Natural Glycoside Salidroside Ameliorates Orthopedic Surgery-Induced Cognitive Dysfunction Through Activating Adenosine 5‘-Monophosphate-Activated Protein Kinase Signaling in Mice

Background: Perioperative neurocognitive disorders (PND) are the most common postoperative complications with few therapeutic options. Salidroside, a plant-derived compound, has gained increased attention as treatment for various neurological diseases and particularly modier of microglia-mediated neuroinammation. However, the effect of salidroside on orthopedic surgery-induced cognitive dysfunction and the underlying mechanisms are largely unknown. Methods: The Morris water maze test was used to investigate potential effects of salidroside in the animal model of tibia fracturing with intramedullary xation. Therapeutic mechanism and related signaling pathways of salidroside in PND were further investigated with animal tissues and microglial cultures in vitro by molecular biology tests. Results: Here we found that salidroside greatly attenuated cognitive impairment in mice after orthopedic surgery. Neuroinammation in mouse hippocampus were also attenuated by salidroside. Meanwhile, salidroside treatment induced a switch in microglia polarization to the anti-inammatory phenotype. In vitro, salidroside suppressed the expression of pro-inammatory cytokines and induced a switch in microglial phenotype to the anti-inammatory phenotype. Mechanically, molecular docking studies revealed potential AMPK activation activity of salidroside. And salidroside did up-regulated the AMPK pathway proteins. Moreover, AMPK antagonist abolished the effects of salidroside in vivo and in vitro. Conclusions: Taken together, our results demonstrated that salidroside effectively suppressed PND by suppressing microglia-mediated neuroinammation through activating AMPK pathway, and it might be a novel therapeutic approach for PND. Salidroside mice. RT-qPCR analysis was used detect the mRNA expression of inammatory factors TNF-α, IL-1β, IL-6 and IL-10 (A) in hippocampus in mice. Representative immunohistochemical staining graphs (B) and optical density analysis (C) of IBA1 in hippocampus in mice showed the activation of microglia. Representative immunohistochemical staining graphs (D) and optical density analysis of CD86 (E) and CD163 (F) in hippocampus in mice showed the M1 or M2 status of microglia. Scale bars were 1000 μm (2×), 800 μm (4×), 500 μm (6×), 250 μm (10×), and 25μm (80×) in representative graphs. Data the mean standard the groups.


Introduction
Perioperative neurocognitive disorders (PND) are the most common central nervous system complication that occurs during the perioperative period, manifested as postoperative changes in social and cognitive ability, even longer-lasting cognitive decline, especially after cardiac and orthopedic surgery [1]. Compared with young patients (36.6%), elderly patients (41.4%) had a higher incidence of postoperative cognitive dysfunction at hospital discharge and higher risk of long-term cognitive impairment [2]. As global aging situation intensi es, more elderly patients will inevitably be exposed to surgical treatment. PND imposes a heavy burden on individuals and society, including prolonged hospital stay, increased mortality, decreased quality of life of patients after surgery, and consumption of medical resources [3,4].
Neuroin ammation is considered to be one of the important mechanisms leading to PND, in which microglia play a key role [5]. Preclinical and clinical studies have proved that microglia mediate postoperative neuroin ammation and cognitive dysfunction in patients and rodents [6][7][8][9]. Perioperative microglial depletion by the colony-stimulating factor 1 receptor inhibition or minocycline signi cantly protected mice from PND [6,7]. However, there are currently few approved drugs for clinical treatment of PND. Therefore, effective treatment options were urgently needed.
Rhodiola rosea is one of the tonic herbs in traditional Chinese medicine and have long been used for antidepressant, anti-fatigue and anti-stress in Asia and Europe [10,11]. Studies report that Rhodiola rosea has bene cial effects for the treatment of cognitive dysfunction through the suppression of oxidative stress and neuroin ammation in brain [12]. Recent study has demonstrated that the extract of Rhodiola rosea inhibited the neuroin ammation in response to stress through mitogen-activated protein kinases (MAPK) and nuclear factor kappa B (NF-κB) signaling pathway [13]. Salidroside is the main active ingredient of the Rhodiola rosea, with multiple functions, including cardiopulmonary protection, antifatigue, pain relief, anti-in ammation, and neuroprotection [12,14]. For example, salidroside reduced the expression of in ammatory cytokines induced by lipopolysaccharide (LPS), ultraviolet or interleukin-1β (IL-1β) in different cells [15][16][17]. In an animal model of spinal cord injury, salidroside promoted motor function recovery by reducing the in ammatory response, and its mechanism involved several signaling pathways such as adenosine 5'-monophosphate-activated protein kinase (AMPK), NF-қB, and extracellular-signal regulated kinase (ERK)/MAPKs [18,19]. Recently, accumulating researches have shown that salidroside showed a neuroprotective effect against Aβ-induced neurotoxicity and improved behavioral performance in drosophila and mouse Alzheimer's disease (AD) models [20,21]. Salidroside reduced cerebral infarction and improve neurological function in animal models of cerebral ischemia, and prevent cognitive dysfunction caused by cerebral hypoxia or hypoperfusion, which was attributed to modulating the phosphoinositide 3-kinase (PI3K), NF-κB, MAPK, hypoxia-inducible factor α (HIFα) or matrix metalloproteinases (MMPs) signaling pathway [22]. Salidroside improved LPS-induced learning and memory impairments and neuroin ammation by SIRT1-dependent nuclear factor erythroid 2-related factor 2 (Nrf-2) / heme oxygenase-1 (HO-1) /NF-κB pathway [23]. However, few studies have reported the effect of salidroside on orthopedic surgery-induced cognitive dysfunction.
Evidence suggests that the risk of fractures increases sharply in elder [24]. The fracture repair or hip arthroplasty is one of the most common surgical operation for the elderly, and orthopedic surgery is one of the main causes of PND [25]. Therefore, an orthopedic surgery mouse model of open tibial fracture with intramedullary xation was used to investigate potential therapeutic effect of salidroside on PND here. In this study, we aim to systemically examine the inhibitory effects of salidroside on cognitive dysfunction and neuroin ammation in mice received tibia fractures with intramedullary xation, which meeting the characters of clinical surgery such as fracture repair, and further explore the underlying mechanisms.

Molecular docking
Schrödinger (2018) was used to conduct molecular docking between AMPK and salidroside to explore the binding mode. Firstly, crystal structure of AMPK from the (PDB:4CFE) was get from https://www.rcsb.org/. The structure of AMPK was imported into the Protein Preparation Wizard module to optimize protein structure, for removing all crystal water molecules, correcting side chains with missing atoms, adding hydrogen atoms. Then using the OPLS3 force eld to distribute protonated states and partial charges. The crystal structure energy was minimized until the root mean square deviation (RMSD) of the non-hydrogen atoms reached 0.3 Å. After that, the LigPrep module was used to optimize 11 small ligand molecules to generate ionization and tautomerism states between pH 6.8 and 7.2. The OPLS3 force eld was again used to minimize the energy of small molecules, the maximum number of stereoisomers was set to 32. Finally, the Receptor Grid Generation module was used to generate lattice les for docking. Glide module was used for molecular docking, and Extra Precision (XP) was used for analyzing. The docking score was recorded to assess the docking effect between salidroside and AMPK.
Animals and orthopedic surgery model Adult male C57BL/6 mice (15 months of age) were provided by the Experimental Animal Center at Nanjing Medical University, Nanjing, China. All experimental protocols and procedures were approved and licensed by Nanjing Medical University Animal Care and Use Committee in accordance with the National Institutes of Health guide for the care and use of Laboratory animals.
Orthopedic surgery was imitated by the model of tibial fracture with intramedullary xation [26]. The mice were anesthetized with sodium pentobarbital (35 mg/kg). The left hind limb of the mouse was shaved and disinfected. After the tibia was exposed, a 0.3 mm pin was inserted into the tibial medullary cavity.
Then the fracture occurred, and the skin was sutured with 4.0 nylon. Lidocaine cream was used locally for analgesia. The mice in sham group received all perioperative operations, except the tibial fracture and intramedullary xation. The mice in surgery + salidroside group were intraperitoneally injected with different concentrations of salidroside (20, 40, 80 mg/kg, once a day) for 5 days before surgery.

Morris water maze (MWM) test
The MWM test was performed as described previously [27]. Brie y, a transparent cylindrical escape platform with a diameter of 10 cm was placed in one quadrant of 1.25 m-diameter circular water pool lled with water, and the platform was 1 cm below the water surface. The latency was recorded, which is the time it takes for the mouse to enter the water and climb onto the platform. If the mouse did not nd the platform within 60 s, it would be guided to the platform, stay for 20 s, and then be taken out. The experiment was carried out at the same time every day for 5 consecutive days. On the 6th day, the platform was removed to start the probe test, and the time for the mouse to enter the target quadrant and the crossing number of the position of the platform within 60 s was recorded.
Section and staining of brain tissue After the mice were sacri ced, their brains were taken out and xed in 4% paraformaldehyde for 24 h.
After the xed mouse brain tissue was dehydrated, it was soaked in xylene and placed in para n melted at 65°C for 4 h. Then, the mouse brain tissue was prepared into a wax block using a para n embedding machine. Mouse brain tissue was sliced into 5 μM slices using a para n rotary microtome. After the para n sections were baked, they were depara nized and then soaked in xylene for 10 min. After washing and soaking with absolute ethanol, the slices were soaked in ethanol solution of decreasing concentration until placed in water.
As for hematoxylin-eosin (HE) staining, mouse brain tissue slices were covered with hematoxylin staining solution, washed after 10 min, differentiated in 1% hydrochloric acid-ethanol solution, and fully turned blue in dilute ammonia. The slices were washed again and covered with eosin staining solution for 1 min.
The excess staining solution on the brain tissue was washed away. After the slices were rapidly dehydrated, they were soaked in xylene for 5 min, and then mounted with a neutral resin slicing solution. The slices were placed under an optical microscope for observation and photographs were taken.
As for immunohistochemical and immuno uorescence staining, mouse brain tissue slices were soaked in 3% hydrogen peroxide solution for 20 min. After washing with PBS, the slices were immersed in a 0.01M sodium citrate solution with a pH of 6.0 after heating to 95°C for 20 min. After natural cooling, the slices were covered with 5% BSA. After 20 min, the BSA solution was removed.
As for immunohistochemical staining, the slices were incubated with the primary antibody, IBA1 (1:100), CD86 (1:100), CD163 (1:100), and incubated overnight at 4°C in the refrigerator. The next day, after washing with PBS, the slices were incubated with diluted HRP-conjugated secondary antibody (1:100) for 2 h. After washing again, the slices were developed with DAB color reagent, and washed with water immediately after full color development. The brain tissue was covered with hematoxylin staining solution, washed after 10 min, and then fully returned to blue in dilute ammonia after differentiation with 1% hydrochloric acid-ethanol solution. The slices were washed again, and after rapid dehydration, the slices were soaked in xylene for 5 min, and then mounted with a neutral resin slicing solution. The slices were placed under an optical microscope for observation and photographs were taken.
As for immuno uorescence staining, the slices were incubated with the rst antibody, goat-IBA1 (1:100) and rabbit-AMPK (1:100), or goat-IBA1 (1:100) rabbit-PPARγ (1:100), overnight at 4°C. On the second day, after the slices were washed with PBS, they were incubated with TRITC-labeled rabbit-anti-goat IgG and FITC-labeled goat-anti-rabbit-IgG for 2 h. After washing, the slices were incubated with DAPI staining solution for 5 min. After washing, the slices were mounted with glycerin gelatin mounting solution. The expression of IBA1 and AMPK or PPARγ was observed with a uorescence microscope, photographed, and analyzed.
Cell culture and viability measurements BV-2 cells were cultured in DMEM containing 10% fetal bovine serum, 100 U/mL penicillin, and 100 mg/mL streptomycin in a 37℃ incubator containing 5% CO2. The cells were treated with LPS (1 μg/ml) with or without salidroside (25, 50, and 100 μM) for 12 h. Cell extracts and supernatants were collected and analyzed.
As for cell viability, after the cells were treated with or without salidroside (25, 50, and 100 μM) for 12 h, LPS (1 μg/ml) was used to induce the BV-2 cells for 24 h, after that the CCK-8 reagent (10 ul) was added to each well. After 4 h, the absorbance value of each well was measured at 450 nm using a microplate reader. According to the formula, cell viability= (OD value of detection well/mean OD of control wells) *100.

Quantitative real-time reverse transcription-polymerase chain reaction (RT-qPCR)
For the brain tissues of the sacri ced mice and the cells collected after treating with or without salidroside for 12 h followed by LPS for 24 h, Trizol reagent was used to obtain the total RNA. After the addition of isopropanol, the supernatant was collected, and the RNA was precipitated with ethanol, and then the mRNA of mouse brain tissue and BV-2 cells was reverse transcribed into cDNA using a reverse transcription kit. The primers of IL-1β, IL-4, IL-6, IL-10, IL-12, TNF-α, IFN-γ, CCL5, CXCL1, CXCL10, CD16, TGF-β, Arg1, YM1, and GAPDH were provided in table 1. These primers were synthesized and provided by Biotech Bioengineering (Shanghai) Co., Ltd. The method was carried out with reference to the instructions of One Step TB Green® PrimeScript™ RT-PCR Kit II (Cat. No. RR086A/B, Baori Biotechnology (Beijing) Co., Ltd., Beijing, China), and the gene expression was analyzed using the ABI7500 system.

Enzyme-linked immune sorbent assay (ELISA)
After the cells were treated with or without salidroside for 12 h, LPS (1 μg/ml) was used to induce the BV-

Western blot
For the brain tissues of the sacri ced mice and the cells collected after treating with or without salidroside for 12 h followed by LPS for 24 h, after lysis with Protein Lysis Buffer, the supernatant was collected after centrifugation at 12000 g/min for 10 min. After the sample was prepared, it was added to the SDS-PAGE gel and the proteins were separated by electrophoresis. The protein was transferred to the PVDF membrane, and after being fully blocked, PVDF membrane was incubated with the diluted antibodies against AMPK (1:2000), p-AMPK (1:1000), PPARγ (1:2000), and GAPDH (1:10000) at 4°C overnight. After the PVDF membrane was washed, it was incubated with the HRC-conjugated Goat Anti-Rabbit IgG for 2 h. Then a gel imaging system and a chemiluminescence kit were used to obtain the western blot photos.
Nuclear translocation of NF-κB p65 BV-2 cells were seeded into a 6-well plate with glass slides for cell climbing. After the cells were treated with or without salidroside for 12 h followed by LPS for 24 h, all cells were xed with paraformaldehyde.
After washing, the cells were covered with diluted anti-NF-κB p65 (1:100) antibody and incubated overnight. After washing the cells with PBS, the cells were covered with diluted Fluorescein (FITC)conjugated Goat Anti-Rabbit IgG (1:100) and incubated for 2 h. After multiple washing, the cells were stained with DAPI and incubated for 5 min. After washing the BV-2 cells for at least 3 times, the cell slide was taken out, and the cell was covered on the glass slide using glycerin gelatin mounting solution. A uorescence microscope was used to take pictures of cells.

Statistics
Image pro plus6.0 was used to analyze immuno uorescence and immunohistochemistry pictures and export measurement data. Quantity One was used to analyze the blot data and perform relative quantitative analysis. After all the data were derived from the measurement data, the SPSS 23.0 statistics were used to perform a one-way analysis of variance on the measurement data rst logarithm, and the measurement data are expressed as the mean ± standard deviation. P<0.05 indicated that the difference was statistically signi cant. Bonferroni's posttest conducted the post hoc test. GraphPad Prism 7.0 (San Diego, CA, USA) was used to graph the measurement data.

Results
Salidroside ameliorates orthopedic surgery-induced memory de cits First, the effect of salidroside on cognitive impairment caused by orthopedic surgery was tested. The Morris Water Maze (MWM) test is used to evaluate the locomotor ability and cognitive behavior of mice. The results showed that the swim speed in 7 to 12 days after orthopedic surgery between the groups has no signi cant different ( Figure 1A), indicating that there is no signi cant difference in the locomotor activities of each group mice. Mice in the surgery group showed a signi cantly increased escape latency compared to the sham group ( Figure 1B). In the probe test, the platform crossing time and the time spent in the target quadrant of the mice in the surgery group were signi cantly reduced, compared with mice in the sham group ( Figure 1C-E). The salidroside treatment signi cantly improved the above behavioral performance in a dose-dependent manner (20, 40 and 80 mg/kg), including the latency, platform crossing time and the time spent in the target quadrant ( Figure 1B-E). We then assessed the histological changes of the mice using HE staining. As shown in Figure 1F, no abnormalities in brain cells were observed in the sham group with or without 80 mg/kg salidroside treated. In contrast, nuclear pyknosis was observed in the surgery group. Moreover, orthopedic surgery mice treated with salidroside showed signi cantly fewer abnormal cells compared to the surgery group.

Salidroside inhibits orthopedic surgeryinduced neuroin ammation in mice
Then, we tested the effect of salidroside on orthopedic surgery-induced neuroin ammation by detecting the expression of in ammatory factors and the activation state of microglia in the hippocampus of mice. The mRNA levels of pro-in ammatory cytokines TNF-α, IL-1β, IL-6 and anti-in ammatory cytokine IL-10 in the hippocampus were detected by RT-qPCR. The results showed that the levels of TNF-α, IL-1β, and IL-6 were signi cantly increased and the levels of IL-10 was signi cantly decreased in the hippocampus of mice after surgery, while the administration of salidroside (80 mg/kg) in orthopedic surgery mice observably prevented the above changing of mRNA expression in in ammatory factors ( Figure 2A). Further, we used immunohistochemical staining to detect the activation and status of microglia, and the results showed that orthopedic surgery induced prominent IBA1 expression in the hippocampus of mice, while salidroside signi cantly reduced the IBA1 expression in surgery mice ( Figure 2B-C). Salidroside signi cantly reduced the expression of CD86 in the hippocampus, indicating that salidroside treatment reduced the number of M1 microglia in orthopedic surgery mice ( Figure 2D-E). We also performed immunohistochemical staining on the M2 microglia marker CD163, which was increased after salidroside treatment ( Figure 2D and 2F). These data indicate that salidroside inhibited neuroin ammation and induced M2 microglia polarization.
AMPK and PPARγ participate in the neuroprotective effect of salidroside in PND mice Next, we examine the possible mechanism of salidroside on the neuroprotection of PND. It has been reported that AMPK and peroxisome proliferator-activated receptor γ (PPARγ) were involved in microglia-mediated neuroin ammation [28,29]. Recent studies have indicated that AMPK signaling is related to the bene cial effects of salidroside in diabetes [30]. Therefore, we hypothesized that salidroside might suppress orthopedic surgery-induced neuroin ammation and cognitive impairment by activating AMPK/PPARγ signaling pathway. Molecular docking con rmed that salidroside formed hydrogen bonds with amino acids H323 and C285 of AMPK, which may partially extract the activation capacity on AMPK, and thus initiating APMK phosphorylation. The docking score between salidroside and AMPK was -8.074 (Fig.3A). Further, AMPK phosphorylation in mouse hippocampus was signi cantly increased after salidroside treatment ( Figure 3B-C). The immuno uorescence results also showed that salidroside remarkably activated AMPK, which is co-localized with IBA1 + microglia ( Figure 3F-G). In addition, the results of western blot showed that salidroside promoted the expression of PPARγ ( Figure 3D-E), which is consistent with the results of immuno uorescence staining, and the activation of PPARγ was co-localized with IBA1 + microglia ( Figure 3H-I). These data indicate that salidroside could activate the AMPK and PPARγ in microglia in mouse hippocampus.
Salidroside inhibits LPS-induced in ammatory response in BV-2 cells  Figure  5A-B). Two clinically commonly used drugs that induce AMPK and PPARγ activation respectively, metformin and rosiglitazone, were used as positive controls. Like salidroside, they both increased the protein expression levels of p-AMPK and PPARγ in LPS induced BV-2 cells ( Figure 5C-D). In LPS induced BV-2 cells, the secretion of TNF-α, IL-6, IL-1β, and IL-10 were also affected by salidroside, metformin and rosiglitazone, they all reduced TNF-α, IL-6, and IL-1β level, and increased IL-10 level in cell culture supernatant ( Figure 5E-H). These results indicated that AMPK and PPARγ activation reduced microglia in ammation, and AMPK and PPARγ were involved in the effect of salidroside on microglia.
Next, we tested the effect of salidroside on the in ammation and differentiation of microglia under the condition of AMPK or PPARγ inhibition. As show in Figure 6A-B, salidroside signi cantly increased AMPK phosphorylation and PPARγ expression in BV-2 cells, the promoting effect of salidroside on PPARγ activity was blocked by AMPK inhibitor compound C, but promoting effect of salidroside on AMPK phosphorylation had no signi cant difference in LPS induced BV-2 cells with or without PPARγ antagonist GW9662 treatment ( Figure 6A Figure 6I, salidroside, rosiglitazone, or metformin decreased the expression of LPS-induced M1 phenotypic marker genes and increased the expression of M2 marker genes. The PPARγ antagonist GW9662 and AMPK inhibitor compound C prevented salidroside-mediated decrease in M1 marker gene expression and increase in M2 marker gene in LPS-induced BV-2 cells ( Figure 6I). These results indicated that AMPK and PPARγ activation reduced microglia in ammation, and salidroside inhibits LPS-induced in ammatory response of microglia and promotes its polarization to the M2 phenotype via activating AMPK/PPARγ pathway.
The effect of salidroside in improving cognitive impairment in mice depends on AMPK Finally, we veri ed in vivo whether the neuroprotective effect of salidroside depends on AMPK. The MWM test is used to evaluate the locomotor ability and cognitive behavior of mice. The results showed that the swim speed in 7 to 12 days after orthopedic surgery between the groups has no signi cant different ( Figure 7A). Salidroside signi cantly inhibited the escape latency of mice in surgery mice and increased the platform crossing numbers and time spent in target quadrants, which could be canceled by AMPK inhibitor compound C ( Figure 7B-E). The immunohistochemical staining was used to detect the PPARγ expression and activation of microglia. The results showed that AMPK inhibitor compound C canceled the suppress effect of salidroside on IBA1 expression, and canceled the promotion effect of salidroside on PPARγ expression in the hippocampus of surgery mice ( Figure 7F-H), which is consistent with in vitro results. In addition, salidroside decreased the expression of M1 phenotypic marker genes (IL-1β, IL-12, TNF-α, IFN-γ, CCL5, CXCL1, CXCL10, and CD16) and increased the expression of M2 marker genes (IL-10, IL-4, TGF-β, Arg1, and YM1) in orthopedic surgery-induced cognitive impairment mice, which was abolished by AMPK inhibitor compound C ( Figure 7I). These results indicate that salidroside improved the cognitive impairment of orthopedic surgery mice by AMPK/PPARγ signaling pathway.

Discussion
Herein, we demonstrated that salidroside prevents PND by inhibiting neuroin ammation in a mouse model of tibia fractures with intramedullary xation. Salidroside shows a further bene cial effect by promoting the polarization of microglia to the anti-in ammatory phenotype. In addition, we also analyze the potential signaling mechanism of salidroside to exert its bene cial effects: promote the activation of AMPK/PPARγ. Salidroside activated the AMPK and PPARγ in the hippocampus of mice and BV-2 cells.
More importantly, the effects of salidroside that mitigated the cognitive de cit in mice induced by orthopedic surgery and promoted PPARγ activation and microglia polarization were abolished by AMPK antagonist. Taken together, salidroside prevented surgery-induced neuroin ammation and cognitive de cits via the AMPK/PPARγ signaling pathway and might be a novel therapeutic approach for PND.
Recently, an increasing number of studies have been conducted on the neuroprotective effects and cognition enhancement of salidroside [14]. It was demonstrated that salidroside ameliorated neurodegeneration and memory impairment in hypoxic rats and Aβ1-40-injected AD rat model [31,32].
Salidroside has shown protective effects in behavior tests and hippocampal long-term potentiation tests in rats with cognitive de cits caused by chronic cerebral hypoperfusion. Salidroside also ameliorated arthritis-induced cognition dysfunction by decreasing pro-in ammatory cytokines in hippocampus and serum [33]. However, the underlying molecular mechanism remains unclear. In the present study, we found that salidroside prevented the cognitive function of orthopedic surgery mice (Fig. 1), which were in line with the ndings of another study which demonstrating that salidroside was effective against in iso urane-induced learning and memory impairment [34]. And it was suggested that the underlying mechanism of the effect of salidroside was related to the inhibition of excessive in ammatory response, decrease of oxidative stress, and regulation of the cholinergic system [34]. Our research proved that the neuroprotective effect of salidroside depends on the activation of AMPK/PPARγ signaling pathway.
Increasing evidence has shown that the AMPK signaling pathway plays a crucial role in the pathophysiology of neurodegenerative diseases [35,36]. AMPK is a Ser/Thr kinase which has a crucial role in the maintenance of energy metabolism and anti-in ammation. AMPK activators have shown bene cial effects in AD models, such as improved cognitive de cits, reduced Aβ production in a transgenic AD mouse model [37]. Moreover, AMPK activation in brain increased microglial M2 gene expression in an LPS-induced neuroin ammation mouse model [38]. Consistent with these studies, our results show that the AMPK agonist metformin signi cantly improved cognitive de cits after surgery and reduced the LPS-induced in ammatory response in BV-2 cells, which mimic the effect of salidroside.
Moreover, PPARγ has been found to play an essential role in cognitive performance [39]. PPARγ activation by rosiglitazone (RSG) was shown to improve the age-dependent decrease of LTP and neurocognitive de cits depending on aging in older animals [40]. Also, activation PPARγ could modulate the activation of microglia and suppress NF-κB-mediated pro-in ammatory gene transcription, which inhibited in ammatory response and was bene cial for AD [41]. In the present study, we used RSG as a positive drug, and demonstrated RSG signi cantly reduced the LPS-induced in ammatory response in BV-2 cells, consistent with the effect of salidroside and AMPK agonist metformin in LPS-induced BV-2 cells. And here salidroside inhibited the AMPK/ PPARγ signaling pathway and thereby exhibited antineuroin ammatory activity in cultured microglial cells in this study.
In addition to the suppression of neuroin ammation, our in vivo and in vitro results suggested that the therapeutic effects of salidroside may also be mediated via the switching of microglial phenotype. There are two different phenotypes of activated microglia: classically activated M1 phenotypes and alternatively activated M2 phenotypes. M1 microglia secretes pro-in ammatory cytokines, such as interleukins-1β (IL-1β), IL-6, and tumor necrosis factor-α (TNF-α), which promote in ammation; while M2 microglia secrete anti-in ammatory cytokines, such as IL-10, IL-4, and transforming growth factor-β (TGFβ), which inhibit in ammation and promote tissue repair [42]. It has been proven that promoting microglia to the M2 phenotype signi cantly improve postoperative cognition [43,44]. Thence, promoting the polarization of microglia to M2 type is an excellent strategy to improve PND. Changes in microglial phenotype were closely associated with the neuroprotective effect of salidroside. Microglial activation and the expression of in ammatory molecules are directly involved in neurodegenerative pathology and in the development of neuroin ammation. We observed an ameliorated in ammatory milieu by salidroside in microglial culture, histological assays of brain tissue from mice also showed that salidroside signi cantly attenuated microglial in ammatory activation in the hippocampus. Interestingly, we further observed that salidroside switched microglial phenotype from the in ammatory phenotype to the anti-in ammatory phenotype in vivo and in vitro. Furthermore, our results suggest that the salidrosideinduced switching of microglial phenotype towards the anti-in ammatory phenotype was mediated by AMPK/ PPARγ signaling pathway in microglia.
Although we demonstrated that salidroside ameliorated behavioral impairments and neuropathological changes in a mouse model of PND by activating AMPK/ PPARγ signaling, the speci c binding site of salidroside remains unknown. In addition, we cannot rule out the possible effect of salidroside on other cells, but we have con rmed through in vitro experiments and immuno uorescence double-labeling tests that the effect of salidroside on PND is mainly through the regulation of microglia polarization.

Conclusions
In summary, we demonstrated that salidroside prevented surgery-induced neuroin ammation and cognitive de cits in the mouse model of PND. Notably, salidroside induced the polarization of microglia to an anti-in ammatory phenotype through regulating the AMPK/PPARγ signaling pathway. These results suggest that salidroside may be a novel therapeutic approach for PND.

Consent for publication
Not applicable.

Availability of data and material
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.