Chemotherapy-induced peripheral neuropathy (CIPN) is a major common adverse event associated with neurological abnormalities. In the present study, we developed a microfluidic device for in vitro neuronal culture, and predict neurotoxicity induced by anti-cancer drugs with different mechanisms of action based on deep learning morphological analysis.
The microfluidic culture device could separate the cell body and neurites, so that the influence on soma or axon can be analyzed independently. COP (Cyclo olefin polymer), which has excellent observability and low drug adsorption, is used as the resin material, and the bottom surface is created thin and flat enough for a clear view by microscope. Next, primary DRG neurons was cultured in the device coated with Poly-L-lysine and Laminin. After culturing with a specific medium containing insulin, neurites grew sufficiently to occupy almost the whole microfluidic channel area, and the axon elongated unidirectional along the horizontal direction.
After administration of several typical anti-cancer drugs, a deep learning AI was trained with immunofluorescence image datasets of either soma or neurites. As results, AI could accurately detected toxicity positive for both soma and neurites, even at low concentrations. Furthermore, the effects of drugs on either soma or neurites could be significantly separated by principal component analysis using toxicity positive rate calculated by AI. Therefore, this method provides an effective in vitro toxicity assessment platform for peripheral neuropathies.