Portal de Periódicos da CAPES

Adverse Outcome Pathways: Moving from a Scientific Concept to an Internationally Accepted Framework

2019; Wiley; Volume: 38; Issue: 6 Linguagem: Inglês

10.1002/etc.4385

1552-8618

Markus Hecker, Carlie A. LaLone,

Computational Drug Discovery Methods

International legislative mandates require chemical safety assessments to increase consumer confidence and protect human health and wildlife. Many of these mandates necessitate that regulators evaluate large numbers of chemicals within short time frames to make informed regulatory decisions (e.g., the European Union's Registration, Evaluation, Authorisation and Restriction of Chemicals and the US Toxic Substances Control Act). To address this challenge and reduce the costs of toxicity testing, a new vision and strategy for toxicity testing has emerged, which aspires to transform testing by making greater use of recent scientific advances (over the past decade) in cell-based and computational methods. A key concept aligning with this vision is the adverse outcome pathway (AOP) framework (Ankley et al. 2010). In brief, AOPs organize available toxicological knowledge and describe the causal linkages between a molecular initiating event (MIE; the first interaction of a chemical with a biological macromolecule such as an enzyme or a receptor) and subsequent measurable responses (termed key events [KEs]) across biological levels of organization, which culminate in an adverse outcome (AO) of regulatory significance—typically at the individual or population level (Box 1). Over the past decade, the AOP framework has matured significantly and has increasingly been recognized as a powerful approach for organizing biological information into a format applicable for chemical safety evaluation in both human health and ecological contexts (Ankley et al. 2010). Further, critical advances in AOP development and technology, including the development of a Web-based AOP Knowledgebase (AOP-KB, https://aopkb.oecd.org/; Box 2), have provided the opportunity to engage stakeholders (users and developers of AOPs) from industry, government, and academia to both evaluate the state of the science and identify next steps in advancing the AOP framework and its uses. ADME = absorption, distribution, metabolism, and excretion; AOP = adverse outcome pathway; qAOP = quantitative AOP. The adverse outcome pathway knowledge base (AOP-KB; image source: https://aopkb.oecd.org/index.html) is the Web-based infrastructure that was created to support the collaborative development, distribution, visualization, and use of AOP knowledge, with the aspiration of serving as the search engine and integration point for the AOP community. The AOP-KB captures effects data from toxicity studies (i.e., integration with Organisation for Economic Co-operation and Development's [OECD's] Harmonized Template 201; http://www.oecd.org/ehs/templates/) and assembles the data to produce quantitative relationships between key events (i.e., through integration with Effectopedia; www.effectopedia.org). The ultimate goal of this KB is to form a comprehensive collection of accessible resources for disseminating AOP knowledge captured by internationally introduced standards. The AOP-KB consists of four separate AOP-related modules. The furthest developed and most widely used module is the AOP-Wiki (https://aopwiki.org/), developed initially by the US Environmental Protection Agency. The other 3 modules include Effectopedia, developed by the OECD and the European Commission Joint Research Centre; AOPXplorer (http://www.aopxplorer.org/), developed by the US Army; and the Intermediate Effects Data Base, which is still under development. Because each module was initially created by a different organization as a stand-alone tool, in various stages of development, to date a complete and well-integrated AOP-KB has not been fully realized. Recently, the OECD launched the e.AOP.Portal, which currently serves as a search engine to mine information from the AOP-Wiki and Effectopedia, bringing the integration of AOP modules a step closer. AOP-Wiki The AOP-Wiki captures manually entered AOP knowledge in a standardized format through the use of crowd sourcing. It constitutes the main source for AOPs to be accessed and evaluated, and it has been instrumental in capturing AOP knowledge and the associated weight of evidence over the past years. The AOP-Wiki has expanded significantly since its initial public release in 2014 with a total of 216 AOP entries, consisting of 6 AOPs that have been endorsed by the OECD after formal review. As the most advanced module of the AOP-KB, the AOP-Wiki was the primary focal point for discussions surrounding advances to the AOP framework throughout the Pellston Workshop. Particularly, work groups explored the strengths and weaknesses of the Wiki in its current state and made recommendations for future iterations as a means to better accommodate multiple AOP stakeholders, which include researchers, risk assessors, and risk managers, with diverse needs and uses for AOP knowledge. A general theme to these discussions included the desire to manipulate the level of detail displayed based on the intended use of the AOP knowledge to address the needs of the stakeholder. DB = database. This manuscript presents a summary and synthesis of the discussions and outcomes from a Society of Environmental Toxicology and Chemistry (SETAC) PellstonTM Workshop, "Advancing the Adverse Outcome Pathway Concept—An International Horizon Scanning Approach," that was held in Cornwall, Ontario, Canada, between April 2 and 6, 2017 (Box 3). The main purpose of this workshop was to begin addressing recognized issues relevant to the development and application of AOPs for chemical risk assessment for both human and ecological health. In preparation for the workshop we engaged the international scientific community via a horizon-scanning effort "to solicit questions concerning the challenges or limitations that must be addressed to realize the full potential of the AOP framework in research and regulatory decision-making," which has been described by LaLone et al. (2017). The questions received during the horizon-scanning stage were subjected to an expert ranking exercise, which identified 4 main themes that were addressed during the workshop by 4 working groups: 1) AOP networks and their applications; 2) quantitative AOPs (qAOPs) and their applications; 3) regulatory use of the AOP framework; and 4) expanding awareness of, involvement in, and acceptance of AOPs to support aspects of predictive toxicology and regulatory decision-making. The present article reports on the outcomes of the actual Pellston Workshop that built on the key questions and associated themes that emerged from the horizon-scanning exercise to develop solutions, strategies, and recommendations to effectively address current challenges and identify critical next steps in realizing the full potential of the AOP framework. Specifically, it focuses on 1) summarizing the overall format, discussions, and cross-cutting themes of this Pellston Workshop, which have culminated in a series of companion journal articles representing the deliberations and outcomes of the 4 workgroups (i.e., Carusi et al. 2018; Knapen et al. 2018; Villeneuve et al. 2018; E.J. Perkins et al. US Army Engineer Research and Development Center, Vicksburg, MS, USA, unpublished manuscript; Coady et al. 2019) providing a perspective on current limitations and roadblocks along with future guidance for the AOP framework to move into mainstream scientific practice. A horizon-scanning, or question solicitation, approach was used to set the stage for the 2017 Pellston Workshop, reaching out to global scientific and regulatory communities (LaLone et al. 2017). Briefly, an online survey was developed asking participants to propose questions that consider key outstanding challenges or limitations that must be addressed to realize the full potential of the adverse outcome pathway framework. From this, approximately 340 valid questions were collected from countries in all continents and across diverse sectors (LaLone et al. 2017). Questions were subjected to an expert ranking exercise and used to develop the themes and charge questions to be addressed at the Pellston Workshop (LaLone et al. 2017). The Pellston Workshop discussions were centered around 4 core themes from this horizon-scanning exercise (Supplemental Data, Table S1). Four work groups, consisting of 8 to 11 experts each, were assigned to address each of the 4 themes and associated questions. In total, 41 international invitees participated in the Pellston Workshop (http://www.saaop.org/workshops/pellston2017.html). These experts were from 9 countries with backgrounds in academia (35%), government (40%), industry (20%), and nongovernmental organizations (5%). AOP = adverse outcome pathway; NGO = nongovernmental organization; qAOP = quantitative AOP. Workgroups were asked to review the state of the science of their respective theme and to focus on demonstrating the application of AOPs in a research and regulatory context (Supplemental Data, Table S1). Charge questions were provided from the top-ranked questions obtained during horizon scanning and the expert ranking exercise to guide deliberations (reviewed by LaLone et al. 2017). During these deliberations it was requested that, when possible, workgroups 1 through 3 use case studies to illustrate the current state of AOP science and how relevant tools for AOP development and evaluation can be applied. Workgroups also were asked to consider current technologies available for capturing, sharing, evaluating, reviewing, and using AOPs, namely the AOP-KB (Box 2). Finally, workgroups 1 through 3 were asked to share challenges associated with communication, collaborative development, and adoption of the AOP framework with workgroup 4, whose charge was to explore how to gain greater awareness, involvement in, and acceptance of AOPs by the international scientific and regulatory/policy communities. The horizon-scanning effort that preceded the Pellston Workshop highlighted the international interest in the AOP framework and its applications for chemical research and regulation (LaLone et al. 2017). Therefore, workgroups addressed the themes from a cross-sector and international perspective, using collaborative discussions with participants across workgroups to ensure that issues attributable to governance/geographic location and entity mission were captured. For example, risk assessors from the other workgroups joined discussions regarding the use of qAOPs in regulatory decision-making. In addition, participants that represented different countries (e.g., Belgium, Canada, China, Germany, Italy, the United Kingdom, the United States) were included in discussions of the regulatory use of AOPs because regulations and regulatory practices can differ significantly from country to country. Workgroups were asked to demonstrate application of the AOP concepts discussed during the Pellston Workshop. It was recognized that case studies would help illustrate the "readiness" of AOP science to address key questions that emerged from the horizon-scanning exercise. Further, case studies help enable the evaluation of whether a particular AOP or AOP network is "fit-for-purpose" (i.e., appropriate for use based on a specific scenario) in addressing regulatory challenges or research questions. Concurrently, workgroups were also asked to develop strategies and recommendations for future initiatives to advance the AOP framework such that it addresses the needs of a broader global stakeholder community (Box 3). From this, with the exception of workgroup 4, each workgroup identified AOP case studies which varied in degree of development and review status (Table 1). Some of the case studies, such as hepatic steatosis or inhibition of aromatase, were used by multiple workgroups because they illustrated diverse characteristics and challenges associated with the development of AOP networks and qAOPs, as well as the application of AOPs in different regulatory contexts (Knapen et al. 2018; E.J. Perkins et al. unpublished manuscript; Coady et al. 2019). Other examples, such as nicotinic acetylcholine receptor (nAChR) "activation leading to colony death/failure in honeybees" (AOP-Wiki identifier 88), were used to make more specific points (e.g., the application of AOPs to inform and support ecological risk assessments; Coady et al. 2019). Depending on workgroup objectives, a variety of case studies were used ranging from individual AOPs illustrating the application to specific questions (e.g., protein alkylation leading to liver fibrosis to demonstrate the application of AOPs to support chemical read-across or targeted ecological risk assessment of nAChR activators in honeybees; Coady et al. 2019) to using the AOP-Wiki as a repository for bioactivity profiling of complex environmental mixtures for prediction of apical hazards by constructing an AOP network (e.g., Knapen et al. 2018; E.J. Perkins et al. unpublished manuscript). Overall, the successful application and illustrative power of the case studies used by the different workgroups demonstrated the utility of the AOP-KB in diverse applications, ranging from chemical design (i.e., the design of chemicals with low hazard to humans and wildlife) and development to ecological risk assessments of complex environmental mixtures. It also was recognized that case studies represent powerful tools for engaging stakeholders, so there is a continual need for further development of case examples that speak to the specific needs of a broader stakeholder community, including those involved in risk management, medicine/health, and policy (Carusi et al. 2018). Furthermore, successful demonstration of relevant case studies is important to gain an overall acceptance of this framework. To date, the majority of activities with regard to AOPs have focused on their development because, out of pragmatic necessity, populating the AOP-KB was the first step in laying the foundation for regulatory application. However, with increasing population of the AOP-KB (Box 2), this initial focus has rapidly turned toward application of the AOP framework in a regulatory context, both to maintain support from current stakeholders and to expand usefulness to new stakeholders. This is reflected in the number of recent workshops that have explored how the AOP concept could be applied to improve regulatory decision-making (reviewed by Coady et al. 2019; and Society for the Advancement of Adverse Outcome Pathways; http://www.saaop.org/workshops/index.html) and identified a number of technical challenges that needed to be addressed. Since these initial workshops, significant progress or suggestions for improvements have been made with regard to developing technical solutions and improvements to address many of these challenges, including improved modeling approaches in support of qAOP development (e.g., Conolly et al. 2017; E.J. Perkins et al. unpublished manuscript), recommendations for optimizing the AOP-Wiki to enable building of AOP networks (Knapen et al. 2018; Villeneuve et al. 2018), and establishing online tools (e.g., the US Environmental Protection Agency's Sequence Alignment to Predict Across Species Susceptibility; https://seqapass.epa.gov/seqapass/) to facilitate cross-species extrapolation (Hecker, 2018), among others. The development and application of this expanding toolbox for AOP development and understanding have been the foundational work of a core group of scientists and practitioners from government, industry, nongovernmental organizations, and academics in North America and Europe with the specific needs of chemical risk assessment in mind. For the AOP framework to expand its applicability, however, it must be broadly accepted by a global stakeholder community, including risk managers and other decision-makers. This requires strategic development of efforts that reach out to, and address the needs of, diverse stakeholder groups across geographic regions (Carusi et al. 2018; Coady et al. 2019). Therefore, the objectives of the 2017 SETAC Pellston Workshop were not only to make progress in addressing technical challenges in the development and application of the AOP framework in regulatory and research contexts but also to explore how to address the needs and gain greater acceptance of the framework by a broader international stakeholder community that includes risk managers and perhaps even those in the biomedical field. Although individual AOPs are a pragmatic unit for development and evaluation, AOP networks composed of multiple AOPs often will be the functional unit for prediction, representing real-world scenarios of pathway perturbation. International efforts to populate the AOP-Wiki have led to the creation of 216 AOPs between 2014 and 2018, providing the basis to begin formally developing and further probing the AOP network concept. The objectives of workgroup 1 were to 1) evaluate the state of the science concerning AOP networks, and 2) recommend best practices to guide future development to ensure meaningful use in research and regulatory decision-making. Discussions surrounding these topics led to the development of 3 manuscripts. The first publication, by Knapen et al. (2018), discussed AOP network development including new recommendations for implementing filters and layers, similar to those used in Geographic Information Systems, depending on intended application to aid in manipulating complex networks and to efficiently extract relevant information depending on the end users' needs. This work used case studies (e.g., hepatic steatosis, thyroid axis disruption, environmental mixtures of chemicals, and polypharmacology) to demonstrate how these filters and layers of networks would be advantageous under specific scenarios. Also described in this publication are approaches for including greater biological detail in AOP networks that may be necessary for applications that require quantitative pathway evaluations. For example, it may be necessary to include greater detail in descriptions of feedback loops, where the product or intermediate in a biological system increases (positive feedback) or decreases (negative feedback) the system as a means to maintain homeostasis and for modulating factors (e.g., external environmental factors or nutritional status) that may influence the biology. In contrast, simplified or less detailed AOP network descriptions may be equally necessary for rapid qualitative evaluations. Therefore, the authors suggested the use of filters and layers that draw on structured information fields captured in the AOP-KB as a way to manage the level of detail extracted for customization of the output based on data needs for specific applications. The publication goes on to introduce the concepts of AOP network topology analysis, which provides a means to evaluate the structure, size, and shape of the network to glean information on pathway interactions, for example. In addition, critical path identification, where a stakeholder focuses attention on a specific pathway because of the research or regulatory question being addressed, was introduced along with descriptions for characterization of interactions among AOP networks. These concepts were then expanded on by Villeneuve et al. (2018), who focused on network analytics or, more simply, the analysis of the AOP network to identify tendencies or patterns that can inform the context or utility of AOPs for a specific end user need. Specifically, this work integrated discussions of how graph theory, which includes mathematical descriptions of biological pathways, can be employed to understand interactions among AOPs and identify pathways of importance (Villeneuve et al. 2018). This publication describes in detail and through examples how network structure can inform assay or model development in considering pathways that meet at a particular KE (i.e., convergent pathways) or alternatively separate after a KE (i.e., divergent pathways). In addition, descriptions of how critical path identification (i.e., identification of the most important path for a specific application) could be used in the selection of the relevant pathways for risk assessment and how interactions between AOP networks can be useful for identification of additive, synergistic, or antagonistic responses were included. The third manuscript builds on concepts from the first 2 publications and demonstrates the utility of graph theory and network analysis, using the AOP-Wiki as a case study, for evaluating large AOP networks and identifying individual AOPs that emerge because of the network analyses (Pollesch et al. 2019). Furthermore, in their article the authors conducted an evaluation of the growth and connectivity, as far as number of AOPs, KEs, and KE relationships, of the AOP-KB since its inception, providing a benchmark of its current state that can be used to understand future growth of the AOP-KB. Overall, the work presented in the series of papers developed from the workgroup 1 discussions at the Pellston Workshop addressed several questions submitted to the horizon-scanning exercise and expanded the ability to more consistently develop, describe, and evaluate AOP networks (Figure 1; Supplemental Data, Table S2). Flowchart of the number of questions collected during the horizon-scanning exercise that were addressed during the Pellston workshop and main themes associated with remaining questions to be addressed by future activities. a The "other" category was represented by a diverse range of questions, from application of the adverse outcome pathway framework to address mixture toxicity to how the framework can be used to support or refute precautionary approaches. A list of all questions can be found in Supplemental Data, Table S2. AOP = adverse outcome pathway; qAOP = quantitative AOP; WG = workgroup. At its most basic, an AOP is a qualitative description of measurable biological endpoints and the relationships between those endpoints. Qualitative AOPs, though appropriate for certain applications such as hazard identification, pose a challenge for risk assessment because the potential risk of chemicals cannot be directly inferred from the inherently descriptive AOP. More quantitative estimates of chemical risk can be achieved using response–response data linking events within an AOP (i.e., a change measured in an upstream KE results in a change in a downstream KE or the AO). This quantitative understanding of the biological pathway can be used to support decision-making. The objective of workgroup 2 was to explore how the AOP framework and KB can be used to develop qAOPs to assess and predict hazards and risks of chemicals or mixtures in support of regulatory decision-making. The manuscript by workgroup 2 (E.J. Perkins et al. unpublished manuscript) focused on characterizing best practices for choosing modeling approaches, model building, qAOP applications, and the necessity for transparent and comprehensive documentation to gain confidence in the use of qAOP models. Different types of quantitative relationships in a qAOP were discussed, including correlations and response–response relationships (i.e., relationship between an upstream and a downstream event). These relationships can be captured by simple mathematical equations or sophisticated biologically based computational models that consider other modulating factors such as compensatory responses or interactions with other biological or environmental variables. Perkins et al. (unpublished manuscript) further explored how chemical agnostic qAOP models can be combined with exposure models or models that describe how chemicals enter and move through the body (e.g., toxicokinetic models), which are by nature chemical-specific. Examples are provided that couple qAOP models with other models to extrapolate the amount of chemical an animal must ingest to have an adverse effect from concentrations causing adverse effects in cell assays (i.e., in vitro to in vivo extrapolation). In addition, the manuscript explores and demonstrates application of these concepts by developing a qAOP network case study for hepatic steatosis using probability-based models (e.g., Bayesian network models) of varying complexity (E.J. Perkins et al. unpublished manuscript). It was shown that Bayesian networks allow for rapid modeling of interactions among AOPs and enable integration of multiple data types collected from cell-based or whole-organism experimentation. The manuscript illustrates the diverse applications of qAOPs ranging from simple qAOP networks that are useful for screening-level decision-making (e.g., to identify and prioritize chemicals for more rigorous testing) to more computationally complex models in support of complex questions and decisions (e.g., understanding the potential hazards of chemical mixtures). The cases presented by workgroup 2 clearly demonstrate the potential of qAOP models to support regulatory risk assessment through predictive assessments based on available mechanistic knowledge. However, examples of mature qAOP models are scarce to date, and there is an urgent need to develop more high-quality, robust, and diverse case studies that have a clearly defined scope and applicability to regulatory risk assessment. Furthermore, it will be crucial to properly define the applicability of specific qAOP models to avoid potential misuse and poor regulatory uptake that would confound confidence and application of such models. The applicability of AOPs to regulatory decision-making processes has been the focus of several recent publications (reviewed in Coady et al. 2019). Although these prior publications concluded that the AOP framework represents a promising and potentially useful approach for this purpose, there is need for increased validation of AOPs, relative to the current practices, for regulatory contexts such as prioritization, categorization, application to integrated testing strategies, and quantitative risk assessment of chemicals (Society for the Advancement of Adverse Outcome Pathways; http://www.saaop.org/workshops/index.html). In recognition of these needs, the objectives of workgroup 3 were to 1) characterize the regulatory decision-making scenarios to which the AOP framework can be applied, 2) establish guidance for decision-makers in determining if and when an AOP is fit-for-purpose, and 3) identify how regulatory needs can inform future AOP development. Coady et al. (2019) focused on the needs of different stakeholder groups including both the regulated community and those involved in regulatory decision-making in context with chemical safety assessments. The workgroup considered various stages of chemical management (e.g., research and development, chemical registrations for manufacture and sale, and postregistration activities) to illustrate the applicability of the AOP framework across different chemical assessment scenarios. Specifically, workgroup 3 explored how the developmental status of a given AOP, from a hypothesized pathway to a quantitative network, could be applied to regulatory scenarios requiring different levels of detail and complexity. They acknowledged that during the different stages of chemical management activities, both stakeholder groups and decision processes can differ significantly. For example, chemical research and development is mostly driven by the regulated community, whereas chemical registration and postregistration processes involve both the regulated and the regulatory communities. Accordingly, fit-for-purpose considerations for AOPs would differ in these different scenarios. Several factors were identified that could inform the fit-for-purpose of an AOP depending on the intended application (Coady et al. 2019). For example, a high degree of confidence and strong weight of evidence for KEs informing regulatory endpoints would be needed if AOPs were to be used in decision-making including chemical management or registration. In contrast, less stringent requirements would apply during early stages of research and development with new chemistries or for prioritization and screening-level assessments. Other scenarios to which AOPs could be applied include emergency situations and bioassay development, as described in the context of the US Endocrine Disruptor Screening Program (EDSP) by Browne et al. (2017), which is mandated to use validated testing methods representing specific KEs to identify and evaluate the potential of chemicals to disrupt the endocrine system of vertebrates including humans. In conclusion, each AOP use scenario has unique requirements with regard to robustness of and confidence in the respective AOPs or AOP networks of interest. Based on these discussions workgroup 3 developed criteria for evaluating the fit-for-purpose of an AOP. Case studies were used to illustrate how AOPs have been, or could be, used in support of regulatory decision-making (Table 1). It was concluded that the AOP framework represents a useful tool in chemical decision-making across the different stages of chemical management but that requirements differ depending on the purpose of the safety assessment. Furthermore, Coady et al. (2019) identified several implementation challenges, including 1) the need for an increased understanding of the AOP framework among regulatory communities, 2) a need for increased demonstration of successful application of AOPs in regulatory decision-making and chemical testing, and 3) the development of validated tools that enable measuring the biological response or effects aligning with relevant KEs in support of chemical screening and prioritization efforts. The AOP framework has been largely developed by a core group of technical and regulatory experts with the specific needs of the chemical risk-assessment and research communities in mind. Despite the momentum gained within the existing AOP community, it became clear during the horizon-scanning exercise that there was limited awareness of the AOP fra