一种用于IO治疗药物疗效和毒性预测测试的芯片上先进3D血管化肿瘤免疫微环境

Introduction: Conventional pre-clinical models for immuno-oncology (IO) therapeutics, such as animal models, suffer from irrelevant biology and ecosystems, while simple 2D/3D in vitro assays lack the complexity of the human tumor microenvironment. These gaps lead to poor prediction of clinical outcomes, including immune-related toxicities. To address this translational gap, we developed and validated an advanced, patient-relevant 3D vascularized tumor immune microenvironment (vTIME) on-a-chip model for the predictive assessment of IO drug efficacy and toxicity-a platform now proven to generate data packages supporting Investigational New Drug (IND) approval. Methods: Within proprietary microfluidic CurioChips, we established the vTIME model by co-culturing patient-derived organoids in hydrogel with stromal cells and endothelial cells (HUVECs). In this co-culture, endothelial cells are self-assembled into a perfusable vascular network. We subsequently introduced human monocytes, which migrated into the tumor bed and differentiated to build an immunosuppressive microenvironment. The transformation of immune microenvironment was characterized by multiple cytokine changes. We then validated the model's predictive utility using a clinical stage T-cell engager (TCE) for dual efficacy and toxicity evaluation. The tumor killing and T cell activations are measured by immunofluorescent images, immune cell profiling and secretome analysis. Results: The model successfully recapitulated a clinically relevant, immunosuppressive TME: monocytes differentiated into tumor-promoting macrophages, fibroblasts converted to CAF-like cells, and metastatic organoids demonstrated vascular invasion. For efficacy testing, TCE-mediated cancer-killing was observed in a dose-dependent manner; however, this efficacy was significantly diminished in the presence of immunosuppressive macrophages, demonstrating the model's ability to avoid the overestimation of efficacy seen in simpler systems. For toxicity assessment, the complete vTIME model captured the key cascade of cytokine release syndrome (CRS). TCE treatment induced a characteristic kinetic, with an initial IFNgamma peak followed by a substantial IL-6 elevation. Critically, we also captured significant, CRS-linked vessel toxicity post-TCE treatment, a key initiating event of organ failure in patients. Conclusion: These findings demonstrate that our advanced 3D vTIME-on-a-chip platform represents one of the powerful New Approach Methodologies (NAMs), which is reliable and predictive for the dual assessment of novel IO therapeutics. It is capable of capturing complex immune-mediated responses, identifying the therapeutic "sweet spot" balancing efficacy and toxicity, and modeling CRS-driven vascular damage not seen in conventional systems.