Dismantling the Stromal‐Metabolic Barrier in TNBC via Machine Learning‐Guided Pan‐Immune Activity Nano‐Regulator

ABSTRACT

Triple‐negative breast cancer establishes a stromal‐metabolic barrier, characterized by dense fibrosis and metabolic dysregulation, to induce immunotherapy resistance. To dismantle this barrier, we propose a P rogrammable R ewiring of I mmuno‐ S tromal M etabolism ( PRISM ) strategy, supported by a POD‐like activity‐oriented machine learning (ML) screening of representative catalytic metals. A metabolic nanoregulator, Au@Fe–Co@HA (ACFH), was engineered to couple glucose starvation with Fe/Co‐mediated oxidative therapy. The Au core of ACFH acted as a glucose oxidase‐like enzyme to deplete glucose and generate endogenous H ₂ O ₂ , thereby powering the peroxidase‐like activity of the iron‐cobalt (Fe/Co) shell. In vivo, ACFH‐mediated combined therapy reduced GLUT1 expression in tumor tissues and suppressed glucose uptake and glycolysis‐associated metabolic activity. The GSH‐responsive release of Fe/Co ions disrupted redox homeostasis and activated immunogenic cell death (ICD), facilitating dendritic cell maturation, M1 macrophage polarization, and CD4 ⁺ /CD8 ⁺ T‐cell infiltration. When combined with immune checkpoint blockade, the therapeutic efficacy was further enhanced. In this study, ML was used to translate the therapeutic requirement for ROS‐amplified ferroptosis and immunogenic therapy into a quantifiable catalytic screening endpoint. By integrating this function‐oriented metal selection with the PRISM strategy, this work provides a rational nanoregulator design to overcome stromal‐metabolic resistance in TNBC.