Anatomical substrates of Rapid Eye Movement sleep rebound in a rodent model of post-sevoflurane sleep disruption

Abstract

Background:

Previous research suggests that sevoflurane anesthesia may prevent the brain from accessing REM sleep. If true, then patterns of neural activity observed in REM-on and REM-off neuronal populations during recovery from sevoflurane should resemble those seen after REM sleep deprivation. In this study we hypothesized that, relative to controls, animals exposed to sevoflurane present with a distinct expression pattern of c-Fos, a marker of neuronal activation, in a cluster of nuclei classically associated with REM sleep, and that such expression in sevoflurane -exposed and REM sleep-deprived animals is largely similar.

Methods:

Adult rats and Targeted Recombination in Active Populations mice were implanted with electroencephalographic electrodes for sleep-wake recording and randomized to sevoflurane, REM deprivation or control conditions. Conventional c-Fos immunohistochemistry and genetically-tagged c-Fos labeling were used to quantify activated neurons in a group of REM-associated nuclei in the midbrain and basal forebrain.

Results:

REM sleep duration increased during recovery from sevoflurane anesthesia relative to controls (157.0 ± 24.8 min vs 124.2 ± 27.8 min, P=0.003), and temporally correlated with increased c-Fos expression in the sublaterodorsal nucleus (SLD), a region active during REM sleep (176.0 ± 36.6 cells vs 58.8 ± 8.7, P=0.014), and decreased c-Fos expression in the ventrolateral periaqueductal gray (vlPAG), a region that is inactive during REM sleep (34.8 ± 5.3 cells vs 136.2 ± 19.6, P=0.001). Fos changes similar to those seen in sevoflurane-exposed mice were observed in REM-deprived animals, relative to controls (SLD: 85.0 ± 15.5 cells vs 23.0 ± 1.2, P=0.004; vlPAG: 652.8 ± 71.7 cells vs 889.3 ± 66.8, P=0.042).

Conclusions:

In rodents recovering from sevoflurane, REM-on and REM-off neuronal activity maps closely resemble those of REM sleep-deprived animals. These findings provide new evidence in support of the idea that sevoflurane does not substitute for endogenous REM sleep.

Introduction

Several lines of evidence suggest that volatile anesthetics may not fully substitute for natural REM sleep. Clinical observations indicate that patients often experience complete or near-complete loss of REM sleep during the first night after surgery, followed by increased REM sleep duration and shorter latency to enter REM sleep in subsequent nights.^1–6 This phenomenon, known as REM sleep rebound, has also been noted in animals exposed to sevoflurane and isoflurane.^7–11

Previous animal studies examining the effects of sevoflurane on sleep recovery theorized that the REM sleep rebound observed after sevoflurane may result from an increased need for REM sleep secondary to a REM debt accumulated during anesthesia.^7,8,10,11 However, the homeostatic nature of sevoflurane-induced REM rebound was not experimentally verified, and the specific cell groups responsible for the rebound were not identified. In this study, we aimed to directly test the hypothesis that sevoflurane-induced REM rebound represents a homeostatic response to a REM sleep deficit accumulated during anesthesia. We also sought to identify the specific cell groups and circuits responsible for sevoflurane-induced REM rebound.

A cluster of nuclei located in the midbrain, pons, and basal forebrain serves as the core circuit for REM sleep generation and regulation. Glutamatergic REM-on neurons in the sublaterodorsal (SLD) nucleus are responsible for initiating REM sleep.^12–13 Inhibitory control over the SLD mainly comes from gamma aminobutyric (GABA)-releasing neurons in the ventrolateral periaqueductal gray (vlPAG).^14,15 The activity of the SLD generator is also modulated by cholinergic neurons in the pedunculopontine tegmental (PPT) and laterodorsal tegmental (LDT) nuclei of the basal forebrain, which send projections to the SLD and receive output from the vlPAG. The primary role of the PPT and LDT nuclei is to maintain REM sleep once it has been initiated by the SLD.¹⁶ During REM sleep rebound, the vlPAG’s inhibitory output to the SLD weakens, leading to SLD disinhibition and onset of REM sleep.

Should sevoflurane anesthesia mimic REM deprivation, during recovery from sevoflurane one would expect to observe patterns of activity in REM-on and REM-off neurons consistent with REM sleep rebound. These activity patterns should resemble those seen in animals deprived of REM sleep. We hypothesized that the expression of the neuronal activation marker c-Fos would increase in the SLD, PPT and LDT nuclei, while decreasing in the vlPAG, of rodents experiencing REM rebound during recovery from sevoflurane anesthesia, relative to controls. We also postulated that c-Fos activity maps in the midbrain and basal forebrain would significantly overlap in animals recovering from sevoflurane anesthesia and REM sleep deprivation.

Materials and Methods

Animals

For our study we used 67 adult animals (43 Sprague-Dawley rats (25 males, 18 females) and 24 Targeted Recombination in Active Populations (TRAP) mice (12 males, 12 females) bred on a C57BL/6J background; rats: Envigo, Indianapolis, IN; TRAP mice: internal breeding colony). An approximately equal number of male and female animals were distributed to control and treatment groups. Animals were kept in a 12 h light: dark schedule (3 AM - 3 PM for rats; 6 AM - 6 PM for mice). All animals were randomized to control and treatment groups using the Random Excel function. All studies were approved by the Institutional Animal Care and Use Committee at the University of Virginia (Charlottesville, Virginia). Figure 1 is a schematic of the experimental design.

Figure 1. Schematic representation of the experimental design.

Panel A: Rats were implanted with EEG/EMG electrodes and randomized to sevoflurane or control conditions. EEG recordings were obtained immediately following sevoflurane, and one week later. In a separate cohort of rats, brains were collected 10 hours after sevoflurane emergence for conventional c-Fos immune-histology. Panel B: TRAP mice were implanted with EEG/EMG electrodes and randomized to sevoflurane or control conditions. EEG recordings were obtained after sevoflurane emergence. Mice were injected with 4-hydroxytamoxifen 10 hours after the end of sevoflurane to measure TRAP c-Fos neuronal activity. Panel C: TRAP mice were implanted with EEG/EMG electrodes and randomized to selective REM sleep deprivation (dREM) or control conditions. EEG recordings were obtained at the end of the REM deprivation protocol. Mice received 4-OHT injection 2.5 hours after sevoflurane emergence. (d: day; h: hours; EEG: electroencephalography; EMG: electromyography; TRAP: Targeted Recombination in Active Populations; dREM: REM sleep deprivation).

Electroencephalography implantation

Animals were implanted with electroencephalographic (EEG) and electromyographic (EMG) electrodes as described previously.^9,17,18 Briefly, they were maintained under anesthesia with 2–2.5% sevoflurane in 97.2% oxygen (O₂) and placed on a stereotactic frame (Kopf Instruments, CA) at a constant temperature of 37 ± 1°C. Four epidural electrodes (left and right frontal cortex, left and right parietal cortex) were implanted via Burr holes, with an additional hole drilled in the right cerebellum for a ground electrode. In mice, electrodes were made of stainless-steel wire, whereas in rats stainless steel screws affixed with stainless steel wire were used. One wire was sutured to the neck muscles for EMG recording. Wires were inserted into a 6-channel pedestal secured to the skull with dental acrylic (Plastics One, Roanoke, VA and Preci-Dip, Delemont, Switzerland). Postoperative analgesia was provided with ketoprofen (2.5–5 mg/kg).

Sevoflurane anesthesia

Animals were given at least 12 days to recover before being randomly assigned to receive 2.8% sevoflurane in 97.2% oxygen for 3h or control conditions (100% O₂ for 3h). Sevoflurane exposure was started between 9 AM and 10 AM (during lights on), and a concentration of 2.8% was chosen as the minimum alveolar concentration that prevents movement to pain in 50% of adult rodents (MAC EC₅₀). Animals were placed in a plexiglass chamber at 37 ± 1°C via a heating pad, and sevoflurane and oxygen were delivered using a calibrated vaporizer, with inspired gas concentrations continuously monitored with an inline Datex Capnomac Ultima analyzer (Datex Ohmeda, Madison, WI).^9,17,18 Control animals were kept in a separate chamber with 100% O₂ at a temperature of 37 ± 1°C.

Arterial blood gas analysis

A subset of sentinel rats underwent surgical access to the femoral artery to obtain blood samples using cut-down and cannulation techniques, as previously described.^19,20 pH, arterial oxygen tension (PaO2), arterial carbon dioxide tension (PacO2) and glucose levels (Glu) were measured in 4 rats at each of the 1, 2 and 3 hour intervals during sevoflurane anesthesia (see Table 1). Arterial blood gas analysis was not feasible in mice due to the technical challenges of arterial cannulation and the large blood volume required for analysis, which would result in animal death.

Table 1. Arterial blood gas analysis during sevoflurane anesthesia in rats.

Data expressed as mean ± SD (N=4 rats at each hour interval during sevoflurane anesthesia, 4 males). PaO2: arterial oxygen tension; PacO2: arterial carbon dioxide tension; Glu: glucose.

Time (hour)	pH	Paco2 (mmHg)	Pao2 (mmHg)	Glu (mg/dl)
1	7.40 ± 0.10	51.42 ± 13.53	499.50 ± 60.22	172.75 ± 31.14
2	7.41 ± 0.01	49.60 ± 10.52	521.25 ± 44.18	169.50 ± 31.10
3	7.41 ± 0.05	45.45 ± 18.07	449.50 ± 43.63	168.75 ± 55.98

Selective Rapid Eye Movement sleep deprivation

Selective REM sleep deprivation was performed for 12 hours (6 AM - 6 PM) by gently handling the mice with a cotton applicator as soon as REM sleep was detected through real-time EEG and EMG monitoring. Control animals rested undisturbed in a separate cage. Food and water were readily available during REM deprivation.

Electroencephalography recording and analysis

EEG signals were recorded for 24 hours following sevoflurane anesthesia and REM deprivation, as described previously.^9,17,18 Animals were placed in individual cages with their head tethered to a commutator via lightweight flexible cables. They were familiarized with the cables and cages for at least 24 hours and were allowed to move around freely. Waveforms were amplified and sampled at 200 Hz using a 16-channel amplifier (model 3500, A-M systems, Sequim, WA). High- and low- pass filters were set at 0.1 Hz and 100 Hz for EEG, and 1.0 Hz and 500 Hz for EMG. Data were processed with LabChart (Biopac Systems Inc., The Netherlands) and analyzed with SleepSign (Kissei, Japan) and MatLab-based software. States of REM sleep, non-rapid eye movement (NREM) sleep and wakefulness were automatically scored with SleepSign, manually verified, and then randomly sampled for scoring accuracy by a board-certified sleep medicine physician on the basis of the predominant state within each 4 second epoch. There was a 95% overall agreement between scorers. Wakefulness was scored when low-amplitude, high-frequency, desynchronized EEG was combined with elevated muscle tone. NREM sleep was scored when high-amplitude, low-frequency EEG was coupled with reduced motor tone compared to wakefulness. REM sleep was scored when the EEG was synchronized with minimal to absent muscular tone. Total time spent in REM sleep, NREM sleep and wakefulness, as well as REM latency, were quantified.^9,17,18

Conventional c-Fos neuronal activity

To test the activation of key neuronal populations classically associated with REM sleep during recovery from sevoflurane anesthesia, we used conventional, antibody-based immunohistochemistry to quantify the expression of c-Fos, an immediate early gene and activity transcription factor widely used to identify neurons active during sleep.^13,14,21 Rats were transcardially perfused with 4% paraformaldehyde (PFA) 10 hours after sevoflurane anesthesia, i.e., at a time when REM rebound was most prominent plus 4 hours to allow for full c-Fos expression (see Supplementary Fig. 1). Brains were cut into 40μm-thick sections with a cryostat (Thermo Cryostar NX50, Epredia, USA) and incubated in primary antibody (1:1000 rabbit anti-c-Fos, Abcam # ab190289; 1:500 mouse anti-GABA, Sigma #A0310; 1:1000 mouse anti-glutamate, Millipore #MAB5304; 1:1000 goat anti-ChAT, Millipore #AB144) overnight. Next, sections were incubated in secondary antibody (1:500 goat anti-rabbit Alexa Fluor 488, Invitrogen #A11034; goat anti-mouse Alexa Fluor 594, Invitrogen #A11005; donkey anti-rabbit Alexa Fluor 488, Invitrogen #A21206; donkey anti-goat Alexa Fluor 594, Invitrogen #A32758) for 2h at room temperature. After addition of nuclear staining (1:1000 Hoechst 33342 solution, Thermo #62249), 20x two-dimensional tile scans were obtained with a Leica thunder microscope (model DMi8, Leica microsystems, Deerfield, IL, USA). Neurons co-labelled with c-Fos and anti-glutamate decarboxylase (GAD, marker for GABAergic neurons), anti-L-glutamate (Glut, marker for glutamatergic neurons) or anti-choline acetyltransferase (ChAT, marker for cholinergic neurons) antibodies were quantified using ImageJ (NIH, Bethesda, MD) by an investigator blinded to experimental conditions. Regions of interest for all nuclei were established based on anatomical maps and maintained constant for all animals. Fos counts were the number of c-Fos positive cells colocalized with GAD/ Glut/ ChAT within the region of interest, obtained as the sum of right and left counts from one 20x tiled section cut across the center of each nucleus.

Targeted Recombination in Active Populations c-Fos neuronal activity

Genetically modified TRAP mice use the c-Fos promoter to drive expression of tdTomato, a red fluorescent protein. The Fos locus is linked to a tamoxifen-dependent recombinase that allows permanent tagging in red of neurons activated by stimuli (in our case REM sleep rebound) during a selective 1 to 2 h window preceding subcutaneous injection of the short-acting tamoxifen metabolite, 4-hydroxytamoxifen (4-OHT; Sigma-Aldrich, St. Louis, MO, USA, 50 mg/Kg).^22–25 TRAP mice were injected with 4-OHT 10 hours after sevoflurane in sevoflurane-exposed mice and their controls, and 2.5 hours after REM deprivation in REM-deprived mice and their controls. This period was chosen as it reflected the time of the most prominent REM rebound (see Supplementary Fig. 1 and Fig. 2). Next, mice were transcardially perfused with 4% PFA and 4% acrylamide and brains were cut into 200μm-thick sections incubated overnight at 4°C in a solution of 1% acrylamide and 0.25% VA044 (Wako, NC0632395), followed by clearing in 8% sodium dodecyl sulfate until transparent. Sections were blocked with overnight incubation at 4°C in 5% normal goat serum and 1% bovine serum albumin. The next day, sections were incubated with mouse anti-NeuN antibody (1:500, Millipore #MAB377) for 5 days, then with secondary antibodies (1:500 goat anti-mouse Alexa Fluor 488, Invitrogen #A11029) for 5 days at 4°C. After addition of a nuclear staining (1:1000 Hoechst 33342 solution, Thermo #62249), slides were cover slipped and imaged with a Nikon C2 microscope (Nikon, NY, USA) at 10X magnification. Three dimensional images (5-μm Z-stack distance between consecutive images) were obtained using a NIS-Elements microscope (Nikon Inc, AR 5.20.02). NeuN/c-Fos positive neurons were counted with Imaris software 9.9 (Bitplane Scientific) using an automated spots module, as described previously.^24,25 Regions of interest for all nuclei were established based on anatomical maps and maintained constant for all animals. Fos counts were the number of c-Fos positive cells colocalized with NeuN within the region of interest, obtained as the average across right and left regions of three Z-stacked 200μm-thick sections.

Figure 2. Raw electroencephalography data.

Panels A, B: Representative EEG and EMG tracings from Sprague-Dawley and TRAP mice, respectively. In each panel, the top row represents the EEG signal and the lower row represents the EMG signal. Panels C-H: raster plots of REM sleep, NREM sleep and wakefulness as a function of time for control (left column) and sevoflurane-exposed (right column) rats, respectively. X axis: time elapsed after sevoflurane emergence expressed in hours. Y axis: animal recordings were obtained from. Each vertical line represents a bout. Symbol ∘ indicates “light on”. Symbol • indicates “light off”.

Statistical Analysis

For EEG studies, power calculations were based on two sample Student’s of the mean difference using preliminary data of time spent in REM sleep. Based on those data, a group size of N=14 animals/group would provide 95% power to detect a 20% increase in REM sleep. For c-Fos studies, power calculations were based on two sample Student’s of mean difference using preliminary data that quantified SLD Fos counts in rats. Based on those data, a group size of N=5 animals per group would provide 95% power to detect an increase in Fos expression level (effect size: 2.7885). Two animals died during anesthesia and were not used for analysis. Our EEG data were found to be consistent with a Gaussian distribution (D’Agostino & Pearson test). Time spent in REM sleep, NREM sleep and wakefulness, and REM latency were compared between sevoflurane-exposed and control animals, and between REM-deprived and control animals, with two sample Student’s. REM sleep in 3-hour blocks was compared between sevoflurane-exposed and control rats using repeated measures analysis of variance and the post hoc Bonferroni test. Fos counts were expressed as the number of c-Fos positive cells localized within the region of interest and compared between sevoflurane-exposed and control animals, and between REM-deprived and control animals, with two sample Student’s. We used linear models to examine the extent to which c-Fos counts differ across the control, sevoflurane, control dREM, and dREM mice, across four brain regions of interest (SLD, PPT, LDT, and vlPAG). The dependent variable, c-Fos count, was log-transformed to correct for skewness. In each model, mice group (control, sevoflurane, control dREM, and dREM) was included as the independent variable, with the control group used as the reference group. As such, results are interpreted as the percent difference in c-Fos counts between the comparison group (sevoflurane, control dREM, and dREM) against the reference group (control). All experimenters were blinded to experimental conditions for all experiments. Statistical analyses were performed using GraphPad Prism 7.0 (La Jolla, CA) and R software. All data values are shown as mean ± S.D. P <0.05 was considered statistically significant.

Results

Raw electroencephalography data

Figure 2 shows representative EEG/EMG tracings obtained from rats (panel A) and TRAP mice (panel B). Panels C-H are raster plots of sleep-wake states as a function of time in sevoflurane-exposed and control rats.

Sevoflurane anesthesia does not cause overt physiological derangements in rats

To determine whether sevoflurane anesthesia produced a disturbance in general physiology of the animals, we monitored pH, PaO2, PacO2 and Glu during sevoflurane anesthesia in rats. Animals did not suffer overt physiology derangements, as indicated in Table 1.

Rats exhibit Rapid Eye Movement sleep rebound during recovery from sevoflurane anesthesia

To set the stage for our studies of the anatomical substrates for REM rebound after sevoflurane, we first quantified REM duration and latency in sevoflurane-exposed rats, and compared it against controls. During recovery from anesthesia, sevoflurane-exposed rats spent on average 30 additional minutes in REM sleep, a 25% increase in REM sleep relative to their peer controls (157.0 ± 24.9 min vs 124.2 ± 27.8 min, P= 0.003, 95% Confidence interval [CI]: 11.85 to 53.79, Fig. 3A). In addition, it took sevoflurane-exposed rats less than half the time of controls to enter REM sleep (22.7 ± 16.7 min vs 55.6 ± 53.3 min, P= 0.044, 95% CI: −64.74 to −1.019, Fig. 3B). No significant differences were observed in time spent awake and in NREM sleep between rats exposed to sevoflurane and controls (data not shown). As the time course of REM changes after anesthesia has not been well characterized, we repeated EEG recordings in a subgroup of rats one week after the initial anesthesia. We found no significant differences in REM sleep duration or latency between rats challenged with sevoflurane and controls (Fig. 3C: 99.8 ± 17.2 min vs 111.4 ± 23.1 min, P= 0.379, 95% CI: −40.00 to 16.79, Fig. 3D: 62.6 ± 56.8 min vs 33.5 ± 42.5 min, P= 0.356, 95% CI: −38.56 to 96.76). Collectively, these data suggest that administration of sevoflurane anesthesia to adult rats leads to transient changes in REM sleep, consistent with a short-term increase in REM sleep drive in response to a loss of REM sleep during anesthesia.

Figure 3. Rats exhibit Rapid Eye Movement sleep rebound during recovery from sevoflurane anesthesia.

Panel A: Sevoflurane anesthesia evoked a significant increase in REM duration compared to controls (P= 0.003; N=14 (8 males, 6 females) control and 13 sevoflurane rats (7 males, 6 females)). Panel B: Latency to enter REM sleep was also significantly shorter in sevoflurane -exposed rats relative to controls (P= 0.044; N=14 (8 males, 6 females) control and 13 sevoflurane rats (7 males, 6 females)). Panels C, D: REM quantity and latency were unchanged in sevoflurane-exposed rats and controls one week after the initial sevoflurane administration (P= 0.379 and P= 0.356, respectively; N= 6 control (3 males, 3 females) and N=5 sevoflurane rats (2 males, 3 females)).

c-Fos is differentially expressed in a cluster of Rapid Eye Movement-active nuclei during recovery from sevoflurane anesthesia

To assess for patterns of activity in REM-on and REM-off neurons consistent with REM sleep rebound during recovery from sevoflurane anesthesia, we quantified c-Fos reactivity in a cluster of brainstem and basal forebrain nuclei classically associated with REM sleep rebound. Specifically, we quantified c-Fos+/Glut+ neurons in the SLD, c-Fos+/GAD+ neurons in the vlPAG, and c-Fos+/ChAT+ neurons in the PPT and LDT of sevoflurane -exposed rats, and compared these counts against controls. The number of SLD c-Fos+ glutamatergic neurons was three times higher in sevoflurane-exposed rats compared to controls at the time of peak REM rebound during recovery after sevoflurane (176.0 ± 36.6 Fos+/Glut+ cells vs 58.8 ± 8.7 Fos+/Glut+ cells, P=0.014, 95% CI: 30.37 to 204.0, Fig. 4C). Conversely, the number of Fos-expressing GABAergic neurons in the vlPAG of sevoflurane-exposed rats was markedly decreased compared to controls (34.8 ± 5.3 Fos+/GAD+ cells vs 136.2 ± 19.6 Fos+/GAD+ cells, P=0.001, 95% CI:−148.1 to −54.67, Fig. 4F). c-Fos+/ChAT+ counts in the PPT and LDT of rats exposed to sevoflurane were also significantly higher than controls (PPT: 115.6 ± 23.1 Fos+/ChAT+ cells vs 36.4 ± 7.2 Fos+/ChAT+ cells, P=0.011, 95% CI: 23.32 to 135.1, Fig. 4I; LDT: 203.6 ± 28.0 Fos+/ChAT+ cells vs 86.2 ± 10.6 Fos+/ChAT+ cells, P=0.004, 95% CI: 48.28 to 186.5, Fig. 4L). These results suggest REM-active neuronal maps consistent with REM sleep rebound in rats recovering from sevoflurane anesthesia.

Figure 4. c-Fos is differentially expressed in a cluster of Rapid Eye Movement-active nuclei during recovery from sevoflurane anesthesia.

Panel C: c-Fos reactivity measured via conventional IHC was increased by threefold in the glutamatergic neurons of the SLD in sevoflurane-exposed rats compared to controls (P=0.014). Panel F: the number of c-Fos+ GABAergic neurons in the vlPAG were significantly decreased in sevoflurane-exposed rats relative to controls (P=0.001). Panels I, L: c-Fos reactivity was markedly increased in the cholinergic neurons of the PPT and LDT nuclei in sevoflurane-exposed rats compared to controls (P=0.011 and P=0.004, respectively). Panels A-K: Representative coronal sections through the SLD, vlPAG, PPT and LDT nuclei at 20X magnification. Subpanels a-k: 40X magnification details of panels A-K. Green: c-Fos; Blue: Hoechst; Red: Glut/GAD/ChAT; Yellow: merge. 4V: fourth ventricle; Aq: aqueduct; Glut: glutamate; GAD: glutamate decarboxylase; ChAT: choline acetyltransferase. SLD: sublaterodorsal nucleus. vlPAG: ventrolateral periaqueductal gray. PPT: pedunculopontine tegmental nucleus; LDT: laterodorsal nucleus. N= 5 control (2 males, 3 females) and 5 sevoflurane rats (2 males, 3 females).

Rapid eye Movement-active maps overlap in sevoflurane-exposed and Rapid Eye Movement-deprived mice

To test the reproducibility of our antibody-labeled Fos findings in rats, we quantified endogenously-tagged Fos+ counts in the SLD, vlPAG, PPT and LDT nuclei of mice exposed to sevoflurane using a TRAP approach, which affords higher cell specificity, better temporal resolution and 3D imaging of thicker sections over 2D conventional Fos immunohistochemistry.^22–25 Consistent with the data obtained from the conventional Fos approach in rats, we found a nearly three-fold increase in Fos+ counts in the SLD of sevoflurane-exposed TRAP mice relative to TRAP controls (22.9 ± 3.8 vs 8.7 ± 1.7, P=0.010, 95% CI: 4.526 to 23.87, Fig. 5C). vlPAG Fos counts were almost half those of controls in sevoflurane-exposed TRAP mice (594.2 ± 97.4 vs 913.5 ± 79.0, P=0.034, 95% CI: −608.6 to −30.02, Fig. 5F). The PPT and LDT nuclei also showed a robust increase in Fos expression relative to controls (PPT: 48.8 ± 4.1 vs 24.0 ± 3.2, P=0.001, 95% CI: 12.82 to 36.78, Fig. 5I; LDT: 74.3 ± 4.4 vs 42.8 ± 5.9, P=0.003, 95% CI: 14.49 to 48.51, Fig.5L). These data establish patterns of activity in REM-on and REM-off neurons that are consistent with REM sleep rebound regardless of rodent species and approach for labeling Fos+ neurons.

Figure 5. Rapid Eye Movement-active maps overlap in sevoflurane-exposed and Rapid Eye Movement-deprived mice.

Panel C: There was a nearly threefold increase in c-Fos expression in the SLD of sevoflurane-exposed TRAP mice compared to controls (P=0.010). Panel F: Fos counts were significantly decreased in the vlPAG of sevoflurane-exposed TRAP mice relative to controls (P=0.034). Panels I, L: c-Fos reactivity was almost doubled in the PPT and LDT nuclei of sevoflurane-exposed TRAP mice compared to controls (P=0.001 and P=0.003, respectively). N= 5 control (3 males, 2 females) and 5 sevoflurane TRAP mice (4 males, 1 female). Panel O: c-Fos expression was more than tripled in the SLD of REM-deprived TRAP mice compared to controls (P=0.004). Panel R: Fos counts were significantly decreased relative to controls in the vlPAG of REM-deprived TRAP mice (P=0.042). Panels U, X: c-Fos expression was robustly increased in the PPT and LDT neurons of REM-deprived TRAP mice compared to controls (P=0.013 and P=0.008, respectively). Panels A-W: Representative coronal sections through the SLD, vlPAG, PPT and LDT nuclei at 20X magnification. Subpanels a-w: 40X magnification details of panels A-W. Green: NeuN; Red: c-Fos. Aq: aqueduct; 4V: fourth ventricle; dREM: REM-deprived. N= 5 control (3 males, 2 females) and 5 dREM TRAP mice (2 males, 3 females).

Next, to test whether midbrain and basal forebrain Fos activity maps were similar in animals recovering from sevoflurane and REM sleep deprivation, we subjected TRAP mice to selective REM sleep deprivation and compared their SLD, vlPAG, PPT and LDT Fos patterns against those of sevoflurane-exposed TRAP mice. EEG recordings obtained following REM deprivation confirmed that our REM deprivation protocol was effective and selective for REM sleep (Supplemental Fig. 2). Upon quantification, the number of c-Fos+ nuclei were more than tripled in the SLD of REM-deprived mice compared to controls (85.0 ± 15.5 Fos+ cells vs 23.0 ± 1.2 Fos+ cells, P=0.004, 95% CI: 26.14 to 97.86, Fig. 5O) and c-Fos counts were significantly decreased in the vlPAG (652.8 ± 71.7 Fos+ cells vs 889.3 ± 66.8 Fos+ cells, P= 0.042, 95% CI: −462.5 to −10.53, Fig. 5R). c-Fos counts in the PPT and LDT nuclei were also significantly higher in REM-deprived mice compared to controls (PPT: 110.3 ± 23.1 Fos+ cells vs 36.3 ± 3.6 Fos+ cells, P=0.013, 95% CI: 19.97 to 128.0, Fig. 5U; LDT: 233.2 ± 54.5 Fos+ cells vs 42.0 ± 5.4 Fos+ cells, P=0.008, 95% CI: 64.83 to 317.6, Fig. 5X).

Next, we plotted the distribution of c-Fos by group (Control, Sevoflurane, Control dREM, dREM) across four brain regions of interest (SLD, PPT, LDT, vlPAG) to determine the extent to which c-Fos counts differed across groups. The boxplots show the distribution of the log-transformed c-Fos values by group, across the four brain regions (Fig. 6). For the SLD region, the Sevoflurane group had 2.76 times the amount of c-Fos than the Control group (P <0.001, 95% CI: 5.4 to 11.5), while the dREM group had 9.94 times the amount of c-Fos, compared to the control group (P< 0.001, 95% CI: 1.7 to 5). For the PPT region, the Sevoflurane group had 2.1 times more c-Fos than the control group (P=0.002), 95% CI: 1.4 to 3.2), while the dREM group had 4.35 times the amount of c-Fos, relative to the control group (P< 0.001, 95% CI 2.8 to 6.7). For the LDT region, the Sevoflurane group had 1.8 times more c-Fos than the control group (P=0.029, 95% CI 1.07 to 3.0) and the dREM group had 4.99 times the amount of c-Fos than the control group (P<0.001, 95% CI 2.96 to 8.4). For the vLPAG region, the Sevoflurane group had 0.61 times the amount of c-Fos relative to the control group (i.e., a reduction of 39.2% compared to the control group) (P=0.024, 95% CI 0.40 to 0.93). Although the dREM group had 0.71 times the amount of c-Fos relative to the control group (i.e., 29.0% less c-Fos than the control group), this difference did not reach statistical significance (P=0.104, 95% CI 0.46 to 1.08).

Figure 6:

c-Fos distribution across the sublaterodorsal nucleus, laterodorsal tegmental nucleus, pedunculopontine nucleus and the ventrolateral periaqueductal gray. We utilized linear models to examine the change of c-Fos counts across four groups (Control; Sevoflurane; Control dREM; dREM) in the four brain regions of interest. dREM: REM-deprived.

These findings suggest that c-Fos counts reflect the total duration of REM sleep deprivation, and that sevoflurane anesthesia is associated with moderate sleep deprivation, as indicated by largely overlapping Fos activity maps in mice recovering from sevoflurane anesthesia and selective REM deprivation.

Discussion:

Our results from adult rodents indicate that: i) anesthesia with sevoflurane causes REM sleep rebound that manifests as a transient increase in REM sleep quantity and shortened latency to enter REM sleep during recovery from anesthesia; ii) during recovery after sevoflurane, Fos activity patterns in a cluster of REM-active neurons in the midbrain and basal forebrain are consistent with REM sleep rebound, regardless of animal species and Fos-labeling method; and iii) Fos activity maps largely overlap in animals recovering from sevoflurane and selective REM sleep deprivation.

Previous animal studies that characterized the effects of sevoflurane on sleep recovery noted REM sleep rebound following cessation of anesthesia with sevoflurane and other similar gases.^7,8,10,11 However, these studies primarily interpreted the acute increase in REM sleep as a potential sign of a homeostatic response. They theorized that the increase might be linked to a deficit in REM sleep that occurred during anesthesia. Nonetheless, these earlier studies didn’t conduct experimental tests to confirm whether sevoflurane-induced REM rebound is indeed a homeostatic response, nor did they pursue the specific brain cell groups responsible for the changes. In our study, we aimed to rigorously examine the hypothesis that the increase in REM sleep following sevoflurane anesthesia is indeed a homeostatic response to a REM sleep deficit that builds up during the anesthesia period. Additionally, we aimed to identify the brain regions responsible for the changes in REM sleep induced by sevoflurane. We found profound activation of a population of glutamatergic REM-on neurons in the SLD, the midbrain REM sleep generator, alongside robust increases in c-Fos expression in the PPT and LDT, two basal forebrain nuclei that are critical for REM sleep maintenance, during recovery from sevoflurane. We also observed a significant decrease in c-Fos expression in the vlPAG, a pontine nucleus known to suppress REM sleep. Animals exposed to sevoflurane presented a distinct expression pattern of c-Fos in a cluster of nuclei classically associated with REM sleep that resembled that seen after REM sleep deprivation. Additionally, c-Fos counts in the REM-active nuclei SLD, PPT and LDT were proportionately lower in mice subjected to 3 hours of sevoflurane anesthesia relative to 12 hours of REM deprivation, consistent with the hypothesis that the 3 hours of sevoflurane anesthesia correspond to time spent in REM sleep deprivation. These findings advance what is already known in the field of post-anesthetic sleep disruption by identifying the cellular and circuit-level substrates for sevoflurane-induced REM rebound and by providing new evidence in support of the idea that sevoflurane anesthesia does not fulfil REM sleep.

Our study differs from previous animal research on sleep recovery after sevoflurane in a few key ways.^7,8,10,11 First, we intentionally adopted an experimental framework devoid of any sleep deprivation prior to anesthesia, allowing us to test homeostatic REM mechanisms perturbed solely by anesthesia. Second, we obtained 24-hour-long post-anesthetic recordings to evaluate the extent of REM disruption across a full sleep-wake cycle. Additionally, we collected physiological data during anesthesia to rule out any potential derangements that might independently influence sleep recovery.

Our data indicate no differences in time spent awake or in NREM sleep between animals exposed to sevoflurane and controls, consistent with previous rodent studies on anesthetic-induced sleep disruption^7,9,10, but seemingly in contrast with optogenetic investigations that observed increased wakefulness and decreased NREM sleep after laser inhibition of the vlPAG.²⁶ Notably, REM sleep increases by 30% over 18–24 hours in animals recovering from anesthesia^7–9, but reaches twelve times baseline levels within seconds of optogenetic stimulation.²⁶ Thus, marked optogenetic-driven fluctuations in REM sleep may set in motion large shifts in wakefulness and NREM states, whereas more modest adaptations in REM sleep after anesthesia may not produce such shifts. Another important possibility is that sevoflurane may interact with neuronal targets outside of the vlPAG-SLD circuit to prevent changes in NREM sleep and wakefulness. Further studies will be needed to test the effects of sevoflurane outside of this circuit.

The duration of postoperative REM sleep disruption is poorly understood. In the only study that measured REM sleep during the 6^th postoperative night, REM duration remained elevated in 75% of patients.⁶ To better define the time course of REM changes in our animal model, we repeated EEG recordings in a small subset of animals one week after anesthetic exposure, and found no differences in REM sleep quantity or latency. These findings support the idea that the changes in REM sleep induced by sevoflurane are acute adaptations in response to the loss of REM sleep during anesthesia, as they subside as soon as there is an opportunity to sleep.^7–9 This phenomenon may be different in young animals, as suggested by our previous findings of long-lasting REM sleep changes after neonatal anesthesia.⁹

Our results appear to contradict a recent study in healthy human volunteers.²⁷ Mashour and colleagues found no changes in rest-activity rhythms measured by actigraphy in the first three days after isoflurane anesthesia. However, actigraphy does not allow for detailed measurement of REM sleep quantity, latency, or sleep architecture. Additionally, while brief anesthesia exposure may be enough to induce sleep adaptations in rodents, it may take repeated or extended sedative medication use in the Intensive Care Unit (ICU) to affect the human brain. Future research should explore different conditions ‒ including surgery and ICU settings with prolonged sedatives/anesthetics ‒ to better understand the clinical relevance of postoperative REM sleep disruption.

Clinicians have long noted that major cardiac, respiratory, and neurological complications tend to happen in the middle of the first week after surgery.^3,6 These complications include myocardial infarction, stroke and hypoxemia, which are more common during this period.^28–30 The reasons for these delayed complications, which occur two or more days after surgery, are not well understood. Importantly, REM rebound, characterized by marked swings in blood pressure and heart rate, increased airway collapsibility and breathing irregularity, could exacerbate some of the physiologic disturbances linked to surgery.^31–33 Additional research that mechanistically links postoperative REM sleep changes with hemodynamic, respiratory and neurological outcomes is needed to better understand the impact of sleep disruption on postoperative recovery.

We used a REM sleep deprivation protocol that involved gentle handling to prevent animals from entering REM sleep, monitored in real-time using EEG/EMG. We chose this method because it is believed to cause less stress to animals than other approaches^34–36 and since it was important to avoid perturbations in NREM sleep that would confound mapping of REM-active neurons. As shown in Supplemental Fig 2, REM sleep was increased in REM-deprived mice compared to controls (Panel A, P=0.031), whereas time spent in NREM sleep and awake remained unchanged (Panel B, P=0.180; Panel C: P= 0.551). REM sleep rebound occurred most prominently in the first hour after REM deprivation, in line with prior findings^10,23 (Suppl. Fig.2, Panel D). The number of handlings required to prevent animals from entering REM sleep also increased progressively during deprivation, reflecting an increase in REM homeostatic pressure (Suppl. Fig.2, Panel E).

Non-GABAergic anesthetics have also garnered attention for their impact on sleep. For example, the effects of the N-Methyl-D-aspartate (NMDA) antagonist ketamine on sleep have been extensively studied, particularly in the context of treatment-resistant depression. In humans, ketamine increases Slow-Wave Activity (SWA), boosts REM sleep on the night following an infusion, reduces REM latency the day after an infusion, and regulates circadian clock genes.^37,38 In rodents, ketamine decreases REM sleep, prolongs REM latency, elevates EEG delta power during NREM sleep, and increases overall gamma power.³⁹ Initial evidence suggests that in mice, dexmedetomidine, an alpha-2 receptor agonist, induces rebounds in both NREM and REM sleep, along with an increase in EEG delta power during NREM sleep.⁴⁰ In humans, dexmedetomidine has been found to promote NREM stage 3 while reducing REM sleep.^41,42

This study has limitations. First, the TRAP approach is recognized as superior to conventional c-Fos histochemistry for detecting neurons activated during complex behaviors like seizures, sensory coding, and memory consolidation.^22–25 Nonetheless, there is limited prior experience in using TRAP mice to identify neurons activated during sleep.²³ To mitigate this concern, we compared our TRAP data with the results obtained via c-Fos histochemistry, a well-established method for identifying neurons active during sleep. Our results support the efficacy of the TRAP approach in detecting REM sleep-active neurons. Second, we chose to administer sevoflurane at the start of the rest phase in rodents. We rationalized this approach as REM rebound would be maximized in our animal model, should sevoflurane anesthesia work similarly to a REM sleep deprivation. Additional studies should assess the effects of sevoflurane delivered during different circadian phases. Third, we used 100% oxygen as a carrier, which may have introduced potential confounding factors. We did not control for the order of measurements or animal cage location, both of which could also have biased our results. Last, we used healthy adult rodents, consistent with previous studies of post-anesthetic REM sleep profiles. However, older animals should be included in the future, as elderly patients are more prone to postoperative sleep disruption.

Supplementary Material

Supplemental Digital Content_1

Supplemental Figure 1. Panel A: Schematic representation of REM sleep duration in blocks of 3 hours in rats. A Repeated Measures Analysis of Variance showed that sevoflurane had a significant effect on time spent in REM sleep between hour 3 and 6 after sevoflurane (P= 0.031; N= 14 control (8 males, 6 females) control and 13 sevoflurane rats (7 males, 6 females)). Arrow: time of brain collection. Panel B: Schematic representation of REM sleep duration in 3 h blocks during recovery from sevoflurane anesthesia in TRAP mice (N= 6 control (2 males, 4 females) and N= 6 sevoflurane-exposed TRAP mice (3 males, 3 females)). Arrow: time of 4-hydroxytamoxifen injection. REM: Rapid Eye Movement.

Supplemental Digital Content_2

Supplemental Figure 2. Selective Rapid Eye Movement sleep deprivation protocol. Panel A: REM sleep duration was significantly increased in REM-deprived mice compared to controls (P= 0.031). Panels B, C: NREM sleep and wakefulness were unchanged in REM-deprived and control mice (P= 0.180 and P= 0.551, respectively). Panel D: Hour-by-hour representation of REM sleep (expressed as % of total sleep) following 12 h of REM deprivation. REM sleep deprivation had a significant effect on time spent in REM during the first and fourth hour of REM sleep recovery (P<0.0001 and P=0.023, respectively). dREM: REM-deprived. Arrow: time of 4-hydroxytamoxifen injection. Panel E: Schematic representation of the number of handlings required to prevent mice from entering REM sleep in blocks of 4 hours. N= 5 control TRAP mice (3 males, 2 females) and N=5 dREM TRAP mice (2 males, 3 females). REM: Rapid Eye Movement; NREM: non Rapid Eye Movement; dREM: REM sleep deprivation.