A Common Neuroendocrine Substrate for Diverse General Anesthetics and Sleep

Summary

How general anesthesia (GA) induces loss of consciousness remains unclear, and whether diverse anesthetic drugs and sleep share a common neural pathway is unknown. Previous studies have revealed that many GA drugs inhibit neural activity through targeting GABA receptors. Here, using Fos staining, ex vivo brain slice recording, and in vivo multichannel electrophysiology, we discovered a core ensemble of hypothalamic neurons in and near the supraoptic nucleus, consisting primarily of neuroendocrine cells, that are persistently and commonly activated by multiple classes of GA drugs. Remarkably, chemogenetic or brief optogenetic activations of these anesthesia-activated-neurons (AAN) strongly promote slow-wave sleep and potentiates GA; whereas conditional ablation or inhibition of AAN led to diminished slow-wave oscillation, significant loss of sleep, and shortened durations of GA. These findings identify a common neural substrate underlying diverse GA drugs and natural sleep, and reveal a crucial role of the neuroendocrine system in regulating global brain states.

INTRODUCTION

The discovery of general anesthetics has revolutionized surgical procedures in medicine over the past centuries (Franks, 2008; Robinson and Toledo, 2012). Each year, millions of patients world-wide undergo general anesthesia (GA) for medical treatments. GA is a global brain and body state characterized by unconsciousness, analgesia, amnesia, and immobility while maintaining vital physiological functions (Brown et al., 2010; Franks, 2008). Despite decades of scientific effort, the target and circuit mechanisms by which extremely diverse groups of GA drugs all can induce sedation and loss of consciousness remain poorly understood (Alkire et al., 2008; Brown et al., 2010; Franks, 2008; Koch et al., 2016; Rudolph and Antkowiak, 2004). Delineating the identities of neurons and the precise neural circuitry that enable GA drugs to produce the unconscious brain state will advance our basic understandings and clinical applications of GA.

During the past decades, GABA (γ-aminobutyric acid) type A (GABA_A) receptor has emerged as a principal target for many anesthetics (Franks, 2008). GABA_A receptor is the major contributor of neuronal inhibition and widely expressed throughout the CNS. Many GA drugs have been shown to potentiate GABA-induced Cl⁻ current; in higher doses, they can directly activate GABA_A receptors (Franks, 2008). On the other hand, there are also non-GABAergic anesthetics such as ketamine, which targets the NMDA receptor to reduce the excitatory action of glutamate; and dexmedetomidine, which binds to alpha-2 (α2) adrenergic receptors and inhibits the norepinephrine release from locus coeruleus (Franks, 2008). Together, these results lead to the idea that different GA drugs work by exerting differentiated inhibitory impacts on the nervous system, although the particular loci (if present) that anesthetics target to induce unconsciousness is still elusive. In recent years, a few studies have identified anesthetic-activated cells in several brain regions, using either immediate early gene markers or ex vivo brain slice recording (Gelegen et al., 2018; Moore et al., 2012; Zhang et al., 2015). However, no study so far has provided conclusive evidence of anesthetic-activated neurons with in vivo electrophysiology, and whether there is a common neural substrate activated by different classes of GA drugs is unknown.

The overlap between GA and endogenous sleep-wake circuitry is also actively debated. Examination of brain oscillations revealed that although different anesthetics produce distinct patterns of electroencephalography (EEG) that may not be observed in natural sleep, one of the shared feature of sleep and GA is the enhancement of slow-delta (0.5–4 Hz) oscillations (Akeju and Brown, 2017; Franks, 2008; Rudolph and Antkowiak, 2004). Notably, sleep deprivation results in increased homeostatic pressure to sleep and enhances the potency of GA (Tung et al., 2002). The recovery process from sleep deprivation can also take place during GA (Tung et al., 2004). However, the exact biological substrate shared between GA and sleep remains to be identified. In recent years, a few studies have revealed that neurons co-release of peptides and small molecule transmitters (glutamate or GABA) participate in regulating sleep across different species (Chung et al., 2017; Jego et al., 2013; Lee et al., 2017). Compared to fast transmitters, neuropeptides exert wider-spreading and longer-lasting effects which are ideally suited to regulate global brain state. However, the potential roles of neuropeptides in GA is largely an uncharted territory, and whether there are common neuropeptides engaged in both sleep and GA is unexplored.

Here, we reasoned that neurons activated by multiple different GA drugs may represent a shared substrate between GA and sleep. Using a combination of immediate early gene expression, ex vivo brain slice, and in vivo extracellular recordings, we uncovered a hypothalamic neuronal population, which unexpectedly consists primarily of neuroendocrine cells, that are surprisingly and persistently activated by multiple distinct GA drugs. With our recently developed capturing activated neuronal ensembles (CANE) technology (Sakurai et al., 2016), we were able to precisely label, characterize and manipulate these anesthesia-activated neurons (AAN). In freely behaving mice, optogenetic and chemogenetic activation of AAN were sufficient to strongly potentiate slow-wave sleep (SWS) and GA. Importantly, conditional ablation of AAN resulted in significant decline of slow-wave power and loss of both SWS and REM (rapid eye movement) sleep, while acute inhibition of AAN shortened the duration of GA. Together, our results revealed a previously unrecognized critical function of neuroendocrine cells, which are known for their role of releasing hormones, in regulating both GA and natural sleep.

RESULTS

Discovery of Anesthesia-Activated Neurons (AAN) In Vivo

Initially, to search for anesthesia activated neurons, or AAN, we subjected the mice to either isoflurane/oxygen anesthesia or to oxygen exposure alone (control) for 2 hours, followed by examining the brain for Fos expression. Fos is an immediate early gene that is generally used as a marker for activated neurons (Morgan and Curran, 1989). As compared to control condition, Fos expression throughout the brain was dramatically diminished under isoflurane GA, we observed a distinct cluster of strong Fos⁺ neurons in the ventral region of the hypothalamus (Figure 1A), a region containing neurons regulating sleep-wake cycles (Scammell et al., 2017; Weber and Dan, 2016). Notably, most of the AAN (>80% of Fos⁺ cells) resided at the upper-corner of the optic chiasm, extending through the anterior-to-posterior axis of the ventral edge of the hypothalamus, which is classically defined as the supraoptic nucleus (SON) (Figure 2A), located posterior to the ventrolateral preoptic nucleus (VLPO) (Kroeger et al., 2018). There was also a sparse population of AAN scattered from ventral preoptic area (POA) to areas dorsal to posterior SON under isoflurane (Figure 1A), we refer to these regions collectively as the paraSON area.

Figure 1. Discovery of General-Anesthesia-Activated Neurons (AAN) in Hypothalamus.

(A) Left, schematics of interested regions on the brain atlas. A, anterior; P, posterior. Right, representative patterns of Fos⁺ neurons (approximately 0 mm to −1.2 mm from bregma) after 2-hour control (oxygen) versus isoflurane exposure (1~1.2% Isoflurane mixed with oxygen) from n = 4 pairs of mice.

(B-G) Simultaneous in vivo extracellular recording of hypothalamic neurons in the AAN region and the brain states before, during, and after GA. n = 89 neurons across 14 sessions from 7 mice. (B) Schematics of recording chamber and electrode placement. Iso, isoflurane; PFC, prefrontal cortex; EMG, electromyography; Gnd, ground. (C) Representative isoflurane-suppressed (Iso-Sup.) and isoflurane-activated (Iso-Act.) neuron. Top two panels, spike-rate of the example neuron; third, frontal cortex LFP (fLFP); bottom, EMG. Black dashed lines mark the duration of isoflurane exposure. Red dashed lines indicate the period of loss-of-consciousness (LOC). (D) Activity profile of all neurons recorded (n = 89). The spike rate of each neuron was normalized by its peak firing rate. (E-F) Activities of isoflurane-activated neurons (raw spike trains convolved with 1-s Gaussian kernel) aligned with the time when mice lose consciousness (E) or emerge from GA (F). Purple square highlights the neurons that increased firing rate before LOC (E), or decreased firing rate ahead of emergence (F). (G) Categorize the neuronal population based on its response toward isoflurane.

See also Figures S1 and S2

Figure 2. Molecular Signatures of AAN.

(A) Experimental procedure of two-color in situ hybridization on AAN. vGat, vesicular GABA transporter; vGlut2, vesicular glutamate transporter 2; AVP, arginine vasopressin; Pdyn, prodynorphin; OXT, oxytocin.

(B) Representative images of two-color in situ hybridization between Fos (green) that marks AAN and following probes (red): vGat, vGlut2, AVP, Pdyn, OXT, and Galanin. Arrow, open arrowhead and solid arrowhead indicating double⁺, probe⁺ only and Fos⁺ only cells, respectively.

(C) Pie chart of percentage of AAN (Fos⁺) colocalized with each probe in SON and paraSON region. Neurons were from 7–23 sections from 2–3 mice for each pair of condition.

See also Figure S3

While Fos is considered as a neuronal activity marker, other cellular events (such as signaling of neurotrophic factors, or activation of protein kinase A) might induce Fos expression without neuronal firing (Lin et al., 2008). To gain more direct evidence that AAN indeed increases firing in vivo under GA, we recorded well-isolated single-unit activity in AAN region with simultaneous monitoring of local field potentials (LFP) in frontal cortex and electromyography (EMG) in neck muscle (as indicators of brain states) before, during, and after GA (Dzirasa et al., 2011) (Figures 1B–G; Figure S1, n = 89 neurons from 7 mice). We classified the identified single units into three categories based on their responsiveness toward isoflurane exposure. While the majority of recorded cells were either suppressed (36%), or weakly modulated (53%) by isoflurane, a small population of neurons (11%, 10 out of 89) with low baseline activity were strongly and persistently activated under GA (Figures 1C–D, 1G). Importantly, a significant portion of these activated cells (60%, 6 of these 10) increased firing rates before the brain state transitioned from wakefulness to the loss of consciousness state (LOC) as determined by the LFP and EMG recording (Figure 1E; Figure S2), and the majority of cells (70%, 7 of the 10) decreased firing before the animal emerged from GA (Figure 1F; Figure S2). These results suggested that GA indeed elicits robust and persistent firing of AAN in vivo.

Molecular Signatures of AAN

We next characterized the neurochemical signatures of AAN. Previous research had suggested that GA drugs and SWS activated hypothalamic GABAergic and/or galaninergic neurons in POA (Moore et al., 2012; Sherin et al., 1998). On the other hand, SON is known to contain peptidergic neuroendocrine cells that mainly produce arginine vasopressin (AVP, also known as antidiuretic hormone, ADH), prodynorphin (Pdyn) or oxytocin (OXT) (Ludwig and Leng, 2006; Mutsuga et al., 2004). Using two-color in situ hybridization in which AAN were identified with Fos RNA probe (induced by isoflurane GA), we found that 45.20% scattered AAN in the paraSON region were vGat⁺, while 44.67% were vGlut2⁺ (Figure 2). Thus, the paraSON AAN contained both inhibitory and excitatory cells. However, AAN located within SON were largely but weakly vGlut2⁺ (Ponzio et al., 2006) (95.39%). Further investigation revealed around 80% AAN in SON expressed AVP and Pdyn (85.95% for AVP; 81.22% for Pdyn), while 10.87% SON AAN were Galanin⁺ (Figure 2). AAN were largely non-overlapping with the high-level (bright) OXT expressing SON cells. However, under GA, there appeared a population of low (dim) level of oxytocin expression SON cells that was not present in control (un-anesthetized) conditions (Figure S3A). Interestingly, 89.26% AAN showed dim OXT expression (Figure 2). It has been known that SON AVP cells express very low level of oxytocin in basal condition (Leng, 2018). These results suggest that GA may induce OXT mRNA expression in AAN. Overall, the hypothalamic AAN population is largely composed of peptidergic neurons (which are also weakly vGlut2⁺) in SON and sparsely distributed GABAergic and glutamatergic cells in paraSON.

AAN Represent a Common Substrate Targeted by Multiple Distinct GA Drugs

To further characterize and manipulate AAN, we employed the recently developed Capturing Activated Neuronal Ensemble (CANE) technology (Sakurai et al., 2016), which is ideally suited to express any desired transgene in Fos⁺ neurons (Figure 3A). This CANE system utilizes engineered pseudotyped Lentivirus (CANE-LV) that selectively infect Fos⁺ neurons in the Fos^TVA knock-in mice. We subjected Fos^TVA mice to 2 hours isoflurane GA, followed by co-injection of CANE-LV-Cre with Cre-dependent AAV-DIO-mCherry into AAN region (Figure 3B). 3~4 weeks later, the mice were re-exposed to 2 hours of isoflurane GA, and the activated neurons were visualized with Fos staining. We found that 92.72% of CANE-mCherry labeled neurons in SON and 73.87% in paraSON re-expressed Fos⁺ after second isoflurane GA, suggesting that the same ensemble of AAN was re-activated by the repeated GA (Figures 3C–D).

Figure 3. A Shared Neuronal Population is Activated by Different Anesthetics.

(A) Schematic diagram of CANE technology.

(B) Viral construct and injection site in Fos^TVA mice.

(C) Illustration of the SON and paraSON. paraSON is defined as a region within 500 μm radius circle with the up-corner of the optic chiasm/tract as the center, extending from anterior to posterior hypothalamus.

(D) Left panels, representative images of CANE-captured isoflurane-activated neurons (red) and Fos⁺ neurons (green) induced by re-exposure to either isoflurane again, or to Propofol, Ketamine (plus xylazine), or dexmedetomidine (Dex). Right panels, pie charts showing the percentage of initial CANE-captured isoflurane-activated neurons that are re-activated (Fos⁺) by different anesthetics in SON as well as in paraSON. Neurons were from 7–30 slices from 2–4 mice for each condition.

(E) Whole-cell patch-clamp recording of CANE-captured isoflurane-activated neurons in acute brain slices following treatments of different classes of anesthetics. Top, representative membrane potential changes after the application of drugs; bottom, statistical summary for all recorded neurons. n = 11 neurons for isoflurane; n = 32 for Propofol; n = 25 for Ketamine; n = 27 for Dex. Wilcoxon signed-rank tests for all drugs.

Data are presented as mean ± s.e.m. **P < 0.01, ***P < 0.001.

See also Figure S3

The CANE-capturing of AAN also afforded us the opportunity to examine whether isoflurane activated AAN can also be activated by other GA drugs. To do this, we first used CANE to express mCherry in isoflurane-activated AAN, and weeks later, we subjected the mice to treatment of either Propofol, Ketamine (with or without Xylazine), or dexmedetomidine (Dex), followed by Fos staining (Franks, 2008) (Figure 3D; Figure S3B). Interestingly, Propofol, Ketamine (plus Xylazine), and Dex all re-activated a large population of AAN that were previously activated by isoflurane in both SON and paraSON (Figure 3D. In SON, 76.64%, 97.85%, and 75.98% Isoflurane-mCherry⁺ cells were Fos⁺ under Propofol, Ketamine, and Dex respectively; in paraSON, 56.32%, 70.28%, and 65.45% isoflurane-mCherry⁺ cells were Fos⁺ under Propofol, Ketamine (plus Xylazine), and Dex, respectively). Ketamine alone also reactivated a shared population of AAN, albeit to a lesser extent (Figure S3B). We further performed whole-cell patch-clamp recording of CANE-mCherry captured AAN (mostly in SON). We discovered that isoflurane, Propofol, Ketamine, and Dex all significantly depolarized the membrane potentials of AAN upon perfusion into the acute brain slices (Figures 3E, ΔV = 11.38 ± 1.82 mV induced by isoflurane, n = 11; ΔV = 5.10 ± 0.98 mV induced by Propofol, n = 32; ΔV = 7.09 ± 1.89 mV induced by Ketamine, n = 25; ΔV = 12.95 ± 1.79 mV induced by Dex, n = 27 recorded AAN), consistent with the fact that these cells became Fos⁺ in vivo under GA. Together, our results strongly support the idea that these hypothalamic AAN represent a common neural substrate activated by distinct classes of GA drugs.

Chemogenetic Activation of AAN Strongly Enhances SWS

We next wanted to determine the causal functions by artificially activating AAN. Since the entire population of AAN (SON plus paraSON) are distributed from ventral preoptic regions to the most posterior end of SON (>1.2 mm along the anterior-posterior axis in the ventral hypothalamus), we exploited chemogenetics (Roth, 2016) to activate this spatially extended population of neurons. The chemogenetic activator hM3Dq-DREADDs (AAN-hM3Dq) or control mCherry (AAN-mCherry) was expressed in AAN using CANE (Figures 4A–C). In acute brain slices, application of Clozapine-n-oxide (CNO), the ligand for hM3Dq, resulted in persistent firing of hM3Dq-expressing AAN (Figure 4C). Systemic injection of CNO also induced robust Fos expression in AAN-hM3Dq mice (Figure 4B). To examine the in vivo effects on brain states, we injected AAN-hM3Dq or control AAN-mCherry mice with either CNO or saline and simultaneously recorded the EEG from frontal and parietal cortex and EMG from the neck muscle for 2 hours immediately after injection (during dark phase) (Figure 4A). In saline injected AAN-hM3Dq, as well as saline or CNO injected AAN-mCherry control mice, we observed normal interspersed wakefulness and short bouts of SWS characterized by higher power oscillations at delta (1–4 Hz) frequency and low EMG activity, and occasional episodes of REM sleep characterized by elevated theta oscillation with very low amplitude of EMG (Scammell et al., 2017; Weber and Dan, 2016) (Figures 4D–E). Remarkably, in CNO treated AAN-hM3Dq mice, the average total duration of SWS nearly doubled that of three control groups, and the average total wake period was concomitantly reduced (Figures 4D–E, on average 56.53 ± 3.70% of time SWS for hM3Dq-CNO, versus 28.30 ± 2.71%, 23.86 ± 2.36%, 24.83 ± 1.05% for hM3Dq-saline, mCherry-saline, mCherry-CNO conditions, respectively). This SWS-enhancing effect of chemogenetic activation of AAN was primarily attributable to extending the average SWS bout duration (Figure 4F, average 48.01 ± 4.85 sec per SWS bout for hM3Dq-CNO, versus 26.95 ± 3.50, 24.52 ± 3.10, 27.35 ± 3.26 sec for hM3Dq-saline, mCherry-saline, mCherry-CNO conditions, respectively, during dark phase with 4-s scoring window). In other words, activating AAN renders mice sleep much longer but does not alter their sleep frequency. We further performed detailed spectral analyses on frontal and parietal EEGs. The power-frequency curves of SWS under AAN-activation almost completely overlapped with those of the naturally occurring SWS with the peak power at delta wave both in frontal and parietal EEG (Figures 4G–H; Figure S4). Taken together, our results strongly suggest that AAN identified here likely represent the shared neural pathway between different GA drugs and natural sleep, and activating AAN is sufficient to promote/extend the duration of SWS.

Figure 4. Chemogenetic Activation of AAN Significantly Potentiates SWS.

(A) Viral-genetic strategy for expressing hM3Dq-mCherry in AAN, and the layout of EEG/EMG recording. fEEG, frontal EEG; pEEG, parietal EEG; Gnd, ground.

(B) Systemic CNO treatment (intraperitoneal injection) induced robust Fos (green) expression in the hM3Dq-mCherry⁺ (red) neurons.

(C) Depolarization of hM3Dq-mCherry⁺ neurons by CNO in acute brain slice.

(D) Representative polysomnographic recording following either saline or CNO treatment in AAN-hM3Dq mouse. Top two panels, representative spectrogram from fEEG and pEEG; third, EMG; bottom, brain state annotated. SWS, slow-wave sleep; REM, rapid-eye movement sleep.

(E) Percentage of time spent in SWS, Wake and REM across 2 hours after injection. Two-way repeated measures ANOVA followed by Sidak’s post hoc test. n = 7 mice for hM3Dq-mCherry and n = 4 mice for mCherry group.

(F) Bout duration and number of SWS across 2 hours after injection. Two-way repeated measures ANOVA followed by Sidak’s post hoc test. n = 7 mice for hM3Dq and n = 4 mice for mCherry group.

(G) Power-frequency analysis across SWS, Wake, and REM from different experimental groups. n = 7 mice for hM3Dq and n = 4 mice for mCherry group.

Data are presented as mean ± s.e.m. ***P < 0.001.

See also Figures S4 and S5

Since the majority of cells among AAN were previously un-suspected SON neurons, which were validated as neuroendocrine cells using the classical Fluorogold uptake assay (Figure S3C) (Weiss and Cobbett, 1992), we next asked whether activating the specific SON cells alone is sufficient to increase SWS. While there is no available Cre-driver line that expresses in all SON-AAN, we decided to utilize AVP^Cre/+ mice since ~85% of SON-AAN are AVP⁺ cells (which also co-express Pdyn (Watson et al., 1982)), keeping in mind the caveat that AVP^Cre mice are homozygous lethal (our observation). We injected Cre-dependent AAV-DIO-hM3Dq bilaterally into SON of heterozygous AVP^Cre/+ mice, and implanted EEG and EMG electrodes in the same mice. Chemogenetic activation of SON AVP⁺ neurons alone was indeed sufficient to potentiate SWS (Figure S5, 27.74 ± 1.89% and 46.87 ± 2.11% of time spent in SWS for saline and CNO conditions), although the extent of SWS duration increase was less than that induced by activating all AAN (99.75% SWS increase for activating all AAN versus 68.96% for activating SON AVP⁺ neurons). Our results thus reveal a previously under-appreciated role of vasopressin⁺/dynorphin⁺ SON neuroendocrine cells in promoting SWS.

Brief Optogenetic Activation of AAN Promotes Subsequent Sleep and Potentiates GA

AAN include a large proportion of neuroendocrine cells, which are able to secrete large quantities of peptides into the cerebrospinal fluid through somatodendritic release (Ludwig and Leng, 2006). Considering that peptide signaling generally lasts longer than classic neurotransmitters, we therefore asked whether a brief stimulation of AAN is sufficient to promote and sustain subsequent sleep. To this end, we expressed the light-activated cation channel channelrhodopsin 2 (AAN-ChR2)(Boyden et al., 2005; Yizhar et al., 2011) or control GFP (AAN-GFP) in AAN using CANE, followed by implantation of fiber optics (200 μm) bilaterally above the AAN (Figures 5A). We validated that blue-light (10 Hz, 10 ms) indeed drove robust neuronal firing of labeled AAN in acute brain slices (Figures 5B). We first performed photo-stimulation in control AAN-GFP mice. To our surprise, we discovered that laser stimulation (10 Hz, 10 ms/pulse, 1-s ON and 1-s OFF, for total of 3 min) had a significant arousal effect on control AAN-GFP animals during the stimulation period (Figures 5C-D). Since the fibers were implanted closely above AAN which are located immediately dorsal to optic chiasm, we reasoned that such deep brain photo-illumination might activate melanopsin-expressing retinal ganglion cells and promote wakefulness. Indeed, a previous study demonstrated that optogenetic deep brain stimulation resulted in retinal activation (Danskin et al., 2015). Remarkably, despite the artifact of light-induced arousal, the same 3 min of laser stimulation in AAN-ChR2 mice drastically increased SWS in the post-stimulation period, signified by a dominant delta oscillation (1–4 Hz) and minimal muscle activity. More interestingly, the sleep-promoting effect lasted up to 10 min after the laser was turned off (Figures 5C–D), indicating that a brief activation of AAN is sufficient to exert a long-lasting effect on brain state. We also observed a complementary decrease in wake time, while there was no significant difference in REM during the post-stimulation period in AAN-ChR2 mice (Figures 5C–D). The optogenetic effect of AAN on SWS also prompted us to investigate the consequence of activating AAN on subsequent GA induction and emergence. Isoflurane (1%) was infused immediately after photo-stimulation (10 Hz, 10 ms/pulse, 1-s ON and 1-s OFF, for total of 3 min), the brain states of animals were determined by EEG/EMG recording. Interestingly, we discovered that photo-activation of AAN did not alter the GA induction but significantly delayed the emergence from GA. In other words, animals would stay longer under GA once we activated AAN (Figure S6), again consistent with the idea of long-lasting peptidergic signaling in sustaining a global sedative state.

Figure 5. Brief Optogenetic Activation of AAN Promotes Subsequent SWS.

(A) Viral-genetic strategy for expressing ChR2 in AAN, the layout of optic fiber implantation, and EEG / EMG recording. Gnd, ground.

(B) Blue laser evokes reliable neuronal spikes in ChR2⁺ AAN.

(C) Representative polysomnographic recording of AAN-GFP (left panel) and AAN-ChR2 (right panel) mice before, during, and after laser stimulation (10 Hz, 10 ms pulses, 1s-On and 1s-OFF, 3 min, 3~4 mW measured at the fiber tips). Top, representative spectrogram of EEG; middle, EMG; bottom, brain state annotated.

(D) Percentage of time spent in SWS (left), Wake (middle), and REM (right) before, during, and after the period of laser stimulation (blue shaded area). n = 28 trials from 5 ChR2 mice (4~6 trials per mouse); n = 24 trials from 4 GFP mice (6 trials per mouse). Permutation test were performed across two groups. Data are presented as mean ± s.e.m. **P < 0.01, ***P < 0.001.

See also Figure S6

AAN Is Activated by Sleep Pressure

We next investigated the relationship between AAN activity and natural sleep. AAN were first labelled with mCherry using CANE in a cohort of mice (Figure 6A). Subsequently, these mice were randomly assigned into three groups: (1) staying in home cage without any disturbance, (2) undergoing sleep deprivation for 5 hours, and (3) undergoing sleep deprivation for 5 hours followed by recovery sleep for 2 hours. Afterwards their brains were collected for Fos staining (Figure 6B). Interestingly, we discovered that sleep deprivation strongly activated AAN. After recovery sleep, the Fos expression in AAN gradually returned to baseline (Figure 6C–D). These results suggest that it is the need for sleep (sleep pressure) that drives the activation of these neurons. This finding also dovetails nicely with our optogenetic experiment result in that activation of AAN can promote subsequent SWS.

Figure 6. AAN Is Activated by Sleep Pressure.

(A) CANE strategy for specifically expressing mCherry in AAN.

(B) Time windows for each experimental group. HC, home cage; SD, sleep deprivation; RS, recovery sleep.

(C) Total percentage of CANE-captured neurons reactivated under HC, SD, and RS. One-way ANOVA followed by Sidak’s post hoc test. n = 3 mice for HC, n = 4 for SD, n = 3 for RS.

(D) Representative Fos staining from HC, SD, and RS.

Data are presented as mean ± s.e.m. ***P < 0.001.

Ablation of AAN Disrupts Natural Sleep

While AAN activation is sufficient to promote SWS, we wanted to know whether AAN are necessary for naturally occurring sleep. To test this, we expressed either the diphtheria toxin receptor (Saito et al., 2001) (DTR) (AAN-DTR) or control mCherry (AAN-mCherry) in AAN using CANE (Figure 7A; Figure S7). We performed chronic EEG and EMG recordings (whole day) before and after diphtheria toxin (DT) injection induced ablation of AAN that expressed DTR (Figure 7B). In AAN-DTR mice, over the course of two weeks post DT injection, the total duration of both SWS and REM sleep gradually and significantly declined (Figure 7C–D, n = 9 mice, SWS: 39.09 ± 0.97, 38.34 ± 1.23, 37.69 ± 1.94, 26.55 ± 2.30, 22.42 ± 3.19% of time; REM: 5.46 ± 0.41 , 5.62 ± 0.43, 2.60 ± 0.43, 1.59 ± 0.39, 1.85 ± 0.49% of time at baseline, day 2, 5, 1-week, or 2-week post DT, respectively). Interestingly, the effect on SWS is largely due to shortened bout duration (but not the total bout numbers), indicating ablating AAN caused the inability to sustain SWS (Figure 7E, average SWS bout duration: 100.23 ± 7.82, 83.24 ± 5.75, 77.34 ± 7.20, 56.93 ± 8.28, 43.94 ± 6.33 sec at baseline, day 2, 5, 1-week, or 2-week post DT, respectively, averaged across the whole day recording with 10-s scoring window). On the other hand, the reduction in REM sleep is mainly attributable to reduced bout numbers (but not per bout duration), suggesting that the AAN-DTR mice had a difficulty in entering REM sleep (Figure 7F, average REM bout number: 86.89 ± 4.97, 102.00 ± 9.16, 42.56 ± 9.83, 26.56 ± 6.88, 32.11 ± 8.57 number of bouts at baseline, day 2, 5, 1-week, or 2-week post DT, respectively). Moreover, we noticed that the delta power of SWS and the theta power of REM sleep were gradually declined in AAN-DTR mice, suggesting that neuronal oscillations which are signatures of normal sleep were impaired (Figures 7E-F, right most panels). One third of animals (3 out of 9 mice) eventually died at the end of experimental session (~ 3 weeks) presumably due to severe sleep deprivation. Importantly, none of these effects were observed in DT treated AAN-mCherry mice (Figure S7). Together, our results indicated that AAN are essential for promoting and stabilizing natural sleep.

Figure 7. Ablation of AAN Disrupts Natural Sleep.

(A) Left, viral-genetic strategy for specifically expressing diphtheria toxin receptor (DTR) in AAN and the layout of EEG and EMG electrodes.

(B) Left, experimental setup of EEG/EMG recording across day and night. Right, top, experimental design and the timing of DT injection. Right, bottom, representative image of AAN ablation after DT injection. DT, diphtheria toxin; 1W, 1-week after DT; 2W, 2-week after DT.

(C) Representative polysomnographic recording before (top) and after (bottom) DT treatment in AAN-DTR mice.

(D) Total percentage of time spent in SWS (left), Wake (middle), and REM (right) across 23 hours recording. One-way repeated measures ANOVA followed by Sidak’s post hoc test. n = 9 mice.

(E) Average bout duration, bout number and delta power of SWS. One-way repeated measures ANOVA followed by Sidak’s post hoc test. n = 9 mice.

(F) Average bout duration, bout number and theta power of REM sleep. One-way repeated measures ANOVA followed by Sidak’s post hoc test. n = 9 mice.

Data are presented as mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001

See also Figure S7

Optogenetic Inhibition of AAN Shortens the Duration of GA

We had wanted to determine whether ablation of AAN could affect GA induction or duration. However, DTR-mediated killing of AAN led to either mortality or chronic sleep loss, which are known to confound GA assay. We therefore turned to perform acute silencing of AAN with optogenetics. We expressed the light-activated cation channel (AAN-eArch) (Chow et al., 2010) or control GFP (AAN-GFP) in AAN using CANE, followed by implantation of fiber optics (200 μm) bilaterally above the AAN (Figure 8A). We validated that yellow-light indeed induced sustained hyperpolarization of AAN expressing proton pump eArch3.0 (Figure 8B). For in vivo experiments, isoflurane (1%) was co-administered with yellow light (5 min square pulse), the brain states of animals were monitored by EEG/EMG. Inhibition of AAN did not alter GA induction, but significantly shortened the duration of GA (Figure 8C–F). These results further underscore the importance of the slow but long-lasting peptidergic (hormonal) signaling in maintaining the GA state.

Figure 8. Optogenetic Inhibition of AAN Shortens the Duration of GA.

(A) Viral-genetic strategy for expressing eArch3.0 in AAN, the layout of optic fiber implantation, and EEG/EMG recording. Gnd, ground.

(B) Yellow light induces a sustained hyperpolarization in eArch3.0⁺ AAN.

(C) Representative EEG/EMG recording across experimental session. Top, spectrogram of EEG; bottom, EMG. Isoflurane (1%) was infused for 10 min. Yellow laser stimulation (5 min square pulse, 5~7 mW measured at the fiber tips) was applied for the first 5 min of Isoflurane infusion. Induction time is identified by occurrence of slow oscillation and reduction of movement. Fully awake time is determined by reduction in slow wave power and raised muscle activity continuously for more than 1 min.

(D-F) Statistical analysis of Induction time (D), fully awake time (E), and anesthesia duration (F). n = 7 for AAN-GFP mice, n = 8 for AAN-eArch3.0 mice. Two-sample t-test.

Data are presented as mean ± s.e.m., *P < 0.05.

Downstream Targets of AAN

Finally, we asked in addition to the pituitary gland, a well-known axonal target of peptidergic SON neurons, whether AAN have any other downstream structure inside the brain. We used CANE technology to express GFP in AAN and mapped their projections throughout the whole mouse brain (Figure S8). We found that AAN project to the septum, a structure known to pace the theta wave (6–10 Hz) of the hippocampus (Buzsaki, 2002), and to the anterior thalamus, a brain region critical for modulating slow-wave activity in EEG (David et al., 2013). We also observed AAN axons in several brain areas engaged in arousal control, including posterior lateral hypothalamus (PLH)(Yamashita and Yamanaka, 2017), tuberomammillary nucleus (TMN) (Yu et al., 2015), supramammillary nucleus (MM) (Pedersen et al., 2017), lateral habenula (LHb) (Gelegen et al., 2018), periaqueductal gray (PAG) (Weber et al., 2018), ventral tegmental area (VTA) (Eban-Rothschild et al., 2016; Taylor et al., 2016), median raphe nucleus (MnR), pedunculopontine nucleus (PPTg), as well as laterodorsal tegmental nucleus (LDTg)(Van Dort et al., 2015). It is interesting to note that lateral habenula (LHb) is one of AAN targets. A recent study discovered that activities of LHb excitatory neurons are required for natural sleep as well as for propofol-induced sedation (Gelegen et al., 2018). Finally, as expected, a large axon bundle was observed that travels towards the arcuate nucleus and median eminence (ARC-ME), which is the known axon pathway for SON axons en route to the posterior pituitary, where hormone are released to the general circulation.

DISCUSSION

The conventional view on GA drugs’ mode of action is that they generally inhibit neuronal activities. Here we demonstrate that chemically distinct GA drugs (isoflurane, propofol, ketamine, dexmedetomidine) all activate a common hypothalamic neural ensemble (AAN) composed mainly of peptidergic (AVP/Pdyn/Galanin) neuroendocrine cells in SON (Figure 1), as evidenced both in slice preparations and in vivo measurements (Figures 2 and 3). Optogenetic and chemogenetic activation of these AAN was sufficient to promote SWS and potentiate GA (Figures 4, Figure 5, and Figures S5), whereas ablating these AAN led to reduced slow-wave power, sleep fragmentation (reduced SWS bout duration), and loss of total sleep (both SWS and REM) (Figure 7). Furthermore, acute silencing of AAN with optogenetics shortened the duration of GA (Figure 8). Together, these results identify a common neural substrate at the intersection of sleep and GA.

In recent years, a couple of studies have identified anesthesia-activated neurons in several brain regions, using either Fos labeling or brain slice recording (Gelegen et al., 2018; Lu et al., 2008; Moore et al., 2012; Zhang et al., 2015). Here we were able to unequivocally demonstrate the presence of AAN in vivo through multi-channel recordings and simultaneous brain state monitoring (Figure 2C–D). Importantly, leveraging the fine temporal resolution afforded by electrophysiology, we found that theses AAN could fire persistently during GA, and even ahead of the time when animals lose consciousness (Figure 2E). In addition, the AAN discovered in this study can be activated by multiple classes of anesthetics (Figure 3), although the underlying mechanisms of depolarization will require further investigation. Interestingly, a recent study showed that GABA is excitatory in adult SON AVP cells due to intracellular Cl⁻ accumulation (Haam et al., 2012). Since a large number of anesthetics target GABA receptors, we speculate SON AAN could be depolarized by other anesthetics as well. On the other hand, several studies discovered that a wide variety of general anesthetics (including halothane, isoflurane, sevoflurane, propofol, ketamine, and dexmedetomidine) could directly interact with G protein-coupled receptors (GPCRs) (Ho et al., 2015; Minami and Uezono, 2013). Since the downstream signaling pathway of excitatory GPCRs can increase intracellular calcium release from stores which in turn can facilitate neuropeptides and hormones release, we speculate that GA drugs may activate some GPCRs in AAN to produce depolarization.

Perhaps the most surprising finding is that the majority of AAN are peptidergic neuroendocrine cells located in SON (vasopressin⁺ / dynorphin⁺ / galanin⁺ neurons) (Figure 1). Magnocellular SON neurons combine the properties of classical neurons and canonical endocrine cells (Leng, 2018) (Figure S3C). Their axons travel all the way to the pituitary gland, while their dendrites cover major areas of the posterior hypothalamus. As neurons, they generate action potentials to trigger hormone release from the pituitary gland into general circulation; as endocrine cells, they directly release peptides from their dendrites and somas in large amount. Within the cerebrospinal fluid, the concentration of peptides detected is high enough to activate targeted receptors in distant brain regions (Ludwig and Leng, 2006). Thus, neuroendocrine cells are ideally positioned to jointly regulate both brain and body state.

Our results indicate that AAN express numerous peptides, mainly AVP, Pdyn, and galanin (Figure 1). Interestingly, these neuropeptides have been implicated in the sleep process. The secretion of SON AVP was previously found to typically increase during the hours of sleep both in human and rodents (Forsling, 1993; Trudel and Bourque, 2010), and enhanced AVP release was implicated in preventing dehydration (Trudel and Bourque, 2010) and lowering body temperature (Hicks et al., 2014) during sleep. AVP also regulate the water channel aquaporin-4, which is highly expressed in astrocytes and is a key component of the glymphatic system that has been implicated in driving metabolism clearance during sleep (Niermann et al., 2001; Xie et al., 2013). Dynorphin (encoded by Pdyn gene), when infused into hypothalamus, has the potential to drastically increase the duration of SWS (Greco et al., 2008). Galanin and its homolog have been shown to promote sleep in several species (Chen et al., 2017; Donlea et al., 2018; Kroeger et al., 2018). Thus, these peptides could potentially regulate multiple aspects of brain and body physiology required for promoting and sustaining sleep. A recent sequencing effort has identified SON as one of the most genetically heterogenous tissues in the body (Mure et al., 2018), thus other as yet unidentified neuropeptides might participate in promoting sleep. On the other hand, AAN also express vGlut2 (albeit at low levels), and glutamate released by these cells could play a role in promoting sleep through activating downstream targets, a possibility awaiting future investigations.

In addition to their activation by multiple general anesthetics (Figure 3), we showed that AAN also respond to sleep pressure (Figure 6). Previous studies had demonstrated that endogenous somnogen prostaglandin D2 (PGD₂) and interleukin-6 could strongly activate SON neurons (Palin et al., 2009; Scammell et al., 1998). Utilizing amino-cupric-silver staining, a previous study also revealed that SON is the most affected region in prolonged sleep deprivation (Eiland et al., 2002). Thus, SON AAN might not be activated by sleep itself, but by the drive to sleep. In addition, SON has been shown to highly express prostaglandin-H2 D-isomerase (PTGDS) (Mure et al., 2018), which convert prostaglandin H2 to PGD_2, and secrete the precursor of adenosine, ATP (Brown et al., 2008). Both PGD₂ and adenosine are elevated in sleep-deprived animals, and intracerebroventricular infusion of these molecules potentiates SWS (Bjorness and Greene, 2009; Urade and Hayaishi, 2011). This is also consistent with our finding that a brief optogenetic stimulation of AAN is sufficient to strongly promote SWS in post-stimulation period (Figure 5). More than a century ago, it was found that the cerebrospinal fluid of sleep deprived animals could promote sleep in other animals (Kubota, 1989). It is conceivable that sleep pressure induced AAN to release peptides and other small molecules into CSF. Since the receptors for these molecules are widely expressed in the brain, they can influence brain-wide neural activities and work on a longer time scale than classic neurotransmitters, which is required for globally coordinating the brain to enter into the sleep/unconscious state.

In fact, activation of AAN by sleep drive instead of sleep per se would be in accordance with other innate homeostatic systems, such as feeding and drinking. The neurons that promote feeding (Agrp⁺ cells in ARC) and drinking (Nos⁺ cells in the subfornical organ) are not activated by eating or drinking itself, but by the desire to eat (hunger) or drink (thirst). Their activity actually decreases immediately after a source of food or water appears (Betley et al., 2015; Zimmerman et al., 2017). Homeostatic regulation of sleep-wake behavior may thus follow the same logic. Future investigation is needed to carefully determine whether SON AAN play a critical role in homeostatic regulation of sleep and to determine which peptides or small molecules the SON AAN release into the brain (or body) once being activated by sleep pressure. Taken together, our results reveal a common and previously unknown neuroendocrine substrate that is targeted by multiple classes of anesthetic drugs, and is also required for natural sleep. These results also bring the neuroendocrine system to the forefront of both practice and future research in GA and sleep medicine.

STAR★METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Fan Wang (fan.wang@duke.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animals

Adult male Fos^TVA knock-in mice (Sakurai et al., 2016) were used in this study (available from JAX stock #027831). Adult AVP-IRES-Cre (JAX stock #023530) and C57BL/6 mice were obtained from Jackson Laboratory. For chemogenetic activation of SON AVP neurons, adult male heterozygous AVP^Cre/+ mice were used. For labelling of anesthesia-activated neurons with CANE technology, adult Fos^TVA mice were single-house for two days to quench the background c-fos expression. Subjects were randomly assigned to experimental and control groups. Animals were housed in standard 12-h dark/12-h light cycle at common facility and all experimental procedures were approved by Duke Animal Care and Use Program.

METHOD DETAILS

Viral Vectors

The CANE-LV-Cre was generated as previously described (Sakurai et al., 2016). Cre-inducible AAV vectors AAV8-hSyn-DIO-hM3D (Gq)-mCherry and AAV8-hSyn-DIO-mCherry, AAV1-EF1a-DIO-hChR2 (H134R)-EYFP were purchased from Addgene. AAV1-CAG-DIO-GFP and AAV1-EF1a-DIO-eArch3.0-EYFP were purchased from the University of North Carolina (UNC) Vector Core. AAV8-hSyn-DIO-DTR was produced by the viral vector core of the Boston Children’s Hospital.

Immunofluorescence

After acutely anesthetized with isoflurane, mice were transcardially perfused with PBS (pH7.4), followed by ice-cold 4% paraformaldehyde (PFA) in PBS. The dissected brains were further post-fixed overnight in 4% PFA at 4 °C, and then tr ansferred into 30% sucrose PBS buffer for 48 hours. Later, the brains were frozen in Tissue-Tek O.C.T. Compound (Sakura) and sliced at 60 μm with cryostat (Leica Biosystems). The sections were washed with PBS, incubated with 1% Triton in PBS at room temperature for 1 hour, and applied with blocking solution (10% Blocking One (nacalai tesque) in PBS with 0.3% Triton X-100) at room temperature for 1 h. Then, the sections were treated with appropriate primary antibodies with desired concentration in blocking solution at 4 °C for overnight. After washed by PBS three times, the sections were incubated with secondary antibody at 4 °C for another night. The sections were further washed, mounted and coverslipped. The primary antibodies used in this study are: goat anti-Fos (Santa Cruz Biotechnology, sc52-g, 1:300). The secondary antibodies are: Alexa Fluor 488 donkey anti-goat (Jackson immunoresearch, 705–545-147 1:400).

Two-Color in situ Hybridization

The cDNA fragments of mouse c-fos, vGlut2, vGat, prodynorphin, and oxytocin, vasopressin, and galanin were amplified by PCR with the antisense primer incorporating the T7 promoter sequence. In vitro transcription was then performed based on the PCR-amplified template using T7 RNA polymerase with DIG-UTP (Roche) or fluorescein-UTP (Roche) for the synthesis of the antisense probes. After hybridization and washing as the protocols we previously described (Bellavance et al., 2017; Zhang et al., 2015), sections were first incubated with alkaline phosphatase–conjugated anti-DIG (1:3500, Roche) and developed with Fast Red substrate (Sigma). Subsequently, the sections were further incubated with POD anti-FITC (1:500, Roche), and developed with FITC-TSA (PerkinElmer).

Capturing Anesthesia-Activated Neurons Using CANE

Adult Fos^TVA (more than 8-week old) mice were single-house for two days to quench the background c-fos expression. On the third day, mice were first put under sustained GA (1~1.2% isoflurane) for 1.5 hours, then a stereotaxic surgery was performed, CANE-LV-Cre (500–750 nl) mixed with desired virus (see below) were co-injected into unilateral or bilateral AAN target area (AP 0.0, ML ±1, DV −5.30~5.40 from the Bregma). For axon-tracing experiment, AAV1-CAG-DIO-GFP (300–500 nl) was co-injected; for AAN labeling and recording, either AAV8-hSyn-DIO-mCherry (500 nl) or AAV1-CAG-DIO-GFP (500 nl) was co-injected; for chemogenetic activation or control experiments, AAV8-hSyn-DIO-hM3D (Gq)-mCherry (500 nl) or AAV8-hSyn-DIO-mCherry (500 nl) was co-injected; for optogenetic activation and inhibition, AAV1-EF1a-DIO-hChR2 (H134R)-EYFP and AAV1-EF1a-DIO-eArch3.0-EYFP were co-injected; for AAN ablation, AAV8-hSyn-DIO-DTR (500 nl) was co-injected. ParaSON is defined as a brain region located within 500 μm radius circle with the up-corner of the optic chiasm/tract as the center, extending from anterior to posterior hypothalamus.

Labelling of Neuroendocrine Cells with Peripheral Fluorogold Injection

To identify neuroendocrine cells, we adapt the protocol as previous described (Kriegsfeld et al., 2003; Oyola et al., 2017). Briefly, mice were injected subcutaneously with 50 μl of 5% Fluorogold (Fluorochrome, Denver, Colorado) in saline. Five to six days later, animals were transcardially perfused with PBS, followed by 4% PFA. The brains were further removed, postfixed overnight, sectioned at 60 μm with cryostat, and the images were taken with Zeiss 700 laser scanning confocal microscopy.

Axon-Tracing and Quantification

CANE-LV-Cre (500 nl) and AAV1-CAG-DIO-GFP (300–500 nl) were co-injected into AAN region of adult Fos^TVA mice. After the viral constructs fully expressed, mice were then sacrificed, the brain were cut into serial 80 μm sections with cryostat (Leica Biosystems). The images were taken with Zeiss 700 laser scanning confocal microscopy. We adopted a similar strategy used by Allen Brain Institute to quantify AAN projection (Oh et al., 2014). In brief, GFP intensity were first normalized and binarized with a custom-written MATLAB code. Region of interests (ROI) were manually defined with respect to standard atlas. GFP positive pixels and total number of pixels in each ROI were quantified.

Multi-channel In Vivo Recording and Data Analysis

Custom-built microelectrode array with 3~4 electrode bundles (30 μm diameter tungsten, California Fine Wire) and two EMG wires (Stablohm 650, California Fine Wire) were constructed as previously described. One wire bundle (with 9 wires) were placed into AAN region (AP: 0.0, ML: ±1.00, DV: −5.40~5.50) to record the single-unit activity. Another bundle (with 6 wires) were placed into frontal cortex (AP: 1.70, ML: ±0.30, DV: −2.00) to capture brain oscillations. EMG wires were inserted into neck muscles. Mice were allowed to recover at least ten days before experiments. For each experimental session, mice were placed in the recording chamber connected with standard Isoflurane vaporizer. After initial 5 min of baseline recording (continuously infused with oxygen), isoflurane (1~1.2%) was applied into the chamber for 10 min, then the mice were allowed to emerge from GA for another 20 min. Neuronal spikes, local field potential, and muscle activity were recorded with CerePlex Direct (BlackRock Microsystem). Brain states were determined by frontal cortex neural oscillations and EMG activity. Spike-sorting were manually done with Offline Sorter (Plexon) based on principal component analysis (PCA). Single-units were identified with distinct clusters (nicely separated from other units and noise) in PCA space and shown clear refractory period in autocorrelation histograms.

Units were classified into three categories based on the responsive property toward isoflurane exposure. Since the mice typically lose consciousness around 400 sec (100 sec after 1~1.2% Isoflurane infusion) and wake up around 1000~1400 sec (100~500 sec after Isoflurane withdraw) with 10 min total duration of isoflurane exposure, we calculated average spike rate for each unit during 400~1200 sec and compared with baseline spike rate of wakefulness 0~400 sec. Units with mean spike rate increase more than double (> 2) were identified as isoflurane-activated cells, while units with mean spike rate decrease more than a half (< 1/2) were classified as isoflurane-suppressed cells. The rest were weak-modulated cells. For illustration purpose, the spike rate of each unit was smoothed with gaussian kernel and normalized with peak firing rate. To capture the neuronal dynamics across brain state transitions (LOC and Emergence), spike density function (SDF) for each neuron is built by convolving Gaussian kernels (1 s-length for main Figure 2, 0.3 s- and 3 s-length for Figure S2) with raw spike train.

Slice Electrophysiology

Coronal slices of hypothalamus were prepared from the adult male AAN-mCherry mice. Animals were briefly anesthetized with 1.2% isoflurane and then killed by decapitation, and brain tissues were immediately dissected out and immersed in ice-cold oxygenated (95% O2 and 5% CO2) slicing solution in which isotonic sucrose was used as a substitute for NaCl (2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 7 MgCl2, 0.5 CaCl2, 7 dextrose, 210 sucrose, 3 sodium pyruvate, 1.3 ascorbic acid). Slices (250 μm in thickness) were cut with a Leica microtome (VT-1000s, Leica, Germany) and immediately transferred to an incubation beaker filled with aerated holding solution: 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 2 MgCl2, 2 CaCl2, 12.5 dextrose, 3 sodium pyruvate, 1.3 ascorbic acid. After about 60-min incubation, we transferred slices to a submerged chamber perfused with aerated normal ACSF containing (in mM): 124 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 2 MgSO4, 2.5 CaCl2, 10 dextrose (315 mOsm, pH 7.4) and visualized by infrared differential interference contrast and fluorescence video microscopy (Examiner.D1, Zeiss). The patch-clamp electrode (4–6 MΩ) was filled with an intracellular solution containing 130 K-gluconate, 5 NaCl, 10 phosphocreatine disodium salt, 1 MgCl2, 10 HEPES, 0.02 EGTA, 0.5 Na2GTP, 2 MgATP, and 0.1% biocytin (pH 7.3, 280–290 mOsm). We employed a MultiClamp 700B amplifier (Molecular Devices) for patch-clamp recording and Spike2 software (Cambridge Electronic Design) for data acquisition. Non-volatile drugs propofol (15 μM, D126608, sigma), Ketamine (100 μM, Henry Schein), dexmedetomidine hydrochloride (40 μM, SML0956, sigma) and CNO (10 μM, Sigma-Aldrich C0832) were diluted from stock in fresh oxygenated aCSF immediately prior to use. Volatile drugs, isoflurane (25 μl) was sonicated into solution in preoxygenated aCSF (50 ml), then stored in a gas-tight syringe (New Era Pump Systems Inc.) to prevent evaporation.

Drug Administration

For chemogenetic activation in freely behaving mice, Clozapine N-oxide (CNO, Sigma-Aldrich C0832) was dissolved in 0.2 ml vehicle solution (PBS with 0.3% DMSO) and administered intraperitoneally (3 mg per kg). Ketamine (100 mg per kg) with or without xylazine (10 mg per kg) were dissolved in PBS and injected intraperitoneally. Dexmedetomidine (Sigma-Aldrich, SML0956) was dissolved in PBS injected intraperitoneally (100 μg per kg). Propofol (Sigma-Aldrich, D126608) was dissolved in intra-lipid solution (I141, sigma) injected intraperitoneally (180 mg per kg). Diphtheria toxin (DT, Sigma-Aldrich, D0564) was dissolved in PBS and intraperitoneally injected (50 μg per kg).

EEG/EMG Recording and Analysis

After the recovery from virus injection surgery (3–4 weeks), mice were further implanted with EEG and EMG. For chemogenetic activation experiment, three stainless steel screws were placed on the frontal, parietal, and cerebellar cortex as EEG electrodes and two thin microwires (Stablohm 650, California Fine Wire) were inserted into bilateral neck muscles as EMG electrodes. For optogenetic activation, two optic fibers (200 μm, NA, 0.39, Thorlab) were inserted on top of AAN (AP 0.0 ~ −0.3, ML ±1, DV −5.00 ~ 5.20 from the Bregma) two stainless steel screws were placed on the parietal cortex and cerebellum, and EMG were inserted into neck muscles. Both EEG and EMG were connected into Omnetics connectors and recorded with CerePlex Direct (BlackRock Microsystem) at 2000 Hz. For chemogenetic activation experiments, after habituation to the recording chambers (Med Associates), experimental sessions typically began at 21:00 (dark phase) after either vehicle or CNO injections. For optogenetic experiment, after habituation to the recording chamber, experiments were performed from 20:30 to 24:00. Two laser trains (10 Hz, 10 ms, 1s-ON and 1s-OFF, 3 min, 3~4 mW from the fiber tip, 473-nm Blue Laser, Cobolt, Sweden) were given per experimental trial with inter-trial interval of 1~1.5 hours. Each animal was tested for 2~3 experimental trials (4~6 pulse trains). Brain state was semi-automatically scored at 4-s epochs with custom-written Matlab code (MathWorks) by the researchers blinded to experimental treatments. Slow-wave sleep was defined as high delta (1–4 Hz) power in the EEG and low EMG activity; REM sleep was defined as high theta (6–10 Hz) power in the EEG with minimal EMG signals; wakefulness was identified as small amplitude and high frequency EEG with tonic EMG activity. Spectrograms were generated with multi-taper approach (Prerau et al., 2017). EEG power spectra were computed for consecutive 4-s windows within the frequency range of 0–50 Hz using the Fast Fourier Transform (FFT). The frequency resolution was set at 0.25 Hz and the power of interested frequency was further divided by the total power.

For optogenetic activation experiments during general anesthesia, mice were first habituated to the recording chamber. Later, one train of laser pulse (10 Hz, 10 ms, 1s-ON and 1s-OFF, 3 min, 3~4 mW from the fiber tip) was given to the animals followed by 1% Isoflurane anesthesia for 12 min. For optogenetic inhibition experiments, yellow laser light (5-min square pulse, 5~7 mW measured at the fiber tip, 561-nm DPSS Laser, OptoEngine LLC, Utah) was delivered for the first 5 min of Isoflurane anesthesia (1% isoflurane for 10 min). The time of Induction was determined by the onset of strong slow-wave power in EEG and minimal muscle activity in EMG; the time point of Fully Awake was determined by diminished slow-wave power and re-appearance of movement continuously for at least 1 min; the total Duration of unconscious time was calculated by the time point of Fully Awake minus the time point of Induction.

For DT/DTR ablation experiment, commercially available EEG/EMG headmounts (Pinnacle Technology, #8201-SS) were used. The electrical signals were acquired at 1000 Hz with EEG/EMG system from Pinnacle Technology. After habituation to the recording chamber, experiments were typically begun at 19:00 and continued toward 18:00 on the next day (23 hours of continuous recordings). Brain states were scored at 10-s epochs by the researchers blinded to experimental treatments with software SIRENIA® SLEEP PRO (Pinnacle Technology). For EEG power bands analysis, the delta (1–4 Hz) power of SWS and the theta (6–10 Hz) power of REM sleep are computed with FFT. For each mouse, the median values of power across all the corresponding windows are used for further statistical analysis.

Mild Sleep Deprivation and Recovery

Mice were kept in their home cages with freely access to water and foods. Sleep deprivation typically began at the light phase from 9:00 to 14:00. Mice were kept awake by introducing novel objects, cage tapping and rotation every 30 min. Recovery sleep was allowed from 14:00 to 17:00. At the end of behavior sessions, Fos staining were performed as described above.

QUANTIFICATION AND STATISTICAL ANALYSIS

For paired observations in slice electrophysiology experiment, Wilcoxon signed-rank test was used to examine the difference between groups. For chemogenetic experiments, mice were randomly assigned to saline or CNO injection in counterbalance order, two-way repeated-measured ANOVA followed by Sidak’s post hoc test (AAN-hM3Dq experiment) or paired t-test (AVP-hM3Dq) was used to assess the statistical difference. For optogenetic experiment, permutation tests (> 10000 iterations) were performed between AAN-ChR2 and AAN-GFP groups across SWS, Wake, REM. For general anesthesia induction experiment, two-sample t-test was used to compare AAN-GFP and AAN-ChR2 groups, or to compare AAN-GFP and AAN-eArch3.0 groups. For DTR-ablation experiment, one-way repeated measures ANOVA followed by Sidak’s post hoc test was used. Statistical analysis was conducted with Prism (GraphPad), SPSS (IBM) and Matlab (MathWorks). All data were presented as mean ± s.e.m. Error bars in figures also represent s.e.m. In addition, the power spectra of SWS and Wake were also plotted with 95% confidence interval in Figure S5. Indications of significance level are as follows: *P < 0.05; **P < 0.01, and ***P < 0.001.

DATA AND SOFTWARE AVAILABILITY

Source data and MATLAB codes are available upon reasonable request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
goat anti-Fos	Santa Cruz Biotechnology	Cat# sc-52-G; RRID: AB 2629503
donkey anti-goat Alexa Fluor 488	Jackson immunoresearch	Cat# 705–545-147; RRID: AB 2336933
Bacterial and Virus Strains
CANE-LV-Cre	Katsuyasu Sakurai et al., 2016	N/A
AAV8-hSyn-DIO-mCherry	Addgene	50459-AAV8
AAV8-hSyn-DIO-hM3D (Gq)-mCherry	Addgene	44361-AAV8
AAV1-CAG-DIO-GFP	UNC Vector Core	N/A
AAV1-Ef1a-DIO-hChR2(H134R)-eYFP	Addgene	20298-AAV1
AAV1-Ef1a-DIO-eArch3.0-eYFP	UNC Vector Core	N/A
AAV8-hSyn-DIO-DTR	Boston Children’s Hospital Viral Core	N/A
Chemicals, Peptides, and Recombinant Proteins
Isoflurane	Baxter Health Corporation	NDC 10019–360-60
Ketamine	Henry Schein	NDC 11695–6835-1
xylazine	Akorn, Inc	NDC 59399–111-50
Propofol	Sigma-Aldrich	D126608
Intralipid	Sigma-Aldrich	I141
Dexmedetomidine	Sigma-Aldrich	SML0956
Clozapine N-oxide (CNO)	Sigma-Aldrich	C0832
Diphtheria toxin (DT)	Sigma-Aldrich	D0564
Fluorogold	Fluorochrome	N/A
Experimental Models: Organisms/Strains
wild type C57BL/6J mice	Jackson Laboratory	664
FosIVA knock-in mice	Jackson Laboratory	027831
AVP-IRES2-Cre-D mice	Jackson Laboratory	23530
Oligonucleotides
See Table S1 for the primers	N/A	N/A
Software and Algorithms
MATLAB 2016a	MathWorks	RRID:SCR_001622
CerePlex Direct	BlackRock Microsystem	N/A
Offline Sorter	Plexon	RRID:SCR_000012
SIRENIA® SLEEP PRO	Pinnacle Technology	N/A
Prism	GraphPad	RRID:SC R_002798
SPSS	IBM	RRID:SC R_002865
Spike2 software	Cambridge Electronic Design	RRID: SC R_000903
ImageJ	NIH Image	RRID:SC R_003070
ZEN	ZEISS	N/A
Other
gas-tight syringe	New Era Pump Systems Inc	N/A