HaloSeeker 1.0, a user-friendly software to highlight halogenated

Subscriber access provided by Macquarie University

Article

HaloSeeker 1.0, a user-friendly software to highlight halogenated chemicals in non-targeted high resolution mass spectrometry datasets Alexis Léon, Ronan Cariou, Sébastien Hutinet, Julie Hurel, Yann Guitton, Celine Tixier, Catherine Munschy, Jean-Phillippe Antignac, Gaud Dervilly-Pinel, and Bruno Le Bizec Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.8b05103 • Publication Date (Web): 13 Feb 2019 Downloaded from http://pubs.acs.org on February 15, 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 10 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

HaloSeeker 1.0, a user-friendly software to highlight halogenated chemicals in non-targeted high resolution mass spectrometry datasets Alexis Léon†,‡, Ronan Cariou*,†, Sébastien Hutinet†, Julie Hurel†, Yann Guitton†, Céline Tixier‡, Catherine Munschy‡, Jean-Philippe Antignac†, Gaud Dervilly-Pinel†, Bruno Le Bizec†. †LABERCA, ‡Ifremer,

Oniris, INRA, F-44307, Nantes, France. Laboratoire Biogéochimie des Contaminants Organiques, F-44311, Nantes, France.

ABSTRACT: In the present work, we address the issue of non-targeted screening of organohalogenated chemicals in complex matrices. A global strategy aiming to seek halogenated signatures in full scan high resolution mass spectrometry (HRMS) fingerprints was developed. The resulting all-in-one user-friendly application, HaloSeeker 1.0, was developed to promote the accessibility of associated in-house bioinformatics tools to a large audience. The ergonomic web user interface avoids any interactions with the coding component while allowing interactions with the data, including peak detection (features), deconvolution and comprehensive accompanying manual review for chemical formula assignment. HaloSeeker 1.0 was successfully applied to a marine sediment HRMS dataset acquired on a LC-HESI(-)-Orbitrap instrument (R=140,000@200). Among the 4532 detected features, 827 were paired and filtered in 165 polyhalogenated clusters. HaloSeeker was also compared to three similar tools and showed the best performances. HaloSeeker ability to filter and investigate halogenated signals was demonstrated and illustrated by a potential homologue series with C12HxBryClzO2 as a putative general formula.

Identifying environmental chemical contaminants in abiotic matrices, along the trophic chains and in human matrices, is a major scientific undertaking supporting risk assessment. In this context, highly specific and sensitive targeted analytical methods involving chromatography and mass spectrometry couplings are efficiently implemented to monitor a wide range of chemical hazards.1 However, targeted compounds represent only a small fraction of the known chemical universe, i.e. compounds referenced in the literature or in databases, as for instance, the Chemical Abstracts Service (CAS) Registry wherein 144 million organic and inorganic unique chemical substances are currently listed.2 This fraction becomes even smaller when compared to the possible chemical space3, which includes synthesized and not yet referenced chemicals.4 Indeed, the possible chemical space of small molecules, i.e., containing a maximum of 30 C, N, O and S atoms, has been estimated to include more than 1060 unique possible structures.5 Many more distinct structures are possible with the addition of other common atoms, such as P or halogens, or the increasing of the number of C, N, O and S atoms. Both examples show how limited our knowledge of the chemical universe is and, by extent, its potential interactions with the environment and human health. Contributing to the comprehensive description of the chemical universe requires approaches that can widen the scope of detected compounds.1,6 In the field of environmental science, three complementary approaches dedicated to chemical analyses are reported7: (i) targeted approaches which focus on “so-called” known knowns, (ii) suspect screening approaches that focus on known unknowns (i.e. compounds that have been, at least, referenced in the literature or/and in databases) and (iii)

non-targeted approaches which focus on unknown unknowns (i.e. compounds that have not been yet identified in the environment nor referenced in the literature or databases). Halogenated compounds, especially chlorinated and brominated ones, are a chemical class of particular concern. Indeed, some of anthropogenic origin like persistent organic pollutants8, together with emerging environmental contaminants such as medium-chain chlorinated paraffins or novel brominated, norbornene-based and organophosphorus flame retardants, exhibit harmful effects to the environment and human health.9-11 Others are naturally produced, especially by marine organisms, and may pose a risk to human health as well.12 Hence, a wide range of halogenated compounds together with their potential transformation products can be found in the environment and non-targeted approaches are needed to undertake a comprehensive description of this class of compounds. Investigating a wide range of compounds requires a compromise on selectivity while keeping high sensitivity. Modern high resolution mass spectrometry (HRMS) technologies allow us to reach such performances and have opened the way to full-scan non-targeted approaches.13 However, the challenge is then shifted to post-acquisition steps since the huge datasets generated require specific and complex bioinformatics tools for rapid and efficient review of the signals of interest. Four main post-processing steps emerged from generalized non-targeted workflows14: (i) peak detection, (ii) deconvolution (ion grouping), (iii) chemical formula assignment and (iv) structural elucidation (MS/MS and database search). Marine Halogenated Compound Analysis (MeHaloCoA)15, Nontarget16 and Dynamic Cluster Analysis

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

(DCA)17 are open-source tools designed to filter halogenated signals. However, these tools only partially cover the postprocessing steps, and need to be used alongside other complementary tools to enhance the automation. Moreover, DCA and MeHaloCoA do not allow global screening of halogenated signals in HRMS datasets, by limiting to monobromo/dichloro halogen distributions (DCA) or by being too dependent on the quality of the datasets (e.g. duplicate hits with MeHaloCoA);17 meanwhile, Nontarget remains complex to optimize (11 rules, 9 parameters), thus lowering its accessibility. Indeed, accessibility is a crucial element to disseminate the tools and workflows to a large audience. Hence, graphical user interfaces are usually implemented to avoid using command lines which otherwise require coding knowledge.18,19 Based on the specific mass difference between isotopes of both atoms, 35Cl/37Cl and 79Br/81Br, our research group previously developed a workflow to filter polyhalogenated signals from complex HRMS fingerprints.20 An in-house deconvolution script (“pairing script”) was developed in Visual Basic for Applications language in order to pair isotopologue ions belonging to the same isotopic clusters. Then, paired clusters (potential halogenated clusters) were manually investigated on a H/Cl-scale mass defect plot (MD plot)21 displayed in an Excel® file (Microsoft Office) and chemical formulae were generated using the Xcalibur software (Thermo Fisher Scientific). However, the manual review remained time consuming and more automation was thus needed. The present paper aims at presenting HaloSeeker 1.0, an allin-one portable user-friendly application developed under the open access R environment. It allows processing and interpreting complex HRMS datasets in order to filter halogenated signals and assign chemical formulae, with a focus on known unknowns and unknown unknowns. HaloSeeker was developed following three main guidelines as follows. First, developing a user-friendly interface that does not require any manual installation nor any modifications directly into the coding component. Second, securing the data and users’ actions (processing parameters, formula assignment …) into an inhouse database, which is part of the application itself. The fact that all modules are interconnected through the database avoids manual manipulations of the data, which may otherwise lead to human errors while exploiting such large datasets. Third, drastically decreasing the time needed for data interpretation via graphical interactivity and processes such as dereplication while allowing the user to interact with the data at every step of the workflow (checking results of each operation and information related to the data). As a proof of concept and for illustration purposes, the whole analytical workflow has been applied to the characterization of a marine sediment sample, an integrative matrix often used to assess contamination in environmental science. The data set has been explored with HaloSeeker 1.0 and several similar open-source applications, such as MeHaloCoA, Nontarget, DCA (deconvolution) and Rdisop22 (formula decomposition), the results of which were then compared to those of HaloSeeker.

EXPERIMENTAL SECTION HaloSeeker 1.0 workflow. The controller part of the back end is processed in the open-source R-environment. The application is composed of 5 main modules described below: conversion, peak-picking (peak detection), pairing

Page 2 of 10

(deconvolution), formula assignment and dereplication (Figure 1). Results from each step are secured in a SQLite database. The front end involves a unique graphical user interface (GUI), operated with the Shiny R package v1.0.5 and displayed in a web browser on the local host. It includes each step of the workflow, from the files import to the Excel export (Figure S1). HaloSeeker was successfully used with Firefox v64.0 (Mozilla) and Chrome v71.0.3578.98 (Google) browsers. The application was processed on a Windows 7 64 bits PC with 8 GB RAM and 3.2 GHz i5 CPU.

Figure 1. HaloSeeker 1.0 workflow.

Conversion. The open access MSConvert23 software (proteowizard version 3.0.9810) was embedded in HaloSeeker to convert the proprietary raw data into the open format mzXML prior to data post-processing. Although MSConvert manages most manufacturer formats, users also have the possibility to import mzXML files directly into HaloSeeker. Peak-picking. Tables containing integrated features characterized by m/z, retention time (tR) and intensity are generated with the centWave algorithm from the xcms package (version 3.2.0).24 The algorithm is based on the detection of “regions of interest” which are characterized by similar m/z in consecutive MS scans. The xcms parameters can be defined by the user. By default, these parameters are set at optimized values for our specific chromatographic conditions (LC system, instrument background noise and accuracy) as follows: ppm = 3, peak width = 5-60, snthresh = 10, prefilter step = 3, prefilter level = 10,000, mzdiff = 0.001. Pairing. Halogenated compounds, due to the occurrence of two natural and stable isomers for both bromine and chlorine with distinct isotopic ratios, lead to specific isotopic patterns in MS spectra. Applying our previously-developed tool for pairing halogenated isotopologue ions20, the exact mass difference between each isotope was used to pair features belonging to the same halogenated cluster. The Visual Basic for Application script was translated into R language and embedded in HaloSeeker. In addition, the script was slightly modified, calculating mass difference from the base peak (A) for all isotopologues (Figure S-2). The script allows discrimination of “paired clusters” (at least an A+2 isotopologue matching the mass difference) from “non-paired clusters” (single peak or


Page 3 of 10 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


A/A+1 clusters). Pairing parameters were optimized as follows: tR tolerance = 1 s, m/z tolerance = 0.5 mDa. Interactive H/Cl-scale MD plot. The H/Cl-scale MD plots, according to Taguchi et al.21, are displayed on the HaloSeeker’s GUI. Briefly, each feature is characterized by its exact mass on the x-axis and the related fractional part on the y-axis. Display options are available to the users to narrow down their research. They include (i) a retention time window, (ii) an intensity threshold and/or a number of most intense clusters, (iii) isotopic filters, (iv) cluster slope filters and (v) cluster assignment status filters. Among the isotopic filters, F0 filter displays all peak-picked features and F1 filter the m/z–paired clusters (containing at least one A+2 isotopologue). F2 and F2+ filters display clusters complying to additional rules to F1, related to ion ratios in order to improve filtering of halogenated and polyhalogenated clusters, respectively (Figure S-3). Indeed, based on theoretical considerations (Table S-1, Figure S-4), isotopologue areas relative to base peak (A) rules are, for F2 filter: 

A-2 ratio = 0 (absence) AND A+2 ratio ≥ 25% OR;



A-2 ratio ≥ 60% AND A+2 ratio ≥ 20% OR;

 A-2 ratio ≥ 27% AND A+2 ratio ≥ 36%. For F2+ filter, F2 clusters suspected to be related to monohalogenated ions are removed according to the following rules: 

A-2 and A+4 ratios = 0 (absence) AND;

 A+2 ratio ∈ [25-39] ∪ [77-117]%. Cluster slope filters specifically display clusters according to the slope of the linear regression curve of paired isotopologues plots (A±n, n being even). Indeed, in H/Cl-scale MD plots, the theoretical slope is -3.3×10-4 and +1.2×10-4, with a correlation coefficient of 1 for (poly)chlorinated and (poly)brominated clusters respectively, whereas mixed clusters exhibit intermediary slopes with lower correlation coefficients. Last, a status filter displays “all”, “all but discarded”, “assigned” or “non-assigned clusters”. Formula assignment. The MD plot graphic is interactive through the GUI. A click on a cluster displays a pop-up window dedicated to formula assignment (Figure S-5). It includes the mass spectrum of the ad-hoc cluster, the extracted ion chromatogram of each isotopologue, a table containing features characteristics and information about the cluster slope. This information supports the chemical formula assignment, chemical formulae being generated by an in-house script adapted from the Rdisop package22 to consider either the monoisotopic or the base peaks. To reduce the number of hits, different filters are also implemented. They include a mass deviation parameter, corresponding to the mean mass deviation of each isotopologue of a cluster, a Double Bond Equivalent (DBE) cut-off (exclusion if lower than -1) and the adapted Seven Golden Rules25. The in-house scoring system compares the observed isotopic pattern with the theoretical one obtained from the R package enviPat26, and applies a matched score. Expertise of the user is then required to select manually the best hit. A cluster considered erroneous or not of interest can be flagged as discarded. Dereplication. Optionally, the dereplication module compares a selected dataset to the in-house database, populated along the software utilizations by the user. It seeks suitable matches between paired clusters from the dataset and chemical

formulae previously assigned in other datasets or to compounds manually included in the database. The module uses mass deviation and the scoring system described previously to match clusters. Users must manually accept each dereplicated cluster. Availability. HaloSeeker 1.0 is freely available on request at [email protected], under the GPLv3 license. The executable file is about 220 MB and runs out of the box without need for computer skills or administrative rights. Compared bioinformatics tools. Other bioinformatics tools were compared to HaloSeeker 1.0 pairing script or its formula decomposition tool. Tools and their version used are listed below: 

Nontarget v1.9 (deconvolution)



DCA v1.09 (deconvolution)



MeHaloCoA v0.99 (deconvolution)

 Rdisop v1.37.1 (formula decomposition) Chemicals. Dichloromethane and n-hexane were supplied by Promochem (trace analysis grade, Wesel, Germany). Water and acetonitrile were purchased from Sigma-Aldrich (LC-MS ChromaSolv grade, St. Louis, MO, USA). Concentrated sulfuric acid (96%), hydrochloric acid (30%), anhydrous sodium sulfate and ammonium acetate were provided by Merck (Emsure grade, Darmstadt, Germany). Copper (fine wires, 4 × 0.5 mm) was obtained from Elemental Microanalysis (Okehampton, United Kingdom). Labelled hexabromocyclododecane isomers (2H18-α- and γHBCDD), tetrabromobisphenol A (13C12-TBBP-A), mixtures of 19 chlorophenols and 19 bromophenols, and 5-chloro-2-(2,4dichlorophenoxy)phenol (triclosan) were purchased from Wellington Laboratories (Guelph, Ontario, Canada). 2'-OH2,4,4'-tribromodiphenyl ether, 6-OH-2,2',3,4,4'pentabromodiphenyl ether , 6-OH-2,2',3,4,4',5hexabromodiphenyl ether and 6-OH-3,5-dichloro-2,2',4,4'tetrabromodiphenyl ether were provided by AccuStandard (New Haven, CT, USA). Preparation of the standard mix. A solution containing all the aforementioned individual standards as well as the halogenophenol mixtures (n=46) was prepared at a concentration of 0,04 ng/µL for each standard, except for ²H18α- and ²H18-γ-HBCDD prepared at 0,05 ng/µL. Sample preparation. A sediment sample was collected in 2002 in the river Seine estuary (France), in proximity to highly industrialized and urbanized sites. The sample was extracted from 5 g lyophilized matter by pressurized liquid extraction (ASE, Dionex, CA, USA) pending two successive cycles with dichloromethane at 100 °C and 100 bar. The organic extract was concentrated to 3 mL with a rotary evaporator and transferred into n-hexane (6 mL) by evaporating dichloromethane under a gentle stream of nitrogen. It was then treated by hydrochloric acid-activated copper to remove sulfurcontaining compounds. A procedural blank was processed similarly. The purified extract was spiked with 10 ng of ²H18-γHBCDD in toluene as external standard, evaporated to dryness under a gentle stream of nitrogen, reconstituted in a mixture of water/acetonitrile 2:8 (v/v, 100 µL) and centrifuged prior to injection (10 µL). LC-HRMS data acquisition. The extract was analyzed with an UltiMate 3000 UHPLC pumping system coupled to an Orbitrap Q-Exactive mass spectrometer equipped with a heated ESI source (Thermo Fischer Scientific, San José, CA, USA).



Page 4 of 10

ion currents), when the organic content of the mobile phase was higher than 70%. This was consistent with the extraction method, which favored relatively lipophilic compounds. Signals observed at 0.72 min and 21.68 min were attributed to the front solvent and column reconditioning, respectively. Thus, the 1-20 min retention time window was selected for further investigations. Table 1 reports the enumeration of features and clusters, obtained after peak picking and pairing processes within 3.7 and 1.5 min, respectively, for the 4 above-described filters (F0 to F2+). The intensity threshold was set at 2×106 AU to remove lower intensity clusters which can show unreliable isotopic patterns. A clear difference in the number of clusters and the cumulated intensities appeared between the S extract and the procedural blank (figures in brackets), suggesting that most of the signals arose from the sample. The peak picking process (F0) led to several thousands of features, overloading the MD plot and hampering an efficient cluster investigation (see Figures S-9,10). The m/z pairing script (F1) drastically reduced the number of potential signals of interest, which were then further refined by adding isotopic ratio rules (F2 and F2+, Figures S-11,12). Both filters are expected to remove potential “false positives” such as silicon and sulfur-containing compounds, as well as irrelevantly peak-picked and/or paired features. Indeed, the mass differences between 28Si and 30Si (1.99684) and 32S and 34S (1.99580), being close to the Br and Cl ones, could overlap with the range of mass scanned by the pairing script, which strongly depends on the mass tolerance parameter and the instrument precision. These filters led to about 200 clusters of interest, which made manual investigation a realistic option. The difference in enumeration of clusters between F2 and F2+ suggested that the sample may contain a significant proportion of monohalogenated clusters (21%), accounting for 10% of the cumulated intensities. Part of the monohalogenated clusters could arise from chlorine adducts, as observed for HBCDD standard, thus increasing “false positive” hits. Thus, we decided to focus on polyhalogenated signals by using the F2+ filter.

The instrument was controlled with Chromeleon Xpress and Xcalibur softwares (Thermo Fischer Scientific). Chromatographic separation was performed on a Hypersil Gold analytical column (100 mm × 2.1 mm, 1.9 µm, Thermo Fischer Scientific) kept at 45 °C. A mobile phase consisting of water (A) and acetonitrile (B), both containing 10 mM ammonium acetate, was used. The gradient started with A/B 80:20 (v/v) for 2 min, was then increased linearly to 20:80 within 5.5 min, and ramped linearly to 100% B over 6.5 min to be further maintained for 6 min. It was then returned to the initial conditions within 2 min and held for 4 min for a total run time of 26 min. The flow rate was set at 0.4 mL/min. Data were recorded in negative mode with HESI parameters as follows: sheath gas flow, 50 arbitrary unit (AU); auxiliary gas flow, 5 AU; capillary temperature, 350 °C; source temperature, 150 °C; spray voltage, 3 kV; s-lens radio frequency, 50 AU. HRMS data were acquired in full scan mode within the m/z range 120-1,000 at a resolving power of 140,000 at m/z 200, using m/z 305.02307 ([CH3COO.(NaCH3CO2)3]-) as lock mass. The automatic gain control (AGC Target) was set at 5×105 and the maximum injection time was set at 250 ms.

RESULTS AND DISCUSSION Quality control. Labelled 2H18-γ-HBCDD was used as external standard to assess the mass deviation (mDa) of the instrument over the 3 analyses performed (sample, blank and standard mix). Observed values ranged from +0.14 to +0.19 mDa. Such low mass deviation drastically narrows the list of chemical formulae generated and is decisive for final formula assignment. The standard mix covered most of the m/z range [120-1000 m/z] of interest (see Figures S-6,7), and showed mass deviation ranging from +0.02 to +0.31 (0.11 ± 0.07, Table S-2). Consequently, a tolerated value of ± 1 mDa was set for the formula generator tool. Overview of datasets. We illustrate the operation and performances of HaloSeeker based on 3 datasets, the S extract, its corresponding procedural blank and the Std injection. Most of the specific compounds of the sediment extract were eluted within the range of 7 to 17 min (see Figure S-8 for total

Table 1. Enumeration of features, paired clusters and cumulated intensities (×109 AU) according to m/z pairing and ion ratio filters for the sediment sample obtained via HaloSeeker. tR range 1-20 min, 2×106 AU intensity threshold. Figures in brackets are for the corresponding procedural blank. Filter

F0

F1

F2

F2+

Features

4532 (1208)

1066 (135)

961 (93)

827 (61)

Clusters

-

245 (32)

211 (17)

165 (9)

Cumulated intensity

131 (31)

6.5 (0.4)

4.2 (0.3)

3.8 (0.3)

Table 2. Comparison between HaloSeeker, Nontarget, MeHaloCoA and DCA, four tools able to highlight halogenated clusters. For each dataset and for each tool, the number of paired clusters (out of 34 for Std and out of 64 for S datasets) and the percentage of clusters exceeding 60% of matching score are recorded. Datasets

Standard mix (n=34) (Std)

Sediment (n=64) (S)

Pairing tool

HaloSeeker

Nontarget

MeHaloCo A

DCA

HaloSeeker

Nontarget

MeHaloCo A

DCA

Paired ion clusters

34

32

24

8

64

64

60

17

>60% of score (%)

94.1

88.2

67.6

2.94

100

100

71.9

15.6


Page 5 of 10 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


Comparison of pairing tools. The pairing script of HaloSeeker was compared to three other open-source tools: Nontarget16, MeHaloCoA15 and DCA17. Two source peak lists obtained via xcms, corresponding to the sediment sample (S) and the standard mix (Std), were used to perform the comparison. It should be noted that only the ability to highlight halogenated clusters was considered, though the tools provide other interesting features. Parameters of each tool were optimized in order to find as many clusters as possible (Table S-3). The capability to retrieve standard clusters from the Std dataset (Table S-4) and clusters from the S dataset (Table S-5) was investigated using each tool. A maximum of 30 compounds out of the 46 injected standards were retrieved in the Std dataset by xcms. The 16 remaining compounds, corresponding to halogenophenols (7 bromophenols and 9 chlorophenols), were not detected by the peak picking step (therefore not present in the Std peak list) likely due to isomer coelutions. Four adducts of ²H18-HBCDD (2 for α and 2 for γ) were also found, leading to a maximum of 34 different ion clusters to be paired. For the S dataset, 64 ions exhibiting a polyhalogenated fingerprint for which a chemical formula was putatively assigned via HaloSeeker were used as a list of paired clusters to be processed by the other tools. It should be noted that the rest of the S dataset was not investigated by the other tools, although they could have retrieved clusters that HaloSeeker did not. For each dataset and each tool, Table 2 shows the number of clusters successfully paired (after manual verifications), alongside the percentage of clusters exceeding 60% of the matching score. Nontarget algorithm paired features according to the mass differences between isotopes of the chosen elements (C, Cl and Br in our case) and to a set of rules and parameters. HaloSeeker and Nontarget showed the most efficient and very similar results, with 34 out of 34 clusters and 32 out of 34 clusters retrieved in the Std dataset, respectively, and 64 out of 64 clusters retrieved in the S dataset for both tools. The only noticeable difference, in the Std dataset, was that Nontarget was not able to pair two ions corresponding to two isomers of dibromophenol, despite parameters optimization attempts. Aside performances, Nontarget appeared less user-friendly than HaloSeeker as it proposes a set of 11 rules and 9 parameters to optimize versus only 2 for HaloSeeker. In addition, Nontarget requires the R console as its user interface. The user-friendly enviMass workflow27, available as a web page, integrates Nontarget but does not yet allow optimization of each parameter, which includes deactivating rule n°7 (clustering stops if 13C isotopologue is not found). In our case, applying this rule would have missed the pairing of some “A+even” isotopologues since some 13C isotopologues (“A+odd”) were not intense enough to be integrated by xcms, degrading the completion of the pairing. MeHaloCoA pairing process relies on two steps. First, the embedded CAMERA package28 groups ions exhibiting similar chromatographic peak shapes and retention times (isotopologues, fragments, adducts) in a so-called “PCgroup” (Peak Cluster group). Second, mass differences and intensity ratios are checked within each PCgroup and features potentially tagged as Cl or Br containing ions. MeHaloCoA was able to retrieve 24 out of 34 clusters and 60 out of 64 clusters in the Std and S datasets respectively. The fact that MeHaloCoA retrieved fewer clusters and less complete clusters (67,6% and 71,9% of matched score >60% for Std and S dataset respectively) was due to the wrong assignment of isotopologues in different PCgroups. CAMERA deconvolution was likely the main cause

as the peak shape could be too different from one isotopologue to another, thus pairing them in different PCgroups. One possible reason could be the relatively low acquisition rate of the Orbitrap mass analyzer which leads to fewer points per peak (scans), altering peak shape, especially for lower intensity peaks. This could also have an impact on peak integration and lead to altered isotopic patterns lowering the calculated score. DCA pairing script was empirically developed in relation to two specific databases (Marinlit and Antibase). Rules of thumb were determined in order to extract halogenated signatures from complex marine microalgae samples. DCA showed less efficient results with only 8 out of 34 clusters retrieved in the Std dataset and 17 out of 64 clusters from the S dataset. Indeed, compounds not present in at least one of the two databases that were used to design DCA may not necessarily pass all the rules of thumb. In particular, one rule, i.e. intensity ratio of [M+n+1]/[M+n] 106 AU). Figure 2 presents a zoomin of the area of interest. Mass differences corresponded to substitution vectors indicating mixed polyhalogenated series (–H/+Cl, -H/+Br and –Cl/+Br). Cluster slopes were also useful to support data interpretation, acting as an additional filter. Indeed, within the H/Cl-scale, there was a slightly positive mass defect between the fractional parts of 81Br (9.14×10-3) and 79Br (8.90×10-3), leading to a slightly positive slope between two consecutive dots of a brominated cluster on the H/Cl-scale MD plot (green circle on Figure 2). Conversely, the mass defect between 37Cl (8.32×10-3) and 35Cl (8.98×10-3) led to a negative slope for chlorinated clusters (blue circle). For mixed halogenated ions (Cl and Br), below a spectrometric resolution of several hundred thousands, it was not possible to discriminate isotopologues showing identical numbers of nucleons. Thus, m/z of centroided peaks resulted from linear combinations of isotopologues, leading to curved plot alignments with intermediate linear regression slopes (orange circles). Most of the clusters (n = 15) were assigned to chlorinated ions, while 4 clusters were assigned to brominated ions and 2 clusters to mixed halogenated ions. Both C12H6Cl3O2 and C12H4Br5O2 were confirmed by standard injections (Figure 3) as triclosan (5-chloro-2-(2,4-dichlorophenoxy)phenol) and 3,4,5tribromo-2-(2,4-dibromophenoxy)phenol (OH-Br5-diphenyl ether). By extension, we suggest that other clusters may be assimilated to potentially brominated, chlorinated or mixed hydroxylated diphenyl ethers. The general formula C12HxBryClzO2 (x ∈ ⟦0-7⟧, y ∈ ⟦0-5⟧ and z ∈ ⟦2-9⟧) was proposed. The presence of hydroxybrominated diphenyl ethers (HOBDEs) is likely due to their natural production by marine species.33 Their occurrence in the marine environment (algae, mussels and sediment) has indeed been demonstrated by several studies.33,34 Interestingly, triclosan was by far the most intense cluster (8.5 times the second most intense one of the series). Despite being phased out in health care antiseptic products in the US35 (including soaps and body washes) and in type-1

H/Cl-scale m/z fractional part

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

Triclosan

Page 6 of 10

product (disinfectants) in the EU36, triclosan is still widely used and several studies have demonstrated its widespread presence in the marine environment (seawater and sediment).37,38 Table 3. Series of paired clusters (n=21) with a C12HxBryClzO2 assigned formula. *Confirmed with standard injection. Ion formula

m/z (Observed base peak)

tR (min)

Cluster intensity (×106 AU)

Deviation (mDa)

C12H7Cl2O2

252.98283

6.99

8.0

0.03

C12H6Cl3O2*

286.94396

7.33

355.0

0.03

C12H5Cl4O2

322.90193

7.94

4.6

0.03

C12H5Cl4O2

322.90194

7.54

7.7

0.03

C12H5Cl4O2

322.90195

7.74

19.3

0.02

C12H7Br2O2

342.87975

7.23

35.7

0.04

C12H4Cl5O2

356.86310

7.81

22.2

0.14

C12H5BrCl3O2

366.85186

7.81

41.3

0.04

C12H3Cl6O2

390.82409

8.18

11.0

0.07

C12H2Cl7O2

424.78516

6.96

5.2

0.10

C12H2Cl7O2

424.78520

8.52

26.6

0.10

C12H4Br2Cl3O2

446.76009

7.92

4.1

0.05

C12HCl8O2

460.74325

8.44

5.2

0.11

C12HCl8O2

460.74325

7.23

22.1

0.11

C12HCl8O2

460.74326

7.77

17.9

0.13

C12HCl8O2

460.74327

8.99

5.7

0.15

C12Cl9O2

494.70414

7.57

11.9

0.02

C12Cl9O2

494.70415

7.92

28.9

0.04

C12H5Br4O2

500.69883

8.04

19.4

0.05

C12H5Br4O2

500.69883

6.42

1.2

0.10

C12H4Br5O2*

580.60710

8.19

3.6

0.07

Mixed halogenated

Hydroxypentabromodiphenyl ether

H/Cl-scale m/z

Figure 2. PNG HaloSeeker export of H/Cl-scale MD plot zoom-in (Sediment dataset), showing the C12HxBryClzO2 series (red rectangle). tR range 1-20 min, filter F2+, 2×106 AU intensity threshold. Colored dots: paired clusters; Light blue, red and green arrows: specific vectors of mixed halogenated homologue series, respectively substitution of 35Cl by 79Br, 1H by 35Cl and 1H by 79Br. Triclosan and OH-Br5-diphenyl ether were confirmed against analytical standards.


Page 7 of 10 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


Besides triclosan, 15 clusters were assigned to potentially other hydroxylated chlorinated diphenyl ethers (HO-CDEs) ranging from di- to nonachlorinated. As HO-CDE standards are not commercially available, it was not possible to confirm their identity. Information about HO-CDEs, as well as their parent compounds (polychlorinated diphenyl ethers - PCDEs)39, are rather scarce in the literature, despite the fact that PCDE levels in environmental samples have been found to be similar to regulated contaminants such as PCBs.40

polyhalogenated clusters filtered, 64 chemical formulae were putatively assigned. The H/Cl-scale MD plot highlighted a series of 21 homologue clusters, with a putative general formula of C12HxBryClzO2. Two compounds were confirmed by standard injections, i.e. triclosan and 3,4,5-tribromo-2-(2,4dibromophenoxy)phenol. Increasing the degree of confidence to potentially link identified halogenated signals to chemical structures requires complementary experiments such as standard injections, MS fragmentation, derivatization or NMR analysis. Finally, we demonstrated the effectiveness of HaloSeeker 1.0 for screening halogenated chemicals in the field of environmental sciences. Its suitability with mass analyzers other than Orbitrap should be further investigated, particularly as regards mass precision. Future applications can be extended to any abiotic or biotic matrix for occurrence or (bio)transformation studies41.

ASSOCIATED CONTENT Supporting Information The Supporting Information is available free of charge on the ACS Publications website.

Figure 3. PNG HaloSeeker export of extracted ion chromatograms (LC-HESI(-)-HRMS) of triclosan (m/z 286.94396; ΔmDa = 0.5; left) and OH-Br5-diphenyl ether (m/z 580.60710; ΔmDa = 0.5; right) in both standard (upper) and sediment sample (bottom).

CONCLUSIONS AND PERSPECTIVES HaloSeeker is a new interactive and easy to use open-source software, specifically developed to process, filter, visualize and identify chlorinated and brominated compounds in full scan HRMS fingerprints. HaloSeeker relies on the manual review of the data, which remains time-consuming, assisted by a GUI that allows the analytical expert to discard any false positives and narrow down the potential chemical formula hits preventively for more accurate results. The deconvolution pairing script was compared to three similar tools and showed the best performances while only requiring two parameters to be optimized, thus facilitating its accessibility. The chemical formula decomposition tool is capable of predicting the right formula from either the suspected monoisotopic peak or the base peak isotopologues, which is crucial for low intensity polyhalogenated signals. The user-friendly web interface gathers all tools in a single entity so that no extensive knowledge in computing is required, encouraging its dissemination to the scientific community. Future directions to upgrade HaloSeeker to v2.0 include implementation of an alignment module to process several datasets at once. This will also open the way to blank subtraction in order to focus on signals originating only from the samples. Another addition will consist in implementing other annotation tools, to further enhance the automation, such as tools which pair signals exhibiting similar chromatographic peak shapes to facilitate the annotation of adducts and fragments. As a proof of concept, a marine sediment sample was investigated through the HaloSeeker workflow. From the 165

Graphical User Interface (Figures S-1,5); Pairing process (Figure S-2); Graphical filters (Figure S-3); Ion ratio considerations (Table S-1, Figure S-4); Total Ion Currents (Figure S-7); H/Cl scale MD plots of standards, procedural blank and sediment (Figures S-6,7,912); List of clusters filtered found in the standard and the sediment extract (Tables S-2,7); Parameters used for the comparisons of the deconvolution and chemical decomposition tools (Tables S-3,6); Comparison of the deconvolution tools for the standard mix and the sediment sample (Tables S-4,5).

AUTHOR INFORMATION Corresponding Author * Laboratoire d’Étude des Résidus et Contaminants dans les Aliments (LABERCA), Oniris, Route de Gachet, Nantes, F44307, France. E-mail address: [email protected]

Author Contributions All authors have approved the final version of the manuscript.

Notes The authors declare no competing financial interest.

ACKNOWLEDGMENTS The authors want to express their acknowledgments to the French General Directorate for Food and the European Union’s Horizon 2020 research and innovation programme HBM4EU under grant agreement No 733032 for their financial support. AL received a PhD grant from the “Région Pays de la Loire, France” and Ifremer, under grant agreement n°16/5210665.

REFERENCES (1) Brack, W.; Ait-Aissa, S.; Burgess, R. M.; Busch, W.; Creusot, N.; Di Paolo, C.; Escher, B. I.; Mark Hewitt, L.; Hilscherova, K.; Hollender, J.; Hollert, H.; Jonker, W.; Kool, J.; Lamoree, M.; Muschket, M.; Neumann, S.; Rostkowski, P.; Ruttkies, C.; Schollee, J.; Schymanski, E. L., et al. Effect-directed analysis supporting monitoring of aquatic environments--An in-depth overview. Sci. Total Environ. 2016, 544, 1073-1118.

7 ACS Paragon Plus Environment


(2) CAS 2018, Chemical Abstract Service. https://www.cas.org/support/documentation/chemical-substances (accessed Jan. 4, 2019), (3) Dobson, C. M. Chemical space and biology. Nature 2004, 432, 824-828. (4) Daughton, C. G. Non-regulated water contaminants: emerging research. Environ. Impact Assess. Rev. 2004, 24, 711-732. (5) Bohacek, R. S.; McMartin, C.; Guida, W. C. The art and practice of structure-based drug design: a molecular modeling perspective. Med. Res. Rev. 1996, 16, 3-50. (6) Andra, S. S.; Austin, C.; Patel, D.; Dolios, G.; Awawda, M.; Arora, M. Trends in the application of high-resolution mass spectrometry for human biomonitoring: An analytical primer to studying the environmental chemical space of the human exposome. Environ. Int. 2017, 100, 32-61. (7) Krauss, M.; Singer, H.; Hollender, J. LC-high resolution MS in environmental analysis: from target screening to the identification of unknowns. Anal. Bioanal. Chem. 2010, 397, 943-951. (8) UNEP, Final Act of the Conference of Plenipotentiaries on the Stockholm Convention on Persistent Organic Pollutants. United Nations Environment Programme, Stockholm, Sweden, 2001. (9) Ballesteros-Gómez, A.; de Boer, J.; Leonards, P. E. Novel analytical methods for flame retardants and plasticizers based on gas chromatography, comprehensive two-dimensional gas chromatography, and direct probe coupled to atmospheric pressure chemical ionization-high resolution time-of-flight-mass spectrometry. Anal. Chem. 2013, 85, 9572-9580. (10) Wang, P.; Zhang, Q.; Zhang, H.; Wang, T.; Sun, H.; Zheng, S.; Li, Y.; Liang, Y.; Jiang, G. Sources and environmental behaviors of Dechlorane Plus and related compounds - A review. Environ. Int. 2016, 88, 206-220. (11) Schinkel, L.; Lehner, S.; Heeb, N. V.; Marchand, P.; Cariou, R.; McNeill, K.; Bogdal, C. Dealing with strong mass interferences of chlorinated paraffins and their transformation products: An analytical guide. Trends Analyt. Chem. 2018, 106, 116-124. (12) Gribble, G. W. Naturally occuring organohalogen compounds - A comprehensive update; Springer: Vienna, 2010; Vol. 91. (13) Hernández, F.; Sancho, J. V.; Ibáñez, M.; Abad, E.; Portolés, T.; Mattioli, L. Current use of high-resolution mass spectrometry in the environmental sciences. Anal. Bioanal. Chem. 2012, 403, 1251-1264. (14) Knolhoff, A. M.; Croley, T. R. Non-targeted screening approaches for contaminants and adulterants in food using liquid chromatography hyphenated to high resolution mass spectrometry. J. Chromatogr. A 2016, 1428, 86-96. (15) Roullier, C.; Guitton, Y.; Valery, M.; Amand, S.; Prado, S.; Robiou du Pont, T.; Grovel, O.; Pouchus, Y. F. Automated detection of natural halogenated compounds from LC-MS profiles-Application to the isolation of bioactive chlorinated compounds from marine-derived fungi. Anal. Chem. 2016, 88, 9143-9150. (16) Loos, M.; Hollender, J.; Schymanski, E.; Ruff, M.; Singer, H. Bottom-up peak grouping for unknown identification from highresolution mass spectrometry data. ASMS 2012, Vancouver 2012. (17) Andersen, A. J.; Hansen, P. J.; Jørgensen, K.; Nielsen, K. F. Dynamic Cluster Analysis: An Unbiased Method for Identifying A + 2 Element Containing Compounds in Liquid Chromatographic HighResolution Time-of-Flight Mass Spectrometric Data. Anal. Chem. 2016, 88, 12461-12469. (18) Misra, B. B.; van der Hooft, J. J. Updates in metabolomics tools and resources: 2014-2015. Electrophoresis 2016, 37, 86-110. (19) Spicer, R.; Salek, R. M.; Moreno, P.; Cañueto, D.; Steinbeck, C. Navigating freely-available software tools for metabolomics analysis. Metabolomics 2017, 13, 106. (20) Cariou, R.; Omer, E.; Léon, A.; Dervilly-Pinel, G.; Le Bizec, B. Screening halogenated environmental contaminants in biota based on isotopic pattern and mass defect provided by high resolution mass spectrometry profiling. Anal. Chim. Acta 2016, 936, 130-138. (21) Taguchi, V. Y.; Nieckarz, R. J.; Clement, R. E.; Krolik, S.; Williams, R. Dioxin analysis by gas chromatography-Fourier transform

Page 8 of 10

ion cyclotron resonance mass spectrometry (GC-FTICRMS). J. Am. Soc. Mass Spectrom. 2010, 21, 1918-1921. (22) Böcker, S.; Lipták, Z. A Fast and Simple Algorithm for the Money Changing Problem. Algorithmica 2007, 48, 413-432. (23) Chambers, M. C.; Maclean, B.; Burke, R.; Amodei, D.; Ruderman, D. L.; Neumann, S.; Gatto, L.; Fischer, B.; Pratt, B.; Egertson, J.; Hoff, K.; Kessner, D.; Tasman, N.; Shulman, N.; Frewen, B.; Baker, T. A.; Brusniak, M. Y.; Paulse, C.; Creasy, D.; Flashner, L., et al. A cross-platform toolkit for mass spectrometry and proteomics. Nat. Biotechnol. 2012, 30, 918-920. (24) Tautenhahn, R.; Böttcher, C.; Neumann, S. Highly sensitive feature detection for high resolution LC/MS. BMC Bioinformatics 2008, 9, 504. (25) Kind, T.; Fiehn, O. Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry. BMC Bioinformatics 2007, 8, 105. (26) Loos, M.; Gerber, C.; Corona, F.; Hollender, J.; Singer, H. Accelerated isotope fine structure calculation using pruned transition trees. Anal. Chem. 2015, 87, 5738-5744. (27) Loos, M. enviMass version 3.5 LC-HRMS trend detection workflow R package. Zenodo https://doi.org/10.5281/zenodo.1213098, 2018. (28) Kuhl, C.; Tautenhahn, R.; Böttcher, C.; Larson, T. R.; Neumann, S. CAMERA: An integrated strategy for compound spectra extraction and annotation of liquid chromatography/mass spectrometry data sets. Anal. Chem. 2012, 84, 283-289. (29) Böcker, S.; Letzel, M. C.; Lipták, Z.; Pervukhin, A. SIRIUS: decomposing isotope patterns for metabolite identification. Bioinformatics 2009, 25, 218-224. (30) Law, R. J.; Allchin, C. R.; de Boer, J.; Covaci, A.; Herzke, D.; Lepom, P.; Morris, S.; Tronczynski, J.; de Wit, C. A. Levels and trends of brominated flame retardants in the European environment. Chemosphere 2006, 64, 187-208. (31) Reineke, N.; Biselli, S.; Franke, S.; Francke, W.; Heinzel, N.; Hühnerfuss, H.; Iznaguen, H.; Kammann, U.; Theobald, N.; Vobach, M.; Wosniok, W. Brominated indoles and phenols in marine sediment and water extracts from the north and baltic seas-concentrations and effects. Arch. Environ. Contam. Toxicol. 2006, 51, 186-196. (32) Granmo, A.; Ekelund, R.; Berggren, M.; Brorström-Lundén, E.; Bergvist, P.-A. Temporal Trend of Organochlorine Marine Pollution Indicated by Concentrations in Mussels, Semipermeable Membrane Devices, and Sediment. Environ. Sci. Technol. 2000, 34, 3323-3329. (33) Malmvärn, A.; Marsh, G.; Kautsky, L.; Athanasiadou, M.; Bergman, A.; Asplund, L. Hydroxylated and Methoxylated Brominated Diphenyl Ethers in the Red Algae Ceranium tenuicorne and Blue Mussels from the Baltic Sea. Environ. Sci. Technol. 2005, 39, 29902997. (34) Fan, Y.; Lan, J.; Zhao, Z.; Zhao, M. Sedimentary records of hydroxylated and methoxylated polybrominated diphenyl ethers in the southern Yellow Sea. Mar. Pollut. Bull. 2014, 84, 366-372. (35) Safety and Effectiveness of Health Care Antiseptics; Topical Antimicrobial Drug Products for Over-the-Counter Human Use, Food and Drug Administration, 2017, 60474-60503. (36) Commission Implementing Decision (EU) 2016/110 of 27 January 2016 not approving triclosan as an existing active substance for use in biocidal products for product-type 1, Official Journal of the European Union, 2016, L21, 86-87. (37) Xie, Z.; Ebinghaus, R.; Flöser, G.; Caba, A.; Ruck, W. Occurrence and distribution of triclosan in the German Bight (North Sea). Environ. Pollut. 2008, 156, 1190-1195. (38) Chen, Z. F.; Wen, H. B.; Dai, X.; Yan, S. C.; Zhang, H.; Chen, Y. Y.; Du, Z.; Liu, G.; Cai, Z. Contamination and risk profiles of triclosan and triclocarban in sediments from a less urbanized region in China. J. Hazard Mater. 2018, 357, 376-383. (39) Yang, J. S.; Lin, S. L.; Lin, T. C.; Wu, Y. L.; Wang, L. C.; Chang-Chien, G. P. Emissions of polychlorinated diphenyl ethers from a municipal solid waste incinerator during the start-up operation. J. Hazard Mater. 2015, 299, 206-214.


Page 9 of 10 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


(40) Domingo, J. L. Polychlorinated diphenyl ethers (PCDEs): environmental levels, toxicity and human exposure: a review of the published literature. Environ. Int. 2006, 32, 121-127. (41) Briels, N.; Løseth, M. E.; Ciesielski, T. M.; Malarvannan, G.; Poma, G.; Kjærvik, S. A.; Léon, A.; Cariou, R.; Covaci, A.; Jaspers, V. L. B. In ovo transformation of two emerging flame retardants in Japanese quail (Coturnix japonica). Ecotoxicol. Environ. Saf. 2018, 149, 51-57.



Page 10 of 10

For TOC Only

Br Br

Br

Br

Br Cl

Br Cl

Cl


10

HaloSeeker 1.0, a user-friendly software to highlight halogenated

Recommend Documents