Comprehensive molecular characterization of gastric adenocarcinoma

Author

The Cancer Genome Atlas Research Network

Doi

PMID: 25079317 · DOI: 10.1038/nature13480 · Journal: Nature (2014)

TL;DR

The Cancer Genome Atlas Research Network profiled 295 primary, treatment-naive [[../cancer_types/STAD.md|gastric adenocarcinoma]] tumours using six molecular platforms (SNP array, whole-exome sequencing, DNA methylation, mRNA-seq, miRNA-seq, and RPPA), with low-pass whole-genome sequencing on 107 tumour/germline pairs. The integrated analysis defines four molecular subtypes: (1) [[../genes/CD274.md|EBV]]-positive (9%), with recurrent PIK3CA mutations, extreme DNA hypermethylation, and 9p24.1 amplification of JAK2/CD274/PDCD1LG2; (2) microsatellite instability (MSI; 22%), with elevated mutation rates and hypermethylation at the MLH1 promoter; (3) genomically stable (GS; 20%), enriched for diffuse histology and RHOA mutations or CLDN18–ARHGAP fusions; and (4) chromosomal instability (CIN; 50%), enriched for intestinal histology, TP53 mutation, and focal amplification of receptor tyrosine kinases. The taxonomy provides a roadmap for biomarker-stratified clinical trials of targeted therapies in gastric cancer.

Cohort & data

  • N = 295 primary gastric adenocarcinoma patients (treatment-naive at surgery), profiled as the TCGA gastric cancer cohort (stad_tcga_pub).
  • Six molecular platforms: array-based somatic copy-number ([[../methods/affymetrix-snp6.md|Affymetrix SNP6]]), [[../methods/whole-exome-seq.md|whole-exome sequencing]], DNA methylation ([[../methods/hm450-methylation-array.md|HM450]]), mRNA-seq ([[../methods/rna-seq.md|RNA-seq]]), miRNA-seq, and reverse-phase protein arrays ([[../methods/rppa.md|RPPA]]); 77% of tumours profiled on all six.
  • Low-pass [[../methods/whole-genome-seq.md|whole-genome sequencing]] (<6× coverage) on 107 tumour/germline pairs.
  • MSI testing on all tumour DNA. Non-malignant gastric tissue controls collected for methylation (n=27) and expression (n=29) analyses.
  • Subtype assignment used a decision tree: EBV-positivity first, then MSI-high status, then CIN vs GS by aneuploidy.

Key findings

  • Four-subtype molecular classification: EBV (9%), MSI (22%), GS (20%), CIN (50%).
  • EBV-positive tumours showed extreme CIMP distinct from MSI-associated CIMP; all assayed EBV-positive tumours had CDKN2A (p16INK4A) promoter hypermethylation but lacked [[../genes/MLH1.md|MLH1]] hypermethylation. [[../genes/PIK3CA.md|PIK3CA]] mutations occurred in 80% of EBV-positive tumours (P = 9×10⁻¹²) vs 3–42% of other subtypes; PIK3CA mutations were dispersed in EBV-positive cancers but localized to the kinase domain (exon 20) in EBV-negative cancers. EBV-positive tumours also had frequent ARID1A (55%) and BCOR (23%) mutations and only rare [[../genes/TP53.md|TP53]] mutations.
  • 9p24.1 amplification of the locus containing [[../genes/JAK2.md|JAK2]], [[../genes/CD274.md|CD274]] (PD-L1) and [[../genes/PDCD1LG2.md|PDCD1LG2]] (PD-L2) was enriched in EBV-positive tumours (15%), with elevated mRNA expression in amplified cases. PD-L1/2 expression was elevated more broadly in EBV-positive tumours.
  • MSI subtype: significantly mutated genes via MutSigCV ([[../methods/mutsig.md|MutSig]]) included [[../genes/TP53.md|TP53]], KRAS, [[../genes/ARID1A.md|ARID1A]], [[../genes/PIK3CA.md|PIK3CA]], ERBB3, PTEN and HLA-B. [[../genes/ERBB3.md|ERBB3]] mutations were found in 16/63 tumours (13 at recurrent or COSMIC sites). Expanded analysis (indels included) yielded 37 significantly mutated genes, adding RNF43, B2M and NF1. BRAF V600E mutations were notably absent (unlike MSI colorectal cancer). One hypermutator harboured an inactivating POLE mutation.
  • GS subtype: enriched for diffuse Lauren histology (40/55 = 73%, P = 7.5×10⁻¹⁷). [[../genes/RHOA.md|RHOA]] mutations in 16 cases (15% of GS, P = 0.0039), clustered in two amino-terminal regions predicted to interface with ROCK1/effectors; not at sites analogous to oncogenic RAS-family mutations. The Y42C mutation attenuates Protein Kinase N activation. [[../genes/CLDN18.md|CLDN18]]–ARHGAP26 (GRAF) interchromosomal fusions identified in 11 cases plus 2 [[../genes/CLDN18.md|CLDN18]]–ARHGAP6 fusions (13 total); fusions use a cryptic splice site in CLDN18 exon 5 to produce in-frame chimaeras retaining the GAP domain. RHOA mutations and CLDN18–ARHGAP fusions were mutually exclusive and together altered 30% of GS cases. CDH1 somatic mutations were enriched in GS (37%).
  • CIN subtype: [[../genes/TP53.md|TP53]] mutation in 71%; recurrent focal amplifications of oncogenes including ERBB2, CCNE1, [[../genes/KRAS.md|KRAS]], MYC, EGFR, CDK6, GATA4, GATA6, ZNF217 and VEGFA; amplification of the gastric stem cell marker CD44; elevated EGFR pY1068 phosphorylation by RPPA.
  • MutSigCV on the 215 non-hypermutated tumours identified 25 significantly mutated genes, including TP53, [[../genes/ARID1A.md|ARID1A]], KRAS, PIK3CA, [[../genes/RNF43.md|RNF43]], APC, CTNNB1, SMAD4, SMAD2, RASA1 and [[../genes/ERBB2.md|ERBB2]] (10/15 ERBB2 mutations at known hotspots; four S310F cases).
  • Splicing: MET exon 2 skipping observed in 82/272 (30%) cases, associated with increased MET expression (P = 10⁻⁴); novel MET exon 18/19 skipping in 47/272 (17%).
  • Pathway analysis ([[../methods/gistic.md|GISTIC]] + integrated mutation/SCNA): elevated mitotic-network signalling (AURKA/AURKB, FOXM1, PLK1) across subtypes; GS subtype distinguished by elevated cell-adhesion and angiogenesis pathways; IL-12 signalling elevated in EBV-positive tumours alongside PD-L1/2 over-expression.

Genes & alterations

  • [[../genes/PIK3CA.md|PIK3CA]] — recurrent mutations in 80% of EBV-positive tumours (mostly non-exon-20, dispersed); 12% across the whole cohort; targetable via PI(3)-kinase inhibition.
  • [[../genes/JAK2.md|JAK2]], [[../genes/CD274.md|CD274]], [[../genes/PDCD1LG2.md|PDCD1LG2]] — co-amplified at 9p24.1 in 15% of EBV-positive tumours; elevated mRNA expression; rationale for JAK2 inhibitors and PD-L1/L2 immune-checkpoint blockade.
  • [[../genes/ARID1A.md|ARID1A]] — 14% overall; especially common in EBV-positive (55%) and GS tumours.
  • [[../genes/BCOR.md|BCOR]] — 4% overall; 23% in EBV-positive tumours.
  • [[../genes/TP53.md|TP53]] — 50% overall; 71% in CIN tumours; rare in EBV-positive.
  • [[../genes/RHOA.md|RHOA]] — mutated in 15% of GS tumours; novel hotspots in the effector-binding region (Y42, D59 etc.) distinct from RAS-family oncogenic sites and from T-cell-neoplasm G17V mutations.
  • [[../genes/CLDN18.md|CLDN18]]–[[../genes/ARHGAP26.md|ARHGAP26]]/[[../genes/ARHGAP6.md|ARHGAP6]] fusions — in-frame chimaeric proteins from t(3;5) and related rearrangements; in 15% of GS tumours; mutually exclusive with RHOA mutation.
  • [[../genes/CDH1.md|CDH1]] — somatic mutations in 11% overall and 26–37% of GS tumours; germline analysis revealed only two non-pathogenic polymorphisms in this cohort.
  • [[../genes/ERBB2.md|ERBB2]] — 17% alteration in CIN (focal amplification + recurrent S310F-class hotspots); 3% mutation overall; established drug target.
  • [[../genes/ERBB3.md|ERBB3]] — 15% alteration (mutations enriched in MSI; 13/16 at COSMIC/recurrent sites).
  • [[../genes/EGFR.md|EGFR]] — 11% alteration (amplification in CIN; elevated pY1068 phosphorylation).
  • FGFR2 — 9% alteration (RTK amplification in CIN).
  • [[../genes/MET.md|MET]] — recurrent exon 2 (30%) and exon 18/19 (17%) alternative splicing; potential therapeutic relevance.
  • [[../genes/KRAS.md|KRAS]] — 6% mutation overall; KRAS/NRAS alterations 17% across CIN/MSI.
  • [[../genes/CCNE1.md|CCNE1]], CCND1, [[../genes/CDK6.md|CDK6]] — recurrent CIN amplifications; suggest CDK4/6 inhibitor evaluation.
  • [[../genes/VEGFA.md|VEGFA]] — recurrent amplification (7%); rationale for VEGFR-pathway inhibition (e.g. ramucirumab).
  • [[../genes/MLH1.md|MLH1]] — promoter hypermethylation drives the MSI subtype (gastric-CIMP).
  • [[../genes/CDKN2A.md|CDKN2A]] — promoter hypermethylation in all EBV-positive tumours; focal deletion in CIN.
  • [[../genes/RNF43.md|RNF43]] — significantly mutated (Wnt pathway negative regulator).
  • [[../genes/APC.md|APC]], [[../genes/CTNNB1.md|CTNNB1]] — β-catenin pathway alterations among non-hypermutated tumours.
  • [[../genes/SMAD4.md|SMAD4]], [[../genes/SMAD2.md|SMAD2]] — recurrent TGF-β pathway mutations.
  • [[../genes/RASA1.md|RASA1]] — negative regulator of RAS, recurrently mutated.
  • [[../genes/B2M.md|B2M]], [[../genes/HLA-B.md|HLA-B]] — MHC class I alterations enriched in MSI tumours; consistent with antigen-presentation escape.
  • [[../genes/NF1.md|NF1]] — added to the significant-gene list in MSI tumours after indel inclusion.
  • [[../genes/PTEN.md|PTEN]], PIK3R1 — PI3K pathway losses (PTEN focally deleted in CIN).
  • [[../genes/POLE.md|POLE]] — one hypermutator case with inactivating POLE mutation, consistent with prior TCGA findings in CRC/endometrial.
  • [[../genes/MYC.md|MYC]], [[../genes/GATA4.md|GATA4]], [[../genes/GATA6.md|GATA6]], [[../genes/ZNF217.md|ZNF217]], [[../genes/CD44.md|CD44]] — focal amplifications in CIN.
  • [[../genes/AURKA.md|AURKA]], [[../genes/AURKB.md|AURKB]] — implicated by elevated mitotic-pathway expression; suggested as candidate therapeutic targets.

Clinical implications

  • EBV subtype: PI3K inhibition warrants evaluation (80% PIK3CA mutation); PD-L1/L2 antagonists (immune-checkpoint inhibition) and JAK2 inhibitors are mechanistically supported by the 9p24.1 amplification and IL-12 signature.
  • MSI subtype: targetable mutations in [[../genes/PIK3CA.md|PIK3CA]], [[../genes/ERBB3.md|ERBB3]], [[../genes/ERBB2.md|ERBB2]] and [[../genes/EGFR.md|EGFR]] at known hotspots. BRAF V600E is notably absent, unlike MSI colorectal cancer.
  • GS subtype: few clear actionable amplifications; [[../genes/RHOA.md|RHOA]] and [[../genes/CLDN18.md|CLDN18]]–ARHGAP alterations point to RHO-signalling and cell-adhesion biology as candidate targets.
  • CIN subtype: recurrent amplifications of RTKs ([[../genes/ERBB2.md|ERBB2]], [[../genes/EGFR.md|EGFR]], [[../genes/MET.md|MET]], [[../genes/FGFR2.md|FGFR2]]) and cyclins/CDKs ([[../genes/CCNE1.md|CCNE1]], [[../genes/CCND1.md|CCND1]], [[../genes/CDK6.md|CDK6]]) amenable to existing or in-development blockade; [[../genes/VEGFA.md|VEGFA]] amplification supports activity of the VEGFR2 antibody [[../drugs/ramucirumab.md|ramucirumab]].
  • The authors propose that existing MSI and EBV testing combined with focused mutation/amplification panels can operationalize the classification for trial enrolment.

Limitations & open questions

  • Initial outcome data showed no survival differences between the four subgroups; longer follow-up needed.
  • H. pylori was only sporadically detected — likely reflects declining bacterial counts in advanced disease and technical loss during processing rather than a true absence of association.
  • The CLDN18–ARHGAP fusions occur downstream of the CLDN18 exon 5 stop codon; the in-frame translation depends on cryptic splice activation — functional consequences for ARHGAP/RHO regulation and CLDN18 adhesion are inferred but not biochemically confirmed in this study.
  • No germline-pathogenic CDH1 carriers were identified in this sporadic cohort, leaving hereditary diffuse gastric cancer biology outside its scope.
  • The decision-tree assignment depends on tumour purity for SCNA detection (CIN vs GS); the authors note no systematic confounding but acknowledge low-purity edge cases.
  • Translational validation of the proposed targets (PI3K inhibitors in EBV, anti-PD-L1 in EBV, CDK4/6 in CIN, RHO-pathway modulators in GS) remains for subsequent trials.

Citations from this paper used in the wiki

  • “We propose a molecular classification dividing gastric cancer into four subtypes: tumours positive for Epstein–Barr virus … microsatellite unstable tumours … genomically stable tumours … and tumours with chromosomal instability” (Abstract).
  • “We observed strong predilection for PIK3CA mutation in EBV-positive gastric cancer … non-silent PIK3CA mutations found in 80% of this subgroup (P = 9×10⁻¹²)” (EBV-associated DNA hypermethylation section).
  • “We identified RHOA mutation in 16 cases, and these were enriched in the genomically stable subtype (15% of genomically stable cases, P = 0.0039)” (Somatic genomic alterations).
  • “CLDN18–ARHGAP fusions were mutually exclusive with RHOA mutations and were enriched in genomically stable tumours (62%, P = 10⁻²³)” (RHOA/ARHGAP section).
  • “Focal amplifications targeted oncogenes such as ERBB2, CCNE1, KRAS, MYC, EGFR, CDK6, GATA4, GATA6 and ZNF217” (SCNA section).
  • “Recurrent amplification of the gene encoding ligand VEGFA was notable given the gastric cancer activity of the VEGFR2 targeting antibody ramucirumab” (Integrated pathway analysis).

This page was processed by crosslinker on 2026-05-11.