Whole-brain MRI and single-cell neural resources underlying sensory-driven temporal expectation in mice

Knowledge of the timing of events is critical for efficient behaviour. Previous work examining the formation of expectations following temporally predictable stimulus sequences has revealed that humans are able to extract regularities, which can improve performance. To determine the underlying neural circuits that drive temporal attention, we established an equivalent behavioural paradigm for mice. We randomize trials with either a long or short delay between a cue and target stimulus, where one delay type occurs with a higher frequency. This allowed for the formation of expectations based on the temporal structure between sensory stimuli. To investigate changes in whole-brain activity patterns with learning, animals underwent a baseline fMRI measurement before training and a second fMRI measurement after training. During training, we recorded single cell and network activity via two-photon calcium imaging in the posterior parietal cortex. This combined approach allowed us to assess and track the allocation of neural resources required for forming temporal expectations. We established differences in learning strategy based on inter-individual variability and differences in licking behaviour, reaction times and task-related changes in pupil diameter. Mice that were task attentive versus task invariant, or showed differing levels of temporal expectation, had differences in sensory-evoked BOLD activation patterns. After training, mice that showed larger levels of temporal expectation recruited fewer sensory-evoked neural resources from brain structures such as prefrontal, posterior parietal, retrosplenial cortex and hippocampus. Furthermore, whether a trial was unexpected or expected in time could be decoded based on the single-cell activity from the posterior parietal cortex.