How neurons encode behavior is a fundamental question. Neuronal ensembles increase or decrease activity during specific behaviors. However, it is unclear whether ensembles encode information solely via changes in activity levels, or whether changes in correlations between neurons carry additional information. We used microendoscopic GCaMP imaging to measure prefrontal activity while mice were either alone or engaged in social interaction. Using neural network classifiers to measure how well prefrontal neurons transmit information about social behavior to downstream neurons, we find that surrogate datasets which preserve dynamic correlations outperform those which preserve ensemble activity but not correlations. Notably, this ability of correlations to enhance the information transmitted by neuronal ensembles is lost in mice lacking the autism-associated gene Shank3. These results show that dynamically modulated correlations create patterns of coactive neurons which are behaviorally-specific and enhance the information transmitted by neuronal ensembles. Furthermore, this process may be disrupted in pathological states.
bioRxiv Subject Collection: Neuroscience