A role for the thalamus in danger discrimination during sleep

A role for the thalamus in danger discrimination during sleep | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A role for the thalamus in danger discrimination during sleep Antoine Adamantidis, Ida Luisa Boccalaro, Mattia Aime, Florence Aellen, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3395895/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Sleep is associated with a sensory disconnection from the environment despite a high vulnerability to danger and predation. Yet, sensory stimuli-evoked responses persist in the brain of flies 1 , rodents 2,3 , primates 4,5 , and humans 6,7 during sleep. Whether discrimination between sensory stimuli occurs in the mammalian brain during sleep remains unclear. Here, we showed that neutral auditory stimuli evoked electrical responses propagate in parallel auditory and non-auditory pathway, some of which awaken sleeping mice. We used a convolutional neural network and identified neural activities of centro-medial thalamic (CMT) neurons as the most discriminant hub for auditory-evoked sleep-to-wake transitions among all recorded structures. Importantly, we found that prior associative learning of danger (conditioned stimulus, CS+) and neutral (CS-) auditory cues resulted in increased awakening events upon CS+ exposure during NREM, but not REM, sleep. These sleep-to-wake transitions were blocked by optogenetic silencing of CMT neurons during CS exposure in sleeping mice. Altogether, these results suggest a central role of the CMT neurons in the residual processing of behaviorally-relevant information in the sleeping brain. Biological sciences/Neuroscience/Circadian rhythms and sleep/Sleep Biological sciences/Neuroscience/Cognitive neuroscience/Perception Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction The sleeping brain maintains a residual ability to respond to external sensory information 8,9 as evidenced by the high probability of awakening to personally relevant auditory stimuli in humans (e.g., one’s own name, baby cry) 7,10–12 , and the sleep-dependent improvement of performance in word pair learning 13 , semantic discrimination 12,14,15 , and oddball auditory paradigm 2 during sleep. Accordingly, in both humans and animals, auditory evoked responses persist in thalamo-cortical circuits during NREM 16–19 and REM sleep 20 and arousal circuits 21 that induce sleep-to-wake transitions depending on the brain state, ongoing activity, oscillations, and cue saliency 22 . Yet, the brain mechanisms underlying the discrimination of auditory cues with content-relevant information – e.g., danger versus safety - during sleep remain unclear. Here, we investigated the brain and behavioral responses of sleeping mice subjected to re-exposure to danger or neutral auditory cues and identified the brain-wide mechanisms responsible for sensory integration during sleep. Hierarchy of brain wide responses to auditory stimuli across sleep-wake states. We first recorded neural responses evoked by neutral auditory stimuli across sleep-wake states in freely-moving mice. Chronic multi-site single-unit, local field potentials (LFPs) electrodes were stereotactically implanted in the primary auditory cortex (Au1), the dorsal medial geniculate (dMG), the central medial thalamus (CMT), and the hippocampus (HP; non-related auditory-processing structure), together with electroencephalogram/electromyogram (EEG/EMG) electrodes (see Methods). After habituation to the recording conditions, neural responses to pure tones were recorded across wakefulness, NREM, and REM sleep (100 ms duration; 1, 2.5, and 5 kHz frequency randomly delivered at 30, 55, and 80 dB over an hour, see Methods; Fig. 1 a). Animals showed rapid habituation to auditory stimulations (Fig. 1 b) and each of these reliably evoked local field potentials (LFPs) and single-unit responses (Fig. 1 c, and d-e, respectively) from all recorded sites (Extended Data Fig. 1 a). Auditory-evoked LFP responses in all regions of interest exhibited short latencies (< 50 ms) during wakefulness, NREM, and REM sleep (Extended Data Fig. 1 b). Interestingly, CMT LFPs showed fast responses to auditory stimuli as compared to HP and dMG in particular during NREM sleep, characterized by high amplitude potentials and increased spiking activities compared to Au1 and HP (Extended Data Fig. 1 b, c). Note that reverse polarity in Au1 accompanied by a decrease in spiking rate during NREM sleep. Importantly, the spiking rate of CMT neurons was significantly higher for auditory stimuli that evoked NREM sleep-to-wake transitions ( A NREMs-WAKE) within a 5-s window latency (Fig. 1 f) than for those that did not wake the animals up ( A NREMs-NREMs; Fig. 1 f). This was selective for CMT neurons as evoked LFPs and spiking rates recorded from other structures were similar for both A NREMs-WAKE and A NREMs-NREMs transitions (Fig. 1 g-i). To test the implication of CMT neurons in auditory-evoked awakenings in a data-driven approach, we trained a convolutional neural network (CNN) using the brain-wide LFP recordings previously acquired (Fig. 1 ) to evaluate the brain structures that show the highest discriminative activity between A NREMs-NREMs and A NREMs-WAKE transitions upon auditory stimulation. After training, the network reached a mean AUC score of 0.80 ± 0.01 on the trained recordings, 0.80 ± 0.02 on the validation sets, and 0.79 ± 0.01 on the test experimental sets, indicating robust differences in local neuronal activity between A NREMs-NREMs vs. A NREMs-WAKE transitions at single-trial level (see Methods; Fig. 2 a). For both A NREMs-NREMs and. A NREMs-WAKE transitions, the sigma-filtered CMT recorded LFPs were mostly relevant for the network’s classification, and showed the highest saliency as compared to other recorded sites, including cortical EEGs (Fig. 2 b). Consistent with this result, we found that CMT showed an activation profile that significantly differed between A NREMs-NREMs vs A NREMs-WAKE transitions upon auditory stimuli, while all other structures tested had similar profiles (Fig. 2 c), confirming that the CMT nucleus is the most discriminant structure in driving awakening upon auditory stimuli. CNN trained on electrophysiological recordings filtered in the slow wave range (0.5-4 Hz) suggested a predominant drive by CMT, but also dMG and frontal EEG as compared to Au1 and parietal EEG (Extended Data Fig. 2 a-b). Consistent with the CNN model, we found that spindle rate showed a significant increase in CMT during A NREMs-NREMs as compared to A NREMs-WAKE transitions, and, to a lesser extent in other recorded structures (HP, dMG) except for the Au1 (Fig. 2 d). Note that similar observations were made for global spindles recorded from parietal EEG electrodes (Fig. 2 d). In contrast, auditory stimuli delivered during periods of low cortical EEG spindle rate resulted in an increased probability of A NREMs-WAKE transitions (Extended Data Fig. 2 c-d). Accordingly, the probability of A NREMs-WAKE transitions was significantly higher when auditory stimuli were phase-locked to the UP phase of a local CMT slow wave (Fig. 2 e), presumably because of CMT neuron hyper-excitability 23 . In contrast, this effect was not significant for other recorded sites (HP, dMG) while it was opposite for Au1 (DOWN phase) when neurons are in a low-excitable state (Fig. 2 e). Silencing of CMT neurons blocks auditory evoked awakening. To assess the role of CMT neurons in mediating auditory-evoked awakenings during sleep, adeno-associated viruses (AAV) AAV5-CamKII-ArchT3.0-eYFP or AAV2-CamKII-eYFP (control) were stereotactically injected into the CMT area and animals were chronically implanted with multi-site tetrodes in the Au1, dMG, CMT, and HP, an optic fiber, EEG and EMG electrodes (see Methods, Fig. 3 a-d). After habituation to the experimental conditions, we observed that optogenetic silencing of ArchT-expressing CMT neurons concomitantly to auditory stimulation significantly decreased the probability of auditory-evoked A NREMs-WAKE (Fig. 3 d-f, respectively, Extended Data Fig. 3 a-b) but had no effect on arousal transitions from REM sleep (Fig. 3 f). Furthermore, we found that auditory stimulation significantly decreased the number of slow waves during NREM sleep in all the recorded regions and both the EEGs, except the HP, an effect that was blocked by the optogenetic silencing of ArchT-expressing CMT neurons (Extended Data Fig. 3 c). Note that spindle rates (Extended Data Fig. 3 d) and sleep architecture (Extended Data Fig. 3 e) remained unchanged upon optogenetic manipulations. Discrimination of danger vs neutral cues during NREM, but not REM, sleep. We next tested whether CMT neurons also contribute to sensory discrimination of behaviorally-relevant stimuli associated with danger vs safety environment during sleep. Mice were implanted with single-unit recordings of CMT neurons combined with EEGs and EMGs (Fig, 4a). After recovery and habituation to the recording conditions, mice were trained to discriminate between two distinct auditory stimuli, with one conditioned stimulus (CS+, 5 kHz) always paired with mild electric foot shock and a neutral auditory CS- (1 kHz; see Methods; Fig. 4 a). Mice showed no preference for either of the two conditioned stimuli (CS + and CS-) during initial habituation (Day 1) and progressively developed discriminative abilities (Day 2) evidenced by the significant increase of freezing behavior in response to CS+, but not CS- (Habituation, Day 1 vs Recall, Day 3; see Methods, Fig. 4 a, b). After conditioning, animals returned to their home cage where they were allowed to sleep. We found that CS + stimuli alone (i.e., without shocks) induced A NREMs-WAKE transitions more frequently than CS- stimuli (Fig. 4 c). The spiking rate of CMT neurons was significantly higher during CS+-evoked A NREMs-WAKE transitions as compared to A NREMs-NREMs transitions (Fig. 4 d, respectively). In addition, we found that CMT neurons exhibited a significant decrease in bursting activity upon CS+-evoked A NREMs-WAKE transitions (post-tone) during these recall sessions as compared to their activity during spontaneous (pre-tone) A NREMs-WAKE transitions (‘CS + Rec’ in Fig. 4 f). This contrasts with their stable bursting activity in all other conditions and suggests that tonic activity of CMT neurons upon CS + exposure induces arousal 23 and supports sensory integration 24 . To test whether this switch of single-cell CMT neuron dynamics after fear conditioning is also associated with a re-organization of population activity during danger vs neutral cue discrimination, we longitudinally recorded the activity of single CMT neurons across the behavioral conditioning and cues exposure during post-sleep using deep two-photon calcium imaging in head-restraint sleeping mice. AAV1-CaMKII-GCaMP8m was stereotactically infused into the CMT area before EEG/EMG electrodes and an aberration-corrected endoscope lens 25 were chronically implanted on the skull of the animals (see Methods; Fig. 5 a). After habituation to the recording conditions, we longitudinally recorded the response of GCaMP8-expressing CMT neuron populations to CS- and CS + during NREM sleep immediately after Habituation (Day 1) and Recall (Day 3) (see Methods; Fig. 5 a). Consistent with the previous experiments, following conditioning to the aversive stimuli, all animals exhibited similar abilities to discriminate between neutral CS- and danger CS + during NREM sleep (Fig. 5 b). Interestingly, we found that the correlation of CS- vs CS + evoked response of the CMT cell populations during NREM sleep decreased following ‘Recall’ as compared to ‘Habituation’ session while it remained stable when the animal was awake (Fig. 5 c-e). Although single-cell calcium transients remained stable across the behavioral task and sleep-wake states (Fig. 5 f, g), after conditioning, we observed that all CMT neurons activity was significantly higher for CS+-evoked A NREMs-WAKE as compared to A NREMs-NREMs transitions (Fig. 5 i). Note that this analysis was restricted to the CS+, as CS- elicited awakening episodes were almost negligible after conditioning (Fig. 5 h). Altogether, these results suggest a spatial reorganization of CMT neural population responses to discriminative cues during NREM sleep (as represented in Fig. 5 d). Finally, we tested the contribution of CMT neurons to the discrimination of danger vs neutral cues during NREM sleep. Animals were stereotactically injected with AAV5-CamKII-ArchT3.0-eYFP and AAV2-CamKII-eYFP (control) into the CMT area and chronically implanted with optic fibers, EEG, and EMG electrodes (see Methods; Fig. 6 a). All animals showed similar performance in discriminating CS- (neutral) from CS+ (danger) stimuli after conditioning (Fig. 6 b). Consistent with previous findings, we observed that optogenetic silencing of ArchT3.0-expressing CMT neurons during NREM sleep blocked the discrimination between CS- and CS + cues in ArchT but not in control (eYFP) animals, as revealed by their significant decrease in CS+-evoked A NREMs-WAKE transitions in Recall as compared to controls (Fig. 6 c). None of the optogenetic manipulation affected the sleep-wake architecture (Fig. 6 d, f). Consistent with our results, no behavioral differences were observed when CMT neurons were optogenetically silenced during REM sleep (Fig. 6 e). These results suggest that CMT neuronal activity contributes to danger discrimination during NREM, but not REM, sleep. Discussion Sleep and its low awareness or responsiveness to environmental stimuli in the animal kingdom have evolved through the adaption of its architecture (or pattern), timing, and habitat relative to the perceived risk of danger including temperature, light, noise, wind, or predation. These trade-offs emerge from awake experience and sensory information storage prior to sleep as an optimization strategy for the physiological needs of the individuals and ultimately the species 26–28 . The state of apparent vulnerability associated with sleep is compensated by the fluctuating threshold of sensory disconnection during successive sleep states 29–31 and the regular awakenings of other individuals acting as sentinels in gregarious species 32–34 . Additional adaptive responses during sleep are likely to engage mechanisms in the central nervous system capable of detecting environmental threats based on various sensory modalities 7,22,28 . Our results revealed that the medio-dorsal thalamic neurons play a crucial role in evoking sensory-driven arousal from NREM sleep 21,23,35–38 . The short latencies of auditory responses observed across all regions of interest during wakefulness are consistent with previous research on sensory processing in cortical and hippocampal circuits 39 . During NREM sleep, the reverse polarity accompanied by a decrease in spiking rate in Au1 suggests the formation of auditory-evoked K-complexes, supporting the notion that these events are a fundamental component of NREM sleep, representing cortical isolated ‘downstate’ 40 triggered by thalamic activity 41 . Although auditory stimuli during sleep primarily trigger the auditory system 2 , parallel circuits are activated and include the locus coerulus 21 , hippocampus 42 and medio-dorsal thalamus 36 through mechanisms that remain to be determined. Prediction from a CNN model and experimental validation expands the role of global spindles in protecting sleep 16,17,31,43–45 to local CMT neuronal activity in the spindle frequency range in auditory-evoked awakening from NREM sleep. These findings reflect the central position of the medio-dorsal thalamus and its engagement in stress or fear reaction 37,46 , sensory encoding 36 and working memory 47–49 . Consistent with a role in the ‘Wakeup call’ hypothesis 24 , the CMT neural activity switch of burst-to-tonic discharge mode appears as a selective response to environmental danger. The absence of sensory-evoked awakenings during REM sleep in our study reflects the higher sensory threshold reported during that state in humans 50,51 and animals 21 in similar experimental paradigms, presumably as a protective mechanism against disruptive environmental stimuli 27 .In addition, the restorative functions of REM sleep are seen as prerequisites for effective waking function with brief awakenings from REM sleep preparing the organism for immediate fight or flight 28,52 , as compared to awakenings from NREM sleep. Our findings, along with previous research 10,11,22,27,28,30,50,52,53 , collectively support the concept of NREM sleep as a vigilant state for monitoring the environment for potential threats, and REM sleep representing a disconnected state, potentially increasing vulnerability. Together with their role in integrating other sensory modalities during sleep (e.g., olfaction 3 ), our findings suggest a ‘sentinel’ role for CMT neurons that integrate behaviorally-relevant subcortical and cortical inputs during sleep 21,36,54 . Collectively, our results advocate for residual processing of self-related information during NREM sleep and expand the repertoire of brain mechanisms at play during sleep such as unihemispheric sleep in marine mammals 30 , sleep while flying in birds 55 , or visual scanning in gulls 56 that optimize sleep over behavioral trade-offs to surrounding threats. These findings open new avenues to further understand information integration during sleep in mice as a reductionist model for the study of consciousness in health and diseases. Methods Animals All experimental procedures, including animal handling, surgery, and experiments followed the Canton Bern Swiss Veterinary Office guidelines (license n. BE 129/2020). Adult male C57BI6 mice from Janvier Labs (FR), 8 to 12 weeks old at the time of the surgery were used for in vivo electrophysiological, optogenetic, and behavioral experiments. Mice were single-housed in Plexiglas cages at constant temperature (20 to 23 C), humidity (40 to 60%), and circadian cycle (12-hour light/dark cycle, starting at 08:00 a.m.). Food and water were available ad libitum. After surgery, mice underwent a recovery period with three days of subcutaneous administration of analgesic (Metacam). On the 6th day, they were chronically tethered to recording cables (and optic fibers, respectively); the experiment started just after 10 days to let the mice recover from the surgery and habituate to the experimental conditions. Viral injections 6 weeks old mice were anesthetized with isoflurane (5% for induction, 1.25–1.75% for maintenance) in oxygen and placed on a stereotaxic frame (Model 940, David Kopf Instruments). Before all surgical procedures, an injection of saline and Metacam was given subcutaneously. After shaving, a midline incision along the skull was made in order to ensure proper positioning of the skull aligning Bregma and Lambda lines (around 4.6 caudally). Under microscopic control, a craniotomy was made using a surgical drill. The injection was done with a Hamilton syringe (7000 series, model 7000.5, 0.5 uL volume) and a syringe pump (Pump 11 Elite Nanomite Infusion/Withdrawal Programmable Syringe Pump, Harvard Apparatus).AAV5-CamK-ArchT (optogenetics groups), AAV2-CamKII-eYFP (optogenetics control groups), or AAV1-CaMKII-GCaMP8m (imaging groups) viruses were injected in CMT (1.58 AP, 0.75 ML, 4.1 DV, 10 ° angle, 200 nl) according to Paxinos’ and Franklin's mouse brain atlas at an injection rate of 50nl/min. All plasmids came from the University of Zurich Viral Vector Facility. Tetrode Implantation Electrodes used for EEG and grounding signals were made of stainless-steel screws, for EMG signals instead were used bare-ended steel wires. Tetrodes were obtained with four strands of 10-um tungsten wire (CFW0010954, California Fine Wire) which were twisted and connected to an electrode interface board by gold pins (EIB-36-PTB, Neuralynx). Anesthesia was induced using isoflurane in oxygen and maintained using a mix of medetomidine (0.27 mg/kg), midazolam (5mg/kg), and fentanyl (0.05 mg/kg). As described above, the animals were placed on the stereotaxic frame and the holes for the position of the electrodes were drilled. The preparation consisted of the implantation of two EEGs, one placed in the skull above the frontal lobe and one above the parietal lobe; the ground screw was placed above the cerebellum and the two EMG wires were sutured to the trapezoid muscle. Tetrodes were implanted unilaterally in CMT (1.58 AP, 0.75 ML, 4.1 DV, 10 ° angle), HP (2.2 AP, 1.5 ML, 1.9 DV), Au1 (2.5 AP, 4.5 ML, 2 DV), and dMG (3.2 AP, 2.2 ML, 3.7 DV). For optogenetic experiments, an optic fiber of 200 um diameter was additionally implanted in CMT via attachment to the respective tetrode in ArchT and eYFP mice. Once the electrodes were fixed by applying Tetric EvoFlow cement by Ivoclar Vivadent, the EEGs (frontal, parietal, and the ground) and the EMG wires were connected to the interface board. To finalize the surgery, Paladur methacrylate cement was applied to fix the implant and protect the skull surface. Anesthesia was terminated by injecting Atipamezole, Naloxone, and Flumazenil subcutaneously. Endoscope Implantation For CMT 2-photon imaging experiments, 6-weeks old male mice were anesthetized with an intraperitoneal (i.p.) injection of a mix containing medetomidine (0.27 mg kg-1), midazolam (5 mg kg-1) and fentanyl (0.05 mg kg-1) in sterile NaCl 0.9% (MMF-mix). Analgesia was achieved by local application of 100 µl of lidocaine (lurocaine, 1%) and subcutaneous (s.c.) injection of metacam (meloxicam, 5 mg kg 1). 40 µl of dexamethasone (Methameson, 0.1mg ml-1) was administered intramuscularly (i.m.) in the quadriceps to prevent inflammation potentially caused by the friction of the drilling. A heating-pad was positioned underneath the animal to keep the body temperature at 37ºC. Eye dehydration was prevented by topical application of eye ointment. The skin above the skull was disinfected with modified ethanol 70% and betadine before an incision was made. An 800-µm-diameter craniotomy was drilled above the CMT. A small track was made with a 0.7 mm sterile needle through the tissue (down to 2.8 mm from the brain surface) to aid endoscope insertion. A custom 500 µm-diameter aberration-corrected microendoscope (type II) 25 was slowly inserted (rate: 1µm/sec) above the CMT, as previously described 56 . Custom microendoscopes were based on a GRIN rod and a corrective polymer lens providing an enlarged field-of-view (FOV) and more homogeneous spatial resolution across the FOV, compared to uncorrected endoscopes 25 . The implant was cemented to the skull with dental acrylic and dental cement. For polysomnographic recordings, three EEG electrodes made of stainless steel screws were placed in the skull to record EEG signals (screw #1: AP: +2.5 ML: ±3.0 mm; screw #2: AP − 2.3 mm, ML ± 2.0 mm; reference screw: AP − 4.3 mm, ML + 0.5mm) and two EMG bare-ended wires were sutured to the trapezius muscle of the neck to record muscle activity signals. In vivo electrophysiological recordings For all the experiments, mice were connected to a tether, digitizing head stage (RHD2132, Intan Technologies), and recordings were done at 20 kHz using an open-source software from Intan Technologies (RHD2000). For the optogenetic experiments additionally to the tetrodes cables, the mice were connected to patch cords coated with black tubing. All the connections within the optic fiber and patch cord, as well as the cement of the implants, were covered by a black varnish in order to reduce the possible excess of light that could disturb the natural sleep of the animals. The experiment started after 10 days, and the auditory stimulation protocol was delivered for 14 sessions after a day of baseline recording. Two sessions have been performed each day, one hour in the morning and one hour in the afternoon. Auditory stimulation All the experiments were conducted in the same recording room with the same researcher working on them. The mice were in their home cage for the duration of the experiment. All sounds were programmed in MATLAB, where one channel was routed before to the PulsePal system and then to 8 speakers (one speaker per each recorded animal) and the other channel of the PulsePal was routed to the electrophysiology acquisition system. The sounds were played free field through a speaker mounted 30 cm above the animal. For optogenetic experiments, an additional channel was connected from the PulsePal to the lasers, in order to have a precise synchronization of all the outputs. Auditory stimuli included pure tones of 100 ms duration and were interrupted by random gaps of the silence of variable duration (from 1 sec to 20 sec). All the stimuli were presented at three different intensity levels (30-, 55-, and 80-dB SPL) and at three different frequencies (1, 2.5, and 5 kHz). In vivo optogenetics For in vivo optogenetic-silencing recording, 2 weeks after viral vector injection, a tetrode was implanted in the CMT (same electrode specifications as in the tetrode implantation section), coupled with a 100-um diameter optic fiber mounted 100-um above the top electrode contact. Light intensity at optic fiber tips was measured with a power meter (Thorlabs PM100D) before optic fiber insertion (output ≃30 mW). The same protocol mentioned above (in vivo electrophysiological recording) was performed with the difference that the laser was delivered 500 ms before and 500 ms after the auditory stimuli onset. 2-photon laser scanning microscopy Head-fixed mice were placed and trained under the microscope every day for at least 6 days prior to the experiment, and then longitudinally imaged using an in vivo 2PLSM (Scientifica HyperScope) equipped with a ×16 objective (0.8 NA, Nikon). ScanImage Software (Vidrio Technologies, LLC) was used to control the microscope, the acquisition parameters, and the TTL-driven synchronization between the acquisition and EEG/EMG recordings. GCaMPs were excited using a Ti: sapphire laser operating at λ = 910 nm (InSight X3, Spectra-Physics) with an average excitation power at the focal point lower than 50 mW. Time-series images were acquired within a field-of-view of 117 x 117 µm (512x512 pixels). All image analyses were performed using Fiji ImageJ and a custom routine in MATLAB. Each imaging session contained a random presentation of intermingled auditory cues (1kHz or 5kHz) during wakefulness or NREM sleep and automatically aligned to the acquired images with a custom-made MATLAB script. No photo-bleaching or photo-toxicity was observed. Acquired images were then corrected by adjusting XY motion artifacts using Suite2p Software 57 . Regions of interest (ROIs) of cell bodies were selected and drawn manually. All pixels within each ROI were first averaged providing a single time-series of raw fluorescence. Raw calcium traces were then normalized to “F” estimated through a Gaussian distribution-based approach, as previously described 56 . Normalized traces are referred to as ‘ΔF/F’ throughout the paper. Ca2 + event detections Calcium events were detected using MATLAB custom scripts. Each ROI ΔF/F trace was analyzed separately after motion correction. Traces were first up sampled via interpolation to be smoother, with cubic spline interpolation (spline function in Matlab). The baseline median ‘M’ and noise level ‘E’ for each trace is estimated by computing the median and standard deviation of the signal in three iterations, the points exceeding 2 times the standard deviation are excluded from the estimation of M and E in the following iteration. Candidate calcium events were identified via the use of the find peaks function in Matlab, using twice the estimated M + E as a threshold for minimum peak height, and M + E as minimum peak prominence. The characteristics of the candidate events, such as duration, peak amplitude, and integral were computed. Each measured characteristic was then compared with minimum and maximum physiological characteristics, and if not respecting these, excluded. The candidate events were then filtered and confirmed as Ca2 + events if their measured characteristics were satisfying some minimum physiological parameters. Physiological parameters used for filtering candidate events were: Minimum Duration of an Event = 1 [seconds]; Maximum Duration of an Event = 20 [seconds]; Maximum Ratio between rise and decay duration = ⅓; Minimum Peak Amplitude = 2 [a.u. (ΔF/F)]; Minimum 10 Integral = 1 [a.u. (ΔF/F) * seconds]. The measured events were grouped and an average measure of events per cell, per sleep state (as defined by the sleep hypnogram obtained from the EEGs and EMG scoring), was computed within 15 seconds of each auditory stimulus presentation. For population analysis, vectors containing ΔF/F neuron response amplitude upon auditory stimulation were computed. Pearson correlation analyses were then computed between vectors representing different auditory stimuli (CS- or CS+) at different times (Habituation or Recall) and during different states (Wake or NREM sleep). Fear conditioning After a baseline of two hours per 14 days (same protocol as mentioned before in the section on auditory stimulation), the mice were habituated by gently handling them for 5 min on 5 consecutive days, then on day 28, the fear conditioning protocol started. On the first day of the procedure (Habituation) at ZTO, a foreign cage with a metal grid as a floor (Context A) was wiped with 70% ethanol and the mice were placed in it. The walls were marked with stars to provide additional contextual information. The ice was first given 3 min of time to explore the novel environment, followed by playing a first auditory stimulus (CS-) consisting of 27 pure tones of 100 ms duration at 1 kHz played over 30 sec for 5 times with a variable interstimulus interval (ISI) between 10s and 30s. Then, a second auditory stimulus (CS+) was played under the same conditions but at 5 kHz. 24 h later, the acquisition of fear memories was performed by wiping context A with 70% ethanol again and subsequently placing the animals in the same cage with the metal grid. After three minutes, CS- and CS + were played intermixed. An unconditional stimulus (US) was paired with CS + by applying an electric foot shock through the metal grid on the floor at 0.5 mA for 1 sec, starting when CS + ended. Another 24h later (Recall), a novel cage (Context B) was wiped with 1% acetic acid, and mice were placed in it. After three minutes of explorations, CS- and CS + were presented to the mice with the same protocol as during habituation. Freezing behavior was measured as a measure of learning performance. It was scored manually as the absence of any movement except breathing. Freezing to CS-/CS + was measured during the time the tone was playing, and only the first two rates of CS- and CS + were considered. Freezing was quantified as the total time of freezing during the total playing time of the tones. If mice generalized on day 3 (Recall), they were excluded from the experiment. The mice that were able to discriminate between the CS- and CS + during recall time, underwent the last part of the experiment. Immediately after the cued recall memory test, the mice returned to their home cage, where they were recorded and re-exposed to the safety (CS-) and dangerous (CS+) sound for 2/4 hours while they were sleeping. We tested two different modulations: the first group of mice was exposed to an NREMs-specific auditory stimulation with CMT optogenetic-silencing (n = 11), and the second group to REMs-specific auditory stimulation with CMT optogenetic-silencing (n = 10). Data analysis Pre-processing of LFP data and peaks analysis. Following the acquisition, LFP row recordings were sampled to 1000 Hz and re-referenced with a common average reference to reduce possible volume conduction. Then, the data were z-scored. After detecting LFP responses to auditory stimulation, a comprehensive analysis was carried out to examine the timing and amplitude of the peaks, calculating the average of all the trials per mouse. This analysis focused on identifying the latency to the first LFP peak, specifically the early activity in response to the stimulation (ranging from 0 to 50 ms) and the late component (ranging from 50 to 100 ms) of the auditory-evoked response, as presented in Extend Fig. 1 b. We also analyzed the strength of the response by calculating the amplitude of the detected LFP peaks for both the early and late components of the response (Extend Fig. 1 c). Sleep scoring. Sleep scoring was performed manually, based on frequency and amplitude characteristics of the EEG and EMG in custom software written in MATLAB. NREMs was identified by high amplitude, synchronous activity in the EEG with a prominent delta (0.5–4 Hz) frequency power, and low EMG activity; REMs was defined by high synchronous theta (6–9 Hz) and flat EMG; wake was characterized by an increase in EMG activity with a low-amplitude, high-frequency (> 6 Hz) EEG. Arousal threshold analysis. We analyzed the averaged LFP activity in response to the stimuli splitting the recording into two groups, based on the behavior of the animals. In blue are the events in which the animals were not waking up from 0 after 5 sec of the auditory stimulations ( A NREMs-NREMs transition), and in red are the events in which the animals were waking up from 0 to 5 sec after the auditory stimulation ( A NREMs-WAKE transitions). As previously described in the section ‘Sleep scoring’, the different states (WAKE, NREMs, and REMs) were defined manually by the EEG/EMG recording. To assess how the animals were awakening from sleep in response to the different tones used in the fear conditioning protocol (CS- and CS+), we used a similar analytical approach as described earlier measuring the events in which the animals woke up within 0 to 5 sec after the stimulation. However, in this case, we calculated the percentage of trials in which the animal woke up in response to each tone (CS- or CS+), which was obtained by dividing the number of trials in which the animal woke up for the specific tone by the total number of stimulations, i.e., the sum of the number of CS- and CS + trials. It is important to mention that the protocol included an equal number of CS- and CS + tones. Next, we computed the difference in awakening between the CS- and CS + tones by calculating the delta of awakening, which is obtained by subtracting the percentage of awakening in response to the CS- tone from the percentage of awakening in response to the CS + tone. This analysis allowed us to determine whether the animals exhibited a differential response to the two tones, as indicated by a higher percentage of awakening for the CS- tone compared to the CS + tone. Slow wave detection. Individual slow waves were detected during NREM sleep using the SWA-MATLAB toolbox. The beginning of the slow wave was marked at the positive to negative zero-crossing before the negative peak and the end of the slow wave was marked at the end of the subsequent positive slope. Spindle detection. With a custom-made script, the spindles were detected for each LFP channel and for the EEGs (spindleDetection_pathLoop). As a first step, the spindle rate was calculated for all the different recordings (14 sessions per animal), then averaged per mouse. Digging more into the analysis we also wanted to analyze the role of spindles in A NREMs-to-WAKE transitions, so we plot all the spindles found in events in which the animals were staying in NREMs ( A NREMs-NREMs in blue) and events in which animals were waking up after 10 sec of auditory stimulation onset ( A NREM-WAKE in red) before (-5;0 sec) and after (-5;10 sec) the auditory stimulation onset. Then we calculated the percentage of spindles in A NREMs-NREMs and A NREMs-to-WAKE transitions (%= (spindles found in the time window of -1 + 1 sec of the stim onset/number of events) *100). Single Unit activity and spike sorting. Single-unit activity was detected using the Offline Sorter Application by Plexon Neurotechnology Research System (version 4.4.1). Raw LFP data was first band-pass filtered (500-4000Hz, Butterworth filter) and then a threshold for multiunit activity was set manually (depending on the recording from − 2 to -6 SD). Single unit activity was then extracted using principal components analysis and manually extracting clusters. The spike ratio for the PSTH was calculated as the total number of spikes found for each trial, divided per number of events, and averaged for all the animals. For the quantification per animal, there were calculated the number of spikes found during the stimulus onset (from 0 to 0.1 sec) divided per event and then averaged per animal. Burst firing of single units was detected as a minimum of three consecutive action potentials with ISI < 6 msec and preceded by a quiescent hyperpolarized state of at least 60 msec. On the basis of the observed pattern of response, a time window (0-0.1, 0-0.2, 0-0.3, 0-0.4, 0-0.5 sec) was selected to analyze the response of a given neuron. For each cell, the values of spontaneous (before the tone) and evoked (during/after the tone) activity obtained during maintained NREM sleep ( A NREMs-NREMs) were compared with the events with the transition to wake ( A NREMs-WAKE), using paired t-test. Decoding and Feature extraction with CNNs. A Convolutional Neural Network (CNN), with an EEGNet architecture 58 was trained to discriminate the conditions A NREMs-NREMs vs. A NREMs-WAKE based on LFP and EEG traces, extracted from a time interval of 10 seconds pre- to 5 seconds post- sound onset, using a tenfold cross-validation scheme. This network has already been successfully used to discriminate electrophysiological responses to auditory stimuli 59 . The network was trained on the raw LFP traces of the HP, CMT, and AU1 as well as the Frontal and Parietal EEG channels. The data of the CMT was filtered in the sigma range (10–40 Hz), as a focus was set specifically for spindles (Extended Data Fig. 2 a-b for a control analysis filtering CMT in the delta range). Accuracy was computed via the area under the receiver operating characteristic curve (AUC). The trained networks were then used to extract condition-specific activation maps 60 , reflecting the features that were most relevant to the network’s decisions, with the use of saliency map, as it is a common practice in the field 61 , Fig. 2 b, which were computed at the single-trial level for all correctly classified trials and then averaged over all mice and overall ten trained networks to increase stability, as in previous work 61 . Statistical analysis. Data were compared via two-way ANOVA, or t tests for parametric data, with post hoc Tukey’s corrections for multiple comparisons. Data distribution was assumed to be normal, but this was not formally tested. Values in the text are reported as mean ± standard error mean (S.E.M.) unless reported otherwise. For each experiment, sample numbers are indicated in the corresponding figure legends. Histology Tissue collection. At the end of all experiments, animals were anesthetized with isoflurane as described above and an electric current was sent through two channels of each tetrode (30 uA, 5 pulses at 2s) in order to induce gliosis at the very end of tetrode placement. After approximately 2h (time to create an internal scar), mice were euthanized with 15mg pentobarbital i.p., and the cardiovascular system was transfused with 30 mL of cold, heparinized phosphate-buffered saline (PBS), followed by 30 mL of 4% Paraformaldehyde (PFA) via puncture in the left ventricle of the heart. Brains were removed and kept in PFA at 4°C overnight and then they were cryoprotected in 30% sucrose for 48h. Next, they were flash-frozen in 2-methyl butane at approximately − 80℃, cut into 40 nm thick sections, and stored in PBS at 4℃. In order to confirm tetrode placement, the tetrodes were stained with 1,1-Dioctadecyl-3,3,3,3-tetramethyl-indocarbocyanin-perchlorat before implantation. After cutting the brain, sections with visible dye traces were selected and either Nissl-stained to reveal the gliosis. Immunohistochemistry. The viral expression specificity and efficacy were checked histologically by double staining of free-floating sections. The brain sections were washed in PBS 0.1% Triton X-100 (PBST) three times for 10 min each, blocked by incubation with 4% bovine serum albumin dissolved in PBST for 45 min and subsequently incubated with anti-GFP antibodies (AB_221569) for 24h at 4℃. Then sections were washed again in PBST (three times for 10 min each) and incubated with a secondary antibody (ab150073) that binds to the primary antibody for 1h at room temperature. Fluorescence microscopy. Images were acquired by a Nikon Ti-E microscope with a 40x resolution. For tetrode placement, a filter Cy3 was used. Instead of fluorescent staining. Methods References 57. Aime, M. et al. Paradoxical somatodendritic decoupling supports cortical plasticity during REM sleep. Science (1979) 376 , 724–730 (2022). 58. Pachitariu, M., Stringer, C. & Harris, K. D. Robustness of spike deconvolution for neuronal calcium imaging . Journal of Neuroscience 38 , 7976–7985 (2018). 59. Lawhern, V. J. et al. EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces . J Neural Eng 15 , 56013 (2018). 60. Aellen, F. M. et al. Auditory stimulation and deep learning predict awakening from coma after cardiac arrest. Brain 146 (2023) doi:10.1093/brain/awac340. 61. Simonyan, K., Vedaldi, A. & Zisserman, A. Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps. (2013). 62. Aellen, F. M., Göktepe-Kavis, P., Apostolopoulos, S. & Tzovara, A. Convolutional neural networks for decoding electroencephalography responses and visualizing trial by trial changes in discriminant features. J Neurosci Methods 364 , (2021). Declarations Acknowledgment We thank all the Tidis Lab members for their insightful discussion of, and comments on, previous versions of the manuscript. This work was supported by the Inselspital University Hospital Bern, and Interfaculty Research Cooperation (A.A. and C.G.H.), Swiss National Science Foundation (A.A.), the University of Bern (A.A.) and the EU H2020-ICT grant (A.S. and T.F.). Author contributions ILB, MA, and CGH performed the experiments. FA with the supervision of AZ trained the neural network and performed the computational approach analysis. ILB, FA, TR, MA, and MB analyzed the data. ILB and AA conceived the studies and wrote the manuscript with the help of all the authors. AA supervised the research. AS and TF provided technical assistance with the fabrication of aberration-corrected endoscopes. Conflict of Interest Statement The authors declare no competing interests. Data availability Statement All raw data used for analysis are made available through the Supplementary Information. Source data are provided in this paper. Code availability Statement All scripts used for analysis will be available through a GitHub repository (https://github.com/ZENLabCode). References French, A. S., Geissmann, Q., Beckwith, E. J. & Gilestro, G. F. Sensory processing during sleep in Drosophila melanogaster. Nature 598 , 479–482 (2021). Nir, Y., Vyazovskiy, V. V., Cirelli, C., Banks, M. I. & Tononi, G. Auditory responses and stimulus-specific adaptation in rat auditory cortex are preserved across NREM and REM sleep. Cerebral Cortex May; 25 (5):1362-78. (2015). Schreck, M. R. et al. State-dependent olfactory processing in freely behaving mice. Cell Rep 38 , (2022). Issa, E. B. & Wang, X. Sensory responses during sleep in primate primary and secondary auditory cortex. Journal of Neuroscience 28 (53):14467-80 (2008). Tasserie, J. et al. Deep brain stimulation of the thalamus restores signatures of consciousness in a nonhuman primate model. Sci. Adv 8 (2022). Hayat, H. et al. Reduced neural feedback signaling despite robust neuron and gamma auditory responses during human sleep. Nat Neurosci 25 , 935–943 (2022). Portas, C. M. et al. Auditory Processing across the sleep-wake cycle: simultaneous EEG and fMRI monitoring in humans. Neuron 28 (2000). Canales-Johnson, A., Merlo, E., Bekinschtein, T. A. & Arzi, A. Neural dynamics of associative learning during human sleep. Cerebral Cortex 30 , 1708–1715 (2020). Colrain, I. M. & Campbell, K. B. The use of evoked potentials in sleep research. Sleep Medicine Reviews 11 277–293. Oswald, I., Taylor, A. M. & Treisman, M. Discriminative responses to stimulation during human sleep. Brain 83 (1960). Poitras, R., Thorkildsen, A., Gagnon, M. A. & Naiman, J. Auditory discrimination during REM and NREM sleep in woman before and after delivery. Can Psychiatr Assoc J 18 (1973). Bastuji, H., Perrin, F., Garcia-Larrea, L. & Bastuji, H. Semantic analysis of auditory input during sleep: studies with event related potentials. International Journal of Psychophysiology 46 (2002). Züst, M. A., Ruch, S., Wiest, R. & Henke, K. Implicit vocabulary learning during sleep is bound to slow-wave peaks. Current Biology 29 , 541-553.e7 (2019). Ibáñez, A., López, V. & Cornejo, C. ERPs and contextual semantic discrimination: degrees of congruence in wakefulness and sleep. Brain Lang 98 , 264–275 (2006). Langford, G. W., Meddis, R. & Pearson, A. J. D. Awakening latency from sleep for meaningful and non-meaningful stimuli. Psychophysiology 11 , 1–5 (1974). Dang-Vu, T. T. et al. Interplay between spontaneous and induced brain activity during human non-rapid eye movement sleep. Proc Natl Acad Sci U S A 108 , 15438–15443 (2011). Schabus, M. et al. The fate of incoming stimuli during NREM sleep is determined by spindles and the phase of the slow oscillation. Front Neurol APR , (2012). Edeline, J., Manunta, Y. & Hennevin, E. Auditory thalamus neurons during sleep: changes in frequency selectivity, threshold, and receptive field size. J Neurophysiol 84 (2000). Simor, P., van der Wijk, G., Nobili, L. & Peigneux, P. The microstructure of REM sleep: Why phasic and tonic? Sleep Medicine Reviews 52 (2020). Hayat, H. et al. Locus coeruleus norepinephrine activity mediates sensory-evoked awakenings from sleep. Sci Adv 6 , (2020). van Kronenberg, P., Milinski, L., Kruschke, Z. & de Hoz, L. Sound disrupts sleep-associated brain oscillations in rodents in a meaning-dependent manner. Sci Rep 12 , (2022). Gent, T. C., Bandarabadi, M., Herrera, C. G. & Adamantidis, A. R. Thalamic dual control of sleep and wakefulness. Nature Neuroscience 21 (2018). Sherman S. Murray. A wake-up call from the thalamus. Nature Neuroscience 4 (2001). Antonini, A. et al. Extended field-of-view ultrathin microendoscopes for high-resolution two-photon imaging with minimal invasiveness. Elife 9 , 1–76 (2020). Lima, S. L., Rattenborg, N. C., Lesku, J. A. & Amlaner, C. J. Sleeping under the risk of predation. Animal Behaviour 70 723–736 (2005). Halperin, J. M. & Iorio, L. C. Responsivity of Rats to neutral and danger-signaling stimuli during sleep. Behavioral and Neural Biology 33 (1981). Tseng, Y. T. et al. The subthalamic corticotropin-releasing hormone neurons mediate adaptive REM-sleep responses to threat. Neuron 110 , 1223-1239.e8 (2022). Andrillon, T. & Kouider, S. The vigilant sleeper: neural mechanisms of sensory (de)coupling during sleep. Current Opinion in Physiology vol. 15 47–59 (2020). Rattenborg, N. C., Lima, S. L. & Amlaner, C. J. Facultative control of avian unihemispheric sleep under the risk of predation. Behavioural Brain Research . 105 (1999). Fernandez, L. M. J. & Lüthi, A. Sleep spindles: Mechanisms and functions. Physiol Rev 100 , 805–868 (2020). Loftus, J. C., Harel, R., Núñez, C. L. & Crofoot, M. C. Ecological and social pressures interfere with homeostatic sleep regulation in the wild. eLife 11 (2022). Karamihalev, S. et al . Social dominance mediates behavioral adaptation to chronic stress in a sex-specific manner. Elife 9 , 1–18 (2020). Samson, R. et al. Species-Specific Information for Enhancing Ecosystem Services. in The Urban Forest , pp. 111–144, Publisher: Springer (2017). McKenna, J. T. et al . Basal forebrain parvalbumin neurons mediate arousals from sleep induced by hypercarbia or auditory stimuli. Current Biology 30 , 2379-2385.e4 (2020). Shin, A. et al. A brainstem-to-mediodorsal thalamic pathway mediates sound-induced arousal from slow-wave sleep. Current Biology 33 (2023). Mátyás, F. et al . A highly collateralized thalamic cell type with arousal-predicting activity serves as a key hub for graded state transitions in the forebrain. Nat Neurosci 21 , 1551–1562 (2018). Tasserie, J. et al. Deep brain stimulation of the thalamus restores signatures of consciousness in a nonhuman primate model. Sci. Adv 8 (2022). Le Merre, P. et al. Reward-Based Learning Drives Rapid Sensory Signals in Medial Prefrontal Cortex and Dorsal Hippocampus Necessary for Goal-Directed Behavior. Neuron 97 , 83-91.e5 (2018). Cash, S. S. et al. The human K-complex represents an isolated cortical down-state. Science (1979) 324 , 1084–1087 (2009). Mak-McCully, R. A. et al . Synchronization of isolated downstates (K-Complexes) may be caused by cortically-induced disruption of thalamic spindling. PLoS Comput Biol 10 , (2014). Cusinato, R. et al. Intrinsic neural timescales in the temporal lobe support an auditory processing hierarchy. JNeurosci 43 (2023). Lecci, S. et al. Coordinated infraslow neural and cardiac oscillations mark fragility and offline periods in mammalian sleep. Sci Adv 3 (2017). Bastuji, H., Lamouroux, P., Villalba, M., Magnin, M. & Garcia-Larrea, L. Local sleep spindles in the human thalamus. Journal of Physiology 598 , 2109–2124 (2020). Andrillon, T. et al. Sleep spindles in humans: Insights from intracranial EEG and unit recordings. Journal of Neuroscience 31 , 17821–17834 (2011). Otake, K., Kin, K. & Nakamura, Y. Fos expression in afferents to the rat midline thalamus following immobilization stress. Neurosci Res 43 (2002). Schmitt, L. I. et al. Thalamic amplification of cortical connectivity sustains attentional control. Nature 545 , 219–223 (2017). Padilla-Coreano, N., Do-Monte, F. H. & Quirk, G. J. A time-dependent role of midline thalamic nuclei in the retrieval of fear memory. Neuropharmacology 62 457–463 (2012). Penzo, M. A. et al . The paraventricular thalamus controls a central amygdala fear circuit. Nature 519 , 455–459 (2015). Rechtschaffen, A., Hauri, P. & Zeitlin, M. Auditory awakening thresholds in REM and NREM Sleep stages. Percept Mot Skills 22 , 927–942 (1966). Pilon, M., Desautels, A., Montplaisir, J. & Zadra, A. Auditory arousal responses and thresholds during REM and NREM sleep of sleepwalkers and controls. Sleep Med 13 , 490–495 (2012). Snyder, F. Toward an evolutionary theory of dreaming. American Journal of Psychiatry 123 , 121–136 (1966). Hellman, K. & Abel, T. Fear Conditioning Increases NREM Sleep. Behavioral Neuroscience 121 , 310–323 (2007). Osorio-Forero, A. et al. Noradrenergic circuit control of non-REM sleep substates. Current Biology 31 , 5009-5023.e7 (2021). Rattenborg, N. C. Somnovolarism. Sleep Med 67 , 246–247 (2020). Beauchamp, G. Sleeping gulls monitor the vigilance behaviour of their neighbours. Biol Lett 5 , 9–11 (2009). Additional Declarations There is NO Competing Interest. Supplementary Files ExtendedDataFigures.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3395895","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":241361353,"identity":"705f7cef-13f6-4ec2-8745-2f20191c75a0","order_by":0,"name":"Antoine Adamantidis","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0003-2531-5175","institution":"[email protected]","correspondingAuthor":true,"prefix":"","firstName":"Antoine","middleName":"","lastName":"Adamantidis","suffix":""},{"id":241361354,"identity":"3e4c0f9b-009c-49b1-a78f-11c2e8107a7c","order_by":1,"name":"Ida Luisa Boccalaro","email":"","orcid":"","institution":"Zentrum für Experimentelle Neurologie, Department of Neurology, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Ida","middleName":"Luisa","lastName":"Boccalaro","suffix":""},{"id":241361355,"identity":"1b465143-4437-446b-8058-638ffc9d25a0","order_by":2,"name":"Mattia Aime","email":"","orcid":"","institution":"Zentrum für Experimentelle Neurologie, Department of Neurology, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Mattia","middleName":"","lastName":"Aime","suffix":""},{"id":241361356,"identity":"ba7dab28-f5ea-4574-aae5-a9b9defa9922","order_by":3,"name":"Florence Aellen","email":"","orcid":"https://orcid.org/0000-0001-9087-8359","institution":"Zentrum für Experimentelle Neurologie, Institute of Computer Science, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Florence","middleName":"","lastName":"Aellen","suffix":""},{"id":241361357,"identity":"ef3f3fcb-d1dd-40c8-9d40-1f98bd3a8beb","order_by":4,"name":"Thomas Rusterholz","email":"","orcid":"","institution":"Zentrum für Experimentelle Neurologie, Department of Neurology, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Rusterholz","suffix":""},{"id":241361358,"identity":"687176d8-0fcd-4403-aa83-fe85418ea8bf","order_by":5,"name":"Micaela Borsa","email":"","orcid":"","institution":"Zentrum für Experimentelle Neurologie, Department of Neurology, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Micaela","middleName":"","lastName":"Borsa","suffix":""},{"id":241361359,"identity":"de28ceb9-eab2-482f-b66e-559e80172869","order_by":6,"name":"Ivan Bozic","email":"","orcid":"","institution":"Zentrum für Experimentelle Neurologie, Department of Neurology, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Ivan","middleName":"","lastName":"Bozic","suffix":""},{"id":241361360,"identity":"dfb9d203-3398-4b5e-b272-0032b33f3d14","order_by":7,"name":"Andrea Sattin","email":"","orcid":"https://orcid.org/0000-0002-1345-2204","institution":"Optical Approaches to Brain Function Laboratory, Istituto Italiano di Tecnologia","correspondingAuthor":false,"prefix":"","firstName":"Andrea","middleName":"","lastName":"Sattin","suffix":""},{"id":241361361,"identity":"12dc4ea5-5ce2-4e2c-a8d8-2a07b064a54a","order_by":8,"name":"Tommaso Fellin","email":"","orcid":"https://orcid.org/0000-0003-2718-7533","institution":"Istituto Italiano di Tecnologia","correspondingAuthor":false,"prefix":"","firstName":"Tommaso","middleName":"","lastName":"Fellin","suffix":""},{"id":241361362,"identity":"bdf79ef2-7516-444e-97c8-4132f6a548e2","order_by":9,"name":"Carolina Gutierrez Herrera","email":"","orcid":"","institution":"Zentrum für Experimentelle Neurologie, Department of Neurology, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Carolina","middleName":"Gutierrez","lastName":"Herrera","suffix":""},{"id":241361363,"identity":"2db3877f-50a9-4786-baa5-f95f9c08f832","order_by":10,"name":"Athina Tzovara","email":"","orcid":"","institution":"Zentrum für Experimentelle Neurologie, Institute of Computer Science, Inselspital University Hospital Bern, University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Athina","middleName":"","lastName":"Tzovara","suffix":""}],"badges":[],"createdAt":"2023-09-28 15:11:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3395895/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3395895/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":49544258,"identity":"d4558e45-475a-4daa-932e-5685fe2b8452","added_by":"auto","created_at":"2024-01-12 17:47:36","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1233203,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAuditory-evoked responses in non-sensory CMT during \u003c/strong\u003e\u003csup\u003e\u003cstrong\u003eA\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003eNREMs-to-WAKE transitions.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(a) Experimental set-up: multi-site tetrode implanted in the central medial thalamus (CMT), hippocampus (HP), primary auditory cortex (Au1), and dorsal medial geniculate (dMG), together with EEG/EMG electrodes. Experimental timeline included recording of 1-h session(360 pure tones randomly distributed) twice a day during wake, NREMs, and REMs (n=14 sessions). (b) Percentage of total duration of wake, NREMs, and REMs. (c) Average data of LFP responses to 80 dB 5 kHz tones (in grey) during NREMs recorded from CMT, HP, Au1, and dMG. Auditory-evoked LFPs in which the animals continue to sleep (\u003csup\u003eA\u003c/sup\u003eNREMs-NREMs) are in blue (n= 1055 stimuli from m = 5 mice), others when the animals wake up (\u003csup\u003eA\u003c/sup\u003eNREMs-WAKE) in red (n= 546 stimuli from m = 5 mice). (d) Representative responses of CMT, HP, Au1, and dMG during \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs (blue) or\u003csup\u003e A\u003c/sup\u003eNREMs-WAKE (red). (e) Averaged peristimulus histogram (PSTH) of recorded units (CMT n = 17; HP n = 6; Au1 n = 12; dMG n = 12) during \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs (blue, CMT n = 255; HP n = 32; Au1 n = 179; dMG n = 165 stimuli) or\u003csup\u003e A\u003c/sup\u003eNREMs-WAKE (red, CMT n = 174; HP n = 19; Au1 n = 115; dMG n = 116 stimuli) from m = 5 mice. (f-i) Quantification of the highest peak LFP amplitude (left) and spike rate (right) of the auditory-evoked during \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs (blue) versus \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE (red) for CMT (f), HP (g), Au1 (h), dMG (i). Data were analyzed using paired t-test (*p\u0026lt;0.05; **p\u0026lt;0.01; ***p\u0026lt;0.001; ****p\u0026lt;0.0001) and presented as means ± S.E.M..\u0026nbsp;\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/6fb509c91d3e94f96ce26fff.png"},{"id":49544259,"identity":"dbfcd6d6-e058-4d49-8985-b72a55f35fa9","added_by":"auto","created_at":"2024-01-12 17:47:37","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":382489,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLocal CMT oscillations predicts\u003c/strong\u003e \u003cstrong\u003eANREMs-to-WAKE.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(a) Modelling pipeline used to train the Convolutional Neural Network (CNN) and extract discriminant features. The training was done with a tenfold cross-validation, on LFP and EEG traces filtered in the sigma band (from 10 to 40 Hz) and extracted from -10 to 5 sec relative to the sound’s onset. (b) Mean activation maps, quantifying the relevance of individual LFP/EEG traces across time into the network’s classification of whether a give trial was \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs (in blue, left) vs \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE (in red, middle) conditions. Right: the difference between the mean activation maps of the \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE (n = 546 trials) and \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs (n = 1055 trials) conditions. (c) Distribution of the mean network activation across trials per region. (d) Percentage of spindle rate recorded during \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs (in blue) and \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions (in red), for frontal and parietal EEGs, CMT, HP, Au1 and dMG. (e) Circular plot of the phase-locking of EEG slow waves (SWs) recorded from EEG, HP, CMT, Au1 and dMG to \u003csup\u003eA\u003c/sup\u003eNREMs-to-WAKE (red, n = 247 SWs from m= 5 mice) and \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs (blue, n = 9397 SWs from m = 5 mice) transitions. Data were analyzed using paired t-test for the slope and two-way ANOVA with Tukey post hoc test for the percentage of spindles (*p\u0026lt;0.05; **p\u0026lt;0.01; ***p\u0026lt;0.001; ****p\u0026lt;0.0001) and presented as means ± S.E.M.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/66948ddd53eba4202ac1ffbb.png"},{"id":49544260,"identity":"9922e850-fd57-45c7-bdbb-c6b978f5add6","added_by":"auto","created_at":"2024-01-12 17:47:37","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":409307,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOptogenetic-silencing of CMT neurons blocks \u003c/strong\u003e\u003csup\u003e\u003cstrong\u003eA\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003eNREMs-to-WAKE transitions.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(a) Experimental timeline and set up showing multi-site tetrodes implanted in CMT, HP, Au1, dMG, together with EEG, and EMG electrodes. (b) Representative recording traces from the different regions, and hypnogram with relative EEG/EMG traces. As shown, the laser was ON from 0.5 sec before and after the stimulus onset (Tone). (c) Representative spiking rate of control and ArchT-expressing CMT neurons upon optogenetic silencing without (left) or with (right) auditory (top left black, n = 90 trials; top right black, n = 100 trials, and bottom left yellow, n = 12 trials; bottom right yellow, n = 39 trials, respectively). (d) Average event-related LFP recorded in CMT in response to the auditory stimuli (in grey box) and to optogenetic silencing (yellow box) without optical stimulation (Laser OFF, blue trace; m = 10 mice), with optogenetic silencing in ArchT mice (Laser ON, red trace; m = 5) and eYFP mice (Laser ON, black trace; m = 5 mice). (e) Average spiking rate of CMT neurons upon optogenetic silencing during NREMs in ArchT (n= 10 units from m=5 mice, average of 110 trials) and eYFP mice (n=8 units from m=5 mice, average of 112 trials). (f) Percentage of awakening from NREMs (left) and REMs (right) in ArchT (n = 7028 during NREM; n=730 during REM from m = 5 mice) and eYFP mice (n = 6429 during NREM; n=553 during REM from m = 5 mice). Data were analyzed using unpaired t-test (*p\u0026lt;0.05; **p\u0026lt;0.01; ***p\u0026lt;0.001; ****p\u0026lt;0.0001) and presented as means ± S.E.M..\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/4431db701811871120e3ca5b.png"},{"id":49544262,"identity":"c32adda7-7d0f-419f-a408-55c033316e45","added_by":"auto","created_at":"2024-01-12 17:47:37","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":212573,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCMT neurons modulate safety vs danger perception during NREMs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(a) Experimental timeframe of fear conditioning and electrophysiological recordings. (b) Percentage of behavioral freezing upon CS- and CS+ exposure during Habituation and Recall sessions (m = 6 mice). (c) Percentage of awakening between the CS- and CS+: Delta values are calculated as the difference between [percentage of awakening for CS+] – [percentage of awakening for CS-] during Habituation (Hab) and Recall (Rec) sessions (m = 6 mice). (d) Average spiking rate of CMT neurons upon CS+ exposure during Habituation (n = 11 units from m = 6 mice) and Recall (n =13 units from m = 6 mice) sessions. (e) Average bursting rate of CMT neurons upon CS- and CS+ exposure during Habituation and Recall sessions for \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs transitions for spontaneous (Pre-tone) vs evoked (Post-tone) burst activity. (f) Average bursting rate of CMT neurons upon CS- and CS+ exposure in Habituation and for \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions for spontaneous (Pre-tone) versus evoked (Post-tone) burst activity. Data were analyzed using two-way ANOVA with Tukey post hoc test and paired t-test (*p\u0026lt;0.05; **p\u0026lt;0.01; ***p\u0026lt;0.001; ****p\u0026lt;0.0001) for the burst analysis and delta awakening and presented as means ± S.E.M..\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/96a6ca49b14a9ff69fde32fa.png"},{"id":49544776,"identity":"f4c969c0-0524-4869-8052-3e7b82ac8f86","added_by":"auto","created_at":"2024-01-12 17:55:37","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":714783,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle cell and population activity of CMT neurons across fear conditioning and sleep\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(a) Schematic of the experimental procedure. (Bottom) Expression profile of GCaMP8 expressing CMT neurons. (b) Freezing percentage upon CS- and CS+ presentation during Habituation and Recall (m = 4 mice). (c) Left: Representative correlation between vectors of ΔF/F response amplitudes; Color-coded population matrices of correlation coefficients upon CS- and CS+ presentation after Habituation or Recall test in wakefulness (middle) or NREM sleep (right). (d) Graphical representation of CMT neurons spatial reorganization after Recall test, selectively during NREM sleep. Circles represent neurons; filled colors represent selectivity to CS- (light red) or CS+ (light blue), respectively. (e) Quantification of correlation coefficients represented in (c). (f) Mean CS- and CS+ evoked ΔF/F activity profiles during wakefulness (light gray) or NREM sleep (dark gray) after Habituation or Recall test (n = 89 neurons from m = 4 mice). (g) Quantification of average integral values (ΔF/F*s) in response to CS- or CS+ during wakefulness (light gray) or NREM sleep (dark gray) after Habituation or Recall test. (h) Percentage of awakening between the CS- and CS+: Delta values are calculated as the difference between [percentage of awakening for CS+] – [percentage of awakening for CS-] during Habituation and Recall sessions (m = 6 mice). (i) Average CS+ evoked ΔF/F activity profiles during awakening from NREM sleep (red) or maintained sleep (blue) after Habituation or Recall test. Data were analyzed using one-way and two-way ANOVA with Tukey post hoc test, and paired t-test (*p\u0026lt;0.05; **p\u0026lt;0.01; ***p\u0026lt;0.001; ****p\u0026lt;0.0001) for delta of awakening and presented as means ± S.E.M..\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/5aa68f7206fe65697abfa4b1.png"},{"id":49545141,"identity":"11d5c0ff-bee6-4569-9c71-96c11b5b404b","added_by":"auto","created_at":"2024-01-12 18:03:37","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":271488,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOptogenetic-silencing of CMT neurons blocks danger vs safety discrimination during NREMs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(a) Experimental schematic showing injection sites and chronic optic fiber implant and EEG/EMG electrodes (left). Right: experimental timeframe including the Habituation, Conditioning, and Recall sessions on separate days and optogenetic silencing window. After memory recall, the mice were re-exposured to CS- and CS+ during NREMs (NREMs-specific auditory stim: eYFP m = 5 mice; ArchT m = 6 mice) or during REMs (REMs-specific auditory stim: eYFP m = 5 mice; ArchT m = 5 mice). (b) Percentage of freezing behaviors upon CS- and CS+ exposure during habituation and recall days (m = 21 mice). (c) Percentage of awakening between the CS- and CS+: Delta values are calculated as the difference between [percentage of awakening for CS+] – [percentage of awakening for CS-] during Habituation (Hab) and Recall (Rec) sessions for the NREM-specific group. (d) Total percentage duration of wake, NREMs and REMs in eYFP (m = 5 mice) and in ArchT (m = 6 mice) mice during optogenetic silencing of ArchT-expressing CMT neurons during Habituation and Recall recordings for the NREM-specific group. (e) Same as in (c) but for the REMs-specific group (eYFP m = 5 mice; ArchT m = 5 mice). (f) Same as in (d) but for the REMs-specific group (eYFP m = 5 mice; ArchT m = 5 mice). Data were analyzed using two-way ANOVA with Tukey post hoc test (*p\u0026lt;0. 05; **p\u0026lt;0.01; ***p\u0026lt;0.001; ****p\u0026lt;0.0001) and presented as means ± S.E.M..\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/cf64d98d886559fee2251470.png"},{"id":49545375,"identity":"e2086570-3910-4588-a715-2b79eb4c55c7","added_by":"auto","created_at":"2024-01-12 18:11:38","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3094893,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/5303e832-6dc9-4d91-9f22-da74583f0c11.pdf"},{"id":49544263,"identity":"b5faf499-2a7c-42de-b3c1-1e99a599725d","added_by":"auto","created_at":"2024-01-12 17:47:37","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2056173,"visible":true,"origin":"","legend":"","description":"","filename":"ExtendedDataFigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-3395895/v1/88b2c488b79bf6daac3e5984.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"A role for the thalamus in danger discrimination during sleep","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe sleeping brain maintains a residual ability to respond to external sensory information\u003csup\u003e8,9\u003c/sup\u003e as evidenced by the high probability of awakening to personally relevant auditory stimuli in humans (e.g., one\u0026rsquo;s own name, baby cry)\u003csup\u003e7,10\u0026ndash;12\u003c/sup\u003e, and the sleep-dependent improvement of performance in word pair learning\u003csup\u003e13\u003c/sup\u003e, semantic discrimination\u003csup\u003e12,14,15\u003c/sup\u003e, and oddball auditory paradigm\u003csup\u003e2\u003c/sup\u003e during sleep. Accordingly, in both humans and animals, auditory evoked responses persist in thalamo-cortical circuits during NREM\u003csup\u003e16\u0026ndash;19\u003c/sup\u003eand REM sleep\u003csup\u003e20\u003c/sup\u003e and arousal circuits\u003csup\u003e21\u003c/sup\u003e that induce sleep-to-wake transitions depending on the brain state, ongoing activity, oscillations, and cue saliency\u003csup\u003e22\u003c/sup\u003e. Yet, the brain mechanisms underlying the discrimination of auditory cues with content-relevant information \u0026ndash; e.g., danger versus safety - during sleep remain unclear. Here, we investigated the brain and behavioral responses of sleeping mice subjected to re-exposure to danger or neutral auditory cues and identified the brain-wide mechanisms responsible for sensory integration during sleep.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHierarchy of brain wide responses to auditory stimuli across sleep-wake states.\u003c/strong\u003e We first recorded neural responses evoked by neutral auditory stimuli across sleep-wake states in freely-moving mice. Chronic multi-site single-unit, local field potentials (LFPs) electrodes were stereotactically implanted in the primary auditory cortex (Au1), the dorsal medial geniculate (dMG), the central medial thalamus (CMT), and the hippocampus (HP; non-related auditory-processing structure), together with electroencephalogram/electromyogram (EEG/EMG) electrodes (see Methods). After habituation to the recording conditions, neural responses to pure tones were recorded across wakefulness, NREM, and REM sleep (100 ms duration; 1, 2.5, and 5 kHz frequency randomly delivered at 30, 55, and 80 dB over an hour, see Methods; Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea). Animals showed rapid habituation to auditory stimulations (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb) and each of these reliably evoked local field potentials (LFPs) and single-unit responses (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ec, and d-e, respectively) from all recorded sites (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea).\u003c/p\u003e\n\u003cp\u003eAuditory-evoked LFP responses in all regions of interest exhibited short latencies (\u0026lt;\u0026thinsp;50 ms) during wakefulness, NREM, and REM sleep (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb). Interestingly, CMT LFPs showed fast responses to auditory stimuli as compared to HP and dMG in particular during NREM sleep, characterized by high amplitude potentials and increased spiking activities compared to Au1 and HP (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb, c). Note that reverse polarity in Au1 accompanied by a decrease in spiking rate during NREM sleep. Importantly, the spiking rate of CMT neurons was significantly higher for auditory stimuli that evoked NREM sleep-to-wake transitions (\u003csup\u003eA\u003c/sup\u003eNREMs-WAKE) within a 5-s window latency (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ef) than for those that did not wake the animals up (\u003csup\u003eA\u003c/sup\u003eNREMs-NREMs; Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ef). This was selective for CMT neurons as evoked LFPs and spiking rates recorded from other structures were similar for both \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE and \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs transitions (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eg-i).\u003c/p\u003e\n\u003cp\u003eTo test the implication of CMT neurons in auditory-evoked awakenings in a data-driven approach, we trained a convolutional neural network (CNN) using the brain-wide LFP recordings previously acquired (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e) to evaluate the brain structures that show the highest discriminative activity between \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs and \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions upon auditory stimulation. After training, the network reached a mean AUC score of 0.80\u0026thinsp;\u0026plusmn;\u0026thinsp;0.01 on the trained recordings, 0.80\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02 on the validation sets, and 0.79\u0026thinsp;\u0026plusmn;\u0026thinsp;0.01 on the test experimental sets, indicating robust differences in local neuronal activity between \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs vs. \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions at single-trial level (see Methods; Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ea). For both \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs and. \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions, the sigma-filtered CMT recorded LFPs were mostly relevant for the network\u0026rsquo;s classification, and showed the highest saliency as compared to other recorded sites, including cortical EEGs (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eb). Consistent with this result, we found that CMT showed an activation profile that significantly differed between \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs vs \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions upon auditory stimuli, while all other structures tested had similar profiles (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ec), confirming that the CMT nucleus is the most discriminant structure in driving awakening upon auditory stimuli. CNN trained on electrophysiological recordings filtered in the slow wave range (0.5-4 Hz) suggested a predominant drive by CMT, but also dMG and frontal EEG as compared to Au1 and parietal EEG (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ea-b).\u003c/p\u003e\n\u003cp\u003eConsistent with the CNN model, we found that spindle rate showed a significant increase in CMT during \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs as compared to \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions, and, to a lesser extent in other recorded structures (HP, dMG) except for the Au1 (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ed). Note that similar observations were made for global spindles recorded from parietal EEG electrodes (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ed). In contrast, auditory stimuli delivered during periods of low cortical EEG spindle rate resulted in an increased probability of \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ec-d). Accordingly, the probability of \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions was significantly higher when auditory stimuli were phase-locked to the UP phase of a local CMT slow wave (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ee), presumably because of CMT neuron hyper-excitability\u003csup\u003e23\u003c/sup\u003e. In contrast, this effect was not significant for other recorded sites (HP, dMG) while it was opposite for Au1 (DOWN phase) when neurons are in a low-excitable state (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ee).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSilencing of CMT neurons blocks auditory evoked awakening.\u003c/strong\u003e To assess the role of CMT neurons in mediating auditory-evoked awakenings during sleep, adeno-associated viruses (AAV) AAV5-CamKII-ArchT3.0-eYFP or AAV2-CamKII-eYFP (control) were stereotactically injected into the CMT area and animals were chronically implanted with multi-site tetrodes in the Au1, dMG, CMT, and HP, an optic fiber, EEG and EMG electrodes (see Methods, Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea-d). After habituation to the experimental conditions, we observed that optogenetic silencing of ArchT-expressing CMT neurons concomitantly to auditory stimulation significantly decreased the probability of auditory-evoked \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ed-f, respectively, Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea-b) but had no effect on arousal transitions from REM sleep (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ef). Furthermore, we found that auditory stimulation significantly decreased the number of slow waves during NREM sleep in all the recorded regions and both the EEGs, except the HP, an effect that was blocked by the optogenetic silencing of ArchT-expressing CMT neurons (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ec). Note that spindle rates (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ed) and sleep architecture (Extended Data Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ee) remained unchanged upon optogenetic manipulations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDiscrimination of danger vs neutral cues during NREM, but not REM, sleep.\u003c/strong\u003e We next tested whether CMT neurons also contribute to sensory discrimination of behaviorally-relevant stimuli associated with danger vs safety environment during sleep. Mice were implanted with single-unit recordings of CMT neurons combined with EEGs and EMGs (Fig, 4a). After recovery and habituation to the recording conditions, mice were trained to discriminate between two distinct auditory stimuli, with one conditioned stimulus (CS+, 5 kHz) always paired with mild electric foot shock and a neutral auditory CS- (1 kHz; see Methods; Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ea). Mice showed no preference for either of the two conditioned stimuli (CS\u0026thinsp;+\u0026thinsp;and CS-) during initial habituation (Day 1) and progressively developed discriminative abilities (Day 2) evidenced by the significant increase of freezing behavior in response to CS+, but not CS- (Habituation, Day 1 vs Recall, Day 3; see Methods, Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ea, b). After conditioning, animals returned to their home cage where they were allowed to sleep. We found that CS\u0026thinsp;+\u0026thinsp;stimuli alone (i.e., without shocks) induced \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions more frequently than CS- stimuli (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ec). The spiking rate of CMT neurons was significantly higher during CS+-evoked \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions as compared to \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs transitions (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ed, respectively). In addition, we found that CMT neurons exhibited a significant decrease in bursting activity upon CS+-evoked \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions (post-tone) during these recall sessions as compared to their activity during spontaneous (pre-tone) \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions (\u0026lsquo;CS\u0026thinsp;+\u0026thinsp;Rec\u0026rsquo; in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ef). This contrasts with their stable bursting activity in all other conditions and suggests that tonic activity of CMT neurons upon CS\u0026thinsp;+\u0026thinsp;exposure induces arousal\u003csup\u003e23\u003c/sup\u003e and supports sensory integration\u003csup\u003e24\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTo test whether this switch of single-cell CMT neuron dynamics after fear conditioning is also associated with a re-organization of population activity during danger vs neutral cue discrimination, we longitudinally recorded the activity of single CMT neurons across the behavioral conditioning and cues exposure during post-sleep using deep two-photon calcium imaging in head-restraint sleeping mice. AAV1-CaMKII-GCaMP8m was stereotactically infused into the CMT area before EEG/EMG electrodes and an aberration-corrected endoscope lens\u003csup\u003e25\u003c/sup\u003e were chronically implanted on the skull of the animals (see Methods; Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ea). After habituation to the recording conditions, we longitudinally recorded the response of GCaMP8-expressing CMT neuron populations to CS- and CS\u0026thinsp;+\u0026thinsp;during NREM sleep immediately after Habituation (Day 1) and Recall (Day 3) (see Methods; Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ea). Consistent with the previous experiments, following conditioning to the aversive stimuli, all animals exhibited similar abilities to discriminate between neutral CS- and danger CS\u0026thinsp;+\u0026thinsp;during NREM sleep (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eb). Interestingly, we found that the correlation of CS- vs CS\u0026thinsp;+\u0026thinsp;evoked response of the CMT cell populations during NREM sleep decreased following \u0026lsquo;Recall\u0026rsquo; as compared to \u0026lsquo;Habituation\u0026rsquo; session while it remained stable when the animal was awake (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ec-e).\u003c/p\u003e\n\u003cp\u003eAlthough single-cell calcium transients remained stable across the behavioral task and sleep-wake states (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ef, g), after conditioning, we observed that all CMT neurons activity was significantly higher for CS+-evoked \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE as compared to \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs transitions (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ei). Note that this analysis was restricted to the CS+, as CS- elicited awakening episodes were almost negligible after conditioning (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eh). Altogether, these results suggest a spatial reorganization of CMT neural population responses to discriminative cues during NREM sleep (as represented in Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ed).\u003c/p\u003e\n\u003cp\u003eFinally, we tested the contribution of CMT neurons to the discrimination of danger vs neutral cues during NREM sleep. Animals were stereotactically injected with AAV5-CamKII-ArchT3.0-eYFP and AAV2-CamKII-eYFP (control) into the CMT area and chronically implanted with optic fibers, EEG, and EMG electrodes (see Methods; Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ea). All animals showed similar performance in discriminating CS- (neutral) from CS+ (danger) stimuli after conditioning (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eb). Consistent with previous findings, we observed that optogenetic silencing of ArchT3.0-expressing CMT neurons during NREM sleep blocked the discrimination between CS- and CS\u0026thinsp;+\u0026thinsp;cues in ArchT but not in control (eYFP) animals, as revealed by their significant decrease in CS+-evoked \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions in Recall as compared to controls (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ec). None of the optogenetic manipulation affected the sleep-wake architecture (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ed, f). Consistent with our results, no behavioral differences were observed when CMT neurons were optogenetically silenced during REM sleep (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ee). These results suggest that CMT neuronal activity contributes to danger discrimination during NREM, but not REM, sleep.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eSleep and its low awareness or responsiveness to environmental stimuli in the animal kingdom have evolved through the adaption of its architecture (or pattern), timing, and habitat relative to the perceived risk of danger including temperature, light, noise, wind, or predation. These trade-offs emerge from awake experience and sensory information storage prior to sleep as an optimization strategy for the physiological needs of the individuals and ultimately the species\u003csup\u003e26\u0026ndash;28\u003c/sup\u003e. The state of apparent vulnerability associated with sleep is compensated by the fluctuating threshold of sensory disconnection during successive sleep states\u003csup\u003e29\u0026ndash;31\u003c/sup\u003e and the regular awakenings of other individuals acting as sentinels in gregarious species\u003csup\u003e32\u0026ndash;34\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAdditional adaptive responses during sleep are likely to engage mechanisms in the central nervous system capable of detecting environmental threats based on various sensory modalities\u003csup\u003e7,22,28\u003c/sup\u003e. Our results revealed that the medio-dorsal thalamic neurons play a crucial role in evoking sensory-driven arousal from NREM sleep\u003csup\u003e21,23,35\u0026ndash;38\u003c/sup\u003e. The short latencies of auditory responses observed across all regions of interest during wakefulness are consistent with previous research on sensory processing in cortical and hippocampal circuits\u003csup\u003e39\u003c/sup\u003e. During NREM sleep, the reverse polarity accompanied by a decrease in spiking rate in Au1 suggests the formation of auditory-evoked K-complexes, supporting the notion that these events are a fundamental component of NREM sleep, representing cortical isolated \u0026lsquo;downstate\u0026rsquo;\u003csup\u003e40\u003c/sup\u003e triggered by thalamic activity\u003csup\u003e41\u003c/sup\u003e. Although auditory stimuli during sleep primarily trigger the auditory system\u003csup\u003e2\u003c/sup\u003e, parallel circuits are activated and include the locus coerulus\u003csup\u003e21\u003c/sup\u003e, hippocampus\u003csup\u003e42\u003c/sup\u003e and medio-dorsal thalamus\u003csup\u003e36\u003c/sup\u003e through mechanisms that remain to be determined. Prediction from a CNN model and experimental validation expands the role of global spindles in protecting sleep\u003csup\u003e16,17,31,43\u0026ndash;45\u003c/sup\u003e to local CMT neuronal activity in the spindle frequency range in auditory-evoked awakening from NREM sleep.\u003c/p\u003e \u003cp\u003eThese findings reflect the central position of the medio-dorsal thalamus and its engagement in stress or fear reaction\u003csup\u003e37,46\u003c/sup\u003e, sensory encoding\u003csup\u003e36\u003c/sup\u003e and working memory\u003csup\u003e47\u0026ndash;49\u003c/sup\u003e. Consistent with a role in the \u0026lsquo;Wakeup call\u0026rsquo; hypothesis\u003csup\u003e24\u003c/sup\u003e, the CMT neural activity switch of burst-to-tonic discharge mode appears as a selective response to environmental danger.\u003c/p\u003e \u003cp\u003eThe absence of sensory-evoked awakenings during REM sleep in our study reflects the higher sensory threshold reported during that state in humans\u003csup\u003e50,51\u003c/sup\u003e and animals\u003csup\u003e21\u003c/sup\u003e in similar experimental paradigms, presumably as a protective mechanism against disruptive environmental stimuli\u003csup\u003e27\u003c/sup\u003e.In addition, the restorative functions of REM sleep are seen as prerequisites for effective waking function with brief awakenings from REM sleep preparing the organism for immediate fight or flight\u003csup\u003e28,52\u003c/sup\u003e, as compared to awakenings from NREM sleep. Our findings, along with previous research\u003csup\u003e10,11,22,27,28,30,50,52,53\u003c/sup\u003e, collectively support the concept of NREM sleep as a vigilant state for monitoring the environment for potential threats, and REM sleep representing a disconnected state, potentially increasing vulnerability.\u003c/p\u003e \u003cp\u003eTogether with their role in integrating other sensory modalities during sleep (e.g., olfaction\u003csup\u003e3\u003c/sup\u003e), our findings suggest a \u0026lsquo;sentinel\u0026rsquo; role for CMT neurons that integrate behaviorally-relevant subcortical and cortical inputs during sleep\u003csup\u003e21,36,54\u003c/sup\u003e. Collectively, our results advocate for residual processing of self-related information during NREM sleep and expand the repertoire of brain mechanisms at play during sleep such as unihemispheric sleep in marine mammals\u003csup\u003e30\u003c/sup\u003e, sleep while flying in birds\u003csup\u003e55\u003c/sup\u003e, or visual scanning in gulls\u003csup\u003e56\u003c/sup\u003e that optimize sleep over behavioral trade-offs to surrounding threats. These findings open new avenues to further understand information integration during sleep in mice as a reductionist model for the study of consciousness in health and diseases.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eAnimals\u003c/h2\u003e \u003cp\u003e All experimental procedures, including animal handling, surgery, and experiments followed the Canton Bern Swiss Veterinary Office guidelines (license n. BE 129/2020). Adult male C57BI6 mice from Janvier Labs (FR), 8 to 12 weeks old at the time of the surgery were used for in vivo electrophysiological, optogenetic, and behavioral experiments. Mice were single-housed in Plexiglas cages at constant temperature (20 to 23 C), humidity (40 to 60%), and circadian cycle (12-hour light/dark cycle, starting at 08:00 a.m.). Food and water were available ad libitum. After surgery, mice underwent a recovery period with three days of subcutaneous administration of analgesic (Metacam). On the 6th day, they were chronically tethered to recording cables (and optic fibers, respectively); the experiment started just after 10 days to let the mice recover from the surgery and habituate to the experimental conditions.\u003c/p\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003eViral injections\u003c/h2\u003e \u003cp\u003e6 weeks old mice were anesthetized with isoflurane (5% for induction, 1.25\u0026ndash;1.75% for maintenance) in oxygen and placed on a stereotaxic frame (Model 940, David Kopf Instruments). Before all surgical procedures, an injection of saline and Metacam was given subcutaneously. After shaving, a midline incision along the skull was made in order to ensure proper positioning of the skull aligning Bregma and Lambda lines (around 4.6 caudally). Under microscopic control, a craniotomy was made using a surgical drill. The injection was done with a Hamilton syringe (7000 series, model 7000.5, 0.5 uL volume) and a syringe pump (Pump 11 Elite Nanomite Infusion/Withdrawal Programmable Syringe Pump, Harvard Apparatus).AAV5-CamK-ArchT (optogenetics groups), AAV2-CamKII-eYFP (optogenetics control groups), or AAV1-CaMKII-GCaMP8m (imaging groups) viruses were injected in CMT (1.58 AP, 0.75 ML, 4.1 DV, 10\u003cb\u003e\u0026deg;\u003c/b\u003e angle, 200 nl) according to Paxinos\u0026rsquo; and Franklin's mouse brain atlas at an injection rate of 50nl/min. All plasmids came from the University of Zurich Viral Vector Facility.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003eTetrode Implantation\u003c/h2\u003e \u003cp\u003eElectrodes used for EEG and grounding signals were made of stainless-steel screws, for EMG signals instead were used bare-ended steel wires. Tetrodes were obtained with four strands of 10-um tungsten wire (CFW0010954, California Fine Wire) which were twisted and connected to an electrode interface board by gold pins (EIB-36-PTB, Neuralynx). Anesthesia was induced using isoflurane in oxygen and maintained using a mix of medetomidine (0.27 mg/kg), midazolam (5mg/kg), and fentanyl (0.05 mg/kg). As described above, the animals were placed on the stereotaxic frame and the holes for the position of the electrodes were drilled. The preparation consisted of the implantation of two EEGs, one placed in the skull above the frontal lobe and one above the parietal lobe; the ground screw was placed above the cerebellum and the two EMG wires were sutured to the trapezoid muscle. Tetrodes were implanted unilaterally in CMT (1.58 AP, 0.75 ML, 4.1 DV, 10\u003cb\u003e\u0026deg;\u003c/b\u003e angle), HP (2.2 AP, 1.5 ML, 1.9 DV), Au1 (2.5 AP, 4.5 ML, 2 DV), and dMG (3.2 AP, 2.2 ML, 3.7 DV). For optogenetic experiments, an optic fiber of 200 um diameter was additionally implanted in CMT via attachment to the respective tetrode in ArchT and eYFP mice. Once the electrodes were fixed by applying Tetric EvoFlow cement by Ivoclar Vivadent, the EEGs (frontal, parietal, and the ground) and the EMG wires were connected to the interface board. To finalize the surgery, Paladur methacrylate cement was applied to fix the implant and protect the skull surface. Anesthesia was terminated by injecting Atipamezole, Naloxone, and Flumazenil subcutaneously.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003eEndoscope Implantation\u003c/h2\u003e \u003cp\u003eFor CMT 2-photon imaging experiments, 6-weeks old male mice were anesthetized with an intraperitoneal (i.p.) injection of a mix containing medetomidine (0.27 mg kg-1), midazolam (5 mg kg-1) and fentanyl (0.05 mg kg-1) in sterile NaCl 0.9% (MMF-mix). Analgesia was achieved by local application of 100 \u0026micro;l of lidocaine (lurocaine, 1%) and subcutaneous (s.c.) injection of metacam (meloxicam, 5 mg kg 1). 40 \u0026micro;l of dexamethasone (Methameson, 0.1mg ml-1) was administered intramuscularly (i.m.) in the quadriceps to prevent inflammation potentially caused by the friction of the drilling. A heating-pad was positioned underneath the animal to keep the body temperature at 37\u0026ordm;C. Eye dehydration was prevented by topical application of eye ointment. The skin above the skull was disinfected with modified ethanol 70% and betadine before an incision was made. An 800-\u0026micro;m-diameter craniotomy was drilled above the CMT. A small track was made with a 0.7 mm sterile needle through the tissue (down to 2.8 mm from the brain surface) to aid endoscope insertion. A custom 500 \u0026micro;m-diameter aberration-corrected microendoscope (type II)\u003csup\u003e25\u003c/sup\u003e was slowly inserted (rate: 1\u0026micro;m/sec) above the CMT, as previously described\u003csup\u003e56\u003c/sup\u003e. Custom microendoscopes were based on a GRIN rod and a corrective polymer lens providing an enlarged field-of-view (FOV) and more homogeneous spatial resolution across the FOV, compared to uncorrected endoscopes\u003csup\u003e25\u003c/sup\u003e. The implant was cemented to the skull with dental acrylic and dental cement. For polysomnographic recordings, three EEG electrodes made of stainless steel screws were placed in the skull to record EEG signals (screw #1: AP: +2.5 ML: \u0026plusmn;3.0 mm; screw #2: AP \u0026minus;\u0026thinsp;2.3 mm, ML\u0026thinsp;\u0026plusmn;\u0026thinsp;2.0 mm; reference screw: AP \u0026minus;\u0026thinsp;4.3 mm, ML\u0026thinsp;+\u0026thinsp;0.5mm) and two EMG bare-ended wires were sutured to the trapezius muscle of the neck to record muscle activity signals.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003eIn vivo electrophysiological recordings\u003c/h2\u003e \u003cp\u003eFor all the experiments, mice were connected to a tether, digitizing head stage (RHD2132, Intan Technologies), and recordings were done at 20 kHz using an open-source software from Intan Technologies (RHD2000). For the optogenetic experiments additionally to the tetrodes cables, the mice were connected to patch cords coated with black tubing. All the connections within the optic fiber and patch cord, as well as the cement of the implants, were covered by a black varnish in order to reduce the possible excess of light that could disturb the natural sleep of the animals. The experiment started after 10 days, and the auditory stimulation protocol was delivered for 14 sessions after a day of baseline recording. Two sessions have been performed each day, one hour in the morning and one hour in the afternoon.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003eAuditory stimulation\u003c/h2\u003e \u003cp\u003eAll the experiments were conducted in the same recording room with the same researcher working on them. The mice were in their home cage for the duration of the experiment. All sounds were programmed in MATLAB, where one channel was routed before to the PulsePal system and then to 8 speakers (one speaker per each recorded animal) and the other channel of the PulsePal was routed to the electrophysiology acquisition system. The sounds were played free field through a speaker mounted 30 cm above the animal. For optogenetic experiments, an additional channel was connected from the PulsePal to the lasers, in order to have a precise synchronization of all the outputs. Auditory stimuli included pure tones of 100 ms duration and were interrupted by random gaps of the silence of variable duration (from 1 sec to 20 sec). All the stimuli were presented at three different intensity levels (30-, 55-, and 80-dB SPL) and at three different frequencies (1, 2.5, and 5 kHz).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003eIn vivo optogenetics\u003c/h2\u003e \u003cp\u003eFor in vivo optogenetic-silencing recording, 2 weeks after viral vector injection, a tetrode was implanted in the CMT (same electrode specifications as in the \u003cspan refid=\"Sec6\" class=\"InternalRef\"\u003etetrode implantation\u003c/span\u003e section), coupled with a 100-um diameter optic fiber mounted 100-um above the top electrode contact. Light intensity at optic fiber tips was measured with a power meter (Thorlabs PM100D) before optic fiber insertion (output ≃30 mW). The same protocol mentioned above (in vivo electrophysiological recording) was performed with the difference that the laser was delivered 500 ms before and 500 ms after the auditory stimuli onset.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2-photon laser scanning microscopy\u003c/h2\u003e \u003cp\u003eHead-fixed mice were placed and trained under the microscope every day for at least 6 days prior to the experiment, and then longitudinally imaged using an in vivo 2PLSM (Scientifica HyperScope) equipped with a \u0026times;16 objective (0.8 NA, Nikon). ScanImage Software (Vidrio Technologies, LLC) was used to control the microscope, the acquisition parameters, and the TTL-driven synchronization between the acquisition and EEG/EMG recordings. GCaMPs were excited using a Ti: sapphire laser operating at λ\u0026thinsp;=\u0026thinsp;910 nm (InSight X3, Spectra-Physics) with an average excitation power at the focal point lower than 50 mW. Time-series images were acquired within a field-of-view of 117 x 117 \u0026micro;m (512x512 pixels). All image analyses were performed using Fiji ImageJ and a custom routine in MATLAB. Each imaging session contained a random presentation of intermingled auditory cues (1kHz or 5kHz) during wakefulness or NREM sleep and automatically aligned to the acquired images with a custom-made MATLAB script. No photo-bleaching or photo-toxicity was observed. Acquired images were then corrected by adjusting XY motion artifacts using Suite2p Software\u003csup\u003e57\u003c/sup\u003e. Regions of interest (ROIs) of cell bodies were selected and drawn manually. All pixels within each ROI were first averaged providing a single time-series of raw fluorescence. Raw calcium traces were then normalized to \u0026ldquo;F\u0026rdquo; estimated through a Gaussian distribution-based approach, as previously described\u003csup\u003e56\u003c/sup\u003e. Normalized traces are referred to as \u0026lsquo;ΔF/F\u0026rsquo; throughout the paper.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eCa2\u0026thinsp;+\u0026thinsp;event detections\u003c/strong\u003e \u003cp\u003eCalcium events were detected using MATLAB custom scripts. Each ROI ΔF/F trace was analyzed separately after motion correction. Traces were first up sampled via interpolation to be smoother, with cubic spline interpolation (spline function in Matlab). The baseline median \u0026lsquo;M\u0026rsquo; and noise level \u0026lsquo;E\u0026rsquo; for each trace is estimated by computing the median and standard deviation of the signal in three iterations, the points exceeding 2 times the standard deviation are excluded from the estimation of M and E in the following iteration. Candidate calcium events were identified via the use of the find peaks function in Matlab, using twice the estimated M\u0026thinsp;+\u0026thinsp;E as a threshold for minimum peak height, and M\u0026thinsp;+\u0026thinsp;E as minimum peak prominence.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eThe characteristics of the candidate events, such as duration, peak amplitude, and integral were computed. Each measured characteristic was then compared with minimum and maximum physiological characteristics, and if not respecting these, excluded. The candidate events were then filtered and confirmed as Ca2\u0026thinsp;+\u0026thinsp;events if their measured characteristics were satisfying some minimum physiological parameters. Physiological parameters used for filtering candidate events were: Minimum Duration of an Event\u0026thinsp;=\u0026thinsp;1 [seconds]; Maximum Duration of an Event\u0026thinsp;=\u0026thinsp;20 [seconds]; Maximum Ratio between rise and decay duration = ⅓; Minimum Peak Amplitude\u0026thinsp;=\u0026thinsp;2 [a.u. (ΔF/F)]; Minimum 10 Integral\u0026thinsp;=\u0026thinsp;1 [a.u. (ΔF/F) * seconds]. The measured events were grouped and an average measure of events per cell, per sleep state (as defined by the sleep hypnogram obtained from the EEGs and EMG scoring), was computed within 15 seconds of each auditory stimulus presentation.\u003c/p\u003e \u003cp\u003eFor population analysis, vectors containing ΔF/F neuron response amplitude upon auditory stimulation were computed. Pearson correlation analyses were then computed between vectors representing different auditory stimuli (CS- or CS+) at different times (Habituation or Recall) and during different states (Wake or NREM sleep).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eFear conditioning\u003c/h2\u003e \u003cp\u003eAfter a baseline of two hours per 14 days (same protocol as mentioned before in the section on auditory stimulation), the mice were habituated by gently handling them for 5 min on 5 consecutive days, then on day 28, the fear conditioning protocol started. On the first day of the procedure (Habituation) at ZTO, a foreign cage with a metal grid as a floor (Context A) was wiped with 70% ethanol and the mice were placed in it. The walls were marked with stars to provide additional contextual information. The ice was first given 3 min of time to explore the novel environment, followed by playing a first auditory stimulus (CS-) consisting of 27 pure tones of 100 ms duration at 1 kHz played over 30 sec for 5 times with a variable interstimulus interval (ISI) between 10s and 30s. Then, a second auditory stimulus (CS+) was played under the same conditions but at 5 kHz. 24 h later, the acquisition of fear memories was performed by wiping context A with 70% ethanol again and subsequently placing the animals in the same cage with the metal grid. After three minutes, CS- and CS\u0026thinsp;+\u0026thinsp;were played intermixed. An unconditional stimulus (US) was paired with CS\u0026thinsp;+\u0026thinsp;by applying an electric foot shock through the metal grid on the floor at 0.5 mA for 1 sec, starting when CS\u0026thinsp;+\u0026thinsp;ended. Another 24h later (Recall), a novel cage (Context B) was wiped with 1% acetic acid, and mice were placed in it. After three minutes of explorations, CS- and CS\u0026thinsp;+\u0026thinsp;were presented to the mice with the same protocol as during habituation. Freezing behavior was measured as a measure of learning performance. It was scored manually as the absence of any movement except breathing. Freezing to CS-/CS\u0026thinsp;+\u0026thinsp;was measured during the time the tone was playing, and only the first two rates of CS- and CS\u0026thinsp;+\u0026thinsp;were considered. Freezing was quantified as the total time of freezing during the total playing time of the tones. If mice generalized on day 3 (Recall), they were excluded from the experiment. The mice that were able to discriminate between the CS- and CS\u0026thinsp;+\u0026thinsp;during recall time, underwent the last part of the experiment. Immediately after the cued recall memory test, the mice returned to their home cage, where they were recorded and re-exposed to the safety (CS-) and dangerous (CS+) sound for 2/4 hours while they were sleeping. We tested two different modulations: the first group of mice was exposed to an NREMs-specific auditory stimulation with CMT optogenetic-silencing (n\u0026thinsp;=\u0026thinsp;11), and the second group to REMs-specific auditory stimulation with CMT optogenetic-silencing (n\u0026thinsp;=\u0026thinsp;10).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eData analysis\u003c/h2\u003e \u003cp\u003e \u003cb\u003ePre-processing of LFP data and peaks analysis.\u003c/b\u003e Following the acquisition, LFP row recordings were sampled to 1000 Hz and re-referenced with a common average reference to reduce possible volume conduction. Then, the data were z-scored. After detecting LFP responses to auditory stimulation, a comprehensive analysis was carried out to examine the timing and amplitude of the peaks, calculating the average of all the trials per mouse. This analysis focused on identifying the latency to the first LFP peak, specifically the early activity in response to the stimulation (ranging from 0 to 50 ms) and the late component (ranging from 50 to 100 ms) of the auditory-evoked response, as presented in Extend Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb. We also analyzed the strength of the response by calculating the amplitude of the detected LFP peaks for both the early and late components of the response (Extend Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec).\u003c/p\u003e \u003cp\u003e \u003cb\u003eSleep scoring.\u003c/b\u003e Sleep scoring was performed manually, based on frequency and amplitude characteristics of the EEG and EMG in custom software written in MATLAB. NREMs was identified by high amplitude, synchronous activity in the EEG with a prominent delta (0.5\u0026ndash;4 Hz) frequency power, and low EMG activity; REMs was defined by high synchronous theta (6\u0026ndash;9 Hz) and flat EMG; wake was characterized by an increase in EMG activity with a low-amplitude, high-frequency (\u0026gt;\u0026thinsp;6 Hz) EEG.\u003c/p\u003e \u003cp\u003e \u003cb\u003eArousal threshold analysis.\u003c/b\u003e We analyzed the averaged LFP activity in response to the stimuli splitting the recording into two groups, based on the behavior of the animals. In blue are the events in which the animals were not waking up from 0 after 5 sec of the auditory stimulations (\u003csup\u003eA\u003c/sup\u003eNREMs-NREMs transition), and in red are the events in which the animals were waking up from 0 to 5 sec after the auditory stimulation (\u003csup\u003eA\u003c/sup\u003eNREMs-WAKE transitions). As previously described in the section \u0026lsquo;Sleep scoring\u0026rsquo;, the different states (WAKE, NREMs, and REMs) were defined manually by the EEG/EMG recording.\u003c/p\u003e \u003cp\u003eTo assess how the animals were awakening from sleep in response to the different tones used in the fear conditioning protocol (CS- and CS+), we used a similar analytical approach as described earlier measuring the events in which the animals woke up within 0 to 5 sec after the stimulation. However, in this case, we calculated the percentage of trials in which the animal woke up in response to each tone (CS- or CS+), which was obtained by dividing the number of trials in which the animal woke up for the specific tone by the total number of stimulations, i.e., the sum of the number of CS- and CS\u0026thinsp;+\u0026thinsp;trials. It is important to mention that the protocol included an equal number of CS- and CS\u0026thinsp;+\u0026thinsp;tones. Next, we computed the difference in awakening between the CS- and CS\u0026thinsp;+\u0026thinsp;tones by calculating the delta of awakening, which is obtained by subtracting the percentage of awakening in response to the CS- tone from the percentage of awakening in response to the CS\u0026thinsp;+\u0026thinsp;tone. This analysis allowed us to determine whether the animals exhibited a differential response to the two tones, as indicated by a higher percentage of awakening for the CS- tone compared to the CS\u0026thinsp;+\u0026thinsp;tone.\u003c/p\u003e \u003cp\u003e \u003cb\u003eSlow wave detection.\u003c/b\u003e Individual slow waves were detected during NREM sleep using the SWA-MATLAB toolbox. The beginning of the slow wave was marked at the positive to negative zero-crossing before the negative peak and the end of the slow wave was marked at the end of the subsequent positive slope.\u003c/p\u003e \u003cp\u003e \u003cb\u003eSpindle detection.\u003c/b\u003e With a custom-made script, the spindles were detected for each LFP channel and for the EEGs (spindleDetection_pathLoop). As a first step, the spindle rate was calculated for all the different recordings (14 sessions per animal), then averaged per mouse. Digging more into the analysis we also wanted to analyze the role of spindles in \u003csup\u003eA\u003c/sup\u003eNREMs-to-WAKE transitions, so we plot all the spindles found in events in which the animals were staying in NREMs (\u003csup\u003eA\u003c/sup\u003eNREMs-NREMs in blue) and events in which animals were waking up after 10 sec of auditory stimulation onset (\u003csup\u003eA\u003c/sup\u003eNREM-WAKE in red) before (-5;0 sec) and after (-5;10 sec) the auditory stimulation onset. Then we calculated the percentage of spindles in \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs and \u003csup\u003eA\u003c/sup\u003eNREMs-to-WAKE transitions (%= (spindles found in the time window of -1\u0026thinsp;+\u0026thinsp;1 sec of the stim onset/number of events) *100).\u003c/p\u003e \u003cp\u003e \u003cb\u003eSingle Unit activity and spike sorting.\u003c/b\u003e Single-unit activity was detected using the Offline Sorter Application by Plexon Neurotechnology Research System (version 4.4.1). Raw LFP data was first band-pass filtered (500-4000Hz, Butterworth filter) and then a threshold for multiunit activity was set manually (depending on the recording from \u0026minus;\u0026thinsp;2 to -6 SD). Single unit activity was then extracted using principal components analysis and manually extracting clusters. The spike ratio for the PSTH was calculated as the total number of spikes found for each trial, divided per number of events, and averaged for all the animals. For the quantification per animal, there were calculated the number of spikes found during the stimulus onset (from 0 to 0.1 sec) divided per event and then averaged per animal.\u003c/p\u003e \u003cp\u003eBurst firing of single units was detected as a minimum of three consecutive action potentials with ISI\u0026thinsp;\u0026lt;\u0026thinsp;6 msec and preceded by a quiescent hyperpolarized state of at least 60 msec. On the basis of the observed pattern of response, a time window (0-0.1, 0-0.2, 0-0.3, 0-0.4, 0-0.5 sec) was selected to analyze the response of a given neuron. For each cell, the values of spontaneous (before the tone) and evoked (during/after the tone) activity obtained during maintained NREM sleep (\u003csup\u003eA\u003c/sup\u003eNREMs-NREMs) were compared with the events with the transition to wake (\u003csup\u003eA\u003c/sup\u003eNREMs-WAKE), using paired t-test.\u003c/p\u003e \u003cp\u003e \u003cb\u003eDecoding and Feature extraction with CNNs.\u003c/b\u003e A Convolutional Neural Network (CNN), with an EEGNet architecture\u003csup\u003e58\u003c/sup\u003e was trained to discriminate the conditions \u003csup\u003eA\u003c/sup\u003eNREMs-NREMs vs. \u003csup\u003eA\u003c/sup\u003eNREMs-WAKE based on LFP and EEG traces, extracted from a time interval of 10 seconds pre- to 5 seconds post- sound onset, using a tenfold cross-validation scheme. This network has already been successfully used to discriminate electrophysiological responses to auditory stimuli\u003csup\u003e59\u003c/sup\u003e. The network was trained on the raw LFP traces of the HP, CMT, and AU1 as well as the Frontal and Parietal EEG channels. The data of the CMT was filtered in the sigma range (10\u0026ndash;40 Hz), as a focus was set specifically for spindles (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea-b for a control analysis filtering CMT in the delta range). Accuracy was computed via the area under the receiver operating characteristic curve (AUC). The trained networks were then used to extract condition-specific activation maps\u003csup\u003e60\u003c/sup\u003e, reflecting the features that were most relevant to the network\u0026rsquo;s decisions, with the use of saliency map, as it is a common practice in the field\u003csup\u003e61\u003c/sup\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb, which were computed at the single-trial level for all correctly classified trials and then averaged over all mice and overall ten trained networks to increase stability, as in previous work\u003csup\u003e61\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003cb\u003eStatistical analysis.\u003c/b\u003e Data were compared via two-way ANOVA, or t tests for parametric data, with post hoc Tukey\u0026rsquo;s corrections for multiple comparisons. Data distribution was assumed to be normal, but this was not formally tested. Values in the text are reported as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard error mean (S.E.M.) unless reported otherwise. For each experiment, sample numbers are indicated in the corresponding figure legends.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eHistology\u003c/h2\u003e \u003cp\u003e \u003cb\u003eTissue collection.\u003c/b\u003e At the end of all experiments, animals were anesthetized with isoflurane as described above and an electric current was sent through two channels of each tetrode (30 uA, 5 pulses at 2s) in order to induce gliosis at the very end of tetrode placement. After approximately 2h (time to create an internal scar), mice were euthanized with 15mg pentobarbital i.p., and the cardiovascular system was transfused with 30 mL of cold, heparinized phosphate-buffered saline (PBS), followed by 30 mL of 4% Paraformaldehyde (PFA) via puncture in the left ventricle of the heart. Brains were removed and kept in PFA at 4\u0026deg;C overnight and then they were cryoprotected in 30% sucrose for 48h. Next, they were flash-frozen in 2-methyl butane at approximately \u0026minus;\u0026thinsp;80℃, cut into 40 nm thick sections, and stored in PBS at 4℃. In order to confirm tetrode placement, the tetrodes were stained with 1,1-Dioctadecyl-3,3,3,3-tetramethyl-indocarbocyanin-perchlorat before implantation. After cutting the brain, sections with visible dye traces were selected and either Nissl-stained to reveal the gliosis.\u003c/p\u003e \u003cp\u003e \u003cb\u003eImmunohistochemistry.\u003c/b\u003e The viral expression specificity and efficacy were checked histologically by double staining of free-floating sections. The brain sections were washed in PBS 0.1% Triton X-100 (PBST) three times for 10 min each, blocked by incubation with 4% bovine serum albumin dissolved in PBST for 45 min and subsequently incubated with anti-GFP antibodies (AB_221569) for 24h at 4℃. Then sections were washed again in PBST (three times for 10 min each) and incubated with a secondary antibody (ab150073) that binds to the primary antibody for 1h at room temperature.\u003c/p\u003e \u003cp\u003e \u003cb\u003eFluorescence microscopy.\u003c/b\u003e Images were acquired by a Nikon Ti-E microscope with a 40x resolution. For tetrode placement, a filter Cy3 was used. Instead of fluorescent staining.\u003c/p\u003e \u003c/div\u003e \n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003eMethods References\u003c/h2\u003e\n \u003cp\u003e57. Aime, M. et al. Paradoxical somatodendritic decoupling supports cortical plasticity during REM sleep. \u003cem\u003eScience\u003c/em\u003e (1979) \u003cstrong\u003e376\u003c/strong\u003e, 724\u0026ndash;730 (2022).\u003c/p\u003e\n \u003cp\u003e58. Pachitariu, M., Stringer, C. \u0026amp; Harris, K. D. Robustness of spike deconvolution for neuronal calcium imaging\u003cem\u003e. Journal of Neuroscience\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 7976\u0026ndash;7985 (2018).\u003c/p\u003e\n \u003cp\u003e59. Lawhern, V. J. et al. EEGNet: a compact convolutional neural network for EEG-based brain\u0026ndash;computer interfaces\u003cem\u003e. J Neural Eng\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 56013 (2018).\u003c/p\u003e\n \u003cp\u003e60. Aellen, F. M. et al. Auditory stimulation and deep learning predict awakening from coma after cardiac arrest. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e146\u003c/strong\u003e (2023) doi:10.1093/brain/awac340.\u003c/p\u003e\n \u003cp\u003e61. Simonyan, K., Vedaldi, A. \u0026amp; Zisserman, A. Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps. (2013).\u003c/p\u003e\n \u003cp\u003e62. Aellen, F. M., G\u0026ouml;ktepe-Kavis, P., Apostolopoulos, S. \u0026amp; Tzovara, A. Convolutional neural networks for decoding electroencephalography responses and visualizing trial by trial changes in discriminant features. J \u003cem\u003eNeurosci Methods\u003c/em\u003e \u003cstrong\u003e364\u003c/strong\u003e, (2021).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank all the Tidis Lab members for their insightful discussion of, and comments on, previous versions of the manuscript. This work was supported by the Inselspital University Hospital Bern, and Interfaculty Research Cooperation (A.A. and C.G.H.), Swiss National Science Foundation (A.A.), the University of Bern (A.A.) and the EU H2020-ICT grant (A.S. and T.F.).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eILB, MA, and CGH performed the experiments. FA with the supervision of AZ trained the neural network and performed the computational approach analysis. ILB, FA, TR, MA, and MB analyzed the data. ILB and AA conceived the studies and wrote the manuscript with the help of all the authors. AA supervised the research. AS and TF provided technical assistance with the fabrication of aberration-corrected endoscopes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll raw data used for analysis are made available through the Supplementary Information. Source data are provided in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll scripts used for analysis will be available through a GitHub repository (https://github.com/ZENLabCode).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eFrench, A. S., Geissmann, Q., Beckwith, E. J. \u0026amp; Gilestro, G. F. Sensory processing during sleep in Drosophila melanogaster. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e598\u003c/strong\u003e, 479\u0026ndash;482 (2021).\u003c/li\u003e\n\u003cli\u003eNir, Y., Vyazovskiy, V. V., Cirelli, C., Banks, M. I. \u0026amp; Tononi, G. Auditory responses and stimulus-specific adaptation in rat auditory cortex are preserved across NREM and REM sleep. \u003cem\u003eCerebral Cortex\u003c/em\u003e May; \u003cstrong\u003e25\u003c/strong\u003e(5):1362-78. (2015).\u003c/li\u003e\n\u003cli\u003eSchreck, M. R. et al. State-dependent olfactory processing in freely behaving mice. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eIssa, E. B. \u0026amp; Wang, X. Sensory responses during sleep in primate primary and secondary auditory cortex. \u003cem\u003eJournal of Neuroscience\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e(53):14467-80 (2008).\u003c/li\u003e\n\u003cli\u003eTasserie, J. et al. Deep brain stimulation of the thalamus restores signatures of consciousness in a nonhuman primate model. \u003cem\u003eSci. Adv\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e (2022).\u003c/li\u003e\n\u003cli\u003eHayat, H. et al. Reduced neural feedback signaling despite robust neuron and gamma auditory responses during human sleep. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 935\u0026ndash;943 (2022).\u003c/li\u003e\n\u003cli\u003ePortas, C. M. et al. Auditory Processing across the sleep-wake cycle: simultaneous EEG and fMRI monitoring in humans. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e (2000).\u003c/li\u003e\n\u003cli\u003eCanales-Johnson, A., Merlo, E., Bekinschtein, T. A. \u0026amp; Arzi, A. Neural dynamics of associative learning during human sleep. \u003cem\u003eCerebral Cortex\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1708\u0026ndash;1715 (2020).\u003c/li\u003e\n\u003cli\u003eColrain, I. M. \u0026amp; Campbell, K. B. The use of evoked potentials in sleep research. \u003cem\u003eSleep Medicine Reviews\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e 277\u0026ndash;293.\u003c/li\u003e\n\u003cli\u003eOswald, I., Taylor, A. M. \u0026amp; Treisman, M. Discriminative responses to stimulation during human sleep. \u003cem\u003eBrain \u003c/em\u003e\u003cstrong\u003e83\u003c/strong\u003e (1960).\u003c/li\u003e\n\u003cli\u003ePoitras, R., Thorkildsen, A., Gagnon, M. A. \u0026amp; Naiman, J. Auditory discrimination during REM and NREM sleep in woman before and after delivery. \u003cem\u003eCan Psychiatr Assoc J\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e (1973).\u003c/li\u003e\n\u003cli\u003eBastuji, H., Perrin, F., Garcia-Larrea, L. \u0026amp; Bastuji, H. Semantic analysis of auditory input during sleep: studies with event related potentials. \u003cem\u003eInternational Journal of Psychophysiology\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e (2002).\u003c/li\u003e\n\u003cli\u003eZ\u0026uuml;st, M. A., Ruch, S., Wiest, R. \u0026amp; Henke, K. Implicit vocabulary learning during sleep is bound to slow-wave peaks. \u003cem\u003eCurrent Biology\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 541-553.e7 (2019).\u003c/li\u003e\n\u003cli\u003eIb\u0026aacute;\u0026ntilde;ez, A., L\u0026oacute;pez, V. \u0026amp; Cornejo, C. ERPs and contextual semantic discrimination: degrees of congruence in wakefulness and sleep. \u003cem\u003eBrain Lang\u003c/em\u003e \u003cstrong\u003e98\u003c/strong\u003e, 264\u0026ndash;275 (2006).\u003c/li\u003e\n\u003cli\u003eLangford, G. W., Meddis, R. \u0026amp; Pearson, A. J. D. Awakening latency from sleep for meaningful and non-meaningful stimuli. \u003cem\u003ePsychophysiology\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 1\u0026ndash;5 (1974).\u003c/li\u003e\n\u003cli\u003eDang-Vu, T. T. et al. Interplay between spontaneous and induced brain activity during human non-rapid eye movement sleep. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e108\u003c/strong\u003e, 15438\u0026ndash;15443 (2011).\u003c/li\u003e\n\u003cli\u003eSchabus, M. et al. The fate of incoming stimuli during NREM sleep is determined by spindles and the phase of the slow oscillation. \u003cem\u003eFront Neurol\u003c/em\u003e \u003cstrong\u003eAPR\u003c/strong\u003e, (2012).\u003c/li\u003e\n\u003cli\u003eEdeline, J., Manunta, Y. \u0026amp; Hennevin, E. Auditory thalamus neurons during sleep: changes in frequency selectivity, threshold, and receptive field size. \u003cem\u003eJ Neurophysiol \u003c/em\u003e\u003cstrong\u003e84 \u003c/strong\u003e(2000).\u003c/li\u003e\n\u003cli\u003eSimor, P., van der Wijk, G., Nobili, L. \u0026amp; Peigneux, P. The microstructure of REM sleep: Why phasic and tonic? \u003cem\u003eSleep Medicine Reviews\u003c/em\u003e \u003cstrong\u003e52\u003c/strong\u003e (2020).\u003c/li\u003e\n\u003cli\u003eHayat, H. et al. Locus coeruleus norepinephrine activity mediates sensory-evoked awakenings from sleep. \u003cem\u003eSci Adv\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, (2020).\u003c/li\u003e\n\u003cli\u003evan Kronenberg, P., Milinski, L., Kruschke, Z. \u0026amp; de Hoz, L. Sound disrupts sleep-associated brain oscillations in rodents in a meaning-dependent manner. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eGent, T. C., Bandarabadi, M., Herrera, C. G. \u0026amp; Adamantidis, A. R. Thalamic dual control of sleep and wakefulness. \u003cem\u003eNature Neuroscience\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e (2018).\u003c/li\u003e\n\u003cli\u003eSherman S. Murray. A wake-up call from the thalamus. \u003cem\u003eNature Neuroscience \u003c/em\u003e\u003cstrong\u003e4\u003c/strong\u003e (2001).\u003c/li\u003e\n\u003cli\u003eAntonini, A. et al. Extended field-of-view ultrathin microendoscopes for high-resolution two-photon imaging with minimal invasiveness. \u003cem\u003eElife\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1\u0026ndash;76 (2020).\u003c/li\u003e\n\u003cli\u003eLima, S. L., Rattenborg, N. C., Lesku, J. A. \u0026amp; Amlaner, C. J. Sleeping under the risk of predation. \u003cem\u003eAnimal Behaviour\u003c/em\u003e \u003cstrong\u003e70\u003c/strong\u003e 723\u0026ndash;736 (2005).\u003c/li\u003e\n\u003cli\u003eHalperin, J. M. \u0026amp; Iorio, L. C. Responsivity of Rats to neutral and danger-signaling stimuli during sleep. \u003cem\u003eBehavioral and Neural Biology\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e (1981).\u003c/li\u003e\n\u003cli\u003eTseng, Y. T. et al. The subthalamic corticotropin-releasing hormone neurons mediate adaptive REM-sleep responses to threat. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, 1223-1239.e8 (2022).\u003c/li\u003e\n\u003cli\u003eAndrillon, T. \u0026amp; Kouider, S. The vigilant sleeper: neural mechanisms of sensory (de)coupling during sleep. \u003cem\u003eCurrent Opinion in Physiology\u003c/em\u003e vol. 15 47\u0026ndash;59 (2020).\u003c/li\u003e\n\u003cli\u003eRattenborg, N. C., Lima, S. L. \u0026amp; Amlaner, C. J. Facultative control of avian unihemispheric sleep under the risk of predation. \u003cem\u003eBehavioural Brain Research\u003c/em\u003e. \u003cstrong\u003e105\u003c/strong\u003e (1999).\u003c/li\u003e\n\u003cli\u003eFernandez, L. M. J. \u0026amp; L\u0026uuml;thi, A. Sleep spindles: Mechanisms and functions. \u003cem\u003ePhysiol Rev\u003c/em\u003e \u003cstrong\u003e100\u003c/strong\u003e, 805\u0026ndash;868 (2020).\u003c/li\u003e\n\u003cli\u003eLoftus, J. C., Harel, R., N\u0026uacute;\u0026ntilde;ez, C. L. \u0026amp; Crofoot, M. C. Ecological and social pressures interfere with homeostatic sleep regulation in the wild. \u003cem\u003eeLife \u003c/em\u003e\u003cstrong\u003e11\u003c/strong\u003e (2022).\u003c/li\u003e\n\u003cli\u003eKaramihalev, S. et al\u003cem\u003e.\u003c/em\u003e Social dominance mediates behavioral adaptation to chronic stress in a sex-specific manner. \u003cem\u003eElife\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1\u0026ndash;18 (2020).\u003c/li\u003e\n\u003cli\u003eSamson, R. et al. Species-Specific Information for Enhancing Ecosystem Services. in \u003cem\u003eThe Urban Forest\u003c/em\u003e, pp. 111\u0026ndash;144, Publisher: Springer (2017).\u003c/li\u003e\n\u003cli\u003eMcKenna, J. T. et al\u003cem\u003e.\u003c/em\u003e Basal forebrain parvalbumin neurons mediate arousals from sleep induced by hypercarbia or auditory stimuli. \u003cem\u003eCurrent Biology\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 2379-2385.e4 (2020).\u003c/li\u003e\n\u003cli\u003eShin, A. et al. A brainstem-to-mediodorsal thalamic pathway mediates sound-induced arousal from slow-wave sleep. \u003cem\u003eCurrent Biology\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e (2023).\u003c/li\u003e\n\u003cli\u003eM\u0026aacute;ty\u0026aacute;s, F. et al\u003cem\u003e.\u003c/em\u003e A highly collateralized thalamic cell type with arousal-predicting activity serves as a key hub for graded state transitions in the forebrain. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 1551\u0026ndash;1562 (2018).\u003c/li\u003e\n\u003cli\u003eTasserie, J. et al. Deep brain stimulation of the thalamus restores signatures of consciousness in a nonhuman primate model. \u003cem\u003eSci. Adv\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e (2022).\u003c/li\u003e\n\u003cli\u003eLe Merre, P. et al. Reward-Based Learning Drives Rapid Sensory Signals in Medial Prefrontal Cortex and Dorsal Hippocampus Necessary for Goal-Directed Behavior. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e97\u003c/strong\u003e, 83-91.e5 (2018).\u003c/li\u003e\n\u003cli\u003eCash, S. S. et al. The human K-complex represents an isolated cortical down-state. \u003cem\u003eScience (1979)\u003c/em\u003e \u003cstrong\u003e324\u003c/strong\u003e, 1084\u0026ndash;1087 (2009).\u003c/li\u003e\n\u003cli\u003eMak-McCully, R. A. et al\u003cem\u003e.\u003c/em\u003e Synchronization of isolated downstates (K-Complexes) may be caused by cortically-induced disruption of thalamic spindling. \u003cem\u003ePLoS Comput Biol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, (2014).\u003c/li\u003e\n\u003cli\u003eCusinato, R. et al. Intrinsic neural timescales in the temporal lobe support an auditory processing hierarchy. \u003cem\u003eJNeurosci \u003c/em\u003e\u003cstrong\u003e43 \u003c/strong\u003e(2023).\u003c/li\u003e\n\u003cli\u003eLecci, S. et al. Coordinated infraslow neural and cardiac oscillations mark fragility and offline periods in mammalian sleep. \u003cem\u003eSci Adv \u003c/em\u003e\u003cstrong\u003e3\u003c/strong\u003e (2017).\u003c/li\u003e\n\u003cli\u003eBastuji, H., Lamouroux, P., Villalba, M., Magnin, M. \u0026amp; Garcia-Larrea, L. Local sleep spindles in the human thalamus. \u003cem\u003eJournal of Physiology\u003c/em\u003e \u003cstrong\u003e598\u003c/strong\u003e, 2109\u0026ndash;2124 (2020).\u003c/li\u003e\n\u003cli\u003eAndrillon, T. \u003cem\u003eet al.\u003c/em\u003e Sleep spindles in humans: Insights from intracranial EEG and unit recordings. \u003cem\u003eJournal of Neuroscience\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 17821\u0026ndash;17834 (2011).\u003c/li\u003e\n\u003cli\u003eOtake, K., Kin, K. \u0026amp; Nakamura, Y. Fos expression in afferents to the rat midline thalamus following immobilization stress. \u003cem\u003eNeurosci Res \u003c/em\u003e\u003cstrong\u003e43\u003c/strong\u003e\u003cem\u003e \u003c/em\u003e(2002).\u003c/li\u003e\n\u003cli\u003eSchmitt, L. I. et al. Thalamic amplification of cortical connectivity sustains attentional control. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e545\u003c/strong\u003e, 219\u0026ndash;223 (2017).\u003c/li\u003e\n\u003cli\u003ePadilla-Coreano, N., Do-Monte, F. H. \u0026amp; Quirk, G. J. A time-dependent role of midline thalamic nuclei in the retrieval of fear memory. \u003cem\u003eNeuropharmacology\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e 457\u0026ndash;463 (2012).\u003c/li\u003e\n\u003cli\u003ePenzo, M. A. et al\u003cem\u003e.\u003c/em\u003e The paraventricular thalamus controls a central amygdala fear circuit. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e519\u003c/strong\u003e, 455\u0026ndash;459 (2015).\u003c/li\u003e\n\u003cli\u003eRechtschaffen, A., Hauri, P. \u0026amp; Zeitlin, M. Auditory awakening thresholds in REM and NREM Sleep stages. \u003cem\u003ePercept Mot Skills\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 927\u0026ndash;942 (1966).\u003c/li\u003e\n\u003cli\u003ePilon, M., Desautels, A., Montplaisir, J. \u0026amp; Zadra, A. Auditory arousal responses and thresholds during REM and NREM sleep of sleepwalkers and controls. \u003cem\u003eSleep Med\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 490\u0026ndash;495 (2012).\u003c/li\u003e\n\u003cli\u003eSnyder, F. Toward an evolutionary theory of dreaming. \u003cem\u003eAmerican Journal of Psychiatry\u003c/em\u003e \u003cstrong\u003e123\u003c/strong\u003e, 121\u0026ndash;136 (1966).\u003c/li\u003e\n\u003cli\u003eHellman, K. \u0026amp; Abel, T. Fear Conditioning Increases NREM Sleep. \u003cem\u003eBehavioral Neuroscience\u003c/em\u003e \u003cstrong\u003e121\u003c/strong\u003e, 310\u0026ndash;323 (2007).\u003c/li\u003e\n\u003cli\u003eOsorio-Forero, A. \u003cem\u003eet al.\u003c/em\u003e Noradrenergic circuit control of non-REM sleep substates. \u003cem\u003eCurrent Biology\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 5009-5023.e7 (2021).\u003c/li\u003e\n\u003cli\u003eRattenborg, N. C. Somnovolarism. \u003cem\u003eSleep Med\u003c/em\u003e \u003cstrong\u003e67\u003c/strong\u003e, 246\u0026ndash;247 (2020).\u003c/li\u003e\n\u003cli\u003eBeauchamp, G. Sleeping gulls monitor the vigilance behaviour of their neighbours. \u003cem\u003eBiol Lett\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 9\u0026ndash;11 (2009).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-3395895/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3395895/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eSleep is associated with a sensory disconnection from the environment despite a high vulnerability to danger and predation. Yet, sensory stimuli-evoked responses persist in the brain of flies\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e1\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e, rodents\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e2,3\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e, primates\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e4,5\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e, and humans\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e6,7\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e during sleep. Whether discrimination between sensory stimuli occurs in the mammalian brain during sleep remains unclear. Here, we showed that neutral auditory stimuli evoked electrical responses propagate in parallel auditory and non-auditory pathway, some of which awaken sleeping mice. We used a convolutional neural network and identified neural activities of centro-medial thalamic (CMT) neurons as the most discriminant hub for auditory-evoked sleep-to-wake transitions among all recorded structures. Importantly, we found that prior associative learning of danger (conditioned stimulus, CS+) and neutral (CS-) auditory cues resulted in increased awakening events upon CS+ exposure during NREM, but not REM, sleep. These sleep-to-wake transitions were blocked by optogenetic silencing of CMT neurons during CS exposure in sleeping mice. Altogether, these results suggest a central role of the CMT neurons in the residual processing of behaviorally-relevant information in the sleeping brain.\u003c/strong\u003e\u003c/p\u003e","manuscriptTitle":"A role for the thalamus in danger discrimination during sleep","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-01-12 17:47:32","doi":"10.21203/rs.3.rs-3395895/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"51df7989-8f7a-4b3f-afd6-83e682f98f9f","owner":[],"postedDate":"January 12th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":25504712,"name":"Biological sciences/Neuroscience/Circadian rhythms and sleep/Sleep"},{"id":25504713,"name":"Biological sciences/Neuroscience/Cognitive neuroscience/Perception"}],"tags":[],"updatedAt":"2024-01-12T17:47:32+00:00","versionOfRecord":[],"versionCreatedAt":"2024-01-12 17:47:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3395895","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3395895","identity":"rs-3395895","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}