Untargeted Screening and Distribution of Organo-IodineCompounds in Sediments from Lake Michigan and the Arctic OceanHui Peng,*,† Chunli Chen,†,‡ Jenna Cantin,† David M. V. Saunders,† Jianxian Sun,† Song Tang,§

Garry Codling,† Markus Hecker,†,§ Steve Wiseman,† Paul D. Jones,†,§ An Li,∇ Karl J. Rockne,○

Neil C. Sturchio,◆ Minghong Cai,*,¶ and John P. Giesy*,†,∥,⊥,#

†Toxicology Centre, University of Saskatchewan, 44 Campus Drive, Saskatoon, Saskatchewan S7N 5B3, Canada‡Key Laboratory of Poyang Lake Environment and Resource Utilization of MOE, School of Resources, Environmental and ChemicalEngineering, Nanchang University, Nanchang 330047, People’s Republic of China§School of Environment and Sustainability, University of Saskatchewan, 117 Science Place, Saskatoon, Saskatchewan S7N 5C8,Canada∥Zoology Department, Center for Integrative Toxicology, Michigan State University, 1129 Farm Lane Road, East Lansing, Michigan48824, United 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 210093,People’s Republic of China∇School of Public Health, University of Illinois at Chicago, Chicago, Illinois 60612, United States○Department of Civil and Materials Engineering, University of Illinois at Chicago, 842 West Taylor Street, Chicago, Illinois 60607,United States◆Department of Geological Sciences, University of Delaware, 255 Academy Street, Newark, Delaware 19716, United States¶SOA Key Laboratory for Polar Science, Polar Research Institute of China, Shanghai 200136, China

*S Supporting Information

ABSTRACT: The majority of halogenated organic compounds present in theenvironment remain unidentified. To address this data gap, we recentlydeveloped an untargeted method (data-independent precursor isolation andcharacteristic fragment; DIPIC-Frag) for identification of unknown organo-bromine compounds. In this study, the method was adapted to enableuntargeted screening of natural and synthetic organo-iodine compounds(NSOICs) in sediments. A total of 4,238 NSOIC peaks were detected insediments from Lake Michigan. Precursor ions and formulas were determinedfor 2,991 (71%) of the NSOIC peaks. These compounds exhibited variationsin abundances (<103 to ∼107), m/z values (206.9304−996.9474), retentiontimes (1.0−29.7 min), and number of iodine atoms (1−4). Hierarchical clusteranalysis showed that sediments in closer proximity exhibited similar profiles ofNSOICs. NSOICs were screened in 10 samples of sediment from the ArcticOcean to compare the profiles of NSOICs between freshwater and marine sediments. A total of 3,168 NSOIC peaks weredetected, and profiles of NSOICs in marine sediments were clearly distinct from Lake Michigan. The coexistence of brominatedand iodinated analogues indicated that some NSOICs are of natural origin. Different ratios of abundances of iodinatedcompounds to brominated analogues were observed and proposed as a marker to distinguish sources of NSOICs.

■ INTRODUCTION

The presence of natural and synthetic halogenated organiccompounds (NSHOCs) in the environment is of specialconcern due to their persistence in various matrices,bioaccumulation, and toxic potencies. Targeted monitoring ofions for known chemicals is currently the main strategy toidentify and quantify NSHOCs, such as perfluorinatedcompounds (PFCs), polybrominated diphenyl ethers(PBDEs), and polychlorinated biphenyl (PCBs), in environ-

mental matrices.1−4 However, mass balance analysis byquantification of total, organically bound halogens revealedthe existence of unidentified NSHOCs in the environment.5,6

Specifically, previously unidentified organo-bromine com-

Received: June 28, 2016Revised: August 21, 2016Accepted: August 29, 2016Published: September 9, 2016

Article

pubs.acs.org/est

© 2016 American Chemical Society 10097 DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

pounds contributed >99.9% of total organic bromine in wildlifeand sediment.5 Thus, identification of these unknownNSHOCs is important for assessment of their potential healtheffects on humans, as well as their potential ecological risks.While chlorinated and brominated compounds have received

much attention, iodinated compounds have been much lessstudied. Despite the limited information on occurrence ofiodinated compounds, three potential sources might beexpected for iodinated compounds. Similar biosynthetic routesshared between iodinated and other halogenated compounds inmicroorganisms7,8 indicated the potential presence of naturallyoccurring iodinated analogues of previously reported bromi-nated/chlorinated compounds, e.g., iodinated PBDEs havebeen recently identified as off-pathway byproducts of PBDEs inmicrobes.9 In addition to potential natural production, organo-iodine compounds have been used since the mid-19th centuryfor various industrial purposes.10 Ioxynil, a hydroxybenzonitrileherbicide, has been widely used for weed control.11 As the thirdsource, the disinfection process can produce iodinateddisinfection byproducts (I-DBPs) in the presence of iodide.12,13

Such multiple sources indicated potential occurrences oforgano-iodine compounds in the environment; indeed, aprevious study has found a relatively large abundance, in theμg/g range, of total organic iodine in environmental matrices.6

However, limited information was available on the environ-mental occurrence of iodinated compounds, except for a fewstudies which reported the existence of iodinated carbazolesand PBDEs.9,14 Previous studies have widely documented thegreater toxicity of organo-iodine compounds than theirbrominated or chlorinated analogues;11,15−17 thus, identifica-tion of unknown organo-iodine compounds would be critical tomore completely understand the occurrence of halogenatedcompounds and the potential risks posed to humans andwildlife.Untargeted screening of iodinated compounds is more

challenging than chlorinated and brominated compounds.Multiple stable isotopes of Cl and Br atoms provide diagnosticinformation for identification of chlorinated and brominatedcompounds.18,19 However, iodine has a single naturallyoccurring stable isotope; thus, mass spectrometric detectionof iodinated compounds cannot rely on the use of diagnosticprofiles of isotopes of iodine. This might affect specificity ofanalyses and give rise to a false positive rate. In a recent study,approximately 2,000 novel natural and synthetic organo-bromine compounds (NSOBCs) were identified in sedimentsfrom Lake Michigan, using an untargeted, data independentprecursor isolation and characteristic fragment (DIPIC-Frag)method.20,21 DIPIC-Frag is a “bottom up” method that firstdetects corresponding peaks with characteristic product ionsand then uses narrow precursor ion isolation windows and achemometric strategy to prospectively evaluate MS1 spectra toidentify precursor ions and calculate chemical formulas. Inprinciple, the DIPIC-Frag method can be used for any organiccompounds with characteristic ions in LC-MS/MS analysis.Thus, the DIPIC-Frag method can be readily extended toscreen other compounds.Biosynthesis of halogenated compounds by marine micro-

organisms has been reported previously,8 but it remainedunclear whether such processes also occurred in freshwaterenvironment. Results of a recent study demonstrated theexistence of numerous organo-bromine compounds in fresh-water sediments from Lake Michigan.21 Thus, the aim of thisstudy was to identify previously unknown natural and synthetic

organo-iodine compounds (NSOICs) in freshwater sedimentsand compare the results to those of marine sediments. Toanalyze multiplexed data sets and increase specificity of theDIPIC-Frag method, a novel data mining strategy wasdeveloped to predict formulas of NSOICs and establish amass spectrometric library of NSOICs. Then, building upon theestablished library, distributions of NSOICs were investigatedin 23 samples of sediments from Lake Michigan and 10 samplesof sediments from the Arctic Ocean.

■ MATERIALS AND METHODSChemicals and Reagents. Iodoacetic acid and triiodo-L-

thyronine were purchased from Sigma-Aldrich Chemical Co.(St. Louis, MO, USA). Florisil (6 cc, 1 g, 30 μm) solid-phaseextraction (SPE) cartridges were purchased from Waters(Milford, MA, USA). Methyl tert-butyl ether (MTBE),dichloromethane (DCM), hexane, methanol, and acetonewere all of omni-Solv grade and were purchased from EMDChemicals (Gibbstown, NJ, USA).

Collection of Sediments. A total of 23 freshwatersediment samples were collected from Lake Michigan inSeptember 2010 as described previously.22 A total of 10 Arcticmarine sediment samples were collected onboard the ice-breaker R/V Xuelong (Snow Dragon) during the fifth ChineseNational Arctic Research Expedition between July andSeptember 2012.23 Samples were taken from the Bering Seaacross the Bering Strait to the Chukchi Sea and then throughthe Northeast Passage to the Amundsen Basin and directly toIceland (54−88°N). 0−2 cm marine sediments were collectedwith a stainless steel grab sampler and scooped using aprecleaned stainless steel scoop and placed into baked andsolvent-rinsed aluminum containers.23 All samples were storedin the dark at −20 °C and shipped on ice in coolers to thelaboratory. Each sample was lyophilized and homogenized byhand before being passed through a 1 mm sieve to removestones and large organic matter.

Sample Pretreatment and LC-Q Exactive Analysis.Approximately 10 g of freeze-dried samples of sediments wasextracted by use of a previously described method thatemployed accelerated solvent extraction and an additionalsolid phase extraction (SPE) cleanup method to removeinterferences.20,21 The use of a simple one-step SPE method(Florisil cartridges) for cleanup was mainly because it couldefficiently remove colored interferences from extracts ofsediments and avoid potential loss of compounds. Aliquots ofsediment extracts were analyzed using a Q Exactive UHRMS(Thermo Fisher Scientific, San Jose, CA, USA) equipped with aDionex UltiMate 3000 UHPLC system (Thermo FisherScientific). A Hypersil GOLD C18 column was used forseparation of NSOICs considering its good separation abilityand sensitivity. Data were acquired in data-independentacquisition (DIA) mode and atmospheric pressure chemicalionization (APCI) ionization source. More details of samplepretreatment and mass spectrometry analysis are provided inthe Supporting Information (SI).A novel chemometric strategy was developed to expand the

detected number of NSOICs and reduce the false positive rateof predicted precursor ions and compound formulas, which aredescribed in the SI.

Quality Control and Assurance. To avoid contaminationof samples, all equipment was rinsed regularly with acetone.One procedural blank (without sediment) was incorporated inthe analytical procedure for every batch of samples. Thirteen

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10098

NSOIC peaks were detected in the blank, partly due toinstrument carryover or contamination during sample pretreat-ment. Background contamination from blanks was subtractedfrom samples for downstream data analysis, and those NSOICswhose abundance was less than three times the backgroundabundance in blanks were considered nondetected.Because most identified NSOICs were novel compounds for

which no authentic standards were available, peak intensitieswere used to semiquantify their abundances in sedimentsamples, as was done previously for NSOBCs.20,21 Methoddetection limits (MDLs) could not be calculated for NSOICs,but a peak intensity cutoff of 1000 was incorporated into theDIPIC-Frag method as described previously21 and used as theMDL for the NSOICs. For the 13 NSOIC peaks detected inblanks, 3× the peak abundance were used as MDLs.Data Treatment and Statistical Analyses. All data

analysis including Pearson correlation, linear regression, andcluster analysis were performed with an in-house R program.For those results that were less than the MDLs, half of theMDL (peak abundance of 500) was assigned to avoid missingvalues in the statistical analysis. Only NSOICs with detectionfrequencies greater than 50% were used for correlation,regression, and cluster analysis. Statistical significance wasdefined as p < 0.05.

■ RESULTS AND DISCUSSION

Workflow of the DIPIC-Frag Method for NSOICs. Useof the DIPIC-Frag method for untargeted screening of NSOICswas inspired by the observation of the fragment peak of iodine(m/z = 126.9045) when mass spectra were evaluated forNSOBCs in Lake Michigan sediment extracts (Figure S1). Forexample, the mixed halogenated carbazole C12H5NBr2I(brominated and iodinated analogues are shown in Figure 1aand 1b) was identified when the fragment of iodine was clearlydetected in the DIA window with similar retention times (rt =10.61 min) as the brominated analogue C12H5NBr3 (rt = 10.76min). Due to the weaker energy of the C−I bond compared tothe C−Br bond, the C−I bond is likely to be preferentiallycleaved in negative ionization mode at relatively high HCDenergies. To test this, a known, natural iodinated compound,triiodo-L-thyronine (T3), and a drinking water disinfectionbyproduct, iodoacetic acid, were analyzed. Both of thesecompounds exhibited an abundant fragment of iodine underrelatively high HCD collisional energies (50 eV) (Figure S2).These results demonstrated that the DIPIC-Frag method andinformation on characteristic peak of iodine could be used foruntargeted identification of NSOICs in the environment.Furthermore, because of the high-resolution of the MS2

spectra, there was no theoretical interference ion at m/z =126.9045, within a 10 ppm mass tolerance, which confirmed thereliability of using the characteristic peak of iodine for detectionof NSOICs.

Figure 1. Typical workflow of the DIPIC-Frag method to identify NSOICs. (a) Bromine fragment peak of brominated carbazole. (b) Iodinefragment peak of iodinated carbazole. (c) 180 successive 5-m/z-width precursor isolation windows were used. (d) Iodine fragment peaks weredetected in each of the data-independent acquisition (DIA) windows. (e) Mass spectra in separated DIA windows at the same retention time asiodine fragment peaks for precursor ion alignment. (f) The chromatographic elution profiles of iodine fragments (top) and precursor ion candidates(bottom). (g) Isotopic peaks of precursor ions were used to predict formulas of precursor ions. (h) A homologue model was used based on apreviously established NSOBCs library. (i) Formulas of the proposed precursor ions were predicted by combining multiple lines of evidence andcalculated scores. (j) The final library containing multiple NSOICs was established with predicted formula, exact mass, retention time, and peakintensity.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10099

Similar to the previous DIPIC-Frag method settings forscreening NSOBCs,20,21 180 successive 5-m/z DIA windowswere used to expand the dynamic range and reduce the falsepositive rate of the method (Figure 1c). When a peakabundance cutoff of 1,000 was used, 4,238 peaks were detectedas NSOICs in 23 Lake Michigan sediments (typical chromato-gram in Figure 1d). The number of NSOIC peaks was greaterthan that of NSOBCs (2,520).20 This was partly due to the useof multiple samples for construction of the NSOIC library(details of the chemometric strategy are provided in the SI), incontrast to two pilot samples for the NSOBC library. NSOICswere distributed across multiple DIA windows and detected in152 of the 180 DIA windows with retention times ranging from1.0 to 29.7 min, which indicated a diversity of NSOICs. Byaveraging abundances of peaks among the 23 Lake Michigansediments, abundances of NSOIC peaks ranged from <103 to∼3 × 106 (Figure 2a). While previous studies have reportedmore than 5,000 chlorinated and brominated compounds in amarine environment,24 to the best of our knowledge, this is thefirst study to provide direct experimental evidence of theexistence of thousands of NSOICs in the environment. Tofurther confirm the presence of high-abundance NSOICs inLake Michigan sediments, concentrations of total organiciodine in three sediments were determined by use ofinstrumental neutron activation analysis (INAA). The totalorganic iodine was 0.21, 0.58, and 0.36 μg/g dry mass (dm) insed-93A, sed-41, and sed-10, respectively, which was 2−10-foldless than total organic bromine (0.8, 1.4, and 3.3 μg/g dm insed-93A, sed-41, and sed-10) in the same samples. Theseresults confirmed the large abundances of NSOICs in LakeMichigan sediments.Similar untargeted screening was performed on 10 marine

sediments from the Arctic Ocean. 3,168 peaks were detected asNSOICs, and abundances of NSOICs in these marinesediments (Figure 2c) were approximately 10-fold less thanthose from Lake Michigan. Although previous studies havereported production of natural chlorinated and brominatedcompounds by marine microorganisms,8,25,26 results of the

present study demonstrated occurrences of numerous NSOICsin both marine and freshwater sediments.

Library of NSOICs in Sediments from Lake Michiganand the Arctic Ocean. Compared to brominated compounds,alignment of precursor ions and prediction of compoundformulas of iodinated compounds was more challenging due tothe lack of multiple stable isotopes of iodine. Based on a newlydeveloped chemometric strategy which recruited exact mass,homologue model, isotopic distribution, and chromatographicelution profiles (Figure 1f-h; further details are provided in theSI), precursor ions were identified and molecular formulas werecalculated for 2,991 NSOICs detected in Lake Michigan(corresponding to 1,976 unique NSOICs after excludingisotopic peaks, and the database is provided in the SupportingInformation), which represented 71% of the 4,238 originallyidentified NSOIC peaks. Most of the unidentified NSOICs hadsmall peak intensities (<104), and precursor ions could not bedetected.Detected NSOICs in Lake Michigan exhibited large

variations in m/z values (206.9304−996.9474) and numbersof I atoms (1−4) (Figure 2b). By searching predicted chemicalformulas against the public database, Chemspider, fewer than200 of the NSOICs were identified, which indicated most of theNSOICs observed in this study had not been previouslyreported. Indeed, previous studies have reported a databasewith more than 2,200 brominated compounds,27 but no largedatabase exists for NSOICs.Of the 3,168 NSOICs detected in sediments from the Arctic

Ocean, precursor ions were identified and molecular formulaswere calculated for 2,243 NSOICs (70%). NSOICs frommarine sediments exhibited greater diversity as indicated by thegreater range of their retention times (Figure 2d), compared toNSOICs detected in sediments from Lake Michigan, whichwere eluted within a relatively narrow range of retention times,between 9 and 20 min. Consistent with this finding, whenextracted with exact mass and retention time, only 387 of theNSOICs detected in marine sediments (17.3% of NSOICs)were detected in Lake Michigan sediments. The largest class ofhigh-abundant NSOICs, in addition to iodophenol and

Figure 2. Distribution of identified NSOICs. (a) Distribution of peak abundances of NSOICs in sediments from Lake Michigan. (b) Distribution ofNSOICs by retention time and m/z values, in sediments from Lake Michigan. (c) Distribution of peak abundances of NSOICs in sediments from theArctic Ocean. NSOICs ID in parts c and a is separated ID, which was based on the order of peak abundance. (d) Distribution of NSOICs byretention time and m/z values, in sediments from the Arctic Ocean. Sizes of the dots are proportional to peak abundances of NSOICs. Colors of dotsrepresent numbers of iodine atoms.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10100

iodoindole, detected in marine sediments, was predicted to belong-chain iodophenols (Figure 2C), but they were seldomdetected in sediments from Lake Michigan. A heatmap of the387 NSOICs showed distinct profiles between marine andfreshwater sediments, and hierarchical clustering showed thatmarine and freshwater sediments were clearly separated intotwo distinct groups (Figure S3). These results indicateddifferent sources of NSOICs between the marine andfreshwater environments, and future studies are warranted toidentify the sources of NSOICs especially in a freshwaterenvironment.Not surprisingly, for the three previously validated model

NSOBCs,21 bromophenol, bromoindole, and bromocarbazole,iodinated analogues were all identified in both the freshwaterand marine sediments (chromatograms are shown in FigureS4), although their relative abundance was different betweenmarine and freshwater sediments (Figure 2a and 2c). Due tothe larger size of iodine relative to that of bromine, retentiontimes of iodophenol and iodoindole on a C18 column were∼0.3 min longer (eluted later) than their Br-containinganalogues, which was consistent with a previous study ofhalogenated chemicals.28 However, retention times of somebromoiodocarbazoles were 0.3 min shorter than the Br-containing analogues (for example, C12H4NBr3I, rt = 13.73min vs C12H4NBr4, rt = 14.06 min; and C12H5NBr2I, rt = 10.61min, vs C12H5NBr3, rt = 10.76 min). These results could beattributed to differences in the position of Br/I in the aromaticring. Recent studies have also reported multiple mixedhalogenated carbazoles in sediments by GC-MS,14,29,30 butdetermination of exact structures of these analogues requiresfurther studies. Consistent with data for the three modelNSOBCs,21 their I-containing analogues also exhibitedrelatively large peak abundances. Particularly, iodinatedcarbazoles were some of the most abundant NSOICs detectedin Lake Michigan (Figure 2a). Due to the lack of commerciallyavailable authentic standards for these novel NSOICs, the threeiodinated phenols, indoles, and carbazoles were used as modelchemicals in subsequent analysis.Spatial Distributions of NSOICs in Lake Michigan

Sediments. Profiles of NSOICs were similar in Lake Michigansediments collected in close proximity. For example, significantcorrelations of abundances of NSOICs were observed betweensed-93A and sed-93B (r = 0.91, p < 0.001) (Figure 3a).However, abundances of NSOICs in sediments collected atgreater distances were frequently poorly correlated. Forexample, abundances of NSOICs in sed-93A exhibited poorcorrelations with those in sed-88 (r = 0.03, p = 0.04) (Figure3b). Poor correlations among sediments indicated greatvariations of NSOICs among surficial sediments collectedfrom the 23 locations, and most compounds exhibited ranges of2 to 3 orders of magnitude in peak abundances. Similarity heatmaps and cluster analyses showed that sediments were groupedinto three clusters according to their Pearson correlation matrix(Figures 4c and 4d). Nearshore sediments from the northernbasin of Lake Michigan (sed-120, sed-93A, sed-93B, sed-113,sed-50, sed-93C, and sed-103) were clustered into the largestgroup (r2 values ranged from 0.31 to 0.83, p < 0.001); offshoresediments from the northern basin of Lake Michigan (sed-48,sed-41, sed-83, sed-47, sed-24, sed-32) also exhibited similarNSOIC profiles, with r2 values ranging from 0.52 to 0.63 (p <0.001); two sediments (sed-2 and sed-10) from the southernbasin were also clustered into a small group and exhibitedstrong correlations (r2 = 0.51, p < 0.001) between abundances

of NSOICs. Such results further indicated that NSOICs havedistinct sources and profiles among locations. Similar patternshave also been observed for NSOBCs in the same samples ofsediments,21 which might be attributed to the presence ofsimilar species of microorganisms at proximal sites that arecapable of producing brominated and iodinated compounds.Analyses of spatial similarities among individual NSOICs

showed distinct spatial distributions of these compounds, butapproximately half of NSOICs were not detected in mostsediments (Figure S5). Most of the high-abundant compounds(especially bromoiodocarbazoles) were clustered, with strongcorrelations, into the same group (bottom left of Figure 5b),which was similar to results for NSOBCs.21

Figure 6 e-h and Figure S6 showed that, similar to thepattern of distribution of NSOBCs, the sites offshore in thenorthern basin exhibited the greatest abundance of totalNSOICs. A similar pattern was also observed for NSOICs inclusters 1 and 2 and three model iodinated chemicals withminor differences (Figure 6 e-h and Figure S6). These resultsindicated similar origins of these brominated and iodinatedcompounds in Lake Michigan. In contrast to the detectedNSOICs and NSOBCs, synthetic compounds such as PFCs andPBDEs were detected at greater concentrations in the southernbasin of the lake, which was located near sites with greaterpopulation densities and industrial activities.22,31 The differentdistribution patterns indicated that anthropogenic sources arenot likely the major sources of most NSOICs and NSOBCs.Previous studies have reported production of iodinatedcompounds by microorganism as byproducts of brominatedcompound biosynthesis.7,8 As such, similar distribution patternsof NSOICs and NSOBCs were consistent with the hypothesisof a common source from natural production.

Spatial Distributions of NSOICs in Sediments from theArctic Ocean. Profiles of NSOICs were similar amongsediments from the Arctic Ocean. This resulted in strongcorrelations between abundances of NSOICs in these sedi-ments. For example, significant correlations of abundances ofNSOICs were observed between BN05 and BN06 (p < 0.001)(Figure 3c), BL03, and BN05 (p < 0.001) (Figure 3d). Suchresults were unexpected, since these locations were relatively

Figure 3. (a) Stronger correlations of abundances of NSOICs amongsediments from Lake Michigan collected in proximity. (b) Weakercorrelations were observed for sediments from Lake Michigancollected further apart. (c-d) Strong correlations of abundances ofNSOICs among sediments from the Arctic Ocean collected inproximity (c) or even in distance (d).

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10101

Figure 4. (a) Sampling sites of 10 marine sediments from the Arctic Ocean. (b) Correlation heatmap and hierarchical clustering of marine sedimentsbased on their Pearson correlation coefficients. (c) Sampling sites of 23 sediments from Lake Michigan. Red, blue, and black dots/circles indicatethree groups of sediments exhibiting similar NSOIC profiles. (d) Correlation heatmap and hierarchical clustering of sediments from Lake Michiganbased on their Pearson correlation coefficients. Red, blue, and black dots/circles indicate three groups of sediments exhibiting similar NSOICprofiles.

Figure 5. Similarity heatmap and hierarchical clustering of NSOICs detected in (a) sediments from the Arctic Ocean and (b) sediments from LakeMichigan. Similarity between NSOICs defined as Pearson correlation coefficients.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10102

distant, compared to the sediments from Lake Michigan.Similarity heat maps and clustering analyses (Figures 4a, 4b,and Figure S5) showed that the pattern of NSOICs insediments from site BL03 was distinct from those of other sites.This might be due to the greater depth of this site (3,613 m)compared to other sites (22.4−223 m). Specifically differentmicrobial communities might exist in the deep sea whichproduced distinct NSOICs. Partly due to the distinct profile ofabundances of NSOIC in sediments from BL03, correlationcoefficients of NSOICs from the Arctic Ocean sediments wererelatively weak compared to Lake Michigan (Figure 5a).Distributions of abundances of NSOICs among 10 marine

sediments were further investigated (Figure 6 a-d). Similarpatterns of distribution were observed for total abundances ofNSOICs and also three model iodinated compounds. Relativelygreat abundances were detected in BN08 and C03. The similarpattern of distributions of NSOICs among sites furtherindicated common sources of these compounds in the marineenvironment.Comparison to Brominated Analogue Compounds.

Significant and positive correlations were observed for two ofthe three model compounds, halogenated phenol (r2 = 0.90, p< 0.001) and halogenated indole (r2 = 0.76, p < 0.001) in Lake

Michigan sediments (Figure S7). These results furtherindicated potential common sources of some iodinated andbrominated compounds, especially for those compounds withanalogues. Further evaluation of the predicted molecularformulas of the 2,991 NSOICs detected in Lake Michiganrevealed that 1,035 (35%) had Cl and 567 (19%) had Br. Theseresults indicated that approximately half of the NSOICs aremixed halogenated compounds, which was consistent withprevious studies that iodinated compounds could be bio-synthesized as byproducts of halogenase enzymes.7

Similar peak abundances were observed for brominated andiodinated analogues of the halogenated phenols and indoles(Figure S8a and S8b). In contrast, abundances of brominatedcarbazole (Log-transformed, 7.49 ± 0.99) were approximately10,000-fold greater than iodinated analogues (Log-transformed,3.30 ± 0.68) (p < 0.001) (Figure S8c). Similar higherabundances of brominated analogues than iodinated analogueswere also observed for halogenated pyrrole, which are producedby a recently identified bacterial enzyme.8 The brominatedpyrrole exhibited 100-fold greater abundances (Log-trans-formed, 4.95 ± 2.04) than its iodinated analogue (Log-transformed, 3.12 ± 2.59) (p < 0.001) (Figure S8d). Thedifferent patterns of halogenation between phenols andpyrroles observed in the present study might be due todifferent halogenation specificities of the enzymes. Bromophe-nol and bromopyrrole were biosynthesized by two differentbrominases, while iodinated byproducts have only beenreported for the phenol brominase,8 which was consistentwith relatively high abundance iodinated phenol in the presentstudy.In the Arctic Ocean, the abundances of iodinated indoles

were about 10-fold greater than those of their brominatedanalogues (p < 0.001), which was different from the lowerabundance of iodinated indoles in sediments from LakeMichigan (Figure S8e). Such results indicated that substrateselectivity of biosynthetic enzymes may vary between marineand freshwater organisms. In addition to the postulation ofdifferences in biosynthetic enzymes, the different ratiosbetween halogenated analogues observed in this study mightbe used as an important marker in future studies to clarifysources of these compounds, as most industrial brominatedchemicals were not likely produced with iodinated or mixedhalogenated analogues.30

Implications. To date, most organohalogens in environ-mental matrices remain unidentified. Detection of ∼4,000NSOICs in sediments of Lake Michigan and ∼3,000 NSOICsin sediments from the Arctic Ocean demonstrated the wideoccurrence of NSOICs in the environment. Although previousstudies have emphasized biosynthesis of halogenated com-pounds in marine organisms, the study presented here reporteda relatively large production of NSOICs in both freshwater andmarine environment, and distinct profiles of NSOICs wereobserved between marine and freshwater environment. Thus,future studies are warranted to clarify sources of NSOICs,especially for the freshwater environment.This study demonstrated that the DIPIC-Frag method was

applicable to detect and identify iodinated compounds withoutdiagnostic isotopic information. In theory, this method couldalso be used for screening chlorinated and perfluorinatedcompounds using characteristic ions (e.g., Cl for chlorinatedcompounds and CF2CF2CF3 for perfluorinated com-pounds),32−34 and future studies are warranted to extend itsapplications. In addition to the natural halogenated compounds

Figure 6. Distribution of NSOICs in sediments from the ArcticOcean: (a) total abundances of NSOICs; (b) iodophenol; (c)iodoindole; (d) iodocarbazole. Distribution of NSOICs in sedimentsfrom Lake Michigan: (e) total abundances of NSOICs; (f)iodophenol; (g) iodoindole; (h) iodocarbazole. Sizes of circles areproportional to abundances of NSOICs.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10103

in sediment reported here, previous studies have highlightedthe potential risk of unknown anthropogenic halogenatedcompounds in drinking water and house dust.13,19,35 Thus,application of the DIPIC-Frag method for untargeted screeningof unknown halogenated compounds in these matrices is ofgreat interest.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.est.6b03221.

Text and Figures S1−S9 (PDF)Table of MS database of iodinated compounds (XLSX)

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: huisci@gmail.com (H.P.)*E-mail: caiminghong@pric.org.cn (M.C.).*Phone (direct): 306-966-2096. Phone (secretary): 306-966-4680. Fax: 306-966-4796. E-mail: jgiesy@aol.com. Correspond-ing author address: Toxicology Centre, University ofSaskatchewan, Saskatoon, Saskatchewan S7N5B3, Canada(J.P.G.).NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis research was part of the Great Lakes SedimentSurveillance Program funded by a Cooperative Agreementfrom the U.S. EPA Great Lakes Restoration Initiative withAssistance #GL-00E00538 (U.S. EPA Program Officer ToddNettesheim), a Discovery Grant from the Natural Science andEngineering Research Council of Canada (Project # 326415-07), and a grant from the Western Economic DiversificationCanada (Projects # 6578, 6807, and 000012711). The ArcticOcean sediment samples were provided by Prof. Cai and thePolar Sediment Sample Chamber of China. C.C.L acknowl-edges financial support from the China Scholarship Council.The authors wish to acknowledge the support of aninstrumentation grant from the Canada Foundation forInfrastructure. Profs. Giesy and Hecker were supported bythe Canada Research Chair program, and Prof. Giesy was alsosupported by the 2014 “High Level Foreign Experts”(#GDT20143200016) program, funded by the State Admin-istration of Foreign Experts Affairs, the P.R. China to NanjingUniversity, and the Einstein Professor Program of the ChineseAcademy of Sciences.

■ REFERENCES(1) Zhu, L. Y.; Hites, R. A. Brominated flame retardants in sedimentcores from lakes Michigan and Erie. Environ. Sci. Technol. 2005, 39(10), 3488−3494.(2) Yang, R. Q.; Wei, H.; Guo, J. H.; Li, A. Emerging brominatedflame retardants in the sediment of the Great Lakes. Environ. Sci.Technol. 2012, 46 (6), 3119−3126.(3) Peng, H.; Wan, Y.; Zhang, K.; Sun, J. X.; Hu, J. Y. Trophictransfer of dechloranes in the marine food web of Liaodong Bay,North China. Environ. Sci. Technol. 2014, 48 (10), 5458−5466.(4) Kwan, C. S.; Takada, H.; Mizukawa, K.; Saha, M.; Rinawati;Santiago, E. C. Sedimentary PBDEs in urban areas of tropical Asiancountries. Mar. Pollut. Bull. 2013, 76 (1−2), 95−105.(5) 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. Contribution of synthetic and naturally occurringorganobromine compounds to bromine mass in marine organisms.Environ. Sci. Technol. 2010, 44 (16), 6068−6073.(6) Kannan, K.; Kawano, M.; Kashima, Y.; Matsui, M.; Giesy, J. P.Extractable organohalogens (EOX) in sediment and biota collected atan estuarine marsh near a former chloralkali facility. Environ. Sci.Technol. 1999, 33 (7), 1004−1008.(7) Blasiak, L. C.; Drennan, C. L. Structural perspective on enzymatichalogenation. Acc. Chem. Res. 2009, 42 (1), 147−155.(8) Agarwal, V.; El Gamal, A. A.; Yamanaka, K.; Poth, D.; Kersten, R.D.; Schorn, M.; Allen, E. E.; Moore, B. S. Biosynthesis ofpolybrominated aromatic organic compounds by marine bacteria.Nat. Chem. Biol. 2014, 10 (8), 640−U182.(9) Agarwal, V.; Li, J.; Rahman, I.; Borgen, M.; Aluwihare, L. I.; Biggs,J. S.; Paul, V. J.; Moore, B. S. Complexity of naturally producedpolybrominated diphenyl ethers revealed via mass spectrometry.Environ. Sci. Technol. 2015, 49 (3), 1339−1346.(10) Stang, P. J.; Zhdankin, V. V. Organic polyvalent iodinecompounds. Chem. Rev. 1996, 96 (3), 1123−1178.(11) Bettiol, C.; De Vettori, S.; Minervini, G.; Zuccon, E.; Marchetto,D.; Ghirardini, A. V.; Argese, E. Assessment of phenolic herbicidetoxicity and mode of action by different assays. Environ. Sci. Pollut. Res.2016, 23 (8), 7398−7408.(12) Plewa, M. J.; Muellner, M. G.; Richardson, S. D.; Fasano, F.;Buettner, K. M.; Woo, Y. T.; McKague, A. B.; Wagner, E. D.Occurrence, synthesis, and mammalian cell cytotoxicity andgenotoxicity of haloacetamides: An emerging class of nitrogenousdrinking water disinfection byproducts. Environ. Sci. Technol. 2008, 42(3), 955−961.(13) Richardson, S. D.; Fasano, F.; Ellington, J. J.; Crumley, F. G.;Buettner, K. M.; Evans, J. J.; Blount, B. C.; Silva, L. K.; Waite, T. J.;Luther, G. W.; McKague, A. B.; Miltner, R. J.; Wagner, E. D.; Plewa,M. J. Occurrence and mammalian cell toxicity of iodinated disinfectionbyproducts in drinking water. Environ. Sci. Technol. 2008, 42 (22),8330−8338.(14) Guo, J. H.; Chen, D.; Potter, D.; Rockne, K. J.; Sturchio, N. C.;Giesy, J. P.; Li, A. Polyhalogenated carbazoles in sediments of LakeMichigan: a new discovery. Environ. Sci. Technol. 2014, 48 (21),12807−12815.(15) Yang, Y.; Komaki, Y.; Kimura, S. Y.; Hu, H. Y.; Wagner, E. D.;Marinas, B. J.; Plewa, M. J. Toxic impact of bromide and iodide ondrinking water disinfected with chlorine or chloramines. Environ. Sci.Technol. 2014, 48 (20), 12362−12369.(16) Plewa, M. J.; Wagner, E. D.; Richardson, S. D.; Thruston, A. D.;Woo, Y. T.; McKague, A. B. Chemical and biological characterizationof newly discovered lodoacid drinking water disinfection byproducts.Environ. Sci. Technol. 2004, 38 (18), 4713−4722.(17) Richardson, S. D.; Plewa, M. J.; Wagner, E. D.; Schoeny, R.;DeMarini, D. M. Occurrence, genotoxicity, and carcinogenicity ofregulated and emerging disinfection by-products in drinking water: Areview and roadmap for research. Mutat. Res., Rev. Mutat. Res. 2007,636 (1−3), 178−242.(18) Pena-Abaurrea, M.; Jobst, K. J.; Ruffolo, R.; Shen, L.;McCrindle, R.; Helm, P. A.; Reiner, E. J. Identification of potentialnovel bioaccumulative and persistent chemicals in sediments fromOntario (Canada) using scripting approaches with GCxGC-TOF MSAnalysis. Environ. Sci. Technol. 2014, 48 (16), 9591−9599.(19) Zhang, H. F.; Zhang, Y. H.; Shi, Q.; Zheng, H. D.; Yang, M.Characterization of unknown brominated disinfection byproductsduring chlorination using ultrahigh resolution mass spectrometry.Environ. Sci. Technol. 2014, 48 (6), 3112−3119.(20) Peng, H.; Chen, C. L.; Saunders, D. M. V.; Sun, J. X.; Tang, S.;Codling, G.; Hecker, M.; Wiseman, S.; Jones, P. D.; Li, A.; Rockne, K.J.; Giesy, J. P. Untargeted identification of organo-brominecompounds in lake sediments by ultrahigh-resolution mass spectrom-etry with the data-independent precursor isolation and characteristicfragment method. Anal. Chem. 2015, 87 (20), 10237−10246.(21) Peng, H.; Chen, C. L.; Cantin, J.; Saunders, D. M. V.; Sun, J. X.;Tang, S.; Codling, G.; Hecker, M.; Wiseman, S.; Jones, P. D.; Li, A.;

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10104

Rockne, K. J.; Sturchio, N. C.; Giesy, J. P. Untargeted screening anddistribution of organo-bromine compounds in sediments of LakeMichigan. Environ. Sci. Technol. 2016, 50 (1), 321−330.(22) 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. Historical trends of inorganicand organic fluorine in sediments of Lake Michigan. Chemosphere2014, 114, 203−209.(23) Ma, Y. X.; Halsall, C. J.; Crosse, J. D.; Graf, C.; Cai, M. H.; He, J.F.; Gao, G. P.; Jones, K. Persistent organic pollutants in oceansediments from the North Pacific to the Arctic Ocean. J. Geophys. Res.-Oceans 2015, 120 (4), 2723−2735.(24) Gribble, G. W. A recent survey of naturally occurringorganohalogen compounds. Environ. Chem. 2015, 12 (4), 396−405.(25) Eustaquio, A. S.; Pojer, F.; Noel, J. P.; Moore, B. S. Discoveryand characterization of a marine bacterial SAM-dependent chlorinase.Nat. Chem. Biol. 2008, 4 (1), 69−74.(26) Galonic, D. P.; Barr, E. W.; Walsh, C. T.; Bollinger, J. M.; Krebs,C. Two interconverting Fe((IV)) intermediates in aliphatic chlorina-tion by the halogenase CytC3. Nat. Chem. Biol. 2007, 3 (2), 113−116.(27) Gribble, G. J. The natural production of organobrominecompounds. Environ. Sci. Pollut. Res. 2000, 7 (1), 37−49.(28) Meng, L. P.; Wu, S. M.; Ma, F. J.; Jia, A.; Hu, J. Y. Tracedetermination of nine haloacetic acids in drinking water by liquidchromatography-electrospray tandem mass spectrometry. J. Chroma-togr. A 2010, 1217 (29), 4873−4876.(29) Wu, Y.; Moore, J.; Guo, J. H.; Li, A.; Grasman, K.; Choy, S.;Chen, D. Multi-residue determination of polyhalogenated carbazolesin aquatic sediments. J. Chromatogr. A 2016, 1434, 111−118.(30) Zhu, L. Y.; Hites, R. A. Identification of brominated carbazolesin sediment cores from Lake Michigan. Environ. Sci. Technol. 2005, 39(24), 9446−9451.(31) Song, W. L.; Li, A.; Ford, J. C.; Sturchio, N. C.; Rockne, K. J.;Buckley, D. R.; Mills, W. J. Polybrominated diphenyl ethers in thesediments of the great lakes. 2. Lakes Michigan and Huron. Environ.Sci. Technol. 2005, 39 (10), 3474−3479.(32) Peng, H.; Zhang, S. Y.; Sun, J. X.; Zhang, Z.; Giesy, J. P.; Hu, J.Y. Isomer-Specific Accumulation of Perfluorooctanesulfonate from (N-Ethyl perfluorooctanesulfonamido)ethanol-based Phosphate Diester inJapanese Medaka (Oryzias latipes). Environ. Sci. Technol. 2014, 48 (2),1058−1066.(33) Benskin, J. P.; Bataineh, M.; Martin, J. W. Simultaneouscharacterization of perfluoroalkyl carboxylate, sulfonate, and sulfona-mide isomers by liquid chromatography-tandem mass spectrometry.Anal. Chem. 2007, 79 (17), 6455−6464.(34) Anger, C. T.; Sueper, C.; Blumentritt, D. J.; McNeill, K.;Engstrom, D. R.; Arnold, W. A. Quantification of triclosan, chlorinatedtriclosan derivatives, and their dioxin photoproducts in Lacustrinesediment cores. Environ. Sci. Technol. 2013, 47 (4), 1833−1843.(35) Allen, J. G.; McClean, M. D.; Stapleton, H. M.; Webster, T. F.Linking PBDEs in house dust to consumer products using X-rayfluorescence. Environ. Sci. Technol. 2008, 42 (11), 4222−4228.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b03221Environ. Sci. Technol. 2016, 50, 10097−10105

10105

S1

Untargeted Screening and Distribution of Organo-iodine Compounds in Sediments from 1

Lake Michigan and the Arctic Ocean 2

3

Hui Peng, Chunli Chen, Jenna Cantin, David M.V. Saunders, Jianxian Sun, Song Tang, Garry 4

Codling, Markus Hecker, Steve Wiseman, Paul D. Jones, An Li, Karl J. Rockne, Neil C. Sturchio, 5

Minghong Cai and John. P. Giesy 6

7

8

Pages 16 9

Words 1344 10

Figures 9 11

12

13

14

15

This supporting information provides text and figures addressing (1) sample pretreatment; 16

(3) total bromine and iodine analysis; (4) LC-Q Exactive data acquisition; (5) chemometric 17

analysis of DIPIC-Frag results; (6) peaks of iodine fragment from sediment extract; (7) 18

production of iodine fragment from two model organo-iodine compounds; (8) detection of three 19

iodinated analogues in sediment; (9) evaluation of method precision; (10) heatmap of peak 20

abundances of 387 NSOICs; (11) heatmap of all NSOICs; (12) distribution of NSOICs in Lake 21

Michigan; (13) regression analysis of peak abundances of brominated and iodinated analogues; 22

(14) Comparison of abundances of iodinated and brominated analogues. 23

24

S2

Total Bromine and Total Iodine Analysis. Total bromine and iodine, and total organic bromine 25

and iodine were determined by instrumental neutron activation analysis (INAA) as described in 26

our previous study.1 In brief, the sediments were directly subjected to INAA for determination of 27

total iodine and bromine. For organic halogens, 3 g (dry weight) of sediment were extracted by 28

accelerated solvent extraction (ASE) method as mentioned above. After evaporation, the extracts 29

were re-dissolved in 3 mL of acetone for total organic bromine and iodine analysis. 30

Sample Pretreatment and LC-Q Exactive Data Acquisition. Approximately 10 g of freeze-31

dried surficial sediments were used for identification and quantification of NSOICs. A two-step 32

organic solvent extraction was used followed by Florisil cartridge cleanup, and the detailed 33

method of extraction has been described previously.1-3

34

Aliquots of extracts were analyzed using a Q Exactive UHRMS (Thermo Fisher Scientific, 35

San Jose, CA, USA) equipped with a Dionex™ UltiMate™ 3000 UHPLC system (Thermo 36

Fisher Scientific).3 Data were acquired in data-independent acquisition (DIA) mode.

3 Parameters 37

of instrumental conditions were described in our previous studies.2, 3

As described in our 38

previous study, nine separate methods were used for each sample to cover the whole mass range 39

(20 windows and100 m/z mass range for each method).3 To reduce potential shift among samples 40

for comparative analysis, we run the same method for all 23 sediment samples, and then run the 41

next method.2 42

Chemometric Analysis of DIPIC-Frag Results. A novel chemometric strategy was developed 43

to expand the detected number of NSOICs and reduce the false positive rates of predicted 44

precursor ions and compound formulas. 45

Peak detection and match across samples. Identification of NSOICs was accomplished by 46

extracting m/z values of iodine fragment (m/z=126.9045) in each of the 180 DIA windows (10 47

S3

ppm mass tolerance). All peaks detected in any of the 23 sediments were included for analysis. 48

Only the peaks exceeding the cutoff of peak abundance at 1000 were used for subsequent data 49

analysis. Because precursor ions and compound formulas for some NSOICs could not be 50

identified, and to include all peaks for subsequent spatial distribution analysis, a strategy similar 51

to selected reaction monitoring (SRM) was used. In this strategy, rather than precursor ion peaks, 52

the iodine peak from each DIA window was used for semi-quantification of NSOICs. DIA 53

windows and retention times were used to match the NSOIC peaks among samples. For isomers 54

with similar retention times and m/z values, a local regression fitting method (LOESS) was used 55

to adjust for shifts between runs, as seen in a previous metabolomics study,4 in order to avoid 56

potential effects of shifts in retention time on alignments of peaks among samples. 57

Precursor ion alignment and formula prediction. While there are 37 known isotopes of iodine 58

with masses ranging from 108 to 144, all but 127

I undergo radioactive decay. To identify 59

precursor ions and propose compound formulas with confidence, we developed a robust scoring 60

system to determine precursor ions and compound formulas of NSOICs. For NSOBCs, two pilot 61

sediment samples were used to establish the mass spectrometry library.2 But for NSOICs, 62

considering the lesser abundances of peaks compared to NSOBCs, samples exhibiting greater 63

abundances of peaks among the 23 samples from Lake Michigan were iteratively used for 64

alignment of precursor ions and calculation of formulas of individual compounds. Because of the 65

better signal-to-noise ratio for isotopic distribution pattern comparison, this strategy was found to 66

increase the number of detected NSOICs, and improve precision of calculated formulas. 67

(1) Elution profiles. The initial step in the analysis required a reduction of the precursor ion 68

list. This was facilitated by correlation of iodine peaks (m/z = 126.9045) with each of the 69

precursor ions from “precursor ion regions” detected at the exact scanning point of the top of the 70

S4

iodine peak. Correlations between abundances of precursor ions and iodine peaks at each 71

scanning point were analyzed and only those precursor ions with a correlation coefficient (r) 72

exceeding 0.83 (which was based on the 99% confidence incidence level based on a previously 73

established brominated compounds library)3 were included for further data analysis. 74

To avoid instances that precursor ions could only be detected under full scan mode, as 75

documented in a previous study,3 the ions detected in full scan at the retention time 0.1 min 76

around the retention times of iodine peaks, were also used for subsequent data analyses. 77

Correlations between abundance of precursor ions and iodine peak at each scanning point were 78

analyzed (the scanning points were adjusted by the retention time of the detected peaks), only 79

the precursor ions with a correlation coefficient (r) exceeding 0.83 were included for further 80

data analysis. 81

Because of the greater peak abundances of precursor ions than product ions after 82

adjustment of the product ion number as described in our previous study,3 the precursor ion 83

candidates having intensities lesser than 20-fold of iodine peaks were excluded from subsequent 84

analyses. 85

(2) Molecular formulas and seven golden standard. Based on results of previous studies,2, 3

86

chemical formulas of NSOICs were set to contain up to 100 C, 200 H, 5 N, 30 O, 8 I, 8 Cl, 8 Br, 87

1 P and 2 S atoms per molecule. Then the mass error (MSerror) of the predicted formulas for 88

each candidate precursor ion was calculated (Equation 1). 89

2exp{ 0.5 ( ) }act pred

error

pred

mz mzMS

mzσ

−= − ×

× (1) 90

where MZact is the observed m/z value of precursor ions, and MZpred is the predicted m/z values 91

of chemical formulas, and δ is the standard derivation of the instrument, which was set to 2.3 92

ppm according to the well curated library for NSOBCs.2, 3

The seven golden standard algorithm 93

S5

was used for further filtering of molecular formulas.5 To derive parameters for the seven 94

heuristic rules, which were derived from multiple databases,5 we used the library of our 95

previous brominated compounds to produce these rules.3 This was found to be more useful to 96

filter incorrect precursor ions and formulas. 97

(3) Isotopic peak. The similarity of the isotopic peaks between actual mass spectra and 98

predicted molecular formula was very useful to filter incorrect formulas, especially for sulfur, 99

chlorinated and brominated compounds. According to previous studies,6 the similarity of the 100

isotopic peaks could be calculated (Equation 2). 101

(2) 102

where ract,i is the actual mass spectra of peaki, rpred,i is the isotopic peak distribution pattern of the 103

predicted formulas, and IM is the abundance of the monoisotopic compound peak with the 104

highest abundance. 105

(4) Homologue model. We noted that many NSOICs in sediment could be detected with 106

brominated homologues documented in our previous study.3 Thus, information regarding 107

iodinated compounds and brominated homologues could be useful to reduce the false positive 108

rate. A score of 1.0 was given to those predicted molecular formulas with homologues detected 109

in the brominated compounds library. 110

(3) 111

(5) Final library output. The total score for each precursor ion and molecular formula was 112

calculated by summing all scores from previous determinations (Equation 4): 113

(4) 114

, ,1 | |

I

act i pred i

M ii

M

Isotope r r

Ir +

= − −

=

∑

0,Homologue

1,

no

yes

=

0.5Score MSerror Isotope homologue= + × +

S6

The precursor ion and molecular formulas with the greatest score for each NSOIC peak were 115

used to establish the final NSOIC library. 116

An iteration procedure was used: the precursor ions and formula prediction for each peak 117

was started at the sample with maximal peak abundance, the searching was stopped if the 118

formula and precursor ion was successfully assigned; but the searching will be continued by 119

using the next sample with 2nd

most abundance of NSOIC peak; the searching will be stopped 120

until the precursor ions and formulas were successfully predicted, or all 23 samples were 121

searched. 122

To utilize the mass spectrometric library for the analysis of multiple sediment samples, a 123

comparative rather than absolute quantitation strategy was used, as we have done previously for 124

NSOBCs.2 Considering that precursor ions of some NSOICs could not be identified, the iodine 125

fragment peak from each DIA window (rather than precursor ions) was used for quantitation 126

(details are described in Supporting Information). Based on such a strategy, even for those 127

NSOICs whose precursor ions have not been identified, patterns of distribution in sediments 128

could still be determined with confidence. Good reproducibility was achieved between 129

injections (r2>0.99, p<0.001) and between two batches of analysis (r

2>0.99, p<0.001) (Figure 130

S9), which justifies use of this method for large-scale comparative investigations. 131

132

133

S7

134

Figure S1. (a) Fragment ion of iodine in MS2 mass spectra (rt=10.41, 405±2.5 m/z DIA window) 135

detected in sediment-32 from Lake Michigan; (b) Detection of multiple iodine peaks 136

(m/z=126.9045) in one of the 180 successive DIA window (405±2.5 m/z DIA window). 137

138

S8

139

Figure S2. Production of fragment of iodine in MS2 spectra from two model iodinated 140

compounds iodoacetic acid (left) and triiodo-L-thryonine (right) under relatively high HCD 141

collision energy (50 eV). 142

143

S9

144 Figure S3. (a) Heatmap of peak abundances of 387 NSOICs in all marine and freshwater 145

sediments. All peak abudances were log-transformed. (b) Similarity heatmap and clustering of all 146

sediments. Marine sediments were clearly separated from freshwater sediments. 147

148

149

S10

150

Figure S4. Detection of three iodinated analogues in sediment-32 from Lake Michigan by 151

extracting their corresponding m/z values at 10 ppm mass tolerance. 152

153

S11

154

Figure S5. Heatmap of peak abundances of NSOICs in sediments from Lake Michigan (left) and 155

the Arctic Ocean (right). 156

157

S12

158 Figure S6. Distribution of (a) total abudances of NSOICs, (b) NSOICs in cluster 1, (c) NSOICs 159

in cluster 2, and (d) NSOICs in cluster 3 in surficial sediments from Lake Michigan. Sizes of 160

circles are proportional to log-transformed abudances of NSOICs. 161

162

S13

163 Figure S7. Log-linear regression of peak abundances between brominated and iodinated 164

analogues of three model chemicals in Lake Michigan sediments. 165

166

S14

167 Figure S8. Comparison of peak abudances of four model brominated compounds and iodinated 168

analogues. (a) halogenated phenol; (b) halogenated indole; (c) halogenated carbazole; and (d) 169

halogenated pyrrole in sediments from Lake Michigan; and (e) halogenated phenol; (f) 170

halogenated indole in sediments from the Arctic Ocean. The left boxplot represents the 171

brominated analogues, and the right boxplot represents the iodinated analogues. Lines between 172

the two boxplot indicate the paired comparison between brominated and iodinated compounds in 173

the same sediment samples. Halogenated carbazole and prrole in marine sediment was not shown 174

due to low detection frequencies. 175

176

S15

177 Figure S9. Evaluation of method precision. (a) Correlations of NSOIC peak abundance between 178

two injections of the same sediment extract; (b) Correlations of NSOIC peaks abundance 179

between two extracts of the same sediment sample. 180

181

182

S16

References 183

(1) Wan, Y.; Jones, P. D.; Wiseman, S.; Chang, H.; Chorney, D.; Kannan, K.; Zhang, K.; Hu, 184

J. Y.; Khim, J. S.; Tanabe, S.; Lam, M. H. W.; Giesy, J. P., Contribution of synthetic and naturally 185

occurring organobromine compounds to bromine mass in marine organisms. Environ. Sci. 186

Technol. 2010, 44 (16), 6068-6073. 187

(2) Peng, H.; Chen, C. L.; Cantin, J.; Saunders, D. M. V.; Sun, J. X.; Tang, S.; Codling, G.; 188

Hecker, M.; Wiseman, S.; Jones, P. D.; Li, A.; Rockne, K. J.; Sturchio, N. C.; Giesy, J. P., 189

Untargeted screening and distribution of organo-bromine compounds in sediments of Lake 190

Michigan. Environ. Sci. Technol. 2016, 50 (1), 321-330. 191

(3) Peng, H.; Chen, C. L.; Saunders, D. M. V.; Sun, J. X.; Tang, S.; Codling, G.; Hecker, M.; 192

Wiseman, S.; Jones, P. D.; Li, A.; Rockne, K. J.; Giesy, J. P., Untargeted identification of organo-193

bromine compounds in lake sediments by ultrahigh-resolution mass spectrometry with the data-194

independent precursor isolation and characteristic fragment method. Anal. Chem. 2015, 87 (20), 195

10237-10246. 196

(4) Smith, C. A.; Want, E. J.; O'Maille, G.; Abagyan, R.; Siuzdak, G., XCMS: Processing 197

mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and 198

identification. Anal. Chem. 2006, 78 (3), 779-787. 199

(5) Kind, T.; Fiehn, O., Seven Golden Rules for heuristic filtering of molecular formulas 200

obtained by accurate mass spectrometry. BMC Bioinformatics 2007, 8. 201

(6) Tsugawa, H.; Cajka, T.; Kind, T.; Ma, Y.; Higgins, B.; Ikeda, K.; Kanazawa, M.; 202

VanderGheynst, J.; Fiehn, O.; Arita, M., MS-DIAL: data-independent MS/MS deconvolution for 203

comprehensive metabolome analysis. Nature methods 2015, 12 (6), 523-+. 204

205

206