Distinguishing mode of action of compounds inducing craniofacial malformations in zebrafish embryos to support dose-response modeling in combined exposures
Graphical abstract
Introduction
On a daily basis, people are exposed to complex mixtures of man-made chemicals, such as pesticide residues which are present in the diet. Many of these chemicals may cause health effects when dosed at a sufficiently high level. However, based on initial risk assessment, exposure can generally be expected to be well below adverse health effect limits with a low associated health risk. Nevertheless, mixtures of chemical substances may cause a combined effect even at a presumed safe level of the individual compounds. This problem of cumulative risk assessment is well recognized in toxicology, and several initiatives have been developed worldwide to address the issue. Some regulations and directives already require mixture assessment, although there is no internationally accepted harmonized strategy for cumulative risk assessment across these initiatives and regulations [1]. In recent years, one of the strategies for cumulative risk assessment was adopted by the European Food and Safety Authority (EFSA).
The EFSA strategy is based on grouping chemicals according to toxicological characteristics, the so-called cumulative assessment groups (CAG) [[2], [3], [4]]. The idea is that chemicals that can induce similar phenotypical effects should be grouped, while chemicals that only act on different systems in the organism are excluded. A first tier in this grouping comprises the target organ or target biological system in the organism (CAG level 1). CAG level 2 then considers the specific toxicological phenotype within that common target organ or system. A further refinement at level 3–4 considers integration about mode and/or mechanism of action (MOA) information. The relevance of MOA information is that simple dose addition (after correction for potency differences) can be safely assumed for chemicals with a similar MOA, whereas chemicals with dissimilar MOA in principle could act independently, or produce positive or negative interaction. This may lead to non-dose additive effects such as antagonism (a lower potency of the mixture compared to the combined single compounds) or synergism (a higher potency of the mixture compared to the single compounds). The absence of information on the toxicological MOA for many chemicals results in grouping under level 2 based on phenomenological effects [5]. A second consequence of the absence of information on toxicological MOA for many chemicals is that EFSA proposed dose addition as a default assumption for cumulative risk assessment [4]. This is considered a conservative approach because dose addition or less-than-dose-addition was more commonly observed compared to more-than-dose-addition in case-studies with mixtures in models relevant to human risk assessment [6,7].
The EU-funded Horizon2020 project “EuroMix” (European Test and Risk Assessment Strategies for Mixtures; https://www.euromixproject.eu/) has elaborated on the EFSA strategy for cumulative risk assessment in proof-of-principle studies, by testing of binary mixtures of chemicals belonging to selected CAGs, and with relevance to exposure of the European population via food. One such study employed the zebrafish embryo as a model for CAG level 2 effects (cleft palate/craniofacial malformations) in CAG level 1 (developmental toxicity) and confirmed dose addition as a common principle [8]. This study used a set of reference compounds with known in vivo developmental effects on the formation of the head skeleton [5].
The transparent appearance of the zebrafish (Danio rerio) embryo enables the microscopic study of skeletal formation. In addition, the zebrafish embryo is a non-licensed experimental organism under current legislation. This renders the zebrafish embryo an excellent model for the study of skeletal development and (toxicant-induced) craniofacial malformations. Despite the differences in skeletal developmental processes (e.g. specific ossification) and the final anatomy of the craniofacial area between and within taxonomic classes, molecular pathways in embryonic development, including the pharyngeal area, are highly conserved among vertebrates.
In the previous study [8], well-defined MOA information was only available for a subset of the tested reference compounds in the present study (Table 1, Table 2), which rendered classification of mixtures as similar or dissimilar MOA approximate in some cases. Therefore, the aim of the present study was to improve this classification with the use of simple experimental models since this could help predict a (newly produced) compound’s contribution to effects in combined exposures.
Full MOA assignment requires comprehensive analysis of genome-wide differential gene expression, preferably of multiple groups of chemically-related substances, in a dose-dependent manner and ideally in a number of models (both simple and complex) relevant to the toxicological phenotype under study. Since this was not achievable within the limitations of the project, we applied an alternative approach, including identification of major MOAs involved in craniofacial malformations and known reference compounds for these, arrangement of these MOAs in an integrated pathway network, and identification of potentially informative marker genes herein. Next, expression of these marker genes was tested for distinctive potential in zebrafish embryos exposed to reference compounds for the identified major MOAs. Expression of the most informative marker genes was then analyzed in zebrafish embryos exposed to test compounds with unknown MOAs, to achieve mapping of these compounds to known MOAs. To support our findings, the results of this marker gene expression were compared to existing activation profiles of reference and test compounds in ToxCast in vitro assays. In addition, the gene expression results were substantiated with in silico analysis of binding affinities (molecular docking modeling) of all compounds to selected biomolecules associated with initiation of the major MOAs.
A literature search to identify MOAs related to craniofacial malformations, including cleft palate, mainly retrieved developmental toxicity studies in classical rodent-based models (Table 1).
Both the disturbance of the retinoic acid (RA) balance and activation of the aryl hydrocarbon receptor (AhR; predominantly AhR-2) are long-established causes of developmental craniofacial malformations [9,10], and could therefore be included as major interactions in the pathway network (Fig. 1). In this line, inhibition of the RA metabolizing enzyme Cyp26 through triazoles and disruption of AhR-signalling through AhR ligands (dioxin-like acting compounds) can be understood as a cause of developmental craniofacial malformations in zebrafish [8,11,12] and rodents [9,13]. Molecular pathways for both MOAs are well described, including the role of inhibition of Cyp26A and B in causing a disbalance of RA and subsequent effects on regulation of hox(n) genes [14], and effects on EGF/TGFβ3 signaling following AhR activation [15]. AhR also regulates sox genes, and in particular involvement of sox9a in cartilage morphogenesis [16] and of sox9b in AhR-related jaw malformations in zebrafish embryos [17] has been established. Additionally detected MOAs include inhibition of histone deacetylases (HDAC) [18] (with a specific role in skull morphogenesis for HDAC8 [19]), modification of folate antagonism [20], induction of oxidative stress [21], inhibition of lysyl oxidase [22], metal chelation [23,24], and activation of peroxisome proliferation activated receptors (PPARs; an important MOA for perfluoroalkyl acids, such as PFOA and PFOS [[25], [26], [27], [28], [29]]).
Valproic acid (VPA), a known HDAC inhibitor, may also interact with coenzyme-A, forming a conjugate which inhibits Cpt1, and thus contributes to VPA-induced malformations [54]. Disruption of cholesterol synthesis may also be a contributory factor, as cholesterol plays a functional role in craniofacial development [55]. 14α-Sterol demethylase (Cyp51) is in vertebrates involved in cholesterol synthesis, and CYP51 knock-out mice also show craniofacial malformations [56]. Cyp51 is the intended (pharmacological) target of triazole fungicides, to inhibit ergosterol synthesis in fungi [57], and its isoform in zebrafish also has an affinity for triazoles [58].
Some reference compounds have a primary MOA related to developmental craniofacial malformations, such as potent AhR activation in the case of TCDD and other dioxin-like compounds. However, other compounds may have additional MOAs to the major mechanism underlying developmental craniofacial malformation (Table 1), such as the dithiocarbamates, which, apart from lysyl oxidase inhibition may affect skeleton formation through heavy metal (copper) chelation, and yet other pathways [22]. Similarly, other teratogens under study may also exhibit multiple MOAs including disruption of the hypothalamus-pituitary-thyroid (HPT) axis with triadimefon in addition to Cyp26 inhibition, or PPARα activation in addition to histone deacetylase inhibition with VPA. In this respect, it is not always clear whether effects on craniofacial formation can be attributed to a primary MOA.
All reference compounds appear to induce further activities, such as effects on fatty acid, steroid, and xenobiotic metabolism pathways [59,60] or inhibition of voltage-gated calcium channels by triazoles [36] or VPA [61]. Although such MOAs may indirectly contribute to craniofacial malformations, these are assumed to be of minor relevance compared to the above MOAs and are therefore not considered further.
Upon identification of major MOAs, key genes from the molecular pathways underlying these MOAs were derived from literature. These served as potential marker genes for further qPCR analysis in the zebrafish embryos. Specifically, this included triazole-related regulation of the RA pathway [30,66,67,71,86,87], dioxin toxicity [15,17,88] and downstream signalling pathways [89], the lysyl oxidase pathway and the related dithiocarbamate-induced disturbance of body axis formation [22,23,51,90], PFAA-induced developmental toxicity [83,84,[91], [92], [93], [94]], thyroid hormone disruption [95], the valproic acid adverse outcome pathway [21,54,[96], [97], [98], [99]], and relevant genes related to oxidative stress [20,97]. Further potential marker genes were derived from literature reviewing molecular epidemiological studies, identifying genes associated with oral clefts in man [82]. Mouse and zebrafish homologs were retrieved for these genes, for comparison with other databases and application in this study. This list was then analyzed for overlap with modes of actions or pathways, which are involved in craniofacial fusion [15], RA signaling [101], and craniofacial development in zebrafish [102]. Based on the overlap and/or specificity among these databases, a subset of 68 genes was selected as a deliverable for the EuroMix project (Table S1) [62]. A final sub selection from this list for actual testing was further fed by re-analysis of our database of previously performed genome-wide expression analysis in zebrafish embryos, with a wide array of reference compounds [66, 86, 100]. The most promising marker genes from the optimized list, which were retained after repeated testing for actual evaluation, were aligned along the assembly of all considered pathways leading to craniofacial malformations in Fig. 1. This arrangement shows that most marker genes have a functional role in the pathway network, although other genes, such as tyr or myod, have no specific or a restricted function in zebrafish head skeleton formation, and just serve as sentinels for pathway activation.
The resulting array of marker genes (Table 3) set the stage for testing and optimization in whole-body mRNA extractions in zebrafish embryos exposed for 72 h to a set of reference compounds with well-known major MOAs, as well as a set of test compounds with unknown MOAs, all known to induce craniofacial malformations. This gene expression was compared to existing ToxCast in vitro activity, and to in silico molecular binding analysis, to accommodate the aim of the study to distinguish similar and dissimilar MOAs in view to support modeling of dose-addition in combined exposure toxicity.
Section snippets
Compound selection
A set of reference compounds was compiled based on data available in the literature linking the compounds to the induction of craniofacial malformations (see Table 1).
Zebrafish
Zebrafish (Danio rerio) were held and bred at the RIVM laboratory under permit NVWA-32600, according to Dutch regulations. Experiments were done with two populations of wild-type (WT) zebrafish. One population was obtained through commercial import from Singapore (referred to as RIVM-WT strain below). The second population of
Gene expression profiles of reference compounds
All reference and test compounds were analyzed for effects on the expression of the final selection of marker genes, as derived from the MOA network associated with craniofacial malformations (see 1.3, Table 3). Some genes showed a consistent pattern of regulation across replicates within a compound class, e.g. cyp26a1 upregulation in triazoles, or sox9b downregulation in TCDD (Table 4a; Supplementary Table 3). Consistent or nearly consistent responses can be considered robust and are marked
Discussion
The basis for risk assessment of mixtures is an informed prediction of the toxicological hazard of the considered combination of chemical compounds. Following the EFSA CAG strategy for an achievable modeling of mixture effects [4], key information for each compound is its target organ or biological process (level 1), the specific phenotype of the effect (level 2), and knowledge about its mode and mechanism of action (level 3/4). The first two levels can be derived from toxicological test
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was part of the H2020-project EuroMix - WP3: Bioassay toolbox and mixture testing (www.euromixproject.eu), which was funded by the European Commission (Grant Agreement 633172), and by the Dutch Ministry of Health, Welfare and Sports (VWS) (project 5.1.2: Knowledgebase and policy advise on CMRS substances). LP and IE were supported by grants from MIUR – Progetto Eccellenza. IE was supported by FFABR 2017 and a Departmental “Linea 2-2019” grant. Shalenie den Braver-Sewradj and Yvonne
References (99)
- et al.
Cumulative risk assessment of pesticide residues in food
Toxicol. Lett.
(2008) - et al.
Approaches for grouping of pesticides into cumulative assessment groups for risk assessment of pesticide residues in food
Regul. Toxicol. Pharmacol.
(2017) - et al.
Dose addition in chemical mixtures inducing craniofacial malformations in zebrafish (Danio rerio) embryos
Food Chem. Toxicol.
(2020) - et al.
TCDD-induced altered expression of growth factors may have a role in producing cleft palate and enhancing the incidence of clefts after coadministration of retinoic acid and TCDD
Toxicol. Appl. Pharmacol.
(1990) - et al.
Stage-dependent abnormalities induced by the fungicide triadimefon in the mouse
Reprod. Toxicol. (Elmsford, N. Y.)
(2011) - et al.
An adverse outcome pathway framework for neural tube and axial defects mediated by modulation of retinoic acid homeostasis
Reprod. Toxicol. (Elmsford, N. Y.)
(2015) - et al.
The role of folate metabolism in orofacial development and clefting
Dev. Biol.
(2015) - et al.
Dithiocarbamates are teratogenic to developing zebrafish through inhibition of lysyl oxidase activity
Toxicol. Appl. Pharmacol.
(2010) - et al.
Dithiocarbamates have a common toxic effect on zebrafish body axis formation
Toxicol. Appl. Pharmacol.
(2006) - et al.
Dithiocarbamates are teratogenic to developing zebrafish through inhibition of lysyloxidase activity
Toxicol. Appl. Pharmacol.
(2010)