Functional determinants of enhanced and depressed inter-areal information flow in NREM sleep between neuronal ensembles in rat cortex and hippocampus

Compared to wakefulness, neuronal activity during non-REM sleep is characterized by a decreased ability to integrate information, but also by the re-emergence of task-related information patterns. To investigate the mechanisms underlying these seemingly opposing phenomena, we measured directed information flow by computing transfer entropy between neuronal spiking activity in three cortical regions and the hippocampus of rats across brain states. State-dependent information flow resulted to be jointly determined by the anatomical distance between neurons and by their functional specialization. We distinguished two regimes, operating at short and long time scales, respectively. From wakefulness to non-REM sleep, transfer entropy at short time scales increased for inter-areal connections between neurons showing behavioral task correlates. Conversely, transfer entropy at long time scales became stronger between non-task modulated neurons and weaker between task-modulated neurons. These results may explain how, during non-REM sleep, a global inter-areal disconnection is compatible with highly specific task-related information transfer.