CX3CL1/CX3CR1-dependent microglial phagocytosis of oligodendrocyte precursor cells contributes to sevoflurane-induced myelination impairments in neonatal mice

Abstract

General anesthesia exposure in early life may disrupt the normal progression of developmental myelination, but the underlying mechanisms remain unclear. Early postnatal microglia in developing white matter exhibit diverse transcriptional and functional states, including a population with pronounced phagocytic activity. This study aims to investigate whether sevoflurane impairs oligodendrocyte myelination by promoting microglial phagocytosis of oligodendrocyte precursor cells (OPCs). Mice received either a single 2-h exposure to 3.3% sevoflurane on postnatal day 2 (P2) or repeated 2-h exposures on P2, P3, and P4. Neurobehavioral tests were used to assess cognitive and fine motor functions. The effects of microglial phagocytosis of OPCs on cerebral myelination were assessed using three-dimensional reconstruction and related biochemical analysis. Our findings indicate that repeated sevoflurane exposure, rather than a single exposure, induced hypomyelination and was associated with persistent cognitive and fine motor deficits. Repeated sevoflurane exposures promoted microglial activation and excessive phagocytosis of OPCs, thereby decreasing the number of oligodendrocytes and impairing developmental myelination. Mechanistically, repeated sevoflurane exposure enhanced CX3CL1/CX3CR1 signaling in neonatal brain tissue. In cultured microglial cells, sevoflurane altered Rac1-related cytoskeletal regulation and increased phagocytic activity. CX3CR1 knockout attenuated the phagocytic ability of microglia and rescued oligodendrocyte myelination and neurobehaviors. This study indicates that CX3CL1/CX3CR1-dependent microglial phagocytosis of OPCs might contribute to sevoflurane-induced myelination impairments in neonatal mice, and provides a potential therapeutic target for preventing anesthesia-induced developmental neurotoxicity.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12974-026-03997-8.

Introduction

Each year, millions of newborns worldwide are exposed to general anesthesia for surgical procedures or diagnostic interventions [1, 2]. Repeated or prolonged exposure to anesthetic agents during early life is recognized as a significant risk factor for abnormal brain development and neurobehavioral impairments, which is termed as anesthesia-induced developmental neurotoxicity [3]. Preterm infants represent a vulnerable subgroup of the neonatal population. Globally, approximately 15 million preterm births occur annually, accounting for more than 10% of all live births [4]. Due to the immaturity of organ systems, preterm infants are highly susceptible to a range of complications and often require surgical interventions under general anesthesia, sometimes on multiple occasions, for conditions such as necrotizing enterocolitis, intraventricular hemorrhage, and other congenital anomalies [5, 6]. Therefore, the effect of general anesthetics on brain development in preterm infants has become an increasing focus of concern.

In preterm infants, the developing brain is highly vulnerable to white matter injury, which can interfere with the normal course of myelination [7, 8]. Clinical evidence suggests a correlation between exposure to anesthesia and cerebral white matter myelination alterations, which might underpin the ensuing cognitive and behavioral abnormalities [9, 10]. The myelin sheath, a lipid-rich insulating layer that encases neuronal axons, is the core functional component of white matter [11]. Its proper formation is essential for the rapid and accurate transmission of nerve impulses and is fundamental to the synchronized activity of neural circuits [12, 13]. Several animal studies have shown that early-life exposure to general anesthesia leads to cerebral hypomyelination [14–16]. In the brain, oligodendrocytes (OLs) are the sole cell type responsible for myelin production [17]. The survival, proliferation, differentiation, and maturation of oligodendrocyte precursor cells (OPCs) are critical determinants of both the timing and integrity of myelination [17, 18]. Dysfunction in any of these key processes may therefore be a central driver of impaired oligodendrocyte development and disrupted myelination [19].

Microglia constitute approximately 10–20% of glial cells in the brain and serve as the primary resident phagocytes in the central nervous system (CNS) [20]. Single-cell transcriptomic analyses have shown that microglia in the developing brain are phenotypically diverse, with early postnatal white matter microglia displaying active metabolic features and strong phagocytic capacity [21–23]. During early postnatal development, a distinct population of amoeboid microglia migrates from the ventricular zone to the corpus callosum, where they actively phagocytose viable OPCs before the onset of myelination [6]. This process is critical for ensuring proper myelination, and microglial dysfunction may lead to developmental abnormalities characterized by dysregulated oligodendrocyte numbers and impaired myelination [24, 25]. Animal studies have demonstrated that early-life exposure to anesthetic agents could induce microglial activations, characterized by increased migration, phagocytosis, and proliferation, along with upregulation of the activation markers [26, 27]. However, the involvement of microglial phagocytosis in OPC changes following sevoflurane exposure remains poorly understood.

Given that brain development in rodent pups at postnatal days (P) 2–5 corresponds to the high-risk window for white matter injury in human preterm infants at 23–32 weeks of gestation [28, 29], we employed a neonatal mouse model to assess the effects of sevoflurane on oligodendrocyte development. Our findings indicate that repeated sevoflurane exposure, rather than a single exposure, led to hypomyelination and persistent impairments in cognition and fine motor function. Repeated sevoflurane exposures promoted microglial activation and excessive phagocytosis of OPCs, thereby decreasing the number of oligodendrocytes and impairing developmental myelination. Microglial phagocytic activity is tightly controlled by surface receptors that sense microenvironmental cues, among which the CX3C chemokine ligand 1 (CX3CL1, also known as fractalkine)/CX3C motif chemokine receptor 1 (CX3CR1) axis plays a key role in maintaining appropriate phagocytic function [30, 31]. Mechanistically, repeated sevoflurane exposure enhanced CX3CL1/CX3CR1 signaling in neonatal brain tissue. Experiments in cultured microglia further suggested that CX3CL1/CX3CR1 signaling may regulate Rac1-related cytoskeletal remodeling and phagocytic capacity toward OPCs. This study indicates that CX3CL1/CX3CR1-dependent microglial phagocytosis of OPCs might contribute to sevoflurane-induced myelination impairments in neonatal mice, and provides a potential therapeutic target for preventing anesthesia-induced developmental neurotoxicity.

Materials and methods

Animals and housing

All animal procedures were conducted in compliance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. The study protocol was reviewed and approved by the Institutional Animal Care and Use Committee of Shengjing Hospital of China Medical University, Shenyang, China (approval No. 2025PS1376K[X1]). C57BL/6 J mice were purchased from Beijing HFK BIOSCIENCE Co. Ltd. Heterozygous CX3CR1 knockout (KO) mice (CX3CR1^−/+) were purchased from Jiangsu GemPharmatech Co. Ltd. The line was maintained by successive backcrossing of heterozygotes. CX3CR-1GFP knock-in/knock-out mice were purchased from the Jackson Laboratory. Male pups were used in this study to reduce variability in later behavioral assessments. All mice were housed under controlled conditions, with free access to standard chow and water. All efforts were made to minimize the total number of animals used and their suffering.

Primary culture of microglia

Primary microglia were prepared from C57BL/6 J mice at P0-3. Briefly, brains were rapidly dissected, and meninges were removed in ice-cold PBS. The brain tissue was enzymatically dissociated with 0.25% trypsin (Procell System) at 37 °C for 15–20 min, followed by gentle trituration in DMEM containing 10% heat-inactivated fetal bovine serum (FBS; Procell System) and 1% penicillin–streptomycin (P/S; Procell System) to obtain a single-cell suspension. Cells were then plated in poly-L-ornithine-coated flasks. After 3 days, the medium was replaced with fresh culture medium supplemented with 10 ng/ml recombinant murine M-CSF (Procell System), 10% heat-inactivated FBS, and 1% P/S, and was subsequently changed every 3 days. After 10–14 days in culture, primary microglia were detached from mixed glial cultures by gently tapping the flasks. The harvested microglia were reseeded into culture plates and maintained at 37 °C in a humidified incubator with 5% CO₂.

Experimental design and exposure to anesthetic

The flow chart of the study protocol is presented in (Fig. 1A). The mice pups were randomly allocated to the sevoflurane or control groups by online simple randomization (https://www.randomizer.org/). The mice were exposed to one of three conditions: Control group (40% oxygen without sevoflurane, 2 h exposure on P2, P3 and P4 to match the total handling time of Sevo × 3 group), a single 2 h exposure (Sevo × 1, 3.3% sevoflurane in 40% oxygen on P2, 40% oxygen without sevoflurane, 2 h exposure on P3 and P4), or three 2 h exposures (Sevo × 3, 3.3% sevoflurane in 40% oxygen on P2, P3 and P4). The 3.3% sevoflurane was defined as 1 Minimum Alveolar Concentration (MAC), as previously reported [32]. Gas concentrations within the chamber, including sevoflurane, oxygen, and carbon dioxide, were monitored using a gas analyzer (Datex-Ohmeda Inc., Tewksbury, MA, USA). Oxygen concentration was maintained at 40% ± 1%, and carbon dioxide concentration was kept below 0.5% by adjusting the gas flow rate if necessary.​

Fig. 1.

Repeated neonatal exposure to sevoflurane led to long-term cognitive and fine motor deficits in mice. A Schematic of the experiment. B Schematic of single-pellet grasping test. C Success rate of single-pellet grasping test. D Schematic of the rotarod test. E Latency to fall on rotarod at constant speed. F Latency to fall on rotarod under uniform acceleration. G Schematic of Y-maze test. H Spontaneous alternation rate in Y-maze test. I Schematic of the new object recognition test. J Discrimination index (based on exploration duration) in novel object recognition test. K Discrimination index (based on number of explorations) in novel object recognition test. L Heat map of movement trajectory in novel object recognition test. n = 12 per group. Values are mean ± S.E.M. ∗ ∗ p < 0.01, compared with Control group; ∗ ∗ ∗ p < 0.001, compared with Control group. One-way analysis of variance followed by Tukey’s post hoc multiple comparison test was used for data analysis

Given the immature thermoregulatory system of neonatal mice, all mice were placed on a thermostatic heating pad (Fine Science Tools, Foster City, CA, USA) throughout the experimental process (including exposure and post-exposure recovery phases), with the pad temperature set to maintain an ambient temperature of 36.5 ± 0.5 °C in the chamber.

Arterial blood gases were measured from the left cardiac ventricle at the end of each 2 h gas exposure (n = 5 per group). Growth patterns of mice body weight were monitored at P2, P4, P7, P14, P28 and P56, respectively. The mice in each group were sacrificed at 6 h after the last gas exposure (n = 5 per group). From P38 to P55, the mice from each group were subjected to behavioral tests (n = 12 per group) by an investigator blinded to the grouping information.

Behavioral studies

All behavioral tests were conducted between 9:00 and 12:00 AM from P38 to P55 in a quiet room consistent with the animal housing conditions. Prior to each test, mice were acclimated to the test room for 30 min to minimize environmental stress. All tests were performed by investigators blinded to group allocation, and subsequent data analysis was conducted by another researcher also blinded to grouping. The test apparatus was thoroughly cleaned with 70% ethanol after each mouse and air-dried for 30 min to eliminate olfactory cues, with wooden objects in the novel object recognition test treated identically.

Single pellet grasping task

The single pellet grasping task evaluated fine motor function and motor learning ability, using a custom-made Plexiglas chamber (15 cm × 8.5 cm × 20 cm) with a 0.5 cm-diameter opening (2 cm above the chamber bottom) on one side, through which a 2 mg standard food pellet was manually delivered. Mild food restriction was initiated on P36 (3 days before training) to maintain mice at 90% of their free-feeding body weight, with free access to water. Body weight was monitored daily to adjust food intake and avoid excessive starvation. Training was conducted for 3 consecutive days (P39-41): each mouse was placed in the chamber for 5 min acclimation, followed by 20 trials (1 pellet per trial, 1 min interval between trials). A trial was recorded as failed if the mouse did not grasp the pellet within 2 min, with no latency recorded for failed trials. "Successful grasp" was defined as the mouse grasping the pellet with its forepaws (oral retrieval excluded) and bringing it to its mouth for consumption. On P42–43 (testing phase), each mouse completed 20 trials per day under identical conditions to training. The primary outcome measures success rate (number of successful grasps/total trials × 100%).

Rotarod test

The rotarod test assessed motor coordination and balance using a computerized rotarod apparatus (ZB-200, Techman, China) with five independent lanes (3 cm diameter). Training was conducted for 3 consecutive days (P45-47): each mouse was placed on the stationary rotarod for 10 s acclimation, followed by three training sessions using an accelerating mode from 4 to 40 rpm over 5 min, with a 10 min rest interval between sessions. On P48-50 (testing phase), mice were tested under both constant-speed and accelerating protocols. For the constant-speed protocol, the rotarod was set at 15 rpm. For the accelerating protocol, the speed increased from 4 to 40 rpm over 5 min. Each mouse completed three trials per day for each protocol, with a 10 min rest interval between trials. "Falling" was defined as the mouse’s four limbs completely detaching from the rotarod and contacting the bottom platform; fall latency was recorded as the time from rotation start to fall, or 300 s if the mouse remained for the full 5 min. The primary outcome measure was the average fall latency across the three trials for each protocol, with longer latencies indicating better motor performance.

Y-maze test

The Y-maze test evaluated spatial working memory using an opaque acrylic three-arm maze (each arm: 30 cm × 6 cm × 15 cm, 120° central angle, labeled A, B, C). To avoid arm preference bias, the blocked arm was balanced across groups (50% of mice with Arm A blocked, 50% with Arm B blocked). On P52 (acclimation phase), each mouse was placed in the central area and allowed to freely explore all arms for 5 min (valid arm entry defined as ≥ 50% of the body entering an arm). On P53 (test phase), the assigned arm was blocked, and the mouse explored the two open arms for 10 min to establish familiarity. After a 1 h intertrial interval, the blocked arm was opened, and exploration was recorded for 5 min. The movement trajectory of the mouse was recorded using a video tracking system (EthoVision XT 16, Noldus, Netherlands). The primary outcome measure was the spontaneous alternation behavior rate, calculated as [(number of consecutive entries into three different arms)/(total valid arm entries—2)] × 100%. A higher alternation rate indicates better spatial working memory.

Novel object recognition test

The novel object recognition test assessed non spatial cognitive memory using an opaque acrylic square arena (50 cm × 50 cm × 40 cm). On P54 (acclimation phase), each mouse was placed in the empty arena and allowed to explore freely for 10 min to adapt to the environment. On P55, the test was divided into two phases: familiarization and test. In the familiarization phase, two identical red wooden cylinders (3 cm diameter × 6 cm height) were placed at (10,10) cm and (40,40) cm (origin at the arena’s bottom-left corner), 10 cm from the nearest wall. The mouse was placed in the arena and allowed to explore the two objects for 10 min, with "exploratory behavior" defined as the nose touching the object (distance ≤ 2 cm, duration ≥ 1 s) or forepaws contacting the object (proximity without contact was excluded). The total exploration time of each object was recorded. After a 24 h retention interval, the test phase was conducted: one familiar cylinder was replaced with a green wooden cube (3 cm × 3 cm × 6 cm, similar volume to the cylinder), and the mouse was allowed to explore for 10 min. The exploration time of the familiar object (F) and novel object (N) was recorded. The primary outcome measure was the discrimination Index (DI), calculated as ((N—F)/(N + F)) × 100%. DI was compared to 50% using a one sample t-test, with DI significantly higher than 50% indicating intact cognitive memory.

Immunohistochemistry

Immunohistochemistry was performed on P56 mice to detect OLIG2 expression in brain tissues: following cardiac perfusion, brain tissues were fixed in paraformaldehyde and postfixed overnight at 4 °C. The tissues were then dehydrated, embedded, and coronally sectioned at 3-μm thickness; paraffin sections were then dewaxed, rehydrated, subjected to antigen retrieval in 0.01 M citrate buffer (pH 6.0) by microwave, quenched for endogenous peroxidase with 3% H₂O₂, blocked with 10% fetal bovine serum in PBS for 40 min, incubated overnight at 4 °C with rabbit anti-OLIG2 primary antibody (1:200, 13,999–1-AP, Proteintech), then with HRP-conjugated goat anti-rabbit IgG secondary antibody (1:200, SA00007-2, Proteintech) for 1.5 h at room temperature, visualized using DAB, counterstained with hematoxylin, dehydrated, cleared, and mounted with neutral balsam; finally, OLIG2⁺ cell images were acquired using a Nikon H600-L light microscope (Nikon, Japan) with a 40 × objective. For each animal, three coronal sections were analyzed, and three non-overlapping fields per region per section were captured from predefined regions of interest within the corpus callosum, striatum, and prefrontal cortex. OLIG2⁺ cells and positive staining areas were quantified using ImageJ with identical threshold settings across groups by investigators blinded to group allocation.

Electron microscopy

Striatal tissue was prepared for electron microscopic examination. The dorsal striatum was dissected into approximately 2 mm × 2 mm blocks and fixed for 4 h in 0.1 M cacodylate buffer (pH 7.4) containing 2% paraformaldehyde, 2.5% glutaraldehyde, and 2 mM CaCl₂. After fixation, samples were dehydrated sequentially in 30% and 50% acetone, followed by incubation overnight in 70% acetone saturated with uranyl acetate. Dehydration was then completed through increasing ethanol concentrations of 89%, 90%, and 100%. The specimens were embedded in Durcupan water-soluble resin (Head Biotechnology, Beijing, China) between ACLAR embedding films (Head Biotechnology) and polymerized at 55 °C for 72 h. Images were acquired using an electron microscope (Hitachi, Tokyo, Japan). Myelin sheaths were identified based on typical ultrastructural features, including compact multilamellar wrapping around axons, regular alternating electron-dense and electron-lucent layers, relatively uniform sheath thickness, and periodic transverse striations.

Immunofluorescence

Following cardiac perfusion, brain tissues were fixed in paraformaldehyde and postfixed overnight at 4 °C. The tissues were then dehydrated, embedded, and coronally sectioned at 4-μm thickness. Sections underwent deparaffinization and antigen retrieval, followed by a blocking step with 10% fetal bovine serum in PBS for 40 min. They were then incubated with primary antibodies overnight at 4 °C; the primary antibodies used include MBP (1:200; 10,458–1-AP, Proteintech, USA), OLIG2 (1:200; 13,999–1-AP, Proteintech, USA), IBA1 (1:100; AB283319, Abcam, USA), CD68 (1:100; AB303565, Abcam, USA), PDGFRα (1:200; 3174S, Cell Signaling Technology, USA) and PLP (1:200; DF13282, Affinity Biosciences, USA). Following primary antibody incubation, sections were rinsed and incubated for 4 h at room temperature with TRITC-conjugated goat anti-mouse IgG (1:200; SA00007-1; Proteintech, USA) and FITC-conjugated goat anti-rabbit IgG (1:200; SA00003-1, Proteintech, USA) secondary antibodies. Nuclei were counterstained with DAPI (C1006; Beyotime Biotechnology), and immunofluorescence images were captured using a Nikon H600-L light microscope (Nikon, Japan). Low-magnification images were acquired using a 20 × objective, and high-magnification images used for colocalization, microglial morphology, and engulfment-related analyses were acquired using a 40 × objective. The same acquisition settings were applied across experimental groups for quantitative comparisons.

Western blotting

Total protein was extracted from corpus callosum (CC), striatum (STR), and prefrontal cortex (PFC) tissue using lysis buffer. Equal amounts of protein were then resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred onto polyvinylidene fluoride membranes. The membranes were incubated overnight at 4 °C with primary antibodies against MBP (1:2000; 10,458–1-AP, Proteintech, USA), PLP (1:2000; DF13282, Affinity Biosciences, USA), CX3CR1 (1:2000; 66,915–1-Ig, Proteintech, USA), CX3CL1 (1:1000; 60,339–1-Ig, Proteintech, USA), Rac1 (1:500; 24,072–1-AP, Proteintech, USA), ACTR3 (1:5000; DF6209, Affinity Biosciences, USA), DIAPH1 (1:1000; 20,624–1-AP, Proteintech, USA), ROCK1 (1:50,000; 21,850–1-AP, Proteintech, USA), WASL (1:1000; AB126626, Abcam, USA), CYFIP2 (1:1000; AB204129, Abcam, USA), WASF1 (1:500; AF9229, Affinity Biosciences, USA), β-actin (1:4000; 66,009–1-Ig, Proteintech, USA), and GAPDH (1:5000; 10,494–1-AP, Proteintech, USA), followed by HRP-conjugated secondary antibodies (RGAR001, Proteintech, USA; RGAM001, Proteintech, USA). Immunoreactive bands were detected with enhanced chemiluminescence reagents, and band intensities were quantified densitometrically using ImageJ software (National Institutes of Health, Bethesda, MD, USA).

Cell processing and RNA sequencing

BV2 cells are easy to culture and exhibit robust proliferative capacity, allowing rapid acquisition of sufficient quantities of homogeneous cells. This effectively avoids contamination by non-microglial cells such as neurons and astrocytes that can occur during primary microglial isolation, thereby reducing confounding signals in sequencing results. BV2 cells were purchased from Procell Life Science & Technology Co. Ltd. BV2 microglia were maintained in DMEM containing 10% FBS and 1% penicillin–streptomycin at 37 °C under 5% CO₂. After random allocation to control and sevoflurane-treated conditions, cells in the sevoflurane group were exposed to 4% sevoflurane for 6 h in a sealed chamber, whereas control cells were kept in the same incubator without sevoflurane exposure [33, 34]. Total RNA was subsequently extracted from each group with TRIzol reagent. RNA purity (A260/A280 = 1.8–2.0) and integrity (28S/18S rRNA ratio≈2.0 via 1% agarose gel electrophoresis) were verified. The cDNA was synthesized from 1 μg of qualified total RNA using a PrimeScript RT Reagent Kit (Takara) to eliminate genomic DNA contamination. RNA sequencing was performed on the DNBSEQ platform by BGI Genomics (Beijing, China) with 3 biological replicates per group.

Bioinformatic analysis of RNA-seq data

Raw sequencing data were quality-controlled with FastQC (v0.11.9) to assess base quality (Q20/Q30), adapter contamination, and sequence duplication. Low-quality reads (Q20 < 90%, Q30 < 80%), adapters, and reads < 36 bp were filtered via Trimmomatic (v0.39; parameters: LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36), and duplicate reads were removed using Picard Tools (v2.27.5). Clean reads were aligned to the mouse GRCm38/mm10 genome (Ensembl) using HISAT2 (v2.2.1, –dta option). Alignment quality was evaluated with SAMtools (v1.15), retaining samples with mapping rate > 90%. Gene expression was quantified by StringTie (v2.2.1) and normalized to FPKM (Fragments Per Kilobase of transcript per Million mapped fragments). Low-expression genes (FPKM < 1 in more than 50% of samples in any group) were excluded before differential expressed genes (DEGs) analysis via DESeq2 (R v4.2.2). DEGs were defined as |log₂FC|> 1 and Padj < 0.05 (Benjamini Hochberg correction for multiple testing). Three DEGs (randomly selected from the top 10 up/down-regulated) were validated by qRT-PCR to confirm sequencing reliability. Functional annotation of DEGs was performed with clusterProfiler (R package): GO (Biological Process, Cellular Component, and Molecular Function; org.Mm.eg.db) and KEGG pathway enrichment analyses, with a significance threshold Padj < 0.05. Top 20 enriched GO terms and KEGG pathways were visualized as volcano plots, bubble plots, and bar charts via ggplot2 (R v4.2.2).

Microglial engulfment

Microglial engulfment was analyzed using IMARIS software (v9.9, Bitplane, Zürich, Switzerland) on immunofluorescent sections of P4 mouse brain (striatum region, n = 5 per group). First, the three-dimensional surfaces of OLIG2⁺ cells and Iba1⁺ cells were reconstructed using the "Surface" function of IMARIS, with parameters set as follows: minimum volume = 50 μm³, intensity threshold = 500, and background subtraction enabled. Subsequently, the IMARIS MATLAB-based (MathWorks) plugin "Split into Surface Objects" was used to assess the spatial association between reconstructed OLIG2⁺ surfaces and Iba1⁺ surfaces. For each animal, three coronal sections were analyzed, and five non-overlapping fields per section were captured at × 400 magnification. Thus, a total of fifteen fields per animal were included in the engulfment-related analysis. All image acquisition and quantification were performed by investigators blinded to group allocation.

ELISA for mouse fractalkine (FK)

The concentration of fractalkine (FK) in mouse brain tissue homogenate supernatants was detected using the Mouse Fractalkine (FK) ELISA Kit (ABclonal, Lot No. 9680007058, China) following the manufacturer’s instructions (https://abclonal.com.cn/). Brain tissues (corpus callosum, striatum, and prefrontal cortex, 50 mg per sample) were homogenized in 500 μL ice-cold PBS (pH 7.4) using a tissue grinder, then centrifuged at 12,000 rpm for 15 min at 4℃. The supernatant was collected and diluted 1:10 with the kit’s sample dilution buffer (within the kit’s detection range: 0.31–20 ng/mL). All samples were assayed in triplicate technical replicates, with blank controls (sample dilution buffer only) and standard controls included. The absorbance at 450 nm (reference wavelength: 630 nm) was measured immediately using a microplate reader (Cytation 5 Imaging Reader, BioTek, USA). The standard curve was constructed by plotting the absorbance values of standards against their concentrations using four-parameter logistic regression (R² > 0.99 to ensure fitting reliability). The concentration of FK in samples was calculated from the standard curve, and results were expressed as pg/mL (mean ± S.E.M, n = 5 per group).

Fluorescence probe in situ hybridization (FISH)

Fluorescence probe in situ hybridization (FISH) was performed on 30-μm fresh frozen brain sections using the RNASweAMI™ Multiplex Fluorescent Reagent Kit according to the manufacturer’s instructions (Servicebio, Wuhan, China). Sections were post-fixed with 4% paraformaldehyde in PBS for 15 min. CX3CL1, NeuN, PDGFRα, and GFAP signals were detected using their corresponding FISH probes. Probe information is provided in Supplementary Table 1. Hybridization with a probe against the Bacillus subtilis dihydrodipicolinate reductase (dapB) gene was used as a negative control. Images were acquired using a Digital Pathology Slide Scanner (Servicebio, China) with a 40 × objective, using identical acquisition settings across groups.

Quantitative reverse transcription polymerase chain reaction (qRT-PCR) array

Quantitative reverse transcription polymerase chain reaction (qRT-PCR) array was performed to validate DEGs from RNA-seq. The qRT-PCR array experiment utilizes PCR array chips (WC-MRNA0037-M, Wcgene® Biotech) supplied by Wcgene Biotechnology Corporation in Shanghai, adhering to the manufacturer's operational procedures. GAPDH was used as the internal control to normalize gene expression. Each sample was assayed in 3 technical replicates, with 3 biological replicates per group. Relative gene expression was calculated using the 2⁻ΔΔCt method. Differential expression was defined as |ΔΔCt|> 1 and p < 0.05. The comprehensive experimental protocol for qRT-PCR array analysis is available on the official website of Wcgene Biotechnology Corporation, Shanghai (http://www.wcgene.com).

Statistical analysis

All statistical analyses were performed using GraphPad Prism version 10.0 (GraphPad Software, USA) and SPSS version 24.0 (IBM, Armonk, NY, USA). The Shapiro–Wilk test was used to verify the normal distribution of data, and Levene’s test was used to test variance homogeneity. Data are presented as mean ± standard error of the mean (S.E.M) for consistency with previous experiments. Statistical methods were selected based on experimental design: (1) Unpaired t-test for comparisons between two groups (e.g., Control vs Sevo × 3 group in ELISA, FISH); (2) One-way ANOVA followed by Tukey’s post hoc test for multiple independent groups (e.g., WB, IHC, qRT-PCR). For in vivo experiments, each group contained 5 mice (non-behavioral) or 12 mice (behavioral); for in vitro experiments, each group contained 3 biological replicates, with 3 technical replicates per sample. There were no missing data, and no data were excluded. Between-group differences were considered statistically significant if two sided p < 0.05.

Result

Repeated neonatal exposure to sevoflurane led to long-term cognitive and fine motor deficits in mice

As shown in Supplemental Table 2, there were no significant differences among the Control group, the Sevo × 1 group, and the Sevo × 3 group on any measured variables for arterial blood gases values. Taking these measures reduced the possibility that sevoflurane-induced neurotoxicity in the immature brains was caused by physiologic side effects (e.g., hypoxia, hypercapnia). As shown in Supplemental Table 3, sevoflurane had no effect on body weight at P2, P4, P7, P14, P28 and P56, respectively.

To investigate the long-term effect of neonatal sevoflurane exposure on cognitive and motor functions in mice, we employed the single-pellet grasping test and rotarod test to evaluate fine motor skills and used the Y-maze test and novel object recognition test to assess cognitive ability (Fig. 1A).

In the single-pellet grasping test, mice in the Sevo × 3 group showed a significant decrease in reaching success rate compared with the Control group (Fig. 1B and C, Control vs Sevo × 3, p < 0.001), while no significant difference appeared between the Control group and the Sevo × 1 group. In the rotarod test at constant speed (15 rpm), latency to fall did not differ between the Control group and the Sevo × 1 group; however, it was significantly lower in the Sevo × 3 group than that in both the Control group and the Sevo × 1 group. In the acceleration test, latency to fall was markedly lower in the Sevo × 3 group (Fig. 1D–F, latency to fall at constant speed: Control vs Sevo × 3, p < 0.001; latency to fall at accelerating speed: Control vs Sevo × 3, p < 0.001). However, there was no statistical difference between the Control group and the Sevo × 1 group in the constant speed and acceleration rotarod tests.

In the Y-maze test, the spontaneous alternation rate did not differ significantly between the Control group and the Sevo × 1 group, whereas it decreased in the Sevo × 3 group (Fig. 1G and H, Control vs Sevo × 3, p < 0.01). Similarly, the novel object recognition test showed that cognitive ability decreased in the Sevo × 3 group compared with both the Control group and the Sevo × 1 group, as indicated by reduced exploration time of the novel object and shorter movement distance within the novel-object region (Fig. 1I–L, discrimination index of exploration duration: Control vs Sevo × 3, p < 0.001; discrimination index of exploration frequency: Control vs Sevo × 3, p < 0.01).

Repeated neonatal exposure to sevoflurane reduced the density of oligodendrocytes and induced hypomyelination in certain brain regions

Immunohistochemistry was used to detect the effects of single or multiple sevoflurane anesthesia on the number of OLIG2⁺ cells in different brain regions at P56 (Fig. 2A and S1). The results showed that in the CC, STR, and PFC, which are crucial brain regions for cognitive and motor abilities, the number of OLIG2⁺ cells in the Sevo × 3 group was significantly reduced (Fig. 2B, Control vs Sevo × 3, p < 0.05, p < 0.01, p < 0.05, in the CC, STR, and PFC, respectively), while there was no significant difference between the Control group and the Sevo × 1 group.

Fig. 2.

Repeated neonatal exposure to sevoflurane reduced the number density of oligodendrocytes and induced hypomyelination. A Representative OLIG2 immunohistochemical staining in the CC, STR, and PFC. B Quantification of OLIG2⁺ cell counts (cells/mm²) in the CC, STR, and PFC (n = 5/group). C Representative transmission electron micrographs of myelin in the STR for each group at P56. D Quantification of myelin area and G-ratio in the STR for each group at P56 (n = 3/group). E Representative MBP immunofluorescence in the CC, STR, and PFC. F Relative MBP fluorescence intensity in the CC, STR, and PFC (n = 5/group). G Representative Western blots for MBP and PLP in the CC, STR, and PFC. H Quantification of MBP protein levels (normalized to β-actin) by Western blot in the CC, STR, and PFC (n = 5/group). I Quantification of PLP protein levels (normalized to β-actin) by Western blot in the CC, STR, and PFC (n = 5/group). CC, corpus callosum; STR, striatum; PFC, prefrontal cortex. Values are mean ± S.E.M. ∗ p < 0.05, compared with Control group; ∗ ∗ p < 0.01, compared with Control group; ∗ ∗ ∗ p < 0.001, compared with Control group. One-way analysis of variance followed by Tukey’s post hoc multiple comparison test was used for data analysis

TEM and immunofluorescence were used to evaluate the effects of single or multiple sevoflurane anesthesia on myelin development. The TEM data showed that the proportion of myelin area in the Sevo × 3 group decreased significantly, and its g-ratio increased significantly (Fig. 2C and D, percentage of myelinated axons: Control vs Sevo × 3, p < 0.05; g-ratio: Control vs Sevo × 3, p < 0.001). There was no statistical difference between the Control group and the Sevo × 1 group. Both the relative fluorescence intensity and relative positive area of MBP and PLP were significantly reduced in the Sevo × 3 group compared with the Control group (Fig. 2E, F and S3A-D, relative fluorescence intensity of MBP, Control vs. Sevo × 3: p < 0.05, p < 0.01 and p < 0.01 in the CC, STR and PFC, respectively; relative positive area of MBP, Control vs. Sevo × 3: p < 0.01, p < 0.05 and p < 0.01 in the CC, STR and PFC, respectively; relative fluorescence intensity of PLP, Control vs. Sevo × 3: p < 0.01, p < 0.01 and p < 0.01 in the CC, STR and PFC, respectively; relative positive area of PLP, Control vs. Sevo × 3: p < 0.01, p < 0.01 and p < 0.001 in the CC, STR and PFC, respectively). There was no difference between the Control group and the Sevo × 1 group in the CC, STR, and PFC.

Quantification of the western blots showed that the protein levels of MBP and PLP were markedly decreased in the Sevo × 3 group. (Fig. 2G–I, MBP/β-actin: Control vs Sevo × 3, p < 0.01, p < 0.01, p < 0.01, in the CC, STR, and PFC, respectively; PLP/β-actin: Control vs Sevo × 3, p < 0.001, p < 0.001, p < 0.01, in the CC, STR, and PFC, respectively). There were no differences between the Control group and the Sevo × 1 group.

Repeated sevoflurane exposures promoted microglial activation and excessive phagocytosis of OPCs

OLIG2/PDGFRα double-immunofluorescence staining at P4 showed extensive co-expression of OLIG2 and PDGFRα, indicating that OLIG2⁺ cells at this developmental stage largely represented OPCs. Accordingly, OLIG2⁺ cells detected in the P4 engulfment-related analyses were interpreted as mainly reflecting OPCs (Supplementary Fig. S2).

Immunofluorescence staining with OLIG2 and CD68 respectively along with IBA1 was performed to detect the changes of microglia and OPCs in different brain regions of neonatal mice at P4 after multiple sevoflurane anesthesia. Statistical results showed that compared with the Control group, in the CC, STR, and PFC brain regions of the Sevo × 3 group, the number of OLIG2⁺ cells decreased significantly, while the number of IBA1⁺ cells increased significantly (Fig. 3A and B, OLIG2⁺ cell density: p < 0.05, p < 0.05, p < 0.05, in the CC, STR, and PFC, respectively; IBA1⁺ cell density: p < 0.01, p < 0.01, p < 0.01, in the CC, STR, and PFC, respectively). In addition, the number of IBA1⁺ CD68⁺ cells increased, and the relative mean fluorescence intensity of CD68 increased significantly in the Sevo × 3 group compared with the Control group (Fig. 3C and D, IBA1⁺ CD68⁺ microglia density: p < 0.05, p < 0.05, p < 0.01, in the CC, STR, and PFC, respectively; CD68 relative mean fluorescence intensity: p < 0.05, p < 0.05, p < 0.05, in the CC, STR, and PFC, respectively).

Fig. 3.

Repeated sevoflurane anesthesia in neonatal mice leads to excessive microglial phagocytosis of oligodendrocyte progenitor cells. A Representative OLIG2 and IBA1 immunofluorescence images from the CC, STR, and PFC. B Quantification of OLIG2⁺ and IBA1⁺ cell densities (cells/mm²) in the CC, STR, and PFC. C Representative CD68 and IBA1 immunofluorescence images from the CC, STR, and PFC. D Quantification of IBA1⁺CD68⁺ cell density (cells/mm²) and relative CD68 area in the CC, STR, and PFC. E Representative three-dimensional reconstructions of OLIG2 and IBA1 immunofluorescence and Sholl analyses of IBA1⁺ microglial morphology. F Quantification of microglial soma size, microglial process radius, and number of OLIG2 in IBA1 for each experimental group. CC, corpus callosum; STR, striatum; PFC, prefrontal cortex. n = 5 per group. Values are mean ± S.E.M. ∗ p < 0.05, compared with Control group; ∗ ∗ p < 0.01, compared with Control group; ∗ ∗ ∗ p < 0.001, compared with Control group. Independent t-test was used for data analysis

Three-dimensional reconstruction after OLIG2 and IBA1 immunofluorescence staining, therefore, was used to explore the spatial relationship between microglia and OPCs at P4 after multiple sevoflurane anesthesia in neonatal mice. Sholl analysis and statistical results showed that microglia were activated in the Sevo × 3 group, accompanied by phenomena such as an increase in soma size and a shortening of the ending radius. Moreover, the OLIG2-positive signals were significantly more abundant in IBA1⁺ cells in the Sevo × 3 group than that in the Control group (Fig. 3E and F, soma size: p < 0.01; ending radius: p < 0.01; OLIG2 in IBA1: p < 0.001, p < 0.05, p < 0.01, in the CC, STR, and PFC, respectively).

Repeated sevoflurane exposures upregulated the expression of CX3CR1 in microglia

To investigate the molecular mechanisms by which repeated sevoflurane exposures induce excessive microglial phagocytosis of OPCs, RNA-seq was performed on BV2 microglial cells after sevoflurane treatment. GO and KEGG enrichment analyses showed that the differentially expressed genes (DEGs) were significantly enriched in pathways related to chemokine-mediated signaling (Fig. 4A). The volcano plot further presented the expression distribution of DEGs in microglia, in which CX3CR1 expression was significantly upregulated (Fig. 4B).

Fig. 4.

Repeated sevoflurane exposure enhanced CX3CR1 expression in microglia. A GO and KEGG enrichment analysis of differentially expressed genes (DEGs) in BV2 microglial cells. B Volcano plot of differentially expressed genes in BV2 microglial cells (|log₂FC|> 1, Padj < 0.05). C Representative Western blot images showing CX3CR1 protein levels (normalized to GAPDH) in distinct brain regions of P4 neonatal mice. D Quantification of CX3CR1 protein expression across brain regions in P4 neonatal mice. E Representative double-immunofluorescence images for CX3CR1 and IBA1 in P4 GFP-CX3CR1 transgenic mice (GFP labels microglia). F Quantification of double-immunofluorescence staining for CX3CR1 and IBA1 in P4 GFP-CX3CR1 transgenic mice. CC, corpus callosum; STR, striatum; PFC, prefrontal cortex. n = 5 per group. Values are mean ± S.E.M. ∗ p < 0.05, compared with Control group; ∗ ∗ p < 0.01, compared with Control group; ∗ ∗ ∗ p < 0.001, compared with Control group. Independent t-test was used for data analysis

Western blot was used to measure CX3CR1 protein levels of P4 neonatal mice to validate the RNA-seq results (Fig. 4C). Quantification of the western blots showed that the gray values of CX3CR1 protein in the CC, STR, and PFC increased significantly in the Sevo × 3 group, indicating that repeated sevoflurane exposure upregulated CX3CR1 expression at the protein level (Fig. 4D, p < 0.05, p < 0.05, p < 0.05, in the CC, STR, and PFC, respectively).

Immunofluorescence staining of IBA1 was performed in CX3CR-1GFP knock-in/knock-out mice to clarify CX3CR1 cellular localization. Statistical results indicated that, in comparison to the Control group, the number of CX3CR1⁺ cells significantly increased in the Sevo × 3 group, and the number of IBA1⁺ CX3CR1⁺ cells also exhibited a significant increase (Fig. 4E and F, number of CX3CR1⁺ cells: p < 0.01, p < 0.01, p < 0.001, in the CC, STR, and PFC, respectively; number of IBA1⁺ CX3CR1⁺ cells: p < 0.05, p < 0.01, p < 0.05, in the CC, STR, and PFC, respectively).

To further characterize the cellular distribution of CX3CR1 and IBA1 signals, we quantified the density and proportion of CX3CR1⁻IBA1⁺, CX3CR1⁺IBA1⁺, and CX3CR1⁺IBA1⁻ cell populations in the CC, STR, and PFC (Supplementary Fig. S4). Among these three subpopulations, CX3CR1⁺IBA1⁺ cells accounted for the vast majority of CX3CR1/IBA1-labeled cells across the examined brain regions.

Repeated sevoflurane exposures enhanced CX3CL1 expression, with CX3CL1 signals detected in OPCs and neurons

We assessed the localization pattern of CX3CL1 signals by multiplex FISH using probes against CX3CL1, PDGFRα, NeuN, and GFAP. CX3CL1 signals were detected in association with PDGFRα probe-positive cells and NeuN probe-positive cells, whereas limited overlap was observed with GFAP probe-positive cells (Fig. 5A and S5). These data showed that CX3CL1 signals were present in cells labeled by OPC- and neuron-related markers under these experimental conditions.

Fig. 5.

Repeated sevoflurane exposures enhanced CX3CL1 expression, with CX3CL1 signals detected within PDGFRα⁺ and NeuN⁺ cells. A Representative in situ hybridization (FISH) images showing CX3CL1 signals in relation to PDGFRα, NeuN, and GFAP signals. B Relative CX3CL1 expression levels measured by qPCR in different brain regions of P4 neonatal mice. C Representative Western blot images for CX3CL1 in different brain regions of P4 neonatal mice. D Quantification of CX3CL1 protein levels (normalized to GAPDH) in different brain regions of P4 neonatal mice. E Relative CX3CL1 content in P4 neonatal mouse brain tissues measured by ELISA (pg/mL). CC, corpus callosum; STR, striatum; PFC, prefrontal cortex. n = 5 per group. Values are mean ± S.E.M. ∗ p < 0.05, compared with Control group; ∗ ∗ p < 0.01, compared with Control group; ∗ ∗ ∗ p < 0.001, compared with Control group. Independent t-test was used for data analysis

Real-time qPCR was used to measure CX3CL1 transcription in the CC, STR, and PFC of P4 neonatal mice. The results showed that relative CX3CL1 mRNA levels in the CC, STR, and PFC were significantly increased in the Sevo × 3 group, indicating that repeated sevoflurane exposure specifically enhanced CX3CL1 transcription (Fig. 5B, p < 0.01, p < 0.01, p < 0.001, in the CC, STR, and PFC, respectively).

Quantitative analysis of western blot indicates that CX3CL1 protein in the Sevo × 3 group was significantly higher than in the Control group (Fig. 5C and D, p < 0.001, p < 0.05, p < 0.01, in the CC, STR, and PFC, respectively). Additionally, we measured soluble CX3CL1 in the supernatant of brain homogenates from CC, STR, and PFC of P4 neonatal mice by ELISA to assess its secretion. The results showed that CX3CL1 levels in the Sevo × 3 group were significantly higher than in the Control group (Fig. 5E, p < 0.01, p < 0.01, p < 0.01, in the CC, STR, and PFC, respectively).

CX3CR1 knockout attenuated sevoflurane-induced neurobehavior deficits, hypomyelination, and excessive microglial phagocytosis of OPCs in mice

CX3CR1 knockout (CX3CR1^−/−) mice were constructed to explore the role of the CX3CL1/CX3CR1 pathway in mice exposed to multiple sevoflurane anesthesia (Fig. 6A). Multiple sevoflurane anesthesia resulted in fine motor dysfunction, as demonstrated by a decrease in the success rate in the single-pellet reaching test and a shortened latency to fall at the constant speed and accelerating speed in the rotarod tests, but these effects were blocked by CX3CR1 knockout (Fig. 6B–D, success rate: WT vs WT + Sevo × 3, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.01; latency to fall at constant speed: WT vs WT + Sevo × 3, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001; latency to fall at accelerating speed: WT vs WT + Sevo × 3, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001). There were no significant differences between the WT group and the KO group. Similarly, multiple sevoflurane administrations induced cognitive impairment, manifesting as a decrease in the spontaneous alternation rate in the Y-maze test and a reduction in the discrimination index (DI) of the exploration distance and frequency in the novel object recognition test. However, this effect was blocked by CX3CR1 knockout (Fig. 6E–H, spontaneous alternation rate: WT vs WT + Sevo × 3, p < 0.05; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001; discrimination index of exploration duration: WT vs WT + Sevo × 3, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001; discrimination index of exploration frequency: WT vs WT + Sevo × 3, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001). There were no significant differences between the WT group and the KO group.

Fig. 6.

Behavioral deficits induced by repeated sevoflurane exposure were ameliorated by CX3CR1 knockout. A Schematic of CX3CR1 knockout (KO) transgenic mouse generation and validation. B Success rate in the single-pellet grasping test for each group. C Latency to fall in the constant speed rotarod test for each group. D Latency to fall in the uniformly accelerated rotarod test for each group. E Spontaneous alternation rate in the Y-maze test for each group. F Discrimination index based on exploration duration in the novel object recognition test for each group. G Discrimination index based on exploration frequency in the novel object recognition test for each group. H Representative heat maps of movement trajectories from the novel object recognition test. n = 12 per group. Values are mean ± S.E.M. ∗ p < 0.05, compared with WT group; ∗ ∗ p < 0.01, compared with WT group; ∗ ∗ ∗ p < 0.001, compared with WT group; ∧p < 0.05, compared with WT + Sevo × 3 group; ∧∧p < 0.01, compared with WT + Sevo × 3 group; ∧∧∧p < 0.001, compared with WT + Sevo × 3 group. One-way analysis of variance followed by Tukey’s post hoc multiple comparison test was used for data analysis

Multiple sevoflurane anesthesia significantly reduced the number of OPCs in the CC, STR and PFC in the WT + Sevo × 3 group, and the number of OPCs was restored after CX3CR1 knockout at P56 (Fig. 7A and B, WT vs WT + Sevo × 3, p < 0.05, p < 0.01, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.05, p < 0.01, p < 0.001, in the CC, STR, and PFC, respectively). There were no differences between the WT group and the KO group. We found that multiple sevoflurane anesthesia in neonatal mice led to myelination disorders, manifested as a decrease in the relative area of myelin sheaths, an increase in the G-ratio in transmission electron microscopy, and a decrease in the relative fluorescence intensity and positive area of MBP and PLP immunofluorescence staining. Knockout of CX3CR1 attenuated the effects caused by multiple sevoflurane anesthesia (Fig. 7C–F and S3E-H, percentage of myelinated axon: WT vs WT + Sevo × 3, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001; G-ratio: WT vs WT + Sevo × 3, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.01; relative mean fluorescence intensity of MBP: WT vs WT + Sevo × 3, p < 0.01, p < 0.01, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.05, p < 0.05, p < 0.01, in the CC, STR, and PFC, respectively; relative positive area of MBP: WT vs WT + Sevo × 3, p < 0.001, p < 0.01, p < 0.05; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.01, p < 0.01, p < 0.01, in the CC, STR, and PFC, respectively; relative mean fluorescence intensity of PLP: WT vs WT + Sevo × 3, p < 0.001, p < 0.001, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.01, p < 0.001, p < 0.05, in the CC, STR, and PFC, respectively; relative positive area of PLP: WT vs WT + Sevo × 3, p < 0.01, p < 0.01, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001, p < 0.001, p < 0.01, in the CC, STR, and PFC, respectively). Similarly, there was no statistical difference between the WT group and the KO group. Quantitative analysis results of Western blot showed that the gray values of MBP and PLP protein in the CC, STR, and PFC regions of the WT + Sevo × 3 group were significantly lower than those in the WT group, and this was improved in the KO + Sevo × 3 group (Fig. 7G–I, MBP/β-actin: WT vs WT + Sevo × 3, p < 0.01, p < 0.001, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001, p < 0.001, p < 0.05, in the CC, STR, and PFC, respectively; PLP/β-actin: WT vs WT + Sevo × 3, p < 0.001, p < 0.05, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.01, p < 0.05, p < 0.001, in the CC, STR, and PFC, respectively). There were no significant differences between the WT group and the KO group.

Fig. 7.

Myelination deficits induced by repeated sevoflurane exposure was ameliorated by CX3CR1 knockout. A Representative OLIG2 immunohistochemical staining in the CC, STR, and PFC. B OLIG2⁺ cell counts (cells/mm²) in the CC, STR, and PFC (n = 5/group). C Representative transmission electron micrographs of myelin in the STR for each group at P56. D Quantification of myelin area and g-ratio in the STR for each group at P56 (n = 3/group). E Representative MBP immunofluorescence in the CC, STR, and PFC. F Relative MBP fluorescence intensity in the CC, STR, and PFC (n = 5/group). G Representative Western blots for MBP in the CC, STR, and PFC. H Quantification of MBP protein levels (normalized to β-actin) by Western blot in the CC, STR, and PFC (n = 5/group). I Quantification of PLP protein levels (normalized to β-actin) by Western blot in the CC, STR, and PFC (n = 5/group). CC, corpus callosum; STR, striatum; PFC, prefrontal cortex. Values are mean ± S.E.M. ∗ p < 0.05, compared with WT group; ∗ ∗ p < 0.01, compared with WT group; ∗ ∗ ∗ p < 0.001, compared with WT group; ∧p < 0.05, compared with WT + Sevo × 3 group; ∧∧p < 0.01, compared with WT + Sevo × 3 group; ∧∧∧p < 0.001, compared with WT + Sevo × 3 group. One-way analysis of variance followed by Tukey’s post hoc multiple comparison test was used for data analysis

Multiple sevoflurane anesthetics led to a decrease in the number of OPCs but an increase in microglia at P4. We verified the spatial relationship between microglia and OPCs using three-dimensional reconstruction and found that in the WT + Sevo × 3 group, microglia were significantly activated and showed increased phagocytosis of OPCs, accompanied by an enlarged soma size and a shortened ending radius. However, the knockout of CX3CR1 blocked this effect of multiple sevoflurane anesthetics (Fig. 8, OLIG2⁺ cell density: WT vs Sevo × 3, p < 0.05, p < 0.05, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.05, p < 0.05, p < 0.001, in the CC, STR, and PFC, respectively; IBA1⁺ cell density: WT vs WT + Sevo × 3, p < 0.001, p < 0.05, p < 0.001; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001, p < 0.01, p < 0.01, in the CC, STR, and PFC, respectively; IBA1⁺ CD68⁺ microglia density: WT vs WT + Sevo × 3, p < 0.01, p < 0.01, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.05, p < 0.05, p < 0.001, in the CC, STR, and PFC, respectively; CD68 relative mean fluorescence intensity: WT vs WT + Sevo × 3, p < 0.001, p < 0.01, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.001, p < 0.05, p < 0.05, in the CC, STR, and PFC, respectively). There was no difference between the WT group and the KO group.

Fig. 8.

Knockout of CX3CR1 inhibited excessive phagocytosis of OPCs by activated microglia following sevoflurane exposure. A Representative OLIG2 and IBA1 immunofluorescence images from the CC, STR, and PFC. B Quantification of OLIG2⁺ and IBA1⁺ cell densities (cells/mm²) in the CC, STR, and PFC. C Representative CD68 and IBA1 immunofluorescence images from the CC, STR, and PFC. D Quantification of IBA1⁺CD68⁺ cell density (cells/mm²) and relative CD68 area in the CC, STR, and PFC. E Representative three-dimensional reconstructions of OLIG2 and IBA1 immunofluorescence and Sholl analyses of IBA1⁺ microglial morphology. F Quantification of microglial soma size, microglial process radius and number of OLIG2 in IBA1 for each experimental group. CC, corpus callosum; STR, striatum; PFC, prefrontal cortex. n = 5 per group. Values are mean ± S.E.M. ∗ p < 0.05, compared with WT group; ∗ ∗ p < 0.01, compared with WT group; ∗ ∗ ∗ p < 0.001, compared with WT group; ∧p < 0.05, compared with WT + Sevo × 3 group; ∧∧p < 0.01, compared with WT + Sevo × 3 group; ∧∧∧p < 0.001, compared with WT + Sevo × 3 group. One-way analysis of variance followed by Tukey’s post hoc multiple comparison test was used for data analysis

The CX3CL1/CX3CR1 pathway regulates Rac1-related cytoskeletal rearrangement in cultured microglial cells

GO and KEGG enrichment analyses were performed on the cytoskeleton-associated DEGs detected by the PCR array. The analyses of DEGs showed that Rac1 may be a core downstream molecule through which the CX3CL1/CX3CR1 pathway regulates microglial functions. Meanwhile, the alterations in cytoskeleton molecules regulated by Rac1 were observed among the 22 cytoskeleton-related genes that exhibited significant differential expression. Specifically, ACTR3, DIAPH1, ROCK1, and WASL showed significant upregulation, whereas CYFIP2 and WASF1 demonstrated significant downregulation which may lead to the transformation of microglia from the resting state to the activated state (Fig. 9A). Sevoflurane exposure led to a decrease in Rac1 expression in primary microglia, as evidenced by a reduction in the relative fluorescence intensity of Rac1 in the immunofluorescence within microglia and a decrease in Rac1 protein content in the Sevo × 3 group. However, there was no decrease in Rac1 expression in microglia of KO + Sevo × 3 group (Fig. 9B, Rac1 fluorescence intensity: WT vs WT + Sevo × 3, p < 0.05; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.05; Rac1/β-actin: WT vs Sevo × 3, p < 0.01; WT + Sevo × 3 vs KO + Sevo × 3, p < 0.05). There was no difference between the microglia in the WT group and the KO group.

Fig. 9.

The CX3CL1/CX3CR1 pathway regulates Rac1-related cytoskeletal rearrangement in cultured microglial cells. A Representative PCR array results showing the expression of cytoskeleton-related genes in cultured microglial cells. B-H Representative immunofluorescence images and Western blot analyses (with quantification normalized to GAPDH) of RAC1 (B), ACTR3 (C), DIAPH1 (D), ROCK1 (E), WASL (F), CYFIP2 (G), and WASF1 (H) in cultured microglial cells. I Schematic diagram of the mechanism by which repeated sevoflurane anesthesia induces excessive microglial phagocytosis of OPCs, thereby affecting myelination development. n = 3 per group. Values are mean ± S.E.M. ∗ p < 0.05, compared with WT group; ∗ ∗ p < 0.01, compared with WT group; ∗ ∗ ∗ p < 0.001, compared with WT group; ∧p < 0.05, compared with WT + Sevo × 3 group; ∧∧p < 0.01, compared with WT + Sevo × 3 group; ∧∧∧p < 0.001, compared with WT + Sevo × 3 group. One-way analysis of variance followed by Tukey’s post hoc multiple comparison test was used for data analysis

In microglia exposed to sevoflurane, ACTR3, DIAPH1, ROCK1, and WASL exhibited significant upregulation, while CYFIP2 and WASF1 displayed significant downregulation, as indicated by alterations in the relative immunofluorescence intensity and protein content of microglia in the WT + Sevo × 3 group. However, in the NSC + Sevo × 3 group, the Rac1 inhibitor NSC 23766 effectively blocked the effects of sevoflurane exposure in microglia (Fig. 9C–H, relative mean fluorescence intensity: WT vs WT + Sevo × 3, p < 0.001, p < 0.001, p < 0.01, p < 0.01, p < 0.001, p < 0.01; WT + Sevo × 3 vs NSC + Sevo × 3, p < 0.01, p < 0.01, p < 0.001, p < 0.05, p < 0.001, p < 0.001, ACTR3, DIAPH1, ROCK1, WASL, CYFIP2 and WASF1, respectively; ACTR3/GAPDH: WT vs WT + Sevo × 3, p < 0.01; WT + Sevo × 3 vs NSC + Sevo × 3, p < 0.01; DIAPH1/GAPDH: WT vs WT + Sevo × 3, p < 0.001; WT + Sevo × 3 vs NSC + Sevo × 3, p < 0.01; ROCK1/GAPDH: WT vs WT + Sevo × 3, p < 0.01; WT + Sevo × 3 vs NSC + Sevo × 3, p < 0.05; WASL/GAPDH: WT vs WT + Sevo × 3, p < 0.05; WT + Sevo × 3 vs NSC + Sevo × 3, p < 0.05; CYFIP2/GAPDH: WT vs WT + Sevo × 3, p < 0.01; WT + Sevo × 3 vs NSC + Sevo × 3, p < 0.05; WASF1/GAPDH: WT vs WT + Sevo × 3, p < 0.01; WT + Sevo × 3 vs NSC + Sevo × 3, p < 0.05). There was no statistical difference between the WT group and the NSC group.

Discussion

Sevoflurane remains one of the most commonly used volatile anesthetics in neonatal and pediatric anesthesia due to its rapid onset and favorable hemodynamic profile. Our findings reveal that repeated, but not single, exposures to sevoflurane during the early postnatal period led to significant impairments in oligodendrocyte maturation and myelination, accompanied by deficits in cognitive performance and fine motor coordination. These behavioral abnormalities underscore the pivotal role of intact white matter architecture in supporting neurocircuit integrity and functional outcomes. White matter is predominantly composed of myelinated axonal fibers, serving as the principal substrate for communication between distributed brain regions. The myelin sheath, synthesized by terminally differentiated oligodendrocytes, facilitates saltatory conduction of action potentials and the synchronization of distributed brain networks. Disruption of myelination, whether due to developmental insults or demyelinating disease, has been shown to compromise learning, memory consolidation, attention, and motor precision in both preclinical models and human populations [35–37]. Importantly, our results align with accumulating evidence that anesthetic-induced developmental neurotoxicity demonstrates marked dependence on administration frequency. Our study extends these findings by identifying impaired white matter myelination as a critical and previously underappreciated pathological substrate contributing to this vulnerability.

OPCs constitute the proliferative pool that generates mature, myelinating oligodendrocytes. In the present study, we found that depletion of OPCs, mediated by excessive microglial engulfment, provides a plausible mechanistic link between repeated sevoflurane exposure and subsequent deficits in myelination and behavior. Immunofluorescence revealed that repeated sevoflurane exposure resulted a marked reduction in OPC density within the corpus callosum, as indicated by decreased OLIG2⁺ cell density, with additional OLIG2/PDGFRα staining supporting that OLIG2⁺ cells at P4 largely represented OPCs. This interpretation is consistent with previous reports that anesthetic exposures impair OPC proliferation and produce hypomyelination. Concurrently, high-resolution confocal analysis revealed a significant increase in microglial engulfment of OPCs, as indicated by increased OLIG2 signals within IBA1⁺ microglia. Microglia are known to regulate neural circuit maturation through synaptic pruning. Anesthetic agents may disrupt microglial-mediated synaptic remodeling, leading to inappropriate pruning or failure of pruning, thereby contributing to circuit dysfunction. It is recently reported that a subtype of microglia named “fountain of microglia”, appears only postnatally in myelinated regions such as corpus callosum and cerebellum, where they actively phagocytose OPCs prior to the onset of myelination [38, 39]. This specific subset of microglia arose from endogenous resident microglia pool but not circulating myeloid cells. Our findings extend this physiological paradigm by demonstrating that repeated sevoflurane exposure amplifies microglial engulfment activity beyond homeostatic limits, producing a net loss of OPCs sufficient to impair downstream myelination.

To interrogate mechanisms underlying the increased phagocytic activity of microglia following anesthetic exposure, we treated cultured BV2 cells with sevoflurane and performed transcriptome sequencing. Transcriptomic profiling revealed differential expression that was significantly enriched for chemokine signaling and phagocytosis/lysosome-associated pathways, implicating the CX3CL1/CX3CR1 signaling axis as a candidate mediator of the observed phenotypic shift. CX3CL1 is the sole member of the CX3C chemokine subfamily and exists in both membrane-anchored and soluble isoforms. It signals via a single known receptor, CX3CR1, which is expressed predominantly by microglia in the CNS. CX3CL1/CX3CR1 interactions orchestrate microglial synaptic pruning during postnatal development and facilitate the clearance of apoptotic cellular elements. Dysregulation of this chemokine pathway has been shown to precipitate aberrant or excessive microglial activation, thereby exacerbating neural injury, impairing synaptic integrity, and disrupting the coordinated progression of developmental neurobiological processes. Notably, recent work demonstrates that fractalkine-dependent microglial pruning regulates oligodendrocyte lineage dynamics and myelination [5, 38, 40]. In the present study, our findings indicate that microglial engulfment of OPCs was mediated, at least in part, by the CX3CL1/CX3CR1 signaling axis.

Our results showed increased CX3CL1 expression and detected CX3CL1 signals within NeuN⁺ and PDGFRα⁺ cells, suggesting a complex intercellular signaling network regulating microglial function during brain development. While CX3CL1 is classically recognized as a neuron-derived chemokine that modulates microglial activation and migration via CX3CR1, the detection of CX3CL1 signals in association with PDGFRα⁺ cells raises the possibility that OPCs may participate in CX3CL1-related communication with microglia [5, 41]. Such signaling may serve to finely tune microglial surveillance and phagocytosis under physiological conditions, maintaining homeostasis in the developing white matter. However, under pathological or stress conditions, such as sevoflurane exposure, increased CX3CL1 expression in this cellular environment may exacerbate microglial activation and promote excessive phagocytosis of OPCs, thereby contributing to white matter injury. These findings also emphasize the importance of OPCs not only as targets but also as active regulators of microglial behavior, suggesting novel avenues for therapeutic intervention that target the CX3CL1/CX3CR1 axis to protect OPCs and preserve white matter integrity during vulnerable developmental windows.

To investigate the intracellular mechanism by which the CX3CL1/CX3CR1 pathway mediates abnormal microglial phagocytosis of OPCs, we identified Rac1, a core Rho-family small GTPase, as a key downstream mediator of microglial regulation. Under normal conditions, Rac1 maintains microglia in a resting state with moderate phagocytic activity by controlling actin polymerization and microtubule assembly. Conversely, pathological stimuli can abnormally activate Rac1, prompting microglia to adopt a pro-inflammatory phenotype [42]. Previous studies have shown that CX3CL1-mediated actin polymerization and chemotaxis depend on the activation of Rac1 [43]. Our results indicated that CX3CR1 knockout effectively blocked these effects and restored Rac1 expression and distribution to physiological levels. Together with the in vitro microglial data, these findings support Rac1-related cytoskeletal remodeling as a downstream process linked to CX3CL1/CX3CR1-dependent microglial phagocytic activation.

Further research indicates that Rac1 activation specifically regulates the expression profiles of downstream cytoskeleton-related genes. After sevoflurane exposure, genes that promote cytoskeletal rearrangement, such as ACTR, DIAPH1, ROCK1, and WASL, were significantly upregulated, while inhibitory genes including CYFIP2 and WASF1 were notably downregulated. These changes produced microglial activation marked by an enlarged soma and shortened processes. They also promoted formation of phagocytic structures through actin polymerization and microtubule assembly, which ultimately enhanced excessive phagocytosis of OPCs. Intervention with the Rac1 inhibitor NSC 23766 reversed the abnormal expression of these cytoskeletal genes and restored microglial phagocytic function to homeostatic levels, further validating Rac1’s mediating role in cytoskeletal rearrangement and regulation of phagocytosis. The proposed mechanism is summarized schematically in (Fig. 9I.)

Our study has several limitations. First, to reduce variability related to the estrous cycle, we restricted behavioral testing to male offspring. Nevertheless, earlier studies indicate that vulnerability to anesthetic-induced neurotoxicity may differ between sexes. Therefore, the generalizability of our findings remains limited [44–46]. Second, OLIG2 was used to quantify oligodendroglial lineage cells, but it is not exclusively specific for OPCs. More stage-specific markers such as NG2, PDGFRα, and CC1 would further refine lineage-stage analysis. However, OLIG2/PDGFRα staining supported that OLIG2⁺ cells at P4 largely represented OPCs. Third, the present study did not distinguish active phagocytosis of viable OPCs from clearance of injured or apoptotic OPCs. Other processes, including altered OPC proliferation, differentiation, maturation, or cell death, may also contribute to OPC reduction. Fourth, FISH localization of CX3CL1 does not define the relative contribution of each cell type to total CX3CL1 expression, and additional cell-type-specific analyses would strengthen this conclusion. Finally, global CX3CR1 knockout may involve developmental compensation. Future studies using inducible or microglia-specific CX3CR1 manipulation would help separate acute signaling effects from developmental effects.

Conclusion

In conclusion, repeated sevoflurane exposures promoted microglial activation and excessive phagocytosis of OPCs, thereby decreasing the number of oligodendrocytes and impairing developmental myelination. Mechanistically, repeated sevoflurane exposure enhanced CX3CL1/CX3CR1 signaling in neonatal brain tissue, and cultured microglial cell data supported Rac1-related cytoskeletal remodeling as a downstream process involved in increased microglial phagocytic capacity toward OPCs.

Authors’ contributions

Study concept and design: Jiangming Lv, Ziyi Wu Acquisition of data: Jiangming Lv, Zhangting Xia, Xin Wang Analysis and interpretation: Jiangming Lv, Zhangting Xia, Xin Wang Writing – review and editing: Jiangming Lv, Ping Zhao, Ziyi Wu.

Funding

This work was supported by Liaoning Provincial Science and Technology Program Joint Plan (National Natural Science Foundation—General Program) (No. 2024-MSLH-599 to Ziyi Wu) and 345 Talent Project to Ziyi Wu.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.