Population dynamics underlying sensorimotor transformation in mouse parietal cortex


January 11, 2017 - 12:00pm - 1:00pm
NW Labs 243
About the Speaker
Gerald Pho
Speaker Affiliation: 

The ability to map arbitrary sensory stimuli to appropriate motor actions is fundamental to cognition and behavior. Posterior parietal cortex (PPC) has been implicated in this process of sensorimotor transformation, but the neural dynamics underlying its function remain poorly understood. Electrophysiological studies in non-human primates have suggested a model by which PPC neurons encode the sensory evidence for a decision in the form of low-dimensional ramping activity. However, recent work in rodents (Harvey et al, Nature 2012) have shown that PPC can exhibit high-dimensional sequential dynamics during a complex navigational task. Whether such dynamics are specific to navigational contexts or fundamental to the process of sensorimotor transformation remains unclear. Here we use population calcium imaging to measure the activity of hundreds of neurons in PPC or primary visual cortex (V1) of mice performing a simple head-fixed visual decision task. We manipulated many aspects of the task to probe the dynamics of V1 and PPC coding, including task engagement, stimulus contrast, and reward contingency. We found that mouse PPC exhibited task-specific sequences that encode the choice of the animal, even in the context of a non-navigational task. Using a cross-temporal decoding analysis, we found that PPC activity patterns dynamically evolved to encode choice, whereas V1 used a largely stationary code. Dynamic PPC coding was specific to choice, as the same PPC population also encoded stimulus contrast but in a stationary manner. Lastly, we re-trained mice on a reversed reward contingency and imaged the same populations before and after reversal. While V1 neurons maintained their stimulus selectivity, we found that the new sensorimotor transformation was reflected in PPC as a novel choice-selective sequence. This sequence included some neurons with pre-reversal choice selectivity, but also recruited new, previously unresponsive neurons. Together these results demonstrate that high-dimensional dynamics specifically emerge in PPC during the execution of a learned sensorimotor transformation, and suggest that such dynamics may be a general mechanism for both navigational and non-navigational tasks.