Exome sequencing identifies frequent inactivating mutations in BAP1, ARID1A and PBRM1 in intrahepatic cholangiocarcinomas

Authors

Yuchen Jiao

Timothy M Pawlik

Robert A Anders

Aldo Scarpa

Ralph H Hruban

Bert Vogelstein

Kenneth W Kinzler

Nickolas Papadopoulos

Laura D Wood

Doi

10.1038/ng.2813

PMID: 24185509 · DOI: 10.1038/ng.2813 · Journal: Nature Genetics (2013)

TL;DR

Jiao et al. performed whole-exome sequencing of 32 intrahepatic cholangiocarcinomas and 9 gallbladder carcinomas (discovery screen), followed by targeted resequencing of 17 candidate driver genes in 32 additional IHCH and 8 additional GBC tumors (prevalence screen). They discovered frequent inactivating mutations in the chromatin-remodeling genes BAP1, ARID1A, and PBRM1 in IHCH — at least one of these was altered in 47% of the discovery cohort and 41% of the combined cohort. IDH1/IDH2 hotspot mutations occurred in 20% of IHCH and were associated with significantly worse survival. By contrast, TP53 was the dominant alteration in GBC, supporting genetically distinct biology between IHCH and gallbladder carcinoma despite their shared anatomic origin.

Cohort & data

  • Discovery screen: 32 intrahepatic cholangiocarcinomas and 9 gallbladder carcinomas, tumor + matched normal, whole-exome sequencing on Agilent SureSelect Paired-End v2.0 Human Exome capture + Illumina HiSeq 2000. Mean coverage 130× over the targeted region; >90% of targeted bases ≥10×. Aligned to hg18 via Eland/CASAVA 1.6. One GBC tumor with a hypermutator phenotype (~3,000 nonsynonymous mutations) was excluded, leaving 8 GBC for analysis. Samples macrodissected; tumor/normal collected under IRB protocols at Johns Hopkins, MSKCC, Mayo Clinic, Fundeni Clinical Institute, and University Hospital Trust of Verona.
  • Prevalence screen: 32 additional IHCH and 8 additional GBC from Verona University Hospital, targeted DNA sequencing of 17 candidate driver genes (AKT1, ARID1A, BAP1, CDKN2A, CTNNB1, IDH1, IDH2, KRAS, NRAS, PBRM1, PIK3C2A, PIK3C2G, PIK3CA, PTEN, SMAD4, TGFBR2, TP53) using a custom Ion AmpliSeq panel on Ion Torrent PGM (318 chip). Mean read depth 1,276×; 88.9% of target bases ≥100×. Aligned to hg19.
  • Validation: 50 mutations confirmed by Sanger sequencing; all 50 verified.
  • Dataset: chol_jhu_2013.
  • Total mutation burden: 1,259 somatic mutations in 1,128 genes across the 32 IHCH (mean 39 per sample, range 13–300); 724 somatic mutations in 695 genes across the 8 evaluable GBC (mean 91 per sample, range 42–252).

Key findings

  • Chromatin-remodeling genes are recurrently inactivated in IHCH. In the discovery screen, BAP1 was mutated in 8/32 (25%), ARID1A in 6/32 (19%), and PBRM1 in 5/32 (17%) IHCH; at least one was altered in 15/32 (47%) of tumors. Mutations were enriched for inactivating types (nonsense, frameshift, splice-site).
  • Significance after multiple-testing correction. Combining single-base substitution counts, indel counts, and nonsynonymous enrichment via Fisher’s combined test with Bonferroni correction, BAP1 P = 7.4 × 10⁻¹², ARID1A P = 4.3 × 10⁻³, PBRM1 P = 1.8 × 10⁻⁵, IDH1 P = 5.4 × 10⁻³, FGFR2 P = 4.8 × 10⁻² — all qualified as significantly mutated genes in IHCH.
  • Combined IHCH cohort (n=64) mutation frequencies: BAP1 13/64 (20%), ARID1A 9/64 (14%), PBRM1 8/64 (13%); 26/64 (41%) of IHCH carried a mutation in at least one of these chromatin-remodeling genes. In the prevalence screen alone, these three genes were mutually exclusive; in the discovery screen 3 tumors carried multiple chromatin-remodeling gene hits.
  • IDH1/IDH2 hotspots cluster as expected and are prognostic. IDH1 was mutated in 4/32 IHCH (codon 132) and IDH2 in 2/32 (codon 172) in the discovery screen (19% combined); 7/32 (22%) in the prevalence screen; 20% overall in the combined 64-tumor IHCH cohort. Subjects with IDH1/IDH2 mutations had 3-year survival of 33% vs 81% for IDH-wild-type (log-rank P = 0.0034); after Cox adjustment for stage, age and sex, HR = 7.37 (95% CI 1.13–48.29, P = 0.037).
  • FGFR2 mutations identified in 13% (4/32) of discovery-screen IHCH (FGFR2), with two at residues previously reported in endometrial carcinoma. Authors note these complement previously reported FGFR2 gene fusions in cholangiocarcinoma.
  • PI3K pathway alterations totaled 22% of discovery-screen IHCH — 2 PIK3CA, 2 PTEN, 2 PIK3C2G, 1 PIK3C2A. TP53 was mutated in only 2/32 (6%); single inactivating CDKN2A, single oncogenic hotspot KRAS, and single NRAS mutations identified. No SMAD4 mutations in IHCH; one TGFBR2 mutation.
  • Gallbladder carcinoma is genetically distinct from IHCH. In 8 GBC discovery samples, TP53 was the most frequently mutated gene (5/8, 63%; only TP53 met the Bonferroni-corrected significance threshold). PBRM1 was mutated in 2/8 (25%) and KMT2C (MLL3) in 2/8 (25%). One PIK3CA and one SMAD4 mutation. No BAP1, ARID1A, IDH1, or IDH2 mutations in discovery-screen GBC.
  • Combined GBC cohort (n=16) mutation frequencies: TP53 44%, PBRM1 25%, BAP1 6%, ARID1A 6%. No IDH1/IDH2 mutations in the additional 8 GBC of the prevalence screen.
  • Genetic heterogeneity. No single gene was mutated in >25% of IHCH tumors in the combined cohort, but pathway-level convergence (chromatin remodeling, IDH metabolism, PI3K/RAS signaling) was clear.

Genes & alterations

  • BAP1 — inactivating mutations (nonsense, frameshift, splice-site predominate) in 13/64 (20%) IHCH combined. Encodes a nuclear deubiquitinase implicated in chromatin remodeling; previously reported in renal cell carcinoma, uveal melanoma, and mesothelioma, but this is the first report in a gastrointestinal cancer.
  • ARID1A — SWI/SNF complex subunit; 9/64 (14%) IHCH; not previously reported in cholangiocarcinoma.
  • PBRM1 — SWI/SNF complex subunit; 8/64 (13%) IHCH and 4/16 (25%) GBC; previously known mainly from clear-cell renal cell carcinoma.
  • IDH1 — codon 132 hotspot missense in IHCH; component of the 20% combined IHCH IDH mutation rate; associated with significantly worse survival in this cohort (HR 7.37, P = 0.037 after adjustment).
  • IDH2 — codon 172 hotspot missense in IHCH; absent in GBC in this study.
  • FGFR2 — 4/32 (13%) IHCH discovery-screen tumors; two mutations at residues mutated in endometrial carcinoma. Flagged as a potential therapeutic target via FGFR inhibitors then in trials.
  • TP53 — dominant driver in GBC (44% of combined 16 GBC); only 6% (2/32) in IHCH discovery screen.
  • KMT2C (MLL3) — chromatin-remodeling gene mutated in 2/8 (25%) GBC discovery-screen tumors; absent in IHCH in this study.
  • PIK3CA, PTEN, PIK3C2A, PIK3C2G — PI3K pathway alterations totaling 22% of IHCH discovery-screen tumors.
  • KRAS, NRAS — single hotspot mutations each in IHCH discovery; additional KRAS and NRAS hits in the prevalence screen confirm RAS signaling involvement.
  • CDKN2A, SMAD4, CTNNB1, TGFBR2, AKT1 — sporadic alterations included in the prevalence-screen panel; no SMAD4 mutations identified in IHCH.

Clinical implications

  • Biomarker — IDH1/IDH2 mutation as a prognostic marker in IHCH. In this Johns Hopkins/Verona cohort, IDH mutation was associated with substantially worse 3-year survival (33% vs 81%, log-rank P = 0.0034; adjusted HR 7.37, P = 0.037). The authors explicitly flag that this direction of effect is opposite to a prior report (ref. 16 in the paper), so the prognostic sign of IDH mutation in IHCH remains open.
  • Early detection — circulating IDH hotspots. Authors propose plasma-based detection of IDH1/IDH2 hotspot mutations as a screening strategy for high-risk populations.
  • Therapy — chromatin-remodeling vulnerability. Authors suggest that BAP1/ARID1A/PBRM1-mutant IHCH may be sensitive to drugs targeting chromatin remodeling, citing HDAC inhibitors then in clinical use or development. No drug was administered in this study.
  • Therapy — FGFR2 alterations as targetable lesions. Authors note multiple FGFR inhibitors in clinical trials at the time as a rationale for testing FGFR2-mutant IHCH.
  • Disease classification. GBC vs IHCH show distinct mutational landscapes (TP53-dominant vs chromatin-remodeling-dominant), supporting the position that they should not be lumped together molecularly despite shared biliary origin.

Limitations & open questions

  • Small GBC sample size. Only 8 evaluable GBC in the discovery screen and 8 in the prevalence screen; authors explicitly state the cohort is too small for definitive genetic-relationship conclusions between GBC and IHCH.
  • Conflicting IDH prognosis. This study’s finding that IDH1/IDH2 mutation is associated with worse survival in IHCH directly contradicts an earlier report showing improved survival; both used relatively small cohorts and the authors call for independent replication.
  • Hypermutator excluded. One GBC tumor with ~3,000 nonsynonymous mutations was removed from the cohort; its underlying mechanism (e.g. MMR deficiency, POLE) was not investigated here.
  • Coding-region bias. Whole-exome sequencing misses non-coding drivers, structural variants, and gene fusions; FGFR2 fusions previously reported in cholangiocarcinoma are explicitly outside the assay’s reach.
  • No therapeutic validation. Sensitivity of chromatin-remodeling-mutant cholangiocarcinomas to HDAC inhibitors and of FGFR2-mutant tumors to FGFR inhibitors is hypothesized but not tested in this paper.
  • Macrodissection but no laser-capture. Tumor purity was enhanced by macrodissection and frozen-section review but not by laser capture; estimates of allele frequency and subclonal architecture should be interpreted accordingly.
  • Reference genome mix. Discovery screen aligned to hg18, prevalence screen to hg19, potentially complicating cross-cohort coordinate reconciliation by downstream users.

Citations from this paper used in the wiki

  • “Through exomic sequencing of 32 intrahepatic cholangiocarcinomas, we discovered frequent inactivating mutations in multiple chromatin-remodeling genes (including BAP1, ARID1A and PBRM1), and mutation in one of these genes occurred in almost half of the carcinomas sequenced.” (Abstract)
  • “Somatic mutations in BAP1 … occurred in 8 of 32 intrahepatic cholangiocarcinomas (25%). … somatic ARID1A mutations in 6 of 32 … (19%). Somatic PBRM1 mutations … in 5 of 32 … (17%).” (Results)
  • “Combining the discovery and prevalence screens, BAP1 was mutated in 13 of 64 tumors (20%), ARID1A was mutated in 9 of 64 tumors (14%), and PBRM1 was mutated in 8 of 64 tumors (13%), adding up to 26 cholangiocarcinomas with mutations in chromatin-remodeling genes out of 64 total cholangiocarcinomas (41%).” (Results)
  • “subjects with IDH1 or IDH2 mutations had 3-year survival of 33% compared with 3-year survival of 81% for subjects with wild-type IDH genes (P = 0.0034) … hazards ratio (HR) = 7.37, 95% confidence interval (CI) = 1.13–48.29, P = 0.037.” (Results)
  • “TP53 was the most frequently mutated gene in gallbladder carcinoma, with somatic mutations in five of eight tumors without a mutator phenotype (63%).” (Results)
  • “the overall mutation prevalence in gallbladder carcinoma for BAP1, ARID1A and PBRM1 was 6%, 6% and 25%, respectively.” (Results)
  • “the P values for their observed mutations were 7.4 × 10⁻¹², 4.3 × 10⁻³ and 1.8 × 10⁻⁵, respectively, using the conservative Bonferroni correction” (BAP1, ARID1A, PBRM1; Results)
  • “these mutations may identify additional therapies for cholangiocarcinomas, as the mutations may cause sensitivity to drugs targeting chromatin remodeling, such as histone deacetylase (HDAC) inhibitors” (Results)

This page was processed by crosslinker on 2026-05-09.