Abnormal release of mitochondrial DNA (mtDNA) and innate immune sensing play key roles in tumor immune regulation, but how they work together with ferroptosis is still unclear. This study aims to clarify the molecular mechanism of how mtDNA reshapes the tumor immune microenvironment in coordination with ferroptosis through activating the cGAS-STING signaling pathway, as well as its anti-tumor potential. By creating tumor cell mitochondrial stress models and using CRISPR-Cas9 gene knockout, lipid peroxidation detection, single-cell transcriptome analysis, and mouse syngeneic tumor models, we systematically assessed how mtDNA release affects the cGAS-STING pathway and ferroptosis. The results showed that mitochondrial stress-induced cytoplasmic mtDNA release significantly activates the cGAS-STING pathway, upregulates type I interferons and chemokines like CXCL10, and promotes lipid peroxidation and ferroptosis by suppressing the xCT/GPX4 antioxidant axis. Ferroptotic tumor cells further release damage-associated molecules like HMGB1 and ATP, enhancing dendritic cell maturation and CD8⁺ T cell tumor infiltration, forming a positive feedback loop of immune activation. In mouse melanoma models, combining mtDNA release with ferroptosis significantly inhibited tumor growth and extended survival, showing better results than either intervention alone. This study reveals a new mechanism where mtDNA sensing collaborates with ferroptosis via the cGAS-STING pathway to reshape the tumor immune microenvironment, providing a theoretical basis and potential targets for developing anti-tumor immunotherapies that jointly regulate innate immunity and ferroptosis.
Gang Liu, Jia Wang, Li-Qiang An· Global Health Care· 0 citations
The effectiveness of immune checkpoint inhibitors (ICIs) heavily depends on the complex spatial interactions of cells within the tumor microenvironment (TME). However, existing predictive biomarkers generally lack spatial resolution and interpretability, making it difficult to guide clinical decisions or reveal actionable targets. This study aims to develop an interpretable spatial AI framework to systematically decode the tumor-immune spatial architecture, achieve high-accuracy predictions of immunotherapy responses, and identify new immunotherapy targets. We collected pre-treatment tumor samples from melanoma and non-small cell lung cancer patients across multiple centers, simultaneously obtaining spatial transcriptomics and multiplex immunofluorescence data. We developed a deep learning framework called “SpaImmune,” which integrates graph attention networks and visual transformer architectures, modeling hundreds of thousands of cells based on their real spatial coordinates as a heterogeneous graph network to automatically learn multi-scale features of the immune microenvironment. The model’s explainability module uses attention weight mechanisms and SHAP values to quantify each spatial component’s contribution to predictions and extract higher-order interaction rules. In three independent validation cohorts, SpaImmune predicted the objective response to ICIs with area under the curve (AUC) values all over 0.91, significantly outperforming methods based on immune cell density, PD-L1 expression, and traditional machine learning. Explainability analysis revealed that the tight spatial coupling of B cells and CD4⁺ follicular helper T cells in tertiary lymphoid structures (TLS) is the strongest feature for predicting treatment response. More importantly, by scanning the cell-gene spatial colocalization network built by the model, we discovered a molecule called SLAMF7, previously unreported as an immune target, which is highly expressed specifically in immune-active regions and associated with good prognosis. Subsequent in vitro functional tests confirmed that activating SLAMF7 can enhance CD8⁺ T cell tumor-killing function, while blocking it weakens immunotherapy effects. Explainable spatial AI can faithfully reveal the intrinsic logic of tumor-immune interactions, not only greatly improving predictive accuracy for immunotherapy but also directly guiding rational discovery of new targets, offering a whole new paradigm for spatial intelligence-driven precision immuno-oncology.
Gang Liu, Jia Wang, Jia Zhu· Intelligent & Human Futu...· 0 citations
This work is the first to use a multi-modal foundation model for neoantigen vaccines, overcoming the tumor-type-specific constraints of conventional approaches and facilitating joint modelling of immunogenicity and clinical efficacy, thereby offering an AI decision engine for precision cancer vaccine design that is applicable to all cancer types.
Gang Liu, Jia Wang, Jia Zhu· Global Health Care· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
