This is the accepted version of the article: Rivas-Ubach, Albert; Liu, Gina; Steiner, Allison L.; [et al.]. «Atmo- ecometabolomics : a novel atmospheric particle chemical characterization methodology for ecological research». Environmental monitoring and assessment, Vol. 191, issue 2 (Feb. 2019), art. 78. DOI 10.1007/s10661-019-7205-x This version is avaible at https://ddd.uab.cat/record/218187 under the terms of the license 1 Atmo-ecometabolomics: a novel atmospheric particle chemical characterization 2 methodology for ecological research 3 4 Albert Rivas-Ubach1, Yina Liu1,2, Allison L. Steiner3, Jordi Sardans4,5, Malak M. Tfaily1, Gourihar 5 Kulkarni6, Young-Mo Kim7, Eric Bourrianne8, Ljiljana Paša-Tolić 1, Josep Peñuelas4,5, Alex 6 Guenther9. 7 8 1. Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, 9 99354, WA, USA. 10 2. Geochemical and Environmental Research Group, Texas A&M University, College Station, 77845, TX, 11 USA. 12 3. Department of Climate and Space Sciences and Engineering, University of Michigan, Ann Arbor, 13 48109, MI, USA. 14 4. CREAF, Campus UAB, Cerdanyola del Vallès, 08913 Catalonia, Spain. 15 5. Global Ecology Unit CREAF-CSIC, Campus UAB, Cerdanyola del Vallès, 08913 Catalonia, Spain. 16 6. Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, 17 99352, WA, USA. 18 7. Biological Sciences Division, Pacific Northwest National Laboratory, Richland, 99354, WA, USA. 19 8. Faculté des Sicences et d’Ingénierie, Université de Toulouse III Paul Sabatier, Toulouse, 31400, France. 20 9. Department of Earth System Science, University of California, Irvine, 92697, CA, USA. 21 22 Author of correspondence: 23 Albert Rivas-Ubach 24 Environmental Molecular Sciences Laboratory, 25 Pacific Northwest National Laboratory, 26 Richland, WA, USA, 99354 27 Tel: 971 319 5962 28 e-mail: albert.rivas.ubach@pnnl.gov // albert.rivas.ubach@gmail.com 29 30 31 32 33 34 35 36 1 37 Abstract. 38 39 Aerosol particles play important roles in processes controlling the composition of the 40 atmosphere and function of ecosystems. A better understanding on the composition of 41 aerosol particles is beginning to be recognized as critical for ecological research to further 42 comprehend the link between aerosols and ecosystems. Whilechemical characterization of 43 aerosols has been practiced in the atmospheric science community, detailed methodology 44 tailored to the needs of ecological research does not exist yet. In this study, we describe an 45 efficient methodology (atmo-ecometabolomics), in step-by-step details, from the sampling to 46 the data analyses, to characterize the chemical composition of aerosols particles, namely 47 atmo-metabolome. This method employs mass spectrometry platforms such as liquid and gas 48 chromatography mass spectrometries (MS), and Fourier transform ion cyclotron resonance MS 49 (FT-ICR-MS). For methodology evaluation, we analyzed aerosol particles collected during two 50 different seasons (spring and summer) in a low biological activity ecosystem. Additionally, to 51 further validate our methodology, we analyzed aerosol particlescollected in a more biological 52 active ecosystem during the pollination peaks of three different representative tree species. 53 Our statistical results showed that our sampling and extraction methods are suitable for 54 characterizing the atmo-ecometabolomes in these two distinct ecosystems with any of the 55 analytical platforms. Datasets obtained from each mass spectrometry instrument showed 56 overall significant differences of the atmo-ecometabolomes between spring and summer as 57 well asbetween the three polliation peak periods. Furthermore, we have identified several 58 metabolites that can be attributed to pollen and other plant-related aerosol particles. We 59 additionally provide a basic guide of the potential use ecometabolomics techniques on 60 different mass spectrometry platforms to accurately analyze the atmo-ecometabolomes for 61 ecological studies. Our method represents an advanced novel approach for future studies in 62 the impact of aerosol particle chemical compositions on ecosystem structure and function and 63 biogeochemistry. 64 65 Keywords: Aerosol particles, metabolomics, ecosystems, biomarkers, mass spectrometry, FT- 66 ICR 67 68 69 70 71 2 72 1. Introduction 73 Aerosols are solids and/or liquids suspended in the atmosphere that are derived from both 74 anthropogenic and biogenic sources (Canagaratna et al. 2007). Primary biogenic aerosol 75 particles (PBAP), which include include pollen, microorganisms, spores, as well as fragments 76 from animal and plant debris, are directly released from biological systems (Després et al. 77 2012). Primary producers also generate large volumes of gas-phase volatile organic 78 compounds (VOCs), which are emitted into the atmosphere and together with anthropogenic 79 gases, are oxidized and can condense to form secondary organic aerosols (SOA)(Després et al. 80 2012; Fuzzi et al. 2006; Pandis et al. 1992) (Figure 1). 81 To date, most research has focused on the impact of aerosols in climate and 82 atmospheric processes (Andreae and Crutzen 1997; Ayers and Gras 1991; Baustian et al. 2012; 83 Carlton et al. 2010; Després et al. 2012; Jokinen et al. 2015; Ramanathan et al. 2001; Zhang et 84 al. 2004). However, many components of the biosphere are constantly in contact with aerosol 85 particles. This aerosol-biosphere interaction can play critical roles in the structure and function 86 of aquatic and terrestrial ecosystems (Baker et al. 2003; Gu et al. 2002; Mahowald et al. 2005; 87 Seco et al. 2007). For example, it is wellknown that plants can absorb deposited particles from 88 the atmosphere (Fageria et al. 2009; Seco et al. 2007; Uzu et al. 2010; Wedding et al. 1975), 89 although the effects of plant particle uptake is still not totally understood.Furthermore, most 90 of such research were focused on trace metals (Achotegui-Castells et al. 2013; Feng et al. 91 2011; Uzu et al. 2010; Xiong et al. 2014) and other nutrients that can play significant roles in 92 agroecosystems (Fernández and Brown 2013). In natural ecosystems, aerosol particles can 93 serve as an important source of nutrients to diverse components of the biosphere. For 94 example, the phyllosphere, which is the microbial community coexisting on plant leaves and 95 have a close relationship with the plysiology of plants (Arnold et al. 2000; Lindow and Brandl 96 2003; Vorholt 2012), can benefit from aerosol particles to acquire nutrients (e.g. nitrogen (N), 97 phosphorous (P), sulphur (S), …). Variations in aerosol particle chemical composition could lead 98 to significant changes on microbial abundance and diversity, affecting their host physiology, 99 and, therefore, the ecosystem structure and function (Peñuelas and Terradas 2014; Whipps et 100 al. 2008). In aquatic ecosystems, the nutritional effects of aerosol deposition for 101 phytoplantkon it has been studied in great extent (Baker et al. 2003; Paerl 1997; Paytan et al. 102 2009; Wang et al. 2015), as itrepresents an important fraction of nutrients for several aquatic 103 primary producers (e.g. cyanobacteria) (Baker et al. 2003; Wang et al. 2015). Furthermore, 104 research in the field of ecological stoichiometry, the study onthe essential elements balance on 105 organisms and ecosystems have proven that changes in nutrient proportions, mainly carbon 106 (C), N P, can lead to substantial alterations in ecosystems (Elser et al. 1996; Sterner and Elser 3 107 2002). Therefore, changes in aerosol particle elemental composition may lead to significant 108 modifications in terrestrial and aquatic ecosystems functioning (Carnicer et al. 2015; Peñuelas 109 et al. 2013; Sardans et al. 2012). Understanding the elemental and molecular compositions of 110 aerosol particles is thus crucial for comprehending processes occuring at the atmosphere- 111 ecosystem interface and for understanding how aerosols can be related to ecosystemic 112 changes. 113 Low molecular weight compounds (~80-1000 Dalton) have not generally identified or 114 taken into account in aerosol particles and could play crucial roles in the ecosystem 115 functioning, especially at the atmosphere-biosphere interface. Here, we describe in detail a 116 novel and simple methodology to collect ambient aerosol particle samples and subsequently 117 characterize their chemical composition using different mass spectrometry (MS) platforms. 118 This methodology represents a useful tool to shed light on novel questions in plant physiology, 119 ecophysiology, and ecology to understand deeper the atmosphere-ecosystem interface. In this 120 manuscript, we first briefly define ecometabolomics and its potential as a technique to 121 characterize the chemical composition of aerosol particles. We then provide a basic overview 122 of the common MS techniques used in metabolomics. Finally, we discuss our methodologic 123 approach and the different recent metabolomics technologies and, briefly, the obtained 124 results from the samples collected and analyzed with the proposed methodology. 125 126 1.1 Atmo-ecometabolomics: metabolomics of aerosols. 127 Metabolomics aims to study the metabolome of entire organisms or specific cells or tissues, 128 and this technique includes all the procedures from sample collection, metabolite extraction 129 and analyses to data filtering, statistical analyses, and interpretation of the results (Figure 2). A 130 metabolome consists of the total set of small (<1,000 Da) compounds (metabolites) present in 131 a sample at a given moment (Fiehn 2002). Such compounds are represented by the substrates 132 and products from cellular primary metabolism such as organic acids, amino acids, and sugars 133 as well as a vast diversity of compounds derived from the secondary metabolism of organisms 134 such as polyphenolics and terpenes, typically synthetized by plants and fungi. All these 135 metabolites are involved in diverse and complex chemical reactions to finally maintain 136 organisms’ homeostasis, reproduction, and growth (Peñuelas and Sardans 2009). 137 Metabolomics has been widely used in human nutrition research (Gibney et al. 2005; Wishart 138 2008), medicine (Claudino et al. 2007; Walsh et al. 2006), plant physiology (Hirai et al. 2004; 139 Kaplan et al. 2004), and more recently in the field of ecology (ecometabolomics) (Bundy et al. 140 2008; Gargallo-Garriga et al. 2014; Rivas-Ubach et al. 2012, 2016a, 2018a). Ecometabolomics 141 has proven to be useful to understand the plasticity of organisms’ metabolomes under specific 4 142 environmental conditions and/or stressor pressures (Gargallo-Garriga et al. 2014; Rivas-Ubach 143 et al. 2016b, 2017; Sardans et al. 2011, 2014). However, ecometabolomic analyses of aerosol 144 particles (atmo-ecometabolomics) have never been reported before and it is a step further to 145 characterize the aerosol particle composition and to improve our understanding of the 146 aerosol-biosphere interface. 147 Metabolomics can detect specific molecular signatures (biomarkers) directly related to 148 organismal stress and the phenological status of ecosystems. Recent climate projections 149 predict increments of extreme climatic events such as drought and warming that increases 150 plant BVOC emissions (Peñuelas and Staudt 2010) and lead to shifts in the phenology of plants 151 (Menzel et al. 2006; Parmesan 2006; Parmesan and Yohe 2003; Walther et al. 2002). 152 Terrestrial plants have proven to show large overall metabolome shifts when exposed to 153 stressors (Gargallo-Garriga et al. 2014, 2015; Leiss et al. 2009; Macedo 2012; Rivas-Ubach et al. 154 2012, 2014) and several stress biomarkers have been already identified (Glauser et al. 2008; 155 Guy et al. 2007; Keltjens and van Beusichem 1998; Shulaev et al. 2008; Thompson et al. 2005). 156 Significant changes in plant phenology have been detected during the last decades (Menzel et 157 al. 2006; Parmesan 2006; Parmesan and Yohe 2003; Walther et al. 2002) and due the directly 158 link between PBAP and ecosystems, such phenoylogical shifts should be detectable with atmo- 159 ecometabolomics. Therefore, long-term atmo-ecometabolomics studies should provide 160 valuble information of phenological changes and succession or recession of ecosystems. Atmo- 161 ecometabolomics can be also applied to assess stress and phenological changes at ecosystem 162 and regional scales through the characterization of atmo-ecometabolomes. 163 164 1.2 Analytical instruments for MS-based metabolomics analyses. 165 Metabolomic profiles are achieved through two main processes: i) obtaining the chemical 166 signature of the sample (metabolomic fingerprint) and ii) identification of metabolites through 167 the information previously obtained from the metabolomic fingerprints (e.g. exact mass, 168 retention time, fragmentation pattern, etc). MS coupled to liquid or gas chromatography (LC- 169 MS and GC-MS respectively) are currently two of the most common techniques used in 170 metabolomics (Fiehn 2002; Sardans et al. 2011; Zhang et al. 2012), demonstrating high 171 performance and sensitivity (Pan and Raftery 2007). LC-MS and GC-MS provide similar data 172 format; i.e. in both instruments, before compound detection, metabolites are separated by 173 chromatography (liquid or gas) producing finally two orthogonal and independent values; 174 retention time (RT) and mass-to-charge ratio (m/z) for each of the detected features. RT and 175 m/z, together with other additional parameters (e.g. fragmentation pattern), can be thus used 176 for metabolite identification (Sumner et al. 2007). Although there is a wide arrange of 5 177 compounds detectably GC-MS and LC-MS platforms, generally, GC-MS is more suitable for 178 detecting highly-volatile and non-polar compounds with mass weight typically <600 Da such as 179 fatty acids, carotenoids, and essential oils (Gullberg et al. 2004). Untargeted GC-MS analyses 180 require previous compound derivatization of the samples to allow polar molecular compounds 181 to be sufficiently volatile so they can elute at moderate temperature without disintegration. 182 Therefore, carbohydrates, organic acids and other polar analytes can be also detected by GC- 183 MS after derivatization (Gullberg et al. 2004). In comparison to LC, GC has proved to have 184 exceptional reproducibility showing little RT shifts between samples. However, derivatization 185 of the samples increases preparation time and provides indirect metabolite detection that 186 complicates the elucidation of unknown compounds. On the other hand, LC-MS can cover a 187 wider range of compounds than GC-MS and allows the detection of compounds with larger 188 molecular weight (typically up to 1200 Da for metabolomics). LC-MS allows analyzing non- 189 volatile and highly polar compounds ranging from primary metabolic compounds such as 190 organic acids, carbohydrates and amino acids to secondary metabolites such as alkaloids, 191 phenolic acids, flavonoids, or saponins (De Vos et al. 2007). LC-MS provides a direct detection 192 of molecular ions because derivatization is not required. However, LC-MS analyses commonly 193 show larger RT shifts between samples in comparison to GC-MS, especially in studies with 194 large number of samples (e.g. >300). Nonetheless, no single mass spectrometry instrument can 195 cover all molecular compound classes (Ding et al. 2007; Zhang et al. 2012), and combining 196 diverse platforms is a common approach for deeper metabolome characterization of the 197 samples (Hall 2006). 198 Mass resolving power of MS instruments is a critical factor to consider in MS-based 199 metabolomics. The high-resolution and mass accuracy (typically up to 250,000 and < 3 ppm, 200 respectively) of Orbitrap mass spectrometers has proven to be suitable for ecometabolomic 201 studies (White et al. 2017) reducing considerably the error of metabolite identification when 202 using high-resolution compound libraries (Rivas-Ubach et al. 2016c). However, Fourier 203 transform ion cyclotron resonance MS (FT-ICR-MS) is currently the MS platform achieving the 204 highest mass resolving power (up to m/Δm50% > 2,700,000 at m/z 400) and mass accuracy (< 205 1ppm mass error after internal calibration) (Smith et al. 2018). Although FT-ICR-MS can be 206 coupled to HPLC, the scan rate of FT-ICR-MS is commonly not rapid enough as to analyze 207 samples at ultra-high resolution (>400,000) when coupled to LC. This is one of the main 208 reasons why direct infusion (DI) has been the most typical approach to analyze samples with 209 the FT-ICR-MS in ultra-high resolution. Furthermore, DI-FT-ICR-MS acquisition time is 210 significantly shorter (typically between 5-15 minutes) than LC or GC methods which can take 211 over 40 minutes per sample. The ultra-high mass resolution and excellent mass accuracy of DI- 6 212 FT-ICR-MS enables the assignment of elemental formula of a wide portion of the detected 213 features based exclusively on their exact mass (Klein et al. 2006; Kujawinski 2002) providing, 214 thus, powerful means to understand the overall chemical characteristics of organic complex 215 samples (Kim et al. 2003; Reemtsma 2009; Roullier-Gall et al. 2014; Schmitt-Kopplin et al. 216 2012; Sleighter and Hatcher 2007; Tfaily et al. 2015). Assessing the diversity of molecular 217 compounds with nutrients such as P, N or S and understanding how the elemental composition 218 in aerosols shifts due environmental changes isof special interest for ecological stoichiometry 219 research and it is, therefore, feasible with DI-FT-ICR-MS. 220 Visualization of DI-FT-ICR-MS data using van Krevelen diagrams (vK) (O:C vs. H:C ratios 221 of the compounds assigned with an elemental formula) has been commonly used for analyzing 222 complex organic matrices (Kim et al. 2003; Schmitt-Kopplin et al. 2012; van Krevelen 1950). vK 223 diagrams provide information of the main chemical reactions such as demethylation, 224 methylation, hydrogenation, condensation, hydration, reduction or oxidation occuring in the 225 samples (Kim et al. 2003). Additionally, plotting O:C vs. H:C ratios of the assigned compounds 226 to elemental formulas can also provide an approximation of the main compound categories of 227 the analyzed samples (e.g. lipids, protein, carbohydrates, etc.) (Kim et al. 2003; Minor et al. 228 2014; Sleighter and Hatcher 2007). However, several compounds in the environment can be 229 transformed or degraded, and thus changing their original O:C and H:C ratios (Rivas-Ubach et 230 al. 2018b). Consequently, while this classification can still provide a general idea of the organic 231 compound compositions in aerosol particles, any compound classification based on 232 stoichiometric constraints should be used with caution and multidimensional stoichiometric 233 compound classification approaches could be considered (Rivas-Ubach et al. 2018b). C number 234 versus mass graph (CvM) is also commonly used to represent DI-FT-ICR-MS data in 235 comparative studies providing important information on oxidation processes or molecular 236 weight shifts (Reemtsma 2009). DI-FT-ICR-MS represents thus a useful tool to obtain high- 237 resolution metabolomic fingerprints and to gain a better comprehension of potential chemical 238 transformations of samples. 239 For more details in metabolomics technologies and their applications, several review 240 articles with special focus on metabolomics technologies have been published (Rochford 2005; 241 Fukusaki and Kobayashi 2005; Shulaev 2006; Lindon et al. 2007; Saito and Matsuda 2010; Lei et 242 al. 2011; Zhang et al. 2012; White et al. 2017; Rivas-Ubach et al. 2018b; Azad and Shulaev 243 2018). 244 245 1.3 Testing metabolomics techniques in aerosol particle samples. 7 246 The techniques to sample and characterize the gas phase of atmo-metabolomes have been 247 already described elsewhere (Smith and Španěl 2011; Tholl et al. 2006). The main aim of this 248 article is to describe a step-by-step method for sampling and characterizing the particle phase 249 of the atmo-ecometabolomes through distinct mass spectrometry techniques. We collected 250 aerosol particles without size cutoff during two distinct seasons (spring and summer) in an 251 ecosystem with low biological activity by following a simple methodology. Unlike traditional 252 aerosol particle studies concerning atmospheric chemistry, size cutoff of particles was avoided 253 in this method because aeresol particles have a broad range of sizes and size exclusion filtering 254 methods employed in atmospheric studies will not yield informative results for ecology 255 studies. This sampling allowed testing how our method perform in ecosystems with low 256 biological activity by detecting significant changes in the composition of aerosols between 257 seasons. Additionally, we further validate the robustness of our method by analyzing aerosols 258 samples collected during three different peak periods of pollination within the same season in 259 a more biologically active area. 260 In this manuscript we describe a suitable protocol to sample aersols, extract 261 metabolites and analyze them with i) LC-MS, ii) GC-MS and iii) DI-FT-ICR-MS. Data from each 262 analytical instrument was thus analyzed following common statistical approaches for 263 ecometabolomics and chemical characterization studies. The methods for aerosol particle 264 sampling, metabolite extraction, and the main results from the collected aerosols are 265 discussed. The application of atmo-ecometabolomics techniques in natural ecosystems 266 represents a novel approach to shed light on the understanding of the link between metabolic 267 composition of aerosols and the ecosystem structure and function. This novel method applied 268 in ecological sciences allows understanding of scientific questions related to ecosystem stress, 269 the phyllosphere, ecological stoichiometry, ecosystem phenology, or global change. 270 271 272 2. Experimental details. 273 In order to optimize the protocol for aerosol characterization, aerosol samples from different 274 locations were collected and analyzed on multiple mass spectrometry platforms. Figure 3 275 summarizes the procedures implemented in each aerosol sampling campaign detailing the 276 sampling period, analytical instruments, the software used for generating the metabolomic 277 datasets and the tables/figures displaying the main results from the aerosol datasets. 278 279 2.1 Study sites. 8 280 The proposed method is demostrated with the aerosol particle samples collected from two 281 distinct seasons (spring and summer) in 2015 from the Pacific Northwest National Laboratory 282 (PNNL) located in the north side of the city of Richland (Washington, USA) (46ᵒ 34´ N, 119ᵒ 28´ 283 W). Additional samples were collected in 2017 at the University of Michigan (UMICH; Ann 284 Arbor, MI, USA) to validate further our methodology by characterizing aerosol particle 285 signatures from different tree pollination peak periods sampled within a single spring season. 286 UMICH is located in the north-east side of the city of Ann Arbor (Michigan, USA) (42ᵒ 29´ N, 83ᵒ 287 70´ W). 288 PNNL site (spring vs. summer): PNNL nearby landscape is represented by an extensive 289 desert covered mainly by steppes, shrubs and herbaceous species with Purshia tridentate, 290 Ericameria nauseosa, Grayia spinose, Chrysothamnus viscidiflorus, Artemisia tripartita, Salsola 291 tragus, Sarcobatus vermiculatous, and Tamarix romosissima as some of the most represented 292 plant species. PNNL campus is covered by grass with planted non-authoctonous tree species 293 such as Platanus sp. The population of the surrounding metropolitan area is of about 250,000 294 and the economy and land use are mainly dominated by diverse crops and the nuclear 295 reservation of Hanford. The climate at PNNL site is semi-arid desert with an averaged yearly 296 rainfall ranging between 180 and 220 mm. Averaged maximum annual temperatures are 297 around 32ᵒC, with sporadic peaks reaching 42-45ᵒC. Averaged minimum annual temperatures 298 are around -2ᵒC with sporadic tempeatures reaching -20ᵒC. 299 UMICH site (pollination peak periods): The surrounding areas are represented by 300 mosaic landscapes mainly covered by urban, agricultural and extended forested areas 301 dominated by several species of maple trees (Acer sp.), oak (Quercus sp.), honeylocust 302 (Gleditsia triacanthos) and, Betula alleghaniensis, Fagus grandiflora, Populus tremuloides and, 303 some populations of Pinus strobus. Ann Arbor is part of the metropolitan area of Detroit with a 304 population of over 4 million. The climate is humid continental with strong influence from the 305 Great Lakes with a mean annual precipitation of 950 mm per year. Average maximum annual 306 temperature is around 28ᵒC, with peaks reaching up to 41ᵒC and the average minimum annual 307 temperature is 4.7ᵒC. 308 309 2.2 Aerosol particle sampling. 310 For the particle phase aerosol collection, we designed a portable and simple system that 311 allows simultaneous collection of multiple aerosol particle samples (Figure S1). For aerosol 312 particle collection, we used high-purity quartz filters (Whatman QM-A 37mm, Whatman 313 International Ltd, Maidstone, UK) precombusted for 5hrs at 450ᵒC for impurity elimination 314 (Schmitt-Kopplin et al. 2012). Each sample was collected by placing a precombusted quartz 9 315 filter into a filter cassette. Flexible PVC tubing of 0.6 cm diameter was used to connect the 316 pump with the filter cassetes. For each site (PNNL and UMICH), the pump operated during 18 317 consecutive hours (09:00am to 03:00am of the following day) at a flow rate of 30 L per minute 318 at the sampling point (filter). Filters were maunally replaced daily at ~08:30am during the 319 sampling periods. Filter samples were thus stored at -80ᵒC until metabolite extraction. 320 PNNL site sampling (spring vs. summer): We collected aerosols during 14 consective 321 days in spring 2015 (May 7 to May 20; spring samples) and for 16 consecutive days in summer 322 2015 (July 15 to July 30; summer samples) (Figure 3). For each day, two samples were collected 323 simultaneously at an 8-meter tower at PNNL campus. According to the US National Weather 324 Service at the Pasco airport (KPSC); daily averaged temperatures ranged from 11 to 21ᵒC 325 (averaged maximum of 14-29ᵒC) for the spring aerosol collection period (May). Daily averaged 326 temperatures for the summer aerosol collection period (July) ranged between 19 and 29ᵒC 327 (averaged maximum of 28-40ᵒC). Humidity ranged from 49 to 78% (daily average)(averaged 328 maximum of 72-100%) for spring sampling period and from 35 to 50% (daily average)(averaged 329 maximum of 57-86%) for summer sampling period. Total precipitation for May and July 330 sampling periods was 28.2 mm and 0 mm, respectively. 331 UMICH site sampling (pollination peak periods): Additional aerosol samples collected 332 at UMICH campus were collected from 24th to 26th April (birch pollination peak), from 10th to 333 12th May (oak pollination peak), and from 30th May to 1st June 2017 (pine pollination peak) 334 (Figure 3). One filter was collected daily during 3 consecutive days for each pollination peak 335 period from a rooftop location. According to weather conditions reported by the US National 336 Weather Service at the local airport (KARB), temperature averages were 15.5ᵒC, 12.2ᵒC, 15ᵒC 337 for sampling periods corresponding to the birch, oak, and pine pollination peaks, respectively. 338 Average humidity ranged between 55-74%, 53-66%, and 59-74% for the sampling periods 339 corresponding to the birch, oak, and pine pollination peaks, respectively. Accumulated 340 recorded precipitation was 3.3 mm, 3.3 mm and 0.3 mm for the sampling sampling periods 341 corresponding to the birch, oak, and pine pollination peaks, respectively. 342 Sampling for sonication time test (PNNL site): The extraction of compounds from the 343 filters was sonication-based. Thus, we collected additional aerosol samples in spring 2015 at 344 PNNL campus for different sonication time testing in GC-MS and LC-MS platforms. Sonication 345 test was not performed in DI-FT-ICR-MS since peak intensity measured with such a method is 346 only semi-quantitative (Kujawinski 2002; Liu and Kujawinski 2015). 347 During two consecutive days, aerosols were simultaneously collected in 3 independent 348 filters (hereafter test-filters) at a flow rate of 30L per minute at the sampling point. The pump 349 operated daily from 09:00am to 03:00am of the following day (18 hours × 2 days = 36 hours 10 350 per filter). A total of 6 rounds of test-filters were collected (6 rounds × 3 filters = 18 filters in 351 total). Sampling of the test-filters was performed from June 5 to June 16 (12 consecutive days) 352 (Figure 3). Test-filters were stored at -80ᵒC until extraction (sonication test details are shown in 353 the supplementary text of the supporting information). 354 355 2.3 Metabolite extraction for mass spectrometry analysis. 356 Three different tube sets were used; set of tubes A, B and C (Figure 4). The number of tubes in 357 each set was the same than the number of samples collected. Set A (8 mL glass tubes) were 358 used to perform the extractions. Set B (15 mL polypropylene centrifuge tubes) were to keep 359 the extracts. Set C tubes (2mL HPLC glass vials) was used to keep the concentrated extract. 360 Glass tubes were combusted at 450°C for 5 hours before use. Each sample (filters) was rolled 361 (Figure 4.1) and inserted into the corresponding tube of set A (Figure 4.2). Subsequently, each 362 tube from the set A received 5 mL of H2O/MeOH (20:80) (this volume of extract may vary 363 depending on the used filter diameter and the diameter of the set A tubes; filters have to be 364 totally covered by the extract solution) (Figure 4.3). Samples were then sonicated at 24ᵒC for 365 10 min (Figure 4.4) (10min was the established time after performing the sonication time tests, 366 see Supporting Information). Four mL of the extract from each tube of the set A were 367 transferred to the corresponding tube of set B (Figure 4.4.1). Two extractions were performed 368 to each sample by repeating the same procedures but adding 4 mL of H2O/MeOH (20:80) 369 instead of 5mL as fresh extract for the second extraction. The resulting extracts from the 370 second extraction were then combined the first extracts (in tubes of set B) (Figure 4.5.1). All 371 extracts (in set B tubes) were evaporated using an ultra-high purity nitrogen evaporator 372 (Figure 4.6). Subsequently, 1 mL of fresh solvent (H2O/MeOH (20:80)) was added to each set B 373 tube and tubes vortexed for 30 s for a proper resuspension of the dried extracts (Figure 4.7). 374 All concentrated extracts (in set B tubes) were centrifuged at 4,500 x g for 5 min (Figure 4.8) 375 and supernatants were transferred into the corresponding tubes of the set C (HPLC 376 vials)(Figure 4.9). Extracts were stored at -80 ºC until MS analyses (Figure 4.10). Experimental 377 blanks (not used combusted filters) were also extracted and analyzed with each of the MS 378 platforms. 379 Preliminary aerosol particle samples collected at PNNL for sonication tests were 380 analyzed by LC-MS and GC-MS. LC-MS and GC-MS provide accurately relatively quantitative 381 data making them suitable for this sonication testing. Spring and summer samples collected at 382 PNNL campus were analyzed by LC-MS, GC-MS and DI-FT-ICR-MS (Figures 3 and 4.11). 383 Additional samples collected at the UMICH campus to contrast different tree pollination peak 384 periods were analyzed exclusively by LC-MS and GC-MS (Figures 3 and 4.11). For DI-FT-ICR-MS 11 385 and LC-MS plaforms, extracts (~200 µL) were directly used for analyses (Figure 4.12). GC-MS 386 required a pre-treatment of the samples prior to the instrumental analyses. 387 For metabolite derivatization for GC-MS analyses, we followed a protocol described 388 elsewhere (Kim et al. 2015). Briefly, for each sample, 500 µL of extract from the set of tubes C 389 (Figure 4.10) was transferred into a clean HPLC vial and dried down in a vacuum evaporator 390 centrifuge. Then, each sample received 20 μL of methoxyamine in pyridine solution (30 391 mg/mL). Samples were vortexed for 30 s and subsequently incubated for 90 min at 37°C and 392 1,000 rpm in a Thermomixer (Eppendorf AG, Hamburg, Germany) to protect carbonyl groups. 393 After the first incubation, all vials were centrifuged for 15 s and 80 μL of N-methyl-N- 394 (trimethylsilyl)trifluoroacetamide (MSTFA) with 1% trimethylchlorosilane (TMCS) was added to 395 each sample. All vials were then vortexed for 10 s and incubated for 30 min at 37°C and 1,000 396 rpm to derivatize caboxyl, hydroxyl, and amine functional groups. After the second incubation, 397 all samples were centrifuged for 5 min at 4,500 x g and derivatized supernatants were 398 transferred into clean glass vials with 200 μL inserts by using glass Pasteur pipettes. All vials 399 were finally tightened with caps with septum. 400 401 2.4 LC-MS analyses and chromatogram processing. 402 A Vanquish ultra-high pressure liquid chromatographer (UHPLC) coupled to a high-resolution 403 mass spectrometer LTQ Orbitrap Velos (HRMS) equipped with a heated electrospray ionization 404 (HESI) source (Thermo Fisher Scientific, Waltham, Massachusetts, USA) was used to obtain the 405 LC-MS chromatograms of aerosol samples. LC was performed with a reversed-phase C18 406 Hypersil gold column (150 × 2.1 mm, 3µ particle size; Thermo Scientific, Waltham, 407 Massachusetts, USA) operating at 30 ᵒC. and at a constant flow rate of 0.3 mL per minute. 408 Chromatography mobile phases consisted of 0.1% formic acid in water (A) and 409 acetonitrile:0.1% formic acid in water (90:10) (B). The injection volume was set at 5 µL.The 410 elution gradient initiated at 90% A (10% B) and hold for 5 min, then the gradient ramped 411 linearly to 10% A (90% B) during the next 15 min. Those conditions were maintained for 2 min 412 before the initial mobil phase proportions (90% A; 10% B) were recovered over the next 2 min. 413 The column was washed and stabilized for an additional 11 minutes at the initial 414 chromatographic conditions (90% A; 10% B). All samples were analyzed in positive (+) and 415 negative (-) electrospray ionization modes. The HRMS operated in FTMS (Fourier Transform 416 Mass Spectrometry) and full-scan mode at high resolution (60,000) and acquired in the mass 417 range of 50-1000 m/z. An experimental blank was injected every 15 samples. Blanks were used 418 to determine the chromatographic background during the dataset filtering. A mixture of 419 standards were injected every 25 samples to test for mass accuracy and sensitivity of the 12 420 instrument. Two injections of fresh H2O/MeOH (20:80) were analyzed right after the standard 421 mixture for column washing purposes and avoid any potential compound carry over. 422 MZmine 2.17 (Pluskal et al. 2010) was used to process the LC-MS RAW files. 423 Chromatograms obtained in positive and negative modes were treated separately. 424 Chromatograms were baseline corrected, deconvoluted, aligned, and “gap-filled” and 425 metabolite identifications were assigned. Datasets were thus exported to CSV format files. The 426 parameters used for generating the LC-MS datasets are detailed in Table S1. The same MZmine 427 parameters were used to generate all datasets spring vs. summer LC-MS dataset (PNNL) and 428 pollination peak period LC-MS dataset (UMICH). 429 Metabolite assignation for LC-MS files was performed by matching the exact mass and 430 RT of the detected features with the corresponding values of our metabolite library including 431 more than 600 metabolites present in organisms commonly present in plants. According to 432 Sumner et al. (Sumner et al. 2007), our LC-MS metabolite assignment is putative since it was 433 based on the measured exact mass of metabolic features and their RT. However, the use of 434 both RT and the high MS resolution achieved with Orbitrap technology reduces substantially 435 the number of false positive assignations. For more detailed information in relation to 436 metabolite library matching see Rivas-Ubach et al. (Rivas-Ubach et al. 2016c). Metabolite 437 matching information (RT and m/z values) for LC-MS files are detailed in Table S2. 438 439 2.5 GC-MS analyses and chromatogram processing. 440 Derivatized samples were analyzed by an Agilent GC 7890A coupled with a MSD 5975C mass 441 spectrometer (Agilent Technologies, Santa Clara, CA). An HP-5MS column (30 m × 0.25 mm × 442 0.25 μm; Agilent Technologies) was used for the GC. The injection port temperature was held 443 at 250 ᵒC. The injecton mode was split-less. The oven with the column was maintained for 1 444 min at 60°C and then ramped at 10°C per minute to 325°C (26.5 min ramp) and hold for 10 445 min. A mixture of fatty acid methyl esters (FAMEs; C8-C28) for retention index calculation (RI) 446 was analyzed prior to sample injections. Experimental blanks were analyzed every 15 samples 447 to determine the chromatographic background. 448 MZmine 2.17 (Pluskal et al. 2010) was exclusively used to generate the metabolomic 449 fingerprint datasets for the test-filters collected at PNNL to be consistent with the LC-MS 450 datasets for the sonication test (Figure 3). Parameters implemented to generate the datasets 451 from the test-filters with MZmine are described in Table S3. Metabolite Detector 2.5 (Hiller et 452 al. 2009) was used to generate the GC-MS datasets for the ambient samples collected at PNNL 453 (spring vs. summer) and UMICH (pollination peak periods) sites (Figure 3). Agilent “.D” files 454 directly obtained from the MS were converted to “netCDF” format with Agilent Chemstation 13 455 software. Subsequently, Metabolite Detector converted “netCDF” files to “bin” files and 456 chromatograms were thus deconvoluted, aligned according to the retention indices (RI) and 457 metabolic features were were assigned to metabolites. Metabolites were identified by 458 matching measured mass spectra to a PNNL improved version of FiehnLib (Kind et al. 2009), 459 containing validated RIs and mass spectra for over 850 metabolites. Metabolite matching 460 probability threshold was set at 0.65. Datasets were finally exported to CSV format files. 461 Metabolite identifications were manually validated by matching measured mass spectra with 462 mass spectra from NIST14 GC-MS library to avoid false identifications and reduce any allegued 463 deconvolution errorduring data-processing. Table S4 includes all the parameters used in 464 Metabolite Detector to generate the GC-MS datasets from the aerosols collected at PNNL and 465 UMICH sites. Table S5 shows metabolite matching information for GC-MS. 466 467 2.6 DI-FT-ICR-MS analyses and spectra processing. 468 DI-FT-ICR-MS analyses were performed exclusively on the spring and summer samples 469 collected at PNNL. Extracts of aerosols were analyzed on a 12 Tesla Bruker SolariX FT-ICR-MS 470 (Bruker daltonics Inc, Billerica, MA, USA). At a flow rate of 3.0 μL/min, aerosol extracts were 471 directly infused into the FT-ICR-MS using an Agilent 1200 series pump (Agilent Technologies, 472 Santa Clara, CA, USA). Compounds were ionizated by a standard Bruker electrospray (ESI) 473 operating in negative mode and equipped with a fused silica tube (30 μm i.d.). Ion 474 accumulation time was optimized for each sample (0.1s). FT-ICR-MS operated at a resolution 475 of 4 MWord, equilavent to a resolving power of 400,000 (m/Δm50% at m/z 400). Experimental 476 conditions were as follows: needle voltage was set at +4.4 kV; Q1 set to 50 m/z; and the 477 heated resistively coated glass capillary was maintained at 180 °C. Solvent blanks were run 478 every 10 samples. 479 A total of 144 individual scans from 100 to 1100 m/z were averaged to produce the 480 final DI-FT-ICR-MS spectrum for each sample. The instrument achieved mass measurement 481 accuracy with < 1 ppm error after external calibration with a standard mixture. All detected 482 ions in the spectra were singly charged as confirmed by the 1.0034 Da spacing between 483 isotopic forms of the same compound (i.e., between 12Cn and 12Cn−1–13C1). Each spectrum raw 484 file was converted to a list of m/z values with DataAnalysis software (BrukerDaltonik version 485 4.2) with the following settings: an absolute intensity threshold of 100 and a signal to noise 486 (S/N) of 7, which is above the minimum detection limit for FT-ICR-MS for natural organic 487 matter (Riedel and Dittmar 2014). The exported peak lists were then internally calibrated using 488 an organic matter homologous series separated by 14 Da (–CH2 groups) using the Formularity 489 software (Tolić et al. 2017). After internal calibration, the mass error was < 0.5 ppm. The 14 490 calibrated peak list was then used to assign with elemental formulas (C, H, O, N, S, and P) using 491 a module within Formularity, based on the compound identification algorithm described 492 elsewhere (Kujawinski and Behn 2006). 493 494 2.7 Datasets filtering. 495 A total of 7 distinct datasets were generated: 3 contrasting spring vs. summer aerosols (LC-MS, 496 GC-MS and DI-FT-ICR-MS; collected at PNNL), 2 comparing three different pollination peak 497 periods (LC-MS and GC-MS; collected at UMICH) and 2 for sonication tests (LC-MS and GC-MS; 498 collected at PNNL). LC-MS and GC-MS datasets were separately filtered through 5 main steps: 499 1) All zero values of the dataset were replaced for missing data (NA). 500 2) For each variable (metabolite feature), outlier values were determined for each season 501 separately and replaced for NA. Outliers were defined as: 502 𝑈𝑝𝑝𝑒𝑟 𝑂𝑢𝑡𝑙𝑖𝑒𝑟𝑠 → 𝑣𝑎𝑙𝑢𝑒 > 𝑄75 + 2.5 × 𝐼𝑄𝑅 503 𝐿𝑜𝑤𝑒𝑟 𝑂𝑢𝑡𝑙𝑖𝑒𝑟𝑠 → 𝑣𝑎𝑙𝑢𝑒 < 𝑄25 − 2.5 × 𝐼𝑄𝑅 504 where Q75 represents the third quartile, Q25 represents the first quartile and IQR is the 505 interquartile range (IQR = Q75-Q25). 506 3) Variables (metabolite features) with less than 50% of data within all factor levels of the 507 dataset (spring and summer for datasets from PNNL aerosol particles; birch, oak and 508 pine pollination peak periods from UMICH aerosols) were removed. 509 4) Variables (metabolite features) with signal to noise lower than 6 were removed from 510 the dataset. Noise level was determined by the experimental blanks run during the 511 analytical sequence. 512 5) To avoid statistical artifacts in multivariate analyses due sample variability, the values 513 for each variable (metabolite feature) and sample were scaled by the total intensity of 514 its chromatogram (LC-MS and GC-MS). This scaling allows comparing the metabolic 515 function between groups of samples independently of the amount of organic matter. 516 It should be noted that DI-FT-ICR-MS data should go through rigorious data quality 517 check before further interpretation, to avoid false discovery. For DI-FT-ICR-MS dataset (PNNL; 518 spring vs. summer), mass features detected from solvent blanks were removed from all the 519 samples. Additionally mass features observed in less than 70% of replicates within all factor 520 levels were also be removed to avoid consideration of noise peaks. Finally, for robust data 521 interpration based on elemental formulas, we only used the assigned formulas with less than 15 522 0.3ppm of error to be conservative; as cutoff values up to 0.5ppm of error are typically used 523 (Osterholz et al. 2016; Rivas-Ubach et al. 2018b). 524 After filtering all databases, our spring vs. summer databases were finally composed 525 by 1,333, 148, and 3,567 metabolomic features detected by LC-MS, GC-MS, and DI-FT-ICR-MS, 526 respectively. For LC-MS and GC-MS datasets, a total of 18 and 15 metabolites were identified, 527 respectively. The pollination peak periods databases were finally composed by 7,832 and 221 528 metabolomic features detected by LC-MS and GC-MS, respectively, with a total of 45 and 20 529 identified metabolites, respectively. 530 531 2.8 Statistical analyses. 532 While the values obtained from the deconvoluted peaks in LC-MS and GC-MS do not represent 533 an absolute concentration (e.g., mg of metabolite per weight of sample), they represent the 534 relative abundance and are suitable for metabolomic comparative analyses as proven in 535 previous studies (Gargallo-Garriga et al. 2015; Lee and Fiehn 2013; Leiss et al. 2013; Mari et al. 536 2013; Rivas-Ubach et al. 2014, 2016a, 2018). We used then the term relative abundance along 537 the article as the relative concentration of metabolic features or metabolites. DI-FT-ICR-MS 538 data is not directly quantifiable (Wozniak et al. 2008), and although not as robust as GC-MS or 539 LC-MS techniques, using the feature intensity can still provide a valuable proxy of their relative 540 abundance (Kellerman et al. 2014; Spencer et al. 2015). 541 All LC-MS and GC-MS datasets, independently if generated from PNNL (spring vs. 542 summer) or UMICH (pollination peak periods) aerosols, were submitted to the same statistical 543 analyses. First, the aerosol metabolome fingerprints obtained from LC-MS and GC-MS were 544 submitted to PERMANOVAs using the Euclidean distance to test for overall metabolomic 545 differences between aerosols from distinct seasons (PNNL site), and between different 546 pollination peak periods (UMICH site) (Table 1). The permutations for the test were set at 547 10,000. 548 LC-MS and GC-MS metabolomic fingerprints were also subjected to principal 549 component analysis (PCA) to graphically represent the natural variability among the analyzed 550 samples reduced in two dimensions (Principal Component (PC) 1 vs. PC2) (Kim et al. 2010; van 551 den Berg et al. 2006) (Figure 5). 552 A heat map was plotted for each entire LC-MS and GC-MS metabolomic fingerprints 553 from the aerosols collected at PNNL and UMICH sites (Figure 6). Additional heat maps were 554 plotted only for the identified features from the spring vs. summer and for the pollination peak 555 periods datasets (Figure 7). All the identified LC-MS and GC-MS features for the spring vs. 556 summer dataset (PNNL) were submitted to Student t-test with season as the categorical factor 16 557 to assess for statistical significance between spring and summer seasons (Table S6). One-way 558 ANOVAs followed by Tukey HSD posthoc tests were performed to asses for significant 559 differences between pollination peak period samples (birch vs. oak vs. pine). (Table S7). 560 For the DI-FT-ICR-MS dataset generated from the PNNL aerosols (spring vs. summer), 561 we counted the proportions of formula classes (CHNO, CHO, CHOS, CHNOS, CHNOP, CHOSP, 562 CHNOSP, and CHOP) that were assigned to each sample. Proportions were then transformed 563 using arcsin (rootsquare) and subsequently submited to Student t-tests with season (spring 564 and summer) as the categorical factor to assess for statistical significance (Figure 8). For each 565 feature detected by DI-FT-ICR-MS, the intensity difference between spring and summer was 566 calculated and used in the C number of each feature vs. m/z (CvM) (Figure 9a) and in the O:C 567 ratio vs. m/z (Figure 9b) plots. Additionally, a Student t-test was performed on the O:C ratios of 568 the detected features with season as the categorical factor to determine whether the 569 oxidation status of the molecular compounds statistically changed significantly between spring 570 and summer aerosols (Figure 9c). 571 All statistical analyses were performed with R (R Core Team 2013). PERMANOVAs were 572 conducted with the function adonis from the “vegan” package (Oksanen et al. 2013). PCAs 573 were plotted using the PCA function from “FactoMineR” package (Husson et al. 2016) with the 574 missing data from the dataset imputed using imputePCA function from package “missMDA” 575 (Husson and Josse 2015). The data matrix was scaled before the PCA by setting SCALE as TRUE 576 from the PCA function. Heat maps were plotted using the function heatmap.2 from “gplots” 577 package (Warnes et al. 2016). Student t-tests and oneway ANOVAs were performed with the 578 functions t.test and aov, respectively, included in the “stats” package (R Core Team 2013). 579 Tukey HSD posthoc tests were performed with the HSD.test function in the “agricolae” 580 package (de Mendiburu 2015). All graphs were first plotted in R and subsequently treated with 581 Adobe Illustrator CS6. Analyses and results from the different sonication times during 582 metabolite extraction are detailed in the supporting information (Supplementary Text). 583 584 3. Results and discussion. 585 3.1 Aerosol sampling. 586 Optimal flow rate for aerosol particles collection is important, as low flow rate may not collect 587 enough aerosol particles and excessive flow rate may damage the filters. After coupling the 588 tubing with the pump and connected two filters, we achieved constant flow rates of 30 L/min 589 at the aerosol sampling point. However, flow rates of ~50 L/min at the aerosol collection point 590 performed properly in 37mm quartz filters without collapsing. Different filter materials, such 591 as polytetrafluoroethylene (PTFE), may present distinct resistances to high flow rates. It is 17 592 important to note that the internal air friction associated with the length of the tube and the 593 sampling of multiple filters simultaneouly can substantially decrease the flow rates at the 594 aerosol sampling point. Larger tube diameters (>0.6cm) could be used if higher flow rates at 595 the sampling point are necessary, especially when sampling in ecosystems with low biological 596 activity. 597 For statistical puproses, our simple sampling methodology allows collecting different 598 number of biological replicates at once. Furthermore, aerosol collection can be easily 599 performed at different heights by extending tubing if the pump performance is sufficient as to 600 ensure acceptable flow rates at the aerosol collection point. The experimental design (number 601 of replicates, filter material, length and diameter of tubing) and the pump performance are 602 thus key elements to consider in atmo-ecometabolomics research. 603 Aerosol particles collection was performed in two distant locations with distinct 604 environments; PNNL and UMICH. PNNL site is semi-urban area surrounded by diverse 605 croplands and a large desert with relatively low biological activity. For this reason we expected 606 to detect a complex variety of compounds which difficult the interpretation of the data at 607 regional scale. Even so, the results obtained contrasting spring vs. summer aerosols were 608 equally valid as to describe in detail the necessary steps to obtain accurately the atmo- 609 metabolomes and to test the sensitivity of each mass spectrometry platform (LC-MS, GC-MS, 610 DI-FT-ICR-MS) by assessing their potential for detecting statistically significant changes 611 between the atmo-ecometabolomes of two contrasted seasons of ecosystems with low 612 biological activity. At the same time, UMICH location provided additional data collected in a 613 more biologically diverse and active ecosystem to further evaluate the sampling and the 614 analytical methodologies. 615 616 3.2 Metabolite extraction in organic solvents. 617 Organic solvents in combination with water are commonly used for metabolite extractions 618 allowing a good extraction range of polar, semi-polar and non-polar metabolites (Kim et al. 619 2010; Lin et al. 2006; Rivas-Ubach et al. 2013; t’Kindt et al. 2008). Although different solvents 620 recover different matrices of compounds with distinct polarities, water/methanol (20:80) 621 solution has been widely used in metabolomics studies showing a wide recovery of polar and 622 semi-polar metabolites (t’Kindt et al. 2008). Solvents such as acetonitrile, methanol, and 623 chloroform could leach plastics especially during sonication, as such MS spectra may show 624 contaminant features derived from plastics (Figure S2). Therefore, the use of silanized glass 625 tubes is advised during sonication (Figure 4.4). Combusting glassware for 5-6 hours at 450ºC or 626 higher is also recommended to prevent organic contaminations. If plastic tubes are used for 18 627 metabolite extraction, especially during the sonication step (set of tubes A), initial tests are 628 recommended to identify any potential contaminant in the extracts. Due the relatively low 629 metabolite concentration in aerosol samples, each sample was extracted twice in 630 water/methanol (20:80) to increase metabolite recovery from aerosols (Böttcher et al. 2007; 631 Nikiforova et al. 2005; Rivas-Ubach et al. 2013) (Figure 4.5). 632 We did not detect overall significant variation in the relative abundance of the 633 observed features between test-filters under different sonication times (Tables S8 and S9) with 634 the water/methanol (20:80) extraction; only days 5th & 6th June showed significant differences 635 between different sonication times (F= 3.21, P = 0.02) in the GC-MS analyses (Table S9). We 636 thus considered that 10 minutes of sonication was enough as to extract the metabolites from 637 aerosol samples in water/methanol (20:80) solution. 638 Filter size is also a crucial factor to consider in atmo-ecoecometabolomic reaseach, 639 especially during the extraction procedures. Extracts of aerosol particles will be more 640 concentrated as lower is the filter size/pump flow rate ratio. Furthermore, smaller filters are 641 easier to handle during extractions and allow higher recovery of the extracts. Quartz filters 642 absorb high volumes of solvent that cannot be easily recovered, even if samples are 643 centrifuged at high rpm. We could recover the 89% of the initial solvent volume with 37mm 644 diameter filters. Larger filters complicate significantly the extraction of metabolic compounds 645 (more filter handling, larger tubes and volumes of extract are required) and decrease 646 remarkably the extract recovery due to the large absorption of solvent. For studies using FT- 647 ICR-MS, a final dissolved organic carbon (DOC) concentration of 15 mg of C per L in 1:1 648 methanol/water in samples has been proven to provide suitable signal for detection of 649 compounds (Medeiros et al. 2017). However, higher concentration of C may be required for 650 other mass spectrometry platforms depending on their sensitivity. Analyses by LC-MS and GC- 651 MS techniques separate the analytes through chromatography; therefore, the amount of C 652 analyzed in each scan is diluted and higher amount of DOC allows a better detection of organic 653 compounds and metabolome characterization. 654 655 3.3 Testing atmo-ecometabolomics contrasting two distinct seasons (PNNL site) 656 Even collecting aerosols in low biologically productive ecosystems, PERMANOVAs showed 657 overall significant differences between spring and summer aerosols for the fingerprints 658 obtained with both LC-MS and GC-MS platforms (Pseudo-F = 13.8, P < 0.0001 and Pseudo-F = 659 9.69, P < 0.0001; respectively) (Table 1). 660 PCAs plotted for both LC-MS and GC-MS figerprtings also showed clear separation 661 between spring and summer aerosol samples (Figure 5a and 5b). The PC1 and PC2 of the PCA 19 662 performed with LC-MS dataset explained the 37.7% and 14.81% of the total variation among 663 samples, respectively. The PC1 and PC2 of the GC-MS PCA explained the 39.72% and 17.2%, 664 respectively, of the total variance. Both PCAs showed similar explained variability for the PC1 665 (37.7% for LC-MS and 39.72% for GC-MS) that primarly segregated spring from summer 666 samples. The results from PERMANOVAs and PCAs indicate, indeed, that the metabolomic 667 fingerprints from both seasons are significantly distinct. Even so, it is important to note that 668 each analytical technique (LC-MS and GC-MS) provide different complementary information 669 (Ding et al. 2007; Zhang et al. 2012). 670 Heat maps of the entire LC-MS and GC-MS metabolomic fingerprints showed large 671 metabolomic variability between aerosol samples within the same season (Figure 6a and 6b). 672 Even so, we still identified several clusters of metabolic features (marked in blue) presenting 673 higher relative abundances in spring or summer aerosol samples. Those heat maps are clearly 674 in accordance with the PERMANOVAs and the PCA by showing major metabolic differences 675 between spring and summer aerosols supporting thus our protocol as sensitive enough as to 676 differentiate the overall aerosol composition between two seasons in a complex and low 677 biologically active environment. 678 Several of the identified metabolites with LC-MS and GC-MS showed statistical 679 significance between seasons after Student t-tests (Figure 7a and Table S6). LC-MS analyses 680 showed spring aerosol particles with significantly higher relative abundance (P < 0.05) of - 681 ketoglutaric acid, acacetin, adenosine, adonitol-ribitol, chrysin, citric acid, glutamine, hexoses, 682 isoleucine, shikimic acid and sorbitol-mannitol (Figure 7a). GC-MS dataset indicated that spring 683 aerosols presented significantly higher relative abundances of glucose and galactose (P < 0.05) 684 and summer aerosol particles showed significantly higher relative abundances of arachidic 685 acid, capric acid, fumaric acid, glyceric acid, linoleic acid, palmitic acid, stearic acid and uracil 686 (Figure 7a). Higher relative abundances of sugars and organic acids related to the Krebs cycle 687 such as citric and alpha-ketoglutaric can be realted to plant growth activity (Rivas-Ubach et al. 688 2012) and atmospheric pollination (Roulston and Cane 2000). Our results from the LC-MS and 689 GC-MS platfroms are in agreement with the DI-FT-ICR-MS dataset showing spring aerosols with 690 significantly (P < 0.05) higher proportions of CHOP and marginally significant (P < 0.1) CHNOSP 691 molecular formulas (Figure 8). High concentrations of sugars and phosphorus in biomass have 692 been previously associated to higher plant activity (Rivas-Ubach et al. 2012, 2014) although 693 sugars can also play multiple physiological functions in plants such as stress tolerance under 694 drought conditions (Ingram and Bartels 1996; Rivas-Ubach et al. 2014, 2016a). Several fatty 695 acid compounds, such as arachidic acid, capric acid, heptadecanoic acid, linoleic acid, oleic 696 acid, palmitic acid, and stearic acid, have been detected by GC-MS (Figure 7a). Fatty acids are 20 697 well represented in cells and can represent up to 20% of dry weight in pollen and other plant 698 material (Roulston and Cane 2000). For example, arachidic acid and linoleic acid, among 699 others, are typically found in pollen (Solberg and Remedios 1980). Our GC-MS analyses 700 indicated that summer aerosols had higher relative abundances of most identified fatty acids 701 (Figure 7a) coinciding with the agricultural production peak of the area, the most extensive of 702 the Pacific Northwest of USA. 703 DI-FT-ICR-MS analyses showed that aerosol particles collected in summer aerosols had 704 significantly (P < 0.05) higher proportions of CHO features than spring aerosols (Figure 8). In 705 addition, we generally observed higher relative intensities in high-mass features in summer 706 aerosols with respect to spring aerosols (Figure 9a). In a CvM plot, at a given carbon number, 707 the increase of nominal mass is contributed by heteroatoms (e.g. N, S, and O). We observed 708 that the higher relative intensities at high-mass features in summer aerosol particles were 709 contributed by heteroatoms as these features appear to possess similar range of carbon 710 numbers as those observed in spring, which were more abundant at lower molecular mass 711 (see region between dashed lines in Figure 9a). In addition, Student t-test on the O:C of the 712 formula-assigned features indicated that summer aerosol particles had significantlyhigher 713 relative intensities in features with higher O:C (more oxidized compounds) compared to spring 714 aerosols (Figure 9b, 9c). This result is in accordance with the larger number of high-molecular 715 weight compounds for a same C-number found in aerosol particles collected in summer 716 compared to those from spring (Figure 9a), suggesting that summer aerosol particles 717 experienced higher oxidation rates. This trend could be related to higher levels of 718 photochemical oxidants associated with hot sunny conditions of the area during summer 719 (Obee and Hay 1997). Additionally, higher relative intensities in compounds over 500 Da were 720 also found in summer (Figure 9a) suggesting higher rates of aerosol condensation and 721 polymerization. These measurements point to a major challenge in atmo-ecometabolomics 722 research; understanding deeply the atmospheric processing of the initial biogenic emissions. 723 724 3.4 Atmo-ecometabolomics for differentiate three different pollination peak periods (UMICH 725 site) 726 PERMANOVAs on the LC-MS and GC-MS fingerprintsshowed overall significant differences 727 between the atmo-metabolomes collected during different pollination peak periods (Pseudo-F 728 = 9.54, P < 0.05 and Pseudo-F = 7.63, P < 0.05; respectively) (Table 1). Additionally, PCAs 729 showed clear clustering of the different sample groups along PC1 and PC2, explaining the 730 39.61% and the 33.83% for LC-MS and the 55.81% and 25.81% for GC-MS analyses, 731 respectively (Figure 5c and 5d). In accordance with PERMANOVAs and PCAs, heatmaps of the 21 732 LC-MS and GC-MS aerosol fingerprints showed clear differences between pollination periods 733 (Figure 6c and 6d). These results indicate thus that the proposed atmo-ecometabolomics 734 methodology was sensitive enough as to accurately accurately extract, characterize and 735 distinguish aerosols collected during three different pollination peak periods (birch, oas and 736 pine) occurring within a continuous 40-day timeframe. 737 Compared to PNNL, the surroundings of UMICH have higher biological and human 738 activity and this difference was distinctly noticeable in the aerosol metabolomic fingerprints. 739 The number of features detected by both LC-MS and GC-MS instruments is considereably 740 larger in UMICH than PNNL aerosol fingerprints (1,333 vs. 7,832 for LC-MS fingerprints and 148 741 vs. 221 for GC-MS fingerprints) (Figure 6a vs. 6c and 6b vs. 6d). Furthermore, a larger number 742 of metabolomics features could be identified at the aerosols collected in UMICH (45 and 20 for 743 LC-MS and GC-MS, respetively) (Figure 7b). Then, those results suggest that the number of 744 detected and identified features in aerosol samples directly depends on the biological and 745 anthrophogenic activity of the surrondings. For this reason, it is crucial to keep relatively high 746 flow rates (≥25-30 L/min) at the aerosol collection point in order to characterize properly the 747 particle component of aerosol metabolomes, especially in ecosystems with low biological 748 activity. 749 750 3.5 Conclusions and future perspectives. 751 · Our methodology was sensitive enough as to detect significant overall differences between 752 atmo-ecometabolomes from aerosol particles sampled in different seasons (spring vs. 753 summer) in an ecosystem with low biological activity. Furthermoe, we detected clear changes 754 between the composition of aerosol particles collected during tree pollination periods. These 755 differences were apparent despite the high complexity of aerosol particle sources in the mix of 756 urban and vegetated areas. 757 · LC-MS and GC-MS showed similar accuracy in terms of sample clustering; both instruments 758 were able to detect clear overall differences between spring and summer aerosols and 759 between aerosols collected in different pollination peak seasons. Which instrument to use will 760 finally depends on the hypotheses of the study and whether a targeted approximation is 761 required. However, no analytical technique alone can still characterize the entire metabolome 762 fingerprint of a sample and combining both platforms is a common approach. In general, the 763 number of detected metabolic features is substantially higher in LC-MS analyses than GC-MS. 764 This is especially useful for overall multivariate analyses in non-targeted metabolomics studies 765 as LC-MS analyses provide a wider representation of the sample metabolome. Both 766 instruments detected several polar and semi-polar metabolites but GC-MS analyses allowed 22 767 the identification of some non-polar compounds in aerosols (e.g. linoleic acid, oleic acid, 768 palmitic acid, stearic acid). On the other hand, DI-FT-ICR-MS obtains ultra high-resolution 769 metabolomic data allowing the assignation of elemental formulas in the detected features in 770 aerosols. This data format allows performing different analyses on the distinct heteroatoms 771 (e.g. O, N, P) present in aerosols. This information is especially useful to study the nutrient 772 transport and deposition in ecosystems. 773 · Coupling environmental variables (e.g., temperature, wind, precipitation, etc.) with atmo- 774 ecometabolomics would allow a more precise interpretation of the aerosol-biosphere 775 interface. 776 · Long term atmo-ecometabolomic research in natural ecosystems would significantly improve 777 our understanding of interannual and seasonal changes of aerosol composition, directly linking 778 the composition of aerosols with plant phenology and physiology, along environmental 779 changes and/or natural spatial gradients. 780 · The use of atmo-ecometabolomic techniques in ecological sciences could improve the 781 detection, identification and quantification of molecular compounds directly related with 782 environmental stressors (biomarkers), providing thus important information of the general 783 status of the ecosystems. A good description of such biomarkers would allow generating 784 libraries of compounds that may serve to asses the ecosystem status or as an environmental 785 monitoring tool. 786 · A better understanding of the direct impacts of aerosols on biological surfaces (e.g. 787 phyllosphere) and on the overall C:N:P stoichiometry of ecosystems can be improved 788 significantly by the overall characterization of the aerosol chemical and molecular 789 composition. 790 · New modern high-resolution mass spectrometry instruments coupled to separation 791 techniques should be implemented in atmo-ecometabolomic research to enable high 792 performance for both mass accuracy and resolution, and retention time. Furthermore, recent 793 advances in metabolomics instruments, such as mass spectrometers coupled to Ion Mobility 794 Spectrometry (IMS-MS), could substantially enhance the number of detected metabolic 795 features in aerosol samples from tens and hundreds to thousands. 796 797 Acknowledgements. 798 The authors thank Rosalie Chu and Therese Clauss for their laboratory support. This research 799 was performed using EMSL, a DOE Office of Science User Facility sponsored by the Office of 800 Biological and Environmental Research at Pacific Northwest National Laboratory. JS and JP 23 801 were supported by the European Research Council Synergy grant SyG-2013-610028 802 IMBALANCE-P, the Spanish Government projects CGL2013-48074-P and the Catalan 803 Government project SGR 2014-274. ALS was supported in part by National Science Foundation 804 grant AGS 0952659. 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 24 825 References. 826 Achotegui-Castells, A., Sardans, J., Ribas, À., & Peñuelas, J. (2013). Identifying the origin of 827 atmospheric inputs of trace elements in the Prades Mountains (Catalonia) with 828 bryophytes, lichens, and soil monitoring. Environmental Monitoring and Assessment, 829 185(1), 615–629. 830 Andreae, M. 0, & Crutzen, P. J. (1997). Atmospheric Aerosols: Biogeochemical Sources and 831 Role in Atmospheric Chemistry. Science, 276(5315), 1052–1058. 832 Arnold, A. E., Maynard, Z., Gilbert, G. S., Coley, P. D., & Kursar, T. A. (2000). Are tropical fungal 833 endophytes hyperdiverse? Ecology Letters, 3(4), 267–274. 834 Ayers, G. P., & Gras, J. L. (1991). Seasonal relationship between cloud condensation nuclei and 835 aerosol methanesulphonate in marine air. Nature, 353(6347), 834–835. 836 Azad K. R., & Shulaev, V. (2018). Metabolomics technology and bioinformatics for precision 837 medicine. Briefings in Bioinformatics, bxx170, doi.org/10.1093/bib/bbx170. 838 Baker, A. R., Kelly, S. D., Biswas, K. F., Witt, M., & Jickells, T. D. (2003). Atmospheric deposition 839 of nutrients to the Atlantic Ocean. Geophysical Research Letters, 30(24). 840 Baustian, K. J., Cziczo, D. J., Wise, M. E., Pratt, K. A., Kulkarni, G., Hallar, A. G., & Tolbert, M. A. 841 (2012). Importance of aerosol composition, mixing state, and morphology for 842 heterogeneous ice nucleation: A combined field and laboratory approach. Journal of 843 Geophysical Research: Atmospheres, 117, D06217. 844 Böttcher, C., Roepenack-Lahaye, E. v., Willscher, E., Scheel, D., & Clemens, S. (2007). Evaluation 845 of Matrix Effects in Metabolite Profiling Based on Capillary Liquid Chromatography 846 Electrospray Ionization Quadrupole Time-of-Flight Mass Spectrometry. Analytical 847 Chemistry, 79(4), 1507–1513. 848 Bundy, J. G., Davey, M. P., & Viant, M. R. (2008). Environmental metabolomics: a critical review 849 and future perspectives. Metabolomics, 5(1), 3–21. 850 Canagaratna, M. R., Jayne, J. T., Jimenez, J. L., Allan, J. D., Alfarra, M. R., Zhang, Q., et al. (2007). 851 Chemical and microphysical characterization of ambient aerosols with the aerodyne 852 aerosol mass spectrometer. Mass Spectrometry Reviews, 26(2), 185–222. 25 853 Carlton, A. G., Pinder, R. W., Bhave, P. V., & Pouliot, G. A. (2010). To What Extent Can Biogenic 854 SOA be Controlled? Environmental Science & Technology, 44(9), 3376–3380. 855 Carnicer, J., Sardans, J., Stefanescu, C., Ubach, A., Bartrons, M., Asensio, D., & Peñuelas, J. 856 (2015). Global biodiversity, stoichiometry and ecosystem function responses to 857 human-induced C–N–P imbalances. Journal of Plant Physiology, 172, 82–91. 858 Claudino, W. M., Quattrone, A., Biganzoli, L., Pestrin, M., Bertini, I., & Di Leo, A. (2007). 859 Metabolomics: Available Results, Current Research Projects in Breast Cancer, and 860 Future Applications. Journal of Clinical Oncology, 25(19), 2840–2846. 861 de Mendiburu, F. (2015). agricolae: Statistical Procedures for Agricultural Research. R package 862 version 1.2-3, http://CRAN.R-project.org/package=agricolae. 863 De Vos, R. C., Moco, S., Lommen, A., Keurentjes, J. J., Bino, R. J., & Hall, R. D. (2007). 864 Untargeted large-scale plant metabolomics using liquid chromatography coupled to 865 mass spectrometry. Nature Protocols, 2(4), 778–791. 866 Després, V. R., Alex Huffman, J., Burrows, S. M., Hoose, C., Safatov, A. S., Buryak, G., et al. 867 (2012). Primary biological aerosol particles in the atmosphere: a review. Tellus B, 868 64(15598), 1–58. 869 Ding, J., Sorensen, C. M., Zhang, Q., Jiang, H., Jaitly, N., Livesay, E. A., et al. (2007). Capillary LC 870 Coupled with High-Mass Measurement Accuracy Mass Spectrometry for Metabolic 871 Profiling. Analytical Chemistry, 79(16), 6081–6093. 872 Elser, J. J., Dobberfuhl, D. R., MacKay, N. A., & Schampel, J. H. (1996). Organism Size, Life 873 History, and N:P Stoichiometry. BioScience, 46(9), 674-684. 874 Fageria, N. K., Filho, M. P. B., Moreira, A., & Guimarães, C. M. (2009). Foliar Fertilization of 875 Crop Plants. Journal of Plant Nutrition, 32(6), 1044–1064. 876 Feng, J., Wang, Y., Zhao, J., Zhu, L., Bian, X., & Zhang, W. (2011). Source attributions of heavy 877 metals in rice plant along highway in Eastern China. Journal of Environmental Sciences, 878 23(7), 1158–1164. 879 Fernández, V., & Brown, P. H. (2013). From plant surface to plant metabolism: the uncertain 880 fate of foliar-applied nutrients. Frontiers in Plant Science, 4, 289. 26 881 Fiehn, O. (2002). Metabolomics - the link between genotypes and phenotypes. Plant molecular 882 biology, 48(1–2), 155–171. 883 Fukusaki, E., & Kobayashi A. (2005). Plant metabolomics: potential for practical operation. 884 Journal of Bioscience and Bioengineering, 100(4), 347-354. 885 Fuzzi, S., Andreae, M. O., Huebert, B. J., Kulmala, M., Bond, T. C., Boy, M., et al. (2006). Critical 886 assessment of the current state of scientific knowledge, terminology, and research 887 needs concerning the role of organic aerosols in the atmosphere, climate, and global 888 change. Atmospheric Chemistry and Physics, 6(7), 2017–2038. 889 Gargallo-Garriga, A., Sardans, J., Pérez-Trujillo, M., Oravec, M., Urban, O., Jentsch, A., et al. 890 (2015). Warming differentially influences the effects of drought on stoichiometry and 891 metabolomics in shoots and roots. New phytologist, 207(3), 591–603. 892 Gargallo-Garriga, A., Sardans, J., Pérez-Trujillo, M., Rivas-Ubach, A., Oravec, M., Vecerova, K., 893 et al. (2014). Opposite metabolic responses of shoots and roots to drought. Scientific 894 reports, 4, 6829. 895 Gibney, M. J., Walsh, M., Brennan, L., Roche, H. M., German, B., & van Ommen, B. (2005). 896 Metabolomics in human nutrition: opportunities and challenges. The American journal 897 of clinical nutrition, 82(3), 497–503. 898 Glauser, G., Guillarme, D., Grata, E., Boccard, J., Thiocone, A., Carrupt, P.-A., et al. (2008). 899 Optimized liquid chromatography–mass spectrometry approach for the isolation of 900 minor stress biomarkers in plant extracts and their identification by capillary nuclear 901 magnetic resonance. Journal of Chromatography A, 1180(1), 90–98. 902 Gu, L., Baldocchi, D., Verma, S. B., Black, T. A., Vesala, T., Falge, E. M., & Dowty, P. R. (2002). 903 Advantages of diffuse radiation for terrestrial ecosystem productivity. Journal of 904 Geophysical Research: Atmospheres, 107(D6), 4050. 905 Gullberg, J., Jonsson, P., Nordström, A., Sjöström, M., & Moritz, T. (2004). Design of 906 experiments: an efficient strategy to identify factors influencing extraction and 907 derivatization of Arabidopsis thaliana samples in metabolomic studies with gas 908 chromatography/mass spectrometry. Analytical Biochemistry, 331(2), 283–295. 909 Guy, C., Kaplan, F., Kopka, J., Selbig, J., & Hincha, D. K. (2007). Metabolomics of temperature 910 stress. Physiologia Plantarum, 132(2), 220-235. 27 911 Hall, R. D. (2006). Plant metabolomics: from holistic hope, to hype, to hot topic. New 912 Phytologist, 169(3), 453–468. 913 Hiller, K., Hangebrauk, J., Jäger, C., Spura, J., Schreiber, K., & Schomburg, D. (2009). 914 MetaboliteDetector: Comprehensive Analysis Tool for Targeted and Nontargeted 915 GC/MS Based Metabolome Analysis. Analytical Chemistry, 81(9), 3429–3439. 916 Hirai, M. Y., Yano, M., Goodenowe, D. B., Kanaya, S., Kimura, T., Awazuhara, M., et al. (2004). 917 From The Cover: Integration of transcriptomics and metabolomics for understanding 918 of global responses to nutritional stresses in Arabidopsis thaliana. Proceedings of the 919 National Academy of Sciences, 101(27), 10205–10210. 920 Husson, F., & Josse, J. (2015). missMDA: Handling Missing Values with Multivariate Data 921 Analysis. R package version 1.9., http://CRA. 922 Husson, F., Josse, J., Le, S., & Mazet, J. (2016). FactoMineR: Multivariate Exploratory Data 923 Analysis and Data Mining. R package version 1.32., http://CRA. 924 Ingram, J., & Bartels, D. (1996). The molecular basis of dehydration tolerance in plants. Annual 925 review of plant biology, 47(1), 377–403. 926 Jokinen, T., Berndt, T., Makkonen, R., Kerminen, V.-M., Junninen, H., Paasonen, P., et al. 927 (2015). Production of extremely low volatile organic compounds from biogenic 928 emissions: Measured yields and atmospheric implications. Proceedings of the National 929 Academy of Sciences, 112(23), 7123–7128. 930 Kaplan, F., Kopka, J., Haskell, D. W., Zhao, W., Schiller, K. C., Gatzke, N., et al. (2004). Exploring 931 the Temperature-Stress Metabolome of Arabidopsis. Plant physiology, 136(4), 4159– 932 4168. 933 Kellerman, A. M., Dittmar, T., Kothawala, D. N., & Tranvik, L. J. (2014). Chemodiversity of 934 dissolved organic matter in lakes driven by climate and hydrology. Nature 935 Communications, 5, 3804. 936 Keltjens, W. G., & van Beusichem, M. L. (1998). Phytochelatins as biomarkers for heavy metal 937 stress in maize (Zea mays L.) and wheat (Triticum aestivum L.): combined effects of 938 copper and cadmium. Plant and Soil, 203(1), 119–126. 939 Kim, H. K., Choi, Y. H., & Verpoorte, R. (2010). NMR-based metabolomic analysis of plants. 940 Nature protocols, 5(3), 536–49. 28 941 Kim, S., Kramer, R. W., & Hatcher, P. G. (2003). Graphical Method for Analysis of Ultrahigh- 942 Resolution Broadband Mass Spectra of Natural Organic Matter, the Van Krevelen 943 Diagram. Analytical Chemistry, 75(20), 5336–5344. 944 Kim, Y.-M., Nowack, S., Olsen, M. T., Becraft, E. D., Wood, J. M., Thiel, V., et al. (2015). Diel 945 metabolomics analysis of a hot spring chlorophototrophic microbial mat leads to new 946 hypotheses of community member metabolisms. Frontiers in microbiology, 6, 209. 947 Kind, T., Wohlgemuth, G., Lee, D. Y., Lu, Y., Palazoglu, M., Shahbaz, S., & Fiehn, O. (2009). 948 FiehnLib: Mass Spectral and Retention Index Libraries for Metabolomics Based on 949 Quadrupole and Time-of-Flight Gas Chromatography/Mass Spectrometry. Analytical 950 Chemistry, 81(24), 10038–10048. 951 Klein, G. C., Rodgers, R. P., & Marshall, A. G. (2006). Identification of hydrotreatment-resistant 952 heteroatomic species in a crude oil distillation cut by electrospray ionization FT-ICR 953 mass spectrometry. Fuel, 85, 2071–2080. 954 Kujawinski, E. (2002). Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass 955 Spectrometry (ESI FT-ICR MS): Characterization of Complex Environmental Mixtures. 956 Environmental Forensics, 3(3), 207–216. 957 Kujawinski, E. B., & Behn, M. D. (2006). Automated Analysis of Electrospray Ionization Fourier 958 Transform Ion Cyclotron Resonance Mass Spectra of Natural Organic Matter. Analytical 959 Chemistry, 78(13), 4363–4373. 960 Lee, D. Y., & Fiehn, O. (2013). Metabolomic response of Chlamydomonas reinhardtii to the 961 inhibition of target of rapamycin (TOR) by rapamycin. Journal of microbiology and 962 biotechnology, 23(7), 923–31. L 963 Lei, Z., Huhman, D., & Sumner, LL. (2011). Mass spectrometry strategies in metabolomics. 964 Journal of Biological Chemistry. doi: 10.1074/jbc.R111.238691 965 Leiss, K. A., Choi, Y. H., Abdel-Farid, I. B., Verpoorte, R., & Klinkhamer, P. G. L. (2009). NMR 966 metabolomics of thrips (Frankliniella occidentalis) resistance in Senecio hybrids. 967 Journal of chemical ecology, 35(2), 219–29. 968 Leiss, K. A., Cristofori, G., van Steenis, R., Verpoorte, R., & Klinkhamer, P. G. L. (2013). An eco- 969 metabolomic study of host plant resistance to Western flower thrips in cultivated, 970 biofortified and wild carrots. Phytochemistry, 93, 63–70. 29 971 Lin, C. Y., Viant, M. R., & Tjeerdema, R. S. (2006). Metabolomics: Methodologies and 972 applications in the environmental sciences. Journal of Pesticide Science, 31(3), 245– 973 251. 974 Lindon, J.C., Nicholson, J.K. & Holmes, E. (eds.) The Handbook of Metabonomics and 975 Metabolomics (Elsevier, Amsterdam, 2007), 55-201. 976 Lindow, S. E., & Brandl, M. T. (2003). Microbiology of the phyllosphere. Applied and 977 environmental microbiology, 69(4), 1875–83. 978 Liu, Y., & Kujawinski, E. B. (2015). Chemical Composition and Potential Environmental Impacts 979 of Water-Soluble Polar Crude Oil Components Inferred from ESI FT-ICR MS. PLOS ONE, 980 10(9), e0136376. 981 Macedo, A. F. (2012). Abiotic Stress Responses in Plants: Metabolism to Productivity. In Abiotic 982 Stress Responses in Plants (pp. 41–61). New York, NY: Springer New York. 983 Mahowald, N. M., Artaxo, P., Baker, A. R., Jickells, T. D., Okin, G. S., Randerson, J. T., & 984 Townsend, A. R. (2005). Impacts of biomass burning emissions and land use change on 985 Amazonian atmospheric phosphorus cycling and deposition. Global Biogeochemical 986 Cycles, 19, GC4030. 987 Mari, A., Lyon, D., Fragner, L., Montoro, P., Piacente, S., Wienkoop, S., et al. (2013). 988 Phytochemical composition of Potentilla anserina L. analyzed by an integrative GC-MS 989 and LC-MS metabolomics platform. Metabolomics : Official journal of the Metabolomic 990 Society, 9(3), 599–607. 991 Medeiros, P. M., Babcock-Adamos, L., Seidel, M., Castelao, R. M., Di Iorio, D., Hollibaugh, J. T., 992 & Dittmar, T. (2017). Export of terrigenous dissolved organic matter in a broad 993 continental shelf. Limnology and Oceanography, 62, 1718-1731. 994 Menzel, A., Sparks, T. H., Estrella, N., Koch, E., Aasa, A., Ahas, R., et al. (2006). European 995 phenological response to climate change matches the warming pattern. Global Change 996 Biology, 12(10), 1969–1976. 997 Minor, E. C., Swenson, M. M., Mattson, B. M., & Oyler, A. R. (2014). Structural characterization 998 of dissolved organic matter: a review of current techniques for isolation and analysis. 999 Environ. Sci.: Processes Impacts, 16(9), 2064–2079. 30 1000 Nikiforova, V. J., Kopka, J., Tolstikov, V., Fiehn, O., Hopkins, L., Hawkesford, M. J., et al. (2005). 1001 Systems Rebalancing of Metabolism in Response to Sulfur Deprivation, as Revealed by 1002 Metabolome Analysis of Arabidopsis Plants. Plant physiology, 138(1), 304–318. 1003 Obee, T. N., & Hay, S. O. (1997). Effects of Moisture and Temperature on the Photooxidation of 1004 Ethylene on Titania. Environmental Science & Technology, 31(7), 2034–2038. 1005 Oksanen, J., Guillaume-Blanchet, F., Kindt, R., Legendre, P., Minchin, P., O’Hara, R., et al. 1006 (2013). vegan: Community Ecology Package. R package version 2.0-9, http ://CRAN.R- 1007 project.org/package=vegan 1008 Osterholz, H., Singer, G., Wemheuer, B., Daniel, R., Simon, M., Niggemann, J., & Dittmar, T. 1009 (2016). Deciphering associations between dissolved organic molecules and bacterial 1010 communities in a pelagic marine system. The ISME Journal, 10(7), 1717-1730. 1011 Paerl, H. W. (1997). Coastal eutrophication and harmful algal blooms: Importance of 1012 atmospheric deposition and groundwater as “new” nitrogen and other nutrient 1013 sources. Limnology and Oceanography, 42(5part2), 1154–1165. 1014 Pan, Z., & Raftery, D. (2007). Comparing and combining NMR spectroscopy and mass 1015 spectrometry in metabolomics. Analytical and Bioanalytical Chemistry, 387(2), 525– 1016 527. 1017 Pandis, S. N., Harley, R. A., Cass, G. R., & Seinfeld, J. H. (1992). Secondary organic aerosol 1018 formation and transport. Atmospheric Environment. Part A. General Topics, 26(13), 1019 2269–2282. 1020 Parmesan, C. (2006). Ecological and Evolutionary Responses to Recent Climate Change. Annual 1021 Review of Ecology, Evolution, and Systematics, 37(1), 637–669. 1022 Parmesan, C., & Yohe, G. (2003). A globally coherent fingerprint of climate change impacts 1023 across natural systems. Nature, 421(6918), 37–42. 1024 Paytan, A., Mackey, K. R. M., Chen, Y., Lima, I. D., Doney, S. C., Mahowald, N., et al. (2009). 1025 Toxicity of atmospheric aerosols on marine phytoplankton. Proceedings of the 1026 National Academy of Sciences of the United States of America, 106(12), 4601–5. 1027 Peñuelas, J., & Sardans, J. (2009). Ecological metabolomics. Chemistry and Ecology, 25(4), 305– 1028 309. 31 1029 Peñuelas, J., Sardans, J., Rivas-ubach, A., & Janssens, I. A. (2012). The human-induced 1030 imbalance between C, N and P in Earth’s life system. Global Change Biology, 18(1), 3-6. 1031 Peñuelas, J., & Staudt, M. (2010). BVOCs and global change. Trends in plant science, 15(3), 1032 133–44. 1033 Peñuelas, J., & Terradas, J. (2014). The foliar microbiome. Trends in Plant Science, 19(5), 278- 1034 280. 1035 Pluskal, T., Castillo, S., Villar-Briones, A., & Orešič, M. (2010). MZmine 2: modular framework 1036 for processing, visualizing, and analyzing mass spectrometry-based molecular profile 1037 data. BMC bioinformatics, 11(1), 395. 1038 R Core Team. (2013). R: A language and environment for statistical computing. Vienna. 1039 Ramanathan, V., Crutzen, P. J., Kiehl, J. T., & Rosenfeld, D. (2001). Aerosols, climate, and the 1040 hydrological cycle. Science, 294(5549), 2119–24. 1041 Reemtsma, T. (2009). Determination of molecular formulas of natural organic matter 1042 molecules by (ultra-) high-resolution mass spectrometry: Status and needs. Journal of 1043 Chromatography A, 1216(18), 3687–3701. 1044 Riedel, T., & Dittmar, T. (2014). A Method Detection Limit for the Analysis of Natural Organic 1045 Matter via Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Analytical 1046 Chemistry, 86(16), 8376–8382. 1047 Rivas-Ubach, A., Barbeta, A., Sardans, J., Guenther, A., Ogaya, R., Oravec, M., et al. (2016b). 1048 Topsoil depth substantially influences the responses to drought of the foliar 1049 metabolomes of Mediterranean forests. Perspectives in Plant Ecology, Evolution and 1050 Systematics, 21. 1051 Rivas-Ubach, A., Gargallo-Garriga, A., Sardans, J., Oravec, M., Mateu-Castell, L., Pérez-Trujillo, 1052 M., et al. (2014). Drought enhances folivory by shifting foliar metabolomes in Quercus 1053 ilex trees. New Phytologist, 202(3). 1054 Rivas-Ubach, A., Hódar, J. A., Sardans, J., Kyle, J. E., Kim, Y.-M., Oravec, M., et al. (2016a). Are 1055 the metabolomic responses to folivory of closely related plant species linked to 1056 macroevolutionary and plant–folivore coevolutionary processes? Ecology and 1057 Evolution, 6(13), 4372-4386. 32 1058 Rivas-Ubach, A., Liu, Y., Bianchi, T. S., Tolić, N., Jansson, C., & Paša-Tolić, L. (2018). Moving 1059 beyond the van Krevelen Diagram: A New Stoichiometric Approach for Compound 1060 Classification in Organisms. Analytical Chemistry, acs.analchem.8b00529. 1061 doi:10.1021/acs.analchem.8b00529 1062 Rivas-Ubach, A., Pérez-Trujillo, M., Sardans, J., Gargallo-Garriga, A., Parella, T., & Peñuelas, J. 1063 (2013). Ecometabolomics: Optimized NMR-based method. Methods in Ecology and 1064 Evolution, 4(5), 464-473. 1065 Rivas-Ubach, A., Poret-Peterson, A. T., Peñuelas, J., Sardans, J., Pérez-Trujillo, M., Legido- 1066 Quigley, C., et al. (2018). Coping with iron limitation: a metabolomic study of 1067 Synechocystis sp. PCC 6803. Acta Physiologiae Plantarum, 40(2), 28. 1068 Rivas-Ubach, A., Sardans, J., Hódar, J. A., Garcia-Porta, J., Guenther, A., Oravec, M., et al. 1069 (2016c). Similar local, but different systemic, metabolomic responses of closely related 1070 pine subspecies to folivory by caterpillars of the processionary moth. Plant Biology, 1071 18(3), 484–494. 1072 Rivas-Ubach, A., Sardans, J., Hódar, J. A., Garcia-Porta, J., Guenther, A., Paša-Tolić, L., et al. 1073 (2017). Close and distant: Contrasting the metabolism of two closely related 1074 subspecies of Scots pine under the effects of folivory and summer drought. Ecology 1075 and Evolution, 7(21), 8976–8988. 1076 Rivas-Ubach, A., Sardans, J., Peŕez-Trujillo, M., Estiarte, M., & Penũelas, J. (2012). Strong 1077 relationship between elemental stoichiometry and metabolome in plants. Proceedings 1078 of the National Academy of Sciences of the United States of America, 109(11), 4181- 1079 4186. 1080 Rochford, S. (2005). Metabolomics reviewed: A new “omics” platform technology for systems 1081 biology and implications for natural products research. Journal of Natural Products, 68, 1082 1813-1820. 1083 Roullier-Gall, C., Boutegrabet, L., Gougeon, R. D., & Schmitt-Kopplin, P. (2014). A grape and 1084 wine chemodiversity comparison of different appellations in Burgundy: Vintage vs 1085 terroir effects. Food Chemistry, 152, 100–107. 1086 Roulston, T. H., & Cane, J. H. (2000). Pollen nutritional content and digestibility for animals. 1087 Plant Systematics and Evolution, 222(1–4), 187–209. 33 1088 Saito, K., & Matsuda, F. (2010). Metabolomics for functional genomics, systems biology, and 1089 biotechnology. Annual Review of Plant Biology, 361, 463-489. 1090 Sardans, J., Gargallo-Garriga, A., Pérez-Trujillo, M., Parella, T. J., Seco, R., Filella, I., & Peñuelas, 1091 J. (2014). Metabolic responses of Quercus ilex seedlings to wounding analysed with 1092 nuclear magnetic resonance profiling. Plant biology, 16(2), 395–403. 1093 Sardans, J., Peñuelas, J., & Rivas-Ubach, A. (2011). Ecological metabolomics: overview of 1094 current developments and future challenges. Chemoecology, 21(4), 191–225. 1095 Sardans, J., Rivas-Ubach, A., & Peñuelas, J. (2012a). The elemental stoichiometry of aquatic and 1096 terrestrial ecosystems and its relationships with organismic lifestyle and ecosystem 1097 structure and function: A review and perspectives. Biogeochemistry, 111(1–3). 1098 Schmitt-Kopplin, P., Liger-Belair, G., Koch, B. P., Flerus, R., Kattner, G., Harir, M., et al. (2012). 1099 Dissolved organic matter in sea spray: a transfer study from marine surface water to 1100 aerosols. Biogeosciences, 9(4), 1571–1582. 1101 Shulaev, V. (2006). Metabolomics technology and bioinformatics. Briefings in Bioinformatics, 1102 7(2), 128-139. 1103 Seco, R., Peñuelas, J., & Filella, I. (2007). Short-chain oxygenated VOCs: Emission and uptake by 1104 plants and atmospheric sources, sinks, and concentrations. Atmospheric Environment, 1105 41(12), 2477–2499. 1106 Shulaev, V., Cortes, D., Miller, G., & Mittler, R. (2008). Metabolomics for plant stress response. 1107 Physiologia Plantarum, 132(2), 199–208. 1108 Sleighter, R. L., & Hatcher, P. G. (2007). The application of electrospray ionization coupled to 1109 ultrahigh resolution mass spectrometry for the molecular characterization of natural 1110 organic matter. Journal of mass spectrometry : JMS, 42(5), 559–74. 1111 Smith, D. F., Podgorski, D. C., Rodgers, R. P., Blakney, G. T., & Hendrickson, C. L. (2018). 21 1112 Tesla FT-ICR Mass Spectrometer for Ultrahigh-Resolution Analysis of Complex Organic 1113 Mixtures. Analytical Chemistry, doi:10.1021/acs.analchem.7b04159 1114 Smith, D., & Španěl, P. (2011). Direct, rapid quantitative analyses of BVOCs using SIFT-MS and 1115 PTR-MS obviating sample collection. TrAC Trends in Analytical Chemistry, 30(7), 945– 1116 959. 34 1117 Solberg, Y., & Remedios, G. (1980). Chemical composition of pure and Bee-Collected pollen. 1118 Meldinger fra Norges landbrukshogskole, 59, 1–13. 1119 Spencer, R. G. M., Mann, P. J., Dittmar, T., Eglinton, T. I., McIntyre, C., Holmes, R. M., et al. 1120 (2015). Detecting the signature of permafrost thaw in Arctic rivers. Geophysical 1121 Research Letters, 42(8), 2830–2835. 1122 Sterner, R., & Elser, J. (2002). Ecological Stoichiometry: The Biology of Elements from 1123 Molecules to the Biosphere. Princetion University Press. 1124 Sumner, L. W., Amberg, A., Barrett, D., Beale, M. H., Beger, R., Daykin, C. A., et al. (2007). 1125 Proposed minimum reporting standards for chemical analysis. Metabolomics, 3(3), 1126 211–221. 1127 t’Kindt, R., De Veylder, L., Storme, M., Deforce, D., & Van Bocxlaer, J. (2008). LC-MS metabolic 1128 profiling of Arabidopsis thaliana plant leaves and cell cultures: optimization of pre-LC- 1129 MS procedure parameters. Journal of chromatography. B, Analytical technologies in 1130 the biomedical and life sciences, 871(1), 37–43. 1131 Tfaily, M. M., Chu, R. K., Tolić, N., Roscioli, K. M., Anderton, C. R., Paša-Tolić, L., et al. (2015). 1132 Advanced Solvent Based Methods for Molecular Characterization of Soil Organic 1133 Matter by High-Resolution Mass Spectrometry. Analytical Chemistry, 87(10), 5206– 1134 5215. 1135 Tholl, D., Boland, W., Hansel, A., Loreto, F., Röse, U. S. R., & Schnitzler, J.-P. (2006). Practical 1136 approaches to plant volatile analysis. The Plant Journal, 45(4), 540–560. 1137 Thompson, H., Heimendinger, J., Gillette, C., Sedlacek, S., Haegele, A., O’Neill, C., & Wolfe, P. 1138 (2005). In Vivo Investigation of Changes in Biomarkers of Oxidative Stress Induced by 1139 Plant Food Rich Diets. Journal of Agricultural and food Chemistry, 53(15), 6126–6132. 1140 Tolić, N., Liu, Y., Liyu, A., Shen, Y., Tfaily, M. M., Kujawinski, E. B., et al. (2017). Formularity: 1141 Software for Automated Formula Assignment of Natural and Other Organic Matter 1142 from Ultrahigh-Resolution Mass Spectra. Analytical Chemistry, 1143 doi:10.1021/acs.analchem.7b03318 1144 Uzu, G., Sobanska, S., Sarret, G., Muñoz, M., & Dumat, C. (2010). Foliar Lead Uptake by Lettuce 1145 Exposed to Atmospheric Fallouts. Environmental Science & Technology, 44(3), 1036– 1146 1042. 35 1147 van den Berg, R. A., Hoefsloot, H. C., Westerhuis, J. A., Smilde, A. K., & van der Werf, M. J. 1148 (2006). Centering, scaling, and transformations: improving the biological information 1149 content of metabolomics data. BMC Genomics, 7(1), 142. 1150 van Krevelen, D. (1950). Graphical-statistical method for the study of structure and reaction 1151 processes of coal. Fuel, 29, 269–284. 1152 Vorholt, J. A. (2012). Microbial life in the phyllosphere. Nature Reviews Microbiology, 10(12), 1153 828–840. 1154 Walsh, M. C., Brennan, L., Malthouse, J. P. G., Roche, H. M., & Gibney, M. J. (2006). Effect of 1155 acute dietary standardization on the urinary, plasma, and salivary metabolomic 1156 profiles of healthy humans. The American journal of clinical nutrition, 84(3), 531–9. 1157 Walther, G. R., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T. J. C., et al. (2002). 1158 Ecological responses to recent climate change. Nature, 416(6879), 389–395. 1159 Wang, R., Balkanski, Y., Bopp, L., Aumont, O., Boucher, O., Ciais, P., et al. (2015). Influence of 1160 anthropogenic aerosol deposition on the relationship between oceanic productivity 1161 and warming. Geophysical Research Letters, 42(24), 10745–10754. 1162 Warnes, G. R., Bolker, B., Bonebakker, L., Gentleman, R., Huber Andy Liaw, W., Lumley, T., et 1163 al. (2016). gplots: various R programing tools for plotting data. R package version 3.0.1, 1164 http://CRAN.R-project.org/package=gplots 1165 Wedding, J. B., Carlson, R. W., Stukel, J. J., & Bazzaz, F. A. (1975). Aerosol deposition on plant 1166 leaves. Environmental Science & Technology, 9(2), 151–153. 1167 Whipps, J. M., Hand, P., Pink, D., & Bending, G. D. (2008). Phyllosphere microbiology with 1168 special reference to diversity and plant genotype. Journal of Applied Microbiology, 1169 105(6), 1744–1755. 1170 White, R. A., Rivas-Ubach, A., Borkum, M. I., Köberl, M., Bilbao, A., Colby, S. M., et al. (2017). 1171 The state of rhizospheric science in the era of multi-omics: A practical guide to omics 1172 technologies. Rhizosphere, 3, 212–221. doi:10.1016/J.RHISPH.2017.05.003 1173 Wishart, D. S. (2008). Metabolomics: applications to food science and nutrition research. 1174 Trends in Food Science & Technology, 19(9), 482–493. 36 1175 Wozniak, A. S., Bauer, J. E., Sleighter, R. L., Dickhut, R. M., & Hatcher, P. G. (2008). Technical 1176 Note: Molecular characterization of aerosol-derived water soluble organic carbon 1177 using ultrahigh resolution electrospray ionization Fourier transform ion cyclotron 1178 resonance mass spectrometry. Atmospheric Chemistry and Physics, 8(17), 5099–5111. 1179 Xiong, T.-T., Leveque, T., Austruy, A., Goix, S., Schreck, E., Dappe, V., et al. (2014). Foliar uptake 1180 and metal(loid) bioaccessibility in vegetables exposed to particulate matter. 1181 Environmental Geochemistry and Health, 36(5), 897–909. 1182 Zhang, A., Sun, H., Wang, P., Han, Y., & Wang, X. (2012). Modern analytical techniques in 1183 metabolomics analysis. The Analyst, 137(2), 293–300. 1184 Zhang, Q., Stanier, C. O., Canagaratna, M. R., Jayne, J. T., Worsnop, D. R., Pandis, S. N., & 1185 Jimenez, J. L. (2004). Insights into the Chemistry of New Particle Formation and Growth 1186 Events in Pittsburgh Based on Aerosol Mass Spectrometry. Environmental Science & 1187 Technology, 38(18), 4797–4809. 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 37 1201 Table 1. PERMANOVAs of the atmo-metabolome fingerprints generated by LC-MS and GC-MS 1202 instruments for overall metabolome comparison between seasons. Spring vs. Summer dataset (aerosols collected at PNNL) LC-MS Sum of Squares Mean Square F P Season 1 0.12 0.12 13.8 < 0.0001 Residuals 28 0.24 0.009 Total 29 0.36 GC-MS Sum of Squares Mean Square F P Season 1 0.027 0.027 9.69 < 0.0001 Residuals 28 0.079 0.0028 Total 29 0.11 Pollination Peak Periods dataset (aerosols collected at UMICH) LC-MS Sum of Squares Mean Square F P Season 2 0.12 0.06 9.54 0.0039 Residuals 6 0.037 0.006 Total 8 0.16 GC-MS Sum of Squares Mean Square F P Season 2 0.14 0.07 7.63 0.0041 Residuals 6 0.05 0.009 Total 8 0.19 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 38 1215 Figure with captions. 1216 1217 1218 1219 1220 Figure 1. Schematic diagram of the aerosol emissions and deposition on ecosystems. 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 39 1238 1239 1240 Figure 2. Diagram of the main procedures of a general ecometabolomic study combined with 1241 complementary measurements. 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 40 1260 1261 1262 Figure 3. Sampling year, collection period, and analytical instruments and software used for 1263 generating the datasets from each of the sampling site and campaign. Table numbers with the 1264 parameters used for dataset generation for each of the software are indicated. Table and 1265 figure numbers containing the main results derived from aerosols analyzed from each 1266 sampling campaign and analytical technique are also indicated. 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 41 1284 1285 1286 Figure 4. Experimental procedures performed on filters to obtain the metabolomic extracts 1287 from aerosol samples for subsequent analyses with distinct mass spectrometry techniques. 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 42 1299 1300 1301 Figure 5. Case plots of the Principal Component (PC) 1 versus the PC2 of the PCAs conducted 1302 on the aerosol metabolomic fingerprints collected at PNNL site contrasting spring vs. summer 1303 aerosols (a for LC-MS; b for GC-MS) and collected at UMICH site contrasting different 1304 pollination peak periods (c for LC-MS; d for GC-MS). Each filter sample corresponds to a point 1305 for each of the case plots. For PNNL site (spring vs. summer); spring aerosol samples are 1306 represented by blue triangles and summer aerosol samples are represented by red circles. For 1307 UMICH site (pollination peak periods); samples collected during birch, oak and pine pollination 1308 peaks are represented by violet squares, green diamonds, and orange crosses, respectively. 1309 1310 1311 43 1312 1313 1314 Figure 6. Heatmaps of the metabolomic fingerprints obtained from LC-MS and GC-MS for PNNL 1315 metabolomics datasets (spring vs. summer) (a and b) and for the UMICH metabolomics 1316 datasets (pollination peak periods) (c and d). Cluster dendrogram for the variables is shown for 1317 each heatmap. Blue rectangles point cluster of variables for the specific group of samples 1318 having higher overall relative abundance values. The color gradient represent the relative 1319 abundance of the metabolomic feature between samples. Color red represents the highest 1320 relative abundance. 1321 1322 1323 1324 1325 1326 44 1327 1328 1329 Figure 7. Heat maps of the identified metabolomic features from LC-MS and GC-MS for the 1330 PNNL metabolomics dataset (spring vs. summer) (a) and for the UMICH metabolomics dataset 1331 (pollination peak periods dataset)(b). For each site, heatmaps of the identified features were 1332 plotted using LC-MS and GC-MS data combined. The color gradient represent the relative 1333 abundance of the metabolomic feature between samples. Color red represents the highest 1334 relative abundance. Variables that presented significance difference (P < 0.05) after t-test (a) 1335 or after oneway ANOVA (b) are marked by an asterisk. Blue rectangles indicate the group of 1336 samples having higher average relative abundance value for each specific variable that 1337 presented statistical significance. For the pollination peak period (b), different letters next to 1338 the heatmap indicate significant differences between groups after Tukey HSD posthoc test. 1339 1340 1341 1342 1343 45 1344 1345 1346 Figure 8. Box plots for the proportion (%) of different formula classes (CHNO, CHO and CHOS 1347 (a) and CHNOS, CHNOP, CHOSP, CHNOSP and CHOP (b)) for PNNL site DI-FT-ICR-MS dataset 1348 (spring vs. summer). Boxes show the median value for each formula class and season. Open 1349 dots correspond to extreme values. Asterisks indicate statistical significance between aerosols 1350 collected spring and summer for each formula class (P < 0.05 (*); P < 0.0001 (****)). Black dots 1351 indicate marginal significance between spring and summer aerosols for each formula class (P < 1352 0.1). 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 46 1365 1366 1367 Figure 9. Carbon number versus m/z (CvM) (a) and Oxygen/Carbon ratio versus mass (b) 1368 diagrams plotted from DI-FT-ICR-MS datasets (PNNL site). Different color of dots represent the 1369 relative intensity of spring relative to summer for each metabolic feature; darker red dots 1370 represent higher relative intensity in summer and darker blue dots represent higher relative 1371 intensity in spring. Mean (±SE) of Oxygen/Carbon of the metabolic features detected in 1372 aerosols collected in spring and summer (c). Statistic-t and P value are indicated in the graph. 1373 47