489 Background: Epstein-Barr virus (EBV)-associated gastric cancer (EBVaGC) accounts for 5-10% of all gastric cancers and has a relatively favorable prognosis compared to other subtypes. While the genomic profile of EBVaGC has been previously reported to exhibit extensive CpG island methylation, elevated levels of PD-L1/2, and an absence of TP53 mutations, transcriptomic features remain less explored. The response to immunotherapy of EBVaGC has been various. This study comprehensively investigates the transcriptomic characteristics of EBVaGC in relation to PD-L1 status. Methods: RNA sequencing was conducted on formalin-fixed, paraffin-embedded samples from 27 EBV-positive and 12 EBV-negative patients with HER2-negative gastric cancer. Gene expression was quantified using the RSEM package. Predict-IO, a machine learning-based predictive algorithm developed at Auristone Pte Ltd based on initial work done at the Genome Institute of Singapore (GIS), assessed the likelihood of a patient responding to immunotherapy based on transcriptome expression data. PD-L1 immunohistochemistry was performed using the 22C3 pharmDx assay, and status was classified by CPS. Results: A total of 200 differentially expressed genes (DEGs; log2FC > 2, FDR < 0.05) were identified in EBV-positive versus negative samples, with 157 genes down-regulated and 43 genes up-regulated. Upregulated genes included immune response related genes such as DMBT1, IDO1, PLA2G2A, and IL13RA2, while downregulated genes were primarily associated with cell adhesion, including LAMA1 and CCN6. In the EBV positive group, immune response pathways were enriched; however, within the PD-L1-Low group (CPS < 1), activated signaling pathways such as MYC, E2F, and MTORC1 were observed. Conversely, the PD-L1-High group (CPS ≥ 5) exhibited upregulated snoRNAs. In the EBV negative group, the PD-L1-Low cohort displayed suppressed immune response and ECM remodeling factors, whereas the PD-L1-High cohort showed enrichment of ECM remodeling factors. The Predict-IO Score, which predicts immunotherapy response, was significantly higher in EBV-positive compared to EBV-negative samples ( p = 0.003), and scores were elevated in the PD-L1-High group compared to the PD-L1-Low group. Conclusions: In summary, EBVaGC includes subsets of different PD-L1 expression patterns with distinct activated signaling pathways, differences in ECM remodeling factors, and varying Predict-IO Scores. With these information, we could properly select the responsive patients to immunotherapy among EBVaGC.
Read full abstract