Sleep-Specific Processing of Auditory Stimuli Is Reflected by Alpha and Sigma Oscillations

Abstract

Recent research revealed a surprisingly large range of cognitive operations to be preserved during sleep in humans. The new challenge is therefore to understand functions and mechanisms of processes, which so far have been mainly investigated in awake subjects. The current study focuses on dynamic changes of brain oscillations and connectivity patterns in response to environmental stimulation during non-REM sleep. Our results indicate that aurally presented names were processed and neuronally differentiated across the wake-sleep spectrum. Simultaneously recorded EEG and MEG signals revealed two distinct clusters of oscillatory power increase in response to the stimuli: (1) vigilance state-independent θ synchronization occurring immediately after stimulus onset, followed by (2) sleep-specific α/σ synchronization peaking after stimulus offset. We discuss the possible role of θ, α, and σ oscillations during non-REM sleep, and work toward a unified theory of brain rhythms and their functions during sleep.SIGNIFICANCE STATEMENT Previous research has revealed (residual) capacity of the sleeping human brain to interact with the environment. How sensory processing is realized by the neural assemblies in different stages of sleep is however unclear. To tackle this question, we examined simultaneously recorded MEG and EEG data. We discuss the possible role of θ, α, and σ oscillations during non-REM sleep. In contrast to versatile θ band response that reflected early stimulus processing step, succeeding α and σ band activity was sensitive to the saliency of the incoming information, and contingent on the sleep stage. Our findings suggest that the specific reorganization of mechanisms involved in later stages of sensory processing takes place upon falling asleep.

Keywords: EEG; MEG; auditory; brain oscillations; information processing; sleep.

Figure 1.

Study design. Processing steps from MEG/EEG acquisition to analysis of the data are illustrated. Spoken first names were presented to the participants lying in the MEG scanner. Volunteers were requested to stay awake with eyes open for 20 min, and then close their eyes and try to sleep for another 2 h. Stimuli were presented aurally and in a pseudo-random order every 2.5 to 6 s. For the analysis, stimuli were grouped according to (1) the name type, (2) the uttering (familiar or unfamiliar) voice, and (3) the current sleep stage at stimulus presentation.

Figure 2.

Oscillatory power response to auditory stimuli presented in various stages of NREM sleep. In (A) depicting MEG sensors and (B) depicting EEG sensors, NREM sleep is characterized by synchronization within the α/σ band (∼9-16 Hz) and a lack of α band (∼8-12 Hz) desynchronization. The plots are normalized against absolute baseline between 0.5 and 0.1s before stimulus onset. Areas outlined in black represent significant clusters for prestimulus versus poststimulus contrasts. C, D, Source level topographical distribution of the MEG θ (4-7 Hz) and σ (12-20 Hz) frequency response, respectively. Primary auditory areas contribute to the θ response in wakefulness and in NREM sleep, along with coactivation in sensorimotor areas and the brainstem after sleep onset. Sources of σ modulation in light sleep (N1 and N2) included thalamus, auditory, and sensorimotor areas, while in deep sleep σ was generated in sensorimotor areas, frontal eye field, and parahippocampal regions. Maps represent the highest 60% of the difference between baseline and poststimulus intervals, where y and x coordinates indicate locations in MNI space. Toi = time of interest.

Figure 3.

Difference between sleep stages in brain response to auditory stimuli. In (A) MEG sensors and (B) EEG sensors, the transition from wakefulness to NREM sleep is characterized by significant changes in induced brain responses. The deeper the sleep (N1 to N2 to N3), the stronger the response in the α/σ frequency band (shown in negative values, blue) to auditory stimuli. Additionally, deep N3 sleep was characterized by a weaker response in the late (>50 0 ms) α/σ frequency window compared with N1/N2 (in red). Areas outlined in black represent significant clusters for between-sleep-stage contrasts.

Figure 4.

MEG/EEG brain response differences depending on stimulus saliency. Plots represent (A) MEG and (B) EEG oscillatory brain responses to two types of cues (own vs other names) separately, and the difference between them (subject's own first name minus other first name). Interestingly, the brain differentiated between the presented stimuli even during all NREM sleep stages. During N1 and N2, the own name induced stronger responses (in a broad 5-30 Hz oscillatory band). During N3, on the other hand, it was the other name that induced a stronger response in a 5-17 Hz frequency window. Areas outlined in black represent significant clusters. The depicted brain response to own and other names is normalized against absolute baseline between 0.5 and 0.1 s before stimulus onset.

Figure 5.

Own versus other name brain response difference between sleep stages. (A) MEG as well as (B) EEG activity revealed a more prominent α-σ frequency band response during drowsiness (N1) and light sleep (N2) compared with wakefulness (W). Areas outlined in black represent significant clusters for between-sleep-stage contrasts.

Figure 6.

Brain response differences depending on voice familiarity during sleep. Plots represent (A) MEG and (B) EEG oscillatory brain responses to names uttered by a familiar versus unfamiliar voice and indicate stronger responses to the unfamiliar voice during NREM sleep (which however does not reach statistical significance). The depicted brain response to familiar and unfamiliar voices is normalized against absolute baseline between 0.5 and 0.1 s before stimulus onset. Statistical analysis did not reveal any significantly different brain responses to the two types of voices in either wakefulness or sleep.

Figure 7.

Sleep spindle topography in EEG and MEG. A, Spontaneous sleep spindles detected in the prestimulus time window (−2 to −0.2 s relative to the stimulus onset) peak at ∼14 Hz and have the highest power over medial, frontocentral EEG sensors and lateral MEG sensors. MEG source reconstruction suggests deep sources surrounding thalamus, hippocampus, and pons. B, Histograms of sleep spindle onsets relative to stimulus onsets show accumulations at ∼500 ms after stimulus presentation, which however appears not to be name category-specific. x, y and z coordinates indicate location in MNI space.

Figure 8.

Frequency-specific brain responses evoked by all stimuli. Two top panels represent brain responses evoked by auditory stimuli in wakefulness and the three NREM sleep stages, in (A) MEG and (B) EEG. C, Source reconstruction of the MEG signal revealed involvement of the primary auditory cortices in the generation of the early (0.1-0.4 s after stimulus onset) evoked responses across wakefulness and NREM sleep. Maps represent the difference between baseline and poststimulus interval, and show the highest 60% of the relative change values, where x, y, z coordinates indicate the location in MNI space. D, The plots compare the appearance of ERP peaks with the narrow-band peaks and troughs in each NREM sleep stage separately. Vertical lines indicate peak-to-peak or trough-to-trough alignment within ±2 sample points. θ oscillations (4-7 Hz) generated the first ERP component during wakefulness and drowsiness. During consolidated NREM sleep (N2 and N3), on the other hand, the first ERP component seems to be generated by faster and phase-synchronized α (8-12 Hz) and σ (11-15 Hz) oscillations.