Cell type-specific reward computations in habenular neurons

Flexible reward-guided behavior relies on generating predictions about what actions lead to rewards, calculating how new outcomes measure up to those predictions (prediction errors), and adapting behavior to improve future outcomes. Reward information is represented in many structures across the mammalian brain. One node in this interconnected reward network is the habenula complex, an evolutionarily conserved epithalamic structure and neuroanatomical hub that links the limbic forebrain to midbrain neuromodulatory systems, including dopaminergic and serotonergic pathways. The habenula is known for its role in processing information about stress, aversion, anxiety, and reward. At the same time, the habenula shows rich transcriptional diversity, including many neuromodulatory- and neuropeptide-related genes, but it is unclear if these genetically-defined cell types carry separate information on each of those behavioral variables. Using fiber photometry recording in mice performing a reward-guided task, we find that genetically-defined populations of habenular neurons encode separate features of rewards: reward-predictive cues, reward history, and reward omissions. Omission responses in one lateral habenula (LHb) cell type were bidirectionally modulated by reward probability, recent reward history, and expected reward size, in agreement with prior work on LHb neurons encoding reward prediction errors. Furthermore, we find that these neurons respond preferentially to expectation-driven computations; innately aversive stimuli produce only weak responses. Taken together, these data build a framework of cell type-specific reward computations in habenular neurons and potentially provide genetic inroads to better understand the anatomical and functional organization of habenular neurons and the role of habenular reward prediction error in flexible decision-making. None.