The importance of simulated lung fluid (SLF) extractions for a more relevant evaluation of the oxidative potential of particulate matter

Particulate matter (PM) induces oxidative stress in vivo, leading to adverse health effects. Oxidative potential (OP) of PM is increasingly studied as a relevant metric for health impact (instead of PM mass concentration) as much of the ambient particle mass do not contribute to PM toxicity. Several assays have been developed to quantify PM oxidative potential and a widely used one is the acellular dithiothreitol (DTT) assay. However in such assays, particles are usually extracted with methanol or Milli-Q water which is unrepresentative of physiological conditions. For this purpose, OPDTT measurements after simulated lung fluids (SLF) extraction, in order to look at the impact of simulated lung fluid constituents, were compared to Milli-Q water extraction measurements. Our major finding is a significant decrease of the OPDTT when the artificial lysosomal fluid (ALF) solution was used. Indeed, ligand compounds are present in the SLF solutions and some induce a decrease of the OP when compared to water extraction. Our results suggest that the effect of ligands and complexation in lining fluids towards PM contaminants probably has been underestimated and should be investigated further.


Results and Discussion
Rate of DTT loss for reference materials in the four extracting solutions. The rate of DTT loss (nmol.min −1 ) or DTT depletion was assessed for different concentrations of selected individual compounds. Further, the rate of DTT loss vs molar concentration relationship (e.g. the dose-response relationship) was established for each compound and each extracting fluid by calculating the regression equation. Only DTT depletions above the LOD (limit of detection) and DTT depletion by Zn 2+ were considered. The LOD in each media, was defined as three times the standard deviation (SD) of DTT depletion in media blanks 9 (detailed information concerning the LOD determination is available in the supplementary information 1, Tables S1 and S2).
Some compounds induced very low or slight DTT depletion (regression slope near zero and/or with R 2 < 0.6) for all the studied media and all tested concentrations. These compounds are benzo(a)pyrene (BaP) and PAH (Polycyclic aromatic hydrocarbon) derivatives: 7 H-benz(de)anthracene-7-one (BA), Benzo(b)naphtho(2,1-d) thiophene (BNT), 3-methylchrysene as well as levoglucosan, NH 4 + , Ca 2+ + SO 4 2− and Zn 2+ + SO 4 2− . Since PAHs are not redox active, they should not induce an oxidation of DTT 13 . However, it is known that PAHs can indirectly induce oxidative stress, through in vivo biotransformation by the P450 cytochrome, epoxide hydrolase and the dihydrodial dehydrogenase detoxification pathway 8,19 . Further, studies assessing the OP of ambient PM have shown strong correlations between DTT losses and PAH contents of the samples. However, these correlations can be attributed to the correlation between PAH and quinones or other organic compounds such as secondary organic aerosol (SOA), which are redox-active 13,19,[35][36][37][38] . In the case of the other non-reactive to DTT compounds, the reaction between the redox couple DTT ox /DTT (E = −0.33 V) and the redox couple Ca 2+ /Ca = −2.87 V or Zn 2+ /Zn = −0.76 V is not favored thermodynamically (i.e. is redox inert). In the case of CaSO 4 , SO 4 2− was the ion in the solution tested towards the DTT assay, and it showed no reactivity with DTT. Zinc sulfate was also tested and allowed to evaluate Zn 2+ complexation phenomena by DTT. Conversely, MnCO 3 , Cu 2+ (+2 Cl − ), CuO and the quinone 1,4-NQ induced a DTT depletion during the assay. The results are summarized in the Fig. 1 and Tables 1 to 4 (Table 1: chemicals tested in Milli-Q water, Table 2: ALF solution, Table 3: Gamble solution, Table 4: Gamble + DPPC solution).
Concerning the extractions in Milli-Q water (Table 1 and Fig. 1A), a linear dose-response relationship was observed for 1.4 NQ and Zn 2+ . A power dose-response relationship was observed for MnCO 3 and Cu 2+ . Such first and second order reactions have already been observed by Charrier and Anastasio (2012) 13 with 1,4-NQ, Mn 2+ and Cu 2+ .
However, while Charrier and Anastasio (2012) 13 observed higher DTT depletion by Cu 2+ than by Mn 2+ , in the present work, a quasi-similar DTT loss was observed for Cu 2+ and MnCO 3 (y = 0.17x 0.20 and y = 0.17x 0.23 , respectively). This difference may be due to the different speciation, Mn 2+ for Charrier and Anastasio 13 , whereas in this study, an insoluble species of Manganese (MnCO 3 ) was used. This may have induced a surface reaction, as already observed in DTT assay 39,40 , rather than a redox cycling reaction.
Copper oxide induces DTT depletion when extracted with Milli-Q water but, due to the difficulty to maintain CuO in stable suspension in Milli-Q water, no dose-response relationship was observed (no regression found).
For a theoretical concentration of 1 µM, the DTT depletion by 1,4-NQ was lower than those by MnCO 3 and Cu 2+ . Because of the power law dose-response relationships found for MnCO 3 and Cu 2+ , this pattern tends to change at higher concentrations, i.e. above 2 µM. However, quinones concentrations in ambient air are often lower than trace metal concentrations. Indeed, concentration levels of oxygenated PAHs in the environment are in the pg.m −3 to lower ng.m −3 range (depending on the compound of interest, sampling period and place) [41][42][43] 13 . Finally, a negative dose-response relationship was observed for Zn 2+ (Table 1 and Figure S5). The same observation has been made for PbO ( Figure S11) measured in a preliminary step of this study.
In the case of the ALF solution (Table 2), which presents lower pH (4.5) and higher citrate content among the different SLF solutions, a dose-response relationship was obtained for 1,4-NQ, Cu 2+ , MnCO 3 , and CuO (regression with R 2 > 0.6). A linear dose-response relationship was observed for all these compounds with the exception of CuO, for which a power law dose-response relationship was calculated. For 1,4-NQ, MnCO 3 , Cu 2+ and CuO positive relationships were obtained with DTT depletion (normalized to 1 µM) ranging as follows: 1,4-NQ > CuO > Cu 2+ > MnCO 3 (Fig. 1B). For Zn 2+ , negative results were again observed at all tested concentrations (Table 2 and Figure S5).
For the Gamble solution (Table 3), a clear dose-response relationship was established for 1,4-NQ, MnCO 3 , Cu 2+ , Zn 2+ and CuO (regression calculated with R 2 > 0.6). A negative linear response was again observed for Zn 2+ ( Figure S5). For CuO, DTT depletion during the experiment was over 70% for the highest CuO concentration (100 µM). Consequently this concentration was not used in the regression, indeed over 70% the catalytic redox reaction rate cannot be simplified as linear 46 . For the same theoretical concentration of 1 µM, DTT depletion ranging as follows: Cu 2+ > CuO > MnCO 3 ~1,4-NQ. Except for 1,4-NQ (linear dose-response relationship) these patterns tend to change at higher concentrations because of the power-law dose-response relationships observed for Cu 2+ , CuO and MnCO 3 (Fig. 1C).
As observed in Milli-Q water and ALF solution, the rates of DTT loss by BaP, BA, BNT, 3-methylchrysene, levoglucosan, NH 4 + , and SO 4 2− were insignificant. Finally, for the Gamble + DPPC solution (Table 4), a positive linear dose-response relationship was observed for 1,4-NQ, MnCO 3 while a positive power law dose-response relationship was observed for Cu 2+ , CuO.
For the same theoretical concentration of 1 µM, the DTT loss measured in the Gamble + DPPC solution, were similar to those measured in the Gamble solution and ranged in the following order: Cu 2+ > CuO > MnCO 3 ~1,4-NQ. A slight evolution can be seen in the highest concentrations due to the observed non-linear dose-response relationships (Fig. 1D). In opposition to the other fluids, in this medium a negative logarithmic dose-response relationship was obtained for Zn 2+ ( Figure S5).
Considering the highest R² values, the best regressions among the extraction solutions were observed in the Gamble + DPPC solution. This is consistent with the findings of Foucaud et al. (2007) 47 who showed an improvement of suspension stability in the presence of DPPC. Impact of the SLF extraction towards DTT loss results. We compared the rate of DTT loss in the four extracting solutions for 1,4-NQ, MnCO 3 , Cu 2+ , CuO and Zn 2+ . All these compounds presented a clear dose-response relationship.
For 1,4-NQ, DTT losses were not significantly different in Milli-Q water, ALF or Gamble + DPPC solutions. The following linear relationships were respectively determined: y = 0.13x −0.026, y = 0.13x −0.015 and y = 0.11x −0.026 and the determination coefficients were always higher than 0.9. Conversely, in the Gamble solution, DTT depletion was lower than in the other solutions (y = 0.058x −0.016) ( Figure S1).
For Cu 2+ normalized at 1 µM, the higher DTT depletion was found in the Gamble + DPPC solution (y = 0.20x 0.23 ) then in Milli-Q water and in the Gamble solution which presented equivalent losses (y = 0.17x 0.20 and y = 0.17x 0.22 ). Finally the lowest depletion was observed in the ALF solution (y = 0.013x + 0.037) ( Figure S4).
The mechanisms that would permit to explain all these differences are not yet fully understood, although some hypotheses can be made: DTT depletions are affected by pH. There is no issue of acidity with the DTT assay since a phosphate buffer was used to maintain a constant and physiological pH value. No absorbance differences of blanks were observed between the four solutions (supplementary information 2, Figure S6). Moreover, no differences have been found between DTT depletion in the four media blanks ( Figure S7). However, pH can affect the dissolution during the extraction phase of the chemicals tested. Metals carried by PM can undergo increased dissolution when pH is lowered and the ALF solution, with the lowest pH among SLF solutions, is known to allow better dissolution of insoluble compounds such as CuO or ZnO 33, 48 . Ligand content and type of reactions (surface/solution reactions, complexation etc.). Complexing anions and functional groups known to chelate metals are present in the tested SLF solutions: orthophosphates, carbonates, carboxyls (from glycine, citrate) and amines (from glycine). Detailed

Concentrations range (µM) Regression equation R² N n
Benzo(a)pyrene (BaP) 7.7 × 10 −5 -9.5 y = −0.0011x + 0.0088 0.56 6 18 7H-Benz(de)anthracene-7-one (BA) 9.7 × 10 −5 -6.6 y = 0.0019x + 0.0091 0.25 6 18   information concerning the composition of the SLF solutions is available in Table S3. The Gamble and Gamble + DPPC solutions present higher glycine contents than the ALF solution. However, the ALF solution presents the higher citrate content (20 mg.L −1 ) and so, the strongest chelating potential among the tested extracting fluids. The chelating strength of these solutions can be ordered as follows: ALF solution > Gamble + DPPC solution ~Gamble solution. The impact of the added DPPC on the chelating strength has not been studied here, consequently no assumption regarding this point was made.
As an organic compound, 1,4-NQ was not expected to form strong complexes with such species from SLF solutions, and was thus supposed to behave similarly in all fluids. However, in the Gamble solution, 1,4-NQ had a lower response towards DTT than in the other media. It can be assumed that this medium is not as favorable to electron transfer as the other solutions. This highlights the possible role of DPPC in electron transfer since DTT depletions are higher in the Gamble + DPPC solution compared to the Gamble solution. Furthermore, the transformation of 1,4-NQ in solution cannot be ruled out, as 1,4-NQ aging has been observed during storage but also during the extraction phase ( Figure S11), thus possibly explaining the weaker DTT depletion by 1,4-NQ in the Gamble solution.
Chemical speciation calculations conducted with the PhreeqC 49 code showed that Cu 2+ forms several complexes in the Gamble solution. For a 25 µM concentration of CuCl 2 in the Gamble solution calculations showed that Cu 2+ primarily forms complexes with glycine and to a lesser extent with citrate and carbonate. Cu 2+ , the free species, should not account for more than 0.0006%. However, DTT depletion by Cu 2+ was very similar in Milli-Q water and the Gamble solution, likely because glycine-Cu complexes present the same redox potential as Cu 2+ , as shown by Mohamed et al. (2012) 50 . The higher Cu 2+ reactivity in the Gamble + DPPC solution can be explained by the potential role of DPPC as an electron transfer facilitator. Cu 2+ speciation in the ALF solution was also calculated with PhreeqC: Cu 2+ primarily forms citrate complexes, which can explain the observed lowest DTT depletion in the ALF solution as citrate-Cu complexes redox potential is reduced as compared to Cu 2+ 51 .
Concerning Zn 2+ , a negative dose-response relationship was observed whatever the extraction media indicating that more DTT was oxidized in blanks than in samples (Table 1 to Table 4 and Figure S5). As already observed previously by Held et al. (1996) 52 , this can be explained by a complexation process between DTT and Zn 2+ , which may have stabilized DTT from oxidization 40 . The same assumption can be applied to PbO 39 ( Figure S11). However and compared to the other tested solutions, the ALF solution exhibited different patterns of DTT depletion rate, with fewer negative linear dose-response relationships. It shows that, in this fluid, DTT is less protected by its complexation with Zn 2+ , which preferentially reacts with the other ligands of the ALF solution (citrate etc.). Once again differences were observed between the Gamble and Gamble + DPPC solutions (linear vs logarithmic dose-response relationship).
Rate of DTT loss by the insoluble MnCO 3 in the tested media decreases (when MnCO 3 < 50 µM) as the concentration of ligands in the SLF solution increases: Milli-Q water > Gamble solution ~Gamble + DPPC solution > ALF solution. These results can be explained by a minor or absence of constituent effect of the Gamble and Gamble + DPPC solutions on MnCO 3 dissolution. The main mechanism involved is likely a surface reaction and a surface occupation of reactive sites by SLF constituents. In the case of the ALF solution, which presents the lowest pH, MnCO 3 could be partially dissolved during the extraction phase and further Mn 2+ complexation with citrate decreases DTT depletion.
DTT losses by the insoluble CuO were very close in both Gamble and Gamble + DPPC solutions, DTT loss by CuO in the ALF solution was lower than in the other fluids and can probably be explained in the same manner as for MnCO 3 , i.e. through a partial dissolution followed by interactions/complexations with citrate. However, this effect is less pronounced than for MnCO 3 probably because of the even lower solubility of CuO.
In summary, DTT depletion measured within the four extracting solutions varied among the tested compounds and also with their interaction with the chemicals present in solution, which have variable chelating strengths (citrate -glycine -DTT). It should be noted that, citrate was primarily used to replace the proteins of the lung lining fluid 53 (such as iron or copper-binding proteins), which can act as a local defense system against oxidative reactions 54 . As illustrated with CuO results, it should also be noted that SLF solutions facilitate the OP measurement when compared with Milli-Q water results. To go further, oxidative potential of ambient PM has been measured after extractions in the four media.

OP DTT results from ambient PM and implications according to the extraction solution. OP DTT
is defined as the mass-normalized DTT depletion for a given substance (nmol DTT.min −1 .µg −1 ). The OP DTT of ambient PM was evaluated for the three SLF solutions and for Milli-Q water to look at the effect of the SLF composition on the OP of complex PM. In France, particulate matter sanitary alert are based on PM 10 measurement. PM 10 studied in this paper, were collected on filters by the certified associations for the monitoring of the air quality (AASQA) and came from the urban background of Nice (southeast of France), from the city of Passy (French Alpine Valley) and from a subway station of Toulouse (southwest of France). Two certified reference materials (ERM CZ 120 and NIST 1649 b) were also tested.
The filters tested were chosen randomly in terms of sampling dates and came from various locations. The chemistry involved is thus quite different and non-specifically representative of the collection sites (average concentration and composition). The objective was not to compare the different locations but rather to evaluate the oxidative potential of very different PM 10 in the three SLF, to start exploring the role of their composition on the PM 10 toxicity as measured within the DTT assay.
DTT depletion per cubic meter (nmol DTT.min −1 .m −3 ) was not calculated as for the CRMs these calculations were impossible. Figure 2 presents the OP DTT of the different tested aerosols (realized in duplicate) in the four model fluids. Figure 3 shows the chemical composition of the main compounds for the three locations and for ERM CZ120 (such complete chemical composition was not available for NIST 1649b).
The OP DTT measured in Milli-Q water, Gamble and Gamble + DPPC solution does not present statistically significant differences. Conversely, the ALF solution led to OP results significantly lower than in the other solutions (Man-Whitney statistical test, p-value < 0.05) (Supplementary information 3, Table S4 and Figure S8). However, it has to be noted that the results obtained in Milli-Q water were systematically higher than in the other extraction solutions.
The samples present a different chemical composition (Fig. 3) which can provide some clarifications about the previous OP results.  The overall lower OP results obtained in the ALF solution can be linked to citrate, a strong metal chelator present in this solution. PM from Toulouse's subway presented the lowest OP variations among all SLF solutions and the highest DTT depletion observed in the ALF solution. Dissolution is increased in the ALF solution due to the low pH during extraction (see bioaccessibility percentage for ERM CZ 120 in the Table S7) and this is combined with a strong complexation of citrate leading to the poor available of metal for DTT reaction. However, high OP results from subway PM after ALF extraction could have been triggered by a greater concentration of free metallic ions than for other PM (highest content of metals, 71%) due to the low pH of ALF solution and further less complexed by citrate since high concentrations of dissolved species were available. It could also be assigned to metal surface reaction 39 due to high metal-rich content partially insoluble (mainly oxides). With 29% metal content, the ERM CZ120 sample also presented a noticeable DTT depletion after ALF extraction supporting the previous hypotheses. Moreover, to end with discussion regards ALF extraction, result for the PM from Nice whose composition is driven by high OC, presents a strong diminution of OP compared to other extractions solutions. High content of OC that usually correlates with high OP values which, presently, do not drive anymore OP results when accounting for ALF extraction. Such result can also add in favor the importance of metals when using ALF extractions solution. In the case of Passy, the lower OP were observed among all locations, the results can be explained by the high nitrate content (45% wt) which is non-reactive to DTT and by the lowest metal content observed among the different aerosols (2%). Finally, for Nice, significantly higher OPs for all extraction solutions (except ALF) were observed and may highlight the importance of organic compounds (48% in such samples), such as SOA, in the OP of ambient PM 19,[35][36][37][38] .
However, the result for the PM from Nice after the ALF solution extraction, which presents a strong diminution of the OP, cannot rule out an inhibition of the DTT oxidation by the organic compounds in presence of ALF ligands (since organic reference compounds presented also < LOD OP results after extraction). Studies on the OP of environmental samples often use metal chelator agents as DTPA in order to inhibit metal ion-induced ROS 55,56 . However, Charrier and Anastasio (2012) 13 showed that the addition of EDTA suppresses DTT response from metals and surprisingly, also from quinones. But, no other metal chelating agents (DTPA, citrate etc.) have been studied. This result may support our observation of an inhibition of DTT depletion by organic compounds in presence of chelating agents.
In this study, DTT assay has been applied in conditions more closely mimicking lung physiology. Interactions between the lung fluids constituents and OP have been observed. The study using reference compounds highlighted the importance and role of specific interactions between chemical species (Cu 2+ …), their speciation, and SLF organic ligands (such as glycine or citrate): they were shown to be able to decrease the OP of these elements through mechanisms involving chelation and/or surface reactions. The decrease or increase of OP values for reference compounds after SLF extraction was unpredictable compared to Milli-Q water extraction. This could be a significant source of biases in correlations with health outcomes studies that could be addressed by further improvements in OP assays. Our results with the CuO also suggest that the organic constituents of the SLF solution allowed a better suspension and as a consequence facilitate measurement of the OP within the DTT assay.
The ALF solution simulates inflammatory conditions as may be encountered during PM pollution peaks. Studying OP of PM 10 with DTT assay in this particular media has shown interesting results. Extractions in the ALF solution induce a significant difference in comparison with others. OP results with the DTT assay is significantly lower in this media and can be explained by complexation phenomena (mainly with citrate). Then, no significant differences have emerged between Milli-Q water, Gamble or Gamble + DPPC solution extractions. However, a constant pattern is observed with always higher OP values in Milli-Q water than in the other extraction solutions. This trend cannot rule out a possible statistically significant difference on a more important number of samples (here, only 5 samples) or in other environments, as such differences exist among reference compounds. However, all these observations have been made for the DTT assay and cannot be generalized to other acellular assays.
This alternative protocol using extractions in SLF solutions before OP measurement allow to focus on PM compounds really available for ROS generation. All these results suggest that the effect of ligands and complexation in lining fluids probably has been underestimated and should be investigated further. Moreover, our results indicate that combining information from OP measurements and PM chemical composition could be important for understanding the complex mechanisms that can bias OP measurements. OP measurements through the DTT assay appeared correlated with some health outcomes 28,29 , other OP measurements were not always good indicators of PM toxicity 28,29,57,58 . Such differences could be assigned to difference into assay designs. In the future, results of such new design, with SLF extraction before OP measurement, could be compared to epidemiological data in order to evaluate if it can improve the links between OP measurement and observed pathologies. Finally, OP assays standardization is compulsory to go forward, so as to allow epidemiology-OP measurement cross-studies.
Samples. PM sampling: Ambient particles were collected by filtration during 24 h (24 × 30 m 3 .h −1 ) with DIGITEL DA-80 (high volume samplers) on quartz filters (Tissuquartz Pallflex). The filters were calcined at 500 °C for 8 h before use. After sampling, the filters were folded, wrapped in aluminum foils sealed in polyethylene bags and stored at −15 °C until analysis 44 . Samples from different locations with diverse compositions were used in this study: from Passy known to present a large contribution of domestic biomass burning to the PM mass 59 , ambient PM from the urban background of Nice with significant marine contribution to the PM mass, and ambient PM collected at a subway station (Toulouse) with high metal content. ERM CZ 120 and NIST 1649 b were used as certified reference material (CRM).
Chemicals species from PM: Six PAH and PAH derivatives (Benzo(a)pyrene (BaP), 7H-Benz(de)anthracene-7-one (BA), Benzo(b)naphtho(2,1-d)thiophene (BNT(2,1-d), 3-methylchrysene, 1,4 naphtoquinone (1,4-NQ), one common anhydro sugar (levoglucosan), two ions (NH 4 + and SO 4 2− ), and four metals (Manganese (II) carbonate, Copper (II) chloride, Copper (II) oxide, Zinc sulfate heptahydrate) were tested. Different concentrations of soluble or insoluble compounds have also been tested for different speciation (CuCl 2 and CuO). As much as possible, the tested concentrations were in the range of atmospheric concentrations 42, 59-61 taking into account the average volume of air inhaled by an adult (around 10 m 3 .day −1 ) and the average volume of lung fluid (20 ml) 32 . All these conditions, as well as justification of compounds selected, are summarized in Table S5. Simulated lung fluid (SLF). Three simulated lung fluids (SLF) have been tested. These fluids were designed to model the interactions of PM with extracellular lung fluids 53 . The Gamble's solution represents the interstitial fluid deep within lungs and is mixture of salts 53 (pH: 7.4). Lung macrophages are one of the main producers of ROS in the early inflammation phase. To compensate ROS production, and possibly to modulate ROS release, macrophages use antioxidant defenses 62,63 . ALF's solution mimics fluid after phagocytosis by alveolar and interstitial macrophages (pH: 4.5) and simulate inflammatory conditions 64,65 . The third SLF is the Gamble's solution supplemented with dipalmitoylphosphatidylcholine (DPPC) (pH: 7.4). This major phospholipid of lung surfactant helps reducing surface tension at the air-water interface of the terminal airways and improves lungs defensive function 66 . The citrate in the SLF solutions replaces proteins that can be found in lung lining fluid and acetate was used instead of organic acids 53 For the ambient PM and CRM study, the PM mass tested (1 punch in 2 ml of extraction solution) are 0.49 µg, 1.01 µg, 0.68 µg respectively for Nice, Toulouse and Passy and 0.86 µg for the two CRM (supplementary information on the filters is available in supporting information 6).
DTT assay. Experimentation for DTT depletion induced by reference materials was conducted in 48 well plates whereas the experimentation for the DTT depletion induced by ambient PM was conducted in 96 well plates.
This test consisted of monitoring DTT depletion when in contact with PM; the rate loss of DTT (when in excess) is proportional to the concentration of redox-active species present in PM 9, 39 . A semi-automated procedure was used for the DTT assay with a plate-reader TECAN spectrophotometer Infinite ® M 200 pro, and 48 and 96 well CELLSTAR ® multiwall plates from Greiner bio-one ® . DTT depletion was monitored for 30 minutes with the following procedure: first, measurement of the intrinsic absorbance (Abs int ) of particles and substrate at 412 nm was made. Then, for experiments conducted in the 48 well plates, 25 nmol of DTT (50 µL of 0.5 mM DTT solution in phosphate buffer) was injected, then for each sample (blank included) DTT was quantified immediately (t = 0) and after 15 and 30 minutes (t = 15 and t = 30) of exposure, with samples in triplicate (100 nmol of 5,5′-dithiobis(2-nitrobenzoic acid) -100 µL of 1 mM DTNB in phosphate buffer). Because of the high repeatability between the triplicates (average CV = 3.2%) (Supplementary information 5, Table S6), positive controls (50 µL of 24.5 ± 0.5 µM 1,4-NQ solution) were quantified only once at each measurement time. For triplicate samples, if the coefficient of variation (CV) was higher than 5% the extreme value between the triplicates was removed.
The rate of DTT loss (nmol.min −1 ) was determined from the slope of the linear regression of calculated nmol of consumed DTT vs time. The amount of remaining DTT was calculated as follows: where n DTT, i is the amount of DTT (nmol) at t = i, Abs i is the absorbance at t = i, n DTT,O is the initial amount of DTT (nmol) and Abs t0 is the absorbance at t = 0.
The quality of the linear regression was estimated with the coefficient of regression R². The linear regression was considered acceptable when R² > 0.90 and when less than 70% of the initial amount of DTT had been oxidized 46 . The repeatability of measurement was estimated using the positive control (1, and was estimated below a 5% difference (Supplementary information 5, Figures S12 and S13). The intrinsic absorbance was subtracted from the final absorbance and the DTT loss in the blanks (oxidation of 100 µM DTT by dissolved oxygen and no added redox-species) was subtracted from the DTT loss of the samples in order to obtain the actual DTT depletion of the samples.
Mass-normalized DTT depletion rates were also calculated and referred to as OP (nmol DTT.min −1 .µg −1 ), by dividing the DTT depletion of the sample by the PM mass (µg) added in the wells.
Statistical analyses. All statistical analyses were carried out using the R statistical software 3.4.0.
Non-parametric Mann-Whitney tests were used in order to look at the statistical significance between the different extraction solutions and for the environmental samples. This non-parametric test was chosen because of the small size of the dataset.