The Mesoscale Functional Organization of Mouse Cortex During Behavior

Abstract

Visually-guided navigation engages a network of cortical areas to process sensory information and plan movements, including visual, parietal, and retrosplenial regions. These regions have been mapped based on anatomy and responses to simple sensory stimuli, and studied individually during behavior. However, how the network as a whole is functionally organized during navigation-related behavior remains poorly understood. Here, we densely sampled the activity of single neurons across the mouse posterior cortex during a navigation task. We developed unsupervised analysis approaches to relate neural activity to behavior and to compare activity across anatomical space. We found that behavioral information was highly distributed across cortex. Although cortical regions were discriminable based on behavior-related functional properties at a resolution similar to the areas defined by retinotopy, most neural activity patterns could be found in most areas. Further, encoding properties varied smoothly across secondary visual and association areas. The discriminability between regions was due to variations in the frequency of distributed encoding properties, rather than region-specific properties. Behavioral variables such as visual and locomotion velocity were encoded in overlapping regions. Where these regions overlapped, multimodal representations emerged. Our results are consistent with recent reports of ultra-dense connectivity between cortical areas, and theoretical models that show that this connectivity can be explained by organizational rules on the level of single neurons, rather than cortical areas. We suggest that during behavior, the functional architecture of cortex is better described by overlapping gradients of task-dependent representations than by discrete, specialized modules.