If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
To date, over 1000 melanocytic neoplasms, spanning all stages of tumorigenesis, have been sequenced, offering detailed views into their -omic landscapes. This has coincided with advances in genetic engineering technologies that allow molecular biologists to edit the human genome with extreme precision and new mouse models to simulate disease progression. In this review, we describe how these technologies are being harnessed to provide insights into the evolution of melanoma at an unprecedented resolution, revealing that prior models of melanoma evolution, in which pathways are turned ‘on’ or ‘off’ in a binary fashion during the run-up to melanoma, are oversimplified.
Melanoma is driven by mutations that result in the uncontrolled proliferation of a melanocyte. Melanocytes have checkpoints, so that a single mutation is unable to drive their transformation into a fully malignant tumor, and thus, they require approximately 5-10 pathogenic alterations, which are spread across several signaling pathways, to trigger their transformation to melanoma (
Mutations that activate the mitogen-activated protein kinase (MAPK) pathway are ubiquitous, and most melanomas also harbor somatic alterations that upregulate telomerase and disrupt cell-cycle checkpoint control. Finally, many melanomas have somatic alterations that perturb the p53 pathway, activate the phosphatidyl-inositol 3–kinase (PI3-kinase) signaling cascade, and disturb chromatin remodeling complexes. These constitute the main signaling pathways with genetic perturbations in melanoma and will be the main focus of this review (Figure 1).
Other factors also help drive the progression of melanoma. In particular, signaling pathways can be rewired via non-genetic mechanisms, and cell non-autonomous factors, including the immune system, cell-cell signaling, and the composition of the local extracellular matrix play critical roles in shaping melanocytic neoplasms. These factors are not covered in this review but are covered elsewhere (
Melanocytic neoplasms are staged by their clinical and histopathologic features. Melanocytic nevi, also known as common moles, occupy the benign end of this spectrum. It is estimated that only 1 in 10,000 nevi eventually transform into melanoma (
). There are also intermediate lesions, defined as having worrisome clinical or histopathologic features but nonetheless falling short of an unequivocal diagnosis of melanoma. Some of the lesions in this category are likely nevi or melanomas, masquerading as something in between; however, genetic studies indicate that there are true biologically intermediate neoplasms, harboring more oncogenic alterations than common nevi but less than melanoma (
). Finally, melanomas occupy the malignant end of this spectrum, and they are further staged by their thickness, ulceration, and the extent to which they have (or have not) spread to other parts of the body (
). During the course of tumor progression, melanocytes accumulate pathogenic mutations, eliminating barriers to transformation, and allowing the ensuing neoplasms to transition through the clinical stages of melanocytic neoplasia described above.
It is tempting to describe the progression of melanoma as a series of binary events, in which specific pathways are turned “on” or “off”, ultimately producing a melanoma, but this is an oversimplification. For this review, we emphasize the specific genetic events that titrate discrete levels of activation (or inactivation) of the critical pathways during the course of progression of melanocytic neoplasms. Reviews covering other aspects of melanoma progression are elsewhere (
). We focus primarily on the MAPK pathway and the Rb pathway because these pathways are the most thoroughly studied with respect to gene dosage.
MAPK signaling ramps up during the course of progression
Activation of the MAPK pathway is likely required to form a melanocytic neoplasm. Mutations that activate BRAF and NRAS, two critical nodes of the MAPK signaling cascade, cumulatively occur in 75% of melanomas (
). The term “wild-type” melanoma has been widely adopted to describe the melanomas without these mutations; however, more comprehensive genomic studies, geared specifically toward these melanomas, have found that “wild-type” melanomas are, themselves, riddled with a “long-tail” of relatively uncommon mutations in the pathway (Figure 1) (
). Currently, melanomas without recognized mutations in the MAPK pathway are rare, and we propose that the ones that do seem to exist probably reflect cases in which a mutation was missed or unappreciated rather than a true biological state.
Mutations in the MAPK-pathway genes are also ubiquitous in earlier stages of melanocytic neoplasia (
). Nevertheless, the levels of pathway activation do not stay the same during the course of progression. This is reflected, in part, by changes in the zygosity of oncogenic mutations in these genes. For instance, most nevi have a heterozygous BRAFV600E mutation (
). Altogether, increases in the gene dosage of oncogenic mutations imply that the MAPK-pathway output ramps up during the evolution of melanoma.
Many melanocytic neoplasms also acquire more than one mutation in the MAPK pathway during the course of progression. To be sure, some combinations of mutations are rarely observed. For example, BRAFV600E mutations do not typically coincide with NRAS codon 61 mutations, and this finding is often interpreted as evidence that these mutations are functionally redundant, precluding their selection (
). However, there are many combinations of mutations in genes in the MAPK pathway that can co-exist. For example, hypoactive BRAF mutations, NF1 mutations, MAP2K1 mutations, and mutations associated with RASopathy syndromes commonly overlap among one another or alongside BRAFV600E and NRAS codon 61 mutations (
Transcriptomic and proteomic data further indicate that MAPK signaling strengthens during the evolution of melanoma. RNA-sequencing can measure total levels of gene expression, as well as the expression from each allele. The RNA-sequencing of nevi and melanomas indicates that the oncogenic allele of BRAF is somewhat repressed in nevi and preferentially expressed in melanoma when compared with the wild-type allele (
). This holds true in tumors without changes in BRAF copy number, suggesting that there may be epigenetic mechanisms to specifically recognize and modulate the expression of the oncogenic allele. Immunostaining with an antibody that specifically recognizes the V600E form of the BRAF protein confirms that BRAFV600E is more highly expressed in melanomas than nevi at the protein level (Fig. 2A) (
In aggregate, genomic, transcriptomic, and proteomic data each indicate that MAPK signaling is turned ‘on’ at the initial stages of melanocytic neoplasia but progressively ramps up during the course of evolution. Based on the composite evidence, we propose a model whereby pathway activation more than doubles during the course of progression (Fig. 2B).
Molecular studies also illustrate the importance of the levels of MAPK signaling in melanocytes. In particular, there are notable phenotypic differences that are dependent on BRAFV600E expression levels in both mice and humans. In mouse models, the introduction of a single allele of BrafV600E in melanocytes results in highly pigmented lesions and melanocytic hyperplasia within a month (
). These observations indicate that the BrafV600E gene dosage exerts a physiologically relevant impact on mouse melanocyte proliferation in vivo – an experimental conclusion that may rationalize the observation of increased copy number of oncogenic BRAF in human melanomas.
The importance of gene dosage is also reflected by in vitro studies of human cells. The ectopic expression of BRAFV600E in fibroblasts or melanocytes induces rapid growth restriction within 3-7 days (
). However, this experimental approach results in supra-physiological levels of the mutant protein that are immune to endogenous transcriptional regulation. Another approach is the precision engineering of a BRAFV600E mutation into its endogenous locus with CRISPR/Cas9-mediated gene editing. In contrast to overexpression, primary human melanocytes engineered in this way acquire a proliferative advantage over sibling melanocytes, which are wild-type for BRAF, and this proliferative advantage is sustained for 2 months (
). Altogether, these observations illustrate that modulating the dosage of BRAFV600E produces different phenotypes in human melanocytes. This is important to appreciate because the mechanisms underlying the growth arrest that follows an initial period of proliferation in BRAFV600E-mutant melanocytes are poorly understood, and the molecular consequences of gaining copies of the oncogenic allele are also not known. Moving forward, accurate methods for modeling the precise changes in BRAFV600E allele dosage, both in vitro and in vivo, will be critical for understanding how gene dosage contributes to melanoma initiation and progression.
The importance of gene dosage in the context of oncogenic RAS is well established through molecular studies in other cancers. For example, an increase in KRAS-mutant allele dosage drives both tumorigenesis and metastasis in a mouse model of pancreatic ductal adenocarcinoma (
In melanoma, Pederson et. al. (2013) developed a mouse model of NrasG12D that can be selectively induced in melanocytes. All mice with homozygous NrasG12D/G12D mutations show a darkening of the skin within 2 months, whereas only half of the mice with a heterozygous NrasG12D mutation show a darkening of the skin that is also much weaker (
). Interestingly, when NrasG12D is expressed in a non-inducible setting, leptomeningeal melanoma forms during development and occurs earlier in homozygous NrasG12D/G12D mice than in heterozygous NrasG12D/+ mice (
), further investigations into the importance of physiologically relevant NRAS-mutant allele dosages are warranted in the human setting.
Disruption of cell-cycle checkpoint control occurs in a stepwise fashion
Normal cells have mechanisms to halt the cell-cycle progression at the transition from the G1 to the S phase of the cell cycle, but abrogation of this checkpoint occurs in most, if not all, melanomas (
). The p16INK4A protein primarily governs this regulation, and it is encoded by the CDKN2A gene, which harbors bi-allelic, loss-of-function alterations in nearly 50% of melanomas. Deletions of CDKN2A often encompass the neighboring gene, CDKN2B, which encodes p15INK4B – another critical checkpoint protein. Finally, somatic alterations also affect additional genes involved in checkpoint control, including CDK4, CCND1, PPP6C, and FBXW7, among others (
), whereas, intermediate neoplasms and melanomas in situ tend to have heterozygous mutations affecting CDKN2A or mutations affecting genes encoding less critical components of the checkpoint apparatus (
). Stepwise loss of cell-cycle-checkpoint control is further reflected by immunostaining studies, in which reduction of the p16INK4A protein occurs at the transition from nevus to melanoma in situ, and complete elimination of the protein occurs at the transition to invasive melanoma (Fig. 2C) (
). Based on the composite evidence, we propose a model whereby cell-cycle control is gradually lost during the course of progression (Figure 2D).
Molecular studies also illustrate the importance of the precise levels of p16INK4A in melanocytes. Bi-allelic loss of CDKN2A promotes invasion and metastasis in vivo in genetically engineered mice, as well as similar phenotypes in vitro in primary human melanocytes (
). In an especially informative experiment, the contribution of loss-of-one versus loss-of-two copies of CDKN2A to metastatic potential was directly assessed using a pair of related melanoma cell lines – a parental line (WM793) and a subclone of this line (1205Lu). Genetic sequencing of the two lines revealed that complete loss of CDKN2A is the only additional pathogenic mutation in the subclone (1205Lu). Both cell lines proliferate in culture and as primary tumors when injected subcutaneously into immune-compromised mice; however, only the subclone (1205Lu), which has complete ablation of the CDKN2A gene, readily metastasizes in the same setting (
). These phenotypes can be toggled by re-expression of p16INK4A in the derivative line (1205Lu) or knock-down of p16INK4A in the parental line (WM793), respectively eliminating or inducing the metastatic phenotype (
). This is a useful model, but in some circumstances, it is oversimplified, and it appears that the role of CDKN2A in melanoma progression is one of those circumstances. Different dosages of CDKN2A influence several melanocyte phenotypes, including growth arrest, proliferation, and invasion (
). The mechanisms underlying how CDKN2A dosage affects these phenotypes are undefined, but they are almost certainly tied to RB1 phosphorylation. One likely mechanism is that by varying RB1 phosphorylation, key lineage-restricted transcription factors that bind to phosphorylated-RB1, such as MITF and TBX2, alter their transcriptional targets (
). Altogether, since CDKN2A bi-allelic loss occurs in melanomas, and mono-allelic loss occurs in earlier stages of neoplasia, further investigation into how RB1 phosphorylation and RB1 binding partners are influenced by specific levels of p16INK4A will provide valuable insights into the progression of melanoma.
Other “hits” to the Rb pathway can likely cooperate with, or even substitute for, the loss of CDKN2A to promote melanoma progression. In particular, p15INK4B loss was functionally shown to reverse the growth-arrest phenotype in nevus cells, resulting in the progression toward melanoma (
). Mutations in additional genes, such as CDK4, PPP6C, CCND1, and RB1, are less common, and therefore, we hypothesize that their phenotypes are attenuated as compared with CDKN2A loss.
Perturbation of other pathways tends to also be incremental
More studies are needed to fully resolve whether other signaling pathways are disrupted in an incremental or a binary fashion during the evolution of melanoma, but genetic observations provide some clues.
The SWI/SNF chromatin-remodeling-complex is likely perturbed in an incremental fashion during the evolution of melanoma. Loss-of-function mutations affect several members of the complex, including ARID2, ARID1A, ARID1B, PBRM1, and SMARCA4 (
). More studies are needed to resolve the spectrum of these mutations in melanoma, but their co-occurrence implies continual selection to perturb chromatin remodeling, even after acquisition of the first mutation.
It is unclear whether the p53 pathway is disrupted incrementally during progression. TP53 mutations occur in approximately 20% of melanomas (
), but the p53 pathway may be disrupted in a much higher percentage of melanomas as a result of CDKN2A loss. The CDKN2A locus encodes two protein products – p16INK4A (described above) and p14ARF, which operates in the p53 pathway (Figure 1). In melanoma, p16INK4A is thought to be the dominant tumor suppressor, indicated by the fact that germline and somatic alterations can affect p16INK4A, while sparing p14ARF (
). Nevertheless, most genetic alterations do impact both proteins, arguing that subsequent TP53 mutations could function as secondary hits to the pathway—this would imply incremental disruption of the pathway.
Lineage-specific transcription factors, including MITF and SOX10, are also critical regulators of melanoma, and the activity of these proteins likely increases throughout progression. MITF and SOX10 are required in the earliest stages of tumorigenesis (
), implying selection to upregulate MITF activity in later stages of tumorigenesis. Moreover, the activity of these genes’ products are carefully modulated at the transcript and protein levels (reviewed elsewhere [
), whereby levels of activity from MITF and its collaborators elicit different phenotypes.
Up-regulation of telomerase is likely an exception to the pattern of incremental disruption described in other pathways. In melanoma, the TERT gene has a high frequency of somatic mutations affecting its promoter (
). Most melanomas have a single TERT promoter mutation, without any other mutations in this pathway, arguing that telomerase is simply turned “on”, rather than incrementally upregulated. Mechanistic studies also support this view. Despite having TERT promoter mutations, melanomas have short telomeres (
). In aggregate, there seems to be no selective advantage to increase telomere lengths beyond the levels necessary to forestall telomeric crisis, and a single mutation in the TERT promoter is sufficient to do this. Overall, these findings argue that a single ‘hit’ is sufficient to turn ‘on’ telomerase and immortalize melanoma cells.
Conclusions and Next Steps
There is a prevailing sentiment that multiple mutations in the same pathway rarely co-occur in cancers because they are functionally redundant (
). In melanoma, this is generally true for BRAFV600E and NRAS codon 61 mutations, but there are numerous exceptions, and the concept that multiple mutations in the same pathway cannot co-occur should not be treated as dogma. It is especially important to pay attention to the zygosity of mutations, as many mutations are themselves subject to changes in their gene dosage. The recent revolution in genetic engineering tools, most notably CRISPR/Cas9, now enable the modeling of different combinations of pathogenic mutations or different mutant allele dosages under endogenous transcriptional regulation. Such approaches should be considered when performing functional and mechanistic studies aimed at understanding the early stages of melanoma progression.
Finally, knowing the precise levels of pathway activation (or inactivation) that are optimal for tumor cells will help guide therapeutic interventions. The importance of this issue is exemplified by patient responses to MAPK-pathway inhibitors, as most patients have an initial response to treatment yet ultimately develop resistance – thereby illustrating how tumors are able to adapt and find their optimal levels of signaling. In extreme scenarios, tumors can even acquire dual BRAFV600E and NRASQ61L mutations (
). Moving forward, studying how patients with different dosages of oncogenic MAPK-pathway mutations respond to MAPK-pathway inhibitors may provide novel biomarkers of drug sensitivity, and we anticipate that new drug targets will emerge by understanding how different dosages rewire cellular signaling pathways and influence cell behavior.
In conclusion, we encourage the melanoma research community to think beyond binary models of melanoma evolution, in which pathways are either ‘on’ or ‘off’, and to consider the precise levels of pathway perturbation in their studies.
AHS was supported by grant NCI K22 CA17997 from the Melanoma Research Alliance and grants from the Melanoma Research Foundation , Dermatology Foundation , and PhRMA Foundation . This work was originally presented at the joint meeting between the Pan American Society of Pigment Cell Research and the Montagna Symposium on the Biology of Skin (Gleneden Beach, Oregon). This scientific retreat was supported by grants R13 AR009431-53 to Molly F. Kulesz-Martin and R13 AR074279-01 to Sancy A. Leachman from the NIAMS, NIA, and NIEHS, which also defrayed the publication costs.
Conceptualization: AHS; Visualization: AHS; Writing – original draft; AHS. Writing – reviewing and editing: HZ, RLJT, and AHS.
Human tumor genomics and zebrafish modeling identify SPRED1 loss as a driver of mucosal melanoma.