Potential Adverse Outcome Pathways of Chlorinated Organophosphate Flame Retardants

Chlorinated organophosphate flame retardants (Cl-OPFRs) have frequently been detected at high levels in environment and in human biological specimens. As a result, the toxic effects associated with Cl-OPFR exposure warrant increased attention. It is crucial to move beyond merely evaluating carcinogenic or non-carcinogenic risks (1) and engage in a comprehensive discussion on potential adverse health effects and toxic mechanisms of Cl-OPFRs. The adverse outcome pathway (AOP) concept comprises molecular initiating events (MIEs), key events (KEs), and adverse outcomes (AOs), offering a mechanistic understanding of crucial events and biological pathways leading to AOs, thereby enhancing the efficacy of toxicity risk evaluations. Recent studies have confirmed the practicality of this framework (2). This study utilizes the AOP framework to assess the health implications and mechanisms of Cl-OPFRs by integrating existing toxicity data. The findings suggest that Cl-OPFR exposure can result in multi-system toxicity, with a particular emphasis on reproductive issues. Through molecular investigations using tools such as Cl-OPFRs-gene-phenotype-AO framework and AOP-helpFinder, key molecular events (IGF1, BAX, AR, MTOR, and PPARG) linked to hormonal processes and reproductive system development were identified, indicating potential reproductive toxicity induction. This research enhances the understanding of the toxic effects, reproductive toxicity, and mechanisms associated with Cl-OPFRs.

Candidate genes, Gene Ontology (GO) terms, and pathways associated with Cl-OPFRs such as tris-(2-chloroethyl)-phosphate (TCEP), tris-(1-chloro-2-propyl)-phosphate (TCIPP), and tris-(1,3-dichloropropyl)-phosphate (TDCIPP), were identified from the Comparative Toxicogenomics Database (CTD, http://ctdbase.org) in February 2023 using the keywords “TCEP”, “TCIPP”, and “TDCIPP”. Target genes were those with more than three interactions. Target phenotypes were determined by overlapping phenotypes obtained from GO enrichment analysis with a significance threshold of P−3 and relevant phenotypes in the CTD database. The target phenotypes were categorized into three levels — subcellular, cellular, and systemic — based on the hierarchical structure of GO terms (2). The study first identified AOs affecting individuals or populations as a result of exposure to Cl-OPFRs, prioritizing both biological importance and phenotype classification. Phenotypes that showed strong correlations with AOs were then selected as potential intermediate KEs. Additionally, genes associated with these KEs were identified as MIEs to construct an AOP framework. To prioritize MIEs and KEs, we developed chemical-gene-phenotype-disease frameworks utilizing Cytoscape software (version 3.9.1, Boston, MA, USA). We utilized AOP-helpFinder (http://aop-helpfinder-v2.u-paris-sciences.fr/) to identify relevant knowledge linking stressors automatically and events within an AOP and to assess potential candidate genes simultaneously. The assessment was conducted according to Organization for Economic Co-operation and Development (OECD) guidelines utilizing Weight of Evidence (WoE) methodology based on Bradford Hill’s causal considerations, composite score, and confidence score. The aim was to bolster credibility by consolidating evidence from PubMed, Web of Science, and the AOP Wiki (https://aopwiki.org/). WoE criteria primarily focus on biological plausibility and empirical support, incorporating mode of action analysis in chemical regulatory practices. Biological plausibility underscores mechanistic relationships, while empirical support leans on experimental data, particularly dose-response concordance.

Given the widespread presence and long-lasting nature of Cl-OPFRs in the environment, the AOP framework extensively outlined the harmful effects of Cl-OPFRs. A flow diagram depicting this is presented in Figure 1A. Seventy-four interactive genes with Cl-OPFRs effects and 531 shared phenotypes (comprising 483 GO terms and 48 pathways) were identified. These phenotypes were categorized into three levels based on the GO ancestor chart. Notably, at the system level, reproductive toxicity-related phenotypes constituted 43.59% of the total GO terms, followed by organ growth and development at 28.21%, Motor system at 7.69%, and others, indicating potential multi-system toxicity from Cl-OPFRs exposure (Figure 1B). Moreover, an in-depth analysis highlighted that genes such as IL1B, BAX, and BCL2 showed higher frequencies within toxic pathways related to reproduction, while the IGF1 gene emerged as a crucial factor across all levels (Supplementary Table S1).

Reproductive toxicity phenotypes were notably prevalent and thus selected as the AO for the establishment of an AOP framework. Reproductive toxicology terms were classified into 14 categories across three levels of biological organization based on thematic similarities (Supplementary Table S2). At the cellular and subcellular levels, the principal categories included hormone-related phenotypes encompassing biological processes and hormonal stimulation, phenotypes associated with cell damage pertaining to the regulation of cell proliferation and cell cycle, and oxidative stress. At the systemic level, out of 17 reproductive system-related phenotypes, 6 were specifically linked to female reproductive health, whereas only 2 pertained to male reproduction. This discrepancy indicates a potentially higher reproductive toxicity risk from Cl-OPFRs for females. The interconnected phenotypes across the three organizational levels constituted the KEs, and 74 genes identified as interacting with these phenotypes were designated as MIEs, thus forming the foundational structure of the AOP (Figure 2).

Figure 2. 

AOP framework of Cl-OPFRs-induced potential reproductive toxicity.

Abbreviation: AOP=adverse outcome pathway; Cl-OPFRs=chlorinated organophosphate flame retardants; MIEs=molecular initiating events; KEs=key events; AO=adverse outcome; ROS=reactive oxygen species.

To assign priority to MIEs and KEs, we utilized a dataset comprising 74 genes and 8 phenotypic metrics to develop a Cl-OPFRs-gene-phenotype-AO network, consisting of 84 nodes and 434 connections (Figure 3A). The relevance of each gene and phenotype within the network was determined by tallying the number of their connections. The AOP-helpFinder tool leveraged PubChem to compile alternate names for the three Cl-OPFRs identified as stressors, as well as the 11 genes that featured prominently in the gene-phenotype network analysis due to high connectivity. Notably, the genes IGF1, BAX, AR, MTOR, and PPARG showcased significant associations with Cl-OPFRs exposure (Supplementary Table S3). Subsequently, we crafted an AOP model delineating the putative role of Cl-OPFRs in reproductive toxicity, structured around the hierarchical and biological interplay between these components (Figure 3B). The proposed AOP model posits that the expression of the five aforementioned MIEs is disrupted upon exposure to Cl-OPFRs, which in turn perturbs biological processes and hormone-mediated pathways, potentially compromising reproductive development and culminating in reproductive toxicity. To evaluate the robustness of this AOP model, we conducted a WoE assessment in accordance with the OECD Handbook, which entailed scrutinizing the AOP Wiki and relevant literature. As indicated in Supplementary Tables S4 and S5, the significance of the KEs and the validity of the inter-KE relationships were judged to fall within the “moderate” to “high” range, based on criteria such as biological plausibility and supportive experimental and epidemiological studies. In summary, the presented AOP model is characterized by a relatively high degree of credibility.

Figure 3. 

Cl-OPFRs-gene-phenotype-AO framework. The green hexagon node represents Cl-OPFRs; the green diamond node represents potential reproductive toxicity; the round nodes represent target genes, with their proximity to the center of the circle indicating their relative contribution to the framework; the blue rectangle nodes represent system phenotypes, while orange rectangle nodes represent cellular phenotypes, and red rectangle nodes represent subcellular phenotypes. The size of the rectangle reflects the magnitude of their impact on the framework. In total, 434 links extracted from the CTD, and GO and KEGG pathway enrichment analyses results are presented as different connections among the nodes.

Abbreviation: Cl-OPFRs=chlorinated organophosphate flame retardants; CTD=Comparative Toxicogenomics Database; MIEs=molecular initiating events; KEs=key events; GO=geng ontology; AO=adverse outcome; ROS=reactive oxygen species.