Genomic Landscape of Late-Stage Gastric Cancer: Analysis From KEYNOTE-059, KEYNOTE-061, and KEYNOTE-062 Studies

Abstract

Purpose: The Cancer Genome Atlas (TCGA) classifies gastric cancer (GC) into four molecular subtypes: Epstein-Barr virus-positive, microsatellite instability-high (MSI-H), genomically stable (GS), and chromosomal instability (CIN). This exploratory analysis compared the genomic landscape of late-stage GC from KEYNOTE-059, KEYNOTE-061, and KEYNOTE-062 studies with early-stage GC from TCGA and evaluated the genomic characteristics of late-stage GC in patients of Western and Asian origin.

Materials and methods: Using pretreatment tumor samples, bulk DNA was analyzed via whole-exome sequencing (WES; KEYNOTE-059/KEYNOTE-061) and FoundationOneCDx (KEYNOTE-062) to determine TCGA-defined molecular subtypes (only MSI-H is determinable from FoundationOneCDx), genomic alterations, homologous recombination deficiency (HRD), and tumor mutational burden (TMB); gene expression signatures were analyzed using RNA sequencing.

Results: When comparing KEYNOTE-059/061/062 combined WES and FoundationOneCDx data with data from TCGA, the MSI-H subtype prevalence was numerically lower in patients of Western (5% v 22%) and Asian origin (5% v 19%). When comparing KEYNOTE-059/061 WES data with the TCGA data set, the GS subtype prevalence was numerically higher (36% v 21%) in patients of Western or Asian origin. Among subtypes in KEYNOTE-059/061, HRD scores and TMB trended highest in CIN and MSI-H subtypes, respectively. TP53 mutation was the most prevalent genomic characteristic per KEYNOTE-059/061/062 combined analysis in patients of Western or Asian origin. Gene expression signature distributions were generally similar between patients of Western and Asian origin.

Conclusion: Numerical differences in the prevalence of MSI-H and GS subtypes were observed between early-stage and late-stage GC. Genomic characteristics of late-stage GC were generally similar between patients of Western and Asian origin.

Trial registration: ClinicalTrials.gov NCT02494583 NCT02335411 NCT02370498.

FIG 1.

OncoPrint for gastric cancer data from (A) the KEYNOTE-059/061 combined analysis using WES (n = 464) and (B) the KEYNOTE-062 data analyzed using FoundationOneCDx (n = 279). The OncoPrints were generated using the ComplexHeatmap (version 2.6.2) package of R (version 4.0.5). CIN, chromosome instability; CPS, combined positive score; dMMR, mismatch repair deficient; EBV, Epstein-Barr virus; GS, genomically stable; HRD, homologous recombination deficiency; MMR, mismatch repair; MSI-H, microsatellite instability–high; pMMR, mismatch repair proficient; Tcell_infGEP, T-cell–inflamed gene expression profile; TMB, tumor mutational burden; WES, whole-exome sequencing.

FIG 2.

Genomic scores by TCGA subtype in the KEYNOTE-059/061 combined analysis: (A) HRD score and (B) TMB score. CIN, chromosome instability; EBV, Epstein-Barr virus; GS, genomically stable; HRD, homologous recombination deficiency; MSI-H, microsatellite instability–high; TCGA, The Cancer Genome Atlas; TMB, tumor mutational burden.

FIG 3.

Prevalence of select genomic features: (A) KEYNOTE-059/061 combined analysis and (B) KEYNOTE-062 analysis. amp, amplification; MSI-H, microsatellite instability–high; TMB-H, high tumor mutational burden.

FIG 4.

CLDN18 gene expression. (A) Expression by subtypes for TCGA data set (n = 232). (B) Expression in the presence of CLDN18 fusion and wild-type genes in the overall KEYNOTE-059/061 combined analysis (n = 333) and TCGA data set (n = 284). The AUROC analysis was performed using pROC package in R and plots were visualized using the ggplot2 package. AUROC, area under the receiver operating characteristics curve; CIN, chromosome instability; EBV, Epstein-Barr virus; GS, genomically stable; MSI-H, microsatellite instability–high; TCGA, The Cancer Genome Atlas; WT, wild-type.

FIG 5.

Gene expression signatures in gastric cancers from TCGA (n = 169) and KEYNOTE-059/061/062 combined analysis (n = 472) by race. This scatter plot presents the AUROC of patients of Western origin versus patients of Asian origin in TCGA (x-axis) and combined KEYNOTE (y-axis) data. Analysis of the TCGA data set was restricted to stage 3 and 4 tumors. AUROC >0.5 means the signature has higher expression in patients of Western origin versus patients of Asian origin. Statistically significant nominal P values were observed for the expression of the hypoxia signature (red) in both the combined KEYNOTE (AUROC, 0.67; P < .001) and TCGA (AUROC, 0.64; P = .008) data sets. Statistically significant nominal P values were also observed in the KEYNOTE-059/061/062 combined analysis data set for the expression of the glycolysis (blue; AUROC, 0.63; P < .001), angiogenesis (AUROC, 0.60; P < .001), and gMDSC (blue; AUROC, 0.59; P = .003) signatures (blue). The dotted box around AUROC = 0.5 separates variables that did not show a significant nominal P value (shown in black) versus those that showed a significant nominal P value (shown in red and blue). AUROC, area under the receiver operating characteristics curve; EMT, epithelial-mesenchymal transition; gMDSC, granulocytic myeloid-derived suppressor cell; mMDSC, monocyte myeloid-derived suppressor cell; Tcell_infGEP, T-cell–inflamed gene expression profile; TCGA, The Cancer Genome Atlas; TGF-β, transforming growth factor-beta.