Brain state dependency in the auditory thalamocortical system: mutual information and Bayesian decoding

The brain is never silent; throughout the thalamocortical system, spontaneous firing activity occurs in different patterns that reflect ongoing behavioural state. For example, during deep sleep and wakefulness, the thalamocortical system operates in the inactivated state in which neurons collectively aternate slowly between periods of high and near-zero firing activity, respectively referred to as up and down phases. During wakefulness and REM sleep, however, neuromodulatory mechanisms lead to the suppression of this slow oscillation, and neurons fire in a desynchronised manner in what is known as the activated state. Sensory evoked activity also is affected by brain state, and brain states may be the signatures of different modes of processing. Indeed, subtle and/or local changes in ongoing spontaneous are thought to be involved in attention. However the full implications of the effects of brain state on sensory processing remain unknown.
Here we investigate the brain state dependency of sensory processing in the auditory thalamocortical system of the rat, quantifying the amount of information about stimuli carried in the post-stimulus spike counts in individual neurons, and comparing between brain states.
Recordings used were taken simultaneously in medial geniculate body (MGB) and down a cortical column of the primary auditory cortex (A1), and auditory stimuli were presented in both the inactivated state (natural under the anaesthesia) and the activated state (induced through electrical stimulation of the basal forebrain). Quantification of stimulus information was performed through the use of the information theoretic measure of mutual information and a Bayesian decoding method known as maximum likelihood decoding.
Stimulus encoding strategies using spike counts were diverse even among cell-types and locations, as was the effect of brain state on these encoding strategies. However, information theoretic and decoding analysis revealed a shift in the amount of information carried in spike counts, with the least informative neurons tending to increase their information content in the
inactivated state, and the most informative neurons tending to decrease their information content in the inactivated state. Considering a range of window sizes and locations after stimulus presentation, a shift is also observed in the timing of when neurons were most informative.
Summarising, the quantification of stimulus information carried in spike counts allows for the brain state dependency of sensory processing to be seen at this earliest stage of cortical sensory processing, with changes observed in both the amount and timing of information conveyed.