Untargeted Identification of Organo-Bromine Compoundsin Lake Sediments by Ultrahigh-Resolution Mass Spectrometrywith the Data-Independent Precursor Isolation and CharacteristicFragment MethodHui Peng,*,† Chunli Chen,† David M. V. Saunders,† Jianxian Sun,† Song Tang,‡ Garry Codling,†

Markus Hecker,†,‡ Steve Wiseman,† Paul D. Jones,†,‡ An Li,⊗ Karl J. Rockne,$ and John P. Giesy*,†,§,∥,⊥,#,○

†Toxicology Centre, University of Saskatchewan, 44 Campus Drive, Saskatoon, Saskatchewan Canada, S7N 5B3‡School of Environment and Sustainability, 117 Science Place, Saskatoon, Saskatchewan Canada, S7N 5C8§Department of Veterinary Biomedical Sciences, University of Saskatchewan, Saskatoon, Saskatchewan Canada S7N 5B3∥Zoology Department, Center for Integrative Toxicology, Michigan State University, East Lansing, Michigan 48824United States⊥School of Biological Sciences, University of Hong Kong, Hong Kong Special Administrative Region, Peoples Republic of China#State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing 210023,People’s Republic of China○Biology Department, Hong Kong Baptist University, Hong Kong, Special Administrative Region, China⊗School of Public Health, University of Illinois at Chicago, Chicago, Illinois 60612, United States$Department of Civil and Materials Engineering (MC 246), University of Illinois at Chicago, 842 West Taylor Street, Chicago,Illinois 60607-7023, United States

*S Supporting Information

ABSTRACT: While previous studies have found that unknown natural and synthetic organo-bromine compounds (NSOBCs)contributed more than 99% of the total organic bromine (Br) in the environment, there was no efficient method for untargetedscreening to identify NSOBCs in environmental matrixes. A novel untargeted method for identifying NSOBCs, basedon ultrahigh-resolution mass spectrometry (UHRMS) with the Q Exactive instrument was developed. This method includeda data-independent precursor isolation and characteristic fragment (DIPIC-Frag) procedure to identify NSOBCs. A total of180 successive 5-m/z-wide windows were used to isolate precursor ions. This resulted in a sufficient dynamic range andspecificity to identify peaks of Br fragment ions for analysis. A total of 2520 peaks of NSOBC compounds containing Br wereobserved in sediments from Lake Michigan, United States. A new chemometric strategy which combined chromatographicprofiles, isotopic peaks, precursor isolation window information, and intensities was used to identify precursor ions and chemicalformulas for detecting NSOBCs. Precursor ions for 2163 of the 2520 NSOBCs peaks (86%) were identified, and chemicalformulas for 2071 NSOBCs peaks (82%) were determined. After exclusion of isotopic peaks, 1593 unique NSOBCs wereidentified and chemical formulas derived for each. Most of the compounds identified had not been reported previously and had

Received: April 16, 2015Accepted: September 17, 2015Published: September 17, 2015

continued

Article

pubs.acs.org/ac

© 2015 American Chemical Society 10237 DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

Natural and synthetic, organo-bromine compounds (NSOBCs)are of concern due to their environmental persistence,

bioaccumulation, and potential for toxicity.Well-knownNSOBCs,such as polybrominated diphenyl ethers (PBDEs) and theirhydroxylated (OH-BDEs) and methoxylated (MeO-BDEs)analogues have been reported to be ubiquitous in environmentalmatrixes,1−3 wildlife,4,5 and humans.6−8 Results of both epidemio-logical investigations and controlled experiments suggest thatexposure to synthetic brominated compounds can cause variousadverse effects to humans and wildlife.7,9,10 Some naturally occurr-ing, brominated compounds exhibit even greater toxic potenciesthan synthetic compounds. For instance, OH-BDEs, which havebeen reported to bind to the thyroid hormone receptor (THR),exhibit greater potencies for neurotoxicity than their analogousPBDEs.11,12 Identification and quantification of these NSOBCs inthe environment is therefore important for assessment of potentialeffects on health of humans and/or wildlife.In natural waters, sediment can be a large depository of

persistent environmental pollutants. For chemicals that arepersistent and relatively hydrophobic, enrichment from waterinto sediment often enables their detection and quantitation attrace levels.3,13,14 Various NSOBCs such as PBDEs and someother brominated flame retardants have been detected in sedi-ments with concentrations commonly in the ∼ng/g, dry mass(dm) range.13−16 However, concentrations of total organicbromine (TOB) in samples of marine organisms and sedimenthave been found to be in the ∼μg/g range,17 which suggests thatcurrently known and concerned NSOBCs contribute <0.1% toTOB. Thus, identities of most TOB in sediments were unknown.17

Targeted ion monitoring, using single or triple quadrupolemass spectrometry (MS) coupled to liquid (LC) or gas (GC)chromatography is currently the main strategy to identify andquantify organic compounds for which standards are avail-able.18−20 To provide maximum selectivity and sensitivity, intargeted ionmonitoring, only characteristic ions or ion transitionsof targeted analytes are monitored. In contrast, ion-trap and time-of-flight (TOF) MS techniques are superior for screening ofunknown compounds.21 For example, an untargeted-methodusing a scripting approach combined with GC × GC-TOFMS has been developed and successfully used to identify newchloro/bromo-carbazole compounds.22 However, in addition tosynthetic brominated compounds and their byproducts, morethan 2 200 natural brominated compounds produced by marineorganisms have been identified.23 Because of the number ofpotential NSOBCs and the current difficulty in identifying novelcompounds, a more robust, untargeted method to identifyNSOBCs in the environment is needed. However, screeningNSOBCs in environmental samples faces several major challengesincluding (1) NSOBCs exhibit a wide range of physical-chemicalproperties which may lead to poor volatility for GC/MS orlow ionization efficiency for LC−MS, thus making it difficultto develop a single, robust mass spectrometric method for allpotential compounds (coverage); (2) concentrations of individ-ual NSOBCs can span several orders of magnitude (dynamicrange), thus necessitating a method with a large dynamic range;

and (3) due to the presence of large numbers of potentialinterferences in environmental samples, specificity of methodswould need to be enhanced, which would increase the complexityof data analysis (specificity).To address these challenges, the goal of this study was to

develop a data-independent precursor isolation and characteristicfragment (DIPIC-Frag) method to screen NSOBCs in environ-mental samples. This method was incorporated in the operationof liquid chromatography (LC) coupled to an ultrahigh-resolutionmass spectrometer (UHRMS), the Q Exactive quadrupole,Orbitrap MS, with atmospheric pressure chemical ionization(APCI). To analyze multiplexed data sets produced by UHRMS,a novel data mining strategy was developed to identify precursorions and predict chemical formulas or chemical structures ofNSOBCs. The method exhibited comprehensive coverage ofchemical structure diversity, large dynamic range, and specificityand was successfully used to detect and identify 2 520 peaksassociated with NSOBCs in sediments from Lake Michigan,the sole Laurentian Great Lake lying completely within theUnited States.

■ MATERIALS AND METHODSChemicals and Reagents. Authentic standards of 10 native

PBDEs, threeOH-BDEs, threeMeO-BDEs, three diastereoisomesof hexabromocyclododecane (HBCDs), tetrabromobisphenol A(TBBPA), bis (2-ethylhexyl)-2,3,4,5-tetrabromophtalate (TBPH),and 2-ethylhexyl-tetrabromobenzene (TBB) were purchasedfrom Wellington Laboratories Inc. (Guelph, ON, Canada).5-Bromoindole and 4-bromophenol were purchased from Sigma-AldrichChemical Co. (St. Louis,MO). 1,3,6,8-Tetrabromocarbazolewas purchased from Toronto Research Chemicals Inc. (Toronto,ON, Canada). Hydroxylated TBB (OH-TBB) andOH-TBPHwerepurified from BZ-54 technical product as previously described.24

Florisil (6 cm3, 1 g, 30 μm) solid-phase extraction (SPE) cartridgeswere purchased from Waters (Milford, MA). Methyl tert-butylether (MTBE), dichloromethane (DCM), hexane, methanol, andacetone were all of omni-Solv grade and were purchased fromEMD Chemicals (Gibbstown, NJ).

Collection of Sediments. Surface sediment samples werecollected from two locations in LakeMichigan in September 2010(sampling map is shown in Figure S1, Supporting Information)using a PONAR grab sample, as described previously.25 Sampleswere separated into aliquots and stored in amber glass jars withaluminum foil liner caps. Samples were transported on iceand stored in the dark at −20 °C. Each sample was lyophilized,manually homogenized, and passed through a 1 mm sieve.

Sample Pretreatment. Approximately 10 g of sediment wasextracted for identification of NSOBC. Methods for extractionhave been described previously.17 Briefly, samples were extractedby use of an accelerated solvent extractor (Dionex ASE-200,Sunnyvale, CA). Two solvents were used in the extraction: (1)n-hexane/DCM (1:1) at 100 °C and 1500 psi, and (2) n-hexane/MTBE (1:1) at 60 °C and 1000 psi. Two extraction cycles(10 min each) were performed for each solvent per sample(approximately 50 mL for each solvent). Following extraction,

intensities which were 100- to 1000-fold greater than the congeners of polybrominated diphenyl ethers (PBDEs). In extracts ofsediments, these compounds exhibited variations in intensities (<103 to ∼108), m/z values (170.9438−997.5217), retention timeson a C18 column (1.0−29.3 min), and the number of Br atoms (1−8). Generally, compounds with greater m/z values had longerretention times and greater numbers of Br atoms. Three compounds were used in a proof-of-concept experiment to demonstratethat structures of some of the screened NSOBCs could be further predicted by combining searching of database libraries and high-resolution MS2 spectra.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10238

fractions were combined. Volumes of extracts were reduced to∼1 mL by rotary evaporation and loaded onto Florisil cartridgeswhich had been previously conditioned by 6 mL of DCM.NSOBCs were eluted from Florisil cartridges by use of 5 mL ofDCM and then 5 mL of methanol. Final extracts were blown todryness under a gentle stream of nitrogen and reconstituted in400 μL of acetone. Acetone was selected as the reconstitutionsolvent, considering the hydrophobicity of identified NSOBCs.Because a limited number of NSOBCs were detected inmethanol fractions (data not shown), only the DCM fractionfrom the cartridges were collected for screening of NSOBCs. Theuse of Florisil cartridges, which have also been used in previoussample pretreatments for halogenated compounds analysis,24

allowed for the removal of most of the yellow interferences inextracts of sediment. Such a simple one-step sample cleanupmethod was useful for untargeted screening of NSOBCs insediments to avoid potential loss of compounds, but moreefficient sample pretreatment methods such as gel permeationchromatography (GPC) would be warranted if the DIPIC-Fragmethod was applied to more complicated matrixes such as bioticsamples. To avoid potential background contamination duringsample pretreatment, all equipment rinses were carried out withacetone and hexane, and procedural blank experiments were per-formed along with each batch of samples. A total of 113 NSOBCpeaks were detected in the blank, partly due to instrument carry-over, but the peak abundances of these NSOBCs were at least10-fold less than those of sediments samples. The backgroundcontamination from the blanks was subtracted from sedimentsamples for downstream data analysis.LC-Q Exactive Data Acquisition. Aliquots of extracts were

analyzed using a Q Exactive UHRMS equipped with a DionexUltiMate 3000 UHPLC system (Thermo Fisher Scientific).Separation of NSOBCs was compared among different typesof HPLC columns, and the Hypersil GOLD C18 column (3 μm;2.1 mm × 50 mm; Thermo Fisher Scientific) was selected forthe present method considering the good separation abilityand sensitivity achieved with its use. Injection volume was 5 μL.Ultrapure water (A) and methanol (B) were used as mobilephases. Initially 20% B was increased to 80% in 3 min, thenincreased to 100% at 8 min and held static for 19.5 min, followedby a decrease to initial conditions of 20% B and held for 2 minto allow for equilibration. Rate of flow was 0.20 mL/min. Thecolumn and sample compartment temperatures were maintainedat 30 and 10 °C, respectively.Data were acquired in data-independent acquisition (DIA)mode.

Parameters for DIA were one full MS1 scan (150−2 000 m/z)recorded at resolution R = 70 000 (at m/z 200) with a maximumof 3 × 106 ions collected within 100 ms, followed by six DIAMS/MS scan recorded at a resolution R = 35 000 (at m/z 200)with maximum of 1 × 105 ions collected within 60 ms. DIA datawere collected by use of 5-m/z-wide isolation windows perMS/MS scan, although different combinations of isolationwindows could be used in future work. Each DIA MS/MS scanwas chosen for analysis from a list of all 5 m/z isolation windows.In these experiments, 180, 5-m/z-wide windows between 100 and1 000 m/z were grouped into nine separate methods, each ofwhich contained 20 windows. Small overlaps with neighboringwindows were used to reduce the likelihood of placing windowedges on critical target peaks. Mass spectrometric settings forAPCI (−) mode were as follows: discharge current, 10 μA;capillary temperature, 225 °C; sheath gas, 20 L/h; auxiliary gas,5 L/h; probe heater temperature, 350 °C.

Instrument detection limits (IDLs) for the model chemicalswere defined as 5 times within 20% relative standard derivationfor the standards. Method detection limits (MDLs) for themodel chemicals were calculated based on six replicate analysesof sample extracts at a concentration of approximately five timesthe corresponding IDLs, and then MDLs were calculated aspreviously described.26 Recoveries were determined by spiking25modelNSOBCs into samples of sediment at 50 ng/g dw (n = 3).The recoveries for these compounds ranged from 73 ± 7%(BDE-183) to 98 ± 12% (OH-TBPH).

Formula Elucidation. Elemental compositions of detectedNSOBCs were calculated using a program written for R softwarein which the mass tolerance was set to 5 ppm for compoundsgreater than 200 m/z. Chemical formulas were set to containup to 100 C, 200 H, 5 N, 30 O, 5 I, and 2 S per molecule. Thenumber of Br or Cl atoms was constrained based on informationfrom isotopic peaks. All assigned formulas were required to meetbasic chemical criteria as described previously.27

Distribution of Intensities of Isotopic Peaks. Since thepattern of isotopic peaks is important to narrow the list ofpotential formulas for a given exact mass, numbers of bromineand chlorine atoms in detected NSOBCs were calculated basedon patterns of isotopic peaks.28 Details of the method for makingthe calculations are provided in the Supporting Information.

■ RESULTS AND DISCUSSIONPrinciples andWorkflow of the DIPIC-Frag Method. To

address challenges for identifying NSOBCs, an untargetedDIPIC-Frag method was developed by combining severaltechniques to address the challenges including coverage, dynamicrange, and specificity (workflow was shown in Figure 1). To testthe performance of the DIPIC-Frag method, 25 model NSOBCsincluding 19 synthetic chemicals and 6 natural products were used(Supporting Information Table S1). These compounds haddiverse chemical structures ranging from hydrophilic phenoliccompounds (e.g., TBBPA) to highly hydrophobic compounds(e.g., BDE-209). To increase coverage of NSOBCs, APCI (−)was used. Following optimization, all 25 model NSOBCs showedsufficient sensitivity (IDLs ranged from 5 ng/L to 20 μg/L),including the hydrophobic BDE-209. The MDLs were in therange of <1 to 10 000 pg/g, dm and were less than concentrationsof the most well-known NSOBCs (typically ng/g) and also totalorganic bromine (typically μg/g) in sediments. Thus, the DIPIC-Frag method would likely be sufficiently sensitive to identifyunknown NSOBCs in environmental samples. Compared tothe traditional GC/MS method, which is not compatible withcompounds of lesser volatility, and LC−ESI-MSmethod, which isnot compatible with less polar compounds, the use of APCI (−)increased coverage of unknown NSOBCs of diverse chemicalstructures. In addition, compared with the electron impact (EI)ionization source used in GC/MS in which molecules are oftencleaved to fragments, the LC compatible APCI source is relatively“softer” and preserves parent molecular ions, which allowed forthe calculation of molecular mass and chemical formulas.To address the issue of specificity of NSOBCs and distinguish

from other interferences, characteristic fragment ions wereused to identify peaks associated with NSOBCs. Although the25 model NSOBCs had diverse chemical structures, they allproduced a Br fragment ion (m/z = 78.9171) in negative mode atrelatively high HCD collision energies (>30 eV) (typical production spectra of OH-TBB was shown in Figure S2). Preferentialcleavage of Br and its greater electronegativity has been welldocumented in traditional GC/MS methods and enables greater

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10239

sensitivity with electron capture negative ionization (ECNI)compared to EI ionization,29,30 which has also been observedfor MeO-BDEs and OH-BDEs using LC−MS/MS.31 Thus, the[Br] ion can be used as a characteristic product ion to specificallyscreen compounds containing Br with the use of APCI (−).Based on this strategy, NSOBCs peaks could be easily dis-tinguished from other interferences and the method providedgreat convenience in subsequent data analysis.To increase the dynamic range of the method, multiple

successive MS2 windows during the data-independent acquis-ition (DIA) were used in the DIPIC-Frag method (Figure 1a andFigure S3). One full scan with a mass resolution of 70 000followed by six cycles of DIA scans with mass resolutions of35 000 were performed in the DIPIC-Frag method. A detailedmass scanning scheme is shown in Figure S3. Performance ofDIPIC-Frag was closely related to the width of precursorisolation windows. In principle, to reduce coeluted interferences,narrower isolation windows (e.g., 50 2-m/z-width windows tocover the mass range, 500−600m/z) are preferable, compared towider isolation windows (e.g., 2 50-m/z windows to cover themass range, 500−600 m/z). However, because of the limitedscanning rates of the Q Exactive instrument, use of narrowerisolation windows would limit coverage of masses and decreasethroughput. Following optimization, the width of isolationwindows was set at 5-m/z (Figure 1a and Figure S3) and nineindividual methods with 100m/z range with 20 windows for eachmethod and 180 windows for all nine methods with 900-m/zmass ranges (100−1000 m/z) were used. Because primaryscreening experiments showed that few NSOBCs detected in

sediments had m/z > 1 000, the maximal mass range was set to1 000 m/z.Performance of the DIPIC-Frag method was evaluated by its

use to identify NSOBCs in sediments from Lake Michigan.When extracted with a 7 ppm mass width, multiple characteristicpeaks for Br (m/z = 78.9171) were detected in multiple pre-cursor isolation windows (Figure 1b). Confirmation was obtainedbymonitoring a second isotopic peak of Br for them/z = 80.9150.Windows for isolation of precursors in the DIPIC-Frag methodwere necessary to deconvolute peaks for NSOBCs, since multipleNSOBCs exhibited similar retention times and thus their Brfragment peaks could not be efficiently deconvoluted by usingsingle precursor isolation mode, as has been used previously inuntargeted methods. An intensity cutoff of 1000 was used in theDIPIC-Frag method, and Br peaks exceeding this thresholdwere identified as NSOBCs for subsequent data analysis.Detected NSOBCs were distributed across multiple precursorisolation windows and retention times. NSOBCs were detectedin all precursor isolation windows greater than 165 m/z withretention times of 1.0−29.3 min. The elution of several NSOBCsat ∼1.0 min indicated they could not be efficiently retainedby a C18 column. Future optimization of HPLC condition iswarranted to enhance coverage of these polar NSOBCs.Finally, an average of 15 NSOBC peaks were detected in each ofthe 180 5-m/z-width precursor isolation windows, with a total2 520 peaks detected. The DIPIC-Frag method detected moreNSOBCs than did previous untargeted methods, which typicallyidentified fewer than 100.22

Figure 1.Typical workflow of the DIPIC-Frag method to identify brominated compounds: (a) 180 successive 5-m/z-width precursor isolation windowswere used in the DIPIC-Frag method. (b) Bromine fragment peaks were detected in each precursor isolation window, with an average of 15 peaksdetected for each window. (c) Mass spectra in separated DIA windows at the same retention time as the bromine fragment peaks for precursor ionsalignment. Collisional energy was set to a wide range (10, 30, 60 eV) to produce both bromine fragments (left red dashed cycle of the spectra) andprecursor ions (right red dashed cycle of the spectra). (d) The chromatographic elution profiles of bromine fragments (top) and precursor candidates(bottom) were used to identify precursor ions for each bromine fragment peak. Typically, 1−3 precursor ions were identified from 20 to 30 candidateions in the precursor ion region. (e) Isotopic peaks of the proposed precursor ions were further determined to confirm the identity of the precursor ions.Intensity for each isotopic peak is shown in the top right of the figure. (f) The chemical formula was calculated based on multiple lines of evidenceincluding number of bromine atoms, intensity information and exact m/z values. (g) Chemical structures of some compounds were identified bycombining database searches and high-resolution MS2 spectra.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10240

Precursor Ion Alignment. Identification of precursor ionsfor each of the 2 520 potential NSOBCs peaks was accomplishedby use of a novel data mining strategy, which was developed bycombiningmultiple lines of evidence (Figure S4). First, a steppedcollision energy procedure was used at lesser (10 eV) and greatervalues (30 and 60 eV) during DIA scans. With this procedure,precursor ions were fragmented with different energies butinjected to the C-trap for detection at the same time, whichallowed for collection of information on both precursor ionsand Br fragments simultaneously in the same spectrum. In theportion of spectra (right side of the mass spectra from Figure 1c)identified as the “precursor ion region”, relatively large signalsfrom precursor ions between the 5-m/z-width isolation windowswere observed. In addition to an expanded dynamic range,the use of precursor isolation windows also reduced the timeto identify precursor ions because the width of the windowinherently limits the number of ions to be identified, whichtypically contained 20−30 precursor ions candidates. Profiles ofretention times and shapes of peaks during chromatographicelution of the 20−30 ions in the precursor ion region wereinvestigated, and the ions showed similar chromatographicprofiles with Br peaks were identified as potential precursor ionsof the corresponding NSOBC peaks. Use of chromatographicelution profiles for alignment of precursor ions has been appliedpreviously in metabolomics and proteomics studies.32,33 In theexample shown in Figure 1 (bromoindole), the precursor ionwith m/z of 193.9599 was specifically identified as havingthe same chromatographic elution profile as the correspondingBr fragment peaks (Figure 1d). Information from the precursorion region in DIA windows and information from full scanspectra was integrated to obtain the most accurate precursor ioninformation (Figure S4). The m/z values identified in full scanmode were used because of the greater mass resolution of themethod (R = 70 000 at 200m/z) compared to separated windows

(R = 35 000 at 200 m/z). Most precursor ions were detectedby both full scan and DIA scan. However, approximately 15%of lesser-abundance precursor ions were observed only in theprecursor ion region in separated DIA windows (Figure 2a). Thisresult was expected due to interferences in extracts that exceededthe dynamic range of the full scan mode. Approximately 10% ofprecursor ions were only observed in full scan mode (Figure 2b),due likely to the greater maximal injection of ions, 3 × 106, infull scan, which was 30-fold greater than for separated windows(1 × 105). Following the first two steps, precursor ions werespecifically detected for most NSOBCs. Third, because they hadrelatively large abundances and have been previously used tocharacterize brominated compounds, isotopic peaks of NSOBCswere used to further confirm results.22 For bromoindole, twoisotopic peaks (m/z = 193.9599 and 195.9579) were specificallydetected with similar chromatographic elution profiles, whichfurther confirmed that the ion at m/z = 193.9599 was theprecursor ion to the corresponding Br peaks (Figure 1e). Finally,theoretically, intensities of precursor ions should be greater thanor similar to that of product ions divided by product ion number(text in Supporting Information). Thus, intensities of precursorions in full scan were also calculated. This calculation can beuseful to identify overlaps of NSOBCs peaks which could notbe completely deconvoluted by DIA windows in a few cases.For instance, a potential precursor ion with m/z 435.7738 and atretention time 12.40 min was observed in the window centered at435 ± 2.5 m/z and had a similar chromatographic elution profileto corresponding Br fragment peaks (Figure 2c). However,intensity of the proposed ion was less than that of the Br fragmentdivided by Br number and monoisotopic peaks number (detailsof the calculation are provided in Supporting Information).Following inspection of the mass spectra at 12.40 min, anotherion with a m/z of 433.7586 (isotopic ion with m/z of 435.7565)was observed (Figure 2d), which overlapped with the ion at

Figure 2. Identification of precursor ions by combining full scan spectra and precursor ion region information from separated DIA windows. (a) Theprecursor ion at 248.8549 m/z was detected in the precursor ion isolation DIA window but could not be detected in full scan mode due to the highabundance of interferences (total ion intensity was 6.08 × 109 in the bottom spectra) and limited dynamic range of full scan spectra; (b) ion at247.8711m/zwas detected in full scan spectra but could not be detected in the precursor ion isolation DIA window; (c) twoNSOBC compounds with amonoisotopic peak at 435.7738 m/z and 433.7586 m/z produced overlapping bromine fragment peaks at a retention time (rt) = 12.40 min. (d)Consistent with part c, mass spectra at rt = 12.40 min showed two isotopic peak clusters with mass derivations of ∼2 Da. Red arrows indicatecorresponding bromine fragment or precursor ions peaks.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10241

m/z 435.7738 with a retention time of 12.40 min. This peak wasexpected to be the dominant compound producing a Br fragmentat 12.40 min. By combining multiple lines of evidence, precursorions for 2 163 peaks corresponding to NSOBCs were identified,which represented 86% of the 2 520 peaks originally identified asbeing due to brominated compounds. Most of the unidentifiedNSOBCs had relatively small peak intensities (<1 × 104) andgreat molecular mass (>600m/z). Therefore, their precursor ionswere expected to be obscured by coeluting greater-abundanceinterferences. These unidentifiedNSOBCpeaks could potentiallybe identified by narrowing the range of the full scan or increasingthe amount of ions injected. Intensities of observed compoundsexhibited a large dynamic range from <103 to ∼108 (Figure 3a).Such a dynamic range of NSOBCs has posed challenges topreviously used untargeted screening methods but was achievedby the novel DIPIC-Frag method.Because of the lack of commercial standards for most observed

compounds, it was not possible to accurately determine theirconcentrations. However, peak intensities of NSOBCs were100−1000-fold greater than several congeners of PBDEs andOH-PBDEs. Another interesting result from the method is thatmultiple isomers were observed for some NSOBCs. Similarm/z values but with different retention times (Figure 2c) wereobserved for some NSOBCs. Considering the resolution of theQ Exactive, these NSOBCs should be multiple isomers with Bratoms in various positions or even different chemical backbones,and further studies are warranted to clarify their exact chemicalstructures.34

Determination of Chemical Formulas. To further deter-mine formulas of NSOBCs, a multiple-step strategy was used.The first step was to use isotopic peaks to calculate numbersof Br/Cl and to distinguish isotopic peaks from the primary

monoisotopic peak. For example, on the basis of distributionsof isotopes (Figure S5), relative intensities of the two isotopicpeaks in Figure 1e were ∼1:1, which indicated that there was1 Br atom in the molecular formula. Isotopic peaks of Brfragments from DIA precursor isolation windows also providedimportant information on isotopic peaks observed for precursorions. For the Br fragment peak (retention time was 10.52 min)from the precursor window at 350 ± 2.5 m/z, only the mono-isotopic ion at m/z = 78.9171 was detected, which indicated thatthe primary monoisotopic ion of the NSOBC precursor ionshould be detected in the second half of the precursor ion region(350−352.5 m/z) (Figure S6). On the basis of this information,the precursor ion at m/z 351.8977 was identified. From the nextprecursor isolation window of 355 ± 2.5 m/z, the Br fragmentpeak was still observed but showed similar intensities for the twoBr fragment ions (m/z = 78.9171 and 80.9151), which indicatedthe presence of two or three isotope peaks in these windows(ions at m/z = 353.8959 and m/z = 355.8939). Taken together,this information allowed for identification of the monoisotopicion of the detected brominated compound, which containedtwo Br atoms (C13H8NOBr2), at m/z 351.8977. Similarly, onthe basis of the detection of both the Br isotopic peak of m/z78.9171 and 80.9151 with similar intensity, the primarymonoisotopic peak of the precursor ion of the NSOBC inFigure 1e was narrowed to the first half of the isolation window(192.5−195 m/z), further confirming that the ion with m/z193.9599 was the primary monoisotopic peak of the NSOBCcompound (bromoindole). Information on numbers of Cl andBr atoms of the compound was used in a program written by theauthors to calculate the elemental composition of the compounds.Generally, after constraining the numbers of Br and Cl atoms andmass tolerance (5 ppm), there were few compounds (<4) included

Figure 3. Distribution of identified NSOBC compounds. (a) Distribution of intensities of 2 520 NSOBCs and their comparison to previously knownbrominated compounds. Multiple analogues were detected for brominated carbazole, and the bromine number (x + y) ranged from 1 to 7, where x and yindicate the number of bromines. (b) Distribution of intensities of the 2 520NSOBC compounds with different ranges ofm/z values. (c) Distribution ofthe 2 520 NSOBC compounds by retention time and m/z values. The sizes of the dots are proportional to intensities. The colors of dots representnumbers of bromines. Red represents numbers of bromines, and gray represents those precursor ions whose formula could not be identified. AllNSOBCs were determined in a surficial sediment sample (sed-32) from Lake Michigan.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10242

on lists with m/z less than 600 m/z and fewer yet (≤2) on listswith m/z less than 400 m/z. However, the list which includedcompounds with larger m/z values (>600) was larger. Thus,compounds with the least mass errors were used to predictchemical formulas of NSOBCs. For example, after constrainingthe number of Br atoms to 1 and for an m/z of 193.9599,C8H5NBr was identified as the single formula with mass error of−3.0 ppm (Figure 1f). Finally, by use of this data mining strategy,chemical formulas were successfully identified for 2 071 of the2 520NSOBCs (82%), corresponding to 1593 of uniqueNSOBCcompounds after excluding isotopic peaks. Detected NSOBCshad great variations in m/z values (170.9438−997.5217),retention times (1.0−29.3 min), and numbers of Br atoms(1−8); yet, these values were clearly associated, since compoundswith larger m/z generally had greater retention times andnumbers of Br atoms (Figure 3c). The smallest NSOBC wasidentified as bromophenol (C6H5OBr) withm/z of 170.9438 andretention time at 6.44 min. As shown in Figure S7, the MS1

spectra, retention time, and MS2 spectra for the DIA windowof the identified bromophenol peak were consistent with resultsfrom the commercially available standard, 4-bromophenol.Among the nine methods applied to various ranges of masses,the number of NSOBCs and their intensities increased from themass range 100−200m/z to 400−500m/z and declined from themass range 500−600 m/z to 900−1000 m/z. This trend wasparticularly apparent for precursor windows greater than 700m/z(Figure 3b). Most compounds detected had m/z values between300 and 700 and retention times between 8 and 15 min. Massesand retention times were similar to those of BDE-47 (molecularmass = 564.7, retention time = 8.7 min), which is one of the mostbioaccumulative synthetic brominated compounds, indicatingthe potential bioaccumulation of these unknown NSOBCs inorganisms.Although identities of dominant NSOBCs were similar in

sediments from the two locations in Lake Michigan, profiles ofdetected NSOBCs showed great differences. For example, takethe NSOBCs peaks extracted from the 405 ± 2.5 m/z DIAwindow, compound peak 2, peaks 3/4, peak 5, and peak 6were prominent in both sediment samples (Figure 4a,b), butthe relative contributions of the compounds varied. The relative

contribution of peak 5 was greater in sediment-32 (Figure 4a),while the relative contributions of peaks 3/4 and peaks 8/9 weregreater in sediment-44 (Figure 4b). Ratios of intensities of all the1593 detected NSOBCs compounds between the two sedimentsamples showed great variation which ranged from 10−5 to 103

(Figure 4c), especially for compounds with retention timesbetween 10 and 15 min (Figure 4d). Previous studies havereported possible natural and anthropogenic emission sourcesof NSOBCs.35,36 The variation of the NSOBC profiles amongsediment locations indicated different emission source patterns(e.g., different microorganism communities), and future studiesare warranted to clarify the source emission of NSOBCs.

Determination of Chemical Structure by CombiningMS/MS Spectra and Chemical Database Information.Most of the compounds detected had not been reportedpreviously. While the present study focused on screening forthe presence of compounds, chemical structures of some of thenovel NSOBCs could be further predicted by combiningdatabase search information and high-resolution MS2 spectra.For example, the chemical formula of themolecular ion [M−H]−

of one of the NSOBCs was determined to be C8H5NBr(Figure 1f). The most likely structure for the chemical with theformula C8H6NBr was determined to be bromoindole by useof the Chemspider database (Figure 1g). By use of a commercial5-bromoindole standard, the NSOBC peak was successfullyvalidated as bromoindole (Figure S8). The class of chemicalsto which this NSOBC belonged was recently reported to beproduced by amarinemicroorganism.35,36 The chemical structureof another detected NSOBC peak with m/z of 399.7975 andretention time at 10.7 min was also identified. The chemicalformula, which was determined to beC12H5NBr3, was expected torepresent the molecular ion of the compound. Thus, the chemicalformula, C12H6NBr3, and the Chemspider database were used toidentify the compound as tribromo-9H-carbazole. On the basisof a similar strategy, we have identified more than 50 isomers/analogues of halogenated carbazoles with different chlorine,bromine, and iodine atom numbers. To validate the results, one ofthe identified halogenated carbazole peaks (tetrabromocarabazole,with similar structure to PCDF) was compared with the com-mercial standard, 1,3,6,8-tetrabromocarbazole. As shown in

Figure 4. Comparison of profiles of NSOBCs between two sediment samples from Lake Michigan: sediment-32 (a) and sediment-44 (b)(chromatogram was extracted from 405± 2.5m/zDIA window). (c) Intensity ratios of NSOBCs in sediment-44 to those in sediment-32. (d) Intensityratios of NSOBCs compounds between two sediment samples and their relationship between retention time.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10243

Figure S9, the MS1, retention time, and MS2 spectra from theDIA window were consistent between the standard and sedimentsample. Since Zhu et al. reported the occurrence of brominatedcarbazoles in sediment cores from Lake Michigan in 2005,37 arecent study has also identified iodine analogues of this class ofcompounds in sediment samples from Lake Michigan.22,38

In most cases, a direct search of publicly available databasesusing predicted formulas did not yield results. Therefore, in thesecases, a different strategy was adopted to achieve greater hit ratesfor chemical structures. Taken the compound from Figure S6,for example, from the isomer clusters of compounds in the350 ± 2.5 m/z window, with retention times of 9−11 min,the monoisotopic peak of the compound was determined to be351.8977 m/z (Figure 5). There were at least four compoundsdetected with the same m/z values and similar retention times,which indicated the potential existence of isomers for thesecompounds. Formulas of compounds were predicted to beC13H8NOBr2 with a mass error of 1.1 ppm. Compounds withthis formula have not been previously reported to occur in theenvironment. Several ions with similar m/z values to theprecursor ions with m/z = 336.8743, 338.8721, and 340.8700were detected in the same DIA mass spectra (Figure 5a). Thesemight be product ions of the compounds. To further confirmthese results, the chromatographic elution profiles of the threeions (Figure 5b), were investigated. All three ions had the samechromatographic profiles as Br fragment peaks and the precursorion at m/z = 351.8977. These results indicated that the threepeaks are likely isotopic peaks of a fragment from the compound.The chemical formula of the fragment was predicted to beC12H5NOBr2, which would represent the loss of a methyl group

(−CH3) from the precursor compound in the HCD collisioncell. To further identify chemical structures, the Chemspiderdatabase was used to further identify potential structures forC13H9NOBr2. Reasonable structures that contained a −CH3

group were not obtained by direct search of Chemspiderdatabase. Thus, the chemical formula was changed to excludethe Br atom (query of C13H11NO rather than C13H9NOBr2), andmethyl- and hydroxylated polybrominated carbazoles wereidentified (Figure 5c). On the basis of the chemical structure,the 336.8743 m/z fragment with loss of a methyl group from theprecursor, it was identified as bromo-carbazole. One limitation ofthis type of identification is the potential displacement reac-tion that has been reported to occur in the APCI (−) source.39This reactionmight result in formation of the [M−Br +O]− ion.Therefore, there are two potential chemical structures forchemicals which incorporate oxygen, as shown in Figure 5c. Inprinciple, the two possible chemical structures could be furtherdistinguished by use of ESI ionization or methylation by use of a1,4-diazabicyclo[2.2.2]octane (DABCO) catalyst or dansylationderivatization, but this is beyond the scope of the present paper.These three examples have demonstrated that the combinationof database searches and high-resolutionMS2 spectra can be usedto identify chemical structures of previously unknown NSOBCs.However, the identification of chemical structures of compoundsis time-consuming and challenging, and it is recommendedto first limit the number of NSOBCs based on intensity oreffects observed in bioassays and then identify structures of targetNSOBCs.Overall, the present study proposed a novel, untargeted

DIPIC-Frag method to screen NSOBCs in the environment and

Figure 5. Determination of structures of a novel compound by combining MS/MS spectra and information from a publicly available compounddatabase. Shown are (a)mass spectra in 350± 2.5m/z and 355± 2.5m/zwindow at rt = 10min (precursor ion was 351.8977m/z, potential product ionwas 336.8743 m/z); (b) chromatographic elution profiles and isotopic peaks to confirm product ions; (c) proposed chemical structures andfragmentation routes. The precursor ion might represent the displacement of [M − Br + O] with addition of oxygen in the APCI source.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10244

presented numerous advantages in coverage, specificity, anddynamic range. A systematic data mining strategy was developedto identify the precursor ions, chemical formulas, and chemicalstructures of detected NSOBCs. On the basis of the results ofthis study, the largest known mass spectrometry library ofNSOBCs (2520 NSOBC peaks, Supplementary Data Table) wasestablished which could be adopted in low-resolution massspectrometry. Future studies are warranted to investigate emissionsources, environmental behaviors, and potential ecological risksof novel NSOBCs identified in this study. Additionally, theircontribution to total organic Br in the environment should beinvestigated.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on the ACSPublications website at DOI: 10.1021/acs.analchem.5b01435.

Calculation of isotopic peaks intensities distribution,information on 25 model NSBCs, sampling map, production spectra of OH-TBB, scanning scheme of the DIPIC-Fragmethod, workflow to identify precursor ions of NSBCscompounds, distribution of isotopic peaks ofNSBCs, isotopicpeaks of bromine fragment to help to identify brominenumber and precursor ions, and validations of bromophenol,bromoindole, and tetrabromocarbazole (PDF)Precursor ion DIA window, retention time, peak intensity,exact mass, predicted formula, and calculated mass error ofall the 2520 NSOBC peaks identified by the DIPIC-Fragmethod (XLSX)

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: huisci@gmail.com.*Phone: 306-966-2096; 306-966-4680 (secretary). Fax: 306-966-4796. E-mail: jgiesy@aol.comNotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis research was part of the Great Lakes Sediment SurveillanceProgram funded by a Cooperative Agreement from the U.S. EPAGreat Lakes Restoration Initiative with Assistance Grant GL-00E00538 (U.S. EPA Program Officer Todd Nettesheim), aDiscovery Grant from the Natural Science and EngineeringResearch Council of Canada (Project No. 326415-07) and agrant from the Western Economic Diversification Canada(Project Nos. 6578, 6807, and 000012711). The authors wishto acknowledge the support of an instrumentation grant from theCanada Foundation for Infrastructure. Prof. Giesy was supportedby the Canada Research Chair program, and the 2012 “HighLevel Foreign Experts” (Grant GDT20143200016) program,funded by the State Administration of Foreign Experts Affairs,the P.R. China to Nanjing University, and the Einstein ProfessorProgram of the Chinese Academy of Sciences.

■ REFERENCES(1) Kohler, M.; Zennegg, M.; Bogdal, C.; Gerecke, A. C.; Schmid, P.;Heeb, N. V.; Sturm, M.; Vonmont, H.; Kohler, H. P. E.; Giger, W.Environ. Sci. Technol. 2008, 42 (17), 6378−6384.(2) Wan, Y.; Wiseman, S.; Chang, H.; Zhang, X. W.; Jones, P. D.;Hecker, M.; Kannan, K.; Tanabe, S.; Hu, J. Y.; Lam,M.H.W.; Giesy, J. P.Environ. Sci. Technol. 2009, 43 (19), 7536−7542.

(3) Zhang, K.; Wan, Y.; Jones, P. D.; Wiseman, S.; Giesy, J. P.; Hu, J. Y.Environ. Sci. Technol. 2012, 46 (4), 2148−2155.(4) Lam, J. C. W.; Lau, R. K. F.; Murphy, M. B.; Lam, P. K. S. Environ.Sci. Technol. 2009, 43 (18), 6944−6949.(5) Bustnes, J. O.; Yoccoz, N. G.; Bangjord, G.; Polder, A.; Skaare, J. U.Environ. Sci. Technol. 2007, 41 (24), 8491−8497.(6) Zhou, S. N.; Buchar, A.; Siddique, S.; Takser, L.; Abdelouahab, N.;Zhu, J. P. Environ. Sci. Technol. 2014, 48 (15), 8873−8880.(7) Chen, A.M.; Yolton, K.; Rauch, S. A.;Webster, G.M.; Hornung, R.;Sjodin, A.; Dietrich, K. N.; Lanphear, B. P. Environ. Health Perspect.2014, 122 (8), 856−862.(8) Sjodin, A.; Jones, R. S.; Caudill, S. P.; Wong, L. Y.; Turner, W. E.;Calafat, A. M. Environ. Sci. Technol. 2014, 48 (1), 753−760.(9) Zhou, T.; Ross, D. G.; DeVito, M. J.; Crofton, K. M. Toxicol. Sci.2001, 61 (1), 76−82.(10) Li, Z. H.; Liu, X. Y.; Wang, N.; Chen, J. S.; Chen, Y. H.; Huang, J.T.; Su, C. H.; Xie, F. K.; Yu, B.; Chen, D. J. Environ. Health Perspect.2012, 120 (4), 541−546.(11) Li, F.; Xie, Q.; Li, X. H.; Li, N.; Chi, P.; Chen, J. W.; Wang, Z. J.;Hao, C. Environ. Health Perspect. 2010, 118 (5), 602−606.(12) Dingemans, M. M. L.; van den Berg, M.; Westerink, R. H. S.Environ. Health Perspect. 2011, 119 (7), 900−907.(13) Peng, H.; Wan, Y.; Zhang, K.; Sun, J. X.; Hu, J. Y. Environ. Sci.Technol. 2014, 48 (10), 5458−5466.(14) Zhu, L. Y.; Hites, R. A. Environ. Sci. Technol. 2005, 39 (10), 3488−3494.(15) Yang, R. Q.; Wei, H.; Guo, J. H.; Li, A. Environ. Sci. Technol. 2012,46 (6), 3119−3126.(16) Kwan, C. S.; Takada, H.; Mizukawa, K.; Saha, M.; Rinawati;Santiago, E. C. Mar. Pollut. Bull. 2013, 76 (1−2), 95−105.(17) Wan, Y.; Jones, P. D.; Wiseman, S.; Chang, H.; Chorney, D.;Kannan, K.; Zhang, K.; Hu, J. Y.; Khim, J. S.; Tanabe, S.; Lam, M. H. W.;Giesy, J. P. Environ. Sci. Technol. 2010, 44 (16), 6068−6073.(18) Trouborst, L.; Chu, S.; Chen, D.; Letcher, R. J.Chemosphere 2015,118, 342−349.(19) Okonski, K.; Degrendele, C.; Melymuk, L.; Landlova, L.;Kukucka, P.; Vojta, S.; Kohoutek, J.; Cupr, P.; Klanova, J. Environ. Sci.Technol. 2014, 48 (24), 14426−14434.(20) Letcher, R. J.; Bustnes, J. O.; Dietz, R.; Jenssen, B. M.; Jorgensen,E. H.; Sonne, C.; Verreault, J.; Vijayan, M. M.; Gabrielsen, G. W. Sci.Total Environ. 2010, 408 (15), 2995−3043.(21) Chen, D.; Hale, R. C.; Letcher, R. J. Environ. Toxicol. Chem. 2015,34 (4), 687−99.(22) Pena-Abaurrea, M.; Jobst, K. J.; Ruffolo, R.; Shen, L.; McCrindle,R.; Helm, P. A.; Reiner, E. J. Environ. Sci. Technol. 2014, 48 (16), 9591−9599.(23) Gribble, G. J. Environ. Sci. Pollut. Res. 2000, 7 (1), 37−49.(24) Peng, H.; Saunders, D.M. V.; Sun, J. X.; Codling, G.;Wiseman, S.;Jones, P. D.; Giesy, J. P. Environ. Sci. Technol. 2015, 49, 2999−3006.(25) Codling, G.; Vogt, A.; Jones, P. D.; Wang, T. Y.; Wang, P.; Lu, Y.L.; Corcoran, M.; Bonina, S.; Li, A.; Sturchio, N. C.; Rockne, K. J.; Ji, K.;Khim, J. S.; Naile, J. E.; Giesy, J. P. Chemosphere 2014, 114, 203−209.(26) Ruan, T.; Wang, Y. W.; Zhang, Q. H.; Ding, L.; Wang, P.; Qu, G.B.; Wang, C.; Wang, T.; Jiang, G. B. J. Chromatogr. A 2010, 1217 (26),4439−4447.(27) Zhang, H. F.; Zhang, Y. H.; Shi, Q.; Zheng, H. D.; Yang, M.Environ. Sci. Technol. 2014, 48 (6), 3112−3119.(28) Kind, T.; Fiehn, O. BMC Bioinf. 2007, 8, 105.(29)Wan, Y.; Hu, J. Y.; Zhang, K.; An, L. H. Environ. Sci. Technol. 2008,42 (4), 1078−1083.(30) Stapleton, H. M.; Kelly, S. M.; Allen, J. G.; McClean, M. D.;Webster, T. F. Environ. Sci. Technol. 2008, 42 (9), 3329−3334.(31) Kato, Y.; Okada, S.; Atobe, K.; Endo, T.; Matsubara, F.; Oguma,T.; Haraguchi, K. Anal. Chem. 2009, 81 (14), 5942−5948.(32) Bond, N. J.; Shliaha, P. V.; Lilley, K. S.; Gatto, L. J. Proteome Res.2013, 12 (6), 2340−2353.(33) Tsou, C. C.; Avtonomov, D.; Larsen, B.; Tucholska, M.; Choi, H.;Gingras, A. C.; Nesvizhskii, A. I. Nat. Methods 2015, 12, 258.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10245

(34) Cortes-Francisco, N.; Caixach, J. Environ. Sci. Technol. 2013, 47(17), 9619−9627.(35) Suarez-Castillo, O. R.; Beiza-Granados, L.; Melendez-Rodriguez,M.; Alvarez-Hernandez, A.; Morales-Rios, M. S.; Joseph-Nathan, P. J.Nat. Prod. 2006, 69 (11), 1596−1600.(36) Agarwal, V.; El Gamal, A. A.; Yamanaka, K.; Poth, D.; Kersten, R.D.; Schorn, M.; Allen, E. E.; Moore, B. S.Nat. Chem. Biol. 2014, 10, 640.(37) Zhu, L. Y.; Hites, R. A. Environ. Sci. Technol. 2005, 39 (24), 9446−9451.(38) Guo, J. H.; Chen, D.; Potter, D.; Rockne, K. J.; Sturchio, N. C.;Giesy, J. P.; Li, A. Environ. Sci. Technol. 2014, 48 (21), 12807−12815.(39) Zhou, S. N.; Reiner, E. J.; Marvin, C.; Helm, P.; Riddell, N.;Dorman, F.; Misselwitz, M.; Shen, L.; Crozier, P.; MacPherson, K.;Brindle, I. D. Anal. Bioanal. Chem. 2010, 396 (3), 1311−1320.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.5b01435Anal. Chem. 2015, 87, 10237−10246

10246

1

Untargeted Identification of Organo-bromine Compounds in Lake Sediments by Ultra-1

High Resolution Mass Spectrometry with Data-Independent Precursor Isolation and 2

Characteristic Fragment (DIPIC-Frag) Method 3

4

Hui PENG, Chunli CHEN, David M.V. Saunders, Jianxian SUN, Song TANG, Garry Codling, 5

Markus Hecker, Steve Wiseman, Paul D. Jones, An Li, Karl J. Rockne, John. P. Giesy

6

7

8

Tables 1 9

Figures 9 10

11

This supporting information provides text, figures and tables addressing (1) Calculation 12

of isotopic peaks intensities distribution; (2) Information on 25 model NSBCs; (3) Sampling 13

map; (4) Product ion spectra of OH-TBB; (5) Scanning scheme of the DIPIC-Frag method; 14

(6) workflow to identify precursor ions of NSBCs compounds; (7) distribution of isotopic 15

peaks of NSBCs; (8) isotopic peaks of bromine fragment would help to identify bromine 16

number and precursor ions; (9) Validation of bromophenol; (10) Validation of bromoindole; 17

(11) validation of tetrabromocarbazole. 18

2

Isotopic Peaks Intensities Distribution. Intensities of isotopic peaks for precursor ions and Br

fragments were compared semi-quantitatively to further confirm the identities of precursor ions.

Since the relative abundances of chlorine and bromine isotopes were much greater than those of

other elements, only isotopic peaks of bromine and chlorine were considered for semi-

quantitative calculation of isotopic peak. For a given NSOBC with formula CxHyOzNiClnBrm, (x,

y, z, and i are usually not available), isotopic peaks of the compound were assumed to have a

known distribution which follow Pascal’s triangle (Equation 1).

( ) ( )m na b c d+ + (1)

Where: a = 0.51 and b = 0.49 are the relative abundances of Br isotopes 79

Br (m/z=78.9183) and

81Br (m/z=80.9163), respectively. m indicates the number of bromine contained in the compound.

c = 0.76 and d = 0.24 are the relative abundances of Cl isotopes 35

Cl (m/z=34.9689) and 37

Cl

(m/z=36.9659), respectively. n indicates the number of chlorine contained in the compound.

Based on the binomial distribution of the isotopic peaks from Equation 1, the relative

abundance of each isotopic peak (with k 79

Br and j 35

Cl) to total abundance could be calculated

as Equation 2.

! !

! !

k m k j n j

i

m nAbundance a b c d

k j

− −= × (2)

Where: Abundancei indicates the relative abundance of the ith

isotopic peak of the compound, k

is the number of 79

Br and j is the number of 35

Cl in the isotopic peak.

Since the isotopic peaks of bromine fragments were also useful information for precursor

ion alignment, the relative abundance of the peak for monoisotopic 79

Br at m/z=78.9183 was

calculated for each isotopic peak by multiplying the relative abundance of the monoisotopic peak

(Equation 2).

3

Br =Abundance ( 1) 2 /i i k m× + × (3)

Where Bri is the relative abundance of 79

Br from the isotopic peak of the compound, (k+1)×2/m

is the relative proportion of 79

Br in the isotopic peak to total 79

Br. Based on Equation 3, the

relative abundance of 79

Br from each isotopic peak could be calculated.

Intensities of precursor ion and bromine fragments ions. Since the relative abundances of

precursor ion candidates and bromine fragments are also important information for precursor ion

alignment, the threshold of the ratio between abundances of precursor ions (indicated by peak

intensities in the present study) and the 79

Br fragment at m/z=78.9183 was predicted. Intensity of

the precursor ions should be greater than that of product ions divided by fragment number

(Equation 4).

precursor productIntensity >Intensity / m (4)

Where: Intensityprecursor is the intensity of precursor ions, Intensityproduct is the intensity of product

ions, m is the number of product ion fragments contained in the formulae (the number of Br atom

in the NSOBC).

Because the bromine fragments were monitored from a 5-m/z precursor isolation window,

which might contain up to 3 isotopic precursor ions for the same NSOBC, the summed

intensities of isotopic precursor ions and bromine fragment were predicted according to Equation

5.

precursor bromine

k=1,2,3

Intensity > Intensity / Intensity /i m m=∑ ∑ (5)

Where: ∑Intensityprecursor is total intensity of isotopic precursor ions from the same NSOBC in

the precursor isolation window. Intensityi is intensity of Br fragment ions from the ith

isotopic

4

precursor ion, which was expected to be less than the intensity of the corresponding precursor

ion, and Intensitybromine is the summed intensity of total Br fragments detected in the

corresponding precursor isolation window. Because the precursor isolation window was 5-m/z,

which allowed at most 3 isotopic peaks for brominated compounds, the intensity of the maximal

precursor ion in the window should be greater than one third of the summed intensity of all

isotopic peaks of precursor ions (Equation 6).

max bromineIntensity >Intensity / (3 )m× (6)

Where: Intensitymax is the intensity of the most abundant isotopic precursor ion peak in the

isolation window.

5

Table S1. Name, molecular weight, KOW value, and instrument sensitivity of 25 model

brominated chemicals.

Compounds Formula MW Log KOWa IDLs (µg/L)

HBCD-α C12H18Br6 635.6509 5.07 0.2

HBCD-β C12H18Br6 635.6509 5.12 0.3

HBCD-γ C12H18Br6 635.6509 5.47 0.1

TBBPA C15H12Br4O2 539.7571 4.5 0.05

TBB C15H18Br4O2 545.8040 8.8 2.0

OH-TBB C15H19Br3O3 483.8884 - 0.005

OH-TBPH C24H35Br3O5 640.0035 9.56 0.01

TBPH C24H34Br4O4 701.9191 11.95 5.0

6-OH-BDE-47 C12H6O2Br4 497.7101 6.4 0.05

4’-OH-BDE-49 C12H6O2Br4 497.7101 6.4 0.05

2-OH-BDE-123 C12H5O2Br5 575.6206 7.2 0.05

6-MeO-BDE-47 C13H8O2Br4 511.7258 7.3 0.3

4’-MeO-BDE-49 C13H8O2Br4 511.7258 7.3 0.3

4’-MeO-BDE-99 C13H7O2Br5 589.6363 8.2 0.3

BDE-47 C12H6Br4O 481.7152 6.8 3.0

BDE-49 C12H6Br4O 481.7152 6.8 4.0

BDE-66 C12H6Br4O 481.7152 6.8 2.0

BDE-85 C12H5Br5O 559.6257 7.7 0.8

BDE-99 C12H5Br5O 559.6257 7.7 0.9

BDE-100 C12H5Br5O 559.6257 7.7 2.0

BDE-153 C12H4Br6O 637.5362 8.6 2.0

BDE-154 C12H4Br6O 637.5362 8.6 3.0

BDE-183 C12H3Br7O 715.4467 9.4 5.0

BDE-209 C12Br10O 949.1783 12.1 20

a KOW values were from references

1-5

6

Figure S1. Sampling locations of two sediment samples (sed-32 and sed-44) from Lake

Michigan.

7

Figure S2. Product ion of a bromine fragment from brominated compounds (hydroxylated TBB

in this sample) under relatively high collision energy (>30eV).

8

Figure S3. Scheme for data independent precursor isolation and characteristic fragment (DIPIC-

Frag) method. Nine different methods (Method1 - Method 9) were performed for a single

sample, each method covered a mass range of 100 Da. For each method, the full scan was used

for each 7 cycles, and then 6 following successive data independent isolation (DIA) windows (5

m/z) was scanned. 20 DIA window was used for each method to cover the 100 Da mass range.

Stepped collision energy at 10, 30 and 60 eV was used for the DIA scanning to simultaneously

record information on bromine fragment and precursor ions in the same mass spectra.

9

Figure S4. Workflow to identify precursor ions of NSOBCs. The first step is to get the

chromatographic profiles of candidate ions from precursor ion regions in separated precursor

isolation DIA windows (there are typically 20-30 ions). If we detected an ion with the same

chromatographic profile with the bromine fragment, we further used full scan spectra to get more

accurate m/z values. If we could not detect precursor ion in the precursor ion region, we

searched precursor ion from full scan spectra. For the potential precursor ions detected in full

scan or precursor ion region, we calculated the intensity of the ions to further make sure that

intensities of precursor ions were greater than that of the fragment divided by number of atom.

Then, we also checked the isotopic peaks of the precursor ions. If the potential precursor ions’

intensities were low or no isotopic peaks were detected, we moved to the next candidate ions in

the precursor ion region for the next round of data analysis. Finally, the list of likely precursor

ions was produced.

11

Figure S5. Distribution of isotopic peaks of brominated compounds with different compositions

of bromine/chlorine. Y-axis indicated the relative intensities of the isotopic peaks to the maximal

intensity of the peak.

12

Figure S6. Isotopic peaks of bromine fragment in separated DIA windows could help to identify

bromine numbers and molecular ions of brominated compounds. As shown in the bottom figure,

only the bromine ion at m/z=78.9171 was detected in the window at 350±2.5 m/z, which meant

that the primary monoisotopic ion of the compounds should be between 350-352 m/z. If the

molecular ion was lower than 350 m/z, we should have observed the isotopic peaks of bromine at

m/z=80.9151 because the mass span of brominated compounds is ~2 m/z. As shown in the top

figure, two isotopic peaks were observed at m/z=353.8959 and m/z=355.8939 respectively. In

this figure, isotopic peak of bromine at m/z=80.9151 were detected at similar intensity to the ion

at m/z=78.9171. By combining the isotopic peaks of the bromine fragment in different precursor

isolation windows and the distribution of the isotopic peaks of the precursor ions, the compound

was identified as C13H8NOBr2 with mass error of 1.1 ppm and a monoisotopic molecular ion at

m/z=351.8977.

13

Figure S7. Validation of an identified bromophenol by use of the commercially available

standard 4-bromophenol. (a) Extracted ion chromatogram at m/z=170.9445 (10 ppm mass width)

for a standard of 4-bromophenol (100 µg/L). (b) MS2 spectra for 4-bromophenol from a 170±2.5

DIA window. (c) Extracted ion chromatogram at m/z=170.9445 (10 ppm mass width) from

sediment extract. (d) MS2 spectra for a bromophenol peak in sediment extract from a 170±2.5

DIA window.

14

Figure S8. Validation of an identified bromoindole by use of commercially available standard of

5-bromoindole. (a) Extracted ion chromatogram at m/z=193.9605 (10 ppm mass width) for a

standard of 5-bromoindole (100 µg/L). (b) MS2 spectra for 5-bromoindole from a 195±2.5 DIA

window. (c) Extracted ion chromatogram at m/z=193.9605 (10 ppm mass width) from sediment

extract. (d) MS2 spectra for a 5-bromoindole peak in sediment extract from a 195±2.5 DIA

window.

15

Figure S9. Validation of an identified brominated carbazole by use of the commercially available

standard of 1,3,6,8-tetrabromocarbazole. (a) Extracted ion chromatogram at m/z=477.7077 (10

ppm mass width) for standard of 1,3,6,8-tetrabromocarbazole (100 µg/L). (b) MS2 spectra for

1,3,6,8-tetrabromocarbazole from a 480±2.5 DIA window. (c) Extracted ion chromatogram at

m/z=477.7077 (10 ppm mass width) from sediment extract. NSOBC at 13.94 min was an isomer

of 1,3,6,8-tetrabromocarbazole with different bromine positions on the aromatic ring. (d) MS2

spectra for a tetrabromocarbazole peak in sediment extract from 480±2.5 DIA window.

16

References

1. He, M. J.; Luo, X. J.; Yu, L. H.; Wu, J. P.; Chen, S. J.; Mai, B. X., Diasteroisomer and

enantiomer-specific profiles of hexabromocyclododecane and tetrabromobisphenol A in an

aquatic environment in a highly industrialized area, South China: Vertical profile, phase partition,

and bioaccumulation. Environ. Pollut. 2013, 179, 105-110.

2. Kelly, B. C.; Ikonomou, M. G.; Blair, J. D.; Gobas, F., Hydroxylated and methoxylated

polybrominated diphenyl ethers in a Canadian Arctic marine food web. Environ. Sci. Technol.

2008, 42 (19), 7069-7077.

3. Wu, J. P.; Guan, Y. T.; Zhang, Y.; Luo, X. J.; Zhi, H.; Chen, S. J.; Mai, B. X., Several

current-use, non-PBDE brominated flame retardants are highly bioaccumulative: Evidence from

field determined bioaccumulation factors. Environ. Int. 2011, 37 (1), 210-215.

4. Saunders, D. M. V.; Higley, E. B.; Hecker, M.; Mankidy, R.; Giesy, J. P., In vitro

endocrine disruption and TCDD-like effects of three novel brominated flame retardants: TBPH,

TBB, & TBCO. Toxicology Letters 2013, 223 (2), 252-259.

5. Peng, H.; Saunders, D. M. V.; Sun, J. X.; Garry, C.; Wiseman, S.; Jones, P. D.; Giesy, J.

P., Detection, identification, and quantification of hydroxylated bis(2-ethylhexyl)-

tetrabromophthalate isomers in house dust. Environ. Sci. Technol. 2015, 49, 2999-3006.