Next-generation human vascularized tumor models reveal HER2-independent efficacy and reduced vascular toxicity of T-DXd

Purpose: Antibody-drug conjugates (ADCs) elicit complex responses that conventional 2D assays fail to capture, especially in HER2-low and HER2-negative tumors. We developed human vascularized 3D tumor models to assess the efficacy of HER2-targeting ADC, their mechanism of action, and vascular toxicity with improved physiological relevance. Methods: Vascularized 3D breast cancer microtissues (HER2-positive and HER2-negative) were generated using human breast cancer cell lines and microvascular networks. Trastuzumab deruxtecan (T-DXd) or Trastuzumab-emtansine (T-DM1) treatment was performed for four days. Real-time monitoring was performed to assess anti-tumor efficacy and vascular toxicity by quantifying the tumor cell-associated fluorescent signal and vascular area density. Results: T-DXd showed significant cytotoxicity in both HER2-positive and HER2-negative vascularized tumors, whereas in 2D monocultures, T-DXd cytotoxicity against HER2-negative cells was not detected. T-DM1 was effective only in HER2-positive vascularized tumors. T-DM1, unlike T-DXd, induced a strong vascular toxicity in both HER2-positive and HER2-negative tumor models with vascularization. Conclusions: Human vascularized tumor models capture key pharmacodynamic features of ADCs, including the HER2-independent mechanism of action, extracellular payload activity, and vascular toxicity, which are not recapitulated in 2D systems. HER2-independent efficacy of T-DXd, but not of T-DM1, has also been reported in in vivo breast cancer xenograft models in agreement with these findings. Furthermore, real-time monitoring enabled quantification of tumor regression and vascular remodelling, underscoring the functional relevance and predictive performance of the platform. This platform offers a rapid, quantitative, and predictive approach for evaluating next-generation ADCs, thereby enhancing translational alignment with in vivo outcomes. AI disclosure: AI-assisted text generation was used to revise this abstract.