The 27-gene IO score is associated with molecular features and response to immune checkpoint inhibitors (ICI) in patients with gastric cancer.

Abstract

4058 Background: Gastric cancer (GC) is the 3rd leading cause of cancer-related death worldwide. Unfortunately, most gastric cancer patients are asymptomatic until the cancer has progressed to an advanced stage. ICIs have improved patient outcomes in a variety of cancers, including GC. A variety of biomarkers have been used to identify patients most likely to benefit from ICI therapies such as high PD-L1 expression, MSI-high, Epstein Barr virus (EBV) positive, or TMB. Despite these potential biomarkers, most patients with advanced GC do not respond to ICI treatment Thus, there remains an unmet clinical need for a biomarker that can better predict response to ICI therapies. Herein, we demonstrate that the 27 gene IO score, a tumor immune microenvironment (TIME) classifier is associated with the existing molecular markers of gastric cancer and with objective response to ICI therapy in a clinical cohort. Methods: RNA-seq expression data was obtained from 3 independent cohorts including TCGA (STAD), ACRG (GSE84437, GSE84426), and clinical cohort with ICI response data (PRJEB25780, PRJEB40416). The 27 gene IO algorithm was applied to all available patient data to derive IO scores. Fisher’s exact test was used to examine the associations between IO score and clinical features and molecular subtypes in each cohort. R (version 4.1.2) was used to calculate ORs with 95%CIs, and ordinal logistic regression modeling. Results: From the TCGA cohort, the IO score was associated with the molecular features of EBV, MSI, TMB, and PD-L1 (n = 135, p < 0.05 for all). Similarly, in the ACRG cohorts, the IO score was significantly associated with EBV, MSI, and PD-L1 ( n = 294, p < 0.001 for all). To determine whether the IO score was associated with response to ICIs, we examined a cohort of Korean patients with advanced stage GC curated by Kim et. al. In this cohort of 59 patients, the IO score was associated with ICI response (Fisher’s exact test, p < 0.05). When response was grouped by responders vs. non-responders (CR/PR vs SD/PD), the odds ratio for the association between IO score and response was 5.3 (95% CI: 1.3 to 23.92, p = 0.01). The linearity of continuous value of the IO score was indicative of a direct relationship between IO score and improved objective response (ordinal logistic regression, t = 2.59, p < 0.01). Conclusions: PD-L1 and TMB have shown marked levels of both spatial and temporal heterogeneity in GCs, thus there exists a need for a more comprehensive biomarker that can fully assess the TIME. The 27 gene IO score is associated with many existing biomarkers in GC and has now been shown to also be associated with response to ICIs. As such, further studies are warranted to demonstrate that the 27 gene IO score may be a more comprehensive biomarker for assessing the TIME and provide complementary data to tumor-specific biomarkers, which together could aid in clinical decision making for ICI treatment of GCs.