In this pilot study, we observed subtle changes in the concentrations of oxidative stress biomarkers (mainly MDA and 8-OHdG) in EBC and urine of workers exposed to subway particles. These positive or negative changes were associated with the concentration of transition metals in PM (Cu and Zn) as well as with metals in EBC (Ba, Co, Cr and Mn) and in urine (Ba, Cu, Co, Mo, Ni, Ti and Zn). The direction of these associations was metal- and time-dependent. The associations between Cu or Zn in PM and MDAEBC generally reached statistical significance only after a delayed time of either 12 or 24 h after exposure. On the contrary, changes in metal concentrations in EBC or in urine induced rapid changes in MDAurine or 8-OHdGurine, observed mainly within the same exposure day.
Exposure concentrations and metals measured in the biological matrices
As described previously (Canu et al.
2021a), the PM metal concentrations presented in Table
2 are in line with other reports related to subway PM characterisation (Moreno et al.
2017; Mugica-Alvarez et al.
2012). The strong correlation observed between Fe–Mn suggests similar sources, probably originating from abrasion of rail/wheels (Font et al.
2019). In addition, these PM present variable oxidative properties as measured with the FOX assay. The FOX assay presents an important selectivity to Fe and Mn content of PM, possibly explaining the highest oxidative properties measured for the locomotive operators (median OP
partFOX: 941 pmol H
2O
2eq/m
3; IQR: 774–1329). Indeed, this group of workers had the highest exposure to Fe based on its fraction in PM
10 or in PM
2.5 (Canu et al.
2021a). Similar OP
partFOX values and associations with Fe have been measured for the particulate fraction in shops using metalworking fluids (Sauvain et al.
2021). The lack of association between OP
FOX and the three oxidative stress biomarkers (Supplementary material, Table
S3) suggests that the Fe concentration is well controlled in EBC and urine and cannot participate to the generation of radicals through Fenton reactions (Valko et al.
2016).
Once deposited in the lung, some elements from metal-rich subway particles might dissolve in the lung lining fluid. The water solubility in decreasing order for metals are Zn (~ 65% of total Zn
PM) > Mn (~ 25% of total Mn
PM) > Ni (~ 15% of total Ni
PM) > Cu (< 10% of total Cu
PM) (Mugica-Alvarez et al.
2012). It is thus, not surprising that we detected these four metals in addition to Ba in our EBC samples. Since Cu, Mn, and Zn in EBC were inter-correlated (Supplementary material, Figure
S1), we suspect that these elements result from the simultaneous partial dissolution of subway particles in the lung lining fluid. Except for Zn, whose EBC concentration was rather high, all other concentration of metal elements in EBC are similar as those reported in Ghio et al. (
2018) for healthy subjects. The difficulty to detect Fe in EBC in our study could result from the low solubility of subway iron-rich PM (Loxham and Nieuwenhuijsen
2019), which could be cleared from the lung through macrophage internalization and through mucociliary clearance (Ghio et al.
2018). In addition, iron homeostasis is influenced by Zn and Cu (Ghio et al.
2020) or Mn (Aguirre and Culotta
2012), which favour the iron uptake through augmented expression of metal importer and ferrireduction (reduction of Fe
3+ to Fe
2+) (Deng et al.
2009; Ghio et al.
2020). There is some suggestive evidence that oxidants like superoxide are produced following increase of ferrireduction (Ghio et al.
2020). Such a process could participate to the observed increased concentration of MDA
EBC. The delayed change in MDA
EBC concentration 12 or 24 h after exposure to Cu or Zn in PM suggests that the release of these elements in the lungs might take some time.
The soluble PM-associated elements are thought to be quickly (within 4–24 h) distributed systemically relative to those in an insoluble state (Wallenborn et al.
2007). The urinary metal content depends on this solubility in addition to other factors like the renal function or nutrition (Fréry et al.
2017). The urinary excretion half-lives of metals are around one day for Ni (28 h, ((ATSDR)
2023), 9.6 days for Ba-sulphate nanoparticles (Konduru et al.
2014), about 10 days for Zn (Poddalgoda et al.
2019), 129 months for Cr (Petersen et al.
2000) and several decades for Cd (Suwazono et al.
2009). Comparing these half-lives in our workers with chronic exposure to subway particles (i.e., 15 years in average) suggests that a steady state should be attained in the excretion of these metals in urine. The urinary metal concentrations measured in this pilot study are quite similar to a representative sample of the general population from Spain (Domingo-Relloso et al.
2019), except for Ba, which is about 60 times lower in our study. A relatively low metal exposure might explain the lack of clear relationship between metal content in PM and urinary metal concentrations. As occupational exposure to metals among our workers were low, the urinary metal concentrations might originate from other sources.
Concentrations and origin of biomarkers measured in the biological matrices
The concentrations of the different oxidative stress biomarkers quantified in EBC and urine (Table
2) are consistent with reference ranges proposed in the literature (Graille et al.
2020a,
b; Toto et al.
2022; Turcu et al.
2022).
This pilot project assumes that the presence of metals in subway particles (biomarker of external exposure) or in EBC or urine (biomarkers of internal exposure) might induce changes in the redox homeostasis of the lung or other organs. Such modifications would result in changes of oxidative stress biomarkers either in the target organ (the lung) or systemically. As hypothesised, positive associations were observed for MDA
EBC with the PM
2.5 content of Zn as well as with Co, Mn, Ba and Cr levels in EBC. This finding suggests that ROS generation and oxidative stress processes can take place in the lungs of the exposed subway workers. Once dissolved, Cr
3+ and Mn
2+ cations can change their oxidation state and act as catalyst for ROS generation through Fenton-like reactions (Forti et al.
2011; Valko et al.
2016). Ba is reported to bind to sulfhydryl groups in proteins altering the redox homeostasis in favour of an oxidative stress (Elwej et al.
2016). Cu in PM
10 is the only element whose association with MDA
EBC was negative. Whereas signs of oxidative stress are present in EBC, the redox homeostasis in the lungs of the study participants appears to be well maintained, because MDA
EBC concentrations are low compared to values reported for healthy individuals (Turcu et al.
2022). In a previous study conducted with the same participants, we observed that exposure to subway particles induced changes in the anion pattern (mainly acetate, lactate and nitrogen oxides) in EBC (Sauvain et al.
2022). We interpreted such changes as an attempt of the cell/organ to maintain redox and/or metabolic homeostasis.
We observed that MDA
urine and 8-OHdG
urine were often negatively associated with transition elements in PM or urine, with the notable exception of Cu, which presents positive coefficients (Tables
4,
5). The association between exposure to metals and oxidative stress in EBC and/or urine has seldom been observed and our results are partly in line with previous studies. Workers exposed to iron oxide nanoparticles presented increased MDA
EBC levels compared to controls, whereas no changes in urinary MDA, 8-isoprostane or 8-OHdG could be observed for these workers (Pelclova et al.
2016). Whereas this study on nanoparticle exposure did not measure the metals content in EBC, it suggests that evaluation of oxidative stress could be more informative using EBC than urine. Few studies reported negative associations between metal exposures and 8-OHdG
urine. In a randomized exposure-crossover study, levels of 8-OHdG
urine of volunteers exposed to steel mill emissions were smaller than for the same volunteers exposed to particles originating from traffic/urban emissions (Pelletier et al.
2017). A similar conclusion was reported for New York City subway workers, who presented a lower concentration of 8-OHdG compared to office workers (Grass et al.
2010). These negative associations might have multiple reasons, as the simultaneous presence of different metals in the lung and other organs might induce complex biological responses. Non-essential metals like chromium have been reported to inhibit the expression of the enzyme glycosylase 1 (OGG1), whose function is to remove adducts from DNA (Hartwig
2013). Such inhibition could reduce the excretion of 8-OHdG
urine. On the other hand, increased concentrations of Zn, Cu (Krezel and Maret
2021) or Ba (Elwej et al.
2016) as well as interactions between Ni and Zn (Nemec et al.
2009) upregulate the production of different proteins participating to the antioxidant response element, like metallothioneins (Krezel and Maret
2017). Metallothionein functions are related to the homeostasis of essential metal ions (Zn
2+, Cu
2+) through metal binding with thiol functions and as a radical scavenger when these thiols functions are oxidised (Krezel and Maret
2017). The upregulation of metallothionein has been shown following exposure of epithelial cells to the ultrafine fraction from underground railways (Loxham et al.
2020). The observed negative associations between Cu in PM
10 and MDA
EBC (Table
3) and Zn in PM
10 and MDA
urine (Table
4) could be in line with such a protective action of both elements. Mn is also an important cofactor for manganese-superoxide dismutase (Sheng et al.
2014) and complexes between Mn and small ligands (phosphate, lactate) were reported to act as antioxidants (Aguirre and Culotta
2012). The observed negative association between Mn in PM
10 and the concentrations of MDA
urine and 8-OHdG
urine (Tables
4,
5) could be a sign of such antioxidant effect of Mn.
Nevertheless, the scientific literature also comprises conflicting results with positive associations between subway PM or metal exposure and urinary oxidative stress biomarkers. Short-term exposures to subway PM
1 have been shown to correlate with increase of 8-OHdG
urine in male volunteers (Zhang et al.
2019). Similarly, 8-OHdG
urine concentrations were greater in subway workers performing activities inside tunnels compared to those who were not (Mehrdad et al.
2015). Regarding metals, volunteers living near a closed zinc smelter with the highest urinary Cd presented an increased 8-OHdG
urine compared to volunteers with the lowest urinary Cd (Ellis et al.
2012). Boilermakers exposed to high concentrations of PM
2.5 residual oil fly ash (median: 440 µg/m
3) containing elevated levels of Mn, Ni, Pb and V also presented positive associations with 8-OHdG
urine (Kim et al.
2004). In the general population, urinary Zn [median (IQR): 198 (102–367) µg/g creatinine] was also positively associated with 8-OHdG
urine and MDA
urine (Domingo-Relloso et al.
2019). This is in contradiction with our results. The positive association between urinary Cu and MDA and 8-OHdG observed in our pilot study is in line with results from a Chinese cohort (Xiao et al.
2018) and among subjects exposed to ambient particles (Liu et al.
2018).
Significant associations between metals in particles and MDA
EBC or MDA
urine were observed with a time latency (labelled 12 or 24 h). This lag could correspond to the time needed for metal dissolution. Once present in the biological milieu, some metals (Ba, Cr and Mn in EBC; Ba in urine) have a rapid influence on the MDA concentrations. Whereas not related directly to metal exposure, an increased concentration of MDA
EBC immediately and 20 h after exposure to wood smoke was reported (Barregard et al.
2008). Also, the percentage change in MDA
urine associated with an increase of ambient PM
2.5 was the highest at lag 0 and gradually decreased with increasing lag day (Gong et al.
2013). Similarly for 8-OHdG
urine, associations with metals were mostly observed on the same day (Table
5), suggestive of a fast systemic response to subway particle exposure for this biomarker. This is in line with the fast increase of this DNA oxidative stress marker (within 1 h) observed after exposure of healthy volunteers to concentrated urban PM (Liu et al.
2018) and the reported half-lives of urinary excretion for 8-OHdG (6–35 h in (Chen et al.
2020; Loft et al.
1992)). In their quasi-experimental study during the 2008 Beijing Olympics, Gong et al. reported significant associations between exposure to PM
2.5 and excretion of 8-OHdG
urine from the exposure day until 72–96 h after exposure with a peak association at 24–48 h (Gong et al.
2014).
Study limitations and strengths
As all observational studies, this pilot study has some limitations that deserve discussion. The first is the low number of workers included in the study sample. The sample size was optimized considering the research and participant burden, the extensive and costly nature of measurements performed and the feasibility of their conduct without compromising the usual work of study participants and metro operation. However, this issue was partially solved by the prospective design of the study with repeated measurements, which allowed us obtaining 144 values per oxidative stress biomarker. The continuous individual exposure measurement over the work shift and twice a day measures of the outcomes for each worker decreased the inter-individual variability, improving by this way the possibility to observe modifications of the biomarker levels in EBC and urine. To improve the robustness of the statistics, we imposed stringent constrains on the data used. Only transition elements and biomarkers with more than 50% of their values above the LOQ were considered in the regression analysis. The statistical approach chosen might raise a concern of multiple comparisons and spurious associations. We studied the effects of 11 metallic elements on 4 biomarkers; with such figures, only 2 results with a
p value below 0.05 could be simply due to chance. Given the exploratory nature of this study and in order not to miss a possible effect (i.e., to avoid type II errors), we did not correct the main analysis for multiple testing, as recommended by Armstrong (
2014). However, to minimize this issue, we limited the analysis to the most salient correlational spots based on the heat-map visualisation (Supplementary material, Figure
S1). Therefore, our conclusions are based on a cautious interpretation of study results and using a triangulation between different analyses performed. In this study, we did not measure the bioavailable metal fraction from PM due to constraints of the budget. It is possible that associations with the soluble fraction instead of the total metal content would have been stronger than what we observed. This aspect deserves further investigation and should be considered in future studies.
The strength of our study lies in the fact that we considered simultaneously two different biological matrices. EBC can be considered as the biological matrix associated with the target organ, where oxidative stress potentially starts. Measurement of biomarkers of oxidative stress in urine, on the contrary, corresponds to the systemic response of the body to the initial insult resulting from the subway PM deposition in the lungs. Considering the association between different exposure components and effect biomarkers in a time-dependent way using three different lag times is another important strength of this study. Such analyses are quite unique and inform upon the possible mechanism of subway particle toxicity. Finally, a prospective design enabling this analysis is a methodologically strong feature enabling the temporality condition to be met when investigating the exposure–effect associations. The results of this study will inform the large epidemiological study protocol and allow a judicious choice of the most relevant effect biomarkers to be measured. For instance, this study motivated decision to exclude the 8-OHdG and 8-isoprostane measurement in EBC in the future study (Canu et al.
2021b) and thus limit the EBC volume and study budget considerably.