Population analyses reveal heterogenous encoding in the medial prefrontal cortex during naturalistic foraging

Abstract

AbstractForaging in the wild requires coordinated switching of critical functions, including goal-oriented navigation and context-appropriate action selection. Nevertheless, few studies have examined how different functions are represented in the brain during naturalistic foraging. To address this question, we recorded multiple single-unit activities from the medial prefrontal cortex (mPFC) of rats seeking a sucrose reward in the presence of a robotic predator (Lobsterbot) that posed periodic threats. Simultaneously recorded ensemble activities from 10-24 neurons were analyzed in reference to various behavioral indices as the animal moved freely between the nest (N) and the goal (E) across the foraging (F) area. In the E-zone, the rat initially received and gradually learned to avoid unpredictable attacks by the Lobsterbot. An artificial neural network, trained with simultaneously recorded neural activity, estimated the rat’s current distance from the Lobsterbot. The accuracy of distance estimation was the highest in the middle F-zone in which the dominant behavior was active navigation. The spatial encoding persisted in the N-zone when non-navigational behaviors such as grooming, rearing, and sniffing were excluded. In contrast, the accuracy decreased as the animal approached the E-zone, when the activity of the same neuronal ensembles was more correlated with dynamic decision-making between food procurement and Lobsterbot evasion. A population-wide analysis confirmed a highly heterogeneous encoding by the region. To further assess the decision-related activity in the E-zone, a naïve Bayesian classifier was trained to predict the success and failure of avoidance behavior. The classifier predicted the avoidance outcome as much as 6 s before the head withdrawal. In addition, two sub-populations of recorded units with distinct temporal dynamics contributed differently to the prediction. These findings suggest that the mPFC neurons may adopt at least two modes of heterogenous encoding that reflect the processing of relevant spatial context and the imminent situational challenge.