Document resource
Background Despite improved survival outcomes with immune checkpoint inhibitor (ICI)-based chemotherapy, currently available biomarkers insufficiently capture the immunological heterogeneity of treatment response. Most patients eventually develop progressive disease (PD). This prospective cohort study aimed to comprehensively characterize the dynamic immune landscapes associated with response and resistance in patients with metastatic gastric cancer (GC) treated with ICI-based chemotherapy.Methods Serial baseline and on-treatment peripheral blood mononuclear cells (PBMCs) and endoscopic tumor biopsies were obtained from gastric cancer patients treated with first-line ICI-based chemotherapy (N=90) or chemotherapy alone (N=10). Based on progression-free survival, patients treated with first-line ICI-based chemotherapy were classified into early-progressors (EP) vs. non-early progressors that were further divided into stable controller and long-term responder (LR). We performed single-cell RNA sequencing (scRNA-seq) of PBMCs (n=30), single-nuclear RNA sequencing (snRNA-seq) of tumor tissues (n=20), and spatial transcriptomics (ST) with Visium (n=60) and Xeinum 5K (n=16) platforms to interrogate immune-related mechanisms of response and resistance. Flow cytometry was used for validation of specific immune cell subsets.Results scRNA-seq analyses of PBMCs revealed enhanced interferon-γ signaling and CD8 + T cell activation in LR, particularly enriched in GZMK+ effector memory CD8+ T cells (Tem), which proliferated upon ICI exposure. In contrast, EP exhibited upregulation of inflammatory pathways and enrichment of S100A8/9 + classical monocytes, both in scRNA-seq and validated by flow cytometry. snRNA-seq analyses of tumor biopsies showed distinct transcriptional programs and immune cell interactions across response groups. Spatial transcriptomics identified discrete cellular communities (‘spot clusters’) reflecting the immune contexture of response: In EP tumors, neutrophil- and IL1B + macrophage-enriched regions adjacent to tumor cells were linked to resistance and epithelial-mesenchymal transition. Among the four gene programs (GPs) identified by cNMF of snRNAseq data, the metastatic EMT program (GP2) was enriched in EP and spatially co-localized with neutrophil-enriched regions forming inflammatory niches, supporting the link between inflammatory myeloid niches and tumor invasiveness. In contrast, tumors from non-early progressors were characterized by expanded lymphoid aggregates expressing markers of tertiary lymphoid structures (TLS), harboring GZMK + CD8+ Tem and CD8+ T cell-recruiting tumor-associated macrophages (CD8+T_TAM), forming a distinct CD8+ T-immune hub. Notably, the abundance of the CD8+ T-immune hub inversely correlated with that of the inflammatory niche, highlighting a spatial dichotomy underlying therapeutic response.Conclusions The identification of IFN-γ-driven CD8 + T cell responses and inflammatory monocyte-neutrophil niches, together with metastatic EMT gene programs, provides a refined framework for predicting therapeutic outcomes and guiding future biomarker development in metastatic gastric cancer.Acknowledgements This research was a collaborative effort made possible through support from F. Hoffmann-La Roche Ltd.Ethics Approval The protocol was approved by the institutional review board of Asan Medical Center (No.2023-0896).