Exome sequencing of desmoplastic melanoma identifies recurrent NFKBIE promoter mutations and diverse activating mutations in the MAPK pathway

PMID: 26343386 · DOI: 10.1038/ng.3382 · Journal: Nature Genetics (2015)

TL;DR

Shain et al. performed low-coverage whole-genome (13×) and high-coverage exome sequencing (89×) on a discovery cohort of 20 fresh-frozen desmoplastic melanomas, followed by targeted sequencing of 293 genes (216×) on a validation cohort of 42 FFPE tumors (62 total). Desmoplastic melanoma carries an exceptionally high mutation burden (median 62 mutations/Mb) with a strong UV-radiation C>T signature, yet is conspicuously lacking the canonical melanoma drivers BRAF V600E and NRAS Q61K/R. Instead, MAPK/PI3K activation is achieved through a diverse set of alterations in NF1, CBL, ERBB2, MAP2K1, MAP3K1, BRAF (non-V600), EGFR, PTPN11, MET, RAC1, SOS2, NRAS Q61H, and PIK3CA — affecting 73% of tumors. The paper identifies novel recurrent promoter mutations of NFKBIE in 14.5% of samples, with evidence for biallelic selection.

Cohort & data

• 62 desmoplastic melanomas total: 20 fresh-frozen tumors with matched normal in the discovery cohort (10 from MSKCC, 10 from Melanoma Institute Australia); 42 FFPE primary desmoplastic melanomas in the validation cohort (UCSF dermatopathology repository); 4 of the 42 validation samples lacked non-lesional normal tissue.
• Cancer type: DESM (desmoplastic melanoma), an OncoTree subtype of MEL; both pure and mixed subtypes included and found to be genetically similar.
• Dataset: desm_broad_2015 (cBioPortal studyId).
• Assays:
  • Discovery cohort: low-coverage whole-genome sequencing (whole-genome-seq, 13×) + high-coverage exome sequencing (whole-exome-seq, 89×) using Agilent SureSelect Exome V4 + UTR.
  • Validation cohort: targeted-dna-seq of 293 genes at 216× using Nimblegen SeqCap EZ Libraries.
  • Copy number: high-resolution array-cgh-agilent-1m on 180K/244K/1M-feature arrays in 44 samples; CNVkit on sequencing data in all 62.
  • Sequencing: Illumina HiSeq 2500. Pipeline: bwa → GATK INDEL realignment/deduplication/recalibration → mutect for point mutations → Unified Genotyper and gatk-somatic-indel-detector for INDELs → oncotator annotation.
  • Validation: sanger-sequencing and deep PCR amplicon resequencing (>900×) confirmed 329/336 mutations (97.9%).
• Data deposition: raw aCGH copy number data in GEO under accession GSE55150.

Key findings

• Extreme mutation burden: median 62 mutations/Mb across 62 tumors, placing desmoplastic melanoma among the most highly mutated cancers — versus ~15 mutations/Mb for conventional cutaneous melanoma and ~2 for most solid tumors (PMID:26343386).
• UV-radiation mutational signature: 88% of mutations were C>T transitions, favoring di-pyrimidine contexts, implicating UV-radiation as the dominant mutagen and pointing to a superficially located cell of origin near the epidermis.
• Age effect: older patients (>55 years) carried significantly more mutations than younger patients (p = 2×10⁻³, t-test); the two tumors from sun-shielded sites had the lowest mutation burdens, and one carried a germline CDKN2A mutation.
• Conspicuously absent canonical melanoma drivers: BRAF V600E and NRAS Q61K/R were completely absent. One tumor had an atypical NRAS Q61H; three had kinase-impaired BRAF G466E, BRAF G469E, BRAF D594N substitutions that paradoxically activate MAPK via CRAF.
• Other oncogenic hotspots observed: ERBB2 S310F (n=4); MAP2K1 P124S/L, PTPN11 E76A/K, PPP6C R264C, RAC1 P29S (n=2 each); EZH2 Y641S, IDH1 R132C, PIK3CA E542K, NRAS Q61H, BRAF G469E, BRAF G466E, BRAF D594N (n=1 each).
• TERT promoter mutations in 85% (17/20) of samples where the locus could be sequenced; overall TERT activation in 90% (promoter mutations plus amplifications).
• Novel recurrent NFKBIE promoter hotspot: the single most recurrent mutational hotspot in the cohort, observed 9 times at one site plus 2 more at a nearby site, totaling 20 clustered mutations in 15 tumors (14.5% of samples). 5/15 tumors had two NFKBIE mutations affecting opposing alleles, and 4 of the 10 single-mutation tumors had undergone LOH at the mutant allele — strong evidence for selection for biallelic inactivation.
• NFKBIE isoform biology: the hotspot resides in the coding region of the long isoform but the promoter region of the short isoform; the short isoform is ubiquitously expressed (melanoma TCGA RNA-Seq, Illumina Human Bodymap 2.0), the long isoform is brain-restricted. The hotspot is highly conserved and predicted to disrupt binding sites for 32 transcription factors including GABPA and ELF1.
• NFκB signaling consequence: two non-desmoplastic melanoma cell lines with NFKBIE hotspot mutations (M257, M375) lacked nuclear NFκB translocation, consistent with a proposed gain-of-function for NFKBIE promoter mutations that inactivates downstream NFκB signaling.
• NFKBIE mutations beyond desmoplastic melanoma: an extension cohort identified 6 non-desmoplastic, UV-driven, BRAF/NRAS-wild-type melanomas with NFKBIE mutations — not found in COSMIC or TCGA studies of any cancer.
• Focal copy number alterations (despite overall low CNA burden):
  • Amplifications: EGFR, CDK4, MDM2, TERT, MAP3K1 (3 cases each); MET, YAP1, NFKBIE (2 each); CCND1, MYC, SOS2 (1 each). IHC confirmed protein-level upregulation for EGFR, CDK4, MDM2, MET, YAP1, CCND1.
  • Deletions: CDKN2A (11 cases) and NF1 (4 cases). p16 (CDKN2A) loss confirmed by IHC.
• LOF-burden tumor suppressor candidates (using the lofsigrank method; FDR 0.5–1.0 cutoffs): TP53, CDKN2A, NF1, ARID2, FSIP1, CBL, FBXW7, CDH2, IL36A, PAK3, RB1, ARID1A. IHC confirmed protein-level loss for representative TP53, CDKN2A, RB1, ARID1A mutants.
• CBL harbored frequent truncating and damaging missense mutations with no synonymous mutations, supporting a tumor-suppressor role; CBL targets RTKs for degradation and is implicated in Noonan syndrome alongside PTPN11, NRAS, KRAS, BRAF.
• MAP3K1 was focally amplified in 3 tumors — a novel melanoma alteration consistent with prior Sleeping Beauty screens identifying MAP3K1 as a BRAF-substituting oncogene in mouse melanoma.
• FBXW7 truncating or WD-domain damaging mutations affected 11% of tumors with no synonymous/conservative-missense mutations.
• RTK→Ras→MAPK / PI3K pathway activation in 73% of tumors, with alterations not mutually exclusive — suggesting cooperation. Phospho-ERK by IHC was uniformly positive across tumors.
• Other recurrently affected pathways: p53/Rb (via CDKN2A, TP53, RB1 inactivation) and SWI/SNF chromatin remodeling (ARID2, ARID1A inactivation).

Genes & alterations

• NFKBIE — novel recurrent promoter hotspot mutations in 14.5% of samples (9 at the primary site, 2 at a nearby secondary site); plus focal amplification in 2 tumors. Biallelic selection demonstrated. Loss of NFκB nuclear translocation in mutant cell lines, suggesting NFKBIE promoter mutation is a gain-of-function event that suppresses NFκB signaling.
• BRAF — V600E absent; three tumors with kinase-impaired G466E, G469E, D594N substitutions that paradoxically activate MAPK via CRAF.
• NRAS — canonical Q61K/R absent; one tumor with atypical Q61H.
• NF1 — loss-of-function mutations and focal deletions in 4 cases.
• CBL — frequent truncating/damaging missense mutations, no synonymous mutations; supports tumor-suppressor role in desmoplastic melanoma.
• ERBB2 — recurrent S310F hotspot in 4 tumors.
• MAP2K1 — P124S/L hotspot in 2 tumors.
• MAP3K1 — focal amplifications in 3 tumors (novel in melanoma).
• EGFR — focal amplifications in 3 tumors, with IHC-confirmed overexpression.
• PTPN11 — E76A/K hotspot in 2 tumors.
• MET — focal amplifications in 2 tumors, IHC-confirmed.
• RAC1 — P29S hotspot in 2 tumors.
• SOS2 — focal amplification in 1 tumor.
• PIK3CA — E542K hotspot in 1 tumor.
• TP53 — top LOF-burden tumor-suppressor candidate; IHC-confirmed protein-level alteration.
• CDKN2A — focal deletions in 11 cases; LOF mutations; one germline mutation in a sun-shielded tumor; p16 loss by IHC.
• FBXW7 — truncating/WD-domain damaging mutations in 11% of tumors.
• ARID2 and ARID1A — inactivating mutations in the SWI/SNF complex; ARID1A IHC-confirmed.
• RB1 — LOF mutations; IHC-confirmed protein loss.
• TERT — promoter mutations in 85% (17/20) of evaluable samples; overall activation (promoter + amplification) in 90% of tumors.
• CDK4, MDM2, YAP1, CCND1, MYC — focal amplifications (1–3 cases each); IHC confirmed for CDK4, MDM2, YAP1, CCND1.
• EZH2 — Y641S hotspot in 1 tumor.
• IDH1 — R132C hotspot in 1 tumor.
• PPP6C — R264C hotspot in 2 tumors.
• Additional LOF-burden TS candidates: FSIP1, CDH2, IL36A, PAK3.

Clinical implications

• Genetically distinct from cutaneous melanoma. Absence of BRAF V600E and NRAS Q61K/R means desmoplastic melanomas are not candidates for BRAF V600-directed inhibitors (vemurafenib/dabrafenib + trametinib regimens) — clinically important because many of these tumors would otherwise be treated under the standard cutaneous-melanoma paradigm.
• Targetable RTK/MAPK alterations. The authors explicitly call out the existence of small-molecule inhibitors against products of MET, EGFR, ERBB2, IDH1, MAP2K1, PIK3CA, and CDK4 as potential therapeutic avenues for desmoplastic melanoma.
• Immune checkpoint blockade candidacy. The exceptionally high mutation burden (median 62 mutations/Mb) positions desmoplastic melanoma as a promising candidate for immune checkpoint inhibitor therapy, by analogy with prior work linking TMB to anti-CTLA-4 response in melanoma.
• Diagnostic biomarker potential. The novel recurrent NFKBIE promoter hotspot is not found in COSMIC or TCGA across any cancer type, giving it potential diagnostic value for distinguishing desmoplastic melanoma from histological mimics — including a small set of UV-driven, BRAF/NRAS-wild-type non-desmoplastic melanomas.
• Ontogeny inference. The strong UV signature in deep dermal tumors implies cell-of-origin in or near the epidermis with most mutations accumulated prior to dermal invasion.

Limitations & open questions

• The validation cohort (n=42) used FFPE material with low neoplastic cellularity and stromal contamination; somatic calls in the 4 normal-less samples relied on population-database filtering (1000 Genomes, ESP5400 NHLBI) which may retain rare germline variants.
• Discovery and validation are dominated by TERT promoter calls only where coverage allowed — the locus was not initially tiled for early samples; reported TERT promoter mutation frequencies are restricted to samples with 90–95% statistical power.
• The mechanism by which NFKBIE promoter mutations modulate NFκB signaling — and the functional contribution of bi-allelic selection — is asserted but not fully dissected; the authors call out the need for future mechanistic studies.
• Targetable alterations are predicted therapeutic vulnerabilities; no clinical-response data is reported in this paper.
• Immune checkpoint blockade candidacy is inferred from mutation burden alone; no neoantigen prediction, immune-infiltrate characterization, or treatment outcome data is presented.
• Cooperation among the multiple non-mutually-exclusive RTK/MAPK/PI3K alterations is observed but not functionally validated.
• The MAP3K1 amplification finding rests on 3 cases and analogy to a Sleeping Beauty mouse screen; functional validation in human desmoplastic melanoma is not provided.

Citations from this paper used in the wiki

• “A high mutation burden (median 62 mutations/Mb) ranked desmoplastic melanoma among the most highly mutated cancers.” (p. 2)
• “Commonly mutated oncogenes in melanomas, in particular BRAFV600E and NRASQ61K/R, were absent. Instead, other genetic alterations known to activate the MAPK and PI3K signaling cascades were identified in 73% of samples…” (p. 2)
• “Novel alterations included recurrent promoter mutations of NF-kappa B inhibitor epsilon, NFKBIE (IkBε), in 14.5% of samples.” (p. 2)
• “Overall, we found 20 clustered mutations in 15 tumors… five of the 15 tumors had two mutations each, and in all five cases the mutations affected opposing alleles… Collectively, these results strongly suggest selection for bi-allelic NFKBIE mutations.” (p. 4)
• “The clustered mutations are not found in COSMIC nor TCGA studies of any cancer…” (p. 4)
• “Some of the alterations found may have clinical implications. For example, there are small-molecule inhibitors directly targeting products of several oncogenes, such as MET, EGFR, ERBB2, IDH1, MAP2K1, PIK3CA, and CDK4. Furthermore, the exorbitantly high mutation burden found in desmoplastic melanomas makes them promising candidates for immune checkpoint blockade therapy.” (p. 6)

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