Abstract
SummaryRecent studies using intracellular recordings in awake behaving mice revealed that cortical network states, defined based on membrane potential features, modulate sensory responses and perceptual outcomes. Single cell intracellular recordings are difficult to achieve and have low yield compared to extracellular recordings of population signals, such as local field potentials (LFPs). However, it is currently unclear how to identify these behaviorally-relevant network states from the LFP. We used simultaneous LFP and intracellular recordings in the somatosensory cortex of awake mice to design and calibrate a model-based analysis method, the Network State Index (NSI), that enables network state classification from the LFP. We used the NSI to analyze the relationship between single-cell (intracellular) and population (LFP) signals over different network states of wakefulness. We found that graded levels of population signal faithfully predicted the levels of single cell depolarization in non-rhythmic regimes whereas, in delta ([2-4 Hz]) oscillatory regimes, the graded levels of rhythmicity in the LFP mapped into a stereotypical oscillatory pattern of membrane potential. Finally, we showed that the variability of network states, beyond the occurrence of slow oscillatory activity, critically shaped the average correlations between single cell and population signals. NSI-based characterization provides a ready-to-use tool to understand from LFP recordings how the modulation of local network dynamics shapes the flexibility of sensory processing during behavior.Significance statementSensation during behaviour is strongly modulated by the animal’s internal state. Such context-dependent modulation of sensory processing is believed to largely stem from top-down control of network states in sensory cortices, with different network states being associated with distinct computational properties of the circuit. So far, a detailed characterization of network states in the awake cortex has mostly been achieved through single-cell intracellular recordings, which however cannot be easily recorded. Here, we developed a new method to classify network states from the easily accessible extracellular LFP recordings of population activity. Given the widespread use of LFPs, our work provides a critical methodology to greatly expand our understanding of the mechanisms underlying state-dependent computations in neocortex.
Publisher
Cold Spring Harbor Laboratory