Abstract Background Subway systems reduce traffic congestion, air pollution, and carbon dioxide emissions in cities but the impacts of subway air pollution on the health of subway users remain obscure. We conducted a randomized controlled trial involving 83 healthy adults, with 80 included in the final analysis, randomly grouped to spend 2 h daily for 5 consecutive days either in an office or on a subway platform. The fine (PM[2.5]) and thoracic (PM[10]) particles concentrations, temperature, and humidity were monitored. Measurements of health parameters were assessed, including lung function and levels of fractional exhaled nitric oxide (FeNO), inflammatory and oxidative stress biomarkers, and metabolites in serum. Results The subway platform exhibited significantly high pollutant levels, with mean PM[2.5] and PM[10] concentrations of 193.4 ± 39.4 µg/m^3 and 311.5 ± 64.3 µg/m^3 respectively. After the 5-day subway exposure, significant declines were observed in lung-function index values, including forced expiratory volume in the first second (FEV1)/forced vital capacity (FVC), maximal voluntary ventilation (MVV) and peak expiratory flow rate (PEFR) as well as serum levels of glutathione peroxidase (GPX)-1 (p < 0.05). Conversely, somatosensory symptom scores, FeNO levels, and serum levels of tumor necrosis factor (TNF)-α and interleukin (IL)-8 were strongly elevated (p < 0.05). Results indicated increased arsenic and cobalt and decreased selenium in urine after the subway exposure (p < 0.05). Finally, the subway exposure was associated with disruptions in seven metabolic pathways and nine metabolites, particularly the depletion of L-cysteine, pretyrosine and O-acetyl-L-serine. Conclusions This study provides the first evidence that repeated exposure to subway airborne particles is associated with reduced lung function and increased respiratory and systemic inflammation in healthy adults. Our results underscore the need to develop strategies to mitigate exposure risks, ultimately protecting public health in urban environments. Supplementary Information The online version contains supplementary material available at 10.1186/s12989-025-00638-5. Keywords: Subway, Particulate matter, Health impacts, Inflammation, Oxidative stress Introduction Subway systems, with their safety, convenience, efficiency, and high transportation capacity, have become an important mode of transportation in many megacities worldwide [[54]1]. However, the generation of airborne microparticles by wheels and brakes, combined with the closed environment, poor ventilation, and high crowd density of subways readily leads to accumulation of pollutants [[55]2]. Indeed, levels of particulate matter (PM) are significantly higher in subway platforms than in the outdoor air, also exceeding those in train carriages and station halls [[56]3]. In one study, exposure to fine PM (PM[2.5]) was three to eight times greater in the London Underground than in other transport microenvironments such as cycling, bus, and car [[57]4]. Measurements from subway platforms in cities such as New York and Beijing have indicated that PM[2.5] levels are significantly higher than the updated air-quality guidelines from the World Health Organization (WHO) [[58]5, [59]6]. Levels of PM[10] in the Seoul subway reach 359 µg/m³, also surpassing air-quality standards [[60]7]. As the primary pollutant in subways, PM (PM[2.5] and PM[10] in particular) may have a significant impact on the health of riders [[61]8]. As noted above, the main source of PM in subways is friction between wheels, tracks, and brakes during operation. Trains in motion produce huge amounts of high-temperature metals via mechanical friction that are rapidly oxidized [[62]9]. Consequently, metals such as magnetite, hematite, and chromium are the predominant components in subway air [[63]10]; these are very different from the types of PM found in the atmosphere. Previous studies have shown that subway PM exhibits higher oxidative potential than urban ambient PM, primarily due to its high content of redox-active transition metals like iron, which can catalyze ROS generation and induce greater DNA damage in vitro [[64]11]. Thus, the toxicity of subway PM may be greater than that of atmospheric PM. Large numbers of people, both young and old, are exposed to such pollutants regularly [[65]12]. Intratracheal instillation of subway PM[2.5] in mice results in inflammatory cell infiltration into lung tissues, accompanied by significant elevations of inflammatory cytokines such as interleukin (IL)-6, IL-13, and tumor necrosis factor (TNF)-α in both blood and bronchoalveolar lavage fluid [[66]13]. That is, it triggers systemic inflammatory responses and causes inflammatory damage to the respiratory system. Additionally, subway environmental dust can elicit oxidative stress responses in animals or cells by generating reactive oxygen species (ROS) [[67]14]. Cellular experiments have demonstrated that when human lung cells (A549) are exposed to subway particles, genotoxicity is approximately eight times higher than that of street air PM, and the likelihood of inducing oxidative stress in lung cells is four times greater [[68]11], indicating toxicity that significantly exceeds that of atmospheric PM. Subway PM affects the health of staff such as subway drivers, conductors, and cleaners; the research indicates that platform workers with high exposure tend to have higher levels of risk markers for cardiovascular diseases than ticket sellers and train drivers with medium to low exposure [[69]15]. However, only a few epidemiological studies have explored the health effects on commuters. Acute exposure to subway PM for 2 h can cause lung and systemic inflammatory responses, while no significant clinical changes in the respiratory system was found [[70]16, [71]17]. After 4 h, compared with the facemask-wearing group, the cardiovascular parameters such as heart rate and heart rate variability of the healthy male volunteers in the non-facemask-wearing group were significantly reduced, while urinary levels of the oxidative stress biomarker 8-hydroxy-2’-deoxyguanosine (8-OHdG) were notably increased [[72]18]. However, most commuters use the subway every day and thus are exposed repeatedly and for long durations to high levels of pollutants. For instance, Canadian commuters spend up to 70 min daily in the subway environment [[73]19], while in Chinese megacities such as Shanghai, daily subway times can exceed 120 min [[74]20]. This highlights the need for research on the cumulative health effects of prolonged and repeated exposure to subway PM. Therefore, we conducted a subacute randomized controlled trial with healthy adult participants to assess the impacts of repeated exposure to subway PM on pulmonary function, inflammation, and oxidative stress markers. We also performed untargeted metabolomics analysis of serum samples to identify key differential metabolites, and explored potential correlations between these metabolites and indicators of respiratory dysfunction, inflammation, and oxidative stress. Results Descriptive statistics A total of 80 participants were analyzed in the study and randomly assigned to the control group in an office (n = 39) or the exposure group in a subway platform (n = 41). The average ages, expressed as mean ± standard deviation (SD), were 23.8 ± 1.0 y for the office control group and 24.2 ± 1.2 y for the subway exposure group. The proportion of female participants was 82.1% (32/39) in the office group and 80.5% (33/41) in the subway group. No significant differences in age, BMI, and sex distribution were observed between the two groups (Table [75]1). Table 1. Characteristics of participants in the office control and the subway exposure groups Characteristic Office (n = 39) Subway (n = 41) p-Value Age (y) 23.8 ± 1.0 24.2 ± 1.2 0.077 Sex Female 32 (82.1%) 33 (80.5%) 0.858 Male 7 (17.9%) 8 (19.5%) BMI (kg/m^2) 21.4 ± 2.6 22.1 ± 3.0 0.269 [76]Open in a new tab Note: BMI, body mass index; Data are n (%), means ± SD During 16:30 − 18:30 from Monday to Friday, the mean concentrations of PM[2.5] and PM[10] were measured in both the office and subway platform. In the office, the mean PM[2.5] concentration was 5.7 ± 1.6 µg/m^3, while the mean PM[10] concentration was 13.7 ± 4.8 µg/m^3. In contrast, the subway platform exhibited significantly higher pollutant levels, with mean PM[2.5] and PM[10] concentrations of 193.4 ± 39.4 µg/m^3 and 311.5 ± 64.3 µg/m^3 respectively. Additionally, the relative humidity on the subway platform was notably higher compared to that in the office. Detailed exposure data are shown in Table [77]2. Table 2. Exposure conditions of the office control and the subway exposure groups Variable Group Mean SD Min P25 P50 P75 Max PM[2.5] (µg/m^3) Office 5.7 1.6 2.0 5.0 6.0 6.0 12.0 Subway 193.4 39.4 98.0 166.8 192.0 219.0 324.0 PM[10] (µg/m^3) Office 13.7 4.8 3.0 10.0 13.0 16.0 33.0 Subway 311. 5 64.3 142.0 266.0 309.0 357.3 527.0 Temperature (℃) Office 26.4 0.9 21.0 26.0 26.5 27.0 28.6 Subway 23.2 0.6 20.4 22.8 23.2 23.5 29.6 Relative humidity (%) Office 51.9 7.3 36.0 46.5 51.4 58.1 67.4 Subway 79. 5 3.1 59.1 78.1 79.9 81.7 87.2 [78]Open in a new tab Note: PM[2.5] and PM[10], particulate mass of particles with diameter < 2.5 and < 10 μm, respectively; min, minimum; max, maximum; P25, 25th percentile; P50, 50th percentile; P75, 75th percentile; SD, standard deviation Somatosensory symptoms To evaluate somatosensory symptoms associated with the exposure on the subway platform, we scored the self-perceived severity of nasal, ocular, respiratory, and head symptoms (Table [79]S1). In the office control group, there were no significant differences in total symptom scores between before and after the exposure. However, in the subway exposure group, the total symptom scores were significantly increased after exposure compared to before exposure from Monday to Friday (p < 0.001). These findings suggested that exposure to the subway platform was associated with a notable rise in self-perceived somatosensory symptoms. Pulmonary function and exhaled nitric oxide As shown in Table [80]3, no significant differences were observed in baseline pulmonary function indices or FeNO between the office and subway group. There were no significant changes in pulmonary function and FeNO between before and after exposure in the office control group, with the exception of a significant increase in VC (3.04 L vs. 3.14 L; p < 0.05). However, the multiple pulmonary function indices, including FEV1 (2.88 L vs. 2.75 L; p < 0.001), FEV1/FVC (88.98% vs. 87.40%; p < 0.01), MVV (112.25 L vs. 103.54 L; p < 0.001), MMEF (4.17 L/s vs. 4.00 L/s; p < 0.001), and PEFR (6.20 L/s vs. 5.45 L/s; p < 0.001) showed significant declines after 5 days exposure, compared to baseline measurements in the subway exposure group. Additionally, FeNO (9.00 ppb vs. 11.00 ppb; p < 0.05) was significantly elevated after 5 days of exposure on the subway platform compared to the pre-exposure value. These findings indicated that prolonged exposure to the subway platform may adversely affect both pulmonary function and airway inflammation. Table 3. The lung function and fractional exhaled nitric oxide of participants before and after the subway exposure Office (n = 39) Subway (n = 41) Lung function VC (L) Before 3.04 (2.77, 3.65) 3.31 (2.97, 3.70) After 3.14 (2.81, 3.62)^* 3.24 (2.83, 3.83) Change 0.04 (-0.08, 0.21) -0.03 (-0.15, 0.08)^§ FEV1 (L) Before 2.82 (2.55, 3.12) 2.88 (2.71, 3.35) After 2.72 (2.50, 3.02) 2.75 (2.44, 3.19)^††† Change -0.06 (-0.12, 0.06) -0.11 (-0.30, 0.03)^§ FEV1/FVC (%) Before 88.85 (84.08, 93.16) 88.95 (84.68, 93.18) After 91.92 (86.39, 95.43) 87.40 (79.68, 90.45)^††‡‡ Change 1.20 (-1.30, 4.24) -2.67 (-8.47, 1.69)^§§§ MVV (L) Before 105.36 (89.96, 115.79) 112.25 (96.63, 125.91) After 103.11 (90.94, 112.76) 103.54 (86.56, 117.83)^††† Change -0.35 (-6.87, 4.76) -8.71 (-15.54, -3.02)^§§ MMEF (L/s) Before 3.95 (3.67, 4.77) 4.17 (3.59, 4.79) After 4.18 (3.72, 4.61) 4.00 (3.24, 4.42)^††† Change -0.06 (-0.24, 0.26) -0.33 (-0.58, 0.00)^§§ PEFR (L/s) Before 5.85 (5.58, 6.68) 6.20 (5.20, 7.30) After 5.95 (5.25, 6.55) 5.45 (4.50, 6.23)^††† Change -0.10 (-0.63, 0.58) -0.75 (-1.40, -0.13)^§§§ FeNO (ppb) Before 8.00 (6.00, 14.00) 9.00 (6.00, 11.00) After 8.00 (5.00, 11.50) 11.00 (8.00, 13.00)^†‡ Change -1.00 (-3.00, 2.00) 2.00 (-1.00, 5.00)^§ [81]Open in a new tab Note: VC, vital capacity; FEV1, forced expiratory volume in the first second; FEV1/FVC, the ratio of FEV1 to forced vital capacity; MVV, maximal voluntary ventilation; MMEF, maximal mid-expiratory flow curve; PEFR, peak expiratory flow rate; FeNO, fractional exhaled nitric oxide. Compared to the office before, ^*p < 0.05; Compared to the subway before, ^†p < 0.05, ^††p < 0.01, ^†††p < 0.001; Compared to the office after, ^‡p < 0.05, ^‡‡p < 0.01; Compared to the change in the office, ^§p < 0.05, ^§§p < 0.01, ^§§§p < 0.001. Data are presented as median (P25, P75), P25, 25th percentile; P75, 75th percentile Cell species, inflammatory and oxidative stress markers No significant changes were observed in the percentages of monocytes, neutrophils, lymphocytes, eosinophils, and basophils in peripheral blood of participants after 5 days of exposure in either the office control or subway exposure groups (Table [82]4). Similarly, there were no differences in serum biomarker indicators such as TNF-α, IL-8, and GPX1 between before-exposure and after-exposure in the office control group. However, serum concentrations of the inflammatory cytokine TNF-α (0.35 pg/mL vs. 0.36 pg/mL; p < 0.05) and IL-8 (2.02 pg/mL vs. 2.64 pg/mL; p < 0.05) were increased, alongside GPX1 (118.43 pg/mL vs. 100.26 pg/mL; p < 0.001) level in serum was decreased after 5 days of exposure, compared to before exposure in the subway platform. Additionally, urinary levels of 8-OHdG, a marker of oxidative stress, showed no significant differences between before-exposure and after-exposure in either the office control or subway exposure groups (Table [83]4). These findings suggested that exposure to the subway platform may induce systemic inflammation and reduce antioxidant capacity, without significantly affecting immune cell distribution or systemic oxidative stress as measured by urinary 8-OHdG. Table 4. Cell species, inflammatory and oxidative stress markers of participants before and after the subway exposure Office (n = 39) Subway (n = 41) Blood differential (%) Monocytes Before 6.30 (5.85, 7.70) 6.40 (5.40, 7.30) After 6.50 (5.90, 7.85) 6.50 (5.50, 7.40) Change 0.40 (-0.50, 0.95) -0.10 (-0.50, 0.60) Neutrophils Before 57.40 (51.45, 61.55) 60.20 (54.00, 63.90) After 54.80 (53.60, 61.60) 58.80 (56.10, 63.30) Change -1.00 (-4.05, 3.45) 0.70 (-3.70, 5.70) Lymphocytes Before 32.30 (28.65, 37.20) 30.80 (28.20, 35.60) After 33.70 (29.70, 37.90) 31.90 (28.90, 35.50) Change 0.80 (-3.10, 3.15) 0.50 (-3.50, 3.20) Eosinophils Before 1.60 (1.20, 2.25) 2.10 (1.00, 2.80) After 1.60 (1.25, 2.30) 1.90 (1.20, 2.80) Change 0.10 (-0.25, 0.40) 0.10 (-0.30, 0.40) Basophils Before 0.40 (0.25, 0.65) 0.40 (0.20, 0.60) After 0.40 (0.30, 0.50) 0.40 (0.30, 0.50) Change 0.00 (-0.15, 0.10) 0.00 (-0.10, 0.10) Biomarkers TNF-α (pg/mL) Before 0.38 (0.29, 0.46) 0.35 (0.26, 0.42) After 0.30 (0.22, 0.42) 0.36 (0.29, 0.43)^* Change 0.00 (-0.12, 0.05) 0.02 (-0.03, 0.08)^† IL-8 (pg/mL) Before 2.46 (1.26, 4.77) 2.02 (0.22, 5.35) After 3.01 (1.04, 6.72) 2.64 (0.90, 5.07)^* Change -0.53 (-1.66, 1.21) 0.62 (-0.30, 2.42) 8-OHdG[urine] (µg/g) Before 308.84 (194.72, 452.37) 244.54 (128.19, 427.57) After 331.79 (194.80, 401.31) 247.53 (191.22, 354.86) Change -59.10 (-180.87, 215.87) 0.24 (-165.26, 107.87) GPX1 (pg/mL) Before 113.79 (84.87, 137.87) 118.43 (100.17, 138.11) After 116.21 (90.96, 136.31) 100.26 (85.57, 128.63)^** Change 5.37 (-6.31, 23.58) -16.53 (-50.21, 3.16)^†† [84]Open in a new tab Note: TNF-α, tumour necrosis factor-alpha; IL-8, Interleukin-8; 8-OHdG[urine], 8-hydroxy-2’-deoxyguanosine in urine; GPX1, glutathione peroxidase 1. Compared to the subway before, ^*p < 0.05, ^**p < 0.01; Compared to the change in the office, ^†p < 0.05, ^††p < 0.01. Data are presented as median (P25, P75) Urine metals In this study, we analyzed the concentrations of 19 urinary metals (µg/g creatinine), as shown in Table [85]S2. In the office control group, only Mo levels (31.94 µg/g vs. 38.37 µg/g; p < 0.01) showed a significant increase after 5 days of exposure compared to baseline measurements. In the subway exposure group, urinary levels of As (9.71 µg/g vs. 11.13 µg/g; p < 0.01) and Co (0.29 µg/g vs. 0.36 µg/g; p < 0.01) were significantly elevated after 5 days of exposure compared to before exposure levels. Conversely, the urinary concentration of Se (22.44 µg/g vs. 20.29 µg/g; p < 0.01) was significantly reduced after 5 days of exposure in the subway group compared to baseline. These findings indicated that exposure to the subway platform may influence the body’s retention or excretion of specific metals, potentially reflecting differences in environmental exposure or metabolic responses. Untargeted serum metabolome analysis In this study, we identified a total of 2,809 and 1,863 metabolite features from positive and negative ions, respectively, across all 160 samples from both office and subway groups before and after exposure. The primary metabolites detected were amino acids and their metabolites (18.75%) (Fig. [86]S1). The PCA plot revealed that principal components PC1 and PC2 explained 16.8% and 5.7% of the variation in the office, and 13% and 6.51% of the variation in the subway platform, respectively. The serum metabolite profiles showed partial separation between before and after exposure timepoints in both the office and subway groups (Fig. [87]S2a and b), though clustering was not fully distinct. OPLS-DA further demonstrated that the metabolite profiles of participants before and after exposure were distributed in distinct regions for both the office and subway groups (Fig. [88]S2c and e). The goodness-of-fit values and predictive ability metrics (office R^2X = 0.291, R^2Y = 0.988, Q^2 = 0.729, and p < 0.05; subway platform R^2X = 0.229, R^2Y = 0.989, Q^2 = 0.775, and p < 0.05) confirmed that the OPLS-DA model possessed a satisfactory fit with strong predictive power (Fig. [89]S2d and f). Furthermore, the PCA plots of all groups also revealed that the metabolites in the subway group and the office group caused different clusters after exposure, and it was further proved by OPLS-DA that they were distributed in different regions, with good fitting effect and predictive ability (Fig. [90]S2g-i). The results revealed that 710 metabolites significantly increased, and 317 metabolites significantly decreased after exposure in the office (Fig. [91]S3a). The top 20 differential metabolites primarily exhibited a downward trend (Fig. [92]S3c). In the subway group, 613 metabolites significantly increased, while 377 metabolites significantly decreased after exposure (Fig. [93]S3b). Among these differential metabolites, the top 20 mainly showed an upward trend (Fig. [94]S3d). The Venn diagram result revealed 279 key differential metabolites significantly related to environmental exposure to the subway platform (Fig. [95]S4a). K-means clustering further identified 107 metabolites with significantly altered expression levels after subway exposure compared to subway before and office after group. These metabolites belonged to classes 1, 2, 3, 5, and 6 (Fig. [96]S4b). KEGG results indicated that the 107 identified metabolites were involved in key metabolic pathways, including sulfur metabolism, sulfur relay system, glutathione metabolism, mannose type O-glycan biosynthesis, biosynthesis of amino acids, pantothenate and CoA biosynthesis, and ascorbate and aldarate metabolism (Fig. [97]1a). Among these enriched metabolic pathways, 9 metabolites were strongly associated with subway exposure. Specifically, D-arabinono-1,4-lactone, dehydroascorbic acid, 3-aminopropionaldehyde, dimethyl sulfoxide, D-ribitol-5-phosphate and L-threonine were significantly up-regulated and pretyrosine, L-cysteine, and O-acetyl-L-serine were notably down-regulated after exposure in the subway platform (Fig. [98]1b). These findings suggested that exposure to the subway platform environment may disrupt specific metabolic pathways, particularly those related to oxidative stress, amino acid metabolism, and sulfur metabolism. Fig. 1. [99]Fig. 1 [100]Open in a new tab Untargeted metabolome-wide association study (MWAS) of serum metabolomics. (a) KEGG enrichment analysis of differential metabolites. The x-axis indicates pathway-specific rich factor values, while the y-axis displays metabolic pathways ranked by ascending p-value. Dot coloration reflects statistical significance (red gradient indicating lower p-values), with dot size proportional to the number of enriched metabolites per pathway. (b) Hierarchical clustering heatmap of differential metabolites. Rows represent annotated metabolites and columns denote sample groups. Color intensity reflects z-score normalized relative abundance (red: elevated levels; green: reduced levels) Correlations between the key differential metabolites and the outcome indicators Spearman correlation analysis revealed several significant associations between metabolites and physiological indicators. L-cysteine showed positive correlations with VC (r = 0.19, p < 0.05), MVV (r = 0.32, p < 0.001), and PEFR (r = 0.28, p < 0.01), and pretyrosine was positively correlated with both MVV (r = 0.19, p < 0.05) and PEFR (r = 0.19, p < 0.05) (Fig. [101]2a). Fig. 2. [102]Fig. 2 [103]Open in a new tab Correlation between key differential metabolism and outcome indicators related to PM exposure in subway platform. (a) Spearman correlation matrix between key differential metabolites (dimethyl sulfoxide; O-acetyl-L-serine; L-cysteine; dehydroascorbic acid; D-ribitol-5-phosphate; pretyrosine; L-threonine; D-arabinono-1,4-lactone; 3-aminopropionaldehyde) and pulmonary function parameters (VC, FEV1, FEV1/FVC, MVV, MMEF, PEFR), *p < 0.05, **p < 0.01, ***p < 0.001. (b-f) The relationship between the key differential metabolites and FeNO levels, inflammatory markers (TNF-α, IL-8), andoxidative stress indicators (8-OHdG, GPX1). Statistically significant correlations (p < 0.05) are highlighted in red In terms of airway and systemic inflammation, D-ribonol-5-phosphate (r = 0.24, p < 0.01), dehydroascorbic acid (r = 0.18, p < 0.05), and O-acetyl-L-serine (r=-0.25, p < 0.01) were significantly correlated with FeNO (Fig. [104]2b and Table [105]S3). Additionally, O-acetyl-L-serine was negatively correlated with the systemic inflammatory factor TNF-α (r=-0.34, p < 0.001) (Fig. [106]2c and Table [107]S3). Regarding oxidative stress, pretyrosine (r = 0.19, p < 0.05) and dimethyl sulfoxide (r=-0.23, p < 0.05) were correlated with urinary 8-OHdG, a marker of oxidative DNA damage (Fig. [108]2e and Table [109]S3). Furthermore, pretyrosine (r = 0.19, p < 0.05) and L-cysteine (r = 0.18, p < 0.05) were positively correlated with the antioxidant enzyme GPX1, suggesting their involvement in antioxidant defense mechanisms (Fig. [110]2f and Table [111]S3). These correlations highlighted the complex interplay between metabolic changes, pulmonary function, inflammation, and oxidative stress in response to environmental exposure. Discussion In all, 80 participants were analyzed and randomly assigned to either the office or subway group. The concentrations of PM[2.5] and PM[10] were approximately 34 times and 23 times higher on the subway platform than in the office. The increase in humidity led to an increase in the adsorptive mass concentration of PM indoors [[112]21], resulting in prolonged exposure to highly concentrated, humid air pollution. The subway group showed a decline in lung function, airway inflammation associated with increased FeNO, TNF-α, and IL-8 levels, and a marked decrease in levels of the antioxidant enzyme GPX1. This is the first randomized controlled trial to investigate the association between repeated exposure to subway air and its health effects, providing valuable insights into the physiological impacts of such exposure. The subway environment is relatively enclosed and often crowded with a high number of passengers, particularly during peak commuting hours. This combination exacerbates air pollution levels, creating a challenging environment with significant exposure to pollutants. PM levels are significantly higher in subways than in outdoor air [[113]22], which may pose potential health risks to commuters and staff. In a previous epidemiological study, exposure to subway stations in Stockholm (with an average PM[2.5] concentration of 71–77 µg/m^3) for only 2 h resulted in a significant increase in T-cells and a systemic inflammatory marker in the blood of 20 healthy volunteers [[114]17]; it also led to elevated levels of activated T-cells in the bronchoalveolar lavage fluid of 16 asthmatic volunteers [[115]16]. In another study, a 2-h exposure of 28 healthy volunteers to PM[2.5] (with an average concentration of 293.6 ± 65.7 µg/m^3) on a New York City subway platform did not cause measurable cardiopulmonary effects apart from reductions in heart rate variability and increases in self‑reported symptoms [[116]23]. Building on these findings, our study provides a more in-depth analysis by examining the effects of repeated exposure to high concentrations of PM (PM[2.5]: 193.39 µg/m³, PM[10]: 311.45 µg/m³) on a subway platform during peak commuting hours over five consecutive working days. We observed compromised lung function, characterized by significant decreases in FEV1, FEV1/FVC, PEFR, and MMEF. These indicators are related to airway obstruction and small-airway injury [[117]24, [118]25]. Intensification of inflammatory reactions and increased airway exudation can further obstruct the airways, leading to changes in lung-function parameters [[119]26]. Collectively, these findings imply that repeated exposure to subway PM may be associated with early physiological alterations and potential risks of pathological changes in the small airways of healthy adults, potentially mediated by inflammatory responses. TNF-α, a critical early pro-inflammatory factor, plays a pivotal role in the inflammatory response, and can induce production of IL-8 [[120]27]. These in turn lead to elevated FeNO levels, which are a reliable biomarker of airway inflammation [[121]28, [122]29]. We also found a significant decline in serum levels of the antioxidant enzyme GPX1 (with selenocysteine as its active center), which catalyzes the reduction of H[2]O[2] (and organic hydroperoxides) into water (and corresponding alcohols), participating in glutathione-based antioxidant defense [[123]30]. Consistent with this, we found a significant reduction in urinary levels of Se, potentially reflecting the substantial consumption of GPX1 during oxidative stress. However, 8-OHdG levels, a key biomarker produced by oxidative damage to DNA [[124]31], did not substantially change. We hypothesize that it takes more than 5 days, the duration of exposure that we tested, to trigger DNA damage via oxidative stress. Furthermore, the levels of metals As and Co in urine significantly increased after exposure to the subway, although subway PM contains elevated levels of metals such as Fe and Cr [[125]10], their urinary concentrations showed limited changes. This discrepancy may reflect differences in metal bioavailability, as previous studies indicate that while As, Ba, Co and Pb have high bioaccessibility, Fe and Cr demonstrate relatively low solubility in biological systems [[126]32]. These findings highlight the importance of considering metal speciation and bioaccessibility when evaluating the relationship between environmental exposure and internal dose. Untargeted metabolomics analyses of serum samples identified key metabolic changes associated with exposure to the subway platform. Specifically, significant decreases were observed in the levels of pretyrosine, O-acetyl-L-serine, and L-cysteine. The consumption of pretyrosine interferes with the metabolism of glutathione and aggravates oxidative stress [[127]33]. O-acetyl-L-serine is an essential substrate for L-cysteine synthesis [[128]34]. Besides its inherent, potent antioxidant properties, L-cysteine serves as a precursor, participating in antioxidant defense systems such as glutathione metabolism, which protect intracellular macromolecules from oxidative stress damage [[129]35, [130]36]. L-cysteine is vital for protecting cells from free radical attack and maintaining redox balance. The substantial decrease in pretyrosine, O-acetyl-L-serine, and L-cysteine implies that exposure to the subway platform disrupts redox balance, due to overproduction of ROS and depletion of key antioxidant metabolites. These results highlight the potential health risks associated with prolonged exposure to subway air pollution. We also found that L-cysteine and pretyrosine were associated with decreased lung function and reduced levels of the antioxidant enzyme GPX1. These metabolites serve as vital substrates, synergistically stimulating glutathione metabolism and effectively inhibiting oxidative stress [[131]33]. For instance, exposing bronchial epithelial cells to silica nanoparticles significantly disrupts glutathione metabolism, involving key metabolites such as glutathione and cysteine, leading to oxidative stress and inevitable damage to the respiratory system [[132]37]. Meanwhile, we found that O-acetyl-L-serine was negatively correlated with inflammation indicators such as FeNO and TNF-α, while dehydroascorbic acid showed a positive correlation with FeNO. Animal models have indicated that increasing L-cysteine levels can effectively reduce serum levels of the inflammatory biomarker TNF-α [[133]38]. In addition, long-term exposure to high levels of atmospheric PM[2.5] has been linked to changes in amino acids, cofactors, and antioxidant vitamins, which are associated with markers of systemic inflammation [[134]39]. The reduction of dehydroascorbic acid to ascorbic acid largely depletes intracellular glutathione and induces outflow of its oxidized form [[135]40]. Therefore, we propose that changes in these key metabolites synergistically amplify inflammatory responses and subsequently impair lung function. These metabolic changes highlight the complex interplay between oxidative stress, inflammation, and lung health in response to exposure to environmental contaminants such as subway air pollution. This study had many strengths. It was a randomized controlled trial, enhancing the reliability of causal inferences. We considered a larger sample size and a longer intervention period than in previous studies, providing more robust data and insights. We also conducted untargeted metabolomics analyses of serum, allowing identification of subtle and sensitive biomarkers and offering novel insights into the mechanisms underlying the health effects of subway PM exposure. We also performed pathway analyses to uncover the potential associated metabolic pathways involved. However, there were also some limitations. Despite pre-study screening, baseline PCA clustering suggested inherent differences between the subway and office groups (e.g., long-term subway habits), though within-group (before-after) comparisons inherently control for such variation. Additionally, serum and urine samples were only collected at baseline and the day after the trial. More frequent sampling could have provided a more comprehensive understanding of various indicators and metabolomic changes during exposure, although it would also have increased the physical invasiveness for participants and potentially reduced their compliance. In addition, because all study participants were healthy students, the generalizability of the findings is somewhat limited. The higher proportion of female participants may reflect greater exclusion of males due to smoking criteria, which further limits population generalizability. Finally, optical PM measurements may underestimate subway PM mass concentrations due to higher particle density such as metal-rich aerosols, though this does not affect the relative comparison between subway and office environments [[136]41]. Conclusion In conclusion, we found that repeated exposure to high concentrations of PM on a subway platform led to significant health impacts associated with inflammation and oxidative stress. We found evidence of impaired lung function, consistent with elevated levels of inflammation indicators including FeNO, TNF-α, and IL-8, as well as reduced levels of the antioxidant enzyme GPX1. Metabolomics analysis of serum further revealed that exposure was associated with metabolic disturbances, particularly depletion of L-cysteine, pretyrosine, and O-acetyl-L-serine, which play critical roles in maintaining redox balance and mitigating oxidative stress. Our results provide robust evidence for the health impacts of subway platform exposure and identify sensitive and effective biomarkers, along with potential biological mechanisms underlying these effects. They also underscore the urgent need for further research into the long-term health consequences of subway air pollution and the development of strategies to mitigate exposure risks, ultimately protecting public health in urban environments. Methods Study design and participants We conducted a randomized controlled trial on a subway platform of Line 2 in Shenyang, China, from July 3 to July 15, 2023. Initially, 94 healthy volunteers were recruited from China Medical University (Fig. [137]S5). The participants had no history of chronic cardiopulmonary disease, had not used the subway in the 2 weeks prior to the study, and had no respiratory infections (such as colds, COVID-19, or bronchitis) within 4 weeks prior to the study; they also did not use any anti-inflammatory drug in the 4 weeks preceding the study. We measured their height and weight to calculate body mass index (BMI), and collected demographic characteristics during screening visits, including age and sex, and excluded participants with a history of smoking, cardiopulmonary diseases, or allergic conditions. Pulmonary function tests were performed as recommended by the American Thoracic Society/European Respiratory Society (ATS/ERS) [[138]42] using a portable lung function detector (BIOBASE K-LFT-I, China). In all, 14 participants were excluded either because their pulmonary function test results were below the normal range (i.e., the ratio of forced expiratory volume in the first second [FEV1] to forced vital capacity [FVC] < 0.65 or the predicted FEV1 < 75%) or they refused to participate in subsequent trials. Eventually, 80 participants were included in analysis. Participants were randomly divided into an exposure group and a control group, and completed 2 h of daily exposure, either at rest on a subway platform or in an office, for 5 consecutive days, from Monday to Friday, during the peak commuting hours of 16:30–18:30. Participants were requested to rest without strenuous exercise during the exposure period, and participants in the subway group were asked to wear masks on their way to the subway platform to avoid interference from exposure to other traffic particles. All participants underwent a physical examination on Sunday before exposure on Monday, and then within 24 h of their last exposure on Friday. Blood and urine samples were collected for follow-up analysis. The study protocol was approved by the review committee of the School of Public Health, China Medical University (approval No. 202060, Shenyang, China) and was conducted in accordance with the 1964 Declaration of Helsinki. All participants provided signed informed consent before any study procedure. During the exposure period from 16:30 to 18:30 every day, a 5-in-1 PM detector (DT9881M, CEM Technology Co., LTD. Shenzhen, China) was used to monitor the airborne concentrations of PM[2.5] and PM[10], as well as the temperature and humidity, both in the office and on the subway platform. The instrument operates with a 21 s sampling duration followed by a 2 s standby interval (display resolution: 1 µg/m^3), executing this duty cycle continuously throughout the monitoring period from 16:30 to 18:30, totaling two hours of measurement. The subway monitoring point was 1 m away from the train track and the monitoring height was 1.5 m. The office monitoring height was 1 m, corresponding to the seated breathing zone of participants in the office environment. A stability test was carried out for 15 min before each monitoring session. After the instrument had stabilized, a staff member measured the mass concentration at the front, middle, and rear of the stopped train on the platform for approximately 40 min at each location (total ~ 2 h). The average concentration of these three points was used as the final exposure concentration. Health measurements and sample collection Subjective symptom assessment To measure the subjective somatic symptoms caused by exposure, we referred to the Swedish Performance Assessment System questionnaire [[139]43], which scores self-perceived severity of nasal, ocular, respiratory, and head symptoms on a scale of 0–5 (Table [140]S4) [[141]44]. A higher score indicates more severe symptoms and 0 indicates no symptoms. The questionnaires were completed by the participants before and immediately after exposure each day. Pulmonary function and exhaled nitric oxide test Lung-function assessments included vital capacity (VC), FEV1, FEV1/FVC, maximal voluntary ventilation (MVV), maximal mid-expiratory flow (MMEF), and peak expiratory flow rate (PEFR), and were measured using a portable lung-function detector (BIOBASE K-LFT-I, China). The levels of fractional exhaled nitric oxide (FeNO) were measured using a nitric oxide detector (WLD801, China). All participants were tested for these indicators before exposure and within 24 h after their last exposure. All participants were asked to avoid strenuous exercise; to avoid nitrate-rich foods including salted fish, cured meat, shrimp sauce, and other preserved meat; and to try to avoid nitrate-containing vegetables such as beets, celery, spinach, or leftover vegetables before the test to ensure accurate FeNO measurements. Blood sampling All participants had 10 ml blood samples taken during their physical examination on Sunday, and within 24 h after their last exposure on Friday. All of the blood samples were taken at the same time of day. A 3 ml aliquot was immediately used for five-class blood-cell analysis (Siemens, Germany) to determine percentages of monocytes, neutrophils, lymphocytes, eosinophils, and basophils. The remaining 7 ml blood samples were centrifuged at 4℃ for 15 min at 1500 ×g; serum samples were extracted and stored at -80℃ for subsequent analysis. The expression levels in serum samples of the inflammatory cytokine TNF-α (R&D Systems Inc, USA), IL-8 (Absin, shanghai, China), and the oxidative stress marker glutathione peroxidase 1 (GPX1) (Novus Biologicals, USA) were detected using an ELISA kit. Urine sampling All participants provided urine samples on Sunday and within 24 h after their last exposure on Friday, with immediate storage at -80 °C for subsequent analysis. Inductively coupled plasma mass spectrometry (ICP-MS, Perkin Elmer NexION 350x, USA) was used to measure the concentrations of 19 metal elements in urine samples, including magnesium (Mg), calcium (Ca), titanium (Ti), vanadium (V), chromium (Cr), manganese (Mn), iron (Fe), nickel (Ni), cobalt (Co), copper (Cu), zinc (Zn), arsenic (As), selenium (Se), rubidium (Rb), strontium (Sr), molybdenum (Mo), cadmium (Cd), antimony (Sb), and barium (Ba), as detailed previously [[142]45]. 8-OHdG (Abcam Plc, USA) was detected using an ELISA kit. All metals and 8-OHdG were corrected against creatinine concentrations. Untargeted serum metabolome analysis Serum samples were processed for liquid chromatography-mass spectrometry (LC-MS) analysis using acetonitrile–methanol extraction and centrifugation. Ultra-high-performance liquid chromatography (Waters ACQUITY UPLC HSS T3C18, USA) was used for quantification, and an HPLC column (1.8 μm, 2.1 mm × 100 mm, Agilent Technologies, USA) was used for separation. Mobile phase A was a water solution with 0.1% formic acid, and mobile phase B was a mixture of acetonitrile and isopropanol containing 0.1% formic acid. The column temperature was 40℃, the flow rate was 0.4 mL/min, and the injection volume was 4 µL. To verify the quality of measurements, the sample extracts were mixed and prepared into quality-control samples. The same pretreatment method was applied to both control and treatment samples. Mixed quality-control samples were included in each analysis to ensure repeatability. The original data were converted into mzXML format using Proteo Wizard, and peak extraction, alignment, and retention time corrections were carried out using the XCMS program. The SVR method was used to correct the peak area, and peaks with > 50% miss rates in each group of samples were filtered out. After correcting the screened peaks, metabolites were identified by searching our own laboratory database and integrating the results with information from a public database and metDNA. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were conducted using the R program to assess metabolic changes. Differential metabolites were identified based on variable importance in projection (VIP) > 1.0 and Student’s t-test (p < 0.05), and visualized as plots. Key metabolites were screened using Venn diagrams, and K-means clustering was used to analyze their trends. Pathway enrichment analysis of differential metabolites was performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) ([143]http://www.genome.jp/kegg) database. Statistical analysis Statistical analysis was carried out using SPSS 22.0. Data normality was assessed using the Shapiro-Wilk test. The results are presented as means ± standard deviations (SD) for normally distributed data or medians with interquartile ranges (IQR) for non-normally distributed data. Wilcoxon’s nonparametric rank sum tests were used for analysis. Correlations between key differential metabolites and outcome parameters were examined using the Spearman r test, with p-values of less than 0.05 considered statistically significant. Electronic supplementary material Below is the link to the electronic supplementary material. [144]Supplementary Material 1^ (10.9MB, docx) Acknowledgements