Sustained microglial activation and accelerated elimination of dendritic spines during acute sleep deprivation and restoration

Sustained microglial activation and accelerated elimination of dendritic spines during acute sleep deprivation and restoration. The Innovation

■ A substantial number of metabolites and pathways fail to recover after recovery sleep.

INTRODUCTION
Sleep is a highly conserved behavior encountered throughout the animal kingdom.Almost all known species with nervous systems exhibit sleep or sleeplike states which are essential for maintaining optimal physical and mental health. 1 In humans, due to various sources of stress imposed by modern society, decreasing sleep duration and sleep quality has become common place and contributes to a host of health issues.Generally, sleep debt is closely associated with metabolic disorders,2-4 cardiovascular disease 5 as well as neurological and cognitive impairment. 6Despite extensive studies over the past few decades, the mechanisms underlying sleep need and the impact of sleep insufficiency have not been well characterized.The brain is one of the organs most affected by lack of sleep, which manifests in cognitive impairment, 7 altered mood states, 8 and the aggravation of psychiatric and neurodegenerative disorders. 9,106][17] However, the underlying pathophysiological mechanisms responsible for the lack of sleep-induced dysfunction of the central nervous system (CNS) remain largely unclear.
2][13][14] However, there is also increasing evidence for critical roles being played by other CNS cell types, notably glial cells.Microglia are the resident immune cells distributed throughout the CNS.Over the past decades, microglia have been increasingly implicated in the regulation of development, function and homeostasis of the CNS. 18,19Especially due to the intimate inter-actions between neurons and microglia 20,21 and the modulated neuroimmune responses induced by sleep insufficiency, [22][23][24][25] microglia have attracted great attention in the area of sleep biology.For example, sleep disturbance results in increased levels of proinflammatory cytokines 22,26,27 and significant morphological and functional changes of microglia. 23,28Furthermore, inhibition of microglial activity prevents sleep pressure build-up, as well as spatial memory defects following sleep deprivation. 24,25Thus, microglia seem to respond to sleep deprivation by changing their state/activity and thereby confer adverse effects on CNS homeostasis.However, various studies which have employed different protocols for sleep manipulation have led to diverse results and conclusions.Therefore, the precise roles of microglia during sleep insufficiency and in particular the recovery sleep after sleep deprivation remain poorly understood.In addition to their well-characterized neuroimmune function, microglia also participate in the modulation of synaptic pruning and refinement, 29 which represent distinctive features of neurodevelopment and are critical for the establishment of neural circuits across brain regions. 30,31Recently, there have been intense controversies over whether microglia-neurons exhibit discrepant behaviors during periods of wakefulness and sleep, or in more extreme cases such as sleep deprivation.For example, microglia seemed to be more "vigilant" in the sleeping brain, surveilling the ambient microenvironment and playing roles in neuronal plasticity, 32,33 while at the synaptic level, microglial processes were observed to be attracted towards active spines during wakefulness, and this interaction was impeded during sleep. 34Interestingly, microglia were demonstrated to be overactive during sleep debt thus contributing to cognitive deterioration. 35evertheless, experimental paradigms varied between these studies and so consistent results remain scarce.Therefore, evidence for the degree to which microglia perform differential roles during wakefulness, sleep debt and restoration remains inconclusive.
"Omics" tools have been applied to investigate the effects and mechanisms of sleep on brain function and activity, revealing that pathways associated with cellular stress responses, energy metabolism, neuronal transmission and synaptic plasticity can all be affected by sleep insufficiency. 36From a metabolic perspective, the brain undergoes cycles of metabolic activity over the course of the sleep-wake cycle.Sleep appears to be required to generate resources as well as clear waste that accumulates during wakefulness, which to some extent support the notion that sleep fulfills a reparatory function. 37,38herefore, metabolic profiling of brain tissues provides an opportunity to characterize the microenvironment and phenotypes associated with sleep, or the state of lack of sleep.
In this study, we have established an acute sleep deprivation (ASD) model in mice based on a "gentle handling" apparatus, and have investigated the impacts of ASD and subsequent recovery sleep (RS) on the phenotypic characteristics and physiological function of microglia.Specifically, we focused on morphological and quantitative changes of microglia, the motility of microglial processes as well as their phagocytic activities in different brain regions.In addition, the synaptic plasticity of neurons in the cerebral cortex was quantified at different stages.Subsequently, the LC-MS/MS based metabolic profiling of brain tissues was conducted to identify potential mechanisms for the observed persistent microglia activation and accelerated elimination of dendritic spines during ASD and RS.The results obtained in this work may thereby help to elucidate the cellular and molecular mechanisms of how irregular sleep schedules lead to neurological disorders.

MATERIALS AND METHODS Ethics statement and animal maintenance
Animal care and experimental procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Ethics Committee of School of Life Sciences, Lanzhou University (Approval No: EAF2020007).All efforts were made to minimize animal suffering during this work.C57BL/6 mice, transgenic mice (Thy1-YFP line H, JAX #003782) expressing yellow fluorescent proteins (YFP) in the layer 5 pyramidal neurons, and CX3CR1 GFP/+ mice expressing green fluorescent proteins (GFP) in the microglia (male, 8 weeks of age, weighing 22-24 g) were used in this study.All mice were housed under conditions of 22 °C, 40-70% humidity, and 12h:12h light / dark cycles, accessing food and water ad libitum.

Acute sleep deprivation / Recovery sleep model and experimental procedures
Acute sleep deprivation (ASD) of mice in this study is achieved by using the automated sleep deprivation system (SANS -SA109), a "gentle handling" apparatus which is capable of sleep depriving mice while minimizing their exercise.Briefly, mice were placed into autoclaved corncob bedding chambers where they had ad libitum access to food and water.A rotating bar was placed 1 cm above the floor of each chamber, lightly nudging the mice from sleep and encouraging low levels of activity until they maintain awake on their own.To prevent sleep acclimation and habituation, the bar was programmed to rotated at a speed of 1-5 RPM and the direction changed randomly (Figure 1A).0][41] Open field tests were performed afterwards to test the efficiency of ASD in this study (Figure S1).Schematic experimen-tal design in this study is shown in Figure 1B.Briefly, mice were randomly divided into the following 4 groups: normal control group (CTRL); ASD for 24 hours (ASD24h); ASD for 48 hours (ASD48h) and Recovery Sleep for 24 hours (RS24h), i.e., 24 hours recovery sleep after the 48 hours of ASD.Mice in each group were placed in the sleep deprivation chamber one week prior to the start of the experiment to allow for acclimation.Mice in ASD24h, ASD48h and RS24h groups were kept awake in the system for consecutive 24 or 48 hours.CTRL and RS24h mice in the RS stage were placed in a same static chamber copy but without the rotating bar.Animals in each group were imaged or sacrificed at the indicated timepoints.

Metabolomic analyses
See Supplemental Text 1 for details.

Preparation of brain sections and immunofluorescence (IF) labelling
See Supplemental Text 2 for details.

In vivo two-photon imaging and data processing
See Supplemental Text 3 for details.

Microglial skeletonization and 3D-reconstruction
See Supplemental Text 4 for details.

Statistical analyses
See Supplemental Text 5 for details.

Sustained microglial activation during ASD and subsequent RS
In the healthy brain, microglia are typically in a physiological state in which their motile multiple-branched processes continuously scan the surrounding microenvironment in the brain parenchyma for diverse extracellular signals and interact with other cells including neurons and astrocytes. 42,43However, this homeostatic state can be perturbed by external stimulation and pathophysiologic conditions, exhibiting a distinct state of microglial "activation".In order to graphically document the dynamic morphological changes of microglia in different brain regions during acute sleep deprivation (ASD) and subsequent recovery sleep (RS), we obtained brain sections of CX3CR1 GFP/+ mice in normal control group (CTRL), mice subjected to 48 hours of acute sleep deprivation (ASD48h) and those that had just experienced 48 hours of sleep deprivation and subsequent 24 hours of recovery sleep (RS24h), and then imaged microglia in the cortex, hippocampal CA1, polymorph layer of the dentate gyrus (PoDG) as well as striatum regions of the brain by means of two-photon imaging.Interestingly, in contrast with much stronger responses observed in many pathological states, in which they change their morphology dramatically from a multi-branched to an amoeboid state, 44,45 we observed slight morphological changes of microglia after 48 hours of ASD, as well as the subsequent 24 hours of RS.In order to better visualize these morphological changes of microglia, both the ImageJ skeletonization and Imaris 3D-reconstruction approaches were applied to these acquired microglial images (Figure 2A), and a significant and persistent reduction of microglial surface, length of processes, and number of microglial process endpoints was identified in the ASD48h and RS24h groups, based on the statistical analysis of these morphological parameters (Figure 2B-D).However, no significant change in microglial cell density was observed throughout the whole process (Figure 2E).Thus, our results suggest a sustained, low level of microglial activation during ASD and subsequent shortterm RS.

Sustained reduction of microglial process motility during ASD and RS
Acting as surveilling immune cells in the CNS, microglia dynamically respond to changes in the microenvironment with their remarkably motile processes. 32,34,46Despite the importance of microglial process dynamics, currently how these processes behave under diverse physiological conditions remains poorly understood.Therefore, by using in vivo two-photon imaging in CX3CR1 GFP/+ transgenic mice, we investigated microglial process dynamics in the somatosensory cortex during ASD and subsequent RS.Briefly, after time-lapse imaging of microglia upon each treatment for a total duration of 30 min, we quantified changes in process motility and stability by comparing Z-projections time points via a custom algorithm in ImageJ software (as illustrated in Figure 3A).Interestingly, our results demonstrated a sustained increase of the long-term stability index and inversely a reduction of instability index in microglia of mice subjected to 48 hours of ASD (ASD48h) and subsequent 24 hours of RS (RS24h) (Figure 3B-C).More specifically, the basal microglial process motility, retraction and extension levels of processes in every 5-min time interval were compared and calculated cumulatively (Figure 3D and Figure S2), and the results showed that all these parameters including retraction and extension index as well as motility index were significantly and continuously decreased during ASD and subsequent RS as compared to the normal control conditions (Figure 3E-G).In general, these data indicate a sustained reduction in microglial process motility during ASD and subsequent RS.

Sustained enhancement of microglial phagocytosis during ASD and RS
As the resident macrophages and synaptic trimmers of the CNS, microglia contribute to immune defense as well as pruning and remodeling of synapses largely based on their phagocytosis functionality. 29,47To assess the phagocytic activity of microglia during ASD and subsequent RS, we traced the expression levels and localization of CD68, a lysosome-associated membrane protein engaged in the phagocytosis and clearance of cellular debris and dead cells by monocytes/macrophages. 48,49According to our immunofluorescence (IF) staining of CD68 and its colocalization with the CX3CR1-GFP labeled microglia (Figure 4A), in sharp contrast to the normal control group both the expression of CD68 and their M1 Mander´s colocalization coefficient were continuously increased in different brain regions of mice subjected to 48 hours of ASD (ASD48h) and subsequent 24 hours of RS (RS24h) (Figure 4A-B).In addition, the quantification of CD68 content as a L f percentage of cell volume within individual microglia reconstructions verified that the CD68 content % was indeed significantly increased throughout this entire ASD/RS process (Figure 4C).Thus, these data point to a sustained enhancement of microglial phagocytosis across brain regions during both ASD and subsequent RS.

Facilitated elimination of dendritic spines during ASD and RS
Previous studies have demonstrated that microglia activity can influence synaptic plasticity.For example, microglial depletion increases both inhibitory and excitatory synapses 50 and microglia have been observed to remove dystrophic synapses in disease models 51,52 as well as altering neurocircuitry during neurodevelopment in an activity-dependent manner. 53,54In order to investigate the effects of ASD and subsequent RS on synaptic plasticity, we continuously assessed the formation and elimination of dendritic spines in the cerebral somatosensory cortex of mice subjected to the entire ASD/RS processes and compared with those under normal control conditions (Figure 5A-B).Interestingly, by dynamic tracing of the dendrite structure in the same neurons, our live imaging data and the subsequent statistics revealed that there was no significant difference in spine formation rates between sleep deprivation-recovery groups and the normal control group (Figure 5C-E; Pre to 48 h, ASD/RS 1.25 ± 0.4% versus CTRL 1.18 ± 0.2%; Pre to 72 h, ASD/RS 1.59 ± 0.52% versus CTRL 2.56 ± 0.1%; 48 h to 72 h, ASD/RS 0.92 ± 0.42% versus CTRL 1.3 ± 0.33%).However, the spine elimination rates of the cortex in mice subjected to sleep deprivation-recovery were significantly increased compared with those in the normal control mice over different periods (Figure 5C,D,F; Pre to 48 h, ASD/RS 10.79 ± 0.5% versus CTRL 1.86 ± 0.28%, p < 0.001; Pre to 72 h, ASD/RS 15 ± 0.6% versus CTRL 3.19 ± 0.1%, p < 0.001; 48 h to 72 h, ASD/RS 5.58 ± 1.4% versus CTRL 1.76 ± 0.21%, p = 0.003).Notably, the observed elimination rates were significantly higher than the formation rates in mice which were subjected to ASD and RS of different durations, suggesting a sustained, facilitated loss of dendritic spines after sleep insufficiency.Thus, most likely due to the persistent low level of microglia activation, we observed excessive spine loss during ASD and subsequent RS.

The brain metabolome is affected by ASD and subsequent RS
The above neuroimaging data have suggested that microglial activation and neuronal plasticity can be persistently affected during ASD and subsequent RS.To further explore the underlying molecular mechanisms, an untargeted LC-MS/MS approach was applied to evaluate their effects on the whole brain metabolic microenvironment.A total of 2749 metabolites with known identities were detected, among which 2011 were included in further analyses.These identified brain metabolites were mainly classified as lipids and lipidlike molecules (28.1%), organic acids (16%), glycerophospholipids (14.7%) and fatty acyls (9.6%).The heatmap indicated that the levels of a considerable number of metabolites were upregulated or downregulated following ASD and subsequent RS (Figure 6A-B and Table S2; n = 6 for each group).To estimate the divergency of different treatment groups, a partial least squares discriminant analysis (PLS-DA) was conducted, the normal CTRL group and RS24h group were well separated from the ASD groups (R2Y = 0.874, Q2Y = 0.44) (Figure 6C), indicating that our ASD and RS treatments significantly altered the whole brain metabolome in mice.Based on the accuracy and reliability of our model, we next investigated the metabolites in the brain differentially affected in response to ASD and subsequent RS.Our results showed that in comparison with the normal control group, the levels of 358 metabolites were significantly altered after 24 hours of ASD, among them, 122 were up-regulated and 236 were down-regulated.With the extension of ASD to 48 hours, more metabolites (643) changed in their abundance, with 187 being up-regulated and 456 down-regulated.Interestingly, the 24 hours of RS after ASD restored the levels of many metabolites, with only 65 up-regulated and 69 down-regu-  S3).Next, a "Multivariate ROC curve based exploratory analysis" on MetaboAnalyst was conducted to identify biomarkers of acute sleep loss in brain tissues, and the results indicated that metabolites including Glucosylsphingosine, 4a-Hydroxytetrahydrobiopterin, 3,4-Dimethyl-5-pentyl-2-furanheptanoic acid, S-3-oxodecanoyl cysteamine may serve as potential metabolic biomarkers for sleep insufficiency in further studies (Figure 6F).Further, functional and pathway enrichment analyses were carried out to annotate and categorize the biological function and metabolic pathways affected by ASD, and the results showed that the top 25 enriched metabolite sets included phosphonate and phosphinate metabolism, glycerophospholipid metabolism, tyrosine metabolism, steroid hormone biosynthesis, lysine degradation, taurine and hypotaurine metabolism (Figure 6G), while the most significantly affected metabolic pathways in the brain were defined as glycerophospholipid metabolism, nicotinate and nicotinamide metabolism, taurine and hypotaurine metabolism, alanine, asparate and glutamate metabolism and fatty acid elongation et al (Figure 6H).Thus, our results reveal extensive metabolic changes in brain during ASD and restorative effects of RS on metabolic levels.

Partial metabolic pathways may not be able to restore within short-term RS
According to our metabolome data, most of the metabolites (411) in the whole brain restored to normal levels within 24 hours of RS after the period of ASD (Figure 7A and Table S4).However, there were still many metabolites (211) which remained up-or down-regulated even after this recovery sleep period (Figure 7A and Table S5).In order to investigate the mechanisms underlying this observed sustained microglial activation during RS, we focused on those brain metabolites for which the abundances did not recover.According to the pathway enrichment analyses, many metabolites involved in the metabolism of glycerophospholipid, taurine and hypotaurine, nicotinate and nicotinamide, D-glutamine and glutamate were restored to normal levels after 24 hours of recovery sleep (Figure 7B).Instead, metabolic pathways including glycerophospholipid, phosphonate and phosphinate metabolism were significantly enriched in the metabolites classified as unrecoverable (Figure 7C).Next, these recoverable and unrecoverable metabolites were imported into Cytoscape for visualization of their interconnections based on the KEGG pathway database (Figure 7D and 7E).The results revealed that 11 core metabolites were involved in major pathways including lipid metabolism (glycerophospholipid metabolism and fatty acid metabolism), amino acid metabolism (alanine, aspartate and glutamate metabolism), oxidative stress (oxidative phosphorylation and FoxO signaling pathway), immune response (NF-kappa B signaling pathway), cellular energy homeostasis (glycolysis/gluconeogenesis, AMPK signaling pathway and purine metabolism) and neurological function (Neurotrophin signaling pathway and GABAergic synapse) in the brains subjected to ASD and RS.Among them, five core metabolites including ADP, Acetyl-CoA, L-Tryptophan, L-Proline and Citicoline were identified as being recovered, while six core metabolites including DG (16:0/16:0/0:0), L-Serine, L-Glutamine, D-Glucose, LysoPC(20:1(11Z)) and 1-Stearoylglycerophosphoinositol were identified to be unrecovered.Focusing on these unrecoverable metabolites which have been closely linked with neurological functions and diseases, L-Serine, L-Glutamine, D-Glucose, LysoPC(20:1(11Z)) and 1-Stearoylglycerophosphoinositol were down-regulated, whereas DG (16:0/16:0/0:0) and 8-Hydroxyguanine were significantly upregulated during ASD, and all of them were not be able to recover within a  short period of RS (Figure 7F-M).Thus, our results have identified recoverable and unrecoverable metabolites from a short period of RS, and thereby have implicated key metabolites and pathways which might be closely associated with the sustained microglial activation and accelerated elimination of dendritic spines during ASD and subsequent RS.

Microglial responses during sleep deprivation and restoration
Regular and ample sleep is essential for protecting against inflammatory insults and preserving the integrity of immunological function. 55,56Increasing evidence has suggested that insufficient sleep promotes immunological .Metabolites and pathways identified as recoverable/unrecoverable within 24 hours of RS (A) Heatmaps show the recoverable/unrecoverable metabolites in the comparison between CTRL, ASD48h and RS24h groups.(B-C) Pathway enrichment analyses of the identified recoverable and unrecoverable metabolites.The x-axis represents the pathway impact, and the y-axis represents the -log (P-value).(D-E) Connections between the recoverable and unrecoverable core metabolites and metabolic pathways.(F-M) The relative levels of representative metabolites which were significantly down-regulated or up-regulated during ASD but were not significantly recovered during the subsequent 24 hours of RS.One-way ANOVA followed by Tukey's multiple comparisons test results are reported in Table S1 (n≥5 mice per group.***p < 0.001, **p < 0.01, *p < 0.05).
responses that are dysregulated and exhibit enhanced pro-inflammatory signaling, which raises the risk of infection as well as inflammation-related chronic diseases. 57,58Serving as the resident immune cells of the CNS, the activation of microglia is tightly correlated with brain states and behaviors. 21,22,24,25,59According to previous studies, microglia can be activated by sleep deprivation and undergo a reduction in the complexity of their processes and an increase in their production of pro-inflammatory cytokines. 24,60Even though studies have indicated that microglia show regional heterogeneity in their characteristics upon challenge, 60 our results have demonstrated that stimulation with ASD elicits similar microglial responses in various brain regions as indicated by de-ramification and decreased surface area.Indeed, sleep deprivation studies have yielded variable results depending on the species, duration, microglial features quantified and brain regions investigated. 61,62For example, microglial activation was observed in the rat hippocampus after 5 days of chronic sleep deprivation (CSD) induced by forced locomotion 28 while 6 weeks of CSD by forced locomotion also promoted microglia activation in a mice Alzheimer model. 63owever, in another mouse study, 8 hours of ASD by gentle handling which increased slow-wave activity of NREM sleep rebound, did not lead to significant microglial activation. 23Similarly, the few investigations on the microglial reaction during RS have produced inconsistent results.For example, a study in adult mice observed an increase in the amount of microglia in the hippocampus, but not in the cerebral cortex after either 1 day or 1 week of recovery following 24 hours of ASD, while the morphological structures of microglia were not examined. 64Another study reported an increase in the surface area of Iba1-immunoreactive microglia in the hippocampus after 3 weeks of recovery following intermittent bouts of REM sleep deprivation over 3 months, in comparison with non-treated controls, whereas the density of hippocampal microglia remained unchanged. 65Even 7 nights of RS was demonstrated to be insufficient to restore homeostasis of microglial density, surface area and ramification following CSD. 66Interestingly, our study which employed a novel "gentle handling" ASD protocol, showed sustained, lowlevel activation of microglia during 48 hours of ASD and subsequent 24 hours of RS, whereas the density of microglia remained unchanged in all brain regions.Furthermore, microglial migration was not observed by our in vivo two-photon imaging during the entire process.Consistent with previous data, changes of certain immune parameters may largely depend on the forms and duration of sleep loss and recovery. 67,68Importantly, it should be stressed that even without microgliosis, persistent low-level microglial activation may still lead to brain damage and detrimental reactions that promote pathological states. 69,70he highly dynamic microglial processes indicate interactions between microglia and brain parenchymal elements including neurons and vasculature. 46,71Microglial surveillance, in which their processes continuously scan the CNS, is crucial for promoting synapse formation during development 72 and monitoring the functional state of synapses upon maturity. 34,53In addition, it has been demonstrated to play a beneficial role in fine-tuning neuronal activity, by increasing the synchrony of neuronal network activity in the healthy brain, 73 or by reducing neuronal activity during certain pathological process. 74It is tempting to speculate that the significant decline of microglial process motility during ASD and subsequent RS observed in this study reflects a persistent reduction in microglial surveillance, which might to some extent affect the immune defenses 32,75 as well as neural network function 33,74 of the CNS.However, approaches such as continuous video recording through long-term in vivo two-photon imaging in both waking and sleeping mice will be an important tool for future studies aimed to clarify this issue.

Microglial phagocytosis and synaptic plasticity during sleep deprivation and restoration
0][31][32] According to previous studies, synaptic elimination depends on the levels of neuronal activity or sensory experience, and microglia preferentially phagocytose less active synapses; 53,54 in addition, signaling pathways including the CX3CR1 96 and TREM2 77 pathways have been demonstrated to be involved in the synaptic elimination by microglia during development.However, evidence of synaptic eliminations by microglia under different physiological states remains scarce. 78In this study, we have assessed microglial phagocytosis and synaptic plasticity based on CD68 colocalization with microglia and the formation/elimination statistics of dendritic spines.Interestingly, our results have revealed a significant enhancement of microglial phagocytosis and a persistent loss of dendritic spines during ASD and subsequent RS.Generally, microglial phagocytosis exerts diverse effects in both physiological and pathological conditions.On one hand, it may play beneficial roles in the clearance of apoptotic cells, metabolic waste, molecules, and debris such as myelin or amyloid deposits 63,79 and therefore increased phagocytosis may represent a positive coping strategy to counter the adverse effects of sleep loss.On the other hand, microglial phagocytosis is capable of removing or remodeling synapses, thus potentially interfering with neuronal communication. 80Therefore, the continuous loss of dendritic spines observed during ASD/RS suggests an associated impairment of neural networks and irrecoverability within the short term, a characteristic feature of various cases of CNS dysfunction and disease. 81In further studies, apart from the Thy1-YFP line H that was used in this work, we also plan to use other transgenic mice with different labeling patterns to verify whether spines in neurons of other layers/regions display similar responses.Interestingly, our results do differ from the findings of previous studies in terms of the effects of sleep deprivation on microglial phagocytosis.For example, the engulfment of postsynaptic components within the microglia was reduced after 72 hours of ASD in adolescent mice. 60Furthermore, the protein complements and Mer receptor tyrosine kinase (MerTK) receptors which have been demonstrated to be involved in cortical microglial phagocytosis have been shown to be increased to enhance phagocytosis after chronic sleep loss in adolescent mice. 23These effects might vary depending on the models used or the duration of sleep loss and recovery, therefore the states of microglia and their precise effects on neuronal networks during sleep insufficiency or restoration warrant more extensive investigation.

Sleep deprivation alters whole brain metabolic profiles
Many "omics" techniques have been applied to sleep research.Metabolomics has an advantage over others in that it integrates and amplifies information from changes at various levels. 82In addition, its comprehensiveness, sensitivity and unbiased nature make it especially suitable for capturing subtle physiological changes.Previous metabolomic analyses have been more focused on less invasive body fluids such as urine, 83 plasma 84 and saliva 85 to screen for potential metabolic biomarkers for evaluating sleep status.Recent studies have also attempted to explore cerebral metabolic changes caused by heterogeneous sleep disruption protocols, but these have yielded inconsistent results. 86Therefore, additional metabolic profiling is urgently required to understand the effects and mechanisms of sleep debt and recovery on CNS function.In this study, observed changes in metabolite levels within various categories indicate that ASD induces extensive metabolic disruption.After 48 hours of ASD, most of the metabolites were restored to normal levels within 24 hours of RS, such as ADP, Acetyl-CoA, Tryptophan, Proline and Citicoline, whereas a proportion of metabolites remained at the abnormal levels suggesting that some metabolic pathways may not be easily restored in the short term.D-Glucose in the brain was significantly downregulated during both ASD and RS.Reduced glycolytic flux and impaired glucose metabolism have been strongly correlated with the severity of Alzheimer's disease (AD). 87According to previous studies, increased neuronal activity due to sleep deprivation may account for glycolytic abnormalities that contribute to greater Aβ release in AD.Specifically, glycolytic inhibition may lead to the abnormal build-up of intermediate metabolites and promote microglia mediated neuroinflammation. 639][90] 8-OHG in the mitochondrial DNA of neurons causes calpain-dependent neuronal loss, while its accumulation in microglia results in PARP-dependent activation of apoptosis-inducing factor and exacerbated microgliosis. 91Sulfatide is a class of multifunctional molecule implicated in various biological processes in the nervous and immune systems and its abnormal metabolism has been demonstrated to be present in the earliest clinically recognizable stage of AD. 92 3-O-Sulfogalactosylceramide, an acidic sulfated glycosphingolipid was significantly down-regulated during ASD and its levels did not recover during short-term RS, suggesting the potential correlation between sleep loss and AD.The glycerophospholipid metabolic pathway was significantly enriched in the ASD-induced differential metabolites, as well as the identified unrecoverable metabolites during RS.Previous studies have shown that lysophospholipids of glycerophospholipid metabolism can exhibit properties similar to those of extracellular growth factors or signaling molecules which activate the PPARγ pathway and are involved in diseases including atherosclerosis, vascular dementia and spinal cord injury. 93For example, lysoPC exhibits multiple properties under pathological conditions.It promotes antioxidative capacity of high-density lipoprotein and protects low-density lipoprotein from oxidation 94 and chemoattracts monocytes, T cells and NK cells to areas of inflammation thereby mediating an inflammatory reaction. 95Furthermore, it can stimulate macrophages, boost their phagocytic activity 96 and polarize them toward an M1-like phenotype. 97Thus, we speculate that the observed, disrupted cerebral metabolic homeostasis is closely linked to the sustained microglial activation and accelerated elimination of dendritic spines we have documented following ASD and subsequent RS.
By employing various transgenic fluorescent mouse models and neuroimaging techniques, we have studied the morphological and functional changes of microglia and have documented sustained microglial de-ramification, reduction of process motility and enhancement of microglial phagocytosis across brain regions during ASD and RS.Consistent with the close links between microglial activity and neuronal plasticity, we revealed an accelerated elimination of dendritic spines during this entire process.Given that mice are nocturnal animals, confirming the basic phenomena and mechanisms revealed in this study in other diurnal model species and humans will be an important goal for future studies.However, in combination with our metabolome data, our results should help to elucidate how sleep schedules affect homeostasis and function of the CNS.

■■■■
Microglial de-ramification occurs during acute sleep deprivation and restoration.Reduced microglial processes mobility during sleep deprivation and restoration.Acute sleep deprivation and restoration increase the phagocytosis of microglia.Acute sleep deprivation and restoration induce elimination of dendritic spines.

Figure 1 .
Figure 1.Acute sleep deprivation (ASD) model and experimental design in this study (A) Diagram of experimental setup to induce ASD in mice.(B) Mice were randomly divided into normal CTRL group, ASD24h, ASD48h and RS24h groups to assess the effects of ASD and subsequent recovery sleep (RS).In vivo two-photon imaging was applied to examine the changes of microglial dynamics and synaptic structures.Immunofluorescent staining, brain sections fluorescence imaging and whole brain metabolome analysis were applied at the indicated timepoints to explore the potential cellular and molecular mechanisms.

Figure 2 .
Figure 2. Microglia display sustained activation during ASD and subsequent RS (A) Representative images of CX3CR1-GFP labeled ramified microglia and 3D tracing reconstructions by using Image J skeletonization and Imaris filament tracker across different brain regions of mice in normal control group (CTRL), acute sleep deprivation group (ASD48h) and recovery sleep group (RS24h).(B-D) The microglial superficial area, the total length of microglia processes and the total number of microglial process endpoints were significantly reduced after ASD and RS.(E) Quantification of microglial density in normal control group, ASD group and RS group.One-way ANOVA followed by Tukey's multiple comparisons test results are reported in Table S1 (n ≥ 5 mice per group.***p < 0.001, **p < 0.01, *p < 0.05).

Figure 3 .
Figure 3.In vivo imaging of microglial process motility during ASD and subsequent RS (A) Diagram illustrating the evaluation of microglial mobility via in vivo time-lapse imaging.The sequential images are pseudo-colored in red or green and merged.Between the two time points, yellow pixels represent stable in the merged images, green pixels indicate newly extended processes and red pixels are retraction.During the whole imaging session (30 min), this procedure was repeated for all sequential time points.(B,C) Quantification of the stability index and instability index of microglial processes.(D) Merging maximal intensity projection (MIP) images of microglial morphology at 5-min time intervals before (control) and after ASD and subsequent RS. (E-G) Quantification of microglia in basal retraction, extension and motility indices.One-way ANOVA followed by Tukey's multiple comparisons test results are reported in Table S1 (n=36 cells from 6 mice.***p < 0.001, **p < 0.01, *p < 0.05).

Figure 4 .
Figure 4. Phagocytic activity of microglia is persistently increased after ASD and subsequent RS (A) Representative images showing lysosomal marker CD68 co-localized with CX3CR1-GFP labeled microglia as a 2D projection of the optical image stack and in the Imaris 3D volume reconstruction in different brain regions of mice subjected to ASD and subsequent RS. (B) M1 Mander´s coefficient showing colocalization between CX3CR1-GFP labeled microglia and CD68.(C) Percentages of CD68 positive signals within the volume of single microglia.One-way ANOVA followed by Tukey's multiple comparisons test results are reported in Table S1 (n = 6 mice per group.***p < 0.001, **p < 0.01, *p < 0.05).

Figure 5 .
Figure 5. Effects of ASD and subsequent RS on the turnover of dendritic spines (A) Timeline for the two-photon in vivo imaging.The animals were either subjected to ASD-RS or left undisturbed.(B) Repeated live imaging showing dendritic structures before ASD, after ASD for 48 h, and after 24 h of RS. (C, D) High-resolution images of dendritic branches showing spine formation and elimination in both normal control (C) and ASD/RS (D) mice.The dendritic segments in panel D are derived from the indicated regions in panel B. Arrows indicate newly formed spines, hollow triangles indicate eliminated spines, and filled triangles indicate stable spines.(E) The percentage of newly formed spines was not significantly changed in mice subjected to ASD/RS compared with the normal control mice.(F) The spine elimination rate was significantly increased in mice subjected to ASD/RS compared with the normal control mice.Two-way ANOVA followed by Sidak's multiple comparisons test results are reported in Table S1 (n≥5 mice per group.***p < 0.001, **p < 0.01, *p < 0.05).

y g e n c o m p o u n d s 2 3 2 . 5 %O rg a n ic n it ro g e n c o m p o u n d s 9 Figure 6 .
Figure 6.Metabolic profiling of whole brain during ASD and subsequent RS (A) Brain-specific metabolites heatmap.Each group is represented by a column, and each metabolite, by a row.The abundance of each metabolite is represented by a specific color indicating the degree of up and down regulation.(B) Classification of metabolites identified in brain tissues.(C) PLS-DA of the metabolome data indicates the clustering of CTRL, ASD and RS samples.(D) The number of up-and down-regulated metabolites in ASD24h, ASD48h and RS24h groups compared with the CTRL group.(E) Venn diagrams depicting the unique and shared differential metabolites in various comparisons.(F) Multivariate ROC curve-based exploratory analysis suggesting potential metabolic markers of sleep loss.(G) The top 25 enriched metabolite sets affected by ASD.Circles represent metabolite sets.The colors define the P-value and the diameters indicate the enrichment ratio.(H) Pathway enrichment analysis of the metabolites affected by ASD in whole brain.The xaxis represents the pathway impact, and the y-axis represents the -log (P-value).
Figure 7. Metabolites and pathways identified as recoverable/unrecoverable within 24 hours of RS (A) Heatmaps show the recoverable/unrecoverable metabolites in the comparison between CTRL, ASD48h and RS24h groups.(B-C) Pathway enrichment analyses of the identified recoverable and unrecoverable metabolites.The x-axis represents the pathway impact, and the y-axis represents the -log (P-value).(D-E) Connections between the recoverable and unrecoverable core metabolites and metabolic pathways.(F-M) The relative levels of representative metabolites which were significantly down-regulated or up-regulated during ASD but were not significantly recovered during the subsequent 24 hours of RS.One-way ANOVA followed by Tukey's multiple comparisons test results are reported in TableS1(n≥5 mice per group.***p < 0.001, **p < 0.01, *p < 0.05).