Tracing the Biotransformation of Polycyclic Aromatic Hydrocarbons in Contaminated Soil Using Stable Isotope-Assisted Metabolomics

Abstract

Biotransformation of organic pollutants may result in the formation of oxidation products more toxic than the parent contaminants. However, to trace and identify those products, and the metabolic pathways involved in their formation, is still challenging within complex environmental samples. We applied stable isotope-assisted metabolomics (SIAM) to PAH-contaminated soil collected from a wood treatment facility. Soil samples were separately spiked with uniformly ¹³C-labeled fluoranthene, pyrene, or benzo[a]anthracene at a level below that of the native contaminant, and incubated for 1 or 2 weeks under aerobic biostimulated conditions. Combining high-resolution mass spectrometry and automated SIAM workflows, chemical structures of metabolites and metabolic pathways in the soil were proposed. Ring-cleavage products, including previously unreported intermediates such as C₁₁H₁₀O₆ and C₁₅H₁₂O₅, were detected originating from fluoranthene and benzo[a]anthracene, respectively. Sulfate conjugates of dihydroxy compounds were found as major metabolites of pyrene and benzo[a]anthracene, suggesting the potential role of fungi in their biotransformation in soils. A series of unknown N-containing metabolites were identified from pyrene, but their structural elucidation requires further investigation. Our results suggest that SIAM can be successfully applied to understand the fate of organic pollutants in environmental samples, opening lines of evidence for novel mechanisms of microbial transformation within such complex matrices.

Graphical Abstract

INTRODUCTION

Bioremediation is a well-accepted technology for the treatment of PAH-contaminated soil, but it might not be sufficient to reduce soil genotoxicity due to the formation of oxidation products.^1–3 In a recent study we identified a previously unknown bacterial metabolite of pyrene in contaminated soil after bioremediation,⁴ illustrating that current knowledge on PAH biodegradation pathways gathered from the study of isolated strains on single compounds can be limited. Interactions among organisms in soil microbial communities or between components of complex PAH-mixtures^5,6 can result in cometabolism,⁷ competitive metabolism,^8–10 or cooperation; ^11,12 such phenomena can enhance, inhibit, or modify the extent and/or profiles of metabolite formation.

An effective strategy to trace the biotransformation of specific pollutants within environmental samples is using stable or radioactive isotope-labeled compounds.^13–16 For radioactive isotopes, their special safety requirements can limit wider application. Conversely, ¹³C-labeled substrates could be a suitable tool to shed light on the biodegradation pathways of organic pollutants.^17,18 However, the identification of labeled metabolites has been restricted so far by the resolution of mass spectrometry and by data analysis methods. Recently, isotope labeling combined with high-resolution mass spectrometry (HRMS) has been applied to track the abiotic transformation of pollutants in aqueous mixtures.^19,20 Adapting this combination of techniques to understanding biotransformation of the components of complex environmental mixtures, such as coal or oil derivatives, requires more powerful analysis of the HRMS data.

Stable isotope-assisted metabolomics (SIAM) could provide new insights on the biotransformation of pollutants and unleash the inherent power of isotope labeling. Such methods have already been established in the field of metabolomics, ^21–24 and various automatic workflows are available.^25–28 The basic principle of SIAM is that when ¹³C-labeled substrates are added to mixtures containing natural-abundance (unlabeled) counterparts and are biotransformed into labeled metabolites, they can be detected in groups with unlabeled metabolites by high performance liquid chromatography (HPLC) coupled with HRMS. Algorithms enable such group selection by picking up features with (1) the same chromatographic retention time; (2) certain exact mass differences; and (3) diagnostic intensity ratios.^29,30 The SIAM approach differs in this respect from previous applications of HRMS with isotope-labeled compounds in environmental science cited above. Among other differences, SIAM cannot be used with deuterated compounds because they elute at different retention times than their unlabeled counterparts. Here, we applied SIAM to trace the biotransformation of PAHs in contaminated soil samples. Our objectives were (1) to check the feasibility of this method for organic pollutants in complex mixtures using uniformly ¹³C (U-¹³C) labeled PAHs as a model, and (2) to apply this method to identify the compound-specific metabolite profiles and degradation pathways for three four-ring PAHs in real contaminated soil samples during an active, laboratory-scale bioremediation process.

METHODS

Chemicals.

Anhydrous sodium sulfate and HPLC-grade solvents were purchased from Fisher Scientific (Pittsburgh, PA, U.S.A.). U-¹³C labeled fluoranthene, pyrene, and benzo[a]anthracene (B[a]A) were synthesized according to previous publications.^31,32

Contaminated soil and spiking.

Creosote-contaminated soil was collected from a wood-treatment facility in Andalucía, Spain with a 100-year history of pollution.³³ PAH concentrations are listed in Table S1. For the preliminary method assessment, treated soil from a different site was collected from a lab-scale aerobic bioreactor.⁴

For each U¹³C-labeled PAH (fluoranthene, pyrene, or B[a]A), the desired mass in acetone solution was spiked into 0.4 g (20% of a 2 g sample) of dry contaminated soil. After solvent evaporation overnight, the spiked soil was mixed with the remaining 1.6 g of unspiked soil in successive 0.4 g increments.³⁴ The amount of spiked U-¹³C PAH was adjusted to maintain a 1:3 (mass:mass) proportion with respect to the corresponding native PAH concentration in the soil. Unspiked samples were amended with the same level of acetone as the spiked samples.

Microcosm incubations.

Aliquots of 1 g of soil (dry wt) were placed in sterile 125 mL Erlenmeyer flasks containing 30 mL of 10 mM phosphate buffer (pH 7.5) supplemented with urea (1 mM) as nitrogen source.⁴ All flasks were incubated at 25 ºC under agitation at 150 rpm. Triplicate unspiked flasks were sacrificed initially and after 7, 14, and 21 days of incubation. Triplicate killed controls were prepared by acidifying (pH 2) using phosphoric acid, and sacrificed after 21 days. Duplicate flasks containing U-¹³C fluoranthene, pyrene or B[a]A were sacrificed after 7 or 14 d. Optimal incubation times for each target PAH were determined in a preliminary incubation using unspiked soil as described in SI (Figure S1).

After incubation, the flask contents were centrifuged to recover the solid and aqueous phases. The solid phase was processed as described previously.¹ The aqueous phase was acidified (pH 2) and extracted with ethyl acetate. Further details on sample extraction and processing are provided in SI.

Instrumental analysis.

Extracts were analyzed using an Agilent 1200 series HPLC interfaced with a 6520 accurate mass quadrupole time-of-flight mass spectrometer (qTOF-MS, Agilent Technologies, Santa Clara, CA). Full scan (m/z 100–1000) with electrospray ionization (ESI) was used for SIAM analysis, and product ion scanning (MS/MS) was applied to elucidate the putative structures of metabolites. Details and parameters are provided in SI.

Data analysis.

Raw data from qTOF-MS were converted into. mzData or. mzXML format for in-depth analysis. Peak detection and alignment were achieved using the R package XCMS³⁵ and the publicly available Scripps XCMS Online platform (https://xcmsonline.scripps.edu/).36 Isotope group picking was conducted in R using the package X13CMS.²⁵ Detailed processing methods and parameter settings are provided in SI.

RESULTS AND DISCUSSION

Assessing performance of the method.

Considering the previous identification of a genotoxic metabolite (2H-naphtho[2,1,8-def]chromen-2-one; NCO) during the incubation of a bioreactor-treated soil with pyrene,⁴ we first assayed three different ratios of U-¹³C to natural abundance (native) pyrene in the bioreactor slurry to test the feasibility of the method. Ratios of 1:1 and 1:3 produced ¹³C-NCO in sufficient quantity for its analytical detection, while 1:9 could not. Considering the cost and scarcity of U-¹³C compounds, and to minimize their impact on the microbial community as an enrichment substrate, we chose the 1:3 ratio to perform our experiments on the creosote-contaminated soil.

To detect the natural abundance and ¹³C-enriched pairs of metabolites, we applied the “getIsoDiffReport” function of the X13CMS package to select the peak groups that shared the same retention time but presented m/z values that differed in N × 1.00335, where N is the number of ¹³C atoms (N varied depending on the metabolite from a given PAH).²⁵ However, with this approach most of the selected groups were false positives that only contained ions corresponding to [M-H]- and [M-H+1]-, assigned to the background of natural abundance ¹³C.

To restrict the number of selected groups, we first restricted the ratio of labeled and unlabeled metabolites according to the 1:3 ratio between the spiked ¹³C PAHs and the native compounds, using the “enrichmentLvsU” parameter in the X13CMS package. We then compared spiked and unspiked samples obtained from the same incubation time using XCMS Online. These two samples differed only in the presence of isotope labeling, and the features significantly higher in the spiked samples should correspond to labeled metabolites. By integrating the results from both approaches we generated the final list of prioritized features (Table 1).

Table 1.

Detected metabolites from the four-ring PAHs

RTa	Phaseb	Ionization	m/z>12C	m/z13C	Formula	Confidence levelc	Tentative identification	PubChem Identifier
			Fluoranthene
23.1	W	ESI-	217.0652	233.1192	C16H10O	3	Hydroxy-fluoranthene	3014645
20.1	S	ESI+	233.0567	249.1094	C16H8O2	3	Fluoranthene o-quinone	21481
14.2	W	ESI-	249.0562	265.1068	C16H10O3	4	9-fluorenone-1-(2-propene)-carboxylic acid
14.1	W	ESI-	195.0459	208.0874	C13H8O2	3	Hydroxy-fluorenone	80663
7.3	W	ESI-	237.0398	248.0695	C11H10O6	4	(2-Carboxybenzyl) malonic acid
			Pyrene
20.0	S	ESI+	233.0567	249.1094	C16H8O2	3	Pyrene o-quinone	160814
18.1	W	ESI-	313.0179	329.0716	C16H10O2·SO3	3	Dihydroxy-pyrene sulfate conjugate
20.4	W	ESI-	358.0043	374.0572	C16H9NO4·SO3	4
21.9, 22.3	W	ESI-	369.9945	385.0452	C15H5N3O9	4
21.2	W	ESI-	342.0005	356.0448	C14H5N3O8	4
21.4	W	ESI-	386.9854	401.0332	C14H4O4·2N2O3	4
			Benzo[a]anthracene
19.9	W	ESI-	339.0331	357.0932	C18H12O2·SO3	3	Dihidroxy benzo(a)anthracene
15.9	W	ESI-	289.0178	303.0645	C14H10O2·SO3	3	Dihydroxy-anthracene; dihydroxy-phenanthrene
11.1	W	ESI-	271.0614	286.1104	C15H12O5	4	1-Hydroxy-2-carboxy-anthracene dihydrodiol
2.6, 5.9	W	ESI-	247.0246	259.0640	C12H8O6	3	Dihydroxynaphthalene-dicarboxylic acid (2 isomers)	69414683

^aRT, retention time

^bW, aqueous (water) phase; S, solid phase

^cConfidence levels for metabolite identification were assigned according to the criteria established by Schymanski et al.⁵³

Metabolite identification.

Because we tested only ESI in positive and negative ionization modes, it is possible that some metabolites escaped analysis. Metabolites present in low abundance, with very low ionization potential, or which would be better detected with a different ionization source could have been missed. However, our primary objective was to establish proof-of-concept of SIAM for applications in complex environmental systems, rather than an exhaustive search for all possible metabolites.

For each target PAH, several labeled features were detected. Tentative structures of metabolites (Table 1) were based on their exact mass, MS/MS fragmentation patterns, deduced number of labeled carbons, and further comparison with data available in the literature and pathway prediction systems such as EAWAG-BBD³⁷ and enviPath³⁸. Standards were not available for any of the proposed metabolites. With a formula and MS/MS pattern, each structure was processed with the MetFrag³⁹ platform and searched in the PubChem database (https://pubchem.ncbi.nlm.nih.gov/), and five of them matched the record (Table 1).

In the solid phase, labeled metabolites were detected only for fluoranthene and pyrene. The identified products with the formula C₁₆H₈O₂ corresponded to the respective o-quinones (see Figure S2 for the fluoranthene quinone), which were previously found to accumulate during incubation with pure bacterial strains ^40,41. The low number of products detected in the solid phase suggested that most of the metabolites were formed in or transferred to the aqueous phase due to their increased polarity relative to the parent compounds.

Two major groups of compounds were detected in the liquid phase: highly oxygenated ring-cleavage products and PAH conjugates. Within the first group, we detected several formulas originating from fluoranthene (C₁₁H₁₀O₆; Figure 1) and B[a]A (C₁₅H₁₂O₅, Figure S3; and C₁₂H₈O₆) corresponding to ring-cleavage products containing one or more carboxyl group (mass spectra consistent with loss of at least one fragment of m/z 44). The fluoranthene product C₁₁H₁₀O₆ did not correspond to any previously identified metabolite, and is proposed as (2-carboxybenzyl) malonic acid (Figure 1a and Figure S6d). The B[a]A metabolite C₁₅H₁₂O₅ has not been previously described, and is proposed as a dihydrodiol of the 1-hydroxy-2-anthracene carboxylic acid. For the formula C₁₂H₈O₆, the MS/MS spectral characteristics (Figure S6e) are identical to those of 1,6-dihydroxynaphthalene-2,7-dicarboxylic acid, a known ring-fission product of B[a]A from Sphingobium sp. strain KK22.⁴²

Figure 1.

Mass spectra, extracted ion chromatograms (EIC), and proposed structures of two representative metabolites. a: C₁₁H₁₀O₆ from fluoranthene; b: C₁₆H₁₀O₂·SO₃ from pyrene. The green lines represent ¹²C metabolites, and the red lines represent U-¹³C labeled metabolites.

A second group of products were conjugates of PAH metabolites, mainly sulfate conjugates of dihydroxylated PAHs. We detected several peaks corresponding to the sulfate conjugates of dihydroxy-pyrene (Figure 1b), dihydroxy-B[a]A, and either dihydroxy-anthracene or -phenanthrene derived from B[a]A (Figure S3). Each of these dihydroxy-compounds was always associated with a co-eluting peak with m/z corresponding to [M-H+80]^- (Figure S5). According to the exact mass (m/z 79.9570), this fragment was assigned to an SO₃ group, indicative of a sulfate conjugate.⁴³ To confirm this observation, we repeated the analysis decreasing the fragmentor voltage to 100V, which increased the sulfate conjugate signal while that of the dihydroxy compounds disappeared, confirming the sulfate conjugates as the actual metabolites. Such sulfate conjugation of dihydroxy-PAHs is common in fungi,^44,45 but has never been observed in bacteria. These results suggest that fungi might have played a significant role in PAH biotransformation and detoxification in the contaminated soil.

Besides the sulfate conjugates, we observed a number of other candidate conjugation products in the incubations containing U-¹³C pyrene (m/z 342.0005, 358.0003, 369.9945, and 386.9854, Figure S4). Although their formulas could not be unambiguously assigned, all of these features had an m/z over 300, and some of the deprotonated ions presented nominal masses in even numbers, suggesting the existence of nitrogen atoms. Their MS/MS spectra revealed fragments with exact masses compatible with the loss of different nitrogen oxides (NO or NO₂) but especially of m/z 75.99, corresponding to N₂O₃ (Figure S6). The formulas of the conjugates were therefore assigned according to their fragmentation patterns. PAH metabolites with such fragments, as well as conjugation reactions with nitrogen oxides, have not previously been described. Based on current evidence, it is difficult to elucidate the identities of these unknown pyrene metabolites. Considering their novelty and diversity, these nitrogen-containing metabolites should be interesting targets for future research.

Proposed compound specific degradation pathways in soil.

Based on the detected ¹³C-labeled metabolites, pathways for B[a]A and fluoranthene degradation in the PAH-contaminated soil are proposed (Figure 2 and Figure S7, respectively). It is important to note that these pathways do not necessarily result from metabolism of the parent compound by an individual organism, and therefore the specific metabolites and their accumulation over time are less meaningful than for incubation of a single substrate with a pure culture. B[a]A biodegradation mainly followed the preferential degradation route previously described for Sphingobium yanoikuyae B1, (formerly Beijerinckia),^46,47 which is initiated by dioxygenation of the molecule in positions 1,2. Further meta-cleavage and pyruvate release would lead to the formation of 1-hydroxy-2-anthracenoic acid, an intermediate also observed in other Sphingobium strains,⁴² followed by oxidation to a dihydrodiol (C₁₅H₁₂O₅) that is reported here for the first time. The identification of two isomers of dihydroxynaphthalene dicarboxylic acid (Table 1) indicated the dehydrogenation and ring cleavage of the latter.⁴² Several fluoranthene metabolites are not consistent with established pathways, but the identification of hydroxy-fluorenone suggests the processing of one of the two fused aromatic rings.^48,49 We observed evidence for both initial mono- (C₁₆H₁₀O) and dioxygenation (C₁₆H₈O₂) of the fluoranthene molecule (Figure S2), but none of the previously described ring-cleavage products was detected. The MS/MS analysis of the formula C₁₆H₁₀O₃, corresponding to a metabolite not previously identified, presented a fragment of m/z 44 indicative of a carboxyl group (Figure S6), suggesting its formation as a product from the initial ring fission. Because this metabolite contained 16 C atoms and only three oxygen atoms, instead of the expected four for a common meta- or ortho-cleavage product, it could not be assigned to any previously described ring fission metabolites.^48,50 Finally, the product with formula C₁₁H₁₀O₆, identified here for the first time, could be consistent with a product from the cleavage and further degradation of the biphenylene derivative originating from angular attack on fluorenone.⁴⁸

Figure 2.

Proposed degradation pathway for benzo[a]anthracene based on detected metabolites and published literature.⁴² Note that substituent positions are either inferred from known mass spectra or are hypothesized. Dashed arrows indicate more than one successive reaction. Structures in brackets are putative and were not observed directly. Reactions follow the canonical reactions in aerobic fungal and bacterial metabolism of PAHs summarized in Cerniglia⁵¹ and Kweon et al.,⁵² respectively. The levels indicated next to structures are the level of confidence as defined in Schymanski et al.⁵³ Level 3: tentative candidates (compounds identified by molecular formula and substructures), based on accurate mass, MS/MS pattern, and database search. Level 4: unequivocal molecular formula based on accurate mass, but no other supporting data.

Time-course evolution of the identified metabolites.

We traced the detected metabolites through the course of the incubations according to their m/z and retention time (Figure S1). Due to the lack of standards, metabolites were quantified relative to their peak areas. Consistent with PAH removal (Figure S1), fluoranthene metabolites peaked at 7 d (except C₁₆H₈O₂, which peaked at 14 d), then decreased in abundance, clearly indicative of their subsequent metabolism. B[a]A metabolites, including two sulfate conjugates, appeared at 14 d and increased in abundance after 21 d incubation. Sulfate conjugates from pyrene also accumulated over the course of incubation, suggesting the potential persistence of such conjugates even in biologically active soil. The recalcitrance of analogous PAH conjugates accumulated by the fungus Cunninghamella elegans to soil microbial communities was previously demonstrated.⁴⁵

Environmental implications and further applications.

In this study, we demonstrated that SIAM can be successfully applied to the biotransformation of pollutants. This method allowed us to trace the biotransformation of individual pollutants in the context of a real contaminated sample. SIAM in combination with MS/MS enabled the detection of novel metabolites and degradation pathways that could result from interactions within complex microbial communities and contaminant mixtures. Despite the analytical limitations associated with the use of a single ionization source, our method allowed us to highlight major mechanisms implicated in PAH biodegradation during biostimulation of a contaminated soil. The SIAM workflow could be applied to a wider range of studies in environmental sciences, such as the biotransformation of micropollutants in aquatic systems and in water treatment processes. In addition, the combination of SIAM with compound specific isotope analysis (CSIA) may become a powerful tool for field-based bioremediation studies.