Minimizing potential human health risks from endocrine disruptors (EDs) has been a highly prioritized issue in the European Union (EU) for several years (European Parliament, 2013, 2019; European Commission, 2018b) and is specifically highlighted in the European Chemicals strategy for sustainability (European Commission, 2020). EDs are exogenous substances that interfere with the normal function of the endocrine system to cause adverse health effects in an intact organism, or its progeny, or in (sub)populations (WHO/IPCS, 2002). Since the endocrine system regulates tissue and organ development and function, disruption can lead to a range of serious health effects, including reproductive, developmental, and metabolic disorders, cardiovascular diseases, and cancers (WHO/UNEP, 2013; Gore et al., 2015; Duh-Leong et al., 2023). Thus, endocrine disruption represents a mode of action (MoA) that can lead to various toxicities. EDs exert their toxicity through mechanisms such as direct interference with hormone receptors, as well as interference with hormone synthesis, transport, or metabolism. While the endocrine system comprises many different hormones, signalling pathways and organ systems, regulatory assessment of EDs in the EU currently focuses on substances that interact with the estrogen, androgen, thyroid or steroidogenesis (EATS) modalities. The reason is that there is relatively good mechanistic understanding of how perturbations of the EATS modalities may result in adverse effects and which types of effects may arise (ECHA/EFSA/JRC, 2018). There are also several standardized in vivo and in vitro test guidelines available from the OECD and US Environmental Protection Agency (EPA) that include relevant endpoints to investigate EATS-related mechanisms and EATS-mediated adverse effects.

Scientific criteria for identifying EDs have been implemented for biocides and plant protection products in the EU (European Commission, 2017; 2018a) and recently in the European legislation for Classification Labelling and Packaging (CLP) (European Commission, 2023). These criteria stipulate that a substance should be considered an ED if: 1) it shows an adverse effect in an intact organism or its progeny, 2) it has an endocrine MoA, and 3) the adverse effect is a consequence of the endocrine MoA. Consequently, identification of EDs requires a high degree of mechanistic understanding, including understanding of the links between early mechanisms and adverse health effects, to establish an endocrine MoA and perform a MoA analysis (ECHA/EFSA/JRC, 2018). The Adverse Outcome Pathway (AOP) framework is particularly well suited for this purpose, as it provides information on all three criteria.

The AOP framework provides a harmonized approach to describe a causal chain of events linking a molecular initiating event (MIE) to key events (KE) at different levels of biological organization, ultimately leading to an adverse outcome (AO) through defined key event relationships (KER) (Ankley et al., 2010; Villeneuve et al., 2014; Ankley and Edwards, 2018). AOPs thus provide a structured format to summarise available evidence for, and confidence in, the causal links between MIEs and AOs; i.e., they are valuable tools for MoA analysis in ED assessment and for developing Integrated Approaches to Testing and Assessment (IATA) (Browne et al., 2017; OECD, 2017; Audouze et al., 2021; Svingen et al., 2022; Bajard et al., 2023; Wiklund et al., 2023). Development of AOPs is currently overseen by the OECD’s Advisory Group on Emerging Science in Chemicals Assessment (ESCA) in collaboration with the Society for the Advancement of AOPs (SAAOP) and AOP Knowledge Base Coordination Group. Fully developed AOPs can be endorsed by the OECD Working Party on Hazard Assessment (WPHA) and/or Working Group of the National Coordinators for the Test Guidelines Program (WNT) after technical review according to a process described in the AOP Developers’ Handbook (OECD, 2023). For the AOPs to be endorsable, it is important that the peer review is overseen either by the OECD AOP Development Programme or by a scientific journal that has a written agreement with the OECD to handle AOP Reports. With the latter option, an acceptance of the journal submission would make it automatically eligible to be considered for endorsement by the WNT and WPHA. To take full advantage of the potential of AOPs for practical applications, such as ED identification and assessment, the need for more complex AOP networks (AOPNs) of inter-connected AOPs is recognized (Knapen et al., 2018). AOPNs provide further insight into different pathways that are connected in some way, either by contributing to the same AO or by sharing common KEs. This methodology also allows for identifying core KEs and KERs, i.e., events that are central to the development of a specific AO and/or shared by many AOPs. Core KE(R)s may be especially interesting to identify in a MoA analysis, target for testing or use as basis for development of new test methods. AOPNs can also support hazard and risk assessment of chemical mixtures, by providing frameworks for assessing the combined effects of chemicals affecting the same target tissue via different mechanisms (Villeneuve et al., 2018; Beronius et al., 2020).

For AOPs to have practical application, they should be described in a harmonized way and be publicly available. The AOP wiki (https://aopwiki.org/) provides such a repository of AOPs, and several AOPs describing perturbations to hormonal pathways leading to adverse effects are already included in the wiki (Audouze et al., 2021; Ravichandran et al., 2022; Wiklund et al., 2023). However, there is a need for further refinement of existing AOPs to improve overall confidence, as well as to develop additional AOPs describing relevant MoAs that are still missing. This includes reproductive toxicity where effects of EDs on both male and female reproductive systems have been well documented (Schwartz et al., 2019; 2021; Duursen van et al., 2020; Sharma et al., 2020; Rodprasert et al., 2021; You and Song, 2021; Delbes et al., 2022; Marlatt et al., 2022), yet where we still lack fundamental knowledge allowing us to predict with high confidence in vivo outcomes based on data from alternative test methods (Svingen et al., 2022).

The aim of this study was to conduct a systematic inventory of the AOP wiki to identify currently available AOPs that are relevant for mammalian reproductive toxicity and human health. The work was limited to AOPs relevant for EAS modalities. The study also set out to identify core KEs and KERs, i.e., KE(R)s shared by several of the identified AOPs and provide a basis for further AOP development to close gaps and support regulatory identification and assessment of EDs.

We applied an inclusive approach to capture all AOPs in the AOP wiki with relevance for EAS-mediated reproductive toxicity in mammals. The applied methodology was based on that developed and described by Wiklund et al. (2023), as outlined in Figure 1.

This paper describes an AOPN for mammalian reproductive toxicity mediated by the EAS modalities. The construction of the network was based on a systematic inventory of the AOP wiki, where a total of 58 single AOPs were identified as relevant and 42 of these contained sufficient information to be connected into one large network. The network includes 23 MIEs, 22 AOs and 145 other KEs, demonstrating the complexity and interconnectivity of mammalian reproductive toxicity mediated by endocrine mechanisms. The AOPN provides a basis for different applications, such as the identification of needs for future development of KEs, AOs or KERs that are relevant for ED assessment but are currently missing. To this end, we made an initial organisation of the AOPN and identified several subnetworks that describe effects on different hormone pathways and on specific AOs, including effects on male and female reproduction and endocrine-related cancers. Additional filters can be added to the AOPN to refine it further for different applications, for example, to highlight areas of the network relevant for specific taxa or life stages (Knapen et al., 2018), or focus on AOs that are of specific relevance for a certain assessment.

We specifically used the AOPN to identify core KEs and KERs that are common for several existing AOPs and that may be considered central events in EAS-mediated reproductive toxicity. Out of the 26 identified core KEs, 19 are already included as parameters measured in current OECD and US EPA test guidelines, indicating that they are relevant for regulatory ED testing and assessment. Further analyses of the core events for which standard tests are lacking are needed to ascertain their relevance, as well as the practicality and feasibility of measuring them, before any recommendations can be made regarding development and standardization of new tests. Additional potential applications of the AOPN include use as a framework for investigating connections between different pathways within the network that can influence effects in target organs, as well as for assessing combined effects of mixtures containing EDs.

A rigorous multi-step process was developed to identify relevant AOPs for inclusion in the network. This methodology should be useful for creating other AOPNs. It includes searches of the AOP wiki by two different methods, i.e., manually screening each AOP in the wiki by two reviewers and searching the wiki pages using pre-defined search terms, carried out by one reviewer. The results were then cross-checked by a third party and discussed until consensus was reached. Both methods identified the same AOPs in this case and it may seem unnecessarily complex to do both. However, exploring both methods is merited from a wider perspective. Manual review of each AOP in the wiki by an expert can reduce the risk of missing potentially relevant AOPs, as well as prevent the inclusion of irrelevant or unsuitable AOPs. The current AOP wiki is small enough to allow for a manual screening process. As more AOPs are added in the future, however, an efficient search methodology using key terms will be necessary. One limitation in searching the AOP wiki using key terms is that the present format of the wiki does not allow for applying strategies to search for several key terms or phrases simultaneously. Instead, the separate AOP, KE and KER pages must be searched individually using all search terms; a time-consuming process that also increases the risk for potential errors. Future developments to the AOP wiki should enable efficient search strategies to make more efficient use of all the available information (Wiklund et al., 2023).

The identification of AOPs for inclusion in the network relied to a significant extent on expert judgment regarding the relevance of specific AOPs (or parts of AOPs) for describing EAS-mediated reproductive toxicity. For example, the screening of the AOP wiki initially resulted in several differences in the AOPs identified as relevant by the two reviewers, which had to be resolved by discussions. It is therefore likely that AOPNs created by different people will identify slightly different AOPNs, even if developed for the same toxicity outcome(s). This underscores the importance of applying a structured approach and transparent reporting.

To visualize the AOPN, we used a computational workflow to automatically process, filter, and format the AOP data for use in Cytoscape (Wiklund et al., 2023). Visualization in Cytoscape allowed us to identify KEs that were similar or identical, as well as KE(R)s that were missing or incorrectly described in the wiki. The automated workflow facilitates updating and regenerating networks once such issues have been resolved, as it standardises and speeds up the process. Here, similar KEs were combined by reviewing the similarity between KE titles and descriptions, and application of expert judgment. This approach was considered appropriate for the purposes of this project, but it should be noted that whenever this practice is used, it needs to be carefully considered to be fit for purpose to ensure that no critical information from individual KE descriptions is lost. In this project the information in the AOP wiki for each combined KE will be analysed and used in further development of AOPs. Similarly, single KEs that were identified as relevant for this project, but were not part of an AOP, were excluded in the final AOPN but will be considered in during continued AOP(N) development.

New AOPs are continuously being added to the wiki and an additional manual screening of the AOP wiki was performed in August 2023 to identify any relevant AOPs that had been added after the initial AOP search. Four new relevant AOPs were identified: AOP 495 “Androgen receptor activation leading to prostate cancer,” AOP 496 “Androgen receptor agonism leading to reproduction dysfunction in zebrafish,” AOP 503 “Activation of uterine estrogen receptor-alfa leading to endometrial adenocarcinoma, via epigenetic modulation,” and AOP 504 “SULT1E1 inhibition leading to uterine adenocarcinoma via increased estrogen availability at target organ level.” All four AOPs include some KEs and AOs that were not previously described in the AOPN and will be considered when the AOPN is further developed.

Other AOPNs describing perturbations to different endocrine pathways have recently been described, albeit for different purposes and with different scopes. Several studies describe the development of AOPNs containing novel pathways as an alternative to postulating single AOPs, for example, related to disrupted female pubertal onset (Franssen et al., 2022), male reproductive tract abnormalities (Palermo et al., 2021), and for effects on male and female reproductive function in fish (Ankley et al., 2023; Morshead et al., 2023). Others describe AOPNs for specific types of endocrine mediated toxicities and stressors, such as exploring male reproductive toxicity of silver nanoparticles (Kose et al., 2023) and inorganic arsenic (Chai et al., 2021), or the impact of phthalates on female reproduction (Pogrmic-Majkic et al., 2022). In contrast, the AOPN developed here combines existing AOPs related to mammalian reproductive toxicity and provides a broad view of mechanisms and effects related to disruption of the EAS modalities for the purpose of further AOP development. Similar networks have been described by Wiklund et al. (2023) and Ravichandran et al. (2022), however these works focus primarily on development of methods for AOPN construction, filtering, and analyses.

There are limitations that need to be considered when using this and other AOPNs for further scientific or regulatory applications. Importantly, many AOPs in the AOP wiki are not yet fully developed, peer-reviewed, or finally endorsed by the OECD. This is also reflected in the AOPs included in this network; only four of the AOPs have been endorsed by WPHA/WNT and three are currently under review, although many are mature with respect to their development status. Consequently, there is no guarantee that all AOPs included in a network are of sufficient quality to justify using them as basis for conclusions on toxicological hazards. Any information in the AOP wiki, including the AOPN described herein, must therefore be carefully assessed for reliability before the information is used for scientific or regulatory purposes. Such assessment has not yet been conducted for the AOPN we have elaborated here, and it should thus only be viewed as a starting point for further analyses and AOP developments, activities that are currently being conducted.

Another limitation of using AOPNs is that the AOP wiki does not include AOPs for all mechanisms and AOs relevant for different toxicities. Any AOPN must therefore be assumed to be missing information necessary to, for example, reliably identify all relevant core KE(R)s. Indeed, we acknowledge that there are several putative or complete AOPs relevant for EAS-mediated reproductive toxicity published in the open literature but not necessarily entered into the wiki. For example, there are several AOPs describing effects on spermatogenesis and the development of the male reproductive tract, as well as nipple retention in rodents, or prostate and testis tumours caused by androgen receptor (AR) antagonism, estrogen receptor (ER) agonism or reduced testosterone levels via different mechanisms such as inhibition of aromatase, oxidative stress, AhR signalling, and retinoic acid signalling (Christiansen et al., 2020; Draskau et al., 2020; Gray et al., 2020; Johnson et al., 2020; Kortenkamp, 2020; Yokel, 2020; Palermo et al., 2021; Silva da et al., 2021; Myden et al., 2022; Pedersen et al., 2022; Kose et al., 2023). A partial AOP for decreased gonadotropin-releasing hormone (GnRH) expression in hypothalamus leading to reduced testosterone production has been postulated (Lu et al., 2023). There are also published putative AOPs for female reproductive toxicity, including disrupted folliculogenesis and follicle maturation, decreased ovulation, disrupted puberty onset in girls, uterine tumours, and breast cancer causally linked to disrupted hypothalamic–pituitary–gonadal (HPG) axis signalling, ER activation, estrogen metabolism, and aromatase inhibition (Del’haye et al., 2022; Franssen et al., 2022; Hernandez-Jerez et al., 2023; Johansson et al., 2020; Myden et al., 2022; Wikoff et al., 2016). Apart from the challenge of finding AOPs published in the literature, such AOPs cannot be included in the AOPN using the computational automated process and must be manually added.

In conclusion, the AOPN developed here summarises AOPs for mammalian reproductive toxicity via disruption of the EAS modalities currently available in the wiki. It will form the basis for further research and AOP development, including deeper analyses of the subnetworks and core KE(R)s to identify specific areas of prioritization. Such future activities entail, for example, identifying central KERs and developing them further to increase the quantitative understanding, as well as analysing the extent to which EAS-mediated effects relevant for regulatory ED assessment are covered by current AOPs to inform further AOP development. The AOPN is being further refined and applied in ongoing projects with the aim to provide a more complete AOPN and practical examples showcasing how this methodology can be used to support the identification and assessment of EDs.