Spatial goals trap hippocampal replay

Abstract

Hippocampal replay is proposed to support memory and planning by preferentially reactivating behaviorally significant locations. Although learning can alter which locations are reactivated and how frequently, whether it also alters the dynamics by which replay propagates through the cognitive map remains unknown. To address this, we recorded large-scale CA1 activity while rats collected identical rewards at multiple spatial locations, only one of which was consistently rewarded and served as the Goal. Replay events occurred nearly twice as often during reward consumption at the Goal compared to Non-Goal sites and were accompanied by reduced inhibition. Goal-associated replays were also slower, shorter, and more locally confined. Replays originating elsewhere in the arena slowed and preferentially terminated when nearing the Goal, indicating that the Goal locally resists replay propagation regardless of where the replay was initiated. Increasing evidence suggests that hippocampal sequences arise from adaptation-like mechanisms that propagate neural activity away from recently active states. We hypothesized that the slowing and spatial confinement of Goal-associated replay arises from reduced effective adaptation near the Goal. A recurrent network model showed that reducing adaptation strength reproduces these effects and correctly predicted increased population firing rates and weaker avoidance of recently traveled or reactivated paths. Finally, we found evidence for reduced interneuron-mediated feedback inhibition near the Goal, providing a potential circuit mechanism for locally modulating adaptation. Our findings show that learning locally reshapes the dynamics governing internally generated activity, causing replay to linger near behaviorally significant locations and providing a new mechanism for their preferential reactivation.