Recent evidence shows that Streptococcus anginosus (S. a) promotes the development of gastric cancer (GC). However, its role in GC progression, particularly its impact on tumor immune microenvironment (TIME), remains unclear. This study aims to investigate the role of S. a in advanced GC and test the hypothesis that S. a upregulates PD-L1 expression in GC cells, thereby inhibiting CD8⁺ T cell function.

Based on in-house RNA sequencing (RNA-seq) raw data (n=98), a computational pipeline was established to determine the relative abundance of S. a (S. a score). In parallel, a machine learning-based signature gene model was constructed to infer S. a abundance from gene expression matrices of public cohorts. Using these approaches, GC patients were stratified into S. a-high (S. a^high) and S. a-low (S. a^low) groups according to the median abundance. Single-cell RNA sequencing (scRNA-seq) data (n=23) were integrated to analyze differences in cellular composition within the TIME between the 2 groups. For in vitro validation, human (NUGC-3) and murine (TKM) GC cell lines were treated with S. a for 8 h, followed by quantification of PD-L1 expression via qPCR and Western blotting. Cytotoxic function of CD8⁺ T cells against S. a-pretreated GC cells was evaluated by flow cytometry in co-culture systems. where S. a-pretreated cancer cells were incubated with CD8⁺ T cells, and cytotoxicity was measured by flow cytometry. For in vivo validation, a C57BL/6 mouse orthotopic GC model was established; the experimental group received S. a bacterial suspension via oral gavage, while the control group received sterile culture medium, with continuous intervention for 6 weeks (n=6). Tumor burden was evaluated by in vivo imaging, tumor volume, and stomach weight.

The abundance of S. a was significantly higher in GC tissues than in normal gastric mucosa (P=0.000850), and high abundance was associated with poor prognosis (HR=1.49, 95%CI: 1.09 to 2.04, P=0.0139). Analysis of the TIME using scRNA-seq data revealed distinct cellular compositions between GC tissues with varying S. a levels. Notably, S. a^high GC tissues exhibited a significant increase in CD8⁺ T cell infiltration (P<0.0001), yet these cells predominantly displayed an exhausted phenotype (P<0.0001). Further analysis identified a subset of GC cells whose abundance is correlated positively with both S. a levels (Spearman’s rho=0.288, P<0.0001) and proportion of exhausted CD8⁺ T cells (Spearman’s rho=0.522, P<0.0001), suggesting their potential roles in modulating CD8⁺ T cell function. Mechanistically, S. a treatment significantly upregulated PD-L1 expression in GC cells (P<0.0001). In co-culture assays, S. a-pretreated cancer cells impaired the cytotoxic function of CD8⁺ T cells (P=0.000526). In vivo results demonstrated that the mice in the S. a gavage group exhibited significantly higher tumor bioluminescence intensity (P=0.0306), larger tumor volume (P=0.0453), and increased stomach weight (P=0.0306) compared to the control group.

Enriched S. a abundance not only reshapes the TIME of GC, but also mediates CD8⁺ T cell exhaustion and facilitates tumor immune evasion by inducing high PD-L1 expression in GC cells.