Mutational landscape of gastric adenocarcinoma in Chinese: Implications for prognosis and therapy

Authors

Kexin Chen

Da Yang

Xiangchun Li

Baocun Sun

Fengju Song

Wenfeng Cao

Daniel J. Brat

Zhibo Gao

Haixin Li

Han Liang

Yanrui Zhao

Hong Zheng

Miao Li

Jan Buckner

Scott D. Patterson

Xiang Ye

Christoph Reinhard

Anahita Bhathena

Deepa Joshi

Paul S. Mischel

Carlo M. Croce

Yi Michael Wang

Sreekumar Raghavakaimal

Hui Li

Xin Lu

Yang Pan

Han Chang

Sujuan Ba

Longhai Luo

Webster K. Cavenee

Wei Zhang

Xishan Hao

Doi

10.1073/pnas.1422640112

PMID: 25583476 · DOI: 10.1073/pnas.1422640112 · Journal: PNAS (2015)

TL;DR

Chen et al. performed whole-exome sequencing on 78 primary gastric adenocarcinomas (GCs) from a northern Chinese cohort (Tianjin Medical University Cancer Institute and Hospital), whole-genome sequencing on two cases with multiple primary-tumor and lymph-node-metastasis samples each, and targeted resequencing of 103 recurrently mutated Wnt, ERBB, and homologous-recombination pathway genes in a 216-case validation cohort. Clonality analysis (SciClone) separated GC into a high-clonality (HiC, n=9) subtype — older patients, enriched for TP53 mutations and C>G transitions, with significantly shorter survival (adjusted HR 4.69, 95% CI 1.62–13.6, P=0.0043) — and a low-clonality (LoC, n=68) subtype — younger, enriched for ARID1A mutations, longer survival. The targetable NRG1–ERBB4 ligand-receptor pair was mutated in 11.6% (34/294) of cases, and BRCA2 mutations (8% of the discovery cohort, validated in TCGA) marked a subgroup with significantly longer survival (pooled log-rank P=0.03; adjusted HR 0.37, 95% CI 0.13–0.96, P=0.05).

Cohort & data

  • Discovery cohort: 78 frozen primary gastric adenocarcinoma samples (treatment-naive, no prior chemo/radiotherapy) from northern Chinese patients with matching germline blood DNA; whole-exome sequencing on Illumina HiSeq 2000 with Agilent SureSelect capture; mean coverage 167× tumor / 170× normal; 85% of exons ≥20×.
  • WGS sub-cohort: Two patients (Pt1, Pt2) each with three primary-tumor regions plus two matched lymph-node metastases (5 samples per case) by whole-genome sequencing.
  • Validation cohort: 216 additional GC samples from the same population by targeted sequencing of 103 recurrent genes on the Ion Torrent PGM platform.
  • Full study population: 294 patients (after excluding 5 with discrepant pathology). Anatomic location: antrum 64 (21.77%), body 118 (40.14%), cardia 112 (38.10%). Histology: 124 intestinal-type, 152 diffuse-type, 18 mixed. Stage: I 6 (2.05%), II 85 (28.91%), III 97 (32.99%), IV 106 (36.05%).
  • Follow-up: Median 13.84 months (WES series 25.08 months, 32 deaths/41.03%; targeted-seq series 12.16 months, 61 deaths/28.24%).
  • Cancer type: Gastric adenocarcinoma (STAD).
  • Dataset: egc_tmucih_2015; validation comparisons against the Russian TCGA gastric cohort (stad_tcga_pub, n=289).
  • Variant calling: MuTect and VarScan 2.2.5; driver discovery with MutSigCV; clonality with SciClone; multi-region phylogenetic tree reconstruction; drug-gene interactions via DGIdb. Sequence data deposited at EGA under accession EGAS00001001056.

Key findings

  • Mutation burden: 13,866 somatic mutations across 78 WES tumors (3,421 synonymous, 8,558 missense, 576 nonsense, 241 splice-site, 967 frameshift, 103 non-frameshift indels). Median 82.5 nonsilent mutations per tumor (range 1–1,486); 11,768 protein-coding subclonal SNVs (92.3% of detected mutations, median 106 per sample).
  • HiC vs. LoC subtypes: SciClone clustering identified 9 HiC (>4 clones) and 68 LoC (≤4 clones) tumors. HiC tumors were associated with older diagnosis age (Pearson r=0.27, P=0.02), significantly higher TP53 mutation frequency, and significantly higher ARID1A frequency in the LoC group (both P=0.003). HiC GCs had shorter survival (log-rank P=0.02); the HiC–survival association remained significant after adjusting for age, sex, stage, Lauren subtype, TP53 and ARID1A mutation status (adjusted HR 4.69, 95% CI 1.62–13.6, P=0.0043), establishing clonal subtype as an independent prognostic factor.
  • Mutational signatures: HiC tumors had a smaller fraction of C>T mutations (45% vs. 52% in LoC) and a higher fraction of C>G mutations (25% vs. 9% in LoC, P=0.002), suggesting distinct mutagenic etiologies.
  • TP53 in HiC is subclonal: All six TP53-mutant HiC tumors carried TP53 only in minor clones (mutation frequency <15%), implying that single-agent therapy could select for TP53-wild-type subclones.
  • Driver discovery: MutSigCV identified 16 significantly mutated genes (q<0.2); 13 mutated in ≥5% of tumors, including previously reported drivers TP53, ARID1A, CDH1, APC, RHOA, PIK3CA, SMAD4, MYC, and KRAS. Pathway enrichment highlighted TP53, Wnt, ERBB, and homologous-recombination pathways.
  • Validation: 86/103 (83%) recurrent genes selected from Wnt/ERBB/HR pathways were re-identified in the 216-case validation cohort.
  • Anatomic-location associations (WES): Antrum tumors enriched for mutations in CSMD3 (30.7%), TSHZ3 (15.4%), PCDHA11 (19.2%), ARHGAP28 (15.4%), DST (23.1%), SORCS1 (19.2%), MUC17 (15.4%), PCDH20 (19.2%), USP9X (19.2%), CACNA2D1 (15.4%); body tumors enriched in FLG2 (15.4%), GRM1 (15.4%), SYNE2 (15.4%); cardia tumors enriched in CDH1 (23.1%), GFRAL (15.4%), SBNO1 (15.4%), STAB2 (15.4%) (all P<0.05). After validation cohort expansion, 7 of these (underlined in original) remained significant. None of these anatomic-location associations were observed in the TCGA Russian cohort, suggesting population-specific mutational distribution.
  • Lauren histology: CDH1 somatic mutations (21/294, 7.1%) were enriched in diffuse GC, with 8 of 21 being loss-of-function (nonsense/frameshift). RHOA mutations were likewise enriched in diffuse GC (consistent with Wang et al. 2014 and Kakiuchi et al. 2014). Wnt pathway mutations showed moderate enrichment in intestinal GC. CDH1-mutated GCs had shorter survival while Wnt-mutated GCs had longer survival (log-rank P=0.04); RHOA-mutated cases were not associated with shorter survival, suggesting RHOA-driven and CDH1-driven diffuse GCs are clinically distinct.
  • NRG1–ERBB4 axis: ERBB4 mutated in 7 discovery + 13 validation = 20 cases; NRG1 mutated in 5 discovery + 16 validation = 21 cases. Combined, NRG1 or ERBB4 was mutated in 34 of 294 GC samples (11.6%), with 3 dual-mutated and 31 mutated in only one of the pair. 7/21 (33%) ERBB4 mutations fell in the kinase domain and 4 (20%) in receptor domains. ERBB4 p.R50C overlaps with a melanoma hotspot; TCGA STAD also harbored ERBB4 kinase-domain mutations (p.V744L, p.774N/G) and receptor-domain mutations (p.R106H/C). NRG1 EGF-like-domain mutations (p.A221T, p.A225P, p.E223G, p.R224Q, p.S226P) cluster near the predicted ERBB4-binding site. In 1,000 permutations, ERBB-family and NRG-family mutations were mutually exclusive (P=0.02).
  • BRCA2 and survival: 5/78 discovery + 12/216 validation = 17 BRCA2-mutant cases in Tianjin; no BRCA1 mutations observed. TCGA STAD carried BRCA2 mutations in 28/289 cases. In Tianjin alone, BRCA2 mutation was marginally associated with longer survival (P=0.07, HR=0.29, 95% CI 0.07–1.19); pooling Tianjin + TCGA (n=569, 45 BRCA2 carriers) yielded log-rank P=0.03. After adjusting for age, stage, and cohort, BRCA2 mutation remained an independent predictor of longer survival (HR=0.37, 95% CI 0.13–0.96, P=0.05).
  • BRCA2 mutation profile: Most BRCA2 mutations in GC were missense, contrasting with the predominantly nonsense/frameshift mutations seen in triple-negative breast and ovarian cancer.
  • Clonal evolution (WGS): Pt1 had 4,082 nonsilent mutations (2,809 missense, 143 nonsense, 132 splice-site, 958 frameshift, 40 in-frame indels); Pt2 had 287 nonsilent mutations. Pt1 carried mismatch-repair and chromatin-remodeling mutations in MSH6, TGFBR2, KDM5A, and KMT2D (reported as MLL4); Pt2 carried SETBP1 p.S944N together with AKAP9 and CDK6 amplifications. CACNA1D p.V529G was found in both patients. CNA spanned 516.6 Mb (Pt1) and 706.8 Mb (Pt2); breakpoints 3,812 (Pt1) and 3,390 (Pt2). Phylogenetic tree analysis showed substantial divergence among the three primary-tumor regions, with two lymph-node metastases derived from a common clonal ancestor sharing only one primary-tumor region.

Genes & alterations

  • TP53 — significantly enriched in HiC subtype; present only as minor (<15% VAF) subclones in all 6 positive HiC cases. Defines a poor-survival, older-onset GC subtype.
  • ARID1A — significantly enriched in LoC subtype; associated with younger onset and longer survival.
  • NRG1 — somatic mutations reported here for the first time in GC; 21/294 cases (~7.1%); EGF-like domain hotspot (p.A221T, p.A225P, p.E223G, p.R224Q, p.S226P). Likely activates ERBB4 signaling via increased ligand availability.
  • ERBB4 — mutated in 20/294 cases (~6.8%); kinase-domain (33%) and receptor-domain (20%) mutations; p.R50C shared with melanoma hotspot. With NRG1, defines a candidate lapatinib-targetable subset.
  • ERBB2 — amplification and mutation confirmed in this cohort; standard trastuzumab target.
  • BRCA2 — mutated in 17/294 Tianjin cases (5.8%) and 28/289 TCGA cases; independent predictor of longer survival in pooled cohort (HR 0.37, P=0.05). Predominantly missense, unlike breast/ovarian.
  • BRCA1 — no mutations observed in the 78-case discovery cohort.
  • CDH1 — 21/294 cases (7.1%) somatic mutations, enriched in diffuse-type and cardia tumors; LoF events common; associated with shorter survival.
  • RHOA — enriched in diffuse GC (consistent with PMID:24816253 and PMID:24816255 if present); not associated with shorter survival in this cohort, distinguishing it clinically from CDH1-driven diffuse GC.
  • APC, PIK3CA, SMAD4, MYC, KRAS — recurrently mutated drivers consistent with prior GC studies; participate in Wnt, PI3K-ERBB, and TP53 pathways. Pt1 carried KRAS p.G13D, PIK3CA p.E542K, and BRCA2 p.E3002D/p.G602fs across separate clones.
  • MSH6, TGFBR2, KDM5A, KMT2D — mismatch-repair and chromatin-remodeling defects in the hypermutated Pt1.
  • SETBP1 p.S944N, AKAP9, CDK6 — mutation/amplifications in Pt2; SETBP1 mutation matches the recurrent myeloid-malignancy site.
  • CACNA1D p.V529G — shared between Pt1 and Pt2; voltage-dependent calcium channel.

Clinical implications

  • High-clonality GC needs combination therapy. Subclonal TP53 in HiC tumors implies single-agent regimens may select for resistant minor clones; the authors used DGIdb to enumerate drug targets across subclones (e.g., ALK, ABL2, SMAD4, FANCG, NRG1, KRAS in PGM71 and PGM32). For Pt1 specifically, combined targeting of PI3K, CDK7, and Notch plus immunotherapy is proposed; for Pt2, combined AURKC and CDK6 inhibitors.
  • NRG1–ERBB4 as a lapatinib target. Mutated ERBB4 activates ERBB4 and PI3K-AKT signaling and is inhibitable by the dual TKI lapatinib, currently in a phase III trial in ERBB4-mutated melanoma (NCT01264081). NRG1 overexpression has been shown to confer lapatinib sensitivity in prior work; 11.6% of GCs carry NRG1 and/or ERBB4 mutations, defining a candidate biomarker-selected population beyond the <15% of GC patients eligible for trastuzumab.
  • BRCA2 as a prognostic and predictive biomarker. BRCA2 mutation independently predicts longer survival in pooled Tianjin + TCGA GC (HR 0.37, P=0.05). The authors note that platinum-based chemotherapy is standard front-line for GC, paralleling ovarian cancer where BRCA2 tumors have improved platinum response; this suggests BRCA2 status could guide platinum and PARP-inhibitor decision-making in GC.
  • Anatomic location matters. Antrum/body/cardia tumors carry distinct mutation enrichments; cardia GCs are enriched for CDH1, connecting to the diffuse-type/poor-prognosis axis.
  • Clonal subtype as an independent prognostic factor (adjusted HR 4.69, P=0.0043) — clonality analysis from WES could potentially be developed as a prognostic assay.

Limitations & open questions

  • Single-population bias: All Tianjin cases are northern Chinese; the authors note that the anatomic-location mutation enrichments seen in their cohort did not replicate in the Russian TCGA cohort, suggesting population (genetic, dietary, environmental) effects on mutation distribution. Generalizability to other Asian or Western populations is unknown.
  • HiC group is small: Only 9 HiC tumors define the lethal subtype. The translational significance of TP53 and ARID1A as HiC/LoC markers needs prospective validation.
  • BRCA2 functional impact unknown: Most GC BRCA2 mutations are missense (unlike breast/ovarian), so deleteriousness is presumed but not functionally validated. The mechanism linking BRCA2 status to longer survival in platinum-treated GC is inferred from ovarian cancer biology.
  • NRG1/ERBB4 mutations not functionally tested: No experimental demonstration in this paper that the GC-specific NRG1 or ERBB4 mutations are activating or lapatinib-sensitive; functional validation is required before clinical actionability.
  • Follow-up duration is short: Median follow-up was 13.84 months overall (25.08 months for WES). Long-term survival differences may be under-estimated.
  • Phylogenetic conclusions rest on two patients: Multi-region WGS was done in only Pt1 and Pt2; broader clonal-evolution claims are extrapolations.
  • Treatment heterogeneity: Although the authors note no significant treatment differences between HiC and LoC, treatment regimens are not detailed; residual confounding is possible.

Citations from this paper used in the wiki

  • “we identified a new lethal subtype of GC that is characterized by high levels of clonal heterogeneity” — abstract / Results.
  • “This association remained significant after adjusting for age, sex, stage, Lauren subtype, TP53 mutation, and ARID1A mutation [adjusted hazard ratio (HR), 4.69; 95% confidence interval (CI), 1.62–13.6; P = 0.0043]” — Results, p. 1108.
  • “HiC GCs had a significantly smaller fraction of C-to-T transversion mutation (45% in HiC vs. 52% in LoC) but greater fraction of C-to-G transition mutation (25% in HiC vs. 9% in LoC, P = 0.002)” — Results, p. 1108.
  • “NRG1 and ERBB4 were mutated in 34 GC samples (11.6%), with 3 samples having mutations in both genes and 31 samples having mutations in either NRG1 or ERBB4” — Results, p. 1110.
  • “In 1,000-times permutation, the ERBBs and NRGs’ mutations were mutually exclusive of each other (P = 0.02)” — Results, p. 1111.
  • “After adjusting for age, stage, and different population, BRCA2 mutation remained a significant factor predicting longer survival (P = 0.05, HR = 0.37, 95% CI: 0.13–0.96)” — Results, p. 1111.
  • “all six positive HiC cases” carried TP53 mutations only “in the minor clones (mutation frequency <15%)” — Results, p. 1108.
  • “86 (83%) [of 103 genes] were found to have at least one mutation in the validation cohort” — Results, p. 1110.

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