The Use of Metabolomic Tool in Assessing Environmental Exposure

The term metabolome, similarly to genome and proteome, means the set of all metabolites (lipids, carbohydrates, vitamins, fatty acids, secondary metabolites, signaling molecules, hormones and others) presenting molecular mass up to 1500 daltons in a given biological sample ^{17, 25, 27}. Thus, it is worth noticing that metabolome structure is divided into segments, i.e., the profile of plasma metabolite differs from that of urine, which differs from that of saliva, and so on. Nicholson et al., (2012)²⁶ claim that humans have more than 500 different metabolites, since approximately 500 different cell types are known to produce several metabolites. The Human Metabolome Database - HMDB (2018) recorded approximately 114.500 identified metabolites²⁸. This amount is relatively smaller than that of genes and proteins, which makes metabolomics even more attractiv e as scientific field ²⁹.

Although metabolomics and metabonomics are commonly regarded as synonyms, there is a subtle difference between them: metabonomics is the study of metabolite interactions in a complex system over time; it was introduced by Nicholson; Lindon and Holmes (1999)³⁰ as the quantitative measurement taken from the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification. On the other hand, metabolomics was defined by Fiehn (2001)³¹ as the comprehensive analysis of all metabolites in a given biological system under certain conditions. The first manuscript using the word metabolome was published in 1998; since then, an increasing number of publications on this topic have been reported. The Figure 1 shows the comprehensiveness of metabolomics studies and chemical exposure described by publications from the last 11 years.

Figure 1.Number of publications located in Pubmed on metabolomics and metabolomics approach using analytical platforms - LC-MS, GC-MS, NMR and CE-MS in toxicological studies in the period 2009-2020.

Exposure to environmental toxins leads to changes in metabolism and/or gene expression (deletion or overexpression). Such an alteration may indicate a specific pattern called “metabolic signature”, which has the ability to signal this exposure ¹⁶. Several studies on metabolomics and environmental exposure to toxins have been proposed based on this concept.

Metabolomics has been proven to be an efficient strategy to assess exposure to chemical agents. It showed great potential for studying the toxicity of chemical substances ³², since the metabolite profile can provide a broad view of the physiology of a given organism ³¹. Accordingly, Holmes et al., (2007)³³ proposed the term “xenometabolomics” to characterize the profile of metabolites derived from the biotransformation of xenobiotics.

Two metabolomics-study approaches have been used to detect metabolites: untargeted and targeted metabolomics²⁵. Untarget metabolomics is the comprehensive analysis of all measurable metabolites, and it has been conducted to understand environmental toxin mechanisms ^{18, 34, 35, 36, 37, 38, 39}. Biomarker candidates detected in the untargeted analysis can be validated by targeted metabolomics. Targeted metabolome analysis is based on biological hypothesis and quantitatively investigates differential gene expression from pre-selected metabolite candidates. Metabolites presenting differential gene expression associated with toxin exposure and/or disease must be subjected to pre-clinical and clinical validation, as elucidated by Figure 2⁴⁰.

Figure 2.Stages of the discovery of new biomarkers in metabolomic

Pre-clinical validation can be done through in vitro testing by comparing control and sickle cell lines of pre-selected metabolite(s) or by adding metabolites to a cell line in order to check whether they can induce the development of pathological phenotypes. Next, clinical validation through biomarker detection in real biological samples must be performed in order to determine biomarker robustness and reproducibility ⁴¹. Analysis of such data reveals that the untargeted approach and GC/LC-MS-based metabolomics are the most common me tabolomics methods. However, there is no prevalent study design — animal, cell culture and real biological samples have all been applied to metabolomics studies.

Yet, the robustness and reproducibility of metabolite profile analysis are impaired by the following factors: highly variable metabolite quality and quantity; wide range of metabolite physicochemical properties ⁴². Capillary electrophoresis, chromatography methods and nuclear magnetic resonance (NMR) — all associated with mass spectrometry — have been widely applied to metabolomics studies. However, none of these analytical methods is able to fully detect and quantify metabolites ^{25, 42, 43}. Alternatively, several procedures and mixed methods, such as GC/LC-MS-based metabolomics data, can be performed to broaden both mass spectrometry and physicochemical property coverage ^{44, 45}.

Biomarkers traditionally applied to environmental exposure assessments are mostly related to recent exposure rather than to adverse health effects, which are most likely related to health risks; therefore, environmental risks are underestimated ¹⁶. Broad assessment strategies able to detect several toxin exposures and to reveal the cumulative interaction (synergism and antagonism) among them are key in assessing the impact of environmental toxin exposure on human health.

Metabolomics is aimed at overcoming such limitations and fostering comprehensive analysis strategies, since it enables a new perspective on the first molecular mechanisms triggered by environmental toxin exposure ⁴⁶. Changes in the metabolite profile caused by exposure to toxins interacting with the biological system assumingly indicate toxicity biomarkers that can be applied to future studies ⁴⁷.

Therefore, studies point towards advances in environmental toxicology and to a new outlook for assessing environmental contamination risks: “Omic” biomarker assessment should be advanced enough to detect both environmental toxins and their adverse health effects (disease biomarkers) in a single diagnosis. Biomarker data analysis should become a tool for clinical decision-making in order to protect, diagnose and recover human health.

Metabolomics studies require rigorous experimental planning due to the wide range of factors leading to biased and/or inaccurate results. Dudzik et al., (2018)⁴⁸ state that a good experimental design must cover different variables to ensure both impartial sample analysis and solutions to specific biological issues. The literature presents several pre-analytical biological factors, both intrinsic and extrinsic, affecting the basal metabolic rate. Intrinsic factors include age, sex, genotype, ethnicity, chronobiology, body mass index (BMI) and sample volume, whereas extrinsic factors include eating habits, stress, physical activity, microbiota, environmental toxin exposure and consumption of drugs, alcohol and tobacco ^{49, 50, 51, 52, 53, 54, 55, 56, 57}. Individuals can be evenly distributed according to weight, age and sex across all different groups in order to limit the effect of the aforementioned factors on statistical analysis^{57, 58}. Moreover, results should be normalized to minimize inaccuracy caused by both metabolite concentration variability and method detection limit. Normalization is particularly important for urine samples, since urine metabolome is influenced by urinary volume and renal clearance. The following urinary biomarkers should be normalized for best results: abundant metabolic signals in each sample, urinary volume, creatinine levels and urine osmolality. Creatinine normalization is a common method, since the impact of external factors on urine concentration can be determined by measuring physiological substances in urine samples, such as creatinine. In other words, urinary creatinine levels determine urinary concentration. Accordingly, the concentration of a given metabolite must be divided by its creatinine levels for normalization purposes ^{59, 60}.

Studies must be well controlled through experimental design in order to minimize analytical errors and avoid result bias ^{42, 58}. Therefore, biological issue, study population and biological matrices of choice must be well known to ensure the following: sample homogeneity and representativeness; proper performance of analytical methods, data processing and statistical analysis ⁴⁸.

Metabolome analysis uses one or more combined analytical methods and bioinformatics tools to assess research results. This analysis is key in understanding the biochemical mechanisms exposed to environmental toxins, since it provides qualitative and semi-quantitative data from affected metabolites through untargeted and targeted approaches ⁶¹. These approaches have similar workflows, since experimental procedure standardization and result publication are high priority for result comparisons and further studies. Accordingly, the Metabolomics Standards Initiative (MSI) (http://www.metabolomics-msi.org/), conceived in 2005, defined the procedures to be adopted in studies and publications to achieve such a harmonization. Both targeted and untargeted technologies involve the following steps: defining the biological issue of interest, experimental design, biological sample collection and preparation; data acquisition and analysis; biological interpretation of results ^{17, 62, 63}.

Both untargeted and targeted metabolomics workflows differ in the first step and in the required sample preparation: the untargeted workflow starts by defining the biological issue of interest, which represents the question(s) that must be answered at the end of the study, whereas targeted metabolomics starts with a biochemical hypothesis and focuses the quantitative metabolite analysis based on one or more previously defined metabolic pathways of interest ^{17, 63}, as shown in Figure 3

Figure 3.Workflow summarized in untargeted and target metabolomics.

Experimental design requires systematic planning and the performance of procedures such as defining the biological matrix of interest, sample number per group, sample collection and storage. Plasma and urine are the most common samples, whereas other samples such as saliva, sweat, exhaled air, cerebrospinal fluid and seminal fluid are the least common ones ⁶⁴. Samples should be collected with due attention, since several lifestyle-related pre-analytical factors can affect the metabolite profile. Thus, data such as collection time, need of fasting, and anticoagulant and preservative use should be previously defined. Slupsky et al., (2007)⁵⁷ demonstrated that eating habits can affect urinary metabolite concentration. Therefore, they recommend that urine should be collected in early morning hours to avoid variations. Urinary metabolite profile findings can be altered by factors such as collection and fasting periods, centrifugation conditions, filtration, food additives (e.g. sodium azide), normalizing procedures and repeated freeze-thaw cycles. Plasma is assumingly more stable than serum, whereas serum can provide greater sensitivity than plasma ⁶⁵. Thus, samples must be chosen based on the required sensitivity-reliability analysis results.

Yet, control groups should be selected to allow comparisons between their results and those of exposed groups. Although this approach has been traditionally applied to toxicological studies, it has been adapted: the exposed group is classified based on exposure intensity and comparison is made between the lowest and highest exposure groups ^{19, 37}. Thus, some factors must be considered prior to analysis procedures, such as whether the result comparison of studies with different approaches is valid and whether it should be performed with caution. Therefore, assessments involving discussion and comparison of both traditional and modern result comparison approaches should be conducted in order to guide the scientific community and ensure result reproducibility.

Metabolic processes must be stopped after sample collection in order to preserve samples — this procedure is known as metabolic quenching. This step is normally accomplished by adding organic solvent to the samples and immediately cooling or freezing them with dry ice or liquid nitrogen ^{17, 66}. Samples must be stored at low temperatures, preferably at -80ºC, in order to minimize potential bacterial contamination and sample degradation when instrumental analysis is performed long after sample collection ⁴⁸. Results of studies assessing metabolomic profile stability in different biological samples have demonstrated that metabolite stability depends on several factors, as follows: time period between collection, processing and storage; number of freeze-thaw cycles; storage time and temperature. Pinto et al., (2014)⁶⁷ reported a 2.5-year stability of the plasma metabolomic profile, whereas the literature review conducted by Stevens et al., (2019)⁶⁵ reported a 26-month stability for urinary metabolite samples stored at room temperature (-25 ºC). Semren et al. (2018)⁶⁸ reported insignificant decrease in urinary metabolites stored at -80 ºC for 6 months.

Metabolite sample preparation is one of the most important steps of the procedure. It should be based on desired approaches, biological matrices and analytical methods of interest. Sample preparation for untargeted metabolomics must be minimal in order to avoid metabolite loss during the process, since the comprehensive metabolic analysis is the aim of the preparation. This step usually involves the following procedures: deproteinization to minimize matrix effects and clogging of ionic compounds ⁶³; solvent delipidation; refrigerated centrifugation to separate unnecessary particles such as cells, waste and high-molecular-weight metabolites; solvent dilution and precipitation ^{56, 68, 69}. Next, liquid-liquid (LLE) and solid phase (SPE) extractions are performed. Urine samples require previous urease preparation to minimize pre-analytical and analytical factors of samples analyzed through the GC-MS method. Urea can interfere with derivatization and lead to incomplete reactions, since it is the most abundant physiological metabolite in urine. Accordingly, concentrated urea can overload chromatographic columns and cause chromatographic peak distortions due to co-elution. Therefore, it is recommended to add 30 µL of urease to each 100 µL of urine and incubate at 37 ºC for 30 to 60 minutes ^{24, 70, 71, 72}. On the other hand, targeted metabolomics is based on selective sample preparation, since it aims at quali-quantitative analysis of one or more metabolites with similar physicochemical properties. Therefore, targeted metabolomics requires both highly selective extraction methods and preconcentration and clean-up procedures to eliminate distortions ⁶⁴.

Derivatization must be performed in order to improve metabolome coverage through the GC-MS method, since most natural metabolites contain non-volatile polar functional groups (organic acids, amino acids, monosaccharides, disaccharides and steroids). Thus, methoxymination reactions followed by silylation — the most common reactions — allow improving analyte properties by making analytes volatile and thermally stable^{70, 73, 74}. Instrumental analysis through capillary electrophoresis and nuclear magnetic resonance methods requires full organic solvent evaporation, subsequent waste resuspension in water and dilution of extracts in deuterated solvents ¹⁷.

A fraction of the sample is collected at the end of the preparation step and subjected to analytical platforms for the next step: data acquisition. It is worth noticing that the aim of metabolome analysis is the study of all metabolites (untargeted) of molecular weight up to 1500 Da or the study of part of them (targeted) - these analytes have different concentrations and great chemical diversity. Therefore, metabolites should be assessed through multiplatform analyses, as they can cover the entire metabolome spectrum. Combined chromatography and mass spectrometry (LC-MS and GC -MS) have been the most widely performed metabolomic methods in the last decade. Nuclear magnetic resonance (NMR) and capillary electrophoresis combined with mass spectrometry (CE-MS) are also reported in the literature [⁷⁵,⁷⁶,⁷⁷].

Combined chromatography and mass spectrometry considerably improves qualitative-quantitative analysis of complex biological samples, since the analytes of a given sample can be separated in the chromatograph and moved into the mass spectrometer where they will be ionized, identified and quantified. When these analytical methods are combined, molecular entities can firstly be identified through chromatographic retention time and accurate mass measurement⁷⁷. Moreover, fragmentation pattern can be the parameter to identify molecular entities when data are obtained through MS-MS (MS tandem)²⁰.

Mass spectrometry plays a leading role in metabolomics research due to its high analytical sensitivity and specificity, which generate ions to be separated in the mass spectrometer based on their mass (m) to charge (z) ratio (m/z). Next, the generated mass spectrum is compared to spectral databases or patterns of known substances⁷⁷,⁷⁸. Mass analyzers should be able to analyze a wide range of masses at high resolution, i.e., they should distinguish two m/z peaks differing by only a single atomic mass unit for metabolome analysis⁷⁸. Hybrid mass spectrometers, which use more than one mass analyzer, such as triple quadrupole, and quadrupole time-of-flight (Q-TOF-MS) analyzers, improve the resolution and precision of ion mass measurements. Therefore, they are the most appropriate analyzers for metabolome analysis^{44, 79, 80}. Different ionization processes must be performed for the coupling of mass spectrometers, due to structural differences between LC and GC methods. GC-MS uses Electron Ionization (EI) and chemical ionization (CI), whereas LC uses Electrospray Ionization (ESI) — the most common LC method —, Atmospheric Pressure Ionization (API) and Atmospheric Pressure Chemical Ionization (APCI)⁸¹.

Selection of analytical ionization methods must be based on desired polarity and molecular mass range. GC-MS is assumingly the most suitable method for low-polar, low-molecular-weight compounds, whereas LC-MS-ESI is the most suitable for highly polar, high-molecular-weight compounds. Furthermore, these analytical methods are highly complementary, which explains why they are combined in order to cover the entire metabolome ⁸².

Yet, both methods have pros and cons. GC-MS is a very robust method, whose sensitivity, selectivity and reproducibility are suitable for simultaneous metabolite analysis ^{61, 83, 84}. Moreover, it stands out for its retention-time reproducibility and its low-cost, comprehensive databases. However, it has a limitation: lack of comprehensive MS-MS fragmentation data ⁴⁵. Comprehensive GC-MS-based metabolomics protocols on sample preparation and analytical conditions have been described by Fiehn (2017)⁸⁵, Garcia and Barbas (2011)⁷⁰ and Mastrangelo et al., (2015)⁸³ .

The LC-MS method is particularly important to solve complex matrices ⁷⁷. It is also suitable for metabolomics studies due to its great versatility (variety of mobile and stationary phases) ^{17, 44, 77, 86}. Hydrophilic interaction liquid chromatography (HILIC) has been reported for polar metabolite analysis, polyhydroxy metabolite analysis and reverse phase chromatography (relatively non-polar metabolite analysis) ^{38, 77, 86}. The advantages of this method are high precision and resolution; analytical sensitivity and specificity. Furthermore, LC-MS requires simplified sample preparation and is applicable to complex polar and nonpolar mixtures. The only disadvantage of LC-MS lies on the high cost of replacing consumables by highly pure and compatible mobile phases ⁴⁵.

Procedures to assess data acquisition quality should be performed in sequence analysis. Such a quality is assessed through quality management processes like Quality Assurance (QA) and Quality Control (QC) ⁴⁸,⁸⁷,⁸⁸. QA involves all the planned and systematic activities implemented in the pre-analytical phase to ensure that preset quality requirements will be fulfilled by subsequent analytical processes. Currently, there are no recommended guidelines for the QA process. QC can be defined as the operational techniques and activities used to measure and report quality requirements during and after data acquisition. Quality control samples can be collected by solutions presenting authentic chemical standards, samples pooled with equal aliquots of biospecimens from different biological sources or samples pooled with equal volumes ⁴⁸,⁸⁷,⁸⁸. Pooling samples are commonly used in QC because they show the complete metabolite structure of all biological samples in a given study. Thus, significant quality control deviations indicate analytical variability resulting from sample preparation or instrumental analysis phases. This strategy has been applied to assess the stability and performance of control systems ^{43, 89}. Dudzik et al., (2018)⁴⁸ recommend intermittent QC analysis after every 5 to 10 samples throughout the analytical run and state that variation coefficients lower than 30% in each peak should be considered acceptable. Graphical representation of QC results through PCA models is a practical way to assess control stability, since intense clustering indicates good accuracy ^{48, 90}.

However, the 2017 questionnaire developed by the Metabolomics Society Data Quality Task Group (DQTG), assigned to 97 employees from 84 metabolomics companies, evidenced that there is no consensus on QC procedures and decision-making regarding data quality in the international community⁸⁷. Thus, further research should be encouraged to foster the development of reliable QA and QC strategies. These strategies should be commonly applied to metabolomics workflows in order to improve the overall quality of results.

Metabolome analysis captures a huge amount of exceedingly complex biochemical data (disease or chemical exposure) that makes manual inspection impractical. Thus, bioinformatics tools are indispensable for data processing. Untargeted metabolomics data analysis is laborious and involves several steps, such as: converting raw data into datasets; data collection and alignment; retention time and baseline correction, spectral deconvolution, data normalization; metabolite identification¹⁷. There are several software for analyzing data collected through combined analytical methods, such as XCMS^®, AMDIS^®, Mass Professional Profile^® and MetAlign® ^{25, 70, 85, 86, 91, 92}. On the other hand, targeted metabolomics data analysis is focused on the quantitative assessment of previously selected metabolites ¹⁷.

The normalized dataset is subjected to multivariate chemometric analysis through unsupervised methods, such as principal component analysis (PCA), and supervised methods, such as partial least squares-discriminant analysis (PLS-DA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA). These methods have clear purposes: PCA preliminarily identifies sample clusters to reduce the dimensionality of large datasets, in addition to allowing the identification of outliers. On the other hand, PLS-DA is a regression method that reveals the variables mostly causing result variability and identifies differences between assessed groups. Finally, OPLS-DA identifies potential biomarkers ¹⁷,⁹³. Several tools are available for multivariate data analysis, including: MetaboAnalyst^{® 94, 95 -} a set of free online tools that can be used in both pre-processing and multi- and univariate statistical analysis and metadata interpretation; SIMCA-P^® paid software developed by Umetrics; Matlab^{® 17}.

Univariate statistical tests such as Student’s t-test, Analysis of Variance (ANOVA) and Mann-Whitney U test are commonly performed to assess the statistical significance of each peak (molecular entity) ^{17, 36, 74, 92}.

Commercial databases such as NIST, HMDB, Blood Exposome and METLIN have been accessed to help identifying metabolites ^{86, 91}. However, it should be noticed that most metabolites are not detectable. Hounoum et al. (2016)⁸⁶state that approximately 25% of them are identified by attempt, since many metabolites are not added to metabolite databases and repositories. To harmonize metabolomics studies, in 2005, the Society of Metabolomics created the Metabolomics Standards Initiative (MSI), which advocates the standardization of the entire experimental procedure, including 4 confidence levels for identification of metabolites ¹²².

Yet, biological data interpretation is as important as the previous steps. This procedure is performed by placing the identified metabolites into a metabolic pathway or network in order to identify their variation rate and compare it to that of the control group. This step aims at finding the answer to the biological question established at the beginning of the workflow to ensure meaningful data interpretation¹⁷,⁹². Pathway mapping, visualization and enrichment analyses are common and accessible tools in databases⁸⁶ such as Kyoto Encyclopedia of Genes and Genomes (KEGG)⁹⁶, MetaCyc ⁹⁷, MarVis-Pathway⁹⁸ and RaMP⁹⁹used to find the mechanistic view of identified metabolites¹⁰⁰. Quantification of identified metabolites is a common procedure of targeted metabolomics ¹⁷.

Piovezan (2014)⁹⁰ states that metabolomics results should be systematically reviewed at the end of each workflow step. Analytical validation is performed through the following procedures: visual inspection of all analyses and of LC analysis’ pressure curves; evaluation of reprocessed data according to the number of molecules extracted per sample, their total signals and QC results. On the other hand, chemometric models are evaluated through statistical tests such as R² and Q², or through cross-validation and permutation testing. Finally, biological validation must be performed in targeted metabolomics studies⁹⁰.

When mixed chemicals (xenobiotics) enter the body, they can either act in single metabolic pathways or overlap in common metabolic pathways, according to Vineis (2018)¹⁰¹. Thus, estimating of low-level chronic chemical exposure risks through conventional epidemiological studies is insufficient, since it cannot cover all the complexity of cumulative exposure effects. In view of this limitation, new strategies for cumulative risk assessment should be developed in order to provide real estimate of such risks¹⁰². Therefore, new technologies have been developed to provide new methods to assess cumulative exposure and its health risks. Metabolomics has been used to assess cellular disorders induced by chemical exposure, estimate disease risk and detect new biomarkers^{24, 40}, since environmental pollutants lead to early biochemical changes in metabolite levels, which can reveal disease onset and prognosis. These changes in metabolite levels are called metabolomic signatures or metabolomic profiles. Thus, metabolomics allows qualitative and quantitative analysis of the metabolomic profile in body fluids, before the onset of clinical symptoms. Moreover, molecular profiling can be used to identify exposure, diagnosis and/or prognosis biomarkers¹⁰³. Such a comprehensiveness significantly increases the opportunities for toxicological and environmental health actions. It also encourages discussions on a new biomarker classification, whose metabolomic profile (metabolomics biomarker) embodies the role of exposure and/or effect biomarker.

In view of the high risk of developing chronic diseases due to environmental chemical exposure^{16, 102}, the lack of publications on metabolomics approach to environmental chemical exposure reveals that this knowledge field has been little explored. Therefore, further research should be encouraged in order to unveil the poisoning pathways of chemicals and develop measures to cure, protect and recover the health of exposed individuals. Some relevant studies proving metabolomics’ potential to assess environmental chemical exposure are described below.

Studies carried out in China have assessed exposure to arsenic¹⁹,³⁷—  an environmental threat to human health, since exposure to it can lead to several diseases, such as cancer, cardiovascular diseases and peripheral neuropathy¹⁰⁴. Zhang et al., (2014)¹⁹ assessed the urinary metabolome from Chinese men environmentally exposed to arsenic through the HPLC/ Q-TOF-MS method. Individuals were classified according to urinary arsenic level and metabolite profile (comparison between highest and lowest exposure groups). Untargeted metabolomics detected five dose-dependent arsenic, differentially expressed metabolites (testosterone, guanine, hippuric acid, N¹-acetyl-N²-formyl-5-methoxykynuramine (AFMK) and serine). Based on such results, the authors stated that endocrine disruption and oxidative stress are the likely causes of arsenic poisoning. They assessed the biomarker potential (selectivity and specificity) of candidates through the ROC (Receiver Operating Characteristic) curve. Results showed that testosterone, guanine, hippuric acid and the combination of them all were gene signatures of arsenic-induced metabolic disturbances. AUC (Area Under the Curve) measures the overall potential of a test to discriminate whether a specific condition is present or not. AUC biomarker values greater than 0.85 are usually considered acceptable for clinical application ^{19, 105}.

Chemical exposure during pregnancy has attracted the attention of the scientific community. The effects of arsenic on pregnant women’s health have also been assessed. The urinary metabolite profile of Chinese pregnant women was analyzed for the first time through the UPLC/Q-TOF-MS analytical method in order to identify the adverse effects of low-level arsenic exposure. The urinary arsenic of all participants was measured and used as reference to group urine samples according to exposure level. Metabolome analysis aimed at assessing metabolite profile and comparing the lowest and highest exposure groups. Results indicated the detection of 9 significantly altered metabolites. The predictive capacity of these metabolites was assessed through the ROC curve, which demonstrated that all 9 metabolites had enough predictive capacity (AUC > 0.8) to be considered potential biomarkers ³⁷. Metabolic disturbances in pregnant women and their fetuses resulting from exposure to perfluoroalkyl substances (PFASs), persistent organic pollutants, were reported by Li et al (2021) ¹⁰⁶. The authors identified that exposure to PFAS led to changes in steroid hormone biosynthesis and acid metabolism fatty acids, vitamins, amino acids and lipids. The pathway enrichment analysis showed that 3-fatty acid metabolism and retinol were significantly correlated with exposure to PFAS in maternal blood, and sterol metabolism was found to be significant in both maternal serum and umbilical cord serum. In addition, serum metabolomics from 102 Chinese pregnant women revealed disruption of thyroid hormone metabolism and glyceraldehyde metabolism. These disturbances have been associated with birth impacts resulting from prenatal exposure to pesticides such as mecarbam and β-hexachlorocyclohexane (β-HCH) ¹⁰⁷.

A recent study conducted by Wang et al. (2018) ¹⁰⁴ assessed metabolite changes induced by combined exposure to polycyclic aromatic hydrocarbons (PAHs) and short-chain chlorinated paraffins (SCCPs) in Hep-G2 cells. Significant changes have been detected in several metabolic pathways, such as phospholipid metabolism, fatty acids, tricarboxylic acid cycle, glycolysis and purine metabolism. Wang et al. (2018)¹⁰⁴ highlighted that lipid metabolism disorder was induced by combined exposure to PAHs and SCCPs. Accordingly, this finding requires further research due to its severe impact on human health.

PAHs are organic contaminants having at least two condensed or fused aromatic ring structures. They are ubiquitous in the environment and stem from the incomplete combustion of organic compounds. PAHs are present as vapors or adhered to particulate matter. Exposure to PAHs can cause cancer, teratogenesis and mutations^{20, 105, 106}. The harmful effects of chronic exposure to PAHs were assessed by Wang et al., (2015)¹⁰⁷ in 566 Chinese volunteers — children and elderly individuals. Untargeted urinary metabolome analysis detected 18 discriminating metabolites between the exposed group and the control group, of which dodecadienylcarnitine and 1-hydroxyphenanthrene (1-OHPh) were potential biomarkers to assess the exposure of the general population to PAHs. The authors suggest that these metabolic changes are caused by PAH-induced oxidative stress, and that supplemental antioxidants should be administered to neutralize or reverse the harmful effects of these hydrocarbons.

Heavy metals are chemicals of great public health concern. Cadmium metal comes from both natural sources and human environmental impacts, and it is suitable for a wide variety of industrial applications. This metal is a human carcinogen and exposure to high doses of it can lead to adverse health effects such as itai-itai disease, which causes osteomalacia, osteoporosis and renal tubular dysfunction ^{108, 109}. Cases of death from cadmium poisoning have been reported in Thailand and Japan ^{24, 109, 110}. Contaminated individuals living in Mae Sot City, Tak province, Thailand were selected for research on the metabolite profile of individuals intoxicated by cadmium. The study aimed at identifying cadmium biomarkers (urinary cadmium) and the urinary metabolite profile of intoxicated and healthy patients (control) through GC-MS analysis. Results indicated that urinary citrate and myo-inositol may be potential biomarkers for Thais exposed to Cd. Moreover, urinary citrate has also been proven to aid in early nephrolithiasis diagnosis and prevention ²⁴. Cadmium-induced nephrotoxicity was investigated in another recent study, conducted by Zeng et al (2021) ¹¹⁶. The urinary metabolic profile of 149 individuals (99 women and 45 men) residing in areas contaminated with different levels of cadmium. The metabolic profile showed that exposure to this metal caused alternations in the creatine pathway, amino acid metabolism, especially tryptophan metabolism, aminoacyl-tRNA biosynthesis and purine metabolism, regardless of gender. Therefore, these findings should be explored to identify early-effect biomarkers that can be used to predict risk and prevent cadmium-induced nephrotoxicity.

Another environmental contaminant of toxicological importance is particulate matter (PM), which is among the main air pollutants. PM consists of liquid or solid particles suspended in the air, whose composition is quite heterogeneous and includes toxic substances such as polycyclic aromatic hydrocarbons (PAHs) and metals (Pb, Hg, V, Cd, Cr and others) ^{117, 118}. Several studies have shown the association between prolonged exposure to PM, especially PM_2.5 (<2.5 μm in aerodynamic diameter) and an increased risk of cardiovascular and metabolic diseases ¹¹⁹,¹²⁰. Chu et al. (2021)¹²¹ performed a prospective cohort study to investigate metabolic changes in plasma of 78 Chinese university students, 36 men and 42 women; exposed to PM_2.5 and PM₁₀ (mean concentration of 53 μg/m³ and 93 μg/m³ respectively), through untargeted metabolomics. All participants underwent 8 rounds of physical examination to assess cardiopulmonary function and collect plasma samples for metabolomic analysis. Air purifiers were installed in 40 dormitories of the 78 participants for 14 days. Plasma metabolomics identified 25 differential metabolites associated with exposure to PM_2.5 and none associated with exposure to PM₁₀. Of the 78 students, 9 were considered susceptible to exposure to PM_2.5 because they had diastolic pressure (DP) and forced levels vital capacity (CVF) significantly associated with the variation of PM_2.5. By comparing their plasma metabolic profile to that of healthy individuals, 6 differential metabolites were identified: (lysoPC (P-20: 0), lysoPC (P-18: 1 (9z)), lysoPC (20: 1), lysoPC ( O -16:0), choline and found 1,3-diphenylprop-2-en-1-one). These findings pointed out that long-term exposure to PM_2.5 leads to changes in phospholipid catabolism. Furthermore, LysoPC (P-20: 0) and LysoPC (P-18: 1 (9z)) increased significantly after the air purification intervention. Thus, the authors suggested the use of the six discriminating metabolites as potential biomarkers to identify individuals sensitive to exposure to PM 2.5; and LysoPC (P-20:0) and LysoPC (P-18:1 (9z)); as biomarkers of exposure to PM_2.5.

The herein described studies highlight the importance of metabolomics as tool to assess the impact of environmental toxin exposure. Thus, this research approach is expected to replace epidemiological questionnaires that oftentimes underestimate environmental factors assessed by epidemiological studies.