T-bet+CD8+ T cells govern anti-PD-1 responses in microsatellite-stable gastric cancers

Abstract

More than 90% of advanced gastric cancers (GC) are microsatellite-stable (MSS). Compared to the high response rate of immune checkpoint inhibitors (ICI) in microsatellite-instability-high (MSI-H) GCs, only 10% of unstratified MSS GCs respond to ICIs. In this study, we apply semi-supervised learning to stratify potential ICI responders in MSS GCs, achieving high accuracy, quantified by an area under the curve of 0.924. Spatial analysis of the tumor microenvironment of ICI-sensitive GCs reveals a high level of T-bet+ CD8 + T cell infiltration in their tumor compartments. T-bet+ CD8 + T cells exhibit superior anti-tumor activity due to their increased ability to infiltrate tumors and secrete cytotoxic molecules. Adoptive transfer of T-bet+ CD8 + T cells boosts anti-tumor immunity and confers susceptibility to ICIs in immune-ignorant MSS GCs in a humanized mouse model. Spatial RNA sequencing suggests a positive-feedback loop between T-bet+ T cells and PD-L1+ tumor cells, which eventually drives T cell exhaustion and can therefore be leveraged for ICI therapy. In summary, our research provides insights into the underlying mechanism of anti-tumor immunity and deepens our understanding of varied ICI responses in MSS GCs. Microsatellite-instability-high gastric cancers typically respond to immune checkpoint inhibitors, but only a small proportion of patients in the microsatellite-stable group responds. Here authors stratify these patients via a semi-supervised learning approach using multi-omics input data and functionally validate T-bet+ CD8 + T cells as crucial factor in the immunotherapy response.