Untargeted metabolomics to study ethanol-induced hepatotoxicity in HepaRG cells

Despite the high prevalence of alcoholic liver disease, its identification and characterization remain poor, especially in early stages such as alcoholic fatty liver disease and alcoholic steatohepatitis. This latter implies diagnostic difficulties, few therapeutic options and unclear mechanisms of action. In this thesis, LC-MS-based metabolomics was used in an in vitro set-up to identify biochemical biomarkers able to elucidate the mechanism of ethanol-induced hepatotoxicity at a mechanistic cellular level. HepaRG, a human hepatocyte cell line was used to elucidate metabolic alterations and pinpoint affected metabolic pathways after exposure to ethanol in order to simulate alcoholic fatty liver disease. Combined exposure to ethanol and tumor necrosis factor alpha was used to simulate alcoholic steatohepatitis in vitro.
Part A of this thesis (i.e., chapter 3-5) was dedicated to development of analytical methods using liquid chromatography-quadrupole-time-of-flight high-resolution mass spectrometry (LC-QTOF-HRMS) and the hyphenation to drift tube ion mobility spectrometry (DTIMS) was explored. In addition, a multidimensional library for untargeted MS-based metabolomics was constructed and guidelines and consideration were formulated. Part B of this thesis (i.e., chapter 6-8) describes the application of the optimized metabolomics methods to study ethanol-induced hepatotoxicity in an in vitroset-up.
In chapter 3, a metabolomics platform was optimized to be able to analyze polar metabolites in HepaRG extracts. The analysis of polar metabolites based on LC-MS methods should take into consideration the complexity of interactions in LC columns to be able to cover a broad range of metabolites of key biological pathways. Therefore, in chapter 3, different chromatographic columns were tested for polar metabolites including reversed-phase and hydrophilic interaction liquid chromatography (HILIC) columns. Based on a column screening, two new generations of zwitterionic HILIC columns were selected for further evaluation. A tree-based method optimization was applied to investigate the chromatographic factors affecting the retention mechanisms of polar metabolites with zwitterionic stationary phases. The results were evaluated based on a scoring system which was applied for more than 80 polar metabolites with a high coverage of key human metabolic pathways. The final optimized methods showed high complementarity to analyze a wide range of metabolic classes including amino acids, small peptides, sugars, amino sugars, phosphorylated sugars, organic acids, nucleobases, nucleosides, nucleotides and acylcarnitines. Optimized methods were applied to analyze different biological matrices, including HepaRG extracts, human urineand plasma using an untargeted approach. The number of high-quality features (< 30% median relative standard deviation) ranged from 3,755 for urine to 5,402 for the intracellular metabolome of HepaRG cells, showing the potential of the methods for untargeted purposes.
In chapter 4, a lipidomics platform was optimized to be able to analyze lipids in HepaRG extracts. There are thousands of lipids in most biological samples, and therefore separation methods before introduction to the mass spectrometer are key for relative quantitation and identification. Chromatographic methods differ across laboratories, without any consensus on the best methodologies. Therefore, we designed an experiment to determine the optimal LC methodology, and assessed the value of ion mobility as an additional dimension of separation. To apply an untargeted method for hypothesis generation focused on lipidomics, LC-HRMS parameters were optimized based on the measurement of 50 panel lipids covering key human metabolic pathways. Similar to the approach in chapter 3, reversed-phase liquid chromatography columns were compared based on a quality scoring system considering the signal-to-noise ratio, peak shape, and retention factor. DTIMS was implemented to increase peak capacity and confidence during annotation by providing collision cross section (CCS) values for the analytes under investigation. However, hyphenating DTIMS to LC-HRMS may result in a reduced sensitivity due to impaired duty cycles. To increase the signal intensity, a Box-Behnken design (BBD) was used to optimize four key factors; drift entrance voltage, drift exit voltage, rear funnel entrance, and rear funnel exit voltages. Application of a maximized desirability function provided voltages for the above-mentioned parameters resulting in higher signal intensity compared to each combination of parameters used during the BBD. In addition, the influence of single pulse and Hadamard 4-bit multiplexed modes on signal intensity was explored and different trap filling and release times of ions were evaluated. The optimized LC-DTIM-HRMS platform was applied to extracts from HepaRG cells and resulted in 3912 high-quality features. From these features, 436 lipid species could be annotated (i.e., matching based on accurate mass <5 ppm, isotopic pattern, MS/MS fragmentation, and CCS database matching <3%).
As feature annotation is crucial in untargeted metabolomics and remains a major challenge, chapter 5 was dedicated to the construction of multidimensional libraries for untargeted MS-based metabolomics. The large pool of metabolites collected under various instrumental conditions is underrepresented in publicly available databases. Retention time (RT) and CCS measurements from liquid chromatography ion mobility high-resolution mass spectrometers can be employed in addition to MS/MS spectra to improve the confidence of metabolite annotation. Recent advancements in machine learning focus on improving the accuracy of predictions for CCS and RT values. Therefore, high-quality experimental data are crucial to be used either as training datasets or as a reference for high-confidence matching. Chapter 5 provides an easy-touse workflow for the creation of an in-house metabolite library, offers an overview of alternative solutions, and discusses the challenges and advantages of building in-house libraries. A total of 100 metabolite standards from various classes were analyzed and subjected to the described workflow for library generation. The outcome was an openaccess available NIST format metabolite library (.msp) with multidimensional information. The library was used to evaluate CCS prediction tools, MS/MS spectra heterogeneities (e.g., multiple adducts, in-source fragmentation, and radical fragment ions using collision-induced dissociation), and the reporting of RT.
Chapter 6 provides a summary of the optimized analytical methods that were used to study ethanol-induced hepatotoxicity in HepaRG cells, in addition to a description of experimental exposure conditions, procedures for sample preparation, data processing, statistics and metabolite annotation. Sample preparation was based on a liquid-liquid extraction with H2O/MeOH/CHCl3 and was used to divide each biological sample in 2 polar and 2 apolar subfractions, which were analyzed using separate corresponding analytical methods to increase metabolite coverage. Throughout the analytical workflow, comprehensive quality assurance and quality control measures were implemented to ensure high reproducibility. These latter measures included, for example, usage of standardized acquisition sequences, pooled quality control samples, and system suitability samples.
Alcoholic fatty liver disease was simulated in HepaRG cells in chapter 7 by exposure of these cells to ethanol at different concentrations and exposure times. Excessive ethanol consumption is known to alter lipid metabolism, followed by progressive intracellular lipid accumulation, resulting in alcoholic fatty liver disease. In chapter 7, HepaRG cells were exposed to ethanol at IC10 and 1/10 IC10 for 24 and 48 h. Metabolic alterations were investigated intra-and extracellularly with LC-HRMS. Ion mobility was added as an extra separation dimension for untargeted lipidomics to improve annotation confidence. Distinctive patterns between exposed and control cells were consistently observed, with intracellular upregulation of di- and triglycerides, downregulation of phosphatidylcholines and -ethanolamines, sphingomyelins, and S-adenosylmethionine, among others. Several intracellular metabolic patterns could be related to changes in the extracellular environment, such as increased intracellular hydrolysis of sphingomyelins, leading to increased phosphorylcholine secretion. Carnitines showed alterations depending on the size of their carbon chain, which highlights the interplay between β-oxidation in mitochondria and peroxisomes. Potential new biomarkers of ethanol-induced hepatotoxicity have been observed, such as ceramides with a sphingadienine backbone, octanoylcarnitine, creatine, acetylcholine, and ethylated phosphorylcholine. The combination of the metabolic fingerprint and footprint enabled a comprehensive investigation of the pathophysiology behind ethanol-induced hepatotoxicity.
In chapter 8, HepaRG cells were exposed for 24 h to both ethanol (IC10, 368 mM) and tumor necrosis factor alpha (TNF-α, 50 ng/mL), in order to improve in vitro simulation of alcoholic steatohepatitis. This combined exposure was compared to solely ethanolexposed as well as -nonexposed cells. As in chapter 7, LC-(DTIMS)-HRMS was used to elucidate both intracellular and extracellular metabolic alterations. Some of the key findings include the influence of TNF-α in the upregulation of hepatic triglycerides and the downregulation of hepatic phosphatidylcholines and -ethanolamines. Sadenosylmethionine showed to play a central role in the progression of alcoholic steatohepatitis. In addition, fatty acyl esters of hydroxy fatty acid (FAHFA)-containing triglycerides were detected for the first time in human hepatocytes and their alterations showed a potentially important role during the progression of alcoholic steatohepatitis. As in chapter 7, ethylated phosphorylcholine was observed as a potential new biomarker of ethanol exposure. In order to evaluate the biomarker potential of this latter compound in humans, a targeted method was developed. As a proof-of-concept, the presence of ethylated phosphorylcholine was confirmed in whole blood samples of heavy drinkers. Details on these latter findings are described in the supplementary information of chapter 7.