CELL INTELLIGENCE · 肿瘤细胞治疗研究

Chimeric antigen receptor T-cell (CAR-T) therapy has achieved unprecedented success in hematological malignancies but faces formidable challenges in solid tumors. These limitations include severe "on-target, off-tumor" toxicity, antigen heterogeneity, and the immunosuppressive tumor microenvironment, where the dense extracellular matrix acts as a physical barrier hindering T-cell infiltration. To address these hurdles, this review proposes a comprehensive Efficacy, Safety, and Accessibility (ESA) framework for engineering next-generation "Smart" CAR-T cells. We explore the implementation of programmable Boolean logic gates (AND, OR, NOT) and conditional activation systems (e.g., synNotch, focused ultrasound) that allow T cells to compute antigen patterns and precisely sense tumor-specific cues. Furthermore, we examine the development of off-the-shelf allogeneic platforms that utilize advanced gene editing technologies-such as CRISPR-Cas9 and base editing-to eliminate endogenous receptors and prevent graft-versus-host disease. Crucially, we highlight the transformative potential of Artificial Intelligence/Machine Learning in accelerating the "Design-Build-Test-Learn" cycle, from optimizing single-chain variable fragment (scFv) affinity to predicting clinical toxicity risks. By integrating these multi-dimensional strategies, we outline a new paradigm in precision immunotherapy, aiming to transform CAR-T cells into intelligent, controllable, and universally accessible living drugs capable of eradicating complex solid tumors.