Exposures were conducted using immortalized murine alveolar macrophages (MH-S, ATCC^®CRL-2019TM) as they are the first line of defense against environmental insults (Oberdörster, 1993; Oberdörster et al., 1992). The particular cell line also retains many properties of primary alveolar macrophages, including phagocytosis as well as the production of ROS/RNS and cytokines (Sankaran and Herscowitz, 1995; Mbawuike and Herscowitz, 1989). MH-S cells were cultured in RPMI-1640 media supplemented with 10 % fetal bovine serum (FBS, Quality Biological, Inc.), 1 % penicillin–streptomycin, and 50 µM β-mercaptoethanol (BME) at 37 ∘C and humid air containing 5 % CO2. For exposure experiments, MH-S cells were seeded at a density of 2 × 104 cells well-1 onto 96-well plates pretreated with 10 % FBS in phosphate-buffered saline (PBS, Cellgro). For seeding and all assay procedures thereon, FBS-supplemented cell culture media without BME addition was used as BME is a reducing agent that may interfere with inflammatory measurements.

SOA formed from the photooxidation of biogenic and anthropogenic precursors were generated in the Georgia Tech Environmental Chamber (GTEC) facility. Details of the facility have been described elsewhere (Boyd et al., 2015). Briefly, the chamber facility consists of two 12 m3 Teflon chambers suspended within a 6.4 m × 3.7 m (21 ft × 12 ft) temperature-controlled enclosure. Black lights and natural sunlight fluorescent lamps surround the chambers, and multiple sampling ports allow for injection of reagents as well as gas- and aerosol-phase measurements. Gas-phase O3, NO2, and NOx concentrations were monitored using an O3 analyzer (Teledyne T400), a cavity attenuated phase shift NO2 monitor (Aerodyne), and a chemiluminescence NOx monitor (Teledyne 200EU), respectively, while hydrocarbon decay was monitored using a gas chromatography–flame ionization detector (Agilent 7890A). Hydrocarbon decay was also used to estimate hydroxyl radical (OH) concentrations. For aerosol-phase measurements, a scanning mobility particle sizer (SMPS, TSI) was used to measure aerosol volume concentrations and distributions, while a High-Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS, Aerodyne; henceforth referred to as the AMS) was used to determine bulk aerosol composition (DeCarlo et al., 2006). AMS data was analyzed using the data analysis toolkit SQUIRREL (v. 1.57) and PIKA (v. 1.16G). Elemental ratios, including O : C, H : C, and N : C, were obtained using the method outlined by Canagaratna et al. (2015) and used to calculate the average carbon oxidation state (OS‾c; Kroll et al., 2011). Temperature and relative humidity (RH) were also monitored using a hydro-thermometer (Vaisala HMP110).

Experiments were designed to probe the effects of humidity, RO2 fate, and precursor identity on cellular inflammatory responses induced by different SOA formed under these conditions (Table 1). All chamber experiments were performed at ∼ 25 ∘C under dry (RH < 5 %) or humid (RH ∼ 45 %) conditions. Chambers were flushed with pure air (generated from AADCO 747-14) for ∼ 24 h prior to each experiment. During this time, chambers were also humidified for humid experiments by means of a bubbler filled with deionized water. Seed aerosol was injected by atomizing a 15 mM (NH4)2SO4 seed solution (Sigma Aldrich) to obtain a seed concentration of ∼ 20 µg m-3. It should be noted that experimental conditions deviate for experiment 7 (isoprene SOA under RO2 + HO2 dominant, “humid” conditions) due to low SOA mass yields. For this experiment, an acidic seed solution (8 mM MgSO4 and 16 mM H2SO4) and a dry chamber were used to promote SOA formation via the isoprene epoxydiol (IEPOX) uptake pathway. This pathway has been shown to contribute significantly to ambient OA and has a higher SOA mass yield compared to the IEPOX + OH pathway (Surratt et al., 2010; Lin et al., 2012; Xu et al., 2015).

SOA precursor was then introduced by injecting a known amount of hydrocarbon solution (isoprene, 99 %; α-pinene, ≥ 99 %; β-caryophyllene, > 98.5 %; pentadecane, ≥ 99 %; m-xylene, ≥ 99 %; naphthalene, 99 %; Sigma Aldrich) into a glass injection bulb and passing pure air over the solution until it fully evaporated. For pentadecane and β-caryophyllene, the glass bulb was also heated gently during hydrocarbon injection to ensure full evaporation (Tasoglou and Pandis, 2015). Naphthalene was injected by passing pure air over solid naphthalene flakes as described in previous studies (Chan et al., 2009). OH precursor was then introduced via injection of hydrogen peroxide (H2O2) for RO2 + HO2 experiments or nitrous acid (HONO) for RO2 + NO experiments. For H2O2, a 50 % aqueous solution (Sigma Aldrich) was injected using the same method described for hydrocarbon injection to achieve an H2O2 concentration of 3 ppm. This amount yielded OH concentrations on the order of 106 molec cm-3. For HONO injections, HONO was first prepared by adding 10 mL of 1 % wt aqueous NaNO2 (VWR International) dropwise into 20 mL of 10 % wt H2SO4 (VWR International) in a glass bulb. Zero air was then passed over the solution to introduce HONO into the chamber (Chan et al., 2009; Kroll et al., 2005). Photolysis of HONO yielded OH concentrations on the order of 107 molec cm-3. NO and NO2 were also formed as byproducts of HONO synthesis. Once all the H2O2 evaporated (RO2 + HO2 experiments) or NOx concentrations stabilized (RO2 + NO experiments), the UV lights were turned on to initiate photooxidation.

Aerosol samples were collected onto 47 mm TeflonTM filters (0.45 µm pore size, Pall Laboratory). The total mass collected onto each filter was determined by integrating the SMPS time-dependent volume concentration over the filter collection period and multiplying by the total volume of air collected. SMPS volume concentrations were converted to mass concentrations by assuming a density of 1 g cm-3 to facilitate comparison between studies. To account for potential H2O2 or HONO uptake, background filters were also collected. These filters were collected when only seed particles and OH precursor (H2O2 or HONO) were injected into the chamber under otherwise identical experimental conditions. All collected samples were placed in sterile petri dishes, sealed with Parafilm M^®, and stored at -20 ∘C until extraction and analysis (Fang et al., 2015b). Collected particles were extracted following the procedure outlined in Fang et al. (2015a) with modifications for cellular exposure. Briefly, filter samples were submerged in cell culture media (RPMI-1640) and sonicated for two 30 min intervals (1 h total) using an ultrasonic cleanser (VWR International). In between sonication intervals, the water was replaced to reduce bath temperature. After the final sonication interval, sample extracts were filtered using 0.45 µm PTFE syringe filters (FisherbrandTM) to remove any insoluble material and supplemented with 10 % FBS (Fang et al., 2015b).

ROS/RNS were detected using the assay optimized in Tuet et al. (2016). Briefly, the assay consists of five major steps: (1) pretreatment of 96-well plates to ensure a uniform cell density, (2) seeding of cells onto pretreated wells at 2 × 104 cells well-1, (3) incubation with ROS/RNS probe (carboxy-H2DCFDA, Molecular Probes C-400) diluted to a final concentration of 10 µM, (4) exposure of probe-treated cells to samples and controls for 24 h, and (5) detection of ROS/RNS using a microplate reader (BioTek Synergy H4, excitation/emission: 485/525 nm). Positive controls included bacterial cell wall component lipopolysaccharide (LPS, 1 µg mL-1), H2O2 (100 µM), and reference filter extract (10 filter punches mL-1, 1 per filter sample, from various ambient filters collected at the Georgia Tech site), while negative controls included blank filter extract (2 punches mL-1) and control cells (probe-treated cells exposed to media only, no stimulants).

A previous study on the ROS/RNS induced by exposure to ambient PM samples found that ROS/RNS production was highly dose dependent and could therefore not be represented by measurements taken at a single dose (Tuet et al., 2016). Here, we utilize the dose-response curve approach described in Tuet et al. (2016). For each aerosol sample, ROS/RNS production was measured over 10 dilutions and expressed as a fold increase in fluorescence over control cells. A representative dose-response curve is shown in Fig. 1. For comparisons to other inflammatory endpoints and chemical composition, ROS/RNS production was represented using the area under the dose-response curve (AUC), as AUC has been shown to be the most robust metric for comparing PM samples (Tuet et al., 2016).

Secreted levels of TNF-α and IL-6 were measured post-exposure (24 h) using enzyme-linked immunosorbent assay (ELISA) kits following the manufacturer's specifications (ThermoFisher). This time point was chosen to enable comparison with ROS/RNS levels (also measured at 24 h, optimized in Tuet et al., 2016) and to ensure a high signal for both cytokines. Previous literature has shown that TNF-α and IL-6 production peak around 4 and 24 h, respectively (Haddad, 2001). However, while TNF-α production peaks earlier, the signal at 24 h is well above the detection limit of the assay, and previous studies have utilized this time point to measure both cytokines (Haddad, 2001; Matsunaga et al., 2001). Nonetheless, it should be noted that these measurements represent a single time point in the cellular response. All measurements were carried out using undiluted cell culture supernatant. For each aerosol sample, TNF-α and IL-6 were measured over seven dilutions and represented as a fold increase over control. Similarly, the AUC was used to represent each endpoint for comparison purposes.

The MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay (Biotium) was used to assess cellular metabolic activity. Briefly, supernatants containing sample extracts were removed after the exposure period and replaced with media containing MTT. Cells were then returned to the incubator for 4 h, during which the tetrazolium dye was reduced by cellular NAD(P)H-dependent oxidoreductases to produce an insoluble purple salt (formazan). Dimethyl sulfoxide was then used to solubilize the salt, and the absorbance at 570 nm was determined using a microplate reader (BioTek Synergy H4).

Linear regressions between bulk aerosol composition and cellular inflammatory responses were evaluated using Pearson's correlation coefficient, and the significance of each correlation coefficient was determined using multiple imputation, which calculated the total variance associated with the slope of each regression. Details of this method are described in Pan and Shimizu (2009). Briefly, response parameters (i.e., AUCs for each endpoint) were assumed to follow a normal distribution. Ten “estimates” were obtained for each response using the average and standard deviation determined from the dose-response curve fit. These estimates were then plotted against bulk aerosol composition (e.g., O : C, H : C, and N : C) to obtain 10 fits, and the slopes and variances generated from these fits were used to calculate the between and within variance. Finally, a Student's t test was used to calculate and evaluate the associated p values using a 95 % confidence interval.

To investigate whether SOA formed from different precursors elicited different inflammatory responses, levels of ROS/RNS, TNF-α, and IL-6 were measured after exposing alveolar macrophages to SOA generated from six volatile organic compounds generated under three formation conditions (Table 1). The AUC per mass of SOA (µg) in the extract for ROS/RNS, TNF-α, and IL-6 are shown in Fig. 2, shaped by SOA formation condition. It should be noted that all responses were normalized to probe-treated control cells to account for differences between endogenous levels of ROS/RNS produced in cells (Henkler et al., 2010). Uncertainties associated with AUC were determined by averaging the AUCs obtained by fitting dose-response data with each point removed systematically, following the methodology described in Tuet et al. (2016). ROS/RNS production was also measured for background filters and found to be within the uncertainty of control cells, indicating that there was no evidence for significant H2O2 or HONO uptake onto seed particles (Fig. S1 in the Supplement). Furthermore, exposure to filter extract did not result in decreases in metabolic activity as measured by the MTT assay for all SOA systems investigated (Fig. S2). Since results from MTT may represent the number of viable cells present, changes in inflammatory endpoints did not likely result from changes in the number of cells exposed (i.e., decreases in response cannot be attributed to cell death).

For all inflammatory responses measured (levels of ROS/RNS, TNF-α, and IL-6), SOA precursor identity and formation condition influenced the level of response, as demonstrated by the range of values obtained from different SOA precursors and different formation conditions (Fig. 2). Despite having a clear effect, no obvious trends were observed for each variable (precursor or formation condition) on individual responses. This is in contrast to that observed for the oxidative potential as measured by DTT (OPWS-DTT) for these samples, where only precursor identity influenced OPWS-DTT substantially (Tuet et al., 2017). However, this may not be surprising as DTT is a chemical assay, which only accounts for the potential of species to participate in redox reactions (Cho et al., 2005), whereas cellular assays account for many complicated cellular events involved in intricate positive and negative feedback loops. Due to the considerably different classes of compounds chosen as SOA precursors, aerosol compositional changes between different precursors were generally larger than those between different formation conditions of the same precursor (see Fig. 3a; Tuet et al., 2017). DTT may only be sensitive to larger differences arising from different precursors, whereas cellular assays could also be sensitive to differences between different formation conditions and chemical composition of the same precursor. Moreover, while Tuet et al. (2017) showed that the intrinsic OPWS-DTT spanned a wide range, with isoprene and naphthalene SOA generating the lowest and highest OPWS-DTT, these bounds were less clear for cellular responses. While isoprene and naphthalene SOA still generated the lowest and highest inflammatory responses in general, a few exceptions exist (e.g., ROS/RNS levels induced by pentadecane SOA formed under dry, RO2 + HO2 dominant conditions, Fig. 2).

Though no apparent trends in individual inflammatory responses were observed as a function of SOA precursor identity or formation condition, several patterns among all three inflammatory responses were observed for SOA precursors whose products share similar chemical structures (i.e., similar carbon chain length and functionalities). Exposure to isoprene SOA induced the lowest levels of TNF-α and IL-6 among the aerosol systems studied (Fig. 2). Furthermore, isoprene SOA generated from different pathways (i.e., photooxidation under different RO2 fates and reactive uptake of IEPOX; Surratt et al., 2010; Xu et al., 2014; Chan et al., 2010) produced similar responses for each inflammatory endpoint. These results suggest that different isoprene SOA products (Surratt et al., 2010; Xu et al., 2014; Chan et al., 2010) may induce similarly low inflammatory responses and are consistent with the intrinsic OPWS-DTT obtained for these SOA samples, where isoprene SOA generated the lowest OPWS-DTT of all SOA systems studied and the OPWS-DTT was similar for all SOA formation conditions explored (Tuet et al., 2017). This finding is in contrast to a previous study by Lin et al. (2016), where methacrylic acid epoxide (MAE)-derived SOA was found to be substantially more potent than IEPOX-derived SOA. However, while exposure to MAE-derived SOA induced the upregulation of a larger number of oxidative stress response genes than IEPOX-derived SOA, the fold changes of several genes reported in Lin et al. (2016) are actually similar (e.g., ALOX12, NQO1). Several of these genes directly affect the production of inflammatory cytokines measured in this study. For instance, studies have observed that arachidonate 12-lipoxygenase (ALOX12) products induce the production of both TNF-α and IL-6 in macrophages (Wen et al., 2007). As such, a similar response level regardless of SOA formation condition may be observed depending on the biological endpoints measured. Thus, it is possible that the inflammatory cytokines measured in this study are involved in pathways concerning those genes, resulting in a similar response level regardless of SOA formation condition.

Similarly, exposure to SOA generated from the photooxidation of α-pinene and m-xylene resulted in similar inflammatory responses for all three formation conditions (Fig. 2). These cellular assay results are consistent with results from the DTT assay where the OPWS-DTT was not significantly different between SOA formed under different formation conditions (Tuet et al., 2017). Response levels induced by these two SOA systems are also similar across all three inflammatory measurements investigated (Fig. 2). This suggests that products from both precursors may induce similar cellular pathways resulting in the production of similar levels of inflammatory markers. Indeed, there are several similarities between products formed from the photooxidation of α-pinene and m-xylene. For instance, a large portion of α-pinene and m-xylene oxidation products under both RO2 + HO2 and RO2 + NO pathways are ring-breaking products with a similar carbon chain length (Eddingsaas et al., 2012; Vivanco and Santiago, 2010; Jenkin et al., 2003). As a result of this similarity, products from both SOA systems may interact with the same cellular targets and induce similar cellular pathways, resulting in a similar response regardless of precursor identity and formation condition. These observations further imply that the chemical structures (e.g., carbon chain lengths and functionalities) of oxidation products may be important regardless of PM source or precursor.

A different pattern was observed for β-caryophyllene and pentadecane SOA, where the IL-6 response spanned a much larger range than ROS/RNS and TNF-α (Fig. 2). This is in contrast to the trends observed for the OPWS-DTT for β-caryophyllene and pentadecane SOA, where OPWS-DTT was similar regardless of formation condition (Tuet et al., 2017). This suggests that there are differences between organic peroxides and organic nitrates formed from certain precursors that influence cellular responses, but they are not captured by redox potential measurements. Less is known about the effects of humidity on SOA formation and chemical composition for all SOA systems investigated, as most laboratory chamber studies in literature have been conducted under dry conditions. Specifically here, very high levels of IL-6 were observed post-exposure to pentadecane SOA formed under humid conditions. Prior studies reported opposing findings with some showing a significant effect of water on aerosol formation and chemical composition (Nguyen et al., 2011; Wong et al., 2015; Healy et al., 2009; Stirnweis et al., 2016), while others found little influence (Edney et al., 2000; Boyd et al., 2015; Cocker III et al., 2001). It is clear that humidity effects are highly hydrocarbon-dependent and further studies into the specific products formed under humid conditions are required to understand how these differences in chemical composition may translate to different cellular endpoints. Nonetheless, the known products formed from the photooxidation of these hydrocarbons may provide some insight into the inflammatory responses observed. While there are no prior studies involving pentadecane oxidation products, it is expected that the oxidation products will be similar to those reported in the oxidation of dodecane (i.e., same functionalities with a longer carbon chain; Loza et al., 2014). It is therefore likely that pentadecane oxidation products resemble long chain fatty acids and could potentially insert into the cell membrane (Loza et al., 2014), as previous studies have shown that fatty acids can feasibly insert into the cell membrane bilayer (Khmelinskaia et al., 2014; Cerezo et al., 2011). This insertion could potentially affect membrane fluidity, which is known to affect cell function substantially although the specific effect depends strongly on the particular modification and cell type of interest (Baritaki et al., 2007; Spector and Yorek, 1985). In some cases, these alterations lead to the induction of apoptosis, which involves pathways leading to the production of TNF-α (Baritaki et al., 2007; Wang et al., 2003). TNF-α can then induce the production of IL-6, which once produced can also inhibit the production of TNF-α in a feedback loop (Kishimoto, 2003; Wang et al., 2003). These cellular events are consistent with the observed inflammatory response induced by pentadecane SOA exposure, where there is a high IL-6 response and a lower TNF-α response. The low ROS/RNS response observed is also in line with these cellular events, as IL-6 exhibits anti-inflammatory functions, which can neutralize ROS/RNS production. These responses are less pronounced for β-caryophyllene aerosol, which may be due to the shorter carbon chain observed in known products (Chan et al., 2011). While β-caryophyllene and pentadecane are both C15 precursors, β-caryophyllene is a bicyclic compound and many SOA products retain the four-membered ring, resulting in a shorter carbon backbone (Chan et al., 2011). As a result, fewer products may insert into the cell membrane, leading to a lower response compared to pentadecane SOA exposure. These observations, particularly those for pentadecane SOA, suggest that aerosols from meat cooking may have health implications, as fatty acids comprise a majority of these aerosols (Mohr et al., 2009; Rogge et al., 1991).

Naphthalene exhibits a different, more distinct pattern compared to the rest of the SOA systems investigated, with a large range observed for both TNF-α and IL-6 under different formation conditions (Fig. 2). Higher levels of ROS/RNS were also observed as a result of exposure to naphthalene aerosol irrespective of SOA formation condition. Similarly, the OPWS-DTT of naphthalene SOA previously measured by Tuet et al. (2017) was an outlier among all SOA systems investigated, as the measured OPWS-DTT was at least twice that of the next highest SOA system. These observations are consistent with the formation of specific SOA products such as naphthoquinones, which are known to induce redox cycling in cells and are formed under both RO2 + HO2 and RO2 + NO pathways (Henkler et al., 2010; Kautzman et al., 2010). Consequently, aerosol generated from naphthalene may induce higher levels of inflammatory responses than other SOA due to this process (Henkler et al., 2010; Lorentzen et al., 1979). However, as shown by the high levels of IL-6, exposure to naphthalene SOA may also induce anti-inflammatory pathways not captured by OPWS-DTT measurements. Moreover, a clear increasing trend is apparent for TNF-α and IL-6 produced upon naphthalene SOA exposure, with a higher level of both cytokines observed for aerosol formed under RO2 + NO dominant and humid conditions. Previously, the effect of different RO2 fates on SOA OPWS-DTT was attributed to the different products known to form under both pathways (Tuet et al., 2017). The same explanation applies for cellular measurements as SOA products that promote electron transfer reactions with antioxidants can result in a redox imbalance as measured by OPWS-DTT and the induction of related cellular pathways such as ROS/RNS and cytokine production (Tuet et al., 2017). Finally, naphthalene SOA induced cellular responses outside of those observed for other aerosol systems, with higher levels of all inflammatory markers than other SOA systems. As shown previously for OPWS-DTT, naphthalene may be an outlier due to aromatic-ring-containing products, which may then induce different cellular pathways compared to other aerosol systems investigated, the products of which do not contain aromatic rings. Additionally, many known aerosol products formed from the photooxidation of naphthalene have functionalities that resemble those of dinitrophenol, which is known to decouple phosphorylation from electron transfer (Terada, 1990). It is therefore possible that the aromatic functionality present in the majority of naphthalene SOA products results in the involvement of very different cellular pathways, leading to outlier inflammatory endpoint responses. Various products of naphthalene oxidation such as nitroaromatics and polyaromatics are known to have mutagenic properties and may induce the formation of DNA adducts (Baird et al., 2005; Helmig et al., 1992). As such, it is possible that these products may induce health effects via other pathways as well and naphthalene SOA exposure may have effects beyond redox imbalance and oxidative stress.

Bulk aerosol elemental ratios (O : C, H : C, and N : C) were determined for each SOA system investigated. Different types of organic aerosol are known to span a wide range of O : C ratios, which may be utilized as an indication of oxidation, and the van Krevelen diagram was used to visualize whether changes in O : C and H : C ratios corresponded to changes in levels of inflammatory response (Figs. 3a and S3; Chhabra et al., 2011; Lambe et al., 2011; Ng et al., 2010). Changes in the slope within the van Krevelen space provide information on SOA functionalization (Heald et al., 2010; Van Krevelen, 1950; Ng et al., 2011). Beginning from the precursor hydrocarbon, a slope of 0 indicates alcohol group additions, a slope of -1 indicates carbonyl and alcohol additions on separate carbons or carboxylic acid additions, and a slope of -2 indicates ketone or aldehyde additions.

As seen in Fig. 3a, the laboratory-generated aerosols span a large range of O : C and H : C ratios. Both SOA formation condition and precursor identity influenced elemental ratios; however, precursor identity generally had a larger effect as is evident by the clusters observed for different SOA precursors. Despite these differences in chemical composition, there were no obvious trends between O : C or H : C and any inflammatory endpoint measured. This is similar to that observed for chemical oxidative potential as measured by DTT, where a higher O : C did not correspond to a higher oxidative potential for both laboratory-generated and ambient aerosols (Tuet et al., 2017). This is likely due to the different formation conditions used to generate SOA, which may not be directly comparable. Nevertheless, a significant correlation (p < 0.05) was observed between ROS/RNS and OS‾c (Fig. 3b). This positive correlation is not surprising, as a higher average oxidation state would likely correspond to a better oxidizing agent. Future studies should evaluate the effect of the degree of oxidation for SOA formed from the same SOA precursor under the same formation conditions to investigate whether atmospheric aging of aerosol (which typically leads to increases in the degree of oxidation) affects inflammatory responses. Finally, the N : C ratio was also determined for SOA systems formed under conditions that favor the RO2 + NO pathway (Fig. S4) and were found to span a large range. Similarly, there was no obvious trend between N : C ratios and the inflammatory endpoints measured.

To visualize whether there exists a relationship between inflammatory markers measured, levels of TNF-α and IL-6 are shown in Fig. 4, sized by ROS/RNS. With the exception of naphthalene SOA, the inflammatory cytokine responses for all aerosol systems investigated follow an exponential curve (Fig. 4, shown in black) where there appears to be a plateau for TNF-α levels. Along this curve, ROS/RNS levels also appear to increase with increasing inflammatory cytokine levels to a certain point, after which ROS/RNS levels decrease. These observations are in line with the interconnected effects of both cytokines. While both TNF-α and IL-6 have pro-inflammatory effects that may lead to the increase of ROS/RNS production, the individual pathways are also involved in many complicated stimulation and inhibition loops and there is extensive cross talk between both pathways. For instance, TNF-α induces the production of glucocorticoids, which in turn inhibit both TNF-α and IL-6 production (Wang et al., 2003). IL-6 also directly inhibits the production of TNF-α and other cytokines induced as a result of TNF-α (e.g., IL-1) and stimulates pathways that lead to the production of glucocorticoids (Kishimoto, 2003). As a result, increases in IL-6 may be accompanied by decreases in TNF-α, resulting in the observed plateau. Furthermore, ROS/RNS levels may represent a fine balance between anti-inflammatory and pro-inflammatory effects. Both cytokines are involved in the acute-phase reaction and can affect ROS/RNS levels via pro-inflammatory pathways. IL-6 also exhibits some anti-inflammatory functions and may thus lower ROS/RNS levels as well. These interconnected pathways could account for the observed parabolic pattern for ROS/RNS production. Exposure to naphthalene SOA resulted in responses outside of those observed for other aerosol systems, likely due to the formation of aromatic-ring-retaining products as discussed in the previous section.

To evaluate how the oxidative potential and ROS/RNS production of the SOA systems investigated compare in the context of ambient samples, the measurements obtained in this study were plotted with those obtained in our previous study involving ambient samples collected around the greater Atlanta area (Fig. 5; Tuet et al., 2016). These ambient samples were analyzed using the same methods for determining oxidative potential (DTT assay; Cho et al., 2005; Fang et al., 2015b) and ROS/RNS production (cellular carboxy-H2DCFDA assay; Tuet et al., 2016). Furthermore, the same extraction protocol (water-soluble extract) was followed in both studies (Tuet et al., 2016). Results from both studies are therefore directly comparable. Previously, a significant correlation between ROS/RNS production and oxidative potential as measured by DTT was observed for summer ambient samples. In the same study, correlations between ROS/RNS production and organic species were also observed for summer ambient samples, and it was suggested that these correlations may reflect contributions from photochemically produced SOA (Tuet et al., 2016).

Figure 5 shows that laboratory-generated SOA oxidative potential is comparable to that observed in ambient samples, with the exception of naphthalene SOA, which produced higher DTT activities due to its aromatic-ring-retaining products (Tuet et al., 2017; Kautzman et al., 2010). Laboratory-generated SOA also induced similar or higher levels of ROS/RNS compared to ambient samples. There are many possible explanations for the observed higher response for some SOA samples. For instance, individual, single-precursor SOA systems were considered in this study, whereas ambient aerosol contains SOA from multiple precursors as well as other species that are not considered in this study (e.g., metals). Interactions between SOA from different precursors is likely to occur and may result in different response levels. Complex interactions between SOA and other species present in the ambient (e.g., metals or other organic species) are also likely involved (Tuet et al., 2016). Previous studies have also suggested the possibility of metal–organic complexes. For instance, Verma et al. (2012) showed that certain metals were retained on a C18 column, which is utilized to remove hydrophobic components, suggesting that these metals were likely complexed and removed in the process. Further chamber studies involving photochemically generated SOA and metals may elucidate these interactions. Furthermore, there are likely species present in the ambient that do not contribute to ROS/RNS production. That is, while certain species contribute to the mass of PM, there is little to no ROS/RNS production associated with these species. Ambient samples where these species comprise a significant fraction will have a low per mass ROS/RNS production level. Finally, only three SOA formation conditions were investigated in this study. There are multiple other possible oxidation mechanisms that lead to the formation of SOA in the ambient, which were not accounted for in this study. Nonetheless, despite the low ROS/RNS levels observed post SOA exposure, there is an association between ROS/RNS production and DTT activity (Fig. 5). These results suggest that our previous findings based on ambient filter samples may be extended to SOA samples. That is, while the relationship between ROS/RNS production and DTT activity is complex, DTT may serve as a useful screening tool as samples with low DTT activities are likely to produce low levels of ROS/RNS (Tuet et al., 2016).