Comprehensive genomic characterization of head and neck squamous cell carcinomas

Author

The Cancer Genome Atlas Network

Doi

10.1038/nature14129

PMID: 25631445 · DOI: 10.1038/nature14129 · Journal: Nature (2015)

TL;DR

The Cancer Genome Atlas Network profiled 279 head and neck squamous cell carcinomas (HNSC) with whole-exome sequencing, low-pass whole-genome sequencing, RNA-seq, miRNA-seq, DNA methylation, RPPA, and SNP6.0 arrays. The study defined two largely non-overlapping disease biologies: human-papillomavirus-positive (HPV+, n=36) tumours dominated by helical-domain PIK3CA mutations, loss of TRAF3, and amplification of E2F1; and smoking-related HPV-negative tumours with near-universal TP53 loss-of-function, CDKN2A inactivation, and recurrent amplifications at 3q26/28 and 11q13/22. A favourable-outcome subgroup of mostly oral-cavity tumours combined infrequent CNAs with activating HRAS or PIK3CA mutations and inactivating CASP8, NOTCH1, and TP53 mutations. Additional subgroups featured loss-of-function alterations of the chromatin modifier NSD1, WNT-pathway genes AJUBA and FAT1, and activation of oxidative-stress factor NFE2L2 (mainly in laryngeal tumours). Candidate therapeutic alterations were identified in most HNSCCs.

Cohort & data

  • n=279 head and neck squamous cell carcinomas with complete multi-platform data, drawn from a larger TCGA target of 500 tumours.
  • Anatomic sites: oral cavity 172/279 (62%) — OCSC; oropharynx 33/279 (12%) — OPHSC; larynx 72/279 (26%) — LXSC; plus rarer HPHSC (hypopharynx) cases.
  • Patients predominantly male (203/279, 73%) and heavy smokers (mean pack-years = 51).
  • HPV status by RNA-seq read mapping to viral genes E6/E7 (threshold >1,000 mapped reads): 36 HPV(+), 243 HPV(−). 64% of oropharyngeal tumours were HPV(+) vs 6% of non-oropharyngeal tumours.
  • Cancer type: HNSC.
  • Dataset / cohort: hnsc_tcga_pub (TCGA HNSC 2015 publication).
  • Assays: whole-exome-seq (mean coverage 95×; 82% of target bases ≥30×); high-coverage whole-genome-seq on n=29; rna-seq; mirna-seq; affymetrix-snp6 (CNAs); hm450-methylation-array; rppa.
  • Analytic tools: mutsig (MutSigCV) for significantly mutated genes; gistic for focal CNAs; targeted re-sequencing of 394 unique regions validated 99% of mutations.

Key findings

  • Genomic instability. Mean 141 CNAs (amplifications or deletions) per tumour from microarray data and 62 structural aberrations per tumour by WGS (n=29). 39 regions of recurrent copy-number loss and 23 of recurrent copy-number gain (q < 0.1).
  • Squamous-cancer chromosomal pattern. Recurrent losses of 3p and 8p, gains of 3q, 5p, and 8q — resembling lung squamous cell carcinoma (LUSC). Focal 3q26/28 amplifications (in both HPV(+) and HPV(−) tumours) span the squamous lineage transcription factors TP63 and SOX2 and the oncogene PIK3CA.
  • HPV(+)-specific alterations. Novel recurrent deletions (5/36, 14%) and truncating mutations (3/36, 8%) of TRAF3 — the first evidence linking TRAF3 inactivation to HPV-associated carcinomas. Focal amplification of E2F1. Intact 9p21.3 (CDKN2A region) that is commonly deleted in HPV(−) tumours. HPV(+) tumours showed clear evidence of host-genome integration (usually a single locus per sample, almost always with amplification of the host genome); however, integration target genes were non-recurrent (no MYC integrations as previously reported in HPV(+) cell lines).
  • HPV(−)-specific alterations. Co-amplifications of 11q13 (CCND1, FADD, CTTN) and 11q22 (BIRC2, YAP1) — implicating cell-death/NF-κB and Hippo pathways. Novel focal deletions in NSD1 and tumour-suppressors (FAT1, NOTCH1, SMAD4, CDKN2A). Recurrent focal RTK amplifications (EGFR, ERBB2, FGFR1).
  • Mutation rates. 12,159 synonymous and 37,061 non-synonymous somatic variants across 279 samples. Mutation rates did not differ by HPV status, but transversions at CpG sites were more frequent in HPV(−) tumours and a TpC-mutation predominance (APOBEC-like signature) was noted in HPV(+) cases.
  • Significantly mutated genes (MutSigCV, q < 0.1). Eleven genes statistically significant: TP53 72%, CDKN2A 22%, FAT1 23%, PIK3CA 21%, NOTCH1 19%, KMT2D (MLL2) 18%, NSD1 10%, CASP8 9%, AJUBA 6%, NFE2L2 6%, HLA-A 3%. Trend-significant (q < 1): HRAS 4%, FBXW7 5%, RB1 3%, PIK3R1 1%, TRAF3 1%, CUL3 4%, PTEN 2%, TGFBR2 4%.
  • TP53 by HPV status. TP53 mutated in 86% of HPV(−) samples (higher than previously reported); only 1/36 HPV(+) cases had a non-synonymous TP53 mutation.
  • PIK3CA hotspots. 73% of PIK3CA mutations localised to E542K, E545K, and H1047R/L hotspots. ~25% of mutated PIK3CA cases had concurrent amplification; an additional 20% of tumours had focal amplification without mutation. PIK3CA mutation frequency was 56% in HPV(+) vs 34% in HPV(−) tumours, and HPV(+) PIK3CA mutations were enriched in the helical domain (exon 9).
  • HRAS hotspots. Recurrent activating mutations at GTPase-domain residues 11–13 (q ≈ 0.2).
  • CNA cluster (“M class”) subgroup. Unsupervised clustering of CNAs identified a mutually exclusive subset of predominantly oral-cavity tumours with reduced CNAs (mutation-driven, “M class”). This subset contained a novel three-gene pattern of activating HRAS mutations co-occurring with inactivating CASP8 mutations on a wild-type TP53 background — confirming a previously reported favourable clinical outcome in HNSCCs with few CNAs.
  • Structural variants. Known fusion oncogenes involving ALK/ROS/RET were not observed. Previously reported FGFR3–TACC3 fusions detected in 2 HPV(+) tumours. EGFRvIII (type III isoform) in only 1/279 patients. Alternative MET exon-14 skipping transcript in 2 HPV(−) tumours. Structural alterations (homozygous deletions, intra-/inter-chromosomal fusions) more commonly drove loss of function in tumour-suppressors (CDKN2A, TP53, RB1, NOTCH1, FAT1) than protein-coding fusions.
  • NSD1 inactivation in 33 HNSCCs. Inactivating mutations (n=29) and focal homozygous deletions (n=4). NSD1 is a histone H3K36 methyltransferase, similar to SETD2 (mutated in ccRCC); germline NSD1 inactivation causes Sotos syndrome with predisposition to squamous carcinoma. NSD1 loss associated with DNA hypomethylation.
  • Four mRNA expression subtypes confirmed (atypical 24%, mesenchymal 27%, basal 31%, classical 18%) — replicating prior HNSCC classifications. Co-occurring features per subtype:
    • Classical — TP53 mutation, CDKN2A loss-of-function, chromosome 3q amplification, oxidative-stress gene alterations (KEAP1, NFE2L2, CUL3), heavy smoking, larynx sub-site — resembling LUSC.
    • Basal — NOTCH1 inactivation with intact oxidative-stress signalling and few 3q alterations; marked relative decrease of SOX2 expression; enriched for HRAS–CASP8 co-mutation and co-amplified 11q13/q22 tumours.
    • Atypical — lack of chromosome 7 amplifications; enriched in HPV(+) tumours with PIK3CA helical-domain mutations; hypomethylation and LOF mutations of NSD1 (with wild-type NOTCH1, shared with classical subtype).
    • Mesenchymal — high levels of innate-immunity gene alteration; high NK-cell marker CD56; low frequency of HLA class I mutations.
  • Anatomic-site differential mutations. TP53 (P<0.001), CASP8 (P=0.01), NSD1 (P=0.01), and CDKN2A (P=0.06) were most differentially mutated across sites. 22/24 CASP8 mutations (92%) occurred in oral-cavity tumours; TP53, NSD1, and CDKN2A had reduced mutation rates in oropharyngeal tumours.
  • NFE2L2 oxidative-stress axis as a tobacco signature. Alterations of NFE2L2, KEAP1, and CUL3, with heavy smoking history and larynx sub-site, co-occurred mainly in classical-subtype tumours — paralleling LUSC.
  • miRNA / copy-number networks. let-7c-5p and miR-100-5p were decreased in tumours vs normal; their deletion strongly associated with increased expression of target oncogenes CDK6, E2F1, PLK1, and HMGA2.
  • Survival stratification. HPV(+) tumours and HPV(−), TP53 wild-type tumours both demonstrated favourable outcomes compared with TP53-mutant and 11q13/CCND1-amplified tumours.

Genes & alterations

  • TP53 — mutated in 72% overall (86% in HPV(−), 1/36 in HPV(+)); predominantly loss-of-function. Distinguishes HPV(−) from HPV(+) biology; central to classical subtype.
  • CDKN2A — inactivation in 22% (58% in HPV(−), 0% in HPV(+) on pathway analysis); co-occurs with TP53 loss in smoking-related tumours; falls in significant deletion peak at 9p21.
  • PIK3CA — mutated in 21% (56% in HPV(+), 34% in HPV(−)); 73% of mutations at hotspots E542K/E545K/H1047R/L; helical-domain enriched in HPV(+); often co-amplified.
  • TRAF3 — recurrent deletions (14%) and truncating mutations (8%) in HPV(+); first link to HPV-associated cancer; loss promotes aberrant NF-κB signalling.
  • E2F1 — focal amplification predominantly in HPV(+); also increased expression via let-7c-5p/miR-100-5p loss.
  • HRAS — activating mutations (4%) at GTPase residues 11–13; defines an M-class oral-cavity subset with CASP8 co-mutation and wild-type TP53.
  • CASP8 — clustered missense and inactivating mutations (9% overall; 92% of mutations in oral cavity); anti-correlated with 11q13 amplification.
  • NOTCH1 — inactivating mutations in 19%; NOTCH2 (9%) and NOTCH3 (5%) non-significant; convergent with FAT1/AJUBA on β-catenin signalling.
  • AJUBA — previously unreported somatic mutations and deletions (6%); primarily 5′-inactivating events plus clustered missense in the LIM domain; regulates Hippo/Wnt/β-catenin.
  • FAT1 — mutated in 23%; novel SMG in HNSCC; aberrant Wnt activation.
  • NSD1 — inactivating mutations (n=29) and focal homozygous deletions (n=4) in 33 tumours (10%); H3K36 methyltransferase; loss associated with DNA hypomethylation; defines atypical/classical mRNA-subtype enrichment.
  • NFE2L2 — significantly mutated in 6%; activating; tobacco-associated; co-occurs with KEAP1 and CUL3 alterations in classical subtype.
  • KEAP1, CUL3 — oxidative-stress complex partners of NFE2L2; tobacco signature.
  • CCND1, FADD, CTTN — 11q13 co-amplification cassette (31% in HPV(−); 3% in HPV(+)).
  • BIRC2, YAP1 — 11q22 co-amplification (rarely amplified without 11q13); selection inferred from BIRC2–FADD–caspase cascade interaction inhibiting cell death; majority of 11q13-amplified tumours had large telomeric 11q22 deletions including ATM, CASP1, CASP4, CASP5, CASP12.
  • KMT2D (MLL2) — significantly mutated (18%); contributes to defective immunosurveillance.
  • HLA-A — significantly mutated (3%); immune-evasion role.
  • TGFBR2 — sporadic inactivating mutations and deletions; primarily oral-cavity tumours.
  • FBXW7 — recurrent Arg505Gly/Leu substitutions (n=14); falls in 4q31.3 significant deletion peak; targets cyclin E and NOTCH.
  • PIK3R1 — mutated in 1%; one of multiple genes with at least one identical COSMIC mutation.
  • PTEN — mutated/altered in 2% overall; component of RTK/RAS/PI(3)K alteration in 62% of HPV(−) and 61% of HPV(+) tumours.
  • RB1 — mutations in 3%; structural alterations more often inactivating than coding fusions.
  • EGFR — focal amplification in 15% of HPV(−) vs 6% of HPV(+); rare EGFRvIII isoform (1/279).
  • ERBB2 — focal amplification in HPV(−) (3%).
  • FGFR1 — focal amplification in HPV(−) (10%) vs 0% HPV(+).
  • FGFR3 — amplification in 2% HPV(−) vs 11% HPV(+); FGFR3–TACC3 fusions in 2 HPV(+) tumours.
  • FGFR2, IGF1R, DDR2, EPHA2, MET — additional candidate RTK alterations across HPV groups (Fig. 3).
  • MYC — amplification in 14% HPV(−) and 3% HPV(+); not the target of HPV integration in this cohort.
  • SOX2, TP63 — 3q26/28 co-amplified squamous-lineage transcription factors; TP63 RNA splice variants identified.
  • ZNF750 — TP63 target; mutation in 4% within significant 17q25.3 deletion peak.
  • SMAD4 — focal deletions in HPV(−) tumour-suppressor set.
  • NF1 — alteration in 3% HPV(−).
  • CTNNB1 — terminal node of Wnt convergence implicated via FAT1/AJUBA/NOTCH inactivation.
  • B2M — implicated in HLA-A/B antigen presentation pathway alterations (7% HPV(−), 11% HPV(+)).
  • PTCH1, CHEK2, SCN9A — genes with at least one identical previously COSMIC-reported mutation.

Clinical implications

  • Therapeutic candidate alterations identified in most HNSCCs (Fig. 3 and pathway analyses).
  • HPV(+) oropharyngeal cancers — aberrant NF-κB activation via TRAF3 loss, PIK3CA activating mutations, and E2F1 amplification suggest opportunities for PI(3)K-pathway and cell-cycle-targeted therapy.
  • HPV(−) oncogenic axes — mutually exclusive subsets defined by 11q13/CCND1/FADD/BIRC2/YAP1 co-amplification or by CASP8–HRAS co-mutation; both target cell cycle, cell death, NF-κB, and Hippo pathways and motivate BIRC-, PI(3)K-, Wnt/β-catenin-, and NOTCH-targeted strategies (under investigation).
  • Favourable outcomes in HPV(+) and HPV(−)/TP53 wild-type tumours vs poorer outcomes in TP53-mutant and 11q13/CCND1-amplified tumours — supporting TP53 status and 11q13 amplification as candidate prognostic biomarkers beyond HPV.
  • Oral-cavity M-class subset (HRAS + CASP8 + wild-type TP53, few CNAs) — favourable outcome subgroup that may warrant de-escalated therapy.
  • Wnt/β-catenin axis — combined inactivation of AJUBA, FAT1, and NOTCH1 predicted to converge on unchecked β-catenin signalling (potential therapeutic target).
  • Tobacco-associated oxidative-stress axis (NFE2L2, KEAP1, CUL3) — highlights this complex as a therapeutic target in laryngeal/classical-subtype HNSCC, paralleling LUSC.
  • RTK targeting — EGFR/ERBB2 or FGFR1/FGFR3 alterations are the most frequent among RTKs; HPV(−) tumours enriched for EGFR and FGFR1 amplification; HPV(+) tumours enriched for FGFR3 alterations (including FGFR3–TACC3 fusions).
  • Immune escape — significantly mutated HLA-A and KMT2D plus B2M/HLA-B pathway alterations may contribute to defective immunosurveillance, with implications for immune-checkpoint strategies.

Limitations & open questions

  • Cohort reflects an early TCGA freeze (279 of an a-priori 500-tumour target); statistical power for rare alterations remains limited.
  • HPV(+) subset is modest (n=36), limiting power for HPV(+)-specific significantly mutated genes and structural variants.
  • High-coverage whole-genome sequencing was performed in only 29 tumours; structural-variant detection sensitivity in the remainder relies on RNA-seq and SNP-array signals.
  • HPV-integration target genes were non-recurrent, suggesting no single driver mechanism related to HPV integration — biological consequences of host-locus amplification at integration sites remain to be defined.
  • Mechanistic significance of co-amplification of BIRC2/YAP1 at 11q22 with 11q13 in the context of telomeric 11q22 deletions (which remove ATM, CASP1, CASP4, CASP5, CASP12) needs functional dissection.
  • Whether candidate therapeutic alterations identified across most tumours translate to clinical benefit requires prospective trials targeting the specific HPV+/HPV− and CNA-class subsets defined here.
  • Biological roles of agents targeting BIRCs, PI(3)K, Wnt/β-catenin, and NOTCH in HNSCC are described as “under investigation” by the authors.

Citations from this paper used in the wiki

  • “The Cancer Genome Atlas profiled 279 head and neck squamous cell carcinomas (HNSCCs) to provide a comprehensive landscape of somatic genomic alterations.” (Abstract)
  • “Human-papillomavirus-associated tumours are dominated by helical domain mutations of the oncogene PIK3CA, novel alterations involving loss of TRAF3, and amplification of the cell cycle gene E2F1.” (Abstract)
  • “Smoking-related HNSCCs demonstrate near universal loss-of-function TP53 mutations and CDKN2A inactivation with frequent copy number alterations including amplification of 3q26/28 and 11q13/22.” (Abstract)
  • “A subgroup of oral cavity tumours with favourable clinical outcomes displayed infrequent copy number alterations in conjunction with activating mutations of HRAS or PIK3CA, coupled with inactivating mutations of CASP8, NOTCH1 and TP53.” (Abstract)
  • “HPV(+) tumours were distinguished by novel recurrent deletions (n=5 out of 36, 14%) and truncating mutations (n=3 out of 36, 8%) of TNF receptor-associated factor 3 (TRAF3).” (p. 576)
  • “We observed TP53 mutation among HPV(−) samples at higher rates (86%) than have been previously reported, while only 1 out of 36 HPV(+) cases had a non-synonymous TP53 mutation.” (p. 578)
  • “Seventy-three percent of PIK3CA mutations localized to Glu542Lys, Glu545Lys and His1047Arg/Leu hotspots… Approximately one-quarter of the mutated PIK3CA cases displayed concurrent amplification, with an additional 20% of tumours containing focal amplification without evidence of mutation.” (p. 578)
  • “A frequently mutated novel gene, the nuclear receptor binding SET domain protein 1 (NSD1), was identified in 33 HNSCCs. Alterations included inactivating mutations (n=29) and focal homozygous deletions (n=4).” (p. 578)
  • “TP53 mutation, CDKN2A loss of function, chromosome 3q amplification, alteration of oxidative stress genes (KEAP1, NFE2L2 or CUL3), heavy smoking history and larynx sub-site co-occurred in most classical subtype tumours… similar to LUSC.” (p. 579)
  • “Co-amplification of FADD ± BIRC2, or CASP8 ± HRAS mutations define exclusive HPV(−) subsets, whereas TRAF3 loss characterizes an HPV(+) subset.” (p. 580)
  • “Loss of TRAF3, activating mutations of PIK3CA, and amplification of E2F1 in HPV(+) oropharyngeal cancers point to aberrant activation of NF-κB, other oncogenic pathways, and cell cycle, as critical in the pathogenesis and development of new targeted therapies for these tumours.” (p. 581)

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