Bridging scales from spiking activity to the local field potential through latent dynamics

Abstract

AbstractDuring wakefulness, cortical networks exhibit diverse states of activity arising from fluctuations in modulatory inputs, any of which may shape the recorded neural activity across a range of scales, from local field potential (LFP) to the spiking of individual neurons. Multiple studies have uncovered state-dependent changes in single neuron activity specific to cell type. These changes are heterogeneous across cell types, suggesting that there is a link between changes in states of wakefulness and the dynamic patterns of spiking activity across the population, yet this is currently not well understood. Here, we use an unsupervised statistical modeling framework to link fluctuations in the state of the LFP to the dynamics of spiking activity of populations of neurons in the whisker area of primary somatosensory cortex of awake male mice. By identifying latent states from population spiking activity, we show that these spiking states are more informative of states of the LFP than a direct readout of the spiking activity of single neurons or of the population. Further, we find that the latent spiking states predict trial-by-trial variability in sensory responses as well as one aspect of behavior, whisking activity. Taken together, our approach captures the latent variables that underlie the dynamics of the spiking activity of cortical neuronal populations, providing a framework to relate cortical activity across different scales.Significance statementDuring wakefulness, variability in cortical activity can arise from shifts in arousal or behavior. We analyze the dynamics of two measures of cortical activity, the spiking activity of local populations of neurons and the local field potential (LFP), typically used to track overall levels of arousal. We show that the latent states that capture variability in population spiking activity are highly predictive of LFP state changes, even more so than spiking activity directly. By linking latent states of spiking activity to changes in the LFP, our novel approach bridges fine-scale microcircuits ensemble activity and the meso-scale LFP. Our latent-state framework opens new avenues for extracting information from the LFP as well as identification of subject-level sources of variability.