We report here a meta-analysis of genomic data from cutaneous melanoma, offering more reliable gene mutation prevalence than data derived from individual studies, as previously demonstrated for renal cancer and in breast cancer brain metastases [8, 11].The most frequently mutated genes were BRAF (particularly BRAF V600E), NRAS, TERT, TP53, NF1, ROS1, CDKN2A, and PIK3CA. We compared our findings with data from The Cancer Genome Atlas and did not include other datasets such as the International Cancer Genome Consortium and TFRI-MOH melanoma dataset, as these datasets are more heterogeneous and less suitable for a standardized prevalence comparison. Our findings are consistent with data from The Cancer Genome Atlas in some key alterations, particularly BRAF (including BRAFV600E) and NF1, as well as recurrent alterations in TP53 and PTEN, particularly for less frequent alterations such as ARID2, IDH1, and PPP6C (Table S11 of the ESM) [12]. These discrepancies may reflect differences in sample size and cohort composition, as our meta-analysis included a substantially larger and more heterogeneous population. These frequently mutated genes enable the identification of three major signaling pathways leading to cell proliferation, invasion, and metastatic process: the RAS-RAF-MAPK pathway, the PIK3-AKT-mTOR pathway, and the CDKN2A-TP53 axis (Fig. 3).

Schematic of signaling pathways involved in cutaneous melanoma carcinogenesis. CDK cyclin-dependent kinase, mRNA messenger RNA

The BRAF and RAS genes, and more broadly the MAPK pathway, are at the cornerstone of cutaneous melanoma carcinogenesis for more than half of patients. The MAPK pathway plays a crucial role in the pathogenesis of many cancers, including melanoma. Among the key oncogenic drivers, RAS and RAF—located at 1p13.2 and 7q34, respectively—are among the most frequently altered genes. In our meta-analysis, we observed high heterogeneity in the reported prevalence of BRAF mutations, indicating considerable variability across studies. Therefore, a comprehensive synthesis of existing data is necessary to provide more accurate estimates. Our study addressed this need by conducting a systematic meta-analysis to generate reliable pooled data. The BRAF gene, located at 7q34, encodes a serine/threonine kinase that functions downstream of RAS in the MAPK/ERK signaling pathway. The BRAF V600E mutation, resulting from the substitution of valine (V) with glutamic acid (E) at position 600, leads to constitutive activation of BRAF as a monomer and to an approximately 500-fold increase in kinase activity, driving uncontrolled cell proliferation [14]. Specific inhibitors such as vemurafenib, encorafenib, and dabrafenib have shown efficacy in BRAF V600-mutant melanoma, particularly in patients with the V600E and V600K variants [15, 16]. The higher prevalence of BRAF and BRAF V600E mutations, as well as BRAF amplification in metastatic lesions compared with primary tumors suggests an important role during melanoma progression. Because of the differences in mutation frequencies observed between metastases and primary tumors in our study, a re-biopsy of metastatic lesions is crucial for optimizing therapeutic strategies and guiding precision medicine approaches.

The RAS superfamily encodes small GTP-binding proteins-HRAS, NRAS, and KRAS, which function as monomeric membrane-localized G proteins regulating critical cellular processes such as proliferation, migration, senescence, differentiation, apoptosis, and survival [17]. In its active GTP-bound state, RAS interacts with and activates RAF kinases, subsequently triggering the MAPK signaling cascade. In melanoma, NRAS is the most frequently mutated RAS isoform and is considered the most biologically significant [18]. Notably, KRAS mutations have been reported to possess lower oncogenic potential in melanoma compared with NRAS [19]. In our analysis, HRAS and KRAS mutations were rarely detected in primary melanomas (3% and 1%, respectively) and were present in fewer than 1% of metastatic tumors. In contrast, NRAS mutations were linked to the metastatic process, particularly brain metastases. Interestingly, this increase was not explained by the hotspot mutations Q61K and Q61R. In another meta-analysis we have performed on breast cancer genomic data, NRAS mutations were more frequent in brain metastases than in primary tumors, but not in extracerebral metastases [11]. Whilst NRAS is an attractive target for melanoma therapy, the development of direct RAS protein inhibitors has seen limited success, likely because RAS binds GTP with very high affinity, making it difficult to effectively target the binding site [20]. In NRAS-mutant melanoma cells, the NRAS protein transmits signals to MEK primarily through CRAF rather than BRAF [21]. As a result, treatment with BRAF inhibitors is ineffective in this context [22]. In such patients, MEK inhibitors such as binimetinib have shown promising therapeutic efficacy [23]. In contrast to other tyrosine kinase inhibitors such as osimertinib or tucatinib that have been developed for their good biodistribution to the central nervous system, there are no such data yet for binimetinib.

The NF1 gene, located at 17q11.2, encodes neurofibromin, a tumor suppressor protein that negatively regulates RAS signaling. Neurofibromin functions as a GTPase-activating protein, facilitating the conversion of active RAS-guanosine triphosphate to its inactive form, RAS-guanosine diphosphate, thereby suppressing downstream RAS signaling [24]. BRAF V600E, NRAS Q61, and NF1 mutations are usually exclusive [25]. In our data, NF1 mutation prevalence was stable during disease progression, suggesting its involvement in early tumor development but also the importance of targeting this pathway when there is only NF1 biallelic inactivation.

TERT, located on 5p15.33, plays a critical role in tumorigenesis by maintaining chromosomal stability through telomere length regulation, thus preventing cellular senescence [26]. Similar to BRAF and NRAS, in our study, the frequency of TERT promoter mutations significantly increased from primary to metastatic melanoma, suggesting a major role for TERT alterations in disease progression. BRAF, NRAS, and TERT promoter mutations frequently co-occur and function synergistically to sustain cellular immortality by maintaining TERT expression and activity. These insights suggest that co-targeting BRAF and TERT may represent a promising therapeutic strategy to overcome resistance to BRAF inhibitors in melanoma [27].

The PIK3-AKT-mTOR pathway is closely linked to the RAS-RAF-MAPK pathway, as RAS can directly activate the PI3K/AKT pathway, further contributing to oncogenic signaling [28]. PIK3CA, located at 3q26.32, is an important gene in numerous cancers while much less frequently mutated in metastatic melanomas than in breast or renal cancers [8, 11]. In our meta-analysis, the prevalence of PIK3CA mutations and of PTEN LOH decrease in metastases, suggesting a less prominent role of the PI3K-AKT-mTOR pathway in the metastatic process. This pathway is a potential therapeutic target currently under investigation in melanoma. Everolimus, an mTORC1 inhibitor, has demonstrated the ability to reduce invasiveness and induce apoptosis in melanoma cells in preclinical studies [29, 30]. However, everolimus either as monotherapy or in combination with BRAF inhibitors or chemotherapy did not lead to clinical efficacy with the limitation that patients were not selected with PIK-AKT-mTOR pathway alterations [31,32,33].

MYC, located at 8q24.21, encodes the MYC protein, a transcription factor that plays a central role in tumorigenesis by suppressing key cellular safeguards, including apoptosis, cell-cycle arrest, differentiation, and senescence [34]. MYC can be activated downstream of RAS signaling via two major pathways: the ERK/MAPK pathway and the PI3K/AKT pathway [35]. In our meta-analysis, MYC amplification is frequently observed in melanoma. Although direct targeting of MYC remains challenging, several therapeutic strategies aiming to inhibit MYC activity are currently under development and hold promise as a potential approach for the treatment of cancer, including melanoma [36].

The two signaling pathways are closely linked to transmembrane receptors with tyrosine kinase domains. Typically, rare mutations and copy number variations are identified in RTK such as KIT, ALK, MET, or ERBB4, with potential therapeutic implications. In fact, melanoma cells expressing mutant ERBB4 had reduced cell growth after treatment with the ERBB inhibitor lapatinib [37]. Among RTKs, ROS1 is of particular interest in melanoma carcinogenesis. While ROS1 fusions are well-established oncogenic drivers in multiple malignancies [38], they remain exceedingly rare in melanoma. In contrast, ROS1 mutations are frequently observed [39], in 14% of metastatic lesions in our meta-analysis. However, only 25% of these mutations are activating mutations in the tyrosine kinase domain, while the remaining 75% are located outside this domain. Mutations outside the tyrosine-kinase domain are predictive of response to immunotherapy, with a longer overall survival. This favorable responsiveness of an ROS1 mutation to immune checkpoint inhibitor therapy may be related to the elevated tumor antigenicity, as the ROS1 mutation was associated with a higher tumor mutation burden relative to the wild-type counterpart in melanoma. [39]. Although ROS1-targeted therapies have demonstrated clinical efficacy in ROS1 fusion-positive non-small cell lung cancer, [40,41,42], their role in melanoma harboring ROS1 mutations remains unclear. Given these findings, the inclusion of ROS1 in a systematic molecular testing panel should be considered to guide potential treatment with immunotherapy. Mutations in other RTKs such as KIT, ALK, or MET were rare, identified in less than 5% of metastatic lesions, but preliminary data indicate a promising therapeutic response in melanoma [43,44,45]. Therefore, inclusion of these genes in molecular testing panels for patients with melanoma is warranted.

The CDK2NA-TP53 pathway is the third major pathway involved in melanoma carcinogenesis. CDKN2A, located on 9p21.3, plays a pivotal role in melanoma pathogenesis. It encodes the p16 protein, a key regulator of the cell cycle that inhibits abnormal cell growth and proliferation by binding to cyclin-dependent kinase (CDK) 4/6 complexes and cyclin D, thereby inducing cell-cycle arrest in the G1 phase [46]. Germline CDKN2A mutations are identified in approximately 20–40% of familial melanoma cases [47]. In our meta-analysis, CDKN2A mutations were identified in approximately 10% of both primary and metastatic melanoma tumors. CDKN2B, located at the same chromosomal locus (9p21.3), encodes the tumor suppressor protein p15, which functions similarly to p16 by inhibiting CDK4/6. These two genes, with MTAP, are frequently co-deleted because of their close genomic proximity [48]. In vitro, p15 and p16 cooperate to suppress the progression from benign melanocytic nevi to malignant melanoma [48]. Additionally, 9p21.3 loss, which contributes to p16, p15, and MTAP inactivation, was observed in 41.7% of primary tumors and 26.3% of metastatic lesions. This may have both prognostic and therapeutic implications, particularly in the context of using CDK4/6 inhibitors in melanoma treatment [49]. Combining palbociclib, a CDK4/6 inhibitor, with the BRAF V600E inhibitor vemurafenib resulted in promising efficacy and manageable toxicity in patients with metastatic melanoma harboring BRAF V600E mutations, CDKN2A loss, and preserved RB expression [50]. Moreover, MTAP loss presents a significant opportunity for therapeutic targeting. MTAP deletion causes an accumulation of methylthioadenosine, disrupting cellular methylation. Recent studies suggest that MAT2A inhibitors, which reduce the methyl donor S-adenosylmethionine, may selectively target MTAP-deleted tumors [51]. This approach, which exploits the synthetic lethality of combining MTAP loss with MAT2A inhibition, could represent a novel and effective therapeutic strategy in melanoma. In the CDKN2A pathway, RB1 is a key tumor suppressor protein that inhibits E2F transcription factors and directly regulates cell-cycle arrest, thereby preventing uncontrolled proliferation [52]. RB1 is predominantly implicated in retinoblastoma, where its inactivation plays a central role in tumorigenesis. However, in melanoma, RB1 mutations are relatively rare. In our meta-analysis, their prevalence was 4% in metastatic lesions, while no mutations were detected in primary tumors.

TP53, located on 17p13.1, is a key tumor suppressor gene that plays a crucial role in cancer pathogenesis by regulating cell-cycle arrest, apoptosis, and genomic stability [53]. In cutaneous melanoma, TP53 mutation prevalence does not significantly vary between primary tumors and metastases, suggesting that TP53 alterations are early carcinogenetic events not specifically linked to the metastatic process. In addition, TP53 mutation prevalence is much lower than in other cancers despite TP53 being particularly vulnerable to ultraviolet-induced DNA damage. Ultraviolet-induced lesions typically give rise to C>T transitions and CC>TT tandem substitutions, patterns encompassed within COSMIC Signature 7 [54], a point that we could not address here because of a lack of individual data. Neither could we address the TP53 genotype in matched normal skin/melanoma in the same patients, TP53 mutations being very early events detected in keratinocytes of histologically normal sun-exposed skin [55]. TP53 mutations are a predictive factor for poor response and shorter survival in patients treated with anti-cytotoxic T-lymphocyte antigen-4 therapy [56]. In addition to mutations, functional inactivation of the p53 pathway can occur through alternative mechanisms, such as MDM2 and PPM1D overexpression or CDKN2A loss, leading to an impaired apoptotic response and increased tumor survival. The MDM2 gene, located on 12q15, encodes the MDM2 protein, which acts as an E3 ubiquitin ligase for p53, thus suppressing its transcriptional activity [57]. In our study, a 12q15 gain was observed in 15.2% of primary tumors, decreasing to 6.2% in metastatic tumors (P = 0.07). Additionally, our meta-analysis identified a 17q23.2 gain, potentially involving PPM1D, a wild-type p53-induced phosphatase 1 that negatively regulates TP53 function by dephosphorylating MDM2 at serine 395, thereby enhancing MDM2 interaction with p53 [57]. In addition, pharmacological blockade of the Hedgehog pathway with the SMOOTHENED antagonist cyclopamine acts synergistically with inhibition of PPM1D in reducing the growth of melanoma and breast cancer cells in vitro [58]. Given its pivotal role in tumor suppression, strategies aimed at restoring p53 function or targeting its downstream pathways hold potential for future therapeutic development in melanoma.

Acral melanoma arises on the palms, soles, or nail unit, which are typically sun-protected sites. This subtype accounts for only a small proportion of cutaneous melanomas in White populations but represents more than 50% of melanoma cases in non-White individuals, including Asian individuals, Hispanic individuals, and Black Americans. In our study, among the 14 studies reporting Asian ethnicity, acral melanoma accounted for 51.5% of the analyzed samples [59]. In addition, acral melanoma is characterized by a distinct molecular profile, with lower mutation rates compared with non-acral cutaneous melanoma [60]. However, to date, few studies have properly compared the mutation rates between these two subtypes. Our study provides reliable evidence of these differences, showing that acral melanoma harbored significantly fewer BRAF mutations, as well as lower frequencies of TERT and NF1 mutations, than non-acral melanoma.

A key finding from our meta-analysis for daily care is that targeted NGS panels provided results comparable to those of broader NGS approaches for the most frequently mutated genes, such as BRAF and NRAS. This was not the case for TERT promoter mutations, which were less consistently detected by NGS approaches. Our results suggest that targeted NGS panels focusing on commonly mutated genes may represent an efficient and cost-effective strategy for routine clinical practice. However, whole exome sequencing/whole genome sequencing are of particular interest for precision molecular medicine, ideally combined with transcriptomic and/or epigenomic data. In addition, formalin-fixed samples provide reliable data on gene mutation prevalence for the most frequently mutated genes, supporting their use in routine practice. However, one should keep in mind that fixation-induced DNA fragmentation and RNA degradation may limit the use of some technologies such as RNA sequencing. Indeed, RNA from formalin-fixed tissues yields lower RNA quality, lower proportions of reads, and a lower concordance between RNA and DNA variants compared with fresh or frozen samples [61]. For personalized medicine, anti-cancer treatments guided by large genomic analyses and particularly multiple data sources, including transcriptomic data, are more likely to result in successful treatments [62]. In contrast to what we had previously reported for renal carcinoma [63], multiple sampling did not significantly increase mutation prevalence, except for BRAF mutations.

Our meta-analysis has some limitations. Our meta-analysis did not employ a paired primary-metastasis design and therefore cannot address genomic evolution or clonal changes within individual patients. In another meta-analysis on 1220 patients, paired samples showed mutational concordance for key driver gens, BRAF and NRAS, between primary melanomas and their matched metastases [64]. Our study on 13,655 patients enables an accurate estimation of the pooled prevalence of numerous gene mutations and CNAs across large cohorts of primary and metastatic cutaneous melanomas derived from heterogeneous real-world studies.

For tumor suppressor genes, such as TP53, because of heterogeneity in reporting across studies, monoallelic and biallelic inactivation events could not be systematically distinguished in the present meta-analysis. In addition, our study was conducted on published rather than individual patient data, with missing information preventing us from performing correlative analyses with treatment response and survival. Consequently, subgroup analyses according to systemic therapies, age, sex, and ethnicity could not be performed. Furthermore, the interpretation of locus-specific CNA frequencies, including LOH, should be made with caution, as the included studies employed heterogeneous sequencing platforms with different sensitivities for CNA detection.

In addition, we could not fully explain the question of heterogeneity, like for our previous meta-analyses, except that it was less marked for metastatic samples. The substantial heterogeneity observed across analyses could not be fully accounted for by methodological differences alone. Therefore, the pooled genomic estimates in our study should be interpreted in the context of underlying clinical heterogeneity. Cutaneous melanoma encompasses biologically distinct subtypes, including superficial spreading, nodular, and lentigo maligna melanoma, each characterized by differences in anatomic distribution, cumulative ultraviolet exposure, and mutational profile. Lentigo maligna melanoma and nodular melanoma are both commonly associated with cumulative sun damage. However, lentigo maligna melanoma is more frequently associated with TP53 mutations, whereas nodular melanoma is more often linked to NRAS mutations. In contrast, superficial spreading melanoma is considered a lower cumulative sun damage subtype and more commonly harbors BRAF V600 mutations [65]. Beyond subtype, factors such as anatomic site, cumulative ultraviolet exposure, and Breslow thickness may further influence the genomic profile. In addition, genomic testing in routine clinical practice is more frequently performed in high-risk primary or metastatic melanomas, potentially resulting in under-representation of lower risk lesions. Accordingly, the substantial heterogeneity observed in this analysis likely reflects both methodological variation and intrinsic clinicogenomic diversity across included cohorts. The present findings should therefore be interpreted as descriptive summaries of reported genomic prevalence rather than definitive epidemiologic estimates.

One methodological consideration of our study lies in the comparison of pooled prevalence estimates derived from separate random-effects meta-analyses of primary and metastatic melanoma. These analyses included partly different sets of studies and therefore do not reflect paired or within-study comparisons. In addition, the Z-test does not fully account for variability within and across studies. Therefore, the observed differences should be interpreted with caution. However, as paired primary-metastatic datasets remain limited, this approach allows the use of a larger body of evidence and provides a robust overview of mutation patterns in primary and metastatic melanoma.