Resting-state cross-frequency coupling networks in human electrophysiological recordings

Neuronal oscillations underlie temporal coordination of neuronal processing and their synchronization enables neuronal communication across distributed brain areas to serve a variety of sensory, motor, and cognitive functions. The regulation and integration of neuronal processing between oscillating assemblies at distinct frequencies, and thereby the coordination of distinct computational functions, is thought to be achieved via cross-frequency coupling (CFC). Although many studies have observed CFC locally within a brain region during cognitive processing, the large-scale networks of CFC have remained largely uncharted. Critically, also the validity of prior CFC observations and the presence of true neuronal CFC has been recently questioned because non-sinusoidal or non-zero-mean waveforms that are commonplace in electrophysiological data cause filtering artefacts that lead to false positive CFC findings. We used a unique dataset of stereo-electroencephalography (SEEG) and source-reconstructed magnetoencephalography (MEG) data to chart whole-brain CFC networks from human resting-state brain dynamics. Using a novel graph theoretical method to distinguish true inter-areal CFC from potentially false positive CFC, we show that the resting state is characterized by two separable forms of true inter-areal CFC: phase-amplitude coupling (PAC) and n:m-cross-frequency phase synchrony (CFS). PAC and CFS large-scale networks coupled prefrontal, visual and sensorimotor cortices, but with opposing anatomical architectures. Crucially also directionalities between low- and high-frequency oscillations were opposite between CFS and PAC. We also found CFC to decay as a function of distance and to be stronger in the superficial than deep layers of the cortex. In conclusion, these results provide conclusive evidence for the presence of two forms of genuine inter-areal CFC and elucidate the large-scale organization of CFC resting-state networks.