Unbiased preclinical phenotyping reveals neuroprotective properties of pioglitazone

Abstract

Animal models are essential for assessing the preclinical efficacy of candidate drugs, but animal data often fails to replicate in human clinical trials. This translational gulf is due in part to the use of models that do not accurately replicate human disease processes and phenotyping strategies that do not capture sensitive, disease-relevant measures. To address these challenges with the aim of validating candidate neuroprotective drugs, we combined a mouse prion (RML scrapie) model that recapitulates the key common features of human neurodegenerative disease including bona fide neuronal loss, with unbiased and machine learning-assisted behavioural phenotyping. We found that this approach measured subtle, stereotyped, and progressive changes in motor behaviour over the disease time course that correlated with the earliest detectable histopathological changes in the mouse brain. To validate the utility of this model system, we tested whether the anti-diabetic drug pioglitazone could slow prion disease progression. Pioglitazone crosses the blood-brain-barrier and has been shown to reduce neurodegenerative disease severity in other mouse models. We found that in addition to significantly slowing the emergence of early-stage clinical signs of neurodegeneration, pioglitazone significantly improved motor coordination throughout the disease time course and reduced neuronal endoplasmic reticulum stress. Together, these findings suggest that pioglitazone could have neuroprotective properties in humans, confirm the utility of the scrapie mouse model of neurodegeneration, and provide generalisable experimental and analysis methods for the generation of data-rich behavioural data to accelerate and improve preclinical validation.