Evaluation and identification of dioxin exposure biomarkers in human urine by high-resolution metabolomics, multivariate analysis and in vitro synthesis

Highlights

• Untargeted metabolomics by UHPLC-QTOF to assess toxic exposure.

• Analysis of human urinary biomarkers of dioxin exposure.

• Multivariate analysis and data filtering as biomarker selection tool.

• In vitro metabolic syntheses as a tool for identifying biomarkers.

Abstract

A previous high-resolution metabolomic study pointed out a dysregulation of urinary steroids and bile acids in human cases of acute dioxin exposure. A subset of 24 compounds was highlighted as putative biomarkers. The aim of the current study was (i) to evaluate the 24 biomarkers in an independent human cohort exposed to dioxins released from the incineration fumes of a municipal waste incinerator and; (ii) to identify them by comparison with authentic chemical standards and biosynthesised products obtained with in vitro metabolic reactions.

An orthogonal projection to latent structures discriminant analysis built on biomarker profiles measured in the intoxicated cohort and the controls separated both groups with reported values of 93.8%; 100% and 87.5% for global accuracy; sensitivity and specificity; respectively. These results corroborated the 24 compounds as exposure biomarkers; but a definite identification was necessary for a better understanding of dioxin toxicity.

Dehydroepiandrosterone 3β-sulfate, androsterone 3α-glucuronide, androsterone 3α-sulfate, pregnanediol 3α-glucuronide and 11-ketoetiocholanolone 3α-glucuronide were identified by authentic standards. Metabolic reactions characterised four biomarkers: glucuronide conjugates of 11β-hydroxyandrosterone; glycochenodeoxycholic acid and glycocholic acid produced in human liver microsomes and glycoursodeoxycholic acid sulfate generated in cytosol fraction.

The combination of metabolomics by high-resolution mass spectrometry with in vitro metabolic syntheses confirmed a perturbed profile of steroids and bile acids in human cases of dioxin exposure.

Introduction

Metabolomics aims to identify and quantify all metabolites (mass < 1000 Da) in a biological system and therefore constitutes a potent approach for assessing phenotype modifications caused by disease or environmental influences (Dettmer et al., 2007). Two major strategies were developed in the last years for metabolomic studies. Targeted metabolomic studies were focused on the detection of a set of pre-selected metabolites, whereas untargeted approaches provided the highest coverage of metabolites in a sample. Untargeted data acquisition is generally achieved as a methodological starting point to derive data-driven hypotheses through the unbiased compound measurement. While first untargeted metabolomic studies relied mainly on nuclear magnetic resonance spectroscopy (NMR), hyphenated methods involving separation techniques and mass spectrometry (MS) have now been demonstrated to be particularly effective (Nicholson and Lindon, 2008). Recent developments in Liquid Chromatography (LC) and especially ultra-high pressure LC (UHPLC) have provided new perspectives regarding chromatographic performance. UHPLC has become a gold standard in less than 10 years in numerous research fields, including metabolomics, as it results in well-resolved peaks with narrow peak widths leading to either an increased peak capacity or a shorter analysis time without a loss of resolution (Guillarme et al., 2007, Guillarme et al., 2010).

The dysregulation of urinary conjugated steroids and bile acids was recently reported in a previous metabolomic study on human cases of acute dioxin exposure (Jeanneret et al., 2014). This study was performed using UHPLC coupled to quadrupole time-of-flight mass spectrometry (QTOF). More specifically, a subset of 24 compounds was highlighted as a putative relevant set of biomarkers, but their definitive identification remained to be performed. Although bile acid and steroids originate from the same precursor, i.e., cholesterol, metabolites retrieved in the urine are numerous due to many successive enzymatic reactions. Among them, two classes of enzymes are particularly involved in the steroidogenesis: the cytochrome P450 family (CYP) and the hydroxysteroid dehydrogenases (HSD), the latter being separated in two groups, the short-chain alcohol dehydrogenase/reductase and the aldo-keto reductase (Miller and Auchus, 2011, Payne and Hales, 2004). Briefly, for steroids, hydroxylation and cleavages are catalysed by CYPs, while conversions from secondary alcohol groups to ketone, or vice versa, are made by HSD. Steroids are then further metabolised by phase II reactions. For example, sulfations or glucuronidations of the parent compounds can be achieved at various positions. From 15 major steroids often illustrated in the classical steroidogenesis scheme, more than 70 metabolites were measured in urine in a study comparing healthy men with patients with prostatic hyperplasia (Moon et al., 2009). Correspondingly, bile acids also undergo several sequential steps of metabolism. Cholic and chenodeoxycholic acids, defined as primary bile acids, are synthesised in the liver from cholesterol by different CYPs and HSD enzymes (Sundaram et al., 2008). They are then partially dehydroxylated to secondary bile acids (respectively deoxycholic acid and lithocholic acid) by the action of bacteria in the intestine. After reabsorption in the liver, these four bile acids are conjugated with either a glycine or a taurine, giving eight conjugated bile acids (glycocholic acid, taurocholic acid, glycodeoxycholic acid, taurodexycholic acid, glycolithocholic acid, taurolithocholic acid, glycochenodeoxycholic acid, and taurochenodeoxycholic acid). Similarly to steroids, these compounds may undergo further sulfation and glucuronidation reactions leading to numerous bile acid metabolites in urine.

Due to the formation of numerous steroid and bile acid metabolites that are closely structurally related, identification (ID) remains highly challenging. The unambiguous identification of a biomarker is generally achieved by comparing various physicochemical properties with authentic chemical standards, such as retention time (t_R), exact mass and fragmentation spectrum in the case of UHPLC-MS/MS methods. The gold-standard for definite structural identification remains NMR, but this remains generally inapplicable for these very low concentration compounds. Moreover, even if chemical synthesis of glucuronides (Stachulski et al., 2006, Stachulski and Meng, 2013) and sulfate conjugates (Al-Horani and Desai, 2010) are well described for numerous small molecules, commercial sources of phase II metabolites are scarce for steroids and bile acids. An identification by comparison with an authentic chemical standard was defined as the highest level of confidence in metabolomics studies, i.e.,“level of confidence 1”; additional confidence levels of features identification were also proposed for the cases where authentic chemical standards were missing (for original definition see (Dunn et al., 2013, Sumner et al., 2007)). Hence, a lower level of identification can be obtained by comparing physicochemical properties and/or spectral similarity with public or commercial spectral libraries (level of confidence 2). For human studies, various open source repositories are available such as the Human Metabolome Database (HMDB) (Wishart et al., 2013), the METLIN Metabolite Database (Tautenhahn et al., 2012) or LIPID MAPS (Fahy et al., 2009). When no database match is found but some characteristic physicochemical properties match with a chemical class of compounds, a lower level of identification is assigned (level of confidence 3). Finally, the last level of identification (level of confidence 4) corresponds to unknown compounds, for whom only a differentiation based upon spectra data is available.

The first aim of this study was to evaluate the discrimination properties of the set of steroid-related biomarkers, with an independent human cohort exposed to dioxins. For this purpose, the analysis of the 24 biomarkers was undertaken with samples from a new independent cohort (Maincy cohort). The latter was composed of people in contact with incineration fumes by living and working for 8 to 40 years near the municipal solid waste incinerator (MSWI) of Vaux-le-Pénil (France), exploited between 1974 and 2002. A 1994 European directive (EC 94/67/CE) established a limit value of 0.1 ng TEQ/m³ of polychlorinated dibenzodioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs) in the emissions of hazardous waste incinerators. PCDDs/PCDFS values of 226 ng TEQ/m³, corresponding to levels approximately 2000 times higher than the norm, were measured (ORF and IAURIF, 2005), and therefore the activity of the MSWI was stopped in June 2002. It has to be noted that a study reported increased PCDDs/PCDFs concentrations in soils and eggs near Vaux-le-Pénil compared to control samples, indicating a contamination of the surrounding environment from the MSWI (Pirard et al., 2005).

The second aim of this study was to identify these biomarkers of dioxin exposure. According to the fact that authentic chemical standards are scarce or missing, synthesis driven approaches were used to obtain identification with a level of confidence 1. Because metabolism involves successive biochemical reactions mainly in the liver prior to excretion in urine, a strategy using different liver fractions (Caron et al., 2006, Jantti et al., 2007) was assessed to generate standards of candidate biomarkers. Identification was needed to gain insights into the interaction between the biomarkers for this independent human cohort and to allow for a better understanding of dioxin exposure.