The assessment of the potential hepatotoxicity of new drugs by in vitro metabolomics

Abstract

Drug hepatotoxicity assessment is a relevant issue both in the course of drug development as well as in the post marketing phase. The use of human relevant in vitro models in combination with powerful analytical methods (metabolomic analysis) is a promising approach to anticipate, as well as to understand and investigate the effects and mechanisms of drug hepatotoxicity in man. The metabolic profile analysis of biological liver models treated with hepatotoxins, as compared to that of those treated with non-hepatotoxic compounds, provides useful information for identifying disturbed cellular metabolic reactions, pathways, and networks. This can later be used to anticipate, as well to assess, the potential hepatotoxicity of new compounds. However, the applicability of the metabolomic analysis to assess the hepatotoxicity of drugs is complex and requires careful and systematic work, precise controls, wise data preprocessing and appropriate biological interpretation to make meaningful interpretations and/or predictions of drug hepatotoxicity. This review provides an updated look at recent in vitro studies which used principally mass spectrometry-based metabolomics to evaluate the hepatotoxicity of drugs. It also analyzes the principal drawbacks that still limit its general applicability in safety assessment screenings. We discuss the analytical workflow, essential factors that need to be considered and suggestions to overcome these drawbacks, as well as recent advancements made in this rapidly growing field of research.

Keywords: HepG2 cells; biomarkers; drug hepatotoxicity; in vitro; mechanisms of hepatotoxicity; metabolomics; primary hepatocytes.

FIGURE 1

Experimental workflow in metabolomic analysis.

FIGURE 2

Metabolite annotation. Several considerations should be accounted for besides accurate mass (AM) and retention time (RT). This includes elimination of background signals (Kennedy, 2002), deisotoping (Joseph, 2017), recognition of adduct peaks in spectra (Patti et al., 2012) and mass spectrometry/mass spectrometry (MS/MS) fragmentation pattern that is compared to in house or publicly available databases or internal standards, and the matching degree is estimated (Johnson et al., 2016). Figure adapted from (Kim and Kang, 2021).

FIGURE 3

Sources of variability in metabolomics data from in vitro hepatotoxicity studies.

FIGURE 4

Within and between batch effect correction using QC samples. PCA scores as a function of the injection order calculated for QCs in the example data set before (left) and after (right) QC-SVRC correction (A) for within batch effect elimination or QC normalization (B) for inter batch effect elimination.

FIGURE 5

Suggested connection between different identified metabolite markers of toxicity (orange squares) and the most reported metabolic pathways of toxicity (green squares). Created with BioRender.com.

FIGURE 6

Schematic workflow for the functional correlation analysis of results from metabolic pathway analysis. The Mantel’s test is based on the estimation of the statistical significance of the correlation between two matrices summarizing the results from pathway analysis. Figure adapted from (Liu et al., 2015; Moreno-Torres et al., 2021; Martínez-Sena et al., 2023).

FIGURE 7

Building prediction models. Incubation of cells with a set of known hepatotoxic compounds for the generation of a metabolic fingerprint for different toxicity mechanisms. Based on this, build predictive models that enable to evaluate the likeliness of a novel/unknown compound to be hepatotoxic and through which mechanism.