Neuronal mechanisms of adaptive value coding in the amygdala

To address their needs, animals must optimize behavior by integrating external cues with internal state signals, such as hunger or thirst, environmental conditions, and past experiences. Accordingly, the perceived value of a reward varies depending on an animal’s current needs. Updating the specific value of rewards predicted by external cues depends on the basolateral amygdala (BLA). However, neurophysiological investigations have centered around the assumption of stimulus-invariant value representations and have not been able to account for behavioral findings that support the BLA’s ability to represent stimulus-specific value of distinct appetitive or aversive outcomes. To address how the BLA encodes and updates stimulus-specific value representations, we exposed head-fixed mice to distinct sets of gustatory rewards while tracking perceived value using lick microstructural analysis. Two-photon calcium imaging of BLA principal neurons revealed that the magnitude of the BLA population response scaled with perceived reward value and that the values of different rewards were encoded by distinct neuronal subpopulations. Moreover, reward representations rapidly re-scaled when mice were exposed to a reward that was larger than all previous rewards. Finally, value representations depended on an animal’s internal state as thirst selectively increased the responses to water rewards whereas aversive experiences strongly attenuated responses to sucrose. Our findings demonstrate that value representations in the BLA are stimulus-specific and highly adaptive to account for changes in relative reward value and to reflect an animal’s current affective and homeostatic state. This mechanism enables sensory-specific value updates necessary for state-adapted decision making and learning.