Computation by recurrent circuits of the cerebral cortex

National Institute Of Mental Health

Investigators

Linked publications & trials

Abstract

Cortical neurons fire in complex patterns during behavior and cognition. In sensory regions of the cortex, animals' interactions with the sensory world evoke neural activity in the cortex. But not all patterns of cortical activity can be elicited by sensory stimuli. We study how non-sensory, artificially induced activity patterns can be used by animals to make behavioral responses. We insert a range of activity patterns into the brain using optogenetic stimulation. These studies shed light on the limits of cortical function, and how circuit properties like neural connections constrain the set of activity patterns the cerebral cortex can process. In this project period, we have made progress understanding the learning rules brains use. Since we know brain networks must change with learning, but we know little about the learning rules used in the brain in vivo, this question is vitally important for understanding brain function. We use external inputs, laser stimulation, to change the activity of neurons during behavioral tasks to determine how networks in the brain reorganize over time and with learning. A priori, it could be that brain networks reorganize to select or amplify added inputs that are useful for behavior. If networks were limited by noise, or otherwise needed to boost weak signals, this might be the way the brain learns to work with new input. On the other hand, it could be that networks change to compensate for added input, reducing their responses in a way to hold inputs constant. If noise is not a major limitation, and other factors are instead being optimized for, this might be the outcome. We find evidence for the second possibility. Early results from this work were submitted to the annual Computational and Systems Neuroscience (Cosyne) meeting. We also have made progress understanding the ongoing changes in brain responses that happen on a daily basis. This type of change, often called representational drift, is mysterious both in purpose and in origin. It is widely observed that neural representations of the sensory world change, or “drift”, over days to weeks. What circuit changes create representational drift? As recurrent networks in the cortex shape cortical responses, plastic changes in recurrent connectivity — resulting from e.g. learning, or homeostatic changes — could underlie representational drift. This plasticity could be random, resulting in undirected, uniform drift rates across different inputs. Alternatively, plasticity could be biased to favor sensory features, resulting in directed, varying rates of drift. We studied how representational drift relates to local, recurrent network interactions using longitudinal two-photon holographic stimulation across weeks. We stimulated patterns of neurons, choosing patterns based on visual responses or randomly, in the mouse visual cortex, tracking changes in evoked network responses as drift occurs. We found that network responses to fixed input patterns do change over days, indicating a reorganization in the local, recurrent network. The amount of change varied in a pattern-dependent manner, where patterns based on visually-evoked responses drifted more than random patterns. This suggests subnetworks activated by visual stimuli undergo more plasticity over time. Together, these results suggest that representational drift is driven by ongoing synaptic changes in the local recurrent network of sensory cortical areas. These findings contribute to our understanding of how cortical networks compute — that is, how the connectivity within and between brain areas transforms inputs to process information —a function that is at the core of how brains work.