Genomic analysis of metastatic cutaneous squamous cell carcinoma

Authors

Yvonne Y. Li

Glenn J. Hanna

Alvaro C. Laga

Robert I. Haddad

Jochen H. Lorch

Peter S. Hammerman

Doi

10.1158/1078-0432.CCR-14-1773

PMID: 25589618 · DOI: 10.1158/1078-0432.CCR-14-1773 · Journal: Clinical Cancer Research (2015)

TL;DR

Li and colleagues performed targeted sequencing of 504 cancer-associated genes on 29 lymph-node metastases from patients with cutaneous squamous cell carcinoma (cSCC) — a rare, FDA-orphaned, poor-prognosis subset of an otherwise curable disease. The cohort was uniformly HPV-negative and dominated by a UV mutational signature (67% C>T transitions, 87% at pyrimidine dinucleotides) with a high coding mutation rate (~33 mut/Mb across the panel). Beyond the previously known cSCC tumor suppressors TP53 (79%), CDKN2A (48%), and NOTCH1/2/4 (24–69% depending on inclusion criteria), the authors found activating mutations in the RAS/RTK/PI3K pathway (BRAF, KRAS, HRAS, EGFR, FGFR3, KIT, ERBB4, PIK3CA, MTOR) and recurrent inactivation of chromatin remodellers (CREBBP, EP300, KMT2D/MLL2, ARID2). RAS/RTK/PI3K pathway activation (excluding EGFR/ERBB4) and chromatin-remodeling mutations were independently associated with worse progression-free survival; their combination was the strongest correlate of poor outcome. The authors argue that targeted agents being developed for lung and head-and-neck SCC (MEK, mTOR, FGFR, BRAF, PI3K inhibitors) should be evaluated in metastatic cSCC across sites of origin.

Cohort & data

  • n = 29 patients with lymph-node metastatic cSCC, 26 with matched normal skin, 3 unpaired (PMID:25589618).
  • Demographics: 19 male / 10 female; median age at metastatic diagnosis 74 (range 48–92); 4 immunocompromised; 12 smokers; primary tumors predominantly head and neck, parotid gland the most frequent nodal site.
  • Median progression-free survival 37 months (range 1–130); median overall survival 60 months (range 7–155); 11/29 patients recurred (3-year recurrence-free rate 41%, vs literature 56%).
  • Dataset: cscc_dfarber_2015. BAM files submitted to dbGaP.
  • Assay: DFCI-ONCOPANEL-1 (OncoPanelv2) — targeted Illumina hybrid-capture of 504 cancer-associated genes, sequenced 100bp on HiSeq 2500. Mean tumor coverage 82× (range 25–166×); normal 69× (range 15–219×). Reads aligned to b37 with Picard/Firehose.
  • Variant calling pipeline: Mutect (SNVs/indels) with OxoG artifact filtering, Oncotator annotation against dbSNP 134 and 1000 Genomes; significance via MutSig (1.5, 2.0, and CV — most significant of the three taken). CNAs called with Nexus 7.5 and recurrence assessed via GISTIC 2.0 with arm-level peel correction; minimum segment size raised from 4 to 6.
  • HPV status independently confirmed negative by p16 IHC plus hybrid-capture sequencing of HPV E6/E7 (Pathogenica).

Key findings

  • High mutation burden consistent with UV damage. In 26 paired samples, mean 59 nonsynonymous mutations/sample (range broad); ~33 mutations per Mb of the panel (range 4–117/Mb). C>T transitions account for 67% of substitutions and 87% of those occur 3’ to a pyrimidine — the canonical UV signature. The TpC>G hypermutator signature seen in HPV-driven cancers was absent (PMID:25589618).
  • TP53 is the most recurrently mutated gene (23/29, 79%). Concordant with HPV-negative status — HPV E6-mediated TP53 degradation is not operating in this cohort (PMID:25589618).
  • CDKN2A altered in 14/29 (48%) by point mutation and homozygous loss; the 9p21 locus (CDKN2A/CDKN2B) was the most significantly recurrent GISTIC deletion, with homozygous loss in 6 samples (21%) (PMID:25589618).
  • NOTCH1/2/4 inactivation in 24% of samples when restricted to truncating or COSMIC-annotated mutations; rises to 69% if all nonsynonymous SNVs of unknown effect are included — bracketing the ~75% rate reported in primary cSCCs (PMID:25589618).
  • Activating oncogene mutations in 11/29 (38%) samples spanning 12 oncogenes across 13 samples; the events are nearly mutually exclusive across the cohort (Fig 3A). Table 2 enumerates: BRAF G464R, BRAF G469R, KRAS G12C, FGFR3 G380R (achondroplasia allele, constitutively active), KIT exon 11 E562D, HRAS G13D, EGFR S720F, ERBB4 E563K, EZH2 Y641S, MTOR S2215F, PIK3CA P471L, HGF E199K, CARD11 E24K, and CARD11 D199N (PMID:25589618).
  • Chromatin remodeling is recurrently disrupted (48% of samples). Truncating mutations: CREBBP (n=6), EP300 (n=3), KMT2D/MLL2 nonsense (n=5), ARID2 truncating (n=3, ~10%). ARID1A, ARID5B, and SMARCA4 each contribute additional inactivating events. NOTCH1/2/4 and EP300 truncations are mutually exclusive across samples (PMID:25589618).
  • Recurrent copy-number events (GISTIC q < 0.25). 25 amplified and 11 deleted gene peaks. TP63 amplified in 7/29 (24%); MYC amplified in 10 samples (one high-level); CCND1 amplified in 4 samples; high-level CDK4 and CDK6 gains in 2 samples each; EGFR is both significantly amplified (one high-level) and harbors an activating S720F. LAMA5 is the most recurrently high-level amplified oncogene (4 samples). PIK3CA and SOX2 high-level amplifications — common in lung SCC — were not observed (PMID:25589618).
  • Prognostic correlation: RAS/RTK/PI3K + chromatin remodeling = worst outcome. RAS/RTK/PI3K pathway activation (excluding EGFR/ERBB4) and chromatin-remodeling mutations each correlated independently with shorter PFS, and their combination was the strongest predictor (Fig 4). Mean PFS: RAS/RTK/PI3K-mutated (excluding EGFR/ERBB4) 12 months; non-mutated 50 months; EGFR/ERBB4-mutated samples 79 months. Cell cycle (TP53, CDKN2A) and squamous differentiation (TP63, NOTCH1/2) alterations did not correlate with prognosis (PMID:25589618).
  • Immunocompromised status trends with worse prognosis (log-rank p = 0.017, n=4 — underpowered) (PMID:25589618).
  • ARID5B and CARD11 alteration associations with PFS are hypothesis-forming; notably CARD11 activating mutation/amplification correlated with better PFS — opposite the canonical NF-kB→differentiation logic that would predict tumor suppression (PMID:25589618).

Genes & alterations

Tumor suppressors recurrently inactivated:

  • TP53 — mutated in 23/29 (79%), the dominant event; mostly UV-pattern missense plus truncations.
  • CDKN2A / CDKN2B — 9p21 the top GISTIC deletion peak; CDKN2A altered by mutation + homozygous loss in 14/29 (48%).
  • NOTCH1, NOTCH2, NOTCH4 — 24% inactivating (truncating/COSMIC) or up to 69% including missense; truncating NOTCH and EP300 events are mutually exclusive.
  • RIPK4 — squamous-differentiation regulator, mutated in 7/29 (24%), two truncating; previously reported in head-and-neck SCC.
  • SMAD4 — COSMIC mutations in 2/29 (7%); one sample with SMAD4 nonsense was the lone TP53/NOTCH/oncogene-wild-type case.
  • STK11, PTEN, NF1 — PI3K/RAS negative regulators, each inactivating events in 1–3 samples.
  • RB1, ATR, NF2 — additional cell-cycle / DNA-damage / chromatin-related tumor suppressor hits.
  • KMT2D (MLL2 in legacy nomenclature), CREBBP, EP300, ARID1A, ARID2, ARID5B, SMARCA4, SMARCB1 — chromatin remodelers truncated or hit at COSMIC sites in 48% of cohort.

Oncogenic activating events (Table 2):

  • BRAF G464R and G469R — non-V600 activating alleles; G469R reported in ~1% of BRAF-mutant melanoma.
  • KRAS G12C; HRAS G13D.
  • EGFR S720F — rare activating allele also seen in NSCLC; one separate sample carried high-level EGFR amplification.
  • FGFR3 G380R — transmembrane domain, constitutively active, achondroplasia allele.
  • KIT E562D — exon 11 (canonical GIST hotspot region).
  • ERBB4 E563K — reported in 19% of melanomas.
  • EZH2 Y641S — gain-of-function lymphoma allele.
  • MTOR S2215F — validated activating.
  • PIK3CA P471L and HGF E199K — recurrence between this cohort and a prior 11-tumor cSCC NGS cohort suggests functional role.
  • CARD11 E24K, D199N — CARD-domain mutations that disrupt autoinhibition; CARD11 activation correlated paradoxically with longer PFS.

Recurrent copy-number gains: TP63 (7/29 amplified), MYC (10/29 amplified, one high-level), CCND1 (4 samples), CDK4 and CDK6 (2 high-level each), EGFR (significant amplification peak; one high-level), NFKBIA, LAMA5 (top high-level oncogene amp, 4 samples), and passenger amplifications of FANCC and SDHB. High-level PIK3CA and SOX2 amplifications — common in lung SCC — were absent.

Discussed but not assayed in this cohort: KNSTRN (recently described cSCC driver), STK19 and additional SCC-associated genes (ASCL4, FOXP1) outside the panel; VEGFA cited from mouse lymphatic-metastasis models.

Clinical implications

  • Biomarker-driven trial design. The authors argue cSCC patients with activating events in MEK-dependent (RAS/RAF), mTOR-dependent (PI3K/AKT/MTOR/STK11/PTEN), FGFR (FGFR3 G380R), and BRAF (non-V600 G469R) pathways are candidates for trials of agents already in development for lung and head-and-neck SCC, and that patients with SCCs from multiple sites of origin should be co-enrolled given the convergent biology (PMID:25589618). This parallels the lung-SCC landscape (PMID:22960745) and the esophageal-SCC landscape (PMID:24686850).
  • EGFR therapy in cSCC remains underspecified. Two of 29 samples had potential EGFR pathway activation (one S720F, one high-level amplification). The single phase II of gefitinib in cSCC (Lewis 2012) reported an 18% CR rate — the authors interpret their data as supporting further EGFR biomarker refinement rather than uniform EGFR targeting (PMID:25589618).
  • Prognostic stratification within metastatic cSCC. Two-axis stratification — RAS/RTK/PI3K activation (excluding EGFR/ERBB4) and chromatin-remodeling inactivation — separates patients with mean PFS of 12 months from those with ~50–79 months. The authors stop short of proposing a validated risk score and emphasize that no validation cohort exists (PMID:25589618).
  • Mechanistic rationale for combinations. Convergence on MEK and mTOR downstream of diverse upstream activators motivates evaluating MEK/mTOR/FGFR/BRAF/PI3K inhibitors (and combinations such as BRAF+MEK that block proliferation in mouse cSCC models) in metastatic cSCC (PMID:25589618).

Limitations & open questions

  • Small, single-institution cohort (n=29). Underpowered for single-gene survival association; the authors explicitly avoid Bonferroni-passing claims at the single-gene level and frame the RAS/RTK/PI3K + chromatin signal as exploratory and unvalidated.
  • No validation cohort. The PFS associations are hypothesis-generating; the authors call this out repeatedly and recommend external replication before clinical use.
  • Targeted panel coverage. 504 cancer-associated genes — non-coding regions, structural variants outside designed intronic baits, methylation, and transcriptome are not captured. Known SCC-relevant genes outside the panel (e.g., ASCL4, FOXP1, STK19, KNSTRN) cannot be assessed.
  • FFPE DNA quality. Variable sequencing coverage (25–166× tumor) likely limited sensitivity in lower-coverage samples; the authors note this is a possible source of undetected events.
  • Paired vs unpaired samples. 3 of 29 lacked matched normal; these had ~2× the SNV count of paired samples, indicating residual germline contamination in the unpaired set.
  • Chromatin remodeling functional interpretation. Mutations annotated as “likely functional” rely on COSMIC recurrence and truncation status — for missense variants in chromatin genes (e.g., EZH2, SMARCB1), functional validation is implied rather than demonstrated.
  • Open question — site-of-origin generalization. Whether mixing cSCC, head-and-neck SCC, and lung SCC in pathway-stratified trials produces convergent benefit, or whether tissue-of-origin context (e.g., differential UV burden, immune microenvironment) demands separate strata, is unresolved.
  • Open question — primary vs metastatic biology. Activating mutations the authors found (BRAF G469R, FGFR3 G380R, MTOR S2215F, etc.) were not seen in prior primary-tumor cohorts including the contemporaneous aggressive-primary exome cohort (PMID:25303977). It remains unclear whether these alterations are enriched specifically at the metastatic stage or simply rare events that prior cohorts missed.
  • Immunocompromised subgroup. Strong trend toward worse outcome (p=0.017) in n=4 patients is underpowered; whether immunosuppression-driven cSCC has a distinct genomic profile is unresolved.

Citations from this paper used in the wiki

  • “We performed targeted sequencing of 504 cancer-associated genes on lymph node metastases in 29 patients with cSCC and identified mutations and somatic copy number alterations associated with metastatic cSCC.” (Abstract).
  • “TP53 was mutated in 23 of 29 (79%) samples and CDKN2A was altered by both mutation and homozygous loss in 14 samples (48%).” (Results, Overview of SNV alterations).
  • “NOTCH genes showed inactivating mutations in seven samples (24%) but if we included nonsynonymous SNVs of unknown functional significance, the rate increased to 69%.” (Results).
  • “There were known gain-of-function oncogene mutations in 11 of 29 samples (38%), though recurrent events were rare in our cohort.” (Results).
  • “Half (48%) of the samples had truncating or COSMIC mutations in one or more chromatin remodeling genes.” (Results).
  • “The average PFS for RAS/RTK/PI3K pathway-mutated samples without EGFR/ERBB4 mutations was 12 months, for non-RAS/RTK/PI3K pathway-mutated samples was 50 months, and for the EGFR/ERBB4 samples was 79 months.” (Genomic or clinical factors correlated to prognosis).
  • “67% of the mutation spectrum… 87% of the C>T transitions occurred after a pyrimidine.” (Results, Landscape of genomic alterations).
  • “Samples were sequenced using the OncoPanelv2 platform… 504 genes with relevance to cancer.” (Methods).
  • Table 2 — enumeration of activating mutations: BRAF G464R/G469R, KRAS G12C, FGFR3 G380R, KIT E562D, HRAS G13D, EGFR S720F, ERBB4 E563K, EZH2 Y641S, MTOR S2215F, PIK3CA P471L, HGF E199K, CARD11 E24K/D199N.

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