Hydroxyacetone sulfate and glycolic acid sulfate (potassium salts, >95 % purity) were synthesized according to Hettiyadura et al. (2015); lactic acid sulfate (24.9 % purity) was synthesized according to Olson et al. (2011); methyltetrol sulfates was synthesized according to Budisulistiorini et al. (2015) and Bondy et al. (2018). Ultra-pure water was prepared on site (Thermo, Barnsted Easypure-II; 18.2 MΩcm-1 resistivity, with total organic carbon (OC) <40 µgL-1). Other reagents include acetonitrile (Optima^™, Fisher Scientific), ammonium acetate (≥99 %, Fluka, Sigma Aldrich), and ammonium hydroxide (Optima, Fisher Scientific).

PM2.5 samples were collected in Atlanta, GA, from 29 July to 27 August 2015. A medium volume sampler (3000B, URG Corp.) operating at a flow rate of 90 Lmin-1 was used to collect PM2.5 on pre-baked (550 ∘C for 18 h) quartz-fiber filters (90 mm, Pallflex^® Tissuquartz^™, Pall life science). The PM2.5 sampler was placed on the roof top of the School of Earth and Atmospheric Sciences building at the Georgia Institute of Technology (33∘46′44.2′′ N, 84∘23′46.2′′ W; height ∼30–40 m). A detailed description of the sampling site is provided by Verma et al. (2014). Samples were collected daily from 13:30 to 12:30 the next day (local time). One filter blank was collected for every five PM2.5 samples. Samples from 29 July, 3, 11, and 19 August were not analyzed for organosulfates, as the filters were used for a different purpose. The collected samples were placed in aluminium-lined (pre-baked at 550 ∘C for 18 h) petri dishes, sealed with Teflon tape, and stored at -20 ∘C until extracted.

Organosulfates were extracted according to the method described in Hettiyadura et al. (2015) that has been demonstrated to efficiently recover 83 %–121 % of organosulfates with aliphatic, aromatic, carbonyl, hydroxyl, and carboxyl acid groups. Briefly, subsamples of filters (averaging ∼ 3 cm2) were extracted with 10.0 mL of acetonitrile and ultra-pure water (95:5, by volume) for 20 min by ultra-sonication (5510, Branson). The sample extracts were filtered using polypropylene membrane syringe filter discs (0.45 µm pore size, Puradisc^™ 25 PP, Whatman^®). The extracts were evaporated to dryness under ultra-high purity nitrogen gas at 50 ∘C (Turbovap^® LV, Caliper Life Sciences, Reacti-Therm III TS 18824, and Reacti-Vap I 18825, Thermo Scientific). Dried extracts were reconstituted in 600 µL of acetonitrile and ultra-pure water (95:5 by volume).

Organosulfates were quantified using HILIC and tandem mass spectrometry (MS/MS) in negative (-) ion mode, using an ultra-performance liquid chromatography system (UPLC, ACQUITY UPLC H-Class, Waters) coupled with a triple quadrupole (TQ) mass spectrometer (AQCUITY, Waters) and an electrospray ionization (ESI) source. The separation of organosulfates was performed on an ethylene-bridged hybrid amide column using an acetonitrile-rich mobile phase (acetonitrile and ultra-pure water; 95:5) and an aqueous mobile phase (ultra-pure water; 100 %). Both mobile phases were buffered at pH 9 with 10 mM ammonium acetate and ammonium hydroxide. Organosulfates were eluted using a stepwise gradient as described in Hettiyadura et al. (2015). Briefly, the acetonitrile-rich mobile phase was held at 100 % from 0 to 2 min, and then decreased to 85 % from 2 to 4 min and held constant at 85 % until 11 min. Targeted analysis was performed in multiple reaction monitoring mode. Hydroxyacetone sulfate and glycolic acid sulfate were quantified using authentic standards. Lactic acid sulfate and methyltetrol sulfate were quantified using their response factors determined previously using authentic standards. Notably, these prior experiments had response factors (determined as the slope of the calibration curve) for glycolic acid sulfate and hydroxyacetone sulfate that were within 10 % of the current experiments, indicating that instrument performance and ionization were consistent within 10 %. The optimized ESI(-)-MS/MS conditions used for each of these organosulfates are given in Hettiyadura et al. (2015, 2018), respectively.

Semi-quantitation of other organosulfates was based upon the MS/MS response of authentic standards and matched to the sulfur-containing fragment ions observed. For semi-quantitation of organosulfates that fragmented to the bisulfate anion (m/z 97, Fig. 1a), one of three surrogate standards were used: for m/z 211, 213, and 260 the response factor of methyltetrol sulfate was used; for other organosulfates eluting prior to 4 min hydroxyacetone sulfate was used, and for those retaining more than 4 min glycolic acid sulfate was used. For the semi-quantitation of organosulfates that fragmented only to the sulfate radical anion (m/z 96, Fig. 1b), methyl sulfate was used. For organosulfates with m/z 137, 139, and 296 that fragmented to the sulfite radical anion (m/z 80, Fig. 1d) hydroxyacetone sulfate was used. The cone voltage and collision energy used for the organosulfates that were semi-quantified using surrogate standards were the same as the ESI(-)-MS conditions used for corresponding precursor ion scans (given in Sect. 2.5.1). The uncertainty of the organosulfate concentrations was calculated accounting for relative errors in air volume, extraction efficiency, and instrumental analysis according to the method described in Hettiyadura et al. (2017). The relative error in the instrument analysis was propagated using the limit of detection and the relative standard deviation for each organosulfate standard given in Hettiyadura et al. (2015). For methyltetrol sulfate and the other organosulfates that did not have authentic standards, the analytical uncertainty was estimated as 30 % of their concentration values (Hettiyadura et al., 2018). This uncertainty does not account for any bias introduced by the use of a surrogate standard, which can only be evaluated by using an authentic standard. All data were acquired and analyzed using MassLynx and QuanLynx softwares (Waters Inc., version 4.1).

Percent contributions of organosulfates to PM2.5 OC were compared to determine the relative abundances of the major organosulfates in Atlanta and Centreville. OC in the PM2.5 samples was measured on 1 cm2 filter punches using a thermal–optical analyzer (Sunset laboratory) according to Schauer et al. (2003). Filter-based measurements of other PM2.5 components, gas-phase measurements, and meteorological conditions were used in correlation analysis to provide insight to precursors and formation pathways of organosulfates. Isoprene SOA tracers (2-methylthreitol, 2-methylerythritol, 2-methylglyceric acid, cis-2-methyl-1,3,4-trihydroxy-1-butadiene, 3-methyl-2,3,4-trihydroxy-1-butene, and trans-2-methyl-1,3,4-trihydroxy-1-butene), cis-pinonic acid, β-caryophyllinic acid, meso-erythritol, 2,3-dihydroxy-4-oxopentanoic acid, aromatic dicarboxylic acids (phthalic acid, terephthalic acid, isophthalic acid, and 4-methylphthalic acid), and mononitroaromatic compounds (4-nitrophenol, 2-methyl-4-nitrophenol, 4-methyl-2-nitrophenol, 4-nitrocatechol, 3-methyl-6-nitrocatechol, and 3-methyl-5-nitrocatechol) were measured by gas chromatography (GC)-MS according to the methods described in Al-Naiema and Stone (2017). Sulfate was measured by ion chromatography following Jayarathne et al. (2014). The hourly based measurements of O3, NOx (nitrogen oxides such as, NO and NO2), and solar radiation were obtained from the Southeastern Aerosol and Research Characterization network monitoring site at Jefferson Street (JST) located 2 km west of the sampling site and were averaged across the sample collection time. Detailed descriptions of their quantification methods are described in Hansen et al. (2003).

Pearson correlation coefficients were assessed using a statistical analysis software (IBM^® SPSS^® statistics, version 25). Correlations were interpreted as very strong (0.9–1.0), strong (0.7–0.9), moderate (0.5–0.7), weak (0.3–0.5), or negligible (0.0–0.3) (Hinkle et al., 2003). The correlations were considered as statistically significant at the 95 % confidence level.

Quantitative information about the organosulfates observed in Atlanta is summarized in Table 1, with time series of selected species shown in Fig. 2. Methyltetrol sulfate is the most abundant quantified organosulfate, contributing 12.6 % of PM2.5 OC, followed by m/z 211 (0.93 %), 213 (0.80 %), glycolic acid sulfate (0.24 %), 2-methylglyceric acid sulfate (0.32 %), and lactic acid sulfate (0.20 %) (Table 2). The remaining 26 organosulfates were estimated to contribute 1 % of PM2.5 OC. Altogether, the 32 measured organosulfates in Table 1 account for 16.5 % of PM2.5 OC. These results indicate that organosulfates in Atlanta during August 2015 were dominated by methyltetrol sulfate, with minor contributions from many other organosulfate species derived from isoprene, monoterpenes, and anthropogenic sources.

Organosulfates were identified by precursors to m/z 97 (HSO4-), 96 (SO4-⚫), 81 (HSO3-), and 80 (SO3-⚫) in three PM2.5 samples collected on 30–31 July and 1 August 2015. Results were similar for all three samples; therefore, the results obtained only for the 30 July sample are shown in Fig. 1. Major organosulfur compounds were defined in one of the two following ways: (1) as having a minimum relative intensity in the MS/MS spectra (≥1.0 % for precursors to m/z 97, >12 % for m/z 96, >5 % for m/z 81, and >3 % for m/z 80 in any of the three samples) or (2) by retaining more than 4 min. Despite the observation that organosulfates eluting after 4 min often have higher concentrations than early eluting species, their MS response is observed to be lower because of the increased water content of the mobile phase as water does not desolvate as efficiently as acetonitrile in the ESI source (Hettiyadura et al., 2017). The absolute MS signal for precursors to m/z 97 was 52, 10, and 8 times greater than MS signals for precursors to m/z 96, 81, and 80, respectively; however, due to differing ionization efficiencies and stabilities among these fragment ions, the strength of the MS signal is not indicative of the relative concentrations of species that form these fragments. Table 1 summarizes the major organosulfates' elemental composition, monoisotopic mass, proposed or known structures, and precursor gases. Of the major organosulfates, 26 of the 32 consisted of C, H, O, and S, while 6 of 32 consisted of C, H, O, S, and N. Structures were proposed based on elemental composition, double bond equivalence (DBE), retention time, and prior studies.

The strongest organosulfate signals observed in m/z 97, 80, and 81 precursor ion scans are associated with isoprene (Fig. 1, Table 1). Methyltetrol sulfate (m/z 215), the most abundant organosulfate observed, is produced from the acid catalyzed nucleophilic addition of sulfate to IEPOX ring (Surratt et al., 2010). Organosulfates with m/z 211 (hydroxymethyl-tetrahydrofuranone sulfates) and 213 (dihydroxymethyl-tetrahydrofuranyl sulfates), of the next-highest abundance, have been observed during the photooxidation of isoprene (Surratt et al., 2008) and are suggested to derive from the oxidation of primary alcohols in methyltetrol sulfates (Hettiyadura et al., 2015). In addition, 14 other major organosulfates identified are known to derive from isoprene and isoprene oxidation products (Table 1). Many of these organosulfates have also been identified as SOA products from diesel and biodiesel fuel emissions (e.g., 2-methylglyceric acid sulfate, lactic acid sulfate, hydroxyacetone sulfate, m/z 167, 183, 197, 211, 213, 237, 239, and 253) (Blair et al., 2017), monoterpenes (m/z 239 and 253) (Surratt et al., 2008), and/or MBO (199; C5H11SO6-) (Zhang et al., 2012). However, their moderate to strong correlations with methyltetrol sulfate (Table S1 in the Supplement) and 2-methyltetrols (Table S2) suggest that they are mainly derived from isoprene.

Among the major organosulfate signals are those associated with isoprene oxidation under high-NOx conditions such as 2-methylglyceric acid sulfate, m/z 260 and 274. 2-Methylglyceric acid sulfate is a tracer for isoprene high-NOx SOA that is formed by the acid-catalyzed nucleophilic addition of sulfate to methacrylic acid epoxide (MAE) and/or hydroxymethyl-methyl-α-lactone (HMML) (Lin et al., 2013). The organosulfate with m/z 260 is a nitrooxy organosulfate that derives from the photooxidation of isoprene under high-NOx conditions (Surratt et al., 2008; Gómez-González et al., 2008). Two isomers of m/z 260 were identified in this study, while up to four isomers of m/z 260 were reported in Centreville (Surratt et al., 2008). The m/z 260 also correlated moderately with methyltetrol sulfate (r=0.539, p value =0.005, Table S1), supporting its formation from isoprene. The organosulfate with m/z 274 is also a nitrooxy organosulfate that is derived from isoprene photooxidation under high-NOx conditions (Nestorowicz et al., 2018). The organosulfate with m/z 274 has multiple isomers, whereas only the two isomers retaining greater than 4 min are considered to be major ones as described in Sect. 3.2 (Fig. 3o). Their longer retention times (5.6 and 5.8 min), three additional oxygen atoms, and one unit of unsaturation suggest the presence of a carboxylic acid functional group and a hydroxyl group. Plausible structures for these two organosulfates are diastereomers of 2-carboxy-3-hydroxy-4-(nitrooxy)butan-2-yl sulfate (Table 1), which could form by the oxidation of a primary hydroxyl group in 1,3-dihydroxy-2-methyl-4-(nitrooxy)butan-2-yl sulfate (an isomer of m/z 260, C5H10SO9-, proposed by Darer et al., 2011) to a carboxylic acid. The strong correlation of these two signals at m/z 274 with the less-oxidized isoprene nitrooxy organosulfate (m/z 260) (r=0.860, p value <0.001, Table S1) supports this prediction. Overall, these results indicate that isoprene is the major precursor of the most abundant organosulfates in this study.

Isoprene-derived organosulfates explain a significant fraction of isoprene-derived organic aerosol observed in Atlanta that had not previously been identified on a molecular level. By factor analysis of aerosol chemical speciation data (using the multilinear engine – ME-2), IEPOX-derived OA was estimated to account for 29 % (3.3 µgm-3) of PM1 OA at the nearby JST monitoring site in summer 2014, whereas the IEPOX-OA tracers measured in PM2.5 (2-methyltetrols, C5-alkene triols, and 3-methyl-hydrofuran-3,4-diols) accounted for 3 % of PM1 OA (Rattanavaraha et al., 2017), assuming negligible differences between PM1 and PM2.5. The remaining IEPOX-derived OA corresponded to 10 %–18 % of PM1 OC (considering an OM:OC ratio of 2.05±0.57) (Xu et al., 2017), and is comparable to the contribution of isoprene-derived organosulfates to PM2.5 OC in this study (15.7 %). Additionally, the isoprene-derived organosulfates observed in this study account for more than half of the PM2.5 secondary organic carbon coming from isoprene, which is estimated as 27 % following the SOA tracer method (Al-Naiema et al., 2019; Kleindienst et al., 2007). These results indicate that more than half of the isoprene-derived OA in Atlanta during summer is comprised by organosulfates, mainly methyltetrol sulfate.

Seven of the thirty-two major organosulfates identified in Atlanta (Table 1) were previously detected among the SOA produced from monoterpenes in the presence of NOx and acidic sulfate seed aerosols (Surratt et al., 2008). Of these, nitrooxy organosulfates at m/z 342, 294, and 296 are derived from monoterpenes either by photooxidation in the presence of NOx or from nitrate radical-initiated oxidation (Surratt et al., 2008; Iinuma et al., 2007). The estimated contribution of these seven monoterpene-derived organosulfates is 0.5 % of PM2.5 OC. However, the accuracy of this estimate is limited by the lack of authentic standards for monoterpene organosulfates and the large differences in molecular structure between the monoterpene organosulfates and the standards utilized in this study. The absence of significant correlations among nitrooxy organosulfates with other organosulfates (Table S1) and biogenic SOA tracers that predominantly derive from photooxidation reactions (Table S2) suggest that these nitrooxy organosulfates likely formed by nitrate radical-initiated oxidation. Organosulfates with m/z 223, 279, and 281 have been identified as SOA products of α-pinene, as well as from other monoterpenes (m/z 279 and 281), in the presence of NOx and highly acidic sulfate seed aerosol (Surratt et al., 2008). The organosulfate with m/z 251 has been identified in SOA from the photooxidation of β-caryophyllene (a sesquiterpene) and limonene (a monoterpene) in the presence of NOx and sulfate seed aerosols (Chan et al., 2011; Surratt et al., 2008). These species did not correlate with β-caryophyllinic acid (Table S2), a SOA tracer for β-caryophyllene formed under high-NOx conditions (Jaoui et al., 2007), suggesting that m/z 251 mainly forms from monoterpenes. Organosulfates with the same m/z were also detected among the organosulfates generated from diesel and biodiesel fuel emissions (Blair et al., 2017) and photooxidation of n-alkanes such as decaline (m/z 281) and cyclodecane (m/z 279 and 281) (Riva et al., 2016b), but these species are expected to be biogenic in nature due to the dominance of biogenic VOC in Atlanta during summer (Geron et al., 1995; Al-Naiema et al., 2019; Rattanavaraha et al., 2017).

Five organosulfates that were previously reported only in photooxidation of diesel and/or biodiesel fuel in the presence of SO2 were identified among the thirty-two major organosulfates. These include m/z 137 and 151 that were generated from diesel fuel emissions and m/z 195, 209, and 265 that were generated from both diesel and biodiesel emissions (Blair et al., 2017). The organosulfate with m/z 265 corresponds to dodecyl sulfate, a widely used surfactant in detergents that can also come from wastewater treatment plants (Hettiyadura et al., 2017). The concentrations of m/z 209 and 195 are at least 3 times higher compared with other organosulfates derived from diesel and/or biodiesel emissions in this study (Table 1). These organosulfates (m/z 209 and 195) were also detected with a high abundance in urban Shanghai and Los Angeles (Tao et al., 2014). The organosulfates with m/z 209 and 195 are homologs, differing by one methylene. Both compounds have two units of unsaturation and two additional oxygen atoms. Further, their retention times (Fig. 3f and h), which were less than 1 min, suggest that they do not contain a carboxylic acid group, but may contain two carbonyl groups. Additional work is required to determine the position of carbonyl and sulfate groups in these compounds. As m/z 209 and 195 are highly abundant in other urban locations and are only known to derive from diesel and/or biodiesel fuel, they may be useful as tracers for SOA derived from diesel and biodiesel emissions.

Aromatic sulfur-containing compounds were not detected among the major organosulfate species (Table 1), although some were observed by HILIC-ToF. Two sulfur-containing compounds had large DBEs indicating aromatic groups: m/z 185 (tR 1.06 min, C7H5SO4-, DBE 5.5, error 3.7 mDa) and 201 (tR 7.56 and 8.17 min, C7H5SO5-, DBE 5.5, error 3.5 mDa). The MS data matched the molecular formula reported by Riva et al. (2015), who detected m/z 185 in naphthalene and 2-methylnaphthalene photooxidation experiments and identified it as formylbenzenesulfonate by MS fragmentation. Riva et al. (2015) also reported m/z 201 in SOA generated by the photooxidation of 2-methylnaphthalene and identified it as 4-sulfobenzoic acid using an authentic standard. In the Atlanta PM2.5, two isomers of m/z 201, likely conformational isomers of 4-sulfobenzoic acid, are observed. The presence of a carboxylic acid group in m/z 201 is evident by the retention time >7 min in the HILIC method (Hettiyadura et al., 2015). None of the aromatic organosulfates reported in Staudt et al. (2014) (phenyl sulfates and benzyl sulfates) were detected in HILIC-ToF. This may be due to the lower retention times and higher detection limits for aromatic organosulfates in HILIC compared to reversed-phase LC (Hettiyadura et al., 2015). These results suggest that aromatic organosulfates have low PM2.5 concentrations in comparison to biogenic organosulfates in Atlanta during the summertime.

Three organosulfates that have not been previously reported in laboratory smog chamber experiments were detected among the major organosulfate signals: m/z 155 (C3H7SO4-), 165, and 242. These signals were previously detected in PM2.5 in Centreville, AL (Hettiyadura et al., 2017), while new insights to their possible precursors and structures are gained here. The species with m/z 155 was previously identified as a mono-hydroxy propyl sulfate (Hettiyadura et al., 2017); in Atlanta, it correlated with most of the isoprene-derived organosulfates (Table S1), suggesting that it was derived from isoprene.

The organosulfate at m/z 165 has an elemental composition of C4H5SO5-, indicating the presence of sulfate, an additional oxygenated functional group, and two DBEs. The ToF chromatograms (Fig. 3d) indicate two isomers of m/z 165 that eluted in less than 2 min. While both isomers fragmented to m/z 80, only the first isomer fragmented into m/z 96, which was quantified. Its elemental composition and DBE suggest a dihydrofuran ring structure (Table 1). The strong correlations of m/z 165 with methyltetrol sulfate (r=0.720, p value <0.001; Table S1) and 2-methyltetrols (r=0.670 and 0.768, p value <0.001; Table S2) suggest that it is also derived from isoprene.

The organosulfate at m/z 242 has an elemental composition of C5H8NSO8-, indicating the presence of sulfate, nitrooxy, an additional oxygenated functional group, and two DBEs. Its short retention time of 0.5 min (Fig. 3k) suggests that it contains a carbonyl group as organosulfates with hydroxyl and carboxylate groups retain more than 1 and 4 min, respectively (Hettiyadura et al., 2015, 2017). A possible formation pathway for this nitrooxy organosulfate can be loss of a water molecule from 2,3-dihydroxy-3-methyl-4-(nitrooxy)butyl sulfate (an isomer of m/z 260, C5H10SO9-, proposed by Gómez-González et al., 2008) forming an enol that tautomerizes to a carbonyl forming 3-methyl-4-(nitrooxy)-2-oxobutyl sulfate (Table 1). Only a few atmospherically relevant isoprene-derived nitrooxy organosulfates have been identified in previous studies. These include m/z 244, 260, 274, and 305 that are derived from isoprene photooxidation under high-NOx conditions (Surratt et al., 2008; Gómez-González et al., 2012). It is expected that m/z 242 is an additional nitrooxy organosulfate that has not been previously identified in isoprene photooxidation experiments. As m/z 242 nitrooxy organosulfate is expected to derive from m/z 260, it may provide insight to the atmospheric aging of isoprene-derived SOA, although further evaluation is needed.

To better understand the extent to which anthropogenic pollutants influenced biogenic SOA formation in urban Atlanta during August 2015, the concentrations of the major organosulfates were compared to those measured in rural Centreville, AL, during summer 2013 analyzed by similar methodology (Hettiyadura et al., 2017). Although the major organosulfates identified at both sites were similar and mainly derived from isoprene, their concentrations were 2–6 times higher in Atlanta than in Centreville, with the greatest enhancement obtained for 2-methylglyceric acid sulfate (Table 2). As the absolute concentrations of these organosulfates vary with time due to changes in meteorology, which affect isoprene emissions, transport, and mixing of biogenic and anthropogenic pollutants, their relative contributions to PM2.5 OC were compared across the two sites (Table 2). In total, 12 organosulfates quantified or semi-quantified in both studies contributed 7 % of PM2.5 OC in Centreville, and 16 % in Atlanta. These 12 organosulfates accounted for 95 % of the total organosulfate mass in Atlanta and 58 %–78 % of the total bisulfate ion signal in Centreville (Hettiyadura et al., 2017), indicating that these were the dominant species at both sites. Similarly, the IEPOX-OA in Atlanta during August 2012 (31 % of PM1 OA) was ∼ 2 times greater than IEPOX-OA in Centreville in summer 2013 (18 % of PM1 OA) (Xu et al., 2015a, b). Overall, these results suggest isoprene SOA is 2 times higher in Atlanta compared with Centreville during summer.

Correlations of major organosulfate species were examined at both the Atlanta and Centreville sites to gain insight into their sources and formation pathways. Organosulfates at both sites show moderate to strong correlations with isoprene, isoprene oxidation products, and/or isoprene SOA tracers (Table S2; Table S6 in Hettiyadura et al., 2018), supporting that they mainly derive from isoprene. The correlations of inorganic sulfate with most of the organosulfates were weak or negligible in Atlanta (Table S4), but were moderate to strong in Centreville (r=0.5–0.8) (Table S6 in Hettiyadura et al., 2018). This is likely due to the consistently high levels of sulfate observed in urban Atlanta (ranging from 0.82 to 3.24 µgm-3 and averaging 1.70±0.58 µgm-3) compared with more variable sulfate concentrations in rural Centreville (ranging from 0.42 to 4.17 µgm-3 and averaging 1.78±0.81 µgm-3) (Hettiyadura et al., 2017). Overall these results suggest isoprene and sulfate are important factors influencing the organosulfate formation in both urban Atlanta and rural Centreville.

Isoprene-derived organosulfates indicated a stronger influence of NOx on their formation in Atlanta compared to Centreville. NOx influence is evident by the elevated levels of high-NOx isoprene oxidation products such as 2-methylglyceric acid sulfate, which was 6 times higher in Atlanta than in Centreville, and the isoprene-derived nitrooxy organosulfate at m/z 260 being the eighth strongest organosulfate signal in Atlanta. These results are consistent with the average NOx concentration in urban Atlanta in August 2015 (10.5 ppb) that was 15 times greater than the average NOx concentration in rural Centreville during summer 2013 (0.7 ppb) (SOAS, 2013). Methyltetrol sulfate, the most abundant organosulfate at both sites, is thus expected to derive from low-NOx oxidation pathway in Centreville as described in Surratt et al. (2010) and by high-NOx oxidation pathway in Atlanta as described in Jacobs et al. (2014). The moderate and strong correlations obtained for isoprene-derived organosulfates with high-NOx SOA products (Table S3) such as meso-erythritol (Angove et al., 2006) and nitroaromatic compounds (Al-Naiema and Stone, 2017), as well as with ozone (Table S4) that is formed by the photochemical reactions of NOx and VOC (Blanchard et al., 2014), also support that NOx plays a key role in isoprene-derived organosulfate formation in Atlanta. However, organosulfate formation from ozonolysis (Riva et al., 2016a) cannot be ruled out. While these findings are consistent with other studies that indicate a substantial influence of anthropogenic SO2 and NOx on biogenic SOA formation in the southeastern US during summer (Rattanavaraha et al., 2016; Xu et al., 2015a), this study provides evidence for a greater influence of NOx on isoprene-SOA formation in urban Atlanta, GA, compared to rural Centreville, AL, in summer.