Ozone-induced cognitive deficits are mediated by the liver–brain axis: peripheral complement C3 triggers microglial synaptic phagocytosis

Abstract

Ozone (O₃) is a significant global air pollutant. Recent epidemiological studies have established a correlation between O₃ exposure and an increased risk of neurological disorders. However, the underlying mechanisms by which O₃ induces cognitive deficits remain unclear. This study demonstrated that exposure to environmentally relevant O₃ levels resulted in significant cognitive impairment in mice. These deficits arose from hippocampal synaptic injury, characterized by reduced dendritic spine density, disrupted synaptic ultrastructure, and impaired long-term potentiation. Mechanistically, O₃ activated the liver complement pathway, leading to increased levels of complement component 3 (C3) and its subsequent release into the bloodstream. Furthermore, O₃ compromised the integrity of the blood–brain barrier, allowing peripheral C3 to infiltrate the hippocampus. Notably, C3 served as a key signal that triggered local pro-inflammatory microglial activation and enhanced their phagocytosis of excitatory synapses, ultimately resulting in synaptic loss and cognitive decline. Importantly, both the microglial inhibitor minocycline and liver-specific C3 knockdown suppressed pro-inflammatory microglial activation and restored synaptic plasticity and cognitive function. These findings systematically reveal a novel liver–brain axis in O₃ neurotoxicity, whereby peripheral C3 drives central microglial phagocytosis of excitatory synapses, offering new mechanistic insights and potential therapeutic targets for O₃-related neurological diseases.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03746-x.

Introduction

Air pollution continues to be a significant global public health challenge, contributing to approximately 8.1 million deaths each year and ranking as the second leading risk factor for mortality worldwide (https://www.stateofglobalair.org/). Over the past decade, while levels of particulate matter pollution have improved due to effective air pollution control measures [1], the concentration of tropospheric ozone (O₃) has been rapidly increasing [2]. This rise is driven by changes in environmental pollution sources and climate change [3], a phenomenon referred to as the “climate penalty.” As a highly active and strongly oxidative gaseous pollutant, O₃ has emerged as a significant global public health issue affecting human health [4]. In 2021, nearly 500,000 premature deaths worldwide were attributed to O₃ pollution, resulting in a loss of 7.14 million disability-adjusted life years [5, 6]. Traditionally, research has concentrated on the cardiopulmonary effects of O₃ exposure. However, emerging evidence suggests that O₃ also induces systemic consequences that affect extrapulmonary organs. Notably, associations between O₃ and central nervous system (CNS) disorders, such as cognitive decline, have been observed, which are potentially mediated by oxidative stress and neuroinflammation [7]. Despite these findings, the precise biological pathways that link O₃ exposure to neurotoxicity remain poorly defined, highlighting a significant knowledge gap in our understanding of the systemic health impacts of this pervasive pollutant.

The pathogenic mechanisms of O₃, a gaseous pollutant, differ significantly from those associated with particulate matter [8, 9]. Recent epidemiological and experimental evidence indicates that exposure to O₃ contributes to metabolic diseases, such as diabetes, by disrupting the regulatory functions of key metabolic organs [10]. The liver—a central hub for metabolic homeostasis—is particularly vulnerable to environmental stressors. Studies on animal models have demonstrated that prolonged exposure to O₃ induces progressive liver injury, alters serum metabolomes, and disrupts systemic metabolic equilibrium [11, 12]. The hepatic dysfunction is increasingly recognized within the context of inter-organ communication, particularly the liver–brain axis. The bidirectional crosstalk along this axis suggests that liver injury may have neurological manifestations, which can result in cognitive impairment, cerebral edema, and alterations in glial cells [13, 14]. These effects are likely mediated by circulating liver-derived factors and the disruption of the blood–brain barrier (BBB) [15, 16]. Clinical data from patients with nonalcoholic fatty liver disease have established a connection between metabolic dysregulation and cognitive deficits, as well as a reduction in brain volume [17, 18]. Additionally, animal studies have indicated that impaired hepatic clearance of amyloid-β exacerbates Alzheimer’s pathology [19], while circulating factors such as IL-17 A mediate alcohol-induced injury along the liver–brain axis [20]. However, the role of the liver–brain axis in O₃-induced neurotoxicity remains largely unexplored. Elucidating these mechanisms is essential for identifying novel biomarkers and developing therapeutic strategies that target multiple organs.

Within the CNS, microglia—the resident immune cells—play a crucial role as sentinels of brain homeostasis [21]. Microglia respond to various disturbances by modulating neuroinflammation and enhancing phagocytic activity [22, 23]. It has been demonstrated that the activation of microglia establishes a significant connection between peripheral insults, such as liver dysfunction, and CNS pathology [24, 25]. Specifically, in models of acute liver failure and metabolic dysfunction–associated steatotic liver disease, microglial activation significantly correlated with the severity of encephalopathy and cognitive decline [26–28]. Microglia are highly dynamic cells that constantly survey the brain environment [29, 30]. They play a crucial role in maintaining brain homeostasis and regulating neuronal circuits through their dynamic processes of surveillance and subsequent interactions with neurons. Microglia monitor neuronal activity by contacting both pre- and postsynaptic elements, assessing their activity, and pruning synapses [31–33]. Interestingly, not only do microglia interact with excitatory neurons during neurodevelopment and certain neurodegenerative diseases, particularly at axonal boutons and dendritic spines [34, 35], but they also eliminate inhibitory synapses in these contexts [36, 37]. Under pathological conditions, microglia may engage in aberrant synaptic pruning through various phagocytic signaling pathways, potentially leading to synaptic loss and neuronal dysfunction [38, 39]. This process is often mediated by the complement system, particularly complement component 3 (C3) [40], which is primarily synthesized in the liver and circulates in the peripheral system [41]. Evidence from autoimmune and neuropsychiatric disorders, such as systemic lupus erythematosus and depression, indicates that peripherally derived complement proteins can infiltrate the brain, activate microglia through the C3/CR3 axis, and promote excessive synaptic phagocytosis [42–44]. However, it remains unclear whether this mechanism operates under environmentally triggered conditions, such as O₃ neurotoxicity, and whether complement derived from the liver acts as a crucial mediator in the crosstalk between the liver and the brain.

This study aimed to elucidate the mechanism by which O₃ exposure induces cognitive dysfunction through the liver–brain axis regulation of the complement C3 signaling pathway while also exploring potential intervention targets. We established an animal model with varying concentrations of O₃ exposure and employed a combination of behavioral tests, molecular biology, and multi-omics techniques. Our findings showed that O₃ exposure compromised the integrity of the BBB, thereby facilitating the entry of liver-derived C3 into hippocampal tissue. By assessing the neuroinflammatory response, microglial phagocytic function, and the expression changes of pre- and postsynaptic markers, we confirmed that C3 activation of microglia led to abnormal phagocytosis of excitatory synapses, resulting in synaptic loss and cognitive dysfunction. Furthermore, through minocycline intervention and liver-specific C3 knockdown experiments, we verified that blocking this pathway effectively ameliorated neurological damage. Ultimately, our study revealed the extensive impact of O₃ on the CNS via peripheral organs, providing new targets and theoretical foundations for the prevention and treatment of cognitive impairments associated with environmental pollutants.

Materials and methods

Animals and exposure conditions

Male C57BL/6 mice (7–8 weeks old, weighing 20–25 g) were obtained from the Hubei Provincial Center for Laboratory Animals and housed at the Laboratory Animal Center of Wuhan University of Science and Technology. CX3CR1-GFP mice (JAX: 005582) were obtained from Cavens Biogle (Suzhou) Model Animal Research Co. Ltd. and bred at the same facility. CX3CR1-GFP mouse genotyping was performed using PCR analysis of tail DNA, and Supplementary Table S1 shows the specific sequence. All of the mice were maintained under controlled environmental conditions (22–23 °C, 50–70% relative humidity) with a 12-hour light/dark cycle. This study was conducted in strict adherence to the guidelines for the care and use of laboratory animals and received approval from the Ethics Committee of Wuhan University of Science and Technology (Issue No. WUST-2022076).

Mice were placed in a glass chamber measuring 50 × 50 × 50 cm (length × width × height) and were exposed to either filtered air or an O₃ environment generated by a KTB portable O₃ generator (0.6 ± 0.2 ppm or 1.2 ± 0.2 ppm) (CH-KTB, CHUAVG, Guangzhou, China). The O₃ concentration was adjusted using a mass flow controller and monitored in real-time with an MS400 O₃ sensor (Eranntex, Shenzhen, China). The experimental environment was maintained at a temperature of 21 ± 2 °C. Mice were exposed to O₃ for 4 h daily (fixed time period: 11:00–15:00, corresponding to the resting phase) for a continuous duration of 30 days. The rationale for dose selection was based on several factors: (1). Toxicology studies indicate that stationary rodents require exposure to O₃ concentrations 3–5 times higher than those of exercising humans to induce similar levels of inflammation [45–47]. This implies that the toxicity of 0.6 ppm in resting rodents is equivalent to that of 0.12 ppm in exercising humans; (2). The dose of 0.12 ppm is comparable to summer O₃ pollution levels in Chinese cities (200 µg m^-3) [48]; (3). In areas with severe O₃ pollution, concentrations often reach 0.15–0.25 ppm [49, 50], which justifies setting 0.6 ppm as a relative environmental dose and 1.2 ppm as a high dose to investigate the dose-response effect; (4). O₃ peaks during summer (especially in June) typically occur from noon to afternoon, lasting approximately 4 h [51], which is why this study established a daily exposure of 4 h for 30 consecutive days. These concentrations were chosen to simulate real-world exposure scenarios and have been widely employed in previous toxicological studies.

Drug treatments

Minocycline was dissolved in sterile physiological saline. During the 30-day O₃ exposure period, mice in the experimental group received intraperitoneal injections of minocycline (13614-98-7, MedChemExpress, New Jersey, USA) at a dosage of 30 mg Kg^-1 every 2 days, followed by exposure to O₃ 1 h after each injection for a total of 30 days. In contrast, mice in the control group received injections of an equivalent volume of physiological saline.

Adeno-associated virus construction and infection

To investigate the effects of C3 knockdown in the liver on neuropathic-related pathology induced by O₃ exposure, we injected two types of viruses (200 µL, dosage: 5 × 10¹² vg kg^-1) into 7- to 8-week-old C57BL/6J mice using an intravenous visual mouse tail injection fixator (XR-YLS-Q9G, Shanghai Xinruan Information Technology Co., Ltd., China). The viruses included a liver-specific C3 knockdown virus [rAAV-TBG-EGFP-5‘miR-30a-shRNA2(C3)-3’-miR30a-WPREs] and a control virus [rAAV-TBG-EGFP-5‘miR-30a-shRNA2(Scramble)-3’-miR30a-WPREs] (BrainVTA, Wuhan, China). After 2 weeks of viral expression, the mice were exposed to O₃ to establish the model, followed by behavioral and molecular experiments. Supplementary Table S2 lists the detailed sequences used in the viral design.

Behavior tests

To evaluate the effects of O₃ exposure on mouse behavior, we conducted a series of behavioral assessments, including the Morris water maze, open field test, novel object recognition test, and Y-maze test. Supplementary Text S1 provides detailed experimental protocols for each behavioral assessment.

Pathological staining

In this experiment, we used hematoxylin–eosin (H&E) staining and toluidine blue staining to investigate the histopathological characteristics of the hippocampus and liver in the examined mice. Supplementary Text S2 presents detailed experimental methods.

Immunofluorescence staining (IF)

The hippocampal tissue sections of mice were incubated with the appropriate primary antibodies, followed by IF staining to localize and analyze the expression of target proteins. The experiment utilized isoflurane to induce deep anesthesia in mice, followed by cardiac perfusion with pre-cooled 0.9% saline solution. The brain tissue was then rapidly excised and fixed in 4% paraformaldehyde (PFA) for a duration of 24 h, after which it underwent gradient sucrose dehydration at concentrations of 15% and 30%. Once the tissue had sunk to the bottom, it was embedded in Tissue-Tek Oct Compound (Sakura, USA) and frozen at -20 °C for equilibration. Coronal serial sections of 40 μm thickness were prepared using a cryostat (CM1950, Leica Biosystems, Germany). The sections were washed with phosphate-buffered saline (PBS) and permeabilized with 0.3% Triton X-100 (prepared in PBS) at room temperature for 30 min to enhance antibody penetration. Subsequently, the sections were blocked for 1 h in PBS containing 10% normal goat serum (Jackson, USA) to minimize non-specific binding. Following blocking, the sections were incubated with specific primary antibodies overnight at 4 °C. The next day, after thorough washing with PBS, fluorescently labeled secondary antibodies were added and incubated at room temperature in the dark for 1 h. The nuclei were counterstained with DAPI (1:100, SL7101, Coolaber, China) for 5 min. Finally, the sections were rinsed with PBS, mounted with an anti-fade mounting medium, and coverslipped to prevent bubble formation. Fluorescence signals were captured using a confocal laser scanning microscope (FV3000, Olympus, Japan), and high-resolution imaging was performed in Z-stack mode (1 μm thickness) to ensure a clear presentation of the three-dimensional structure. All image acquisition parameters (laser intensity, gain, exposure time, etc.) were kept consistent to ensure data comparability. Supplementary Table S3 lists the specific antibodies used. Supplementary Text S3 describes detailed experimental procedures and analytical methods.

Fluorescence micro-optical sectioning tomography (fMOST) whole-brain imaging and quantification

After anesthetizing CX3CR1-GFP mice, perfusion was sequentially performed using ice-cold saline, followed by 4% paraformaldehyde (PFA). The brain was then post-fixed in 4% PFA for 24 h and subsequently dehydrated in a gradient ethanol solution. Following resin embedding, the sample was transferred to a container containing 0.01 M phosphate-buffered saline (PBS) and 0.01 M Na₂CO₃. Fluorescence signals were acquired using structure illumination-based whole-brain fMOST, achieving a voxel resolution of 0.32 μm × 0.32 μm × 2 μm. The sample surface was scanned layer by layer in a line-scanning mode through the movement of a three-dimensional precision platform. Finally, the preprocessed data were visualized and converted into graphical representations using Imaris software (v.9.7.2, Bitplane, Switzerland). To quantify the total number of microglia (GFP) in the mouse hippocampus, we employed Imaris software. The Surfaces-Mask function was used to isolate the hippocampus, followed by the application of the Spots function to calculate the number of microglial cells present. Additionally, the Snapshot and Animation functions in Imaris were used to generate images.

Golgi-Cox staining

The FD Rapid Golgi Stain™ Kit (FD Neuro Technologies, Columbia, MD, USA) was used to conduct Golgi staining on mouse brain tissue to evaluate potential changes in the density and morphology of dendritic spines in hippocampal neurons. The experiment commenced with the induction of deep anesthesia in the mice using isoflurane. All experimental procedures strictly adhered to the instructions provided in the kit, with detailed operational steps available in Supplementary Text S4. The staining results were captured using a high-resolution microscope (VS120, Olympus Corporation, Japan), and the analysis of dendritic spine density and morphology was conducted using ImageJ software.

Transmission electron microscopy (TEM)

After anesthesia induction, the mice were immediately subjected to intracardiac perfusion with 0.9% saline, followed by a fixative for TEM (Servicebio). The target tissues were rapidly excised and trimmed into 1-mm³ cubes using a sharp blade in precooled fixative. Subsequently, the samples were post-fixed in the fixative and stored at 4 °C for 4 h. The tissue blocks were then transferred to a 1% osmium tetroxide (OsO₄) solution in 0.1 mol L^-1 PBS and fixed at room temperature in the dark for 2 h. Next, the tissues were dehydrated using a graded series of ethanol (30%, 50%, 70%, 80%, 95%, and 100%), followed by immersion in 100% acetone. Finally, the tissues were embedded, sectioned, and cut into 60-nm ultrathin sections. The ultrastructure was observed using a TEM (HT7700, HITACHI, Japan).

Serological testing

Serological testing in the mice primarily involved the determination of serum complement C3, along with alanine aminotransferase (GPT/ALT) and aspartate aminotransferase (GOT/AST). Supplementary Text S5 presents detailed information.

Western blot analysis

The expression levels of relevant proteins in the hippocampus and liver of the mice were assessed using western blot analysis. Supplementary Text S6 and Supplementary Table S4 present a comprehensive description of the experimental procedures.

Quantitative real-time PCR

Gene expression levels in mouse tissue samples were measured using quantitative reverse-transcription polymerase chain reaction (qRT-PCR) technology. Supplementary Text S7 and Supplementary Table S5 detail the relevant experimental parameters.

RNA-sequencing (RNA-seq) analysis and 4D-DIA quantitative proteomic analysis

RNA-seq transcriptome sequencing technology was employed to analyze gene and transcript expression in the hippocampal and liver tissues of the mice from different groups. Concurrently, 4D-DIA quantitative proteomics technology was used for the quantitative analysis of proteins in serum samples. The Majorbio Cloud platform facilitated bioinformatics analysis of the transcriptomic and proteomic data. Supplementary Text S8 provides detailed experimental methods.

Microelectrode array experiments

The experiment employed a medium-throughput 64-channel MED64-Quad II system (Alpha MED Scientific, Panasonic, Japan) to record extracellular field potentials. Field excitatory postsynaptic potentials (fEPSPs) from CA1 neurons were obtained by stimulating the Schaeffer fibers in the CA2 region. Long-term potentiation (LTP) was induced through high-frequency stimulation (HFS) at 100 Hz for a duration of 1 s. The magnitude of LTP was determined as the average of the responses recorded 50–60 min following the conditioning stimulus, normalized to the baseline. All signals were captured using the MED64 system. Supplementary Text S9 presents detailed experimental methods.

Statistical analysis

Data analysis was performed using GraphPad Prism 8 software (GraphPad Software, USA). The experimental results are presented as mean ± standard error of the mean (mean ± SEM). In the statistical graphs for specific experimental results, such as immunofluorescence, the primary data (based on animal means, where n represents the number of animals) are illustrated as violin plots, with corresponding individual measurements displayed in scatter plots on the right (indicated by small dots). All statistical analyses were performed exclusively on the basis of animal means; individual measurements in scatter plots were not included in the statistical tests. Differences between two groups were statistically analyzed using Student’s t test, while comparisons among multiple groups were performed using one-way analysis of variance. P values lower than 0.05 were considered statistically significant.

Result

O₃ exposure damages synapses in the mouse hippocampus, leading to cognitive decline

To investigate the effects of subchronic O₃ exposure on cognitive behavior, a mouse model was established using a gradient concentration protocol (0, 0.6, 1.2 ppm, 4 h/day for 30 days) (Fig. 1A). Behavioral tests revealed that exposure to 1.2 ppm O₃ significantly reduced the movement distance ratio in the center zone of the open field (Supplementary Fig. S1D) without affecting the total distance moved (Supplementary Fig. S1E), indicating anxiety-like behavior. In cognitive assessments, both the 0.6 and 1.2 ppm groups exhibited significantly prolonged escape latency on day 5 in the Morris water maze test (Fig. 1B). However, only the 1.2 ppm group demonstrated significant reductions in target quadrant residence time, path ratio, and platform crossings on day 7 (Fig. 1D–F), suggesting impaired spatial memory. In the Y-maze and novel object recognition tests, the 1.2 ppm group showed significantly reduced spontaneous alternation rates and preference indices (Supplementary Fig. S1G, K), with no differences in motor activity (Supplementary Fig. S1H, J, L), further confirming deficits in working and recognition memory. Although the body weight of mice in the 1.2 ppm group was lower than that of the control group on day 30 (Supplementary Fig. S1A), their weekly average feed consumption exhibited no significant difference (Supplementary Fig. S1B), and basic motor function remained unaffected. Consequently, the 1.2 ppm concentration was selected for subsequent experiments.

Fig. 1.

O₃ exposure induces synaptic plasticity impairment and cognitive dysfunction in the hippocampus of mice. A Timeline representation of O₃ exposure and experimental procedures. B Statistical graph of escape latency in mice during the first 5 days of the learning stage (n = 12 per group). C Representative swim path trace images in the Morris water maze probe trial. D–G Statistical graphs of time in the target quadrant D, platform crossings E, proportion of path length in the target quadrant F, and swimming velocity G during the testing phase (n = 12 per group). H, I Protein expression levels of spinophilin, PSD95, and synaptophysin in the hippocampus (n = 5 per group). J, K TEM images of the mouse hippocampal (scale bar, 2 μm) and the associated quantitative analysis, with red pentagrams indicating synaptic locations (n = 5/8 per group). Yellow-boxed areas show higher magnification views of presynaptic (blue) and postsynaptic (red) terminals (scale bar, 500 nm). L Schematic of the placement of MED64 probes on a coronal hippocampal slice. M Representative fEPSP traces before and after LTP induction (n = 4 per group). N Amplitude of fEPSP after HFS recorded on hippocampal slices. O Quantitative analysis of normalized fEPSPs during the last 10 min after HFS (n = 4 per group). *P < 0.05, **P < 0.01, and ***P < 0.001

H&E staining results indicated that the cells in the CA1, CA3, and DG regions of the hippocampus in the control group exhibited normal morphology and an orderly arrangement (black arrows). Conversely, in the ozone group, cells in these regions displayed a sparse and disorganized arrangement characterized by loosely distributed cells (black arrows), along with nuclear pyknosis and vacuolation (red arrows) (Supplementary Fig. S1M). Toluidine blue staining further revealed that the cells in the CA1, CA3, and DG regions of the hippocampus in the control group maintained normal morphology and an orderly arrangement (black arrows), with distinct, granular bluish-purple Nissl bodies visible within the neurons (red arrows). In contrast, cells in the corresponding regions of the ozone group exhibited a sparse and disorganized arrangement (black arrows), alongside pale staining and a reduced number of Nissl bodies (red arrows) (Supplementary Fig. S1N). Furthermore, Golgi staining illustrated a decrease in the dendritic complexity of neurons in the CA1 region in the O₃-exposed group (Supplementary Fig. S1P, Q), accompanied by a significant reduction in dendritic spine density (Supplementary Fig. S1R). At the synaptic level, western blot analysis indicated that the expression of three critical synaptic proteins, namely PSD95, synaptophysin, and spinophilin, was downregulated in the hippocampal tissue of O₃-exposed mice (Fig. 1H, I). Moreover, we used TEM to assess ultrastructural changes in synapses, and the results revealed a significant decrease in both synaptic density and postsynaptic density (PSD) thickness (Fig. 1J, K). LTP of fEPSPs in the hippocampus is considered fundamental to learning and memory. Electrophysiological recordings demonstrated a significant reduction in LTP within the CA3–CA1 pathway of the hippocampus in the ozone group (Fig. 1L–O), suggesting a specific impairment in synaptic plasticity. These findings indicate that exposure to 1.2 ppm O₃ can cause structural damage to hippocampal neurons, synaptic loss, and a decline in synaptic plasticity in mice, which is associated with various cognitive behavioral deficits.

O₃ exposure triggers complement C3-mediated microglial phagocytosis of excitatory synapses

To investigate the molecular mechanisms underlying hippocampal damage induced by O₃ exposure, we conducted transcriptome sequencing on mouse hippocampal tissues. Principal component analysis (PCA) demonstrated robust sample reproducibility (Fig. 2A). A total of 361 differentially expressed genes (DEGs) were identified, comprising 146 upregulated and 215 downregulated genes (Fig. 2B). The volcano plot revealed that several DEGs, including Il6ra, Slc385a, and Lcn2, were associated with inflammation and synaptic function (Fig. 2C). KEGG analysis indicated that these DEGs were significantly enriched in pathways related to serotonergic synapses, cholinergic synapses, and GABAergic synapses, along with the IL-17 signaling pathway, among others (Fig. 2D; Supplementary Fig. S2A). The integration of KEGG enrichment analysis results with the DEG expression heat map further illustrated the distribution of gene expression differences (Supplementary Fig. S2B). Through heatmap visualization analysis of phagocytosis-related pathways, we observed a significant upregulation of multiple phagocytosis-related genes in the ozone group, which exhibited distinct differences when compared to the control group (Supplementary Fig. S2C, D). The results of RT-qPCR and western blot analyses further confirmed that O₃ exposure led to an increase in the expression of pro-inflammatory factors in the hippocampus (Supplementary Fig. S2E–F), suggesting that neuroinflammation plays a role in O₃-induced hippocampal damage.

Fig. 2.

O₃ exposure leads to impairment of excitatory synapses in the hippocampus. A PCA of hippocampal transcriptome sequencing samples in mice (n = 3 per group). B DEGs in the hippocampus. C A volcano plot illustrating DEGs in the hippocampus. D KEGG enrichment analysis of hippocampal tissues. E Representative immunofluorescence images of VGLUT2 (purple) and PSD95 (green) co-localization in the hippocampus, along with orthogonal views and Imaris spot analysis images (scale bar, 5/0.5 μm). Yellow dashed circles indicate VGLUT2⁺/PSD95⁺ co-localized puncta; red arrows point to the corresponding positions of the selected co-localized puncta in the orthogonal views. F Quantification of the percentage of VGLUT2⁺ puncta, PSD95⁺ puncta, and VGLUT2⁺/PSD95⁺ co-localized puncta in the hippocampus (n = 5 per group). G Representative immunofluorescence images of Gephyrin (green) and VGAT (purple) co-localization in the hippocampus, along with orthogonal views and Imaris spot analysis images (scale bar, 5 μm). Yellow dashed circles indicate Gephyrin⁺/VGAT⁺ co-localized puncta; red arrows point to the corresponding positions of the selected co-localized puncta in the orthogonal views. H Quantification of the percentage of Gephyrin⁺ puncta, VGAT⁺ puncta, and Gephyrin⁺/VGAT⁺ co-localized puncta in the hippocampus (n = 5 per group). ****P < 0.0001

To investigate the impact of O₃ exposure on synaptic plasticity, we employed IF staining to assess the effects of O₃ exposure on both excitatory and inhibitory synapses. The analysis of excitatory synapse markers (Fig. 2E) revealed a significant decrease in the number of PSD95⁺ and co-localized VGLUT2⁺/PSD95⁺ synapses in the O₃-exposed group, while the number of VGLUT2⁺ synapses remained unchanged (Fig. 2F). This indicated that O₃ exposure primarily affected the postsynaptic components, leading to a notable loss of excitatory synapses. In contrast, the analysis of inhibitory synapse markers showed that O₃ exposure did not result in significant changes to the number of Gephyrin⁺ puncta, VGAT⁺ puncta, or Gephyrin⁺/VGAT⁺ co-localized puncta (Fig. 2G, H). These findings suggested that O₃ exposure predominantly affected the postsynaptic components, resulting in a significant loss of excitatory synapses.

Based on previous sequencing data indicating that O₃ exposure can activate pathways related to phagocytosis and inflammation, we subsequently directed our research towards microglia. To investigate the effects of O₃ exposure on microglia, in this study, we used fMOST technology to conduct whole-brain imaging on CX3CR1-GFP transgenic mice (Fig. 3A). Selected cognitive-related brain regions, including the hippocampus, olfactory bulb, and medial prefrontal cortex, were identified using brain atlases, and a volume of 200 × 200 × 50 μm was randomly selected in each region. The number of cells was quantified using Imaris software (Fig. 3B, C; Supplementary Fig. S3A, B). The results demonstrated a significant increase in the number of microglia in the hippocampus of the O₃-exposed group. IF staining further confirmed a significant elevation in the number of microglia in the hippocampus of the ozone group (Fig. 3D). Three-dimensional reconstruction and Sholl analysis revealed a reduction in branch complexity and an increase in soma size, consistent with pro-inflammatory microglial activation state (Fig. 3E–G; Supplementary Fig. S3C). Astrocytes exhibited no significant changes in either quantity or morphology (Fig. 3H–J; Supplementary Fig. S3D), and there were no notable alterations in the expression of A1/A2 activation markers (Supplementary Fig. S3E). These findings indicate that O₃ exposure can specifically activate pro-inflammatory microglia in the hippocampal region and induce morphological changes, suggesting that abnormal pro-inflammatory microglial activation playmicroglials a critical role in O ₃-induced cognitive dysfunction.

Fig. 3.

O₃ exposure induces pro-inflammatory microglial activation in the hippocampus of mice. A Schematic of the workflow for the collection and preparation of CX3CR1-GFP mouse brain tissue and the fMOST whole-brain imaging procedure. B Representative images of GFP fluorescence imaging in the whole brain and hippocampus of CX3CR1-GFP mouse by fMOST (scale bar, 500/30 μm). C Quantitative analysis of cx3cr1+ cell density in the CX3CR1-GFP mouse hippocampus (n = 3 per group). D Representative images of GFAP (red) and NEUN (blue) immunofluorescence staining in the hippocampus of CX3CR1-GFP (green) mice (scale bar, 30/10 μm). Representative pictures of 3D reconstruction of astrocytes and microglia (scale bars, 10 μm). E Branching structure of microglia demonstrated by Sholl analysis (n = 4 per group). F Analysis and quantification of microglial density (n = 4 per group). G Analysis and quantification of total branch points in microglia (n = 4 per group). H Branching structure of astrocyte demonstrated by Sholl analysis (n = 4 per group). I Analysis and quantification of astrocyte density (n = 4 per group). J Analysis and quantification of total branch points in astrocyte (n = 4 per group). *P < 0.05, and **P < 0.01

Based on the activation of phagocytosis-related pathways suggested by transcriptome sequencing results (Supplementary Fig. S2C, D), and considering the complement system as a classical pathway of microglial activation, we further investigated the role of complement components in synaptic loss through immunofluorescence colocalization experiments (Fig. 4A, D, I). The results indicated that the fluorescence intensity of PSD95⁺ in the hippocampal region of mice decreased following O₃ exposure (Fig. 4B, K), while the fluorescence intensity of complement component C3 significantly increased (Fig. 4G, J). Morphological analysis of microglia revealed an increase in cell body volume and a reduction in total branch length, indicating a typical pro-inflammatory microglial activation state (Fig. 4E, F). Colocalization analysis revealed that, compared to the control group, the ozone group exhibited a significantly higher number of Iba1⁺/PSD95⁺ colocalized puncta (Fig. 4C), C3⁺/Iba1⁺ colocalized puncta (Fig. 4H), and C3 deposits within PSD95⁺ structures (Fig. 4L). However, immunofluorescence detection demonstrated that O₃ exposure did not cause significant changes in the expression level of the complement initiation component C1q (Supplementary Fig. S4A, B). These results suggest that C3 may specifically accumulate around excitatory synapses, thereby promoting the phagocytosis of synapses by pro-inflammatory microglia.

Fig. 4.

Complement C3 mediates microglial phagocytosis of excitatory synapses. A Representative immunofluorescence images of Iba1 (red) and PSD95 (green) co-localization in the hippocampus (scale bar, 20 μm), orthogonal views (scale bar, 5 μm), and 3D reconstruction analysis images (scale bar, 2 μm). White arrows indicate PSD95 puncta on microglia. B Quantitative analysis of PSD95 intensity in the hippocampus of mice (n = 5 per group). C Quantitative analysis of the number of PSD95⁺ puncta within Iba1 cells in the hippocampus (n = 5 per group). D Representative immunofluorescence images of Iba1 (purple) and C3 (green) co-localization in the mouse hippocampus, with corresponding 3D reconstructions (scale bar, 10/8/2 µm). White arrows indicate C3 puncta within microglia. E, F Analysis and quantification of microglial morphological parameters including soma size and total dendritic length (n = 5 per group). G Quantitative analysis of C3 intensity in the hippocampus of mice (n = 5 per group). H Quantitative analysis of normalized C3⁺/Iba1⁺ colocalized puncta in the hippocampus of mice (n = 5 per group). I Representative immunofluorescence images of C3 (green) and PSD95 (red) co-localization in the hippocampus, with corresponding orthogonal views (scale bar, 5 μm). White arrows indicate the positions of the selected C3⁺/PSD95⁺ co-localized puncta in the orthogonal views. J–L Quantification of the percentage of C3 deposits, PSD95⁺ puncta, and C3⁺/PSD95⁺ co-localized puncta in the hippocampus (n = 5 per group). *P < 0.05, **P < 0.01 and ***P < 0.001

Minocycline alleviates O₃-induced cognitive deficits by inhibiting microglial synaptic phagocytosis

To investigate the role of microglia in O₃-induced neurotoxicity, we administered minocycline, a broad-spectrum agent known to effectively inhibit the activation of pro-inflammatory microglia (Fig. 5A). Minocycline, a derivative of tetracycline, is capable of crossing the blood-brain barrier (BBB). Behavioral experiments, including the open field test, Y-maze, novel object recognition, and Morris water maze, demonstrated that minocycline significantly ameliorated cognitive dysfunction induced by O₃ exposure (Fig. 5B–G; Supplementary Fig. S5B–K). Additionally, RT-qPCR, western blot, Golgi staining, and electron microscopy results demonstrated that minocycline reduced the expression of pro-inflammatory factors, restored the expression of synaptic-related proteins, enhanced dendritic complexity and spine density, and improved synaptic ultrastructure (Supplementary Fig. S5L–R; Fig. 5H–L). LTP experiments confirmed that minocycline significantly improved synaptic plasticity in O₃-exposed mice (Fig. 5M–O).

Fig. 5.

Minocycline ameliorates O₃-induced cognitive dysfunction and hippocampal synaptic impairment. A Timeline representation of O₃ exposure, minocycline administration, and experimental procedures. B Statistical graph of escape latency in mice during the first 5 days of the learning stage (n = 13/17 per group). C Representative swim path trace images in the Morris water maze probe trial. D–G Statistical graphs of the proportion of path length in the target quadrant D, time in the target quadrant E, platform crossings F, and swimming speed G during the testing phase (n = 13/17 per group). H–L TEM images of the mouse hippocampal (scale bar, 2 μm) and the associated quantitative analysis, with red arrows indicating synaptic locations (n = 5 per group). Yellow-boxed areas show higher magnification views of presynaptic (blue) and postsynaptic (red) terminals (scale bar, 500 nm). M Representative fEPSP traces before and after LTP induction (n = 4 per group). N, O Amplitude of fEPSP after HFS recorded on hippocampal slices N, and quantitative analysis of normalized fEPSPs during the last 10 min after HFS O (n = 4 per group). *P < 0.05, **P < 0.01, and ***P < 0.001

IF and three-dimensional reconstruction results revealed that O₃ exposure led to an increase in the number of hippocampal microglia, morphological activation, and enhanced phagocytic activity, as evidenced by elevated Iba1⁺/CD68⁺ co-localization. In contrast, minocycline intervention markedly suppressed this activation (Supplementary Fig. S6A–I; Fig. 6A–C). Further analysis indicated an increase in PSD95⁺ granules within microglial cells in the O₃ group, alongside a decrease in the overall PSD95 fluorescence signal in the hippocampus. This indicates that pro-inflammatory microglia mediate synaptic loss through enhanced synaptic phagocytosis. In contrast, minocycline treatment substantially attenuated the phagocytic activity of microglia toward synaptic structures and effectively alleviated synaptic loss (Fig. 6D–F).

Fig. 6.

Minocycline alleviates the enhanced phagocytic capacity of pro-inflammatory microglia induced by O₃ exposure. A Representative immunofluorescence images of Iba1 (red) and CD68 (green) co-localization in the hippocampus (scale bar, 30/7 µm), along with 3D reconstruction analysis (scale bar, 7 μm). B Quantification of Iba1⁺ cell density in the hippocampus (n = 5 per group). C Statistical chart of the volume proportion of CD68 in Iba1 fluorescence of microglia (n = 5 per group). D Representative immunofluorescence images of Iba1 (red) and PSD95 (green) co-localization in the hippocampus (scale bar, 10 μm), orthogonal views (scale bar, 5 μm), and 3D reconstruction analysis images (scale bar, 2 μm). White arrows indicate PSD95 puncta on microglia. E Quantitative analysis of PSD95 intensity in the hippocampus of mice (n = 5 per group). F Quantitative analysis of the number of PSD95⁺ puncta within Iba1 cells in the hippocampus (n = 5 per group). G Relative mRNA levels of C1q and C3 in the mouse hippocampus (n = 6 per group). H, I Protein expression levels of C3 in the mouse hippocampus. (n = 6 per group). *P < 0.05, **P < 0.01 and ***P < 0.001

To further investigate the regulatory effect of minocycline on the complement system, we assessed the mRNA and protein expression levels of complement C3 in the hippocampus of mice following drug administration. The results indicated that O₃ exposure did not affect the mRNA expression level of hippocampal C3 (Fig. 6G); however, it significantly elevated C3 protein content. Importantly, minocycline treatment did not mitigate the increase in C3 protein induced by O₃ exposure (Fig. 6H, I). These findings suggest that the elevated hippocampal C3 protein resulting from O₃ exposure may not stem from local synthesis but rather from the peripheral system.

In summary, the findings demonstrate that minocycline significantly alleviates synaptic damage, declines in synaptic plasticity, and cognitive dysfunction induced by O₃ exposure. This effect is achieved by inhibiting the abnormal activation of pro-inflammatory microglia and their synaptic phagocytic activity. Furthermore, the elevated levels of hippocampal complement C3 protein induced by O₃ may originate from peripheral sources and are not regulated by minocycline.

Liver-derived complement C3 is a crucial factor in the activation of hippocampal microglia in O₃-exposed mice

Based on our previous findings, the elevated levels of hippocampal C3 protein induced by O₃ exposure may not stem from local synthesis; rather, they are more likely derived from peripheral sources. To verify this hypothesis, we first conducted a systematic evaluation of the structural and functional changes in the BBB. The BBB restricts the entry of blood-borne substances into brain tissue through structures such as tight junctions (Fig. 7A). TEM results demonstrated that O₃ exposure disrupted the ultrastructure of tight junctions (Fig. 7B, C) and significantly increased the number of endothelial cell vesicles (Fig. 7C), indicating enhanced endocytosis. Additionally, RT-qPCR analysis revealed a significant decrease in the mRNA expression levels of Occludin and PECAM-1 in the hippocampal region of the O₃-exposed group (Fig. 7D), suggesting an impairment in BBB structure and function. However, after minocycline treatment, no significant improvement was observed in the O₃-induced ultrastructural damage to the BBB or in the downregulation of related gene mRNA expression (Supplementary Fig. S8A–D). Thus, O₃ exposure impairs BBB integrity in mice, and this impairment is not fully restored by minocycline intervention.

Fig. 7.

O₃ exposure induces BBB damage and increases serum C3 protein levels in mice. A The schematic diagram of the structural composition of the BBB in mice. B, C TEM images of the mouse hippocampal (scale bar, 2 μm/500 nm/500 nm) and the associated quantitative analysis (n = 5 per group). Yellow indicates astrocyte end-feet, green represents pericytes, blue denotes the vascular basement membrane, red signifies endothelial cells, and red arrows indicate tight junctions. D Relative mRNA levels of ZO-1, Occludin, Claudin-5 and PECAM-1 in the mouse hippocampus (n = 6 per group). E Schedule of peripheral experiments in O₃-exposed mice. F Differential scatter plot analysis of mouse serum proteomics. (n = 4 per group). G GSEA of mouse serum proteomics. H Combination plot of heat map and differential bubble for the complement pathway in mouse serum proteomics. I Reactome enrichment analysis plot of mouse serum proteomics. J C3 levels in mouse serum in each group (n = 8 per group). K Relative mRNA levels of C3 in the liver, lung, spleen, kidney, and heart of mice. (n = 6 per group). *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001

To investigate peripheral immune responses, in this study, we performed a multi-omics analysis on the mice exposed to O₃ (Fig. 7E). The serum proteomics results indicated that O₃ exposure resulted in 370 differentially expressed proteins (DEPs), comprising 221 upregulated and 149 downregulated proteins (Supplementary Fig. S7C, D). PCA demonstrated robust intra-group sample clustering and pronounced inter-group separation (Supplementary Fig. S7A), and the sample correlation heat map further substantiated the inter-group differences (Supplementary Fig. S7B). The differential protein scatter plot depicted the distribution of DEPs (Fig. 7F). Reactome enrichment analysis revealed significant enrichment of DEPs in pathways such as “Classical Antibody-Mediated Complement Activation” (Fig. 7I), and gene set enrichment analysis (GSEA) further confirmed the significant activation of the complement system pathway (Fig. 7G). The KEGG enrichment heat map visually illustrated the functional distribution of DEPs (Supplementary Fig. S7E). Notably, analyses of the cluster heat map and bubble plot related to complement signaling pathway proteins demonstrated a significant increase in serum complement C3 expression in the O₃-exposed group (Fig. 7H), and enzyme-linked immunosorbent assay (ELISA) results further validated the upregulation of serum C3 protein (Fig. 7J).

Complement protein C3 is a fundamental component of the complement system, primarily synthesized in the liver. Additionally, macrophages and peripheral organs, such as lung tissue and kidneys, serve as significant sources of C3. To trace the origin of C3, we investigated its expression in various peripheral organs and found that only the liver exhibited an increase in C3 mRNA levels (Fig. 7K). Additionally, the level of C3 protein in the liver (Fig. 8A, B) was significantly elevated. Liver function indicators revealed that O₃ exposure led to a decrease in the liver-to-body weight ratio (Fig. 8C) and an increase in serum ALT and AST activity (Fig. 8D, E). H&E staining demonstrated cytoplasmic staining, vacuolization, and cellular edema in hepatocytes (Fig. 8F), indicating that O₃ exposure induced liver injury. It is noteworthy that minocycline treatment also failed to significantly reverse the O₃ exposure-induced elevation of complement C3 levels in both the liver and serum (Supplementary Fig. S8E–G).

Fig. 8.

O₃ exposure induces liver injury by activating the hepatic complement pathway and increasing C3 expression. A, B Protein expression levels of C3 in the liver (n = 5 per group). C Liver weight-to-body weight ratios in mice (n = 12 per group). D, E ALT and AST levels in mouse serum in each group (n = 7 per group). F Representative images of H&E staining of mouse liver tissue (scale bar, 2.5 mm/100 µm/50 µm; n = 3 per group). G Volcano plot of differentially expressed genes identified from mouse liver transcriptome sequencing data (n = 4 per group). H GSEA revealed the complement cascade signaling pathway in the mouse liver transcriptome. I Reactome enrichment analysis of mouse liver transcriptome sequencing data. *P < 0.05, and ***P < 0.001

Liver transcriptomic analysis identified 494 DEGs, comprising 284 upregulated and 210 downregulated genes (Supplementary Fig. S7G). The volcano plot illustrated significant differential expression of genes involved in the complement pathway (Fig. 8G). Both KEGG and Reactome enrichment analyses revealed substantial enrichment of DEGs in complement-related pathways (Fig. 8I; Supplementary Fig. S7H, J), and GSEA further corroborated the activation of the complement cascade pathway (Fig. 8H). The radar plot depicting the expression changes of the complement pathway genes provided a detailed overview (Supplementary Fig. S7I). Collectively, these findings suggest that the complement cascade system plays a pivotal role in O₃-induced liver injury.

Joint analysis revealed 1,043 common elements between the serum proteome and liver transcriptome (Supplementary Fig. S7M). Association clustering indicated consistent trends in changes across both omics (Supplementary Fig. S7L), with most genes exhibiting concordant alterations in mRNA and protein levels (Supplementary Fig. S7O). Venn analysis demonstrated the compositional similarity of differential elements between the two omics (Supplementary Fig. S7N). KEGG enrichment analysis indicated that O₃ exposure simultaneously activated the complement signaling pathway in both the liver and serum (Supplementary Fig. S7K).

In summary, exposure to O₃ activated the complement signaling pathway in the liver of mice, resulting in an abnormal elevation of complement protein C3 levels in both the liver and serum, which subsequently induced liver injury. Concurrently, ozone exposure compromised the integrity of the BBB. These findings suggest that C3, derived peripherally, may enter the central nervous system through the impaired BBB, thereby playing a critical role in O₃-induced neurotoxicity. This discovery further supports the significant connection of the liver-brain axis in this pathological process.

Downregulation of hepatic C3 expression can improve behavioral abnormalities and synaptic plasticity in mice exposed to O₃

Based on previous findings, we hypothesized that hepatic complement C3 levels increase following O₃ exposure, subsequently entering the hippocampus through the compromised BBB. This process activates microglia and initiates synaptic phagocytosis, ultimately resulting in cognitive impairment. To validate this proposed mechanism, we performed tail vein injections of AAV-shRNA containing the liver-specific promoter thyroxine-binding globulin (TBG) to selectively knock down C3 expression (Fig. 9A). Two weeks after injection, both IF and western blot analyses of mouse hippocampus tissue confirmed the effective knockdown of hepatic C3 (Supplementary Fig. S9A–C), with no observed effect on mouse body weight (Supplementary Fig. S10A). Furthermore, behavioral experiments demonstrated that the knockdown of C3 significantly alleviated O₃ exposure–induced anxiety-like behaviors and cognitive dysfunction in the mice. This was evidenced by an increased proportion of central path exploration in the open field test (Supplementary Fig. S10B, C), restoration of the spontaneous alternation rate in the Y-maze, and an improved novel object recognition preference index (Supplementary Fig. S10D–H), as well as enhanced learning and memory capabilities in the water maze test (Fig. 9B–F).

Fig. 9.

Effects of hepatic C3 virus knockdown on synaptic plasticity and cognitive function in O₃-exposed mice. A Schedule of liver C3 knockdown mouse model establishment and O₃ exposure. B Statistical graph of escape latency in mice during the first 5 days of the learning stage in the Morris water maze (n = 9 per group). C Representative swim path trace images in the Morris water maze probe trial. D–G Statistical graphs of platform crossings D, proportion of path length in the target quadrant E, proportion of time in the target quadrant F, and swimming speed G during the testing phase in the Morris water maze (n = 9 per group). H–L TEM images of the mouse hippocampal (scale bar, 2 μm) and the associated quantitative analysis, with red pentagrams indicating synaptic locations (n = 5 per group). Yellow-boxed areas show higher magnification views of presynaptic (blue) and postsynaptic (red) terminals (scale bar, 500 nm). M Representative fEPSP traces before and after LTP induction (n = 5 per group). N Amplitude of fEPSP after HFS recorded on hippocampal slices. O Quantitative analysis of normalized fEPSPs during the last 10 min after HFS (n = 5 per group). *P < 0.05, **P < 0.01, and ***P < 0.001

Through Golgi staining, we observed that the knockdown of hepatic C3 in O₃-exposed mice significantly mitigated the reduction in neuronal complexity and dendritic spine density associated with O₃ exposure (Supplementary Fig. S10K–M). Furthermore, western blot analysis and TEM confirmed that the knockdown of C3 restored the impaired PSD95 protein expression levels resulting from O₃ exposure (Supplementary Fig. S10I, J) and reversed the decrease in hippocampal synaptic density as well as the reduction of PSD thickness (Fig. 9H–L). Electrophysiological results from brain slice preparations indicated that the knockdown of hepatic C3 significantly enhanced hippocampal LTP in O₃-exposed mice, thereby restoring synaptic plasticity (Fig. 9M–O). Additionally, we found that under non-exposure conditions, liver-specific C3 knockdown did not significantly affect behavioral performance, synaptic protein expression, dendritic spine morphology, or synaptic ultrastructure in mice (Supplementary Fig. S9D–R). Therefore, liver-specific knockdown of C3 can alleviate neural damage induced by O₃ exposure, further supporting the critical role of liver-derived C3 in O₃-induced cognitive impairment.

Liver-derived C3 drives microglia-mediated excitatory synaptic phagocytosis

To evaluate the impact of hepatic C3 knockdown on synapses following O₃ exposure, we conducted immunofluorescence staining for excitatory synaptic markers VGLUT2⁺/PSD95⁺ and inhibitory synaptic markers Gephyrin⁺/VGAT⁺ (Supplementary Fig. S11A; Fig. 10A). The results demonstrated that O₃ exposure resulted in a significant decrease in the number of VGLUT2⁺/PSD95⁺ co-labeled signals and PSD95⁺ signals, which could be restored to normal levels by C3 knockdown (Fig. 10B, D). However, there was no significant change in the number of VGLUT2⁺ signals (Fig. 10C). Regarding inhibitory synapses, there were no significant differences in the co-localization ratio or the number of spots of Gephyrin⁺/VGAT⁺ among the groups (Supplementary Fig. S11B–D). These results indicate that hepatic C3 knockdown can specifically reverse the excitatory synaptic damage induced by O₃ exposure without significantly affecting inhibitory synapses, suggesting that O₃ exposure primarily disrupts the integrity of excitatory synapses through a complement C3-dependent pathway.

Fig. 10.

Knockdown of hepatic C3 inhibits pro-inflammatory microglial activation and excitatory synapse loss. A Representative immunofluorescence images of VGLUT2 (purple) and PSD95 (green) co-localization in the hippocampus, along with orthogonal views and Imaris spot analysis images (scale bar, 5/0.5 μm). Yellow dashed circles indicate VGLUT2⁺/PSD95⁺ co-localized puncta; red arrows point to the corresponding positions of the selected co-localized puncta in the orthogonal views. B–D Quantification of the percentage of PSD95⁺ puncta, VGLUT2⁺ puncta and VGLUT2⁺/PSD95⁺ co-localized puncta in the hippocampus (n = 5 per group). E Representative immunofluorescence images of C3 (green) and PSD95 (red) co-localization in the hippocampus, with corresponding orthogonal views (scale bar, 5 μm). White arrows indicate the positions of the selected C3⁺/PSD95⁺ co-localized puncta in the orthogonal views. F–H Quantification of the percentage of C3 deposits, PSD95⁺ puncta, and C3⁺/PSD95⁺ co-localized puncta in the hippocampus (n = 5 per group). I Representative immunofluorescence images of Iba1 (red) and PSD95 (green) co-localization in the hippocampus (scale bar, 20 μm), orthogonal views (scale bar, 5 μm), and 3D reconstruction analysis images (scale bar, 5 μm). White arrows indicate PSD95 puncta on microglia. J, K Quantitative analysis of PSD95 intensity and the number of PSD95⁺ puncta within Iba1 cells in the hippocampus (n = 5 per group). L Representative images of H&E staining of mouse liver tissue (scale bar, 2.5 mm/100 µm/50 µm; n = 4 per group). M C3 levels in mouse serum in each group (n = 6 per group). N, O ALT and AST levels in mouse serum in each group (n = 9 per group). *P < 0.05, **P < 0.01, and ***P < 0.001

To further investigate the impact of hepatic C3 knockdown on microglial synaptic function, we conducted multiplex immunofluorescence colocalization analysis on hippocampal tissues, examining the colocalization of C3/PSD95, Iba1/CD68, Iba1/PSD95, and Iba1/C3 (Supplementary Fig. S11E, I; Fig. 10E, I). The results indicated that O₃-exposed mice exhibited increased C3 deposition in the hippocampal region, a reduced number of PSD95-positive puncta, and a significantly elevated colocalization ratio between C3 and PSD95 (Fig. 10F–H). Concurrently, the colocalization signals of Iba1⁺/CD68⁺, Iba1⁺/PSD95⁺, and Iba1⁺/C3⁺ were markedly enhanced. These changes could be effectively reversed by C3 knockdown (Supplementary Fig. S11E–K; Fig. 10I–K). These findings suggest that C3 is involved in regulating the phagocytic function of microglia, and that liver-derived complement C3 plays a crucial mediating role in the abnormal synaptic phagocytosis of hippocampal microglia induced by O₃ exposure. Furthermore, peripherally derived C3 may contribute to O₃-induced neuroinflammation and synaptic loss by impairing the integrity of excitatory synapses.

Next, we evaluated the effect of liver-specific knockdown of the virus on the expression of complement C3 in the hippocampus, serum, and liver of mice. The results from Western blot, ELISA, and qRT-PCR demonstrated that C3 knockdown effectively alleviated the aberrant expression of C3 induced by O₃ in the hippocampus, serum, and liver tissues (Fig. 10M; Supplementary Fig. S11M–P). Concurrently, it reduced the gene expression of pro-inflammatory factors in the hippocampus triggered by O₃ exposure (Supplementary Fig. S11L). Furthermore, C3 knockdown significantly ameliorated the O₃-induced elevation of serum ALT and AST, as well as the vacuolation and cellular edema in liver tissue (Fig. 10N, O). These findings indicate that the knockdown of liver complement C3 alleviates O₃-induced liver injury.

Discussion

The systemic toxicity of O₃ extends beyond the lungs [52]; however, the mechanisms underlying its extrapulmonary damage, particularly to the brain, remain unclear. Although a functional liver-brain axis has been implicated in various metabolic and neurological disorders [53, 54], its role in neurotoxicity induced by environmental pollutants is largely unexplored. We demonstrate that subchronic exposure to environmentally relevant levels of O₃ initiates a novel pathogenic cascade along the liver-brain axis. Specifically, we found the homeostasis of hepatic complement signaling is a primary target of O₃, and its disruption serves as a critical upstream event that propagates pro-inflammatory signals to the brain. This, in turn, activates hippocampal pro-inflammatory microglia, leading to the specific loss of excitatory synapses and, consequently, cognitive impairment. This newly identified pathway offers a broader perspective on how environmental pollutants exert systemic effects, shifting the focus from direct oxidative insult in the brain to a systemic disorder of inter-organ communication.

While our previous research confirmed the role of O₃ in exacerbating pulmonary inflammation [55], recent epidemiological studies have linked increasing O₃ levels to heightened anxiety and cognitive decline [56–58]. Our experimental model directly validates this association, demonstrating that O₃ exposure is sufficient to induce significant cognitive deficits. These deficits are mechanistically rooted in hippocampal damage, a brain region essential for learning and memory [59, 60]. Histological analysis revealed a specific disorganization of the hippocampal architecture, indicative of tissue injury. Given that synaptic plasticity represents the fundamental cellular basis of cognitive function and is frequently disrupted in neurodegenerative and neurotoxic processes [61, 62], we further explored whether O₃ exposure impacts synaptic integrity, a key determinant of synaptic plasticity. Our findings indicate that O₃ exposure compromises synaptic function by downregulating the expression of pivotal synaptic plasticity-related proteins and altering dendritic spine morphology in the hippocampus, which represents morphological and molecular changes tightly linked to cognitive decline. Collectively, these findings reveal significant impairments in synaptic plasticity, corroborating previous reports and identifying deficits in this process as the critical mechanism underlying O₃-induced cognitive decline.

To elucidate the cellular mechanisms underlying synaptic loss, we focused on microglia, the resident immune sentinels of the CNS [63], known to mediate the neurotoxic effects of various environmental stressors, including air pollution [64, 65]. However, research on the responses of whole-brain microglia to O₃ remains relatively limited. In this study, we employed fMOST technology to achieve whole-brain fluorescence imaging of microglia. The results of quantitative statistical analysis of cognition-related brain regions demonstrated a significant increase in microglial activation within the hippocampus. Additionally, findings from pharmacological intervention using minocycline further underscored the critical role of pro-inflammatory microglial phagocytosis in O₃-induced neurotoxicity. This shifts the perspective from viewing pro-inflammatory microglia as passive responders to recognizing them as active mediators in O₃ neurotoxicity, thereby raising the critical question of the specific signaling pathway that drives their deleterious phagocytic activity. Among the myriad signals that can influence pro-inflammatory microglial phagocytosis, the complement cascade, particularly the C1q/C3 axis, emerges as a prime candidate due to its well-defined role in synaptic pruning [37, 66]. Under physiological conditions, complement proteins, including C3, contribute to synaptic tagging in the brain, facilitating the microglia-mediated pruning of labeled synapses [67]. In epilepsy, C3-assisted microglia selectively engulf inhibitory synapses, thereby disrupting the excitatory-inhibitory balance and promoting epileptogenesis [37]. Conversely, in a murine fracture surgery model, microglial phagocytosis of C3-tagged excitatory synapses results in synaptic dysfunction within excitatory neurons [68]. Our study demonstrates that O₃ exposure induces a C3-dependent yet C1q-independent activation signature in the hippocampus. Notably, we establish that this C3 activation directs pro-inflammatory microglia to phagocytose excitatory neurons. In O₃-exposed mice, either microglial inhibition or interference with hepatocyte-derived complement C3 mitigates synaptic injury. These findings suggest that complement C3 acts not only as a synaptic tag but also as an enhancer of microglial phagocytic activity in this context, although the precise mechanisms remain to be fully elucidated.

In our study, the lack of astrocyte activation (the primary source of C3 in the CNS) and the ineffectiveness of minocycline-mediated pro-inflammatory microglial inhibition in restoring hippocampal C3 levels strongly suggest that central production of C3 is not the main driver of hippocampal complement activation. These critical observations prompted us to hypothesize that peripheral, rather than central, production of C3 mediates O₃-induced hippocampal complement activation, thereby shifting our investigative focus to the peripheral-brain axis [69]. To validate this hypothesis, we first assessed the integrity of the BBB, as blood-derived C3 cannot directly penetrate the central nervous system when the BBB is intact, unless its integrity is compromised [70]. Our findings confirmed that O₃ exposure impairs BBB function through the downregulation of tight junction proteins, such as ZO-1, and induces ultrastructural damage to cerebral endothelial cells. This anatomical breach facilitates the influx of systemic factors into the brain parenchyma. The circulating complement protein C3, a key effector of immune and inflammatory responses, is primarily synthesized in the liver [71]. We subsequently observed elevated serum C3 levels in O₃-exposed mice, and analyses of hepatic tissue revealed O₃-induced activation of hepatic complement signaling, along with increased hepatic C3 expression and signs of liver injury. The most compelling evidence for a functional liver-brain axis driving O₃ neurotoxicity emerged from liver-specific C3 gene knockdown experiments, which demonstrated that this intervention abrogated O₃-induced C3 upregulation in both serum and hippocampus, suppressed microglial activation and pathological synaptic pruning, and restored cognitive function in the exposed mice. Furthermore, building on previous studies that have shown peripheral complement components, including C3 and C1q, can cross a compromised BBB into the central nervous system via systemic circulation [42, 44, 69], we propose that the liver-brain axis, mediated by hepatically derived complement C3, represents a novel mechanistic pathway underlying cognitive impairment induced by O₃ exposure.

Prior studies have demonstrated that O₃ exposure activates both microglia and astrocytes in the central nervous system, with astrocyte activation specifically documented in animal models of Alzheimer’s disease and in aged mice following O₃ exposure [72–74]. In contrast, we observed robust aberrant pro-inflammatory microglial activation in healthy 2-month-old mice after only 30 days of O₃ exposure, without concomitant astrocyte activation. This finding further supports the notion that microglia may serve as more sensitive innate immune sentinels in response to O₃ exposure than their astrocyte counterparts within the central nervous system. Given that the severity of neurotoxicity increases with O₃ concentration and exposure duration [7], elucidating how O₃ modulates glial crosstalk to govern neuronal function and survival represents a promising avenue for clarifying the mechanisms underlying O₃-induced neurotoxicity in future studies. Furthermore, while O₃ exposure resulted in reduced body weight, our measurement of weekly food intake revealed no significant decrease. Literature indicates that O₃-induced weight change is a transient, metabolically driven adaptation distinct from persistent neural damage [10]. This suggests that the observed weight modulation represents a parallel systemic metabolic response, likely involving hepatic lipid dysregulation as a homeostatic adjustment, rather than a nutritional confound that drives the specific neurotoxic cascade of hepatic complement activation, BBB disruption, and microglial synaptic pruning.

Despite the novel insights gained from this study, several limitations exist that underscore important avenues for future research. First, although minocycline-mediated inhibition strongly implicates microglial involvement, its well-established pleiotropic effects prevent the definitive exclusion of its direct neuroprotective or vascular contributions. Therefore, to conclusively delineate the functional mechanism of complement C3 in O₃-induced microglial activation, it is essential to employ microglia-specific genetic models (e.g., conditional receptor knockouts or depletion strategies) alongside quantitative analyses of key complement cleavage products (e.g., C3a, iC3b) and their downstream receptor signaling. Second, while hepatic C3 knockdown attenuates hippocampal C3 levels and associated neuropathological changes, and we have demonstrated O₃-induced BBB impairment, direct evidence that O₃-induced circulating C3 protein traverses the compromised BBB into the brain parenchyma remains lacking. Therefore, tracking the dynamic changes in hepatic C3 expression and hippocampal microglial activation at key time points (including 1, 7, 15, and 30 days post-O₃ exposure) is essential to delineate the critical time window and mechanistic cascade linking hepatic C3 upregulation to subsequent neurotoxicity.

Conclusion

In summary, this study integrates animal exposure, multi-omics sequencing, and toxicological mechanisms to demonstrate that subchronic O₃ exposure induces liver injury through the activation of the hepatic complement system, particularly complement C3. As a potent oxidizing pollutant, O₃ compromises organ barriers, facilitating the translocation of liver-derived complement C3 from the bloodstream to the CNS. This translocation subsequently triggers pro-inflammatory microglial activation and specific phagocytosis of excitatory synapses in the hippocampal region, ultimately leading to cognitive impairment in mice (Fig. 11). From the perspective of the liver–brain axis, these findings provide novel insights into the neurotoxic mechanisms of O₃, emphasizing the need for further investigation into this pathway to clarify its role in neurological impairment, identify potential strategies for mitigating O₃-related harm, and inform public health measures aimed at reducing O₃ exposure.

Fig. 11.

Graphical abstract of this study

Authors’ contributions

Y.W., H.Q., and W.D. contributed equally to this work. Study Conceptualization: Y.W., H.Q., D.N., Y.Z., and J.L.; Methodology: Y.W., W.D., Y.C., and J.L.; Investigation: Y.W., H.Q., W.D., and Y.C.; Formal analysis: Y.W., H.Q., W.D., Y.C., and J.L.; Writing-Original Draft: Y.W., W.D., and Y.Z.; Writing-Reviewing and Editing: Y.W., H.Q., D.N., Y.Z., and J.L.; Funding acquisition: W.D., and J.L.; Project administration: Y.Z., and J.L.; Resources: Y.Z.; Software: Y.Z.; Validation: Y.Z.; Visualization: Y.Z.; Supervision: Y.Z., and J.L. All the authors have read and approved the paper.

Funding

This work was supported by grants Brain Science and Brain-like Intelligence Technology - National Science and Technology Major Project (2022ZD0211600), “The 14th Five-year Plan” Hubei provincial advantaged characteristic disciplines (groups) project of Wuhan University of Science and Technology (2023C0106) and College Students Innovative Entrepreneurial Training Plan Program (202510488001).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in strict adherence to the guidelines for the care and use of laboratory animals and approved by the the Ethics Committee of Wuhan University of Science and Technology. (Approval No. WUST-2022076).

Consent for publication

All the authors agree with the submission of this manuscript.

Competing interests

The authors declare no competing interests.