Determination of Chemical Mixtures in Environmental, Food, and Human Samples Using High-Resolution Mass Spectrometry-Based Suspect Screening Approaches

Chemical contaminants are pervasive across environmental, food, and biological matrices, representing a growing public health concern. To better understand human exposure, a suspect screening strategy based on large spectral libraries was developed using combined LC- and GC-HRMS approaches.

This methodology was applied to 16 samples spanning environmental, food, and human domains. A total of 547 compounds were likely identified, with 63 confirmed at the highest confidence level. Environmental and wastewater samples contained the most diverse range of contaminants, while fish had fewer. Pharmaceuticals, pesticides, and personal care products were the most commonly detected substances, particularly in water and serum samples. Natural and endogenous compounds were also consistently found across all matrices, highlighting the broad and interconnected nature of chemical exposure.

The original article

Determination of Chemical Mixtures in Environmental, Food, and Human Samples Using High-Resolution Mass Spectrometry-Based Suspect Screening Approaches

Solène Motteau, Gaud Dervilly, Ronan Cariou, Maria Margalef, Marja Lamoree, Timo Hamers, Maria König, Beate I. Escher, Anne Marie Vinggaard, Christina Rørbye, Bruno Le Bizec, and Jean-Philippe Antignac*

Environ. Sci. Technol. 2025, 59, 39, 21265–21277

https://doi.org/10.1021/acs.est.4c12608

licensed under CC-BY 4.0

Selected sections from the article follow. Formats and hyperlinks were adapted from the original.

Environmental chemical pollution is a major concern in terms of public health. The number of molecules to which humans are exposed is considerable; however, only a very small number is known in proportion to reality. The CAS Registry contains over 250 million entries, and the European Chemical Agency lists over 100,000 substances under the REACH regulation (https://echa.europa.eu/fr/information-on-chemicals/registered-substances). Contaminants of concern (CCs) refer to synthetic or natural chemicals that have recently been identified and detected in the environment and potentially or already determined as hazardous to the environment and human health. (1−5) CCs are diverse and numerous, including pharmaceuticals, personal care products (PCPs), pesticides, surfactants, persistent organic pollutant (POP)-like substances, legacy pollutants, flame retardants, and various industrial chemicals such as per- and polyfluoroalkylated substances (PFAS) or plasticizers. (6,7) CCs should be understood as either new chemical substances/alternatives or already known chemicals for which new concerns arise based on new exposure routes, data on specific populations, or new toxicological data. The presence of CCs has been reported in various environmental matrices, (8,9) in food, (10,11) and in humans through biomonitoring studies. (12,13) Due to the widespread distribution of CCs and legacy contaminants in the environment, biota are exposed to complex mixtures of chemicals with varying concentrations, physicochemical properties, and toxicological effects. Moreover, the environmental (bio)degradation and metabolic processes lead to the formation of (bio)transformation products (TPs), which introduce an additional layer of complexity with regard to the characterization of real-life chemical mixtures to which humans are exposed. (14−16) The complexity of these mixtures requires a multidisciplinary and integrated approach to document, understand, simulate, and manage the associated risk. (17) This holistic perspective aligns with the “One Health” concept, which acknowledges the connections between human, animal, and environmental health. By adopting this framework, it is possible to address the challenges posed by complex chemical mixtures and their impacts on the broader ecosystem more effectively.

To determine the occurrence and fate of chemicals contributing to this human exposure, targeted analysis remains the gold standard for sensitive and specific analysis, enabling reliable quantification. However, they are restricted to a relatively low number of contaminants defined a priori, usually between fewer than 10 and fewer than 50 of the same chemical family for accredited methods. Multiresidue methods enable expanding the number and the diversity of the compounds analyzed, although they still rely on reduced, predefined lists, (18−20) generally between 20 and 200 of multiclass compounds, in some case above 500 contaminants. (21) Over the past 15 years, the so-called suspect and nontargeted screening analysis (SS/NTS) has become increasingly popular, supported by the development of high-resolution mass spectrometry (HRMS) associated with specific and advanced data processing software linked to increasingly large spectral databases. These databases can contain over 15,000 unique molecules (Massbank Europe) to more than 900,000 in the METLIN Metabolite and Chemical Entity Database.

Unlike targeted analysis, SS/NTS allows for the detection without the need of standards and, when possible, identification of an extended range of exposure markers, without prior knowledge of their identity and/or presence for NTS studies. (22−24) SS/NTS may also allow for the retrospective data mining of archived data, enabling the characterization of past exposures and temporal trends. Since its emergence, SS/NTS has been applied to various fields, especially in the environmental area (25,26) and more recently in food (27,28) and human matrices. (29,30) For example, these approaches have enabled the annotation of over 300 compounds in surface waters, with the observation of temporal variations. (31) Another illustration of the benefit of SS/NTS is provided by Tran et al. (2020), (32) who prioritized 40 halogenated compounds in breast milk, 34 of which were not typically monitored. SS/NTS has high potential and evokes huge expectations for documenting complex real-life chemical mixtures and for further supporting exposure and risk assessment. (33)

While these SS/NTS approaches have made significant progress, challenges remain in the field, particularly in the comparison and interpretation of results for the assessment of chemical mixtures. (34) Since the development of SS/NTS, liquid chromatography (LC) has been the preferred analytical technique, with gas chromatography (GC) being used less frequently. According to Manz et al. (2023), (35) 51% studies referenced in their review were conducted using LC–HRMS and 32% using GC–HRMS, limiting the chemical space of compounds detectable with those approaches. Additionally, communities working on SS/NTS have been historically separated according to their focuses on different environmental matrices, food safety, and human biomonitoring. (35) This separation has led to the development of distinct guidelines, protocols, and analytical approaches, leading to compartmentalized results, lacking the acknowledgment of the existing chemical continuum from external environmental to human internal exposures. (36,37) Therefore, there is a need to break down barriers and look at the continuum as a whole to identify and control the sources of contaminants that contribute to exposure to chemical mixtures.

The PANORAMIX project (EU H2020 Green Deal, 2021–2025) aims to develop, assess, and provide innovative strategies for improved risk assessment of complex real-life mixtures for the protection of European citizens and the environment. (38) While the present work aimed at identifying real-life chemical mixtures of concern and exposure drivers, this whole multidisciplinary project also integrated in vitro bioassays, effect-directed analyses, and mixture modeling, which have conditioned the present component, particularly regarding the developed and applied sample preparation strategies. The interdisciplinary framework is based on a set of identical samples to ensure a comprehensive analysis, from chemical profiling to toxicity end points, to how the data generated can be integrated into a common mixture assessment framework. In this context, the specific objective of the present study was to characterize and compare the chemical mixtures present in selected pooled samples from the environment (wastewater)–food (fish, cow milk, drinking water)–human (serum, milk) continuum by screening approaches using both LC– and GC–HRMS.

2. Materials and Methods

2.2. Chemical Profiling

2.2.1. LC–HRMS Settings

Extracts were analyzed by ultrahigh performance chromatography coupled to high-resolution mass spectrometry (UHPLC–HRMS). The chromatographic separation was carried out on a Dionex UltiMate 3000 system (Thermo Fisher Scientific) coupled to a Thermo Q Exactive (Thermo Fisher Scientific), equipped with a heated electrospray ionization source. The separation of analytes was performed on an Acquity Premier HSS T3 C18 instrument (2.1 mm × 100 mm, 1.8 μm, Waters) or on a Hypersil Gold instrument (2.1 mm × 100 mm, 1.9 μm, Thermo Fisher Scientific). Two distinct acquisition methods were run for one-dimensional and MS/HRMS (top 5 data-dependent acquisition).

2.2.2. GC–HRMS Settings

Extracts were injected on a Trace 1310 gas chromatography system equipped with a TriPlus RSH autosampler (Thermo Fisher Scientific) and coupled to a Thermo Q Exactive (Thermo Fisher Scientific), equipped with an electronic impact ionization source. The separation was carried out on a DB-5MS (30 m × 0.250 mm, 0.25 μm, Agilent Technologies) column.

3. Results and Discussion

3.3. Signatures of the Observed Real-Life Chemical Mixtures According to Use Classes

In addition to the identified compounds, the annotated compounds in the samples allow for expansion of the picture of the presence of CCs and their multiple origins. Eleven substance classes and/or use classes were considered to classify each detected compound and to determine exposure profiles according to the different types of analyzed samples (Table S14). The assignment was performed using the use- and manufacturing-related information collected from the PubChem and HBMD databases. Some of the chemical substances detected may belong to up to three different use classes (Table S14). In particular, some substances may be natural compounds that can also be used in PCP or food additives, for example, as fragrances. Compounds that could not be classified according to this grid were grouped together under “miscellaneous”. The chemical profiles depended on the type of sample but samples of similar origins clustered (Figure 2).

Environ. Sci. Technol. 2025, 59, 39, 21265–21277: Figure 2. Classification of the compounds identified in each sample extract (Table S13) according to their known substance class and/or use relative to the total number of compounds detected per sample, and resulting chemical exposure profiles (use 1 to 3, Table S14). WW: wastewater treatment plant influent; EFF: wastewater treatment plant effluent; SW: surface water; TW: tap water; BW: bottle water; LFW: lean fish (wild catch); FFW; fatty fish (wild catch); FFA: fatty fish (aquaculture); OCM: organic cow milk; CCM: conventional cow milk; BM: human breast milk; S1, S2, and S5: human adult (male/female) serum; S3: women of child-bearing age serum, S4: human umbilical cord serum.

The profiles obtained show the complex nature of the extracts characterized, especially the water and serum samples (Figure 2). Natural and endogenous compounds appeared as the most represented use class (299), followed by pharmaceuticals with 164 different compounds, PCPs (95), industrial compounds (72), pesticides (58), and food additives (49) (Table S14). The other use classes were found <20 times, and illicit drugs, lifestyle compounds, and flame-retardants were found ≤10 times (Table S14).

3.4. Common Chemical Detected across the Environment–Food–Human Continuum

A wide range of chemical exposure markers were detected in extracts analyzed through the applied suspect screening chemical profiling approach, reflecting the complexity of real-life coexposures and resulting chemical mixtures. The comparison of chemical exposure profiles observed in the different analyzed samples enabled the discussion of the fate of certain detected chemicals across the environment–food–human continuum (Figure 3). Fifty-one common chemicals (compound detected in at least two different matrices) were in the three environmental samples, 16 exclusively between EFF and WW, and 39 between SW and EFF (Figure 3A, Table S15).

Environ. Sci. Technol. 2025, 59, 39, 21265–21277: Figure 3. Upset graphs of the common chemicals identified: (A) in drinking water (TW and BW) and the environmental (WW, EFF, and SW) extracts, (B) in food extracts (F, CM, DW, and DW) and serum, and (C) in human extracts (BM and S) and WTTPs (EFF and WW).

In total, 54 pharmaceuticals, 12 pesticides, and 9 industrial use-related compounds were detected across the environmental samples. Therefore, WWTPs are hot-spot sources of environmental contamination, loading a large panel of chemicals into the receiving compartments. In addition, 22 chemicals were detected in drinking water (BW and TW) and the environmental water (SW, EFF, and WW) samples, among them carbamazepine with four related metabolites, venlafaxine, omelsartan, chlorotoluron, sucralose, hexa(methoxymethyl)melamine, 2-benzothiazolesulfonic acid, and benzoylecgonine (Figure 3A and Table S15). Figure 3A also highlights 13 common compounds among TW, SW, EFF, and WW and 19 between the first three. Thus, contaminants that may originate from WWTPs can subsequently be detected in drinking water. In our case, the tap water was particularly of concern, and it may be related to the type of water used in the production process (surface water versus groundwater). Therefore, not only one specific type of compound but also a broad spectrum of compounds is of concern, as environmental contamination can lead to exposure to complex mixtures.

4. Limitations of the Study

The first limitation of the applied methodology is the chemical space covered by the whole method, with each step applied tending to reduce this space. (34) During sample preparation, choices were made depending on the nature of the matrix, which has presumably resulted in loss of compounds during the procedure depending on the physicochemical properties of the analytes. (79) The chromatographic separation and ionization conditions used will also be affected by the initial choices of sample preparation and instrumentation. In our case, the analyses were carried out using reversed-phase LC and highly polar molecules are not retained on the column, although the column used in this study retains certain compounds more strongly than other C18 columns (according to the manufacturer specifications).

The second limitation comes from the employed annotation databases, which include only a subset of substances that form a part of the real human exposome. MoNA and Massbank, accessible in open-access, contain at best 20,000 molecules but lack metabolites and transformation products. Thus, a high number of detected features with MS/HRMS spectra remain unidentified at this stage by using these spectral databases.

These two limitations could explain the lack of chemical overlap between the serum and food samples, as the chemical spaces captured by the sample preparations applied to the different sample types may differ, and some chemicals of interest remain out of the chemical space amenable to the instrumental techniques. As explained in the second limitation, common compounds present in different matrices may still be unidentified features.

The implementation of RTI was not achievable in the present study that could have permitted to upgrade the identification confidence level associated with some of the detected markers. Although the present number of reported markers with highest confidence levels (1 and 2) is already high and suitable for further interpretation, this RTI workline would certainly be a relevant add-on to limit the number of false positives and improve this annotation of marker with a higher confidence level.

The lack of field blank samples for checking and managing possible external contamination with particular chemicals in some of the analyzed samples also appears as a limitation. As an example, the field blank available in the present study for some of the serum samples (collection blood bag) showed the presence of propylparaben, pointing out that the results should be interpreted with caution for that compound. In a study on maternal serum, urine, and amniotic fluid, Bräuner et al. (2022) (80) also reported field blanks contaminated by propylparaben, compromising the calculated concentrations. The same attention should be paid to other chemicals like plasticizers, 4-nitrophenol, as well as the N,N-diethyl-m-toluamide.

Finally, the relative characteristics of SS/NTS analyses still do not allow one to draw more quantitative conclusions and comparisons. In the present study, relative abundances were used to assess the intervariability between samples of the same type and across samples, normalizing by the usually nonmimetic labeled standards. In addition, the instrumental response factors differ depending on the compound and do not permit a comparison of the levels observed between different compounds and the concentration factors.