Overexpression of Sirt6 ameliorates sleep deprivation induced-cognitive impairment by modulating glutamatergic neuron function

Abstract

Sleep benefits the restoration of energy metabolism and thereby supports neuronal plasticity and cognitive behaviors. Sirt6 is a NAD⁺-dependent protein deacetylase that has been recognized as an essential regulator of energy metabolism because it modulates various transcriptional regulators and metabolic enzymes. The aim of this study was to investigate the influence of Sirt6 on cerebral function after chronic sleep deprivation (CSD). We assigned C57BL/6J mice to control or two CSD groups and subjected them to AAV2/9-CMV-EGFP or AAV2/9-CMV-Sirt6-EGFP infection in the prelimbic cortex (PrL). We then assessed cerebral functional connectivity (FC) using resting-state functional MRI, neuron/astrocyte metabolism using a metabolic kinetics analysis; dendritic spine densities using sparse-labeling; and miniature excitatory postsynaptic currents (mEPSCs) and action potential (AP) firing rates using whole-cell patch-clamp recordings. In addition, we evaluated cognition via a comprehensive set of behavioral tests. Compared with controls, Sirt6 was significantly decreased (P < 0.05) in the PrL after CSD, accompanied by cognitive deficits and decreased FC between the PrL and accumbens nucleus, piriform cortex, motor cortex, somatosensory cortex, olfactory tubercle, insular cortex, and cerebellum. Sirt6 overexpression reversed CSD-induced cognitive impairment and reduced FC. Our analysis of metabolic kinetics using [1-¹³C] glucose and [2-¹³C] acetate showed that CSD reduced neuronal Glu₄ and GABA₂ synthesis, which could be fully restored via forced Sirt6 expression. Furthermore, Sirt6 overexpression reversed CSD-induced decreases in AP firing rates as well as the frequency and amplitude of mEPSCs in PrL pyramidal neurons. These data indicate that Sirt6 can improve cognitive impairment after CSD by regulating the PrL-associated FC network, neuronal glucose metabolism, and glutamatergic neurotransmission. Thus, Sirt6 activation may have potential as a novel strategy for treating sleep disorder-related diseases.

Introduction

Sleep enables the restoration of neuronal energy that is needed for synaptic plasticity (Tononi and Cirelli, 2014), which underpins alertness, concentration, episodic memory, and working memory (Kuang et al., 2021). Sleep loss leads to altered levels of key metabolites consumed by cells for energy production, such as glucose (Knutson, 2007), amino acids (Davies et al., 2014), and lipids (Chua et al., 2015) in brain tissues. This results in a significant decrease in the relative metabolism of the prefrontal cortex, thalamus, and striatum (Wu et al., 2006). Clarifying the mechanisms underlying sleep-dependent neuronal-astrocytic metabolic coupling and the brain energy budget will significantly advance our understanding of how sleep benefits various cognitive functions.

Most of the brain energy budget appears to be expended on maintaining glutamatergic synaptic activity, which is critically important for learning and memory formation. In this respect, glutamate homeostasis is essential for brain activity. Neuronal glutamate levels can be affected by extracellular glutamate reuptake, glutamine conversion in astrocytes, the glutamine shuttle, and glutamate synthesis in neurons (Cooper and Jeitner, 2016; Hertz and Chen, 2017). A previous study reported that abnormal glucose metabolism in the brain may affect glutamate synthesis and even glutamatergic neurotransmission (Chowdhury et al., 2017). In this regard, the altered kinetics of energy/glucose metabolism after sleep deprivation (SD) is a potential mechanistic approach to understanding SD-associated abnormal cerebral mechanisms.

Sirtuins (Sirt) are a family of seven NAD⁺-dependent protein deacetylases (SIRT1–7) that are involved in various biological processes including transcription and metabolism (Carafa et al., 2012; Yang et al., 2021). Sirt6, an essential regulator of glucose metabolism, might regulate metabolic kinetics in the brain, and adjust the functional brain network after SD. Sirt6 is well characterized as a deacetylase of H3K9 and H3K56, and a corepressor of several transcription factors (e.g., the nutrient stress regulator hypoxia-inducible factor 1-alpha) (Zhong et al., 2010). Through this molecular function, Sirt6 has been shown to play a critical role in glucose homeostasis and genomic stability. For example, the loss of Sirt6 could cause a shift from mitochondrial respiration to anaerobic glycolysis, even with adequate oxygen supplies. Although a conditional Sirt6 knockout in the brain is not lethal, it can cause early growth retardation because of decreased levels of the growth hormone insulin-like growth factor-1, which can lead to impairments in learning and other forms of cognition (Schwer et al., 2010; Zhang et al., 2018). Emerging evidence has demonstrated that Sirt6 is involved in maintaining energy metabolism in the neocortex, and the loss of Sirt6 may trigger premature aging (Kanfi et al., 2012), leading to Alzheimer’s disease (Kaluski et al., 2017) and other age-related neurodegeneration (Garcia-Venzor and Toiber, 2021). Thus, Sirt6 is a good target for bridging cerebral functions and sleep.

The prelimbic cortex (PrL), one of the cytoarchitectonic parts of the medial prefrontal cortex, is critical for complex behaviors, such as attention, memory, and behavioral flexibility (Alexander and Brown, 2011; Minxha et al., 2020; Jiang et al., 2022). These complex behaviors require functional interactions between the PrL and other brain areas such as the hippocampus (Gordon, 2011; Brincat and Miller, 2015) and amygdala (Likhtik and Paz, 2015; Burgos-Robles et al., 2017). These cognitive behaviors are susceptible to SD (Acosta-Peña et al., 2015), indicating that sleep loss may perturb the PrL. Indeed, one study (Noorafshan et al., 2017) reported that 21 days of SD in rats led to decreases in PrL volume, the total neuronal and glial count, dendrite length, and the number of spines per dendrite, and these changes were associated with impaired working and reference memory. Although the effect of sleep loss on the PrL and other brain regions has been well studied as a neurophysiological and behavioral event, the potential molecular targets for improving PrL function after SD remain limited.

In the present study, we examined whether Sirt6 serves as a glycolytic regulator to improve SD-induced cognitive impairment, as this topic has been seldom investigated. We proposed that Sirt6 modulates neuronal-astrocytic glucose metabolism to restore glutamate synthesis and synaptic plasticity, leading to strengthened functional connectivity (FC) between the whole brain and PrL, and ultimately cognitive improvement. To test this hypothesis, we assessed animal behaviors, FC, metabolic kinetics, and glutamatergic neurotransmission to assess the effect of Sirt6 overexpression in the PrL on the restoration of brain function after chronic SD (CSD).

Methods

Animals

All animal experiments were conducted following the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals (National Research Council, 2011), and the procedures were approved by the Animal Ethics Committee of the Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences (approval No. APM22022A; approved on July 29, 2022). Nine-week-old male C57BL/6J mice (~25 g, Liaoning Changsheng Biotechnology Co., Benxi, Liaoning, China) were group-housed in a specific-pathogen-free room maintained at 23 ± 1°C with a 12/12-hour light-dark cycle. The mice were randomly divided into three groups (n = 28/group) as follows: control group (Con + eGFP), in which mice were transduced with AAV-eGFP virus and allowed to sleep naturally; CSD group (CSD + eGFP), in which mice were transduced with AAV-eGFP virus and subjected to CSD for 14 days; and CSD + Sirt6 overexpression group (CSD + Sirt6), in which mice were transduced with AAV overexpressing Sirt6-eGFP and subjected to CSD for 14 days.

Virus injections

Seven days before the onset of CSD, the mice were anesthetized with ketamine (dissolved in 1% pentobarbitone, intraperitoneal injection, 40 mg/kg, Jiangsu Hengrui Pharmaceuticals Co., Ltd., Lianyungang, China). Then, the head of each mouse was fixed in place using a stereotaxic frame (RWD Life Science, Shenzhen, China), and a skin incision was made to expose the skull. The virus was bilaterally injected into the PrL (Paxinos and Franklin, 2001) (anterior-posterior: +2.68 mm, medial-lateral: ±0.5 mm, dorsal-ventral: –1.52 mm) at a rate of 20 nL/min using a 10 µL microsyringe (Gaoge, Shanghai, China) connected to a micropump (UMP3, WPI, Worcester, MA, USA). In the mice in the CSD + Sirt6 group, the microsyringe was preloaded with 200 nL AAV2/9-CMV-bGlobin-MCS-Sirt6-EGFP-3FLAG-WPRE-hGH-polyA (titer 7.65 × 10¹² vg/mL; Shanghai Genechem Co., Ltd., Shanghai, China), which was bilaterally delivered to the PrL. The control virus, i.e., AAV2/9-CMV-bGlobin-MCS-EGFP-3FLAG-WPRE-hGH-polyA, was bilaterally injected into the PrL of the mice in the Con + eGFP and CSD + eGFP groups. At the end of the injection, the microsyringe was kept in place for 10 minutes to prevent virus backflow upon withdrawal.

To capture the dendritic spine density of individual neurons, we employed sparse labeling based on the recombinase system-dependent co-packaging strategy (n = 3/group). The AAV-sparse-RFP virus was produced by co-packaging the AAV-CMV-cre plasmid and AAV-DIO-mCherry plasmid at a ratio of 1:200,000 in a single rAAV production step (AAV2/9-sparse-RFP, total virus titer = 5.45 × 10¹² vg/mL, BrainCase Co., Ltd., Shenzhen, China). Glass micropipettes preloaded with a mixture of AAV2/9-sparse-RFP and AAV2/9-CMV-bGlobin-MCS-Sirt6-EGFP-3FLAG-WPRE-hGH polyA (1:2, 200 nL) or a mixture of AAV2/9-sparse-RFP and AAV2/9-CMV-bGlobin-MCS- EGFP-3FLAG-WPRE-hGH polyA (1:2, 200 nL) were gradually lowered into the unilateral PrL of mice in the CSD + Sirt6, Con + eGFP, and CSD + eGFP groups to deliver the virus at a rate of 20 nL/min. The microsyringe was kept in place for 10 minutes to prevent virus backflow upon withdrawal.

CSD

According to previous studies, CSD was established via the modified multiple platform technique (Machado et al., 2004; Holth et al., 2019). Multiple platforms (outer diameter: 2.5 cm; height: 5.0 cm) were placed in a water tank (19 cm × 28 cm × 40 cm) and spaced ~5.0 cm apart. The tank was filled with water to a depth of ~2.0 cm below the surface of the platforms. During CSD, the mice were allowed to move freely between the platforms. At the beginning of each rapid eye movement (REM) episode, the mice lost muscle tonus and fell into the water, which caused awakening (Zager et al., 2007). The mice in the control group were placed in the same tank, which had been fitted with a stainless wire mesh fixed at a height of ~2.0 cm above the water. This allowed them to lie down without touching the water. The CSD mice were placed in the chamber for 20 hours a day (14:00–10:00) and allowed to sleep naturally in their home cage for 4 hours (10:00–14:00) a day for 2 weeks. The control mice had a similar schedule. The whole experimental procedure is illustrated in Figure 1.

Figure 1.

Schematic diagram showing the experimental design.

The AAV2/9-Sirt6 or AAV2/9-eGFP virus was injected into the bilateral PrL 7 days before the beginning of sleep deprivation. Then, mice were subjected to chronic sleep deprivation from day 1 to day 14. ¹H-NMR: Proton nuclear magnetic resonance spectroscopy; CSD: chronic sleep deprivation; GFP: enhanced green fluorescent protein; fMRI: functional magnetic resonance imaging; IF: immunofluorescence; NOR: novel object recognition; OFT: open filed test; S: sleep; SD: sleep deprivation; Sirt6: Sirtuins 6; SL: sparse labeling; TCT: three-chamber social test; WB: western blot analysis; YM: Y-maze.

Electrophysiological recording and sleep architecture analysis

Mice (n = 4/group) were exposed to isoflurane (1.5–2.5% during the whole surgical period; RWD Life Science) and placed in a stereotaxic frame. Four wire electrodes were screwed into the skull surface for electroencephalogram (EEG) recordings. Two were bilaterally inserted into the left and right frontal regions, and the other two were inserted into the left and right parietal regions, respectively. An additional pair of insulated wire electrodes was planted into the nuchal muscle for electromyogram (EMG) recording. The electrodes were connected to a micro-connector and fixed onto the skull surface with dental acrylic resin. After 7 days of postoperative recovery, the EEG and EMG signals were recorded at 1000 Hz using the Medusa small animal electrophysiology recording system (Medusa, Bio-Signal Technologies, Nanjing, China).

For sleep architecture analysis, the data were converted to the European data format (edf) using Bio-utility software (Bio-Signal Technologies). The sleep states were categorized as wake, REM sleep, or non-rapid eye movement (NREM) sleep using automated sleep-scoring software (LunionData, Shanghai, China). The wakefulness state was defined by the presence of small amplitude and high-frequency EEG with high muscle tone. REM sleep was identified by the presence of high theta (6–10 Hz) power in the EEG with muscle atonia. NREM sleep was marked by the presence of high-voltage, low-frequency EEG signals (1–4 Hz) with low EMG activity (Jiang-Xie et al., 2019).

Neurobehavioral tests

The mice (n = 8/group) were subjected to different behavioral tests from day 16 to day 20. To address ethical concerns regarding the number of mice used in the study, the same mice were subjected orderly to perform the open field test (OFT) on day 16, the novel object recognition (NOR) test on day 18, the Y-maze test on day 19, and the three-chamber social test on day 20, which could eliminate the influence of the stress from each task transition as much as possible. Descriptions of the behavioral tests are provided in the following sections.

OFT

Based on our previous study (Zhu et al., 2020), mice were transported to the testing room at least 1 hour prior to the experiment. The open field apparatus consisted of a chamber (40 cm × 40 cm × 40 cm) made of gray polyvinyl chloride. Each mouse was randomly placed in the center of the chamber. The total distance moved and the amount of time spent in the center zone were automatically recorded using the Any-maze video-tracking system (Stoelting Inc., Kiel, WI, USA). The arena was cleaned with ethanol between trials to limit the presence of olfactory cues. All OFT sessions were performed during the light phase of the sleep/wake cycle.

NOR test

The NOR test was carried out according to a previous study (Zhu et al., 2020). Before the training sessions, mice were placed in the apparatus for 3 minutes to adapt to the environment. After acclimation, the mice were subjected a training session in which two identical objects were introduced to the apparatus for 5 minutes. For the test session, performed 2 hours after training, one of the two objects from the training session was replaced with a novel object with a different shape but other identical properties. The mice were allowed to explore these two objects for 5 minutes. The time spent exploring each object was recorded by the Any-maze video-tracking system, and the recognition index and novel object entries were calculated. The recognition index value was obtained as the ratio between the time spent in the novel object and the total duration of the test.

Y-maze test

The protocol was based on a previous study (Kraeuter et al., 2019). The Y-maze apparatus consisted of three arms of equal size. One of the three arms was randomly selected as the novel arm during the training session and was blocked with a gray shield. The mice were allowed to freely explore the remaining two arms for 10 minutes. For the test session, carried out 1 hour after training, the novel arm was opened and the mice were allowed to explore the three arms for 5 minutes. The time spent in the novel arm, number of novel arm entries, and average speed were recorded and analyzed using the Any-maze video-tracking system. The novel arm preference index was calculated as the ratio between the time spent in the novel arm and the total trial duration (the time spent in the novel arm and the familiar arms).

Three chamber social test

We used the three chamber social test to assess social interactions and social novelty preferences (Moy et al., 2004). The test was carried out across three sessions using a three-chambered apparatus with openings between the chambers. During the habituation session, the openings were blocked with two gray shields, and each mouse was placed in the central chamber for 5 minutes. For the social interaction session, the mouse encountered an unfamiliar gender-matched mouse (Stranger I) under one wire cup and an empty wire cup on the left or right side of the chamber. The shields were removed, and the mice were allowed to explore for 10 minutes. In the social novelty session, the mice encountered the first intruder as well as another never-before-met mouse (Stranger II). This session was 10 minutes long. The time spent sniffing each wire cup and the time spent in each chamber were recorded by the Any-maze video-tracking system. The social preference index was calculated as the ratio between the time spent interacting with the cup containing Stranger I and the total trial duration (the time spent interacting with the cup containing Stranger I and the empty wire cup) in the social session, and the social novelty preference index was calculated according to the time spent interacting with the cup containing Stranger II and the total trial duration (the time spent interacting with the cups containing Stranger II and Stranger I).

Magnetic resonance imaging

To eliminate the effect of the acute SD introduced by the last round of CSD, mice (n = 8/group) were allowed to recover for 2 days before functional magnetic resonance imaging (fMRI). On the 16^th day after CSD onset, each mouse was initially anesthetized with 4.0–5.0% isoflurane, followed by maintenance of the anesthesia state using 1.0–1.5% isoflurane. After the mouse was anesthetized, it was gently placed on the animal bed of a 7.0T Biospec small animal system (Bruker, Saarbrücken, Germany) for in vivo fMRI. The animal’s body temperature was maintained via a cycling water heating system located under the animal bed. The mouse head position was manually fixed in place via two ear bars and a tooth bar for imaging acquisition, and the animal’s respiratory rate was continuously monitored and maintained at 70–90 breaths/min to maintain light anesthesia.

A 20-cm birdcage coil was used for transmission during magnetic resonance imaging (MRI) and a 10 mm surface coil was used for receiving. T2-weighted high-resolution imaging was obtained using a fast spin-echo sequence (turbo spin echo-rapid acquisition with refocusing echoes), and blood-oxygen-level-dependent signals were obtained via a single-shot spin-echo planar imaging sequence. The parameters of turbo spin echo-rapid acquisition with refocusing echoes and the spin-echo planar imaging sequence were similar to our previous studies (Cai et al., 2021). All MRI studies were performed between 9:00 and 17:00 to generate consistent MRI data for all animals.

The raw MRI data were converted to the NIFITI format using Bru2anz software (Bruker, Saarbrücken, Germany). The EPI data were preprocessed and analyzed for despiking, slice timing, motion correction, smoothing, and registration using free spmmouseIHEP software (Nie et al., 2013). To analyze the resting-state fMRI data, the preprocessed data were detrended, filtered (0.01–0.1 Hz), and regressed with the signals from the white matter and cerebrospinal fluids in Dpabi.

To analyze the region of interest (ROI)-based whole-brain network between the groups, the PrL region was automatically defined as the ROI according to the mouse brain template (Nie et al., 2019). FC was calculated via a seed-based correlation analysis on a voxel-by-voxel basis. Briefly, the time courses of all voxels in the PrL were averaged and used as the reference time course. Then, the Pearson cross-correlation coefficients between the reference time course and the other time courses of individual voxels were calculated to quantify the FC in the whole brain. Following this, the FC underwent Fisher’s Z-score transformation within the mask, and a FC map was constructed for each mouse. Given the limited number of animals and to avoid false negative results, the FC between the Con + eGFP, CSD + eGFP/CSD + Sirt6, and CSD + eGFP groups was compared using a one-tailed two-sample t-test. The significant regions were calculated based on a voxel-threshold of P < 0.05 (uncorrected), followed by GRF correction (Bardeen et al., 1986) with P < 0.05. Furthermore, the cluster-extern threshold of contiguous voxels was set to 100 to reduce the false positive rate. The whole procedure was conducted using a toolbox plug-in in SPM (Nie et al., 2019).

[¹H-¹³C]-nuclear magnetic resonance study

As illustrated in Figure 1, the assessments of metabolic kinetics were performed on the 17^th day after the fMRI study using [¹H-¹³C]-nuclear magnetic resonance (NMR) technology. The experiments were conducted similarly to our previous study (Fang et al., 2022), except that two different ¹³C labeled chemicals, namely [1-¹³C] glucose (Glc₁) and [2-¹³C] Acetate (Ace₂), were used in the current study. While Ace₂ is taken up and metabolized exclusively by astrocytes before being converted to [2-¹³C]-acetyl CoA, Glc₁ can be transported and metabolized in both astrocytes and neurons (de Graaf et al., 2003). The whole metabolic processes of Ace₂ and Glc₁ are illustrated in Figure 2. Previously, a series of metabolites, including several neurotransmitters, was successively labeled with the ¹³C isotope through glutamine-glutamate or glutamine-γ-aminobutyric acid (GABA) metabolic cycles and quantified by NMR. It is also possible to distinguish the contribution of neuronal-astrocytic metabolism by analyzing changes in Ace₂- or Glc₁-derived differential metabolites together. Therefore, we used these two different ¹³C labeled probes to investigate the influence of CSD on neuronal and astrocyte metabolism. The experimental procedure is summarized below.

Figure 2.

Schematic diagram of Glc₁ and Ace₂ metabolism in glutamatergic neurons, GABAergic neurons, and astrocytes.

Ace₂: [2-¹³C] acetate; CoA: coenzyme A; GABA: γ-aminobutyric acid; Glc₁: [1-¹³C] glucose; Gln: glutamine; Glu: glutamate; Pyr: pyruvate; TCA: tricarboxylic acid cycle; α-KG: α-ketoglutarate.

To decrease the influence of endogenous glucose on ¹³C labeling, all animals were fasted from 17:00 before the day of the experiment. On the test day, each mouse was initially anesthetized with 2.0% isoflurane, and tail vein catheterization was performed with a PE10 tube (Instech, Plymouth Meeting, PA, USA) for infusion of ¹³C-labeled probes (Qingdao Tenglong Weibo Technology Co., Ltd., Qingdao, China). The animal was allowed to recover and move freely for 15 minutes. Afterwards, the ¹³C labeled probe was infused through the tail vein for 2.0 minutes at a constant rate. The total amounts of Glc₁ (0.75 M) and Ace₂ (1 M) were 7.65 and 14 mmol/kg, respectively. After the infusion procedure, the mice moved freely in their cages and showed no signs of stress. After 30 minutes, the mice were deeply anesthetized with isoflurane and euthanized using a head-focused microwave machine (PJ21CAU, Midea, Foshan, China), similar to a previous study (Liu et al., 2020). The prefrontal cortex was collected and the metabolites were extracted with 60% ethanol according to the published protocol (Liu et al., 2020). The extracted metabolites were dissolved in phosphate buffer (pH 7.2, 0.6 mL D₂O with 0.2 M Na₂HPO₄/NaH₂PO₄) and 5 mM 3-(trimethylsilyl) propionic-2, 2, 3, 3-d4 acid sodium salt was used as the inner standard chemical in the buffer. The mixture was centrifuged and the supernatant (~0.53 mL) was collected for [¹H-¹³C]-NMR analysis.

All samples were detected at 25°C using a 500-MHz NMR spectrometer (BrkerBiospin, Saarbrücken, Germany). We used the pulse sequence for proton-observed carbon-edited spectroscopy (Spec, [¹H-¹³C]-NMR) to differentiate and detect the metabolites labeled with ¹³C in the different groups. This method has been widely used in metabolic kinetics studies (Mishra et al., 2020; Fang et al., 2022). We used the following acquisition parameters. Echo time: 8 ms; sweep width: 20 ppm; repetition time: 20 seconds; number of scans: 64; and data acquisition: 64 K.

There were two spin echoes in the [¹H-¹³C]-NMR Spec. The first one (Spec1) reflected the concentration of total metabolites including ¹²C and ¹³C labeled chemicals. The second (Spec₂) was related to the different concentrations between the ¹²C- and ¹³C-labeled chemicals. Thus, we calculated the concentrations of ¹³C-labeled metabolites using the differences in the peak areas in the associated NMR regions between the two spectra: A_C13 = (A_Spec1 – A_Spec2)/2. The ¹³C enrichment was calculated using the ratio of A_C13/A_Spec1. All NMR data were processed using Topspin commercial software and custom-made NMRSpec software, which is widely used in metabolomics (Sun et al., 2021; Liu et al., 2022) and metabolic kinetics (Guo et al., 2021; Fang et al., 2022).

The NMR data were preprocessed with phase and baseline correction in Topspin and the processed spectra were automatically imported to NMRSpec for peak alignment and integration (Guo et al., 2021). To calculate the ¹³C enrichment of the metabolites, the peak area of the same region related to the pure metabolite in Spec₂₁ and Spec₂ was automatically integrated using NMRSpec.

Gene analysis in the PrL region by quantitative polymerase chain reaction

The mice (n = 3/group) were deeply anesthetized with isoflurane and the PrL regions were quickly dissected, frozen in liquid nitrogen, and stored at –80°C for further analysis. The total RNA was obtained using TriZol (Invitrogen, Carlsbad, CA, USA). The first-strand complementary DNA was synthesized using a reverse transcription kit (Invitrogen). Specific primers were used as follows: mouse-Glu1 (glutamate synthase 1), (forward) 5′-AAT TTG ATT ATG ATG GAC CAC TCA T-3′ and (reverse) 5′-GTC CAC TAG GTA GTT ACC TCT GCT C-3′; mouse-Abat (GABA aminotransferase), (forward) 5′-AGC AGG AAT ATA CTC TTA TGG GAA C-3′ and (reverse) 5′-GTA ATC TTG ACT CCA GCA TAC AAG C-3′. The gene expressions of Glu1 and Abat were detected using the Bestar quantitative polymerase chain reaction Master Mix (DBI, Ludwigshafen, Germany), and the expression of β-actin ((forward) 5′-CGT TGA CAT CCG TAA AGA CCT C-3′ and (reverse) 5′-TAG GAG CCA GGG CAG TAA TCT-3′) was used as the internal control.

Western blot analysis

The mice (n = 3/group) were anesthetized with isoflurane and the frontal cortices were collected for further examination of Sirt6 protein expression. The total proteins were obtained via radioimmunoprecipitation assay with a protein lysate buffer (Aspen Biological, Wuhan, China), and subjected to 8–15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis for separation. Then, the proteins were transferred to a polyvinylidene fluoride membrane (Aspen Biological) and incubated with 5% skimmed milk powder (AS1033, Aspen Biological). The membranes were then incubated with rabbit anti-Sirt6 antibody (1:1000, Cell Signaling Technology, Boston, MA, USA, Cat# 12486S, RRID: AB_2636969) and rabbit anti-tubulin polyclonal antibody (1:1000, Thermo Fisher Scientific, Waltham, MA, USA, Cat# 10094-1-AP, RRID: AB_2210695) at 4°C overnight. After incubation with horseradish peroxidase-conjugated goat anti-rabbit secondary antibody (1:1000, Aspen Biological, Cat# SA003) for 1.5 hours at room temperature (~26°C), the membranes were detected using a chemiluminescent imaging system (Tanon, Shanghai, China). The relative optical density of the bands was quantified using ImageJ software (v 1.8.0_322, National Institutes of Health, Bethesda, MD, USA) (Schneider et al., 2012).

Immunohistochemistry

Mice (Sirt6 protein immunopositivity: n = 4/group, two groups; Sirt6 transfection efficiency: n = 4/group, two groups; Sirt6 transfection specification: n = 3/group, one group; vesicular glutamate transporter 1 (VGLUT1) immunopositivity: n = 3/group, three groups; sparse labeling: n = 3/group, three groups) were deeply anesthetized with isoflurane and then perfused sequentially with saline and 4% paraformaldehyde (Merck, Darmstadt, Germany). The whole brain was dissociated and post-fixed in 4% paraformaldehyde for 2 days, followed by dehydration in 30% sucrose solution. The frozen brains were cut into 40-µm sections using a Thermo Fisher cryostat microtome (NX50). The brain sections were stored in an antifreeze solution at –20°C. For staining, the sections were washed 3× in phosphate buffer saline for 10 minutes and blocked with 10% normal goat serum in phosphate buffer saline containing 0.01% Triton X-100. Next, all sections were washed 3× in phosphate buffer saline for 10 minutes and incubated with rabbit anti-Sirt6 antibodies (1:500, CST, Cat# 12486S, RRID: AB_2636969), mouse anti-NeuN antibodies (1:100, Merck, Darmstadt, Germany, Cat# MAB337, RRID: AB_2313673), rabbit anti-ionized calcium binding adapter molecule 1 (IBA1) antibodies (1:500, Fujifilm, Wako Chemicals, Tokyo, Japan, Cat# 019-19741, RRID: AB809504), anti-glial fibrillary acidic protein (GFAP) antibodies (1:500, Abcam, Cat# ab7260, RRID: AB_305808), or pig anti-VGLUT1 antibodies (1:500, Synaptic Systems, Göttingen, Germany, Cat# 135304, RRID: AB_887878) at 4°C overnight. Thereafter, the sections were washed 3× in phosphate buffer saline for 10 minutes and incubated with Cy3 goat anti-mouse (1:1000, Jackson ImmunoResearch Laboratories Inc., Lansing, MI, USA, Cat# 115-165-003, RRID: AB_2338680), Alexa Fluor® 647 goat anti-rabbit (1:1000, Jackson ImmunoResearch Laboratories Inc., Cat# 111-607-008, RRID: AB_2632470), Cy3 goat anti-guinea pig (1:1000, Thermo Fisher, Cat# A_11076, RRID: AB_141930), or Alexa Fluor® 488 goat anti-rabbit (1:1000, Jackson ImmunoResearch Laboratories Inc., Cat# 111-545-045, RRID: AB_2338049) antibodies for 1 hour at room temperature (~26°C). Finally, all sections were counterstained with DAPI (Cat# C1002; Beyotime Biotechnology, Shanghai, China) for 10 minutes and washed three times.

We used a TCS SP8 confocal microscope (Leica) for image acquisition. Images were obtained with a resolution of 1024 pixels in the X-Y dimension. Z dimensions were variable. To assess dendritic spine density and VGLUT1 expression, z-stack images were acquired at a step size of 2 µm per slide under a 63× oil objective lens. The dendritic spines and terminal VGLUT1 expression of self-expressing PrL neurons were quantified by a researcher blinded to the condition. For dendritic spine calculation, we used the semi-automagical spine estimation method in ImageJ, which has been successfully applied to spine counting in previous studies (Wang et al., 2016; Soler et al., 2018). Briefly, the photograph was converted into an 8-bit image to minimize the effect of counter-staining (Additional Figure 1 ^{(362.6KB, tif)} ). Then, the 8-bit image was binarized and the skeleton of the dendrite was obtained based on the standard skeletonization function (Additional Figure 1 ^{(362.6KB, tif)} ). Finally, the number of skeleton endings was automatically calculated using the ImageJ plug-in for skeleton analysis. To determine the specific viral infection, co-staining of enhanced green fluorescent protein (eGFP) with NeuN, GFAP, or IBA1 was manually counted by a researcher who was blinded regarding the aim of the experiment.

Whole-cell patch-clamp recording

Mice (n = 3/group) were anesthetized with isoflurane and the whole brain was quickly dissociated into pre-chilled and oxygenated dissection fluid containing (in mM) 213 sucrose, 10 glucose, 26 NaHCO₃, 3 KCl, 1 NaH₂PO₄·2H₂O, 10 MgCl₂, and 0.5 CaCl₂. Acute brain slices (300 µm) containing the PrL were acquired in chilled dissection fluid using a microtome (VT1000S, Leica). Sections were transferred to the incubation chamber and immersed in artificial cerebrospinal fluid containing (in mM; 125 NaCl, 26 NaHCO₃, 5 KCl, 1.2 NaH₂PO₄, 2.6 CaCl₂, 1.3 MgCl₂, and 10 glucose) at 30°C for 1 hour. After incubation, sections were placed in the slice chamber for electrophysiological recording with continuous perfusion of artificial cerebrospinal fluid (saturated with 95% O₂/5% CO₂).

The PrL neurons were identified by eGFP and visualized for recording using an infrared-differential interference contrast microscope (Olympus, Tokyo, Japan). We used patch pipettes (4–8 MΩ, WPI, Sarasota, FL, USA) for the whole-cell patch clamp recordings. The signal was amplified by a MultiClamp 700B amplifier (Molecular Devices, San Jose, CA, USA). The miniature excitability postsynaptic current (mEPSC) was sampled in the presence of tetrodotoxin (1 μM) and picrotoxin (100 μM) at –70 mV. The patch pipettes were filled with intracellular solution containing (in mM) 122 potassium-gluconate, 5 NaCl, 2 MgCl₂, 0.3 CaCl₂, 10 HEPES (2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid), 5 EGTA (ethylene glycol-bis (β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid), 4 Mg-ATP, and 0.3 Na₃-GTP (guanosine-5′-triphosphate, pH 7.4; osmolarity: 300 mOsm). The action potentials were recorded using the injected current (from 0 pA to 300 pA in 50 pA increments, 500 ms duration) without any synaptic transmission blockers. We calculated the number of action potentials induced by each injected current (shown as firing rate – injected current (f-I) curves), frequency (f-I), and amplitude of the mEPSCs using pCLAMP10.7 software.

Statistical analysis

All data are represented using the mean ± standard error of mean (SEM). Normally distributed data with two groups were tested using a Student’s t-test. Datasets with more than two groups were analyzed using a two-way analysis of variance followed by Tukey’s post hoc test. Statistical significance was assessed based on P values (P < 0.05) in Matlab R2019b (Mathworks, Torrance, CA, USA).

Results

Validation of SD protocol

To investigate the influence of the SD protocol, we obtained the sleep architecture from the EEG/EMG monitoring system from 14:00 to 14:00 of the next day. During the SD period (14:00 to 10:00 of the next day), the SD induced by the modified multiple platform technique had a significant effect on the sleep stages (Additional Table 1). Specifically, compared with the control mice (Figure 3), SD led to an increase of 46% for the waking state (P = 0.03), decrease of 90% for REM sleep (P = 0.009), and a slight decrease of 31% for NREM sleep, although this was not significant (P = 0.09). During the day-night cycle (Figure 3, Additional Figure 2 ^{(386KB, tif)} and Additional Table 2), the amount of time spent in REM sleep significantly decreased (P = 0.01) from 1.95 ± 0.24 hours to 0.53 ± 0.08 hours, and the amount of wake time significantly increased (P = 0.04) from 11.04 ± 0.79 hours to 14.77 ± 0.54 hours in the CSD group. However, the amount of time spent in NREM slightly decreased without any significant differences (P = 0.11). Collectively, these results demonstrated that the modified multiple platform protocol served as an effective method for REM SD (Machado et al., 2004).

Additional Table 1.

Different brain states in a day-night cycle for normal and SD mice

States	14:00 (day 1)-10:00(day2), 20h		10:00 (day 2)-14:00(day2), 4 h		14:00 (day 1)-14:00 (day2), 24h
	Con	SD	Con	SD	Con	SD
Wake (h)	9.32±0.77	13.58±0.60*	1.72±0.10	1.19±0.08	11.04±0.79	14.77±0.54*
REM(h)	1.75±0.23	0.18±0.06**	0.20±0.02	0.36±0.03*	1.95±0.24	0.53±0.08*
NREM(h)	8.94±0.64	6.21±0.63	2.08±0.09	2.45±0.06	11.02±0.65	8.65±0.58

Data are presented as mean ± SEM, and were analyzed by Student’s t-test. *P < 0.05, **P < 0.01. Con: control group; NREM: non-rapid eye movement; REM: rapid eye movement; SD: sleep deprivation group.

Figure 3.

Influences on sleep stages by sleep deprivation.

(A) Representative signals for EEG, EMG, and sleep state analysis during the whole SD period for mice in the Control or SD groups. (B) Representative EEG and EMG signals for different sleep states, such as wake, REM, and NREM. (C) Statistical analysis of the influence of SD on sleep stages. SD had a significant effect on sleep stages compared with control mice. Specifically, SD increased the waking state by 46%, decreased REM sleep by 90%, and slightly decreased NREM sleep by 31%, although this was not significant. All data are represented by the mean ± SEM and were analyzed by Student’s t-test. Con: Control group; EEG: electroencephalogram; NREM: non-rapid eye movement; REM: rapid eye movement; SD: Sleep deprivation group.

Additional Table 2.

Information of different clusters during fMRI comparisons.

ROI	CSD+eGFP >Con+eGFP					CSD+eGFP >Con+eGFP					CSD+Sirt6>CSD+eGFP					CSD+eGFP >CSD+Sirt6
	Voxel	t-value	X	Y	Z	Voxel	t-value	X	Y	Z	Voxel	t-value	X	Y	Z	Voxel	t-value	X	Y	Z
Cluster1	6421	5.512	-1.295	5.285	1.901	1115	2.949	-0.918	3.715	3.291	7550	5.358	-1.027	1.026	-5.915	662	3.227	-3.622	3.403	-5.057
Cluster2	2356	3.281	1.778	1.422	-7.091	513	3.136	1.995	5.678	-5.392	5729	5.012	1.585	2.279	4.092	504	3.083	-0.231	5.223	-1.865
Cluster3	2162	3.611	-0.356	0.651	2.546	376	2.783	-1.095	6.005	-5.852	2659	5.115	-1.589	4.603	2.481	242	4.598	2.958	5.215	-4.341
Cluster4	385	2.685	-2.657	2.999	-7.769	288	2.535	3.342	5.080	-1.639	678	3.513	-2.662	2.873	2.696	201	2.627	2.075	3.106	-2.014
Cluster5	325	2.607	3.239	3.442	-7.539	267	2.106	0.627	3.699	3.406	637	3.832	2.844	2.020	-7.203	145	2.897	-0.724	4.046	5.177
Cluster6	172	2.638	-0.351	1.766	5.265	215	3.520	3.431	3.446	-5.893	480	2.486	-1.319	1.287	4.085	143	3.027	-3.611	5.191	-6.211
Cluster7	154	2.631	0.416	0.337	1.130	210	3.436	-2.481	0.416	-1.451	356	2.368	0.436	3.926	2.116	139	3.632	-3.817	3.032	-6.002
Cluster8	110	2.388	-2.839	5.352	-2.094	179	2.242	0.934	5.824	-8.093	293	2.284	-0.040	4.546	-7.284
Cluster9						162	2.301	-1.008	4.170	-6.934	206	3.479	-2.460	3.864	-5.642
Cluster10						146	2.774	-2.453	5.046	-6.568	174	3.207	-2.664	1.881	-5.901
Cluster11						125	2.414	-1.290	5.595	-7.386	141	4.441	-0.358	0.000	-2.519
Cluster12						113	2.222	0.723	2.883	-7.189	138	3.411	2.087	5.298	-0.222
Cluster13											115	2.328	-1.889	2.466	-3.191
Cluster14											104	2.556	-0.343	2.206	-8.254
Totalsize	12085					3709					19260					2036

Cluster: The different fMRI signals were clustered together, and generated different clusters during the comparisons; CSD: chronic sleep deprivation; eGFP: enhanced green fluorescent protein; fMRI: functional magnetic resonance imaging; ROI: region of interest; Sirt6: Sirtuins 6; X, Y, Z: the zero location is bregma

Sirt6 expression in the PrL after CSD

A previous study demonstrated that CSD could lead to a relative decrease in cerebral glucose metabolism (Wu et al., 1991). Sirt6 is important for glucose homeostasis in the liver (Kim et al., 2010), and is highly expressed in the immature brain (Garcia-Venzor and Toiber, 2021). Thus, we measured Sirt6 protein expression in the PrL to determine the influence of CSD. As shown in Figure 4A, the Sirt6 protein was primarily restricted to the cell nuclei. Compared with the animals in the control group, the CSD mice exhibited significantly reduced (P < 0.05) Sirt6 expression in the PrL, as demonstrated by immunohistochemistry (Figure 4A and B) and western blotting (Figure 4C and D).

Figure 4.

Sirt6 expression in the PrL after chronic sleep deprivation treatment via immunohistochemistry (A and B) and western blot assay (C and D).

Chronic sleep deprivation reduced Sirt6 expression (green, Alexa Fluor 488) in neurons (marked by NeuN, red, Cy3). Data are presented as mean ± SEM, and were analyzed by Student’s t-test. White arrows indicate Sirt6 expression in neurons. CSD: Chronic sleep deprivation; DAPI: 4′,6-diamidino-2-phenylindole; PrL: prelimbic cortex; Sirt6: Sirtuins 6.

Sirt6 overexpression in PrL neurons

To generate the Sirt6 overexpression animal model, we delivered adeno-associated virus serotype 9 (AAV2/9) encoding Sirt6-eGFP fusion protein under the control of the CMV promoter to the PrL (Figure 5A). We sacrificed the mice for brain collection 21 days after the virus infection, and assessed virus overexpression and cell tropism. As expected, exogenous Sirt6 was dramatically expressed and restricted in the nuclei (Figure 5B). The normalized transfection efficiency of Sirt6 was significantly increased to 3.46 ± 0.11 in the Sirt6-eGFP group compared with the fluorescence intensity (1.00 ± 0.08) in the control eGFP group (P < 0.01). Within the virally transduced region, the Sirt6 overexpression was mainly observed in neurons (Figure 5C) with high specificity (Figure 5D–F). Co-staining with the microglial marker IBA1 (Zhang and Cui, 2021) or the astrocytic marker GFAP (Yu et al., 2020) showed that AAV9 infrequently underwent glial transduction (Figure 5D–F).

Figure 5.

Sirt6 overexpression in PrL neurons.

(A) Schematic diagram of the viral vector construction used to target the cells in the PrL. The Sirt6 transgene was placed under the control of the CMV promoter (top panel). Schematic of the area injected with AAV2/9-Sirt6 or AAV2/9-eGFP (below panel, left). Bilateral double injection of AAV2/9-Sirt6 or AAV2/9-eGFP in the PrL, verified by eGFP expression (green; right, below panel). (B) Overexpression of the Sirt6 protein according to a 3-fold change in vectors. Sirt6 expression (red, Cy3) in the PrL cells was transduced with empty vectors (left) and with the Sirt6 transgene (right). Data are presented as mean ± SEM, and were analyzed by Student’s t-test. (C) eGFP (green) was co-stained with the neuronal nuclear marker NeuN (red, Cy3). White arrows indicate the body of eGFP⁺ cells. (D, E) The specificity of eGFP in astrocytes and microglia. The astrocytes and microglia were seldom infected with the virus. (F) Analysis of the tropism of AAV9 for different kinds of cells, such as neurons (NeuN), astrocytes (GFAP), and microglia (IBA1). AAV2/9: Adeno-associated virus serotype 9; CMV: cytomegalovirus; DAPI: 4′,6-diamidino-2-phenylindole; eGFP: enhanced green fluorescent protein; GFAP: glial fibrillary acidic protein; IBA1: ionized calcium-binding adapter molecule 1; PrL: prelimbic cortex; Sirt6: Sirtuins 6.

Sirt6 overexpression ameliorates the neurobehavioral deficits induced by CSD

To explore the role of Sirt6 in CSD-induced neurobehavioral deficits, we conducted a series of animal behavioral tests and compared the results between groups. Neither CSD nor Sirt6 overexpression affected motor function in the OFT, as indicated by a comparable total movement distance (Figure 6A). Compared with the control mice, CSD mice stayed in the center zone for a significantly shorter period of time (P < 0.05), indicating that CSD could decrease cognitive ability and induce anxiety-like behaviors. However, Sirt6 overexpression in the PrL reversed the CSD-induced anxiety-like behaviors (P < 0.05; Figure 6A) compared with the CSD mice. Likewise, in the NOR test, CSD reduced the recognition index (P < 0.05) and the times of novel object entry (P < 0.05) compared with the control mice. At the same time, forced Sirt6 expression diminished CSD-induced abnormalities in memory processes (P < 0.05; Figure 6B). CSD also induced spatial memory impairments, as evidenced by a decrease in the novel arm preference index in the Y-maze test compared with control mice (P < 0.05). Furthermore, Sirt6 overexpression prevented the decline in the novel arm preference index (P < 0.05). We found no significant differences in the average speed between the three groups (Figure 6C). When examining social interactions and preferences for social novelty among the groups, we found that the CSD mice spent a similar amount of time interacting with the cup containing Stranger I and the empty wire cup. However, Sirt6-overexpressing CSD mice behaved similarly to control mice in that they tended to explore the cup containing Stranger I in the sociality session (Figure 6D). In the social novelty session, the CSD mice spent a similar amount of time interacting with Strangers I and II, but both the CSD mice overexpressing Sirt6 and the control mice exhibited a stronger preference for Stranger II over Stranger I (P < 0.05; Figure 6D).

Figure 6.

Sirt6 overexpression ameliorates neurobehavioral deficits in chronic sleep deprived mice.

(A) Left: Representative traces for the open field test. Right: Quantitative results of total distance moved and the amount of time spent in the center zone. Chronic sleep deprived mice stayed for less time in the center zone, and Sirt6 overexpression in the PrL reversed CSD-induced anxiety-like behaviors. (B) Left: Representative traces for novel object recognition. Right: Qualification of the recognition index and entries into a novel object in each group. Chronic sleep deprivation reduced the recognition index and number of entries into a novel object compared with the control mice. Forced Sirt6 expression diminished chronic sleep deprivation-induced abnormalities in memory processes. (C) Left: Representative paths for the Y-maze test. Right: Quantitative results of averaged speed and novel arm preference index. Chronic sleep deprivation induced spatial memory impairments, as evidenced by a decrease in the novel arm preference index compared with control mice. Sirt6 overexpression prevented the decline in the novel arm preference index. (D) Left: Representative tracing heatmaps. Right: Quantitative results of the social preference index in the social session and the social novelty preference index in the social novelty session. Chronic sleep deprived mice spent a similar amount of time investigating the cup containing Stranger I and the empty wire cup. However, Sirt6-overexpressing chronic sleep deprived mice behaved similarly to control mice in that they tended to explore the cup containing Stranger I in the sociality session. Data are presented as mean ± SEM, and were analyzed by two-way analysis of variance followed by Tukey’s post hoc test. Red dot: start recording point; blue dot: end recording point; yellow line: the recorded areas; purple line: the movement locus of animals. CSD: Chronic sleep deprivation; eGFP: enhanced green fluorescent protein; F: familiar arm; N: novel arm; PrL: prelimbic cortex; S1 and S2: Stranger SI and II; Sirt6: Sirtuins 6.

Sirt6 overexpression improves CSD-diminished functional connectivity

We explored changes in FC in terms of the relationship between Sirt6 and CSD using resting-state fMRI. To investigate FC variations in the PrL, we analyzed PrL-based FC in the whole brain among the Con + eGFP, CSD + eGFP, and CSD + Sirt6 groups. Then, we compared the FC between the Con + eGFP, CSD + eGFP/CSD + Sirt6, and CSD + eGFP groups to assess the influences of CSD and Sirt6 overexpression. The results are illustrated in Figure 7, and the data for different clusters are shown in Additional Table 2. To investigate the brain regions related to CSD or Sirt6 overexpression, we segmented the different clusters into different brain regions. The results are collected in Table 1. Compared with the Con+eGFP group, the FC in the PrL was significantly decreased in several brain regions after CSD, including the accumbens nucleus (voxel size 2245), piriform cortex (voxel size 2070), motor cortex (voxel size 1551), somatosensory cortex (voxel size 727), olfactory tubercle (voxel size 750), insular cortex (voxel size 628), and cerebellum (voxel size 534). After the overexpression of Sirt6 in the PrL, the FC between the PrL and the above brain regions such as cerebellum (voxel size 2107), piriform cortex (voxel size 1348), insular cortex (voxel size 501), and motor cortex (voxel size 318) was almost reversed (Figure 7 and Table 1). These data indicate that CSD could significantly decrease the FC of the PrL, and that the overexpression of Sirt6 almost reversed these changes.

Figure 7.

Sirt6 overexpression improves functional connectivity damaged by chronic sleep deprivation.

(A) CSD + eGFP versus Con + eGFP groups; (B) CSD + Sirt6 versus CSD + eGFP groups. The colored bar in the image presents the t-values of the comparisons (P < 0.05). 5Cb: 5^th cerebellar lobule; 6Cb: 6^th cerebellar lobule; AI: agranular insular cortex; AO: anterior olfactory nucleus; Crus 1 & 2: crus 1 & 2 of the ansiform lobule; CSD: chronic sleep deprivation; eGFP: enhanced green fluorescent protein; FrA: frontal cortex; M1: primary motor cortex; PrL: prelimbic cortex; S1: primary somatosensory cortex 1; Sirt6: Sirtuins 6.

Table 1.

Comparisons of different brain regions between Con+eGFP and CSD+eGFP/CSD+Sirt6 and CSD+eGFP using functional magnetic resonance imaging

Brain regions	Sirt6>Csd		Csd>Sirt6		Con>Csd		Csd>Con
	Voxel size	t-value	Voxel size	t-value	Voxel size	t-value	Voxel size	t-value
Accumbens nucleus	188	3.349			2245	5.342
Amygdala	102	3.411	63	2.091	76	2.388	164	2.332
Anterior olfactory nucleus	642	4.617			306	4.429	361	2.949
Caudate putamen	143	2.833			696	4.357
Cerebellum	2107	5.358			534	3.11	132	2.948
Cingulate cortex	14	1.943			62	3.142
Dentate gyrus	57	2.262
Ectorhinal cortex			10	2.46
Entorhinal cortex			693	4.598
Hippocampus	21	2.238	27	2.124
Hypothalamus			308	3.083
Insular cortex	501	3.067			628	3.611	46	2.342
Primary motor cortex	45	2.6			287	3.295
Secondary motor cortex	47	2.854			286	3.108
Motor cortex	318	3.356			1551	3.611
Olfactory bulb	668	4.828			41	2.429	312	2.917
Olfactory tubercle	67	4.076			750	5.115
Orbital cortex	980	3.291			43	3.401
Piriform cortex	1348	5.115	7	1.923	2070	5.512	179	2.535
Prelimbic cortex	221	2.383			187	3.544
Preoptic nucleus	17	2.813
Retrosplenial granular cortex	160	4.441
Septal	8	1.939			361	3.563
Somatosensory cortex	45	2.387			727	3.823	159	3.436
Subiculum			85	3.021
Tdalamus			22	2.486
Visual cortex	264	3.819					47	2.912

Con: Con + eGFP group; Csd: CSD + eGFP group; Sirt6: CSD + Sirt6 group. CSD: Chronic sleep deprivation; eGFP: enhanced green fluorescent protein; Sirt6: Sirtuins 6.

Sirt6 overexpression recovers metabolic kinetics in the neurons and astrocytes of CSD mice

In the current study, we evaluated the metabolic kinetics of both neurons and astrocytes using forms of ¹³C labeled Glc₁ and Ace₂. Thirty minutes after the isotope probe infusion, the mice were euthanized and FC was assessed to analyze the metabolic kinetics. As demonstrated in Figure 2, Ace₂ could only be taken up and metabolized in astrocytes, while Glc₁ could be metabolized in both neurons and astrocytes. With the infusion of Glc₁ or Ace₂, Gln₄, Glu₄, and GABA₄ were labeled with ¹³C through the first round of the tricarboxylic acid cycle in both astrocytes and neurons (glutamatergic and GABAergic neurons). Then, Gln (Gln₂ and Gln₃), Glu (Glu₂ and Glu₃), and GABA (GABA₂ and GABA₃) were isotopically labeled at different positions in the following round of the tricarboxylic acid cycle. Because the NMR signals of Gln₂, Glu₂/Gln₃, and Glu₃ always overlapped, Glx₂ and Glx₃ were adopted to represent the overlapping signals, respectively. Analysis of the metabolic kinetics of these two ¹³C probes provided a valuable window for assessing the metabolism of neurons and astrocytes, given that Gln and Glu/GABA are mainly generated in astrocytes and neurons, respectively.

The representative ¹H-NMR spectra for the metabolic kinetics of Glc₁ and Ace₂ are illustrated in Figure 8A. The ¹³C enrichment in most metabolites of Glc₁ is higher than that of Ace₂, except for glutamine (Gln₄). The relevant ¹³C enrichments in major metabolites are collected in Figure 8B and C. For the analysis of metabolic kinetics of Ace₂, we found that relevant ¹³C enrichments of Glu (Glu₃, Glu₄, and Glx₃) and GABA (GABA₂, GABA₄) were significantly decreased (P < 0.05) after CSD compared with the control group, while Sirt6 overexpression in the PrL partially reversed CSD-induced declines in Glu₃ and Glx₃ (P < 0.05). The ¹³C-labelled enrichment of Gln₄ did not change significantly among these three groups. In terms of the metabolic kinetics of Glc₁, the ¹³C-labelled enrichments of Glu₄, Gln₄, GABA₂, Glx₃, and GABA₄ significantly fell (P < 0.05) after CSD compared with the control group. Remarkably, most of these metabolites were recovered after the overexpression of Sirt6, especially for Glu₄ and GABA₂ (P < 0.05).

Figure 8.

Sirt6 overexpression recovers metabolic kinetics in neurons and astrocytes of chronic sleep deprived mice.

(A) Representative NMR spectra for functional connectivity according to [1-¹³C] glucose (left) and [2-¹³C] acetate (right). Black line: NMR signals for all metabolites (¹²C + ¹³C); red line: 2-¹³C-labeled metabolites (2-¹³C). (B, C) Comparisons of ¹³C enrichment for [1-¹³C] glucose (B) or [2-¹³C] acetate (C) in various positions of different metabolites in functional connectivity. (D, E) The mRNA expression levels for Glu₁ and Abat for the PrL region in the three different groups. Data are presented as mean ± SEM, and were analyzed by a two-way analysis of variance followed by Tukey’s post hoc test. Abat: GABA aminotransferase; Ala: alanine; ASP: aspartic acid; CSD: chronic sleep deprivation; eGFP: enhanced green fluorescent protein; GABA: γ-aminobutyric acid; Gln: glutamine; Glu: glutamate synthase; Glx: glutamine + glutamate; Lac: lactate; NAA: N-acetylaspartate; Sirt6: Sirtuins 6.

Sirt6 overexpression restores glutamate and GABA-related gene expression in the PrL of CSD mice

We quantified the gene expressions of Glu₁ and Abat using quantitative polymerase chain reactions for PrL tissue to monitor the activities of glutaminase and GABA transaminase among the Con + eGFP, CSD + eGFP, and CSD + Sirt6 groups. The results indicated that CSD significantly reduced the mRNA levels of Glu₁ and Abat (both P < 0.05). However, these changes were significantly reversed after the overexpression of Sirt6 in the PrL region (P < 0.05; Figure 8D and E), compared with the animals in the CSD + eGFP group. Thus, these data indicate that the activities of glutaminase and GABA transaminase were almost consistent with the metabolic kinetics data.

Sirt6 overexpression ameliorates synaptic dysfunction in CSD mice

Previous evidence has suggested that impairment of the PrL glutamatergic system is strongly correlated with the progression of cognitive decline (Kashani et al., 2008). Meanwhile, CSD was demonstrated to decrease the number of synapses in pyramidal neurons (Bellesi et al., 2017), indicating that it might also suppress synaptic transmission. Vesicular glutamate transporter VGLUT1 is a specific marker for the presynaptic terminals of pyramidal neurons (Fujiyama et al., 2001). In the current study, we observed a decrease in glutamate in the prefrontal cortex of CSD mice. Sirt6 overexpression might have reversed the CSD-induced decrease in the number of presynapses.

To test our hypothesis, we stained sections containing PrL regions with VGLUT1 antibody. As shown in Figure 9A, VGLUT1 appeared as red fluorescent puncta in brain sections (Figure 9A), and the puncta co-labeled with eGFP were considered to be VGULT1 from AAV-infected neurons. Compared with the control group, CSD significantly reduced (P < 0.05) the number of VGLUT1-positive puncta overall and that in AAV infected neurons. This reduction was diminished in neurons overexpressing Sirt6, as indicated by a significant increase in eGFP-VGLUT1 double-positive puncta in the CSD + Sirt6 group compared with the CSD + eGFP group (P < 0.05; Figure 9A).

Figure 9.

Sirt6 overexpression ameliorates synaptic dysfunction in chronic sleep deprived mice.

(A) Left: Representative images of co-stained eGFP (green) and VGULT1 (Cy3, red) in PrL neurons among the three groups. The presynapses (VGULT1, a pre-synaptic protein) might have been diminished by chronic sleep deprivation, and then recovered by Sirt6 overexpression. Right: Summarized data for eGFP⁺ VGULT1⁺ numbers per 100 µm² (n = 3/group). (B) Left: Illustrations of the virus injection. Middle: Remodeling of randomly selected neurons co-stained with eGFP and RFP, which were further subjected to dendritic spine density analysis. Representative images of neuronal dendrites by sparse labeling among the three groups. Chronic sleep deprivation reduced the number of spines, and overexpression of Sirt6 reversed this reduction. Right: Summarized data for spine numbers per 10 µm. (C) Representative traces (left) and summarized data regarding mEPSCs amplitude (middle) and frequency (right) from eGFP⁺ neurons in the PrL among the three groups (3–4 cells/per mouse). (D) Left: Trains of APs evoked by depolarizing current steps of 100, 200, and 300 pA. Right: Mean frequency of AP firing at each depolarizing current step for eGFP⁺ neurons among the three groups (3–4 cells/per mouse). Data are presented as mean ± SEM, and were analyzed by a two-way analysis of variance followed by Tukey’s post hoc test. AP: Action potential; CSD: chronic sleep deprivation; DAPI: 4′,6-diamidino-2-phenylindole; eGFP: enhanced green fluorescent protein; mEPSC: miniature excitatory postsynaptic current; RFP: red fluorescent protein; Sirt6: Sirtuins 6; VGLUT1: vesicular glutamate transporter 1.

Next, we examined the dendritic spine densities in the eGFP-expressing neurons using the sparse-labeling method. The results showed that the CSD mice had a significantly decreased number of dendritic spines compared with the control group (P < 0.05). Meanwhile, mice overexpressing Sirt6 had significantly higher spine densities than CSD mice (P < 0.05; Figure 9B).

In the patch-clamp experiment, eGFP⁺ neurons in CSD mice showed a significant reduction in the frequency and amplitude of mEPSCs (P < 0.05) compared with those in the control group. Sirt6 overexpression was sufficient to prevent falls in the frequency and amplitude of mEPSCs after CSD (P < 0.05, Figure 9C). Furthermore, we recorded trains of action potentials evoked by applying depolarizing currents at 50 mV increments to assess the firing rate of the eGFP⁺ neurons in the three groups. Compared with the control animals, the total numbers of action potentials generated by the stimulation currents (150, 200, 250, and 300 pA) were significantly lower (P < 0.05) in the neurons from CSD mice, as shown in the f-I plot in Figure 9D. Consistently, Sirt6 overexpression could reverse the CSD-induced decrease in action potential firing rates in eGFP⁺ neurons (Figure 9D).

Discussion

There is considerable evidence that sleep plays a vital role in memory and cognitive processes. The present work corroborates the previous findings that SD can cause cognitive impairment in animal models. We found that CSD significantly reduced Sirt6 levels in the PrL, a region strongly associated with cognitive functions. Furthermore, we found that Sirt6 overexpression was sufficient to reverse impaired cognitive behavior, and that this was accompanied by the strengthening of FC networks after CSD, including the FC of the PrL and accumbens nucleus, piriform cortex, motor cortex, somatosensory cortex, olfactory tubercle, insular cortex, and cerebellum. We also demonstrated that Sirt6 effectively enhanced glutamate-glutamine cycling and glutamatergic neurotransmission. Our data suggest that restoring glutamate metabolic kinetics in the prefrontal cortex through Sirt6 activation is a promising approach for synaptic homeostasis and maintenance of FC in brain networks. Thus, this study has important implications for the emerging therapeutic potential of Sirt6 modulators for cognitive enhancement after CSD.

With its high transport capacity, acetate is almost exclusively metabolized in astrocytes (Waniewski and Martin, 1998). After uptake by astrocytes, acetate isotopomers (Ace₂) created ¹³C-labelled glutamate at the fourth carbon to produce Glu₄, and then quickly converted to Gln₄ along the tricarboxylic acid cycle and α-ketoglutarate-glutamate-glutamine conversion route. This was followed by the generation of Glu₄/GABA₂ (first turn) and Glu₃/GABA₃ (further turn) in neurons via the glutamine-glutamate/GABA shuttle between neurons and astrocytes (de Graaf et al., 2003). In contrast, ¹³C-labeled glucose can be taken up by both astrocytes and neurons, in which glutamate, glutamine, and GABA are successively tagged with isotopes through tricarboxylic acid cycles, glutamate-glutamine interconversion, and the glutamine-glutamate/GABA shuttle (de Graaf et al., 2003). In this regard, Ace₂ can be used to reflect metabolic kinetics in astrocytes under different brain states, while Glc₁ is related to the overall metabolic kinetics in the specific brain region tested. In this study, Ace₂ infusion did not result in different relative Gln₄ levels between groups, indicating that CSD did not affect metabolic kinetics in astrocytes. Combined with the Ace₂ analysis, the observed reduction in Gln₄ levels after Glc₁ infusion suggests that the neuronal metabolism or glia-neuron shuttle was impaired after CSD. Using both tracers, we found that ¹³C-labeled glutamine was reduced by CSD, while Sirt6 overexpression was sufficient to replenish glutamine synthesis via the upregulation of the Glu₁ gene. Note that Sirt6 overexpression was mainly observed in neurons (98% of the eGFP cells were positive for NeuN). Therefore, we conclude that CSD can interfere with glutamate synthesis and neuronal Sirt6 function to compensate for such defects.

Previous studies have highlighted the role of Sirt6 as a critical modulator of energy metabolism given its role in gluconeogenesis and glycolysis (Yuan et al., 2022). By deacetylating H3K9 on hypoxia-inducible factor 1-alpha target genes, Sirt6 downregulates pyruvate dehydrogenase kinase 4 expression to maintain the catalytic activity of the mitochondrial pyruvate dehydrogenase complex, favoring the tricarboxylic acid cycle in mitochondria. In contrast, the loss of Sirt6 shifts mitochondrial respiration to aerobic glycolysis. A whole-brain measurement of resting oxygen and glucose metabolism suggested that the energy metabolism pathway varies in a region-dependent fashion, with the prefrontal cortex as one of the regions that prefers aerobic glycolysis (Vaishnavi et al., 2010). The decrease in Sirt6 implies that PrL neurons might diminish Sirt6 expression to satisfy the high demand for aerobic glycolysis. Unfortunately, dysregulated Sirt6 is associated with insufficient glutamate levels, which could then impair the glutamatergic neurotransmission required for cognitive function. However, our study of metabolic kinetics suggested that Sirt6 activation after CSD might promote recovery from glutamate insufficiency and thus improve cognitive function.

Pyramidal neurons comprise the main neuronal population in the PrL that sends projections to other brain regions, such as the nucleus accumbens (Green et al., 2020), dorsal striatum (Green et al., 2020), basolateral nucleus of the amygdala (Cho et al., 2013), and periaqueductal gray (Cheriyan and Sheets, 2018). Altered excitability of projecting neurons and altered functional connectivity between two brain regions are often associated with diseases or symptoms. For example, the excitability of periaqueductal gray-projecting pyramidal neurons in the PrL is significantly reduced in an animal model of chronic pain (Cheriyan and Sheets, 2018). In line with this, FC between the PrL and periaqueductal gray was negatively correlated with pain intensity in patients with chronic low back pain (Yu et al., 2014). These studies suggest that a reduced neuronal firing could be the basis for disrupted FC. In the present study, we found that CSD significantly decreased the FC between the PrL and several brain regions. The number of action potentials triggered by a range of currents was significantly reduced in PrL pyramidal neurons after CSD, representing a potential cellular mechanism of CSD-induced FC disruption. Furthermore, this decreased firing rate was accompanied by a significantly reduced mEPSC frequency, indicating a presynaptic mechanism. The probability of glutamate release and the regulation of glutamate loading in synaptic vesicles can profoundly affect the mEPSC frequency (Zhou et al., 2000; Sharma and Vijayaraghavan, 2003). Therefore, the decreased mEPSC frequency after CSD might be attributable to reduced glutamate concentration. In addition, we found that CSD significantly reduced the number of VGLUT1 puncta. It is well known that VGLUT1 determines the refilling of synaptic vesicles with glutamate, and thereafter, the quantal size of mEPSCs (Wilson et al., 2005; Nakakubo et al., 2020). Therefore, a reduction in VGLUT1 puncta could also have caused mEPSC suppression. We propose that decreased glutamate synthesis and repressed VGLUT1 expression could serve as two parallel mechanisms underlying the CSD-induced suppression of presynaptic glutamate release and mEPSCs. Alternatively, decreased glutamate synthesis could trigger a feedback effect that reduces VGLUT1 expression, hence the diminished VGLUT1 puncta. Finally, we found that Sirt6 overexpression completely restored CSD-induced changes in mEPSCs and action potentials. Sirt6 promoted glutamate synthesis via the facilitation of mitochondria respiration or the epigenetic activation of VGLUT1 via an unknown mechanism that could drive these therapeutic effects. Our study highlights Sirt6 activation as a strategy to restore PrL activity and thereafter improve cognitive function.

However, there were some limitations to the current study. First, to reduce the number of animals used, we examined FC on the 16^th day of the experiment and then used the same animals to measure the metabolic kinetics on day 17. We tested animal behavior from day 16 to day 20. Given that the cognitive function and FC data were obtained from different animals, it was not possible to conduct a correlation analysis, which is one method for addressing the present hypothesis. Second, the current findings may only be applicable to male animals because we only used male animals in the study. Third, we believe that the FDR or FEW correction tools in SPM are the best choice for conducting multiple comparisons in fMRI studies. However, these are somewhat conservative and could cause a negative result. To avoid that outcome, we used GRF correction in the current study. However, this approach was too lenient to control the false positive results, and we set the cluster-extern threshold of contiguous voxels to 100 to reduce the false positive rate. Finally, the FC values for the PrL region were varied in several brain regions for the CSD+eGFP and CSD+Sirt6 groups. Given that we did not deeply investigate these variations in the current study, these findings require further verification using other technologies, such as electrophysiology. Despite these limitations, we believe that our findings represent a valuable foundation for future sleep-related research.

The current work reports that Sirt6 ameliorates CSD-associated impairments in PrL FC and cognitive behavior by regulating glutamatergic synaptic homeostasis. Thus, small-molecule Sirt6 agonists might be a viable strategy to restore cognitive function in people with chronic sleep problems.

Additional files:

Additional Table 1: Different brain states in a day-night cycle for normal and SD mice.

Additional Table 2: Information of different clusters during fMRI comparisons.

Additional Figure 1 ^{(362.6KB, tif)} : Scheme of the semi-automatic method for spine counting.

Additional Figure 1

Scheme of the semi-automatic method for spine counting.

(A) An example dendrite fragment. (B) The photograph was converted into an 8-bit image. (C) Skeleton calculation. (D) Analysis of the terminal endings of the skeleton. End-point voxels were displayed in blue, slab voxels in orange and junction voxels in purple. Black arrow indicates the processing steps from left to right; white arrow indicates the spine.

Additional Figure 2 ^{(386KB, tif)} : Influences of the sleep stages by CSD during the 4-hour home cages period (10:00–14:00).

Additional Figure 2

Influences of the sleep stages by CSD during the 4-hour home cages period (10:00–14:00).

(A) Represented signals for EEG, EMG and sleeping states during the whole SD period for the animals in Con or SD groups. (B) Statistical analysis of the influences of the sleep stages by CSD. Data are presented as mean ± SEM, and were analyzed by Student's t-test. Con: Control; CSD: chronic sleep deprivation; EEG: electroencephalogram; NREM: non-rapid eye movement; REM: rapid eye movement; SD: sleep deprivation.