Skip to content

Melanoma molecular subtypes and genomics

TL;DR — The Cancer Genome Atlas established the working genomic taxonomy of cutaneous melanoma from 333 tumours in 331 patients: four subtypes defined by the most prevalent significantly mutated genes — mutant BRAF, mutant RAS, mutant NF1 and triple-wild-type — with KIT mutations, focal amplifications and complex structural rearrangements enriched in triple-WT. Critically, genomic subtype did not correlate with outcome; what did was a transcriptomic immune subclass associated with lymphocytic infiltrate and high LCK protein (Cancer Genome Atlas Network 2015, PMID 26091043). Melanoma's defining genomic feature is a UV mutational signature present from benign precursor through invasive disease, with point-mutation burden rising monotonically along that path and copy-number alterations appearing only at invasion (Shain 2015, PMID 26559571). The 2018 WHO classification reorganised melanoma into nine types by evolutionary pathway rather than by histology, split first on sun-exposure relatedness and then on degree of cumulative solar damage (Elder 2020, PMID 32057276). The practically load-bearing genotype remains BRAF V600, because only it has a licensed matched therapy in cutaneous melanoma — see targeted therapy.

The TCGA framework

Subtype Defining alteration Notable features
BRAF BRAF hotspot mutation Most common subtype; matched inhibitors available
RAS NRAS/HRAS/KRAS mutation No matched licensed agent in melanoma
NF1 NF1 loss-of-function High mutation burden; MAPK activation without a hotspot
Triple wild-type None of the above Enriched for KIT mutations, focal amplifications and complex structural rearrangements

Source: 333 primary and/or metastatic melanomas from 331 patients, DNA/RNA/protein integrated (PMID 26091043). The key negative finding — no significant outcome correlation with genomic class — is often lost when the four-way scheme is used as a prognostic device. What predicted survival was the transcriptomic immune subclass: samples with high immune gene expression, lymphocytic infiltrate on pathology review and high LCK protein had improved survival, implicating tumour stroma immunobiology rather than driver genotype in the prognosis of regionally metastatic melanoma (PMID 26091043).

Driver frequencies in practice

Population BRAF NRAS Other Source
German routine care, 217 patients 40.1% 24.4% 2.3% concurrent; 33.2% wild-type for tested exons Heppt 2017, PMID 28797232
Japanese cutaneous, 37 patients 76% 8% — Hida 2024, PMID 39564955
Japanese acral, 52 patients 9% 17% KRAS 8%, KIT 19%, NF1 7% PMID 39564955
Japanese mucosal, 15 patients — major drivers in NRAS, KRAS, NF1, PTEN, GNAQ, KIT — PMID 39564955
KIT across 32 studies, 5,224 patients — — KIT mutated in 9.5% (497/5,224) Gong 2018, PMID 29746316
TERT promoter, sporadic melanoma — — 74% of metastatic cell lines, 85% of metastatic tissues, 33% of primaries Horn 2013, PMID 23348503

KIT mutation defines a distinct clinical subtype: significantly associated with older age (OR 1.296, 95% CI 1.025–1.641, P = .031), mucosal site (OR 1.363, 1.094–1.697, P = .006), acral site (OR 1.374, 1.123–1.682, P = .02) and chronically sun-damaged skin (OR 1.880, 1.127–3.136, P = .016), and negatively associated with non-CSD skin (OR 0.562, 0.392–0.805, P = .002) and extremity location. It was not associated with sex, histological type, Breslow thickness, ulceration, mitotic rate or stage (PMID 29746316). Genotype in melanoma therefore tracks anatomical site and UV history, not tumour aggressiveness.

Prognostically, NRAS-mutant tumours behaved more aggressively in routine-care data: higher frequency of nodal relapse (P = .013), of metastatic disease (P = .021), shortest time to loco-regional nodal relapse (P = .002), and NRAS mutation was an independent risk factor for progression (HR 2.01, 1.02–3.98). BRAF mutation associated with younger age at diagnosis (P < .001) and truncal primary site (P = .002) (PMID 28797232).

The evolutionary sequence

Sequencing 293 cancer-relevant genes across 150 areas of 37 primary melanomas and their adjacent precursors produced the first ordered account of melanoma genesis:

Stage Genetic events
Unequivocally benign precursor BRAF V600E exclusively
Intermediate lesion Enriched for NRAS mutations plus additional drivers; >1 pathogenic alteration defines the category
Intermediate lesion / melanoma in situ TERT promoter mutation in 77% — selected unexpectedly early
Invasive melanoma Biallelic CDKN2A inactivation appears exclusively here; copy-number alterations become prevalent
Advanced primary melanoma PTEN and TP53 mutations appear

Source: Shain 2015, PMID 26559571. Point-mutation burden rose from benign through intermediate to melanoma with a strong UV signature detectable at every stage, and genetically distinct subpopulations emerged as tumours progressed (PMID 26559571). A follow-up analysis of 230 histopathologically distinct areas from 82 patients, including matched primaries and regional metastases, ordered the pathway disruptions as: MAPK activation → telomerase upregulation → chromatin modulation → G1/S override → MAPK ramp-up → p53 disruption → PI3K activation. No mutation was specifically associated with metastatic progression — all these pathways were already perturbed during evolution of the primary, and UV-induced point mutations increased steadily until invasion, at which point copy-number alterations became prevalent (Shain 2018, PMID 29990500).

The clinical consequence is uncomfortable: the genomic events that distinguish melanoma from its precursor are the same events already present in some benign and intermediate lesions, which is the molecular restatement of the diagnostic-reproducibility problem in histopathology and prognostic factors and the mechanistic core of the overdiagnosis argument. The WHO responded by proposing melanocytoma as a term for intermediate tumours with increased but still low progression probability (Elder 2020, PMID 32057276).

Mutational signatures and UV

Analysis of 4,938,362 mutations from 7,042 cancers extracted more than 20 distinct mutational signatures, established the association of specific signatures with known mutagenic exposures, and described localised hypermutation ("kataegis") across cancer types (Alexandrov 2013, PMID 23945592). The expanded PCAWG/TCGA analysis of 84,729,690 somatic mutations across 4,645 whole genomes and 19,184 exomes identified 49 single-base-substitution, 11 doublet-base-substitution, 4 clustered and 17 indel signatures (Alexandrov 2020, PMID 32025018). Melanoma is the archetype of an exposure-attributable signature, and mouse work shows the interaction is genotype-dependent: after a single neonatal UVB exposure, melanomas from Braf-mutant mice carried on average twice as many single-nucleotide variants and five times as many dipyrimidine variants as tumours from Nras-mutant mice, and the UVB signature mirrored COSMIC human skin-cancer signatures more strongly in Braf-driven tumours; UVA did not accelerate tumour onset (Bowman 2021, PMID 34210801). This offers a mechanistic account of why BRAF-mutant melanomas arise preferentially at intermittently rather than chronically exposed sites.

The 2018 WHO reorganisation

The 4th-edition WHO classification abandoned the purely histological scheme in favour of nine types placed at the end of evolutionary pathways rooted in their respective precursors, split first into sun-related and non-sun-related and then, within sun-exposed skin, by histopathological degree of cumulative solar damage measured through solar elastosis:

Category Types
Low cumulative solar damage Superficial spreading melanoma
High cumulative solar damage Lentigo maligna melanoma, desmoplastic melanoma
Non-solar Acral melanoma, melanoma in congenital nevi, melanoma in blue nevi, Spitz melanoma, mucosal melanoma, uveal melanoma

Source: Elder 2020, PMID 32057276. This is the scheme that makes the anatomical–genomic correlations above coherent: BRAF with low-CSD trunk lesions, NF1 and KIT with high-CSD and acral/mucosal sites, GNAQ/GNA11 with uveal disease (see uveal melanoma).

Intratumoural and intertumoural heterogeneity

  • Between paired tumours. Across 271 tumours from 99 patients, mutations were detected in BRAF (39%), NRAS (21%) and TERT (78%); 18 of 99 patients (18%) showed mutational heterogeneity, including 13 with discordant TERT status, 7 with both TERT −124C>T and −146C>T at different allele frequencies in a single tumour, and one patient whose BRAF-mutant primary was not represented in at least one metastasis (Chang 2020, PMID 32087194).
  • Within tumours, at single-cell resolution. Single-cell RNA-seq of 4,645 cells from 19 patients found malignant cells within the same tumour in two distinct transcriptional states — MITF-high and MITF-low/AXL-high — in all tumours profiled, alongside T-cell exhaustion programs and patient-variable microenvironmental structure (Tirosh 2016, PMID 27124452). MITF/AXL phenotype switching is the transcriptional, non-genetic axis of resistance, distinct from the mutational resistance mechanisms in targeted therapy.
  • Resistance heterogeneity defeats single-biopsy sequencing. Among 59 BRAF-V600-mutant metastases progressing on dabrafenib or vemurafenib, resistance mechanisms were identified in 58%, occurred within and between patients and tumours, were present in some pre-treatment biopsies, and MAPK activity remained inhibited in 21% of resistant tumours with poor outcomes. No patient responded to a subsequent targeted therapy even when the progressing biopsy carried a mechanism predicted to be sensitive (Rizos 2014, PMID 24463458).
  • In Japanese melanoma cohorts, 11 of 21 patients with both primary and metastatic material showed distinct mutations in each (PMID 39564955).

Microenvironment as the prognostic axis

TCGA's outcome signal came from immune biology, not driver genotype (PMID 26091043), and subsequent work localised part of that signal to organised lymphoid structure. In metastatic melanoma samples, co-occurrence of CD8⁺ T cells and CD20⁺ B cells was associated with improved survival independently of other clinical variables; CXCR5/CXCL13/CD20 staining showed tertiary lymphoid structures in these tumours, a derived TLS gene signature predicted outcomes in checkpoint-treated cohorts, and B-cell-rich tumours had more TCF7⁺ naive/memory T cells while T cells in TLS-negative tumours had a dysfunctional molecular phenotype (Cabrita 2020, PMID 31942071). This is the mechanistic bridge from genomics to immunotherapy in advanced disease.

Late-stage genome architecture

Whole-exome and whole-genome sequencing of 88 tumour samples from 13 patients sampled at multiple sites and times found only limited gain of point mutations as disease progressed — with net mutational loss in some metastases — while whole-genome doubling and large-scale aneuploidy dominated. Widespread loss of heterozygosity sculpted the burden of point mutations, neoantigens and structural variants, in some patients already in treatment-naive primary cutaneous melanoma (Vergara 2021, PMID 33664264). Melanoma therefore has two distinct genomic phases: a UV-mutagenesis phase that builds the point-mutation and neoantigen load, and a genomic-instability phase that reshapes it. Mucosal melanoma, by contrast, is dominated by structural variation from the outset (Newell 2019, PMID 31320640) — see acral and mucosal melanoma.

Genomics of immunotherapy response and resistance

Mechanism Evidence
IFN-γ signalling as the response axis Transcriptomes of baseline and on-therapy biopsies from 101 advanced melanoma patients on nivolumab ± ipilimumab: T-cell infiltration and IFN-γ signatures correlated most strongly with response, with reciprocal decreases in cell-cycle and WNT pathways; modelling in 58 cell lines showed a conserved IFN-γ transcriptome response unless the IFN-γ receptor was altered (Grasso 2020, PMID 32916126)
T-cell-inflamed gene expression profile Learn-and-confirm across pembrolizumab studies from 19 melanoma patients to a pan-tumour signature in 220 patients with 9 cancers; IFN-γ-responsive genes covering antigen presentation, chemokines, cytotoxicity and adaptive resistance — necessary but not always sufficient for benefit (Ayers 2017, PMID 28650338)
Acquired JAK1/JAK2/B2M loss Paired baseline and relapse biopsies from four patients with initial response then late progression on pembrolizumab: loss-of-function JAK1 or JAK2 with wild-type allele deletion in two, truncating B2M in a third; JAK truncation abolished IFN-γ response, B2M truncation removed surface MHC class I (Zaretsky 2016, PMID 27433843)
PTEN loss Inhibits T-cell-mediated killing and reduces T-cell trafficking in preclinical melanoma; in patients correlates with lower T-cell infiltration, lower success of TIL expansion from resected tumours and inferior PD-1 inhibitor outcomes; PI3Kβ inhibition improved anti-PD-1 and anti-CTLA-4 efficacy in mice (Peng 2016, PMID 26645196)
Three-program resistance taxonomy Genome, transcriptome and high-dimensional flow profiling of short-term cell lines and matched tumours progressing on checkpoint inhibitors: (1) loss of wild-type antigen expression via tumour-intrinsic IFN-γ signalling and dedifferentiation, (2) disrupted antigen presentation through several independent MHC mechanisms, (3) immune exclusion associated with PTEN loss (Lim 2023, PMID 36934113)

The recurring theme is that IFN-γ signalling is simultaneously the mechanism of response and, when chronically engaged or genetically disrupted, a route to resistance — the same axis read in two directions (PMID 32916126; PMID 27433843; PMID 36934113). Therapeutic consequences are developed in cellular therapy and resistance.

Assays derived from tumour genomics

  • 31-gene expression profile. In an independent cohort of 523 primary cutaneous melanomas, 5-year recurrence-free survival was 88% for Class 1 versus 52% for Class 2 and distant metastasis-free survival 93% versus 60% (P < .001). The GEP remained significant in multivariate models alongside thickness and sentinel-node status (RFS HR 2.1; DMFS HR 2.7) and identified 70% of stage I–II patients who ultimately developed distant metastasis (Zager 2018, PMID 29402264). Prospective prognostic validation exists (Hsueh 2017, PMID 28851416; Keller 2019, PMID 30950242), and two prospective multicentre studies show that adding the assay changed sentinel-node-biopsy decisions and reduced procedures (Yamamoto 2023, PMID 36617959; Guenther 2025, PMID 39754143). Those decision-impact studies did not test recurrence or survival benefit, and no prospective study was located using the assay to select adjuvant systemic therapy.
  • Methylation-based classification of ambiguous melanocytic tumours. Parallel sequencing of 611 gene exons plus 850k EPIC methylation arrays showed Spitz nevi cluster independently of both nevi and melanoma with a distinct mutation profile; multiple copy-number alterations and TERT promoter mutations were found only in melanomas; histologically difficult spitzoid cases clustered between the two groups (Zaremba 2022, PMID 35737508). This is the most direct molecular attempt to date on the diagnostic problem that morphology cannot solve.

What genomics does and does not deliver clinically

Use Status
BRAF V600 testing to select BRAF/MEK inhibition Established; recommended from stage IIB/C by the European guideline (Garbe 2025, PMID 39700658)
KIT testing in acral/mucosal/CSD melanoma Defines a subtype with a candidate targeted approach; see acral and mucosal melanoma (PMID 29746316)
TCGA four-way class as a prognostic tool Not supported — no outcome correlation in the defining study (PMID 26091043)
Single progressing biopsy to choose next targeted agent Not supported — heterogeneity defeats it (PMID 24463458)
Immune/TLS signatures as response predictors Promising, not yet a clinical assay (PMID 26091043; PMID 31942071)
Genomic separation of indolent from progressive early lesions Absent — the events overlap across the benign–intermediate–malignant boundary (PMID 26559571; PMID 29990500)
31-gene expression profile Prognostically validated prospectively and shown to alter sentinel-node-biopsy decisions; no prospective evidence that its use improves recurrence or survival, or that it should select adjuvant systemic therapy (PMID 30950242; PMID 36617959; PMID 39754143)
Methylation profiling of ambiguous spitzoid tumours Research-grade discriminator, not a validated clinical test (PMID 35737508)

Interpretation rules for this page

  • Driver frequency is population-specific. BRAF is 40.1% in a German routine-care series and 76% in a Japanese cutaneous series with the same assay class; report the population, not "the" frequency (PMID 28797232; PMID 39564955).
  • Genotype tracks site and UV history, not aggressiveness. KIT associates with age and anatomical site but with no histological or stage variable (PMID 29746316).
  • A subtype scheme is not a prognostic scheme unless its defining study showed outcome separation; TCGA's did not (PMID 26091043).
  • Precursor-stage findings constrain diagnostic claims. A BRAF V600E or TERT promoter mutation does not establish malignancy — both occur in benign or intermediate lesions (PMID 26559571).
  • Heterogeneity is the default, not the exception. 18% of patients show discordant driver status between tumours (PMID 32087194) and every tumour profiled at single-cell resolution contained two transcriptional states (PMID 27124452).

Open questions

  • Can any molecular marker separate indolent from progressive melanoma in situ and thin invasive melanoma? The staged-evolution data say the current driver set cannot (PMID 26559571; PMID 29990500).
  • Why does TERT promoter mutation arise so early (77% of intermediate lesions and in situ melanoma) and what does that imply for its use as a diagnostic marker (PMID 26559571; PMID 23348503)?
  • Does the lower tumour mutational burden reported in East Asian cutaneous melanoma explain lower checkpoint-inhibitor efficacy, or is site and subtype mix the real variable (PMID 39564955)?
  • Is MITF-low/AXL-high phenotype switching therapeutically addressable, or only descriptive (PMID 27124452)?
  • Can tertiary lymphoid structures be induced, and would induction improve checkpoint response (PMID 31942071)?
  • Given that no mutation is specific to metastatic progression, what determines which primaries metastasise (PMID 29990500)?

References

  1. Cancer Genome Atlas Network. Genomic Classification of Cutaneous Melanoma. Cell. 2015;161:1681-96. PMID 26091043
  2. Shain AH, et al. The Genetic Evolution of Melanoma from Precursor Lesions. The New England journal of medicine. 2015;373:1926-36. PMID 26559571
  3. Elder DE, et al. The 2018 World Health Organization Classification of Cutaneous, Mucosal, and Uveal Melanoma: Detailed Analysis of 9 Distinct Subtypes Defined by Their Evolutionary Pathway. Archives of pathology & laboratory medicine. 2020;144:500-522. PMID 32057276
  4. Heppt MV, et al. Prognostic significance of BRAF and NRAS mutations in melanoma: a German study from routine care. BMC cancer. 2017;17:536. PMID 28797232
  5. Hida T, et al. Genetic Characteristics of Cutaneous, Acral, and Mucosal Melanoma in Japan. Cancer medicine. 2024;13:e70360. PMID 39564955
  6. Gong HZ, et al. The clinical significance of KIT mutations in melanoma: a meta-analysis. Melanoma research. 2018;28:259-270. PMID 29746316
  7. Horn S, et al. TERT promoter mutations in familial and sporadic melanoma. Science (New York, N.Y.). 2013;339:959-61. PMID 23348503
  8. Shain AH, et al. Genomic and Transcriptomic Analysis Reveals Incremental Disruption of Key Signaling Pathways during Melanoma Evolution. Cancer cell. 2018;34:45-55.e4. PMID 29990500
  9. Alexandrov LB, et al. Signatures of mutational processes in human cancer. Nature. 2013;500:415-21. PMID 23945592
  10. Alexandrov LB, et al. The repertoire of mutational signatures in human cancer. Nature. 2020;578:94-101. PMID 32025018
  11. Bowman RL, et al. UVB mutagenesis differs in Nras- and Braf-mutant mouse models of melanoma. Life science alliance. 2021;4. PMID 34210801
  12. Chang GA, et al. TERT, BRAF, and NRAS Mutational Heterogeneity between Paired Primary and Metastatic Melanoma Tumors. The Journal of investigative dermatology. 2020;140:1609-1618.e7. PMID 32087194
  13. Tirosh I, et al. Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq. Science (New York, N.Y.). 2016;352:189-96. PMID 27124452
  14. Rizos H, et al. BRAF inhibitor resistance mechanisms in metastatic melanoma: spectrum and clinical impact. Clinical cancer research : an official journal of the American Association for Cancer Research. 2014;20:1965-77. PMID 24463458
  15. Cabrita R, et al. Tertiary lymphoid structures improve immunotherapy and survival in melanoma. Nature. 2020;577:561-565. PMID 31942071
  16. Vergara IA, et al. Evolution of late-stage metastatic melanoma is dominated by aneuploidy and whole genome doubling. Nature communications. 2021;12:1434. PMID 33664264
  17. Newell F, et al. Whole-genome landscape of mucosal melanoma reveals diverse drivers and therapeutic targets. Nature communications. 2019;10:3163. PMID 31320640
  18. Grasso CS, et al. Conserved Interferon-γ Signaling Drives Clinical Response to Immune Checkpoint Blockade Therapy in Melanoma. Cancer cell. 2020;38:500-515.e3. PMID 32916126
  19. Ayers M, et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. The Journal of clinical investigation. 2017;127:2930-2940. PMID 28650338
  20. Zaretsky JM, et al. Mutations Associated with Acquired Resistance to PD-1 Blockade in Melanoma. The New England journal of medicine. 2016;375:819-29. PMID 27433843
  21. Peng W, et al. Loss of PTEN Promotes Resistance to T Cell-Mediated Immunotherapy. Cancer discovery. 2016;6:202-16. PMID 26645196
  22. Lim SY, et al. The molecular and functional landscape of resistance to immune checkpoint blockade in melanoma. Nature communications. 2023;14:1516. PMID 36934113
  23. Zager JS, et al. Performance of a prognostic 31-gene expression profile in an independent cohort of 523 cutaneous melanoma patients. BMC cancer. 2018;18:130. PMID 29402264
  24. Hsueh EC, et al. Interim analysis of survival in a prospective, multi-center registry cohort of cutaneous melanoma tested with a prognostic 31-gene expression profile test. Journal of hematology & oncology. 2017;10:152. PMID 28851416
  25. Keller J, et al. Prospective validation of the prognostic 31-gene expression profiling test in primary cutaneous melanoma. Cancer medicine. 2019;8:2205-2212. PMID 30950242
  26. Zaremba A, et al. Genetic and methylation profiles distinguish benign, malignant and spitzoid melanocytic tumors. International journal of cancer. 2022;151:1542-1554. PMID 35737508
  27. Garbe C, et al. European consensus-based interdisciplinary guideline for melanoma. Part 1: Diagnostics - Update 2024. European journal of cancer (Oxford, England : 1990). 2025;215:115152. PMID 39700658
  28. Yamamoto M, et al. The 31-gene expression profile test informs sentinel lymph node biopsy decisions in patients with cutaneous melanoma: results of a prospective, multicenter study. Current medical research and opinion. 2023;39:417-423. PMID 36617959
  29. Guenther JM, et al. A prospective, multicenter analysis of the integrated 31-gene expression profile test for sentinel lymph node biopsy (i31-GEP for SLNB) test demonstrates reduced number of unnecessary SLNBs in patients with cutaneous melanoma. World journal of surgical oncology. 2025;23:5. PMID 39754143