Ksenia Groh (EAWAG), Tilman Gocht (University of Tübingen) and Knut Erik Tollefsen (NIVA)
This session focused on the use of the adverse outcome pathway (AOP) concept in basic toxicological and ecotoxicological research, and on the application of AOPs in support of hazard and risk assessment. The presentations in this session had reemphasized and expanded on the three major themes particularly crucial for further research progress in the AOP field.
AOPs provide a framework to collect and organize the knowledge on the progression of toxicity across different levels of biological organization, leading to adverse outcomes of regulatory relevance and on the predictive relationships between the events occurring at different levels.
Two presentations discussed the practical tools and conceptual approaches that support AOP development. In particular, Steven Enoch described a chemoinformatics approach for chemical category formation based on structural alerts for mitochondrial toxicity. Such in silico methods promise to be of great use in defining the chemicals able to trigger a particular molecular initiating event. Geoffrey Hodges presented the outcomes from a recent workshop that discussed the roles of “-omics” data for enhancement of species sensitivity predictions and for supporting the extrapolation from molecular initiating events to population-relevant endpoints.
AOPs provide a tool to critically assess the available knowledge and identify data gaps to guide further research.
Mathieu Vinken presented an illustrative example, an AOP linking the inhibition of bile salt export pump to cholestatic injury and the experimental steps taken to validate this AOP. Gene expression analysis linked to an AOP allowed identification of potential additional key events that in turn themselves may lead the research towards the establishment of novel biomarkers for detection of cholestasis.
Dries Knapen presented the usefulness of AOP framework for basic research. The AOP concept was applied to elucidate the mechanism of narcosis through defining specific types of membranes that can be targeted in this pathway. It was suggested that interpretation of non-specific mechanisms should be carried out within a specific and well-delineated biological context. This will likely improve the ability to develop the individual steps of the AOP continuum and assist in improving the accuracy of predictions made relying on AOP knowledge.
Predictive relationships defined by AOPs provide the foundation for the use of alternative testing methods for hazard assessment. Currently, several groups are working on developing quantitative AOPs as these are intuitively the most useful for application in predictive toxicology. However, the maturation process from a qualitative to a quantitative AOP is commonly associated with various challenges.
Luigi Margiotta-Casaluci provided an example of synthetic glucocorticoid effects in fish. The insight into the approaches for obtaining experimental data necessary for development of quantitative AOPs was enlightening. He demonstrated that chemicals with similar in vitro profiles can sometimes lead to largely different effects in vivo, emphasizing that it is important to consider both chemical-specific toxicokinetic profiles as well as potential polypharmacology (i.e., different AOPs triggered by the same chemical).
Daniel Villeneuve provided another example of AOP application in predictive ecotoxicology. He described the development of a quantitative AOP linking aromatase inhibition to impaired reproduction in fish. Using a case study, the paper addressed the question “How close are we to making quantitative predictions of in vivo adverse outcomes based purely on in vitro data?” A computational model was built, and its predictions of reproductive output, made using in vitro data from ToxCast assays for aromatase inhibition, were compared to experimental measurements of fecundity in fathead minnows. Although overall this approach appeared promising, several limitations were identified especially for quantitative predictions at low effect concentrations. While currently this computational model is being further optimized, it can be concluded that model results are strongly influenced by a number of variables including the type of input data used for estimating toxicity equivalents, the model parameters and model construct itself. Furthermore, the models should be tested against a variety of exposure scenarios, and chemical-specific toxicokinetics should be accounted for when used for quantitative predictions.
When developing quantitative AOPs, careful consideration should be given to prioritize the activities that provide the greatest return of value since these research and development costs can be high. However, not all AOPs would need to be fully quantitative. As has been demonstrated in this session, simple qualitative AOPs are often able to support hazard evaluation and risk assessment efforts.
Authors’ contact information: firstname.lastname@example.org, email@example.com and firstname.lastname@example.org