Understanding the neuronal mechanisms of closed-loop olfaction

Project Summary In nature, sensory perception and motor processing operate in closed-loop. Self-generated movements impact sensory input, and sensory inputs guide future motor commands. Through experience, the brain may learn the reciprocal relationship between sensory inputs and movements in the form of generative sensorimotor models that predict the sensory consequences of upcoming actions. In vertebrates, olfaction is intrinsically linked to motor action through sniffing, and just as for other sensory modalities, via head and body movements. Due to technical challenges, however, most studies in laboratory settings have probed olfactory processing during passive odor sampling. Even when investigating odor-driven navigation, the effect of movements on odor responses has rarely been analyzed. Here we will test the central hypothesis that, in closed-loop olfaction, mice generate olfacto-motor predictions on the sensory consequences of their actions, which further guide odor sampling movements. At the circuit level, we hypothesize that specific olfactory cortex circuits represent olfacto- motor prediction errors, computed by comparing odor input and movement-related predictions of the expected odor input. We plan to test these hypotheses using a novel closed-loop odor localization task (Smellocator) developed in our group, together with a rich repertoire of sensorimotor perturbations, state-of-the-art recordings and cell-type circuit analysis tools with increasing levels of specificity. ● To this end, we will first investigate whether under closed-loop coupling of movements and odor sensing, mice detect olfacto-motor errors, and further compensate for them. In the Smellocator task, head-fixed mice learn to steer a lightweight lever with their paws to control the lateral location of an odor source according to a fixed-gain sensorimotor mapping rule. In catch trials, we will transiently alter the relationship between lever movement and odor displacement via a range of precise, unexpected sensorimotor perturbations. Preliminary data indicate that expert mice successfully compute sensorimotor prediction errors, and quickly engage in fine corrective movements triggered by these perturbations in an individual specific manner. • Then, we will investigate whether the olfactory cortex (piriform vs. anterior olfactory nucleus) represents olfacto- motor prediction errors in face of transient surprises. We will check whether brief sensorimotor perturbations trigger sudden changes in cortical activity (mismatch responses). We will refine our analysis to determine if different semilunar and pyramidal cells types (e.g. Netrin+, Cux1+, Tbr1+, Tle4+) represent primarily sensory inputs vs. olfacto-motor errors by combining distributed recordings and modern genetic labeling strategies. • Finally, we will investigate whether the olfactory cortex enables adaptation in the presence of persistent olfacto- motor errors. We will change the sensorimotor mapping rules in blocks of trials, and across behavioral sessions, and compare the roles of specific cell types in supporting sensorimotor adaptation taking advantage of flexible optogenetic suppression strategies.