Electrophysiological low-frequency coherence and cross-frequency coupling contribute to BOLD connectivity

Abstract

Brain networks are commonly defined using correlations between blood oxygen level-dependent (BOLD) signals in different brain areas. Although evidence suggests that gamma-band (30-100 Hz) neural activity contributes to local BOLD signals, the neural basis of interareal BOLD correlations is unclear. We first defined a visual network in monkeys based on converging evidence from interareal BOLD correlations during a fixation task, task-free state, and anesthesia, and then simultaneously recorded local field potentials (LFPs) from the same four network areas in the task-free state. Low-frequency oscillations (<20 Hz), and not gamma activity, predominantly contributed to interareal BOLD correlations. The low-frequency oscillations also influenced local processing by modulating gamma activity within individual areas. We suggest that such cross-frequency coupling links local BOLD signals to BOLD correlations across distributed networks.

Figure 1

Robust resting-state FMRI functional connectivity across a visual thalamo-cortical network. Correlation map showing the connectivity of a right V4 seed with the rest of the brain of monkey BS. V4 was significantly connected with the LIP, TEO and pulvinar, among other regions. ROI boundaries are color coded: red, LIP; blue, V4; green, TEO; cyan, pulvinar.

Figure 2

FMRI functional connectivity between pulvino-cortical ROIs. There were significant correlations between all ROIs during epochs without eye movements. Bars show mean functional connectivity +/− standard error of the mean.

Figure 3

Slow fluctuations in the power of low frequencies, not gamma, contributed most to functional connectivity. There were significantly higher correlations in slow-waves of alpha power between ROIs compared with slow-waves of theta, beta, and gamma power (p < 0.001). We found the lowest correlations for gamma power. Bars show mean functional connectivity +/− standard error of the mean. Mean connectivity of ROI pairs across frequency bands fits a Gaussian function, f_fitted(x) = 0.54 × e^{(−1)×((x−1.92)/2.51)2}, with a peak at the alpha band (R² = 0.989).

Figure 4

Fast-wave coherence at low frequencies best predicts BOLD connectivity. Population mean LFP-LFP coherence (+/− standard error of the mean) calculated for all ROI pairs during stable eye epochs of 500 ms duration. The peak coherence occurred at low frequencies below 20 Hz.

Figure 5

Cross frequency coupling between neural oscillations in the alpha and gamma frequency bands. (Left column) Normalized synchronization index (SI) between the alpha frequency band (8–13 Hz) and higher frequencies across each of the 58 sessions, for the pulvinar, TEO, LIP and V4. The color bar (bottom) shows the SI range. The SI was highly consistent across sessions for each ROI. (Right column) Population average SI (+/− standard error of the mean) shows high coupling between alpha and the gamma frequency band (predominantly 40–80 Hz) for all four ROIs.