Activation of the Hippocampal LXRβ Improves Sleep-Deprived Cognitive Impairment by Inhibiting Neuroinflammation

Sleep deprivation (SD) leads to cognitive impairment due to neuroinflammation associated with impaired hippocampal neuronal plasticity and memory processes. Liver X receptors (LXRs), including LXRα and LXRβ isoforms, are crucial for synaptic plasticity and neuroinflammation. However, the potential roles of LXRs in the pathogenesis of cognitive impairment induced by SD remain unclear. We revealed that SD resulted in LXRβ reduction in the hippocampus, which was associated with upregulated expression of high mobility group box 1 (HMGB1)/toll-like receptor 4 (TLR4)/NF-κB p65, and knockdown of hippocampal LXRβ by shRNA (shLXRβ) led to cognitive impairment. GW3965, a dual agonist for both LXRα and LXRβ, ameliorated SD-induced cognitive impairment by inhibiting microglia activation, suppressing HMGB1/TLR4/NF-κB p65 pathway, and ultimately affecting the hippocampal expression of inflammatory cytokines in SD mice. LXRβ knockdown by shLXRβ abrogated the GW3965-mediated inhibition of the HMGB1/TLR4/NF-κB p65 pathway, therefore, abolishing the cognitive improvement. Moreover, inhibition of HMGB1 by glycyrrhizin (GLY) synergistic promoted GW3965-mediated anti-inflammation in activated microglia after lipopolysaccharide (LPS)/ATP stimulation and facilitated the cognitive improvement after GW administration by activating LXRβ. All the data suggested that GW3965 ameliorated impaired cognition in SD mice by suppressing the HMGB1/TLR4/NF-κB p65 pathway followed LXRβ activation. This study correlates a deficit of LXRβ in cognitive dysfunction in SD associated with HMGB1 inflammatory pathway in hippocampus, and LXRs may serve as a potential therapeutic target for cognitive impairment with anti-inflammation.


Introduction
Sleep benefits proper cognitive functioning, whereas sleep disruption or deprivation (SD), getting sleep less than 4 h a day, leads to cognitive deficits, such as impaired attention, decision making, learning, and various types of memory [1,2]. SD pervasively links to numerous neurological and psychiatric disorders [3], and the high clinical incidence of SD seriously endangers public health and brings a great burden to society [4]. Therefore, it is necessary to determine how SD impairs cognition and may thus be a target for disease treatment and/or prevention.
Hippocampus, a neural structure, is crucially involved in many cognitive functions including the formation and optimization of learning and memory and the consolidation of declarative memory [5,6] and is particularly sensitive to sleep loss. Evidence showed that SD substantially decreased the ability to induce long-term potentiation (LTP) in the hippocampus [7] and also reduced the synthesis of proteins related to hippocampal neuroplasticity and impaired neurogenesis in the hippocampus [8]. SD resulted in learning and memory impairments, especially when the hippocampus was required Chen Qiu and Min Wang contributed equally to this work. for these learning and memory processes [9]. Furthermore, the immunological and redox system was affected upon SD injury, leading to neuroinflammation and oxidative stress, and the hippocampus was rich in microglia and inflammatory factor receptors which responded to inflammatory reaction [10].
As the resident immune cells in the central nervous system (CNS), microglia exert critical physiological functions in regulating neuronal activity and synaptic plasticity, thus emotions can be processed and integrated correctly by pruning synapses and secreting active substances [11,12]. Microglia were observed activated under SD stress and followed by the release of a large number of proinflammatory factors, leading to cognitive impairment [9]. Therefore, it is of great importance to correct microglia dysfunction and following issues to treat cognitive impairment induced by SD [13]. High mobility group box 1 (HMGB1), a proinflammatory cytokine, is secreted by activated microglia [14], acting as an alarm protein for neuroinflammation [15]. HMGB1 is translocated to extracellular space from the nucleus and cytoplasm, resulting in the increased NF-κB nuclear translocation by binding to toll-like receptor 4 (TLR4), a specific receptor for HMGB1 expressed exclusively in microglia in the CNS [16]. Subsequently, the expression of proinflammatory cytokines is induced including interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α) [17]. It is possible to rescue the cognitive deficit by correcting the abnormal neural plasticity mediated by neuroinflammation.
Activation of liver X receptor β (LXRβ), a nuclear receptor, increases hippocampal neurogenesis, regulates synaptic plasticity, and inhibits neuroinflammation. Knockout of LXRβ leads to microglia activation and supports that LXRβ is critical in neuroinflammation [18,19]. GW3965 (GW), an LXRs full agonist for both LXRα and LXRβ isoforms, can readily cross the blood-brain barrier to exert its specific actions in the brain [20]. Studies have proved that GW can inhibit the release of inflammatory factors in lipopolysaccharide (LPS)-induced inflammation and improve cognitive impairment caused by Alzheimer's disease (AD) by activating LXRs [21,22]. However, whether GW can ameliorate SD-induced cognitive impairment and the mechanisms of LXRs involved remains unknown.
In this study, we found that the reduction of LXRβ expression was associated with SD-induced cognitive impairment. Activation of LXRβ by GW alleviated SD-induced cognitive impairment through inhibiting the activation of microglia and subsequently decreased the expression levels of HMGB1, TLR4, NF-κB p65, TNF-α, and IL-1β in hippocampal microglia. We demonstrated the important role of LXRβ in SDinduced cognitive impairment and presented a new therapeutic approach to mitigate SD-induced cognitive impairment by giving GW administration.

Animals
Eight-to ten-week-old adult male C57BL/6 mice (weighed 20-25 g) were obtained from the Experimental Animal Center of Fourth Military Medical University (certificate no. 201000082, grade II). The mice were kept as one group in a 380 × 325 × 180-mm plastic box made of opaque polypropylene (By-DSLH380). They were placed in a 12-h light/dark cycle environment (temperature of 24 ± 2°C and humidity of 50-60%) with food and water available ad libitum. All experimental procedures were approved by the Ethics Committee of Fourth Military Medical University (approval reference number no. KY20193145) in full accordance with the ethical guidelines of the National Institutes of Health for the care and use of laboratory animals. The animals were habituated to laboratory conditions for at least 1 week before testing. All the behavioral tests were taken between 8 a.m. and 6 p.m. Every effort was made to minimize the number of animals used and their suffering.

Experimental Procedure
An experimenter, who was blind to the design of the experiment, predesignated 18 mice from 1 to 18, and a computer randomization table was used to randomly divide the numbered mice into 3 groups: control (Ctrl), SD, and SD + GW (10 mg/kg). The experimental procedure was shown in Fig.  1a. The mice in the SD and SD + GW groups were deprived of sleep 20 h a day for 28 days. Ctrl mice were kept in the same environment but allowed to sleep at will. After 14 days of SD, the mice received GW (10 mg/kg) or vehicle (saline, 10 ml/ kg) by intraperitoneally (i.p.) injection once a day for 14 days (from day 14 to day 28) consecutively. Drugs were given between 9 a.m. and 10 a.m. Animal learning and memory behaviors were tested at a fixed time after SD. Mice from each group were sacrificed to collect the hippocampus for immunofluorescence staining and western blot analysis after behavior tests.

Sleep Deprivation Procedure
The SD model was established by using the small platform water environment method [23]. SD was performed in a 50 × 40 × 20-cm water tank with several cylindrical platforms (5 cm in height, 3 cm in diameter, and 1 cm on the horizontal surface), and mice were free to eat and drink on the platform. The platforms were spaced at a distance of 4 cm so that mice could move freely from one platform to another. When the mice entered rapid eye movement (REM) sleep, they showed a drop in muscle tone that caused them to drop their heads to touch the water and woke up, thus keeping the mice awake and unable to sleep. Before the experiment, the mice were adapted to the platform environment for 3 days. During the experiment, the indoor temperature was maintained at 22-25°C, and the 40-W fluorescent lamp was continuously illuminated. The mice were sleep deprived for 20 h a day and rested for 4 h (8 a.m. to 12 a.m.).

Morris Water Maze
Morris water maze (MWM) was conducted as described previously [24]. The MWM consisted of a 100-cm diameter pool with nontoxic skim milk powder as an opaque background and divided into equal-sized quadrants. A circular platform (10 cm in diameter) was used as a target, and the experimental water temperature was 20 ± 2°C. There were different special visual cues around the pool. All behaviors were recorded in real time using an automatic tracking system. During the experiment, the mice were put into the water with their heads facing the wall. The entry quadrant varied, but the platform location remained constant. During the training, if the animal could not find the platform within 60 s, it was guided to the platform by the experimenter and allowed to sit on the platform for 15 s before being removed. All mice were trained for 4 days, four times a day, and their swimming paths were recorded. One hour after the final learning trial, a single 60-s probe trial was conducted without the platform. The latency to the target area (the previous platform location), the time spent in the target quadrant, the number of platform crossings, and the distance traveled were calculated. The analysis results were used to evaluate the spatial learning and memory ability of different groups of mice.

Novel Object Recognition
The novel object recognition (NOR) test was performed as described previously [25]. The apparatus was composed of a 25 cm × 25 cm × 20-cm soundproof device equipped with a digital camera on the roof. The cylinder and cube shared the same volume. Habituation trial was taken on the first day, and mice were placed in the behavioral chamber to adapt to the new environment. In the absence of objects, each mouse acclimated to the instrument for 10 min. The training trial was taken on the second day. The mice were placed in the instrument with two identical cylinders placed in symmetrical corners of the box and allowed to explore freely for 10 min. The testing trial was taken on the third day. One cylinder was replaced with a cube of similar volume in the same place and the mice were allowed to explore freely for 10 min, and the time spent exploring each object was recorded. After each phase, the objects and the box were wiped with 70% ethanol to remove the odor of the previous experimental animal. The mice's exploratory tracks were recorded by a camera fixed above the floor and analyzed with a video-tracking system. Interaction parameters were defined as contact with the object (tail only excluded) or facing the object (distance< 2 cm). The time spent exploring familiar objects (TF) and the time spent exploring the novel objects (TN) were recorded. The proportion of the exploration time exploring the novel object was defined as the "recognition index" expressed by the ratio of TN/ (TF + TN).

Western Blot Analysis
Protein samples were harvested from mice hippocampus or cultured N9 microglial cells after various treatments. Total proteins were lysed by M-PER Protein Extraction Buffer. Nuclear and cytoplasmic proteins of the hippocampus were extracted with nuclear and cytoplasmic protein extraction kit according to the manufacturer's instructions. Protein concentrations were determined using a BCA kit. Thirty micrograms of protein per sample were run on 12% SDS-PAGE to separate target proteins and were subsequently transferred to the PVDF membrane. The membrane was blocked with 5% nonfat milk in Tris-buffered saline containing 0.1% Tween 20 for 60 min at room temperature. Subsequently, the membranes were incubated in the primary antibody solutions overnight at 4°C. The primary antibodies included anti-LXRα , and anti-β-actin (1:10,000) served as loading controls. After incubation with the appropriate HRP-conjugated secondary antibody, the proteins were visualized and digitized using ECL solution and ImageJ program.

Immunohistochemistry
The immunohistochemistry was conducted as described before [27]. The animals were anesthetized with sodium pentobarbital (40 mg/kg. i.p.) after the behavioral studies and perfused with PBS first and then with 4% paraformaldehyde. Brains were quickly removed and fixed overnight. After dehydration and cryoprotection in 30% sucrose at 4°C for 48 h, the hippocampus segments of the frozen brain were sectioned into 30-μm slices. All sections were washed with 0.1% Triton X-100 in PBS for 30 min and blocked in 10% goat serum for 1 h. Then the brain slices were incubated with anti-LXRβ antibody (1:200) and anti-Iba-1 antibody (1:100) overnight at 4°C, followed by Alexa Fluor secondary antibody incubation. Nuclei were counterstained with Hoechst 33258. Fluorescent signals were photographed and analyzed using confocal fluorescence microscopy (Olympus, Japan).

Cell Culture and Treatment
One cell line of murine microglia N9 was provided by the Shanghai Sixin Biotechnology Ltd. (Shanghai, China) and cultured in RPMI1640 supplemented with 10% FBS, 100 IU/ml penicillin, and 100 μg/ml streptomycin. The cells were cultured at 37°C in the presence of 5% CO 2 with the culture medium changed daily. Cells were pretreated with GW (0, 1, 10, 100 μM) for 2 h and then subjected to the stimulation of LPS (1 μg/ml) for 4 h and ATP (2.5 mM) for 0.5 h. For the experiment utilizing shLXRβ (1 × 10 9 gc/ml), cultured microglia N9 cells were transfected with the AAV shRNAs for 24 h according to the manufacturer's instruction. The cells were cultured for 72 h after infection and then treated as before. The knockdown efficiency of shLXRβ was determined at 72 h after infection using western blot analysis. For the experiment utilizing inhibitors, cells were pretreated with GLY (50 μM) for 10 h and then treated with or without GW (10 μM) for 2 h. Then cells were stimulated with LPS (1 μg/ml) for 4 h and ATP (2.5 mM) for 0.5 h. Samples were collected by centrifugation for subsequent experiments.

Enzyme-Linked Immunosorbent Assay
The supernatant of cultured N9 cells was harvested after various treatments and subjected to enzyme-linked immunosorbent assay (ELISA) to detect the changes of factors involved. The supernatant was centrifuged at 2-8°C for 20 min (2000-3000 rpm), and the supernatant was carefully collected. The contents of HMGB1, TNF-α, and IL-1β in the supernatant of cells were measured by the ELISA kits according to the manufactures' protocols. The absorbance value was measured After SD for 14 days, the mice received GW3965 (GW, 10 mg/kg) or 0.9% saline by intraperitoneally (i.p.) injection once a day for 14 days consecutively. b Western blot analysis showed no changes of LXRα and reduced LXRβ in the hippocampus after SD. c-f Morris water maze (MWM) test showed that SD induced cognitive impairment, which was relieved by the treatment of GW. c Mean daily escape latencies (time to find the hidden platform). d Total swimming distance during the learning phase. e Percentage of time spent in the target quadrant. f Numbers of crossing platform site spent in the target quadrant. g, h The results of the novel object recognition (NOR) test. g Recognition index toward a novel object. h Total distance traveled in the NOR test. n = 6 in each group, ** p < 0.01, *** p < 0.001, compared with the Ctrl group; # p < 0.05, ## p < 0.01, ### p < 0.001, compared with the SD group with an enzyme-labeled instrument (Bio-Rad, USA). The concentrations of samples were determined by absorbance value and standard curve.

Statistical Analysis
Results were presented as the mean and SEM (mean ± SEM). Unpaired Student's t test was used to test the difference between the two groups. One-way ANOVA followed by a posthoc Tukey test was used to test the difference among multiple groups. In all cases, p< 0.05 was considered statistically significant. All statistical analyses were performed using GraphPad Prism 7.03 and SPSS statistical software package version 20.0.

The Expression of LXRβ was Decreased in Hippocampus of SD Mice
To determine whether LXRs signaling was required in SD, the SD model was established using a small platform water environment, in which the mice were deprived of a 20-h sleep a day for 28 days. The expression changes of two subtypes of LXRs in the hippocampus were detected after SD. Interestingly, the expression level of LXRβ in the SD group was substantially decreased to 37.94% ± 5.23% of Ctrl (p< 0.001, SD group vs. Ctrl; Fig. 1b), while the expression level of LXRα in the hippocampus was not altered after SD (p> 0.05, SD group vs. Ctrl; Fig. 1b). These data suggested that SD elicited the reduction of LXRβ expression in the hippocampus and activation of LXRβ with an agonist might alleviate SD-induced symptoms.

LXRs Agonist GW3965 Alleviated SD-Induced Cognitive Impairment
Cognitive impairment is one of the most important symptoms induced by SD [28]. In this study, two classic behavioral methods, MWM and NOR, were used to examine the cognitive changes of mice from each group. The experimental process was shown in Fig. 1a. As expected, SD mice spent more time to find the platform in the MWM test, which was manifested as an increase in escape latency of 39.01 ± 2.61 s at day 4 compared with the Ctrl group of 10.89 ± 2.24 s (p< 0.001 at day 4, SD group vs. Ctrl; Fig. 1c). The swimming distance in the SD group was 14.82 ± 7.47 m at day 3 and 12.68 ± 1.26 m at day 4, while the Ctrl group was 8.43 ± 0.98 m at day 3 and 3.70 ± 0.90 m at day 4 (p< 0.01 at day 3, p< 0.001 at day 4, SD group vs. Ctrl; Fig. 1d). This result indicated that SD mice swam longer distances than the Ctrl group before finding the platform. However, the escape latency and swimming distance in GW-treated group were significantly decreased to 14.90 ± 1.79 s at day 4 (p< 0.001, SD + GW group vs. SD group; Fig. 1c) and 5.69 ± 0.58 m (p< 0.01, SD + GW group vs. SD group; Fig. 1d). This result indicated that GW treatment ameliorated SD-induced learning impairment. During the probe trials, SD induced impaired memory, which was manifested as a decrease in time spent in the target quadrant and platform crossings compared with Ctrl group (p< 0.01, SD group vs. Ctrl; Fig. 1e; p< 0.001, SD group vs. Ctrl; Fig. 1f). However, the impaired memory was improved after GW administration (p< 0.05, SD + GW group vs. SD group; Fig. 1e, f). In the NOR test, SD decreased the time spent exploring the novel object. The recognition index in the SD group was 0.49 ± 0.026 compared with the Ctrl group of 0.69 ± 0.027 (p< 0.001, SD group vs. Ctrl; Fig. 1g), indicating that SD impaired the memory of mice. As expected, GW treatment alleviated this decrease to 0.61 ± 0.018 (p< 0.01, SD + GW group vs. SD group; Fig. 1g). Finally, to exclude the interference of locomotor activity, the locomotor abilities of different groups were analyzed, and there were no differences among the three groups (p> 0.05; Fig. 1h). Consistent with previous studies, SD-induced cognitive impairment was observed in our study. The result also proved that activating LXRs by GW alleviated cognitive impairment after SD.

Knockdown of LXRβ by shRNA in Hippocampus Led to Cognitive Impairment
It has been proven that SD led to a significant decrease of LXRβ expression in the hippocampus. This result suggested that hippocampal LXRβ defects might be associated with cognitive impairment. Given the lack of selective antagonists for the LXRβ isoform, we used shLXRβ stereotactic injection into the hippocampus to knock down endogenous LXRβ. Western blot data showed that the level of LXRβ was decreased to 33.61% ± 2.28% of shNC (p< 0.001, shLXRβ group vs. shNC; Fig. 2a) after shLXRβ transfection. The MWM result indicated that the mice in the shLXRβ group spent more time and traveled longer distance to find the platform (p< 0.001 at d 4, shLXRβ group vs. shNC; Fig. 2b, c). Compared with the shNC group, the time spent in the target quadrant and platform crossings of the shLXRβ group was also decreased (p< 0.05, shLXRβ group vs. shNC, Fig. 2d; p< 0.01, shLXRβ group vs. shNC, Fig. 2e). In the NOR test, the shLXRβ group showed a decrease in recognition index (p< 0.01, shLXRβ group vs. shNC; Fig. 2f). These results indicated that reduced hippocampal LXRβ expression was associated with cognitive impairment. However, when LXRα expression was decreased to 43.38% ± 3.78% of shNC (p< 0.001, shLXRα group vs. shNC; Fig. 2a) after shLXRα infection in the hippocampus, the mice showed no cognitive impairment compared with shNC mice (p> 0.05, shLXRα f Recognition index toward a novel object. g There was no significant difference in the locomotor ability of mice in each group. n = 6 in each group, * p < 0.05, ** p < 0.01, *** p < 0.001, compared with the shNC group group vs. shNC; Fig. 2b-f). Furthermore, combining shLXRα + shLXRβ infection could not lead to further reduction of LXRα and LXRβ expression and more severe cognitive impairment (Fig. 2).

GW3965 Improved SD-Induced Cognitive Impairment by Activating LXRβ Isoform
Based on previous results, it seemed that the LXRβ isoform played a much more important role in cognition than the LXRα isoform. Since GW was a nonselective agonist for both LXRα and LXRβ subtypes, shRNA-mediated knockdown of LXRα and LXRβ in the hippocampus was also used to verify the subtype of GW-mediated cognitive improvement. The SD model was set up 7 days after shRNA infection, and mice were treated as before (Fig. 3a). After shLXRα infection, GW could still improve SD-induced learning disability (p< 0.001 at day 4, shLXRα + SD + GW group vs. shLXRα + SD group; Fig.  3b, c) and memory impairment (p< 0.05, shLXRα + SD + GW group vs. shLXRα + SD group, Fig. 3h; and p< 0.01, shLXRα + SD + GW group vs. shLXRα + SD group, Fig. 3i). This result indicated that the knockdown of hippocampal LXRα expression did not affect GW-mediated cognitive improvement in SD mice. However, the knockdown of hippocampal LXRβ prevented GW-mediated cognition improvement (p> 0.05, shLXRβ + SD + GW group vs. shLXRβ + SD group; Fig. 3d, e, h, i). Furthermore, when the expression levels of LXRα and LXRβ were both downregulated, GW also failed to exert its effects (p> 0.05, shLXRα + β + SD + GW group vs. shLXRα + β + SD group; Fig. 3f-i). These results suggested that GW relieved SD-induced cognitive impairment by activating the LXRβ isoform.

GW Relieved SD-Induced Microglia Activation, Followed by Suppressing HMGB1/TLR4/NF-κB p65 Pathway in Hippocampus
It has been shown that GW improved SD-induced cognitive impairment through LXRβ, but the mechanisms involved were still unclear. Evidence showed that the knockout of LXRβ led to microglia activation, suggesting LXRβ was responsible for neuroinflammation (Colonna & B u t o v s k y , 2 0 1 7 ; D a i , T a n , W u , W a r n e r , & Gustafsson, 2012). In this study, we found that hippocampal microglia were activated with a decreased LXRβ expression (Extended data Fig. 1), suggesting that the status of microglia might be related to the expression level of LXRβ upon SD injury. Besides, the presence of LXRβ in the microglia was also observed (Extended data Fig. 2). To verify whether GW could inhibit microglia activation after SD, the status of microglia in the SD + GW group was tested. As expected, activated microglia in SD mice were significantly reduced after GW administration (Fig.   4a). Western blot results also showed that SD robustly enhanced the hippocampal expression level of Iba-1 to 235.30% ± 14.22% of Ctrl (p< 0.001, SD group vs. Ctrl; Fig. 4b, c), while GW treatment decreased this level to 121.30% ± 5.67% of Ctrl (p< 0.001, SD + GW group vs. SD group; Fig. 4b, c). These results confirmed that GW could mitigate microglia activation induced by SD in the hippocampus.
HMGB1 is a typical damage-associated molecule involved in various neuroinflammatory and neurodegenerative diseases, and the activation of the HMGB1/TLR4/NF-κB p65 pathway is highly associated with microglial activation and leads to elevated levels of inflammatory factors. In this study, we found that the hippocampal expression of HMGB1 and TLR4 was increased to 151.00% ± 7.50% and 205.30% ± 7.70% of Ctrl after SD (p< 0.001, SD group vs. Ctrl; Fig.  5a, b, d, e). Besides, SD also increased the nuclear level of NF-κB p65 and decreased cytoplasmic NF-κB p65 expression correspondingly in the hippocampus (p< 0.001, SD group vs. Ctrl; Fig. 5g-i), leading to elevated TNF-α and IL-1β (p< 0.001, SD group vs. Ctrl; Fig. 5c, f). However, GW treatment could significantly reduce the protein levels of HMGB1, TLR4, NF-κB p65, TNF-α, and IL-1β (SD + GW group vs. SD group; Fig. 5a-i), suggesting that GW could ameliorate neuroinflammation in the hippocampus by inhibiting the HMGB1 pathway.

GW Suppressed HMGB1/TLR4/NF-κB p65 Pathway in Activated Microglia by Activating LXRβ
The inhibitory effects of GW on the HMGB1 pathway in activated microglia were further confirmed in cultured murine N9 microglia in vitro. GW pretreatment dose dependently decreased the induction of Iba-1 and CD68, a marker of activated microglia, induced by LPS/ATP stimulation shown in Extended data Fig. 3. On account of no obvious differences between 10 and 100 μM groups, 10 μM of GW was used in the following trials.
To verify whether the inhibitory effect of GW was through LXRβ, shLXRβ was further applied in cultured N9 microglia. Western blot data showed that the expression of LXRβ was successfully downregulated to 35.92% ± 4.45% of shNC after shLXRβ infection for 3 days (p< 0.001, shLXRβ group vs. shNC; Fig. 6a). LPS/ATP stimulation robustly enhanced the expression levels of HMGB1, TLR4, NF-κB p65, TNF-α, and IL-1β, which were relieved by the pretreatment of GW as shown in Fig. 6. However, the proteins in the HMGB1 pathway were increased in the shLXRβ + LPS/ATP + GW group compared with the shNC + LPS/ATP + GW group and showed little differences compared with the shLXRβ + LPS/ ATP group (Fig. 6b-h). The results indicated that GW inhibited HMGB1/TLR4/NF-κB p65 pathway in activated microglia due to the presence of LXRβ. HMGB1, TNF-α, and IL-1β were released from cells to exert the proinflammatory roles, then the content changes in the supernatant were also detected by ELISA assay. The ELISA data, shown in Extended data Fig. 4, exhibited a similar tendency to that of western blot results as shown in Fig. 6. The expression of HMGB1, TNF-α, and IL-1β in the supernatant showed little differences between the shLXRβ + LPS/ATP + GW group and shLXRβ + LPS/ATP group, which meant GW inhibited inflammatory cytokines released from N9 microglia by activating LXRβ.

Inhibition of HMGB1 Facilitated the Anti-Inflammatory Effect of GW upon LPS/ATP-Induced Inflammatory Responses in Murine N9 Microglia
To further confirm that the HMGB1 pathway exerted an important role in GW-mediated anti-inflammatory effects, an HMGB1 inhibitor GLY was used, which bound directly to HMGB1, interacting with two shallow concave surfaces formed by the two arms of both HMG boxes [29]. The results showed that GLY (50 μM) significantly inhibited nuclear translocation of NF-κB p65 upon LPS/ATP stimulation (LPS/ATP + GLY group vs. LPS/ATP group; Fig. 7a, b), followed by the expression reductions of TNF-α and IL-1β (LPS/ATP + GLY group vs. LPS/ATP group; Fig. 7a, c) and the release reduction into the supernatants (LPS/ATP + GLY group vs. LPS/ATP group; Fig. 7d). Furthermore, GLY administration enhanced the GW-mediated reduction of NF-κB p65 upon LPS/ATP injury and the release of TNF-α and IL-1β (LPS/ATP + GW + GLY group vs. LPS/ATP + GW group; Fig. 7). This result indicated that GLY could promote the anti-inflammatory effects of GW in activated microglia.

Inhibition of HMGB1 Pathway Effectively Alleviated SD-Induced Cognitive Impairment and Enhanced GW-Mediated Cognitive Improvement
Finally, the role of the HMGB1 pathway in SD-induced cognitive impairment was tested. GLY (20 mg/kg) and/or GW (10 mg/kg) were injected intraperitoneally starting from day 14 after SD for 14 days consecutively (Fig. 8a). The behavior test showed that treatment with GLY alone for 14 days effectively alleviated SD-induced learning and memory deficit (SD + GLY group vs. SD group; Fig. 8b-f). Furthermore, GW and GLY had synergistic effects in improving cognitive impairment, which was manifested as a tendency to increase learning and memory functions compared with using GW or GLY alone (SD + GW + GLY group vs. SD +GW or SD + GLY group; Fig. 8b-f). Collectively, all the results confirmed that the HMGB1 pathway played an important role in cognitive deficit, and the inhibition of HMGB1 enhanced GW-mediated cognitive improvement after SD.

Discussion
In the present study, we found that SD led to cognitive impairment associated with the hippocampal reduction of LXRβ expression, and the knockdown of LXRβ by s h L X R β a l s o r e s u l t e d i n a c o g n i t i v e d e f i c i t . Furthermore, it was confirmed that GW-mediated cognitive improvement was through LXRβ activation, followed by the inhibition of the HMGB1/TLR4/NF-κB p65 pathway in microglia both in vitro and in vivo, thus alleviating neuroinflammation. These results suggested a link between hippocampal LXRβ deficiency and cognitive impairment and proved the important roles of LXRβ activation for the treatment of cognitive impairment induced by SD. It also provided new insights into the neuroinflammatory mechanisms of cognitive impairment.
Sleep benefits neuronal recovery and plasticity, which supports brain function, and ultimately, cognition and emotion, whereas sleep loss leads to cognitive impairment, such as learning and memory deficits [30]. Interestingly, when the hippocampus is needed for these learning and memory processes, such impairments appear to occur, suggesting that this area of the brain may be particularly sensitive to the consequences of SD. SD significantly decreased the LTP in the hippocampus and impaired hippocampal neurogenesis [31]. In addition, functional magnetic resonance imaging indicated that the hippocampal activity was significantly lower after SD than that of the normal volunteers [32]. Exploring the impact of SD may find a target for disease treatment and/or prevention under all major neurological and psychiatric Representative results of the Western blot analysis for Iba-1. GW notably suppressed the expression of Iba-1 in SD mice. n = 6 in each group, *** p < 0.001, compared with the Ctrl group; ### p < 0.001, compared with the SD group conditions, including AD, schizophrenia, anxiety disorders, and addiction in which comorbid sleep disruption presents [3].
Available evidence suggested that neuroinflammation impaired hippocampal neuronal plasticity and memory processes, leading to cognitive impairment [33]. It is critical to investigate the mechanism involved in SD-induced neuroinflammation. LXRs, a nuclear receptor which can inhibit the expression of proinflammatory genes in immune cells, are increasingly concerned because of their critical role in regulating neuroinflammation [34,35]. In this study, we found that SD led to the reduction of hippocampal LXRβ expression; moreover, downregulation of LXRβ in the hippocampus by AAV shRNA resulted in significant cognitive impairment in the mice. This suggested that the LXRβ deficit might be responsible for the pathogenesis of SD-induced cognitive impairment. Furthermore, this study also confirmed that GW, a non-selective LXRs agonist, improved the cognitive function of SD mice. Due to the lack of LXRβ isoform-specific agonist, GW-mediated improvement by activating LXRβ instead of LXRα was verified using shRNA-mediated knockdown of LXRβ.
Previous studies have found that the knockout of LXRβ led to an increase of activated microglia, which are the cellular mediators of neuroinflammation and the main source of proinflammatory cytokines [18,19]. Overactivated microglia produce neurotoxicity, damage neurons, and are involved in a range of neurodegenerative diseases characterized by cognitive impairment such as AD, Parkinson's diseases, and Huntington's diseases [36]. HMGB1, a typical damage-associated molecule, has been shown to be released by hippocampal microglia upon unpredictable stress [14]. The release of HMGB1 from microglia is decreased after the inhibition of microglia activation, thus the HMGB1 protein level is restored accordingly [14]. HMGB1 serves as a risk factor for TLR4, and f IL-1β and h, i nuclear translocation of NF-κB p65 in the hippocampus after SD. n = 6 in each group, *** p < 0.001, compared with the Ctrl group; ## p < 0.01, ### p < 0.001, compared with the SD group memory impairment, chronic neurodegeneration, and progression of neuroinflammation in neurodegenerative diseases [37]. Released HMGB1 binds to TLR4, a specific receptor expressed by microglia in rodents, and induces the nuclear translocation of NF-κB p65 from the cytoplasm, leading to the induction of inflammatory factors [ 38,39 ]. Consiste nt with previous literatures, overactivated microglia and activation of HMGB1/ TLR4/NF-κB p65 pathway upon SD injury were also observed in this study, which led to a significant increase of inflammatory factors TNF-α and IL-1β. Elevated levels of inflammatory factors in the hippocampus led to impaired neural plasticity, resulting in hippocampusdependent learning and memory impairments. Therefore, LXRβ-mediated cognition improvement was due to the inhibition of microglia activation and HMGB1 pathway, Fig. 6 GW3965 suppressed the HMGB1 pathway in the murine N9 microglia activated by LPS and ATP through activating LXRβ. a After a 3-day infection of shLXRβ adeno-associated virus (AAV), the level of LXRβ was reduced in cultured N9 microglial. b Representative results of the western blot analysis. GW (10 μM) treatment inhibited elevated levels of c HMGB1, d TLR4, g TNF-α, and h IL-1β and e, f nuclear translocation of NF-κB p65 upon LPS/ATP stimulation. Knockdown of LXRβ expression prevented GW from inhibiting the HMGB1 pathway. n = 6 in each group, *** p < 0.001, compared with the shNC group; ## p < 0.01, ### p < 0.001, compared with the shNC + LPS/ATP group; $$ p < 0.01, $$$ p < 0.001, compared with the shNC + LPS/ATP + GW group which suppressed the induction of inflammatory factors upon SD stress, followed by the correction of abnormal neural plasticity.
The inhibition of HMGB1 by GLY, a direct inhibitor of HMGB1, dampened the activation of the subsequent TLR4-NF-κB pathway and reduced the levels of TNF-α and IL-1β in the inflammatory pain model [40]. Similar results were also obtained in our study. The critical role of HMGB1/TLR4/ NF-κB p65 pathway in SD-induced cognitive deficit and in GW-mediated cognitive improvement was further confirmed by GLY as shown in Figs. 7 and 8. This result expanded the understanding of the mechanisms involved in GW-mediated cognitive effect. In addition to exerting anti-inflammatory effects, GW improved learning and memory by promoting neuronal survival and neural stem cell proliferation [24] and increased the clearance of Aβ which aggravated cognitive impairment [41]. All these beneficial contributions to cognition may be related to the expression pattern of LXRβ in different kinds of cells in the CNS. Therefore, more studies of LXRs-mediated cognitive improvement need to be explored further in the future.
Collectively, activating LXRβ offered the improvement in SD-induced cognitive impairment by inhibiting hippocampal microglia activation, followed by suppressing the pathway HMGB1/TLR4/NF-κB p65, and ultimately affecting the release of inflammatory factors. Furthermore, our data validated Fig. 7 HMGB1 inhibitor GLY synergistic promoted GW3965-mediated effects in inhibiting inflammatory responses induced by LPS and ATP in murine N9 microglia. a The expression levels of nuclear (Nuc) and cytoplasmic (Cyto) NF-κB p65, TNF-α, and IL-1β were determined by western blot. HMGB1 inhibitor GLY promoted the inhibitory effect of GW on b nuclear translocation of NF-κB p65 and c the expression of TNF-α and IL-1β. d The levels of TNF-α and IL-1β in supernatants from each group were evaluated by ELISA assay. n = 6 in each group, *** p < 0.001, compared with the Ctrl group; # p < 0.05, ## p < 0.01, ### p < 0.001, compared with the LPS/ATP group; $ p < 0.05, compared with the LPS/ATP + GW group the critical role of the hippocampal HMGB1 pathway that contributed to cognitive impairment induced by SD. These findings provided a new mechanism and target for the treatment of cognitive impairment induced by SD. Fig. 8 Inhibition of HMGB1 facilitated GW3965-mediated improvement of cognitive impairment induced by SD. a Schematic illustration of the experiment procedure. GLY (20 mg/kg) and/or GW (10 mg/kg) were injected intraperitoneally (i.p.) starting from 14 days after SD for 14 days consecutively. b-f GLY could alleviate the cognitive dysfunction induced by SD, and the effect tends to be enhanced when combined with GW. b Mean daily escape latencies. c Total swimming distance during the learning phase. d Percentage of time spent in the target quadrant. e Numbers of crossing platform site spent in the target quadrant. f Recognition index toward a novel object. g There was no significant difference in the locomotor ability of mice in each group. n = 5 in each group, *** p < 0.001, compared with the Ctrl group; # p < 0.05, ## p < 0.01, ### p < 0.001, compared with the SD group