Exome sequencing of hepatocellular carcinomas identifies new mutational signatures and potential therapeutic targets

PMID: 25822088 · DOI: 10.1038/ng.3252 · Journal: Nature Genetics (2015)

TL;DR

Schulze et al. performed whole-exome sequencing on 243 European hepatocellular carcinomas (HCC) and matched normal liver, defining the most comprehensive mutational landscape of HCC at that time. They identified 161 putative driver genes spanning 11 recurrently altered pathways, discovered two novel mutational signatures (signatures 23 and 24, the latter linked to aflatoxin B1 / HBV exposure), and showed that 28% of tumors harbor an alteration potentially targetable by an FDA-approved drug. Mutation associations clustered around three hubs (CTNNB1 with alcohol exposure, TP53 with HBV, and AXIN1), and TERT promoter mutations emerged as the earliest event during cirrhosis-to-HCC progression while CDKN2A loss and FGF3/FGF4/FGF19/CCND1 amplification appeared late and predicted poor outcome.

Cohort & data

• 243 surgically resected liver tumors with matched non-tumor liver from France (n=193, Créteil and Bordeaux), Italy (n=41, Milan) and Spain (n=9, Barcelona); cohort registered as cBioPortal study hcc_inserm_fr_2015.
• Cancer type: HCC.
• Fibrosis distribution: F4 cirrhosis n=118; F2–F3 n=46; F0–F1 non-fibrotic n=79.
• Risk factors: alcohol 41%, HCV 26%, NASH/metabolic syndrome 18%, HBV 14%, hemochromatosis 7%, no known etiology 11%.
• Cirrhosis progression substages represented: 7 dysplastic macronodules, 7 early HCC, 17 small-and-progressed HCC, 58 classic HCC, 29 poor-prognosis HCC.
• Assay: Agilent SureSelect whole-exome capture (v2/v3/v4/v5+UTRs) sequenced on Illumina HiSeq 2000 (paired-end 75 bp) to a mean depth of 72× (~92.6% targeted bases at ≥10×).
• Mutation calling benchmarked by Sanger sequencing of 11 genes in 155 tumors: sensitivity 88% (95% CI 82–92%; 95% excluding GC-rich ARID1A exon 1 and large CTNNB1 exon 3 deletions), specificity 99%.
• Variants annotated with Alamut and Oncotator; driver discovery via MutSigCV plus a focal-CNA integration pipeline; mutagenic processes deciphered with the Wellcome Trust Sanger Institute mutational-signatures NMF framework.
• Sequence data deposited in EGA (EGAS00001000217, EGAS00001000679, EGAS00001001002) and ICGC data portal (release 18, 10 December 2014).

Key findings

• 28,478 somatic mutations were called; excluding one hypermutated tumor (CHC892T, 6,184 mutations), the median was 21 silent and 64 non-silent mutations per tumor (range 1–706), giving a mean coding mutation rate of 1.3/Mb.
• Eight mutational signatures were detected across the 243 exomes (signatures 1A, 1B, 4, 5, 6, 16, 23, 24); signatures 23 and 24 are novel. Signature 24 has a high C>A rate, was found in 5 tumors from sub-Saharan African migrants infected with HBV, and 3 of 5 carried the canonical AFB1-associated TP53 R249S mutation — strongly implicating aflatoxin B1 exposure as the mutagen behind signature 24.
• Hierarchical clustering on signature contributions defined 6 MSig groups plus 4 singletons. MSig3 (n=19) was enriched for alcohol + tobacco co-exposure (67%, P = 6×10⁻⁵), CTNNB1 mutation (74%, P = 8×10⁻⁴) and non-cirrhotic livers (79%, P = 0.02), suggesting a genotoxic alcohol-tobacco synergy underlying ~8% of HCC.
• Validation in 452 ICGC-Japan and 198 TCGA exomes recapitulated AFB1-related signature 24 in 6 TCGA tumors (all African or Asian, P = 0.002; 3 of 6 R249S-positive, P = 8×10⁻⁴).
• MutSigCV identified 14 significantly mutated genes (q < 0.05): TP53, CTNNB1, AXIN1, ALB, ARID1A, ARID2, ACVR2A, NFE2L2, RPS6KA3, KEAP1, RPL22, CDKN2A, CDKN1A, and RB1.
• Recurrent homozygous deletions: CFH locus, IRF2, CDKN2A, PTPN3, PTEN, AXIN1, RPS6KA3. Recurrent focal amplifications: TERT, VEGFA, MET, MYC, the FGF3/FGF4/FGF19/CCND1 locus, JAK3, CCNE1.
• Integrating mutations and focal CNAs yielded 161 putative driver genes mapping to 11 pathways altered in ≥5% of HCC: TERT-promoter/telomerase activation (60%), WNT/β-catenin (54%), PI3K/AKT/mTOR (51%), TP53/cell cycle (49%), MAP kinase (43%), hepatic differentiation (34%), epigenetic regulation (32%), chromatin remodeling (28%), oxidative stress (12%), IL6/JAK/STAT (9%), TGFβ (5%).
• Newly implicated HCC drivers include the β-catenin inhibitors ZNRF3, USP34, and MACF1; the hepatocyte-secreted proteins APOB and FGA; and the TGFβ receptor ACVR2A.
• Mutation co-occurrence/exclusivity analysis defined three association clusters centered on CTNNB1, AXIN1, and TP53 (Fig. 4).
• Etiology-specific enrichments (P < 0.05): alcohol-related HCC enriched for CTNNB1, TERT, CDKN2A, SMARCA2, and HGF alterations; HBV-related HCC enriched for TP53 mutation; IL6ST mutations exclusive to HCC of unknown etiology. HCV, metabolic syndrome, and hemochromatosis showed no significant gene associations.
• Progression analysis: mutation counts and chromosome-aberration counts both rose monotonically from dysplastic macronodule → poor-prognosis HCC (P = 1.2×10⁻³ and P = 1.3×10⁻⁵, Jonckheere-Terpstra). TERT promoter mutations were already frequent in dysplastic and early lesions; CTNNB1 and TP53 mutation frequencies rose with progression; CCND1/FGF focal amplifications were largely confined to poor-prognosis HCC (P < 0.01, chi-square trend). In non-fibrotic livers, TERT promoter mutations appeared later. HNF1A and IL6ST mutations were mostly restricted to hepatocellular adenomas, arguing that most HCC in non-fibrotic liver does not derive from adenoma transformation.
• Multivariate Cox analysis (n=216 R0-resected patients): CDKN2A inactivation and FGF3/FGF4/FGF19/CCND1 amplification were independently associated with poor overall survival at 60 months beyond classical clinicopathologic features.

Genes & alterations

• TP53 — significantly mutated (MutSigCV q < 0.05); a TP53/cell-cycle pathway is altered in 49% of HCC. HBV-related HCC are significantly enriched for TP53 mutation. The R249S hotspot is the AFB1 signature and tracks with the novel signature 24.
• CTNNB1 — significantly mutated; central node of the WNT/β-catenin pathway (54% altered) and of the alcohol/tobacco-driven MSig3 cluster (74% CTNNB1-mutated). Mutation frequency rises during progression.
• AXIN1 — significantly mutated and recurrently homozygously deleted; defines its own mutation-association cluster.
• ALB, APOB, FGA — hepatocyte-secreted protein genes recurrently mutated; newly highlighted in the hepatic-differentiation pathway (34% pathway alteration).
• ARID1A, ARID2, SMARCA2 — chromatin remodelers (pathway altered in 28%); SMARCA2 alteration is enriched in alcohol-related HCC.
• ACVR2A — TGFβ receptor, newly implicated as recurrently mutated HCC driver (previously seen in chondrosarcoma).
• NFE2L2 and KEAP1 — oxidative-stress pathway (altered in 12%); both significantly mutated by MutSigCV.
• RPS6KA3 — significantly mutated (7% of tumors); inactivating mutations shown in vitro to activate RAS/MAPK signaling with elevated phospho-ERK1/2, nominating RPS6KA3-mutated HCC for ERK/MEK inhibition.
• RPL22 — newly identified as significantly mutated in HCC.
• CDKN2A, CDKN1A, RB1, CCND1, CCNE1 — cell-cycle/RB axis; CDKN2A loss is enriched in alcohol-related HCC and independently associated with poor survival.
• TERT — promoter mutation and focal amplification together produce telomerase activation in 60% of HCC; TERT promoter mutation is the earliest event in cirrhotic progression.
• FGF19, FGF3, FGF4, CCND1 — co-amplified 11q13.3 locus; focal amplification (~4% FDA-targetable, 6% of HCC for FGF3/4/19) appears at advanced stages and is independently associated with poor survival. Aligns with PMID:24798001, which previously linked RB1/FGF19 to patient stratification.
• VEGFA — focal amplification in 1% (FDA-targetable).
• MET, HGF — MET amplified in 1%; HGF alteration in 3% and enriched in alcohol-related HCC.
• MYC, JAK3 — recurrently amplified.
• PTEN, MTOR, PIK3CA — components of PI3K/AKT/mTOR (altered in 51%); MTOR mutations 2%, FDA-targetable.
• BRAF, KRAS, ERBB2, EGFR, KIT, KDR, TEK, EPHA4, JAK1 — low-frequency but FDA-druggable alterations in the MAP kinase, RTK, and JAK/STAT pathways.
• IL6ST, IL6R — IL6/JAK/STAT pathway (altered in 9%); IL6ST mutations exclusively found in HCC of unknown etiology and mostly restricted to hepatocellular adenomas in non-fibrotic settings.
• HNF1A — mutated mostly in hepatocellular adenomas (HCA), not classic HCC; supports the conclusion that most HCC arising in non-fibrotic liver does not originate from HCA transformation.
• ZNRF3, USP34, MACF1 — newly highlighted β-catenin–pathway inhibitors recurrently mutated in HCC.
• CFH, IRF2, PTPN3 — recurrent homozygous deletions.
• NQO1 — not a driver, but a modifier of HSP90-inhibitor sensitivity. In 29 liver cancer cell lines, GI50 for 17-AAG / 17-DMAG was inversely correlated with NQO1 expression (Pearson r = −0.56, P = 0.0015). Two of three KEAP1-mutant lines were highly sensitive; the third (MHCC97H) was resistant because it is homozygous for the NQO1 P187S (NQO1*2, rs1800566) variant that destabilises the NQO1 protein.

Clinical implications

• Druggable landscape: 28% of HCC harbored at least one damaging alteration potentially targetable by an FDA-approved drug; 86% by a drug in phase I–III trials. FDA-targetable hits: FLT* family (6%), FGF3/4/19 (4%), PDGFR family (3%), EPHA4 (3%), JAK3 (3%), HGF (3%), MTOR (2%), VEGFA (1%), EGFR (1%), FGFR family (1%), IL6R (1%), KIT (1%), MET (1%), TEK (1%), BRAF (<1%), ERBB2 (<1%), JAK1 (<1%), KDR (<1%). Advanced-stage tumors were enriched for targetable alterations, especially FGF/CCND1 amplifications.
• RPS6KA3-mutant HCC as candidates for MEK/ERK inhibition: siRNA knockdown of RPS6KA3 in liver cancer cells increased phospho-ERK1/2, suggesting RAS/MAPK pathway activation is a vulnerability in the ~7% of HCC with RPS6KA3 loss-of-function.
• HSP90 inhibitors (tanespimycin / alvespimycin) as a candidate strategy for KEAP1/NFE2L2-mutated (oxidative-stress) HCC, modulated by NQO1: NQO1 reduces 17-AAG to a more potent inhibitor; tumors with KEAP1 or NFE2L2 mutations have elevated NQO1 and may be more 17-AAG sensitive, except in patients homozygous for the NQO1 P187S (rs1800566) variant. This nominates NQO1 expression + NQO1*2 genotype as a combined predictive biomarker.
• Prognostic biomarkers: CDKN2A inactivation and FGF3/FGF4/FGF19/CCND1 amplification are independent predictors of poor overall survival at 60 months in R0-resected HCC, controlling for classical clinical and histological features.
• Etiology-driven trial design: alcohol-related HCC (CTNNB1/TERT/CDKN2A/SMARCA2/HGF-enriched), HBV/AFB1-related HCC (TP53 R249S, signature 24), and idiopathic HCC (IL6ST) represent distinct molecular subgroups that should be stratified separately in targeted-therapy trials.

Limitations & open questions

• Exome-only mutational catalogues may underestimate the number of operative signatures (the authors note that 8 signatures were detected by whole-genome analysis of 88 prior HCC vs. 4 de novo signatures here, requiring re-introduction of consensus signatures to recover the full set).
• MSig5 (n=46) had low mutation rates and early histological stage but also lower tumor cell content — the authors flag possible confounding by normal-cell contamination.
• The hypermutated tumor CHC892T (signature 23) is a single case in a 71-year-old female with non-fibrotic liver and anthracotic pigment deposition; the underlying mutagen remains unknown.
• The Sanger-validation benchmark exposes systematic blind spots in exome calling: GC-rich ARID1A exon 1 (10 missed variants) and large CTNNB1 exon 3 deletions (n=3) — relevant when comparing this cohort to others.
• Drug-sensitivity inferences for HSP90 inhibitors rest on 29 cell lines (3 KEAP1-mutant) and have not been validated in vivo or clinically. The NQO1 P187S genotype-by-drug interaction is a single-cell-line observation (MHCC97H).
• Druggability calls were made by literature/database review (ClinicalTrials.gov, NCI Drug Dictionary) rather than direct functional screens; many “FDA-targetable” alterations are at <5% frequency and the actionability of single-hit amplifications in HCC remains to be demonstrated in trials.
• The cohort is exclusively European-resected HCC; etiologic distribution (41% alcohol, 14% HBV, 7% hemochromatosis) differs substantially from Asian and African HCC, limiting direct transfer of frequency estimates and the AFB1/signature-24 effect was validated in only 6 TCGA tumors.
• The conclusion that most non-fibrotic-liver HCC does not arise via hepatocellular-adenoma transformation rests on the HNF1A/IL6ST distribution and would benefit from orthogonal molecular timing data.

Citations from this paper used in the wiki

• “Exome sequencing analysis of 243 liver tumors revealed mutational signatures associated with specific risk factors, mainly combined alcohol/tobacco consumption, and aflatoxin B1.” (Abstract)
• “We identified 161 putative driver genes associated with 11 recurrent pathways.” (Abstract)
• “In 28% of the tumors we identified genetic alterations potentially targetable by FDA-approved drugs.” (Abstract)
• “Fourteen genes were significantly enriched for damaging mutations (q < 0.05, Fig. 2a): TP53, CTNNB1, AXIN1, ALB, ARID1A, ARID2, ACVR2A, NFE2L2, RPS6KA3, KEAP1, RPL22, CDKN2A, CDKN1A and RB1.” (Results)
• “11 pathways altered in ≥5% of HCC … TERT promoter mutations activating telomerase expression (60%), WNT/ß-Catenin (54%), PI3K/AKT/mTOR (51%), TP53/cell cycle (49%), MAP kinase (43%), hepatic differentiation (34%), epigenetic regulation (32%) chromatin remodeling (28%), oxidative stress (12%), Il6/JAK/STAT (9%), and TGFß (5%).” (Results)
• “Alcohol-related HCC were significantly enriched in CTNNB1, TERT, CDKN2A, SMARCA2, and HGF alterations (P < 0.05) … HBV-related HCC were frequently mutated for TP53, and IL6ST mutations were exclusively identified in HCC with no known etiology.” (Results)
• “The growth inhibition of 50% (GI50) was significantly inversely correlated with the expression of NQO1 (Pearson’s r = −0.56, P = 0.0015).” (Results, Fig. 5b)
• “We identified that CDKN2A inactivation and FGF/CCND1 amplification were associated with poor prognosis in our cohort of resected HCC, independently of classical prognostic clinical and histological features.” (Results, Fig. 6c)
• “Although TERT promoter mutations were already frequent at early stages, CTNNB1 and TP53 mutation frequencies increased significantly with progression, and focal amplifications at the CCND1/FGF locus were mostly encountered in poor prognosis HCC.” (Results)

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