Abstract

Metastatic dissemination is initiated by tumor cells interpreting spatially organized biochemical and biophysical cues that remain difficult to reproduce using conventional migration assays. Here, we developed a computationally guided metastasis-on-a-chip (MET-on-a-chip) platform based on the concept of the Minimally Functional Unit (MFU), in which only the biological components required to answer a defined experimental question are incorporated. The platform consists of two independent culture chambers connected through an array of confined microchannels that permits diffusion of soluble factors while constraining tumor cell migration. Rather than relying on empirical optimization, finite-element COMSOL simulations were first used to predict molecular transport, define growth factor loading conditions, identify biologically relevant exposure regions, and guide the rational design of the microfluidic assay. Computational predictions were experimentally validated using 70-kDa FITC-dextran diffusion and VEGF release studies, confirming the formation of stable spatial concentration gradients across the microfluidic platform. The simulations further demonstrated that both growth factor loading and cell positioning relative to the predicted gradients critically influenced assay performance, leading to the optimization of the platform through spatial reconfiguration of the tumor compartment. Using the optimized configuration, we compared the migratory responses of neuroblastoma, Ewing sarcoma, and osteosarcoma cells to vascular (VEGF-A165) and lymphatic (VEGF-C) chemotactic cues. VEGF-C significantly increased migration through the microchannel array in Ewing sarcoma and osteosarcoma cells, whereas VEGF-A165 produced no significant effect. In contrast, neuroblastoma cells exhibited minimal migration under either condition, revealing tumor-specific differences in responsiveness to VEGF signaling. Together, these findings establish a computationally guided workflow for the rational design of metastasis-on-a-chip assays, in which predictive modeling informs experimental design before biological validation. By substantially reducing empirical trial-and-error while enabling quantitative control over growth factor exposure, this strategy provides a robust framework for developing minimally functional microphysiological systems capable of dissecting individual steps of the metastatic cascade under experimentally defined conditions.