tube (NT). These cells will give rise among others to neurons of the dorsal root ganglia (DRG). Before migrating along the dorso-lateral pathway (green arrow), melanoblast precursors stall in the migration staging area (MSA), then move between the somites (S) and the ectoderm to ultimately give rise to melanocytes of the first wave. A second wave of melanocytes (red arrow) arises from the Schwann cell precursors (SCPs) associated with the peripheral nerves (PN). Melanocytes are found near the nerve endings. N = notochord.

Genetic and molecular aspects of cell migration during development

Subtle variations in pigmentation traits are readily observable and recognizable. For centuries, crosses and selections, based on coat colour, have been performed by animal breeders in various species including rodents. In mice, study of pigmentary genetics began more than 100 yr ago (Lamoreux et al., 2010). Since then, many spontaneous coat colour mutants have been isolated and genetically engineered mutants have been produced (see http://www.espcr.org/micemut/ for a complete list). Most of the genes causing pigmentary phenotypes have now been cloned. A simple examination of coat colour indicates the type of cellular mechanism affected. For instance, the presence of a white belly spot in mice usually points to a potential defect in melanoblast migration, although we cannot exclude that it could be due to a reduction in melanoblast proliferation. Studies of several vertebrates – mice, chicken, Xenopus and zebrafish – have made a major contribution to our understanding of the molecular mechanisms underlying melanoblast migration during development. We will describe here some of the genes and proteins involved in the migration of cells of the melanocyte lineage. We will focus on the relative importance of cell surface receptors (kit, EphB, Ednr) and their ligands in the various species. The critical role of the transcription factors Mitf, Sox10 and Pax3 in the migration of mouse melanocytes will be highlighted, together with that of β-catenin, Rac1 and P-Rex1 proteins.

Several aspects of the migration process are common to all vertebrates, but there are also marked differences between species. This probably reflects the complexity of pigmentation in fish and amphibians, in which at least four types of unpigmented precursors (melanoblasts, leucoblasts, xanthoblasts and iridoblasts) and their respective pigmented derivatives (melanophores, leucophores, xanthophores and iridophores), migrating along different pathways (Kelsh et al., 2009), are required. Unlike the melanoblasts of mice and chickens, those of fish (zebrafish or medaka) reside principally in the dermis and do not invade the epidermis (Hirata et al., 2005). Some of the key genes responsible for migration and survival in mice, such as the Kit gene, have also been implicated in the migration and survival of melanoblasts in zebrafish. However, signalling through the Kit orthologue (kita) appears to be more transient in fish than in rodents (Rawls and Johnson, 2003). Zebrafish kita (sparse)-homozygous null mutant embryos are viable. kita mutants have fewer (about 50%) and less motile melanoblasts than the wild type. These cells fail to migrate to the periphery of the embryo, instead remaining in the dorsal region, where they eventually undergo apoptosis, after becoming pigmented (Parichy et al., 1999). Interestingly, there seems to be no equivalent role for the kit receptor in birds. In chickens, two signalling systems functioning in an additive manner are involved in melanoblast migration: the ephrin B ligand/ephrin B2 (ephB2) receptor complex and the endothelin-3 (EDN-3) ligand/endothelin receptor B2 (EDNRB2) complex (Lecoin et al., 1998; Pla et al., 2005). The ephrin/ephB2 system initially prevents the migration of NCCs along the dorsolateral pathway [Hamburger–Hamilton (HH) stage 12, with 16 somites, corresponding to 45–49 h after fertilization (Hamburger and Hamilton, 1992)], but it stimulates the dorsolateral migration of melanoblasts after stage HH 18 (with 26–36 somites, 51–68 h after fertilization) (Santiago and Erickson, 2002). The ephB2 and EDNRB2 genes are thought to be up-regulated in melanoblasts before migration, allowing them to overcome the effect of inhibitor molecules (Kelsh et al., 2009). No equivalent mechanism has been found in either zebrafish or mouse. The absence of the EDNRB2 gene from these species may account for these differences [see (Pla and Larue, 2003)]. Natural mouse mutants with a phenotype of white belly spots, caused by the absence of melanoblasts in these areas, have been instrumental to our understanding of melanoblast migration and homing. Studies of the microphthalmia (Mitf), dominant white spotting (Kit), steel (Kitl), piebald (Ednrb) and lethal spotting (Edn-3) mutants have led to the identification of genes and proteins essential for the first step of melanoblast migration following delamination from the neural tube (Lamoreux et al., 2010). For example, the Edn-3 ligand, produced by the ectoderm, binds to the Ednrb receptor expressed on melanoblasts. Signalling through this complex is essential for migrating melanoblasts to reach their destination (Lee et al., 2003). The kit ligand produced in the dermomyotome, dermis and hair follicle binds to the c-kit receptor tyrosine kinase expressed by melanoblasts. Only melanoblasts expressing the c-kit and Ednrb receptors can respond to the attractive and repulsive signals controlling cell migration (Reid et al., 1996).

The Mitf (microphthalmia-associated transcription factor) gene and, in particular, the M-Mitf isoform specific to the melanocyte lineage, plays a key role in melanoblasts and melanocytes, not only ensuring their specification and survival, but also contributing to the primary (dorsolateral) and secondary (ventral) stages of migration through regulation of numerous target genes (Tsao et al., 2012). Among them, Slug (Snai2) was shown to be directly regulated by M-Mitf, which has been involved in pseudo-EMT of NCC and melanoblast migration (Sanchez-Martin et al., 2002). Unsurprisingly, spontaneous heterozygous mutations in two of the genes synergistically regulating M-Mitf expression, Sox10 and Pax3, give rise to a white-spotted phenotype (dominant megacolon/Sox10 and Splotch/Pax3). Another regulator of M-Mitf, the CTNNB1 gene, encoding β-catenin, was identified as a regulator of melanoblast migration through the use of transgenic techniques. When located in the nucleus, β-catenin functions as a co-activator of the LEF/TCF family of proteins, to target, among others, M-Mitf promoter. The specific production of an activated form of β-catenin (bcat*) in the nuclei of mouse melanoblasts is associated with a white belly spot, due to a defect of melanoblast migration during development (Delmas et al., 2007; Gallagher et al., 2012). The M-Mitf and Csk (Src inhibitor) genes have been shown to be involved in the defective melanocyte migration observed in this model (Gallagher et al., 2012). The relative importance of M-Mitf in cell migration is certainly due to the fact that it is regulating the expression of other crucial genes involved in this cellular mechanism. Interestingly, M-Mitf is regulated by BRN2, which is itself regulated by β-catenin (Goodall et al., 2004, 2008; Kobi et al., 2010; Wellbrock et al., 2008). A mutant of BRN2, BRN2AA, represses melanoblast migration (Berlin et al., 2012). Another key gene controlling the migration of melanocytes in vivo is the Rac1 gene, encoding the Rac1 small GTPase. In zebrafish, Rac1 regulates the migration of neural crest cells towards the chemo-attractant stromal cell-derived factor 1 (Sdf-1)/CXCL4 (Theveneau et al., 2010). Specific inactivation of the Rac1 gene in mouse melanoblasts results in white belly spots. Absence of the Rac1 protein markedly decreases the speed of melanoblast migration, but does not prevent these cells from crossing the dermal/epidermal junction or colonizing hair follicles (Li et al., 2011). A recurrent activating somatic mutation in the human RAC1 gene promoting melanoblast migration was recently identified in melanoma samples (Krauthammer et al., 2012). Further evidence for the crucial role of Rac1 in migration was provided by specific inactivation of the P-Rex1 gene in the mouse melanocyte lineage. P-Rex1 encodes a member of the GEF (Rho GTPase guanine nucleotide exchange factor) family responsible for activating Rac1-GTP formation. P-Rex1-deficient mice have a white belly spot, due to defective melanoblast migration, and high levels of P-Rex1 have been found in several human melanoma cell lines forming metastases (Lindsay et al., 2011).

Modes of migration and associated molecular mechanisms

Little is known about the mode of melanoblast migration in vivo due to the small number of these cells and the difficulty to follow their movement in the pregnant mouse. Most studies have therefore been carried out in vitro, in 2D systems, with melanoma cell lines. The recent use of 3D culture systems, recreating an environment mimicking the situation in vivo, has revealed marked differences between these two systems.

2D migration and molecular mechanisms

Cells migrating in a 2D system must form plasma membrane extensions (protrusions) at the migration front, in a manner coordinated with the rest of the cell body. Adhesive cells migrating on uniform 2D surfaces have two types of protrusion – lamellipodia and filopodia – with different structures and molecular compositions. Lamellipodia were originally described in fibroblasts, as thin sheet-like regions containing microfilaments (Abercrombie et al., 1970), and filopodia are finger-like extensions from the plasma membrane containing parallel bundles of actin filaments (Ridley, 2011). These protrusions may appear spontaneously or may be induced by external signals originating from the microenvironment. Migration is thus dependent on successive cycles of membrane extension, adhesion, retraction of the rear end of the cell and translocation of the cell body. These processes require asymmetry of the front (leading edge) and back (trailing edge) of the cell, resulting in antero-posterior polarity (Spiering and Hodgson, 2011).

Table 1. Type of protrusions involved in single-cell locomotion and associated migration mode
SystemDimensionType of protrusionsActin dependencyMigration modeReferences
  1. NM, non-mesenchymal; NA, non-amoeboid.

  2. This table refers exclusively to single-cell motility. Collective migration also exists in several types of cells including melanoma [see (Hegerfeldt et al., 2002)]. On the opposite, melanoblasts in vivo migrate only individually using Rac1-independent short stubs and Rac1-mediated long protrusions (Li and Machesky, 2012).

In vitro 2D

Lamellipodia

Filopodia

Invadopodia

Blebs

Yes

Yes

Yes

No

Mesenchymal

Mesenchymal

Mesenchymal

Amoeboid (pseudopodal)

(Machesky, 2008)

(Faix and Rottner, 2006)

(Chen, 1989)

(Yoshida and Soldati, 2006)

In vitro 3D

Lamellipodia

Lobopodia

Blebs

Yes

Yes

No

Mesenchymal

Lobopodial

Amoeboid (blebby)

(Friedl and Wolf, 2009)

(Petrie et al., 2012)

(Charras and Paluch, 2008)

In vivo 3D

Short stub protrusions

Long protrusions

Yes

Yes

NM and NA

Mesenchymal

(Li et al., 2011)

(Li et al., 2011)

The actin cytoskeleton and the molecules regulating its polymerization and dynamics are involved in these protrusions. Actin assembly and disassembly are therefore key elements in cell migration. Signals are transmitted from the membrane receptors to the actin cytoskeleton by small Rho GTPases, oscillating between an active GTP-bound form and an inactive GDP-bound form. These GTPases are regulated by two families of proteins – GEFs and GAPs (GTPase accelerating proteins) – which interact directly with the GTPase protein, at the same binding site. GEFs catalyse the formation of the active form of the GTPase by releasing the bound GDP in exchange for GTP, whereas GAPs mediate inactivation by GTP hydrolysis after the insertion of a water molecule into the catalytic pocket of the GTPase (Raftopoulou and Hall, 2004; Spiering and Hodgson, 2011). The Rac1 and Cdc42 GTPases regulate lamellipodial and filopodial formation, respectively, by inducing the polymerization of actin filaments. This leads to extension of the migration front, through the successive activation of proteins of the WASP (Wiskott–Aldrich syndrome protein) family, acting as nucleation factors binding to the Arp2/3 complex to generate actin filament networks (Chesarone and Goode, 2009). Dia1 and Dia2, which belong to the formin family, also play a role in migration, by promoting actin filament elongation in lamellipodia and filopodia, respectively. These proteins protect the barbed ends of actin filaments, ensuring their continuous growth (Ridley, 2011).

It has been shown that M-Mitf regulates directly the transcription of mDia1 (Diaph1) and MET (Met), in melanoma cells (Carreira et al., 2006; McGill et al., 2006). In some melanoma cell lines, lower levels of M-Mitf and mDia1 productions were found to induce cell invasion (Carreira et al., 2006).

3D migration and molecular mechanisms

During the migration of individual cells in 3D systems, certain melanoma cells switch between two interconvertible morphologies and types of migration – mesenchymal migration and amoeboid migration – allowing them to adapt to changes in the microenvironment (Friedl and Wolf, 2010; Sanz-Moreno et al., 2008) (Figure 2). Mesenchymal migration is typical of elongated, bipolar adherent cells interacting with the extracellular matrix (ECM) at focal adhesion points. This type of migration roughly resembles 2D migration, but the constraints imposed by the stiffness and elastic behaviour of the ECM necessitate changes in cell morphology. Lamellipodia are replaced with pseudopodia oriented in three dimensions, and the proteases required for ECM remodelling are secreted (Friedl and Wolf, 2010; Sahai and Marshall, 2003). These cells migrate slowly, at an estimated speed of 0.1 μm/min (Sanz-Moreno and Marshall, 2010).

Figure 2.

Schematic representation of the opposite patterns of signal transduction from the Rho GTPases on the cell cytoskeleton and the movement of melanoma cells. Rho GTPases (RhoA, Cdc42 and Rac1) are the main proteins of a central signalling hub. Rac1 activation through the Nedd9/Dock3 complex decreases the contractility of actomyosin, promoting mesenchymal movement, whereas the activation of RhoA increases actomyosin contractility, generating amoeboid movement. The scaffold protein Nedd9 also binds to phosphorylated Src Y416. Phosphorylation of the tyrosine 416 residue of Src, through integrin β3 signalling, down-regulates the kinase activity of ROCK by phosphorylating its tyrosine 722 residue (Ahn et al., 2012; Sanz-Moreno et al., 2008). p21-activated kinase (PAK) binds specifically to the GTP-bound form of Rac 1. It then activates the LIM-motif-containing protein kinase (LIMK) to drive cofilin-mediated actin turnover. IRSp53 (insulin receptor tyrosine kinase substrate p53) is an I-BAR domain-containing adaptor protein that binds to Rac1 and Wave 2, thus contributing to lamellipodial extension (see text for details).

Amoeboid migration is associated with rounded or ellipsoid cells lacking focal adhesion. This mode of migration is integrin independent, as motility in a 3D system does not necessarily require the cell to be attached to its substrate (Lammermann and Sixt, 2009). Movement is achieved by the creation of ‘blebs’, short spherical extensions of the plasma membrane devoid of actin with a very short lifetime. ‘Blebbing’ is initiated by the local detachment or rupture of cortical actin. An increase in intracellular pressure leads to an influx of cytosol into the membrane, extending the bleb. A new actin cortex is then formed under the bleb, triggering its retraction. Blebs mostly occur at the migration front (Charras and Paluch, 2008), and they initiate migration by enabling the blebby round cell to insert itself into the natural cavities of the matrix without the need for proteolysis. This type of propulsive cell migration has a speed of between 1 and 10 μm/min (Sanz-Moreno and Marshall, 2010).

An unusual mode of 3D cell motility based on the presence of large blunt cylindrical protrusions called lobopodia has been recently described in fibroblasts (Petrie et al., 2012), (see Table 1). Lobopodial migration seems to combine aspects of lamellipodial and blebbing locomotion (Sixt, 2012). However, its role in the motility of melanoblasts and/or melanoma cells remains unknown.

The signalling pathways used in mesenchymal and amoeboid movements are distinct (Figure 2). The mesenchymal 3D migration of melanoma cell lines involves the formation of a complex between Nedd9 (a member of the CAS family of adaptor proteins) and DOCK3 (a member of the DOCK subfamily of GEFs), resulting in activation of the Rac1 protein. Rac1-GTP activates the Wave 2 protein (a member of the WASP family), which regulates the Arp2/3 complex involved in actin polymerization (Sanz-Moreno et al., 2008). By contrast, amoeboid migration is controlled by the Rho/ROCK complex, which is responsible for the phosphorylation of myosin light chain II (MLC2). The resulting actomyosin complex generates the contractile force required for cell movement. The transition from mesenchymal to amoeboid migration is associated with a decrease in Rac1-GTP levels and an increase in MLC2 phosphorylation (Figure 2), demonstrating the opposite effects of Rac1 and Rho on the migration of melanoma cells (Sanz-Moreno et al., 2008).

Migration and melanoma invasion

Melanomas are highly metastatic skin tumours resulting from the malignant transformation of melanocytes. The presence of lymph node and visceral metastases is directly related to poor outcome, with a mean survival time of 7.5 months from diagnosis (Chudnovsky et al., 2005). Melanomagenesis can be defined at the epidemiological, clinical, histopathological, cellular and molecular levels (Whiteman et al., 2011). At the cellular level, melanoma invasion results from a combination of several mechanisms: a pseudoepithelial–mesenchymal transition, a loss of cell-to-cell adhesion, a loss of cell–matrix adhesion, matrix degradation, chemo-attraction/repulsion and migration. All these cellular events are closely regulated and tightly interconnected using various signalling pathways including MAPK, PI3K and Wnt/β-catenin (Larue and Beermann, 2007). During the radial growth phase (RGP), melanocytes proliferate in an aberrant manner and their interaction with keratinocytes decreases, partly due to a loss of E-cadherin. This initial stage is followed by a vertical growth phase (VGP) marking the start of the transmigration process characterized by the passage of melanoma cells from the epidermis, across the basal lamina, to the dermis. This transmigration, which runs in the opposite direction to that of normal melanoblast migration during embryonic development, is a key step in metastasis. Migrating melanoma cells make use of cellular and molecular processes commonly associated with melanoblasts during normal development. This observation is entirely consistent with Virchow's assertion that ‘Neoplasms arise in accordance with the same law, which regulates embryonic development’ (Virchow, 1859). Support for this assertion was recently provided by a comparison of gene expression between transplanted metastatic melanoma cells and primary melanocytes in a chick embryo model (Bailey et al., 2012). However, in vivo analysis of mouse melanoblast migration suggested that the acquisition of basal lamina-degrading activity by melanoma cells does not correspond to a simple reversion to the normal melanoblast genetic programme (Li et al., 2011).

In addition to lamellipodia, filopodia, blebs and lobopodia, a fifth type of membrane protrusion, the invadopodium, has been characterized as an actin-rich matrix-degrading finger-like protrusion (Buccione et al., 2009; Chen, 1989). The presence of invadopodia has been shown to be essential for the crossing of the basal membrane by melanoma cells (Figure 3). The formation of invadopodia in cells with a mesenchymal morphology requires the presence of cortactin, a scaffolding protein that stabilizes branched actin networks and regulates the vesicular trafficking of matrix metalloproteinases to invadopodia (Figure 3) (Clark et al., 2007). Cortactin contributes to these two processes by strongly promoting cell invasiveness (Kirkbride et al., 2011). Cortactin was originally identified as a Src kinase substrate, but its SH3 domain also interacts with many different kinases and it binds to and buffers cofilin (Oser and Condeelis, 2009). The invasion process of elongated mesenchymal melanoma cells also involves the elongation of actin filaments through cofilin dephosphorylation and the stabilization of these filaments by fascin, which is normally produced by melanocytes (Li et al., 2010). The degradation of the ECM by invadopodia clearly distinguishes these structures from lamellipodia and filopodia. Vesicles containing metalloproteases, including MT1-MMP/MMP14, MMP2 and MMP9, are targeted to the invadopodia by the exocyst complex, mediating the tethering of post-Golgi vesicles to the plasma membrane (Poincloux et al., 2009) (Figure 3). Metalloproteases are released at the extreme end of the invadopodia (Ridley, 2011; Schoumacher et al., 2010). They degrade the ECM, facilitating the progress of melanoma cells (Gaggioli and Sahai, 2007). However, the absence of invadopodia from normal melanoblasts/melanocytes indicates that these cells use membrane protrusions other than invadopodia for motility in vivo (Li et al., 2010, 2011; Pichot et al., 2010). Two types of melanoblast protrusion (short stubs and long pseudopods) have been described in vivo (Li et al., 2011) and associated with a mesenchymal mode of migration in wild-type cells. Interestingly, melanoblasts from Rac1-deficient mice formed only actin-rich short stubby protrusions inducing mesenchymal and amoeboid-independent slower migration. Whether this type of motility could correspond to the lobopodial migration (using large blunt cylindrical protrusions called lobopodia) recently observed in a 3D system (Petrie et al., 2012; Table 1) remains to be assessed. Melanoma cells do not destroy the basal membrane completely. Instead, the invadopodia generate small membranous perforations, facilitating the dissemination of cells in the dermis and interactions with endothelial cells preceding the intravasation of blood vessels (Figure 3). Transcriptomic studies have identified a number of genes displaying abnormal expression in invasive melanoma cells (Hoek et al., 2006). Several of these genes encode growth factors, including TGFβ. This protein is thought to play an important role in invasion by poorly pigmented cells with high levels of the transcription factor Brn2/N-Oct3/Pou3f2. Upon stimulation with TGFβ, melanoma cells acquire an elongated mesenchymal phenotype, facilitating their passage across the ECM (Pinner et al., 2009). Brn2 regulates directly M-Mitf and promotes melanoma invasiveness in vitro (Goodall et al., 2008), being itself negatively regulated by M-Mitf through miR-211 (Boyle et al., 2011). The many molecules actively contributing to the invasion processes of melanoma cells include the adaptor protein Nedd9. This scaffold protein is often produced in large amounts in metastatic melanomas and has been shown to be involved in 3D migration in vitro, promoting the transition between the amoeboid and mesenchymal modes of migration (Kim et al., 2006; Sanz-Moreno et al., 2008). However, it remains unclear whether this transition occurs in vivo (Sabeh et al., 2009). This review would not be complete without some mention of the small GTPase ARF6, which is present in invadopodia and involved in the invasion of the lungs by melanoma cells (Ridley, 2011). It activates Rac1 by phosphorylating ERK (Muralidharan-Chari et al., 2009). Finally, reversible epigenetic events linked to the microenvironment are also thought to contribute to the process of metastasis (Shackleton and Quintana, 2010).

Figure 3.

Formation, molecular characteristics and function of invadopodia. (A) Interaction between a tumour cell and the basal membrane (BM) triggers the formation of invadopodia, which start degrading the BM (stage 1). (B) Invadopodia elongate and infiltrate the stromal compartment (stage 2). High magnification of an invadopodium showing a molecular overview of the key proteins involved in actin assembly and ECM degradation. Actin filaments are branched at the cell surface and unbranched at the tip of the protrusion (Schoumacher et al., 2010). (C) Maturation of the invadopodia involves cortactin and actin polymerization via N-Wasp (stage 3). Nck is an upstream regulator of N-Wasp and Tks5 (tyrosine kinase substrate with 5 SH3 domains). This scaffold protein can recruit and phosphorylate cortactin. (D) ECM degradation by invadopodia (stage 4). During this process, cortactin is involved in the vesicular trafficking of matrix metalloproteinases (MMPs) to invadopodia via the regulation of post-Golgi trafficking or vesicle capture at the plasma membrane by means of an exocyst complex. The scaffold protein IQGAP1 coordinates actin assembly and the exocytosis machinery. MMPs are released into the ECM through small perforations in the plasma membrane.

Conclusion and perspectives

Molecular genetics, biochemistry, developmental and cell biology have generated a large body of information, highlighting different aspects of melanocytic cell migration. In vitro studies in 2D and 3D systems have facilitated elucidation of the biochemical mechanisms controlling the migration of melanoma cells. We now know that melanoma cell movement in vitro is based on mesenchymal and amoeboid types of migration. A third type of migration based on lobopodia may also occur in melanocytes and/or melanoma cells. These types of migration are mediated by the plasticity of the cell body and the presence of plasma membrane protrusions (lamellipodia, filopodia, blebs and invadopodia) at the cell front. Migration clearly involves both cell-autonomous and cell-non-autonomous mechanisms. Thus, the microenvironment (including at least cell–matrix and cell–cell interactions) is clearly of importance for both cell migration and protrusions.

Despite considerable advances towards an understanding of the processes underlying protrusion formation and cell motility in a 2D environment, the impact of the physical forces generated by the ECM on melanoma cell movement in 3D or in vivo remains unclear. Primary fibroblasts migrating within two different 3D environments were recently shown to adopt two matrix elasticity-dependent modes of migration involving lamellipodia and lobopodia (Petrie et al., 2012). Similarly, melanoma cells may react to the stiffness and elasticity of the tissue in which they migrate, by converting their mechanical perception of matrix properties into variable intracellular signals, resulting in differences in cell movement.

Another critical question concerns the way by which small numbers of melanoma cells colonize tissues located large distances away from the primary tumour during metastasis? The importance of STAT3 activation in primary tumours and tumour-infiltrating myeloid cells in the colonization process was recently highlighted (Lee et al., 2010). STAT3 signalling by tumour-associated myeloid cells in premetastatic niches before the arrival of metastatic melanoma cells would render distant organs hospitable for future colonization by disseminating tumour cells (Deng et al., 2012). The microenvironment also plays a key role. Stromal cells surrounding tumour cells have been shown to mediate directional cancer cell migration, by producing the Sdf-1 ligand activating the chemokine receptor 4 or CXCR4 (Joyce and Pollard, 2009). A blockade of stromal cell function through the use of ligand inhibitors or receptor antagonists might constitute a promising strategy for preventing metastases.

An understanding of cell movement in vivo, during development, homoeostasis and transformation, is of crucial importance for deciphering the mechanisms underlying these processes. The transparency of zebrafish embryos is a considerable advantage for developmental studies. Attempts to carry out studies in vivo in mammals have been hindered by the difficulty to identify and visualize migrating melanocytes, which are highly dispersed and poorly accessible. The emergence of new ex vivo culture techniques, together with real-time imaging techniques with high-resolution microscopes, will undoubtedly lead to substantial progress in the near future (Li et al., 2011; Mort et al., 2010). Likewise, the ability to determine the full molecular content (RNA, protein) of a single melanocyte should also shed new light on the migration of single cells in vivo.

Acknowledgements

MD was supported by a fellowship from MENRT. This work was supported by the Ligue Nationale Contre le Cancer (Equipe labellisée), INCa and ARC.

Ancillary