Resting state brain networks arise from electrophysiology-invisible signals.

Resting-state brain networks (RSNs) have been widely applied in health and disease, but their interpretation in terms of the underlying neural activity is unclear. To systematically investigate this cornerstone issue, here we simultaneously recorded whole-brain resting-state functional magnetic resonance imaging (rsfMRI) and electrophysiology signals in two separate brain regions in rats. Our data show that for both recording sites, band-specific local field potential (LFP) power-derived spatial maps can explain up to 90% of spatial variance of RSNs obtained by the blood-oxygen-level dependent (BOLD) signal. Paradoxically, the time series of LFP band power can only explain up to 35% of temporal variance of the local BOLD time course from the same location even after controlling for the factors that may affect apparent LFP-BOLD correlations such as contrast-to-noise ratio. In addition, regressing out LFP band powers from the rsfMRI signal does not affect the spatial patterns of BOLD-derived RSNs, collectively suggesting that the electrophysiological activity has a marginal effect on the rsfMRI signal. These findings remain consistent in both light sedation and awake conditions. To reconcile this contradiction in the spatial and temporal relationships between resting-state electrophysiology and rsfMRI signals, we propose a model hypothesizing that the rsfMRI signal is driven by electrophysiology-invisible neural activities that are active in neurovascular coupling, but temporally weakly correlated to electrophysiology data. Meanwhile, signaling of electrophysiology and electrophysiology-invisible/BOLD activities are both constrained by the same anatomical backbone, leading to spatially similar RSNs. These data and the model provide a new perspective of our interpretation of RSNs.