Variations in value and outcome processing along the longitudinal axis of the hippocampus

Abstract

The hippocampus is involved in numerous brain functions, including memory, imagination, and emotional processing. Although ample evidence indicates anatomical, transcriptional, physiological, and functional variations along the longitudinal axis of the hippocampus, the exact roles played by different parts of the hippocampus remain unclear. To obtain insights on this matter, I investigated how value and outcome processing varies along the longitudinal axis of the hippocampus. Specifically, I compared inactivation effects and neuronal discharge characteristics of the dorsal (DH) and ventral hippocampus (VH) in mice performing probabilistic classical conditioning tasks. Three different odor cues were paired with water (reward) or air-puff (punishment) delivery in a probabilistic manner. The animal’s anticipatory licking response was used as an index for cue-outcome contingency learning. Inactivating the DH or VH by muscimol infusion impaired value-dependent anticipatory licking behavior. Furthermore, value-coding neurons were found in both areas. These results indicate the involvement of both the DH and VH in value processing. However, characteristics of value- and outcome-dependent neural activity differed between the two regions. First, the DH neuronal population increased its activity as a function of value, while VH neuronal population was preferentially responsive to the highest-value sensory cue. Second, neural signals related to outcome-dependent value learning, namely value and outcome signals during the outcome period, signals related to the history of past cues and outcomes, and signals related to updated value, were stronger in the DH. Third, VH neurons showed rapid responses to punishment and strongly biased responses to negative prediction error. These results indicate that the DH faithfully represents the external value landscape, whereas the VH preferentially represents behaviorally relevant, salient features of experienced events. I propose that differentiating DH and VH functions by the level of extracting behaviorally-relevant information rather than the modality of processed information may better account for their functional differentiation.