Automated, High-throughput Analysis of Neurite Dynamics in Neurodegenerative Disease

Despite decades of research, there are no effective therapies for neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) and Parkinson's Disease (PD). These diseases are marked by a progressive loss of the neuronal processes referred to as neurites and the cell bodies. One obstacle to developing therapies is the uncertainty in translating improvements in disease model system into humans. The measures in disease models cannot match the scope of measures in a patient, and they are further limited by practical considerations on the number of timepoints and measurements to explore a larger space of possible therapeutic interventions. The failure of the previous approaches makes it important to find a new approach with the promise of more relevant measures. We have developed a fully-automated system for high-throughput robotic microscopy and an automated image analysis pipeline for our longitudinal datasets that incorporates supervised machine learning. We use this system to extend our measurements to include the quantification of neurite dynamics. The positional accuracy of our robotic microscopy combined with supervised machine learning enables us to measure neurite extension and retraction over time. In an ALS model, we have shown that the decrease in neurite area is mainly due to an increase in neurite retraction. We compared our measures of neurite dynamics to cell survival and demonstrated that measures of neurite dynamics are superior at separating the healthy and disease conditions, and they can quantify differences between the conditions that would not be detectable using survival analysis. We investigated the ability of measures of neurite dynamics to distinguish between an ALS and PD disease model. While accuracy of the PD model was lower than the ALS model, the parameters that contributed to the accuracy were very differently ranked, indicating there is not a single useful neurite measure for differentiating diseases. We believe the robustly and longitudinally detecting neurites will give us valuable insights into mechanistic and therapeutic studies of neurodegeneration