Electron capture negative ionization (ECNI; soft ionization) or electron ionization (EI; energetic ionization) sources were fitted at the interface between the gas chromatograph and the high-resolution mass spectrometer (Trace 1310 and Q Exactive GC Orbitrap, Thermo Scientific San José, CA, USA). Briefly, the GC fraction (1 μL) was injected onto a DB-5MS column (30 m × 0.25 mm, 0.25 μm, Agilent, Palo Alto, CA, USA) and carried with helium (100°C to 325°C, 10°C.min⁻¹) to the mass spectrometer operated in full-scan mode. The mass spectrometer parameters are detailed in Table S3.

An electrospray ionization source (ESI), operated in positive and negative polarities during distinct analytical sequences, was fitted at the interface between the liquid chromatograph and the high-resolution mass spectrometer (Ultimate 3000 UHPLC and Q Exactive Orbitrap, Thermo Scientific). Briefly, the LC fraction (5 μL) was injected on a reverse phase C₁₈-like column (Hypersil Gold, 100 × 2.1 mm, 1.9 μm fitted with a Hypersil Gold guard column, 10 × 2.1 mm, 1.9 μm, Thermo Fisher Scientific) maintained at 45°C. The mobile phase consisted of water and acetonitrile containing 10 mM ammonium acetate with a flow rate set at 0.4 mL.min⁻¹. The mass spectrometer parameters are detailed in Table S4.

Data mining was performed on the full scan GC/ECNI-HRMS and LC/ESI-HRMS data for the sentinel sample set using HaloSeeker 2.0,⁴⁷ in order to pinpoint chlorine- and/or bromine-containing signals. The proprietary raw data were converted in the mzXML open format using MSConvert software (Proteowizard version 2.0.3.3) before being imported to HaloSeeker for deconvolution. The software then proceeded to peak picking, pairing—that is isotopic pattern (cluster) reconstitution according to the precise mass differences between naturally occurring C, Cl and Br isotopes—and the alignment of the clusters across the sequences. The data filtering was carried out using the F2+ filter, based on the triplet A (base peak), A–2, A + 2 and ion ratio rules, restricting to the polyhalogenated clusters. Furthermore, an additional filter related to the intensity differences of the aligned cluster between samples and procedural blanks was also applied. The deconvolution and filtering parameters are detailed in Table S5.

The filtered clusters were then reviewed in decreasing order of intensity. In order to discard false positives (non- or monohalogenated clusters), each cluster was manually reviewed using Xcalibur software (version 3.0.63, Thermo Scientific). The coelution, the precise mass difference and the isotopic pattern of the clustered features were subjected to examination.

Chemical formulas were manually annotated using HaloSeeker and Xcalibur in a double-blind procedure considering as first intention H, C, O, Cl, Br, N, P, S and Si up to 50, 30, 10, 10, 10, 10, 10, 5 and 5 elements, respectively. The computed ion formulas were deemed unsuitable if (i) the difference between the precise and the exact m/z (measured and theoretical m/z, respectively) was greater than ±1 mDa, (ii) the computed isotopic pattern did not align with the observed one, or (iii) the double-bound equivalent was inconsistent (integer in the absence of adduct formation and float in the presence of adducts).

After confidence levelling according to Schymanski et al.⁴⁹ identified Cl/Br-containing compounds were included in a suspect list. Each compound in the suspect list was identified with the t_R, 1 quantifier m/z (most intense) and at least 1 confirmation m/z (second most intense) across the first analytical stage.

Among the 126 identified Cl/Br-containing compounds, 73 were identified as legacy known and characterized contaminants in environment and food. They were grouped in 3 congener series (#1 to #3) and 3 individual compounds, and were all detected by GC/HRMS.

Further details on their identification are provided Figures S4–S11 and Tables S9–S12. The molecular formula and/or the match with libraries indicated with little doubt the proposed structures. Considering that the aim of this study was not to identify legacy POPs congeners at the highest confidence level, upgrading the identification to level 1 was not deemed a valuable use of resources.

Series #1, including 35 congeners, was annotated C₁₂H_10–xCl_x (5 ≤ x ≤ 10) and was then related to polychlorinated biphenyls (PCBs) at level 2a (MS Search library match) (Figure S4, Tables S9 and S10).

Series #2, including 17 congeners, was annotated C₁₂H_10–xOBr_x (4 ≤ x ≤ 9) and was then related to polybrominated diphenyl ether (PBDEs) at level 2a (MS Search library match) or level 2 b (diagnostic evidence) (Figure S5, Tables S11 and S12).

Series #3 was annotated C₁₈H_14−xCl_x (7 ≤ x ≤ 10) and tentatively related to polychlorinated terphenyls (PCTs) at level 3 (tentative candidate) due to a lack of sensitivity using the EI source (Figures S6–S8). Because of the exceptionally high signal content resulting from the vast number of theoretically possible congeners for this family of compounds (n = 8557), only 18 congeners, for which there was no ambiguity in chlorine number considering in source fragmentation, were included in the suspect list.

The 3 individual compounds were identified as POPs organochlorine pesticides and their transformation product. The first one was annotated C₆Cl₆ (m/z_monoisotopic = 281.814, score = 93.9%, Δm/z = –0.05 mDa) and was further confirmed as hexachlorobenzene at level 1 (injection of the analytical standard, Figure S9). The second one was annotated C₁₀H₄OCl₆ (m/z_monoisotopic = 349.840, score = 91.9%, Δm/z = –0.04 mDa) and was then identified as heptachlor epoxide at level 2a (Figure S10). The last one was annotated C₁₂H₈OCl₆ (m/z_monoisotopic = 377.871, score = 93.21%, Δm/z = 0.0 mDa) and was further identified as dieldrin (level 2a) since the endrin analytical standard did not match retention time criteria (Figure S11).

The identification of legacy POPs and POP-like contaminants, which are usually quantified in human diet and plasma in France, demonstrated that this chemical space can be explored by the sampled sentinel matrices.^51-53 This observation indicated that sentinel matrices could be suitable for the identification of POP-like ECs.

In contrast with the legacy contaminants previously identified, all the analytical techniques enabled identifying potential POP-like ECs, some of which were part of 6 congener series (#4 to #9). Further details on their identification are provided Figures S12–S21 and Tables S13–S15.

A unique compound, detected by LC/ESI(–)-HRMS, was annotated [C₂₄H₁₂N₂Cl₄ — H]^– (m/z_monoisotopic = 466.969) and showed a nitrate adduct (Figures S12–S13). A fragment was then annotated [C₁₂H₆NCl₂]^– from the MS² data, indicating a symmetric structure for the parent ion (Figure S14). Among the 19 candidates within PubChem, only two compounds, related to dichlorocarbazole dimers, showed structures that could fragment as observed (Figure 3A). These isomeric dichlorocarbazole dimers differing by the binding sites between the monomer units were proposed as tentative candidates (level 3). To the best of our knowledge, the presence of dichlorocarbazole dimers in the environment or in human matrices has not been reported yet, but at least one has been synthesized by chemical engineering.^54-56 Polyhalogenated carbazoles are POP-like ECs: they exhibit dioxin-like activity, are persistent in soil, are biomagnified in the aquatic food web, and are widely distributed.^57-60 The origin of these substances remains unclear. Some studies suggest an anthropogenic origin while others suggest a natural biotransformation process. Despite the growing interest in polyhalogenated carbazoles over the past decade, no study has yet assessed their risk to human health. The identified dichlorocarbazole dimer may exhibit properties similar to those of polyhalogenated carbazoles, as the present study may demonstrate. The toxicity of the dimer may be different due to its non-planar structure.

Series #4, including 4 hexahalogenated congeners detected by LC/ESI(–)-HRMS, was annotated [C₁₆H₆N₂Br_6–xCl_x — H]^– (0 ≤ x ≤ 2) (Figures S15–S17). The elution pattern, that was attributed to –Br/+Cl substitution, and the close t_R resulted in the classification of these compounds as congeners. Unfortunately, the MS² data showed only HBr or HCl successive losses. The origin of the samples provided a clue to the tentative identification of this series, as congeners were only detected in gull egg yolk, which are closely related to the marine environment. The molecular formula C₁₆H₆N₂Br₆ was attributed to potential candidates from PubChem: 3 bromobiindole congeners and 2 diynyl-bis bromoaniline congeners. While no information was found on the origin of diynyl-bis bromoaniline, bromobiindoles are halogenated natural products that were extracted from marine cyanobacteria and algae.^61,62 According to Vetter (2006), mixed Cl/Br-containing compounds are rarely produced from anthropogenic sources, except through incineration. They are rather more frequently originated from natural sources, such as marine organisms.⁶³ This tends to identify this series as hexahalogenated biindole congeners (BIs) at confidence level 3 as MS² data could not indicate a potential structure (Figure 3B). Recently, the lipophilic cyanobacterial neurotoxin aetokthonotoxin (AETX), a pentabrominated biindole congener (C₁₇H₆Br₅N₃), was identified as responsible of a bald eagle mass mortality.⁶⁴ The tentatively identified BIs in the present study may exhibit similar bioactivity and be at risk to the animal and human populations.

Series #5, including 7 heptahalogenated congeners detected by GC/HRMS, was annotated C₉H₃N₂Cl_7–xBr_x (0 ≤ x ≤ 2) and was further identified as heptahalogenated methylbipyrroles (MBPs) (Figure S18–S21). The relative retention times were found to be in accordance with the –Cl/+Br substitution, which resulted in the formation of less volatile compounds. The EI mass spectra of the heptachlorinated congener matched with 2,3,3',4,4',5,5'-heptachloro-1'-methyl-1,2'-bipyrrole (Cl₇-MBP), thus upgrading the confidence to level 2a for this congener (Figure 3C). The mixed Cl/Br congeners were identified as mixed-MBPs at level 2 b. The –Cl/+Br substitution was previously reported for MBPs.^65,66 MBPs are suspected to be halogenated natural products due to their identification in samples originating from marine ecosystems across various locations.^65,67 While the biomagnification of MBPs in the marine food web and human exposure has been reported,^68-71 no toxicological studies have been carried out, to the best of our knowledge and despite growing interest.

Data from 42 clusters (n = 23 for GC/HRMS, Table S13; n = 16 for LC/ESI(–)-HRMS, Table S14; n = 3 for LC/ESI(+)-HRMS, Table S15) did not meet the criteria for identification at a confidence level lower than 3 for various reasons. For some clusters identified using the ECNI source, it was not possible to carry out comparison with spectral libraries due to a lack of sensitivity using the EI source. For some other clusters detected using these two ionization sources, the matching with spectral libraries was not satisfying. The LC/ESI-MS/HRMS data were sometimes meaningless due to consecutive loss of halogen atoms, which provided no structural information. Besides, the MS² data may have provided information that was not consistent with the annotation. Furthermore, structural elucidation using SIRIUS was often a failure due to the confusion between the monoisotopic and the most intense signals. The identification of Cl/Br-containing compounds assisted with bioinformatics software has been identified as an issue.⁷² The development of dedicated software that can take into account the full isotopic pattern is required. Among these clusters, 4 series were established from GC/HRMS acquisitions. Series #6, including 2 congeners, was annotated C₈H₂Cl₆ without tentative structure(s). Series #7, including 2 congeners, was annotated C₁₀H₅Cl₉ with a proposed structure analogous to nonachlor. Series #8, including 4 congeners, suggested a pentachlorinated isotopic pattern but annotation remained equivocal. The annotation of series #9, including 4 congeners, remained equivocal as well, although the isotopic pattern suggested hexachlorinated congeners. Series #8 and #9 may be related to each other due to a Δm/z = 33.961 Da, corresponding to the –H/+Cl substitution. The relative retention times supported this hypothesis. Despite the low confidence level associated with these 42 compounds, they were included in the suspect list and semi-quantified in both sample sets. They will be denominated from their feature information [m/z_monoisotopic@t_R] in the following paragraphs.

PCBs, PBDEs, PCTs and the 3 organochlorine pesticides were quantified in the sentinel matrices with orders of magnitude ranging from pg.g⁻¹ to µg.g⁻¹ w.w., which appeared consistent with the literature.^{20-25,30,31,33,36,73} For many of these POPs, the highest concentrations were observed in gull egg yolks. This could be explained by an easier access to fish and seafood, which are known to be a major source of POPs exposure.^5-8 In pigeon, adipose tissue seemed to concentrate much more PBDEs than liver and egg yolk, a finding that has also been reported for herons.⁷⁴

The dichlorocarbazole dimer was identified in almost all the samples at concentrations in the range of ng.g⁻¹ w.w. The concentrations in gull egg yolks did not demonstrate local trend. On the other hand, the pigeon egg yolks from Montpellier, a town located a few kilometers from the coast, evidenced higher concentrations than the samples obtained in Paris and Tours. Estimated concentrations for pigeon adipose tissue could indicate bioaccumulation of the dimer, considering that the adults concentrated more than the young did.

BI and MBP congeners were only identified in gull egg yolks at concentrations ranging from a dozen of pg.g⁻¹ w.w. to hundreds of ng.g⁻¹ w.w. Their identification in gull egg yolks exclusively may support the natural marine origin, as previously suggested.

The other compound concentrations ranged from pg.g⁻¹ w.w. to ng.g⁻¹ w.w. The highest concentration was estimated for [297@17.9], with more than 200 ng.g⁻¹ w.w. in 2 pigeon adipose tissue samples. Apart from a few exceptions, these other compounds were present in all gull samples, regardless of the colony’s location, suggesting a wide spatial distribution across the French littoral.

The suspect screening resulted in the identification of 71 out of the 126 Cl/Br-containing compounds in the FOAO sample set. Among them, 23 PCBs, 7 PBDEs and the 3 organochlorine pesticides were identified with estimated concentrations ranging from pg.g⁻¹ w.w. to ng.g⁻¹ w.w. The highest estimated concentrations were observed for seafood products. The orders of magnitude of concentrations and contamination patterns were consistent with those reported in the literature,^5-8 suggesting a reliable semi-quantification.

The dichlorocarbazole dimer was identified in 4 meat and meat products at lower level than those observed in the sentinel samples (<0.1 ng.g⁻¹ w.w.).

Some BI and MBP congeners were identified in the FOAO sample set. While 4 BIs were included in the suspect list, Br₆-BI was the only identified congener in 6 seafood products with estimated concentrations that did not exceed a few ng.g⁻¹ w.w. for the Atlantic mussel. These concentrations were in the same range as in gull egg yolks, but below the most contaminated samples. The 7 MBPs included in the suspect list were all observed in seafood products. In addition, Cl₇-MBP was also identified in terrestrial foodstuffs but with lower concentrations. The estimated concentrations ranged from pg.g⁻¹ w.w. to hundreds of ng.g⁻¹ w.w. The highest estimated concentration was observed for great scallops at one order of magnitude above the most contaminated gull egg yolk. Once more, the detection of BIs and MBPs mostly in seafood products supports the identification of halogenated natural products of marine origin.

Among the 42 compounds identified at confidence level 3 and above, 28 were detected in the FOAO sample set. The estimated concentrations were mostly below a dozen ng.g⁻¹ w.w. although samples reached concentrations above hundreds ng.g⁻¹ w.w for [297@17.9]. While this latter compound was quantified in 2 sentinel samples, it was semi-quantified in 51% of the FOAO samples, including almost all meat and meat products at the higher concentrations. This observation supports the hypothesis of a ubiquitous contaminant.

Furthermore, the cross exploration of the data acquired for the 2 sample sets provided a basis for formulating of new proposals regarding the origin of potential POP-like ECs. The recurrent detection of [354@22.3], [318@7.5], [454@14.1], [542@15.4] and [542@18.8] in seafood products, as well as in gull eggs, suggested that these compounds are of marine origin, or at the very least, that the marine environment is highly contaminated by them. Conversely, the detection in multiple FOAO categories indicates the ubiquitous presence of potential POP-like ECs, including [297@17.9], [160@7.7], [627@28.5] and [406@19.7], as well as series #6.