Probing the chemical transformation of seawater- soluble crude oil components during microbial oxidation

Studies assessing the environmental impacts of oil spills focus primarily on the non-water-soluble components, leaving the fate of the water-soluble fraction (WSF) largely unexplored. We employed u...


ABSTRACT 21
Studies assessing the environmental impacts of oil spills focus primarily on the non-water-soluble 22 components, leaving the fate of the water-soluble fraction (WSF) largely unexplored. We 23 employed untargeted chemical analysis along with biological information to probe the 24 transformation of crude oil WSF in seawater, in the absence of light, in a laboratory experiment. 25 Over a 14-day incubation, microbes transformed WSF into various metabolic intermediates, 26 without significantly altering the dissolved organic carbon concentrations. Microbial 27 transformation processes increased the chemical diversity and overall oxygen content of WSF 28 compounds, concomitant with an increase in dioxygenase gene abundances. While the majority of 29 metabolites formed from the transformation of WSF could not be structurally identified with 30 existing databases, elemental formulas suggest that many of these compounds could be oxidation 31 products of water-soluble non-polar compounds such as PAHs. In particular, metabolites with 32 three oxygen atoms may represent a key transition point for WSF degradation. One such 33 compound, salicylic acid, likely provides a route for complete WSF remineralization, as it is labile 34 to non-oil degrading marine bacteria. The environmental persistence and toxicity of WSF 35 metabolic products are still unknown, but results from this study provide a framework for further 36 exploration of the fate of WSF in marine ecosystems. 37 38 39

INTRODUCTION 40
Millions of barrels of crude oil are released to the ocean each year from unintentional spillages 41 and natural seepage. 1 A small, but significant, fraction of the oil dissolves in the water and behaves 42 differently than the bulk oil. The composition of dissolved oil is distinct from the total oil and is 43 enriched in small (<1000 Da), polar molecules. 2 Despite decades of study on the fate of oil in the 44 environment, the water-soluble fraction (WSF) is vastly understudied because its components are 45 not resolved in traditional gas-chromatography (GC)-based analytical methods. Consequently, we 46 know much less about the factors affecting the fate and transport of crude oil WSF in marine 47 ecosystems, despite evidence suggesting that this fraction is enriched during weathering and is 48 more toxic to aquatic organisms than the parent oil. 3-6 49 WSF can be preferentially enriched in the aqueous phase at any oil-water interface, such as in 50 surface waters in contact with oil slicks, seawater around oil seeps or deep-sea oil spills, as well 51 as in water-inundated oil-contaminated soil. The best-studied case for such oil-water partitioning 52 phenomena was at the Deepwater Horizon (DWH) drill site in 2010, where 3.19 million barrels of 53 This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license.
at -20°C until mass spectrometry analysis. Immediately before analysis, we dried down the eluents 154 to near dryness and reconstituted them in 250 μL of 95:5 water:acetonitrile. We added deuterated 155 biotin (5 μL; final concentration 0.05 μg/mL) to each sample as an internal standard. PPL resins 156 preferentially capture the aromatic compounds in WSF, but do not retain very small, highly polar 157 molecules such as succinic acid. 51 Therefore, to ensure similar extracted carbon concentrations 158 across all treatments and within the pooled samples, we diluted the WSF treatment eluents 50× 159 due to high concentrations of extracted carbon relative to the non-WSF treatments. We combined 160 50 μL of each sample to create a pooled sample as a reference for data quality control and 161

processing. 162
We divided all samples into two equal volumes, one for targeted analysis and one for untargeted 163 analysis, following methods described by Kido Soule et al. 52 The untargeted approach, using liquid WSF components compared to other organic substrates over the 14-day experiment. Here, we 170 define a metabolite profile as all the features in a given sample, where a feature is defined as a 171 unique combination of mass to charge ratio (m/z) and retention time (RT). Each feature 172 corresponds to a specific metabolite or to a group of co-eluting isomers. 173 We subjected the top four features in each mass scan to tandem mass spectrometry (MS/MS or 174 MS2) for compound identification. We applied rigorous data quality control procedures to ensure 175 data robustness, e.g. removing features whose variability is driven by instrumental parameters This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license. and/or are consistently present in all samples at similar intensities (see Supporting Information). 177 After these measures, approximately 40-48% of features in the dataset had associated MS2 spectra. 178 We used a step-wise approach to classify metabolites of interests based on the Metabolomics 179 media. Among the three treatments, the chemical and microbial community compositions were 239 different between non-WSF (i.e. VSW and succinic acid) and WSF incubations ( Figures S2a and  240 b). The use of succinic acid as a control carbon source allowed us to determine that this observed 241 shift in microbial community is due specifically to the addition of WSF and not due to 242 opportunistic microbes. Indeed, we observed known hydrocarbon degraders uniquely in the WSF 243 treatment, including Cycloclasticus, Oceaniserpentilla, and Rhodospirillales, consistent with field 244 observations of microbial diversity during the DWH spill (Table S5 and S6). 11, 24, 57

WSF is Transformed in Incubation Experiments. We used changes in DOC concentrations 246
across time points to evaluate complete remineralization of organic carbon to CO2 in each 247 treatment. DOC values were not statistically significantly different across the 14-day incubation 248 in the WSF treatments, based on a one-way ANOVA test at 95% confidence level ( Figure S3). In 249 contrast, DOC concentrations decreased more than 100 μM-C by T = 14 in the succinic acid 250 treatment, suggesting catabolism of succinate for energy ( Figure S3). DOC concentrations in the 251 VSW treatment were also not statistically significantly different across the three time points, which 252 was likely due to the overall lower biomass and microbial activity as a result of lower DOC 253

concentrations (Figures S4). 254
We used LC-FT-ICR MS to examine detailed chemical changes as the microbes altered WSF. 255 We define polar WSF-derived chemical features (P-WSFTotal), resolved by LC-FT-ICR MS, to be 256 those found in WSF treatment but not in any non-WSF treatments. We culled the list of features 257 to those that were found in all replicates at a time point. We then divided P-WSFTotal into P-WSF0, 258 or the polar WSF compounds found at T = 0, and P-WSFM, or the polar metabolites produced 259 during microbial degradation of WSF compounds ( Figure S5). 260 To understand the chemical dynamics within WSF over the course of our experiment, we further 261 subdivided P-WSF0 into four groups: P-WSF0-C, or features that were likely consumed completely 262 (present only at T = 0); P-WSF0-U, or features that were unaltered (similar relative abundance 263 values over 14 days); P-WSF0-I, or features whose relative abundances increased over 14 days; and 264 P-WSF0-D, or features whose relative abundances decreased over 14 days. The last three groups 265 (P-WSF0-U, P-WSF0-I, and P-WSF0-D) include features present in WSF treatment samples at all 266 time points. We based our feature classifications on pair-wise one-tailed Student's t-test of feature 267 intensities. A visual overview of our classification scheme is shown in Figure S5.
Overall, the features in P-WSFTotal increased from 449 at T = 0 to 741 at T = 14 ( Figure 1; Table  269 S1), indicating formation of new compounds as a result of microbial transformation of WSF crude 270 oil. Only 80 of 449 (<18 %) P-WSF0 features were missing by the end of the experiment, due 271 either to complete degradation or to reduction below the detection limit ( Figure 1; Table S1). In 272 contrast, P-WSFM accounted for 41% and 50% of the total features observed in T = 7 and 14, 273 respectively (Tables S1 and S2). The similarity in DOC concentrations in WSF treatment samples 274 across time points, together with the increase in the number of P-WSFTotal and the small fraction 275 of P-WSF0-C, implies that the majority of compounds initially found in WSF were transformed 276 into metabolic intermediates by microbial degradation rather than completely remineralized to 277 CO2, within the time frame of the experiment. 278 Changes in the relative abundances of features in P-WSF0-I and P-WSF0-D were significant 279 between T = 0 and T = 7 but not significant between T = 7 and T = 14, based on paired one-tailed 280 Student's t-tests ( Figure 2 and Table S3). In contrast, increases in relative abundances for P-WSFM 281 features were significant between T= 0 and T = 7, and between T = 7 and T = 14 (Figure 2 and 282 Table S3). These findings suggest that processes that lead to a decrease or increase in abundances 283 of P-WSF0 features proceed at slower rates, or not at all, after T = 7; while production of P-WSFM 284 continues after T = 7 and potentially beyond T = 14. The source of P-WSFM may be P-WSF0 and/or 285 low-molecular weight, water-soluble non-polar compounds, such as PAHs. We can discount the 286 first possibility because the change in relative abundance in P-WSF0-D is minimal between T = 7 287 and T = 14. In contrast, the second possibility could only be explained by microbial oxidation of 288 compounds that were originally outside the LC-FT-ICR MS analytical window, such as the non-289 polar compounds in WSF. 290 This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license.

Degradation of PAHs is a Likely Source of Polar Metabolites in WSF.
One of the possible 291 sources of P-WSFM is the oxidation of non-polar compounds. To investigate whether non-polar 292 compounds in WSF are the sources of polar compounds, we focused our data interpretation on the 293 degradation of water-soluble PAHs, i.e. typically those with less than 3-rings, which should 294 account for ~10% of the WSF. 3 If all or a major fraction of PAHs were completely remineralized 295 into CO2, such a process should result in statistically significant changes in DOC concentrations. 296 However, DOC concentrations did not change significantly while total PAH concentration 297 decreased and the number of P-WSFTotal increased (Figures 1 and S6). These observations support 298 the notion that the source of the increasing P-WSFM was likely the non-polar compounds that were 299 not detected by LC-FT-ICR MS at T = 0. 300 We identified a total of 56 metabolites that occur within known aerobic aromatic compound 301 degradation pathways, such as xylene, toluene, naphthalenes, and PAHs with three or fewer rings 302 (Table S4). Twenty-four of the 56 compounds from P-WSFTotal were assigned a level-2 putative 303 annotation based on exact mass and fragmentation pattern matches (Table S4). Of the 24 Level-2 304 metabolites, 7 were associated with the degradation of naphthalene and methyl-naphthalenes 305 (Table S4), even though naphthalene was not detected in PAH analysis. Multiple naphthalene 306 degradation products were observed across all time points with either negligible abundances at T 307 = 0 and elevated abundances at T=7 and/or T=14 or with the highest abundances at T = 7 (e.g. 308 Figures S7 and S8). We observed multiple metabolites from the degradation of aromatic 309 compounds with 1-3 rings, but no metabolites associated with the degradation of high molecular 310 weight PAHs. This is consistent with the absence of high molecular weight PAHs in WSF based 311 on their low aqueous solubilities. 2 312 This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license.

O3 and O4 Metabolites are Important Intermediates in WSF Transformations. The number 313
of putatively annotated compounds related to WSF degradation accounted for only a small 314 percentage (<12%) of P-WSFTotal in each sample. To better understand the characteristics of the 315 broad range of compounds observed in WSF, existing chemical reaction information available in 316 the KEGG database and metagenomics data from this study was used to guide the interpretation 317 of the remaining P-WSFTotal features. 318 We focused on P-WSFTotal features with assigned formulas (Levels 1, 2, and 3) that contained 319 oxygen atoms, since oxidation is a well-known mechanism for oil degradation. Aerobic 320 biodegradation of PAHs is catalyzed by either monooxygenases or dioxygenases, enzymes that 321 add one or two oxygen atoms, respectively, in the initial steps.   In this experiment, three O3 compounds, salicylic and 3-and 4-methylsalicylic acids, were the 331 most abundant Level-1 metabolic intermediates (Table S4 and Figure S7). These compounds, 332 however, only account for a small fraction of the DOC (<0.1%), suggesting that: (1) a large 333 proportion of the WSF compounds were unknown and (2) these compounds likely represent a pool 334 of WSF organic carbon that has low concentrations but high flux.
To further examine oxygen-containing compounds beyond the level-1 identification, we 336 compared the oxygen distributions of compounds with CxHyOz formulas. Comparisons of relative 337 abundance between each oxygen compound class across time points were based on pair-wise one-338 tailed Student's t-test. Within the oxygen number distributions of P-WSFTotal features, the relative 339 abundances of O4 compounds were significantly higher at T = 7 and 14 than at T = 0 (Figure 4a). 340 Changes in relative abundance of P-WSF0-I features show that most of the production occurred in 341 the O3 and O4 classes (Figure 4b). The relative abundance of O3 P-WSFM also increased 342 significantly at T = 14, compared to T = 7 (Figure 4d). The oxygen number distribution 343 characteristics from a broader range of compounds in P-WSFTotal features are similar to known 344 PAH degradation pathways. Therefore, the high relative abundance of O4 compounds in WSF is 345 likely due to multiple reactions from the degradation of polar and non-polar compounds. 346 The observed oxygen number pattern and complementary genomics data suggest that the 347 degradation of naphthalene and other PAHs in the experiment was likely attributed to dioxygenase-348 catalyzed reactions. We observed enrichments of multiple genes encoding for dioxygenases in the 349 WSF samples ( Figure 5). Although evidence of monooxygenase-catalyzed PAH metabolism is 350 present in some bacteria, 35, 63 such reactions are predominantly observed in eukaryotic cells such 351 as fungi, yeast, and mammalian cells. 64-67 Unlike dioxygenase-encoding genes, we did not observe 352 genes that encode for monooxygenases. The increase in O3 class relative abundance was, 353 therefore, not related to monooxygenases, and is instead more likely to be breakdown products 354 from O4 compounds after the first ring opens. We found that both P-WSF0-I and P-WSFM features 355 include many O3 compounds (Figure 4b and d). The high abundance of O3 compounds may reflect 356 the production and accumulation of intermediates similar to salicylic acids (i.e. salicylic acid, its 357 methylated forms, and other modifications), which are key metabolites from naphthalene and methyl-naphthalene degradation. Interestingly, the abundance of a subset of O3 compounds also 359 decreased substantially over the course of the experiment within P-WSF0-D (Figure 4c), suggesting 360 some of these compounds were rapidly degraded. These findings suggest that the O3 compound 361 class is a dynamic group of compounds with production, accumulation, and degradation occurring 362 simultaneously. 363 can impact ecosystems, 3, 5, 75 e.g. in deep-sea oil spills or seepages, in the water column beneath a 398 surface oil slick ( Figure 6). The molecular-level understanding of the fate of the polar compounds 399 in crude oil WSF, however, is still scarce. This study offers the first multi-omics insights into 400 microbial degradation of WSF, achieved by integrating a broad suite of chemical features with 401 hydrocarbon-degrading genes. 402

O3 Metabolites Such as Salicylic Acids Represent a Key Transition Point for WSF
Although this is a controlled laboratory experiment, our results set important reference frames 403 for future oil spill studies and provide new compound targets for field monitoring. The wide range This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license.  (Table S7), suggesting these compounds may be used as markers for PAH degradation in the field. 413 Additional work with fresh samples or surface oil slicks is needed to confirm these results. This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license. striped bars highlighted features that persisted from T = 0 to T = 14 and whose relative abundance 455 did not change (P-WSF0-U). Red striped bars highlighted features whose relative abundance 456 increased over the course of the experiment (P-WSF0-I). Blue striped bars highlighted features 457 whose relative abundance decreased the course of the experiment (P-WSF0-D). Green bars 458 highlighted new compounds present at T = 7 and T = 14, but not at T = 0 (P-WSFM). The difference 459 between the initial striped bar and the sum of the three striped bars at T=7 and T=14 constitutes 460 the features that were lost (P-WSF0-C). 461 This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license. This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license. identification. Chemical names in green indicate a level-2 putative annotation. Chemical names in 474 blue indicate a level-3 putative characterization. Chemical names in black indicated that we did 475 not observe these compounds. BLAST identities of dioxygenase genes detected against an 476 experimentally validated database 76 are listed in Table S12. 477 This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license. This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license. 27 486 Figure 5. Heatmap of genes that encode for aromatic compound metabolisms identified from un-487 replicated metagenomic data (described in Supporting Information). Only genes encoding for the 488 degradation of oil-derived compounds are included. The darker color indicates higher Z-Score 489 values for specific genes compared to the light yellow. Abundance and percentile of the gene are 490 listed in Table S11. BLAST identities of dioxygenase genes detected against an experimentally 491 validated database 76 are listed in Table S12. 492 This manuscript is a non-peer reviewed EarthArXiv preprint. This is a revised manuscript to be resubmitted to the Journal of Earth and Space Chemistry. The copyright holder for this non-peer reviewed preprint is the author/funder, who has granted EarthArXiv a license to display the preprint in perpetuity. It is made available under a CC-By Attribution 4.0 International license. 28 493 494 Figure 6. Proposed occurrence and fate of the water-soluble fraction (WSF) in the environment. 495 Through dissolution, WSF is expected in the water around the surface oil slicks during an aquatic 496 oil spill. The chemical fingerprint of WSF is distinct from the water-accommodated fraction 497 (WAF), which contained emulsified oil droplets. In the water column, WSF is expected to contain 498 a larger fraction of BTEX, low molecular weight PAHs, and polar compounds. Based on our 499 results, non-polar compounds such as BTEX and PAHs can contribute to polar compounds in WSF 500 through bio-transformation in the water column. At the surface, BTEX may readily volatilize while 501 PAHs are transformed to polar compounds through photo-and bio-mediated pathways. 502