Identification of Potentially Toxic Transformation Products Produced in Polycyclic Aromatic Hydrocarbon Bioremediation Using Suspect and Non-Target Screening Approaches

Abstract

Polycyclic aromatic hydrocarbons (PAHs) are a class of ubiquitous environmental contaminants that can be remediated through physical, chemical, or biological means. Treatment strategies can lead to the formation of PAH-transformation products (PAH-TPs) that, despite having the potential for adverse ecological and human health effects, are unregulated and understudied in environmental monitoring and remediation. Unavailability of reference standards for PAH-TPs limits the ability to identify PAH-TPs by targeted methods. This study utilized suspect and non-target screening approaches to identify PAH-TPs produced by a bacterial culture, Rhodococcus rhodochrous ATCC 21198, using liquid chromatography-high resolution mass spectrometry. Open-source tools were used to predict biotransformation products, predict potential PAH-TP structures from mass spectra, and estimate health hazards of potential PAH-TPs. The workflow developed in this study allowed for the tentative identification of 16 PAH-TPs (confidence levels 2a to 3), seven of which were not previously detected by targeted analysis. Several new potential transformation pathways for our bacterial pure culture were suggested by the PAH-TPs, including carboxylation, sulfonation and up to three hydroxylation reactions. A computational toxicity assessment indicated that the PAH-TPs shared many hazard characteristics with their parent compounds, including genotoxicity and endocrine disruption, highlighting the importance of considering PAH-TPs in future PAH studies.

Graphical Abstract

1.0. Introduction

Polycyclic aromatic hydrocarbons (PAHs) are a class of persistent environmental pollutants associated with many adverse health effects, including endocrine disruption, genotoxicity, and carcinogenicity.^1,2 PAHs are ubiquitous in the environment due to their many natural and anthropogenic sources, such as forest fires and the burning of fossil fuels.^3,4 Sixteen PAHs are recognized as priority pollutants by the United States Environmental Protection Agency (US EPA), therefore, areas impacted by significant PAH contamination must undergo remediation.^5,6 A variety of remediation strategies have been applied to PAH contaminated media, generally falling into one of three treatment categories: physical, chemical, or biological treatment.^7,8 Physical treatment involves separating and removing PAHs from the environmental matrix (soil, water, etc.) by physical processes such as adsorption or soil washing.^9-11 Chemical treatment relies on breaking down PAHs via chemical reactions with the use of reagents such as Fenton’s reagent or ozonation.^8,12 Finally, biological treatment or bioremediation of PAHs uses living organisms such as bacteria or fungi, to metabolize or cometabolize PAHs in situ.^13-15 Bioremediation of PAHs often occurs without any engineering intervention, referred to as natural attenuation.^16,17 In breaking down PAHs by chemical or biological processes, PAH transformation products (PAH-TPs) are produced, either as intermediate or dead-end products.¹⁸

PAH-TPs have gained interest in recent years due to their prevalence in the environment and their toxicity.^1,19 However, PAH-TPs are still a largely unregulated class of environmental contaminants.²⁰ PAH-TPs are typically more polar than parent PAHs and are thus more bioavailable and mobile in the environment.^6,21,22 In addition, several studies have demonstrated that PAH-TPs can be equally or even more toxic than their parent compounds.^6,23-25 Despite their toxicity and presence in natural and engineered systems, PAH-TPs remain an understudied and vastly unregulated class of environmental pollutants. Studies that have investigated the production and toxicity of PAH-TPs often report hinderances related to the detection and identification of PAH-TPs, including a lack of PAH-TP libraries and availability of reference standards to confirm compound structures analytically.^26-28

Alternatives to traditional, targeted methods that rely on standard matches are non-targeted analysis (NTA).^29,30 Environmental monitoring approaches generally fall under three categories: target analysis, suspect screening, and non-target analysis, described in detail in Schymanski et al. (2015).³¹ While targeted analysis focuses on measuring specific, well-known compounds with established reference standards, non-target analysis measures all detectable compounds in a sample. Suspect screening serves as a middle ground between targeted and non-targeted approaches. In this approach, we screen for a list of “suspect” compounds based on a priori knowledge, such as chemical databases or related compounds that are expected to be present. Unlike targeted analysis, suspect screening doesn’t require pure reference standards for each compound, but it does rely on computational methods and mass spectrometry libraries to tentatively identify likely compounds, or features, in the sample. Suspect screening is especially useful for identifying potential transformation products of known contaminants or emerging pollutants.

Identification of detected features are assigned confidence levels on the widely accepted Schymanski scale of 1 to 5, based on the availability and quality of information considered.³² Several studies have investigated the prevalence of PAH-TPs in environmental samples using NTA, and have revealed PAH-TPs as a ubiquitous yet largely unmonitored class of environmental contaminants in both natural and engineered systems. Schemeth et al. (2019) utilized suspect and non-target screening approaches to characterize hydroxylated-, carbonylated-, and carboxylated-PAHs present in environmental water samples from two different sites and those produced in a engineered system simulating photooxidation.³³ Similarly, Luo and Schrader (2021) identified a wide variety of PAH heterocycles and highly aromatized and alkylated PAHs with non-target approaches, highlighting the vast chemical space that targeted methods measuring the 16 priority PAHs ignore.³⁴ Titaley et al. (2021) paired non-target analysis with an in vitro bioassay to identify PAHs and their derivatives with the potential to bind to the aryl hydrocarbon receptor (AhR), ultimately identifying 12 PAH compounds that should be prioritized in future monitoring studies.²⁶

A previous study investigated the biotransformation of a mixture of four parent PAHs: fluorene, anthracene, phenanthrene, and pyrene by the bacterial pure culture Rhodococcus rhodochrous ATCC 21198.³⁵ The four PAHs studies are among the 16 priority pollutant PAHs designated by the US EPA and each represent a different structural class of PAHs (nonalternant, linear, bent, and cluster for fluorene, anthracene, phenanthrene, and pyrene, respectively). After 5 days of incubation, 21198 transformed approximately 91% of fluorene, 50% of phenanthrene, and 96% of anthracene, however no pyrene transformation was observed.³⁵ Limited target analysis revealed that 21198 transformed PAHs primarily into oxygenated transformation products. Incorporation of oxygen is known to be a primary initial reaction for xenobiotic cometabolism with 21198, and is often carried out by a non-specific short-chain alkane monooxygenase (SCAM).^36-39

Although several potential PAH transformation products were identified in the previous study, the number of compounds identified by target screening were significantly outnumbered by chemical features detected that did not match a reference standard. This finding, combined with the enhanced toxicity of the PAH-TP mixture produced by 21198, motivated further investigation of potentially toxic transformation products using NTA and suspect screening. The objective of the present study was to use NTA and suspect screening to fill the gaps in our understanding of the chemical makeup of the PAH-TP mixture produced during bioremediation to better understand the biotransformations occurring, and identify potentially toxic PAH-TPs that contributed to the increased toxicity post-remediation observed previously. In doing so, a workflow was established for processing the high-resolution mass spectrometry data using a variety of data processing software and predictive tools to not only tentatively identify novel PAH-TPs, but also predict their toxicity.

2.0. Materials and Methods

2.1. Chemicals

The parent PAHs used in this study were purchased from AccuStandard (New Haven, CT): phenanthrene (100%), anthracene (99.6%), pyrene (99.1%), and Sigma Aldrich (St. Louis, MO): fluorene (98%). Microbial growth substrates, isobutane (99.99%) and 1-butanol (99.4%), were purchased from Airgas Inc. (Randor, PA) and Acros Organics (MA), respectively. Organic solvents used were purchased from various vendors and were of analytical grade purity.

2.2. Cell Culturing

Rhodococcus rhodochrous ATCC 21198 was maintained on minimal agar plates in a sterile airtight container with isobutane as the sole carbon source provided at 1.5% of the headspace volume. Cells were growth in batch reactors as described in Rolston et al. (2019).⁴⁰ Briefly, 21198 was grown in mineral salt media (MSM)⁴¹ with isobutane provided at 10% of the headspace volume as the sole carbon source. Cells in the late exponential phase were harvested with centrifugation, and biomass was measured using a total suspended solids analysis as described previously.²⁵ Cells were stored at 4°C used in batch experiments within 2 days of harvesting.

2.3. PAH Bioremediation Experimental Design

PAH bioremediation experiments were prepared in 125 mL glass media bottles with butyl septa caps. All batch bottles contained 100 mL 10 times diluted MSM and were amended with 1 mg/L phenanthrene and fluorene, 0.5 mg/L anthracene, and 0.1 mg/L pyrene from concentrated dimethyl sulfoxide (DMSO) stock solutions. The initial PAH concentrations reflect the aqueous solubility limits of the PAHs. PAHs were added to their solubility limit to maximize the mass of PAH available for transformation by 21198 while ensuring the entirety of the PAH mass added was dissolved in the media. 40 mg/L of 1-butanol was also added as an energy source, as it has been shown previously to support aromatic hydrocarbon transformation by 21198.³⁹ Active batch bottles were prepared in triplicate and contained 30 mg isobutane grown 21198 cells. Autoclaved cell control and abiotic control batch bottles were prepared in duplicate and contained 30 mg autoclaved isobutane grown 21198 cells or no cells, respectively. Batch bottles were incubated at 30°C in the dark for 5 days and were removed daily to refresh oxygen. An additional dose of 1-butanol was added after 3 days of incubation.

After the 5-day incubation period, the liquid volume of the batch bottles replicates was pooled into 3 samples: active (A), autoclaved cell control (C), and abiotic control (X). Samples were stored at −20°C until extraction. The solid phase extraction procedure used has been described previously.²⁵ Briefly, Bond Elut Plexa (500 mg, 6mL) cartridges (Agilent Technologies, New Castle, DE) were preconditioned with dichloromethane and methanol, loaded with the liquid samples, eluted with a mixture of acetone and dichloromethane, concentrated under a nitrogen stream, and reconstituted in 250 μL 9:1 methanol: isopropyl alcohol. An extraction blank was prepared by extracting 300 mL of MSM alongside the A, C, and X samples.

2.5. Chemical Analysis

Extracts were analyzed using ultra-high performance liquid chromatography (UPLC) with the Sciex Exion LC AD equipped with an ACE Excel column (1.7 um; 2.1 x100mm, Hichrom Limited, UK). All samples were run in a single batch with a blank injected in between each sample injection. Three replicate injections (technical replicates) were injected per sample. Sample injection volume was 2 μL. Eluents were nanopure water with 0.1% formic acid (eluent A) and acetonitrile with 0.1% formic acid (eluent B) flowing at 0.3 mL/min with the following gradient: 5% B for 3 minutes, 9-minute gradient to 98% B, hold for 5 minutes, 0.5-minute gradient to 5% B, hold for 5.5 minutes. UPLC was coupled with high resolution mass spectrometry (HRMS) using the Sciex ZenoTOF 7600. Ionization was achieved with electrospray ionization (ESI) in both positive and negative mode. Full scan MS and TOFMSMS scans were collected in data dependent acquisition (DDA) mode over a m/z range of 40 to 1000 Da.

2.6. PAH-TP Suspect List Curation

A suspect list of potential PAH-TPs was prepared from three sources: theoretical transformation products based on previous published work, transformation products predicted with the EAWAG biocatalysis/biodegradation pathway prediction software (EAWAG-BBD-PPS), and an in-house library of potential PAH-TP standards run on our UPLC-HRMS system. For all sources, transformation products of the four parent PAHs were considered, along with transformation products for naphthalene. Naphthalene TPs were considered, despite naphthalene not being added to batch reactors initially, because its structure is present as a substructure in anthracene, phenanthrene, and pyrene, therefore transformations of these parent PAHs may lead to the formation of naphthalene TPs. MS1 data was compiled for PAH-TPs predicted from theoretical transformations based on previous work and the EAWAG-BDD-PPS, whereas MS1 and MS2 data was compiled for the standard reference materials run on our instrument.

2.6.1. Theoretical PAH-TPs Produced by 21198

The collection of theoretical transformation products was based on the work presented in Huizenga et al. (2024).³⁵ It was found that 21198 produces several classes of oxygenated PAH-TPs, namely hydroxylated PAH-TPs, quinones, and ring fission products. The formation of these compounds can be explained by transformations catalyzed by enzymes annotated in 21198’s genome, such as SCAM, epoxide hydrolase, and ring cleaving dioxygenase.⁴² Due to the broad substrate specificity of these enzymes, it is possible that one PAH compound can undergo several transformation events. This hypothesis is supported by the identification of hydroxylated PAH-quinones in the previous study, which could only be formed by multiple transformation steps. Therefore, in preparing the theoretical TP suspect list, three transformation events were considered. The first transformation event, acting on a parent PAH, could be hydroxylation, dihydro-dihydroxylation, quinone formation, or ring fission. The second transformation event, acting on a primary transformation product could be either hydroxylation, dihydro-dihydroxylation or quinone formation. Only one ring fission event was considered due to the significant change in structure that arises from a ring fission event. The third and final transformation event was only considered to be hydroxylation, acting on secondary TPs. This was chosen due to the steric hinderances of the functional groups added during the first two transformation events, that would lead dihydro-dihydroxylation or quinone formation events to be more unlikely than a single hydroxylation event. These theoretical transformations generated 28 structures (22 unique structures) per parent PAH, 140 structures total.

2.6.2. Predicted PAH-TPs from EAWAG-BBD-PPS Tool

The EAWAG-BBD-PPS tool is an open source, rule-based system that predicts microbially mediated reactions using a combination of substructure searching and atom to atom mapping ⁴³. The Simplified Molecular Input Line Entry System (SMILES) for the 5 parent PAHs considered were input into the tool, and the SMILES from all very likely, likely, or neutral predicted structures were compiled into a list. SMILES were turned into chemical formulae and ultimately exact mass using a MATLAB script (Martin, 2024). In total, 97 structures were included in the suspect list from this source, including 13 for fluorene, 26 for phenanthrene, 22 for anthracene, 15 for pyrene, and 21 for naphthalene.

2.6.3. In-House PAH-TP Library

An in-house PAH-TP library prepared for the previous study was also included in the suspect list. These standards of potential PAH-TPs included 16 each for fluorene, anthracene, and naphthalene, 18 for phenanthrene, 4 for pyrene, and one potentially formed from all (benzoic acid). A full list of the reference standards, exact masses, and retention times (RTs) are provided in Table S1.

2.7. Data Processing Procedure

A combination of data processing software and prediction tools were used to interpret the raw data into proposed PAH-TP structures. A schematic of the workflow is shown in Figure 1.

Figure 1.

Workflow diagram of the non-targeted data analysis process from raw data acquired with the UPLC-HRMS to proposed PAH-TP structures. Note that all tools are open source with the exception of Sciex OS.

2.7.1. Suspect Screening Procedure

The first step in data processing, carried out in Sciex OS, was to trim the preliminary suspect list down to suspects that had a high potential to be in the active extract based on their exact mass, RT, isotope ratio, chromatography, and intensity. This streamlined the tentative identification of suspects in further processing steps by removing poor-quality suspect matches early on. Acceptable error ranges for mass error and percent difference in isotope ratio were 5 ppm and 20 percent, respectively. Annotations that passed the acceptable error and isotope ratio filtering were then manually checked for acceptable chromatography and height intensity above 1000 a.u.. Annotations that passed this second step were then exported with their exact mass and expected RT to form the trimmed suspect list.

This trimmed suspect list was then used in the second step of data processing using MS-Dial (v4.9.221218), an open-source software for identification of small molecules ⁴⁴. Raw data files from the instrument were converted using Reifycs ABF converter. The trimmed suspect list, along with an open-source MS-Dial metabolomics spectral library were used to annotate features in the raw data from the UPLC-HRMS. Parameters for the MS-Dial alignment are provided in Table S2. After processing, filtering steps were performed to reduce the list of all features to the most relevant features. The filtering steps included the following:

1. Removing features that had an average abundance less than three times the average abundance in the method blanks, extraction blanks, C controls, and X controls

2. Removing features that were detected in fewer than 2/3 of the active extract technical replicates

3. Removing features without MS2 data

4. Removing features annotated as natural products

5. Removing features with a mass error > 5 ppm

The first filtering step was used to isolate features that were distinct to the active extract, as all PAH-TPs would be exclusively in the active extract. The second and third filtering steps ensured that the features were reproducibly detected, and that there was enough data to assign at least a level 3 confidence to PAH-TPs. The final two filtering steps removed natural products that were likely from cell material and features that exceeded an acceptable mass error. These filtering steps were intended to favor PAH-TPs, as these would only be present in the active samples. After filtering, annotations were manually assessed for retention time match (measured RT within 2 minutes of expected RT), and MS1/MS2 match to library match where applicable.

2.7.2. Non-Target Screening Procedure

Features that had MS1 and MS2 data but did not match a library MS2 spectra and/or RT in MS-Dial were considered unknowns. In order to tentatively identify these unknowns, several additional tools were used, as outlined in Figure 1. To start, MS1 isotope spectra were input into the SIRIUS software platform (v5.8.6) to determine a chemical formula.⁴⁵ Chemical formulas were computed in SIRIUS referencing all available databases. All possible chemical formula matches were exported and filtered based on included elements.

Chemical formulas containing halogens and boron were removed, leaving chemical formulas containing C, H, O, P, S, and N. These elements were considered based on their prevalence in phase I and phase II PAH metabolites. Phase I metabolites, involving a phase I enzyme such as a monooxygenase, would incorporate oxygen into the PAH. Chemical formulas that contained a carbon number between 13 and 16, and oxygen were considered phase I PAH metabolites. Phase II enzymes ligate a polar group to a phase I metabolite to increase its solubility and promote excretion. Several such groups were considered in the data analysis based on their prevalence in the literature for bacteria, listed in Table S3.⁴⁶ Ratios of N:S, along with N:P and P:S for coenzyme A (CoA), were used to determine potential matches for phase II PAH metabolites. Chemical formulas with the proper ratios were then investigated further for their carbon number. If the carbon number of the chemical formula minus the carbon number of the matched phase II group was between 13 and 16, this chemical formula was considered further as a phase II PAH metabolite. As with the features that were annotated in MS-Dial, only chemical formulae with a mass error less than 5 ppm were considered to match the detected features.

To generate potential structures for the chemical formulas of the unknowns, chemical formulas and MS2 data were input into MetFrag to generate potential structures and annotate the MS2 peaks. MetFrag is an in silico fragmentation prediction tool that references compound databases and uses bond dissociation energies to predict the fragmentation of candidate compounds and compare this to the user input fragments.⁴⁷ Up to 6 MS2 peaks above 100 abundance units were included in this analysis. Potential structures that contained a parent PAH substructure were prioritized for further investigation. SMILES of potential structures from MetFrag were input into CFM-ID along with the MS2 data to support MetFrag’s annotation of the MS2 spectra based on the proposed structure. CFM-ID uses single energy competitive fragmentation modeling to predict spectra based on chemical SMILES, assign substructures to MS2 fragments, and identify compounds referencing databases.⁴⁸ Spectra prediction and peak assignment features were primarily used to corroborate or supplement MetFrag outputs. The number of peaks annotated and in agreement between MetFrag and CFM-ID were considered in the confidence assessment.

2.7.3. Confidence Level Assessment

Confidence levels were assigned to the tentatively identified compounds based on the Schymanski scale. However, the Schymanski scale leaves room for defining confidence levels in the context of each specific study, particularly for lower confidence levels such as Level 2b and Level 3. Table SXThe final assessment of the confidence in structure assignment to the unknown compounds evaluated the likelihood of the structure formation based on 21198’s annotated enzymes, 21198’s known PAH transformation pathways, thermodynamic favorability, and reports of similar compounds in literature. Structures that would result from feasible enzymatic transformations, structures that had upstream-TPs identified, and transformations that were thermodynamically favorable were assigned higher confidence. Several literature reviews of microbial degradation of PAHs were referenced to identify potential TPs that have been reported previously.^13,14,49,50

2.8. Hazard Assessment

An assessment of the toxicity of potential PAH-TPs was conducted in the US EPA Hazard Cheminformatics Module (HCM). The HCM compiles hazard information from a wide variety of sources, including public databases, literature, and generated quantitative structure-activity relationships (QSAR) using the US EPA Toxicity Estimation Software Tool (T.E.S.T.). Reported endpoints span human health effects, ecotoxicity, and environmental fate of the chemicals. Toxicity is reported qualitatively as inconclusive (I), low (L), medium (M), high (H), and very high (VH). The distinction of the toxicity information source is reported as well, either as authoritative, screening, or QSAR. This hazard module has successfully been applied as a screening tool for prioritizing compounds tentatively identified with NTA in other studies.^51,52 The HCM accepts a wide variety of chemical identifiers, such as SMILES, CAS number, and International Chemical Identifiers (InChI). SMILES were used as inputs for the PAH-TPs of interest in this study because they are available for all structures and are provided in Table S6. All PAH-TPs were assessed together as a batch and the full hazard assessment report was exported to Excel after it was generated.

3.0. Results and Discussion

3.1. Suspect and Non-Target Screening Features

After processing the raw data in MS-Dial with the open-source library and the trimmed suspect list, containing 92 of the original 285 suspect compounds, 4507 features were detected. The number of features was reduced by more than tenfold by eliminating those with abundances less than three times the levels observed in the blanks and controls. The final filtering step further reduced the number of remaining features by a factor of 10, removing those that failed to meet the chemical formula requirements outlined in section 2.7.

Only 19 features remained after all processing steps. Of these 19 features, 13 were tentatively identified using the suspect screening approach described in section 2.7.1. The remaining six were further investigated with the non-target screening approach described in section 2.7.2, after which three features were tentatively identified, and three features remained unknowns. A summary of the PAH-TPs tentatively identified is provided in Table 1. Note that even for compounds that were confirmed with a reference standard, the confidence level is reported at Level 2a instead of Level 1 on the Schymanski scale due to inability to confirm the specific isomer detected. Using an orthogonal technique, such as nuclear magnetic resonance (NMR) or ion mobility may help to distinguish between isomers.^32,53 RT was not able to sufficiently distinguish isomers due to co-elution. RT, mass error, isotope spectra, and MS2 spectra for all compounds listed in Table 1 are provided in Table S4.

Table 1.

PAH-TPs tentatively identified with the NTA workflow. Literature matches indicate reports of the exact compound or structurally similar compound. All matches of a structurally similar compound are specified in more detail in the discussion. SL = Suspect List.

PAH-TP Label	PAH-TP Compound Name	Chemical Formula	Annotation Source	Confidence Level	Literature Matches
F1	2-Hydroxyfluorene	C13H10O	SL: Reference Standard	Level 2a	54-57
F2	9-Fluorenone	C13H8O	SL: Reference Standard	Level 2a	56,58-60
F3	1-Carboxyfluorene	C14H10O2	SL: Reference Standard	Level 2a	61,62
F4	?,?-Dihydroxyfluorene	C13H10O2	SL: Predicted Theoretical TP	Level 2b	55,63
F5	?,?,?-Trihydroxyfluorene	C13H10O3	SL: Predicted Theoretical TP	Level 2b	N/A
F6	?-Carboxy-?-Hydroxyfluorene	C14H10O3	Non-Targeted Tools	Level 2b	62,64,65
F7	?-hydroxy-?H-fluoren-?-yl hydrogen sulfate	C13H10O5S	Non-Targeted Tools	Level 3	N/A
F8	(?S,?S)-?,?-dihydroxy-?,?-dihydro-?H-fluoren-?-yl hydrogen sulfate	C13H12O6S	Non-Targeted Tools	Level 3	N/A
P1	1-Hydroxyphenanthrene	C14H10O	SL: Reference Standard	Level 2a	25,66,67
P2	Diphenic acid	C14H10O4	SL: Reference Standard	Level 2a	61,68-71
P3	3,4-Benzocoumarin	C13H8O2	SL: Reference Standard	Level 2a	72-74
P4	9-Hydroxyphenanthrene	C14H10O	SL: Reference Standard/Open-Source Library	Level 2a	25,66,75,76
AP1	?,?'-((?E,?Z)-naphthalene-?,-diylidene) diacetic acid	C14H10O4	SL: Predicted Theoretical TP	Level 2b	72,77-79
A1	1-Hydroxyanthraquinone	C14H8O3	SL: Reference Standard/Open-Source Library	Level 2a	26
A2	2-Hydroxyanthraquinone	C14H8O3	SL: Reference Standard/Open-Source Library	Level 2a	80
A3	1,2-Dihydroxyanthraquinone	C14H8O4	SL: Reference Standard	Level 2a	81

Compound names that are not bolded were compounds that were detected in the previous study. Bolded compound names were tentatively identified for the first time in the present study.

Of the 16 PAH-TPs listed in Table 1, nine compounds (indicated in non-bolded text) were detected in the previous study.³⁵ These compounds consisted of three mono-hydroxylated parent PAHs, one PAH-quinone, three hydroxylated quinones, and two ring fission products. These classes of PAH-TPs are among the most commonly reported in the literature, as their formation relies on oxygenases that incorporate oxygen into the aromatic ring structures and ultimately cleave ring structures.^13,50 Their formation by 21198 is described in detail in Huizenga et al. (2024).³⁵ There were several compounds detected in the previous study that were not identified using the non-targeted workflow, including trans-9,10-dihydrophenanthrenediol, dihydroxy-naphthalene, hydroxy-fluorenone, and phenalene-1-one. This is likely due to the decrease in sensitivity of the non-targeted method using DDA, compared to the targeted method using multiple reaction monitoring (MRM) mode.⁸² The seven remaining PAH-TPs were new discoveries in this complex mixture produced by 21198 and are discussed further in the following sections.

3.2. Newly identified PAH-TPs

3.2.1. Hydroxylated fluorenes

The majority of new tentatively identified PAH-TPs were from the parent PAH fluorene. This is likely because the total mass of fluorene transformed was the highest among the mixture of four parent PAHs (approximately 0.27 mg fluorene transformed compared to 0.15 mg phenanthrene, 0.14 mg anthracene, and 0 mg pyrene transformed). Hydroxylation is known to be a primary initial reaction for xenobiotic cometabolism with 21198, which is often carried out by SCAM.^36-39 Because of the broad substrate specificity of SCAM, multiple hydroxylation steps, acting on different positions of a substrate, are possible. This supports the tentative identification of di-hydroxylated and tri-hydroxylated fluorene TPs with level 2b confidence. This confidence level was based on multiple lines of evidence. The presence of a mono-hydroxylated fluorene, the precursor to di- and tri-hydroxylated fluorenes, was confirmed with a reference standard. Furthermore, the order of elution of the tentatively identified mono-, di-, and tri-hydroxyfluorenes matches the elution order estimated from Sciex (di-hydroxyfluorene first, followed by tri-hydroxyfluorene and mono-hydroxyfluorene). Hydroxylation of the parent compound or hydroxylated TP is also thermodynamically favorable, as shown in Figure 3. Di-hydroxylated fluorenes are rarely reported in the literature; however, their production has been documented in previous studies. Finkelstein et al. (2003) reported the production of 2,7-dihydroxyfluorene by two different Rhodococcus strains, also indicating that this transformation product was first formed from 2-hydroxyfluorene.⁵⁵ Similarly, Meyer and Steinhart (2000) reported the production of dihydroxy-fluorenone by a microbial consortia, and proposed that dihydroxyfluorene was an undetected intermediate.⁶³ Reports of the biological formation of trihydroxyfluorene transformation products were not found in the literature.

Figure 3.

Potential transformation pathways for the production of the tentatively identified PAH-TPs produced by 21198 displaying thermodynamic favorability and abundance of each compound.

3.2.2. Carboxylated Fluorenes

A new category of 21198’s potential PAH transformation pathways, carboxylation, was suggested by the detection of two carboxylated fluorene TPs. The identification of 1-carboxyfluorene was confirmed with a reference standard, while a carboxy-hydroxy-fluorene was tentatively identified with MetFrag as a proposed structure. The output from MetFrag also annotated the most abundant fragment (after M-H) in the MS2 spectrum of this feature as M-COOH, supporting the presence of a carboxy group on the structure. The presence of the carboxy-hydroxy-fluorene is supported by the presence of the precursors, hydroxy-fluorene and carboxy-fluorene, as shown in Figure 3. Although the formation of carboxylated fluorenes is thermodynamically favorable, carboxylation of PAHs is predominantly reported at an anaerobic pathway for PAH transformation.^83,84 Few reports on aerobic formation of carboxylated PAHs were found in the literature. Mishra et al. (2020) reported the formation of 1-carboxyphenanthrene produced by the aerobic bacterium Zhihengliuella sp. ISTPL4, but did not comment on the mechanism.⁶¹ Grifoll et al. (1995) reported the formation of a carboxy-hydroxy-fluorene by Pseudomonas sp. strain F274, however this was reported to form from methyl-fluorene as the parent compound, rather than fluorene.⁶² Similarly, carboxy-hydroxy-fluorene was also reported as an aerobic transformation product of fluoranthene.^64,65 The purity of the fluorene standard was the lowest out of all parent PAHs in this study (98%), therefore it is possible that the carboxylated fluorenes were produced from a different parent PAH, such as methylfluorene or fluoranthene. Impurities should be characterized where possible in future work to support this hypothesis.

3.2.3. Sulfonated Fluorenes

Another new PAH transformation pathway for 21198 suggested by tentatively identified PAH-TPs is sulfonation, a phase II transformation that incorporates a sulfonic acid group to a hydroxylated PAH-TP. This is a common detoxification mechanism reported for other organisms, such as aquatic animal models, rodent models, and humans.^18,85,86 However, sulfonation by bacteria is rarely discussed in literature. Sulfonation has been implicated in sulfur metabolism or the formation of natural products such as sulfated glycolipids, but not as a predominant detoxification strategy for bacteria.^87,88 Because of this, the confidence level for the two sulfonated PAH-TPs was determined to be level 3. The isotope spectra match with the assigned chemical formula, sulfonation reactions are thermodynamically favorable, and precursors such as a dihydroxyfluorene have been tentatively identified. The structure prediction tool MetFrag output sulfonated aromatic compounds, but it did not predict sulfonated fluorenes. However, predictive tools such as MetFrag that pull structures from databases such as PubChem would not propose structures for sulfonated fluorenes, as they do not exist in these databases.

3.2.4. Anthracene and Phenanthrene Ring Fission Products

A ring fission product with two fused benzene rings is proposed as a potential new transformation product of anthracene and/or phenanthrene with level 2b confidence, as shown in Figure 4. This PAH-TP would be formed from the same enzymatic steps that produce diphenic acid, initial oxidation with a monooxygenase to form an epoxide, hydrolysis of the epoxide via an epoxide hydrolase, and ring cleavage with a ring cleaving dioxygenase. However, the location of the ring cleavage would be on one of the ends of phenanthrene or anthracene, rather than the k-region of phenanthrene. Ring fission products of this variety have been reported in many studies reported in Table 1, all of which formed through a similar pathway of ortho (intradiol) ring cleavage of dihydroxy- anthracene or phenanthrene.

Figure 4.

Potential transformation pathways for the production of the tentatively identified ring fission products of anthracene and phenanthrene.

Another ring fission product, 3,4-benzocoumarin was tentatively identified in this study, which agrees with the initial study. Different benzocoumarin isomers such as 6,7-benzocoumarin and 7,8-benzocoumarin have been reportedly formed during PAH biotransformation as well.⁸⁹ However, despite the abundance of benzocoumarin production reports, the mechanism for benzocoumarin production from PAHs is not well understood.

3.3. Unidentified potential PAH-TPs

Three features remained after all processing steps that were unable to be tentatively identified. Collected data for unknowns is provided in Table S5. Unknown 1, with the assigned chemical formula C₁₄H₉NO₂, was originally annotated as an aminoanthraquinone by the open-source library. However, a reference standard was purchased to confirm this identification, and the collected MS2 spectra of the reference standard did not match that of unknown 1. Furthermore, although anthraquinone was produced by 21198, the addition of an amine group to anthraquinone is not a documented transformation by aerobic bacteria. Unknown 2 was assigned the chemical formula C₃₅H₄₅N₈O₁₈P₃S and represented the only potential PAH-TP that had a coenzyme A (CoA) group substituted. Interestingly, the chemical formula for unknown 2 appears to be CoA plus unknown 1. However, the addition of CoA onto a compound typically involves the removal of an oxygen atom from the substrate⁹⁰, therefore the precursor to unknown 2 would have a chemical formula of C₁₄H₉NO. The final unknown, unknown 3, was assigned the chemical formula C₁₃H₉O. Despite the difference of this chemical formula to a hydroxyfluorene by only one hydrogen atom, no feasible structures were predicted. It is possible that rather than an M+H adduct, unknown 3 was a radical M+ adduct of hydroxy-fluorene, which can form from PAHs during ESI.⁹¹ This is supported by the nearly identical retention time between 2-hydroxyfluorene (11.846 min) and unknown 3 (11. 841 min).

3.4. PAH-TP Toxicity Screening Results

The tentatively identified compounds listed in Table 1, along with their parent compounds, were uploaded to the HCM using SMILES (provided in Table S6). The hazard summary provided by the HCM was used to develop the heatmap shown in Figure 5.

Figure 5.

Heatmap of the predicted toxicity of the PAH-TPs and their parent compounds by the HCM. Note that six endpoints were omitted: acute mammalian toxicity from inhalation or dermal exposure, neurotoxicity for single and repeat exposures, and systemic toxicity for single and repeat exposures. These endpoints were omitted due to no conclusive information available for any PAH-TPs compounds. The full output of the HCM is provided in Table S7.

Several endpoints are predicted for most, if not all potential PAH-TPs investigated. These include acute mammalian toxicity via oral exposure route, genotoxicity/mutagenicity, endocrine disruption, developmental toxicity, and acute aquatic toxicity. Developmental toxicity and acute aquatic toxicity are directly supported by the embryonic zebrafish toxicity assessment conducted in the previous study.³⁵ Furthermore, all aforementioned common endpoints have been established in the literature as potential effects of PAH exposure.^1,21,92 Other commonly reported endpoints for the PAH-TPs were carcinogenesis and skin sensitization, associated with anthraquinones, mono-hydroxylated phenanthrene and fluorene, diphenic acid, and carboxylated fluorene. Note that the endpoints predicted for the PAH-TPs are generally also shared with the parent PAHs. Therefore, despite being unregulated like their parent compounds, the PAH-TPs tentatively identified in this study share similar human and environmental health implications.

Three metrics of environmental fate; persistence, bioaccumulation, and exposure potential, were also reported by the HCM. Unlike the health effect trends, that tended to be similar between parent PAHs and their TPs, environmental persistence and bioaccumulation tended to decrease for TPs compared to parent PAHs. This is presumably due to the increased solubility of PAH-TPs, allowing them to be more mobile, susceptible to further degradation, and easier to excrete. Thus, the more polar groups added to the parent PAHs, the less persistent and less prone to biodegradation the PAH-TP is. However, this trend is not reflected in the exposure potential, as parent PAHs and PAH-TPs show greater similarity where data was available. This means that although the environmental residence time of PAH-TPs may be shorter, the potential for exposure and the subsequent health effects associated with exposure remain similar to their parent PAHs. The similarities between parent PAHs and PAH-TPs in regard to their toxicity and exposure potential suggest that PAH-TPs should be considered in risk assessments of PAH contaminated sites. Although PAH-TPs present similar health hazards as parent PAHs, their lower persistence and bioaccumulation potential indicate that the long-term environmental prevalence of PAH-TPs is of less concern than parent PAHs.

A significant limitation of the reported health effects for the selected compounds reported in Figure 5 is that they are for a single isomer of a PAH-TP. The toxicity of a PAH-TP can be isomer dependent, meaning that the position of the functional groups on the aromatic rings changes bioactivity.^21,93,94 However, experimental information is often needed to distinguish isomer effects, and most of information reported from the HCM for the PAH-TPs was generated from QSARs. QSARs are generally not robust enough to predict isomer effects.⁹⁵ For the HCM, a test case was conducted with 5 isomers of dihydroxyfluorene to determine if the HCM would report different health effects for different isomers. It was observed that the predicted level of effect (L to VH) varied between isomers where data was available. However, the predicted endpoints were the same across all isomers, as shown in Table S8. Since the HCM only reports a summary of qualitative information, a representative isomer for each compound was chosen to report in Figure 5. Further experimental work is required to not only identify which isomers of the proposed PAH-TPs are present in the sample, but also investigate the impact of the functional group positions on health outcomes.

3.5. Challenges and Future Directions

The NTA workflow presented in this study elucidated 16 PAH-TPs within a complex mixture produced by bacteria cometabolism, seven of which were not identified in the initial target screening conducted previously. Despite this success, several challenges limited the ability to tentatively identify novel PAH-TPs. Because PAH-TPs are a class of environmental contaminants underrepresented in public databases, the literature, and mass spectrometry libraries, the ability of predictive tools like MetFrag and CFM-ID to generate feasible PAH-TP structures is limited. Similarly, the availability of reference standards for PAH-TPs is limited, hindering the ability to confirm structures by comparing to a reference standard analyzed on the same instrument. Of the seven PAH-TPs newly tentatively identified in this study, only one is commercially available (1-carboxyfluorene). The cost of synthesizing reference standards to confirm the identification of the proposed PAH-TPs tentatively identified by NTA constrains this path towards structure confirmation and highlights the need for reference free methods for confirming structures. However, orthogonal techniques such as NMR or ion mobility could be applied in future studies to identify structural isomers.

Another challenge specific to the bioremediation system explored in this study is that the majority of tools that predict biotransformation (e.g. EAWAG-BBD-PSS, KEGG, etc.) favor metabolic pathways over cometabolic pathways. Similarly, PAH degradation pathways reported in literature are often metabolic pathways, which differ significantly from cometabolic pathways in that complete mineralization is typically not achieved. The inability to mineralize a contaminant leads to the accumulation of cometabolic transformation products, which have the potential to negatively impact both environmental and human health.

Overall, the suspect and non-target screening workflow presented in this study combined a wide variety of tools to identify PAH-TPs produced from bioremediation with a bacterial pure culture. Applying this workflow expanded the potential transformation pathways used by 21198 to cometabolize PAHs to include carboxylation and sulfonation. Furthermore, novel PAH-TPs were tentatively identified that have not been previously known to be produced by bacteria, including trihydroxyfluorene and sulfonated fluorenes. All PAH-TPs tentatively identified were predicted by the US EPA Hazard Cheminformatics Module to have toxic effects similar to their parent compounds, highlighting the urgent need for more research, and possibly regulatory measures addressing PAH-TPs in the environment.

Figure 2.

The number of detected features remaining after each filtering step, defined to the right of the bar graph and described in detail in section 2.7. The criteria for chemical formulas included formulas only containing C, H, O, P, S, and N, ratios of elements that matched phase II metabolites listed in Table S3, and a carbon number between 13 and 16. Note that the x-axis is displayed on a logarithmic scale.

Synopsis.

This study uses new approaches to identify and estimate the toxicity of contaminants produced during the breakdown of PAHs that are not able to be identified by traditional methods.

Data Availability Statement

Data will be made available upon request.