Sevoflurane-induced disruption of critical period Arc signaling drives aberrant microglial synaptic pruning and cognitive deficits Authors: Bai-hui Chen, Ye-ru Chen, Liang-yu Zheng, Hua-jing Cai, Xin-long Ke, Shi-yu Li, Zheng-jie Chen, Xiang-nan Zhang, Feng-quan Zhou, Gang Chen Published: 2026 ABSTRACT Early-life sevoflurane exposure is associated with long-term cognitive deficits. Given that hippocampal development relies on precise critical windows, disruption of developmental processes during these periods is likely the origin of these impairments. However, the molecular mechanisms underlying sevoflurane-induced perturbations during critical development periods and their progression to lasting cognitive dysfunction remain elusive. Here, we determined that the first three postnatal weeks are a critical window of vulnerability to early-life sevoflurane exposure. Mechanistically, sevoflurane exposure suppressed the physiological upregulation of activity-regulated cytoskeleton-associated protein (Arc; also known as Arg3.1) during the critical period of hippocampal development, a process driven by glycogen synthase kinase-3 beta (GSK3β)-mediated protein degradation. We demonstrated that transient suppression of Arc via hippocampal injection of Arc antisense oligonucleotide (ASO) during the third postnatal week was sufficient to recapitulate the sevoflurane-induced phenotype, impairing microglial synaptic pruning and causing initial synaptic redundancy. This early pathology subsequently evolved into aberrant microglial phagocytic activation in adolescence and adulthood, resulting in excessive synaptic loss and cognitive deficits. Crucially, restoration of Arc expression specifically during this critical period using a doxycycline (Dox)-inducible Tet-On system successfully reversed this pathological pruning trajectory and prevented long-term cognitive impairment. Our findings highlight that Arc upregulation during this critical period is essential for microglial function and synaptic homeostasis, establishing Arc as a time-sensitive therapeutic target for preventing the developmental neurotoxicity associated with pediatric anesthesia. This diagram illustrates the role of the Arc protein in microglial synaptic pruning and cognitive development under physiological conditions (upper panel, blue; Arc is highly expressed during the critical period) and sevoflurane exposure (lower panel, red; Arc expression during the critical period is downregulated). Under normal conditions, Arc is expressed during the critical period and is localized to synapses, where it facilitates the tagging of redundant synapses for elimination. Synaptic pruning peaks during the critical period, followed by further refinement during adolescence, leading to mature cognitive function in adulthood. Neonatal sevoflurane exposure leads to aberrant activation of GSK3β during the critical period. This promotes Arc degradation, thereby disrupting its synaptic localization. This leads to microglial dysfunction in terms of synaptic engulfment. In adolescence, this disruption leads to excessive microglial phagocytosis and significant synaptic loss, ultimately resulting in cognitive deficits. Graphical elements: GSK3β, green circles; Arc, orange circles; phosphorylation site, pale yellow circle containing the letter “P”; C1q, yellow umbrella shapes. Introduction Multiple or prolonged general anesthesia in infants and young children is linked to long-term neurological impairments, including cognitive and behavioral dysfunction [1, 2]. Among clinical agents, sevoflurane is a widely utilized inhalational anesthetic in pediatric anesthesia. Animal experiments have demonstrated that repeated exposure to sevoflurane during early developmental stages leads to hippocampus-dependent cognitive deficits in adolescence or adulthood [3, 4]. Postnatal brain development is characterized by critical periods, with transient windows of heightened plasticity where experience profoundly shapes neural circuits [5]. Specifically, the critical period for hippocampal cognitive development, which spans the third postnatal week in mice, is indispensable for hippocampal network maturation, the disruption of which leads to persistent cognitive deficits [6]. However, whether early-life exposure to sevoflurane disrupts these core critical period mechanisms remains an important yet unanswered question. Recent investigations into sevoflurane-induced neurotoxicity have targeted diverse pathogenic mechanisms, including mitochondria-related ferroptosis driven by iron overload and SIRT1-mediated deficits in neurogenesis, with synaptic alterations serving as a critical point of convergence [7, 8]. Moreover, substantial evidence indicates that excessive synaptic loss during adolescence or adulthood serves as the neurostructural basis for long-term cognitive impairments following such early sevoflurane exposure, a pathology involving microglial overactivation [9, 10]. Crucially, emerging evidence indicates that synaptic homeostasis is disrupted before this delayed microglial hyperactivity emerges, suggesting that adolescent microglial overactivation is a consequence, rather than the origin, of earlier synaptic dysfunction [11]. Synaptic pruning in the developing brain occurs during two major peak periods: the widely-studied adolescent period, which coincides with the delayed pathology, and a much earlier, distinct peak period that is aligned with the critical period of the hippocampus [5, 6]. This suggests that the delayed pathology may be linked to a disruption of this critical early pruning. Elucidating how sevoflurane disrupts synaptic development during this early critical period is essential for understanding how the long-term pathological cascade is programmed. The immediate early gene activity-regulated cytoskeleton-associated protein (Arc) offers a potential molecular link [12]. Its expression in the hippocampus shows a remarkable, transient peak during the third postnatal week, and its dysregulation is known to correlate with adult memory dysfunction [13]. This, combined with Arc’s role in synaptic elimination in the cerebellum, underscores its universal and time-sensitive role in circuit refinement [14]. Therefore, whether sevoflurane specifically targets this fundamental Arc-dependent process and whether this disruption serves as the molecular switch that initiates the long-term maladaptive trajectory of microglia remain unclear. The purpose of this study was to investigate the toxicological mechanism through which early-life sevoflurane exposure drives long-term cognitive impairment via its impact on the critical period of hippocampal development. We first employed distinct exposure paradigms to pinpoint the specific developmental window of vulnerability. We revealed that sevoflurane induces Arc downregulation during this critical period via the overactivation of upstream glycogen synthase kinase-3 beta (GSK3β). Subsequently, utilizing Arc antisense oligonucleotide (ASO)-mediated transient suppression, we delineated the causal timeline of synaptic pathology and microglial dysfunction. Finally, we evaluated whether time-specific restoration using a doxycycline (Dox)-inducible Tet-On system could rescue the pathological pruning trajectory and prevent long-term cognitive deficits. Materials and methods Animals All the experimental procedures were conducted in accordance with the guidelines of Zhejiang University and were approved by the Animal Care and Use Committee of Zhejiang University (ZJU20240792). C57BL/6 mice were procured from Hangzhou Qizhen Biological Technology Co., Ltd. (Hangzhou, China), while Thy1-EGFP mice were obtained from Jackson Laboratory (strain #007788, USA). The animals were housed in a controlled environment with a 12-h light/dark cycle, 45%–50% relative humidity, and a temperature range of 22–24 °C and were provided with a standard diet. Efforts were made to minimize the use of animals, with littermates being randomly assigned to experimental groups for further study. Both male and female mice were used. For key biochemical and behavioral tests, the initial sex-separated analyses revealed highly consistent trends. Thus, the datasets were pooled to account for variations in the natural birth sex ratio and ensure statistical power. To maintain transparency, these individual data points are annotated as circles (females) and triangles (males). For other experiments, data from equal numbers of both sexes were pooled and are presented as diamonds. The specific numbers of animals used are provided in each figure legend. Sevoflurane anesthesia The randomly grouped littermate mice were placed in the same type of anesthesia chamber. The experimental group was continuously exposed to a gas mixture containing 3% sevoflurane (composed of 55% O₂ with the remainder as balance gas), while the control group was exposed to 55% O₂ alone, both at a constant flow rate of 3 L/min. The chamber temperature was consistently maintained at 37 °C using a heating pad. Detailed information regarding specific exposure durations and the ages of the animals is provided in the accompanying figure. Previous studies have demonstrated that these exposure conditions do not lead to significant alterations in blood gas parameters [3]. After treatment, the mice were returned to standard housing cages for routine care. Y-maze spontaneous alternation test The Y-maze spontaneous alternation task was utilized to assess spatial working memory. The Y-maze apparatus was constructed from gray polyvinyl chloride and consisted of three symmetrical arms, each 30 cm in length, 10 cm in width, and 15 cm in height. Spatial cues were strategically placed around the maze to facilitate navigation. The mice were introduced at the end of one arm and allowed to explore the maze freely for 5 min. The percentage of spontaneous alternations was determined manually. Novel object recognition (NOR) test The experiment was conducted using an open field apparatus with dimensions of 45 cm × 45 cm × 45 cm. Prior to the initiation of the experiment, the mice were allowed to acclimate to the testing environment for 1 h. During the initial phase, two identical plastic objects were placed in the corners of the apparatus, and the mice were given 10 min to explore these objects. Afterward, the mice were removed from the apparatus and returned to their home cages. Twenty-four hours later, the mice were again placed in the test apparatus, where one of the original plastic objects was replaced with a novel object of similar size but differing in color and shape. The mice were then allowed to explore the objects for 5 min. The statistical area was defined as a circular region with a radius of 2.5 cm surrounding the novel object. Exploration time and the frequency of entries into the statistical area were automatically recorded using ANY-maze behavioral analysis software (version 4.3; Stoelting Co.). Contextual/Cued fear conditioning (CFC) test These experiments utilized the conditional fear system developed by Coulbourn Instruments. On the first day, the mice were acclimated to the chamber for 120 s, after which they were subjected to three cycles of a conditioned stimulus. This stimulus consisted of a 30-s, 85 dB, 5000 Hz tone, immediately followed by a 0.5 mA foot shock administered 2 s prior to the conclusion of the tone. The intervals between stimuli were maintained at 30 s. On the second day, a memory test was conducted to evaluate contextual memory by measuring the duration of freezing behavior (defined as immobility lasting at least 0.75 s) as the mice were reintroduced to the chamber for 300 s. Following a 2-h intermission, the environment of the behavior chamber was modified for the cue memory test, which included a 120-s adaptation phase followed by a 120-s tone stimulus, during which the freezing time was recorded. These tests were repeated on Day 5. Long-term memory retention was quantified using the forgetting index, which was calculated as (1- [freezing time on Day 5]/[freezing time on Day 2]) × 100%. Freezing time was automatically recorded using behavioral analysis software in accordance with the standardized conditioned fear paradigm. Stereotaxic injection The procedure for viral injection into neonatal mice was adapted from established protocols [15]. Neonatal mice at P1-2 underwent hypothermic anesthesia and were immobilized in a stereotaxic apparatus (RWD Life Science, China) on a chilled metal plate. A glass micropipette connected to a nanoliter syringe was used to penetrate the scalp and skull above the hippocampus at the following coordinates: 0.95 mm anteroposterior (AP) and ±1.05 mm mediolateral (ML) relative to the lambda, which served as the reference point. The micropipette was advanced to a dorsoventral (DV) depth of −1.25 mm from the skull surface, and 250–350 nL of adeno-associated virus (AAV) was delivered over a period of 3 min. Following the injection, the position of the micropipette was maintained for an additional 3 min to facilitate proper diffusion. The entire procedure was completed within 15 min. After the injection, the pups were placed on a heating pad set to 40 °C and continuously monitored until they exhibited movement. The pups remained on the heating pad until all surgical procedures were completed, after which they were returned to the dam. The following viral constructs were used: ssAAV9-hsyn-tet on 3G-Arc-eGFP (1 × 1012 GC/ml), pSCAAV9-hSyn-Arc-P2A-eGFP (1.09 × 1013 vgs/ml) and pSCAAV9-hSyn-egfp (1.15 × 1013 vgs/ml). In P16 mice, deep anesthesia was induced using pentobarbital, followed by a midline scalp incision. The mice were then positioned in a stereotaxic apparatus (RWD Life Science, China). The stereotaxic coordinates were 0.8 mm AP and ±1.8 mm ML relative to the lambda. Two holes were subsequently drilled into the skull, and a micropipette was inserted to a DV depth of −1.5 mm from the skull surface. A total of 250 nL of antisense oligonucleotides (ASOs) was administered at a rate of 50 nL/min per hemisphere, and the needle was retained in the same position for 5 min following injection. Upon completion of the bilateral microinjections, the scalp was sutured, and the mice were transferred to clean cages for recovery. Throughout the procedure, a heating pad was used to maintain the body temperature of the anesthetized mice at 37 °C. Immunofluorescence staining Mice were deeply anesthetized with sodium pentobarbital (50 mg/kg) and transcardially perfused with ice-cold phosphate-buffered saline (PBS; pH 7.4), followed by 4% paraformaldehyde (PFA). The brains were postfixed in 4% PFA at 4 °C for 12 h and cryoprotected in 20% and 30% sucrose solutions at 4 °C until they sank (approximately 72 h). The tissues were embedded in optimal cutting temperature (OCT) compound and coronally sectioned at a thickness of 20 μm or 40 μm using a cryostat (Cryostar NX50, Thermo Scientific, USA). The sections were subsequently washed three times in PBS and blocked for 1 h at room temperature in PBS supplemented with 10% normal donkey serum and 0.1% Triton X-100. Primary antibodies were applied for incubation overnight at 4 °C according to specific experimental requirements, with antibody combinations selected on the basis of target validation. Primary antibodies against the following factors were used: Arc (1:500, Synaptic Systems, cat# 156003; 1:400, Synaptic Systems, cat# 156005), Homer1 (1:500, Synaptic Systems, cat# 160004), VGLUT1 (1:500, Synaptic Systems, cat# 135011), GSK3β (1:100, Santa Cruz, cat# sc-377213), phospho-GSK3β (Ser9) (1:50, Santa Cruz, cat# sc-373800), PSD-95 (1:200, Invitrogen, cat# 7E3-1B8), Synaptophysin (1:500, ABclonal, cat# A19122), C1q (1:500, Abcam, cat# ab11861), MAP2 (1:500, Abcam, cat# ab254264), GFAP (1:500, Cell Signaling Technology, cat# 3670), Iba1 (1:400, Wako, cat# 019-19741), and CD68 (1:200, Invitrogen, cat# MA5-16674). After being washed, the sections were incubated for 1 h at room temperature with species-specific secondary antibodies (1:500; Abcam), including goat anti-rabbit Alexa Fluor 594, goat anti-pig Alexa Fluor 647, goat anti-mouse Alexa Fluor 488, goat anti-rat Alexa Fluor 488, and goat anti-mouse Alexa Fluor 594. Finally, the sections were mounted with DAPI-containing anti-fade mounting medium (Abcam, UK) and imaged using a confocal microscope (IX83-FV3000; Olympus, Japan). For certain experiments employing superresolution imaging, STED microscopy was performed using the Abberior Facility Line system (Abberior Instruments GmbH, Göttingen, Germany). Western blotting Following isoflurane anesthesia, the mice were decapitated, and the hippocampal tissues were rapidly dissected on ice. The tissues were subsequently rinsed in cold PBS and homogenized in RIPA buffer supplemented with protease and phosphatase inhibitor cocktails. The homogenates were then centrifuged at 12,000 × g for 15 min at 4 °C, after which the supernatants were collected. The protein concentrations were quantified using a BCA protein assay kit (Beyotime, China). Equal amounts of protein (40 μg per lane) were mixed with 5× SDS loading buffer, denatured at 95 °C for 5 min, and stored at -80 °C until use. Proteins were separated by 10%–15% SDS‒PAGE and transferred to PVDF membranes using the wet transfer method (ice bath, constant current of 250 mA; transfer time adjusted on the basis of molecular weight). The membranes were blocked in 5% nonfat milk prepared in TBST for 90 min at room temperature and incubated overnight at 4 °C with primary antibodies diluted in TBST. Primary antibodies against the following factors were used: Arc (1:1000, Synaptic Systems, cat# 56003), Homer1 (1:1000, Synaptic Systems, cat# 160004), PSD-95 (1:2000, Invitrogen, cat# 7E3-1B8), Synaptophysin (1:1000, ABclonal, cat# A19122), SAP102 (1:1000, ABclonal, cat# A16831), CSF1R (1:1000, ABclonal, cat# A3019), C1QA (1:2000, ABclonal, cat# A1821), GSK3β (1:100, Santa Cruz, cat# sc-377213), phospho-GSK3β (Thr216) (1:1000, Invitrogen, cat# 44-601 G), GAPDH (1:100000, ABclonal, cat# A19056), and β-Tubulin (1:5000, ABclonal, cat# A12289). After three 5-min washes with TBST, the membranes were incubated for 1.5 h at room temperature with HRP-conjugated secondary antibodies (1:5000; ABclonal). After additional washes, the immunoreactive bands were visualized using enhanced chemiluminescence (ECL) reagent and captured with a Tanon imaging system (Tanon, China). Densitometric analysis was performed using ImageJ software, and protein expression levels were normalized to those of internal controls and compared across groups. A general immunoprecipitation kit (Abbkine, KTD104-CN, China) was used for coimmunoprecipitation (Co-IP) analyses. Mouse hippocampal tissue was homogenized in nondenaturing lysis buffer, incubated on ice at 4 °C for 5 min, and subsequently centrifuged at 12,000 × g for 10 min to obtain the supernatant. The protein concentration was quantified using a BCA assay. To eliminate nonspecific binding, 25 μL of protein A/G magnetic beads were washed three times with 1× wash buffer and incubated with 1 mg of the protein sample at 4 °C for 30 min. After magnetic separation, the supernatant was discarded. Another 25 μL aliquot of magnetic beads was washed and incubated with 2 μg of antibody at room temperature for 30 min, followed by an additional three washes. The pretreated protein sample was then added to the beads for incubation overnight at 4 °C, followed by three additional washes. Finally, 1× SDS‒PAGE loading buffer was added, and the samples were denatured at 95 °C for 5 min. The supernatant was subsequently collected for Western blot analysis. Golgi staining The mouse brains were subjected to Golgi staining using the FD Rapid GolgiStain™ Kit (FD NeuroTechnologies, USA). The brains were rinsed with double-distilled water and immersed in a 1:1 mixture of solutions A and B at room temperature in the dark. Twenty-four hours later, the solution was replaced, and the tissue was incubated for 2 more weeks, followed by incubation in solution C for 3 d. The tissue was sectioned to a thickness of 100 μm using a cryostat, placed on gelatin-coated slides, and dried in air for 4 d. The sections were rinsed, stained with working solution (solutions D and E and water at a 1:1:2 ratio) in the dark for 10 min, and then they were washed twice with water for 4 min each. They were then dehydrated through a graded ethanol series (50%, 75%, 95%, and 100%), cleared in xylene three times (4 min each), and mounted with neutral resin. Dendritic spine morphology in the hippocampal CA1 region was observed using an Olympus BX61 microscope (Olympus, Japan) with a 100× oil immersion objective and analyzed with ImageJ software. Filopodia-like spines were defined as protrusions with a length >3 μm, a width <0.6 μm, and a length/width ratio >1. Thin spines were defined as protrusions with a length of 1–3 μm, a width <0.6 μm, and a length/width ratio >1. Stubby spines were defined as protrusions with a length <1 μm, a width >0.6 μm, and a length/width ratio <1. Mushroom-shaped spines were defined as spines with a head width >0.6 μm. Real-time quantitative PCR (RT‒qPCR) Total RNA was isolated from the hippocampal tissue of mice aged P6 to P30 using the FastPure Cell/Tissue Total RNA Isolation Kit (RC112-01; Vazyme, China). The quality and concentration of the RNA were assessed using a NanoDrop spectrophotometer (Thermo Scientific, USA). Following the protocol provided by the Reverse Transcription Premix Kit (MR05101; Monad, China), reverse transcription was carried out in a 20 μL volume. Quantitative real-time PCR was subsequently performed using a SYBR Green I Premix qPCR Kit (QP-01012, FOREGENE, China) with a total reaction volume of 20 μL. The reaction mixture contained 1 μL of cDNA, 1.6 μL of each primer pair, 7.4 μL of nuclease-free water, and 10 μL of 2× Real PCR EasyTM Mix-SYBR. The PCR cycling conditions were as follows: initial denaturation at 95 °C for 10 s, followed by denaturation at 60 °C for 10 s and annealing at 72 °C for 20 s for a total of 40 cycles. The expression level of the target gene was normalized to that of β-actin and analyzed using the 2−ΔΔCT method for relative gene expression quantification. The primer sequences are detailed in Supplementary Table 1. Bulk RNA-seq RNA sequencing was conducted on mice that received hippocampal injections of either Arc ASOs or a scrambled control on P16 via stereotaxic injection. Fresh tissue samples from three mice in each group were collected on P21 and P56, ensuring equal representation of both sexes. The samples were promptly placed in labeled Eppendorf tubes and rapidly frozen in liquid nitrogen prior to mRNA sequencing, which was performed by LC-Bio Technology. Following the generation of the final transcriptome, transcript expression levels were estimated using StringTie and Ballgown, and mRNA expression levels were quantified by calculating FPKM values. Differential gene expression analysis was performed utilizing the DESeq2 software package. Genes whose P value was less than 0.05 and whose absolute fold change was greater than zero were classified as differentially expressed. These genes were further analyzed for pathway enrichment using the KEGG or GO databases. Bioinformatic analyses were executed using OmicStudio tools (available at https://www.omicstudio.cn/tool). All the mice were included in the analyses. Microglial engulfment IMARIS software (Bitplane, Zurich, Switzerland) was used to quantitatively analyze microglial phagocytosis. The “Surface” module was used to reconstruct microglia labeled with Iba-1 and lysosomal regions labeled with CD68, whereas the “Spots” function was used to identify Homer1-positive puncta. After threshold optimization, the “Shortest distance to Surface” filter was applied to select spots completely contained within microglia that colocalized with CD68-positive regions (distance≤0 μm). The number of colocalized Homer1 puncta was quantified to evaluate the amount of synaptic components engulfed and transported by microglia to lysosomes. Synaptic protein colocalization analysis Three-dimensional analyses of protein interactions, including Synaptophysin, Homer1, PSD95, VGLUT1, ARC, and C1q, were conducted using the “Spots” function to identify puncta. After threshold optimization, the “Shortest Distance to Spots” function was used to assess spatial colocalization for pre- and postsynaptic pairings (e.g., Synaptophysin with Homer1 and PSD95 with VGLUT1) and functional complex pairings (e.g., ARC with C1q). The colocalization distance threshold was 0.35 μm, and the number of spots within this distance was quantified. Microglial cell morphology Morphological quantification of microglia was conducted utilizing the NeuronJ plugin within ImageJ software. First, the semiautomated tracing function was employed to manually delineate the two-dimensional spatial distribution of the cell branches, thereby generating the branch topology of the microglia. The Sholl Analysis plugin was subsequently used to evaluate the complexity of the branches. Centering on the soma, the initial radius of the concentric circles was established at 2 μm, with a step size of 1 μm, expanding outward. The number of intersections between each concentric circle and the cell branches was determined to quantify the radial distribution characteristics of the branch extension. Statistical analysis All the statistical analyses were performed using GraphPad Prism version 10.2. The data are presented as the mean ± standard error of the mean (SEM). The sample size (n) represents the number of independent biological replicates per experimental group, as indicated in the figure legends. The normality of the data distribution was assessed using the Shapiro‒Wilk test. For comparisons between two groups, an unpaired t test was used when the data were normally distributed with equal variances; if the Brown–Forsythe test indicated unequal variances, an unpaired t test with Welch’s correction was applied. For nonnormally distributed data, the Mann‒Whitney U test was used. For comparisons involving three or more unpaired groups with normally distributed data and equal variances, one-way ANOVA followed by Tukey’s post hoc test was performed. When the data were not normally distributed, a Kruskal‒Wallis test followed by Dunn’s multiple comparisons test was applied. Two-way comparisons involving two independent variables were analyzed using two-way ANOVA, followed by Tukey’s multiple comparisons test. A P value < 0.05 was considered to indicate statistical significance. Significance levels are denoted as follows: P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), and P < 0.0001 (****). Results The first three postnatal weeks represent a vulnerable window during which sevoflurane exposure disrupts critical period Arc expression and long-term cognitive deficits We first employed a well-established neonatal exposure paradigm in which mice were exposed to 3% sevoflurane for 2 h daily from postnatal day (P)6 to P8 (Fig. 1a, b). In control mice, hippocampal mRNA expression of Arc gradually increased during the third postnatal week, peaking at P21, followed by a rapid decrease to baseline levels, which is consistent with previous reports [13]. Notably, compared with control conditions, repeated sevoflurane exposure significantly reduced Arc mRNA levels at P21 (Fig. 1c). Western blot analysis confirmed decreased ARC protein levels in both female and male mice at P21 following sevoflurane exposure (Fig. 1d, e). Cognitive performance was assessed in adulthood (P56) using hippocampus-dependent tasks: the Y-maze spontaneous alternation test (working memory), the NOR test with a 24-h retention interval (intermediate-term memory), and the CFC test with a 5-day forgetting index (long-term memory). Compared with control mice, sevoflurane-exposed mice exhibited a significant increase in the forgetting index for both contextual and cued memory in the CFC test (Fig. 1f). In the NOR test, sevoflurane-exposed mice showed a significantly reduced exploration time for the novel object and fewer entries into the 20 mm exploration zone (Fig. 1g). In the Y-maze task, these mice tended toward reduced spontaneous alternation, although the difference did not reach statistical significance (Fig. 1h).Fig. 1Neonatal sevoflurane exposure induces both a reduction in Arc expression at P21 and cognitive deficits in adults, whereas exposure during the fourth week reduces Arc expression at P30 but does not cause cognitive deficits. a, b, i Experimental design. c–e Molecular analysis following sevoflurane exposure at P6, P7, and P8, comparing the Control and Sevoflurane groups. (c) RT‒qPCR analysis of Arc mRNA expression at P6, P9, P14, P21, and P30. Control vs. Sevoflurane [two-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001; n = 3–6 mice]. d, e Representative Western blot and quantification of ARC protein expression at P21 in female and male mice. Control vs. Sevoflurane [male: two-tailed unpaired t test, P = 0.032; female: two-tailed unpaired t test, P = 0.019; n = 3 mice]. f–h Cognitive behavioral tests in adulthood following sevoflurane exposure at P6, P7, and P8, comparing the Control and Sevoflurane groups. f Forgetting index for both contextual and cued recall in the CFC test. Control vs. Sevoflurane [context: Mann‒Whitney test, P = 0.0447; cue: two-tailed unpaired t test, P = 0.0141; Control: n = 12 mice (7 males, 5 females); Sevoflurane: n = 13 mice (7 males, 6 females)]. g Exploration trajectories and percentage of time and entries toward the novel object between groups [time: two-tailed unpaired t test, P = 0.019; entries: two-tailed unpaired t test, P < 0.0001; n = 11 mice per group (6 males, 5 females)]. h Spontaneous alternation rate in the Y-maze test in sevoflurane-treated vs. Control mice [two-tailed unpaired t test, P = 0.0531; n = 11 mice per group (6 males, 5 females)]. j–k Representative Western blot and quantification of ARC protein expression following sevoflurane exposure on P24–P26, comparing the Sevoflurane and Control groups [two-tailed unpaired t test, P = 0.016; n = 3 mice]. l–n Cognitive behavioral tests in adulthood following sevoflurane exposure at P24, P25, and P26, comparing the Control and Sevoflurane groups. l Spontaneous alternation rate in the Y-maze test. Control vs. Sevoflurane [two-tailed unpaired t test, P = 0.4108; Control: n = 9 male mice, Sevoflurane: n = 10 male mice]. m Forgetting index for both contextual and cued recall in the CFC test. There was no significant difference between the groups [context: Mann‒Whitney test, P = 0.366; cue: Mann‒Whitney test, P = 0.5338; Control: n = 6 male mice; Sevoflurane: n = 7 male mice]. n Exploration trajectories and percentage of time and entries toward the novel object in both groups [time: two-tailed unpaired t test, P = 0.2537; entries: two-tailed unpaired t test, P = 0.9308; Control: n = 9 male mice; Sevoflurane: n = 10 male mice]. Sev, Sevoflurane. ns, P > 0.05. * and **** indicate P < 0.05 and 0.0001, respectively. The data are presented as the mean ± SEM. Neonatal sevoflurane exposure induces both a reduction in Arc expression at P21 and cognitive deficits in adults, whereas exposure during the fourth week reduces Arc expression at P30 but does not cause cognitive deficits. a, b, i Experimental design. c–e Molecular analysis following sevoflurane exposure at P6, P7, and P8, comparing the Control and Sevoflurane groups. (c) RT‒qPCR analysis of Arc mRNA expression at P6, P9, P14, P21, and P30. Control vs. Sevoflurane [two-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001; n = 3–6 mice]. d, e Representative Western blot and quantification of ARC protein expression at P21 in female and male mice. Control vs. Sevoflurane [male: two-tailed unpaired t test, P = 0.032; female: two-tailed unpaired t test, P = 0.019; n = 3 mice]. f–h Cognitive behavioral tests in adulthood following sevoflurane exposure at P6, P7, and P8, comparing the Control and Sevoflurane groups. f Forgetting index for both contextual and cued recall in the CFC test. Control vs. Sevoflurane [context: Mann‒Whitney test, P = 0.0447; cue: two-tailed unpaired t test, P = 0.0141; Control: n = 12 mice (7 males, 5 females); Sevoflurane: n = 13 mice (7 males, 6 females)]. g Exploration trajectories and percentage of time and entries toward the novel object between groups [time: two-tailed unpaired t test, P = 0.019; entries: two-tailed unpaired t test, P < 0.0001; n = 11 mice per group (6 males, 5 females)]. h Spontaneous alternation rate in the Y-maze test in sevoflurane-treated vs. Control mice [two-tailed unpaired t test, P = 0.0531; n = 11 mice per group (6 males, 5 females)]. j–k Representative Western blot and quantification of ARC protein expression following sevoflurane exposure on P24–P26, comparing the Sevoflurane and Control groups [two-tailed unpaired t test, P = 0.016; n = 3 mice]. l–n Cognitive behavioral tests in adulthood following sevoflurane exposure at P24, P25, and P26, comparing the Control and Sevoflurane groups. l Spontaneous alternation rate in the Y-maze test. Control vs. Sevoflurane [two-tailed unpaired t test, P = 0.4108; Control: n = 9 male mice, Sevoflurane: n = 10 male mice]. m Forgetting index for both contextual and cued recall in the CFC test. There was no significant difference between the groups [context: Mann‒Whitney test, P = 0.366; cue: Mann‒Whitney test, P = 0.5338; Control: n = 6 male mice; Sevoflurane: n = 7 male mice]. n Exploration trajectories and percentage of time and entries toward the novel object in both groups [time: two-tailed unpaired t test, P = 0.2537; entries: two-tailed unpaired t test, P = 0.9308; Control: n = 9 male mice; Sevoflurane: n = 10 male mice]. Sev, Sevoflurane. ns, P > 0.05. * and **** indicate P < 0.05 and 0.0001, respectively. The data are presented as the mean ± SEM. To further investigate whether cognitive vulnerability to sevoflurane is confined to early developmental stages, male mice were exposed to 3% sevoflurane for 2 h per day from P24 to P26 (Fig. 1i). We measured ARC protein expression at P30 in these mice. Compared with those in control mice, the ARC protein levels in the hippocampus were reduced (Fig. 1j, k). Subsequent behavioral assessments, including the NOR test, Y-maze spontaneous alternation test, and CFC test, revealed no statistically significant differences between sevoflurane-exposed mice and controls (Fig. 1l–n). To more accurately simulate clinical exposure in infants, where anesthetic procedures are typically scheduled with extended intervals, we modified the exposure timing in our paradigm. Specifically, neonatal mice were exposed to 3% sevoflurane for 2 h on P6, P11, and P16, which is once per week during the first three postnatal weeks (Fig. S1a). Consistent with the behavioral outcomes, hippocampal ARC expression at P21 was significantly reduced compared with that in the control group (Fig. S1b, c). Behavioral assessments in adulthood revealed that compared with control mice, mice that had repeated exposure to sevoflurane exhibited significant reductions in both the novel object exploration time and frequency in the NOR test (Fig. S1d), along with marked impairments in working memory, as measured by the Y-maze spontaneous alternation test (Fig. S1e). In contrast, mice that received a single 2-h exposure at P7 (Fig. S2a) did not exhibit significant cognitive impairments in adulthood. The performance of these mice in the CFC, NOR, and Y-maze tasks remained comparable to that of the controls (Fig. S2b–d), which aligns with previous findings [16]. Western blot analysis revealed no change in ARC expression at P21 compared with that in the controls (Fig. S2ae–f). These results collectively indicate that the timing and repetition of sevoflurane exposure during a sensitive postnatal period are critical factors in the induction of long-term cognitive impairment. In comparison, the cumulative dose of anesthesia and the proximity of the exposure intervals appear to play smaller roles. Taken together, these findings establish that the first three postnatal weeks serve as the vulnerable window for repeated sevoflurane exposure that leads to cognitive impairment, which coincides with the critical developmental period of the hippocampus. Notably, sevoflurane exposure significantly reduces ARC expression during the critical period, which is associated with long-term cognitive deficits. Reductions in ARC expression outside this window do not produce detectable behavioral impairments. Reduced Arc expression during the critical period leads to long-term cognitive impairments Immunofluorescence analysis revealed that the ARC protein was predominantly localized to the CA1 region of the hippocampus at P21, with significantly lower expression in other hippocampal subregions (Fig. 2a, b). Within the CA1 region, the ARC signal was observed across multiple layers, including dendrite-rich nonsomatic areas. In mice exposed to sevoflurane, a substantial reduction in the fluorescence intensity of ARC was observed specifically in the CA1 region compared with that in control mice (Fig. 2c, d). To identify the cellular sources of ARC expression, we performed double immunofluorescence staining to assess the colocalization of ARC with MAP2, IBA-1, and GFAP. The results revealed strong colocalization of the ARC with MAP2, while minimal overlap was observed with IBA-1 and GFAP (Fig. 2e, f), indicating that sevoflurane predominantly affects ARC expression in CA1 pyramidal neurons. Given the temporal and spatial specificity of ARC expression during the critical period, we stereotaxically injected Arc ASO into the hippocampi of P16 mice (Fig. 2g). This intervention transiently suppressed ARC protein translation and facilitated mRNA degradation at the third postnatal week but did not affect Arc levels at other time points. Compared with those in the scramble group, fluorescence imaging at P21 confirmed a significant reduction in the expression of the ARC protein in CA1 neurons (Fig. 2h). Western blot analysis validated these results, revealing a pronounced decrease in the level of the ARC protein at P21, with recovery to baseline levels by P28 (Fig. 2i, j). Behavioral assessments conducted in adulthood revealed that in the Y-maze spontaneous alternation test, no significant differences were observed between mice with transient Arc knockdown and those injected with scrambled controls (Fig. 2k). In the fear conditioning test, compared with mice injected with scrambled control, Arc ASO-injected mice exhibited an increased forgetting index for both contextual and cued memory (Fig. 2l). In the NOR test, compared with control mice, Arc ASO mice displayed significant reductions in the exploration time and frequency (Fig. 2m). These results indicate that transient disruption of Arc expression using Arc ASO during the critical developmental period is sufficient to induce hippocampus-dependent cognitive deficits comparable to those observed in sevoflurane exposure models.Fig. 2Transient Arc knockdown during the critical period leads to cognitive deficits in adulthood. a, b Representative immunofluorescence images of ARC (red) in different subregions of the hippocampus (CA1, DG, CA2, and CA3) and magnified views of the CA1 region at P21. DAPI (blue) was used to stain the nuclei. Scale bars = 200 μm, 10 μm or 5 μm. Quantification of mean ARC intensity in various hippocampal subregions [one-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001 for CA1 vs. DG/CA2/CA3, P = 0.9573 for DG vs. CA2, P = 0.8679 for DG vs. CA3, P = 0.9934 for CA2 vs. CA3]. c, d Representative immunofluorescence images of the ARC protein (red) and DAPI (blue) in the hippocampal CA1 region at P21 following sevoflurane exposure at P6-8. Scale bars = 10 μm. The mean ARC intensity was quantified and compared between the Control and Sevoflurane groups [two-tailed unpaired t test, P = 0.0482; quantification was based on the average of three fields per mouse; n = 6 mice per group]. e, f Colocalization analysis of ARC (red) with distinct cellular markers in the hippocampal CA1 region at P21. Neurons were labeled with MAP2 (green), microglia with IBA1 (green), and astrocytes with GFAP (green). Representative colocalization images and corresponding 2D histograms of selected ROIs are presented to evaluate the spatial overlap between ARC and each cellular marker. Scale bars = 10 μm [one-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001 for MAP-2 vs. IBA-1 and MAP-2 vs. GFAP; n = 6 cells per group]. g Experimental design. h–j Histological analysis following the hippocampal injection of Arc ASO at P16, comparing the Scramble and Arc ASO groups. h Representative immunofluorescence images of ARC protein (red) and DAPI (blue) in the hippocampal CA1 region at P21. Scale bars = 10 μm. i, j Representative Western blot and quantification of ARC protein expression at P21 and P28. Scramble vs. Arc ASO [P21: two-tailed unpaired t test, P = 0.0241; Scramble: n = 3 mice; Arc ASO: n = 4 mice; P28: two-tailed unpaired t test, P = 0.555; n = 4 mice per group]. k‒m Results of the cognitive behavioral tests in adulthood following the hippocampal injection of Arc ASO at P16, comparing the Scramble and Arc ASO groups. k Spontaneous alternation rate in the Y-maze test between the Scramble and Arc ASO groups [Mann‒Whitney test, P = 0.6189; scramble: n = 10 mice (5 males, 5 females); Arc ASO: n = 9 mice (6 males, 3 females)]. l Forgetting index for both contextual and cued recall in the CFC test. Scramble vs. Arc ASO. [context: two-tailed unpaired t test, P = 0.0251; cue: two-tailed unpaired t test, P = 0.0126; Scramble: n = 10 mice (5 males, 5 females), Arc ASO: n = 8 mice (6 males, 2 females)]. m Exploration trajectories and percentage of time and entries toward the novel object in both groups [time: two-tailed unpaired t test, P < 0.0001; entries: two-tailed unpaired t test, P < 0.0001; Scramble: n = 14 mice (7 males, 7 females); Arc ASO: n = 17 mice (8 males, 9 females)]. ROI, region of interest. ASO, antisense oligonucleotide. ns, P > 0.05. * and **** indicate P < 0.05 and 0.0001, respectively. The data are presented as the mean ± SEM. Transient Arc knockdown during the critical period leads to cognitive deficits in adulthood. a, b Representative immunofluorescence images of ARC (red) in different subregions of the hippocampus (CA1, DG, CA2, and CA3) and magnified views of the CA1 region at P21. DAPI (blue) was used to stain the nuclei. Scale bars = 200 μm, 10 μm or 5 μm. Quantification of mean ARC intensity in various hippocampal subregions [one-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001 for CA1 vs. DG/CA2/CA3, P = 0.9573 for DG vs. CA2, P = 0.8679 for DG vs. CA3, P = 0.9934 for CA2 vs. CA3]. c, d Representative immunofluorescence images of the ARC protein (red) and DAPI (blue) in the hippocampal CA1 region at P21 following sevoflurane exposure at P6-8. Scale bars = 10 μm. The mean ARC intensity was quantified and compared between the Control and Sevoflurane groups [two-tailed unpaired t test, P = 0.0482; quantification was based on the average of three fields per mouse; n = 6 mice per group]. e, f Colocalization analysis of ARC (red) with distinct cellular markers in the hippocampal CA1 region at P21. Neurons were labeled with MAP2 (green), microglia with IBA1 (green), and astrocytes with GFAP (green). Representative colocalization images and corresponding 2D histograms of selected ROIs are presented to evaluate the spatial overlap between ARC and each cellular marker. Scale bars = 10 μm [one-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001 for MAP-2 vs. IBA-1 and MAP-2 vs. GFAP; n = 6 cells per group]. g Experimental design. h–j Histological analysis following the hippocampal injection of Arc ASO at P16, comparing the Scramble and Arc ASO groups. h Representative immunofluorescence images of ARC protein (red) and DAPI (blue) in the hippocampal CA1 region at P21. Scale bars = 10 μm. i, j Representative Western blot and quantification of ARC protein expression at P21 and P28. Scramble vs. Arc ASO [P21: two-tailed unpaired t test, P = 0.0241; Scramble: n = 3 mice; Arc ASO: n = 4 mice; P28: two-tailed unpaired t test, P = 0.555; n = 4 mice per group]. k‒m Results of the cognitive behavioral tests in adulthood following the hippocampal injection of Arc ASO at P16, comparing the Scramble and Arc ASO groups. k Spontaneous alternation rate in the Y-maze test between the Scramble and Arc ASO groups [Mann‒Whitney test, P = 0.6189; scramble: n = 10 mice (5 males, 5 females); Arc ASO: n = 9 mice (6 males, 3 females)]. l Forgetting index for both contextual and cued recall in the CFC test. Scramble vs. Arc ASO. [context: two-tailed unpaired t test, P = 0.0251; cue: two-tailed unpaired t test, P = 0.0126; Scramble: n = 10 mice (5 males, 5 females), Arc ASO: n = 8 mice (6 males, 2 females)]. m Exploration trajectories and percentage of time and entries toward the novel object in both groups [time: two-tailed unpaired t test, P < 0.0001; entries: two-tailed unpaired t test, P < 0.0001; Scramble: n = 14 mice (7 males, 7 females); Arc ASO: n = 17 mice (8 males, 9 females)]. ROI, region of interest. ASO, antisense oligonucleotide. ns, P > 0.05. * and **** indicate P < 0.05 and 0.0001, respectively. The data are presented as the mean ± SEM. Sevoflurane exposure disrupts Arc protein upregulation during the critical developmental period via the activation of GSK3β Extensive studies have demonstrated that repeated sevoflurane exposure leads to persistent hyperactivation of GSK3β [4, 16] and previous studies have shown that GSK3β promotes Arc degradation via the ubiquitin‒proteasome pathway [17, 18]. To determine the activation state of GSK3β in various sevoflurane exposure models at P21, we assessed its phosphorylation level. Compared with control treatment, repeated exposure to sevoflurane from P6 to P8 significantly increased GSK3β phosphorylation at threonine 216 (GSK3β-T216) and upregulated total GSK3β expression in the hippocampus (Fig. 3a, b). In contrast, compared with control treatment, a single sevoflurane exposure did not significantly alter the GSK3β or GSK3β-T216 levels (Fig. S2e-f). Moreover, compared with control mice, male mice exposed to sevoflurane from P24 to P26 presented elevated GSK3β-T216 levels, indicating that repeated exposure during later developmental stages also sustains GSK3β activation (Fig. 3c, d). Together, these results demonstrate an inverse correlation between GSK3β activation and Arc expression, suggesting that repeated sevoflurane exposure may impair Arc stability during the critical period by promoting its degradation through the activation of GSK3β.Fig. 3Neonatal sevoflurane exposure activates GSK3β, disrupting Arc upregulation through its interaction with the Arc protein during the critical period. a, b Representative Western blot and quantification of GSK3β-T216 expression relative to total GSK3β expression at P21 following sevoflurane exposure at P6–P8, comparing the Control and Sevoflurane groups [total GSK3β: two-tailed unpaired t test, P = 0.0003; GSK3β-T216/Total GSK3β: two-tailed unpaired t test, P = 0.0315; n = 3 mice per group].c, d Representative Western blot and quantification of GSK3β-T216 expression relative to total GSK3β expression at P30 following sevoflurane exposure at P24–P26 in the Control vs. Sevoflurane groups of mice [total GSK3β: two-tailed unpaired t test, P = 0.2921; GSK3β-T216/Total GSK3β: two-tailed unpaired t test, P = 0.04; n = 3 mice per group]. e Immunofluorescence staining of the hippocampal CA1 region showing the distribution of GSK3β (green), ARC (red), DAPI (blue), and GSK3β-Ser9 (far-red) in control and neonatal sevoflurane-exposed mice. Scale bar = 10 μm. Magnified views indicate regions of colocalization (yellow arrows). Scale bar = 1 μm. f Representative Western blot results of the coimmunoprecipitation experiments showing the interaction between ARC and GSK3β at P21 in control and sevoflurane-exposed mice. g Experimental design. h, i Representative Western blot and quantification of ARC protein expression at P21 after daily intraperitoneal injection of CHIR99021 from P14 to P21 following neonatal sevoflurane exposure [one-way ANOVA followed by Tukey’s multiple comparisons test, P = 0.1797 for Control + DMSO vs. Sevoflurane + DMSO, P = 0.0249 for Sevoflurane + DMSO vs. Sevoflurane + CHIR; n = 3 mice per group]. CHIR, CHIR99021. DMSO, dimethyl sulfoxide. ns, P > 0.05. *, **, and *** indicate P < 0.05, 0.01, and 0.001, respectively. The data are presented as the mean ± SEM. Neonatal sevoflurane exposure activates GSK3β, disrupting Arc upregulation through its interaction with the Arc protein during the critical period. a, b Representative Western blot and quantification of GSK3β-T216 expression relative to total GSK3β expression at P21 following sevoflurane exposure at P6–P8, comparing the Control and Sevoflurane groups [total GSK3β: two-tailed unpaired t test, P = 0.0003; GSK3β-T216/Total GSK3β: two-tailed unpaired t test, P = 0.0315; n = 3 mice per group].c, d Representative Western blot and quantification of GSK3β-T216 expression relative to total GSK3β expression at P30 following sevoflurane exposure at P24–P26 in the Control vs. Sevoflurane groups of mice [total GSK3β: two-tailed unpaired t test, P = 0.2921; GSK3β-T216/Total GSK3β: two-tailed unpaired t test, P = 0.04; n = 3 mice per group]. e Immunofluorescence staining of the hippocampal CA1 region showing the distribution of GSK3β (green), ARC (red), DAPI (blue), and GSK3β-Ser9 (far-red) in control and neonatal sevoflurane-exposed mice. Scale bar = 10 μm. Magnified views indicate regions of colocalization (yellow arrows). Scale bar = 1 μm. f Representative Western blot results of the coimmunoprecipitation experiments showing the interaction between ARC and GSK3β at P21 in control and sevoflurane-exposed mice. g Experimental design. h, i Representative Western blot and quantification of ARC protein expression at P21 after daily intraperitoneal injection of CHIR99021 from P14 to P21 following neonatal sevoflurane exposure [one-way ANOVA followed by Tukey’s multiple comparisons test, P = 0.1797 for Control + DMSO vs. Sevoflurane + DMSO, P = 0.0249 for Sevoflurane + DMSO vs. Sevoflurane + CHIR; n = 3 mice per group]. CHIR, CHIR99021. DMSO, dimethyl sulfoxide. ns, P > 0.05. *, **, and *** indicate P < 0.05, 0.01, and 0.001, respectively. The data are presented as the mean ± SEM. To investigate whether GSK3β directly interacts with Arc during this critical period, we performed immunofluorescence staining of CA1 pyramidal neurons at P21. ARC and GSK3β exhibited strong colocalization in the cytoplasm of sevoflurane-exposed neurons, particularly within dendritic regions, while minimal presence in the nuclei was detected. Additionally, the level of inactive GSK3β (phosphorylated at Ser9) decreased in response to sevoflurane exposure, whereas total GSK3β expression increased and this protein colocalized with Arc (Fig. 3e). To confirm the physical association between GSK3β and Arc, we performed coimmunoprecipitation (Co-IP) assays. These experiments revealed a direct interaction between Arc and GSK3β, with the interaction between Arc and GSK3β in sevoflurane-exposed mice being greater than that in control mice (Fig. 3f). Finally, to test whether pharmacological inhibition of GSK3β can rescue Arc levels, we administered CHIR99021 (a selective GSK3β inhibitor) via daily intraperitoneal injection from P14 to P21 following sevoflurane exposure. While the baseline reduction in Arc expression between the Control + DMSO and Sevoflurane + DMSO groups was not significantly different (likely because of the biological variability and stress associated with repeated daily injections), Western blot analysis revealed that compared with vehicle treatment (DMSO), CHIR99021 treatment significantly restored Arc protein levels (Fig. 3g–i). These results collectively support that repeated sevoflurane exposure activates GSK3β, promoting Arc degradation and impairing its developmental upregulation during the critical period. Disruption of Arc expression during the critical period promotes redundant synapses and delays morphological maturation To elucidate the role of Arc in hippocampal function during the critical period, we performed bulk RNA sequencing on hippocampal tissues collected at P21 from mice that received Arc ASOs or Scramble control ASOs at P16 (Fig. 4a). Differential expression analysis revealed that 556 genes were upregulated and 683 genes were downregulated in the Arc ASO group compared with the Scramble group (Fig. 4b). Notably, the expression of several genes associated with neuronal development and synaptic function, such as Ephb3, Pcdha9, Sema3d, Egr2, Grin2a, and Syt7, was significantly altered in the Arc ASO group compared with the Scramble control group (Fig. S3a). In addition, microglia and immune-related genes, such as Mmp12, Itgb2, Fcgr2b, Tnf, and Cst7, were differentially expressed following Arc knockdown (Fig. S3b), suggesting that Arc may influence neuroimmune signaling pathways. Furthermore, the expression of genes involved in myelination and structure, such as Olig2, Sirt2, Opalin, Gpr17, and Ndrg1, also substantially changed compared with that in the Scramble controls (Fig. S3c). KEGG pathway enrichment analysis revealed significant alterations in several key developmental and signaling pathways, including “MAPK signaling,” “axon guidance,” “glutamatergic synapses,” “neuroactive ligand–receptor interactions,” and “lysosomes,” in the Arc ASO group compared with the Scramble group (Fig. 4c). Together, these findings demonstrate that Arc knockdown during the critical period disrupts transcriptional programs essential for neuronal maturation, synaptic connectivity, glial function, and circuit formation.Fig. 4Transient Arc knockdown during the critical period leads to synaptic redundancy and microglial dysfunction. a Experimental design. b, c Bulk RNA-seq results at P21 following transient Arc knockdown via hippocampal injection of Arc ASOs at P16, comparing the Scramble and Arc ASO groups. b Volcano plots showing differentially expressed genes. Upregulated genes are shown in red, downregulated genes are shown in blue, and genes with no significant change in expression are shown in gray. Scramble vs. Arc ASO. |FC | ≥ 1, P < 0.05. c KEGG pathway enrichment analysis of the DEGs between groups. d–e Representative Western blots and quantification of the expression of the synaptic proteins PSD95, Homer1, SAP-102, and Synaptophysin in the hippocampus at P21 following the injection of Arc ASOs at P16. Scramble vs. Arc ASO [PSD95: two-tailed unpaired t test, P = 0.0329; Homer1: two-tailed unpaired t test, P = 0.0195; SAP-102: two-tailed unpaired t test, P = 0.162; Synaptophysin: two-tailed unpaired t test, P = 0.0247; n = 3 mice per group]. f, g Representative immunofluorescence images and quantification of the colocalization of the presynaptic marker Synaptophysin (red) and the postsynaptic marker Homer1 (green) in the CA1 region of the hippocampus at P21 following the hippocampal injection of Arc ASOs at P16, comparing the Scramble and Arc ASO groups. Scale bar = 5 μm [Synaptophysin spots: two-tailed unpaired t test, P = 0.1313; Homer1 spots: two-tailed unpaired t test, P = 0.0009; colocalized spots: two-tailed unpaired t test, P = 0.0247; quantification was based on the average of six fields per mouse; n = 3 mice per group]. h–i Representative Golgi-stained images and quantification of dendritic spine morphology in CA1 pyramidal neurons at P21 following hippocampal Arc ASO injection at P16. Analyses were performed on secondary apical dendrites. Spines were classified as mushroom, stubby, long, or filopodia-like. Scale bar = 10 μm [Scramble vs. Arc ASO: two-tailed unpaired t test; mushroom, P = 0.002; stubby, P = 0.0738; long, P = 0.1976; filopodia-like, P = 0.0002; n = 29–32 dendritic segments from 3 mice per group]. j–m Analysis of hippocampal microglia at P21 following Arc ASO injection at P16, comparing the Scramble and Arc ASO groups. j Representative skeletonized images of IBA-1-labeled microglia. Scale bar = 20 μm. k Quantification of microglia number, soma area and branching complexity via Sholl analysis in the hippocampal CA1 region, as assessed by IBA-1 immunostaining. Scramble vs. Arc ASO [cell number: two-tailed unpaired t test with Welch’s correction; P = 0.9574; quantification is based on the average of six fields per mouse; n = 3 mice per group; soma area: two-tailed unpaired t test, P = 0.3348; n = 31–52 cells from 3 mice per group; Sholl analysis: two-way ANOVA followed by Tukey’s post hoc test, P = 0.1956, n = 12 cells from 3 mice per group]. l Representative immunofluorescence images showing the merged signals of IBA-1 (red), CD68 (green), Homer1 (far red), and DAPI (blue), a magnified view of CD68 (green)/Homer1 (far red) colocalization, and a 3D-rendered magnified image of Homer1 puncta within CD68 in IBA-1-labeled microglia. White arrows highlight representative examples of synaptic engulfment, demonstrating the localization of Homer1 within microglial CD68-positive compartments. Scale bars = 10 μm or 5 μm. m Quantification of Homer1 puncta localized within the CD68 compartments of microglia. Scramble vs. Arc ASO [Mann‒Whitney test, P < 0.0001; n = 18 cells from 3 mice per group]. n–o Representative Western blots and quantification of CSF1R and C1qa expression at P21 following hippocampal Arc ASO injection at P16 [CSF1R: two-tailed unpaired t test, P = 0.0442; C1qa: two-tailed unpaired t test, P = 0.7185; n = 3 mice per group]. ASO, antisense oligonucleotide. FC, fold change. DEGs, differentially expressed genes. ns, P > 0.05. *, ***, and **** indicate P < 0.05, 0.001, and 0.0001, respectively. The data are presented as the mean ± SEM. Transient Arc knockdown during the critical period leads to synaptic redundancy and microglial dysfunction. a Experimental design. b, c Bulk RNA-seq results at P21 following transient Arc knockdown via hippocampal injection of Arc ASOs at P16, comparing the Scramble and Arc ASO groups. b Volcano plots showing differentially expressed genes. Upregulated genes are shown in red, downregulated genes are shown in blue, and genes with no significant change in expression are shown in gray. Scramble vs. Arc ASO. |FC | ≥ 1, P < 0.05. c KEGG pathway enrichment analysis of the DEGs between groups. d–e Representative Western blots and quantification of the expression of the synaptic proteins PSD95, Homer1, SAP-102, and Synaptophysin in the hippocampus at P21 following the injection of Arc ASOs at P16. Scramble vs. Arc ASO [PSD95: two-tailed unpaired t test, P = 0.0329; Homer1: two-tailed unpaired t test, P = 0.0195; SAP-102: two-tailed unpaired t test, P = 0.162; Synaptophysin: two-tailed unpaired t test, P = 0.0247; n = 3 mice per group]. f, g Representative immunofluorescence images and quantification of the colocalization of the presynaptic marker Synaptophysin (red) and the postsynaptic marker Homer1 (green) in the CA1 region of the hippocampus at P21 following the hippocampal injection of Arc ASOs at P16, comparing the Scramble and Arc ASO groups. Scale bar = 5 μm [Synaptophysin spots: two-tailed unpaired t test, P = 0.1313; Homer1 spots: two-tailed unpaired t test, P = 0.0009; colocalized spots: two-tailed unpaired t test, P = 0.0247; quantification was based on the average of six fields per mouse; n = 3 mice per group]. h–i Representative Golgi-stained images and quantification of dendritic spine morphology in CA1 pyramidal neurons at P21 following hippocampal Arc ASO injection at P16. Analyses were performed on secondary apical dendrites. Spines were classified as mushroom, stubby, long, or filopodia-like. Scale bar = 10 μm [Scramble vs. Arc ASO: two-tailed unpaired t test; mushroom, P = 0.002; stubby, P = 0.0738; long, P = 0.1976; filopodia-like, P = 0.0002; n = 29–32 dendritic segments from 3 mice per group]. j–m Analysis of hippocampal microglia at P21 following Arc ASO injection at P16, comparing the Scramble and Arc ASO groups. j Representative skeletonized images of IBA-1-labeled microglia. Scale bar = 20 μm. k Quantification of microglia number, soma area and branching complexity via Sholl analysis in the hippocampal CA1 region, as assessed by IBA-1 immunostaining. Scramble vs. Arc ASO [cell number: two-tailed unpaired t test with Welch’s correction; P = 0.9574; quantification is based on the average of six fields per mouse; n = 3 mice per group; soma area: two-tailed unpaired t test, P = 0.3348; n = 31–52 cells from 3 mice per group; Sholl analysis: two-way ANOVA followed by Tukey’s post hoc test, P = 0.1956, n = 12 cells from 3 mice per group]. l Representative immunofluorescence images showing the merged signals of IBA-1 (red), CD68 (green), Homer1 (far red), and DAPI (blue), a magnified view of CD68 (green)/Homer1 (far red) colocalization, and a 3D-rendered magnified image of Homer1 puncta within CD68 in IBA-1-labeled microglia. White arrows highlight representative examples of synaptic engulfment, demonstrating the localization of Homer1 within microglial CD68-positive compartments. Scale bars = 10 μm or 5 μm. m Quantification of Homer1 puncta localized within the CD68 compartments of microglia. Scramble vs. Arc ASO [Mann‒Whitney test, P < 0.0001; n = 18 cells from 3 mice per group]. n–o Representative Western blots and quantification of CSF1R and C1qa expression at P21 following hippocampal Arc ASO injection at P16 [CSF1R: two-tailed unpaired t test, P = 0.0442; C1qa: two-tailed unpaired t test, P = 0.7185; n = 3 mice per group]. ASO, antisense oligonucleotide. FC, fold change. DEGs, differentially expressed genes. ns, P > 0.05. *, ***, and **** indicate P < 0.05, 0.001, and 0.0001, respectively. The data are presented as the mean ± SEM. On the basis of the results of the transcriptomic analysis, we investigated how the transient suppression of Arc expression during the critical developmental period affects hippocampal synaptic structure. Western blot analysis revealed significantly higher protein levels of PSD-95, Homer1, and Synaptophysin in the Arc ASO group than in the Scramble group, whereas the protein level of SAP-102 remained unchanged (Fig. 4d, e). Immunofluorescence staining supported these findings, revealing a trend toward increased expression of the presynaptic marker Synaptophysin in the Arc ASO group compared with the scramble control group, although the difference was not statistically significant. In contrast, the expression of the postsynaptic marker Homer1 was significantly upregulated, and the colocalization signal between Synaptophysin and Homer1 markedly increased, suggesting an increase in the synaptic contact density under conditions of Arc reduction (Fig. 4f, g). To determine whether these molecular changes were accompanied by structural remodeling, we performed Golgi staining of the hippocampal CA1 region. The results revealed significant alterations in dendritic spine morphology in the Arc ASO group, with a reduction in mature mushroom-shaped spines and a marked increase in filopodia-like immature spines compared with those in the scramble group (Fig. 4h, i). Given the increased synaptic pruning activity at this developmental stage, these findings may reflect that Arc knockdown affects processes related to synaptic maturation or elimination. To assess whether this phenotype could be recapitulated by sevoflurane exposure, we performed the same analysis in the CA1 region at P21 in mice treated with sevoflurane. Like those in the Arc ASO group, these mice exhibited increased numbers of filopodia-like and long spines; however, compared with those in the control group, the number of mature mushroom spines were unchanged (Figure. S4a, b). Together, these results suggest that the transient suppression of Arc expression during the critical period induces the accumulation of immature spines, a hallmark also observed in the sevoflurane exposure model. Arc suppression impairs microglia-mediated synaptic pruning during the critical period Synaptic pruning is mediated primarily by microglia and preferentially targets immature synapses. To explore whether Arc downregulation directly affects microglial state and function, we systematically evaluated the morphological features and functions of microglia in the hippocampal CA1 region at P21. We found no significant differences in microglial number, soma size, or branching complexity between the Arc ASO and Scramble control groups (Fig. 4j, k). However, the Arc ASO group showed a significant reduction in microglial synaptic engulfment, as determined by the quantified number of Homer1 puncta within CD68-positive compartments (Fig. 4l–m), indicating that Arc knockdown impaired microglia-mediated synaptic clearance during the critical period. Further analysis revealed that CSF1R expression was significantly lower in the Arc ASO group than in the scramble control group (Fig. 4n, o). Similarly, compared with the control group, the sevoflurane-exposed group also exhibited a significant reduction in CSF1R expression (Fig. S4c, d). Since CSF1R signaling is essential for maintaining microglial proliferation and pruning capacity, this reduction supports the notion that microglia may adopt a pruning-deficient state, with impaired synaptic clearance despite unaltered morphology [19, 20]. Given that the Arc protein exhibits a unique “inverse synaptic tagging” property and is selectively enriched at functionally weak or immature synapses, we used STED superresolution microscopy in Thy1-EGFP mice at P21 to examine its synaptic localization [21, 22]. Arc was predominantly localized to dendritic spines with immature morphology (Fig. S5a, b), suggesting that it may serve as a postsynaptic signal to label underdeveloped synapses for clearance. To investigate whether Arc participates in microglial recognition of immature synapses, we examined its spatial colocalization with C1q, a key initiator of the complement-mediated synaptic pruning cascade that preferentially binds to weak or immature synapses [23]. Immunofluorescence staining revealed robust colocalization of Arc and C1q in the hippocampus during the critical period, with some signals were also detected within microglia (Fig. S5c). Notably, the intensity of Arc-C1q colocalization was significantly lower in sevoflurane-exposed mice than in control mice (Fig. S5d), indicating that this interaction may be disrupted by sevoflurane-induced Arc downregulation. To confirm this causal relationship, Arc expression was knocked down, which significantly reduced the number of C1q and Homer1 colocalized puncta, suggesting decreased synaptic localization of C1q (Fig. 5e, f). Moreover, Western blot analysis revealed that Arc knockdown did not affect overall C1q expression levels in microglia (Fig. 4n, o), suggesting that Arc regulates the localization and synaptic targeting of C1q rather than its production.Fig. 5Transient Arc knockdown during the critical period leads to synaptic loss and increased microglial phagocytic activity in adolescence/adulthood. a Experimental design. b, c Bulk RNA-seq results at P56 following transient Arc knockdown via hippocampal injection of Arc ASOs at P16, comparing the Scramble and Arc ASO groups. b Volcano plots showing genes that were differentially expressed between groups. Upregulated genes are shown in red, downregulated genes are shown in blue, and genes with no significant change in expression are shown in gray. Scramble vs. Arc ASO. |FC | ≥ 1, P < 0.05. c GO pathway enrichment analysis of the DEGs. d, e Representative Western blots and quantification of the expression of the synaptic proteins PSD-95 and Homer1 in the hippocampus of adults following hippocampal injection of Arc ASO at P16 comparing the Scramble and Arc ASO groups [PSD-95: two-tailed unpaired t test, P = 0.043; Homer1: two-tailed unpaired t test, P = 0.675; n = 3 mice per group]. f, g Representative immunofluorescence images and quantification of the colocalization of the postsynaptic marker PSD95 (red) and the presynaptic marker VGLUT1 (green) in the CA1 region of the hippocampus in adults following the injection of Arc ASO at P16. Scramble vs. Arc ASO. Scale bar = 5 μm [PSD95 spots: two-tailed unpaired t test, P = 0.0286; VGLUT spots: two-tailed unpaired t test, P = 0.0157; colocalized spots: two-tailed unpaired t test, P = 0.0791; quantification was based on the average of six fields per mouse; n = 3 mice per group]. h, i Representative immunofluorescence images of IBA-1, CD68, and DAPI staining in the CA1 region of the hippocampus in adult mice following Arc ASO injection at P16. High-magnification 3D-rendered images illustrate the CD68 signal within IBA-1-labeled microglia. Scale bars = 50 μm or 10 μm. The volume of CD68 per cell is shown. Scramble vs. Arc ASO [two-tailed unpaired t test, P < 0.0001; n = 15 cells from 3 mice per group]. j, k Representative Western blots and quantification of the expression of the synaptic proteins PSD-95, Homer1 and Synaptophysin in the hippocampus at P28 following the hippocampal injection of Arc ASOs at P16 comparing the Scramble and Arc ASO groups [PSD-95: two-tailed unpaired t test, P = 0.0367; Homer1: two-tailed unpaired t test, P = 0.0361; Synaptophysin: two-tailed unpaired t test, P = 0.0021; n = 3 mice per group]. l–o Analysis of hippocampal microglia at P28 following Arc ASO injection at P16 comparing the Scramble and Arc ASO groups. l Representative skeletonized images of IBA-1-labeled microglia. Scale bar = 20 μm. m Quantification of microglia number, soma area and branching complexity via Sholl analysis in the hippocampal CA1 region, as assessed by IBA-1 immunostaining. Scramble vs. Arc ASO [cell number: two-tailed unpaired t test, P = 0.0227; quantification was based on the average of six fields per mouse; n = 3 mice per group; soma area: two-tailed unpaired t test, P < 0.0001; n = 22–23 cells from 3 mice per group; Sholl analysis: two-way ANOVA followed by Tukey’s post hoc test; P < 0.0001; n = 12–13 cells from 3 mice per group]. n Representative immunofluorescence staining images showing merged signals of IBA-1 (red), CD68 (green), Homer1 (far red), and DAPI (blue), a magnified view of CD68 (green)/Homer1 (far red) colocalization, and a 3D-rendered magnified image of Homer1 puncta within CD68 in IBA-1-labeled microglia. White arrows highlight representative examples of synaptic engulfment, demonstrating the localization of Homer1 within CD68-positive microglial compartments. Scale bars = 10 μm or 5 μm. o Quantification of Homer1 puncta localized within CD68-positive microglial compartments. Scramble vs. Arc ASO [two-tailed unpaired t test, P < 0.0001; n = 15 cells from 3 mice per group]. ASO, antisense oligonucleotide. FC, fold change. DEGs, differentially expressed genes. ns, P > 0.05. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively. The data are presented as the mean ± SEM. Transient Arc knockdown during the critical period leads to synaptic loss and increased microglial phagocytic activity in adolescence/adulthood. a Experimental design. b, c Bulk RNA-seq results at P56 following transient Arc knockdown via hippocampal injection of Arc ASOs at P16, comparing the Scramble and Arc ASO groups. b Volcano plots showing genes that were differentially expressed between groups. Upregulated genes are shown in red, downregulated genes are shown in blue, and genes with no significant change in expression are shown in gray. Scramble vs. Arc ASO. |FC | ≥ 1, P < 0.05. c GO pathway enrichment analysis of the DEGs. d, e Representative Western blots and quantification of the expression of the synaptic proteins PSD-95 and Homer1 in the hippocampus of adults following hippocampal injection of Arc ASO at P16 comparing the Scramble and Arc ASO groups [PSD-95: two-tailed unpaired t test, P = 0.043; Homer1: two-tailed unpaired t test, P = 0.675; n = 3 mice per group]. f, g Representative immunofluorescence images and quantification of the colocalization of the postsynaptic marker PSD95 (red) and the presynaptic marker VGLUT1 (green) in the CA1 region of the hippocampus in adults following the injection of Arc ASO at P16. Scramble vs. Arc ASO. Scale bar = 5 μm [PSD95 spots: two-tailed unpaired t test, P = 0.0286; VGLUT spots: two-tailed unpaired t test, P = 0.0157; colocalized spots: two-tailed unpaired t test, P = 0.0791; quantification was based on the average of six fields per mouse; n = 3 mice per group]. h, i Representative immunofluorescence images of IBA-1, CD68, and DAPI staining in the CA1 region of the hippocampus in adult mice following Arc ASO injection at P16. High-magnification 3D-rendered images illustrate the CD68 signal within IBA-1-labeled microglia. Scale bars = 50 μm or 10 μm. The volume of CD68 per cell is shown. Scramble vs. Arc ASO [two-tailed unpaired t test, P < 0.0001; n = 15 cells from 3 mice per group]. j, k Representative Western blots and quantification of the expression of the synaptic proteins PSD-95, Homer1 and Synaptophysin in the hippocampus at P28 following the hippocampal injection of Arc ASOs at P16 comparing the Scramble and Arc ASO groups [PSD-95: two-tailed unpaired t test, P = 0.0367; Homer1: two-tailed unpaired t test, P = 0.0361; Synaptophysin: two-tailed unpaired t test, P = 0.0021; n = 3 mice per group]. l–o Analysis of hippocampal microglia at P28 following Arc ASO injection at P16 comparing the Scramble and Arc ASO groups. l Representative skeletonized images of IBA-1-labeled microglia. Scale bar = 20 μm. m Quantification of microglia number, soma area and branching complexity via Sholl analysis in the hippocampal CA1 region, as assessed by IBA-1 immunostaining. Scramble vs. Arc ASO [cell number: two-tailed unpaired t test, P = 0.0227; quantification was based on the average of six fields per mouse; n = 3 mice per group; soma area: two-tailed unpaired t test, P < 0.0001; n = 22–23 cells from 3 mice per group; Sholl analysis: two-way ANOVA followed by Tukey’s post hoc test; P < 0.0001; n = 12–13 cells from 3 mice per group]. n Representative immunofluorescence staining images showing merged signals of IBA-1 (red), CD68 (green), Homer1 (far red), and DAPI (blue), a magnified view of CD68 (green)/Homer1 (far red) colocalization, and a 3D-rendered magnified image of Homer1 puncta within CD68 in IBA-1-labeled microglia. White arrows highlight representative examples of synaptic engulfment, demonstrating the localization of Homer1 within CD68-positive microglial compartments. Scale bars = 10 μm or 5 μm. o Quantification of Homer1 puncta localized within CD68-positive microglial compartments. Scramble vs. Arc ASO [two-tailed unpaired t test, P < 0.0001; n = 15 cells from 3 mice per group]. ASO, antisense oligonucleotide. FC, fold change. DEGs, differentially expressed genes. ns, P > 0.05. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively. The data are presented as the mean ± SEM. Taken together, these findings demonstrate that reduced Arc expression impairs microglial synaptic clearance capacity without affecting microglial morphology. Moreover, Arc may act as a postsynaptic signal that identifies immature synapses and facilitates complement-mediated pruning by microglia during the critical period of synaptic development. Neonatal sevoflurane exposure does not trigger alterations in microglial function or structure before the critical period To assess whether repeated sevoflurane exposure from P6 to P8 triggers early microglial responses prior to the onset of the critical period, we examined the status of hippocampal microglia at P14 (Fig. S6a). RT‒qPCR analysis revealed that the expression levels of several microglia-related genes (Aif1, Arg1, Cx3cr1, Csfr1, Tmem119, Trem2, Pu.1, and P2ry12) did not significantly differ between the control and sevoflurane-exposed groups (Fig. S6b). Immunofluorescence staining of IBA-1 and CD68 revealed no significant differences in microglial morphology or phagocytic function between the two groups (Fig. S6c). Specifically, there were no changes in microglial density, soma area, or branching complexity (Sholl analysis) (Fig. S6d). Quantitative analysis confirmed that the CD68 volume within microglia was not significantly different (Fig. S6e). Taken together, these findings suggest that sevoflurane exposure during P6–P8 does not elicit detectable microglial activation or structural remodeling by P14, indicating that subsequent changes in function may not result from an acute immune response but rather emerge progressively during the critical period. Transient Arc knockdown during the critical period leads to abnormal microglial phagocytic activity and synaptic loss in adulthood Given the presence of long-term cognitive deficits, we sought to characterize the transcriptional and cellular alterations in the adult hippocampus following transient Arc knockdown during the critical period. We conducted bulk RNA-seq on the hippocampi of adult mice (P56) that had undergone transient Arc knockdown during the critical period (Fig. 5a). The results revealed notable transcriptional changes: compared with the Scramble control group, 539 genes were upregulated, and 572 genes were downregulated (Fig. 5b). GO enrichment analysis highlighted significant alterations in pathways such as “immune system process,” “lysosome,” “synapse,” “microglial cell activation involved in immune response,” and “axon” (Fig. 5c). The differentially expressed genes related to microglial function included Cd68, Trem2, C1qc, Spp1, and Csf3r (Fig. S3d). The genes involved in neurotransmission and signaling included Homer1, Grin2c, Grm3, Nptx2, and Gabre (Fig. S3e). Additionally, genes associated with cytoskeletal organization and neuroprotection included S100b, Ralgds, Plek, Mag, and Pdlim4 (Fig. S3f). These transcriptional changes suggest that transient Arc knockdown during the critical period may have long-lasting effects on neuroimmune signaling pathways and the maintenance of synaptic homeostasis. To validate the transcriptomic signatures suggesting disrupted synaptic homeostasis and immune activation, we next performed histological and biochemical analyses of the adult hippocampus following transient Arc knockdown. Western blot analysis revealed a significant reduction in PSD95 protein levels in the hippocampus, whereas Homer1 expression remained unchanged (Fig. 5d, e). Immunofluorescence analysis confirmed a significant reduction in the number of PSD95- and VGLUT1-positive puncta compared with that in the Scramble control group, and the number of colocalized puncta tended to decrease (Fig. 5f, g). These results indicate a persistent reduction in synaptic integrity in adulthood, aligning with the cognitive impairments observed at this stage. To investigate the functional status of microglia, we analyzed CD68 expression in adult IBA-1⁺ microglia. Compared with the Scramble group, the Arc ASO group exhibited a significant increase in the intracellular volume of CD68 per microglial cell, indicating increased phagocytic activity (Fig. 5h–i). This sustained microglial hyperactivity may contribute to the observed synaptic loss in adulthood. Taken together, these findings reveal that transient Arc knockdown disrupts the development of neuroimmune balance, leading to enduring synaptic loss and aberrant microglial activation in the adult hippocampus. Transient Arc knockdown during the critical period disrupts adolescent microglial synaptic pruning Given that adolescence is another critical window for microglia-mediated synaptic pruning that is crucial for the fine-tuning of neural circuits, we evaluated the synaptic and microglial status in the hippocampus at P28 in mice subjected to transient Arc knockdown during the early critical period (Fig. 5a). Compared with those in the Scramble control group, Western blot analysis revealed that the expression of PSD95, Homer1, and Synaptophysin in the hippocampi of Arc ASO mice was significantly lower, whereas the expression of SAP-102 tended to increase (Fig. 5j, k). Compared with those in the scramble control group, the number of IBA-1⁺ microglia was significantly greater in the Arc ASO group, which was accompanied by an enlarged soma area and increased branching complexity, both of which are morphological indicators of an activated state (Fig. 5l, m). Additionally, the number of Homer1⁺ puncta engulfed by CD68⁺ microglia was significantly greater in the Arc ASO group than in the scramble control group, indicating increased synaptic phagocytosis (Fig. 5n, o). These findings indicate that the synaptic loss and microglial overactivation observed in adulthood begin to emerge as early as adolescence (P28), a period that represents a critical window for the second wave of synaptic pruning. These results further suggest that Arc knockdown during the early critical period may initiate a trajectory of neuroimmune dysregulation that starts in adolescence and persists into adulthood. Time-specific restoration of Arc during the critical period reverses both the dysregulation of microglial pruning and the impairment of cognition induced by repeated sevoflurane exposure To determine whether Arc overexpression could alleviate the abnormal phenotypes induced by repeated sevoflurane exposure, we constructed a doxycycline-inducible, neuron-specific Arc overexpression viral vector (ssAAV9-hSyn-Tet-on 3G-Arc), which was injected into the hippocampi of mice at P1. The mice were then subjected to repeated sevoflurane exposure from P6 to P8, and Arc expression was induced during the critical period (P14–P21) via maternal lactation following the administration of Dox in the drinking water (Fig. 6a). To directly assess Arc protein levels during the critical period, we performed Western blot analysis on hippocampal tissues across all four experimental groups. Our results demonstrated that the Dox-inducible Tet-On system effectively restored Arc expression in sevoflurane-exposed mice (Fig. 6b, c).Fig. 6Transient restoration of Arc expression during the critical period rescues the sevoflurane-induced dysregulation of microglial synaptic pruning and long-term cognitive deficits. a Experimental design. b–c Representative Western blots and quantification of ARC protein levels at P21, following transient Arc restoration at P14–P21 via the Tet-On system to compensate for the sevoflurane-induced reduction in Arc expression. Vehicle-Control vs. Dox-Control [one-way ANOVA, P = 0.0184 for Vehicle-Control vs. Dox-Control, P = 0.0486 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0049 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; n = 3 per group]. d–f Results of cognitive behavioral tests in adults following transient Arc restoration at P14–P21 via the Tet-On system to compensate for the sevoflurane-induced reduction in Arc expression. d Exploration trajectories and percentage of time and entries toward the novel object [time: one-way ANOVA, P = 0.2218 for Vehicle-Control vs. Dox-Control, P = 0.0076 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0095 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; entries: one-way ANOVA, P = 0.2146 for Vehicle-Control vs. Dox-Control, P = 0.0009 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0069 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; Vehicle-Control: n = 12 mice (5 males, 7 females), Dox-Control: n = 12 mice (7 males, 5 females), Vehicle-Sevoflurane: n = 9 mice (5 males, 4 females), Dox-Sevoflurane: n = 14 mice (8 males, 6 females)]. e Forgetting index for both contextual and cued recall in the CFC test [context: one-way ANOVA, P = 0.5760 for Vehicle-Control vs. Dox-Control, P = 0.0128 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0223 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; cue: one-way ANOVA, P = 0.4005 for Vehicle-Control vs. Dox-Control, P = 0.4796 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0904 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; Vehicle-Control: n = 8 mice (3 males, 5 females), Dox-Control: n = 14 mice (8 males, 6 females), Vehicle-Sevoflurane: n = 8 mice (5 males, 3 females), Dox-Sevoflurane: n = 14 mice (7 males, 7 females)]. f Spontaneous alternation rate in the Y-maze test [one-way ANOVA, P = 0.0005 for Vehicle-Control vs. Dox-Control, P = 0.0631 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0002 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; Vehicle-Control: n = 10 mice (4 males, 6 females), Dox-Control: n = 14 mice (8 males, 6 females), Vehicle-Sevoflurane: n = 9 mice (5 males, 4 females), Dox-Sevoflurane: n = 17 mice (9 males, 8 females)]. g–j Analysis of microglial synaptic pruning in the hippocampal CA1 region at P21 or P30 following transient Arc restoration from P14 to P21 via the Tet-On system to compensate for the sevoflurane-induced reduction in Arc expression. h, j Representative immunofluorescence staining images showing merged signals of IBA-1 (red), CD68 (green), Homer1 (far red), and DAPI (blue), a magnified view of CD68 (green)/Homer1 (far red) colocalization, and a 3D-rendered magnified image of Homer1 puncta within CD68 in IBA-1-labeled microglia. Scale bars = 10 μm or 5 μm. g, i Quantification of Homer1 puncta localized within CD68-positive microglial compartments [P21: Kruskal‒Wallis test followed by Dunn’s multiple comparisons test, P = 0.0115 for Vehicle-Control vs. Vehicle-Sevoflurane, P < 0.0001 for Vehicle-Sevoflurane vs. Dox-Sevoflurane, P > 0.9999 for Vehicle-Control vs. Dox-control; P = 0.2233 for Vehicle-Control vs. Dox-Sevoflurane; n = 12–19 cells from 3 mice per group; P30: one-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001 for Vehicle-Control vs. Vehicle-Sevoflurane, P < 0.0001 for Vehicle-Sevoflurane vs. Dox-Sevoflurane, P = 0.5835 for Vehicle-Control vs. Dox-Sevoflurane; n = 16 cells from 3 mice per group]. White arrows indicate representative examples of synaptic engulfment, demonstrating the localization of Homer1 within CD68-positive microglial compartments. Dox, doxycycline. ns, P > 0.05. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively. The data are presented as the mean ± SEM. Transient restoration of Arc expression during the critical period rescues the sevoflurane-induced dysregulation of microglial synaptic pruning and long-term cognitive deficits. a Experimental design. b–c Representative Western blots and quantification of ARC protein levels at P21, following transient Arc restoration at P14–P21 via the Tet-On system to compensate for the sevoflurane-induced reduction in Arc expression. Vehicle-Control vs. Dox-Control [one-way ANOVA, P = 0.0184 for Vehicle-Control vs. Dox-Control, P = 0.0486 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0049 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; n = 3 per group]. d–f Results of cognitive behavioral tests in adults following transient Arc restoration at P14–P21 via the Tet-On system to compensate for the sevoflurane-induced reduction in Arc expression. d Exploration trajectories and percentage of time and entries toward the novel object [time: one-way ANOVA, P = 0.2218 for Vehicle-Control vs. Dox-Control, P = 0.0076 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0095 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; entries: one-way ANOVA, P = 0.2146 for Vehicle-Control vs. Dox-Control, P = 0.0009 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0069 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; Vehicle-Control: n = 12 mice (5 males, 7 females), Dox-Control: n = 12 mice (7 males, 5 females), Vehicle-Sevoflurane: n = 9 mice (5 males, 4 females), Dox-Sevoflurane: n = 14 mice (8 males, 6 females)]. e Forgetting index for both contextual and cued recall in the CFC test [context: one-way ANOVA, P = 0.5760 for Vehicle-Control vs. Dox-Control, P = 0.0128 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0223 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; cue: one-way ANOVA, P = 0.4005 for Vehicle-Control vs. Dox-Control, P = 0.4796 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0904 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; Vehicle-Control: n = 8 mice (3 males, 5 females), Dox-Control: n = 14 mice (8 males, 6 females), Vehicle-Sevoflurane: n = 8 mice (5 males, 3 females), Dox-Sevoflurane: n = 14 mice (7 males, 7 females)]. f Spontaneous alternation rate in the Y-maze test [one-way ANOVA, P = 0.0005 for Vehicle-Control vs. Dox-Control, P = 0.0631 for Vehicle-Control vs. Vehicle-Sevoflurane, P = 0.0002 for Vehicle-Sevoflurane vs. Dox-Sevoflurane; Vehicle-Control: n = 10 mice (4 males, 6 females), Dox-Control: n = 14 mice (8 males, 6 females), Vehicle-Sevoflurane: n = 9 mice (5 males, 4 females), Dox-Sevoflurane: n = 17 mice (9 males, 8 females)]. g–j Analysis of microglial synaptic pruning in the hippocampal CA1 region at P21 or P30 following transient Arc restoration from P14 to P21 via the Tet-On system to compensate for the sevoflurane-induced reduction in Arc expression. h, j Representative immunofluorescence staining images showing merged signals of IBA-1 (red), CD68 (green), Homer1 (far red), and DAPI (blue), a magnified view of CD68 (green)/Homer1 (far red) colocalization, and a 3D-rendered magnified image of Homer1 puncta within CD68 in IBA-1-labeled microglia. Scale bars = 10 μm or 5 μm. g, i Quantification of Homer1 puncta localized within CD68-positive microglial compartments [P21: Kruskal‒Wallis test followed by Dunn’s multiple comparisons test, P = 0.0115 for Vehicle-Control vs. Vehicle-Sevoflurane, P < 0.0001 for Vehicle-Sevoflurane vs. Dox-Sevoflurane, P > 0.9999 for Vehicle-Control vs. Dox-control; P = 0.2233 for Vehicle-Control vs. Dox-Sevoflurane; n = 12–19 cells from 3 mice per group; P30: one-way ANOVA followed by Tukey’s multiple comparisons test, P < 0.0001 for Vehicle-Control vs. Vehicle-Sevoflurane, P < 0.0001 for Vehicle-Sevoflurane vs. Dox-Sevoflurane, P = 0.5835 for Vehicle-Control vs. Dox-Sevoflurane; n = 16 cells from 3 mice per group]. White arrows indicate representative examples of synaptic engulfment, demonstrating the localization of Homer1 within CD68-positive microglial compartments. Dox, doxycycline. ns, P > 0.05. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively. The data are presented as the mean ± SEM. In the adult behavioral assessments, Arc overexpression during the critical period fully rescued sevoflurane-induced deficits. In the NOR test, both exploration time and number of entries for the novel object in the Arc-rescue group were significantly greater than those in the vehicle-sevoflurane group. Additionally, compared to the vehicle control group, the Dox-Sevoflurane group did not significantly differ (Fig. 6d). Similarly, in the contextual fear conditioning test, the forgetting index was significantly lower in the Dox-Sevoflurane group than in the vehicle-sevoflurane group and returned to normal values that were comparable to those in the control group (Fig. 6e). However, in terms of the cue memory component of the CFC task, compared with the sevoflurane group, the Arc overexpression group exhibited a mild trend toward improvement, although the difference did not reach statistical significance (Fig. 6e). In the Y-maze test, Arc overexpression increased spontaneous alternation in both sevoflurane-exposed and nonexposed mice, restoring performance to the level of vehicle controls (Fig. 6f). Immunofluorescence analyses at P21 revealed that compared with the vehicle-sevoflurane group, the Arc-overexpressing group exhibited complete restoration of microglial synaptic phagocytic activity, as indicated by a significant increase in the number of Homer1 puncta within CD68-positive compartments to levels statistically indistinguishable from those in the vehicle-control group (Fig. 6g, h). By P30, after the cessation of Dox, Arc-overexpressing mice also showed normalized synaptic pruning activity, reflected by sustained Homer1/CD68 colocalization that was comparable to that in control animals, along with a marked reduction in sevoflurane-induced microglial hyperactivation (Fig. 6i, j). These findings indicate that transient Arc reintroduction during the critical period fully reverses the sevoflurane-induced abnormalities in microglial pruning function. To further investigate the cell-type specificity and sufficiency of Arc in regulating microglial synaptic pruning during the critical period, we constructed a neuron-specific ASO-resistant Arc-overexpressing viral vector (pSCAAV9-hSyn-Arc-P2A-eGFP). This vector was injected into the hippocampus at P1, followed by Arc ASO administration at P16. Microglial synaptic pruning was assessed at P21 (Fig. S7a). The results revealed that Arc overexpression alone did not alter microglial synaptic phagocytic activity at P21 (Fig. S7b, c), which is consistent with findings from the Dox-inducible Arc overexpression model. Importantly, after Arc ASO treatment, neuronal Arc overexpression effectively reversed the phagocytic deficits induced by transient Arc knockdown during the critical period (Fig. S7b, c). These findings confirm that restoring Arc expression in neurons is sufficient to reverse abnormalities in microglial synaptic pruning caused by the suppression of Arc expression during the critical period. Discussion Recent advances in anesthetic neurotoxicity research have highlighted the importance of defining developmental windows of vulnerability and identifying the molecular mechanisms underlying long-term cognitive impairments [24–27]. Our study provides clear and compelling evidence that the third postnatal week represents a critical period during which sevoflurane induces cognitive deficits. We demonstrate that a transient decrease in Arc protein levels during this period leads to long-lasting impairments in microglial function. Mechanistically, repeated exposure to sevoflurane induces prolonged abnormal activation of GSK3β, which alters the protein level of Arc during the critical period, impairing microglial synaptic phagocytosis and resulting in excessive premature synaptic retention during early development. This initial pruning deficit is associated with abnormal C1q synaptic recognition. As development progresses, the initially suppressed microglial pruning function becomes pathologically overactivated during adolescence, leading to excessive synaptic elimination, eventually leading to persistent cognitive deficits. Importantly, transient overexpression of Arc during the critical period successfully reversed both the microglial and synaptic abnormalities. Together, these findings establish a novel developmental mechanism of anesthetic neurotoxicity and suggest that Arc is a promising therapeutic target for preventing long-term neurodevelopmental damage. In 2016, the FDA issued a warning stating that repeated exposure to general anesthetics in children under three years of age could lead to serious consequences, a view that was supported by earlier clinical studies [1, 28, 29]. Subsequent research has extended this concern to preschool-aged children. Ramsaran et al. (2023) reported that the complete maturation of precise memory in mice occurs between P20 and P24 via a process critically dependent on neuronal activity in the hippocampal CA1 region during the third postnatal week [30]. This developmental window aligns with the typical emergence of episodic memory precision in humans of preschool age, reinforcing the biological plausibility of a critical period for anesthesia vulnerability [31, 32]. Our findings provide compelling experimental evidence to support this critical window. Some studies have suggested that an earlier developmental period may be more vulnerable [24]. However, the long-term impairments observed following sevoflurane exposure during postnatal week three underscore the importance of this particular window [33, 34]. Importantly, our data show that the cognitive deficits associated with sevoflurane exposure are no longer evident when individuals are exposed at P24-26. This time frame was selected in part because earlier suggestions that anesthetic sensitivity may correlate with the onset of sex hormone production [35]. However, notable fluctuations in hippocampal sex hormone levels were not detected at this stage, suggesting that circuit maturation, rather than endocrine changes, underlies the observed resilience [35, 36]. GSK3β, a serine/threonine kinase and one of the two isoforms of glycogen synthase kinase-3, plays a central role in neural development [37]. Dysregulation of GSK3β is associated with several neurodevelopmental and neurodegenerative disorders, including schizophrenia, autism, and Alzheimer’s disease [38–40]. Prior studies have shown that even transient activation of GSK3β during early development can lead to long-term cognitive deficits and synaptic abnormalities, which aligns with our findings on Arc regulation [41]. GSK3β has long been considered a key molecule in sevoflurane-induced cognitive dysfunction. Its prolonged activation contributes to multiple pathogenic pathways, including Tau hyperphosphorylation, the disruption of Wnt signaling and myelination defects [4, 16, 42, 43]. However, our results show that sevoflurane exposure during the fourth postnatal week alters GSK3β activity but does not impair cognitive function in adults. These findings suggest that the harmful effects of GSK3β are dependent on a critical period. We observed a strong interaction between GSK3β and Arc during the critical period, suggesting that GSK3β may promote Arc degradation by facilitating its phosphorylation [17]. Although GSK3β activation persists, the regulation of Arc during the critical period appears to be more directly linked to long-term cognitive and synaptic outcomes. This highlights the GSK3β-Arc pathway as a time-sensitive mechanism that might be essential in anesthesia-related neurodevelopmental impairment. Arc is a well-established “reverse tag” that is localized to inactive dendritic spines in response to reduced neuronal activity and promotes AMPA receptor endocytosis, thereby decreasing synaptic strength [21, 22, 44]. Arc has also been implicated in neuron-intrinsic mechanisms of synaptic elimination, such as autophagy [45]. However, whether glial involvement contributes to Arc-mediated synapse removal has not been clearly established. Our findings demonstrate that the manipulation of Arc expression during the critical period significantly alters microglial engulfment of synapses, suggesting that Arc may influence neuron‒microglia communication. C1q, the component that initiates the classical complement cascade, plays a critical role in activity-dependent synaptic pruning during early brain development. C1q binds to structurally or functionally weakened synapses and facilitates their removal by microglia via the receptor CR3 [23, 46]. While the downstream mechanisms of complement-mediated pruning are relatively well understood, the upstream signals that guide C1q to specific synaptic targets remain unclear. On the basis of the shared tendency of Arc and C1q to accumulate at weak or inactive synapses, we hypothesize that Arc may function as a postsynaptic mediator that facilitates C1q recruitment. Consistent with this idea, we observed strong colocalization of Arc and C1q at dendritic spines around P21 (Fig. 5j), which represents a peak period of synaptic pruning in the CA1 region, and both proteins were subsequently detected within microglia. These findings support the hypothesis that Arc contributes to the formation of a permissive synaptic environment for C1q deposition and complement-dependent synapse elimination. Although the precise molecular mechanism linking Arc to C1q remains to be elucidated, several possibilities can be considered. Arc may promote the exposure of “eat-me” signals such as phosphatidylserine, alter the synaptic glycoprotein landscape to favor complement binding, or modulate the release of chemotactic cues that recruit microglia [23, 47, 48]. Rather than interacting with C1q directly, Arc may create synaptic conditions that are conducive to complement tagging in low-activity states. In summary, Arc may serve not only as a key regulator of synaptic weakening but also as a molecular bridge connecting neuronal activity with immune-mediated synaptic remodeling. By interfering with Arc expression during the critical period to model the effects of neonatal sevoflurane exposure, we identified a biphasic pattern of synaptic remodeling in the hippocampal CA1 region: impaired synaptic pruning during early development, followed by excessive synaptic loss during adolescence. Before P21 and following sevoflurane exposure, microglia exhibit reduced pruning activity, leading to the retention of structurally unstable or premature synapses. From P28 onward, however, microglia became markedly hyperactive, accompanied by intensified synaptic elimination and a pronounced decrease in synaptic density. This maladaptive synaptic pruning pattern in microglia, while not previously described in detail, has been hinted at by prior observations in sevoflurane-exposed models [11, 49]. It also shows conceptual resemblance to findings in fragile X syndrome and early-life stress (ELS) models, suggesting that disruptions in the early neural microenvironment may exert sustained, cross-stage effects on brain structural development [50–52]. This process is likely driven by multiple interconnected mechanisms [52–56]. First, in our Arc knockdown model, microglia exhibited a hyporesponsive state during early development, leading to insufficient synaptic pruning and the accumulation of a premature synaptic burden. As development progresses, this delayed maturation may induce compensatory hyperactivation during adolescence. Second, the premature synapses retained because of early pruning deficits may gradually expose “eat-me” signals, such as phosphatidylserine externalization or desialylated glycans, increasing their susceptibility to complement tagging and microglial recognition. These aberrant synapses are likely to be preferentially eliminated during adolescence, a period marked by hormonal fluctuations and increased demand for synaptic remodeling. While the elimination of these aberrant synapses likely acts as the trigger that initiates microglial overactivation during adolescence, their transient presence cannot fully explain the long-term persistence of this state. Third, early-life disturbances may epigenetically “preprogram” microglia to enter a durable sensitized state, increasing their transcriptional sensitivity to subsequent inflammatory, complement-related, or neuronal signals and thereby maintaining the hyperactive microglial phenotype and excessive synaptic pruning from adolescence well into adulthood. This study has several limitations that should be acknowledged. First, transcriptome analysis revealed significant changes in gene expression levels and pathways related to myelination at both P21 and P56 following transient Arc knockdown during the critical period. Future studies to explore whether and how Arc influences oligodendrocyte development or the process of axonal myelination would be very interesting. Although Arc knockdown caused lasting abnormalities in microglia-mediated synaptic pruning, our findings were limited to cellular phenotype observations. The underlying molecular mechanisms, particularly the signaling interactions between neurons and microglia initiated by altered Arc expression, remain unclear. Further investigation into how neuronal Arc affects microglial behavior during the critical period could improve our understanding of the role of neuron–glia communication in synaptic development. In addition, differences between neonatal sevoflurane exposure and isolated Arc modulation were observed in the Y-maze spontaneous alternation test. This behavioral discrepancy is likely due to the mechanistic differences between the two models. Although Arc acts as a key mediator of the observed cognitive deficits, compared with the highly specific nature of isolated Arc modulation, sevoflurane exposure inherently induces a broader spectrum of molecular and cellular disruptions across multiple brain networks. Future studies incorporating additional spatial learning paradigms, such as the Morris water maze, may further clarify how these broad, early-life network alterations specifically impair hippocampus-dependent cognition. Finally, while we observed significant synapse loss in the hippocampus during adolescence and adulthood, which has been widely reported to correlate with cognitive impairments, we did not directly assess the functional consequences of these structural changes. We highlight the need for future studies combining electrophysiological or functional imaging approaches to directly establish the relationship between structural changes and circuit-level functional consequences. In summary, this study links the spatiotemporal specificity of anesthetic neurotoxicity to interactions between neurons and microglia, identifying Arc as a central mediator that connects neuronal activity with immune-driven synaptic maturation during development and acts as a decisive switch for long-term cognitive outcomes. Clinically, these findings provide novel mechanistic evidence for extending the window of vulnerability to the preschool years and suggest that temporally targeted modulation of Arc may offer a promising therapeutic strategy to prevent or mitigate anesthesia-related neurodevelopmental disorders early in life. Supplementary information Supplementary Figure 1 Supplementary Figure 2 Supplementary Figure 3 Supplementary Figure 4 Supplementary Figure 5 Supplementary Figure 6 Supplementary Figure 7 Supplementary information Supplementary Figure 1 Supplementary Figure 2 Supplementary Figure 3 Supplementary Figure 4 Supplementary Figure 5 Supplementary Figure 6 Supplementary Figure 7 Supplementary information