Comparison between five acellular oxidative potential measurement assays performed with detailed chemistry on PM10 samples from the city of Chamonix (France)

Abstract. Many studies have demonstrated associations between exposure to ambient
particulate matter (PM) and adverse health outcomes in humans that can be
explained by PM capacity to induce oxidative stress in vivo. Thus, assays
have been developed to quantify the oxidative potential (OP) of PM as a more
refined exposure metric than PM mass alone. Only a small number of studies
have compared different acellular OP measurements for a given set of ambient
PM samples. Yet, fewer studies have compared different assays over a year-long period and
with detailed chemical characterization of ambient PM. In
this study, we report on seasonal variations of the dithiothreitol (DTT),
ascorbic acid (AA), electron spin resonance (ESR) and the
respiratory tract
lining fluid (RTLF, composed of the reduced glutathione (GSH) and ascorbic
acid (ASC)) assays over a 1-year period in which 100 samples were analyzed.
A detailed PM10 characterization allowed univariate and
multivariate regression analyses in order to obtain further insight into
groups of chemical species that drive OP measurements. Our results show that
most of the OP assays were strongly intercorrelated over the sampling year
but also these correlations differed when considering specific sampling
periods (cold vs. warm). All acellular assays are correlated with a
significant number of chemical species when considering univariate
correlations, especially for the DTT assay. Evidence is also presented of a
seasonal contrast over the sampling period with significantly higher OP
values during winter for the DTT, AA, GSH and ASC assays, which were assigned
to biomass burning species by the multiple linear regression models. The ESR
assay clearly differs from the other tests as it did not show seasonal
dynamics and presented weaker correlations with other assays and chemical
species.



Section 1 Details on the chemical compounds analyzed
Briefly, the instrumental techniques used for PM 10 chemical characterization are: -A Sunset instrument and the EUSAAR2 protocol for elemental and organic carbon (OC, EC). 15 -An AE33 Aethalometer for BC and the distinction between wood burning and fossil fuel BC.
-A combination of separation on a DEAE resin and quantification with a DOC analyzer for humic like substances (HULIS°) after a Milli-Q water.
-∑ PAHs correspond to the sum of the following analyzed PAH: phenanthrene, anthracene, fluoranthene, pyrene,

Section 2 DTT and AA assays
Linear response to PM concentrations : DTT and AA assays: In both cases, 1 to 6 punches from 2 samples (different year period) were extracted using Gamble + DPPC solution and in order to investigate assay response pattern towards PM concentration. Non-linear response was observed in the case of the 15 DTT assays (Figure S1 A) whereas for the AA assays linear pattern was observed (Figure S1 B).

Field vs lab blank filters :AA assays:
Field blank sampling dates were identical to those employed in the DTT assay. For the AA assay, no difference in AA depletion was found between field and lab blanks ( Figure S3).

Section 3 The ESR assay
For ESR assay, used parameters were: Microwave power, 1.7 mW; Modulation amplitude, 0.18 mT; Modulation frequency, 100 KHz. The spectra exhibit the 4-line pattern with ratio of intensity 1:2:2:1 that is reminiscent of the HO • adduct of DMPO. Before quantification we ensured that the spectra were recorded under non-saturating conditions. For the measurement of intensity two methods were tested: Double integration of the whole spectra or measurement of the height of 5 the central lines (the linewidth was constant within the series). The second method was preferred since more accurate and reproducible.
We investigate the influence of the extraction time on the intensity of the ESR signal. Results are depicted in the Figure S4, a plateau was reached around 40 min. After approximately one hour the signal slightly decreases, presumably due to the 10 decomposition of the OH-DMPO adduct.
The compatibility between the assay and the use of Gamble + DPPC solution was tested as follows: filter punches from the same sample (03/12/2013) were placed in 1 mL tubes to which were added 125 µL of H 2 O 2 (0.5 M) and 250 µL of DMPO (0.05 M). Then, 125µL of either Milli-Q water or a Gamble + DPPC solution was added to the tubes. Tubes were vortexed during 15 seconds before being placed in an agitator plate for incubation at 37° C for 40 min. After incubation, the 15 suspension was vortexed again for 15 s. 35 µL of the suspension was then transferred into a capillary tube to measure hydroxyl radical (HO • ) formation catalyzed by PM 10 by ESR spectroscopy. The results showed much weaker signal intensity for the samples prepared in the Gamble + DPPC medium in comparison to Milli-Q water ( Figure S5).
Calibration curve was obtained as follows: Half, 1 and 2 filter punches deriving from the same sample (03/12/2013) were placed in 1 ml tubes, 125 µL Milli-Q water, 125 µL H 2 O 2 (0.5 M) and 250 µl of DMPO (0.05 M) were added. The 3 sub-20 samples were vortexed for 15s, placed in incubation at 37°C for 40 min, vortexed again and finally, 35 µl were transferred to capillary tubes for analyses. This relationship shows that the intercept = 0 was not reached (y= 2484.4x + 32 700). We made the supposition that the relationship between 0 µg and the 1 st measuring point (2.5 µg) was linear. For samples overpassing 2.5 µg of PM in the capillary tube, the signal was corrected using the linear curve. Thirty-seven samples were corrected out of the 75 analyzed samples. 25

Section 4 Cumulative score of correlations
Spearman correlations were employed to relate the chemical composition of PM 10 (48 species/ compounds) with OPv measurements. A score of 1 was attributed when the correlation was >0.45 (i.e. moderate). A cumulative score of correlations was then calculated by adding the number of correlation > 0.45.
As an example, 5/5 cumulative score indicates that the 5 OPv measurements exhibit moderate to strong correlations with the chemical compounds of interest. A 37/48 cumulative score indicates that 37 over 48 chemical species show moderate to strong correlations with OPv.

Section 5 Multivariable linear regression model
An hypothesis on the additive effect of the chemical species on the OP measurement has been assumed (Charrier et al., 5 2015).

Data set preparation:
Extreme values of chemistry have been removed by performing monthly boxplots for each chemical compound over the sampling year. Samples (sampling days) showing values that lie an abnormal distance (Q75th +1.5 IQR, with IQR Q75th-10 Q25th) from other values (i.e. outliers) were manually discarded.
Variable transformations were then made to obtain as close as possible, normal distribution. Table S2 summarizes information on the observations used for the models, the variable transformation used, as well as the p value from the Shapiro-Wilk test to test the null hypothesis that OP data or transformed OP data come from a normally distributed population. 15 For the ESR assay, no transformation was necessary, since the initial distribution is already quasi normal. For DTT, AA and ASC assay log transformations were done. It has to be noted that for the AA assay (single compound assay), two distinct models were performed since the overall data set is distributed among 2 normal distributions, one in a colder period (19 observations from November to mid-March), and other in a warmer period (40 observations from mid-March to October).
Finally, for the GSH assay, square root transformations were done, but normal distribution was not really reached. However, 20 analysis was pursued for this assay (GSH assay) as a first rough investigation.

Model realizations and validations:
BIC number allows selecting variables that are significant to explain observed OP values. For models evaluation, residues and collinearity between predictors (chemical species) were taken into account (Table S3). All residue models are quite well randomly and normally distributed around zero. In the case of OP GSHv, the normal distribution was, once again, not fully 25 reached (p-value = 0.05). The residual plot for log (AAv) for the cold period shows that residues are not evenly distributed.
Because of the small dataset for this period (n=19), it is difficult to conclude whether it is the effect of the very small dataset or the presence of two different groups than cannot be modeled similarly. This second assumption for OP AAv cannot be excluded since two groups have already been identified for the two seasons.
To evaluate collinearity, VIF (variance inflation factors) were calculated and were under 10 indicating no collinearity between chemical species used in the model.

Application to the overall data set: 5
For some observations, the models, when applied to the general data set, overestimated OP values (more than 2 times higher than observed OPv). These values represent only between 6 % (OP AAv) to 13% (OP GSHv) of the number of observations and are mostly found during the cold period, indicating that PM of very different nature can happen at this time. The smaller number of overestimated observations in the case of the AA assay is due to the two models used, with one centered on the cold period, such combination of 2 models leading to the best overall model explaining OP variance. 10

Contribution of each predictor:
The daily contribution of predictors was calculated by using coefficients obtained in the different model and daily atmospheric concentration of predictors. Log(OPv) and √OPv were back-transformed in order to obtain predicted OPv and daily contribution of predictors to the predicted OPv. Cold and warm period means were calculated, without taking 15 overestimated values.