The Expression of Interleukin-1β is Regulated by DNA Methylation in Microglia to Mediate Postoperative Cognitive Dysfunction in Aged Mice

Postoperative cognitive dysfunction (POCD) is a common postoperative complication in elderly individuals. Neuroinammation is closely related to its occurrence. However, the exact molecular mechanism underlying this link is undetermined. This study aimed to establish a mouse model of POCD to explore the role of DNA methylation in regulating the expression of interleukin-1β (IL-1β), which mediates the occurrence of POCD in aged mice. The POCD model was established by exploratory laparotomy and evaluated by novel object and Y-maze tests. We also assessed IL-1β production in the dorsal hippocampus and the expression of the DNA methylation-related proteins DNA methyltransferase 3a (DNMT3a), DNA methyltransferase 3b (DNMT3b), and methyl CpG binding protein 2 (MeCP2). Methylation specic PCR (MSP), methylated DNA immunoprecipitation (MeDIP) and DNA methylation sequencing in IL-1β promoter were used to explore the regulation of IL-1β by DNA methylation in this model. Finally, Golgi-Cox staining and Western blotting were used to further explore the role and potential mechanisms of IL-1β in POCD. Cognitive impairment was observed in aged but not adult mice. In aged mice, the microglia cells in the dorsal hippocampus were activated, while the DNA methylation in the IL-1β promoter was decreased. Interestingly, the global DNA methylation in the dorsal hippocampus was unchanged. IL-1β inhibition prevented surgery-induced cognitive decline and dysfunction of synaptic plasticity. Overall, these results indicated that DNA methylation regulation of IL-1β expression may be an important mechanism increasing the susceptibility to POCD. C57/BL mice (3 m and 18 m) were subjected to laparotomy in a chamber lled with 1.4-2% isourane for 20 min. Open eld, Y-maze and novel object tests were performed at different time points. b The mice showed a normal total distance traveled in the open eld test. c-d The DI of aged mice was decreased in the OLM task but not the ORM task in the 1d group. The spontaneous alternation index (e) and the time spent in the novel arm (f) in the Y-maze test were decreased in aged mice in the 1d group. g-j Simple anesthesia had no effect on the cognitive function of adult and aged mice. n = 9. The data are presented as the mean ± SEM values. **P < 0.01, ***P < 0.001 versus the corresponding group.


Introduction
Postoperative cognitive dysfunction (POCD) is a clinical phenomenon that is characterized by disturbances in memory, intellectual ability and executive function which adversely affects quality of life, social dependence and mortality [1]. The incidence of POCD in middle-aged patients is 19.2% [2] and up to 52% in aged patients [3,4]. Increased susceptibility to POCD has also been found in aged animals [5][6][7][8]. However, the molecular mechanism for the high incidence of this complication in elderly individuals remains controversial.
During aging, excessive release of interleukin-1β (IL-1β) and microglial priming are usually considered the main causes of POCD, as IL-1β and microglia are at the nexus of crosstalk between neuroin ammation and abnormal neuronal function and synaptic injury [9,10]. IL-1β, IL-1β receptors and the naturally occurring interleukin-1 receptor antagonist (IL-1ra) are expressed at comparatively high levels in the hippocampus, a brain structure critical for spatial and contextual memory [11,12]. However, the mechanisms by which anesthesia and surgery cause excessive release of IL-1β in elderly individuals and by which aberrantly elevated levels of IL-1β in the hippocampus may impair synaptic plasticity and memory are unclear.
It's established that epigenetic regulation plays a key role in in ammatory and cognitive behavioral changes. DNA methylation is a reversible epigenetic mechanism associated with gene silencing [13,14]. This modi cation is catalyzed by DNA methyltransferases (DNMTs), including DNMT1, DNMT3a, and abdominal cavity through the skin, abdominal muscles, and peritoneum. Then, a 5-cm segment of the small intestine was pulled out of the abdominal cavity and gently massaged for 30 s. Finally, sterile 4-0 chromic gut sutures were used to suture the incision layer by layer. The operation time for each mouse was approximately 20 min, during which the body temperature of the mouse was monitored and kept between 36°C and 37°C by a heating pad. Simultaneously, the operator monitored the concentration of iso urane and ow of oxygen, maintaining a pinkish hue in the mouth, nose, and limbs of the mouse and a breathing rate of 70-80 breaths/min. After the mice recovered from anesthesia/surgery, they were returned to their home cages. Compound lidocaine cream (2.5% lidocaine and 2.5% prilocaine) was applied locally to treat postoperative pain.

Open eld test (OFT)
An open eld apparatus consisting of a black Plexiglas chamber with a white oor (35 cm × 23 cm × 23 cm) was positioned in a dimly lit room. The OFT was used to measure the locomotor activity of the mice at different time points, as we described previously [20]. In brief, mice were placed individually into the center of the apparatus and were allowed to freely explore the environment for 5 min. The total distance traveled by the mouse was automatically recorded with ANY-maze software (ANY-maze, Stoelting Co., IL, USA). The apparatus was wiped with 75% ethanol after each test to avoid olfactory cues (animal scent, stool, urine, etc.).

Object location memory (OLM) and object recognition memory (ORM) tasks
The novel object test apparatus was adjusted according to the method described by Annie Vogel-Ciernia [21][22][23][24][25][26]. The novel object test was performed using a 35 × 23-cm open arena with 23-cm-high walls. For habituation, a researcher transported the mouse to the chamber and lowered it to the bottom. Then, the researcher gently turned his or her hand so that the mouse could step out of his or her hand into the chamber and allowed the mouse to explore freely for 30 min. During the training session, the mouse was exposed to two familiar objects and allowed to explore them for 15 min. Then, the mouse was returned to its cage. For the OLM task, one of the two familiar objects was moved to a new location 2 hours after the training session. For the ORM task, the object locations remained constant, but one of the objects was replaced with a new item. After the object was exchanged, the mouse was allowed to freely explore the area for 10 min. An exploration was scored when the mouse oriented its head toward the object and came within 1 cm or when the nose of the mouse touched the object. Preference for the novel item was expressed as a discrimination index (DI) = (t novel -t familiar) / (t novel + t familiar). Mice that explored both objects for less than 2 s during testing or 3 s during training were excluded from further analysis.
Mice that showed a preference for one object during training (DI > ± 0.2) were also excluded.

Y-maze test
The Y-maze consisted of three opaque arms (8 × 30 × 15 cm, width × length × height) oriented 120° with respect to each other. Mice were handled for 2 min/day for 4 days. For the spatial reference memory test, the mouse was allowed to explore the maze with the novel arm blocked for 15 min during training. After a 2-h interval, the mouse was placed back into the maze with the block removed, and the time spent in the novel arm was recorded. For the spontaneous alternation test, the mouse was allowed to explore the maze for 15 min during training. After a 2-h interval, the mouse was allowed to explore the maze undisturbed for 8 min, and the numbers of all arm entries and alternations were simultaneously recorded. An alternation was de ned as a consecutive entry into all three arms. The alternation percent (%) was calculated using the following formula: alternation % = number of alternations / (total number of arm entries -2) × 100 [27].
Enzyme-linked immunosorbent assay (ELISA) Dorsal hippocampal homogenate was obtained on days 1, 3, and 7 after surgery and day 1 after iso urane anesthesia. The homogenate was centrifuged at 12000 × g and 4°C for 15 min to collect the supernatant. IL-1β (ABclonal, RK00006, USA) was measured according to the manufacturer's protocols. Immuno uorescence staining Mice were perfused transcardially via the left ventricle with 0.9% saline followed by 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Brain tissues were harvested, post xed with 4% paraformaldehyde overnight and dehydrated in 30% sucrose, as we reported previously [29]. Sections (35 µm thick) were sliced with a freezing microtome (Leica CM1950, Germany). The brain tissue was blocked with 10% normal donkey serum in phosphate-buffered saline supplemented with Tween 20 (PBST) for 1 h at room temperature and was then incubated with primary antibodies against Iba1 overnight. The brain tissue was then incubated with donkey anti-rabbit Alexa Fluor 594 (1:400; A21207; Invitrogen, USA) and donkey anti-mouse Alexa Fluor 488 (1:400; A32766; Invitrogen, USA) as the secondary antibodies at room temperature in the dark for 2 h. Then, the sections were mounted with DAPI Fluoromount-G (Southern Biotech, USA). The uorescence was visualized under an FV-1000 confocal uorescence microscope (Olympus, Japan). For each animal, the uorescence intensities from three slides (three visual elds per slide) were averaged [30].
Golgi-Cox staining and dendritic spine counting Golgi-Cox impregnation is an effective method for studying the morphology of neurons and visualizing structural synaptic plasticity [31]. An FD Rapid GolgiStain TM Kit (FD Neurotechnologies, PK401, Columbia) was employed according to the manufacturer's instructions. In brief, mice were deeply anesthetized by iso urane and rapidly sacri ced. The brains were removed as quickly as possible, washed with double-distilled water, immersed in impregnation solution (a mixture of solutions A and B), and stored in the dark at room temperature for 14 days. Next, the samples were transferred to solution C and stored at 4°C in the dark for at least 72 h. Finally, the brains were sliced at a thickness of 150 µm with an oscillating tissue slicer (chamber temperature -22°C), stained and mounted on gelatin-coated slides. After alcohol dehydration, the tissue sections were cleared in xylene and coverslipped. The dendrites from hippocampal neurons in the CA1 region were imaged with a confocal microscope (100× oil objective). The dendritic spine density was determined along CA1 secondary dendrites starting from their point of origin on the primary dendrite, and counting in each sample was performed by an experimenter blinded to the group allocation.

Dot blot analysis
Total DNA (250 ng, 3 µl) was diluted with incubation buffer (65.7% ammonium acetate, 7.77% formaldehyde (37-40%), and 6.8% MOPS) to 9 µl and incubated at 65°C for 5 min. Then, 12 µl of 20× saline-sodium citrate was added. The above mixture (5 µl) was spotted onto a Hybond-N + membrane (GE Amersham, RPN303B, UK). DNA was hybridized to the membrane by a 10 min incubation with UV crosslinking. Then, the membrane was immersed in a solution of methylene blue and mixed liquor (0.02% methylene blue and 0.3 M sodium acetate) for 5-10 min. The membrane was then incubated with a 1:1000 dilution of an anti-5-methylcytosine (5mC) antibody (Active Motif, 61255, USA) at 4°C overnight. After three rounds of washes with 1× PBST, the membrane was incubated with the corresponding secondary antibody for 2 h. Finally, an ECL detection system (Beyotime) and ImageJ were used for data analysis and statistical calculations [32].
Cell culture and treatment, lentivirus production and lentiviral transduction For culture of BV2 microglia and HEK293T cells, the medium was changed every other day using a 1:1 mixture of fresh DMEM (Gibco, RNBJ5984, USA) containing 10% fetal bovine serum (FBS; SE OU Biology, C100-900, China). The cells were incubated at 37°C in a humidi ed incubator with 5% CO 2 / 95% air and used for experiments after treatment. For lentivirus production and lentiviral transduction, the DNMT3agreen uorescent protein (GFP) overexpression and empty-GFP plasmids were rst constructed by JingMai (Nanjing JingMai Co., Ltd., Nanjing, China). The constructed core plasmid (16 µg) and the packaging and envelope plasmids psPAX2 (12 µg) and pMD2.G (4.8 µg), respectively, were cotransfected into HEK293T cells in a 6-well plate according to the manufacturer's instructions for Lipofectamine 6000 (Beyotime, C0526, China). Then, the supernatant was collected 48 h after transfection and concentrated through a Centricon Plus-70 lter unit (UFC910096, Millipore, USA). Lentivirus with a titer of 10 8 TU/ml was used in the experiment. Lentiviral transduction into BV2 microglia was performed as described in a previous study [33].

Microglia isolation
Animals were anesthetized with iso urane after treatment, and the dorsal hippocampi were harvested on ice and used immediately for microglial isolation via a procedure adapted from Astrid E. Cardona [34]. To ensure that enough cells were recovered, dorsal hippocampi from 2 mice from a given experimental group were pooled. Samples were enzymatically digested using a Papain Dissociation System (Worthington, LK003150, USA). Tissue debris was removed by passing the cell suspension through a 40-mm cell strainer. After myelin removal using 30% Percoll (VICMED, VIC1555, China), cells in PBS supplemented with 0.5% bovine serum albumin (BSA) were incubated for 30 min with an anti-CD11b-APC antibody (5 µl, 17-0112-82, eBioscience, China) and an anti-CD45-FITC antibody (5 µl, 11-0451-82, eBioscience, China) for ow cytometry. The number of isolated CD11b + /CD45 low microglia was approximately 1×10 6 cells per pool of 2 brains and did not differ between the treatment groups. and a nal extension step at 72°C for 10 s followed by real-time melt curve analysis to verify product speci city. In addition, agarose gel electrophoresis was used to verify product speci city, and the raw image is shown in supplemental Methylation-speci c real-time PCR (MSP) Sodium bisul te modi cation and MSP were used to analyze DNA methylation changes in promoter regions. Bisul te modi cation of genomic DNA converts unmethylated cytosine residues into uracil residues. In contrast, methylated cytosine residues remain unconverted. Sodium bisul te modi cation of DNA was performed using a CpGenome TM Turbo Bisul te Modi cation Kit (Millipore, 2906402, USA) according to the manufacturer's protocol. PCR ampli cation was carried out according to the instructions of the Episcope MSP Kit (TaKaRa, R100A, China). The primer pairs targeted a methylated and an unmethylated CG dinucleotide in DNA associated with the promoter region of IL-1β or unmethylated βtubulin-4 as a reference gene, as previously published [37]. The primer sequences were as follows: IL-1β methylated, F: 5′-TTTTAGTTTAAGTATAAGGAGGCGA-3′ and R: 5′-ACACATTCGCAAATATATCATCGTA-3′; IL-1β unmethylated, F: 5′-TTTTAGTTTAAGTATAAGGAGGTGA-3′ and R: 5′-AACACATTCACAAATATATCATCATA-3′; β-tubulin-4 unmethylated, F: 5′-GGAGAGTAATATGAATGATTTGGTG-3′ and R: 5′-CATCTCCAACTTTCCCTAACCTACTTAA-3′. Product speci city was determined by melt curve analysis. For MSP data analysis, the methylation index was calculated by dividing the fold change value for the methylated primer pair by the fold change value for the unmethylated primer pair as previously described [38,39].

Intracerebroventricular cannulation and administration of recombinant IL-1ra
IL-1ra functions as a competitive inhibitor of IL-1α and IL-1β, binding to the IL-1 type 1 receptor with an a nity equal to that of IL-1β and without any agonist activity [40]. Administration of IL-1ra directly into the cisterna magna has been reported to effectively block the action of IL-1β within the central nervous system (CNS), including the hippocampus [41]. To con rm the impact of blocking the central action of IL-1β before surgery, 18 m mice were randomly divided into four groups: the sham + vehicle (veh) group, sham + IL-1ra group, 1d + veh group, and 1d + IL-1ra group. An intracerebroventricular cannula was inserted into each mouse as described previously [42,43]. In brief, mice were anesthetized by intraperitoneal injection of Avertin (isoamyl alcohol, 250 mg/kg) and positioned in a stereotaxic frame so that the plane formed by the frontal and parietal bones was parallel to horizontal zero. Sequenom MassARRAY® methylation analysis of the IL-1β promoter The data are presented as the mean ± SEM values and were analyzed with GraphPad Prism 8.0 (GraphPad Software, Inc.). Statistical comparisons between two groups were made with an independent samples t test. Multiple comparisons were performed with one-way ANOVA followed by the Bonferroni test for multiple comparisons. Differences were considered signi cant when P < 0.05.

Results
Surgery but not anesthesia impaired hippocampusdependent cognition in aged mice We rst tested whether either surgery or anesthesia alone can affect cognition in aged (18 m) and adult (3 m) mice (Fig. 1a). In the open eld test, the total distance traveled did not differ signi cantly among the groups (aged: F (3, 32) = 1.383, P = 0.2657; adult: F (3, 32) = 1.593, P = 0.2103, Fig. 1b), meaning that locomotor activity was unaffected. Next, we investigated the effects of surgery on learning and memory. As shown in the novel object test, there was no signi cant preference in any group during training (Supplemental Fig. 2a, 2b). OLM testing revealed that surgery induced object-place recognition de cits in aged mice 1 day after surgery (F (3, 32) = 7.934, P = 0.0004, Fig. 1c). However, ORM was normal (F (3, 32) = 0.7731, P = 0.5176, Fig. 1d). Adult mice showed no signi cant differences among the groups in either OLM (F (3, 32) = 0.9032, P = 0.4504, Fig. 1c) or ORM (F (3, 32) = 0.3802, P = 0.7679, Fig. 1d). OLM and ORM appear to be dependent on different brain regions. OLM requires the hippocampus for encoding, consolidation, and retrieval [44,45] and is particularly sensitive to manipulations in the dorsal CA1 region [46], whereas ORM requires the participation of different brain regions, including the perirhinal cortex, insular cortex and medial prefrontal cortex [47,48]. Therefore, we think that the effect of surgery on the cognitive function of aged mice depends mainly on the dorsal hippocampus. In addition, in the Y-maze test, the spontaneous alternation index (F (3, 32) = 13.07, P < 0.0001, Fig. 1e) and the time spent in the novel arm (F (3, 32) = 9.374, P = 0.0001, Fig. 1f) were decreased in the 1d group but not in the 3d and 7d groups compared with the sham group for aged mice. In contrast, adult mice showed no signi cant In the present model, anesthesia and surgery often occur concurrently; thus, we explored the effect of iso urane anesthesia alone on the cognitive function of adult and aged mice. In the OLM task, ORM task, and Y-maze test, the cognitive function of both adult and aged mice did not change compared with that in the control group 1 day after iso urane anesthesia alone (aged: t = 0.  Fig. 1j). These results suggested that surgical trauma during the perioperative period may play a more important role than anesthesia in POCD. Collectively, these results indicated that surgery induced hippocampus-dependent cognitive decline in aged mice.

Surgery activated microglia in the dorsal hippocampus
Since IL-1β mediates the in ammatory response and affects hippocampal memory function [49], we focused on neuroin ammation induced by IL-1β in the present model. Our data showed that the mRNA expression level of IL-1β (F (4, 25) = 6.836, P = 0.0007, Fig. 2a) as well as its protein expression level (F (4, 15) = 7.163, P = 0.002, Fig. 2b) in the dorsal hippocampus were higher in the 1d group than in the sham group in aged mice, while adult mice showed no signi cant differences among the groups (F (4, 35) = 0.3623, P = 0.8337, Fig. 2a; F (4, 15) = 1.558, P = 0.2363, Fig. 2b). Anesthesia alone did not change the IL-1β levels in the 1d group and sham group. Nuclear factor-kappa B (NF-κB) is a transcription factor that plays a crucial role in the expression of proin ammatory cytokines in microglia. After translocation into the nucleus, it upregulates the expression of IL-1β [50]. We next evaluated the nuclear expression level of p65, a subunit of NF-κB, in dorsal hippocampal cells and found that it was increased 1 day after surgery, as expected, in aged mice (t=5.029, P = 0.0125, Fig. 2c).
IL-1β expression has been documented in neurons, microglia and astrocytes [51][52][53][54]. Thus, we assessed the changes in microglia and astrocytes. Immuno uorescence staining showed that ionized calcium DNMTs (DNMT3a and DNMT3b) add a methyl group to the 5-carbon of the cytosine ring, particularly in cytosine nucleotides located next to a guanine nucleotide, to form 5mC, a stable repressive regulator of promoter activity [55]. To this end, we assessed the level of 5mC by dot blot analysis. The global 5mC level in the dorsal hippocampus did not signi cantly differ in the 1d group compared with the sham group (t = 0.06296, P = 0.9528, Fig. 3c), indicating that the global level of DNA methylation in the dorsal hippocampus remained unchanged. In ammatory factors are released mainly from microglia. Thus, we next examined the regulation of gene expression by DNA methylation in microglia. Dorsal hippocampal cells from aged mice after surgery or sham operation were isolated by ow cytometry using anti-CD11b-APC and anti-CD45-FITC antibodies. These cells were identi ed as CD11b + /CD45 low and were thought to be predominantly microglia (Fig. 3d). Then, the mRNA levels of several enzymes that participate in DNA methylation were measured. Notably, the mRNA levels of DNMT3a, DNMT3b and MeCP2 were decreased in aged mice 1 day after surgery (DNMT3a: t = 32.08, P < 0.0001; DNMT3b: t = 6.409, P = 0.0007; MeCP2: t = 8.632, P = 0.0001, Fig. 3e). These results showed that alterations in epigenetic regulators (DNMT3a, DNMT3b and MeCP2) in microglia in the dorsal hippocampus were induced by surgery in aged mice.
Surgery decreased DNA methylation of IL-1β promoter in aged mice The expression of NF-κB P65 was increased in aged mice 1 day after surgery. Moreover, other studies have reported that NF-κB binding activity might play a role in IL-1β promoter demethylation [56]. Therefore, in this study, we investigated methylation sites adjacent to a canonical NF-κB binding site (CpG sites -229 bp and -215 bp with respect to the translation start site in the IL-1β promoter). As expected, the levels of DNA methylation at the above two sites were reduced by surgery in aged mice (MSP: t = 2.661, P = 0.0375; MeDIP: t = 3.753, P = 0.0095, Fig. 4a). In contrast, most other CpG sites except for the abovementioned two sites were highly methylated and not affected by surgery; we examined >10 CpG sites in the region between -1439 bp and 760 bp with respect to the ATG translation start codon (Fig. 4b).
To demonstrate whether DNA methylation plays a direct role in IL-1β expression, the BV2 immortalized murine microglial cells were transduced with lentiviral vectors expressing DNMT3a tagged with GFP (Lenti-DNMT3a-GFP) or empty vector (Lenti-Empty-GFP). Then, the cultures were harvested after treatment with LPS. BV2 cells were effectively transduced with Lenti-GFP (Fig. 4c), and the cell morphology did not change. We rst detected the expression of DNMT3a, which was signi cantly reduced by LPS treatment (t = 3.846, P = 0.0049, Fig. 4d). After transduction with Lenti-DNMT3a, the expression of DNMT3a was restored compared to that in the Lenti-Empty + LPS (-) group (t = 3.501, P = 0.0081, Fig. 4d).

IL-1β impaired synaptic plasticity after surgery in aged mice
To speci cally con rm the role of IL-1β in memory de cits, IL-1ra was injected bilaterally into the lateral cerebral ventricle to suppress the expression of IL-1β in the dorsal hippocampus (Fig. 5a). The level of IL-1β in the dorsal hippocampus of aged mice was reduced by administration of IL-1ra via the lateral cerebral ventricle just before surgery (F (3, 12) = 11.30, P = 0.0008, Fig. 5b). The Golgi-Cox staining results showed that IL-1ra injection reversed the reduction in spine density in the dorsal hippocampal CA1 region (F (3, 12) = 4.841, P = 0.0197, Fig. 5d). Regarding synaptic functional plasticity, various genes are related to IL-1β and synaptic function, for example, brain-derived neurotrophic factor (BDNF), postsynaptic density 95 (PSD95), and synaptophysin (Syn), which are involved in learning and memory. The protein expression levels of hippocampal BDNF, PSD95 and syn were signi cantly decreased in mice in the surgery groups, while IL-1ra injection similarly reversed this change (F (3, 16) = 4.045, P = 0.0256, Fig. 5f, F (3, 16) = 4.582, P = 0.0169, Fig. 5g, F (3, 12) = 4.877, P = 0.0192, Fig. 5h). Furthermore, the number of c-Fos + neurons in the CA1 region was decreased on postoperative day 1 in aged mice, but IL-1ra injection signi cantly attenuated the loss of c-Fos + neurons (F (3, 16) = 17.53, P < 0.0001, Fig. 5i). These results indicated that IL-1β impaired synaptic plasticity in the dorsal hippocampus of aged mice in the POCD model.
Inhibiting IL-1β in the dorsal hippocampus rescued memory de cits induced by surgery Next, we further investigated whether inhibiting IL-1β in the dorsal hippocampus of aged mice by bilateral administration of IL-1ra into the lateral cerebral ventricle rescues the memory de cits induced by surgery.
The above data indicated that the surgery-induced impairment of memory performance was alleviated by suppression of IL-1β.

Discussion
In the present study, we showed that aged mice but not adult mice exhibited hippocampus-dependent postoperative cognitive decline and increased gene expression of IL-1β 1 day after surgery. In aged mice, microglia in the dorsal hippocampus was activated and DNA methylation level of IL-1β promoter was decreased. The increased expression of IL-1β impaired synaptic plasticity and hippocampus-dependent memory formation in aged mice in the POCD model.
However, impairment of OLM and spatial memory did not occur with a single exposure to iso urane alone in either adult or aged mice, suggesting that a single exposure to iso urane alone would not induce memory de cits in patients undergoing abdominal surgery. Previous studies have demonstrated that iso urane exposure in aged rats leads to impaired spatial memory and neuroin ammation [57]. The discrepancies among these studies involving iso urane inhalation alone might result from the different doses and times. Consistent with this hypothesis, general anesthetics may impair developing neurons and induce cognitive dysfunction in young adult male C57BL/6 mice in a dose -and time -dependent manner [58,59]. We also found that the LPS-induced impairment of object recognition and fear memory was not aggravated by iso urane anesthesia in adult mice [29].  [70], which is consistent with our results. In addition, astrocytes undergo morphological changes that are not strictly dependent on the GFAP protein level during neuroin ammation [71], Thus, the role of astrocytes in POCD needs to be determined by further experiments.
We demonstrated that age and surgical trauma can alter the expression levels of genes important for establishing and maintaining DNA methylation, such as DNMT3a, DNMT3b and MeCP2 in microglial cells isolated from surgery-treated mice. Interestingly, the global DNA methylation in the dorsal hippocampus was unchanged suggesting that DNA hypomethylation may not be a global phenomenon.
In fact, many studies have observed an age-related increase in DNA methylation in the promoters of neuronal activity-related and synaptic marker genes [72,73] and an age-related decrease in DNA methylation in the promoters of immune-related genes [74]. DNA methylation consistently appears to play a central role in cognitive function and manipulation of DNA methylation-related mechanisms consistently impacts memory processes [75]. The present study suggested that DNA methylation regulated the expression of genes with functions related to neuroin ammation in memory-critical brain regions. However, the mechanism by which DNMTs are targeted to speci c genomic sites is unknown.
Although studies have suggested that noncoding RNAs [76,77] or chromatin marks [78, 79] may be responsible for that targeting, these mechanisms are still far from being understood.
The methylation sites investigated in this study are adjacent to a canonical NF-κB-binding site and are individual CpG sites near the transcription start site (TSS). The reason for choosing this site is that IL-1β promoter region in mouse has no CpG island and there is considerable variability in methylation in promoter regions without CpG islands [55]. At the same time, studies have pointed out that several NF-κB (p65/RelA and cRel) consensus sequences have been identi ed within the Gadd45β gene promoter, and these sites are involved in the regulation of Gadd45β expression and DNA demethylation in hippocampal neurons during fear memory formation [80]. Evidence in microglia indicates that epigenetic regulation of IL-1β in aged mice could be regulated by increased NF-κB binding, which inhibits proper DNA methylation [56]. Interestingly, Transcriptional regulation of IL-1β through methylation in mouse myeloid cells is very consistent with ndings during normal aging in humans and in patients with dementia with tauopathy.
One speci c CpG site 215 bp upstream of the TSS is found whose methylation status affects the IL-1β transcript level in mouse, which is consistent with our results. Two CpG sites (cg01290568 and cg15836722) are located in the IL-1β gene near the TSS whose methylation negatively correlates with aging and dementia in human [81]. This nding suggests selective regulation of speci c CpG sites in the IL-1β gene in both mice and humans could change their cognitive performance. Therefore, our research could provide a new target for the clinical prevention of POCD in elderly individuals.
IL-1β is a key proin ammatory cytokine associated with age-related cognitive decline [82]. IL-1β exerts its biological activity by binding to IL-1 receptor type 1 (IL-1R1). IL-1ra, the endogenous inhibitor of IL-1β, competing with IL-1R1 [40]. Accumulating evidence indicates that synaptic plasticity [83], learning and memory [9] become more vulnerable to IL-1β-mediated impairment. The expression levels of synapseassociated proteins, especially synaptophysin and PSD95, are decreased in the hippocampus of mice treated with LPS. Galantamine not only prevents the LPS-induced reductions in synaptophysin and PSD95 expressions but also increases the dendritic spine density, which can prevent the activation of microglia and improve neuroin ammation by inhibiting in ammatory signaling molecules (NF-κB p65) and cytokines (TNF-α, IL-1β and IL-6) in the hippocampus [84].

Conclusion
In summary, the current study indicated that DNA methylation may be an important mechanism mediating susceptibility to POCD in aged mice by regulating the expression of IL-1β, which could induce synaptic plasticity dysfunction and cognitive decline (Fig. 7). Our results may provide a direction for further research on the relationship among DNA methylation, in ammation and POCD. Our research also provided a new target for the clinical prevention of POCD in elderly individuals.

Declarations
Gao, Yue You, Yin Gao, Yue-Ying Liu, Hu-Yi Wang, Jing-Ru Hao and Nan Sun revised the paper; and Jing-Ru Hao and Nan Sun provided technical and material support. All authors read and approved the nal paper.  Surgery activated microglia in the dorsal hippocampus a-b The protein and mRNA levels of IL-1β in the dorsal hippocampus were increased in the aged+1d group. c The nuclear expression of NF-κB P65 in the dorsal hippocampus in aged mice was increased 1 day after surgery. d Representative images of Iba1 and GFAP uorescence in the CA1 region on day 1 after surgery (scale bar = 100 m). e-f Quanti cation of Iba1 and GFAP uorescence. g-h The expression of Iba1 was signi cantly increased in the 1d groups of aged and adult mice, while GFAP showed no signi cant difference in the 1d group compared with the sham group. n=4-6. The data are presented as the mean ± SEM values. *P < 0.05, **P < 0.01, ***P < 0.001 versus the corresponding group.   Surgery decreased DNA methylation of IL-1β promoter in aged mice a The MeDIP and MSP results showed that the DNA methylation level in the IL-1β promoter region in aged mice was signi cantly decreased 1 day after surgery at the -215 bp and -229 bp sites. b However, into decreases were not observed at the other CpG sites in IL-1β. c BV2 cells were effectively transduced with Lenti-GFP. d-f Overexpression of DNMT3a in BV2 cells reversed the LPS-induced increase in IL-1β expression and decrease in the IL-1β promoter DNA methylation level. n = 3-5. The data are presented as the mean ± SEM values. *P < 0.05, **P < 0.01 versus the corresponding group.