K-complexes modulate the processing of relevant sensory information during NREM sleep

Introduction: In as much as the sleeping brain responds differently to auditory stimuli of distinct characteristics, the role of such differential responses is not fully understood. K-complexes represent the most prominent brain response to sensory perturbations during non-rapid eye movement (NREM) sleep. Latest research proposes that K-complexes reflect both sleep protecting as well as arousal inducing processes. We sought to investigate the role of the evoked K-complex to different auditory stimuli, in order to disentangle the function of brain responses to external stimuli during NREM sleep, whether they reflect inhibition and sleep protection or rather sensory processing. Materials and Methods: We recruited 17 healthy humans (14 females) to sleep for a full night (~8h) in the laboratory while acquiring polysomnography data using high-density electroencephalography (EEG). During the night, we presented subjects with their own names as well as two unfamiliar names. These names were spoken by either a familiar voice to the subject (FV) or an unfamiliar one (UFV). We used an automatic algorithm to detect K-complexes, and performed event-related, time-frequency, and phase coherence analyses to unravel the ongoing brain processes during the auditory-evoked K-complex. Finally, we utilized a machine learning approach to differentiate between brain responses to different stimuli during NREM sleep. Results: We show that UFVs evoked more K-complexes than FVs; however, there was no difference in the number of evoked K-complexes between the names. The difference in the number of evoked K-complexes between FVs and UFVs appeared as early as 100ms post-stimulus and disappeared right after the stimulus presentation ends (mean stimulus duration 808ms). Moreover, by contrasting FV and UFV stimuli that evoked K-complexes, we observed that UFVs evoked a larger amplitude of the N550 component of the K-complex. Further analysis revealed that this difference in the amplitude of the N550 does not demonstrate larger amplitudes of the UFV-evoked K-complexes but rather stronger phase synchronization of brain responses to the onset of UFVs as shown by inter-trial phase coherence analysis. Spectral analysis revealed in the presence of the evoked K-complex, UFVs evoked stronger arousal-like response (>16Hz) relative to FVs. Finally, by training a linear discriminant analysis (LDA) classifier to decode between FV and UFV stimuli from post-stimulus brain activity, we show that only in the presence of evoked K-complexes, the classifier was able to decode the presented voice. Conclusions: Our results suggest the presence of time windows in NREM sleep during which the brain continues to respond preferentially to relevant sensory information. Central to such responses is the K-complex which, when evoked by sensory stimuli, reflects underlying brain processes that serve to extract and process relevant information. We propose that such dynamic reactivity to the environment entails the presence of a sentinel-processing mode where the brain remains connected to the environment while engaging in the vital processes that are ongoing during sleep.