1. Pathway-based approaches in toxicity testing

Regulatory toxicology is constantly challenged with the need to assess an ever increasing number of chemicals in accordance with legislative mandates. At the same time, there is a growing pressure to reduce the animal use, but also the time and costs required to perform chemical toxicity testing.

A solution proposed by the U.S. National Research Council in their 2007 report “: A vision and a Strategy” is to reduce the reliance on the traditional whole animal-based descriptive toxicity testing approaches. Instead, a larger emphasis should be placed on the use of mechanistic (pathway-based) data. This novel testing paradigm should incorporate an expanded array of computational in silico methods as well as animal-free in vitro assays, preferably those where testing is performed with material of human origin.

However, it has been recognized that the use of such tools in risk assessment may be hampered by the insufficient understanding of how the endpoints measured or predicted at the molecular/cellular level are linked to the toxic effects manifested at the whole-organism level. As one way of addressing this challenge, a concept of an Adverse Outcome Pathway (AOP) was put forward as a framework to collect, organize, and evaluate the knowledge available on such linkages.

2. AOP definition

An AOP describes a progression of toxicity from the lower to the higher levels of biological organization, culminating in an adverse outcome of regulatory relevance. An AOP starts with a Molecular Initiating Event (MIE), referring to a molecular perturbation caused by a toxicant. MIE is followed by a sequence of causally related Key Events (KEs) of increasing complexity, ultimately linking to an Adverse Outcome of regulatory relevance. AOs occur at organism- or population-level and include, for example, organ toxicity or a disease state in humans, or reduced survival, disturbed growth/development, and reproduction failures in wildlife. AOPs allow organizing the available knowledge within a clear conceptual framework. This in turn allows identifying knowledge gaps, useful in guiding further research efforts. Within an AOP, the data obtained by computational approaches and diverse toxicity assays can be mapped along the sequence of events leading to an AO. This allows for an enhanced understanding of predictive relationships, thus facilitating the replacement of traditional animal-based toxicity testing by animal-free methods and testing approaches.

3. AOP development programs

In 2012, a program on the was launched by the Organization for Economic Co-operation and Development (OECD). AOP development at OECD is overseen by the Extended Advisory Group on Molecular Screening and Toxicogenomics. Together with other players, this organization also oversees the maintenance of the , for example its module used for recording qualitative AOP information, and module supporting the development of AOP-based quantitative models. The first AOP officially endorsed by the OECD was that for . Recently, several more AOPs have been endorsed and published by the OECD in the dedicated outlet , where the new AOPs developed at OECD will appear in the future as well. One current U.S. EPA research project deals with constructing multiple as a means of linking the bioassays used in and programs to the unique biological targets and whole-organism apical effects they aim to predict. As of June 2016, about 760 ToxCast assays have been mapped to about 300 unique biological targets. The outcomes of this project are expected to facilitate the interpretation and use of ToxCast screening data, but also help identifying important deficiencies in the array of ToxCast assays, as well as aid prioritization of toxicity pathways for the development of OECD-compliant AOP descriptions. Moreover, the catalogue of putative AOPs developed by EPA is also expected to help engaging a broader scientific community in the development of new AOPs, increasing the coverage of the common toxicological space.

4. AOP use in diverse research fields

The AOP concept has been originally suggested by Ankley and colleagues from the U.S. Environmental Protection Agency (EPA) for use in . Since then, the application of the AOP framework in other fields has also been advocated, for example, in human and research {Langley, 2015 #471}. In the latter publication, the authors emphasized that many non-communicable diseases, such as cancer, diabetes, immune system disorders, and cardiovascular ailments, are caused by an interplay between a multitude of both genetic and environmental factors. Modern technological advances allow not only to identify the genes contributing to the disease susceptibility, but also understand how the molecular constituents of our organism react to various environmental influences. AOP framework can be used to integrate these diverse sources of information, providing for a more complete understanding of disease development and progression. This knowledge may in turn contribute to better understanding of disease etiology and identification of subjects at risk, supporting the development of prevention programs and novel treatment strategies. Other recent publications provide the introduction to and to be followed when developing an AOP, and discuss how to use the AOP concept to facilitate the research on, for example, , , , and , as well as in design of Integrated Approaches to Testing and Assessment (), and in support of .

5. Current shortcomings of the AOP framework

Some authors and organizations have questioned the appropriateness and usefulness of the AOP concept for regulatory toxicology applications. For example, it has been argued that the use of AOPs could undermine the exercising of precautionary principle in the regulatory actions, particularly with regard to those issues where high scientific uncertainty still exists. Indeed, to regulate a certain hazardous chemical, it should be sufficient to only demonstrate the ability of a certain chemical to cause harm, even when its mechanism of action is not yet understood.  The AOPs, however, appear to have a strong mechanistic focus requiring an in-depth understanding of the biological interactions in order to postulate a particular AOP. However, the OECD guidance emphasized that developed AOPs can differ in regard to the nature of predictions that can be made (qualitative or quantitative), and even some incomplete AOPs (for example, missing the understanding of certain intermediary events) may still be useful for certain applications (e.g. category grouping or read-across). Moreover, AOPs are living documents that are expected to be updated once a deeper scientific understanding emerges. The linear nature of the proposed AOP construct has also been heavily criticized as being too rigid and not allowing to capture the complex interactions common in biology. Moreover, the currently available AOPs do not yet appear to address the important low-dose effects, or effects caused by chemical mixtures. Thus, AOPs are viewed as an extreme version of the “reductionist” approach, being an unrealistic representation of the real life interactions resulting in health effects of concern. The AOP developers countered that the linear construct was adopted for purely practical reasons, i.e. to simplify the presentation of the most essential steps. The interrelationships between the different components of biological systems are expected to be captured through building the so-called ‘AOP networks’ where different events can be shared between multiple pathways, can exert influences on each other, and thus can also collectively influence the final AOs resulting from a particular perturbation or exposure to a particular stressor. It is expected that AOP networks will allow integrating the consequences of activation of multiple pathways by a single agent, as well as understanding the concerted effects of multiple agents on a particular pathway or outcome of interest. However, before such AOP networks can be built, many more AOPs still need to be developed and deposited in a database, such as AOP-wiki, in order to cover a sufficiently high proportion of biological and toxicological space. Currently, this is not yet the case, as the numbers of available AOPs are still rather low, and those of officially validated AOPs even lower. Despite the rapid growth of the interest in the AOPs, the development of sufficient numbers of reliable AOPs still presents a major challenge and requires a more active involvement of a broader scientific community, possibly supported by the dedicated funding schemes.

Moreover, many challenges and open questions still remain on the ability of AOPs to make quantitative predictions and to support in vitro to in vivo extrapolations. The progress in this regard will heavily depend on the progress of the whole field of alternative testing methods in general. However, AOPs have the potential to influence and benefit this field by helping to prioritize the most promising alternative testing approaches for further development.

Read more

OECD (2016). “” EPA (2016). “” EPA (2016). “”

Fay, K., et al. (2015). “” EPA Science Inventory

6. References

Edwards, S., et al. (2016). “” The Journal of Pharmacology and Experimental Therapeutics 356:170-181.

Gerloff, K., et al. (2016). “” Computational Toxicology DOI 10.1016/j.comtox.2016.07.001

Langley, J., et al. (2015). “” Environmental Health Perspectives 123:A268-272.

Knapen, D., et al. (2015). “” Reproductive Toxicology 56:52-55.

Bal-Price, A., et al. (2015). “” Critical Reviews in Toxicology 45:83-91.

Groh, K., et al. (2014). “” Chemosphere 120:764-777.

Tollefsen, K.-E., et al. (2014). “” Regulatory Toxicology and Pharmacology 70:629-640.

Villeneuve, D., et al. (2014). “” Toxicological Sciences 142:312-320.

Villeneuve, D., et al. (2014). “” Toxicological Sciences 142:321-330.

Vinken, M. (2013). “” Toxicology 312:158-165.

OECD (May 4, 2012). “” ENV/JM/MONO(2012)10/PART1 (pdf)

OECD (May 4, 2012). “” ENV/JM/MONO(2012)10/PART2 (pdf)

Ankley, G., et al. (2010). “” Environmental Toxicology and Chemistry 29:730-741.

National Research Council (2007). “” The National Academies Press doi:10.17226/11970