Thalamo-cortical dynamics across natural brain states and their role in perception and learning – AUDIODREAM

Understanding the switch of the brain’s operating mode between wake and sleep remains a major challenge in neuroscience. Sleep indeed leads to an absence of conscious perception but not to an absence of sensory processing even in higher order centers such as sensory or associative cortex. This has been shown first by the presence of sensory responses in cortex during sleep and second by the elaborate filtering of sensory information that can wake up human subjects. For example hearing one’s first name triggers more sleep interruptions than a meaningless sound with similar acoustic content. These observations are at odds with the widespread idea that the thalamus simply blocks cortical access to sensory information during sleep. Hence, new mechanisms explaining the gating of sensory perception during sleep need to be elucidated. The goal of this project is to test a novel hypothesis that can account for lack of perception without relying on thalamic gating. Supported by a range of preliminary results, our new hypothesis proposes that, during sleep, sounds activate the same combinations of neurons that are recruited by spontaneous activity. In contrast to the thalamic gating theory, this hypothesis can account both for the absence of perception since sound responses would be indistinguishable from on-going cortical activity and for continued sound processing since specific sounds can still evoke different patterns. In line with this new hypothesis, we recently found that spontaneous activity and sound-evoked responses recruit clearly distinct combinations of neurons in wakefulness, but that in at least two different anesthesia, sound responses and spontaneous activity recruit the same combinations of neurons. We therefore aim to show that sleep produces a similar phenomenon although sleep and anesthesia states are widely distinct. We also aim to demonstrate that the overlap of spontaneous and evoked activity matters for perception, by showing that if in sleep a given sound recruits combinations of neurons that are not seen in spontaneous activity, then there is a high chance that the animal wakes up with this sound. The key innovations of the project are (i) the use of two-photon calcium imaging and Neuropixels to extensively evaluate the diversity of neuronal combinations recruited by sounds or spontaneous activity (ii) the use of population analysis methods for activity patterns, a keyapproach to evaluate if spontaneous and evoked activity patterns differs or not in the presence of neuronal variability, (iii) the use of state-of-the-art sleep scoring methods to test the effect of new and previously experienced sounds on sleep, (iv) the development of data-driven computational neuronal population models to reproduce observed network dynamics, propose precise network mechanisms and predict responses to novel sounds. The models will account both for the dynamical switch of the thalamo-cortical network between sleep and wake, and for sound-specific responses in both states. Together, this project aims to establish new principles of auditory information processing across sleep/wake states and their relation to behavior. This question was never addressed at the neurophysiological level and will uncover some of the principles by which stimulus information accesses perception.