A refined non-target workflow for the investigation of metabolites

Subscriber access provided by UNIV OF LOUISIANA

Article

A refined non-target workflow for the investigation of metabolites through the prioritization by in-silico prediction tools Lubertus Bijlsma, Marc H.G. Berntssen, and Sylvain Merel Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.9b01218 • Publication Date (Web): 11 Apr 2019 Downloaded from http://pubs.acs.org on April 12, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

A refined non-target workflow for the investigation of metabolites through the prioritization by in-silico prediction tools Lubertus Bijlsma†,‡§,*, Marc H.G. Berntssen‡, Sylvain Merel†§,‡ † Research Institute for Pesticides and Water, University Jaume I, Avda. Sos Baynat s/n, E-12071 Castellón, Spain. ‡ Institute of Marine Research, P.O. Box 2029 Nordness, N-5817 Bergen, Norway § Visiting scientist ABSTRACT: The application of non-targeted strategies based on high-resolution mass spectrometry (HRMS), directed towards the discovery of metabolites of known contaminants in fish is an interesting alternative to true non-target screening. To reduce prolonged and costly laboratory experiments, recent advances in computing power have permitted the development of comprehensive knowledge-based software to predict the metabolic fate of chemicals. In addition, machine-based learning tools allows the prediction chromatographic retention times (RT) or collision cross section (CCS) values when using ion mobility spectrometry (IMS). These tools can ease data evaluation and strengthen the confidence in the identification of compounds. The current work explores the capabilities of in-silico prediction tools, refined by the use of RT and CCS prediction, to prioritize and facilitate non-target LC-IMS-HRMS data processing. The fate of the insecticide pirimiphos-methyl (PM) in farmed Atlantic salmon exposed to contaminated feed was used as a case study. The theoretical prediction of 60 potentially relevant biological PM metabolites permitted the prioritization of screening in different salmon tissues (liver, kidney, bile, muscle and fat) of known and unknown PM metabolites. An average of 43 potential positives was found in the sample matrices based on the accurate mass of protonated molecules (mass error ≤ 5 ppm). The application of different tolerance filters for RT (Δ ≤ 2 min) and CCS (Δ ≤ 6%) based on predicted values permitted to reduce this number up to 66% of the features. Finally, five PM metabolites could be identified, two known metabolites (2-DAMP and N-desethyl PM) were confirmed with a standard, whereas three previously unknown metabolites (2-DAMP glucuronide, didesethyl PM and hydroxy-2-DAMP glucuronide) were tentatively identified in different matrices, allowing the first proposition of a metabolic pathway in fish.

A non-target screening based on liquid chromatography coupled to high-resolution mass spectrometry (HRMS), without any a priori information on the compounds to be detected is a challenging task. The data obtained, generally, contain large sets of features (i.e. retention time (RT), mass-to-charge ratio, peak intensities) characteristic of known and unknown compounds. Revision of this data is time-consuming, especially when analyzing samples from complex matrices, and the success rate of identifying candidates depends strongly on available compound databases1–3. Priorization in non-target approaches is, therefore, of high interest in order to reduce the number of features and facilitate data processing and data interpretation. The application of non-targeted strategies, directed towards the discovery of metabolites or degradation products of known compounds, is less laborious as it is limited by the number of meaningful structures that can be assigned to an unknown peak. In this workflow, ion mobility spectrometry

(IMS) has added an extra-dimension to further improve the identification process, making use of the drift time of an ion (i.e. the time an ion takes to travel through the mobility cell), in addition to the obtained RT and accurate mass4. Furthermore, drift times obtained during the IMS separation enables a much better alignment of (de)protonated molecules and fragment ions, resulting in cleaner and higher-quality spectra5. Directed non-target approaches, however, requires prolonged and costly laboratory experiments following well elaborated study designs6. Recent advances in computing power have permitted the development of comprehensive knowledge-based software to predict the metabolic fate or chemical degradation7,8. One commercial available software package for in-silico prediction is Meteor Nexus developed by Lhasa Limited. This computational tool predicts the metabolic fate of chemicals in mammals by using structure activity relationship (SAR) analysis. Absolute and relative reasoning are then used 1

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

for the selection and ranking of predicted metabolites7,9. The use of this in-silico prediction tool in combination with LC-HRMS analysis has been employed and demonstrated its potential in metabolism studies9–11. Hence, there is a growing focus on using such in-silico predictions to help the discovery of unknown metabolites or transformation products (TPs) by introducing “targets” in the untargeted workflow. In addition, machinebased learning tools allow the prediction of RT and collision cross section (CCS) values12–15. The latter can be derived from the drift time obtained when using IMS, providing information about the shape and size of an ionic molecule. Both RT and CCS prediction allows a reduction in the number of potential positives to be prioritized and evaluated and improves the confidence in the identification of suspects. In practice, non-target screening approaches are applied in multiple fields in order to identify or discover all kinds of contaminants. For instance, the growing use of agricultural products in the production of fish feed raises concern regarding the potential transfer and fate of pesticides. In particular, the insecticide pirimiphos methyl (PM) has been detected in plant ingredients and related salmon feed4,16–18, but its biological metabolism in farmed seafood such as Atlantic salmon remains unknown, while the main metabolites have been identified in mammalian models19. The identification of biological pesticide metabolites in food producing animals is part of the risk assessment of animal health and consumer safety19,20. In the present work, we investigate the potential of computational tools for the prioritization of features in a nontargeted approach using ultra-high performance liquid chromatography (UHPLC) coupled to IMS with quadrupoletime-of-flight mass spectrometry (QTOF MS). As a case study, the biological formation of PM metabolites in farmed Atlantic salmon were assessed. In a first stage, theoretically biological PM metabolites were predicted using a knowledge-based software tool, Meteor Nexus. Subsequently, RT and CCS values were predicted for each theoretical metabolite using Artificial Neural Networks (ANN), a machine-based learning tool. A library was then built, using all predicted information, and applied to screen UHPLC-IMS-QTOF MS data of different tissues of Atlantic salmon that was fed a controlled PM dose during a 3-month period. Finally, the application of this library and different tolerance filters (i.e. for mass accuracy, RT and CCS) permitted to refine data processing and to identify five PM metabolites (known and unknown) in a more time and cost effective manner. EXPERIMENTAL SECTION

Chemicals, Reagents, and Materials. Pirimiphos-methyl (PM), N-desethyl pirimiphos-methyl (N-Desethyl PM), 2diethylamino-6-methyl-4-pyrimidinol (2-DAMP) and leucineenkephalin, used for mass correction, were purchased from Sigma-Aldrich (St Louis, MO, USA). Deuterated pirimiphosmethyl (PM-d6) was acquired from Dr. Ehrenstorfer (Augsburg, Germany). Reference standards had purities higher than 98% (w/w). HPLC-grade acetonitrile (ACN), HPLC-grade methanol (MeOH) and residue analysis-grade acetone were procured from Scharlau (Barcelona, Spain) along with formic acid (HCOOH, > 98%) and extra pure anhydrous magnesium sulphate (MgSO4). HPLC-grade water (H2O) was obtained by purifying demineralized water in a Milli-Q plus system from Millipore (Bedford, MA, USA).

Page 2 of 9

Instrumentation. The chromatographic separation was performed using an Acquity UPLC Cortecs C18 2.1 × 100 mm, 1.7 μm (Waters) at a flow rate of 300 μL/min. Gradient elution was performed using mobile phases of A = H2O and B = MeOH, both with 0.01% HCOOH. The initial percentage of B (10%) was linearly increased to 90% in 14 min, followed by a 2 min isocratic period, then, returned to initial conditions for 2 min. The total run time was 18 min. The column temperature was set to 40 °C and sample temperature at 10 °C. The injection volume was 1 μL. A Waters Acquity I-Class UPLC system (Waters, Milford, MA, U.S.A.) was interfaced to a VION IMS-QTOF MS, using an electrospray (ESI) interface operating in positive ionization mode with a resolution of approximately 40.000 at full width half maximum (FWHM) at m/z 556. A capillary voltage of 0.7 kV and cone voltage of 40 V were established. Nitrogen was used as the drying gas and nebulizing gas. The desolvation temperature was set to 550°C, and the source temperature to 120°C. The cone gas flow was 250 L/h and desolvation gas flow was 1000 L/h. MS data was acquired using the VION in highdefinition (HD) MSe mode, in the range m/z 60−1000, with Nitrogen as the drift gas. Two independent scans with a scan time of 0.3 s but different collision energies were acquired during the run: a collision energy of 6 eV for low energy (LE) and a ramp of 28 − 56 eV for high energy (HE). Nitrogen (≥ 99.999%) was employed as collision-induced dissociation (CID) gas. All data was examined using an accurate mass screening workflow within UNIFI informatics platform from Waters Corporation. The setting of the peak-picking process were as followed: both chromatographic and drift peak width were automatically defined by UNIFI, intensity thresholds were LE ≥ 200 counts, HE ≥ 100 counts and the background filter noise was set “high”. The maximum number of features to keep per sample during the peak-picking process was established at 50,000. Sample Collection and Preparation. During 3 months, Atlantic salmon (Salmo salar) in triplicate tanks were fed respectively a control feed that did not contain PM and a feed spiked with PM at a concentration of 15.2 mg/kg. The feed was spiked by dissolving the PM in fish oil, which was added to PM free commercial feed pellets by vacuum top coating as described by Berntssen et al.21 The feeds were stored at -20 °C before and during the feeding trial. At the end of the feeding trial, fish were terminal anaesthetized with tricaine methanesulfonate (MS-222; ~60 mg/L). Liver, kidney, bile, muscle and fat were sampled flash frozen and subsequently stored at -80 °C. The experiment complied with the guidelines of the Norwegian Regulation on Animal Experimentation and EC Directive 86/609/EEC. The National Animal Research Authority approved the protocol (12091). Once in the laboratory, samples were extracted with a mixture of ACN and acetone (8/2) with 1% HCOOH. After adding MgSO4 and centrifugation, the supernatant was diluted with H2O and spiked with internal standard PM-d6 to a final volume of 1 mL. Extracts were filtered (0.45 µm, nylon) and 1 µL of each extract was directly injected into the UHPLC-IMS-QTOF MS system. More details on each extraction step of the different salmon matrices are described in Table S1. Prediction. The fate of PM, i.e. biological metabolites, was predicted with Meteor Nexus from Lhasa Limited (Leeds, UK), considering a total of 566 bio-transformations but limiting the 2


Page 3 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry pathway to 3 levels. The software tool relies both on expert knowledge base of patterns and reasoning rules to propose the most likely transformations under specific conditions22,23. The Artificial Neural Networks (ANN) predictions of RT and CCS values were made using Alyuda NeuroIntelligence 2.2 (Cupertino, CA). The RT and CCS predictors were previously developed and optimized by using different algorithms and data of 544 and 205 small molecules, respectively. Molecules were partitioned into training:verification:blind test sets in the ratio 68:16:16. The final network designs selected were 4-layer perceptrons, 16-19-9-1 (RT) and 8-2-8-1 (CCS). More details of the predictors can be found elsewhere12,13. Simplified molecular-input line-entry system (SMILES) codes of the predicted metabolites were obtained from ChemDraw Professional v17.1 from PerkinElmer by converting their molecular structures. Relevant molecular descriptors for RT and CCS prediction (16 and 8, respectively) were obtained from PC Client freeware within VCC-Labs24. RT and CCS values of each PM metabolite could then be predicted as described previously. All the results from the predictions of potential metabolites, RT and CCS were gathered to build a library (Table S2) used for data processing. This library included PM and 60 metabolites. Overall, these metabolites mostly consist in multiple cleavage and oxygenation of PM as well as, glucuronide conjugation. Workflow. The usual non-target workflow involves data acquisition followed by peak picking and data processing based on mass filter (≤ 5 ppm) and evaluation of characteristic fragment ions to reach structure elucidation and identification. In this study, we modified such general workflow by the implementation of in-silico prediction (Figure 1). Knowing the parent compound, metabolism knowledge-based software allowed creating an exhaustive library avoiding small metabolites (m/z < 60) and structures impossible or very unlikely to be formed. The library was then enriched with ANN predicted RT and CCS, which allowed further data reduction while improving confidence in structure elucidation and identification.

Figure 1. Schematic presentation of the general workflow applied in the directed non-target approach. RESULT AND DISCUSSION

Prediction. Meteor Nexus uses a comprehensive, expert knowledge base to predict the metabolic fate of chemicals i.e. biological including all enzymatic systems (phase I, II, III)22,23. A total of 203 metabolites were predicted using PM as input. After excluding duplicates and small molecules (m/z < 60) due to the acquisition of spectra in the range m/z 60−1000, 60 metabolites were left of which 41 had a unique molecular formula (Table S2). SMILES codes of the 60 metabolites were generated and introduced in PC Client freeware to generate the molecular descriptors used for RT and CCS prediction based on ANN. Prediction of RT and CCS can play a role in increasing confidence in screening efforts for new metabolites, where experimentally derived data based on reference standards are unavailable25. The robustness of the RT and CCS predictors and selected ANN architectures were evaluated with compounds randomly partitioned into training, validation, and blind test sets12,13. RT and CCS could be predicted within 2 min and 6% relative error, respectively, when considering the 95th confidence interval. RT depends strongly on the reversed-phase LC conditions used. Therefore, practically the same chromatographic conditions used for developing the RT predictor12 were applied to perform the analysis in this study. The influence on the RT prediction caused by the use of a different instrument (possible effect due to dead volume) and chromatographic reversed-phase C18 column (Cortecs versus BEH, both 2.1 × 100 mm, 1.7 μm) was tested by analyzing and running the same blind test set as used in the work of Bade et al.12. RT revealed 95% prediction accuracy within the 2-minute elution interval. Unlike the RT, CCS values are independent of the used LC conditions and are unaffected by the sample matrix. Experimental CCS values (± 2% tolerance) have been proven to be robust across multiple platforms and conditions, when using the same buffer gas4,16,26,27. CCS has demonstrated to be very useful in identifying compounds as it renders, in combination with LC and HRMS data, increasing peak capacity and selectivity in complex samples while constraining false-negative and falsepositive detections16. However, there is currently no empirical CCS library available, thus on the basis of CCS, compounds discovered when applying suspect or non-target strategies cannot be identified. Hence, especially for compounds for which standards are not commonly available, the prediction of CCS, which is completely universal for small molecules, is of interest. The relative error of CCS prediction was within 6% for 95th percentile of all values for protonated molecules13. Predicted RT and CCS values of each PM metabolite, together with their molecular formula, exact mass and SMILES codes were incorporated in the compound library (Table S2) for data processing. Prioritizing Features. Sample extracts of liver, kidney, bile, muscle and fat of Atlantic salmon were analyzed by UHPLCIMS-QTOF MS. Data processing was prioritized and the number of features to evaluate was reduced by employing the library of potential metabolites with tolerance filters for mass accuracy ≤ 5 ppm, RT Δ ≤ 2 min and/or CCS Δ ≤ 6 %. Table 1 gives an overview of the number of features in each matrix based on the distinct filters applied. An average of 43 potential positives (range 30 – 67) were found in the different matrices. 3



Page 4 of 9

This number could be reduced by an average of 39% and 14% prediction, respectively. An average reduction of 52% was when applying mass accuracy and the filter for RT or CCS obtained when Table 1: Effect of data filtration on the number of features to be evaluated as potential positive metabolites of PM in different Atlantic salmon matrices (in between brackets is the % of reduction). Filters

Tolerances

Liver

Kidney

Bile

Muscle

Fat

Mass accuracy

≤ 5 ppm

43

40

67

37

30

Mass accuracy and RT prediction

≤ 5 ppm, ≤ 2 min

21 (51%)

17 (57%)

34 (49%)

23 (38%)

19 (37%)

Mass accuracy and CCS prediction

≤ 5 ppm, ≤ 6 %

39 (9%)

35 (12%)

41 (39%)

35 (5%)

28 (7%)

Mass accuracy, RT prediction and CCS prediction ≤ 5 ppm, ≤ 2 min, ≤ 6% 21 (51%)

16 (60%)

23 (66%)

23 (38%)

17 (43%)

Figure 2. Schematic presentation of the prioritization process for the positive finding of 2-DAMP in a bile sample. In the center are all features observed after extracting m/z 182.12879 ≤ 5ppm. The features are prioritized by the application of RT and CCS prediction (green field in the center). At the top and bottom are the chromatograms corresponding to the bile sample and reference standard, 4


Page 5 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry respectively, which include the predicted retention time of 2-DAMP (red dotted line) and the 2 min tolerance window. At the left and right of the scheme are the mobilograms corresponding to the bile sample and reference standard, respectively, which include the predicted CCS value of 2-DAMP (red dotted line) and the 6% tolerance window. prioritizing features to be reviewed, but should not be applying all three tolerance filters. The larger data reduction considered as an absolute exclusion criterion. when applying the RT filter compared to when applying the CCS filter only was expected. Several previous studies have When searching for metabolite M373 with m/z 374.15579 ≤ shown that the RT of a compound is mainly driven by its 5 ppm, data reduction by means of RT prediction did not occur. hydrophobicity (log D)28–30 while the CCS is mainly driven by Hence, two features with a RT of 1.65 min and 2.81 min were its mass4. As a result, screening for a specific mass will result a potential match. However, in this case, CCS prediction in positive findings distributed across the entire chromatogram, together with the elucidation of corresponding fragment ions, but clustered in a limited area of the mobilogram. Therefore, permitted the tentative identification of hydroxy-2-DAMP data prioritization applying restriction on RT will leave out glucuronide as the chromatographic peak eluting at 2.81 min more potential positives than when applying restriction on CCS (Figure S1). Indeed, the chromatographic peak at 1.65 had a only. 6.39% difference compared to the predicted CCS value (190.7 Å), which makes it out of the 95% confidence interval. A Figure 2 demonstrates the prioritization process for a positive reference standard is, however, needed to confirm the identity finding in a bile sample. When searching for the metabolite M181 also known as 2-diethylamino-6-methyl-4-pyrimidinol of the compound 31,32. (2-DAMP), eight features with m/z 182.12879 ≤ 5 ppm were In general, the prediction tools applied allowed gaining more found, including co-eluting isomers. The different RT observed confidence in the identification process during suspect or nonwere 1.36, 1.59, 1.78, 3.28, 3.70 and 7.34 min, but by applying target analysis. These prediction tools, especially CCS due to the filter to the predicted RT of M181 (4.60 ≤ 2 min), five its experimental robustness across platforms, would give an features were not further considered at that stage. Thus, three extra-dimension to the confidence levels proposed by features with chromatographic peaks eluting at 3.28 and 3.70 Schymanski et al. 32. Obviously, the development of empirical min were prioritized. CCS libraries would empower this even more. Subsequently, the CCS filter was employed (138.15 Å ≤ 6%), Identified Compounds. Table 2 gives an overview of the which permitted the reduction of one more feature and thus left compounds identified in the liver, kidney, bile, muscle and fat solely two features to be further investigated. Based on the of Atlantic salmon. PM, 2-DAMP and N-desethyl PM could be observed fragment ions in the HE spectra, none of the two identified and confirmed using reference standards. Other features prioritized could be excluded as both features comply metabolites, 2-DAMP glucuronide, didesethyl PM and with the criteria for a tentative identification i.e. accurate mass hydroxy-2-DAMP glucuronide were tentatively identified in for protonated molecule and at least one fragment ion (≤ 5 ppm) different matrices based on their accurate mass, predicted RT plus the predicted RT and CCS values. Therefore, a reference and CSS, and at least one explained fragment ion (≤ 5 ppm). As standard was purchased, which allowed the confirmation of the can be observed, the average mass error ranged from 0.2 – 2.1 peak eluting at 3.28 min as 2-DAMP, even though the other ppm, the average deviation compared to predicted RT ranged peak at 3.70 min was more abundant and was fitting better with from 0.40 – 1.89 min, whereas CCS could be predicted within the predicted RT and CCS values. Therefore, when several an average relative error in the range of 0.11 – 3.25%. Although features pass all the filters (mass error ≤ 5ppm, RT ≤ 2 min, high confidence levels were reached, it is worth emphasizing CCS ≤ 6%), no tentative assignment should be made based on that reference standards are required to achieve a Level 1 the abundance or the best fit with predicted RT and CCS values. identification as proposed by Schymanski et al.32 and according In fact, it is important to emphasize that predicted data are to the Metabolite Standards Initiative31. within a 95% of confidence interval. So, the prediction aids Table 2. Overview of the compounds identified in Atlantic salmon matrices

Formula

Matrix*

Mass error (ppm)

Observed RT (min)

Observed CCS (Å²)

ΔRT (min)**

ΔCCS (%)**

C11H20N3O3PS

L, K, B, M, F

0.5

12.03

164.88

1.05

2.81

C9H15N3O

L, K, B, M, F

2.1

3.28

138.35

1.36

0.44

C9H16N3O3PS

L, K, M, F

0.6

8.58

159.01

0.40

0.11

2-DAMP glucuronide (M357)

C15H23N3O7

B

0.4

1.36

182.01

1.89

2.69

Didesethyl PM (M249)

C7H12N3O3PS

L, K, M

0.2

6.07

149.77

0.88

1.51

Hydroxy-2-DAMP glucuronide (M373)

C15H23N3O8

B

2.2

2.81

184.51

0.72

3.25

Compound name Pirimiphos methyl (PM) 2-DAMP (M181) N-desethyl PM (M277a)

* L stands for liver, K for kidney, L for bile, M for muscle and F for fat ** Average deviation with respect to the predicted values

5



Page 6 of 9

Figure 3. The proposed metabolic pathway of PM in Atlantic salmon and relative distribution of the compounds across the tissues.

Metabolic Pathway of PM in Atlantic Salmon. The controlled salmon exposure combined with the application of UHPLC-IMS-QTOF MS and in-silico prediction revealed the transformation of PM into 5 metabolites (Table 2) among which 3 have not been previously characterized. The metabolic pathway (Figure 3) indicates that PM was initially metabolized either in 2-DAMP (M181) by enzymatic hydrolysis of phosphorothionate ester or in N-desethyl PM (M277a) by oxidative N-dealkylation via cytochrome P450 (CYP450). On the one hand, N-desethyl PM was then further metabolized into

didesethyl PM (M249) through a second N-dealkylation. On the other hand, 2-DAMP was further metabolized through the action of glucuronosyltransferase to form 2-DAMP glucuronide (M357). In addition, 2-DAMP was also transformed through the hydroxylation of methyl group on the aromatic ring to form hydroxy-2-DAMP (M197a), an unobserved intermediate subsequently forming the hydroxy-2-DAMP glucuronide conjugate (M373). 6


Page 7 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry The accurate quantification of PM metabolites could not be performed in the current study, due to the lack of analytical standards and to the fact that extraction efficiencies as well as matrix effects were not assessed. However, comparing the responses of the different compounds allows a semiquantification. In this case, PM the insecticide to which salmon was exposed showed 53% relative abundance. In particular, a large fraction of PM (75%) was detected in the kidney extract while another two fractions (10% each) were detected in muscle and fat extracts, which might indicate a potential accumulation in parts used for human consumption. The first phase I metabolites, 2-DAMP and N-desethyl PM showed relative abundances of 19% and 7%, respectively. Both metabolites were mainly detected in liver, kidney and muscle extracts (Figure 3). The last phase I metabolite, didesethyl PM, showed a relative abundance lower than 1% and was distributed between liver (39%), kidney (43%) and muscle (18%). Finally, the phase II metabolites 2-DAMP glucuronide and hydroxy-2DAMP glucuronide represented respectively 20% and less than 1% relative abundance, but both were exclusively detected in bile samples. The fate of PM in salmon presented previously is the first description of the metabolic pathway of this pesticide in fish, therefore providing novel and relevant knowledge for policy makers and expert committees in charge of evaluating pesticides in the food production chain such as the European Food Safety Authority (EFSA). Indeed, previous research focused on the metabolism of PM in terrestrial livestock20 or the chemical degradation of PM during storage of products such as rice grains33, maize grains34 and wheat grains35. Overall, the observation made on salmon are consistent with the existing literature on other food producing animals. For instance, a study on lactating goat20 and another one on chicken36 showed that residual of PM were mostly found in kidney (on average 63% of PM in chicken), followed by liver and muscle. Moreover, the identification of 2-DAMP and its glucuronide conjugate as the most abundant PM metabolites in fish is also consistent with previous studies in other matrices. For instance, 2-DAMP was previously characterized during the photocatalytic degradation of PM37, in cereal grain20 as well as in urine samples from children in farmworkers households potentially exposed to pesticides38. While a study in livestock revealed the occurrence of conjugated 2-DAMP in liver and kidney after sample preparation involving acid hydrolysis20, the current application of UHPLC-IMS-QTOF MS characterized the conjugate as a glucuronide although for salmon it was only detected in bile samples. In addition, N-desethyl PM was also reported as one of the main metabolites in poultry and ruminant, particularly in samples of milk, eggs, muscle, liver and kidney. However, possibly due to different metabolism or their low abundance, didesethyl PM and hydroxy-2-DAMP glucuronide are novel metabolites since they have not been previously characterized in peer-reviewed literature. Finally, the findings from the current study on the fate of PM in salmon also highlight the need for further research on the toxicity of the metabolites and more particularly on 2-DAMP, as requested by the EFSA20. More particularly, quantitative structure–activity relationship (QSAR) with potential toxic pathways as outcome would represent additional in-silico prediction to be implemented in the non-target workflow to further reduce the amount of data to process by not prioritizing compounds predicted to have no toxic effect by several QSAR models.

CONCLUSIONS

In-silico prediction permitted to prioritize and facilitate nontarget UHPLC-IMS-QTOF MS data processing thus aiding the assessment of dietary PM metabolism in farmed Atlantic salmon. The application of different tolerance filters for predicted RT (Δ ≤ 2 min) and CCS (Δ ≤ 6 %) in addition to mass accuracy (≤ 5 ppm) allowed an average reduction of 52% in the number of features to be further investigated. Based on data obtained, a metabolic pathway of PM in salmon could be proposed, revealing two known metabolites of PM (2-DAMP and N-desethyl PM) and the putative identification of three metabolites not previously reported (2-DAMP glucuronide, didesethyl PM and hydroxy-2-DAMP glucuronide). Moreover, the predicted RT and CCS values support and gives more confidence to the procured results. The presented strategy can easily be applied to other research fields such as environment, medicine and forensic studies in order to obtain relevant information in a more time and cost effective manner.

ASSOCIATED CONTENT Supporting information Additional tables and figures (PDF) as noted in the text are available free of charge via the ACS Publications Website at http://pubs.acs.org.

AUTHOR INFORMATION Corresponding author * E-mail: [email protected] Tel.: +34 964 387452

Author contributions The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

ACKNOWLEDGMENTS The authors acknowledge the financial support of the Norwegian Research council project “AQUASAFE” (254807). The authors would like to thank Dr. G. Rosenlund and Tårn Helgøy Thomsen for designing, sampling, and performing the feeding trial at the Skretting ARC, Lerang Research Station. The authors thank Borja Garlito Molina for performing sample extraction and facilitating the UHPLC-IMS-QTOF MS data. L. Bijlsma wish to thank Marc Berntssen and the Institute of Marine Research (Norway) for hosting him as visiting researcher. S. Merel is grateful to Dr. Felix Hernandez for allowing him to visit his research group at the University of Jaume I (Spain).

REFERENCES (1) Hernández, F.; Bakker, J.; Bijlsma, L.; de Boer, J.; BoteroCoy, A. M.; Bruinen de Bruin, Y.; Fischer, S.; Hollender, J.; KasprzykHordern, B.; Lamoree, M.; López, F. J.; Laak, T. L. ter; van Leerdam, J. A.; Sancho, J. V.; Schymanski, E. L.; de Voogt, P.; Hogendoorn, E. A. The role of analytical chemistry in exposure science: Focus on the aquatic environment. Chemosphere 2019, 222, 564–583. (2) Zedda, M.; Zwiener, C. Is nontarget screening of emerging contaminants by LC-HRMS successful? A plea for compound libraries and computer tools. Anal. Bioanal. Chem. 2012, 403 (9), 2493–2502. (3) Merel, S.; Anumol, T.; Park, M.; Snyder, S. A. Application of surrogates, indicators, and high-resolution mass spectrometry to

7



evaluate the efficacy of UV processes for attenuation of emerging contaminants in water. J. Hazard. Mater. 2015, 282, 75–85. (4) Regueiro, J.; Negreira, N.; Berntssen, M. H. G. Ionmobility-derived collision cross section as an additional identification point for multiresidue screening of pesticides in fish feed. Anal. Chem. 2016, 88 (22), 11169–11177. (5) Damen, C. W. N.; Isaac, G.; Langridge, J.; Hankemeier, T.; Vreeken, R. J. Enhanced lipid isomer separation in human plasma using reversed-phase UPLC with ion-mobility/high-resolution MS detection. J. Lipid Res. 2014, 55 (8), 1772–1783. (6) Calza, P.; Medana, C.; Padovano, E.; Dal Bello, F.; Baiocchi, C. Identification of the unknown transformation products derived from lincomycin using LC-HRMS technique. J. Mass Spectrom. 2012, 47 (6), 751–759. (7) Djoumbou-Feunang, Y.; Fiamoncini, J.; Gil-de-la-Fuente, A.; Greiner, R.; Manach, C.; Wishart, D. S. BioTransformer: A comprehensive computational tool for small molecule metabolism prediction and metabolite identification. J. Cheminform. 2019, 11 (2), 1–25. (8) Kern, S.; Fenner, K.; Singer, H. P.; Schwarzenbach, R. P.; Hollender, J. Identification of transformation products of organic contaminants in natural waters by computer-aided prediction and highresolution mass spectrometry. Environ. Sci Technol 2009, 43, 7039– 7046. (9) Ballesteros-Gómez, A.; Erratico, C. A.; Eede, N. Van den; Ionas, A. C.; Leonards, P. E. G.; Covaci, A. In vitro metabolism of 2ethylhexyldiphenyl phosphate (EHDPHP) by human liver microsomes. Toxicol. Lett. 2015, 232 (1), 203–212. (10) Tyrkkö, E.; Pelander, A.; Ketola, R. A.; Ojanperä, I. In silico and in vitro metabolism studies support identification of designer drugs in human urine by liquid chromatography/quadrupole-time-of-flight mass spectrometry. Anal. Bioanal. Chem. 2013, 405 (21), 6697–6709. (11) Valerio, G. L.; Long, A. The In Silico Prediction of HumanSpecific Metabolites from Hepatotoxic Drugs. Curr. Drug Discov. Technol. 2010, 7 (3), 170–187. (12) Bade, R.; Bijlsma, L.; Miller, T. H.; Barron, L. P.; Sancho, J. V.; Hernández, F. Suspect screening of large numbers of emerging contaminants in environmental waters using artificial neural networks for chromatographic retention time prediction and high resolution mass spectrometry data analysis. Sci. Total Environ. 2015, 538, 934–941. (13) Bijlsma, L.; Bade, R.; Celma, A.; Mullin, L.; Cleland, G.; Stead, S.; Hernandez, F.; Sancho, J. V. Prediction of Collision CrossSection Values for Small Molecules: Application to Pesticide Residue Analysis. Anal. Chem. 2017, 89 (12), 6583–6589. (14) Miller, T. H.; Musenga, A.; Cowan, D. a; Barron, L. P. Prediction of Chromatographic Retention Time in HighResolution Anti-Doping Screening Data Using Artificial Neural Networks. Anal. Chem. 2013, 85, 10330–10337. (15) Zhou, Z.; Shen, X.; Tu, J.; Zhu, Z.-J. Large-Scale Prediction of Collision Cross-Section Values for Metabolites in Ion MobilityMass Spectrometry. Anal. Chem. 2016, 88, 11084–11091. (16) Regueiro, J.; Negreira, N.; Hannisdal, R.; Berntssen, M. H. G. Targeted approach for qualitative screening of pesticides in salmon feed by liquid chromatography coupled to traveling-wave ion mobility/quadrupole time-of-flight mass spectrometry. Food Control 2017, 78, 116–125. (17) Portolés, T.; Ibáñez, M.; Garlito, B.; Nácher-Mestre, J.; Karalazos, V.; Silva, J.; Alm, M.; Serrano, R.; Pérez-Sánchez, J.; Hernández, F.; Berntssen, M. H. G. Comprehensive strategy for pesticide residue analysis through the production cycle of gilthead sea bream and Atlantic salmon. Chemosphere 2017, 179, 242–253. (18) Nácher-Mestre, J.; Berntssen, M. H. G.; Serrano, R.; Garlito, B.; Hernández, F.; Ballester-Lozano, G. F.; Portolés, T.; CalduchGiner, J.; Pérez-Sánchez, J. Comprehensive overview of feed-to-fillet transfer of new and traditional contaminants in Atlantic salmon and gilthead sea bream fed plant-based diets. Aquac. Nutr. 2018, 24 (6), 1782–1795. (19) EFSA. EFSA scientific report; 2005.

Page 8 of 9

(20) EFSA. Reasoned opinion on the review of the existing maximum residue levels (MRLs) for 1 naphthylacetamide and 1 naphthylacetic acid according to Article 12 of Regulation (EC) No 396/2005. EFSA J. 2015, 13 (1), 3974. (21) Berntssen, M. H. G.; Hannisdal, R.; Buttle, L.; Hoogenveen, R.; Mengelers, M.; Bokkers, B. G. H.; Zeilmaker, M. J. Modelling the long-term feed-to-fillet transfer of leuco crystal violet and leuco malachite green in Atlantic salmon (Salmo salar). Food Addit. Contam. - Part A Chem. Anal. Control. Expo. Risk Assess. 2018, 35 (8), 1484– 1496. (22) Judson, P. N.; Long, A.; Murray, E.; Patel, M. Assessing Confidence in Predictions Using Veracity and Utility - A Case Study on the Prediction of Mammalian Metabolism by Meteor Nexus. Mol. Inform. 2015, 34 (5), 284–291. (23) Kleinman, M. H.; Baertschi, S. W.; Alsante, K. M.; Reid, D. L.; Mowery, M. D.; Shimanovich, R.; Foti, C.; Smith, W. K.; Reynolds, D. W.; Nefliu, M.; Ott, M. A. In Silico Prediction of Pharmaceutical Degradation Pathways: A Benchmarking Study. Mol. Pharm. 2014, 11 (11), 4179–4188. (24) Tetko, I. V.; Gasteiger, J.; Todeschini, R.; Mauri, A.; Livingstone, D.; Ertl, P.; Palyulin, V. A.; Radchenko, E. V.; Zefirov, N. S.; Makarenko, A. S.; Tanchuk, V. Y.; Prokopenko, V. V. Virtual Computational Chemistry Laboratory – Design and Description. J. Comput. Aided. Mol. Des. 2005, 19 (6), 453–463. (25) Miller, T. H.; Gallidabino, M. D.; MacRae, J. I.; Hogstrand, C.; Bury, N. R.; Barron, L. P.; Snape, J. R.; Owen, S. F. Machine Learning for Environmental Toxicology: A Call for Integration and Innovation. Environ. Sci. Technol. 2018, 52 (22), 12953–12955. (26) Paglia, G.; Angel, P.; Williams, J. P.; Richardson, K.; Olivos, H. J.; Thompson, J. W.; Menikarachchi, L.; Lai, S.; Walsh, C.; Moseley, A.; Plumb, R. S.; Grant, D. F.; Palsson, B. O.; Langridge, J.; Geromanos, S.; Astarita, G. Ion Mobility-Derived Collision Cross Section As an Additional Measure for Lipid Fingerprinting and Identification. Anal. Chem. 2015, 87 (2), 1137–1144. (27) Fiebig, L.; Laux, R. A collision cross section and exact ion mass database of the formulation constituents polyethylene glycol 400 and polysorbate 80. Int. J. Ion Mobil. Spectrom. 2016, 19 (2–3), 131– 137. (28) Bade, R.; Bijlsma, L.; Sancho, J. V; Hernández, F. Critical evaluation of a simple retention time predictor based on LogKow as a complementary tool in the identification of emerging contaminants in water. Talanta 2015, 139, 143–149. (29) Merel, S.; Lege, S.; Yanez Heras, J. E.; Zwiener, C. Assessment of N-Oxide Formation during Wastewater Ozonation. Environ. Sci. Technol. 2017, 51 (1), 410–417. (30) Stanstrup, J.; Neumann, S.; Vrhovšek, U. PredRet: Prediction of Retention Time by Direct Mapping between Multiple Chromatographic Systems. Anal. Chem. 2015, 87 (18), 9421–9428. (31) Sumner, L. W.; Amberg, A.; Barrett, D.; Beale, M. H.; Berger, R.; Daykin, C. A.; Fan, T. W.-M.; Fiehn, O.; Goodacre, R.; Griffin, J. L.; Hankemeier, T.; Hardy, N.; Harnly, J.; Higashi, R.; Kopka, J.; Lane, A. N.; Lindon, J. C.; Marriott, P.; Nicholls, A. W.; Reily, M. D.; Thaden, J. J.; Viant, M. R. Proposed minimum reporting standards for chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI). Metabolomics 2007, 3 (3), 211–221. (32) Schymanski, E. L.; Jeon, J.; Gulde, R.; Fenner, K.; Ru, M.; Singer, H. P.; Hollender, J. Identifying Small Molecules via High Resolution Mass Spectrometry: Communicating Confidence. Environ. Sci. Technol. 2014, 48, 2097–2098. (33) de Ávila, M. B. R.; Faroni, L. R. A.; Heleno, F. F.; de Queiroz, M. E. L. R.; Costa, L. P. Ozone as degradation agent of pesticide residues in stored rice grains. J. Food Sci. Technol. 2017, 54 (12), 4092–4099. (34) Freitas, R. da S. de; Faroni, L. R. D. A.; Queiroz, M. E. L. R. de; Heleno, F. F.; Prates, L. H. F. Degradation kinetics of pirimiphos-methyl residues in maize grains exposed to ozone gas. J. Stored Prod. Res. 2017, 74, 1–5.

8


Page 9 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry (35) Savi, G. D.; Piacentini, K. C.; Bortolotto, T.; Scussel, V. M. Degradation of bifenthrin and pirimiphos-methyl residues in stored wheat grains (Triticum aestivum L.) by ozonation. Food Chem. 2016, 203, 246–251. (36) Mahugija, J. A. M.; Chibura, P. E.; Lugwisha, E. H. J. Residues of pesticides and metabolites in chicken kidney, liver and muscle samples from poultry farms in Dar es Salaam and Pwani, Tanzania. Chemosphere 2018, 193, 869–874.

(37) Herrmann, J. M.; Guillard, C.; Arguello, M.; Agüera, A.; Tejedor, A.; Piedra, L.; Fernández-Alba, A. Photocatalytic degradation of pesticide pirimiphos-methyl Determination of the reaction pathway and identification of intermediate products by various analytical methods. Catal. Today 1999, 54 (2–3), 353–367. (38) Arcury, T. A.; Grzywacz, J. G.; Barr, D. B.; Tapia, J.; Chen, H.; Quandt, S. A. Pesticide urinary metabolite levels of children in Eastern North Carolina farmworkers households. Environ. Health Perspect. 2007, 115 (8), 1254–1260.

9


A refined non-target workflow for the investigation of metabolites

Recommend Documents