Basolateral amygdala circuits in detailed associative reward memory
To make good decisions we often rely on detailed associative memories to infer the availability of prospective rewards from predictive cues within the environment. These stimulus-outcome relationships are essential elements of our cognitive map that links specific outcomes to antecedent cues and the specific actions needed to obtain them. This internal model enables us to project into the future, anticipate the consequences of our actions and adapt our behaviors accordingly (i.e., model-based decision making). The research presented here investigated neural circuitry mediating the encoding and subsequent retrieval of outcome-specific reward memories. Targeted optical manipulation and recording methods revealed the basolateral amygdala (BLA) as a central hub for detailed stimulus-outcome associations. The BLA was robustly activated by the delivery of distinct food outcomes preceded by auditory stimuli and this activity was necessary for encoding sensory-specific cue-reward memories. This function was supported by modulatory inputs from ventral tegmental dopamine neurons (VTADA). Optically manipulating VTADABLA axonal terminals during Pavlovian conditioning and in a novel Pavlovian blocking paradigm revealed these inputs are both necessary and sufficient to drive outcome-specific learning. Excitatory projections from the lateral orbitofrontal cortex (lOFC) were also essential for facilitating the BLA in encoding these associations. Moreover, reciprocal lOFCBLAlOFC connections formed an encoding and retrieval circuit. Associative memories are complex, containing information about stimuli, motivational state, outcome identity, value, etc. A pathway-specific serial disconnection revealed that the component of the associative memory that was encoded through activation of lOFCBLA projections was the same as that which was later accessed through BLAlOFC projections to enable cue motivated adaptive reward seeking. Collectively, these data uncover neural substrates for detailed associative reward memories that enable model-based decision making.