Quantification of Malondialdehyde in Atmospheric Aerosols: Application of the Thiobarbituric Acid Method

Based on available toxicity data, malondialdehyde (MDA; O=CHCH 2 CH=O) has been designated as a potential human carcinogen. A handful of studies suggest that MDA forms in the gas and aerosol phase in the troposphere, potentially contributing to inhalation toxicity, yet it has never been quantified in ambient air. The thiobarbituric acid (TBA) acid assay for MDA has been used as a marker for reactive oxygen species (ROS), oxidative stress, and lipid peroxidation in biological samples for decades. Here we apply the TBA assay to estimate the amount of MDA in ambient fine particulate matter (PM 2.5 ) for the first time, in samples containing biomass burning/urban aerosol from Fresno, CA, and urban aerosol from Los Angeles. We found 0.31–0.75 ng m –3 MDA in the particle phase, similar to the low end, but up to three orders of magnitude lower than the upper end of reported concentrations of the common C 3 oxygenates methylglyoxal and malonic acid. Additionally, we investigated the response in the TBA assay to seven common small oxygenates, and found interference only from acrolein, but only when the acrolein was at millimolar concentrations, well above expected levels in aerosol extracts. In sum, this work suggests that MDA is present at moderate levels in biomass burning and urban aerosols; more may be in the gas phase.


INTRODUCTION
Malondialdehyde (MDA) has been widely used as an indicator of aqueous reactive oxygen species (ROS), lipid peroxidation, oxidative stress, and rancidity in food products (Buege and Aust, 1978;Halliwell and Gutteridge, 1981;Zeb and Ullah, 2016;Agarwal and Majzoub, 2017).MDA is a product of lipid peroxidation (Buege and Aust, 1978;Del Rio et al., 2005) and OH-mediated oxidation of 2-deoxyribose, and it has been widely used to assess oxidation in human, animal, and ecotoxicity applications (Halliwell and Gutteridge, 1981;Gutteridge and Halliwell, 1988;Genaro-Mattos et al., 2009).MDA has also been shown to be mutagenic and carcinogenic, resulting in its classification as a possible human carcinogen (Millar, 1991), and may promote atherosclerosis, suggesting a role more significant than simply an indicator for oxidation in biological systems (Basu and Marnett, 1983;Niedernhofer et al., 2003;Del Rio et al., 2005;Papac-Milicevic et al., 2016).
Two studies have reported formation of gas phase MDA formation in laboratory organic photooxidation experiments.Liu et al. (1999a) observed MDA formation from photooxidation of several aromatic oxidation products including 2-butenedial, 4-oxo-pentenal, and 1,3-butadiene in the gas phase in an environmental chamber.Zhou et al. (2014) found that ozonolysis of polyunsaturated fatty acids at the surface of an aqueous layer produces gaseous MDA.Furthermore, Beeby et al. (1987) found that photolysis of glycolaldehyde in aqueous solutions produced MDA, suggesting MDA can also form in bulk aerosol or cloud water.Destaillats et al. (2002) reported identification of MDA in ambient air in San Francisco, CA, using a derivatization method coupled with High-Resolution Gas Chromatography/Ion Trap Mass Spectrometry.The authors stated that MDA co-eluted with an internal standard that was distinguishable by interpreting a combination of electron ionization, methane chemical ionization and derivative chemical ionization spectra.Their experimental design was not able to quantify its concentration and did not distinguish between gas or particle phase MDA.Okochi and Brimblecombe (2002) used a bond contribution method to estimate a Henry's Law Constant for gas-particle partitioning for MDA, estimating a value of 1.4 × 10 4 M atm -1 , below the value (~10 5 M atm -1 ) needed for the majority of MDA to partition into cloud and fog droplets.Their model predicted that 8-9% of gas phase MDA would partition into a pH 6 fog droplet when [MDA] = 10-10 -4 ppb, but that by pH 2, partitioning would be negligible.This pH dependence is understood by recognizing that aqueous malondialdehyde prefers the enol form Fig. 1(a), pK a = 4.7, which is more soluble than the dicarbonyl; at low pH the dialdehyde is favored, reducing its solubility substantially.Their model further predicts that MDA complexation of Cu(II) and Ni(II) at the droplet surface would enhance MDA partitioning (Okochi and Brimblecombe, 2002).MDA has most commonly been measured in biological and other systems via derivatization with thiobarbituric acid (TBA) (Halliwell and Gutteridge, 1981;Gutteridge and Halliwell, 1988;Genaro-Mattos et al., 2009), the approach used here.The method reacts two TBA molecules with MDA, in the presence of acid and heat to form a TBA 2 -MDA adduct that can be measured with absorption or fluorescence spectroscopy (Fig. 1(b)).Interferences have been reported from MDA precursors such as carbohydrates that can form small amounts of (TBA) 2 -MDA upon heating, and from other carbonyls that form non-MDA adducts with TBA and absorb or fluoresce at similar wavelengths (Waravdekar and Saslaw, 1959;Morales and Munné-Bosch, 2019).Such interferences are greatly reduced by using high performance liquid chromatography (HPLC) or mass spectrometry to detect TBA 2 -MDA (Moselhy et al., 2013;Domijan et al., 2015).
Here, for the first time, we apply the TBA method to estimate MDA concentrations in urban and biomass burning aerosols (BBA) particles smaller than 2.5 microns in diameter (PM 2.5 ).We also characterize the fluorescence of the extracts with detailed excitation-emission (EEM) scans and investigate the potential of seven small oxygenated species common in the atmosphere for their potential to interfere with the MDA signal in the TBA assay.

Aerosol Sample Collection
Four ambient PM 2.5 aerosol samples are tested here, one from Fresno (36.82°N, 119.74°W) and three from Los Angeles, CA (34.07°N, 118.44°W).The Fresno sample contained mixed urban aerosol mixed with biomass burning aerosol from residential wood burning in the surrounding areas, and was collected on an a 406 cm 2 Teflon-coated glass fiber filter from Sept 10-16, 2015 (Gonzalez et al., 2017).Urban PM 2.5 from Los Angeles, CA (Urban LA) was collected on the roof of the Math Sciences Building at UCLA.Urban LA samples were collected on acid washed and preweighed PTFE filters (PALL, 47 mm 2 µm pore size) using an URG cyclone at 92.5 L min -1 , corresponding to a cut size of 2.5 microns.Three samples and three blanks were collected for approximately 24 hours each during March 27 th -30 th 2019.The mass of collected particles was determined immediately after collection using a microbalance (1 µg precision, ME 5, Sartorius).To remove charge on the PTFE filters, a charge neutralizer was passed over the filter for 30 seconds before weighing.The Fresno BBA sample mass 467 µg in -2 , corresponding to average PM 2.5 concentrations of 3.0 µg m -3 , of which about 270 µg in -2 was BBA.The content of BBA was characterized with optical absorption using an aethalometer (Paulson et al., 2019).The fraction of the sample comprised of BBA is at the higher end of observed BBA fraction compared to earlier measurements in Fresno (Paulson et al., 2019).The three urban LA samples had PM 2.5 masses of 201 µg, 551 µg, and 835 µg, corresponding to average PM 2.5 concentrations of 1.5 µg m -3 , 4.1 µg m -3 , and 6.3 µg m -3 respectively (Table 1).These values are on the low end for the West Los Angeles site, but such low values are common in the spring.

Application of the 2-Thiobarbituric Acid Method to
Measure MDA in Ambient PM 2.5 PM 2.5 filter samples and blanks were placed in 15 mL Falcon tubes and extracted in 7.5 mL HPLC-grade methanol for 1 hour at room temperature in the dark.The extraction volume and time were chosen to allow all soluble organic constituents to dissolve.Extracting samples in the  Angeles 3 (Mar 29, 2019) 835 6.3 72 0.55 ± .06 0.09 ± .01 − a Mass on whole 47 mm filter for LA samples; mass on 1" square punch from Fresno filter.dark minimizes the possibility of photochemical reactions that may change the composition of the aerosol extract.Methanol was selected because it is a good solvent for small oxygenates and evaporates easily.The filters were then removed, and the methanol extracts were evaporated to dryness using a gentle stream of N 2 at room temperature and reconstituted in 720 µL of milliQ water (adjusted to pH 3 with H 2 SO 4 ), followed by addition of 4 mM TBA (30 µL of 100 mM TBA), and incubation at 100°C for 1.25 hours.

Estimation of Malondialdehyde Using 2-Thiobarbituric Acid
A wide variety of protocols have been reported for HPLC-fluorescence detection of TBA 2 -MDA in biological samples, but no protocols for the TBA assay applied to PM 2.5 extracts were available.We performed quantification of TBA 2 -MDA using High-Performance Liquid Chromatography (HPLC) with a fluorescence detector (Shimadzu RF-10AXL).A reversed phase C-18 chromatography column (GL Sciences Inc., Intersil ODS-2, 5 µm, 4.6 × 250 mm) and guard column (Thermo Scientific, ODS Hypersil JAVELIN Filter, 5 µm, 4 × 10 mm) separated analytes, and peaks were analyzed with Chromperfect Software (Justice Laboratory Software).Because the TBA 2 -MDA adduct is most stable under acidic conditions (pH 2-3) (Guillén-Sans et al., 1997) and an eluent of 7:3 acetontirile:milli-Q water (18MΩ) acidified to pH 3 (with 0.1 N sulfuric acid) was suggested by Fukunaga et al. (1995), we performed the assay at pH 3. The eluent was continuously degassed with a gentle stream of argon and delivered at a rate of 1.0 mL min -1 .The TBA 2 -MDA adduct eluted at 6 minutes, and fluorescence was measured at E x /E m = 530 nm/550 nm.
The HPLC was calibrated daily with four MDA standards ranging from 0.25 to 2.5 µM.The method detection limit was about 0.1 µM.A typical TBA 2 -MDA calibration curve is shown in Fig. 2(a); calibration slopes were within ± 12% of one another.Calibration standards were prepared from pH 3 stock solutions of 100 mM TBA and 20 mM malondialdehyde tetrabutylammonium salt serially diluted to 20 µM MDA.TBA stock solution was prepared in a Teflon bottle with stirring and heating (90°C) for approximately 15 minutes until all TBA was dissolved.The TBA was used immediately after preparation because precipitants form approximately 20 minutes after the  solution is removed from the hot plate.30 µL TBA was added to 626-711 µL pH3 MilliQ water, then 9.4-94 µL aliquots of the 20 µM MDA stock solution were added for a total volume of 750 µL.The resulting solutions were capped and incubated in a boiling water bath (100°C for 1.25 hours, after which the calibration solutions turned a pink-purple color; blanks did not change color.Solutions were cooled in a refrigerator at 4°C for 15 minutes and analyzed with the HPLC immediately.

Excitation-Emission Matrix Spectra and Interfering Compounds (3D Fluorescence)
The Excitation-Emission Matrix (EEM) scan mode (Lumina Fluorometer, Thermo Scientific) was used to determine fluorescence features of MDA calibrations, BBA extracts, PM samples and potential interfering compounds.Scans were performed every 5 nm in both excitation and emission space, using 10 nm excitation and emission slit widths and 20 ms integration time for each step.The instrument scanned at 60 nm per second.Fig. 3 shows an EEM for a 1 µM MDA standard after reaction with 4 mM TBA. Fluorescence contours indicate a fluorophore with peak fluorescence centered at Ex/Em = 530 nm/550 nm, corresponding to the TBA 2 -MDA adduct (Del Rio et al., 2005;Moselhy et al., 2013;Domijan et al., 2015).
To characterize potential interfering compounds, we made 10 mM solutions of formaldehyde, formate, oxalic acid, malonate, glyoxal, methylglyoxal, and acrolein and incubated them in the presence of 4 mM of TBA adjusted to pH 3 (with H 2 SO 4 ) heated at 100°C for 1.25 hrs.

RESULTS AND DISCUSSION
3.1 Malondialdehyde in Fresno BBA and Los Angeles PM 2.5

Concentrations
All HPLC analyses of the derivatized PM 2.5 extracts exhibited a signal that matched that of the MDA standards, with a retention time of 6 minutes and fluorescence at E x /E m = 530/550 nm, indicating the presence of TBA 2 -MDA.We used varying amounts of the Fresno PM 2.5 sample to test dependence of the signal on aerosol mass and found a linear relationship (Fig. 2(b)).The estimated concentration of MDA in the Fresno sample was 0.31 ± 0.02 ng m -3 or (10.2 ± 0.6) × 10 -3 ng MDA (µg PM 2.5 ) -1 (Fig. 4).Urban LA samples contained 51 to 97 ng MDA corresponding to approximately 0.41, 0.75, and 0.55 ng m -3 or 0.25, 0.18, and 0.087 ng µg -1 respectively (Table 1 and Fig. 4).
While we were unable to find reports of MDA concentrations in urban aerosols, we can compare concentration measurements to the concentrations of similar C 3 oxygenated organic compounds in ambient urban PM 2.5 (Destaillats et al., 2002;Ho et al., 2010;Kawamura et al., 2013;He et al., 2014;Ho et al., 2015;Shen et al., 2018).Reported concentrations for methylglyoxal, the 1, 2-carbonyl isomer of MDA (Fig. 1(c)), range from 0.8-242 ng m -3 in urban PM 2.5 (Destaillats et al., 2002;Ho et al., 2010;Kawamura et al., 2013;He et al., 2014;Ho et al., 2015;Shen et al., 2018).Malonic acid, a structurally similar molecule containing two carboxylic acids instead of two aldehyde groups (Fig. 1(c)); we note that there is no known pathway for oxidation of MDA to form malonic acid in the atmosphere), has been reported at concentrations in the range 17.6-233 ng m -3 in urban PM 2.5 (Ho et al., 2010;He et al., 2014;Ho et al., 2015).Thus, our reported range of 0.31-0.75ng m -3 MDA are similar to the low end of methylglyoxal and up to three orders of magnitude lower than the upper limits of the concentrations of both methylglyoxal and malonic acid.E x /E m = 330/410 nm are characteristic of HULIS, and are similar to Fulvic Acids (Graber and Rudich, 2006;Kuang, 2017).1).The diagonal features in the center and at E x /E m ~680/320 nm and 320/650 nm respectively are scattering artifacts inherent to the spectrometer.
TBA-aldehyde adduct of a different dicarbonyl species.Possible identities for this species are discussed below.
Figs. 6(a-c) show EEM scans of the three concentrated Urban LA PM 2.5 extracts after reaction with 4 mM TBA. Concentrated extracts of Urban LA PM 2.5 without addition of TBA had no observable fluorescence.EEMs for all three samples show the characteristic fluorescence of the TBA 2 -MDA adduct centered at E x /E m = 530/550 nm.Additional trapezoidal-shaped peaks are also observed in these samples, including the same E x /E m = 455/470 nm peak observed in the Fresno sample, and a second satellite peak at E x /E m = 640/665 nm.
While TBA may react with many aldehydes, the formation of a fluorophore requires addition of two TBA molecules, forming a conjugated system connecting the two aromatic rings, a system that is only possible for molecules with odd numbers of carbon atoms in the backbone and two aldehyde groups.However, despite running the TBA assay on multiple C 1 -C 3 oxygenated compounds (Fig. 1(c)) we were not able to produce any fluorescence features other than the one matching MDA, and that was only observed in trace quantities for acrolein (below).Generally, any substitution in the fluorophore will red-shift the peak, suggesting that the peak at 640/655 nm could be from methyl malondialdehyde (methylpropanedial) or another malondialdehyde with a substitution at the center carbon.The explanation for the peak at 455/470 nm is less clear, although its wavelengths might suggest one carbon bridging the two aromatic rings rather than three.

Other small oxygenates
We tested the common small oxygenates expected in ambient samples for their ability to react with TBA and produce a product with the same or similar fluorescence characteristics as the MDA-TBA 2 adduct.Formaldehyde, formic acid, oxalic acid, malonate, glyoxal and methylglyoxal (Fig. 1(c)) produced no measurable fluorescence anywhere in the E x /E m spectrum.Of all compounds tested, only acrolein produced any measurable fluorescence.The signal for acrolein appears at the same retention time in the HPLC and has the same fluorescence features as the TBA 2 -MDA adduct.Triplicate samples of 1 mM and 10 mM acrolein were reacted with 4 mM TBA under oxygenated conditions produced 0.45 ± 0.07 µM and 0.87 ± 0.2 µM MDA respectively, corresponding to 0.004%-0.008%conversion of acrolein to MDA.As it is unlikely that acrolein would make up more than a few % of aerosol mass, acrolein is unlikely to contribute measurably to the TBA 2 -MDA signals for the aerosol extracts.
The MDA associated with acrolein may have been present in the bottle from the manufacturer, or it may have been produced via acid hydration of the acrolein followed by oxidation, as proposed in Fig. 7.Under this mechanism, protonation of the alkene group produces a primary and secondary carbocation, followed hydration that produces 2-hydroxypropanal and 3-hydroxypropanal.Two possible oxidation products of these hydration products are glyoxal and MDA.The hydration of acrolein to 3-hydroxypropanal has been identified under acidic conditions (Pressman and Lucas, 1942;Melicherčík and Treindl, 1981;Campadelli et al., 1983), but we could find no studies identifying MDA as a product of acrolein hydration and oxidation.Furthermore, primary carbocations Fig. 7. Proposed mechanism for conversion of acrolein to MDA under acidic, oxygenated conditions.are known to be less stable than secondary carbocations.Thus, formation of MDA from acrolein should be a minor pathway, consistent with the very low observed yield.

Reactive oxygen species and potential formation of MDA in the assay
There is some potential for formation of MDA in the assay itself.This would most likely happen via an oxidation reaction, mediated by hydroxyl radicals or other reactive oxygen species.Any hydroxyl radicals or other reactive oxygen species that might form during the heating phase of the assay should be scavenged by the large excess (4 mM) of TBA in the solution.Nonetheless, for some calibration samples, a gentle stream of argon was bubbled through the solution for approximately 1 minute prior to TBA addition and incubation to remove oxygen and reduce ROS generation during the assay.No differences were observed (data not shown), indicating that oxygen does not impact the condensation reaction between TBA and MDA, and ROS did not affect the calibration.

Potential Sources of Atmospheric MDA
MDA in particles could arise via reactions on the particles themselves, or partitioning from the gas phase, either directly into the particles or into cloud or fog droplets followed by incorporation into the particles once the droplet re-evaporates.Liu et al. (1999a) reported formation of malondialdehyde from gas phase photo-oxidation of butadiene and unsaturated dicarbonyls.The Henry's law coefficient of methylglyoxal (and several other aldehydes) estimated from recent field measurements indicate that the Henry's Law partitioning coefficient for methylglyoxal could be ~10 8 M atm -1 , much higher than values reported earlier (~10 4 M atm -1 ) (Betterton and Hoffmann, 1988;Lee and Zhou, 1993;Shen et al., 2018).Since MDA and methylglyoxal have similar theoretical Henry's Law constants (~10 4 -10 5 M atm -1 ) (Okochi and Brimblecombe, 2002;Shen et al., 2018) and similar molecular structures, it is possible that they share similar gas-particle partitioning behavior.MDA has a higher boiling point (108°C) than methylglyoxal (72°C), and thus may partition into the aqueous phase even more readily than methylglyoxal.Further potentially enhancing partitioning is the formation of the enol form (above the pK a of 4.7) although this is expected to be more likely to play a role in clouds and fog, as aerosols are believed to often be too acidic (pH 0-2) for this to be relevant (Weber et al., 2016).MDA complexation to Cu(II) and Ni(II) at the droplet surface may also enhance MDA uptake (Okochi and Brimblecombe, 2002).
It is also possible that photochemical reactions (Beeby et al., 1987) or dark oxidation reactions within aerosol waters result in MDA production.MDA is expected as an oxidation product from 1,4 dienes, and 1, 3 unsaturated aldehydes; and it has been observed as an oxidation product of butadiene oxidation (Liu et al., 1999a, b).Further, polyunsaturated fatty acids are MDA precursors in biological systems, and it is well documented that aqueous oxidation of 2-deoxyribose sugar produces MDA (Halliwell and Gutteridge, 1981;Gutteridge and Halliwell, 1988).Atmospheric aerosols contain some biological material, including whole or fragmented bacteria and viruses, and fragments of plant material, material that is composed of a variety of polyunsaturated fatty acids and polysaccharides such as cellulose, hemicellulose, lignin, and free sugars including deoxyribose.

MDA Toxicity
MDA has been classified as a potential occupational carcinogen, although no reference exposure limits (RELs) have been established (CFR, 2011).Many small aldehydes exhibit toxicity, although their reference exposure limits vary widely; the reference limits for chronic exposure range from 0.35, 9 and 140 µg m -3 for acrolein, formaldehyde and acetaldehyde, respectively (OEHHA, 2021).While the REL for acrolein is in the same range as our measurements of MDA in the particle phase, it has yet to be established that MDA is as toxic as acrolein.The MDA concentrations in the particle phase are significantly lower than the RELs for formaldehyde and acetaldehyde.Future studies should aim to measure both gas and particle phase MDA concentrations for toxicology assessment.

Conclusions
The thiobarbituric acid assay has been successfully applied to measure MDA in ambient aerosol particles.Other compounds found in aerosols do not appear to present significant interferences for the assay.Levels of MDA in urban samples, including one containing a significant contribution from biomass burning, were moderate at ~0.5 ng m -3 .This concentration is at the low end of observed concentrations of similar small carbonyl compounds in ambient aerosols, but it may contribute to toxicity of ambient air in combination with gas-phase MDA and other toxic species.

Fig. 1 .
Fig. 1.(a) Malondialdehyde aqueous equilibrium reaction; (b) Condensation Reaction of MDA and TBA to form TBA 2 -MDA; (c) C 1 -C 3 compounds tested for interference in the TBA assay.

Fig. 2 .
Fig. 2. (a) Calibration curve for TBA 2 -MDA adduct measured with HPLC-Fluorescence.Peaks eluted at a retention time of 6 minutes.(b) Mass of MDA measured from TBA assay for different quantities of Fresno urban/biomass burning aerosol.Error bars indicate ± 1σ of three values measured on the HPLC from the same sample extract.

Fig. 4 .
Fig. 4. MDA measured with the TBA assay for the Fresno biomass burning aerosol (BBA, blue bar) and urban Los Angeles PM 2.5 (Urban LA; red, green, and purple bars) extracts.Error bars indicate ± 1σ of three values measured on the HPLC from the same sample extract.
Fig. 5(a) shows an EEM of the extract of 467 µg of Fresno PM 2.5 , without addition of TBA.While urban samples typically do not exhibit fluorescence, the biomass burning HUmic-Like Substances (HULIS) in the Fresno sample are strongly fluorescent.The sample's two peaks centered at E x /E m = 350/460 nm andEx /E m = 330/410 nm are characteristic of HULIS, and are similar to Fulvic Acids(Graber and Rudich, 2006;Kuang, 2017).Fig.5(b) shows an EEM scan for the same Fresno sample reacted with 4 mM TBA.After processing, the sample retains some of the fluorescence features of HULIS and gains a fluorescent feature matching the TBA 2 -MDA fluorophore centered at E x /E m = 530/550 nm.Interestingly, there is another fluorophore centered at E x /E m = 455/470 nm with a similar trapezoidal shape as the TBA 2 -MDA fluorophore.This fluorophore could arise from a

Fig. 5 .Fig. 6 .
Fig. 5(a) shows an EEM of the extract of 467 µg of Fresno PM 2.5 , without addition of TBA.While urban samples typically do not exhibit fluorescence, the biomass burning HUmic-Like Substances (HULIS) in the Fresno sample are strongly fluorescent.The sample's two peaks centered at E x /E m = 350/460 nm andEx /E m = 330/410 nm are characteristic of HULIS, and are similar to Fulvic Acids(Graber and Rudich, 2006;Kuang, 2017).Fig.5(b) shows an EEM scan for the same Fresno sample reacted with 4 mM TBA.After processing, the sample retains some of the fluorescence features of HULIS and gains a fluorescent feature matching the TBA 2 -MDA fluorophore centered at E x /E m = 530/550 nm.Interestingly, there is another fluorophore centered at E x /E m = 455/470 nm with a similar trapezoidal shape as the TBA 2 -MDA fluorophore.This fluorophore could arise from a