A nutrient-responsive cancer-on-chip model for assessing chemotherapy responses under glucose restriction

Fasting-mimicking diets have emerged as promising adjuvant strategies to enhance chemotherapy efficacy. However, conventional preclinical models do not adequately reproduce the microenvironmental and transport-related conditions associated with nutrient modulation, limiting mechanistic investigation of fasting-related therapeutic responses. Here, we developed a nutrient-responsive cancer-on-chip (CoC) platform integrating 3D human breast cancer models with systemic-like cisplatin administration under defined glucose conditions (high, low, no glucose). Complete glucose deprivation markedly reduced metabolic activity and cell viability, while mild reduction was largely tolerated, reflecting tumor metabolic adaptability. Tumor models were exposed to glucose modulation and cisplatin, with or without 24 h glucose-free pre-conditioning, in static and dynamic 3D systems. Static cultures displayed limited treatment response, whereas CoC exhibited significantly increased cell death, highlighting the limited predictive capacity of conventional models. Glucose-free pre-conditioning enhanced chemotherapy sensitivity, while fasting initiated concurrently with treatment produced minimal impact. Glucose reintroduction after glucose-free pre-conditioning impaired metabolic recovery, revealing a transient vulnerability window. Importantly, in healthy fibroblasts, glucose deprivation reduced metabolic activity without markedly affecting viability. A qualitative comparison with published clinical data from fasting-mimicking diet trials indicated that dynamic CoC models reproduced several treatment-related response trends more consistently than static systems. These findings suggest the potential of nutrient-responsive CoC systems as complementary tools for investigating context-dependent responses to fasting-associated therapeutic strategies.