(188dj) Multi-Omics Analysis Reveals That Co-Exposure to Phthalates and Metals Disturbs Urea Cycle and Choline Metabolism

This study was part of the European project HEALS, which proposes the functional integration of multi–omics in the context of exposome. During the previous years, metabolomics analysis of 600 urine and plasma samples from two pre-existing cohorts (Repro and PHIME) revealed the biomarkers of exposure in relation to the phthalates and heavy metals. REPRO PL is a Polish mother and child cohort study that evaluates the impact of exposure to several pollutants on children’s health and neurodevelopment. Previous work showed an inverse association between prenatal exposure to 11 phthalate metabolites and child motor development. The PHIME cohort consists of children from Italy, Greece, Slovenia and Croatia; the first studies were interested in associations between prenatal exposure to mercury/methyl mercury and either influence of polymorphisms in ABC transporter genes or CYP3A expression and neurodevelopment. The aim of this study was to obtain mechanistic insight into how co-exposure to phthalates and heavy metals causes neurodevelopmental perturbations based on in vitroassays.

The HepaRG cell line, a validated model for xenobiotic metabolism,was used as in vitromodel.More specific the human liver-derived HepaRG cell line was chosen for several reasons: i) after differentiation by DMSO, two cell types coexist in the culture (50% hepatocyte-like cells and biliary-type cells). Bile ducts are present in the culture between hepatocytes, and thus it is one of the best cellular model for human liver. HepaRG are known to express all the xenosensors and xenobiotic metabolizing enzymes which can be implicated in the effects of the pollutants and ii) these cells can be treated for a long time (several weeks) with low concentrations of the selected pollutants to mimic at best a low, semi-chronic exposure.

HepaRG cells were exposed to mixtures of DEHP, DiNP, and BBzP phthalates, methylmercury and total mercury. These were the most abundant pollutants in the REPRO PL and PHIME cohorts that studied the environmental causation of neurodevelopmental disorders in neonates and children across Europe. The effective concentrations of the chemicals in vitrowere estimated through extrapolation from human biomonitoring data through internal dosimetry modeling using the INTEGRA computational platform.

HepaRG cells were seeded at 30,000 cells/cm² and proliferated for 2 weeks in William’s E media containing 10% fetal calf serum, 50µM hydrocortisone hemisuccinate, 4µg/mL insulin, 500UI/mL penicillin streptomycin, and 2mM glutamine. To induce differentiation, 1.5% DMSO was added to the media. Cells were fully differentiated within 15 days of treatment with DMSO. Differentiated cells were trypsinized and seeded into 6 well plates at a density of 210,000 cells/cm². Seventy two hours later, the medium was changed and 48h later, treatments were started. All pollutants were freshly prepared 3 times a week from stock solutions.

Table 1. Concentrations of the pollutants in the human liver (modelled)

Concentration	DEHP	DiNP	BBzP	Hg	Pb
Max (nM)	2.56	4.778	0.192	17.947	27.053
Mean (nM)	0.23	0.478	0.018	4.686	8.213
Min (nM)	0.128	0.096	0.003	0.499	4.831

Table 2. Standard solutions for in vitro experiments

THg	THg	MeHg	MeHg
(µg/mL)	(nmol/mL) (µM)	(µg/mL)	(nmol/mL) (µM)
0.5	2.49	0.5	2.49
1	4.98	1	4.98
3	14.9	3	14.9
10	49.8	10	49.8
30	149	30	149
500	2493	500	2493

The same cells were treated in parallel plates for the various omics procedure. For the preparation of samples for the transcriptomics analysis, media was discarded and 350µL of RLT plus lysis buffer (Qiagen) with 10% β-mercaptoethanol was added (just before the extraction) into each well of a 6 well plate well and frozen at -20°C until RNA extraction. RNA was extracted with RNeasy Mini kit (Qiagen) on a Qiacube automate. The preparation of samples for proteomics and metabolomics included cell washing 5 times with ice cold PBS. In order to prepare the samples for proteomics analysis 150 µL of proteomic buffer ((8M urea, 0.1%SDS, 50mM ammonium bicarbonate) was subsequently added into each well of a 6 well plate. Cells were scrapped with a cell scrapper and transferred into a 1.5mL Eppendorf tube. The lysate was centrifuged at 13000g for 30min at 4°C. Samples were analyzed in a nanoAcquity liquid chromatographer (Waters) coupled to a LTQ-Orbitrap Velos (Thermo Scientific) mass spectrometer. The preparation of samples for metabolomics included the addition of one mL of LS/MS grade methanol, followed by 1mL of LS/MS grade H₂O. Cells were scrapped with a cell scrapper and transferred into a 15mL falcon tube. Four mL of chloroform was added followed by vortex mixing. The mixture was incubated for 30 min on ice and 2mL of LS/MS grade H₂O were added. After centrifugation at 3000g for 20min at 4°C (no brake), the upper and lower phases were transferred respectively into 1.5mL Eppendorf tubes (for each condition, 2 aliquots of 1mL) and frozen in liquid nitrogen. ThermoFisher Scientific (Bremen, Germany) model LTQ Orbitrap Discovery MS with a resolution 30,000 system coupled with an Acquality UPLC HSS T3 column (100 x 2.1 mm, 1.8 μm, Waters, Milford, MA, USA) was used. All samples were stored at -20°C before the analysis.

Raw data was analyzed using GeneSpring GX v.14.9 (Agilent Technologies). In case of transcriptomics results of the raw data files (.cel) of the Affymetrix Human 2_0-st chip was imported into the software. The RAM (Robust Multichip Averaging) algorithm was used to normalize the data where it was subjected to quantile normalization with a median of all samples taken for baseline transformation. The generated .raw data from proteomics analysis were processed with MaxQuant software. Andromeda search engine was used to search against SwissProt Human database under the following parameters: Enzyme: Trypsin; Missed Cleavage: 2; Fixed modifications: Carbamidomethyl; Variable modifications: Acetyl (Protein N-term), Oxidation (M); Peptide Mass Tolerance: 20 ppm; Fragment Mass Tolerance: 0.6 Da. MaxQuant Label-free Quantification (LFQ) was done using non-conflicting and Razor peptides and protein grouping. The software normalization algorithm was applied. Abundances were calculated as the area under the MS peak for every matched ion feature. In total, 3350 proteins were quantified from a mean of 11800 identified peptides per sample. The .csv file of the 3350 identified proteins was imported into the GeneSpring GX software for further statistical analysis. Metabolites identified by LC-MS/MS analysis were determined using MZMine v2.31. After data upload, mass spectral peaks were selected, aligned and annotated. The data were deconvoluted using the Wavelet (XCMS) algorithm to remove the noise, thus providing only the biologically relevant information. Human Metabolome DataBases(HMDB) and Metlin, were used for the identification step under mass tolerance 5ppm. Finally, more than 1000 metabolites were imported into GeneSpring GX for further bioinformatics analysis. The data were analysed using bioinformatics algorithms involving normalization, quality control, advanced statistics, and multi-omics pathway analysis with GeneSpring GX V.14.9. Fold change analysis was used to identify genes with expression ratios or differences between mixture 1 and solvent control, as well between mixture 2 and solvent control, that are outside of a given cut-off or threshold. Differentially expressed genes, proteins and metabolites were mapped on available pathways from WikiPathways, BioCyC, and KEGG databases, using the module Pathway Architect of GeneSpring. Hypergeometric distribution was used to calculate a p-value determining the probability that there is an association between the biomarkers in the entity list and the pathways that cannot be explained by chance alone (p≤0.1, Minimum match ≥ 1).

Integrated pathway-level analysis of transcriptomics and proteomics data revealed that co-exposure to phthalates and heavy metals leads to the perturbation of the urea cycle due to alterations in the expression levels of arginase-1 and -2, argininosuccinate synthase, carbamoyl-phosphate synthase, ornithine carbamoyltransferase, and argininosuccinate lyase. Co-mapping of proteomics and metabolomics data revealed that their common drivers are responsible for the homeostasis of metabolic pathways related to choline, phosphatidylcholine, phospholipases and triacylglycerol metabolism. The identification of the urea, phosphatidylcholine biosynthesis I and phospholipases metabolic pathways is of particular interest since these pathways have been also identified in human samples from the REPRO PL and PHIME cohorts using untargeted metabolomics analysis and have been associated with impaired psychomotor developmentin children at the age of three to six. Our work reveals that co-exposure to plasticizers and metals disturb biochemical processes related to mitochondrial respiration during critical developmental stages that are clinically linked to neurodevelopmental perturbations.