Non-BRAF Mutant Melanoma: Molecular Features and Therapeutical Implications

Melanoma is one of the most aggressive tumors of the skin, and its incidence is growing worldwide. Historically considered a drug resistant disease, since 2011 the therapeutic landscape of melanoma has radically changed. Indeed, the improved knowledge of the immune system and its interactions with the tumor, and the ever more thorough molecular characterization of the disease, has allowed the development of immunotherapy on the one hand, and molecular target therapies on the other. The increased availability of more performing technologies like Next-Generation Sequencing (NGS), and the availability of increasingly large genetic panels, allows the identification of several potential therapeutic targets. In light of this, numerous clinical and preclinical trials are ongoing, to identify new molecular targets. Here, we review the landscape of mutated non-BRAF skin melanoma, in light of recent data deriving from Whole-Exome Sequencing (WES) or Whole-Genome Sequencing (WGS) studies on melanoma cohorts for which information on the mutation rate of each gene was available, for a total of 10 NGS studies and 992 samples, focusing on available, or in experimentation, targeted therapies beyond those targeting mutated BRAF. Namely, we describe 33 established and candidate driver genes altered with frequency greater than 1.5%, and the current status of targeted therapy for each gene. Only 1.1% of the samples showed no coding mutations, whereas 30% showed at least one mutation in the RAS genes (mostly NRAS) and 70% showed mutations outside of the RAS genes, suggesting potential new roads for targeted therapy. Ongoing clinical trials are available for 33.3% of the most frequently altered genes.


INTRODUCTION
Cutaneous melanoma is one of the most aggressive malignancies of the skin. Its incidence is globally growing partly due to the increase of early diagnoses, and contextually, the prevalence is also increasing (Bray et al., 2018;Schadendorf et al., 2018). Until 10 years ago, advanced melanoma was associated with poor survival due to the lack of durable responses to conventional chemotherapy and biochemotherapy (Korn et al., 2008), with a median Overall Survival (OS) of about 6 month in patients with stage IV melanoma. Since 2011, however, the rules of the treatment of stage IV melanoma have been completely rewritten, with the introduction of targeted therapies with BRAF and MEK inhibitors Long et al., 2014;Robert et al., 2016), and immunotherapy with the anti CTLA-4 ipilimumab (Hodi et al., 2010) and the anti-PD-1 nivolumab (Robert et al., 2015) and pembrolizumab (Schachter et al., 2017). These new therapeutic approaches improved melanoma prognosis, resulting in a 5-year survival rate of 34-43% (Hamid et al., 2019;Robert et al., 2019). However, mainly because of primary and acquired resistance to treatments, the majority of patients will ultimately relapse, and only patients harboring a BRAF mutation, observed in about 50% of cutaneous melanoma, can receive a targeted treatment with BRAF and MEK inhibitors (Spagnolo et al., 2015). The current state of molecular-target drugs and the current therapeutic scenario for patients with BRAF mutated melanoma, from the introduction of BRAF inhibitors as single agents to modern clinical practice, has been extensively described in a related minireview . With the purpose of further improving the prognosis of melanoma patients, several preclinical and clinical trials are studying new actionable mechanisms and/or molecules, to simultaneously tackle multiple resistance mechanisms.
The aim of this review is to describe the landscape of mutated non-BRAF melanoma, in light of recent data deriving from Next-Generation Sequencing (NGS) (or Massive Parallel Sequencing -MPS) analysis, focusing on available, or in experimentation, targeted therapies. The advent of MPS, allowing the simultaneous analysis of several genes, led, in the past two decades, to Whole-Exome Sequencing (WES) and Whole-Genome Sequencing (WGS) studies that found several mutated genes in human cancers. The evolution of molecular testing in melanoma, as well as the main techniques and MPS platforms currently in use for BRAF mutation testing, have been recently reviewed .
The first actionable mutation to be targeted by specific drugs in melanoma, BRAF V600, was found in 2002 along several other drivers of human cancers (Davies et al., 2002). Since then, several other genes have been identified as putative drivers of melanomagenesis and/or melanoma progression, and additional candidate drivers are currently being assessed, prompting pharmacogenomics studies on potentially actionable targets (Priestley et al., 2019). However, melanoma is one of the tumors with the highest mutation burden, and results from different studies were frequently not overlapping, possibly due to dissimilar sample size and cohort characteristics (Berger et al., 2012;Hodis et al., 2012;Krauthammer et al., 2012;Snyder et al., 2014;Van Allen et al., 2015). Although this high mutational burden is one of the reason behind the success of immunotherapy in this tumor, it makes it hard to clearly identify novel driver genes that could be used for targeted therapies (Davis et al., 2018).
In 2015, The Cancer Genome Atlas analyzed 333 cutaneous melanoma samples by integrating integrated multi-level genomic analyses, namely WES and low-pass WGS, transcriptome sequencing including miRNA, protein expression, and classified melanoma in four major molecular subtypes: mutant BRAF, mutant RAS, mutant NF1 and triple wild-type (The Cancer Genome Atlas Network, 2015). However, as NF1 mutations can be found in melanomas with concurrent BRAF or NRAS hotspot mutation, a three-group classification of melanoma (mutant BRAF, mutant RAS, non-BRAF mut /non-NRAS mut ) has been proposed (Palmieri et al., 2018). Although providing an unprecedented insight into the complex mutational spectrum of melanoma, the TCGA study cohort did not include acral and mucosal melanomas. Two years later, this issue was addressed with the molecular characterization of 183 melanoma samples through WGS, including the acral and mucosal subtypes (Hayward et al., 2017). Recently, a joint effort by the TCGA and the ICGC resulted in the description of the molecular spectrum of the largest whole genome dataset of 38 different tumor types, which included a subset of 118 melanoma samples previously described (ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium, 2020).
The spectrum of genomic alterations in melanoma involve multiple genes and signaling networks, but the most frequently altered pathways are MAPK, PIK3CA, KIT signaling, and apoptosis/cell senescence pathways (Figure 1). This review focuses on cutaneous melanoma, including the acral melanoma.
To obtain an overview of molecular alterations in skin melanoma, we focused our analysis on all WES or WGS studies on melanoma cohorts or pan-cancer cohorts that included melanoma, for which information on the mutation rate of each gene was available. For selecting TCGA melanoma samples we considered the group's pan-cancer flagship paper (Hoadley et al., 2018), which included the original cohort of the melanoma-only TCGA study, plus an additional set of 60 skin melanoma samples. With these criteria we collected and combined mutational data from 10 studies published from 2012 to 2019, available from either the cBioportal repository (cBioPortal for Cancer Genomics, 2020) or from the tables within the published manuscripts (Hayward et al., 2017;Birkeland et al., 2018;Kontogianni et al., 2018;Wilmott et al., 2019), to obtain cumulative mutational frequencies of all genes assessed by these studies (Table 1) (Berger et al., 2012;Hodis et al., 2012;Krauthammer et al., 2012;Snyder et al., 2014;Van Allen et al., 2015;Hayward et al., 2017;Kontogianni et al., 2018;Wilmott et al., 2019;cBioPortal for Cancer Genomics, 2020). When present, we filtered out uveal and mucosal melanoma samples, as well as synonymous variants. Moreover, only three studies used for this review were performed with WGS (Berger et al., 2012;Hodis et al., 2012;Hayward et al., 2017;Wilmott et al., 2019), while all the others are WES studies. Non-coding regions were not considered for our analysis, except for TERT promoter, whose frequency was calculated on data from a single WGS study (Hayward et al., 2017). Similarly, we used the WGS subset to assess Copy Number Variations (CNVs). An overview of the studies analyzed for this review is found in Table 1.
In the following sections, we provide a description of 33 selected established and candidate melanoma driver genes, as well as the mutational and CNV frequency of each gene (Figures 2, 3) in the melanoma samples analyzed, and we describe available, or in experimentation, targeted therapies for each gene/pathway, excluding immunotherapy. Mutational frequency of these genes in melanoma across each study included in the review is displayed TABLE 1 | Skin melanomas samples from the 10 NGS-based studies evaluated for the mutational and CNV frequency in the present report.  in Supplementary Figures S1-S10; ongoing targeted clinical trials are displayed in Table 2.
Moreover, for each of the 33 genes included in this review and listed here by mutational frequency and/or pathway, we also provide molecular function, frequency alterations (mutations plus CNV) in our dataset, and ranked their effect on melanoma according to the DisGeNET Gene-Disease Association (GDA) Score (Piñero et al., 2015) (Figure 4).

RAS GENES (NRAS/KRAS/HRAS)
Ras is a superfamily of small GTPase proteins that regulate cell growth, survival, differentiation and play a key role in transmitting the signal from Receptor Tyrosine Kinases (RTK) to several downstream signaling pathways, in particularly MAPK (mitogen-activated protein kinase) and PI3K (phosphoinositide 3-kinase). The three RAS tissue specific isoforms (NRAS, neuroblastoma ras viral oncogene homolog; KRAS, Kirsten rat sarcoma viral oncogene homolog; HRAS, Harvey rat sarcoma viral oncogene homolog) are frequently mutated in cancers. In cutaneous melanoma, NRAS is mutated in about 15-30% of cases, while KRAS (∼2%) and HRAS (∼1%) play a minor role (Milagre et al., 2010). NRAS, KRAS, and HRAS, mutations and CNVs in our dataset are reported in Figures 2, 3 and Supplementary Figures S1A-C.

NRAS
The NRAS was the first oncogene identified in melanoma in 1984, and the second most prevalent after BRAF (mutation FIGURE 2 | Somatic mutations frequency in top melanoma driver cancer genes based on 992 skin melanoma samples derived by the 10 selected NGS studies. Somatic coding mutations were considered for the analysis excluding synonymous variants. Red asterisks indicate the presence of clinical trials for the altered gene.
FIGURE 3 | CNV frequency in top melanoma driver cancer genes. Amplifications and deletions are reported in red and blue, respectively. Only four studies with available CNV information were considered for CNV analysis. Red asterisks indicate the presence of clinical trials for the altered gene.
Frontiers in Molecular Biosciences | www.frontiersin.org   frequency of 30%) (The Cancer Genome Atlas Network, 2015). NRAS mutations primarily occur at position 61 (Q61R/K/L accounts for about 80%) and, less frequently, at positions 12 and 13 (G12/13 accounts for about 6%) (Bos, 1989;Lee et al., 2011). These mutations cause an altered GTPase activity that keeps NRAS activated: this alwayson signal induces a constitutional activation of the whole MAPK pathway with cell proliferation, dysregulation of the cell cycle and activation of other pro-survival pathways . Melanoma patients harboring mutated NRAS display different characteristics compared to those harboring mutated BRAF: they are older (>55 years) and have a story of UV exposure, thicker primary tumors and a higher rate of mitosis. Several studies showed that NRAS mutations may result in an inferior clinical outcome with a shorter Melanoma-Specific Survival (MSS) FIGURE 4 | Molecular function (A) and ranking (B) of the 33 genes analyzed. (A) Gene Ontology (GO) molecular function of the 33 genes. GO molecular function was obtained using FunRich v3.1.3, a stand-alone software tool (Pathan et al., 2015). (B) The table reports the frequency of genomic alterations (mutations and CNV) resulting from the 10 selected NGS studies, as well as the Gene-Disease Association (GDA) Score obtained from the DisGeNET (Piñero et al., 2015) database using the keyword "melanoma" (C0025202 result). The GDA Score is computed by integrating evidence from multiple sources. The higher the GDA Score, the more reliable the gene-disease relationship is. (Devitt et al., 2011;Ellerhorst et al., 2011), although partially debated (Omholt et al., 2003;Akslen et al., 2005;Edlundh-Rose et al., 2006;Ugurel et al., 2007;Ellerhorst et al., 2011).
The high number of NRAS mutations in cutaneous melanoma did not allow the development of effective drugs: targeting directly NRAS remains a great challenge, and the target therapy for NRAS mutant melanoma is focused on MEK inhibitors.

KRAS
KRAS encodes for two proteins resulting from alternative splicing of exon 4, KRAS4A and KRAS4B. These proteins have different structures in their C-terminal region and use different mechanisms to localize to cellular membranes (Welman et al., 2000). KRAS4B is the most frequent in human cells and differs from KRAS4A for a C-terminal residue which allows to bind calmodulin and induce its phosphorylation by PKC (Villalonga et al., 2001;Sung et al., 2013). Binding between KRAS4B and calmodulin seems to determine drug resistance, facilitate tumorigenesis and express stem-like markers on the cell surface . KRAS mutations occur most commonly at codon 12 but also at 13 and 61 (Pylayeva- Gupta et al., 2011;Liu et al., 2019). Mutation of G12 interferes with GAP binding and GAP-stimulated GTP hydrolysis and represents ∼12% of all KRAS mutations (Tate et al., 2019;COSMIC, 2020). The mutation in G13 decreases GAP binding and its hydrolysis while mutation in codon 61 has a direct role in inhibiting the hydrolysis reaction (Ostrem and Shokat, 2016). KRAS mutations are found in 15-20% of cancers, mostly in colorectal and pancreatic adenocarcinomas (Chiosea et al., 2011;Hartman et al., 2012;Zhou et al., 2017;Román et al., 2018). In melanoma, KRAS mutations are rare (1.7% of our cases) occurring almost exclusively in codon G12 (Milagre et al., 2010). KRAS mutant remains undruggable.

HRAS
HRAS encodes for a GTPase involved in regulating cell division in response to growth factor stimulation (Wong-Staal et al., 1981). Mutations in HRAS cause cell overgrowth and are implicated in a variety of cancers (Rauen, 2007). HRAS is altered in 0.97% of all cancers and, rarely, in 1.5% of melanoma (AACR Project Genie Consortium, 2017). Moreover, HRAS has been shown to be mutated in Spitz nevi both by CNVs (12/102; 12%) and point mutations (8/12; 67%) (Bastian et al., 1999(Bastian et al., , 2000. The reason why mutations in HRAS lead to Spitz nevi is unclear but could be related to higher affinity for the PI3K-PKB/AKT pathway which would be able to drive the symmetrical overgrowth of cells with an epithelioid morphology without marked activation of the melanizing pathways (Ross et al., 2011). Spitz nevus should not be regarded as a precursor lesion of melanoma. HRAS mutation analysis seems to be useful in the differential diagnosis between Spitz nevus and Spitzoid melanoma, and the presence of HRAS mutations is a marker of benignity and/or favorable clinical outcome (Dimonitsas et al., 2018).

RAS Genes -A Clinical Approach
To date, therapeutic inhibition of RAS remains an unmet need. Indeed, there are no approved therapies that specifically target NRAS, KRAS or HRAS and RAS-mutated patients are usually treated with immunotherapy. However, several trials are ongoing ( Table 2).
Several years ago in vitro studies had already shown that mutated NRAS melanoma cells were sensitive to MEK inhibitors (Solit et al., 2006). However, the efficacy of such drugs on mutated NRAS cells was lower than that observed in mutated BRAF cells, and this lower efficacy could be explained, at least in part, if we consider the complexity of the molecular pathways network involving RAS. From these observations, two hypotheses arose: first, that probably a therapy based on a single drug could be insufficient; second, that the best therapeutic possibility was to target RAS directly (Mandalà et al., 2014). Unfortunately, targeting directly RAS did not give the expected results. Indeed, acting on GTP binding pocket in RAS protein is difficult due to the excessive affinity between RAS and GTP (Baines et al., 2011). In the same way, the inhibition of farnesylation of a cysteine residue, a post translational modification necessary to RAS insertion to the plasma cell membrane, has proven ineffective. Due to these disappointing results, the therapeutic strategies for mutated NRAS melanoma reverted on MEK inhibitors.
Initially, the use of MEK inhibitors led to modest results, with an Overall Response Rate (ORR) of 10% and a high incidence of adverse events (Rinehart et al., 2004;LoRusso et al., 2010). Subsequently, the MEK 1/2 inhibitor selumetinib (AZD-6244) was developed (Yeh et al., 2007). Phase I trial enrolled 11 melanoma patients and showed promising results (Adjei et al., 2008, 142886); on this wave, phase II trials comparing selumetinib and chemotherapy (temozolomide and docetaxel) in BRAF-WT and NRAS-unselected melanoma patients were initiated. The results were unsatisfactory for both trials, with no difference in efficacy outcomes (Kirkwood et al., 2012;Gupta et al., 2014).
In preclinical studies binimetinib inhibited the growth of NRAS and BRAF mutated melanoma tumor cells (Winski et al., 2010). In the subsequent phase II trial (Ascierto et al., 2013) MEK162 obtained an ORR of 14.5% and a Disease Control Rate (DCR) of 56%. Median Progression Free Survival (PFS) was 3.6 months, underlying a rapid development of resistance, and the median OS 12.2 months (van Herpen et al., 2014). Phase III trial (NEMO) (Dummer et al., 2016) compared the efficacy of MEK162 versus dacarbazine in 402 NRAS mutated, melanoma patients. The ORR was 15% in the binimetinib arm versus 7% in dacarbazine arm. Furthermore, binimetib significantly prolonged median PFS, with 2.8 months [3.9 in patients with normal lactate dehydrogenase (LDH)] versus 1.5 months, respectively [hazard ratio (HR), 0.62]. Interestingly, immunotherapy pretreated patients had a longer median PFS (5.5 months). However, no differences in terms of OS were noted, and binimetinib as single agent did not receive regulatory agencies approval for the treatment of NRAS-mutated melanoma.
Finally, pimasertib (PIM; AS703026) has been evaluated in a phase I trial with encouraging results. Phase II study enrolled 194 patients to be treated with pimasertib or dacarbazine. Results showed a significant benefit for pimasertib, with a median PFS of 13.0 versus 6.9 weeks and a DCR of 37.7% versus 26.6%. Unfortunately, no difference in OS was showed (8.9 vs. 10.6 months) (Lebbe et al., 2016).
Combination of MEK inhibition with other targets is currently being evaluated. Among all, the results of a combination trial performed on 14 patients treated with ribociclib (LEE001) and MEK162 are particularly interesting. Indeed, the combination obtained six partial response and six stable disease, with a DCR of 85% (Sosman et al., 2014).
As we mentioned before, NRAS has to undergo some post-translational modifications, like farnesylation (Konstantinopoulos et al., 2007). Farnesyl Transferase Inhibitors (FTIs), like lonafarnib, was developed in an attempt to exploit this phenomenon. FTIs inhibit the function of RAS and seem to be able to sensitizing melanoma cells to RTK inhibitors like sorafenib (Meier et al., 2009). Unfortunately, FTIs failed in clinical trials showing no efficacy against NRAS and KRASdriven cancers (Smalley and Eisen, 2003;Konstantinopoulos et al., 2007;Niessner et al., 2011;Gajewski et al., 2012;Margolin et al., 2012;Cox et al., 2015). The reason of this failure seems to be due to the action of geranylgeranyltransferase I (GGTase I) in the alternative prenylation (Whyte et al., 1997). FTIs in combination with GGTase I inhibitors have been tested in clinical trials but turned out to be too toxic (Brock et al., 2016). Other approaches attempting to inhibit some post-translational modifications are currently being evaluated. However, the great limit remains the toxicity of these drugs and challenges in delivering siRNA using nanoparticles (Davis et al., 2010).

PREX2
The PREX2 gene plays oncogenic roles in human cancers, such as melanoma, since it is involved in PIK3CA-PTEN-AKT signaling pathway (Fine et al., 2009;Srijakotre et al., 2017). It has been demonstrated that ectopic expression of mutant PREX2 accelerates tumor formation of immortalized human melanocytes in vivo (Berger et al., 2012). The upregulation of PREX2 and its mRNA increasing expression led to the AKT activation by PTEN phosphorylation and increases tumor proliferation. PREX2 non-synonymous variants have been found in 44.0% of a 25 patients cohort and in 14.0% of a 107 melanoma samples validation cohort (Berger et al., 2012). Notably, the truncating mutation PREX2 E824 * was further studied to determine its in vivo implications in the context of mutant NRAS (Lissanu Deribe, 2016a). PREX2 truncation E824 * was found to cooperate with NRAS mutations but not with BRAF V600E mutation, to accelerate melanoma development (Lissanu Deribe, 2016a). In 2017, 100 primary melanoma samples were analyzed by targeted NGS for 35 melanoma-related genes, and PREX2 mutations were reported in 14 samples (de Unamuno Bustos et al., 2017). Interestingly, recently eighty patients with conjunctival melanoma were examined by WES, identifying three PREX2 mutations (frequency of 37%) (Demirci et al., 2019). In our skin melanoma dataset, mutations and CNVs in PREX2 gene are found at frequency of 22.3 and 4.9%, respectively, as shown in Although these evidences point to PREX2 as a key player in melanoma, suggesting that PREX2 may be a potential therapeutic target, to date no clinical trials are available.

TP53
The TP53 tumor suppressor gene, which is universally recognized as the "guardian of the genome, " prevents the cell from dividing and promotes apoptosis. Moreover, it is the most frequently mutated gene in human cancer with a significant prevalence of missense mutations (Hainaut and Pfeifer, 2016). The TCGA database analysis identified across 32 different cancer types 3,786/10,225 (36.8%) patients with TP53 somatic mutations (Donehower et al., 2019).
Loss of TP53 was uncommon in thicker, early or in situ melanoma, but is commonly reported as a late event in the development of melanoma (Shain et al., 2015). The Ultra-Violet Radiation (UVR) role on melanoma formation in BRAF V600E mice is often associated with mutations in TP53 (Viros et al., 2014). In our data, mutations and CNVs in the TP53 gene are present at a frequency of 14.9 and 1.3%, respectively (Figures 2, 3 and Supplementary Figure S2A).
Concerning the use and response of immune checkpoint inhibitor therapy, data are controversial. The TP53 mutation had previously been considered a potential negative predictor of metastatic melanoma treated with CTLA-4 blockade (Xiao et al., 2018), while a very recent paper showed that cell cycle regulators, such as TP53 and CDKN2A, do not appear to significantly alter clinical outcomes when immune checkpoint inhibitors are used (DeLeon et al., 2020). At the time of the writing of this manuscript, no clinical trial with TP53-targeting drugs is ongoing.

NF1
NF1 encodes neurofibromin 1, a cytoplasmic protein highly expressed in neurons, Schwann cells, oligodendrocytes, and leukocytes but also involved in RAS pathway as a tumor suppressor through its inhibiting activity as GAP (GTPaseactivating protein) that converts the active RAS-GTP to RAS-GDP (Trovó-Marqui and Tajara, 2006). Inactivating variants in NF1, the most frequently mutated gene in melanoma after BRAF/NRAS/TP53, being reported in 10-15% of cases, were described in up to 46% of wild-type BRAF and RAS melanomas, in particular in male, older or chronically sun-exposed patients, and in copresence of mutations in RASopathy genes, e.g., PTPN11 and RASA2, that enhance its role in melanomagenesis, besides the RAS missing inhibition (Krauthammer et al., 2015). A poorer OS for NF1-mutated subtype melanoma has also been described (Cirenajwis et al., 2017). In BRAF V600E melanomas, loss of NF1, frequently co-occurring with BRAF and RAS alterations, allows a high level of activity of RAS-GTP and resistance to BRAF inhibitors (Nissan et al., 2014). Since 2000, a role for NF1 was proposed for the genesis of desmoplastic neurotropic melanoma, an uncommon melanoma with pathologic features in common with schwannomas (Gutmann, 2001;Kiuru and Busam, 2017;Mahalingam, 2017). Moreover, NF1 together with BRAF and NRAS has been found significatively mutated in melanoma (Hayward et al., 2017). In our series of skin melanomas, mutations and CNVs are reported in Figures 2, 3 and Supplementary Figure S2B, with a frequency of 14.3 and 1.1%, respectively.
Regarding therapy, a study revealed Calpain1 (CAPN1), a calcium-dependent neutral cysteine protease, as a novel NF1 binding partner that regulates NF1 degradation in melanoma cells. ShRNA-mediated knockdown of CAPN1 or treatment with a CAPN1 inhibitor showed to stabilize NF1 protein levels, downregulate AKT signaling and melanoma cell growth. Moreover, combination treatment of Calpain inhibitor I with MEK inhibitor Trametinib in different melanoma cells seemed to be more effective in reducing melanoma cell growth compared to treatment with Trametinib alone, suggesting that this combination may have a therapeutic potential in melanoma (Alon et al., 2019, p. 1). This novel mechanism for regulating NF1 in melanoma provides a molecular basis for targeting CAPN1 to suppress Ras activation. Despite these data, this novel approach is waiting to be tested within clinical trials. Currently, there are no ongoing clinical trials that evaluate NF1-targeted drugs, but two experimentations regard specifically NF1-mutated melanoma patients, treated with either a MEK inhibitor plus a FAK inhibitor or with RMC-4630, a potent and selective inhibitor of SHP2. The results of these trials have not been published yet ( Table 2).

ARID2
ARID2 encodes a subunit of the SWI/SNF chromatin remodeling complexes (polybromo-and BGR1-associated factor [PBAF]), which facilitates ligand-dependent transcriptional activation by nuclear receptors. The SWI/SNF multiprotein chromatin remodeling complex is involved in regulating cell growth and proliferation. Mutations in nine genes encoding for subunits of the SWI/SNF chromatin remodeling complex are found in 20.0% of human cancers (Shain and Pollack, 2013). ARID2 was found frequently mutated in melanoma, with a frequency ranging from 7.0 to 18.0% Krauthammer et al., 2012;Ticha et al., 2019). In our dataset, mutations and CNVs in ARID2 are found in 12.9 and 5.5%, respectively, as shown in Figures 2, 3 and Supplementary Figure S2C.
To date, studies investigating the biological function of ARID2 in melanocytes and its role as tumor suppressor are missing. In 2018, it has been reported that cancers with inactivating mutations in ARID2 are more sensitive to PD-1 blockade as well as other forms of immunotherapy (Pan et al., 2018). In a very recent study, a higher sensitivity to different DNA-damaging therapies in ARID2-deficient nonsmall cell lung cancer cells, likely as a result of the ARID2 involvement in DNA repair, was observed (Moreno et al., 2020). In addition, ARID2 deficiency showed synthetic lethality with PARP inhibition using veliparib, an inhibitor that has shown good results in the treatment of breast cancer and is included in several clinical trials on breast, ovarian and, lung cancer (Moreno et al., 2020). Overall, recent data prompt further investigation of the role of ARID2 in melanomagenesis and direct testing of several potential therapeutic compounds already available.

CDKN2A
The CDKN2A (Cyclin Dependent Kinase Inhibitor 2A) gene encodes two alternatively spliced variants known to function as inhibitors of G1 progression through different mechanisms, e.g., the tumor suppressor p16 INK4A and p14 ARF . p16 INK4A blocks the G1 progression, binding CDK4 and CDK6, and preventing them from interaction with D-cyclins and phosphorylation of RB. p14 ARF , whose name derives from the presence of an Alternate Reading Frame (ARF) in the transcript, binds MDM2 (Mouse double minute 2 homolog or E3 ubiquitin-protein ligase) whose role is the degradation of p53; therefore, it also plays an indirect role in G1 progression. Furthermore, p14 ARF has a role in p53independent cell functions through the interaction with other proteins (Inoue and Fry, 2018). Loss, e.g., by promoter hypermethylation, or alterations, e.g., homozygous deletions of CDKN2A, are described in several tumors. CDKN2A is altered in 21.4% of cutaneous melanomas (AACR Project Genie Consortium, 2017). The loss of the CDKN2A locus is the most common acquired genetic change in precursor lesions, including in situ melanomas (Shain et al., 2015). Over 75% of cutaneous melanoma metastases have lost one or both alleles of CDKN2A (The Cancer Genome Atlas Network, 2015). Recently, a CRISPR-based study (Zeng et al., 2018) demonstrated that CDKN2A suppresses the initiation of melanoma invasion through inhibition of BRN2, a lineage restricted transcription factor which encodes an established regulator of melanocyte and melanoma invasion (Fane et al., 2017). Then the loss of p16 INK4A enhances the motility of melanocytic cells through increased expression of BRN2.

CDKN2A and CDK4: A Clinical Perspective
To date, several clinical trials are ongoing, attempting to find a way to modulate this pathway (Spagnolo et al., 2015) ( Table 2). With regard to CDKN2A alterations, drugs on study include: ilorasertib (ABT-348), a potent and ATP-competitive multitargeted kinase inhibitor that inhibits Aurora C, Aurora B, and Aurora A and that suppresses RET tyrosine kinase, PDGFRβ and Flt1; palbociclib and SHR6390, two selective inhibitor of the cyclin-dependent kinases CDK4 and CDK6.
Phase I trial with ilorasertib showed two clinical responses among 58 treated patients, and confirmed a good tolerability and safety profile of the drug (Maitland et al., 2018). Palbociclib, together with abemaciclib and ribociclib, have already been approved for the treatment of metastatic breast cancer, after several studies showing their activity in a spectrum of solid tumors including melanoma. Palbociclib is currently under investigation among patients affected by acral melanoma with documented gene aberrations in cell cycle pathways, including CDK4 amplification and/or CCND1 amplification and/or CDKN2A loss. Finally, SHR6390 showed a promising activity in preclinical studies performed on cell lines and human tumor xenograft models.
P276-00 is a novel inhibitor for CDK-1, CDK4 and CDK9 that has been tested in 16 tumor cell lines from different human cancers, showing a significant antiproliferative effect compared to cisplatin. Interestingly, some cancers showed to be particularly sensitive to P279-00, like tumors of central nervous system, NSCLC, breast cancer and melanoma (Joshi et al., 2007).
Ribociclib also showed some activity in melanomas with activating mutations of BRAF or NRAS.
A phase Ib/II study with ribociclib plus MEK162 has been performed in 2013. Among 14 NRAS mutant advanced or metastatic melanoma patients, the combinations of drugs allowed to obtain six partial responses and six stable disease (Sosman et al., 2014).
Abemaciclib, another CDK4/6 inhibitor, structurally different from ribociclib and palbociclib, showed to be effective among several human tumors, including melanoma, in preclinical models (Gelbert et al., 2014;Tate et al., 2014). Of note, abemaciclib demonstrated to be particularly effective in BRAF resistant melanoma cells (Yadav et al., 2014). Subsequently, a phase I trial was conducted to evaluate safety and tolerability of abemaciclib and its antitumor activity. A total of 26 melanoma patients were enrolled: one achieved a partial response and the DCR was 27%. Interestingly, abemaciclib was found in the cerebrospinal fluid demonstrating to pass the blood brain barrier and, on this basis, a trial specifically dedicated to patients with brain metastasis was designed. Unfortunately, no information concerning the melanoma cohort are available to date.

PTEN
The PTEN tumor-suppressor gene encodes for a phosphatidylinositol-3,4,5-triphosphate 3-phosphatase which negatively regulate the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) pathway. The PTEN gene, which is also known as MMAC1 (mutated in multiple advanced cancers) and TEP1 (TGF-β regulated and epithelial cell-enriched phosphatase) exhibits both protein and lipid phosphatase activities. Therefore, deletion or inactivation of PTEN results in constitutive AKT activation. Loss of PTEN function through deletion, mutation, and/or decreased expression, has been found in human sporadic cancers, including melanoma (Bonneau and Longy, 2000). PTEN mutations were identified for the first time in 1997 through the analysis of 35 melanoma cell lines (Guldberg et al., 1997). Subsequently, different studies reported that approximately 30% of cutaneous melanoma cell lines harbor PTEN mutations or deletions (Guldberg et al., 1997;Tsao et al., 1998) and in vitro studies were performed attesting the involvement of PTEN LOH in the development of more than 30-40% of melanomas (Robertson et al., 1998). Then, PTEN mutations were identified in 4 of 61 (7.0%) metastatic melanoma tumors (Birck et al., 2000), a lower rate compared to previous studies (Guldberg et al., 1997;Tsao et al., 1998), but the authors explained this disagreement due to technical limitations and the in vitro selection of cells harboring PTEN mutations.
PTEN and NRAS mutations were described as mutually exclusive (Tsao et al., 2000). Moreover, evidence of cooperation between BRAF activating mutations and PTEN loss in melanoma development was found, suggesting that the activation of MAPK and AKT pathways may be required for melanoma progression (Tsao et al., 2004).
In conclusion, PTEN mutations frequently coexist with BRAF mutations, but not with NRAS, which can independently activate the PI3K cascade, with a mutation and deletion rate frequency of 8.5 and 7.0% in our dataset (Figures 2, 3 and Supplementary Figure S3B).
To date, no therapeutic strategy specifically targeting PTEN has been developed. Interestingly, some trial evaluating the safety and efficacy of two Pi3K inhibitor, alone or in combination with pembrolizumab, in patients with loss of PTEN, are ongoing ( Table 2).

PPP6C
PPP6C gene encodes the catalytic subunit of the PP6 serine/threonine phosphatase complex and regulates cell cycle progression in response to IL2 receptor stimulation (Bastians et al., 1997;Filali et al., 1999). Loss of PPP6C function has been known to cause the increase of Aurora A activity that, as the Aurora A amplification, leads to chromosome instability (Zeng et al., 2010;Hammond et al., 2013). PPP6C mutations make melanoma cells susceptible to inhibition by Aurora kinase inhibitors (Gold et al., 2014). Moreover, PPP6C mutations in melanoma cells seems to induce an increased autophagy. Indeed, PPP6C mutants bind to the PPP6Rs leading to its rapidly degradation. This increases wild-type PPP6C stability, sensitizing the cells autophagy induction in response to mTORC1 inhibition (Wengrod et al., 2015).
In our dataset, PPP6C mutations and CNVs are found in 6.9 and 1.9% of cases, respectively (Figures 2, 3 and Supplementary Figure S3C). In conclusion, PPP6C mutations in melanoma may lead to new targeted approaches, such as specific PPP6 inhibitors, but at now, no trials are ongoing.

CTNNB1
CTNNB1 gene encodes β-catenin that is a core component of the adherens junctions that play a key role in maintaining of the epithelial cell layers, and in transmitting the contact inhibition signal, by anchoring the cytoskeleton. CTNNB1 is also part of the Wnt signaling pathway and is involved, e.g., in pre-birth development, the maintenance and renewal of stem cells. Finally, it has a role in the formation of the matrix of hair follicles. Gain-of-function variants of CTNNB1 cause accumulation of the protein in the nucleus, where it promotes cell proliferation by gene-activation, and have been found in several types of cancer. In melanoma cells, growth is promoted by altering expression of MITF (melanocyte inducing transcription factor) (Widlund et al., 2002), although an opposite effect of invasion-blocking MITF mediated has been observed (Arozarena et al., 2011).
The Wnt pathway may be altered in melanomas by different mechanisms, e.g., somatic variants not only in CTNNB1 but also in APC (adenomatous polyposis coli gene) or ICAT (inhibitor of beta-catenin) genes (Reifenberger et al., 2002). However, a major mechanism of promoting melanomagenesis is the cooperation between CTNNB1 and NRAS by suppression of p16 INK4a expression, releasing cells from oncogene−induced growth arrest (Curley and Bosenberg, 2008). Furthermore, the CTNNB1 promotes expression, in a positive feedback loop, of Tspan8, which triggers melanoma cell detachment and invasiveness (El Kharbili et al., 2019).
Interestingly, CTNNB1 (S33C) mutation was found to confer resistance to imatinib in a metastatic melanoma patients harboring KIT L576P mutation (Cho et al., 2017). In summary, in our dataset frequency of CTNNB1 mutations and CNVs is 6.6 and 1%, respectively (Figures 2, 3 and Supplementary Figure S3D). Currently there are no drugs that target beta-catenin right now, nor ongoing clinical trials.
Several studies reported that DDX3X is involved in doublestranded RNA unwinding, pre-mRNA splicing, RNA export, transcription, and translation Lin, 2019). Despite its important roles in several cytosolic steps of mRNA metabolism, its function in tumorigenesis remains incompletely known. In our dataset, DDX3X mutations and CNVs are found with a frequency of 6.4 and 0.4%, respectively (Figures 2, 3 and Supplementary Figure S4A). Moreover, it has been reported that DDX3X loss decreases MITF mRNA levels, leading to a proliferative-to-metastatic phenotype in vivo, and it is implicated in resistance to BRAF inhibitors (Phung et al., 2019).
These studies reveal a DDX3X role in melanoma cancer providing a new therapeutic target that should be elucidated in the future.
Recently, RASA2 and NRAS mutations were confirmed to be mutually exclusive, with NF1 mutations significantly cooccurring with RASA2 mutations in BRAF and NRAS wild-type melanomas, since RASA2 inhibits NRAS and NF1 inhibits KRAS and HRAS . Indeed, RASA2-and NF1mutated genes co-selection in melanoma could be equivalent to oncogenic RAS mutation .
To date, clinical trials direct toward RASA2 are not currently available.
Very interestingly, SF3B1 mutations were found to induce cancer cells to produce an abnormal form of the BRD9 RNA molecule, including noncoding DNA sequences or "junk DNA, " which garbled the genetic message (Inoue et al., 2019). This "junk DNA" originated from a viral element that recently inserted itself into the human genome. The use of CRISPR technology and antisense oligonucleotides to suppresses tumors growth deriving by SF3B1-mutant cells opens for future therapeutic approaches, but to date no clinical trials directed to SF3B1 are available (Inoue et al., 2019).

RAC1
RAC1 encodes for a RHO GTPase that plays a key role in cellular cytoskeleton organization and motility and can induce RAS dependent pathways stimulating cell proliferation (Krauthammer et al., 2012(Krauthammer et al., , 2015. WES studies led to the discovery of a hotspot mutation (P29S) in RAC1 gene, defining it as the most frequent driver mutation in sun-exposed melanomas, with a frequency of 5-7% Krauthammer et al., 2012;The Cancer Genome Atlas Network, 2015). The frequency of RAC1 P29S mutation (overall RAC1 P29S mutation of 4.1%) was found to be more prevalent in male patients and similar between primary (9.2%) and metastatic tumors (8.6%) (Krauthammer et al., 2012). Indeed, mutations in RAC1 enhanced melanoma disease progression: RAC1 P29S analysis in a cohort of 814 primary cutaneous melanomas, with known BRAF and NRAS mutation status, revealed an association with increased tumor thickness, increased mitotic rate, ulceration, and presence of lymph node metastases at the time of diagnosis (Mar et al., 2014). In general, frequency of RAC1 mutations and CNVs in our skin melanomas dataset is 5.1 and 5.9%, respectively (Figures 2, 3 and Supplementary Figure S4D).
Activation of RAC1 may be also promoted by mutated PREX2. Indeed, truncating mutations in PREX2 have been shown to activate the GTPase RAC1 abolishing the binding to PTEN and activating the PI3K (phosphatidyl inositol 3 kinase)/Akt signaling pathway (Lissanu Deribe, 2016b; Lissanu Deribe et al., 2016). Concerning treatment, in vitro studies have shown that melanoma cell lines harboring the RAC1 P29S mutation are resistant to BRAF and MEK inhibitors, but its role in conferring this resistance is still to be elucidated (Watson et al., 2014). However, clinical inhibitors of RAC1 are not currently available although SRF/MRTF inhibitors in combination with BRAF inhibitors have been recently demonstrated to have utility in the treatment of BRAF mutant melanoma with an RAC1 P29S mutation (Lionarons et al., 2019).
High levels of PD-L1 in patients with RAC1 P29S mutations compared to wild-type RAC1 melanoma samples from the TCGA cohort were observed (Vu et al., 2015). These findings could open up individualized therapy based on immunological characteristics of patient tumor and the presence of RAC1 P29S mutations with anti-PD-1 or anti-PD-L1 antibodies treatment in melanoma patients with high expression of PD-L1 harboring RAC1 P29S mutation.

MAP2K1/MAP2K2
MAP2K1, mitogen-activated protein kinase 1, also known as MEK1, and MAP2K2 (MEK2), mitogen-activated protein kinase 2, encode for mitogen-activated protein (MAP) kinase involved in many cellular processes, such as proliferation, transcription regulation, differentiation and development, and represent the downstream targets of the RAS-RAF-MAPK cascade. Activating mutations in MAP2K1 and MAP2K2 represent one of the multiple mechanisms of resistance to BRAF and MEK inhibitors (Spagnolo et al., 2015). Overall MAP2K1/MAPK2 mutational and CNV prevalence in our skin melanomas dataset is 6.7 and 2.5%, respectively (Figures 2, 3). In particular, melanoma patients have a higher frequency of MAP2K1 mutations/CNVs compared to MAP2K2 mutations/CNVs (Figures 2, 3). Notably, the coexistence of MAP2K1 mutations and BRAF or NRAS mutations are often observed. In general, MAP2K1/2 mutations in BRAF V600E melanomas are linked to both intrinsic and acquired resistance to BRAF inhibitors.
To date, no clinical trials having MEK1/2-mutated melanoma patients as population of interest are ongoing but, as we mentioned before, MEK inhibitors are being tested in several studies, alone or in combination with other drugs, and their use with BRAF inhibitors has become a standard in the therapy of BRAF mutated metastatic melanoma.
Trametinib is a small molecule and a selective MEK1/2 inhibitor, non-ATP competitive.
Based on the encouraging results obtained in phase I (Infante et al., 2012) and phase II trial (Kim et al., 2013, 1), a phase III study (METRIC) with trametinib monotherapy compared to chemotherapy was performed . A total of 322 patients affected by previously untreated BRAF V600 E/K stage III or IV melanoma were enrolled. Trametinib was associated with a higher PFS (4.8 months vs. 1.5), a greater OS (6-months OS 81% vs. 67%), and a higher rate of responses (22% vs. 8%). Based on these data, trametinib was approved as single agent by the FDA in May 2013. Despite this, currently MEK inhibitors do not play a role in monotherapy in NRAS-mutated or in WT patients but could play a role in combination with immunotherapy or with other target agents.
Finally, TAK-733, a selective MEK1/2 inhibitor, showed a broad antitumor activity in melanoma cell lines and in 10 out of 11 patient-derived xenografts models (Micel et al., 2015). In phase I study, performed among 51 patients, TAK-733 showed a manageable toxicity profile but a limited antitumor activity, with partial responses obtained just in 5% of patients ( Table 2) (Adjei et al., 2017).

KIT
KIT encodes for a class III tyrosine kinase receptor that is expressed on several cell types, including melanoblasts and differentiated melanocytes, but also hematopoietic progenitors and mast cells. The binding to the stem cell grow factor causes c-kit homodimerization that leads to the phosphorylation of tyrosine residues by activating the MAPK/ERK and the PI3K/AKT/mTOR pathways (Lennartsson and Rönnstrand, 2012). KIT was initially thought to act as a tumor suppressor gene, because its presence in normal melanocytes and benign nevi, and its loss during progression and in metastatic melanoma was reported (Montone et al., 1997;Shen et al., 2003;Isabel Zhu and Fitzpatrick, 2006). Moreover, loss of c-kit expression was observed in different cultured melanoma cells (Lassam and Bickford, 1992;Natali et al., 1992;Zakut et al., 1993) and related to a higher metastatic potential of melanoma xenografts in nude mice (Gutman et al., 1994). However, KIT also acts as an oncogene. Indeed, it has been found an increase of KIT mutations and/or CNVs in mucosal (39%), acral (36%), and melanomas arose on chronically sun-damaged skin (28%) (Curtin et al., 2006). Interestingly, a recent metaanalysis reported KIT mutations in 497 (9.5%) melanoma patients analyzing 5,224 patients from 32 studies selected (Gong et al., 2018). Moreover, a close association with older age, acral, mucosal, or chronic sun-damage sites, but not with any histological features or tumor stage was found (Gong et al., 2018). Approximately 70% of KIT mutations identified in melanoma and leading to constitutive activation of kinase activity are localized in exon 11 (L576P) or exon 13 (K642E) (Shtivelman et al., 2014). Mutations in KIT are generally mutually exclusive with other driver mutations, including NRAS and BRAF. In our skin melanomas dataset, KIT mutations and CNVs are found in 4.5 and 2.3% of patients, respectively (Figures 2, 3 and Supplementary Figure S5C).
The activity of imatinib mesylate in KIT mutated melanoma patients was explored through three single arm phase II trials ( Table 2). Carvajal et al. treated 28 patients affected by KIT mutated advanced or metastatic melanoma with imatinib 400 mg twice daily (Carvajal et al., 2011). Among the 25 evaluable patients, two complete and four partial responses were observed, with a median time to progression of 12 weeks and a median OS of 10.7 months. Best responses were observed among patients with mutations involving recurrent hotspots or with a mutant -WT allelic ratio superior of 1 (40 vs. 0%). Interestingly, particularly good responses were obtained among patients with K642E mutation, that showed an ORR of 50% and a DCR of 100%. Guo et al. (2011) reported data from 43 treated patients who received imatinib 400 mg daily or, in case of disease progression, 800 mg daily. The median PFS was 3.5 months with a 6-month PFS of 36.6%, while median OS was 12 months. Globally, the (DCR) was 53.3%: 10 patients (23.3%) and 13 patients (3.2%) achieved partial response and stabilization of disease, respectively. Overall, 18 patients (41.9%) achieved tumor regression. Specifically, 9/10 partial responses were observed in patients with mutations in exons 11 or 13. The overall 1-year survival rate was 51%.
Finally, in a third study published by Hodi et al. (2013), 25 patients affected by metastatic melanoma of the mucosa, acral or chronically UV-damaged skin with amplifications or KIT mutations received 400 mg of imatinib daily or, in the absence of a clinical response, twice a day. This study confirmed the clinical activity of imatinib, mainly concerning KIT mutations: indeed, all responses were achieved among patients with KIT mutations while the best response observed in patients with amplifications was stable disease. OS was 12.5 month in the overall cohort, similar to the previously reported trials.
Other KIT inhibitors showed activity in KIT mutated advanced melanomas ( Table 2).
Nilotinib is a small molecule that showed activity among KIT mutated advanced or metastatic melanoma patients in four clinical studies. In the first trial published in 2015 (Carvajal et al., 2015), 19 patients, mostly in disease progression after treatment with imatinib, achieved an ORR of 15.8% and a median OS of 9.1 months. In a second trial, 39 patients were treated with nilotinib (Lee S. J et al., 2015): one experienced complete response and six partial response; median PFS was 3.3 months and median OS 11.9 months. The subsequent phase II trial (TEAM), an ORR of 26.2% was achieved among 42 patients naïve for previous KIT inhibition (91% of which in the presence of exon 11 mutations), a median PFS of 4.2 months and a median OS of 18 months. In another phase II study, 25 patients were treated with an ORR of 16% at six months of observation, median PFS of 6 months and median OS of 13.2 months (Guo et al., 2017).
Sunitinib is another molecule tested in 12 patients with advanced or metastatic melanoma, and obtained four clinical responses in a small clinical trial (1 complete response and 3 partial response) (Minor et al., 2012).
Finally, dasatinib was evaluated in 22 patients with advanced or metastatic KIT mutated melanoma. With regard to clinical activity, four partial responses and seven stable diseases were obtained. The median PFS was 4.7 months and median OS 12.3 months (Kalinsky et al., 2017).
In conclusion, to date it is recommended to test KIT mutations (especially exons 11 and 13) in acral, mucosal, and unknown origin melanomas, as well as cutaneous ones arising on chronically damaged skin, to offer an additional therapeutic option.

RB1
RB1 acts as a tumor suppressor gene by regulating cell cycle division: when dephosphorylated, it interacts directly with E2F1 and inhibits its transcriptional activity with cell cycle arrest. Thus, the cell cycle is finely regulated by CDKs with CDKIs varying levels of RB phosphorylation, E2F family and TP53; alterations that disrupt the p16INK4A:cyclinD-CDK4/6:RB functional pathway may the first critical step leading to melanomagenesis (Bartkova et al., 1996;Lee B. et al., 2015). Cells deficient or with low levels of RB go into p53-mediated apoptosis: this phenomenon is particularly important to identify new compounds able to activate p53 (Knudsen and Wang, 2010). In mouse melanoma cells Rb1 cooperates with MITF to activate expression of Tyr and Cdkn1a/p21Cip1 (Carreira et al., 2005). RB1 is also important in maintaining chromatin structure, stabilizing histone methylation (Shao and Robbins, 1995). Summing up, RB is likely a multifunctional protein that binds to at least 100 other proteins (Morris and Dyson, 2001). RB1 is altered in about 4% of all cancers (MCG) 1 . In our dataset, the most common somatic alterations in RB1 are mutations with a frequency of 4.4% while CNVs are reported in 1.9% of skin melanoma patients (Figures 2, 3 and Supplementary Figure  S5D). To date, no clinical trials specifically designed to target RB are ongoing. However, patients with RB1 loss are specifically considered in one clinical trial with LY260636, a checkpoint kinase 1 inhibitor. The trial is active, not recruiting.

FBXW7
FBXW7 is a critical tumor suppressor gene and a member of the F-box protein family, ubiquitin ligase complex, that controls proteasome-mediated degradation of oncoproteins such as cyclin E, c-Myc, Mcl-1, mTOR, Jun, Notch, and AURKA, STAT2 (Minella and Clurman, 2005;Yeh et al., 2018;Lee et al., 2020). 1 https://www.mycancergenome.org/ Inactivating mutations in FBXW7 have been described in a variety of human tumors and cancer cell lines (Akhoondi et al., 2007). Loss of function of FBXW7 in several human cancers has clinical implications and prognostic value: the use of rapamycin has proven to inhibit breast cancer cells with loss of FBXW7 by mTOR inhibition (Mao et al., 2008;Yeh et al., 2018).
FBXW7 expression was reduced in primary and metastatic melanoma compared with dysplastic and normal nevi and an increase in FBXW7 expression was significantly correlated with a better 5-year patient survival. In vitro studies demonstrated that FBXW7 inhibited human cell migration through MAPK/ERK signaling pathway suggesting its prognostic and potential therapeutic role for melanoma treatment (Cheng et al., 2013). Several WES and WGS studies reported a FBXW7 mutations in around 4% of melanomas (Wei et al., 2011;Berger et al., 2012;Hodis et al., 2012;Krauthammer et al., 2012;The Cancer Genome Atlas Network, 2015;Hayward et al., 2017;ICGC Data Portal, 2020). WES of 103 cutaneous melanomas, including 77 tumor samples and 26 cell lines, found FBXW7 mutations (8.1%) independently of BRAF or NRAS mutation (Aydin et al., 2014). In the same study, the authors discovered that inactivation of FBXW7 gene determines enhanced tumorigenesis by NOTCH1 activation (Aydin et al., 2014). These evidences open up to FBXW7 potential therapeutic targeting through modulating NOTCH1 signaling (Aydin et al., 2014). Moreover, another study revealed FBXW7α deficiency leading to HSF1 (Heat shock factor 1) accumulation and subsequent activation of the invasionsupportive transcriptional program and metastatic potential of human melanoma cells (Kourtis et al., 2015).
To date, just one clinical trial designed to enroll specifically FBXW7-mutated patients is active (Table 2). LY2606368, the drug on study, performed well in a phase I trial where it was tested in 45 patients affected by solid tumors, who experienced treatment failure with standard therapies. Among 43 evaluable patients, two partial responses and 15 stable diseases were achieved, with a DCR of 37.7% (Hong et al., 2016).
Alterations in PIK3CA have an oncogenic effect mostly related to activating variants in two hotspots located in the regions of helical and kinase domains. The PI3K-AKT pathway plays a significant role in melanomagenesis, frequently by activating RAS-RAF-MEK-ERK pathway, e.g., for NRAS activating variants or loss of PTEN (Davies, 2012), as demonstrated by studies on resistance to targeted therapies based on BRAF inhibitors (Deng et al., 2012;Penna et al., 2016). Usually, PIK3CA activating mutations are rare in melanoma with a frequency of 5%, despite the ability of activated PIK3CA (H1047R) to cooperate with BRAF V600E promoting melanomagenesis in mouse models (Marsh Durban et al., 2013). PIK3CA mutations frequently co-occurred with either a BRAF or an NRAS mutation (The Cancer Genome Atlas Network, 2015). In our skin melanoma dataset, PIK3CA mutations and CNVs are 4.2 and 1.5%, respectively, as shown in Figures 2, 3 and Supplementary Figure S6B.
Several clinical trials with PI3K inhibitors are ongoing ( Table 2).
Alpelisib (BYL-719) is an oral selective inhibitor of PI3K isoform-α that showed to be active against the somatic PI3Kα mutations and wild-type PI3Kα (Fritsch et al., 2014). Based on these data, 58 patients with BRAF or RAS mutated advanced solid tumors were enrolled in a phase Ib trial and were treated with BYL-179 plus MEK162. A total of five NRAS mutated patients experienced a partial response (Juric et al., 2014).
Parsaclisib is a novel Pi3kδ inhibitor that exhibited an excellent profile in xenograft models (Yue et al., 2019, 0504). To date, one clinical trial is ongoing to assess the efficacy and the safety of the combination between parsaclisib plus pembrolizumab and itacitinib, a JAK1 inhibitor. The association parsalisib plus itacitinib showed good results in a phase I/II trial for hematological malignancies, with an ORR of 67-78%, respectively in mantle cell lymphoma and marginal zone lymphoma. Lower results were obtained among patients with diffuse large B-cell lymphoma (ORR 30%) (Forero-Torres et al., 2019).
The combination between pimasertib, a MEK1/2 inhibitor, and voxtalisib, a dual PI3K/mTOR inhibitor among patients affected by solid tumors, including melanoma, was tested in a phase Ib trial in 146 patients. In December 2018, data from this experimentation were published, showing a DCR of 52% with one complete response and five partial responses. Unfortunately, the toxicity profile of this combination was considered not acceptable (Schram et al., 2018).

EZH2
EZH2 encodes a histone methyltransferase that constitutes the catalytic component of the polycomb repressive complex-2 (PRC2) which has a role in epigenetic silencing during cell differentiation, in particular in the development of the hematopoietic and central nervous systems. EZH2 can also induce an epithelial-to-mesenchymal transition in the cancer cells, increases their metastatic potential (Min et al., 2010) and acts as a coactivator for transcription factors including the androgen receptor (Xu et al., 2012).
More recently it has been described the role of EZH2 as a recruitment platform for DNA methyltransferases in epigenetic repression (Viré et al., 2006;Barsotti et al., 2015;Moran et al., 2018).
The EZH2-dependent expression of genes associated with cell motility contributes to early phases of metastasis (Manning et al., 2015) while activating variants promotes melanoma progression inactivating tumor suppressor genes (Tiffen et al., 2015). The reactivation of tumor suppressors was correlated to increased survival confirming that EZH2-mediated epigenetic repression has a major role in advanced melanoma progression (Zingg et al., 2015). A study suggested the possibility that combined immunohistochemical expression of EZH2, H3K4me2, and H3K27me3 might identify cancer cells with potential stem cell properties; another relevant data that many epigenetic changes are pharmacologically reversible (Kampilafkos et al., 2015;Mahmoud et al., 2016).
EZH2 mutations and CNVs in our dataset are detailed in Figures 2, 3 and Supplementary Figure S6C, with an overall mutational rate of 3.9 and 5.7%, respectively.
Of note, coexistence of BRAF V600E mutation and EZH2 amplification is rather prevalent in melanoma. Indeed, in a cohort of 138 patients with BRAF V600E-mutated melanoma, 40 cases (29.0%) showed EZH2 gain. Moreover, a significant difference in overall survival and disease-free survival between no EZH2 copy number gain and gain groups was reported (Yu et al., 2017).
A recent study highlights that both benign melanocytes and cutaneous melanomas frequently harbor amplifications of EZH2 that silence genes correlated to the integrity of the primary cilium (Zingg et al., 2018).
Vemurafenib and trametinib induce senescence consistent with downregulation of c-MYC but an EZH2 variant limits this effect (Hartman et al., 2019). EZH2 has a role in differentiation of CD4+ T-cells and in the function of T regulatory cells. Its activation causes immune suppression and it has been suggested that EZH2 inhibitors may have a role in combination with immunotherapy and targeted therapies to prevent immunosuppression (Tiffen et al., 2016). It has been further demonstrated that EZH2 controls melanoma escaping mechanisms during T cell-targeting immunotherapies (Zingg et al., 2017) and because the upregulation of EZH2 and its histone modification H3K27me3 seems correlated to melanoma progression and resistance to immune checkpoint blockade, clinical trials based on EZH2 inhibitors are strongly suggested (Hoffmann et al., 2020).

WT1
WT1 encodes a transcription factor implied in the prenatal development of kidneys and gonads, mainly known for its action in cell differentiation and apoptosis. Moreover, WT1 has a role in tumorigenesis controlling several other genes [e.g., Pecam-1 (CD31) and c-KIT (CD117)], and modulating vascularity, immune response and metastasis formation (Wagner et al., 2014). High expression of WT1 is described in leukemias and in solid tumors and it seems correlated to the chemoresistance and poor outcome. Interestingly, inhibition of cell proliferation by shRNA-WT1, cisplatin, and gemcitabine in B16F10 cells induces cell death and potentiates the action of anticancer drugs by inducing synergistic effects both in vitro and in vivo (Zapata-Benavides et al., 2019). Although WT1 could be expressed by Spitz naevi and in up to one third of dysplastic naevi, it is considered as a diagnostic tool in melanoma diagnosis (Wilsher and Cheerala, 2007). Indeed, WT1 is expressed in more than 80% of malignant melanoma cells, but not in normal skin or benign melanocytic nevi (Wagner et al., 2008). WT1 protein expression was associated with shorter overall survival in melanoma (Garrido-Ruiz et al., 2010) and deemed as a target antigen for immunotherapy since a novel signaling mechanism mediated by PPARbeta ligands, which led to melanoma cell growth suppression through the direct repression of WT1, was described (Michiels et al., 2010, 1). The silencing of WT1 through shRNAi has a synergistic effect with doxorubicin and cisplatin, sensitizing B16F10 melanoma cells (Zapata-Benavides et al., 2012). WT1 somatic mutations are described in several cancers, with cutaneous melanoma having the greatest prevalence. Two similarly sized WES analysis of 100 and 114 skin melanomas found WT1 somatic mutations in the 3 and 1.8% of skin melanoma patients, respectively Krauthammer et al., 2012) (Supplementary Figure S6D). In summary, frequency of WT1 mutations and CNVs in our skin melanomas dataset are 3.7 and 1.3%, respectively (Figures 2, 3 and Supplementary Figure S6D).
To date, one trial with DSP-7888 plus nivolumab or pembrolizumab in advanced solid tumors including melanoma, is ongoing ( Table 2). DSP-7888 is a peptide cancer vaccine that includes WT1 derived peptides: it has shown to induce both a CD8+ and CD4+ mediated immune response against WT1 overexpressing tumor cells.

SNX31
SNX31 encodes for the sorting nexins protein involved in membrane trafficking (Worby and Dixon, 2002;Ghai and Collins, 2011;Ghai et al., 2014). It is upregulated in more than 50% of bladder carcinoma transitional cell, but to date its role in cancer is poorly understood. In melanoma, missense mutations in SNX31 were first reported with a 7% frequency . Interestingly, a 9.0% mutation frequency was reported in 46 primary mucosal melanomas . In our skin melanomas dataset SNX31 mutations and CNVs are 3.7 and 6.0%, respectively (Figures 2, 3 and Supplementary Figure S7A). No clinical trials with SNX31 inhibitors are ongoing.

IDH1
IDH1 encodes for a single, soluble, cytoplasmic isocitrate dehydrogenase 1 enzyme, that converts isocitrate to αketoglutarate (also known as 2-oxoglutarate) in an NADP + dependent manner, protecting cells against reactive oxygen species (ROS) and radiations (Minard and McAlister-Henn, 1999;Jo et al., 2002). First described in glioma (Balss et al., 2008, p. 1;Parsons et al., 2008;Yan et al., 2009, p. 2), mutations in IDH1 and IDH2 (isocitrate dehydrogenase 2 (NADP+), mitochondrial) were described in several cancers, including melanoma (Lopez et al., 2010). IDH1 and IDH2 mutations occur primarily in the catalytic domain at residue R132 in IDH1, and R140 and R172 in IDH2. The mutant IDH1/IDH2 proteins lead to a reduction of α-ketoglutarate (α-KG) with a relative production of oncometabolite 2-hydroxylglutarate (D-2HG) that it is likely to play a major role in the pathophysiology of tumors blocking cellular differentiation by competitively inhibiting αKG-dependent dioxygenases involved in histone and DNA demethylation (Mondesir et al., 2016). The first IDH1 mutation in melanoma was reported in 2010 (Lopez et al., 2010). The authors also revealed BRAF V600E mutation, but no NRAS/TP53/CDKN2A/CDKN2B mutations, in one metastatic sample carrying the R132C mutation, analysing 78 patients. Moreover, IDH2 R172 mutations were not detected in any of these samples (Lopez et al., 2010). The next year, in a cohort of 142 primary non-epithelial tumors, it was shown that about 10.0% of metastatic melanoma lesions (4/39) harbored an IDH1 or IDH2 heterozygous mutation, equally distributed (Shibata et al., 2011). This study confirmed the co-occurrence of BRAF mutation with IDH1 mutation previously observed, and demonstrated that mutant IDH1 conferred in a BRAF V600E melanoma cell line acquired in vivo growth activities and enhanced activation of the MAPK and STAT3 pathways (Lopez et al., 2010;Shibata et al., 2011). Mutant IDH1 also reduced the expression of RASSF1, DHRS1, ADH5, whereas it induced the expression of JUN, MYCN, and ATF3 (Lopez et al., 2010).
Thanks to these evidences, future clinical studies could be pursued investigating the efficacy of IDH inhibitors for melanoma cancer.

STK19
STK19, also known as RP1, encodes a nuclear serine/threonine kinase (Shen et al., 1994;Gomez-Escobar et al., 1998). STK19 protein can bind ATP phosphorylated a-casein proteins at Serine/Threonine residues and histone at Serine residues, therefore phosphorylation of STK19 is involved in transcriptional regulation (Gomez-Escobar et al., 1998). STK19 is important for the transcription-related DNA damage response since it is involved in DNA repair during active transcription and in nuclear signal transduction (Boeing et al., 2016). STK19 mutations have been found with a frequency of around 4.1% in melanoma, with D89N mutation present at 3.3%, whereas no SKT19 mutations were detected in nevus associated-melanomas Shitara et al., 2015). In our dataset, STK19 mutations and amplifications are reported in 3.3 and 6.0% of samples selected (Figures 2, 3 and Supplementary Figure S7C).
In a recent study, STK19 was identified as a novel regulator of NRAS function (Yin et al., 2019). STK19 alterations were mutually exclusive with BRAF (Yin et al., 2019). Through in vitro and in vivo experiments STK19 was found to phosphorylate the residue S89 of NRAS activating NRAS signaling via the MEK-ERK and PI3K pathways. In NRAS Q61R transgenic mice the STK19 D89N mutant promoted oncogenic NRASdriven melanomagenesis. An STK19 inhibitor (ZT-012-037-1) was able to inhibit NRAS activation in NRAS-STK19 mutant mice preventing NRAS-driven melanoma development and growth (Yin et al., 2019), revealing a promising therapeutic strategy for NRAS-mutant melanoma tumors treatment (Yin et al., 2019). Based on these last observations, several natural compound libraries were screened using a phosphorylation assay-based approach and chelidonine was identified as a potent and selective inhibitor of STK19, providing a novel option for targeting NRASmutant cancers (Qian et al., 2020). Nevertheless, clinical trials are not available.

MITF
Microphthalmia-associated transcription factor is a basic helixloop-helix (hHLH)-leucine zipper protein that plays a role in the development of neural crest-derived melanocytes and retinal pigment epithelial cells, and is encoded by MITF gene (Fuse et al., 1999). MITF is directly involved in the expression of genes that encode melanin synthesis enzymes (TYR, TYRP1, and DCT), melanosome proteins (PMEL, MLANA, and RAB27A), proteins involved in the cell cycle (CDKN1A, CDKN2A, TBX2, and CDK2) and in cell survival (e.g., BCL2, BIRC7, HIF1A, and MET) (Levy et al., 2006;Cheli et al., 2010;Vachtenheim and Borovanskı, 2010), DNA replication and repair, cell proliferation, and mitosis (Strub et al., 2011;Webster et al., 2014). This wellknown role in melanocyte development led MITF nicknaming as the "master regulator" of melanocytes (Levy et al., 2006). Despite MITF acts together with many transcription factors as SOX10, YY1, TFAP2A, LEF1, RB1, IRF4, and PAX3 (Seberg et al., 2017), its activity level determines the phenotype adopted by melanoma cells, whether invasive, proliferative, or differentiated according to the MITF rheostat model (Goding, 2011). MITF locus is amplified in 7.9% of skin melanomas, while in our dataset, MITF somatic mutations are quite rare, with a frequency of 1.7% (Figures 2, 3 and Supplementary Figure  S7D). MITF amplification is more prevalent in metastatic melanomas and correlates with decreased overall patient survival. It is responsible for resistance to conventional chemotherapy and BRAF inhibition (Garraway et al., 2005). Recently, Wang et al. provided evidence that p300/CBP inhibition suppressed the melanoma-driven transcription factor, MITF, and could be utilized as a potential therapy for treating melanoma (Wang R. et al., 2018;Kim et al., 2019). To date, no clinical trials are attempting to target MITF directly; on the other hand, MITF-mediated pathways targeting through histone deacetylase inhibitors (HDACi) has been investigated with disappointing results. Indeed, despite HDACi showed the ability to silence MITF promoter in vitro and in vivo (Yokoyama et al., 2008), this ability was not observed among treated patients in a phase I trial performed between 2010 and 2012. A total of 16 patients affected by unresectable stage III or IV melanoma received panobinostat, a potent HDACi that previously showed promising results among preclinical studies in melanoma samples and hematologic malignancies (Giles et al., 2006). Results from this phase I clinical trial were published in 2016: among 15 evaluable patients, four achieved stable disease, with an ORR of 0% and a disease control rate of 27%. Due to the absence of partial or complete response, and the heavy toxicity of study drug, the trial was stopped (Ibrahim et al., 2016).

TERT
TERT gene needs to be addressed separately from the other genes, since the most common mutations in TERT are localized in TERT promoter and have not been systematically investigated in most of the studies included in our dataset. TERT gene codes for a telomere reverse transcriptase catalytic subunit, the principal mechanism of telomere maintenance in cancer cells.
Most of human cancers re-activate telomerase (Kim et al., 1994). The first studies that identified cancer related TERT promoter mutations were performed in melanoma patients. The most common TERT promoter mutations are −57A/C, −124C/T, −146C/T, upstream the TERT gene ATG (Horn et al., 2013). From TCGA dataset of 9,127 patients and 31 cancer types emerged that 27% of all analyzed samples harbored one of these promoter mutations (Barthel et al., 2017), placing them among the most frequent cancer mutations (Lorbeer and Hockemeyer, 2020). In particular, 73 out 115 SKCM melanoma samples (frequency of 63.4%) harbored TERT promoter mutations (The Cancer Genome Atlas Network, 2015). Interestingly, WGS studies analyzed TERT promoter reporting a frequency of 81.2% in acral and cutaneous melanomas (Supplementary Figure S9) (Hayward et al., 2017).
Then, a discordancy on the mutational status between the primary and metastatic lesion was found in a cohort of 194 primary nodular melanomas matched with 72 loco-regional metastases, while TERT protein expression was found to correlate with reduced patient survival (Hugdahl et al., 2018). A frequent TERT promoter polymorphism at −245 was associated with an increased rate of metastasis in melanoma (Nagore et al., 2019) and inversely correlated with BRAF mutation (Bruno et al., 2018). Moreover, mutations of the TERT promoter are more frequent in fast-growing melanomas rather than slow-growing ones and this characteristic could be used to identify more aggressive tumors that might benefit from adjuvant therapy (Nagore et al., 2016). Recently, CNV was found in 61.5% of acral melanoma which is consistent with a previous study that reported rates from 44.9 to 75% (Liang et al., 2017;Yu et al., 2018Yu et al., , 2020. In our skin melanoma dataset, TERT coding mutations (3.9%) and CNVs (5.5%) are described in Figures 2, 3 and Supplementary Figure  S8A. Specifically, the 42 TERT coding variants found in our dataset were all not-pathogenic variants in melanoma, underlying TERT as a polymorphic gene.
No clinical trials for TERT mutated patients or with TERT inhibitors are ongoing.

MTOR
The MTOR gene encodes a protein belonging to the phosphatidylinositol kinase-related kinases family, which regulates cell growth, proliferation, motility and survival, protein synthesis and transcription (Wullschleger et al., 2006). mTOR is regulated and responds to growth factors, energy metabolites and/or levels of nutrients (Vogt, 2001). mTOR is the catalytic subunit of two different multiprotein complexes, mTORC1 (highly sensitive to rapamycin) and mTORC2 (relatively insensitive to rapamycin) with different functions. AKT regulates the mTORC1 complex by phosphorylating and inhibiting the TSC-2 gene (Tuberous Sclerosis 2), which is a GTP-ase activating protein (GAP) that binds to TSC-1 (Tuberin) forming a complex and blocking the G Rheb protein. The inhibition of TSC-2 allows the Rheb protein to accumulate in a GTP-bound state and to activate mTORC1. mTORC1 regulates several key steps of protein synthesis, controlling the expression of proteins that promote cell proliferation and survival. The mechanism controlling mTORC2 is not yet well known; the activation of this complex, however, is linked to the PI3K signaling (Pópulo et al., 2012). mTOR, as key protein of the PI3K/AKT pathway, acts both upstream and downstream of AKT, and aberrant mTOR activation of promotes survival and proliferation of tumor cells in several human cancers (Saxton and Sabatini, 2017;Chamcheu et al., 2019). Nonsynonymous MTOR mutations are present in about 3.61-12% of melanoma patients (The Cancer Genome Atlas Network, 2015; AACR Project Genie Consortium, 2017). In a Chinese cohort of 412 melanoma patients, nonsynonymous MTOR mutations were found by NGS analysis in acral, mucosal, cutaneous melanomas (with and without chronic sun-induced damage), and unknown primary subtypes, with a frequency of 11.0, 14.3, 3.4-6.7, and 11.1%, respectively (Kong et al., 2016). In this study, H1968Y, P2213S and S2215Y were established as gain-of-function mutations sensitive to specific inhibitors (Kong et al., 2016). Similarly to P2213S and S2215Y, the same authors found another gain-offunction mutation (H2189Y) that was sensitive in heterozygous at the AKT inhibitor (AZD5363) and the phosphoinositide 3-kinase inhibitor (LY294002), and in homozygous at the mTOR inhibitor (everolimus) and the AKT inhibitors (AZD5363) and (MK-2206 2HCL), and the phosphoinositide 3-kinase inhibitor (LY294002) (Wu et al., 2018). Interestingly, analyzing 31 candidate genes existing in either the PI3K or MAPK pathway in 105 metastatic melanoma patients, the researchers found two novel nonsynonymous mutations (R2443 * and L552F) in MTOR gene (Shull et al., 2012).
In our dataset, mutations and CNVs in MTOR are found in 7.9 and 2.3% of the samples, respectively, as shown in Figures 2, 3 and Supplementary Figure S8B. However, among the 78 MTOR coding mutations found in our dataset, only five were pathogenic (6.4%; 2 E1799K, 2 S2215F, and 1 I2500F). Interestingly, these mutations have previously been identified as activating mutations in cancer patients using publicly available databases of cancer genome sequencing data (Grabiner et al., 2014). This study found 33 MTOR mutations that confer pathway activation, not by reducing sensitivity to mTOR inhibitors, but by altering the response of mTORC1 signaling pathway to nutrient deprivation (Grabiner et al., 2014).
Regarding melanoma therapy, MTOR may be targeted by selective inhibitors (Chamcheu et al., 2019). Rapamycin, a specific mTORC1 inhibitor, inhibits the cell growth and proliferation in several melanoma cell lines (Buscà et al., 1996;Molhoek et al., 2005;Karbowniczek et al., 2008). Two other rapamycin analogs, everolimus, and temsirolimus, also showed promising results in preclinical studies but failed the clinical experimentation in melanoma patients (Table 2).
Everolimus, as single agent, was tested in a phase II trial with 53 metastatic melanoma patients: the drug led to obtain just two clinical responses with a heavy toxicity profile and the study was terminated for futility (Vera Aguilera et al., 2018). A combination of everolimus with carboplatin and paclitaxel as I line treatment for metastatic melanoma was also tested in a phase II trial. Among 70 enrolled patients, just 12 showed objective responses, with a median PFS of 4 months and a median OS of 10 months (Hauke et al., 2013). Bevacizumab and everolimus were tested in a phase clinical II trial with 57 metastatic melanoma patients and led to a disease control rate of 70%. Unfortunately, median PFS and OS were just 4 month and 8.6 months, respectively (Hainsworth et al., 2010). Lastly, a recent study evaluated the combination of everolimus, bevacizumab, carboplatin and paclitaxel (CPB) in a randomized phase II trial. A total of 149 patients affected by stage IV melanoma were treated with CPB or with CPB plus everolimus: no differences in terms of PFS or OS were noted among the two groups, but everolimus strongly increased toxicity (McWilliams et al., 2018).
A new attempt to bring attention back to everolimus comes from a recently published phase I study, that evaluated the safety and the efficacy of vatalanib plus everolimus in patients with advanced solid tumors, including melanoma. Among 70 evaluable patients, nine achieved a partial response and 41 a stable disease, with best results obtained in neuroendocrine tumors, renal cancer, melanoma and NSCLC (Zhu et al., 2020).
Regarding the use of everolimus in mTOR mutated patients, only two clinical trials exist to date, to our knowledge. The first one is a phase II clinical trial on everolimus in cancer patients with TSC1 and TSC2 mutation or activating MTOR mutation. The second is a single-arm, open-labeled and singlecentered study of everolimus in selective patients with metastatic melanoma and mutations (Kinase domain) of mTOR. Both studies are ongoing, and thus results are not yet available.
Temsirolimus was tested in combination with sorafenib in a phase I  and in a phase II clinical trials (Margolin et al., 2012), with disappointing results. Indeed, not only this combination failed to achieve any clinical response, but it induced a significant toxicity at higher dose levels. More encouraging results came from a phase II clinical trial with temsirolimus and bevacizumab: among 17 treated patients, three obtained a partial response and nine a stable disease, with best responses obtained in BRAF wild type population (Slingluff et al., 2013).

TACC1
TACC1 gene belongs to the TACC gene family that encodes centrosomal proteins that may play roles in microtubule regulation and spindle function, and, thus, may be an important driver of genomic instability in cancer cells (Anafi et al., 2000;Conte et al., 2003). Moreover, TACC1 proteins play an important role in transcriptional regulation (Gas41, thyroid hormone receptor, and retinoid acid receptor α) and mRNA processing (LSM7 and SmG) (Conte et al., 2002;Guyot et al., 2010). Many studies reported that TACC1 expression was modified in several cancers (Line et al., 2002;Rhodes et al., 2002;Nguyen et al., 2005;Ghayad et al., 2009). TACC1 has been also reported to stimulate the RAS and PI3K pathways playing an oncogenic role in tumor formation in the murine mammary gland (Cully et al., 2005). Finally, FGFR-TACC fusions, including FGFR1-TACC1, were also described in glioblastomas since 2012 (Di Stefano et al., 2015;Lasorella et al., 2017).
TACC1 mutations were first reported in a WES study of 121 melanoma tumor/normal pairs reporting five novel melanoma candidate genes, including TACC1, with a frequency of 7% which was confirmed in a further similar study of 25 metastatic/germline pairs (Berger et al., 2012;Hodis et al., 2012) (Supplementary Figure S8C). Finally, in a recent study several key genes involved in melanomagenesis, including TACC1, in 115 human melanoma cell lines, 248 patient-derived xenografts, 31 cell lines derived from PDX, and 68 patient tumors, an overall frequency rate of 1.47% was reported (Garman et al., 2017). In conclusion, in our dataset TACC1 mutations and CNVs have a frequency of 2.9 and 2.3%, as shown in Figures 2, 3 and Supplementary Figure S8C. No clinical trials are available.

CNOT9
CNOT9, also known as RQCD1, encodes for the CCR4-NOT transcription complex subunit 9 that plays a role as transcription cofactor in multiple biological processes including cellular differentiation and RNA processing (Mathys et al., 2014). Several NGS studies revealed the presence at low frequency of recurrent mutations in RQCD1/CNOT9 (P131L) in melanoma samples (Dutton-Regester et al., 2014). Interestingly, Wong SQ et al. found that melanoma tumors harboring RQCD1/CNOT9 P131L mutation were associated with increased thickness, head and neck and upper limb location, lentigo maligna melanoma subtype, and BRAF V600K but not V600E or NRAS Q61 mutations (Wong et al., 2015). CNOT9 alterations in our dataset are shown in Figures 2, 3 and Supplementary Figure S8D (frequency mutations and CNVs of 2.9 and 1.3%, respectively).
The functional oncogenic role of CNOT9/RQCD1 in melanoma remains currently unknown although limited studies revealed RQCD1 implication in AKT activation and cell proliferation (Ajiro et al., 2009(Ajiro et al., , 2010. In particular, the CNOT9/RQCD1 mutant has demonstrated to stimulate stronger immune responses than the wild type, implicating the formation of a neoantigen that could be a potential therapeutic target although no drugs or clinical trials are available (Wong et al., 2015).

YAP1
YAP1 encodes for transcriptional coactivators belonging to the Hippo pathway. The core of the Hippo pathway in mammals consists of a kinase cascade, MST1/2 and LATS1/2, as well as downstream effectors, transcriptional coactivators YAP and TAZ (Ma et al., 2019). The Hippo pathway plays a crucial role in organ size control by regulation of cell proliferation, apoptosis, cell differentiation, and cell migration. Moreover, the upregulation of the Hippo pathway downstream effectors, such as YAP and TAZ, are common across various cancers since mutations and altered expression of its core components promote the migration, invasion, malignancy of cancer cells (Moroishi et al., 2015;Yu et al., 2015;Han, 2019). Very recently, 9,125 tumor samples were profiled revealing a widespread dysregulation of Hippo pathway components in multiple human cancer types, including melanoma (Sanchez-Vega et al., 2018). In particular, a strong interaction between Hippo pathway and GNAQ and GNA11 oncogenes was observed. Several studies reported the functional role of YAP in uveal melanoma cells carrying GNAQ/11 mutations (Feng et al., 2014;Yu et al., 2014). In uveal melanoma ATCG samples, only one among 80 uveal melanoma patients revealed a missense mutation in YAP (G369R) co-occurring with GNAQ R183Q and GNA11 R183C (Bakhoum and Esmaeli, 2019). Moreover, no YAP CNV was shown in uveal melanoma samples by a comprehensive molecular characterization of the Hippo core genes in 9,125 tumor samples . In our skin melanoma dataset, YAP1 alterations are reported in Figures 2, 3 and Supplementary Figure S10, with an overall frequency mutation and CNVs of 1.4 and 2.6%, respectively.
Although the central role of YAP in uveal melanomagenesis is deeply investigated, the role of Hippo/YAP signaling in cutaneous melanoma is less understood. Zhang et al. (2019) proved for the first time that YAP hyperactivity is elevated in invasive melanoma cell lines, induce invasion in normally non-invasive melanoma cells, induces spontaneous melanoma metastasis in vivo and promotes melanoma cell invasion by regulating expression of AXL, CYR61 and CRIM1. Moreover, they did not found association between YAP activity and the BRAF/NRAS mutation status. Of note, they found for the first time somatic activating mutations in YAP gene by WES. Indeed, they identified seven independent serine to alanine substitutions in YAP gene (called YAP-7SA) from a primary cutaneous melanoma patient carrying BRAF V600E and RAC1 P29S mutation. In vitro study confirmed that YAP-7SA encodes a hyperactive version of the YAP protein . Despite the great interest of scientific community toward the role of Hippo pathway in cancer, no drugs and clinical trials direct to YAP in melanoma are described.

CONCLUSION AND PERSPECTIVES
The better definition of the genetic mechanisms underlying melanomagenesis allowed to synthesize new molecular targeted drugs that radically changed the prognosis of melanoma patients.
In this review we analyzed more than 30 genes other than BRAF, that can be or become targets for new molecular target therapies. We collected mutation data on a total of 992 melanoma samples analyzed by WES or WGS to prove the genetic heterogeneity of melanoma, beyond the BRAF mutation. Only 11 of them (1.1%) showed no coding mutations, whereas 298 (30%) showed a mutation in one of the RAS genes (mostly NRAS) and most of the samples studied (698, 70%) showed mutations in non-RAS genes (Figures 2, 3) suggesting potential new roads/candidates for targeted therapy.
We also provide an overview of the related clinical trials, ongoing or completed, to better outline the state of molecularderived clinical research ( Table 2). Among the 33 selected established and candidate melanoma driver genes here described, clinical trials have been performed for 11 genes (33.3%) and this percentage could be increased if we considered ongoing trials with immunotherapy although it was not the topic of our review. Moreover, several preclinical studies are showing that some drugs could have anti-tumor ability toward melanoma driver genes opening future clinical studies and new potential therapeutic perspectives.
In this scenario, we strongly believe that in a near future, the use of NGS could extend the cohort of patients that could benefit by this therapeutic approach. Indeed, at least two clinical trials including melanoma patients (NCT02465060 and NCT02645149) are currently ongoing to prove this concept, enrolling hundreds of patients regardless of their neoplasm, and treating them with the specific drug targeting the specific identified mutation. Hopefully, the results of these experimentations will be a milestone in medical oncology because will allow us to put into practice a real personalized therapy, tailored on what is written in tumor genome.

AUTHOR CONTRIBUTIONS
IV and ET contributed equally in conceiving the review focus, conducting the literature review and summarizing the manuscript, analyzed data, wrote the first draft, and finalized the manuscript. PG designed and coordinated the review. BD, LP, VA, WB, AB, FS, and PG reviewed the literature, revised and made corrections to the manuscript. All authors have read and agreed to the final version of the manuscript.

FUNDING
We acknowledge the funds granted by the Italian Ministry of Health RF-2016-02362288, Ricerca Corrente to Ospedale Policlinico San Martino, and University of Genoa, FRA 2017.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmolb. 2020.00172/full#supplementary-material         FIGURE S10 | YAP1 mutations and CNV frequency in the 10 NGS selected studies. Mutations are indicated in green and CNV in purple. Only four studies with available CNV information were considered for CNV analysis.