Anika Liu, Namshik Han, Jordi Munoz-Muriedas, Andreas Bender
Adverse event pathogenesis is often a complex process which compromises multiple events ranging from the molecular to the phenotypic level. In toxicology, Adverse Outcome Pathways (AOPs) aim to formalize this as temporal sequences of events, in which event relationships should be supported by causal evidence according to the tailored Bradford-Hill criteria. One of the criteria is whether events are consistently observed in a certain temporal order and, in this work, we study this time concordance using the concept of “first activation” as data-driven means to generate hypotheses on potentially causal mechanisms. As a case study, we analysed liver data from repeat-dose studies in rats from the TG-GATEs database which comprises measurements across eight timepoints, ranging from 3 hours to 4 weeks post-treatment. We identified time concordant gene expression-derived events preceding adverse histopathology, which serves as surrogate readout for Drug-Induced Liver Injury (DILI). We find known mechanisms in DILI to be time-concordant, and show further that significance, frequency and log fold change (logFC) of differential expression are metrics which can additionally prioritize events although not necessary to be mechanistically relevant. Moreover, we used the temporal order of transcription factor (TF) expression and regulon activity to identify transcriptionally regulated TFs and subsequently combined this with prior knowledge on functional interactions to derive detailed gene-regulatory mechanisms, such as reduced Hnf4a activity leading to decreased expression and activity of Cebpa. At the same time, also potentially novel events are identified such as Sox13 which is highly significantly time-concordant and shows sustained activation over time. Overall, we demonstrate how time-resolved transcriptomics can derive and support mechanistic hypotheses by quantifying time concordance and how this can be combined with prior causal knowledge, with the aim of both understanding mechanisms of toxicity, as well as potential applications to the AOP framework. We make our results available in the form of a Shiny app
Adverse drug reactions are a major reason for compound failure in the clinical trials [1,2] and a significant cause for post-marketing withdrawals. To counter exposing patients to these risks, it is desired to identify adverse events earlier in the individual patient but also in the drug development process. Mechanistic understanding of adverse event pathogenesis is crucial in this regard, i.e. to derive early safety biomarkers or in vitro assays. However, current understanding of toxicity is largely incomplete, in particular for complex phenotypes such as organ injury which can usually be caused by a wide range of compounds perturbing the biological system at different points mediated through multiple biological scales and entities [3,4].
Open TG-GATES data processing
The TG-GATES gene expression data from studies in 6-week-old male Crl:CD Sprague-Dawley (SD) rats with daily repeat-dosing (S1 Fig) was downloaded from the Life Science Data Archive (DOI: 10.18908/lsdba.nbdc00954-01-000). The raw liver gene expression levels were background corrected, log2 transformed, and quantile normalized with the rma function of the affy package per treatment across all doses and timepoints . Quality control was then performed using the ArrayQualityMetrics package  and detected outliers with high distance to other experiments or unusual signal distribution were removed (List of removed outliers summarised in S1 File). The platform information for the Affymetrix Rat Genome 230 2.0 Array was derived from Gene Expression Omnibus  (GEO accession: GPL1355) and was then used to summarise probe IDs to rat gene symbols by median for all probes mapping uniquely to one gene symbol. Only the 360 compound-dose combinations with at least 6 measured timepoints after quality control were included. Out of the these, all eight timepoints were measured in most time series, while only six timepoints were measured in two time series, and only seven timepoints in seven time series.
Results and Discussion
In order to derive the time concordance between cellular events and later adverse histopathology, we use the workflow outlined in Fig 2 with each step being also introduced in the subsequent sections and details on their respective implementation being described in Methods. We first derived TF and pathway activity across expression profiles from the same experiment and subsequently defined the first up- or downregulation TFs or pathways as events. Furthermore, we obtained binary histopathology labels describing the occurrence of each histopathological finding at different levels of severity and frequency from the Toxscores provided by Sutherland et al. . Subsequently, we derive the earliest timepoint of each event, e.g. pathways or adverse histopathology, within each time-series. As last step of the time concordance analysis, we then evaluate which gene expression-derived events are significantly enriched before or at the time where adverse histopathology is found, as well as additional time concordance metrics outlined in Table 1.
In this study, we introduce “first activation” as concept to quantify the strength of temporal concordance between events across time series with the assumption that each activated event may have downstream effects irrespective of whether it is continuously or only transiently activated. With this approach, we study gene expression-based TF and pathway-level events found before adverse histopathology indicating liver injury in repeat-dose studies in rats from TG-GATEs as a case study. We find some known processes in DILI to be highly confident, e.g. bile acid recycling, while others are highly frequent but less specific including adaptive response pathways such as the eIF2α/ATF4 pathway .
Citation: Liu A, Han N, Munoz-Muriedas J, Bender A (2022) Deriving time-concordant event cascades from gene expression data: A case study for Drug-Induced Liver Injury (DILI). PLoSComputBiol 18(6): e1010148. https://doi.org/10.1371/journal.pcbi.1010148
Editor: James Gallo, University at Buffalo - The State University of New York, UNITED STATES
Received: December 8, 2021; Accepted: April 26, 2022; Published: June 10, 2022
Copyright: © 2022 Liu et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Histopathology data was derived from the supplementary information of Sutherland et al. (2018), accessed through https://doi.org/10.1038/tpj.2017.17. TG-GATEs gene expression data was derived from the Life Science Database Archive (https://dbarchive.biosciencedbc.jp/en/open-tggates/download.html). The files and code for the Shiny app are deposited in GitHub (https://github.com/anikaliu/DILICascades_App) and Zenodo (doi:10.5281/zenodo.5767783).
Funding: AL received funding from and JM was a full-time employee of GlaxoSmithKline (https://www.gsk.com) throughout the study. NH is funded by LifeArc (https://www.lifearc.org). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: AL received funding from GlaxoSmithKline and is a consultant at PharmEnable Ltd. JM was an employee of GlaxoSmithKline throughout the study. NH is a cofounder of KURE.ai and CardiaTec Biosciences. AB is a shareholder of Healx Ltd. and PharmEnable Ltd., and CSO at Terra Lumina.