Credit: J. Vallis/Springer Nature Limited

Learning to navigate towards a goal location involves both spatial memory and reward (goal) encoding, and hippocampal pyramidal cells play roles in both. How goal location is encoded is not well understood, but here, Gauthier and Tank identify a subpopulation of cells in hippocampal CA1 and subiculum that are dedicated to this role.

The authors designed a virtual reality task in which head-fixed mice ran on a linear track while a virtual environment was projected on a surrounding screen. This set-up allowed water-restricted mice to be trained to ‘traverse’ this virtual environment towards a particular location where they would receive a water reward (the goal); neural activity in CA1 and subiculum was monitored simultaneously using two-photon imaging.

As expected, in both subiculum and CA1 the authors found numerous cells that respond differently depending on position on the track, with a greater density of place fields occurring near the reward location. Although the place fields of most cells remained the same when the location of the reward was changed, a few cells tracked the new reward location. The authors determined that these cells constituted a small, discrete population of cells separate from place cells, which they termed reward-associated cells (RACs). RACs consistently produced reward-associated responses across different virtual environments and accounted for the increased density of cell responses near the reward location.

Within this population, some fired before the reward (reward-predictive cells, RPCs), and when the reward location and/or environment was changed, the sequence of activation and timing of RPCs remained highly consistent, suggesting that these cells are highly specialized for encoding reward location.

RPC firing in both CA1 and subiculum coincided with anticipatory behaviours, such as slowing down on approach to the reward location and anticipatory licking. The authors found that RPC firing at the current reward location was approximately five times greater than at a previously-trained reward location, whereas anticipatory behaviours were similar at both locations, suggesting that RPCs do not directly encode behaviour.

Overall, these data reveal a population of cells in both subiculum and CA1 that encode the location of a reward with high consistency, even when the location or environment containing the reward changes.