Assessing the dynamics of visual feedback training for the adaptation of oculomotor integration in mice

Neural integration is an essential computation that converts transient inputs into extended neural activity, a computation that supports many brain functions, such as motor control and working memory. One specific example is the oculomotor integrator, that is responsible for allowing a subject to hold the eye in a given position for prolonged periods of time to maintain a stable focus on a visual object. Previous work in goldfish has shown that the oculomotor integrator can be adaptively modified using visual feedback. The purpose of this project was to examine the potential for visual feedback training to modify the oculomotor integrator in mice. To investigate this, experimental paradigms were developed using an adaptive visual field around the subject to create conditions in which the dynamics of the oculomotor integrator must be altered to stabilize the images on the retina. The ratio of image velocity to eye position during training (training gain) was varied, as was the range of eye positions around null position that were excluded from training. For all training gains, visual feedback training designed to induce unstable oculomotor integrator behavior was effective at eliciting eye movements to track the visual stimulus during training and inducing a learned increase in the instability of the oculomotor integrator as tested in the dark after training. In contrast, the tracking eye movements and learning induced by leak training were more variable, although there was a trend for the oculomotor integrator to be more leaky after training with several of the gains and null exclusion range values tested. Both leaky and unstable training were most effective for a gain value of 0.75 and a null exclusion range at or below 1.5 degrees. To identify an approach for testing the role of the cerebellum in this type of visual feedback-driven learning, oculomotor learning experiments were conducted in mice expressing ArchT in Purkinje cells. These mice did not show reliable oculomotor learning, even in the absence of illumination to inhibit the Purkinje cells, suggesting that an alternative approach is needed.