Single-neuron selectivity rules underlying population representations in prefrontal cortex

Abstract

Neurons in prefrontal cortex exhibit selectivity for a range of behaviorally relevant variables, such as sensory stimuli, place, time, choice, context, and category. At the population level, representations of these variables can be organized in various ways--for instance, geometrically or hierarchically. However, how selectivity for each feature is distributed across cells and how this gives rise to such population-level representational structure is unclear. To address these questions, we analyzed coding for space, time, and category in rats performing a navigational task featuring two behavioral categories, starts and goals, each consisting of multiple locations. Population activity was organized hierarchically, with neural distances between elements of a category being smaller than the distances between categories. At the single cell level, individual medial prefrontal cortex (mPFC) neurons were tuned to category and space independently, consistent with hierarchical population geometry emerging from random mixed selectivity, while neurons exhibited preferences for representing space or time across categories, consistent with functional specialization. These distinct selectivity rules across different levels of abstraction suggest that different aspects of prefrontal representations have disparate origins and may enable modularized computations across multiple behaviors and contexts.