Transcriptomic mapping of the inter-individual variability of cellular stress response activation in primary human hepatocytes

Abstract

Background & AimsOne of the early key events of drug-induced liver injury (DILI) is the activation of adaptive stress responses, a cellular mechanism to overcome stress. Given the diversity of DILI outcomes and lack in understanding of population variability, we mapped the inter-individual variability in stress response activation to improve DILI prediction. Approach & ResultsHigh-throughput transcriptome analysis of over 8,000 samples was performed in primary human hepatocytes of 50 individuals upon 8 to 24 h exposure to broad concentration ranges of stress inducers: tunicamycin to induce the unfolded protein response (UPR), diethyl maleate for the oxidative stress response, cisplatin for the DNA damage response and TNF for NF-{kappa}B signalling. This allowed investigation of the inter-individual variability in concentration-dependent stress response activation, where the average of benchmark concentrations (BMCs) had a maximum difference of 864, 13, 13 and 259-fold between different hepatocytes for UPR, oxidative stress, DNA damage and NF-{kappa}B signalling-related genes, respectively. Hepatocytes from patients with liver disease resulted in less stress response activation. Using a population mixed-effect framework, the distribution of BMCs and maximum fold change were modelled, allowing simulation of smaller or larger PHH panel sizes. Small panel sizes systematically under-estimated the variance and resulted in low probabilities in estimating the correct variance for the human population. Moreover, estimated toxicodynamic variability factors were up to 2-fold higher than the standard uncertainty factor of 101/2 to account for population variability during risk assessment, exemplifying the need of data-driven variability factors. ConclusionsOverall, by combining high-throughput transcriptome analysis and population modelling, improved understanding of variability in stress response activation across the human population could be established, thereby contributing towards improved prediction of DILI.