Genomic Classification of Cutaneous Melanoma

PMID: 26091043 · DOI: 10.1016/j.cell.2015.05.044 · Journal: Cell (2015)

TL;DR

The TCGA Network performed integrated multi-platform genomic, transcriptomic, and proteomic characterization of 333 primary and/or metastatic cutaneous melanomas from 331 patients, establishing a framework for genomic classification into four subtypes based on the most prevalent significantly mutated genes: BRAF-mutant (52%), RAS-mutant (NRAS/HRAS/KRAS hot-spot, 28% NRAS alone), NF1-mutant (14%), and Triple-WT (wild-type for all three, n = 46). The Triple-WT subtype was enriched for KIT mutations, focal amplifications of receptor tyrosine kinases (KIT, PDGFRA, KDR), and complex structural rearrangements, with notably few TERT promoter mutations and UV signatures. Although genomic subtype did not correlate with post-accession survival, a transcriptomic “immune” subclass — characterized by high immune-gene expression, high lymphocyte score on pathology review, and elevated LCK protein — was associated with significantly improved survival in patients with regional metastatic disease.

Cohort & data

• 333 cutaneous melanoma samples from 331 adult patients across 14 tissue source sites: 67 (20%) primary cutaneous melanomas (all non-glabrous skin) and 266 (80%) metastases. Of the metastases, 160 were from regional lymph nodes, 52 from regional skin/soft tissue, and 35 from distant sites. Median primary thickness 2.7 mm (mean 4.9 mm).
• Cancer type: SKCM (Skin Cutaneous Melanoma, OncoTree).
• Dataset: skcm_tcga_pub_2015 (cBioPortal study id; raw data at GDAC Firehose stddata__2013_11_14 and the TCGA SKCM Data Portal).
• Assays / platforms:
  • [[whole-exome-sequencing|WES]] on 320 samples (paired tumor + germline for 318 patients; Agilent SureSelect Human All Exon v2.0 44 Mb, Illumina HiSeq 2 × 76 bp PE, ~87× mean exon coverage).
  • DNA copy-number profiling on 333 samples (Affymetrix SNP 6.0).
  • mRNA sequencing on 331 samples (Illumina mRNA TruSeq, HiSeq 2000).
  • microRNA sequencing on 323 samples.
  • DNA methylation profiling on 333 samples (Illumina 450K).
  • Reverse-phase protein array (RPPA, 181 cancer-related proteins/phosphoproteins) on 202 samples.
  • [[whole-genome-sequencing|WGS]] (deep) on 38 samples; low-pass WGS on 119 samples.
  • PCR-Sanger sequencing of TERT promoter C228T/C250T in 115 samples.
• Complete six-platform data available on a core set of 199 samples.
• Significantly mutated genes called by MutSig (Q < 0.1) and InVEx (Bonferroni p < 0.05 or Q < 0.1).
• Copy-number purity/ploidy via ABSOLUTE; integrative subtypes via iCluster; transcriptomic clusters via consensus-hierarchical-clustering; complex structural rearrangements via ShatterSeek; reads aligned with BWA / Picard / Firehose at the Broad.

Key findings

• Mutation burden: mean 16.8 mutations/Mb — the highest reported for any TCGA tumor type at the time. 228,987 total mutations (SNVs + indels) identified across 318 WES samples; 96% targeted-validation rate on 455 SMG SNVs.
• UV signature: 76% (44/58) of primaries and 84% (221/262) of metastases had a UV signature (C>T at dipyrimidines > 60% or CC>TT > 5%); median C>T at dipyrimidines 77.7%; median CC>TT 3.9%.
• 13 InVEx SMGs / 42 MutSig SMGs: BRAF, NRAS, CDKN2A, TP53, PTEN, RAC1 (UV hot-spot, 6.9%), MAP2K1, PPP6C, ARID2, NF1, IDH1 (R132, ~6%, UV hot-spot 6.2%), RB1, and novel candidate DDX3X (RNA helicase). MutSig also flagged MRPS31 and RPS27 ribosomal-protein SMGs with UV-induced 5′ UTR hot-spots (~5% and ~9%).
• Four genomic subtypes (n = 318 WES cases):
  • BRAF subtype (52%, n = 166): 145 V600 alterations — V600E (124), V600K (18), V600R (3); 5 K601 mutations. BRAF V600/K601 anti-correlated with NRAS hot-spots (Fisher’s p < 1e–15); non-hot-spot BRAF mutations co-occurred with NRAS/HRAS/KRAS and NF1 mutations.
  • RAS subtype (28% NRAS, n = 88): NRAS hot-spots Q61R (35), Q61K (28), Q61L (11), Q61H (4), G12R/D/A (4), G13R/D (3); 4 HRAS (G13D, G13S, Q61K×2) and 3 KRAS (G12D, G12R, Q61R) — all mutually exclusive with BRAF V600/K601 and NRAS.
  • NF1 subtype (14%, n = 28): 65 mutations; >50% LoF (27 nonsense, 9 splice, 4 frameshift; InVEx LoF p = 1.8e–11, Q = 9.1e–12). Highest mutation prevalence (39 mut/Mb, ~2× other subtypes). NF1 mutated in 38.7% of non-hot-spot BRAF/NRAS samples (29/75) and ~70% (26/38) of those with UV signature. Anti-correlated with hot-spot BRAF (p = 1.93e–9).
  • Triple-WT subtype (n = 46): no hot-spot BRAF/RAS/NF1. Recurrent COSMIC alterations included KIT (n = 6), CTNNB1 (n = 3), GNA11 Q209L (n = 2), GNAQ Q209P (n = 1), EZH2 (n = 1). Two GNA11 hot-spot tumors co-occurred with SF3B1 R625H but lacked BAP1 mutations (contrast with uveal melanoma).
• UV signature is depleted in Triple-WT: only 30% (14/46) Triple-WT vs 90.7% BRAF / 93.5% RAS / 92.9% NF1 (Fisher’s p = 1e–15).
• Copy-number landscape (Triple-WT-enriched): significantly more CN segments; focal amplifications of KIT / PDGFRA / KDR (4q12), CDK4, CCND1, MDM2, TERT (p < 0.01–0.05, FDR < 0.05). In the BRAF subtype: focal amplifications of BRAF, MITF (melanocyte lineage oncogene; p < 0.01, FDR < 0.05), and CD274 (PD-L1 ligand). NRAS amplifications co-occurred with NRAS mutations.
• Structural rearrangements: 224 candidate fusion drivers identified by integrative CNA/RNA-seq/WGS analysis. Recurrent partner promiscuity around BRAF (ATG7-BRAF, TAX1BP1-BRAF), RAF1 (TRAK1-RAF1, RAF1-AGGF1, CLCN6-RAF1), AKT3 (CEP170-AKT3, AKT3-PLD5, ZEB2-AKT3, ARHGAP30-AKT3), 3 MITF fusions (MITF-FOXP1, CADM2-MITF, FRMD4B-MITF), and 3 HMGA2 fusions. 8 fusions had a predicted intact kinase domain. Driver fusions and complex rearrangements (ShatterSeek) significantly enriched in Triple-WT (fusion enrichment p = 2e–04; complex rearrangement 11/16, Fisher’s p = 0.00098); 7/7 Triple-WT samples with complex rearrangements lacked UV signature.
• TERT promoter mutations: C228T (23.5%) and C250T (40.9%) mutually exclusive (n = 115). Only C228T was associated with elevated TERT mRNA (rank-sum p = 0.001), contrasting with GBM (difg_glass) where both increase expression. Frequencies by subtype: BRAF 75% (39/52), RAS 71.9% (23/32), NF1 83.3% (10/12), Triple-WT only 6.7% (1/15) (p = 8e–5) — suggesting alternative TERT activation (amplification/rearrangement) in Triple-WT.
• CIMP phenotype: strongest associations with IDH1 (OR = 4.05, p = 0.005) and ARID2 (OR = 3.5, p = 0.0003) — both chromatin-related. Higher NRAS hot-spot (OR = 2.3, p = 0.003) and lower BRAF hot-spot (OR = 0.4, p = 0.0008) in CIMP cluster.
• RPPA signaling differences: phospho-MEK1/2 (S217/S221) elevated in BRAF and RAS subtypes; highest phospho-ERK1/2 (MAPK1/MAPK3 T202/Y204) in RAS subtype; highest KIT protein abundance in Triple-WT; highest CRAF expression in NF1 subtype; highest BCL2 in Triple-WT; elevated IGFBP2 in BRAF hot-spot mutants.
• Pathway integration (n = 318): RAS-MAPK-AKT altered in 91%, RB1/CDKN2A cell-cycle in 69%, MDM2/TP53 apoptosis in 19%. 46/49 (93.9%) TP53 mutations and all 12 RB1 mutations occurred in UV-signature samples. PTEN mutations/deletions enriched in BRAF-mutant; AKT3 amplification/overexpression enriched in RAS, NF1, and Triple-WT vs BRAF (p < 0.05).
• Three transcriptomic subclasses (n = 329, top-1,500 variant genes):
  • “Immune” (51%, n = 168) — enriched for T-cell, B-cell, NK-cell, checkpoint, cytokine genes; favorable post-accession survival (log-rank p = 0.003 among regional metastases). Not driven by lymph-node tissue admixture (no signature difference between LN and non-LN samples).
  • “Keratin” (31%, n = 102) — high keratin/pigmentation/epithelial expression; 74% of primaries clustered here; worst regional-metastasis outcome (log-rank p = 0.0007 vs Immune/MITF-low).
  • “MITF-low” (18%, n = 59) — low MITF target / pigmentation / epithelial expression; enriched for nervous-system / neuronal-development genes; highest BRAF hot-spot frequency (66% vs 33%/45% for keratin/immune; Fisher’s p = 0.0003).
• iCluster integration identified an Immune-mRNA + low-CN + normal-like methylation cluster, a hypomethylation + MITF-expression cluster, and a CIMP + keratin + miRNA-cluster-3 group.
• Lymphocyte score (LScore): standardized pathology review by AWG dermatopathologists. Higher in LN-derived tumors (Wilcoxon p = 5.6e–8). Elevated LScore significantly associated with prolonged post-accession survival in regional metastases; strikingly concordant with assignment to the Immune transcriptomic subclass (Fisher’s p < 1e–12).
• Protein markers (LCK, SYK): both T/B-cell tyrosine kinases were highly expressed in tumors enriched for Immune subclass and high LScore. Only high LCK (not SYK) was associated with favorable post-accession survival.
• Tri-feature prognostic model: combining LScore + Immune subclass + LCK expression predicted regional-metastasis outcome better than any single feature (log-rank p = 8.0e–6).
• Bivariate clinical-deployable model: high LScore + high LCK vs low/low had log-rank p = 7.9e–5, HR = 5.5; both terms independently significant in multivariable Cox regression — proposed as a more practical replacement for unsupervised transcriptomic clustering.

Genes & alterations

• BRAF — 52% mutated; V600E (n = 124), V600K (18), V600R (3), K601 (5); recurrent fusion partners ATG7 and TAX1BP1 (kinase domain intact); focal amplifications in BRAF subtype. Anti-correlated with hot-spot NRAS and NF1.
• NRAS — 28% mutated; Q61R/K/L/H dominant; mutually exclusive with hot-spot BRAF V600/K601; amplifications co-occur with NRAS mutations.
• HRAS / KRAS — rare hot-spot mutations (HRAS G13D/G13S/Q61K×2; KRAS G12D/G12R/Q61R); mutually exclusive with NRAS and BRAF V600/K601.
• NF1 — 14% mutated; >50% LoF; defines third MAPK-activating subtype; ~70% of non-hot-spot BRAF/NRAS UV-signature tumors carry NF1 mutations.
• RAC1 — UV hot-spot in 6.9%; previously linked to BRAF-inhibitor resistance (Van Allen et al., 2014; Watson et al., 2014).
• IDH1 — R132 in ~6%; UV hot-spot; enriched in CIMP cluster (OR = 4.05, p = 0.005).
• ARID2 — chromatin-remodeling SMG strongly enriched in CIMP (OR = 3.5, p = 0.0003); frequently co-occurs with NF1 — proposed synthetic-lethal vulnerability for chromatin-modifier targeting.
• PPP6C — hot-spot SMG frequently co-occurring with RAS mutations; proposed Aurora kinase inhibitor combination target.
• MAP2K1 — recurrent SMG in MAPK pathway.
• TP53 — 49 mutations; 93.9% (46/49) in UV-signature samples; enriched in BRAF/RAS/NF1 vs Triple-WT.
• RB1 — 12 mutations, all in UV-signature samples; enriched in NF1 subtype.
• CDKN2A / CDKN2B — alterations evenly distributed across subtypes.
• CDK4, CCND1 — focal amplifications enriched in Triple-WT.
• PTEN — mutations/deletions enriched in BRAF-mutant melanomas.
• AKT3 — amplification + overexpression and recurrent fusions (CEP170-AKT3, AKT3-PLD5, ZEB2-AKT3, ARHGAP30-AKT3) enriched in RAS/NF1/Triple-WT vs BRAF (p < 0.05).
• PIK3CA / AKT1 — recurrent mutations (PIK3CA E545K, H1047L; AKT1/3 E17K) nominated as biomarkers for combination MEK + PI3K/AKT/mTOR therapy.
• KIT — mutations and 4q12 focal amplifications enriched in Triple-WT; highest KIT protein abundance in Triple-WT by RPPA; rationale for imatinib/dasatinib.
• PDGFRA, KDR — co-amplified with KIT at 4q12 in Triple-WT; rationale for sorafenib, crenolanib, regorafenib, pazopanib.
• MDM2 — focal amplifications enriched in Triple-WT; rationale for MDM2 inhibitors (nutlin-3, amg-232).
• TERT — promoter C228T (23.5%) and C250T (40.9%) mutually exclusive; only C228T raises TERT mRNA; depleted in Triple-WT (6.7%, p = 8e–5) — Triple-WT activates TERT via amplification or rearrangement.
• MITF — focal amplifications in BRAF subtype; recurrent fusions (MITF-FOXP1, CADM2-MITF, FRMD4B-MITF); defines low-expression transcriptomic subclass.
• CD274 (PD-L1) — focal amplifications in BRAF subtype; relevant to pembrolizumab/nivolumab biomarker hypothesis.
• PDCD1 (PD-1) — expression elevated in Immune transcriptomic subclass.
• GNAQ / GNA11 — rare Q209 hot-spots in Triple-WT (typically uveal-melanoma driver); co-occur with SF3B1 R625H but not BAP1 mutations in this cutaneous cohort.
• SF3B1 — R625H hot-spot, co-occurring with GNAQ/GNA11 in Triple-WT.
• CTNNB1 — recurrent COSMIC mutations (n = 3) in Triple-WT.
• EZH2 — Y641 hot-spot in Triple-WT (n = 1).
• BCL2 — highest median protein expression in Triple-WT (RPPA); rationale for BH3 mimetics.
• LCK — high RPPA expression in Immune subclass tumors; independent favorable prognostic marker.
• SYK — high RPPA expression in Immune subclass but NOT prognostic — suggests T-cell (LCK) rather than B-cell (SYK) signaling drives the survival benefit.
• RAF1 — recurrent fusion partner (TRAK1-RAF1, RAF1-AGGF1, CLCN6-RAF1) preserving intact kinase domain.
• HMGA2 — 3 recurrent fusions (PCBP2-HMGA2, TSFM-HMGA2, SENP1-HMGA2).
• MAPK1 / MAPK3 — phospho-ERK1/2 (T202/Y204) highest in RAS subtype by RPPA.
• IGFBP2 — elevated in BRAF hot-spot mutants by RPPA.
• DDX3X — novel candidate SMG (RNA helicase), previously unreported in melanoma.
• MRPS31, RPS27 — ribosomal-protein SMGs with UV-induced 5′ UTR hot-spots (~5% and ~9%); RPS27 5′ TOP element regulates translation downstream of PI(3)K/AKT/mTOR.

Clinical implications

• Therapy stratification by subtype:
  • BRAF: targetable with FDA-approved vemurafenib, dabrafenib, trametinib.
  • RAS: combinatorial MAPK + PPP6C-context Aurora kinase inhibition (Gold et al., 2014).
  • NF1: MEK and/or ERK inhibitors (extending trametinib to BRAF-wt NF1-mutant patients); synthetic lethal targeting of chromatin modifiers in co-occurring ARID2-mutant tumors.
  • Triple-WT: imatinib / dasatinib for KIT mutation/amplification (Carvajal et al., 2011; Hodi et al., 2008); combination with sorafenib / crenolanib / regorafenib / pazopanib to target co-amplified PDGFRA / KDR; nutlin-3 / amg-232 / BH3-mimetics for MDM2-amplified, BCL2-high tumors.
• PI3K/AKT/mTOR combination: AKT3 amplification/overexpression in RAS/NF1/Triple-WT and PIK3CA / AKT1 hot-spots in BRAF/RAS subtypes support combination MEK + PI3K/AKT/mTOR trials.
• Immune checkpoint blockade biomarkers: CD274 (PD-L1) and PDCD1 (PD-1) expression highest in the Immune transcriptomic subclass; CD274 focal amplifications observed in the BRAF subtype. The authors caveat that the data do not prove the immune subtype is uniquely responsive to pembrolizumab/nivolumab/ipilimumab, but the parallels with Tumeh et al. (2014) and existing aldesleukin/interferon-alpha data are emphasized.
• Prognostic biomarker for stage III (regional metastasis): a bivariate model of pathology-derived LScore + RPPA LCK expression predicts survival (HR = 5.5 for both-high vs both-low; log-rank p = 7.9e–5). Both features are independent in multivariable Cox regression and the authors propose integration into future AJCC staging and adjuvant-therapy decision frameworks.
• No genomic subtype × outcome correlation: BRAF/RAS/NF1/Triple-WT assignment did not predict post-accession survival — consistent with the absence of preferential anti-tumor responses by BRAF status in checkpoint trials (Ascierto et al., 2014; Robert et al., 2014).
• TERT activation in Triple-WT: targeted-therapy programs for TERT-driven tumors should consider amplification/rearrangement-based (rather than promoter-mutation-based) TERT activation in this subtype.

Limitations & open questions

• Advanced-tumor bias: specimens required sufficient mass and quality, so thick primaries and metastases dominate. Population thickness was higher than registry baselines; early-stage primaries are under-represented.
• Only two matched primary + metastasis pairs had complete multi-platform data — limits inference about clonal evolution and metastatic progression.
• Genomic subtype did not correlate with survival — the authors flag this as a key open question about what does drive outcome differences within and across subtypes.
• Functional validation gap: novel SMG DDX3X and the ribosomal SMGs MRPS31 / RPS27 are nominated but not functionally characterized as melanoma drivers.
• Immune subtype ≠ proven immunotherapy responders: the cohort lacks paired checkpoint-blockade outcomes, so the predictive (rather than prognostic) value of the Immune signature for pembrolizumab / nivolumab / ipilimumab response remains untested.
• Fusion functional validation: 224 candidate driver fusions are nominated; only kinase-domain intactness is reported. Transforming activity, kinase activation, and drug sensitivity of fusions like ATG7-BRAF and TRAK1-RAF1 await follow-up.
• Antigen / neoepitope hypothesis: whether specific mutated melanoma antigens drive the lymphocyte infiltration phenotype is unresolved (cited Robbins et al., 2013; Snyder et al., 2014).
• TERT C250T paradox: C250T was as frequent as C228T but only C228T raised TERT mRNA — the mechanistic explanation differs from GBM (difg_glass) (where both raise expression) and remains unexplained.
• Triple-WT heterogeneity: this is explicitly a residual category; only ~30% had UV signatures and a substantial fraction lacked any of the recurrent COSMIC alterations interrogated.
• Generalizability: the “keratin” subclass had worse outcomes but is partially confounded with primary-tumor origin; whether keratin expression itself is prognostic vs reflecting site of origin is unsettled.

Citations from this paper used in the wiki

• “We describe the landscape of genomic alterations in cutaneous melanomas through DNA, RNA, and protein-based analysis of 333 primary and/or metastatic melanomas from 331 patients. We establish a framework for genomic classification into one of four subtypes based on the pattern of the most prevalent significantly mutated genes: mutant BRAF, mutant RAS, mutant NF1, and Triple-WT (wild-type).” (Summary)
• “Integrative analysis reveals enrichment of KIT mutations and focal amplifications and complex structural rearrangements as a feature of the Triple-WT subtype.” (Summary)
• “We found no significant outcome correlation with genomic classification, but samples assigned a transcriptomic subclass enriched for immune gene expression associated with lymphocyte infiltrate on pathology review and high LCK protein expression, a T cell marker, were associated with improved patient survival.” (Summary)
• “The mean mutation rate was 16.8 mutations/Mb, the highest reported for any cancer type thus far analyzed by TCGA.” (Results, Identification of Significantly Mutated Genes)
• “Of the 318, 52% (n = 166) harbored BRAF somatic mutations. Of those, 145 targeted the well-documented V600 amino acid residue: V600E (n = 124), V600K (n = 18), and V600R (n = 3).” (BRAF Subtype)
• “NF1 subtype (n = 28) had the highest mutation prevalence (39 mutations/Mb, more than double that of the other three subtypes).” (NF1 Subtype)
• “Only 30% (14/46) of samples in the Triple-WT subtype harbored a UV signature, compared to 90.7% of samples with a BRAF hot-spot mutation (136/150), 93.5% with a RAS (N-H-K) hot-spot mutation (86/92), and 92.9% of the NF1 subtype (26/28).” (UV Signature)
• “TERT promoter mutations were observed in 75.0% (39/52) of BRAF, 71.9% (23/32) of RAS, and 83.3% (10/12) of NF1 subtypes but in only 6.7% (1/15) of Triple-WT (p = 8e–5).” (TERT Promoter Mutations)
• “Tumors with high LScore and high LCK expression were associated with significantly improved post-accession survival compared with those having low LScore and low LCK expression (log-rank p = 7.9–5, hazard ratio = 5.5).” (Clinical Significance of the Immune Transcriptomic Subclass)
• “Our classification supports the use of imatinib and dasatinib to treat patients with KIT-mutated/amplified cutaneous melanomas and consideration of combination therapies with sorafenib, crenolanib, regorafenib, and pazopanib to target co-amplified RTKs, PDGFRA, and KDR (VEGFR2).” (Discussion)

This page was processed by crosslinker on 2026-05-14.