Long-term exposure to polystyrene microplastics exacerbates seizure symptoms via lipid metabolic disruption and ferroptosis: insights from multi-omics analyses

Abstract

As a consequence of global industrial growth, microplastics (plastic fragments < 5 mm) have become ubiquitous environmental contaminants, prompting serious questions about their impact on human health. Beyond the established risks of ingestion, the inhalation of these airborne particles is now a primary focus, especially regarding potential effects on the neurological system. Emerging evidence from laboratory and animal models shows that inhaled microplastics can penetrate the brain. Once there, they can trigger a cascade of harmful effects, including neuroinflammation, oxidative stress, and deficits in learning and memory. Among the general population, children are uniquely vulnerable to microplastic exposure due to their developing physiological systems, which are particularly susceptible to the chemical and physical hazards posed by these particles. Nevertheless, current scientific understanding of the health consequences of microplastic exposure remains limited, with a substantial knowledge gap concerning the long-term effects of respiratory microplastic exposure on pediatric populations, particularly those with pre-existing neurological conditions such as epilepsy. This study investigates the relationship between chronic respiratory exposure to polystyrene microplastics and seizure severity, with the aim of establishing a potential exposure-metabolite-gene regulatory network. We hypothesize that prolonged respiratory mciroplatsic exposure induces systemic oxidative stress and inflammation, which subsequently disrupts lipid metabolism, alters gene expression profiles, triggers ferroptosis, and ultimately exacerbates seizure manifestations.This study highlights three key findings. First, chronic respiratory microplastic exposure induces systemic oxidative stress and inflammation, posing substantial health risks. Second, integrated metabolomics and Mendelian randomization analyses reveal that this exposure disrupts lipid metabolism, with metabolic perturbations strongly associated with ferroptosis activation and increased seizure severity. Third, multi-omics approaches coupled with in vivo validation confirm that microplastics disrupt lipid homeostasis, dysregulate ferroptosis-related gene expression, and exacerbate seizure manifestations. Notably, our data identify melatonin as a promising therapeutic candidate for mitigating these adverse effects. Collectively, these findings substantially advance the understanding of microplastic-induced neurotoxicity and reveal actionable molecular targets for potential therapeutic interventions.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04599-5.

Introduction

The global proliferation of plastics has resulted in ubiquitous human exposure to microplastics, defined as fragments smaller than 5 mm present in air, water, and food, through daily inhalation and ingestion. While ingestion and inhalation represent the primary exposure routes, respiratory microplastic exposure has emerged as a critical concern for neurological health. Experimental evidence demonstrates that microplastics can traverse the blood-brain barrier, inducing oxidative stress, neuroinflammation, and neuronal damage [1].However, the long-term neurological consequences of chronic respiratory exposure, particularly for vulnerable populations such as children, remain poorly characterized. Pediatric populations exhibit heightened susceptibility due to ongoing neurodevelopment and higher respiratory rates relative to body weight. Recent studies have detected significantly elevated microplastic levels in infants compared to adults, underscoring the urgency of investigating developmental neurotoxicity [2]. Therefore, this study investigates the relationship between long-term respiratory exposure to polystyrene microplastics and seizure severity exacerbation. We aim to establish a comprehensive exposure-metabolite-gene regulatory network to elucidate the underlying molecular mechanisms governing this pathological process.

Seizures, the leading neurological disorder in children, are abrupt electrical disturbances in the brain that affect behavior, movement, and awareness [3]. Children’s developing nervous systems make them particularly vulnerable to environmental pollutants, which have been linked to neurological disorders [4]. For instance, higher microplastic concentrations have been found in the brains of dementia patients, suggesting a link between exposure and neurological risk [5]. When the body is exposed to pollutants, it can develop conditions like oxidative stress and neuroinflammation. These harmful states may, in turn, worsen an individual’s seizures [6, 7]. Because seizure thresholds are sensitive to subtle changes in the brain’s environment, seizures may be an early sign of toxic microplastic accumulation [8].Even so, we still don’t know much about the prolonged consequences of inhaling microplastics on seizure symptoms, or the biological processes that cause those effects.

Researchers are increasingly linking the development of seizures to a process called ferroptosis. This is a regulated, iron-dependent form of cell death defined by the destructive breakdown of fats in cell membranes [9]. Our previous experimental studies have demonstrated that seizures induce oxidative stress, iron accumulation, and glutathione depletion in the hippocampus, all of which promote hippocampal neuronal ferroptosis [6, 7, 10]. Furthermore, a growing body of research points to ferroptosis as the critical bridge connecting exposure to environmental pollutants with the onset and worsening of neurological disorders [11]. Airborne microplastics can access the central nervous system via two primary routes: (1) direct neuronal translocation through the olfactory nerve, delivering particles to the olfactory bulb and subsequently to connected limbic regions, and (2) systemic circulation following inhalation and translocation across the alveolar–capillary barrier in the deep lung. Due to their small size and lipophilic nature, these particles enter the bloodstream, where they induce cellular damage by generating reactive oxygen species (ROS), disrupting mitochondrial energy metabolism, and promoting membrane lipid peroxidation [12]. This pro-oxidative state triggers robust inflammatory responses, establishing a self-amplifying cycle of oxidative stress, neuroinflammation, and progressive neuronal injurys [9]. Despite these advances, a significant knowledge gap persists regarding how microplastic-induced inflammation and oxidative stress mechanistically intersect with the ferroptosis pathway to influence seizure development and severity. While it is established that both seizures and microplastic exposure independently create conditions conducive to ferroptosis, including iron dysregulation, glutathione depletion, and peroxidation of polyunsaturated fatty acid, the precise molecular crosstalk between inhaled plastic particulates and the core ferroptosis regulatory machinery in the context of epilepsy remains undefined. Elucidating this mechanistic interface is essential to fully understand the neurotoxicological risks posed by chronic microplastic exposure and to identify novel therapeutic targets for environmentally exacerbated seizure disorders.

In this study, we establish a comprehensive exposure-metabolite-gene regulatory network to elucidate the underlying mechanisms. Our research offers three key contributions. First, by integrating data from our observational pilot cohort with human blood biochemical assays, we demonstrate that chronic respiratory exposure to polystyrene microplastics induces systemic oxidative stress and inflammation, posing substantial risks to human health. Second, through the integration of human plasma untargeted metabolomics, targeted lipidomics, and Mendelian randomization analyses, our results reveal that exposure-induced systemic stress profoundly disrupts lipid metabolism. Notably, these metabolic perturbations are strongly associated with the ferroptosis pathway, potentially contributing to increased seizure severity. Third, employing a comprehensive multi-modal approach encompassing observational cohort studies, single-cell transcriptomics, RNA-seq analysis, and in vivo animal experiments, we validated our central hypothesis. Our findings confirm that chronic respiratory microplastic exposure induces systemic oxidative stress and inflammation, which subsequently disrupts lipid homeostasis, dysregulates ferroptosis-related gene expression, and ultimately exacerbates seizure manifestations. Moreover, our data identify melatonin as a promising therapeutic candidate for mitigating these adverse effects. In summary, this study substantially advances the understanding of microplastic-induced neurotoxicity and elucidates actionable molecular pathways that may inform future therapeutic interventions.

Methods and materials

Study population and design of three pilot cohorts

To investigate the potential adverse effects of prolonged respiratory microplastic (MP) exposure, we established three pilot observational cohorts with specific focus on: (1) cognitive function impairments, (2) inflammatory response dysregulation, (3) oxidative stress induction, and (4) their association with seizure symptoms ( Information S1). Our study design and methodologies were adapted from previously published approaches [6, 7, 13, 14].

Cohort 1 (occupational respiratory exposure) :This cohort was designed to gather preliminary evidence. It included 15 plastics manufacturing workers with over a decade of high-level respiratory occupational exposure, a group chosen for their anticipated high cumulative MP burden from inhalation [15, 16]. For the control group, we recruited 15 age- and sex-matched colleagues from the same manufacturing facility who worked in administrative or non-production roles without direct MP exposure. This matching strategy minimized potential confounding variables related to socioeconomic status, lifestyle factors, geographic location, and environmental co-exposures. We assessed anxiety-like behaviors and cognitive function using the Generalized Anxiety Disorder-7 (GAD-7) and Montreal Cognitive Assessment (MoCA) questionnaires, respectively (Information S2 and S3). Key demographic and clinical data are summarized in Table 1.

Table 1.

Baseline characteristics of Cohort 1 participants by occupational respiratory microplastic exposure status

Group	Exposed group	Control group	P-value
Participant profile (Mean ± SD)
Participant age (years)
Total	36.07 ± 2.96	36.07 ± 2.96	> 0.99
Male	36.55 ± 3.23	36.55 ± 3.23	> 0.99
Female	34.75 ± 1.71	34.75 ± 1.71	> 0.99
Gender (count, percentage)
Total	15 (100)	15 (100)	> 0.99
Male	11 (73)	11 (73)	> 0.99
Female	4 (27)	4 (27)	> 0.99
Duration of employment (years)
Total	12.53 ± 1.76	12.67 ± 1.29	0.82
Male	11.91 ± 1.51	12.27 ± 1.11	0.52
Female	14.25 ± 0.62	13.75 ± 0.63	0.59

Cohort 2 (biomarker association)

This cohort was designed to investigate the association between prolonged respiratory microplastic exposure and non-specific systemic biomarkers predictive of seizure severity, thereby providing clinical evidence that chronic microplastic exposure may exacerbate seizure symptoms. The seizure group included all volunteer pediatric patients presenting with seizures at the Affiliated Hospital of Jiangnan University between November 2022 and January 2024. Healthy children undergoing routine check-ups during the same period formed the control group. We analyzed standard clinical biomarkers, including complete blood counts, liver and kidney function, glucose, and serum electrolytes, collected within 24 h of admission. The clinical characteristics of the 284 participants are presented in Supplementary Table 4.

Cohort 3 (mechanistic validation)

This cohort was designed to validate our central hypothesis. We propose that prolonged respiratory microplastic exposure exacerbates seizure severity by inducing systemic oxidative stress and inflammation. This pathological state disrupts lipid metabolism and dysregulates the expression of ferroptosis-related genes, ultimately triggering neuronal ferroptosis. The study cohort included 15 children with a new diagnosis of epilepsy, along with a control group of healthy children who were matched by age and sex. To evaluate the relationship between MP levels and seizure severity, patients were stratified by seizure duration into mild (< 2 min), moderate (2–5 min), and severe (status epilepticus, ≥ 5 min) subgroups, following international guidelines [17, 18]. Baseline characteristics for this cohort are detailed in Table 2.

Table 2.

Baseline characteristics of participants in Cohort 3

Group	Seizure patients	Healthy controls	P-value
Participant profile (Mean ± SD)
Participant age (Years)
Total	9.13 ± 1.99	9.13 ± 1.99	> 0.99
Male	9.66 ± 2.01	9.66 ± 2.01	> 0.99
Female	8.33 ± 1.86	8.33 ± 1.86	> 0.99
Gender (count, percentage)
Total	15 (100)	15 (100)	> 0.99
Male	9 (60)	9 (60)	> 0.99
Female	6 (40)	6 (40)	> 0.99
Frequency of seizure subtypes (count, percentage)#
Mild	/	4 (27)	/
Moderate	/	6 (40)	/
Severity	/	5 (33)	/

# Mild (< 2 min), Moderate (2–5 min), and Severe (status epilepticus,≥minutes)

Ethical approval and participant consent

The Institutional Review Board at the Affiliated Hospital of Jiangnan University reviewed and approved this study’s protocol (ID: LS2024JN110806R1), and all procedures involving human participants were conducted in accordance with the ethical standards of the Declaration of Helsinki. We obtained written informed consent from every participant before their inclusion in the study, ensuring they understood the study’s purpose, methods, potential benefits, and risks. We emphasized the voluntary nature of participation and each individual’s freedom to withdraw from the research at any point without consequence. All personal data were anonymized to protect participant privacy and stored with robust security measures.

Animal model and experimental design

All animal experiments were approved by the Animal Ethics Committee at Jiangnan University (Approval ID: JN.20231030c0400131[496]). This study aimed to validate how airborne microplastic-induced neurotoxicity, particularly the interplay of inflammation and ferroptosis, affects seizure severity.

Model and reagents

We used 4-week-old C57BL/6J mice to model a pediatric population, consistent with our previous work [13]. (1) Microplastic Exposure: To mimic environmental respiratory exposure and induce systemic oxidative stress and inflammation, we administered 5μm polystyrene microplastics (PS-MPs) via intratracheal instillation at a dose of 1 mg/kg. This particle size was selected to represent inhalable airborne particulate matter, while the dose was designed to simulate chronic cumulative human intake and serve as a high-burden stressor capable of triggering robust systemic oxidative stress and inflammatory responses [19]. (2) Seizure Induction: Epileptic conditions were induced with an intraperitoneal (i.p.) injection of kainic acid (KA) at 20 mg/kg [13]. 3.Therapeutic Agents: We tested two treatments: the ferroptosis inhibitor ferrostatin-1 (Fer-1), administered daily at 3 mg/kg (i.p.) and the neuroprotectant melatonin, administered daily at 20 mg/kg via intragastric gavage [20–22].

Experimental groups

Mice (n = 6 per group) underwent a 90-day pretreatment period before seizure induction and were divided into five groups: (1) Control: Received daily intratracheal saline. (2) KA Group: Received daily intratracheal saline, followed by a single KA injection. (3) KA + MP Group: Received daily intratracheal PS-MPs, followed by a single KA injection. (4) KA + MP + Fer-1 Group: Received daily PS-MPs and Fer-1, followed by a single KA injection. (5) KA + MP + Mel Group: Received daily PS-MPs and melatonin, followed by a single KA injection.

Seizure severity assessment

Seizure activity was scored using the 5-stage Racine scale, where a score of 3 or higher indicates a significant motor seizure [10]: Stage 1: Mouth and facial movements (chewing, whisker twitching). Stage 2: Head nodding. Stage 3: Forelimb clonus. Stage 4: Rearing with forelimb clonus. Stage 5: Rearing, falling, and generalized tonic-clonic convulsions.

Data analysis and statistics

All data are presented as mean ± standard deviation (SD), with statistical significance set at P < 0.05. We assessed data normality using the Shapiro-Wilk test. Depending on the distribution and number of groups, we used unpaired t-tests, ANOVA (with Tukey’s test), the Mann-Whitney U test, or the Kruskal-Wallis test to analyze the data in GraphPad Prism (v9.1). To ensure reproducibility, our Python and R analysis scripts are publicly available on GitHub (https://github.com/PediatricLab-Jiangnan/Airborne-singlecell-bulk). Raw data from human plasma untargeted metabolomics and targeted lipidomics analyses have been deposited in the Genome Sequence Archive under accession numbers OMIX009167 and OMIX014782, respectively. Additional methodological details are described in the Supplementary Methods (Methods S1-S30).

.

Results

Microplastic exposure induces oxidative stress and inflammation, leading to lipid metabolism disturbance

To systematically identify and elucidate the potential mechanisms underlying microplastic-induced neurotoxicity, we employed sophisticated bibliometric tools, including VOSviewer and CiteSpace (Method S1) [23]. Keyword co-occurrence network analysis consistently revealed a critical and interconnected triad involving oxidative stress, neuroinflammation, and lipid metabolism dysregulation (Fig. 1A and Figure S1). Furthermore, burst analysis performed in CiteSpace revealed that inflammation and oxidative stress pathways have emerged as predominant research hotspots in the context of respiratory microplastic exposure since 2022 (Fig. 1B). This temporal shift reflects the growing recognition of atmospheric microplastic pollution as a significant route of human exposure and the urgent need to understand its neurological implications. The convergence of these bibliometric findings strongly supports our central hypothesis.

Fig. 1.

Analysis of microplastics in the blood of an occupationally exposed cohort.(A) VOSviewer analysis of keyword co-occurrence in microplastics research (2019–2025). (B) Top 10 keywords with the strongest citation bursts identified by Citespace. (C) Flowchart of the observational cohort study design. (D) Experimental workflow for isolating and analyzing microplastics from blood samples. (E) Raman spectra and corresponding images of identified PP, PS, and PE microplastics. Scale bars are 200 μm–20 μm. (F) Blood microplastic concentrations in the exposed group versus the non-exposed group. *P < 0.05, **P < 0.01

To investigate the metabolic consequences of chronic microplastic inhalation, we initiated a pilot observational study (Cohort 1) involving plastics manufacturing workers. This cohort represented a unique population with well-characterized, long-term respiratory exposure exceeding one decade (Fig. 1C) [3] Employing high-resolution Raman spectroscopy for precise polymer identification and quantification (Method S2–S3 and Fig. 1D), we confirmed that the occupationally exposed group exhibited significantly elevated blood concentrations of three major microplastic polymers: polyethylene (PE), polypropylene (PP), and polystyrene (PS) (Fig. 1E–F, Figure S2, and Table S1).We then assessed cognitive function and inflammatory marker levels following this prolonged exposure. The microplastic-exposed group showed significantly impaired cognitive performance (MoCA score) and higher anxiety levels (GAD-7 score) ( P < 0.05, Fig. 2A–B) [24, 25]. Concurrently, this group exhibited significantly elevated blood levels of pro-inflammatory cytokines, including IL-6, IL-1β, and TNF-α ( P < 0.05 ; Method S4–S5 and Fig. 2C–E). Notably, Spearman correlation analyses indicated a positive correlation between these cytokines and blood microplastic levels, with IL-6 demonstrating the strongest association (Fig. 2F–H). We further quantified oxidative stress levels between the two groups. Results indicated that long-term respiratory microplastic exposure significantly increased MDA levels while decreasing GSH levels and SOD activity (Method S6 and Fig. 2I–K). Similarly, Spearman correlation analyses revealed a positive correlation between microplastic concentrations and MDA levels, and negative correlations with GSH levels and SOD activity. Collectively, these findings provide direct evidence that long-term respiratory microplastic exposure induces systemic inflammation and oxidative stress, posing significant hazards to human health [26].

Fig. 2.

Human peripheral blood biochemistry analysis reveal that prolonged respiratory microplastic exposure induces oxidative stress and inflammation. (A-B) Comparison of questionnaire scores between the Control (Con) and Respiratory microplastics-exposed groups for (A) GAD-7 (anxiety) and (B) MoCA (cognitive function). (C-E) Relative levels of inflammatory cytokines in human blood, including (C) IL-6, (D) IL-1β, and (E) TNF-α. (F-H) Spearman correlation analyses showing the relationship between blood microplastic levels and inflammatory markers: (F) IL-6, (G) IL-1β, and (H) TNF-α. (I-K) Assessment of oxidative stress markers in human blood: (I) MDA levels, (J) GSH levels, and (K) SOD activity. (L-N) Spearman correlation analyses between microplastic levels and oxidative stress indicators: (L) MDA, (M) GSH, and (N) SOD. * P < 0.05, ** P < 0.01

To establish the exposure-metabolite-target gene network underlying respiratory microplastic exposure, we subsequently conducted untargeted plasma metabolomic profiling using liquid chromatography-mass spectrometry. This analysis revealed that chronic microplastic exposure induces profound alterations in systemic metabolite signatures (Method S7 and Fig. 3A). Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) demonstrated clear separation between the exposed and control groups, indicating substantial metabolic reprogramming (Fig. 3B–C). Differential metabolite analysis identified 126 significantly altered metabolites (fold change > 1.5, P < 0.05; Figure S3), encompassing diverse chemical classes including amino acids, lipids, organic acids, and nucleotides.

Fig. 3.

Integrated human plasma untargeted metabolomics and machine learning analysis reveals that prolonged respiratory microplastic exposure disturbs lipid metabolism (A) Workflow for metabolomic analysis and machine learning-based identification of hub metabolites. (B) Principal Component Analysis (PCA) showing distinct metabolic clustering between control (Con) and microplastic-exposed (Micro) groups. (C) Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) plot confirming clear separation between the groups. (D-E) Top 20 enriched metabolic pathways identified by (D) KEGG and (E) Reactome analysis. (F) Gene Set Enrichment Analysis (GSEA) plot highlighting significant alterations in the lipoic acid metabolism pathway. (G-H) SVM-RFE feature selection based on (G) classification error and (H) accuracy, pinpointing 14 optimal metabolites. (I) Architecture of the Multilayer Perceptron (MLP) neural network. (J) Venn diagram showing hub metabolites identified by both SVM-RFE and MLP models. (K-N) Relative intensities of 5 key hub metabolites in the Con and Micro groups. *P < 0.05, **P < 0.01

Pathway enrichment analyses were performed to elucidate the biological impact of these metabolic changes. KEGG pathway analysis revealed that respiratory microplastic exposure significantly perturbs lipid metabolism, particularly glycerophospholipid metabolism, lipoic acid metabolism, and linoleic acid metabolism. Furthermore, Reactome pathway analysis highlighted the dysregulation of glycerophospholipid catabolism, glycerophospholipid biosynthesis, and SLC-mediated transmembrane transport (Fig. 3D–E). Notably, Gene Set Enrichment Analysis (GSEA) revealed that prolonged exposure suppresses lipoic acid metabolism, a pathway critically involved in ferroptosis inhibition (Fig. 3F) [27]. These convergent findings suggest a potential mechanistic link between respiratory microplastic exposure and lipid peroxidation-mediated ferroptosis.

To systematically identify the core molecular drivers underlying these metabolic perturbations, we employed a robust machine learning approach integrating Support Vector Machine-Recursive Feature Elimination (SVM-RFE) and Multi-Layer Perceptron Artificial Neural Network (MLP-ANN) algorithms to enhance feature selection reliability through cross-validation across distinct computational frameworks (Method S8 and Fig. 3G–I) [26]. Through iterative feature ranking, we identified four hub differential metabolites exhibiting the highest discriminatory capacity between groups (Fig. 3J). These key metabolites comprised four lipid-related molecules (Fig. 3K–N): 15(R)-Hydroperoxy-EPE: a bioactive lipid mediator involved in inflammatory resolution. Three distinct phosphatidylcholines (PC(O-18:0/18:2), PC(20:4/0:0), and PC(16:0/18:2), which are essential membrane phospholipids highly susceptible to oxidative damage. Collectively, these findings demonstrate that respiratory microplastic exposure significantly disrupts lipid metabolism, potentially promoting ferroptosis through the dysregulation of specific lipid mediators and antioxidant pathways.

Microplastic exposure induces lipid metabolic disturbance, triggering ferroptosis and exacerbating seizure severity

To further substantiate the hypothesis that respiratory microplastic exposure induces lipid metabolic dysregulation, thereby exacerbating seizure severity and correlating with altered ferroptosis-related gene expression, we first performed Mendelian Randomization analyses (Method S9 and Table S2). These analyses established a causal link between the dysregulation of phosphatidylcholine PC(16:0/18:2) and increased seizure severity (Fig. 4A–B). We subsequently conducted human lipidomic profiling to provide direct evidence of microplastic-induced ferroptosis ( Method S10 and Table S3). As shown in Fig. 4C–F, prolonged respiratory microplastic exposure significantly altered plasma lipid profiles. Specifically, plasma levels of eicosapentaenoic acid, arachidonic acid, and linoleic acid were significantly increased (P < 0.05). Furthermore, plasma concentrations of 8-iso-PGF2α, a well-established biomarker of ferroptosis, were significantly elevated (P < 0.05).

Fig. 4.

Integrated Mendelian randomization and targeted lipidomics reveal that prolonged respiratory microplastic exposure disrupts lipid metabolism and drives ferroptosis.(A) Forest plot of Mendelian Randomization (MR) analysis evaluating the causal association between plasma phosphatidylcholine (PC) levels and focal epilepsy risk. (B) Scatter plot visualizing the MR causal estimate for the effect of PC(16:0/18:2) levels on focal epilepsy.(C-G) Human plasma targeted lipidomics analysis comparing the Control (Con) and Microplastic-exposed (Micro) groups. Results show: Upregulation of key ferroptosis-related lipid markers, including (C) Eicosapentaenoic acid, (D) Arachidonic Acid, (E) Linoleic Acid, and (F) 8-iso-PGF2-alpha. ** P < 0.01

To provide clinical evidence supporting the link between prolonged microplastic exposure and exacerbated seizure severity, we analyzed a pediatric cohort (Cohort 2, Figure S5 and Table S4). Using a machine learning pipeline (Method S11-12) integrating feature selection via multivariate logistic regression and the Boruta algorithm (Figure S6–S8), we compared nine distinct models to identify non-specific predictors of seizure severity [28, 29]. This analysis revealed that the Gradient Boosting Machine (GBM) algorithm demonstrated the highest predictive accuracy for seizure risk based on non-specific peripheral blood markers (Fig. 5A–D, Figures S9–S10) [28, 29]. To interpret the GBM model, we employed Shapley Additive Explanations (SHAP) analysis, which identified Red Blood Cell Distribution Width (RDW) as the most significant independent risk factor (Fig. 5F–G). SHAP values confirmed a strong positive correlation, indicating that elevated RDW values were associated with an increased predicted probability of seizures (Fig. 5H). Critically, RDW levels were significantly higher in the microplastic-exposed group compared to non-exposed controls (Fig. 5I). Given that RDW is an established marker of systemic oxidative stress, inflammation, and lipid metabolic imbalance, this clinical finding strongly supports our central hypothesis: prolonged respiratory microplastic exposure induces systemic oxidative stress and inflammation, leading to lipid metabolic disruption, altered gene expression, ferroptosis activation, and ultimately, exacerbation of seizure manifestations.

Fig. 5.

Interpretable machine learning identifies RDW as a key predictor linking respiratory microplastic exposure to seizure risk. (A) Boruta ridge plot for initial feature importance screening. (B) Receiver Operating Characteristic (ROC) curve analysis for 12 selected clinical features and their combined predictive power. (C-D) We assessed the performance of nine machine learning models using (C) ROC and (D) Precision-Recall (PR) curves, identifying the Gradient Boosting model as the top performer. (E) Gradient Boosting model was then analyzed for feature importance and cumulative AUC to determine the final set of predictive features. (F) Feature ranking by mean absolute SHAP values, identifying Red Blood Cell Distribution Width (RDW) as the top predictor. (G) SHAP summary plot illustrating the impact of top features on model predictions. (H) SHAP dependence plot showing the positive correlation between RDW and predicted seizure risk. (I) Comparison of RDW levels between control (Con) and microplastic-exposed (Micro) groups. **P < 0.01. SAbbreviations: CO₂, Carbon dioxide; Cr, Creatinine; Hb, Hemoglobin; HCT, Hematocrit; MCH, Mean Corpuscular Hemoglobin; MCV, Mean Corpuscular Volume; Neu_Pct, Neutrophil Percentage; PA, Prealbumin; RDW, Red Blood Cell Distribution Width

To elucidate the exposure-metabolite-target gene regulatory network underlying respiratory microplastic exposure, we next investigated expression alterations of ferroptosis-related genes (Method S13). Specifically, we focused on genes involved in lipid metabolism (ACSL4), antioxidant defense and inflammation (NRF2/NFE2L2, SLC7A11, GPX4), and NRF2-mediated iron metabolism (FTH1 and FPN1). As shown in Fig. 6A–F, prolonged microplastic exposure significantly upregulated ACSL4 expression while downregulating NRF2, SLC7A11, GPX4, FTH1, and FPN1 (P < 0.05). Furthermore, Spearman correlation analyses revealed the pivotal role of NRF2 expression in mediating the regulation of SLC7A11, FTH1, and FPN1 (Fig. 6G). To genetically link these ferroptosis-related gene alterations to seizure risk, we performed Mendelian Randomization analysis. This analysis revealed a significant association: genetically predicted lower expression of the ferroptosis-regulatory gene FTH1 correlates with increased risk of focal epilepsy, a finding further validated by colocalization analysis (Fig. 6H–J).To validate the role of ferroptosis in seizure pathology, we conducted an in-depth analysis of single-cell RNA sequencing data from the DISCO database, utilizing surgically resected pediatric epileptic brain tissue and normal hippocampal tissue specimens (Method S14 and Figure S11-S15). UMAP visualization and cell-type proportion analyses revealed significant cellular compositional shifts in seizure tissue, specifically affecting ferroptosis-vulnerable neurons and inflammation-associated microglia (Fig. 7A–B). At the molecular level, we observed significantly altered expression of key ferroptosis-related genes (NFE2L2, SLC7A11, FTH1, and GPX4) (Fig. 7C). Pathway enrichment analyses, including KEGG pathway analysis and programmed cell death activity scoring, identified ferroptosis as a highly prominent and enriched mechanism in the pathophysiology of epilepsy (Fig. 7D–E). Additionally, analysis of publicly available RNA-seq data (GSE143272), comprising peripheral blood samples from 91 epilepsy patients and 50 healthy controls, confirmed the dysregulation of key ferroptosis-related genes in epilepsy (Method S15 and Fig. 8A–B). Critically, our own clinical validation study in newly diagnosed pediatric seizure patients (Cohort 3) employed advanced neuroimaging techniques (MRS and MRI) to demonstrate that hippocampal iron accumulation and lipid metabolic disruption are characteristic neurochemical features of epilepsy pathophysiology, providing direct in vivo evidence of the iron-lipid dysregulation underlying ferroptosis susceptibility (Method S16–S17, Fig. 8C–D). Furthermore, RT-qPCR analysis of ferroptosis-related gene expression in human peripheral blood revealed a similar dysregulation pattern (Fig. 8E–J). Collectively, these comprehensive analyses provide multi-dimensional evidence supporting our central hypothesis: prolonged respiratory microplastic exposure induces systemic oxidative stress and inflammation, leading to lipid metabolic disruption, altered ferroptosis-related gene expression, ferroptosis activation, and ultimately, the exacerbation of seizure manifestations.

Fig. 6.

Respiratory microplastic exposure disturbs ferroptosis-related gene expression and links FTH1 to focal epilepsy risk. (A-F) Relative expression levels of ferroptosis-related genes in human blood. The results show significantly upregulated (A) ACSL4 and downregulated (B) NRF2, (C) SLC7A11, (D) GPX4, (E) FTH1, and (F) FPN1 in the microplastic-exposed group compared to controls. (G) Spearman correlation analysis between NRF2 expression and its downstream targets (SLC7A11, FTH1, and FPN1). (H) Forest plot of Mendelian Randomization (MR) analysis evaluating the causal association between ferroptosis-related genes and focal epilepsy. FTH1 (highlighted in red) shows a significant association. (I) Scatter plot visualizing the MR analysis results for the effect of FTH1 expression on focal epilepsy. (J) Colocalization analysis of FTH1 expression and focal epilepsy susceptibility. * P < 0.05, ** P < 0.01

Fig. 7.

Single-cell analysis of the human epileptic hippocampus reveals altered cell composition and active ferroptosis. (A) UMAP plots of single-cell transcriptomes from control versus seizure-associated brain tissue, showing distinct cell clusters. (B) Absolute (left) and relative (right) cell type proportions in control and seizure tissues, highlighting shifts in neuron and microglia populations. (C) UMAP and violin plots showing altered expression of key ferroptosis-related genes (NRF2, SLC7A11, FTH1, GPX4) in seizure tissue. (D) Top 30 enriched KEGG pathways in seizure tissue, including ferroptosis. (E) Circular bar plot ranking the activity of various programmed cell death pathways, with ferroptosis being prominent. *P < 0.05, **P < 0.01

Fig. 8.

Clinical validation confirms ferroptosis-related gene alterations in the pathophysiology of seizures.(A) Differential expression analysis of key ferroptosis-related genes (IL6R, FTH1, GPX4, NRF2) in the blood of epilepsy patients compared to controls, based on the GSE143272 dataset.(B) Spearman correlation analysis revealing a positive correlation between NRF2 and FTH1 expression.(C) Representative 1 H-MRS images and quantification of relative hippocampal lipid levels. (D) Representative T2-weighted MRI images and quantification of relative brain iron levels. (E-J) Relative expression levels of ferroptosis-related genes in human blood. Compared to the control group, the microplastic-exposed group showed significantly upregulated (E) ACSL4 and downregulated (F) NRF2, (G) SLC7A11, (H) GPX4, (I) FTH1, and (J) FPN1. * P < 0.05, ** P < 0.01

Microplastic exposure worsens cognitive and anxiety-like deficits in epileptic mice

To validate the role of microplastics in the pathophysiology of epilepsy, we conducted a clinical validation study in newly diagnosed pediatric patients. This analysis revealed that patients exhibited significantly higher blood concentrations of prevalent microplastics, including PE, PP, and PS, compared to healthy controls (Fig. 9A–B and Figure S16). Furthermore, this microplastic burden demonstrated a dose-response relationship with seizure severity (Fig. 9C). Cytokine profiling indicated that elevated IL-6 levels were positively associated with the degree of microplastic exposure (Fig. 9D–E), while oxidative stress assays demonstrated that microplastic exposure induced systemic oxidative stress (Fig. 9F–I). These findings suggest that microplastic-induced systemic oxidative stress and inflammation are intimately involved in the pathophysiology of epilepsy, further supporting our proposed exposure-metabolite-target gene regulatory network.

Fig. 9.

Clinical validation demonstrates that microplastic exposure exacerbates inflammation, oxidative stress, and seizure severity. (A) Identification and characterization of different microplastic types (PP, PS, PE) in human blood using high-resolution imaging and Raman spectroscopy. (B) Paired analysis of relative microplastic concentrations (Total, PS, PP, PE) in the control and microplastic-exposed groups. (C) Subgroup analysis demonstrating a positive association between blood microplastic levels and seizure severity. (D) Relative levels of the inflammatory cytokine IL-6 in the blood of control subjects (Con) versus seizure patients (Seizures). (E) Spearman correlation analysis revealing a positive correlation between microplastic levels and IL-6 concentration. (F-H) Assessment of oxidative stress markers in seizure patients compared to controls: (G) MDA, (H) GSH, and (I) SOD activity. (I) Spearman correlation analyses showing that microplastic levels are positively correlated with MDA and negatively correlated with GSH and SOD activity. * P < 0.05, ** P < 0.01

To validate the neurotoxic effects of prolonged microplastic exposure, we established a murine epilepsy model. Mice were intratracheally instilled with 5µm polystyrene microplastics daily for 90 days to simulate inhalation exposure and induce systemic inflammation and oxidative stress, followed by kainic acid (KA)-induced seizures (Method S18; Figure S17 and Fig. 10A) [13, 30–32]. Results indicated that by day 90, before KA injection, microplastic exposure had significantly increased the body weight of mice compared to controls (P < 0.05), a change potentially linked to disruptions in lipid metabolism (Fig. 10B) [33]. Notably, no spontaneous seizures were observed during the 90-day exposure period, suggesting that inhaled microplastics do not directly trigger seizures but rather tend to exacerbate their severity. Following KA injection, the mice exposed to microplastics showed significantly aggravated seizure symptoms, reflected in increased Racine scores, seizure frequency, and duration (P < 0.05; Method S19 and Fig. 10C–E). However, this exacerbated seizure phenotype was markedly alleviated with pretreatment using either the ferroptosis inhibitor Ferrostatin-1 (Fer-1) or melatonin.

Fig. 10.

Exposure to polystyrene microplastics worsens both the seizures and the behavioral problems caused by kainic acid in mice. (A) FE-SEM image of the polystyrene microplastics. (B) Mouse body weight over the 90-day exposure period. (C-E) Seizure severity assessment, including (C) Racine scores, (D) duration, and (E) frequency. (F-G) Evaluation of recognition memory using the Novel Object Recognition (NOR) test, showing (F) exploration time for familial (F) and novel (N) objects, and (G) the discrimination rate. (H-L) Morris Water Maze (MWM) for spatial memory: (H) swim speed, (I) escape latency over 5 training days, where a slower rate of decline indicates impaired spatial learning capabilities, (J) probe trial paths, (K) platform crossings, and (L) time in target quadrant. (M-O) Open Field Test for anxiety-like behavior, with (M) travel paths and quantification of (N) center entries and (O) time in the center. (P-Q) Marble Burying Test for anxiety, showing (P) representative results and (Q) percentage of marbles buried. *P < 0.05, **P < 0.01

We conducted Novel Object Recognition (NOR) and Morris Water Maze (MWM) tests to evaluate the cognitive impact of microplastic exposure on epileptic mice (Method S20-S21). Both assessments confirmed that while KA induced cognitive dysfunction, microplastic co-exposure significantly worsened these deficits. In the NOR test, both KA and KA+Micro groups showed impaired recognition memory, failing to distinguish between familiar and novel objects (P > 0.05; Fig. 10F). However, the discrimination index revealed a significantly more severe cognitive deficit in the KA+Micro group compared to the KA-only group (P < 0.05; Fig. 10G). Similarly, in the MWM test, we first analyzed swimming speeds to rule out the interference of locomotor dysfunction. The swimming speeds were comparable across all groups (P > 0.05; Fig. 10H), confirming that the observed performance differences were due to cognitive deficits rather than motor impairment. In the spatial learning phase, the escape latency served as a key indicator of learning capability. While control mice rapidly learned the platform location, microplastic exposure significantly exacerbated the learning deficits induced by KA, characterized by a much slower rate of decline in escape latencies over the training days (P < 0.05; Fig. 10I). Furthermore, in the probe trial conducted to assess memory retention, the KA + Microplastics group exhibited highly disorganized search patterns (Fig. 10J) compared to the KA group. This spatial memory impairment was quantitatively confirmed by significantly fewer platform crossings (Fig. 10K) and reduced time spent in the target quadrant (Fig. 10L), indicating a failure to recall the platform’s location. Notably, these microplastic-induced cognitive impairments were effectively alleviated by Melatonin or Fer-1 treatment. Anxiety-like behaviors were evaluated using the open-field and marble-burying tests (Methods S22–S23). Both assays indicated that while KA alone increased anxiety-like behavior, microplastic co-exposure further exacerbated this phenotype. In the open-field test, the KA+Micro group spent significantly less time in and made fewer entries into the center zone compared to the KA-only group (P < 0.05; Fig. 10M–O). This finding was corroborated by the marble-burying test, in which the KA+Micro group buried significantly more marbles (P < 0.05; Fig. 10P–Q). Together, these results provide compelling evidence that prolonged respiratory exposure to polystyrene microplastics not only intensifies seizure severity in epileptic mice but also significantly aggravates associated cognitive dysfunction and anxiety-like behaviors, effects that can be mitigated by targeting ferroptosis.

Melatonin protects against microplastic-induced neurotoxicty on epilptice mice

To further test our hypothesis that chronic respiratory exposure to microplastics triggers systemic oxidative stress and inflammation, leading to lipid metabolism dysregulation, altered expression of ferroptosis-related genes, ferroptosis activation, and ultimately exacerbation of seizure phenotypes, we first evaluated key hallmarks of ferroptosis in the brain. T2-weighted MRI revealed that microplastic exposure significantly exacerbated iron accumulation in the brains of KA-treated epileptic mice (P < 0.05; Figs. 11A–B). Consistent with this, MRS demonstrated a further increase in brain lipid levels following microplastic exposure (P < 0.05; Figs. 11C–D). Nissl staining (Method S24) confirmed pronounced neurotoxicity, showing significant neuronal loss in the CA1 and CA3 subfields of the hippocampus (Figs. 11E–H). Critically, pretreatment with either the ferroptosis inhibitor Ferrostatin-1 (Fer-1) or melatonin markedly attenuated these pathological changes, reducing both iron and lipid accumulation and preserving neuronal integrity (P < 0.05). To identify the underlying molecular targets, we performed RNA-seq analysis (Method S25). Comparative transcriptomics between the KA+Micro and KA-only groups revealed significant differential expression of genes associated with inflammation and ferroptosis, including Il6, Fth1, and Nfe2l2 (Figs. 12A–B). GO, KEGG, and GSEA consistently highlighted the ferroptosis pathway as central to the observed pathology (Figs. 12C–E). Furthermore, machine learning approaches, specifically Support Vector Machine-Recursive Feature Elimination (SVM-RFE) and Random Forest-RFE (RF-RFE), identified Nfe2l2 (Nrf2) as a pivotal gene mediating the neurotoxic effects of microplastics on seizure exacerbation (Figs. 12F–I). Collectively, these findings indicate that microplastics exacerbate seizure symptoms by triggering ferroptosis.

Fig. 11.

Microplastic exposure worsens KA-induced iron accumulation, lipid dysregulation, and neuronal loss in the hippocampus.(A-B) Assessment of hippocampal iron accumulation using (A) T2-weighted MRI (arrows indicate T2 hypointensity) and (B) quantification of relative iron levels. (C-D) Measurement of hippocampal lipid levels by (C) 1 H-MRS (arrows indicate lipid peak) and (D) quantification. Scale bars for A & C = 1 cm. (E) Representative Nissl-stained sections of hippocampal subfields (CA1, CA3, DG), with arrows indicating neuronal damage. Scale bar = 500 μm. (F-H) Quantification of surviving neurons in the (F) CA3, (G) CA1, and (H) DG subfields. *P < 0.05, **P < 0.01

Fig. 12.

Hippocampal transcriptomics reveal microplastic-induced dysregulation of ferroptosis pathways in mice. (A) PCA plot of hippocampal gene expression from KA+Micro versus KA groups. (B) Volcano plot identifying differentially expressed genes (DEGs) between the two groups. (C) Gene Ontology (GO) analysis of DEGs showing top enriched terms. (D) KEGG analysis showing top enriched pathways associated with DEGs. (E) GSEA plot demonstrating significant enrichment of the ferroptosis pathway. (F) Feature selection via Support Vector Machine (SVM), identifying 41 optimal DEGs based on model accuracy and error rate. (G) Feature selection via Random Forest, identifying 12 key DEGs based on minimal error rate. (H) Venn diagram showing the overlap of DEGs identified by SVM and Random Forest, with Nfe2l2 highlighted as a hub gene. *P < 0.05, **P < 0.01

To validate the proposed exposure-metabolite-target gene regulatory network and elucidate the neuroprotective mechanism of melatonin, we gathered convergent cellular, biochemical, and molecular evidence. Transmission electron microscopy (TEM; Method S26) revealed ultrastructural features characteristic of ferroptosis in hippocampal neurons of KA-treated mice, including mitochondrial shrinkage and increased membrane density (Fig. 13A). Microplastic exposure intensified this damage, whereas pretreatment with either Fer-1 or melatonin substantially preserved mitochondrial morphology. At the biochemical level, prolonged respiratory microplastic exposure further elevated serum IL-6 levels and disrupted lipid homeostasis in KA-induced epileptic mice, as evidenced by increased phosphatidylcholine (PC) and polyunsaturated fatty acid (PUFA) concentrations, changes that correlated with the clinical severity observed in our pilot cohort. Additionally, elevated levels of 8-iso-PGF2α, a marker of lipid peroxidation, were detected, supporting the notion that microplastic exposure promotes lipid peroxidation and thereby triggers neuronal ferroptosis (Method S27). Notably, both Fer-1 and melatonin pretreatment significantly ameliorated these adverse effects.

Fig. 13.

Melatonin mitigates microplastic-exacerbated seizure severity by inhibiting hippocampal ferroptosis in mice.(A) Representative Transmission Electron Microscopy (TEM) images of hippocampal mitochondria. Red arrows indicate characteristic ferroptotic features, such as shrunken mitochondria and increased membrane density. Scale bar = 1 μm.(B-E) Analysis of serum biochemical markers across groups: (B) IL-6, (C) Phosphatidylcholine (PC), (D) Polyunsaturated Fatty Acids (PUFA), and (E) 8-iso-PGF2-alpha.(F-I) Quantification of oxidative stress and ferroptosis markers in the mouse hippocampus: (F) GSH, (G) MDA, (H) PTGS2, and (I) 4-HNE.(J) Western blot analysis and quantitative normalization of hippocampal proteins involved in inflammation (IL-6, p-NF-κB, NF-κB), the antioxidant response (Nrf2, Keap1), and ferroptosis (Fth1, Fpn1, Slc7a11, Gpx4, Acsl4). * P < 0.05, ** P < 0.01

To further dissect the mechanisms underlying melatonin’s neuroprotection, we quantified key ferroptosis-related biomarkers (Methods S28–S29). Microplastic exposure amplified KA-induced increases in hippocampal Fe²⁺, MDA, PTGS2, and 4-HNE, while concurrently depleting the antioxidant glutathione (GSH) (Figs. 13F–I and Figure S18). Both Fer-1 and melatonin effectively reversed these alterations, restoring redox balance and antioxidant capacity (P < 0.05). Finally, western blot analysis (Method S30) was used to interrogate the molecular pathways within our proposed regulatory network. Microplastic exposure upregulated pro-inflammatory proteins (Il6 and NF-κB) and the pro-ferroptotic enzyme Acsl4, while downregulating key anti-ferroptotic proteins, including Nrf2, Keap1, Fth1, Fpn1, Slc7a11, and Gpx4. In contrast, melatonin treatment robustly reversed this entire expression profile (P < 0.05), suppressing neuroinflammation and enhancing the Nrf2-mediated cellular defense system against ferroptosis (Fig. 13J). Together, these multimodal findings provide compelling evidence that melatonin exerts potent neuroprotective effects against microplastic-induced neurotoxicity in epileptic mice by inhibiting neuroinflammation and activating the Nrf2-dependent anti-ferroptotic pathway.

Discussion

As global plastic production accelerates, microplastic contamination has emerged as a critical environmental and public health challenge. Microplastics, defined as plastic particles smaller than 5 mm, can be readily inhaled or ingested, leading to bioaccumulation [34]. Recent evidence identifies airborne microplastics as emerging neurotoxic hazards, with children representing a particularly vulnerable population [35]. Pediatric susceptibility stems from several physiological factors: higher respiratory rates per unit body mass result in greater inhaled microplastic burdens, while immature respiratory, immune, and blood-brain barrier systems may compromise particle clearance and enhance systemic distribution [36, 37]. Despite growing concerns, the neurotoxicological effects of airborne microplastic exposure under real-world conditions, particularly in children, remain poorly understood [38]. Among pediatric neurological disorders, epilepsy is the most common chronic condition, characterized by sudden, uncontrolled alterations in behavior, movement, consciousness, or sensation [37, 39]. Notably, seizure susceptibility is highly sensitive to perturbations in brain homeostasis, and epileptic episodes may serve as early indicators of microplastic-induced neurotoxicity [40].Previous studies in zebrafish have demonstrated that microplastic exposure increases seizure-like behaviors. Furthermore, prior experimental work utilizing zebrafish models has demonstrated that microplastic exposure may elevate seizure-like behavioral events [41], however, the underlying mechanisms by which airborne microplastics contribute to seizure pathophysiology remain largely unknown. This study addresses this critical knowledge gap by investigating the neurotoxicological mechanisms linking airborne microplastic exposure to seizure severity and associated neurological complications in pediatric populations.

In this study, we first conducted a preliminary observational cohort investigation (Cohort 1) involving 15 plastic manufacturing workers with > 10 years of occupational exposure, selected based on their documented high risk of chronic respiratory microplastic contact. Comprehensive profiling of cytokines, oxidative stress markers, and Raman spectroscopic analysis of blood samples revealed significantly higher levels of prevalent environmental microplastics in the circulation of exposed individuals compared to unexposed controls under real-world exposure conditions. Prolonged respiratory microplastic exposure was associated with elevated systemic oxidative stress and inflammation. Notably, although we observed statistically significant increases in pro-inflammatory cytokines (IL-6, IL-1β, and TNF-α) following long-term exposure, these levels remained below clinical thresholds requiring medical intervention, suggesting that the neurotoxic risk of ambient microplastic exposure in the general population is likely limited under typical environmental conditions. We therefore hypothesize that microplastic exposure may act as a “second hit,” not directly triggering seizures but exacerbating their severity by establishing a subthreshold pro-oxidant and pro-inflammatory microenvironment that promotes neuronal hyperexcitability and lowers seizure thresholds. It is also important to note that Raman spectroscopy identified PE as the predominant microplastic polymer accumulating in human peripheral blood. However, for subsequent animal experiments, we selected PS due to its well-characterized physicochemical properties, which enable more reproducible and controlled experimental exposures. The relatively larger particle sizes detected in human blood, as also reported in prior studies, may result from ex vivo aggregation of smaller nanoplastics during sample processing [42]. Our protocol involved chemical digestion and ethanol washing to remove organic matrices, which can promote rapid clustering of nano-sized particles into micrometer-scale aggregates. While this limits precise size determination, it reliably reflects the total body burden of plastic polymers and supports the observed correlation between microplastic load and symptom severity. Collectively, our findings underscore that the primary neurotoxic mechanism of inhaled microplastics involves the induction of systemic oxidative stress and inflammation, which may pose significant hazards to seizure severity and neurological outcomes.

In this study, bibliometric analysis highlighted the critical interplay among oxidative damage, inflammatory responses, and lipid metabolic dysregulation as central determinants of microplastic-induced neurotoxicity, thereby uncovering a potential exposure-metabolite-target gene regulatory network. To further explore this axis, we performed untargeted metabolomic profiling of human plasma from individuals with chronic respiratory microplastic exposure. Pathway enrichment analysis using the KEGG database revealed significant perturbations in lipid metabolism, particularly in glycerophospholipid metabolism, lipoic acid metabolism, and linoleic acid metabolism. Complementary Reactome pathway analysis further identified dysregulation in glycerophospholipid catabolism, biosynthesis, and SLC-mediated transmembrane transport, processes intimately linked to ferroptosis. Ferroptosis is a recently defined form of regulated cell death driven by iron-dependent accumulation of lethal lipid peroxidation products [43]. In the context of epilepsy, prior animal studies have demonstrated that iron overload and lipid peroxidation trigger ferroptotic death of hippocampal neurons, with oxidative stress serving as the primary mechanistic driver [6, 7, 12]. This process critically depends on the metabolism of polyunsaturated fatty acid (PUFA)-containing glycerophospholipids. Moreover, SLC-mediated transport plays a pivotal role: impaired SLC7A11-dependent cystine uptake and FPN1-mediated iron export create a cellular milieu that favors uncontrolled lipid peroxidation, ultimately promoting neuronal loss in the hippocampus and worsening seizure outcomes. Integrating SVM-RFE and ANN approaches, we identified four key differential metabolites associated with chronic respiratory microplastic exposure: 15(R)-hydroperoxy-EPE and three phosphatidylcholines, PC(O-18:0/18:2), PC(20:4/0:0), and PC(16:0/18:2). Notably, 15(R)-hydroperoxy-EPE is a lipid hydroperoxide derived from eicosapentaenoic acid (EPA, an omega-3 PUFA). Contrary to its typical anti-inflammatory reputation, EPA acts as a pro-ferroptotic substrate due to its penta-unsaturated structure (five double bonds), which renders it highly susceptible to peroxidation and accelerates membrane disintegration [44]. Similarly, all three phosphatidylcholines are established pro-ferroptotic mediators: PC(O-18:0/18:2) represents an ether-linked phospholipid that enhances membrane susceptibility to ferroptosis [45]; PC(20:4/0:0) reflects mobilization of arachidonic acid (AA,20:4), the core substrate for lipid peroxidation [46]; PC(16:0/18:2) serves as a reservoir of linoleic acid (LA,18:2), an upstream precursor for AA synthesis and a direct source of the toxic 4-HNE [47]. Mendelian randomization analysis further supported a causal association between dysregulation of PC(16:0/18:2) and increased seizure severity, while targeted human lipidomic profiling indicated the dysregulation of EPA, AA, and linoleic acid. Importantly, despite their classification as either omega-3 (EPA) or omega-6 (AA, LA) PUFAs, all serve as polyunsaturated substrates for peroxidation. Their collective accumulation and subsequent incorporation into membrane phospholipids via ACSL4 markedly elevate the membrane peroxidizability index, thereby fueling the ferroptotic cascade and driving neuronal death [48]. This conclusion is further corroborated by elevated levels of 8-iso-PGF₂α, a specific downstream product of arachidonic acid peroxidation and a well-established biomarker of ferroptosis, confirming that prolonged respiratory microplastic exposure indeed triggers ferroptosis. Collectively, these findings support our central hypothesis: chronic respiratory microplastic exposure induces systemic oxidative stress and inflammation, which disrupt lipid metabolism, alter ferroptosis-related gene expression, activate ferroptotic neuronal death, and ultimately exacerbate seizure manifestations (Fig. 14).

Fig. 14.

Schematic illustration of the proposed mechanism linking respiratory microplastic exposure to exacerbated seizure symptoms

To provide clinical evidence that prolonged microplastic exposure may exacerbate seizure severity, we conducted a second observational cohort study involving pediatric patients with seizure disorders. This study aimed to determine whether chronic respiratory microplastic exposure acts as an independent risk factor for worsened seizure manifestations. By integrating nine distinct machine learning algorithms with SHAP interpretability analysis, we identified RDW as the strongest independent predictor of seizure risk and found that prolonged microplastic exposure significantly modulated RDW levels, thereby contributing to symptom exacerbation [28, 29]. RDW reflects heterogeneity in red blood cell size and is mechanistically linked to oxidative stress, inflammation, and lipid metabolism through multiple interconnected pathways. ROS impair bone marrow erythropoiesis and damage circulating erythrocytes, leading to increased size variability. Oxidative stress also promotes protein degradation, phospholipid peroxidation, and erythrocyte swelling, all of which elevate RDW [49]. Clinically, RDW correlates strongly with pro-inflammatory cytokines such as IL-6 and IL-18; these cytokines suppress erythropoietin-mediated erythroid maturation, further increasing RDW [50]. Moreover, emerging evidence suggests RDW serves as an intermediary biomarker connecting central adiposity to dyslipidemia, with body mass index influencing RDW indirectly via inflammatory mediators like C-reactive protein [51]. In the context of epilepsy, RDW has been associated with both seizure prognosis and, to a lesser extent, seizure classification, particularly in critically ill and pediatric populations. Prior studies have shown that elevated RDW independently predicts higher in-hospital mortality and prolonged ICU stays among patients primarily diagnosed with seizure disorders, offering prognostic value beyond conventional severity scores such as APACHE IV [52]. Additionally, RDW levels are significantly higher in children with complex febrile seizures compared to those with simple febrile episodes, highlighting its potential utility in differentiating seizure subtypes [53]. Collectively, these findings support our central hypothesis: chronic respiratory microplastic exposure induces systemic oxidative stress and inflammation, which disrupt lipid metabolism and elevate RDW, a key integrative biomarker, thereby exacerbating seizure severity in pediatric patients.

To further validate the roles of microplastic burden and ferroptosis in seizure pathophysiology, we first confirmed significant differences in the expression of ferroptosis-associated genes between workers with chronic respiratory microplastic exposure and unexposed controls. Mendelian randomization analysis further established a causal link between genetic variants in ferroptosis-related genes and seizure severity, highlighting FTH1 as a key mediator. FTH1, a core subunit of ferritin, plays a critical protective role against ferroptosis by sequestering redox-active iron within inert ferritin complexes, thereby preventing iron-catalyzed lipid peroxidation, the hallmark driver of ferroptotic cell death [54]. To dissect the cellular context of this pathway, we performed single-cell RNA sequencing on surgically resected brain tissues from pediatric patients with drug-resistant epilepsy, using histologically normal cortical specimens as controls. Comparative analysis revealed pronounced dysregulation of ferroptosis-related genes, including NRF2, SLC7A11, FTH1, and GPX4, in neurons and glial cells from epileptic tissue. KEGG pathway enrichment and programmed cell death activity scoring consistently identified ferroptosis as the dominant cell death mechanism. Notably, signatures of apoptosis and autophagy were also present, suggesting that neuronal loss in epilepsy involves a coordinated interplay of multiple regulated cell death pathways. Oxidative stress appears to act as a common upstream trigger, simultaneously driving neuroinflammation, lipid metabolism disruption, and ferroptosis activation [55, 56]. These findings were corroborated by bulk RNA-seq analysis of a public peripheral blood dataset comprising 91 epilepsy patients and 50 healthy controls, which confirmed systemic dysregulation of core ferroptosis-related genes in epilepsy. Finally, a third observational cohort of children with newly diagnosed seizures was enrolled to clinically validate our hypothesis that prolonged respiratory microplastic exposure exacerbates seizure severity via ferroptosis activation. Blood concentrations of total microplastics, and specifically PE, PP, and PS, were significantly higher in seizure patients than in healthy controls. Moreover, subgroup stratification by clinical seizure severity revealed a positive dose-response relationship: higher microplastic levels correlated with more severe manifestations. Consistent with a pro-ferroptotic state, seizure patients exhibited elevated levels of the pro-inflammatory cytokine IL-6, which positively correlated with circulating microplastic concentrations. Parallel increases were observed in oxidative stress biomarkers, including MDA, alongside reductions in GSH and SOD activity. Expression profiles of key ferroptosis-related genes in peripheral blood mirrored those seen in brain tissue and occupational cohorts. Collectively, these multimodal human data robustly support our central model: chronic respiratory microplastic exposure induces systemic oxidative stress and inflammation, disrupts lipid homeostasis, dysregulates ferroptosis-related gene networks, and ultimately promotes ferroptotic neuronal injury, thereby exacerbating seizure severity in susceptible individuals.

Finally, to validate the hypothesized exposure-metabolite-gene regulatory network underlying microplastic-induced neurotoxicity in seizure disorders, we employed a KA-induced epileptic mouse model with chronic intratracheal instillation of 5µm polystyrene microplastics to mimic realistic respiratory exposure. Comprehensive multimodal assessments, including behavioral scoring, T2-weighted MRI, MRS, and Nissl staining, demonstrated that chronic PS-MP exposure disrupts cerebral iron and lipid homeostasis, exacerbating seizure severity through ferroptosis-dependent mechanisms. Integration of hippocampal RNA-sequencing data with machine learning approaches (including SVM-RFE and random forest) further identified NRF2 as a central regulatory node orchestrating the neurotoxic response to prolonged atmospheric PS-MP exposure in epileptic mice. Notably, melatonin exhibited potent neuroprotective effects. Mechanistically, melatonin mitigates ferroptosis by simultaneously suppressing neuroinflammation and oxidative stress: it reduces IL-6 levels and inhibits NF-κB activation, while decreasing MDA and restoring GSH pools. Critically, melatonin activates the NRF2 antioxidant defense pathway, leading to coordinated upregulation of System xc⁻ (via SLC7A11) and GPX4, key guardians against lipid peroxidation. Concurrently, melatonin significantly downregulates ACSL4, the enzyme responsible for esterifying polyunsaturated fatty acids into membrane phospholipids, thereby limiting the availability of peroxidation-susceptible substrates. By enhancing GPX4-mediated detoxification while restricting ACSL4-driven pro-ferroptotic lipid remodeling, melatonin shifts the cellular balance toward survival and effectively counteracts PS-MP-induced neuronal damage and seizure aggravation (Figure S19).

Limitations

Our study has several important limitations that warrant acknowledgment. First, the pilot nature of our human cohorts, particularly the small sample size in both the occupational and pediatric groups, limits the statistical power and generalizability of our findings. Although our multimodal approach strengthens internal validity, larger, more diverse populations are needed to confirm these associations across different demographics and exposure settings. Moreover, within the occupational cohort, potential confounders such as variable use of personal protective equipment, co-exposure to other airborne particulates (e.g., metal fumes, volatile organic compounds), and psychological stress were not fully accounted for. Additionally, the cross-sectional design precludes causal inference and prevents assessment of long-term neurological outcomes; longitudinal follow-up is essential to evaluate the chronic effects of microplastic exposure. Second, our experimental models focused exclusively on polystyrene microplastics, which, while useful for controlled mechanistic studies due to their well-defined physicochemical properties, do not reflect the complexity of real-world exposure. In environmental and biological contexts, humans are typically exposed to heterogeneous mixtures of polymers, including polyethylene and polypropylene, along with plastic-associated additives (e.g., phthalates, bisphenols) that may exert independent neurotoxic effects. Our design could not disentangle the contributions of the polymer matrix from those of leachable chemical constituents. Future studies should investigate complex microplastic mixtures to better approximate actual human exposure scenarios. Third, the intratracheal instillation method used in our animal model, while enabling precise dose control and reproducibility, does not fully recapitulate the physiological dynamics of natural inhalation exposure. Inhalation exposure would more accurately simulate real-world human respiratory exposure patterns, including particle deposition kinetics, mucociliary clearance, and dose-rate effects. Furthermore, the PS particles used in this study are too large to cross the blood-brain barrier (BBB), and transmission electron microscopy analyses did not detect microplastic polymers in the hippocampus. Consequently, our study does not address the potential translocation of nanoplastics or their direct neurotoxic effects following BBB penetration. Given growing evidence that nanoplastics can enter the brain, future work must employ inhalation exposure models and investigate the biodistribution, cellular uptake, and direct neuronal impacts of nanoplastics. Fourth, while micro-Raman spectroscopy confirmed the presence of PE, PP, and PS in human blood, whether as primary nanoplastics or ex vivo aggregates, the technique cannot reliably distinguish particle size or quantify mass concentration in biological matrices. Aggregation during sample processing may have inflated apparent particle sizes, potentially masking the true burden of nanoplastics. Advanced analytical methods such as Py-GC/MS or LDIR chemical imaging are needed to accurately characterize the size distribution, polymer composition, and absolute load of micro- and nanoplastics in tissues and biofluids. Fifth, although machine learning identified RDW as a robust predictive biomarker for seizure severity in the context of microplastic exposure, RDW is inherently non-specific and can be influenced by anemia, infection, or chronic inflammation. Its specificity to microplastic-related neurotoxicity requires further validation in controlled clinical settings. Sixth, while ferroptosis emerged as a central mechanism, our focus may have overshadowed contributions from other regulated cell death pathways (e.g., apoptosis, autophagy) or additional ferroptosis regulators such as LPCAT3, which plays a key role in remodeling PUFA-containing phospholipids. A more comprehensive multi-omics approach could uncover complementary mechanisms. Finally, the neuroprotective effects of melatonin observed in our murine model are promising but preliminary. Before clinical translation for microplastic-associated neurological disorders, its efficacy, optimal dosing, long-term safety, and impact on cognitive outcomes must be rigorously evaluated in prospective, placebo-controlled trials. Collectively, these limitations highlight critical avenues for future research and underscore the need for integrated environmental, toxicological, and clinical investigations to fully understand and mitigate the neurological risks of microplastic pollution.

Conclusion

To the best of our knowledge, this study represents the first comprehensive investigation into the effects of prolonged respiratory polystyrene microplastic exposure on seizure disorders. Our findings establish a novel exposure-metabolite-gene regulatory network and provide multimodal evidence, spanning epidemiological, clinical, omics, and experimental domains, that robustly supports a central pathophysiological model: chronic respiratory microplastic exposure triggers systemic oxidative stress and inflammation, disrupts lipid homeostasis, dysregulates ferroptosis-related gene networks (including NRF2, SLC7A11, GPX4, and ACSL4), and ultimately drives ferroptotic neuronal injury, thereby exacerbating seizure severity in susceptible individuals. Critically, we demonstrate a dose-response relationship between blood microplastic burden and clinical seizure severity in pediatric patients, offering compelling human evidence of a direct environmental contribution to neurological disease progression. These results underscore the urgent public health implications of airborne microplastic pollution and highlight the particular vulnerability of children, whose developing nervous and immune systems are disproportionately affected by such exposures. Our work provides strong scientific justification for enhanced environmental monitoring, stricter regulatory controls on plastic emissions, and targeted interventions in high-risk settings, especially those with dense pediatric populations. Given the pervasive and persistent nature of microplastic contamination, immediate policy action is warranted to mitigate long-term neurological risks and safeguard the health of future generations.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

Yueying Liu: Conceptualization, Resource, Methodology, Visualization, Writing – original draftRuijin Xie: Conceptualization, Methodology, Visualization, Writing – original draftWenjing Zhao: Conceptualization, Methodology, Visualization, Writing – original draftHua Xu: Conceptualization, Methodology, Visualization, Writing – original draftJianrui Dou: Data curation, Validation, SoftwareXue Xiao: Data curation, Validation, SoftwareYufan Luo: Formal analysis, SoftwareHen Zhang: Formal analysis, SoftwarePeiwen Wang: Investigation, SoftwareWei Xiao: Investigation, ValidationXiao Tong: Investigation, ValidationShenjie Xu: Investigation, SoftwareDongqin Wu: Investigation, SoftwareXianhui Deng: Investigation, SoftwareYu Wu: Conceptualization, Methodology, Resources, Supervision, Writing – review & editingChenyu Sun: Conceptualization, Funding acquisition, Resources, Supervision, Writing – review & editingShudong Hu: Methodology, Resources, Supervision, Writing – review & editing.

Funding

This research was supported by the National Natural Science Foundation of China (Grant No. 82371462), the Provincial Subsidy Funds for Pairing Assistance (Grant No. DHBFZ202505), Jiangsu Traditional Chinese Medicine (PDJH2024027), and the Medical Key Strategic Project of the Wuxi Health Commission to Yueying Liu. Additional support was provided by the Qing Lan Project of Jiangsu (JS2023-27), the Traditional Chinese Medicine Science and Technology Development Plan of Zhejiang Province (2024ZL140), the Hangzhou Medical and Health Science and Technology Project (B20252321), the Yangzhou Polytechnic College Research Project (2025XJ17), the Yangzhou Basic Research Program (YZ2025162), and the Social Science Research Project on Yangzhou’s Child-Friendly Initiative (2025YFE-057) awarded to Ruijin Xie. Yu Wu receives funding from the University Qinglan Project of Jiangsu Province and the Fundamental Research Funds for the Central Universities (JUSRP221024, JUSRP202406004), while Shudong Hu receives funding from Scientific Research Project of Jiangsu Provincial Health Commission (K2025041).

Data availability

To ensure reproducibility, our Python and R analysis scripts are publicly available on GitHub (https://github.com/PediatricLab-Jiangnan/Airborne-singlecell-bulk). Raw data from human plasma untargeted metabolomics and targeted lipidomics analyses have been deposited in the Genome Sequence Archive under accession numbers OMIX009167 and OMIX014782, respectively.

Declarations

Competing interests

The authors declare no competing interests.