**Contents**



## **About the Editors**

**Kun LU** is an associate professor at the Gillings School of Global Public Health, University of North Carolina at Chapel Hill. He has extensive experience in gut microbiome, metabolomics profiling, DNA adducts, and biomarker development. His research currently has an emphasis on microbiome research, biomarker discovery and exposome mapping. Dr. Lu's lab aims to answer how the gut microbiome interacts with environmental exposure and affects disease susceptibility, and how host factors crosstalk with the microbiome to influence its response. His lab first discovered that exposure to environmental toxins, such as organophosphate pesticides and arsenic, perturbed the gut microbiome community and its associated functions. His lab demonstrated that gut microbiome phenotypes significantly influence toxicological responses to environmental exposure. Likewise, Dr. Lu's lab maps the exposome for human disease, with the goals of characterizing all exposures over the lifespan via high-resolution mass spectrometry, understanding the health impact of the exposome, and designing strategies to reduce exposure-associated adverse effects. Dr. Lu's lab has also been developing protein/DNA adducts as novel and sensitive biomarkers to evaluate the health effects of different environmental chemicals. He has published over 70 peer-reviewed articles and served on a number of external scientific review panels/committees.

**Robert J. Turesky** is the Masonic Chair in Cancer Causation and a professor at the Department of Medicinal Chemistry. His research is devoted to cancer etiology programs at the University of Minnesota. Dr. Turesky received his PhD in nutrition and food science from M.I.T. Prior to his current position, Dr. Turesky served as: the group leader of the Biomarkers Unit, Nestle´ Research Center, Lausanne, Switzerland (1986–2000); Division Director of Chemistry, National Center for Toxicological Research, U.S. Food and Drug Administration, Jefferson, AR, (2000–2004); and Principal Investigator, Wadsworth Center, New York State Department of Health (2004–2013). He investigates the biochemical toxicology of dietary and environmental genotoxicants, and applies mass spectrometry methods to identify and measure biomarkers of these chemicals in molecular epidemiology studies that seek to understand the role of chemical exposures in the etiology of cancer.

## *Editorial* **Biomarkers of Environmental Toxicants: Exposure and Biological E**ff**ects**

**Robert J. Turesky 1,\* and Kun Lu 2,\***


Received: 8 May 2020; Accepted: 22 May 2020; Published: 22 May 2020

Biomarkers of environmental toxicants are measures of exposures and effects, some of which can serve to assess disease risk and interindividual susceptibilities. Metabolites, protein, and DNA adducts also serve to elucidate mechanisms of bioactivation and detoxication of reactive toxicant intermediates. Some environmental chemicals act as modulators of gene and protein activity and induce dysbiosis of the microbiome, which impacts the metabolome and overall health. In this Special Issue on "Biomarkers of Environmental Toxicants", original research and review articles are reported on the latest biochemical, bioanalytical, and mass spectrometry-based technologies to monitor exposures through targeted and nontargeted methods, and on mechanistic studies that examine the biological effects of environmental toxicants in cells and humans.

The exposome, or the totality of environmental exposures, represents both internal exposures originating from host physiology and external exposures deriving from diverse toxicants and chemicals, infectious agents, as well as diet and drugs. Exposome has been proposed to be a critical entity of human disease. Xue et al. reviewed the current chemical exposome measurement approaches with a focus on those based on the mass spectrometry [1]. They further explored the strategies in implementing the concept of chemical exposome and discussed the available chemical exposome studies. Early progress in chemical exposome research was outlined, and major challenges were highlighted. Apparently, efforts towards chemical exposome have only uncovered the tip of the iceberg, and further advancement in measurement techniques, computational tools, high-throughput data analysis, and standardization would allow for more exciting discoveries to be made concerning the role of exposome in human health and disease.

The gut microbiome has emerged as a new mediator in human health. On the other hand, the human gut microbiome can be easily disturbed upon exposure to a range of toxic environmental agents. Environmentally induced perturbation in the gut microbiome is strongly associated with human disease risk. Functional gut microbiome alterations that may adversely influence human health is an increasingly appreciated mechanism by which environmental chemicals exert their toxic effects. Tu et al. defined the functional damage driven by environmental exposure in the gut microbiome as gut microbiome toxicity [2]. The establishment of gut microbiome toxicity links the toxic effects of various environmental agents and microbiota-associated diseases, calling for a more comprehensive toxicity evaluation with an extended consideration of gut microbiome toxicity.

Living organisms respond to environmental changes and xenobiotic exposures via diverse mechanisms. tRNA-mediated mechanisms are only recently emerging as important modulators of cellular stress responses. Huber et al. discussed many ways that nucleoside modifications confer high functional diversity to tRNAs, with a focus on tRNA modification-mediated regulation of the eukaryotic response to environmental stress and toxicant exposures [3]. Additionally, the potential applications of tRNA modification biology in the development of early biomarkers of pathology are also highlighted. Their review highlights a high functional diversity, ranging from the control of tRNA maturation and translation initiation, to translational enhancement through modification-mediated codon-biased translation of mRNAs encoding stress response proteins, and translational repression by stress-induced tRNA fragments. Future work in this area would provide more exciting discoveries to better understand the role of tRNA modification in exposure-induced human disease.

Exposure to heavy metals is ubiquitous and has been associated with a number of human diseases. Biomarkers of heavy metal exposure in children are particularly important for monitoring exposure and health risk assessment. The study by Jursa et al. measured hair Mn, Pb, Cd, and As levels in children from the Mid-Ohio Valley to determine within and between-subject predictors of hair metal levels. Specifically, occipital scalp hair was collected in 2009–2010 from 222 children aged 6–12 years (169 females, 53 males) participating in a study of chemical exposure and neurodevelopment in an industrial region of the Mid-Ohio Valley [4]. They found that hair Mn and Pb levels were comparable (median 0.11 and 0.15 μg/g, respectively) and were ~10-fold higher than hair Cd and As levels (0.007 and 0.018 μg/g, respectively). In addition, metal levels were different between male and female subjects and showed different profiles along hair segments.

Benzene is a known carcinogen and causes hematotoxicity. Benzene exposure occurs through factory occupations, and from emissions of burning coal and oil in the air. Cigarette smoking is another important source of exposure to benzene. In this research article, Tranfo and colleagues employed a targeted LC-MS/MS to biomonitor benzene through the biomarker S-phenyl-mercapturic acid (SPMA), which is formed from the benzene oxide metabolite conjugated with glutathione (GSH), and then excreted in urine as SPMA [5]. The findings reveal that the main source of benzene exposure in a cohort not occupationally exposed to benzene in central Italy was through active smoking; however, nonsmokers were also exposed to airborne concentrations of this carcinogen.

Phthalates are used as plasticizers and additives in many consumer products. Human exposure to phthalates is prevalent and occurs mainly through dietary sources, dermal absorption, and inhalation. Laboratory animal studies reveal endocrine-disrupting and reproductive effects of phthalates. Thus, human exposure to phthalates is a public health concern. Wang and colleagues have compiled a review on the biomonitoring studies of phthalates in populations across the globe and associated adverse health effects. Urine, serum, amniotic fluid, breast milk, semen, and saliva serve as biospecimens to screen phthalate metabolites. Epidemiological studies have linked high exposure to phthalates with sex anomalies, endometriosis, altered reproductive development, early puberty and fertility, breast and skin cancer, allergy and asthma, overweight and obesity, insulin resistance, and type II diabetes [6].

Aldehydes are ubiquitous in the environment, originating from man-made sources, tobacco smoke, and natural processes. Some aldehydes are implicated in diseases, including diabetes, cardiovascular diseases, neurodegenerative disorders (i.e., Alzheimer's and Parkinson's Diseases), and cancer. Aldehydes are strong electrophiles that react with nucleophilic sites in DNA and proteins to form reversible and irreversible modifications. These modifications, if not eliminated or repaired, can lead to alteration in cellular homeostasis, cell death, and contribute to disease pathogenesis. In this review, Dator and colleagues describe the metabolism of aldehydes in vivo, and the bioanalytical, and mass spectrometry-based approaches to characterize aldehydes in cells and biomonitoring in humans [7].

Hemoglobin (Hb) and albumin (Alb) are the most abundant proteins in the blood and form covalent adducts with toxicants and endogenous electrophiles. In this review, Preston and Phillips describe targeted and nontargeted mass spectrometry-based strategies to measure exposures to a wide range of toxicants that alkylate the N-terminal valine residues of the α-chain of Hb, and the cysteine residue (Cys-β93) of the β-chain of Hb, which reacts with many electrophiles, including carcinogenic aromatic amines [8]. The histidine residues of Alb react with epoxides of polycyclic aromatic hydrocarbons, and lysine residues form adducts with aflatoxin B1 dialdehyde. The highly nucleophilic Cys-34 residue of Alb is the only site for which untargeted adductomic methods have

served to screen for electrophiles in humans. Protein adductomics techniques can screen for harmful exposures to causative agents of chronic disease and identifying individuals at risk.

Aasa and colleagues employed the *N*-(2,3-dihydroxypropyl)valine hemoglobin adduct formed with glycidol, a carcinogen present in refined edible oils, to assess internal doses of this genotoxicant in a cohort of children [9]. In this research article, the investigators report the adduct showed a fivefold variation between the children. The estimated mean intake of glycidol (1.4 μg/kg/day) was about two times higher than the estimated intake for children by the European Food Safety Authority. The estimated lifetime cancer risk (200/10<sup>5</sup> ) was calculated by a multiplicative risk model from the lifetime in vivo doses of glycidol in the children and exceeded the acceptable cancer risk estimate. The protein adduct biomarker data, calculated intakes, and corresponding estimated cancer risks emphasize the importance of identifying and mitigating the sources of background exposure to glycidol from foods and other possible sources.

Exposure to environmental chemicals often leads to diverse DNA damage, and the formation of DNA adducts is one of the key events in chemical-induced carcinogenesis. It is critical to determine whether DNA adducts cause mutagenesis. Chemical incorporation of a modification at a specific site within a vector (site-specific mutagenesis) has been a useful tool to deconvolute what types of damage quantified in biologically relevant systems may lead to toxicity and/or mutagenicity, thereby allowing researchers to focus on the most relevant biomarkers that may impact human health. Here, Bian et al. introduced shuttle vector-based methods and reviewed a sampling of the DNA modifications that have been studied by shuttle vector techniques [10].

A significant limitation in biomonitoring cancer-causing agents is the paucity of fresh frozen tissues available for DNA adduct biomarker research. In this review, Yun and colleagues report on the methods commonly used to biomonitor DNA adducts, and the use of formalin-fixed paraffinembedded (FFPE) tissues for the measurements of DNA adducts of genotoxicants found in the diet and tobacco smoke [11]. The authors developed a technique to retrieve the DNA adducts from FFPE tissues under mild conditions that completely reverses the DNA crosslinks while preserving the structures of the DNA lesions. FFPE tissues for which there is a clinical diagnosis of disease present a previously untapped source of biospecimens for molecular epidemiology studies that seeks to assess the causal role of environmental chemicals in cancer etiology.

DNA adducts are believed to play a central role in the induction of cancer in cigarette smokers. Ma and colleagues have summarized the research on DNA adducts formed with carcinogens in tobacco smoke and from oxidative DNA damage [12]. The analytical approaches most commonly used are mass spectrometry (MS), <sup>32</sup>P-postlabeling, and immunohistochemistry. Because of the high selectivity and sensitivity, MS methods are the preferred technique and have largely supplanted immunochemical and postlabeling techniques over the past decade. DNA adducts of different classes of tobacco carcinogens have been identified in human biospecimens. Issues pertaining to the validation of DNA adducts such as biomarkers, mitigation of artifacts, and caveats in the designs of human studies are highlighted.

Aristolochic acids (AAs) are found in *Aristolochia* plants, some of which have been used in the preparation of traditional herbal medicines worldwide. AAs are highly nephrotoxic and carcinogenic to humans and implicated as causative agents of the Balkan endemic nephropathy (BEN) and "Chinese herbs nephropathy" in Asia. In this article, Chan and colleagues provide an overview of the exposure of AAs in the Balkan Peninsula, where the comingling of *Aristolochia* plants with grains and the release of AAs from decayed seeds of *Aristolochia* plants contaminate the agricultural soil, the food crops, and the water supply [13]. The links between exposure to AAs and their biomarkers of DNA damage, mutations in cancer driver genes, and mechanisms of kidney fibrosis in Asian cohorts are reported.

Accelerator mass spectrometry (AMS) is an exquisitely sensitive technique to measure long-lived radionuclides that occur naturally in the environment. AMS has been used for many years in the earth sciences, such as for radiocarbon dating in archaeology. In this review, Malfatti and colleagues describe the approaches and advances employing AMS in human health and risk assessment. The applications of radiocarbon tracer technology in cancer-related studies assessing

low-dose toxicology studies of naphthalene-DNA adduct formation, benzo[a]pyrene pharmacokinetics in humans, and the antibacterial triclocarban exposure and impact on the endocrine system are reported [14]. AMS applications in precision medicine include the use of radiocarbon-labeled cells for better defining mechanisms of metastasis and the use of drug-DNA adducts in the in vivo and ex vivo microdosing strategy of chemotherapeutics as predictive biomarkers of interindividual response to chemotherapy.

In summary, this collection of original research and review articles provides a valuable update of the most recent biochemical and analytical tools that employ biomarkers in toxicology research, biomarker discovery, and exposure and risk assessment in population-based studies.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Towards Mass Spectrometry-Based Chemical Exposome: Current Approaches, Challenges, and Future Directions**

#### **Jingchuan Xue 1, Yunjia Lai 1, Chih-Wei Liu <sup>1</sup> and Hongyu Ru 2,\***


Received: 6 June 2019; Accepted: 14 August 2019; Published: 18 August 2019

**Abstract:** The proposal of the "exposome" concept represents a shift of the research paradigm in studying exposure-disease relationships from an isolated and partial way to a systematic and agnostic approach. Nevertheless, exposome implementation is facing a variety of challenges including measurement techniques and data analysis. Here we focus on the chemical exposome, which refers to the mixtures of chemical pollutants people are exposed to from embryo onwards. We review the current chemical exposome measurement approaches with a focus on those based on the mass spectrometry. We further explore the strategies in implementing the concept of chemical exposome and discuss the available chemical exposome studies. Early progresses in the chemical exposome research are outlined, and major challenges are highlighted. In conclusion, efforts towards chemical exposome have only uncovered the tip of the iceberg, and further advancement in measurement techniques, computational tools, high-throughput data analysis, and standardization may allow more exciting discoveries concerning the role of exposome in human health and disease.

**Keywords:** chemical exposome; biomonitoring; environmental monitoring; mass spectrometry; disease; bioinformatics

#### **1. Introduction**

Studies have demonstrated that environmental factors play an equal or even more significant role in the pathogenesis of human chronic diseases compared with other risk factors such as genetic variants [1–4]. Environmental factors can induce changes in the human genome, transcriptome, epigenome, proteome, and metabolome. In 2005, the concept of exposome was first proposed by Christopher Wild to account for the unexplained risk factors underlying human diseases [5]. The "exposome" concept has shaped the thinking of scientists when studying environment-disease associations by switching the research paradigm from a single exposure-disease model to an agonistic analysis of environmental influences on human health [6–8]. Many research areas are benefiting from this mindset shift, including environmental epidemiology, health risk assessment, biomonitoring and environmental monitoring, and mechanistic biology. A considerable amount of published commentaries and reviews have highlighted the potential benefits of the characterization and integration of exposome in future studies [9–17].

The definition of exposome has been evolving ever since its birth. Wild originally defined it as "the totality of environmental exposures from birth onwards" [5]. Later, he redefined the scope of exposome to include three broad categories of non-genetic exposure: internal (e.g., metabolism, gut microbiome, inflammation), specific external (e.g., environmental pollutants, diet, occupation), and general external (e.g., socioeconomic status, education, and climate) [18]. In 2014, Miller and Jones expanded the Wild definition to include the measures of biological responses to these exposures [19].

Exposome studies aim to accomplish two critical goals: (1) to measure the cumulative exposures throughout the entire life of humans and (2) to evaluate the associations or causal relationships between these exposures and any biological changes. Depending on the specific types of exposure, exposome can be measured through a wide array of techniques including remote sensors, questionnaires, geography information systems, biomonitoring and environmental monitoring, and metabolomics [13,20,21]. As a comprehensive review of exposome is extremely large in scope, in this paper, we focus on the entire chemical exposures from embryo onwards, "chemical exposome".

There is still a long way to go to completely characterize the human chemical exposome. However, efforts are being made to capture a critical portion of it [22–24]. The aim of this paper is to undertake a systematic review of published evidences regarding the measurement techniques for the chemical exposome and available studies linking chemical exposome to human health. We further discuss the challenges confronted in implementing the exposome concept. Furthermore, we propose potential solutions to address the scientific and technological barriers to advancing exposome research.

#### **2. Exposome Measurement Approaches**

Measuring the exposome has been a challenging task because of its complex and dynamic nature. To capture the full spectrum of exposures of interest in an individual's life, many approaches have either been newly developed or transferred from other fields. For ease of description, we group the available approaches into three categories: chemical approaches (directly measure the exposures or early biomarkers using chemistry techniques), biological approaches (measure the biological changes induced by exposures using molecular biology techniques), and other approaches (those techniques which do not belong to either chemical or biological approaches, such as a personal wearable device). This paper focuses on the available mass spectrometry-based measurement techniques, but briefly discusses other approaches.

#### *2.1. Chemical Approaches*

Mass spectrometry based analytical techniques are widely used in biomonitoring and environmental monitoring studies and have become the predominant chemical approach in characterizing chemical exposome [9,25]. Mass spectrometry, coupled with a separation technique such as liquid or gas chromatography, has been the most popular method used in the direct measurement of chemicals (xenobiotics, their metabolites, and the early biomarkers) in biological or environmental samples due to the superiority in sensitivity, specificity, and dynamic range [9,25]. Low-resolution mass spectrometry has been traditionally used in targeted analytical methods to measure one or several classes of known chemicals in the sample [26–28]. High-resolution mass spectrometry-based biomonitoring techniques are considered very promising tools in achieving a more complete understanding of the biological significance of exposome [8,9,25].

#### 2.1.1. Low-Resolution Mass Spectrometry

Low-resolution mass spectrometers (LRMS), such as the triple quadrupole mass analyzer (QqQ), is only able to achieve *m*/*z* accuracy at unit level (~1 amu mass window) and is not capable of distinguishing compounds with very similar molecular mass. Also, LRMS has shown a low sensitivity in full scan mode. Both restrictions limit the ability of LRMS to detect unknowns. However, the selected reaction monitoring (SRM) mode of LRMS allows the quantitative analyses of a list of target precursor-product ion transitions with high sensitivity and broad linear dynamic range. To ensure the identification of a compound with LRMS, the retention time, at least two transitions (two product ions of the precursor ion), and their ratio of intensity are normally required [29]. False positive identifications are possible when only one transition is used [30].

Traditional biological measurement (targeted analysis) of exposures heavily relies on LRMS, which measures target xenobiotic compounds or their metabolites in the biological samples. This analytical platform typically provides validated and reliable quantification for analytes even at trace levels. This is of critical value for exposome studies because concentrations of xenobiotics are normally low in biospecimens [22]. Biomonitoring data in current databases or health surveys are primarily obtained via this approach because of its availability and maturity. For instance, the National Biomonitoring Program (NBP), launched by the Centers for Disease Control and Prevention (CDC) in the U.S., is routinely measuring approximately 300 chemicals which are known to be toxic to human beings [31]. New chemicals are added to the list when sufficient evidence supports the toxicological relevance and occurrence in the human body.

However, LRMS based targeted analysis has several limitations when it comes to exposome: (1) the inability to cover a wide range of chemicals in a single run; (2) the high possibility of missing the compounds which are at high levels in the biospecimens but not in the target analyte list; (3) the limited availability of commercial standards to quantify the target analytes in the samples; (4) the potential of false positives due to certain precursor ions only generating one fragment ion; and (5) the limiting of detection of certain chemicals in analysis due to matrix interferences.

Efforts have been taken to capture the exposome with this targeted approach. To increase the coverage, scientists seek to measure as many chemicals as possible in one sample through multiple runs. In a recent study, 128 persistent and non-persistent endocrine disruptors belonging to 13 chemical classes were measured with both liquid chromatography-tandem mass spectrometry (LC-MS/MS) and gas chromatography mass spectrometry (GC-MS) [32]. Although this approach is costly, laborious, and needs a large volume of biological samples, it can provide validated information about the quantities of a relatively large number of chemicals in the human body. Further, with the technical advances in mass analyzers, the latest generation of QqQ instruments allows a notable increase in the number of transitions acquired within the same run. For instance, based on the triggered multiple reaction monitoring (tMRM) function in Agilent 6400 series QqQ instruments, a multiresidue LC-MS/MS method was established with a coverage of about 450 globally important pesticides within 10 min [33]. Several multi-target screening LC-MS/MS methods have also been established to simultaneously measure up to 700 drugs with the hybrid triple quadrupole linear ion trap technology (QTrap) [34,35].

#### 2.1.2. High-Resolution Mass Spectrometry

High-resolution mass spectrometers (HRMS) overcome the drawbacks of LRMS by providing high-quality mass resolution, exact molecular mass, and high sensitivity in full scan mode [36,37], which allows less strict requirements for the chromatography separation and improves the capability to detect low abundance chemicals in complex samples. Common HRMS typically possess mass-resolving power > 10, 000 (R, defined at full width at half maximum, FWHM), including time of flight (TOF), Fourier Transform Orbitrap (FT-Orbitrap), and Fourier transform-ion cyclotron resonance-mass spectrometer (FT-ICR-MS). FT-ICR-MS is costly and has limited availability and user feasibility, followed by FT-Orbitrap and TOF instruments. Currently, a vast majority of biomonitoring and environmental monitoring studies are based on TOF instruments, followed by FT-Orbitrap, with FT-ICR-MS platform generally not used in this field [25]. The pros and cons of TOF and FT-Orbitrap have been reviewed in multiple review articles [36–38]. Firstly, the FT-Orbitrap typically possesses higher mass-resolving power and mass accuracy than TOF instruments [38]. Mass-resolving power is highly dependent on the molecular composition, molecular mass, and scan speed [38]. Q Exactive Hybrid Quadrupole-Orbitrap is one of the commonly used FT-Orbitrap instruments and its resolution can be up to 1000 K at *m*/*z* of 200. Secondly, FT-Orbitrap instruments under full scan mode can achieve similar sensitivity down to a femtogram as QqQ instruments. As one of the newest versions of TOF instruments, Agilent 6550 iFunnel Q-TOF LC/MS can achieve a mass resolving power greater than 25K at 322 *m*/*z*. However, the sensitivity of TOF instruments under full scan mode is compromised with 1–2 orders of magnitude lower than that of QqQ operating in SRM mode [38]. This limits the capability of TOF instruments

to measure chemicals at trace levels in samples. Thirdly, FT-Orbitrap instruments show an inverse relationship between scanning speed and mass resolution, while TOF instruments can maintain a rapid scanning rate regardless of the resolving power. Thus, the FT-Orbitrap based instrumental method needs to be well optimized to achieve a compromise between required mass resolution and adequate chromatography to adequately serve the research purpose [39]. Further, it needs to be noted that the application of an automatic gain control (AGC) in Orbitrap instruments is employed to prevent the overfilling of C-trap, but it also can affect the detection capability of trace compounds in complex matrices by reducing the absolute number of analyte ions that enter the ion trap [40]. The multiplexing feature of Q-Orbitrap instruments can help overcome this problem and achieve a higher sensitivity and an extended intrascan dynamic range for low-abundance chemicals in the complex matrices [40].

HRMS is typically combined with other mass analyzers to form hybrid instruments in real applications, including quadrupole (Q)-TOF, ion trap (IT)-TOF, Q Exactive, LTQ-Orbitrap, and Orbitrap Fusion Lumos Tribrid MS [36]. These hybrid mass analyzers can offer additional advantages such as increased sensitivity and fragments information for structural elucidation. In general, three analytical approaches have been developed with these hybrid mass spectrometers for exposome studies.

*Targeted analysis*. In HRMS based targeted analysis, standards and information of target analytes are available. There are several advantages of HRMS based targeted analysis over LRMS based analysis. Firstly, the confidence in compound identification increases significantly by using exact mass, especially for those chemicals that have only one transition and non-specific transitions (e.g., neutral loss of H2O or CO2, which are also common for matrix interferences). It has been reported that an excellent selectivity regardless of the matrices could be achieved for most compounds with an R = 30,000, by which the target analytes could be distinguished from interferences of the same nominal *m*/*z* [38]. The separation of isobars generally requires higher mass resolving power, for instance, the isobars glutamine and lysine are separated with a multiple-reflection TOF MS in excess of 70 K after a flight time of 0.2 ms [41]. Except for the transitions and retention times, isotopic pattern and monoisotopic mass of analytes can also be used for the identification. Secondly, a higher coverage of analytes can be achieved in a single run. Many hybrid mass analyzers, like Q-TOF and Q Exactive, offer data-dependent MS/MS acquisitions. For example, if a compound in a target ion list is detected in the full scan mode, an MS/MS analysis will be triggered accordingly. This allows for full scan product ion spectra recording within the same run for a significant number of compounds at superior sensitivity than the limit of most LRMS. The limitation of TOF instruments based targeted analysis lies in the lower sensitivity and linear dynamic range compared with the QqQ [42,43]. However, many studies have achieved better sensitivity employing full scan or full MS-data dependent MS<sup>2</sup> mode with FT-Orbitrap instruments compared with QqQ in targeted analysis [44–46].

*Suspect screening (biased non-targeted analysis)*. This approach is used when two conditions are met: compound-specific information (e.g., molecular formula, chemical structure, and physicochemical properties) of the suspects are known and the reference standards for the suspects are not available. M+1 and M+2 isotopes of precursor ion are critical in identifying the chemical formula and MS/MS spectral information is critical in elucidating the chemical structure. Using either Q-TOFMS or Orbitrap instruments, many studies have employed this analytical approach in successfully identifying suspect transformation products/metabolites of parent compounds in environmental media, such as natural waters and wastewater [47,48]. Recently this suspect screening technique was used for human samples. One study used GC-TOF/MS platform and found the presence of pentachlorobiphenyl in child brain tissue [49]. Another study used LC-QTOF/MS platform for the identification of new environmental organic acids (EOAs) in maternal plasma and discovered 65 suspect EOAs including benzophenone-1 and bisphenol S [50]. The available information of the compounds can help facilitate the acquisition and identification of suspects. Data-dependent acquisition mode is normally used in suspect screening, in which the sample analysis starts with full scan MS acquisition and switches to MS/MS mode when an analyte of interest appears in the run and is recognized by the data software based on pre-determined

criteria [25]. Triggers normally used in the data-dependent acquisition mode have been reviewed in details, including ion intensity, accurate mass inclusion, isotope pattern, pseudo-neutral loss, and mass defect criteria [25]. When a suspect ion is found in the samples, chemical structure and physicochemical property-derived information can be used in the identification and confirmation of the compounds [47].

*Unknown screening (unbiased non-targeted analysis)*. When starting without any prior information about the compounds to be detected, an unknown screening approach is employed. Unknown screening theoretically enables the measurement of an unlimited number of compounds in the sample, making it a promising technique for the full characterization of the exposome. Workflows in analyzing HRMS data in non-targeted metabolomics studies have been reviewed elsewhere [25,51–53]. In exposome studies, the workflow can be borrowed from metabolomics with minor modifications, as detailed below.


The workflow mentioned above works well for identifying chemicals in one single sample. When the comparison across different samples is involved, other analyses are needed, such as peak aligning across samples, data pretreatment, and statistical analyses [25]. There are a wide range of chemometric and bioinformatic tools available at each stage in HRMS data analyses: preprocessing, annotation, and statistical analysis [52,64,65]. A variety of workflows, which encompass all stages of HRMS data analyses, are also available, including workflow4metabolomics, galaxy-M, XCMS online, MetaboAnalyst, MAVEN, MAIT, and MZmine 2 [64]. To standardize the confidence level in unknown identification, a variety of systems have been established to communicate the confidence [66,67]. In general, five levels are present and listed as follows with the increase of confidence, as shown

in Figure 1: exact mass only, unequivocal molecular formula, multiple putative chemical structures available, one putative chemical structure, and validated structure with reference standard match in both retention time and mass spectra [68].

**Figure 1.** Proposed confidence levels in unknown xenobiotic identification with high resolution mass spectrometric analysis and exemplified with mono(2-ethylhexyl) phthalate.

Until now, targeted analysis, suspect screening, and unknown screening based on TOF instruments (mainly QTOF) have been used on a widespread basis to measure human exposures to a broad range of chemicals, including both persistent (e.g., organochlorine pesticides, polychlorinated biphenyls, polybrominated diphenyl ethers) and non-persistent (e.g., drugs, pesticides, surfactants, personal care products) chemicals [25,69]. One study developed a LC-QTOF/MS based suspect screening technique with a library of collision-induced dissociation accurate mass spectra of more than 2500 toxic compounds, including illegal and therapeutic drugs, pesticides, and alkaloids [70]. A large number of studies also reported the application of FT-Orbitrap instruments in xenobiotic screening including drugs and pesticides in both human and environmental samples [71–76]. Most studies are using the three analytical approaches together to capture a wide range of chemicals [39,76].

One research area in which HRMS has gained widespread application is metabolomics. Metabolome refers to the sum of all low molecular weight metabolites present in a living system [77]. Non-targeted metabolomics based on HRMS can simultaneously detect the endo- and exogenous chemicals, directly linking exposure to internal dose, biological effects, and disease pathobiology, which is a critical component of the exposome [78]. Thus, metabolomics has become a critical platform in exposome research.

#### *2.2. Biological Approaches*

Human exposure to xenobiotics can induce changes of the biological functions in many respects, such as gene transcription and protein synthesis. Instead of directly measuring the exposure itself, biological approaches measure these biological changes to understand the influences of exposure on human health. This approach is particularly helpful for those exposures which cannot be measured directly, such as reactive agents. For instance, reactive electrophiles, including reactive oxygen and nitrogen species, aldehydes, oxiranes, and quinones, can rapidly react with DNA and protein once absorbed into the organism [79]. Measuring adducts of these electrophiles with blood electrophiles such as hemoglobin (Hb) and human serum albumin (HAS) can help assess the human exposure to reactive electrophiles [79].

Analytical techniques used in biological approaches can be either instrument-based (e.g., mass spectrometry), or effect-based (e.g., bioassays), or the combination of both [80–82]. As the instrumentbased techniques have been discussed earlier, here we focus on the effect-based and combined approaches. Advantages of effect-based approaches are listed as follows: (1) covers a wide spectrum of modes of action (MOA); (2) provides the opportunity to investigate the actual molecular targets; (3) allows prioritization of individual organisms for further investigation as well as chemical groups for identifying relevant mixture components; and (4) is time/cost effective and can be applied in high through-put screening scenarios [80,83,84]. One recent study screened and evaluated internal exposures of turtles to chemical mixtures based on an in vitro effect-based approach including a battery of sensitive bioassays with different modes of action, including aryl hydrocarbon receptor (AhR)-mediated xenobiotics (AhR-CAFLUX), NrF2-mediated oxidative stress (AREc32), NFκB mediated response to inflammation (NFκB-bla), estrogen binding (VM7Luc4E2), and baseline toxicity (Microtox) [80].

In most cases, routinely measured target compounds only partly account for the biological effects of concern. Thus, a hybrid approach, or effect-directed analysis (EDA), a combination of biotesting, fractionation procedures, and chemical analytical methods, is proposed to identify non-target compounds that cause biological toxicity in the sample [85]. General procedures of EDA are as follows: (1) extract the compounds of interest in an effective and non-selective way; (2) clean-up the extracted samples fraction; (3) conduct bioassay to select the interesting sample fraction; (4) fractionate the selected sample fraction to reduce the complexity; and (5) identify the suspected compounds with instrumental analysis methods. EDA has become a successful strategy in identifying biologically active compounds in environmental samples [82,86]. Although this approach has scarcely been applied in biological samples because of the difficulty in sample preparation, it is a promising approach in identifying the unknown chemical exposure leading to certain specific toxicity endpoint [87]. For instance, by using EDA approach, one study revealed the presence of new environmental contaminants, including di- and two monohydroxylated octachlorinated biphenyls (octaCBs) and linear and branched nonylphenol (NP), in polar bear serum with transthyretin-binding potency bioassay [23].

One well known example of the biological approaches in exposome research is omics profiling, which has found applications in large scale studies at population level [78]. High-dimensional analytical platforms were usually employed in omics profiling [78]. Besides metabolomics, omics profiling approaches also include genomics, epigenomics, transcriptomics, and proteomics [78]. Within the exposome framework, these approaches can provide a deep understanding at system biology-level of how chemical exposure influences the human health [78]. An unprecedented source of information has been produced with respect to the effective biological effects of exposures at omic-level. This omics data can be used in the generation of novel hypothesis to discover the disease etiologies of chemical exposure.

#### *2.3. Other Approaches*

Human exposome encompasses all types of exposures throughout the life course, including those from exo- and endogenous processes at individual level (e.g., environmental contaminants and infection) and general exposures at global level (e.g., climate and social economic status) [20,78]. Biological approaches can identify the influences of these exposures on human health, but are not able to characterize the exposure source, identify the route of exposure, and provide a picture of spatial and temporal variability of the exposure, which are critical in establishing links between exposome

and biological significance [20,78]. While a few external exposures (e.g., environmental contaminants exposure) can be assessed through chemical approaches most external exposures (e.g., air quality and social economic status) need to be measured through other approaches such as questionnaires and static monitors [20,78]. Recently, a variety of novel assessment methods have been employed to quantify the external exposures, including those methods based on geographic information systems, environmental sensors, and personal sensing technology [20,78]. Although this review focuses on the analytical approaches employed in characterizing the chemical exposome, it is highly stressed that these new external exposure measurement methods are an integral part of the entire exposome measurement. With the rapid advancement of technologies, increasing popular applications of these novel methods are expected in exposome studies at both individual and population levels.

#### **3. Measurement-Based Exposome Studies**

Commonly-used strategies for chemical exposome study include top-down and bottom-up approaches [1]. A top-down exposome strategy starts with measuring all chemicals in a subject's (cases and controls) biospecimen, such as blood, at each life stage either through direct measurement or by investigating the physiological effects of exposures. After identifying the variant with significance by a series of data analysis, biological annotation is followed to determine the biological significance of the critical variable. This approach covers both exogenous and endogenous chemicals in the internal chemical environment, offers an efficient means for profiling individual exposures, and is the predominant strategy used in chemical exposome studies to date. A bottom-up strategy starts with a complete measurement of all the chemicals present in each external source (e.g., air, food, water, etc.) of a subject's exposome at each time point. After determining the analyte possessing significant association with health outcomes, uptake and metabolism of the analyte in the human body is evaluated. This approach requires enormous efforts in identifying the important chemicals in various environmental media and would miss important endogenous components in the body generated due to non-chemical factors such as physical activity, noise, inflammation, and social stress. One significant advantage of this approach is that it can provide valuable information regarding the critical chemicals present in each external source, making it complementary to top-down approach. Details of the two approaches as well as the available exposome studies based on these two approaches are discussed below.

#### *3.1. Top-Down Exposome Approach*

The research process of a chemical exposome study employing a top-down approach can be divided into five steps theoretically. (1) sample preparation, including both the collection of samples from groups of population (e.g., health and diseased) and the pretreatment of samples for instrumental analysis; (2) sample analysis, including the selection of instrumental types and methods; (3) data analysis, including chemometric analysis of the raw data and statistical analysis of the processed data; (4) biological annotation, determining the biological significance of the critical exposure identified; and (5) source identification, characterizing the source of the critical exposures identified.

Biomonitoring of biospecimens such as blood has several advantages for the exposome research. Firstly, it allows for the simultaneous measurement of exposures and the metabolic phenotypes. Secondly, it allows for the development and measurement of biomarkers for historic and current exposures. Any exposure including both chemical and non-chemical factors can leave unique signatures in the human body, even in the case of an exposure that has passed a long time ago [88,89]. If the signature is persistent and irreversible, we can assess the health outcomes of historic exposure. Thirdly, it allows for the development and measurement of biomarkers for disease at any stages. There is a long way to go from exposure to observed effects: external dose; internal dose; target organ dose; target organ metabolism; target organ responses; cellular/subcellular dose and interaction; toxic response; observed effects [90]. The biomarker development of diseases can help explain the roles of environmental exposures in the pathogenesis of diseases as well as facilitate the diagnostic and

prognostic of diseases. Lastly, it allows for the retrospective analysis of samples. Biospecimen samples can be stored for a long period of time in appropriate conditions.

Many large cohort studies are using targeted approaches or the combination of both targeted and non-targeted approaches to collect a wide range of exposures, followed by statistical analyses to investigate the relationships between important groups of environmental exposure as well as the associations between exposure and disease. For instance, INfancia y Medio Ambiente (The INMA), a birth cohort study in Spain, investigated the levels of exposure to a wide array of pollutants during pregnancy such as brominated flame retardants, perfluoroalkyl substances, and metals, aiming to examine the role of environmental pollutants during pregnancy and early childhood in relation to child growth and development [91]. The existing exposome initiative in the U.S., the Children's Health Exposure Analysis Resource (CHEAR), is using both targeted and non-targeted analyses through a network of laboratories to provide a comprehensive measurement of myriads of environmental xenobiotics and biological response indicators in various biological samples to better understand the roles of environmental factors in children's health [7].

In addition to the national efforts towards exposome, many individual laboratories in the academic field are also dedicated to the exposome research. Dr. Dean Jones' group established an automated workflow for non-targeted exposome analysis with a dual chromatography (DC)-FTMS, combining a reverse phase C18 chromatography and anion exchange (AE) chromatography [24]. An adaptive processing software package, apLCMS, was exclusively designed for LC-FTMS data analyses [24]. The addition of C18 column increased the *m*/*z* feature detection by 23–36%, yielding a total number of features up to 7000 for individual samples [24]. This exposome workflow was capable of detecting environmental chemicals in the nanomolar and sub-nanomolar concentration ranges [92]. It has been extensively used in multiple studies to simultaneously detect endogenous metabolites with plasticizers, insecticides, fungicides, herbicides, drugs, bacterial products, and correlate environmental chemical exposure with a variety of health outcomes, including tuberculosis disease and neurological development [93–99].

By combining redesigned METLIN Exposome database with XCMS platform and cognitive computing, a newly established nontargeted workflow allows the detection of endocrine disrupting chemicals at low-nanomolar concentrations in human serum and urine and also allows the readout of the biological effect of a chemical [100]. An innovation of this workflow is using artificial intelligence as a potential tool to prioritize findings in exposome studies [100]. It is expected that this workflow will be more extensively used in future exposome research.

#### *3.2. Bottom-Up Exposome Approach*

The complete research process of bottom-up approach-based exposome studies include the following procedures: (1) sample preparation, including sample collection from all the potential external exposure sources and sample pretreatment prior to instrumental analysis; (2) sample analysis, identifying the chemicals present in the samples; (3) data analysis to identify the compounds associated with the disease of interest; and (4) biological validation, studying the metabolism and toxicity of the selected analytes in animal or cell-based models to confirm the relationship hypothesized.

This approach helps identify the critical exposure sources, therefore facilitating the authority or person to take actions to mitigate exposures. It is beneficial for studying those health outcomes which can be mainly ascribed to external exposome, such as allergies [101]. Although genetic factors also contribute to the incidences of allergies, it has been recognized that the increase in allergies observed in the past decades can be explained exclusively by environmental changes occurring in the same time period [101]. A number of studies have suggested that a variety of air pollutants, including volatile organic compounds, formaldehyde, toluene, and polycyclic aromatic hydrocarbons, not only exacerbate but also cause many types of allergies such as atopic dermatitis [102–104]. However, known environmental factors found with a traditional research paradigm cannot explain the increase in the prevalence of allergic diseases worldwide. Therefore, a recent study called for the integration of the

external exposome in the etiopathogenesis of these diseases since a wide range of environmental factors were involved [101].

This approach can also be used to understand the metabolism of xenobiotics in biological organisms. By comparing the mass spectra arising from the environmental media and the organisms living in it, for instance, water and fish, it is possible to differentiate the xenobiotics, metabolized xenobiotics, and endogenous metabolites in the organism [105].

The major limitation of this approach is the laborious workload needed in acquiring the complete exposome from endless external sources. Therefore, careful selection of exposure sources is crucial in the successful identification of the critical exposure related with the target disease. Knowledge of potential mechanisms of disease and the differences between cases and controls are helpful in the selection of appropriate exposure sources.

#### **4. Publicly Accessible Data-Based Exposome Studies**

The ultimate goal of exposome study is to investigate the role of exposome in the pathogeneses of human diseases. Because of the huge challenge in the measurement of exposome, a few studies focused on those human exposures with publicly-available data [14]. Although this data is neither comprehensive across all exposure domains nor longitudinal, it provides an important platform for generating and testing hypothesis between exposome and human health. Such platforms allow the simultaneous analysis of multiple types of exposure (e.g., chemical exposure, social economic status, etc.). One successful example is the National Health and Nutrition Examination Survey (NHANES), a biannual health survey conducted by the U.S. Centers for Disease Control and Prevention, which provides information about the range of representative exposures across the general population. The NHANES includes environmental exposures such as chemicals, nutrients, and infectious agents and the measurement tools include LC/GC-MS, immunological assays, and questionnaires.

Using cross-sectional data from NHANES, Patel et al. employed an Environmental-Wide Association Study (EWAS) approach to investigate the relationships between 266 unique environmental factors and the clinical status for type 2 diabetes [106]. EWAS relies on linear regression models fitted independently for each covariate to separately examine the association between single exposure factor and the health outcome [106]. Then the factors with significant associations were validated across all models [106]. EWAS approach has been used on a widespread basis to assess the comprehensive relationships between a broad range of environmental/behavioral/clinical factors and various types of diseases, including blood pressure [107], type 2 diabetes in the Marshfield Personalized Medicine Research Project Biobank [108], all-cause mortality [109], and telomere length [110]. Besides EWAS, other approaches were also proposed to assess the effects of multiple chemical and non-chemical environmental stressors on health outcomes. For instance, by combining "big data", computational tools, and traditional biostatistics, one research evaluated putative relationships between exposures from natural, built, and social environment domains and lung cancer mortality and mortality disparities across four race and gender groups [111]. A total of 2162 chemical and nonchemical environmental stressor was involved in the study [111].

To move from a single exposure-disease analysis paradigm to cumulative exposure-disease models, advanced biostatistics are required to tackle large, multiple, heterogeneous, and secondary datasets. One recent study compared the performance of commonly-used multiple linear regression statistical methods within exposome context (237 exposure covariates), including EWAS, EWAS-multiple linear regression (MLR), Elastic net, sparse partial least squares regression, Graphical Unit Evolutionary Stochastic Search, and Deletion-Substitution-Addition algorithm (DSA) [112]. Authors found that none of the statistical methods outperformed others across all scenarios and properties examined. However, overall, multivariate methods outperformed univariate approaches in investigating the exposome [112]. Barrera-Gomez et al. extended this work by considering scenarios with statistical interactions and by providing a systematic comparison of methods that have been recommended to search for interactions [113]. In this study, Group-Lasso INTERaction-NET (GLINTERNET) and DSA

had better overall performance than the other methods in detecting two-way interactions, but the sensitivity and false discovery rate was compromised [113]. Therefore, none of the statistical methods has outperformed others in analyzing such a large number of exposures so far. Many factors can affect the selection of the right exposure variant, such as highly correlated exposures and multiplicity. Patel and Ioannidis argued that effects that survive multiplicity considerations and that are large may be prioritized for future scrutiny in the exposome studies [114]. Future efforts may focus on other statistical methodologies such as profile regression, cluster analysis, and even machine learning methods in tackling large exposome datasets [112].

Several metabolome databases also provide essential information about the exposures, such as levels of xenobiotics and their metabolites in the biospecimens. HMDB, currently one of the world's most comprehensive metabolome database, is a great example [115]. Its latest version contains 114,100 metabolites and 21,834 xenobiotics and their metabolites [115]. Many studies have been conducted to interrelate the xenobiotics and their metabolites with endogenous metabolites, metabolic pathway, and health outcomes based on HMDB and other publicly available information. Rappaport et al. obtained human blood concentrations of 1561 small molecules and metals derived from foods, drugs, pollutants, and endogenous processes from the literature [compiled by the Human Metabolome Database (HMDB) and NHANES], and mapped chemical similarities after weighting by blood concentrations, disease-risk citations, and numbers of human metabolic pathways [22]. The results showed that endogenous chemicals, drugs, and food chemicals have similar concentration ranges in human blood, whereas those of pollutants were 1000 times lower [22]. While chemicals in the four classes were equally studied in terms of disease risks, studies of metabolic pathways were dominated by endogenous molecules and essential nutrients [22]. Bessonneau et al. obtained concentrations of 1233 chemicals that had been detected in saliva from the literature integrated into the HMDB, then connected salivary metabolites with human metabolic pathways and PubMed Medical Subject Headings (MeSH) terms, followed by pathway enrichment and pathway topology analyses [116]. The study found that 196 salivary metabolites with KEGG id were mapped into 49 metabolic pathways and associated with human metabolic diseases, central nervous system diseases, and neoplasms [116]. Saliva exposome represents at least 14 metabolic pathways, including amino acid metabolism, TCA cycle, gluconeogenesis, glutathione metabolism, pantothenate and CoA biosynthesis, and butanoate metabolism [116]. These studies offer insights into the roles of environmental factors in the etiology of diseases from a systematic perspective.

#### **5. Challenges in Exposome Research**

The introduction of exposome is expected to have breakthrough changes in uncovering the secrets of human diseases. However, a variety of challenges are present at each step of exposome research, measuring the exposome and linking it with health outcomes [17,117]. These challenges as well as the potential solutions are briefly discussed in this section with a focus on the chemical exposome research.

#### *5.1. Challenges in Measuring the Exposome*

The complexity and heterogeneity nature of exposome and its dynamic variation in both time and space presents a huge challenge in measuring the exposome [18]. The current most comprehensive approach is constructing epidemiological cohort studies with large sample size and long-term follow-ups. Human samples such as blood are collected at each critical stage, including fetal, early postnatal, childhood, teenage, and adult, from the same population. EXPOsOMICS and HELIX projects in the European Union and CHEAR project in the U.S. are exemplary studies regarding exposome research [7,118,119]. However, such studies are usually costly and laborious, and unaffordable for individual laboratories. There is an urgent need for alternative approaches to lower the cost of exposome research. Tooth and hair matrix biomarkers can incorporate the intensity and timing of exposure and has been referred to as "retrospective temporal exposome" [120–122]. This provides an effective approach to study the historical exposures in human beings for individual principal investigators.

Another challenge is that no single analytical technique can exhaust the chemical exposome in one sample because of the remarkable differences of chemicals in a wide range of physicochemical properties including mass, polarity, abundance, lipophilicity and pKa [89,123]. Even with the same technique, different sample processing methods and parameter settings can influence the results significantly. To measure as many chemicals as possible, samples should be carefully portioned and appropriately processed to fit different analytical techniques, which is costly and laborious.

Xenobiotics and their metabolites in the biological samples are usually at trace levels with several orders of magnitude lower than that of endogenous metabolites [22]. This demands high sensitivity for analytical instruments, which is also one of the reasons that mass spectrometry-based analytical platforms are gaining popularity in chemical exposome measurement. Low abundance mass spectra are significantly affected by the instrumental noise. Further, isotope pattern observed in high abundance mass spectra is usually not available for xenobiotics, which increases the difficulty in the identification of these compounds. Recently, efforts are being made to increase the intensity of these low abundance signals. For instance, one research group from Singapore isotopically labelled those xenobiotic biomarkers with common functional groups including phenolic hydroxy, carboxyl, and primary amine [124]. This method has improved sensitivity of 2–1184 fold for xenobiotics compared with other mass spectrometry based methods [124]. It has also been reported that increasing the number of replicate injections can help improve the reliability in low abundance chemicals measurement in high resolution metabolomics [125]. One recent study recommends integrating ion mobility spectrometry into mass spectrometry-based exposome measurements, which can provide increased overall measurement dynamic range and thus result in frequent detections of lower abundance molecules that are previously undetected [68].

HRMS data analysis such as unknown identification is also a huge challenge in HRMS based chemical exposome studies. None of the chemometric and bioinformatic tools available can successfully group and align all the features correctly. Every algorithm has its own pros and cons. In addition, although the compound databases are increasing the coverage annually, they are still far behind the number of chemicals available. More than 60 million chemicals are present in PubChem, however, only around 220,000 MS/MS spectra from 20,000 molecules or so are accessible in the databases [126].

#### *5.2. Challenges in Associating Exposome with Diseases*

In exposome, we are dealing with thousands of environmental risk factors that vary by source, place, and time. These factors affect human health differently depending on the exposure route, exposure timing window, dose, and specific target organ. In addition, these factors can also interact with each other to have synergistic, additive, or antagonistic effects when exerting effects on certain health outcome. To fully understand the effects of exposome on human health, it is necessary to integrate all the exposure factors and evaluate their effects systematically, which poses a great challenge.

In addition, exposome mapping could discover hundreds or even thousands of altered molecular features associated with disease endpoints. However, it is difficult to identify key exposome features that may drive disease or contribute to disease etiology. To address this, advanced statistical methodologies, such as machine learning and artificial intelligence, hold the promise of pinning down molecular features that play a key role in the pathogenesis of human disease.

#### **6. Future Directions**

Many factors can contribute to the further development of exposome research, including advancement in analytical platforms, high-throughput statistical analysis, HRMS data mining algorithms, large database of chemical substances and mass spectra, and biochemical pathways. It is the rapid advancement in the high-resolution metabolomics techniques that provides a high-throughput and affordable platform for the monitoring of environmental exposures in human beings. Instruments with high sensitivity, broad dynamic range, high resolution, high mass accuracy and a low cost are desirable in measuring the chemical exposome. Exciting achievements are also expected from the development of effective bioinformatic computation. Excellent algorithms are needed to help remove the noise in the mass spectra, group peaks, generate MS/MS spectra and retention time information, etc. It is time for exposome to embrace techniques such as machine learning.

With the establishment of exposome ontology through national studies such as CHEAR, more exposome researches are expected from the academic field. Thus, it is necessary to establish standards to allow cross comparison and validation. If every study follows the established standards, the data can be used for the integrative and systematic study in the future contributing to the complete picture of the exposome research. To establish the relationship between exposome and health outcomes, traditional statistical analysis needs to embrace big data analysis techniques to pinpoint the critical exposures. To better correlate the exposome with biological effects, high-throughput analytical techniques are needed to integrate data in exposome with other omics analyses including genome, transcriptome, proteome, and epigenome. This will help the principal investigators to study the mechanistic basis of the exposome in an affordable way.

#### **7. Conclusions**

Exposome research paradigm provides a great opportunity to identify critical non-genetic factors that contribute to the onset and progress of various diseases. This paper discussed the commonly used measurement techniques in chemical exposome research and reviewed the available chemical exposome studies. As the technologies moving forward along with the establishment of exposome ontology, more exciting discoveries are waiting in the journey to uncover the roles of non-genetic factors in the pathogeneses of human diseases.

**Author Contributions:** Conceptualization, H.R.; original draft preparation, J.X.; review and editing, all.

**Funding:** This research was funded in part by National Institute of Environmental Health Sciences under grant number P30ES025,128, R01ES024950, P30ES010126 and the North Carolina State University.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Gut Microbiome Toxicity: Connecting the Environment and Gut Microbiome-Associated Diseases**

#### **Pengcheng Tu, Liang Chi, Wanda Bodnar, Zhenfa Zhang, Bei Gao, Xiaoming Bian, Jill Stewart, Rebecca Fry and Kun Lu \***

Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA; ptu@live.unc.edu (P.T.); liang16@live.unc.edu (L.C.); wanda\_bodnar@unc.edu (W.B.); zhenfaz@email.unc.edu (Z.Z.); wintergb2012@gmail.com (B.G.); bxmroly@uga.edu (X.B.); jill.stewart@unc.edu (J.S.); rfry@unc.edu (R.F.) **\*** Correspondence: kunlu@unc.edu; Tel.: +1-919-966-7337

Received: 4 September 2019; Accepted: 6 March 2020; Published: 12 March 2020

**Abstract:** The human gut microbiome can be easily disturbed upon exposure to a range of toxic environmental agents. Environmentally induced perturbation in the gut microbiome is strongly associated with human disease risk. Functional gut microbiome alterations that may adversely influence human health is an increasingly appreciated mechanism by which environmental chemicals exert their toxic effects. In this review, we define the functional damage driven by environmental exposure in the gut microbiome as gut microbiome toxicity. The establishment of gut microbiome toxicity links the toxic effects of various environmental agents and microbiota-associated diseases, calling for more comprehensive toxicity evaluation with extended consideration of gut microbiome toxicity.

**Keywords:** gut microbiome; environment; chemical toxicity

#### **1. Introduction**

The human gut microbiome including the microorganisms, their genomes, and the surrounding environment in the gut, has received unprecedented attention over the past decade [1]. Mounting evidence suggests that the metabolic activities in the gut microbiome are profoundly intertwined with human health and disease [2]. A number of important functions performed by the gut microbiome are well recognized including the digestion of polysaccharides, biosynthesis of vitamins and nutrients, colonization resistance, and immune system modulation [3–5]. Moreover, the effects of the gut microbiome on host metabolism and physiology extend beyond the gut to distant organs such as the liver, muscle, and brain [2,6]. Due to its crucial role in human fitness, the gut microbiome is now considered as a new organ in the human body [7–9]. It is unequivocal that the gut microbiome functions properly on the premise that a normal gut microbial homeostasis is maintained [2]. However, the constitution and functionality of the gut microbiome can be readily influenced by diverse intrinsic and extrinsic factors [10]. For example, exposure to various xenobiotics leads to functional perturbation in the gut microbiome [11–15]. Mounting studies suggest that these environmentally induced perturbations are potentially linked to elevated disease risks [2,16]. Adverse health outcomes including inflammatory bowel disease (IBD), obesity, diabetes, cardiovascular disease, liver disease, colorectal cancer, and neurological disorders can be at least in part attributed to undesirable functional alterations in the gut microbiome [17–23].

Certain environmental toxins can induce damage and dysfunction in the liver, which is termed as liver toxicity. Exposure to these toxins changes the morphology and functionality of the liver, hence leading to liver diseases. Similarly, exposure to some environmental chemicals causes structural

differences and functional alterations in the gut microbiome, which probably results in a series of adverse health outcomes. For instance, arsenic exposure can perturb the composition and metabolites of the mouse gut microbiome, which potentially contributes to its toxicity [12]. Given the increasingly recognized role of the gut microbiome in human health coupled with its susceptibility to environmental insults, it is of significance to define gut microbiome toxicity. As the gut microbiome is viewed as a new human organ by its importance, we accordingly define the environmentally driven functional damage in the gut microbiome as gut microbiome toxicity. The development of gut microbiota-related diseases can be generalized as environmental factors leading to deleterious alterations in the gut microbiome, which adversely affect human health via host–gut microbiota interactions. In other words, gut microbiome toxicity triggered by environmental exposures contributes to gut microbiota-related adverse outcomes. By including gut microbiome toxicity into the organ toxicity family, we can now discuss the relationship between the environment and gut microbiota-related diseases in the context of toxicology (Figure 1).

**Figure 1.** A potential new member of the organ toxicity family: gut microbiome toxicity. Toxicity of organs including brain, liver, and kidney is well defined and acknowledged. Similarly, the discussion of gut microbiome toxicity encompasses environmental exposures (causes), interactions between the gut microbiome toxicity and human diseases (mechanisms of gut microbiota-related diseases), biomarker and assessment (diagnosis), and modulation (treatment).

Environmental exposure is a significant risk factor for a series of human diseases, overlapping those diseases that are associated with the gut microbiome [24–26]. Thus, gut microbiome toxicity may be the missing link between environmental exposure and microbiome-related human diseases. Moreover, the current toxicity testing system does not include toxic endpoints regarding the effects of environmental chemicals on the gut microbiome [27,28]. Considering the potential involvement of the gut microbiome in human disease, it is imperative to integrate gut microbiome toxicity into the toxicity assessment of environmental exposure. Thus, the establishment of gut microbiome toxicity may offer insights regarding the mechanistic basis underlying the toxicity of environmental chemicals, calling for more comprehensive risk assessment with the integration of gut microbiome toxicity. Additionally, environmentally driven alterations in the gut microbiome are not necessarily always adverse. By functional damage, we refer in particular to those that potentially contribute to adverse health outcomes, for instance, the production of pro-inflammatory metabolites. With the introduction of 'gut microbiome toxicity', we highlight the underappreciated mechanisms by which environmental factors lead to or exacerbate diseases through perturbing the gut microbiome functions.

In this review, we carefully define gut microbiome toxicity as environmentally driven functional damage in the gut microbiome. Functional damage may include changes in bacterial metabolites, loss of bacterial diversity, or effects on energy metabolism and balance. We focus on recent studies in support of the establishment of gut microbiome toxicity, and we accordingly discuss the environmental exposures, metabolic interactions in human disease, biomarkers and assessment, and modulation (Figure 1). Specifically, we review recent studies demonstrating the functional perturbation in the gut microbiome driven by various xenobiotics such as antibiotics, heavy metals, pesticides, and artificial sweeteners. These functional changes include, but are not limited to, alterations in the bacterial production of metabolites, diversity loss in the bacterial community, and interference in energy metabolism, which are further linked to the development of gut microbiota-related diseases. Moreover, microbiome changes including compositional and functional changes can serve as biomarkers for gut microbiome toxicity. Additionally, we briefly summarize current approaches for the assessment of gut microbiome toxicity as well as effective gut microbiome modulation.

#### **2. Environmental Exposures**

The fact that a number of xenobiotics can trigger gut microbiome toxicity suggests the underestimation of the toxic effects of specific chemicals. On one hand, the induction of gut microbiome toxicity may be considered a potential new mechanism by which known toxic chemicals (e.g., heavy metals, pesticides) lead to or exacerbate human diseases. On the other hand, it is of necessity to reconsider the health effects and acceptable daily intake (ADI) of widely-used chemicals such as food additives in the context of their contribution to gut microbiome toxicity. The impact of xenobiotics on the human gut microbiome can be direct or indirect. The human gut microbiome encodes more diverse metabolic enzymes, which greatly expands the repertoire of biochemical reactions within the human body [29]. Some environmental chemicals can directly affect the gut bacteria by interrupting specific metabolic pathway or gene expression, leading to distinct selection pressures, hence shaping the gut microbial community due to the uniqueness of the set of metabolic pathways and genome possessed by different bacterial species [29]. Therefore, the selection of resistant bacteria upon certain exposure could lead to an unbalanced gut eco-system. Additionally, some environmental chemicals can indirectly impact the gut microbiome through the influence on host physiology (e.g., gut mucosa [30]) and cell-to-cell communications of bacteria (e.g., quorum sensing [31]). That being said, the mechanistic basis underlying the microbial perturbation induced by specific chemical exposure remains elusive. Here, we highlight representative xenobiotics such as antibiotics, heavy metals, pesticides, and artificial sweeteners that cause gut microbiome toxicity with significant functional alterations.

#### *2.1. Antibiotics*/*Drugs*

It is commonly accepted that antibiotic administration, especially broad-spectrum antibiotics, severely impacts commensal bacteria. Both short-term and long-term antibiotic treatments lead to gut microbiome toxicity, although partial recovery may occur [32,33]. In many cases, effects of antibiotics on bacterial communities result in diversity loss and compositional imbalance [34]. Moreover, antibiotics

not only disturb the gut microbiome at the compositional level, but also substantially change its functional profiles. For example, a recent study used a multi-omics approach to resolve the changes induced by beta-lactam in human gut microbiome [35]. The results showed that beta-lactam treatment caused both taxonomic and functional alterations in gut microbiome supported by alterations at the metagenomic, metatranscriptomic, metametabolomic, and metaproteomic levels. Antibiotic exposure in mice has been linked to diseases such as obesity and diabetes [36,37]. Aside from antibiotics, non-antibiotic drugs also affect the gut microbiome [38]. For instance, metformin [39], non-steroidal anti-inflammatory drugs [40], proton pump inhibitors [41], and atypical antipsychotics [42] are reported to have effects on the gut microbiome, although the health consequences remain underexplored. A most recent study tested 1200 marketed drugs by in vitro screening to investigate their effects on the gut microbiome [43]. A quarter of human-targeted drugs were discovered to have effects on the gut bacteria to some degree, indicating the potential of medication to induce gut microbiome toxicity.

#### *2.2. Heavy Metals*

Heavy metals continue to be a class of intensely-studied environmental contaminants. However, the role of heavy metals in gut microbiome toxicity still remains underappreciated. In fact, the gut bacteria play an important role in the biotransformation of heavy metals, which may promote or attenuate their toxicity. For example, human gut bacteria are able to transform inorganic arsenic into less toxic organic arsenic species [44], and demethylation of methyl-mercury by gut bacteria can generate more toxic inorganic mercury [45]. Rats exposed to heavy metals including arsenic, cadmium, cobalt, chromium, and nickel exhibited significant changes in their gut microbial compositions [46]. Moreover, the functional profiles in the gut microbiome can be perturbed by heavy metals. Four weeks of arsenic exposure in drinking water (10 ppm) caused significantly different metabolite profiles in the mouse gut microbiome [12]. Likewise, 13 weeks of arsenic exposure with an environment-relevant dose (100 ppb) also perturbed diverse bacterial metabolic pathways [47]. Arsenic-induced gut microbiome toxicity provided a new angle to look at the mechanism of arsenic toxicity. Follow-up studies evaluating arsenic metabolism further demonstrated that arsenic-induced gut microbiome toxicity can be affected by factors including host genetics [48], gender [49], and bacterial infection [50]. Moreover, different arsenic doses (10 ppm and 100 ppb) induced different levels of perturbation in the mouse gut microbiome, indicating the dose-dependent effects of arsenic exposure, which together with toxicity response thresholds of arsenic-induced gut microbiome toxicity need to be further defined. In addition, exposure to manganese and lead disturbs the gut microbial functions of mice with perturbed pathways and metabolites [51,52].

#### *2.3. Pesticides*

Excessive use of pesticides in agriculture has raised concern about their health effects. The argument that certain pesticides are safe to humans because their targeted pathways do not even exist in the human body fails to consider the microbes in the gut [53]. For example, herbicides like 2,4-dichlorophenoxyacetic acid (2,4-D), which impact plant hormones, may affect gut bacteria because not only plants but also bacteria can synthesize plant hormones [54]. Likewise, the shikimate pathway, the target of herbicide glyphosate, is commonly present in human gut bacteria [55,56]. In bacteria, this pathway has an important function linking the metabolism of carbohydrates to the biosynthesis of folates and aromatic amino acids. Several studies have demonstrated the association between gut microbiome toxicity and pesticide exposure. For example, the fungicide imazalil changed the composition of gut microbiome in zebrafish and mice [57,58]. Of interest, a recent study showed that the mouse gut microbiome was perturbed by 13 weeks of diazinon exposure (4 ppm) [13]. Bacterial genes and metabolites involved in neurotransmitter synthesis were significantly perturbed, suggesting that diazinon-induced gut microbiome toxicity with altered bacterial biosynthesis of the neurotransmitter may be partially responsible for the neurotoxicity of diazinon [59,60]. In addition, exposure to diazinon and malathion impacts the quorum sensing of gut bacteria, providing evidence that affecting bacterial communications may be one of the underlying mechanisms of gut microbial perturbations [61,62].

#### *2.4. Artificial Sweeteners*

Food additives have facilitated the development of the modern food industry. Normally, food additives (e.g., artificial sweeteners, emulsifiers, preservatives) are added in food products with an approved safe amount. Nevertheless, gut microbiome toxicity was not taken into consideration when the related standards were determined. Many artificial sweeteners are considered safe because they are poorly metabolized by the human body [29]. However, the gut bacteria are actively involved in the biotransformation. For example, cyclamate, which is currently banned in the USA, can be metabolized by gut bacteria into cyclohexylamine, which is carcinogenic [63]. Artificial sweeteners stevioside and xylitol can also be metabolized by the gut bacteria [64,65]. Several studies have demonstrated that some artificial sweeteners and emulsifiers were able to induce gut microbiome toxicity with potential gut microbiota-related health consequences. For instance, in an elegantly conducted study by Suez and colleagues, consumption of saccharin induced both compositional and functional changes in mouse gut microbiome that might be involved in the development of glucose intolerance [66]. Another study reported similar results with increased inflammatory levels in addition to gut microbial perturbation induced by saccharin in mice [15]. Additionally, artificial sweeteners acesulfame potassium [61], sucralose [67], aspartame [68], and neotame [69] can also perturb bacterial metabolites in concert with health implications including obesity and inflammation. Moreover, another study found that two commonly used emulsifiers altered mouse gut microbial composition together with elevated inflammatory levels [70].

#### *2.5. Others*

The above discussion is not intended to be exhaustive. More information could be referred to in recent reviews regarding the relationship between xenobiotics and the gut microbiome [11,29,71]. We emphasize functional changes in the gut microbiome induced by environmental exposure in the current review, therefore studies were included documenting not only the compositional shifts after exposure, but also functional alterations manifested by functional metagenomics and metabolomics. A rapidly-increasing list of xenobiotics is linked to gut microbiome toxicity. Some are commonly present in our daily life; a typical example is the antibacterial and antifungal agent triclosan. It has been repeatedly reported that triclosan induced changes in the gut microbiome using multiple animal models [14,72–74]. However, the effects of triclosan on human gut microbiome remain controversial [75]. Furthermore, exposure to nicotine (a major toxic component of tobacco smoke) also perturbed the gut microbiome, affecting bacterial production of neurotransmitters in mice [76]. Such a large range of chemicals that may induce gut microbiome toxicity supports the necessity of considering gut microbiome toxicity regarding the toxicity evaluation of environmental agents.

#### **3. Relationship between Gut Microbiome Toxicity and Human Diseases**

The mutually beneficial relationship between the gut microbiome and the host is built on the premise that a well-balanced gut microbiota is maintained [2]. However, when afflicted with gut microbiome toxicity, functional alterations occur in the gut microbiome. Although it is difficult to disentangle these alterations, changes in microbial metabolites, diversity loss, and interference in energy metabolism are three major types of microbial disturbances that may adversely impact the host health via multiple host–microbiota axes, potentially leading to increased disease risks. Therefore, gut microbiome toxicity is a new link between the environment and human diseases. It should be noted that not all changes in the gut microbiome associated with environmental exposure are necessarily adverse. Nevertheless, it is of significance to establish the role of the gut microbiome in the toxicity of a number of environmental toxic agents, which has been largely underappreciated in the chemical research of toxicity. In this part, we discuss the connection between gut microbiome toxicity and human diseases, providing some mechanistic insights regarding environmentally driven gut microbiome-associated diseases.

#### *3.1. Changes in Microbial Metabolites*

Production of functional metabolites by bacteria plays a key role in human health and disease [4]. Gut microbiome toxicity has an altered bacterial metabolite profile, which influences host metabolism and physiology in a significant way. First, numerous bacterial metabolites act as signaling molecules through binding to receptors and activating diverse signaling cascades. Pathogen-associated molecular patterns (PAMPs) including lipopolysaccharide (LPS) and peptidoglycan can bind to Toll-like receptor 4 and nucleotide-binding oligomerization domain, respectively; both of which lead to pro-inflammatory effects [77–79]. Classic metabolites of gut bacteria, short-chain fatty acids (SCFAs), and bile acids can also function as signaling molecules and bind to cellular receptors. Specifically, SCFAs can bind to G-protein-coupled receptors (GPCRs), and bile acids can bind to GPCR TGR5 and nuclear receptor farnesoid X receptor (FXR) [80]. Activation of signaling pathways is implicated in important biological functions; the gut microbiome may therefore contribute to human health and disease by regulating metabolic activities involved in the production of SCFAs and bile acids. For instance, SCFAs and bile acids can modulate the secretion of glucagon-like peptide-1 (GLP-1) by binding to GPR43 [81] and TGR5 [82], respectively, which affects insulin secretion and glucose homeostasis. Perturbation in those bacterial activities may affect the risk of diabetes. In addition, tryptophan metabolites produced by bacteria such as indole 3-propionic acid and indole-3-acetic acid regulate intestinal immune cells and barrier functions through the activation of aryl hydrocarbon receptor (AHR) and the pregnane X receptor (PXR) [83–85]. AHR activation is involved in inflammatory bowel disease (IBD) among other diseases, and it is suggested that a reduction in bacterial tryptophan metabolism may contribute to IBD [85]. Second, some bacterial metabolites are strongly associated with specific diseases and phenotypes. A compelling example is the association of trimethylamine N-oxide (TMAO) and cardiovascular disease [19]. The gut bacteria can convert dietary components choline and L-carnitine to trimethylamine (TMA), which is further metabolized into TMAO in the liver. Gut microbiome-derived TMAO is highly correlated with cardiovascular disease risks. Likewise, products of protein fermentation (e.g., N-nitroso compounds, polyamines) derived by gut bacteria exert carcinogenetic effects and promote colorectal cancer [21]. Third, microbiome-derived metabolites play a role in brain functions through the gut–brain axis, many of which are neurotransmitters or their precursors (e.g., serotonin, gamma-aminobutyric acid) [4]. As mentioned previously, bacterial metabolites that are neurotransmitters were perturbed by environmental chemicals such as organophosphate diazinon and nicotine, which may partially explain their neurotoxicity. Additionally, the gut microbiome is an important source of beneficial vitamins and nutrients, therefore reduction in the bacterial production of those beneficial metabolites could be detrimental to human health [86]. Taken together, these examples support that gut microbiome toxicity can lead to diseases via altered metabolite profiles.

#### *3.2. Diversity Loss*

Diversity loss has been associated with many microbiota-related diseases such as IBD [87,88], irritable bowel syndrome (IBS) [89], acute diarrhea [90], and *Clostridium di*ffi*cile*-associated disease (CDAD) [91]. Trillions of microorganisms residing in the human gut form a complex microbial ecosystem, which is deeply intertwined with human biology [34]. Therefore, it is important to view the gut microbiome from an ecological perspective, although it is formidable due to fluctuations over time and variations between individuals [34]. Resilience is the extent of perturbation that an ecosystem can tolerate before it equilibrates toward a different state [92]. Resilience of the gut microbiome is crucial to colonization resistance to pathogens [34,93]. Species richness and evenness is key to the resilience of the gut microbial community. Gut microbiome with species-rich communities is less susceptible to perturbation and stress because different species are specialized to each potentially-limiting resources [94]. Moreover, high species richness enables alternative species with similar functions to fill a niche and maintain the diversity when the original species is compromised [95]. The diversity of the gut microbial ecosystem can be compromised by environmental factors (e.g., antibiotics), which makes it less resilient and more susceptible to pathogen invasion. For instance, antibiotics can induce

changes in the gut microbiome and metabolic features that increased its susceptibility to *Clostridium di*ffi*cile* infection [96]. In addition, a core set of gut microbial species across individuals does not exist. However, a functional core microbiome is shared with similar functional gene profiles [97]. Maintaining the functional core of the gut microbiome is indispensable because normal functioning of the human biology relies in part on the essential functions performed by the gut microbiome. However, exposure to toxic environmental chemicals possibly reduces species richness and diversity of the gut microbiome, leading to potential dysfunction.

#### *3.3. Interference in Energy Metabolism*

Accumulating evidence suggests that the gut microbiome plays a crucial role in energy metabolism. Humans cannot degrade most plant polysaccharides, which instead, can be utilized by the gut bacteria, producing SCFAs that are important energy substrates [80]. Direct evidence supporting the role of the gut microbiome in energy balance is that germ-free rats have reduced intestinal levels of SFCAs and doubled excretion of calories through urine and feces [98,99]. It is suggested that the capacity for the energy harvest of the gut microbiome is correlated with its microbial composition [100], specifically, the ratio of two major phyla *Firmicutes* and *Bacteroidetes*. Moreover, enriched genes encoding enzymes that are important for the initial steps of complex carbohydrate metabolism were found in the gut microbiome of obese mice [101]. Studies showed that the energy balance and body weight of the host is associated with the gut microbiome types. For instance, germ-free mice with fecal microbiota transplantation from obese mice gained more weight than that from lean mice [100]. Likewise, mice with the microbiota from people afflicted with Kwashiorkor, a form of malnutrition, suffered severe weight loss [102]. Thus, it is possible that gut microbiome toxicity interferes with the energy extraction and harvest, leading to diseases such as obesity or malnutrition.

#### **4. Biomarkers and Assessment of Gut Microbiome Toxicity**

Routine toxicity screening and evaluation of environmental chemicals fail to consider gut microbiome toxicity. There is no toxic endpoint currently established to report the relative toxic effects of certain chemicals on the gut microbiome. Thus, it is imperative to assess the functional alterations induced by various environmental chemicals, or at least the chemicals of frequent and long-term exposure (e.g., artificial sweeteners). Current approaches for the assessment of gut microbiome toxicity mainly comprise an integration of animal models (e.g., mouse, rat, and germ-free animals) and the meta-omics toolkit [10]. The use of animal models enables us to mimic the progress of gut microbiome toxicity under environmental exposures; the meta-omics toolkit comprises sequencing-based gene profiling and mass spectrometry-based metabolite profiling. Meta-omics comprise approaches that reveal both compositional levels and functional levels. Compositional profiling, that is, taxonomic profiling, provides details of the microbial constitution and diversity. However, knowing the taxonomic information alone does not necessarily lead to an accurate understanding of microbiome functions due to the existence of functional redundancy in the microbiota [34]. In the context of gut microbiome toxicity, the functional changes including the genes, mRNAs, proteins, and metabolites are what we should emphasize. Furthermore, humanized gnotobiotic mice with gut microbiota more similar to that of humans allow for better elucidation of the interactions between human gut microbiome and the environment [103]. The use of germ-free mice and in vitro techniques extends the observational studies to causality [10]. The accurate assessment of gut microbiome toxicity provides knowledge of how gut microbes react to environmental exposures, offering insights into the mechanistic basis of chemical-induced microbial perturbations and diagnostic markers for microbiota-associated diseases.

In order to diagnose gut microbiome toxicity, specific and effective biomarkers are needed. The gut microbiome and its functions will change under various environmental pressure at almost all times, however, not all changes are necessarily adverse and lead to adverse outcomes. Therefore, it is imperative to develop strategies identifying alterations that adversely influence human health. Currently the techniques and approaches used for gut microbiome assessment are usually at the

meta-level; thus, the pinpoint of specific bacterial genes or metabolites that can be used to sensitively indicate environmentally induced dysfunction in the gut microbiome is warranted. Additionally, it should be noted that biomarker development may be on a case-by-case basis. Different xenobiotics would induce distinct gut microbiome changes. The elucidation of the role of the gut microbiome in the toxicity of certain exposure is the premise of biomarker development of gut microbiome toxicity.

Biomarkers are commonly used as primary end points in basic and clinical research, connecting environmental exposures to health outcomes [104]. Incorporating gut microbiome toxicity, our understanding of biomarkers should include functional changes in the gut microbiome as critical indicators in progressions from exposure to microbiome-associated diseases [105]. The functional role of the gut microbiome in host metabolism and physiology is largely determined by microbiome metabolic profiles, especially metabolic pathways and products of gut bacteria. Exposure to a range of xenobiotics would lead to perturbation in microbiome profiles, thereby resulting in functional alterations and gut microbiome toxicity. An in-depth look at microbiome changes upon various environmental exposures will provide insights regarding biomarkers of gut microbiome toxicity induced by specific environmental chemicals. While keeping the host in the picture, development and characterization of sensitive and robust biomarkers of gut microbiome toxicity could spur new advances in environment–microbiome interactions and microbiome-related diseases.

Biomarkers of gut microbiome toxicity could be bacterial species, genes, or metabolites, even a combination of several these markers. Signature changes in the gut microbiome upon exposure to certain chemical could be used to indicate an exposure to or the effect of specific xenobiotics, which provides a novel and potentially less invasive method for environmental health monitoring. More importantly, if the underlying mechanisms of chemical toxicity involves perturbation of the gut microbiome with specific functional alterations, then these alterations can also be used as potential biomarkers of environmentally driven health conditions.

Recent studies have documented the functional changes in the gut microbiome upon exposure to diverse xenobiotics (Table 1) [12–15,47,51,52,61,62,66–70,76,106–109]. Consistent changes in gut microbial profiles could be potential biomarkers of gut microbiome toxicity associated with specific chemical exposures. Outlining changing patterns and trajectories of microbial composition offers a sketch of biomarkers for gut microbiome toxicity. The *Firmicutes*/*Bacteroidetes* ratio is suggested to be indicative of energy harvesting capacity in the gut microbiome that is associated with host adiposity [100]. Likewise, *Enterobacteriaceae* is associated with gut inflammation [110]. The ratio of *Firmicutes* and *Bacteroidetes* as well as the abundance of *Enterobacteriaceae*in the gut can be readily changed by chemicals such as carbendazim [108] and aspartame [68]. Thus, such taxonomic characteristics can serve as biomarkers of gut microbiome toxicity associated with health outcomes such as inflammation and obesity. Moreover, distinctive changes in functional profiles such as key metabolites and metabolic pathways could serve as more relevant biomarkers because alterations in functional profiles directly influence the host. For example, arsenic exposure perturbed the gut microbial metabolite profiles, especially indole-containing metabolites, isoflavone metabolites, and bile acids [12]. Alterations in these functional metabolites could be a potential new mechanism of arsenic toxicity, and particularly, changes of these metabolites (e.g., bile acids and indole-containing compounds) can be used as biomarkers of arsenic-induced gut microbiome toxicity. Likewise, consumption of artificial sweeteners is associated with increased levels of pro-inflammatory metabolites and genes in the gut microbiome. This may be used as bioindicators of artificial sweeteners-induced gut microbiome toxicity that consequently leads to inflammation [15,67]. In addition, diazinon changed the bacterial pathways and metabolites involved in neurotransmitters in a gender-dependent manner, indicating that those bacteria-derived neurotransmitters can be biomarkers to probe gut microbiome toxicity arising from chemicals that have neurological toxicity [13]. The gender-dependent effect also indicates individual variation in biomarkers of gut microbiome toxicity resulting from gender differences in the gut microbiome.


chemicalthatcanserveaspotentialbiomarkersofgutmicrobiome

More efforts should be put into the search and validation of biomarkers of gut microbiome toxicity, which would further elucidate the link between environmental chemicals and microbiome-related disease. Delineating these microbial changes and elucidating their biological effects is undoubtedly challenging due to the complexities within the gut microbiome as well as the intertwinement between the gut microbiome and other systems including immune, endocrine, and nervous systems. However, recent advances and emerging approaches are enabling progress toward a better understanding of gut microbiome toxicity biomarkers, which will inform toxicology risk assessment and development of therapeutic interventions via modulation of the gut microbiome.

#### **5. Gut Microbiome Modulation**

It is increasingly acknowledged that one of the crucial mechanisms underlying chemical toxicity is perturbation of the gut microbiome functions. The inclusion of the modulation section corresponds to 'treatment' in traditional organ toxicity and related diseases. Therefore, it is reasonable to include the treatment of gut microbiome-associated diseases—gut microbiome modulation.

The gut microbiome is becoming an attractive therapeutic target, especially now with its role well recognized in human health and disease. Current approaches for gut microbiome modulation including fecal microbiota transplantation (FMT), probiotics, and prebiotics are mainly untargeted without predictable outcomes [10]. To move from untargeted toward targeted modulation, a healthy gut microbiome needs to be defined. A consensus on the healthy endpoints of gut microbiome modulation remains elusive, which is a major challenge [111]. Nevertheless, the potential of targeted, hypothesis-driven gut microbiome modulation has been demonstrated in some recent studies. Use of whole foods or food components as dietary intervention to modulate the gut microbiome has received increasing attention due to their low toxicity profiles and high patient compliance [112]. Even with well acknowledged health benefits and capacity of gut microbiome modulation, it should be noted that there is evidence that dietary fiber could also possibly exacerbate gut conditions [113,114]. Here, we use *Akkermansia muciniphila* (*A. muciniphila*) as an example to review recent progress on attempts at targeted microbiome modulation.

*A. muciniphila*, a mucin-degrading bacterium commonly present in human and mouse gut microbiome, has many probiotic effects in gut barrier function, glucose homeostasis, and inflammation in humans and diverse animal models [115–119]. Several studies reported targeted gut microbiome modulation with increased *A. muciniphila* population via consumption of whole foods or food components. For example, consumption of several berry fruits including cranberries and raspberries promoted increased content and enhanced function of *A. muciniphila* in the gut microbiome in rodent studies. Specifically, cranberry extract improved insulin sensitivity and reduced weight gain in concert with a significant increase of *A. muciniphila* in diet-induced obese mice [120]. Likewise, black raspberries boosted *A. muciniphila* population in the gut microbiome together with profound changes in microbial functions and metabolites [121–123]. The polyphenols abundant in berry fruits could be a reason that *A. muciniphila* thrives. Feeding polyphenols from grapes to mice showed similar results with a drastic increase of *A. muciniphila* [124]. The gut microbiome offers a link between polyphenols and their diverse beneficial effects because polyphenols are poorly absorbed and metabolized by the human body [124]. Meanwhile, *A. muciniphila* uses mucin as carbon, nitrogen, and energy sources [125]. Goblet cells are the major producer of mucin in the intestinal epithelium [126]. It is reported that the number of goblet cells and the thickness of intestinal mucosa were increased in rats fed oligofructose [127]. Therefore, oligofructose may be an alternative factor for the increase of *A. muciniphila* in mice fed berries, which is supported by the study that administration of oligofructose did increase the *A. muciniphila* population in the gut microbiome of mice [128]. Of particular interest, metformin, medication to treat type 2 diabetes, also promotes *A. muciniphila* population in the gut microbiome, which is believed to contribute to its therapeutic effects [129,130]. Perturbation by environmental toxic chemicals and modulation by dietary components regarding the gut microbiome are fundamentally similar, except with different

expected outcomes. Knowledge of how gut microbes react to xenobiotics and dietary components will address gaps in our understanding of both perturbation and modulation of the gut microbiome.

#### **6. Conclusions**

To summarize, exposure to xenobiotics such as antibiotics, heavy metals, and artificial sweeteners induces gut microbiome toxicity. Compositional alterations and functional changes occur along with this process in the gut microbiome, which can serve as potential biomarkers of gut microbiome toxicity. These chemical-induced perturbations lead to human diseases via several mechanisms including changes in the metabolite profiles, diversity loss, and altered energy metabolism (Figure 2).

**Figure 2.** Schematic representation of how gut microbiome toxicity connects the environment and microbiota-associated diseases. Triangles of different colors at the bottom represent functional metabolites produced by a perturbed gut microbiome such as signaling molecules, detrimental metabolites, and neurotransmitters, which could potentially contribute to adverse health outcomes.

Given the continued enthusiasm in gut microbiome research, it is now an opportune time to examine environmentally induced gut microbiome alterations through the lens of toxicology. Although strong connection has already been established between gut microbiome disturbances and environmental exposure, the mechanisms of these disturbances and health implications await future studies. The goal of this paper was to establish and emphasize gut microbiome toxicity with the

definition of chemical-driven functional damage in the gut microbiome and to review the current state of knowledge regarding biomarker, assessment, and modulation of gut microbiome toxicity. Toxic effects of various environmental agents on the gut microbiome must not be underappreciated. The integration of gut microbiome toxicity endpoints into the evaluation of chemical toxicity will provide a better understanding of the associations between the environment and human health and disease, and will facilitate the development of diagnostic markers and therapeutic interventions.

**Funding:** This research was funded by the University of Georgia, the University of North Carolina at Chapel Hill, and the NIH/NIEHS (R01ES024950).

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
