CELL INTELLIGENCE · 肿瘤细胞治疗研究

Chimeric antigen receptor (CAR) T-cell therapy has revolutionised cancer gene therapy, yet its expansion into solid tumours is hindered by a critical vulnerability: the autonomous, "always-on" nature of conventional CAR constructs. This unregulated activity drives severe toxicities, including cytokine release syndrome (CRS) and on-target/off-tumour damage, while constitutive signalling in hostile tumour microenvironments (TMEs) accelerates T-cell exhaustion. Early safety strategies relied on irreversible genetic "kill switches," which sacrifice the therapeutic cell population entirely. This review traces the conceptual evolution of CAR T-cell controllability from binary elimination towards platforms enabling graded, reversible, and spatiotemporally precise regulation. We examine the transition from calibrated signalling architectures and small-molecule-regulated split-CARs to advanced optogenetic and sonogenetic controllers, detailing the biophysics of photoreceptor pairs and their preclinical efficacy. Furthermore, we explore complementary architectures, including autonomous logic-gated receptors. Finally, we propose that the optimal next-generation CAR T product will integrate calibrated signalling, external control, and context-dependent armouring to achieve truly programmable, safe, and durable cellular immunotherapy.