Ampullary cancers harbor ELF3 tumor suppressor gene mutations and exhibit frequent WNT dysregulation

Authors

Marie-Claude Gingras

Kyle R. Covington

David K. Chang

Lawrence A. Donehower

Anthony J. Gill

Michael M. Ittmann

Chad J. Creighton

David A. Wheeler

Richard A. Gibbs

Doi

10.1016/j.celrep.2015.12.005

PMID: 26804919 · DOI: 10.1016/j.celrep.2015.12.005 · Journal: Cell Reports (2016)

TL;DR

Gingras et al. performed whole-exome sequencing, targeted deep sequencing, SNP-array copy number profiling, and (in a 30-patient subset) RNA-seq on 160 periampullary tumors — 98 ampullary adenocarcinomas (AMPAC), 44 distal bile-duct/cholangiocarcinomas (CAC), and 18 duodenal adenocarcinomas (DUOAC) — to build a molecular taxonomy distinct from the subjective AJCC anatomical/IHC classifications used clinically. Across all three anatomical sites, ~46% of tumors carried mutations in the WNT pathway and 10.6% carried predominantly inactivating frameshift/nonsense mutations in ELF3, a transcriptional regulator of TGFBR2 — an ELF3 mutation rate roughly 3× higher than reported in any other cancer in cBioPortal at the time. Microsatellite instability (MSI) was present in 12 patients (with PMS2 the dominant Lynch-syndrome gene in this cohort) and conferred a survival advantage. The authors argue these tumors should be regarded as a single “WNT+/–” entity for therapeutic decision-making, given small-molecule β-catenin inhibitors then in clinical trials.

Cohort & data

  • 160 periampullary tumors from the Australian Pancreatic Cancer Genome Initiative (APGI), Baylor College of Medicine, MD Anderson, and TU Dresden, all from patients who underwent curative-intent pancreatoduodenectomy (PMID:26804919):
  • Assays: whole-exome sequencing (HiSeq 2000) at ~120× average coverage on 152 samples; an 8-sample addendum captured by a 71-gene targeted exome panel; ultra-deep validation sequencing on all called mutations; Illumina HumanOmni2.5-8 SNP arrays for copy number; RNA-seq (TruSeq Stranded Total RNA, 100 bp paired-end, ≥50 M reads) on 28/98 AMPAC and 2/18 DUOAC.
  • Bioinformatics stack: HGSC Mercury pipeline → BWA → GATK recalibration → Atlas-SNP / Atlas-Indel / PInDel for variant calling → ANNOVAR / COSMIC / dbSNP annotation → MutSig-CV for significantly mutated genes → nsNMF for mutation signatures → GISTIC for focal CNAs → deFuse for fusion calling.
  • Reference genome: GRCh37.
  • Dataset: ampca_bcm_2016. Sequence data are deposited in dbGaP under accession PRJNA280134 (PMID:26804919).

Key findings

  • WNT pathway mutated in 46% of all periampullary tumors but at significantly different rates across anatomic sites: 72% in DUOAC, 49% in AMPAC, 30% in CAC (ChiSq p < 0.05). When grouped by histologic subtype, the intestinal subtype was 67% WNT-altered vs 30% in pancreatobiliary tumors (PMID:26804919, Tables S4a–b).
  • ELF3 mutated in 10.6% of periampullary tumors, predominantly via frameshift/nonsense (inactivating) variants — about 3× the rate observed in any other cancer in cBioPortal at the time. ELF3 mutations co-occurred with WNT-pathway mutations 71% of the time across all three groups (Chi-square p = 0.02) (PMID:26804919).
  • MSI in 12 patients (8% overall; AMPAC 3%, DUOAC most enriched). MSI tumors had dramatically elevated mutation rates (AMPAC 68/Mb, CAC 127/Mb, DUOAC 108/Mb vs MSS medians of 3.8, 4.6, 4.7 /Mb). All 6 MSI AMPAC patients were alive 2–8 years post-diagnosis (p = 0.04); across all three sites, MSI predicted significantly better survival (p < 0.0021).
  • Two-thirds of hypermutated WES samples carried germline mutations in Lynch-syndrome genes, with PMS2 mutated in half of MSI patients — strikingly higher than the <5% PMS2 share of Lynch syndrome population-wide (PMID:26804919).
  • Five mutational signatures were prominent (out of 21 nsNMF-derived signatures), led by C>T at CpG (#6). Signature #1 (AC, AT>AN, non-transcribed enrichment) appeared at >20% in 9.6% of tumors and was independently associated with worse outcome (multivariate Cox p = 0.02).
  • Significantly mutated genes (excluding MSI cases): 19 genes by MutSig-CV plus 3 added by inactivation-bias test (PBRM1, RECQL4, KDM6A) (Tables S3a–b). In CAC alone, only TP53, KRAS, SMAD4, and CDKN2A reached significance — TP53 highest. IDH1/IDH2 and BAP1, common in intrahepatic CAC, were essentially absent (single IDH1 hotspot only).
  • TGFBR2 mutated more often in DUOAC, partially driven by an A8 homopolymer that is an MSI target (5 of 12 TGFBR2 mutations were at this site).
  • Survival associations (multivariate Cox, adjusted for stage, gender, subtype, MSI status): TGF-β pathway mutations → better OS (p = 0.0059, HR = 0.42); PI3K pathway mutations → better OS (p = 0.036, HR = 0.43); TP53/KRAS/WNT/chromatin-remodeling mutations were not significant.
  • WNT-mutated tumors had significantly higher RNA expression of WNT antagonists, agonists, and target genes (p < 0.001) — consistent with feedback dysregulation seen in CRC. Some WNT-non-mutated tumors also showed elevated WNT signatures, suggesting alternate activating mechanisms.
  • Two non-recurrent fusions: SLC45A3-ELK4 (a marker first noted in prostate cancer) and a LINE-MET fusion in a patient with no KRAS or TP53 driver and high MET expression.
  • Copy number landscape was dominated by arm-level events; combined GISTIC identified focal 9p23.1 deletion (CDKN2A) and a chromosome 9 deletion removing the promoter and 5′ end of KDM4C (statistically significant in AMPAC), which also reduced expression of upstream UHRF2.

Genes & alterations

  • ELF3 — inactivating frameshift/nonsense mutations in 10.6% of periampullary tumors; tumor-suppressor pattern (3× higher than any other cancer in cBioPortal at the time). ELF3 transactivates TGFBR2 and EGR1 in CRC; loss may suppress TGF-β tumor-suppressor signaling. 75% of ELF3-mutant tumors were lower stage (I/II), and a duodenal adenoma with intraepithelial neoplasia carried an ELF3 mutation, supporting an early-driver role (PMID:26804919).
  • APC and SOX9 — drove the elevated WNT-pathway mutation frequency in DUOAC.
  • CTNNB1 — implicated as a WNT-pathway component; activation associates with WNT target gene up-regulation in mutated tumors.
  • TP53, KRAS, SMAD4, CDKN2A — the four MutSig-CV-significant genes when restricted to the 44 CAC subset; TP53 had the highest mutation incidence.
  • KRAS — major activated RAS-pathway oncogene in all three tumor types; overall RTK/RAS/PI3K activation 84–94% across sites.
  • TGFBR2 — more frequently mutated in DUOAC; 5/12 TGFBR2 mutations targeted an A8 homopolymer (MSI hotspot).
  • SMAD4 — most commonly mutated TGF-β-pathway gene in AMPAC and CAC (sites closest to the pancreas).
  • ARID1A, ARID2 — most frequent SWI/SNF chromatin-remodeling alterations; equally distributed across the three tumor types.
  • PBRM1, RECQL4, KDM6A — added to the significant gene list by the inactivation-bias test.
  • PMS2 — germline Lynch-syndrome gene mutated in half of MSI patients, an enrichment the authors explicitly contrast with the <5% population frequency.
  • KDM4C — focal chromosome-9 deletion removing the promoter/5′ end; statistically significant in AMPAC; deletion lowered expression of both KDM4C and the upstream UHRF2. Both have prior reports in colon-cancer growth and survival, and KDM4C complexes with β-catenin.
  • CDKN2A — focal 9p23.1 deletion (expected GISTIC peak).
  • IDH1, BAP1 — essentially absent in this extrahepatic CAC cohort (single IDH1 hotspot only), in contrast to intrahepatic CAC where these are common (consistent with Nakamura et al. 2015).
  • MET — recipient of a LINE-element insertion fusion in a single tumor lacking KRAS/TP53 mutations, with high MET expression.
  • SLC45A3 – ELK4 — non-recurrent in-frame fusion; previously a prostate-cancer prognostic marker.

Clinical implications

  • WNT-pathway status as a putative therapeutic stratifier. With nearly half of all periampullary tumors WNT-mutated and small-molecule β-catenin/WNT inhibitors then in clinical trials, the authors argue that ampullary, duodenal, and distal bile-duct adenocarcinomas should be regarded as a single “WNT+/–” entity for treatment decisions rather than triaged by anatomical/IHC subtype (PMID:26804919).
  • MSI as a favorable-prognosis biomarker in this disease group: all 6 MSI-AMPAC patients alive 2–8 years post-diagnosis (p = 0.04), and across all three sites MSI predicted better survival (p < 0.0021). The paper does not test immunotherapy response but situates MSI alongside its known prognostic role in other GI cancers.
  • Mutational signature #1 as an independent predictor of poor outcome (multivariate Cox p = 0.02) — a candidate prognostic biomarker.
  • TGF-β and PI3K pathway mutations independently predicted better OS (HR = 0.42 and 0.43 respectively in multivariate Cox).
  • Limited utility of histological/IHC subtyping as a treatment stratifier: pancreatobiliary vs intestinal IHC and morphology agreed only 53–77% of the time and did not segregate the molecular alterations cleanly — molecular profiling outperforms subjective subtyping for therapy selection.
  • PMS2 germline screening may be disproportionately valuable in periampullary tumors compared with its <5% role in general Lynch-syndrome populations.

Limitations & open questions

  • Modest CAC and DUOAC sample sizes (44 and 18 respectively) limit power for site-specific significantly-mutated-gene calls, especially in DUOAC.
  • RNA-seq subset is small and AMPAC-skewed (28 AMPAC + 2 DUOAC; no CAC), constraining cross-site expression comparisons.
  • ELF3’s functional duality is unresolved — the authors note prior reports of both tumor-suppressor and oncogenic (in amplification contexts) roles, and they hypothesize but do not establish that early ELF3 loss-of-function provides a moderate growth advantage by suppressing TGF-β signaling. Mechanism, stage-dependence, and expression-level effects are flagged as needing follow-up.
  • No formal therapeutic testing — the WNT-targeting clinical recommendation is grounded in mutation frequency plus contemporaneous trials of β-catenin inhibitors, not in this study’s own response data.
  • Anatomic-site assignment remains operator-dependent for the 98 AMPAC cases (“epithelium of origin could not be clearly defined morphologically”), which is exactly the inter-observer-variability problem the molecular taxonomy is meant to circumvent.
  • MSI counting depends on a custom homopolymer-slippage caller (“Shinbrot et al., in preparation”) for the targeted-sequencing arm; method details deferred to a then-unpublished manuscript.
  • Cross-paper comparison with intrahepatic cholangiocarcinoma cohorts (e.g., Nakamura et al. 2015, cited but not in this corpus) suggests molecular distinctness from intrahepatic CAC (no IDH1/BAP1 here) but is not formally tested.

Citations from this paper used in the wiki

  • “ELF3 a transcriptional regulator of TGFBR2 was mutated in 10.6% of the periampullary tumors with predominantly inactivating frameshift or nonsense mutations… This mutation frequency is 3 times higher than in any other cancer.”
  • “The WNT pathway was mutated in 46% of patients overall, but was clearly differentially mutated across the 3 tumor types, being more frequently mutated in DUOAC, 72%, than in AMPAC, 49%, or CAC, 30%, (ChiSq p < 0.05).”
  • “ELF3 mutation occurred 71% of the time with WNT pathway mutations in all three periampullary groups (Chisquare test, p=0.02).”
  • “Microsatellite instable phenotypes were observed in 12 patients… all 6 AMPAC patients were alive ranging from 2 to 8 years after diagnosis (p = 0.04)… Taking all three anatomical sites into consideration, MSI have better survival, p<0.0021.”
  • “PMS2, a gene that accounts for less than 5% of Lynch Syndrome patients overall… was mutated in one half of our MSI patients.”
  • “Considering the 44 CAC alone, four genes were significantly mutated in this cancer: TP53, KRAS, SMAD4 and CDKN2A with the highest mutation incidence in TP53.”
  • “Mutations in the TGF-β pathway are associated with better overall survival (multivariate-Cox proportional hazard p=0.0059, HR=0.42)… Mutations in the PI3K pathway were also associated with better overall survival (multivariate-Cox proportional hazards p=0.036, HR=0.43).”
  • “A focal deletion in chromosome 9 removed the promotor and 5′ end of KDM4C… This deletion resulted in a significant decrease in expression of KDM4C as well as the upstream UHRF2.”
  • “Sequence data used for this analysis are located in dbGAP accession number PRJNA280134.”

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