Prediction and Learning for Postural Control in Mice

Abstract

Postural control is crucial for maintaining balance and body position in response to both self-generated and externally triggered perturbations. Studies in humans have shown that anticipation of predictable postural perturbations can modulate postural responses in adaptive ways. These findings indicate that postural control could involve higher-level neural structures associated with predictive functions, rather than being purely reflexive or reactive. However, which neural circuits are involved and how they function remain largely unknown. In this dissertation project, we developed a novel postural task for mice modeled after human postural perturbation studies, in which a dynamic platform was used to create reproducible translational perturbations. In this task, while mice stood on their hind legs atop a round perch in order to receive water rewards, they experienced backward translations that were either unpredictable or were preceded by an auditory cue. We investigated the effect of a preceding cue and learning on postural responses to perturbations across multiple days. For all animals, we found that the cue improved the postural responses, measured as the ability of the mouse to keep its mouth near the lick spout and to continue to receive water droplets. Two out of three mice showed evidence for learning postural responses across days for both predictable and unpredictable perturbations. We did not observe a consistent trend in learning across trials within daily sessions. Such improvements by cue and learning were generally consistent with previous findings in humans, although the time scale of learning was different. In addition, we explored the nature of pre-perturbation kinematic changes: whether they were elicited by the cue and whether they improved the postural responses. Our studies, though limited, provide validation of a new postural control model. This opens the door to the types of neural population recordings and circuit manipulation that are currently possible only in mice. Our work thus establishes a mouse model system with which to explore the neural mechanisms underpinning predictive postural control.