Value Representations in Visual Association Cortex

Abstract

The needs of the body can direct neural processing towards motivationally-relevant sensory stimuli. For example, humans attend more to food cues during hungry versus sated states. Human neuroimaging studies consistently demonstrate hunger-dependent enhancement of neural responses to food cues in visual association cortex and the amygdala, but not primary visual cortex. These studies, however, pool the activity across thousands of neurons, and lack the cellular resolution to dissect the microcircuits that underlie this hunger-dependent food cue bias. This thesis examines how motivational states like hunger bias processing to food cues and how this information is encoded in the visual system. Using two-photon calcium imaging we tracked the activity of individual neurons across days, and across states of hunger and satiety in mice performing a Go/NoGo task involving discrimination of food-predicting cues from other visual cues. We recorded from neurons in visual association cortex (postrhinal cortex; POR), primary visual cortex (V1), or long-range feedback axons from the lateral nucleus of the amygdala to POR. As in humans, neurons in POR, but not V1, demonstrated a response bias to a motivationally relevant food cue that was abolished following satiation. Feedback axons from the amygdala to POR demonstrated even stronger food cue biases and sensitivity to hunger state, suggesting the amygdala may act as a source of feedback that guides sensory processing towards motivationally-relevant cues. In order to dissect the microcircuit underlying the response bias to a food cue, we then trained animals on a reversal paradigm while chronically imaging neurons in visual association cortex. We identified correlated ensembles of neurons that either tracked low-level stimulus features or the predicted outcome (e.g. whichever cue predicted a food reward) across a switch of cue-outcome associations. Visual responses of this latter ensemble encoded “value” as these neurons demonstrated a response bias to the food cue and sensitivity to reward history and hunger state. Strikingly, this “value-coding” was not evident in the neurons that tracked low-level stimulus features. These results suggest intermingled ensembles of neurons in visual association cortex maintain a faithful representation of a visual stimulus and flexibly encode predicted outcomes and their motivational salience.