Single-neuron and population contributions of hippocampal LFPs to spike prediction
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Local field potentials (LFPs) contain signals generated by individual neurons and by coordinated population activity, but distinguishing these contributions remains a challenge. We examined how hippocampal LFPs at different frequencies predict single-neuron spiking during spatial navigation in male rats using two datasets. At each frequency, we assessed the spatial distribution of LFP-based prediction across the electrode array and its generalization across behavioral contexts in which a neuron remained active, but its co-active peers changed.
For pyramidal cells, spatially distributed LFP features, consistent with population-level activity, contributed primarily to spike prediction at theta (∼10 Hz) and its harmonics. In contrast, spatially localized signals, consistent with the recorded neuron’s activity, contributed predominantly at higher frequencies. Notably, gamma-band LFPs (30-80 Hz) provided comparatively little information about pyramidal-cell spiking, while distributed LFP features predicted interneuron spiking across a broader frequency range.
Together, this predictive approach separates local and distributed correlates of spiking within the LFP. In hippocampal CA1, these correlates fell into two spatiotemporal regimes: a distributed regime expressed primarily at theta frequencies and a localized regime reflecting single-neuron activity at higher frequencies.
Significance Statement
Brain waves reflect the activity of neuronal populations across spatiotemporal scales. We asked which features of brain waves recorded in the hippocampus predict an individual neuron’s spiking activity as a rat navigated a maze. Brain waves predicted spiking activity most accurately at two different regimes: a low-frequency, spatially distributed regime and a high-frequency, local regime. The distributed regime was concentrated in the ∼10 Hz theta band, while the local regime appeared to reflect the recorded neuron’s activity at high frequencies (>100 Hz). Surprisingly, the gamma band, often linked to neuronal communication and cell assemblies, was the weakest predictor of place cell activity. Our work provides a way to separate individual-neuron and population-level information in hippocampal LFPs.
