The human hepatocyte TXG-MAPr: WGCNA transcriptomic modules to support mechanism-based risk assessment

Abstract

Mechanism-based risk assessment is urged to advance and fully permeate into current safety assessment practices, possibly at early phases of drug safety testing. Toxicogenomics is a promising source of comprehensive and mechanisms-revealing data, but analysis tools to interpret mechanisms of toxicity and specific for the testing systems (e.g. hepatocytes) are lacking. In this study we present the TXG-MAPr webtool (available at https://txg-mapr.eu/WGCNA_PHH/TGGATEs_PHH/), an R-Shiny-based implementation of weighted gene co-expression networks (WGCNA) obtained from the Primary Human Hepatocytes (PHH) TG-GATEs dataset. Gene co-expression networks (modules) were annotated with functional information (pathway enrichment, transcription factor) to reveal their mechanistic interpretation. Several well-known stress response pathways were captured in the modules, are perturbed by specific stressors and show preserved in rat systems (rat primary hepatocytes and rat in vivo liver), highlighting stress responses that translate across species/testing systems. The TXG-MAPr tool was successfully applied to investigate the mechanism of toxicity of TG-GATEs compounds and using external datasets obtained from different hepatocyte cells and microarray platforms. Additionally, we suggest that module responses can be calculated from targeted RNA-seq data therefore imputing biological responses from a limited gene. By analyzing 50 different PHH donors' responses to a common stressor, tunicamycin, we were able to suggest modules associated with donor's traits, e.g. pre-existing disease state, therefore connected to donors' variability. In conclusion, we demonstrated that gene co-expression analysis coupled to an interactive visualization environment, the TXG-MAPr, is a promising approach to achieve mechanistic relevant, cross-species and cross-platform evaluation of toxicogenomic data.

Mapping the cellular response to electron transport chain inhibitors reveals selective signaling networks triggered by mitochondrial perturbation

“…Revisiting four ER stress-related WGCNA modules (13, 15, 62 and 295; (Callegaro et al 2021)), in the PHH data mentioned above (Fig. 4), supported the observed difference between the UPR stress inducer tunicamycin and the ETC inhibitors, i.e., deactivation of module 13 (including HSPA5) per treatment (Suppl.…”

“…We next moved from the analysis of responses at the individual gene level to analysis of gene networks and pathways, which may enhance translation across model systems. First, we took advantage of previously established weighted gene co-expression network analysis (WGCNA) modules for PHH comprising ~ 400 gene modules that can be visualized as a toxicogenomics map (Callegaro et al 2021 ). The ETC inhibitor TempO-Seq transcriptomics data obtained in HepG2 was projected on these prior established co-expression modules and used to calculate module eigengene scores (EGSs) to monitor chemical-induced changes in these modules.…”

“…D Table of module 13, 15, 62 and 295 describing: the percentage coverage when projecting the S1500 + + + set, pathway enriched pathways of this specific module and the genes measured in that particular module (bold: log2FC > 0.58 upon exposure to 2 µM rotenone and 10 µM antimycin, underlined: hub gene of this module). The modules were selected based on their involvement in the ER stress response (Callegaro et al 2021 ). E Concentration response curves of eigengene scores (EGS) per selected module upon 24 h exposure to rotenone, antimycin or tunicamycin.…”

“…The modules were selected based on being upregulated in both the CI and CIII correlation plot (Suppl. Figure 6A) or being identified as ER module by Callegaro et al (2021). G Concentration response curves of the log2FC ± SE for ASNS upon exposure to rotenone, mepronil or antimycin.…”

“…Primary human hepatocyte (PHH) gene expression data was obtained from the open TG-GATES database: “Toxicogenomics Project and Toxicogenomics Informatics Project under CC Attribution-Share Alike 2.1 Japan” and processed as reported previously (Callegaro et al 2021 ). Briefly, microarrays were jointly normalized using the Robust Multi-array Average (RMA) method (affy R package) (Gautier et al 2004 ) and probes were mapped to gene IDs with BrainArray chip description file (CDF) version 20.…”