Selective coupling between theta phase and neocortical fast gamma oscillations during REM-sleep in mice

Abstract

Background: The mammalian brain expresses a wide range of state-dependent network oscillations which vary in frequency and spatial extension. Such rhythms can entrain multiple neurons into coherent patterns of activity, consistent with a role in behaviour, cognition and memory formation. Recent evidence suggests that locally generated fast network oscillations can be systematically aligned to long-range slow oscillations. It is likely that such cross-frequency coupling supports specific tasks including behavioural choice and working memory.

Principal findings: We analyzed temporal coupling between high-frequency oscillations and EEG theta activity (4-12 Hz) in recordings from mouse parietal neocortex. Theta was exclusively present during active wakefulness and REM-sleep. Fast oscillations occurred in two separate frequency bands: gamma (40-100 Hz) and fast gamma (120-160 Hz). Theta, gamma and fast gamma were more prominent during active wakefulness as compared to REM-sleep. Coupling between theta and the two types of fast oscillations, however, was more pronounced during REM-sleep. This state-dependent cross-frequency coupling was particularly strong for theta-fast gamma interaction which increased 9-fold during REM as compared to active wakefulness. Theta-gamma coupling increased only by 1.5-fold.

Significance: State-dependent cross-frequency-coupling provides a new functional characteristic of REM-sleep and establishes a unique property of neocortical fast gamma oscillations. Interactions between defined patterns of slow and fast network oscillations may serve selective functions in sleep-dependent information processing.

Figure 1. Raw signals during both theta states.

Raw and filtered field potentials recorded from the parietal cortex during active waking and REM-sleep of one representative animal. Filtered traces emphasize theta range (second line, 4–12 Hz), gamma range (third line, 40–100 Hz), and fast gamma ranges (bottom, 120–160 Hz). Note higher fast gamma activity in active waking but more pronounced theta-phase coupling in REM-sleep.

Figure 2. Power-frequency distributions and vigilance state.

A, C, E: Mean power-frequency distributions emphasizing the theta, gamma and fast gamma ranges during active waking (aWk; black lines) and REM-sleep (REM; blue lines). Dotted lines represent SEM (n = 10 mice). Note the different power scales (dB) in C and E compared to A, B, D, F (µV²/Hz): Bar diagrams show mean band power for (B) theta (4–12 Hz), (D) gamma (40–100 Hz) and (F) fast gamma (FG, 120–160 Hz) during aWk (black columns) and REM blue columns). Asterisks in B, D, F indicate significant differences between active waking compared to REM-sleep (B, t-test, theta: p<0.05; D, gamma: p<0.0001; F, fast gamma: p<0.001).

Figure 3. Power spectral densities (PSD) and cross-frequency coupling (CFC) in different vigilance states.

A: PSDs during spontaneous active waking (aWk), quiet waking (qW), NREM- and REM-sleep (n = 10 mice). Arrows point to the power peaks for theta, gamma and fast gamma activity. B: Mean comodulation maps (CFC) in parietal cortex during the four behavioural states (n = 10 mice). The abscissa represents the phase-modulating frequencies (f_p) and ordinate amplitude-modulated frequencies (f_a). Pseudocolour scale indicates the modulation index (see Methods for details). C, D: Theta-gamma (C) and theta-fast gamma (D) CFCs are significantly higher in REM compared to aWk. E: Ratio of modulation indices in REM compared to aWk for gamma and fast gamma oscillations show significantly higher vigilance differences for theta-fast gamma compared to theta-gamma (p = 0.0014).

Figure 4. Theta phase coupling of gamma and fast gamma oscillations in parietal cortex depends on theta power, both in active waking (aWk) and REM-sleep (REM).

A and C: Data from a representative animal. Gamma and fast gamma CFCs were calculated from 30s periods of REM and aWk for bins of 50 theta power units (µV²/Hz). B and D: After normalization of theta power (see Methods for details) 30s periods of aWk and REM were pooled for bins of 0.01 normalized theta power units and CFC for gamma and fast gamma calculated and averaged over animals (n = 8 mice). Note positive correlation in all four data sets (see Results). Power was calculated from 1s windows in steps of 0.5s.

Figure 5. Gamma and fast gamma oscillations are differentially phase-locked to theta.

A and B: Mean amplitudes of neocortical gamma (A, 40–100 Hz) and fast gamma (B, 120–160 Hz) plotted according to the phase of theta in active waking (black dots) and REM-sleep (blue dots). Amplitude maxima of gamma occur at the positive peak of the theta wave in parietal cortex (0 degrees), while the amplitude maxima of fast gamma occur at 30 degrees, on the falling flank of the positive peak of the theta wave. C: Time-frequency plot of mean amplitude distribution time-locked to the theta peak (n = 9 mice). The positive peaks of the averaged theta waves correspond to 0 s. Bottom panel shows the theta peak-locked averaged raw signal.