Molecular pathology of cutaneous melanoma

SUMMARY 

Cutaneous melanoma is associated with strong prognostic phenotypic features, such as gender, Breslow's thickness and ulceration, although the biological significance of these variables is largely unknown. It is likely that these features are surrogates of important biological events rather than directly promoting cutaneous melanoma progression. In this article, we address the molecular mechanisms that drive these phenotypic changes. Furthermore, we present a comprehensive overview of recurrent genetic abnormalities, both germline and somatic, in relation to cutaneous melanoma subtypes, ultraviolet exposure and anatomical localization, as well as pre-existing and targeted therapy-induced mutations that may contribute to resistance. The increasing knowledge of critically important oncogenes and tumor-suppressor genes is promoting a transition in melanoma diagnosis, in which single-gene testing will be replaced by multiplex and multidimensional analyses that combine classical histopathological characteristics with the molecular profile for the prognostication and selection of melanoma therapy.

Practice points.

• Tumor thickness, ulceration, mitotic activity, gender and solar elastolysis are strong independent prognostic factors for cutaneous melanoma prognosis.

• Increasingly knowledge on the molecular underpinnings of these phenotypic correlates demonstrated that the activation of different signaling cascades can have the same end result and that the phenotype cannot be reduced to one or a few causative factors.

• Melanoma is a highly mutated tumor, characterized by an abundance of ultraviolet-induced DNA aberrations. The bulk of recurrent mutations that have been identified as melanoma risk factors or that are detected in sporadic cutaneous melanoma and their metastases are involved in DNA damage repair and proliferation. The frequency of most mutations is low (<1%) and their biological significance as a single marker in combination with others remains largely elusive.

• Although next-generation sequencing of melanoma has demonstrated large intertumor mutation heterogeneity, it has also led to the discoveries of recurrent and drugable driver mutations, passenger mutations and back-seat drivers. This accumulating insight indicates that a tumor should not only be tested for the presence of an actionable mutation, but also for potential pre-existing mechanisms of resistance. This is preferably carried out by parallel targeted sequencing rather than consecutive single-gene testing.

• In cases of melanocytic lesions of uncertain malignant potential, mutation analysis may not be the optimal approach, because many mutations are already present in nevi. Copy number variations and differences in expression levels of a number of genes can support the assessment of their malignant potential.

Melanoma is one of the solid tumors with the best characterized clinical and histopathological prognostic features. Primary lesion thickness, mitotic rate and ulceration are all strong, independent prognostic variables [1,2]. Other robust and easily assessed prognostic factors of melanoma include gender, age at diagnosis, the site of the lesion and the presence of metastasis to lymph nodes [1–4]. The staging system developed by the Melanoma Task Force of the American Joint Committee on Cancer (AJCC) uses the strongest prognostic factors in order to effectively stratify patients based on their risk of disease progression. However, the molecular mechanisms underlying these prognostic factors are poorly understood. Melanoma is also a paradigm in personalized medicine, illustrating how the discovery and further clinical validation of driver mutations can represent a breakthrough in cancer therapy. In this article, we will address the molecular mechanisms that may drive the strongest prognostic features of cutaneous melanoma, describe current knowledge of phenotype–genotype correlations and summarize the current standards in molecular pathology testing. The molecular pathology of uveal melanoma is different and isdescribed in detail elsewhere [5,6].

Phenotype–biology correlations

• Tumor thickness

In 1970, Alexander Breslow reported that thickness, cross-sectional area and depth of invasion were prognostic of cutaneous melanoma recurrence or metastasis rate at 5 years [7]. It is notable that although this report identified the most robust and reliable prognostic feature among all histological prognostic features ever described in cancer, this article is based on an error. Breslow considered maximal thickness to be an indicator of tumor burden and cross-sectional area to be the other important prognostic feature. We now know that the prognostic significance of Breslow's index is actually not related to tumor burden and that cross-sectional area does not predict clinical outcome. In 2009, the AJCC confirmed the prognostic significance of Breslow's thickness: the 10-year survival is 92% among the patients with T1 melanomas (0.1–1.0 mm) and is 50% in patients with T4 melanomas (≥4.1 mm). [2] Several studies have attempted to identify an expression signature associated with Breslow's thickness [8–17]. These studies identified only a few genes whose expressions change with increasing thickness, which include E- and N-cadherin [10,15], cadherin-19 [10,18–22], bcl2a1 [23–27], pcdh7 [28–34], rgs20 [35] and ALCAM/CD166 [36–42]. There are no data to support the view that melanoma thickness and Clark's level of invasion directly promote melanoma metastasis by, for instance, increasing the likelihood for the melanoma cells to encounter vessels. Breslow's thickness is probably an important phenotypic indicator of the biology of the melanoma cells at the leading edge of the tumor. Proteins whose expressions correlate with tumor thickness are commonly involved in cell survival and invasion. E-cadherin is a keratinocyte–melanoma adhesion molecule whose loss is required for the acquisition of an invasive phenotype [20,43–46]. Interestingly, this loss is mediated by the transcription factor Tbx3, which is also involved in suppressing melanocyte senescence through repressing the cyclin-dependent kinase inhibitors p19 (ARF) and p21 (WAF1/CIP1/SDII) [47,48]. The cadherin switch, especially the decreased expression of E-cadherin and the increased expression of N-cadherin, is an early phenomenon during melanoma progression that is associated with increased motility and invasiveness of the tumor and altered signaling, leading to decreased apoptosis and evasion of senescence [18,29,32,43,49–58]. Loss of E-cadherin expression in melanoma can be caused by gene loss, promoter methylation or inhibition of transcription. Promoter methylation and expression of proteins and miRNAs that regulate E-cadherin expression, such as snail, miRNA-200 and Gli2, are similarly associated with thickness and invasion [11,15,59]. Importantly, loss of E-cadherin expression affects β-catenin activity [28,60–64]. β-catenin anchors the actin cytoskeleton to E-cadherin, and loss of the latter causes β-catenin to move towards the nucleus, where it acts as a transcription factor that drives the expression of a wide variety of genes that promote invasion, such as urokinase-like plasminogen activator [65]. However, in contrast to carcinoma cells, the presence of melanocyte-specific microphthalmia-associated transcription factor (MITF) can attenuate β-catenin's proinvasive properties that are otherwise active in nonpigmented tumor cells [66]. In line with this, loss of β-catenin expression is part of a seven-marker signature predicting a high risk of disease recurrence [67]. Therefore, one may regard Breslow's index as a quantitative surrogate of the multifactorial biological machinery that drives melanoma progression and invasion.

• Melanoma ulceration

Tumor ulceration is associated with a poor prognosis in melanoma patients [1,68]. Thick melanomas have a higher incidence of ulceration than thin melanomas [69]. However, the prognostic strength of ulceration is independent of thickness [2]. The survival of patients with an ulcerated melanoma is significantly poorer than that of patients with a nonulcerated melanoma of equivalent T category. Interestingly, this adverse prognostic effect of ulceration remains robust even when the patient has metastatic disease in two or three lymph nodes. This suggests that ulceration is a phenotypic surrogate of an important biological event, rather than directly promoting metastatic evolution.

Several hypotheses have been proposed in order to explain the molecular changes underlying the adverse prognosis of ulceration. Factors related to the cellular biology of the tumor and the immunomodulation of the melanoma cells have been identified. Changes in the expression of cell adhesion molecules not only modulate intracellular adhesion and invasion, but also seem to have an effect on the inflammatory microenvironment. Loss of activated leukocyte cell adhesion molecule-mediated adhesion [42] triggers the expression of genes associated with the innate immune response [70]. Loss of E-cadherin expression induces an imbalance in inflammatory mediators and impairs keratinocyte control of melanocyte proliferation [44,71,72]. The N-cadherin-mediated interaction between fibroblasts and melanoma cells creates an imbalance of growth factor production, especially FGF-β, in a microenvironment that is already rich in melanoma-derived TGF-β [73]. The synergy between both growth factors can result in the recruitment of peripheral blood and bone marrow mesenchymal stem cells [74] and may drive the inflammatory response resulting in ulceration. Furthermore, loss of E-cadherin may result in increased β-catenin-regulated gene expression [21,75–79]. β-catenin is highly expressed in ulcerated melanomas [80], but whether this phenomenon is a cause or effect remains to be determined.

Venous leg ulcers elicit a persistent stimulation of the innate immune response and a strong Th1-like inflammatory response [81]; however, this is not thought to occur in ulcerating melanomas. A retrospective analysis of 537 consecutive micrometastatic sentinel lymph nodes with melanoma demonstrates that ulceration in a primary melanoma is associated with a lower density of mature dendritic cells in the sentinel node compared with sentinel nodes from nonulcerated melanomas [82]. Whether the lack of Th1 inflammation is due to constitutional host characteristics that also favor melanoma ulceration needs further investigations. Post hoc analyses and meta-analyses of several adjuvant interferon (IFN) therapy trials strongly indicate that patients with an ulcerated primary melanoma are far more sensitive to IFN than patients with nonulcerated primaries [44,83–86]. This suggests that melanoma ulceration is associated with a defect in Th1 response, rather than a strong, persistent activation thereof. The efficacy of adjuvant IFN in mounting an antitumor Th1 immune response in ulcerated melanomas is currently being studied in a clinical trial. The interaction between the inflammation-inducing mediators released by melanoma cells and the lack of activation of a Th1 response indicates complex and poorly understood reciprocal interactions between melanoma and inflammatory cells. In this context, the inflammatory response that is associated with improved outcomes [87] is a paradox. Thickness and melanoma cell proliferation, but not ulceration, were inversely associated with grade of inflammation. However, the composition of the inflammatory responses in melanoma with and without ulceration has not been studied extensively, and possible differences in helper T cell polarization in ulcerated and nonulcerated melanomas remain elusive.

• Mitotic activity

Proliferation of the primary melanomas, defined by the mitotic rate, is a powerful and independent predictor of survival [1,3]. As a result, the primary tumor mitotic rate is now a required element for the 2009 edition of the melanoma staging system. Data from the AJCC melanoma staging database demonstrate a highly significant correlation between increasing mitotic rate and declining survival rates (p < 10^-3). In a multifactorial analysis of 10,233 patients with clinically localized melanoma, mitotic rate was the second most powerful predictor of survival, after tumor thickness. Two large, validated gene expression profiling studies in melanoma that predict for the risk of metastasis or death reveal a strong representation of genes associated with replication or DNA repair [17,88]. The expression of proteins involved in the initiation of DNA replication, such as the DNA unwinding protein complex subunit mini-chromosome maintenance helicases MCM4 and MCM6, have a strong prognostic value for progression- and metastasis-free melanoma survival [17]. Firing of the origins of DNA replication is tightly regulated in order to duplicate DNA only once during the S phase and prevent aneuploidy (reviewed in [89]). In human cells, Cdc6 accumulates in the nucleus during G1 and binds to chromatin via the origin recognition complexes that occupy the origins of replication throughout the genome. Subsequent binding of the Cdt1–MCM2–7 complex and phosphorylation by cyclin A/CDK2 results in the release of phosphorylated Cdt1 and phosphorylated Cdc6 from this complex and marks the start of DNA replication [90,91]. Expression of the kinase that phosphorylates Cdc6 (i.e., CDK2) is of high prognostic value [92].

Highly proliferative cells, such as aggressive melanoma, require an effective DNA-repair machinery in order to correct deleterious errors that compromise genomic integrity. Overexpression of DNA-repair genes is associated with metastases or death [93,94]. Increases in postreplicative DNA-repair capacity associated with topoisomerase IIα could explain the spontaneous resistance of most melanomas towards radiotherapy and alkylating agents. In contrast to glioblastoma, promoter methylation of the counteracting DNA-repair enzyme MGMT does not predict clinical outcome of treatment with dacarbazine or temozolomide [95]. Mutations in mismatch-repair and nucleotide excision-repair genes have been described in familial melanomas and are described below (see ‘Constitutional defects in cutaneous melanoma’ section).

• Gender

The gender effect on survival is another unresolved mystery in the melanoma field. The male gender is associated with an adverse outcome that persists even after adjustment for other prognostic variables [2,96–100]. After adjustment, the relative excess risk of dying from melanoma is 1.85 (95% CI: 1.65–2.10) in males [96]. This gender impact on mortality risk is observed at all stages, even in patients with visceral metastases. No biological explanation has been identified so far. In particular, it is difficult to evoke a hormonal influence, as the adjusted risk estimates are similar among patients below 45 years or above 60 years of age. As most of the cancer testis antigens (CTAs) genes are located on the X chromosome, one possibility would be that CTA expression differs between females and males, but in fact, no difference has been observed. Moreover, data regarding the prognostic impact of CTA expression are conflicting. A possible confounding factor could have been differences in behavior towards ultraviolet (UV) exposure. However, the gender effect is unchanged when body site is introduced in the model, strongly suggesting that the survival difference between the sexes is not due to behavioral differences [99]. A possible explanation may be found in X-linked gene expression. Although one of the two X chromosomes in female cells is inactivated, this inactivation is incomplete and can result in a possible dosage effect of X inactivation-escaping genes in females compared with the expression of the same gene from the one X in males [101]. These escaping genes may be melanoma-suppressor genes, such as the UTX gene, coding for a H3-K27 demethylase, and WTX, which is involved in a subset of Wilms tumors [102,103]. Whether this would be sufficient to explain the gender difference is still unclear and requires further study.

• Solar elastosis

An important phenotypic variable in melanoma is the presence of histopathological features of chronic sun exposure damage, such as grade 2 or 3 solar elastosis. Melanomas arising in severely sun-damaged skin have a higher absolute number of mutations compared with those arising in sun-shielded locations. The majority of the nucleotide substitutions present in melanomas with solar elastosis are consistent with UV-induced damage, with a higher proportion of C>T and G>A – and more specifically CC>TT and GG>AA – transitions [102]. In addition to differences in the types of mutations found, solar elastosis correlates with a different spectrum of affected genes and chromosomal aberrations. In the absence of solar elastosis, melanomas show more frequent losses of 10p and 10q, while in the presence of solar elastosis, losses of 17p and gains of 15q are more common [104]. It has been clearly demonstrated that the presence or absence of solar elastosis correlates with the rate and the type of BRAF mutations [105–110]. Most notably, the prevalence of the oncogenic, non-UV-induced BRAF V600E mutation is tenfold higher in tumors arising on the trunk compared with those arising in sun-exposed sites [111]. The correlation with location is not as strong as that with solar elastosis, because not all sun-exposed sites with a melanoma display chronic sun damage marked by elastosis.

Molecular biology of melanoma

• Constitutional defects in cutaneous melanoma

Several constitutional susceptibility loci have been associated with an inherited cutaneous melanoma risk (reviewed in [112]). Germline mutations that affect cell proliferation are the most the common. The CDKN2A locus encodes for two separate proteins [113,114] and is frequently mutated in melanoma families (21% of familial melanoma cases in the Norwegian population) [115]. The estimate of relative risk for the association of any CDKN2A mutation with subsequent primary melanoma, adjusted for age, sex, center and phenotypic risk factors, is 4.3 (95% CI: 2.3–7.7) [116]. Importantly, different mutations in the CDKN2A gene may confer substantially different risks of melanoma. P14 binds to MDM2, thereby preventing p53 ubiquitination by MDM2 and subsequent degradation [117]. P16 prevents CDK4 and CDK6 from phosphorylating RB, which promotes G1–S transition [118,119]. CDK4 mutations that have been described in 17 families [120] similarly interfere with RB-mediated control over the cell cycle progression pathway and promote cell proliferation [121–123]. RB1 mutations predispose individuals to familial melanoma, as well as bilateral retinoblastoma [124,125].

Germline mutations in genes involved in DNA repair are associated with melanoma. Mutations in the deubiquitinase BAP1 increase BRCA1 ubiquitination and reduce BRCA1-mediated cell growth suppression [126]. BAP1 mutations are preferentially associated with cutaneous–ocular melanoma families [127]. Xeroderma pigmentosum patients carry mutations in the nucleotide excision DNA-repair enzymes of the xeroderma pigmentosum complementation genes and have a more than 1000-fold increase in melanoma risk [128]. Although polymorphisms in xeroderma pigmentosum genes have been associated with a mildly increased melanoma risk [129], this was not confirmed in a more recent study [130].

Germline mutations in melanin biosynthesis genes are associated with increased melanoma risk. MC1R codes for a G-protein-coupled receptor-activating adenylcyclase that, after binding of a melanocyte-stimulating hormone, increased MITF expression [131,132]. MC1R variants are strongly associated with skin and hair phenotype due to their effects on pigmentation. Germline mutations in MITF have been described in different families [133].

MITF expression levels in melanoma function as a rheostat with an effect on senescence, stem cell phenotype, proliferation and differentiation with increasing expression levels [134]. MITF controls the transcription of many genes, including pigmentation genes and HIF1A, and predisposes individuals to melanoma [135,136]. Mutations in MITF similarly affect melanocyte development and differentiation. More recently, it has been demonstrated that MITF is involved in the survival of UV-irradiated melanoma cells via the activation of the melanoma inhibitor of the apoptosis gene BRG1 [137].

Similarly, other low-penetrance single-nucleotide variations in pigmentation-related genes have been described. Genetic variation of the OCA2 enhancer is strongly correlated with gene expression and pigmentation [138], and two variants with this coding domain have been observed in melanoma families [139].

A recent study confirmed the association of previously identified nucleotide variations from five different genome-wide association studies in or near genes that potentially have an impact on melanogenesis (TYR, SLC45A2, MC1R and PIGU), cell cycle regulation (CDK10), cell growth and apoptosis (PLA2G6) or tumor suppression (CDKN2A) [140]. Genome-wide association studies identified TYRP1 variants as a risk factor for developing multiple melanomas, but this could not be confirmed in an independent population-based study [141]. Intriguingly, variant rs16891982 in SLC45A2, an iron transporter in melanosomes, shows a stronger association with melanoma in males (odds ratio: 5.5; 95% CI: 2.94–10.28) than females (odds ratio: 2.37; 95% CI: 1.69–3.31), but the potential gender difference and genetic contribution to melanoma risk remain largely elusive.

• Somatic alterations in cutaneous melanoma

Melanoma is the most frequently mutated cancer, with mutations predominantly caused by the mutagenic effect of UV radiation. Whereas some mutations occur frequently in different tumors, most mutations are rare or of unknown biological significance. A handful of genetic events have been demonstrated to be drivers on neoplastic progression. Next-generation sequencing efforts point towards a molecular melanoma classification. Melanomas arising in sun-exposed or sun-shielded sites display different mutation spectra [110]. As such, melanomas can be classified into three molecular groups [110]: sun-shielded melanomas with wild-type BRAF and NRAS are characterized by a high number of copy gains and a low mutation load and display molecular changes that contribute to the activation of the PI3K–AKT–mTOR pathway independent of CDKN2A or PTEN copy loss; sun-exposed melanomas with wild-type BRAF and NRAS with a high mutation burden but few copy number alterations typically originate in older patients and frequently contain deleterious mutations in NF1, TP53, ARID2 and PTPRK; and sun-exposed melanomas with mutations in BRAF or NRAS with frequent copy losses in PTEN and/or CDKN2A, as well as copy gains and point mutations in genes whose products modulate the RAS–RAF–MEK–ERK signaling cascade.

The MEK–MAPK pathway and/or the PI3K pathway are often activated in melanoma (reviewed in [142]). Mutations in NRAS (Q61K [34%], Q61R [35%], Q61L [8%] and G12D [4%]) have been reported in 15–25% of melanomas and seem to be more frequently activated in melanomas due to chronic sun damage [143–145]. The V600 gain-of-function mutation in BRAF (V to E [70%], K [10–30%] or R [<5%]) can be detected in approximately 50% of all melanomas [142]. The frequency of BRAF mutations in primary cutaneous melanomas is not significantly different from that in metastatic lesions. A few cases of discrepancies between the primary tumor and subsequent metastases or between metastases have been reported, but this appears to be a rare situation [146–148]. The incidence rate for BRAF V600K varies by region, and a higher amount of cumulative sun-induced damage is associated with BRAF V600K, but not with BRAF V600E melanomas [149]. BRAF mutations carry a weak adverse prognostic value [150], but these three mutations are associated with sensitivity to the clinically approved BRAF inhibitors vemurafenib and dabrafenib [151]. Concurrent NRAS and BRAF mutations in one melanoma cell appear to be mutually exclusive, but tumors carrying both a mutation in NRAS and BRAF have been observed with a frequency of approximately 5%, suggesting the coexistence of two cell populations in which the RAS–RAF–MAPK cascade is activated at two different levels at least [152].

MEK1 and MEK2 are downstream from RAS and RAF, on the same MAPK pathway [142]. MEK1 is encoded by MAP2K1 and MEK2 by MAP2K2. Activating mutations of MEK1 and MEK2 are found in 8% of melanomas [153].

The PI3K pathway is activated through a PTEN loss-of-function abnormality (most often deletion) in 20–40% of melanomas [154–156]. Activating mutations or amplifications of PI3K or of AKT1 can also be found in some melanomas, although inhibitors of this pathway have not yet shown significant efficacy in any melanoma subtype.

In addition to germline mutations in MITF and BAP1 in familial melanoma or patients with multiple primary melanomas, somatic aberrations have been described. MITF is amplified in approximately 4% of melanomas [157]. Somatic BAP1 mutations have been described in 5% of sporadic cutaneous melanomas, predominantly of the desmoplastic subtype [158].

Mutations in KIT are noted in less than 1% of melanomas overall; their occurrence is more frequent in acral and mucosal sites [159,160]. Activating mutations in and gene amplification of KIT have been initially reported in 39% of mucosal melanomas, 36% of acral melanomas and 28% of sun-exposed skin-asssociated melanomas. However, further studies found much lower proportions of approximately 15% in acral sites and below 5% in mucosal sites [161,162]. It is likely that these variations reflect both an overestimation of the frequency of KIT mutations in acral melanomas in the initial reports and sensitivity differences in the detection technique. Mutations in exons 11 and 13 make up 85% of the KIT mutations reported in melanomas and are associated with sensitivity to imatinib [163,164]. Other mutations are reported in exon 17. The major differences seen in gastrointestinal stromal tumors are that melanomas have more point mutations than deletions or insertions, have more frequent mutations in exons 13 and 17 and have more amplified wild-type KIT. These amplifications seem to be predictive of no response to anti-KIT therapies [165]. KIT mutations that activate the MAPK pathways are reported in 20% of cutaneous melanomas with NRAS mutations.

Whole-exome sequencing of melanoma [110,157] has identified a number of additional potential oncogenic drivers, of which some have been confirmed to have tumor growth-promoting capacities. In addition to the genes described above, the landscape of genetic abnormalities include gains of 5p13 (RICTOR and others), 11q13 (CCND1 and others) and 12q14 (CDK4), losses of 9p21 (CDKN2a and others) and 10q23 (PTEN) and mutations in DCC, TNC, TP53, PTPRK, PPP6C, TLR4, CD163L1, GRM3, NPAP1, SLC15A2, RAC1, MAGEC1, JAKMIP2, RAC1, NF1, SNX31, TACC1, STK19 and ARID2. Recently, a highly recurrent mutation (71–74%) in the promoter of the telomerase reverse transcriptase gene TERT has been described in two studies [166,167]. As a consequence, a two- to four-fold increase in the transcriptional activity of TERT has been observed in vitro [166]. Notably, TERT germline mutations have been observed in one family [167]. A possible functional consequence is supported by the observations of increased telomerase expression and activity with melanoma progression [168].

Recently, it has been demonstrated that although BRAF inhibitor–resistant melanomas display mutations that are not detectable in the pretreatment lesion (e.g., PIK3CA and the deletion, indel or mutation of PTEN), the therapy-resistant lesion can share mutations of the primary lesion that may contribute to early therapy failure, such as mutations in RAC1P29S [169]. Similarly, mutations in MAP2K2 and MITF were present in the pretreatment and therapy-resistant biopsies, which may confer resistance to anti-BRAF therapy [170]. These findings add to a growing list mechanisms that confer BRAF inhibitor resistance (for a review, see [171]).

Molecular testing in melanoma

The identification of the druggable BRAF mutation was an important push towards the molecular classification of melanoma [172,173]. Single-gene testing for the BRAF V600 mutation, such as with real-time PCR and using anti-BRAF antibodies, can select patients that are eligible for receiving vemurafinib or dabrafenib. The real-time PCR cobas^® 4800 assay (Roche Molecular Diagnostics, CA, USA) has little cross-reactivity for non-V600E profiles [174]. Therefore, when this method is used, tumors with negative results should also be screened for V600K mutations with another technique. Similarly, the V600 mutation-specific VE1 antibody has a high sensitivity (97%) and specificity (98%) for the detection of the V600E mutation, but not for V600K [175–180]. Therefore, unbiased sequencing of the BRAF gene is preferred over a specific test for the presence of BRAF V600E on either the DNA or protein level. Although these BRAF inhibitors induce an objective therapeutic response in almost all patients, the duration is short, with almost all patients developing therapy resistance within 1 year [181,182]. With a more complete overview of the molecular landscape of mutations in melanoma, it has become clear that reactivation of ERK can occur at many levels, and that coexisting gene mutations, such as MAP2K2 and MITF, can predict early treatment failure. This supports the use of the multiplex testing of the tumor for druggable targets, as well as the genetic aberrations that are known to mediate a bypass of the RAS–RAF–MEK–ERK pathway. The presence of one or more specific genetic alterations drives clinical decision-making with respect to the choice of therapy [110,157]. Therefore, with our current knowledge, single-gene testing for melanoma has become futile and should be replaced by multiplex testing using a targeted approach in order to obtain a more complete molecular profile of druggable targets, including possible pre-existing mechanisms of BRAF inhibitor resistance.

An algorithm for molecular testing should be based on the present therapeutic situation. Next-generation sequencing technologies are now common for whole-genome, whole-exome and whole-transcriptome sequencing (i.e., RNA sequencing) of tumors in order to identify point mutations, structural or copy number alterations and changes in gene expression in a research setting. In a diagnostic setting, targeted mutation profiling platforms, such as MiSeq™ (Illumina, CA, USA) or Ion Torrent™ platform (Life Technologies, CA, USA), are well suited. At a 500× read depth, one can confidently call a single-nucleotide variant at a frequency of 1%. The flexibility to change the panel of genes that one wants to interrogate, shorter turnaround times and relative ease of interpretation of the data make these technologies more adapted for clinical use. However, major hurdles need to be overcome before these technologies can become part of the routine molecular diagnostic laboratory. A major challenge is the development of the appropriate software that can deal with the very high number of short-sequence reads generated by these platforms. For example, quality filtering settings, variations in read depth and choices of reference affect the calling of a single-nucleotide variant. This is especially important for recognizing low-frequency polymorphisms [183–185]. Therefore, the mutation profile of a tumor should be preferentially compared with the patients' normal DNA (e.g., from blood) in order to identify tumor-associated mutations.

At what stage should a multiplex molecular profile be obtained? Without the existence of an active adjuvant-targeted treatment, testing a patient without metastases for targetable mutations in routine clinical practice is controversial. In clinical practice, it is more logical to test the metastasis. In the case of vemurafenib, a stronger positive impact of progression-free and overall survival was observed for patients with a relatively low tumor burden [186]. This argues for rapid testing, maybe as early as the primary lesion, in cases of high-risk melanoma patients (stage 2B and higher). However, although the recurrence rates for T1b and T2a may be 20%, in breast cancer, HER2 testing is performed on small tumors with much lower recurrence risks. Tissue from a primary melanoma is often scarce and does not allow consecutive testing for individual genes. Therefore, a multiplex molecular profile could be obtained from the primary lesions. If the patient progresses to stage 4 disease, rapid single-gene tests can confirm whether the previously identified druggable target genes are still present in the metastasis in order to gain valuable time for starting the treatment.

Molecular testing in difficult melanocytic lesions

A thorough histopathologic examination is the gold standard for classifying melanocytic lesions as benign or malignant. While this approach is adequate for the majority of melanocytic lesions, in certain situations, ambiguity can arise. Spitz nevi, recurrent nevi, blue nevi, mechanically irritated nevi and nodular proliferations in congenital nevi are all lesions that are prone to having overlapping features of both benign and malignant tumors [187]. The correct identification of the malignant potential of these lesions is crucial for guiding appropriate patient treatment and follow-up care. Currently, single-gene testing is of no added value in ambiguous lesions because robust mutations that are specific to nevi or melanomas have not been identified. Classic melanoma-associated mutations in BRAF, NRAS, HRAS, GNAQ and GNA11 have all been described in nevi [188,189]. However, for spitzoid lesions of uncertain malignant potential, HRAS mutations are strongly associated with a very low risk of recurrence or metastasis [190]. In addition, multiple gene expression testing can be of diagnostic value. A nine-gene expression panel in combination with the AJCC staging system has been reported to have prognostic value for overall survival [191]. This panel of genes shows no overlap with the prognostic seven-immunohistochemical marker signature described in 2012 [67]. The prognostic value of these and other published signatures needs to be validated in the context of auxillary testing for ambiguous lesions. Myriad Genetics (UT, USA) recently launched the myPath™ Melanoma Test that differentiates malignant melanomas from benign skin lesions using a 23-gene expression panel (with a claimed sensitivity of 89% and specificity of 93%).

The most striking genetic difference between nevi and melanomas is the presence of chromosomal aberrations. More than 95% of melanomas harbor chromosomal aberrations, with a mean of 7.5 events. Common gains include 6p, 1q, 7p, 7q, 8q, 17q and 20q, and common losses include 9p, 9q, 10q, 10p and 6q [104,192]. In stark contrast, chromosomal aberrations are found in less than 15% of melanocytic nevi and tend to affect a single chromosome; Spitz nevi account for the majority of these nevi and are associated with an isolated 11p gain [104]. The pronounced genomic instability of melanomas can be exploited by FISH for diagnostic purposes. The standard melanoma FISH assay uses probes in order to detect copy number changes in RREB1 (6p25), MYB (6q23), CCND1 (11q13) and the centromere of chromosome 6 (6p11.1–q11.1). This combination of probes is sensitive (82–94%) and specific (90–98%) for the majority of melanocytic lesions. In particular, the standard melanoma FISH test is well suited for distinguishing a sclerosing nevus from a desmoplastic melanoma and a blue nevus from a blue nevus-like melanoma. [189] In other lesions, such as spitzoid melanomas, it has been shown that adding probes for 9p21 and centromere 9 to the classic cocktail greatly increases the sensitivity of the FISH test [193]. As our understanding of the chromosomal changes that are specific to different types of melanoma evolves, probes for different loci are being investigated that will probably lead to a more tailored approach in the future. Rather than having a common FISH panel in order to determine the malignant nature of all melanocytic lesions, panels could be developed to specifically address common histologic differentials, which has the potential to greatly increase the accuracy of FISH.

Conclusion & future perspective

The melanoma landscape has considerably changed and a large proportion of patients with metastatic melanoma, and hopefully soon with high risk primary melanomas, will directly benefit of these scientific progresses. However, several unmet needs are still hampering the development and rationalization of therapies. One of them is the absence of a strong predictive biomarkers for immunotherapy. Also, in the context of active adjuvant therapies to come, there is a strong need to refine prognostication in melanoma based on the mutation and gene expression profiles. With decreasing costs of multigene sequencing, this technology will most likely replace single-gene testing in a clinical setting in order to identify patients carrying druggable mutations, as well as the presence of possible mechanisms of resistance. The WINTHER trial (NCT01856296) and the SPECTA initiative launched by the European Organization for Research and Treatment of Cancer represent breakthrough concepts for matching tumor biology (gene mutation and expression status) and therapeutics in individual cancer patients, including melanomas. Furthermore, the prognostic and predictive value of combining genetic information with classical histopathological parameters needs further investigation. Importantly, the use of artificial biomarker breakpoints (the presence or absence of mitosis or Breslow thicknesses of 0–1, 1–3 or >3 mm) will probably be replaced by continuous variables in a multidimensional prognostic model. The pace of new biomarker development will quickly make it impossible to update the list of prognostic variables to be assessed each time a new biomarker is identified. Major improvements will come from shared computerized tools, which will help us to generate continuous likelihood scores for diagnosis, prognosis and response to treatment predictions (e.g., [194]). This will lead to the development of platforms that can be used by scientists from different fields in order to integrate and share high-quality data in the precompetitive setting and generate new probabilistic causal models. This is also especially important in the context of the genetic landscape of therapeutic resistance.