Skip to main content
NCBI home page
ACS AuthorChoice logoLink to ACS AuthorChoice
. 2025 Feb 27;59(9):4304–4317. doi: 10.1021/acs.est.4c11757

Leveraging Zebrafish Embryo Phenotypic Observations to Advance Data-Driven Analyses in Toxicology

Paul Michaelis †, Nils Klüver †,*, Silke Aulhorn †, Hannes Bohring ‡,§, Jan Bumberger ‡,∥, Kristina Haase ‡,§, Tobias Kuhnert ‡,§, Eberhard Küster †, Janet Krüger †, Till Luckenbach †, Riccardo Massei †,‡,∥, Lukas Nerlich †, Sven Petruschke ‡,§, Thomas Schnicke ‡,§, Anton Schnurpel †, Stefan Scholz †, Nicole Schweiger †, Daniel Sielaff ‡,§, Wibke Busch †,*
PMCID: PMC11912306  PMID: 40012510

Abstract

graphic file with name es4c11757_0006.jpg

Zebrafish have emerged as a central model organism in toxicological research. Zebrafish embryos are exempt from certain animal testing regulations, which facilitates their use in toxicological testing. Next to the zebrafish embryo acute toxicity test (ZFET) according to the OECD TG 236, fish embryos are used in mechanistic investigations, chemical screenings, ecotoxicology, and drug development. However, inconsistencies in the applied test protocols and the monitored endpoints in addition to a lack of standardized data formats impede comprehensive meta-analyses and cross-study comparisons. To address these challenges, we developed the Integrated Effect Database for Toxicological Observations (INTOB), a comprehensive data management tool that standardizes the collection of metadata and phenotypic observations using a controlled vocabulary. By incorporating data from more than 600 experiments into the database and subsequent comprehensive data analyses, we demonstrate its utility in improving the comparability and interoperability of toxicity data. Our results show that the ZFET can detect toxicity spanning 7 orders of magnitude at the scale of effect concentrations. We also highlight the potential of read-across analyses based on morphological fingerprints and their connection to chemical modes of action, provide information on control variability of the ZFET, and highlight the importance of time for mechanistic understanding in chemical exposure-effect assessments. We provide the full Findable, Accessible, Interoperable, and Reusable (FAIR) data set as well as the analysis workflow and demonstrate how professional data management, as enabled with INTOB, marks a significant advancement by offering a comprehensive framework for the systematic use of zebrafish embryo toxicity data, thus paving the way for more reliable, data-driven chemical risk assessment.

Keywords: morphological effects, dose−response modeling, FAIR data, comparative chemical assessment, read across analysis, effect patterns

Short abstract

This article shows how a novel data management tool for zebrafish embryo toxicity data advances comparative chemical risk assessment by enhancing data and metadata standardization, comparability, and cross-study analyses.

Introduction

The release of chemicals into the environment due to human activities poses significant risks to both the environment and human health. Consequently, different legislation worldwide requires comprehensive risk assessment data for the registration of industrial chemicals, pesticides, biocides, and pharmaceuticals. This includes directives and guidelines from various regulatory bodies.1−4 The required data encompass toxicity information across different trophic levels and necessitate experimental testing with vertebrates, particularly fish. Several methods are particularly effective at determining toxicity thresholds and concentrations such as estimating lethal concentration (LCx) and effective concentration (ECx) values or deriving points of departure (POD). As there is great societal and ethical demand to replace animal testing, the framework of the 3Rs (replacement, reduction, refinement) for animal testing was established. Furthermore, there has been a shift toward information-driven evidence-based risk assessment for humans and the environment using more mechanistic information, e.g., within the framework of the adverse outcome pathways (AOP5,6).

Zebrafish (Danio rerio) are increasingly recognized as a valuable model for studying chemical-induced toxicity, not only in environmental toxicology but also in human health. As zebrafish possess orthologs for 70% of human genes, 80% of human disease-related genes, and 86% of general human drug targets,7,8 the model represents a powerful translational system for human hazard and risk assessment, and disease models in drug discovery.9−13 Zebrafish are conveniently maintained and bred in laboratories, and their development is rapid and observable due to the translucency of the egg chorion. They undergo external fertilization and have high reproduction rates, allowing for waterborne exposure to compounds and medium to high throughput experiments with embryos. Additionally, the small-sized and transparent embryos permit direct observation of developmental delays and malformations. Furthermore, zebrafish embryos (ZFE) up to 5 days post fertilization are considered nonprotected stages and an alternative to animal testing under European legislation.14,15 This offers the development of new approach methods (NAMs) using the zebrafish embryo as a whole organism model in certain regulatory frameworks.16,17

The zebrafish embryo acute toxicity test (ZFET) has already been standardized18 and has been promoted as an alternative method for the acute fish toxicity test (AFT). The ZFET can be used in weight-of-evidence approaches within Registration, Evaluation, Authorization, and Restriction (REACH) regulations and it is already used as an approved method for routine testing of wastewater effluents.19,20 The ZFE also served as the model system in developmental toxicity studies to detect teratogenicity and/or developmental neurotoxicity of chemicals.21,22 Furthermore, detailed responses on different effect levels, e.g., morphological phenotype, behavior, or abundance of genes and proteins at the molecular level determined after chemical exposures of ZFE have been used to identify and define toxicological effect fingerprints for particular modes of chemical action.13,23−26

The use of the ZFE model within these different research areas and its use within the regulatory context have increased the number of scientific publications and respective observational data. However, data often do not comply with the Findable, Accessible, Interoperable, and Reusable (FAIR) principles. For example, they are often provided as analyzed results for a particular research hypothesis but not as raw data. Furthermore, varying nomenclature is used in scientifically published literature for toxicological effects of chemicals using early life stages of zebrafish. These inconsistencies underscore the need for harmonizing terminology, experimental metadata, and observational data. It has been shown that providing phenotypic ontology terms and harmonizing data and descriptions led to an improvement in results and the interoperability of data.27 Partially, information on chemical effects on zebrafish has been added to manually curated, publicly available databases like the ECOTOX database (https://cfpub.epa.gov/ecotox/, (last accessed 2024-10-14)) and the zebrafish information network (ZFIN, https://zfin.org/, (last accessed 2024-10-14)). However, missing detailed information and metadata, e.g., on exposure conditions and the lack of standardized ontologies prevent data from being analyzed jointly, compared with each other, or used in interoperable ways, e.g., for model training.28,29

To overcome the above-mentioned obstacles in comparability and joint assessment of toxicological phenotypes, we developed an integrated effect database for toxicological observations (INTOB) for ZFE toxicity data. The tool supports the documentation of experimental metadata and phenotypic observations with a defined vocabulary for observed toxicological endpoints in a structured and machine-readable format, facilitating joint analyses incorporating data from many experiments. Next to applying the software in performing ZFET in our laboratories since 2020, we also entered ZFET data of the last 20 years into the database that were available and stored at our institute. Based on the available data, which we provide via Zenodo, we performed analyses that provide insights into ZFE control variability, correlations of endpoints over time, and similarities of effect fingerprints across chemicals. A comparison with data from the literature reveals that data retrieved from the literature are, so far, unsuitable to perform meta-analyses of ZFET data, underlining the value of the established INTOB system.

Methods and Materials

Overall Approach

Our modular data management tool INTOB provides a relational database structure and a web-based user interface (UI) for data entry and enables us to follow the FAIR (Findable, Accessible, Interoperable, and Reusable) principles for zebrafish embryo (ZFE) toxicity data. It is a modular software that consists of a UI, where experiments can be set up in a repeatable format and chemicals can be selected through a mirror of the CompTox database, ensuring unique identifiers for tested compounds.30 User-defined samples or chemical mixtures can also be entered. The molecular weight is automatically integrated, so that mass in grams can be converted to the amount of substance in moles and vice versa. Concentrations are defined and assigned to containers, which can be selected from predefined types (vial format or common well-type formats). Finally, relevant metadata are entered and stored alongside the experimental data (Table S1).

Observations can be recorded in a way that for each embryo effects can be assigned based on a set of defined effects that are documented within the database and amendable by the users. All data are stored in a relational database which can be explored through the UI by filtering experiments based on their metadata. Experiments can be downloaded via a comma-separated value (CSV) download functionality, where multiple CSV files comprise all metadata, observations, and physicochemical properties pertaining to the selected experiments. Additionally, an application programming interface (API) access is offered via GraphQL and REST, to allow for an integration of INTOB into computational workflows (https://www.ufz.de/intob/index.php?en=51326).

The INTOB system allows structured storage of phenotypic observations at different time points during the exposure experiments and facilitates data export for subsequent analysis (Figure 1). The experimental metadata entries were defined and entered according to Table S1. Next to an upload option for observations from automated image analysis, a defined vocabulary was used for microscopic observations with ZFE at different stages (Table 1) including the terms “coagulation”, “lack of somite formation”, “non-detachment of the tail”, “lack of heartbeat” of the OECD TG 236.18 With this system, we recorded observations for 638 experiments and analyzed them with regard to data quality, effect concentrations, and phenotypic effect patterns.

Figure 1.

Figure 1

Open in a new tab

Data storage and retrieval structure of INTOB. Experimental metadata are stored alongside observations. Data can be retrieved via the user interface in CSV format or an application programming interface (API).

Table 1. List of Morphological Effects with a Unified Vocabularya.

effect id parent id effect time point start (hpf) time point end (hpf)
1   no embryo 0 120
2   coagulated 0 120
3   normal 0 120
4   abnormal eye 0 120
5   abnormal tail effects 0 120
6   developmental delay effects 0 120
7   deformation/malformation 0 120
8   abnormal behavior 0 120
9   heartbeat 48 120
10   abnormal blood circulation 48 120
11   edema 24 120
12   abnormal pigmentation 48 120
13   abnormal hatching 48 120
14   abnormal swim bladder 72 120
15   necrosis 0 120
16   miscellaneous 0 120
17 4 no eye 0 120
18 4 abnormal eye size 0 120
19 5 abnormal tail 0 120
20 5 abnormal tail tip 0 120
21 5 abnormal tail fin 0 120
22 5 abnormal tail length 0 120
23 5 lack of somites 0 120
24 5 no tail detachment 0 120
25 6 developmental delay 0 120
26 6 hours post fertilization 0 120
27 7 deformation head 0 120
28 7 smaller head 0 120
29 7 deformation yolk sac 0 120
30 7 malformation sacculi/otoliths 0 120
31 7 modified structure of chorda 0 120
32 7 scoliosis 0 120
33 7 abnormal heart 48 120
34 8 shivering 0 120
35 8 tremor 0 120
36 8 no spontaneous tail contraction 0 30
37 8 no movement after hatching 72 120
38 9 decreased heartbeat 48 120
39 9 increased heartbeat 48 120
40 9 no heartbeat 48 120
41 9 beats per minute 48 120
42 10 no blood circulation 48 120
43 10 blood congestion 48 120
44 11 yolk sac edema 24 120
45 11 pericardium edema 24 120
46 12 no pigmentation 48 120
47 12 low pigmentation 48 120
48 12 increased pigmentation 48 120
49 13 early hatching 48 72
50 13 no hatching 72 120
51 14 no swim bladder 72 120
52 14 smaller swim bladder 72 120
Open in a new tab
a

The time window where this effect can be observed is defined. Effects are organized in a tree structure, where a parent effect can entail several more detailed related effects. The effect names “coagulated”, “lack of somites”, “no tail detachment” and “no heartbeat” are compliant with those specified in OECD TG 236.

Fish Embryo Acute Toxicity Test

We used wild-type in-house strains of the zebrafish (D. rerio). Fish were cultured at 26 ± 1 °C at a 14:10 h light/dark cycle in a recirculating tank system and used according to German and European animal protection standards and approved by the Government of Saxony, Landesdirektion Leipzig, Germany (Aktenzeichen 75-9185.64).

To harvest the eggs, trays with artificial plants were placed in tanks with male and female fish in the evening. Spawning took place shortly after the lights were turned on in the morning. Eggs were collected and fertilized eggs were sorted and cultured from 2 h post fertilization (hpf) in the control embryo medium until exposure at 26–28 °C. Experiments were carried out based on OECD TG 236,18 although there were differences in e.g., replicate number, exposure start, observation time points, or recorded effects.

Chemical exposures were conducted at 26–28 °C in different exposure vessels, such as, e.g., glass vials (2 mL), crystallization dishes covered with watchmaker glasses, or multiwell-plates (24-, 96-wells) covered with lids. Every experiment included controls (unexposed embryos) and treatments (exposed embryos); if solvents were used, a solvent control was always considered. Embryos were checked daily for effects on phenotype and survival with a stereomicroscope (Table 1) and data were entered into the INTOB database in parallel using a tablet/web browser to access INTOB. All data (metadata and effect data) of each experiment were recorded using the INTOB software.

Data Integration and Selection for Analysis

Next to the established direct input using INTOB on a tablet in the lab, ZFET data recorded at the UFZ during the last 20 years until December 2022 and documented in lab books and Excel files were manually integrated into INTOB. For data analysis, all data recorded in INTOB before the seventh of March 2024 was used. Microscopic phenotypic observations on ZFE morphology, a summary of phenotypic effects, metadata, physicochemical properties of chemicals, and the list of effects were downloaded through the CSV download functionality in the UI. Information on the container layout including concentrations and coordinates in the well plates or vials for embryos was retrieved via the GraphQL API. All analyses described subsequently were conducted in R (v.4.3.031). GraphQL requests were sent to INTOB from R using ghql (v.0.1.0), and the returned JSON (Javascript Object Notation) was converted into tabular format using jsonlite (v.1.8.8).

Data Processing and Quality Control

Initially, experiments with limited data (fewer than 15 embryos in the experiment or only one tested experimental condition) were not considered for analysis. Subsequently, experiments were assessed for the quality of their control (nonchemical-treated) embryos. The probability of control embryos exhibiting no phenotypic morphological effects at the last time point of observation of each experiment was calculated across all experiments to be 93.6%. Assuming effects in unexposed embryos follow a binomial distribution with p(normal) = 0.936 and p(effect) = 1 – p(normal), we conducted one-sided binomial tests on the final control observations for each experiment. For experiments with p-values <0.05 we accepted the alternative hypothesis (p(normal) < 0.936) and did not include them in subsequent analyses. We want to point out that we consciously decided not to adjust p-values. Here, we are interested in the experiments for which we can not reject the null hypothesis (p(normal) = 0.936), with p-value adjustment there would be more experiments that fall into this group, potentially including more low-quality experiments (false negatives). Thus, without p-value adjustment, we aim to limit the type II error. To additionally limit the number of low-power experiments, experiments with fewer than 9 embryos in the control group were excluded (Figure S1).

Data Analysis

Dose–Response Modeling (DRM)

For all remaining single substance experiments with 0 or 24 hpf exposure start points and observations at 48, 72, 96, or 120 hpf, DRM was performed. For lethal DRM, only “coagulated” was considered as lethal, and all other effects were considered as nonlethal. For sublethal DRM, all effects except for “normal” were considered as sublethal, i.e., the cumulative DRMs were fitted. For each substance, lethal and cumulative dose–response curves were fit for all combinations of exposure start and observation time points using the drc package (v.3.0-132). Two parameter log–logistic, Weibull 1 and Weibull 2 models were fitted, and the best fitting model was chosen using Akaike’s Information Criterion (AIC). Further, models were only considered valid if their AIC was lower than the AIC of a linear model with slope = 0, their predicted EC50 was lower than the maximum tested concentration, and the standard error of the EC50 was smaller than the EC50 itself.

Sensitivity Ratio (SR) Analysis

The sensitivity of all endpoints (sublethal + coagulation) was evaluated by modeling the EC50 values (considering all endpoints) and compared with the LC50 for each chemical. In line with the teratogenic index concept,33,34 we calculated the sensitivity ratio (SR) as the ratio of LC50 to EC50.35 A value close to 1 suggests that sublethal or teratogenic effects occur at concentrations close to the lethal ones.

Phenotypic Fingerprints

To compare effect patterns between substances, compound-specific phenotypic fingerprints based on the prevalence of individual effects were derived. To minimize the effect of differently chosen exposure concentration ranges, data from concentrations between the EC5 and the LC99, or in the case of the lethal DRM failed, data from concentrations larger than the EC5 were used. Normal embryos were excluded, and for the remaining data, for each effect, the proportion of embryos exhibiting this effect within the selected concentrations was calculated. These vectors were clustered using hierarchical clustering (Ward method with Euclidean distance). Additionally, principal component analysis (PCA) was carried out to visualize the fingerprints and identify important variables. Since a PCA creates a set of new variables (principal components) by applying linear transformations to the original variables, the contributions (loadings) of each original variable to the PCs can be used to assess the importance of variables for explaining variance. Uniform Manifold Approximation and Projection (UMAP) was applied to the fingerprints to visualize clusters of chemicals with similar fingerprints. In line with the hierarchical clustering, the Euclidean distance measure was used.

Effect Propagation

Data from 48 and 96 hpf observations were used to assess how different effects are correlated with each other. To ensure that the effects in individual embryos could be tracked over time, only experiments using well-type containers (24-well, 48-well, or 96-well plates) were considered. Only qualitative effect measures (yes/no) were used and uninformative (“miscellaneous”, “no embryo”) effects were excluded. Pearson correlation coefficients (PCCs) and p-values were calculated between effects at 48 hpf and effects at 96 hpf. Correlations with p-values <0.01 were considered significant. For visualization, this matrix was clustered using hierarchical clustering with Euclidean distance.

Literature Review of ZFE Toxicological Phenotypes

To compare our systematic approach with data from the literature, we systematically collected and annotated studies from 2014 to 2023 that describe and assess morphological phenotypes in the ZFE after chemical exposure. Studies were searched using PubMed and Science Direct, search terms consisted of combinations of organism-specific terms (“Zebrafish larva”, “Zebrafish embryo”, “Danio rerio larvae”, “Danio rerio embryo”, “FET”) and effect-specific terms (Table S2). Studies were searched for keywords related to effects (Table S2), and information was entered into a Zotero library. Effects were entered and annotated to match the effects used in INTOB (Table 1). Also, metadata was collected accordingly. This information included substance name, embryo age at observation (hpf), and the fish strain (Table S3). Further, we provide this data set in JSON and CSV format (Data_file S1 and Table S4).

Data and Code Availability

The data used in this study is available at Zenodo (10.5281/zenodo.11030299). This includes the data retrieved from the INTOB database, the code used for the analysis, and the results of the literature research as a Zotero library.

Results

Data extracted from the INTOB database comprised a total of 638 experiments. After initial filtering (see methods for details), 609 experiments remained, which included 509 experiments testing single substances, 65 mixture experiments, 34 experiments with environmental samples, and 10 experiments labeled “other”, which mainly include tests with commercial products. The data set comprises experiments on 136 individual substances; on average, a substance has been assessed in 2.6 experiments. The most often investigated substances by the number of embryos tested are 3,4-dichloroaniline (a commonly used positive control substance), azinphos-methyl, and deltamethrin (Figure 2a). Most experiments made observations at three different time points after exposure; however, some experiments included up to 15 time points of observation. An overview of the most commonly used exposure start and observation time points and the associated numbers of experiments is shown in Figure 2b. On average, per experiment 7 individual exposure concentrations were tested using a total of 95 embryos; however, there is variation, as the number of exposure concentrations ranges from 2 to 20 and the number of embryos ranges from 21 to 480 (Figure 2c). In summary, this data set presents a diverse set of experiments with varying sample sizes, exposure conditions, and coverage of observation time points per substance.

Figure 2.

Figure 2

Open in a new tab

Data overview. (a) Number of tested embryos across all experiments for the 10 most frequently tested substances. (b) Number of experiments per exposure start and observation time points (most common time points depicted). (c) Histogram showing the number of experiments with the number of embryos. (d) Red dots signify experiments where control embryos were identified to not stem from a binomial distribution with p(normal) = 0.936 or had fewer than 9 control embryos, and green dots represent experiments for which the null hypothesis was not rejected.

Control Variability in ZFETs

Control embryos, i.e., embryos that were not exposed to toxic chemicals but exposed to media or solvent vehicles (also called negative controls or solvent controls) can exhibit morphological effects, similar to embryos exposed to chemicals. These result from biological variation and environmental factors that are difficult to control. To account for this, experiments are typically discarded if the control embryo quality does not meet a predefined cutoff. For example, the OECD TG 236 requires that 90% or more of the control embryos survive until the end of the experiment at 96 hpf.18 For experiments with defined numbers of embryos, this is adequate; however, with varying sample sizes, this number might need statistical adjustments. For this reason, we implemented a strategy that utilizes binomial tests to determine whether the distribution of normal and non-normal control embryos in a given experiment could plausibly stem from a binomial distribution for which the probability of an embryo exhibiting no effect was calculated from the 11,080 control embryos available in this data set. Across these embryos, the rate at which unexposed embryos are “normal” at the end of an experiment is 93.6%, and the rate at which they survive until the end of the experiment is 95.9%. Applying these criteria to the 609 experiments, 62 were determined to have insufficient embryo quality and were removed before further analysis, leaving 547 experiments for further investigation (Figure 2d).

Dynamic Range of Effect Concentrations in the ZFET

For analyzing and comparing effect concentrations, we selected data from all single substance experiments that followed a 0–96 or 24–96 hpf exposure design, corresponding to 71 unique substances. Out of 180 experiments fitting these criteria, 166 could be used for cumulative DRM, the remaining experiments only assessed mortality and could thus only be used for lethal DRM. Given that data was available, we estimated EC500–96 hpf and EC5024–96 hpf as well as LC500–96 hpf and LC5024–96 hpf for all substances (Figure 3a). We also calculated the sensitivity ratio (SR), within both exposure scenarios. For 58 substances, at least one model could successfully be fitted; for 55 substances, at least one cumulative model (considering all sublethal and lethal effects) could be fitted; for 43 substances, at least one sensitivity ratio could be calculated; and for 13 substances, data were available; however, none of the models succeeded in fitting (Figure 3b). For most substances, SRs were small, only six substances had SRs above 5 (benzovindiflupyr: 22.4; valproic acid: 16.3; ferbam: 9.8; diclofenac sodium: 6.2 (0–96 hpf) and 8.4 (24–96 hpf); acetaminophen: 7.2; propoxur: 6.7).

Figure 3.

Figure 3

Open in a new tab

Concentration ranges for ZFET data with observations at 96 hpf. (a) EC50 and LC50 values were modeled per compound from observational data of experiments with 0 or 24 hpf exposure start points, error bars indicate their standard error. Violin plots indicate the distribution of measured concentrations. (b) Summary of data availability, and modeling. (c) Sensitivity ratios (SRs) when no SR could be calculated, a gray square is displayed. Substance names from (a) carry over to (b, c). Some substance names were abbreviated or changed to shorter synonyms: n-(1,3-Dimethylbutyl)-N′-phenyl-p-phenylenediamine (6-PPDQ), ethyl 3-aminobenzoate methanesulfonic acid salt (tricaine mesylate), and tris(1,3-dichloro-2-propyl) phosphate (TDCPP).

In most experiments (75%), embryos were exposed immediately after fertilization, and fewer experiments initiated exposure at 24 hpf (21%). If the SR could be calculated for both exposure scenarios for one substance, these ratios were generally in agreement (Figure 3c). In a few cases, the SR was higher in experiments with a 24 hpf exposure start, notably in cyprodinil (0 hpf exposure: 1.1, 24 hpf exposure: 3.1) and genistein (0 hpf exposure: 2.4, 24 hpf exposure: 4.3).

Comparing EC50 and LC50 values across exposure start points, there are substances for which the exposure start does not influence effect concentrations, such as citalopram hydrobromide, naproxen sodium, chlorophene, or triclosan. For other substances, there was a time-dependent effect. For example, for verapamil hydrochloride, diclofenac sodium, diuron, diazinon, cyprodinil, and genistein, toxicity was consistently higher when embryos were exposed at 0 hpf as opposed to a 24 hpf exposure start. Two examples were found (carbamazepine and imidacloprid), where later exposure starts (at 24 hpf) led to higher toxicity compared to early exposure start (0 hpf). Similarly, for nicotine, concentration-dependent effects could be observed when ZFE were exposed at 24 hpf; in experiments with an exposure start at 0–2 hpf, no DRMs could be modeled due to low toxicity across all tested concentrations.

The concentrations across these substances are normal-distributed on a logarithmic scale and span 9 orders of magnitude with a range from 2.9e–04 μmol/L (1.2e–04 mg/L) to 4.6e+04 μmol/L (7.0e+03 mg/L), a mean of 3.7e+02 μmol/L (7.7e+01 mg/L) and a standard deviation of 1.9e+03 μmol/L (3.1e+02 mg/L) (Figure 3a).

Clusters of Chemicals with Similar Morphological Effect Fingerprints

For recording the impact of chemicals on the development of ZFE we defined 52 endpoints, which are listed and annotated within the INTOB database (Table 1). With this unified vocabulary, experiments are comparable with each other, and effect fingerprints for individual substances can be analyzed. Therefore, we created toxicological effect fingerprints for experimental data of single substances that (a) followed an exposure design of 0–96 or 24–96 hpf, (b) where at least a cumulative DRM could successfully be modeled, and (c) that were obtained with concentrations between the EC5 and the LC99 (or above the EC5, if no lethal DRM could be modeled). For each effect, the proportion of embryos exhibiting the effect in the respective concentration range was calculated; the vector of all effect proportions defines the fingerprint for a substance. After filtering out substances where fewer than 30 embryos were available to calculate the fingerprint, as well as effects that occurred on average in less than 1% of embryos, fingerprints could be derived for 47 substances, corresponding to 133 experiments and 6225 embryos (Figure 4).

Figure 4.

Figure 4

Open in a new tab

Phenotypic fingerprinting of chemical effects at 96 hpf. (a) A clustered heatmap of phenotypic fingerprints with values representing the proportion of embryos exhibiting an effect when exposed to concentrations between EC5 and LC99 of the respective substance. Some substance names were shortened for brevity: n-(1,3-Dimethylbutyl)-N′-phenyl-p-phenylenediamine (6-PPDQ), ethyl 3-aminobenzoate methanesulfonic acid salt (tricaine mesylate), and tris(1,3-dichloro-2-propyl) phosphate (TDCPP). (b) UMAP embedding of the fingerprints for visualization of relations between clusters. (c) The 10 most important effects for explaining variance based on the Euclidean norm of their contributions to the first five principal components.

Clustering of the substances based on their fingerprints into six clusters revealed distinct groups ranging from 5 to 12 substances in size (Figure 4a,b). Cluster 3 (including isoproturon, ferbam, and carbendazim, among others) is the most distinct cluster, defined by a coagulation phenotype. Substances in this group exhibit almost no sublethal effects. Evidently, these substances have sensitivity ratios close to 1 (Figure 3c).

“Coagulation” is an important endpoint for explaining variance in a PCA of the fingerprints, only surpassed by “no hatching” (Figures 4c and S5). Interestingly, there are clusters in which substances exhibit “coagulation” but fewer “no hatching” effects (clusters 3 and 6, which include, among others, 3,4-dichloroaniline, naproxen, and diuron) and clusters where “no hatching” is a predominant effect, while “coagulation” is less prevalent (cluster 2, including, among others, bisphenol A and valproic acid, and cluster 1, which includes genistein and diclofenac sodium among other compounds). The third most important effect is “low pigmentation”, which very distinctly defines cluster 4 (including 6-propyl-2-thiouracil (6-PTU), 6-PPDQ, lenacil, acetaminophen, and fipronil), in which few other effects were observed. The fourth most important effect for explaining variance is “developmental delay”, which, however, does not seem to be important for defining the clusters. Cluster 5 (including propiconazole, imidazole, and bromoxynil, among others) represents another group of substances that is defined by a small set of effects, namely, the behavioral effects “no movement after hatching” and “shivering”.

Next to effects that seem to be drivers for particular clusters, it is obvious that substances in clusters 1, 2, and 6 exhibit a wide range of sublethal effects (Figure 4a). Substances of cluster 2 cause “no hatching”, “decreased heartbeat” and, to a lesser extent “developmental delay” and edema effects in ZFE. Edema effects (“yolk sac edema” and “pericardium edema”) which are highly correlated, as can be seen by the direction and magnitude of their loadings in the PCA (Figures 4c and S5), occur often after exposure with substances in clusters 1 and 6, and to some extent in cluster 2. Fingerprints for other time points were also calculated and clustered (Figures S2–S5). Comparing them, it stands out that there is always a cluster of substances defined by coagulation. It should be pointed out that individual substances often exhibit different fingerprints at different time points, indicating an effect of the exposure duration.

Occurrence and Co-occurrence of Morphological Effects Dependent on the Exposure Time, Concentration, and Substance

Morphological effects observed at one time point are not independent of those of another time point in the same experiment. To determine the correlation, we calculated Pearson correlation coefficients (PCCs) between vectors of effects from all non-control embryos with observations at 48 hpf and at 96 hpf and from experiments that used well-type containers (24-well, 48-well, or 96-well containers), equating to a total of 778 embryos, (Figure 5a). Effects at 48 hpf are typically strongly correlated to themselves at 96 hpf. One exception is the “decreased heartbeat”, which is only weakly correlated (PCC = 0.14). Instead, this effect, observed at 48 hpf, correlates with more severe heart-related effects at 96 hpf (“no blood circulation” and “no heartbeat”, among other effects). Other examples are “no tail detachment” and “no pigmentation”, which are not significantly correlated with themselves because they are not yet well pronounced after 48 h.

Figure 5.

Figure 5

Open in a new tab

Dependency of observed effects on the time and concentration. (a) Correlations between effects observed at 48 hpf and effects observed at 96 hpf in the same embryos. Correlated effects are marked in red, anticorrelated effects in blue, and insignificant (p > 0.01) correlations are marked with an x. (b–d) Detailed phenotypic fingerprints of (b) triclosan, (c) bisphenol A, and (d) 6-propyl-2-thiouracil (6-PTU), respectively. Panels on top of the fingerprints display the distribution of applied concentrations and modeled EC50 and LC50 values, error bars indicate the respective standard error. Effects that are not valid at a certain time point are marked with a gray line.

PCCs between effects at 48 hpf and the “coagulation” effect at 96 hpf are almost not significant (p > 0.01), exceptions are a strong and logical anticorrelation with “normal” and weak but significant correlations to the three effects specified as lethal in the OECD TG 236, namely “no heartbeat”, “lack of somites”, and “no tail detachment”.

Among all chemical-exposed ZFE (n = 90972) that were considered in this analysis, 53.4% were found to be not affected by the treatment and were tagged as “normal”. The term “normal” is anticorrelated with all effects except itself, indicating that a large proportion of unaffected embryos do not exhibit any adverse effects over time. Additionally, “normal” at 48 hpf and “coagulated” at 96 hpf have the strongest negative correlation of all combinations of effects (PCC = −0.61), indicating that “normal” embryos rarely transition to “coagulated”. Other prominent correlations were observed in cases where 48 hpf embryos were classified as developmentally delayed. Effects at 96 hpf strongly correlated to this were, apart from “developmental delay” itself, “no pigmentation”, “no tail detachment”, “deformation head”, and “no blood circulation”.

Detailed phenotypic fingerprints reveal that individual effects occur in a time and concentration-dependent manner. Figure 5b–d shows three examples with different effect patterns over time and concentration ranges. Observations at 48, 72, and 96 hpf following exposure to triclosan (Figure 5b) exhibit a pronounced effect of exposure duration on the sensitivity ratio. At 48 hpf, a range of sublethal effects could be observed, determining the EC50 at that time point, before coagulation set in at higher concentrations, determining the LC50. At 96 hpf this changes and the distribution of effects is almost binary, where embryos exposed to low concentrations were normal and embryos exposed to higher concentrations coagulated, resulting in a sensitivity ratio of close to 1 (Figure 5b). In contrast, the effect patterns following bisphenol A exposure did not change over time and were comparatively constant (Figure 5c). A concentration dependency can be observed for sublethal effects, where the prevalence of an effect increases with concentration. With higher concentrations, the prevalence of coagulation increases with the result that sublethal effects decrease. It is also obvious that some sublethal effects are more sensitive than others, for example, “decreased heartbeat” could be observed at lower concentrations compared to the more severe effects like “no heartbeat” or abnormal tail effects (Figure 5c). In exposures with 6-PTU (Figure 5d), concentrations were not high enough to reach lethality, and the observed effect patterns are defined by the same set of sublethal effects across all time points. The most frequent effect was “low pigmentation”, which could be observed at concentrations well below concentrations where other sublethal effects started to emerge. For the two edema effects (“pericardium edema” and “yolk sac edema”), additionally to a concentration dependency, an increase of these effects with time could be observed. This could indicate a time-dependent increase of internal concentrations. However, the EC50 value stays almost constant over time as only the number of embryos exhibiting any sublethal effect is counted. Stable EC50 or LC50 values over time, in turn, can indicate that the internal chemical amount is not drastically changing during the observation period. A very slow uptake, e.g., due to a high lipophilicity, with phenotypes only observable at very late time points was not indicated for the three compounds shown here. The influence of physicochemical properties of substances on the occurrence of phenotypes was not investigated here systematically, also due to the lack of such metadata. However, the detailed comparison of phenotypes occurring during the ZFET in a quantitative and qualitative manner shows how standardized data and a comprehensive and comparable analysis enlarge the amount of interpretable information and open up further perspectives and possibilities to use such data, also when further enriched with more metadata, such as, e.g., physicochemical properties.

Comparison of Chemical Effects in ZFET Based on Data from the Literature

In the final step, we aimed to compare our data with data from literature. Therefore, we conducted an extensive literature review of studies assessing ZFE phenotypic effects within a toxicological context published between 2014 and 2023. We obtained 162 individual publications and a total of 522 distinct chemicals, samples, or mixtures that were investigated in those studies. The heterogeneity across all the studies is extremely large with variation in the considered number of effects, time points and exposure windows, and the selection of investigated concentrations (Data_file S1 and Table S4).

We selected substances (a) for which observations at 72, 96 and/or 120 hpf were available and (b) that were described with observations in at least two publications. In most cases, observations are reported in a “yes/no” manner, not allowing for quantitative prioritization of effects. We transferred descriptions of effects to a standardized vocabulary and performed clustering of compounds according to their phenotypic fingerprints that could be generated from the studies. Hierarchical clustering revealed that substances tested in separate studies rarely clustered together, clusters were rather defined by individual studies (Figure S6).

As this research was done after the INTOB database was filled with the in-house ZFET data and due to the selected criteria that should enable phenotype comparisons, the overlap of substances investigated in our study and those where data from literature was available was small. One example, where comparison to our data was possible, is valproic acid. Brotzmann and colleagues performed two studies investigating the effects of valproic acid and different analogues, one study in 2021 with 9 analogues, and another study in 2022 with 14 analogues.36,37 Although substances from these studies presumably act in similar ways, their effect patterns are hardly comparable due to different methodological approaches within the two different studies. Comparing the fingerprints of valproic acid with our data shows some overlaps but no clear picture (Figure S6). Reported EC50 and LC50 values differ across studies, but whether this is due to data analysis, modeling strategy, or experimental differences is difficult to evaluate (Figure S6).

Discussion

In the present study, we analyzed a data set of ZFET data containing over 600 experiments. We provide a method to assess control variability in ZFETs taking into account the varying sample sizes. We show that morphological effect fingerprints can be used to effectively cluster substances and that factors such as exposure start or observation time points can have an influence on such fingerprints. These types of analyses were made possible with the development of INTOB, a tool that stores ZFET data in a machine-readable and consistent format. For further development and use, we provide both, the ZFET data with observations for 136 chemicals as well as the data analysis workflows via Zenodo (10.5281/zenodo.11030299).

Morphological effect fingerprints might be used in the future to infer mechanisms of toxic action if they can be linked to specific phenotypic patterns. Our study provides evidence that hierarchical clustering effectively groups substances with similar phenotypic fingerprints, highlighting relationships between chemical exposure and observed developmental effects in embryos. Truong and colleagues already demonstrated that simultaneous evaluation of different phenotypic and behavioral endpoints of the zebrafish embryo revealed distinct patterns of chemical responses, aiding in the identification of mechanistic pathways.13 In our study, we found a cluster of substances defined by a low pigmentation phenotype. This cluster includes 6-PTU, a model compound for developmental toxicity and inhibitor of thyroid peroxidase, which is involved in the synthesis of the thyroid hormones (TH) triiodothyronine (T3) and thyroxine (T4). A link between thyroid hormone levels and pigmentation has been observed and a mechanism has been suggested where TH regulates the differentiation of the different pigment cells in zebrafish.38 Walpita and colleagues observed low pigmentation in ZFE, when levels of T3 were lowered by antisense oligonucleotide targeting of the type 2 iodothyronine deiodinase, which catalyzes the conversion of T4 to T3, a phenotype they were able to rescue by exogenous exposure with T3.39 Indeed, other substances of this cluster (see Figure 4a) have also been indicated as disruptors of the thyroid hormone system. 6-PPD and its metabolite 6-PPDQ and their effects on the thyroid hormone system have been described and mechanisms of action were hypothesized.40−42 For acetaminophen low pigmentation phenotypes have been observed and transcriptomic analyses indicate involvement of thyroidal pathways.43,44 Exposure to fipronil decreased T3 and T4 levels in adult zebrafish and their offspring,45 presumably by affecting key genes involved in TH synthesis and regulation.46 Conversely, for lenacil the European Food Safety Authority (EFSA) concluded that endocrine disruption criteria are not met in the T-modality, however no results from ZFE were discussed in this review.47

The disruption of the thyroid system during zebrafish development has not only been shown to reduce pigmentation but also the size of the eyes.48 For acetaminophen and lenacil, we could observe effects on head and eye size, albeit not for the other substances in this cluster (Fipronil, 6-PTU, and 6-PPDQ). Benzovindiflupyr exposures also caused decreased head and eye sizes as well as low pigmentation effects; however, in the clustering, it is grouped with other chemicals as it also exhibits many other effects. These findings show the limitations of the current approach but, at the same time, highlight the potential for further mechanism-related studies and generating mechanistic evidence, e.g., on modes of action by read-across. Many effects are not independent of each other, but instead often co-occur or transition into related effects across observations. Investigation of these trajectories, possibly with higher temporal resolution, could give further insight into the ways at which chemicals affect ZFE. In some cases, like triclosan, these time-dependent changes can result in drastically different phenotypic fingerprints for different observation time points (see Figure 5b). Including information on such dynamics in analyses might be crucial to detecting patterns related to specific mechanisms. In this line, the specificity of phenotypes might also be considered in more detail in the future as less specific effects might obscure MoA-related patterns and obstruct the clustering in the current approach.

During the course of the 96 h ZFET, the ZFE develops a liver and gastrointestinal system, which enables active metabolization and biotransformation of chemicals. This can result in detoxification of a compound or the formation of toxic metabolites. Increasing or decreasing EC or LC values over time or within different exposure windows are indicative of such processes, which provides mechanistic evidence for risk assessments. Here, we showed that lower effect concentrations were more often found when exposures start early after fertilization compared with later exposures, possibly indicating detoxification processes in later embryonic stages due to an active metabolism. The sensitive biological processes in early embryos, such as cleavage, blastulation, and gastrulation during the first 10 h of development, are another explanation. Disruption of those processes has the potential to disrupt dorsoventral patterning and the normal trajectory of development later on. The opposite was observed for some compounds (carbamazepine, imidacloprid, and nicotine) that are known to act on the nervous system. The nervous system is developed rather late in ZFE, which may explain the higher toxicity in later stages.49 This gap of the OECD TG 236, which requires an exposure start directly after fertilization, was also highlighted by Sobanska and colleagues when LC50 values of ZFETs were compared with those of adult fish.50 Therefore, we conclude that adding a second window with an exposure start at 24 hpf to the OECD TG 236 would increase the potential of the assay to provide more mechanistic evidence and to completely avoid animal testing with adult fish in the future.

Comparing sublethal and lethal effect concentrations helps to identify chemicals that have the potential to be teratogenic and/or developmentally toxic.25,33,34 In our study, high SRs were found, e.g., for propoxur and valproic acid, two chemicals where an adverse impact on the health of the fetus due to exposure during pregnancy have been described.51,52 Another chemical with a high ratio between LC50 and EC50 is benzovindiflupyr, a rather novel fungicide for which teratogenicity has not yet been described. Using the INTOB database with additional and quantitative endpoints from e.g., automated imaging of morphology or behavior, in larger systematic studies that include well-known teratogens can provide further insight into the potential of the ZFE as a valid prescreening tool for teratogenicity.

One of the limitations of phenotypic assessments is that even users who received the same training judged observed effects differently. A harmonized vocabulary is a step forward to improve reports, and the implementation of ontologies for toxicological screens can reduce the variability in reported phenotypes and enhance agreement and consistency across laboratories.27 Biases introduced by experimenters due to judgment of effects in microscopic observations can be reduced by incorporating data from automated imaging systems and respective software tools as have already been established for the zebrafish embryo model.25,53,54 Such data sets provide more quantitative features and could be extended further with readouts other than morphology, such as behavioral data. The current version of the INTOB software already enables the upload of such data, providing the potential for larger and more detailed analyses in the future and enabling comparative assessments of chemicals, also in the context of high throughput testing.

Structured and machine-readable data in toxicology are key to accelerating the assessment of chemicals and applying read-across assessments. While databases, such as the US EPA ECOTOX DB or the US EPA CompTox dashboard provide data on EC and LC values, mechanistic information on morphological phenotypes is rarely reported.29,30 Within the ZFIN (https://zfin.org/, (last accessed 2024-10-14)28) a repository of zebrafish phenotypes has been annotated with an ontology that is much broader than the one we used in this study. While phenotypes reported in the literature for chemically exposed zebrafish are listed in the ZFIN repository, information on experimental metadata is not provided. Similarly, our own literature review revealed that gaining insights based on data from literature is tedious, as information is often not available in machine-readable formats, and effects are not described using a unified vocabulary, resulting in poor interoperability and comparability. Using the INTOB software helps to overcome these limitations and enables improved data management of toxicological data, including metadata. Via a GraphQL API, the software enables a quick data export in a machine-readable format, e.g., for uploading them into larger repositories, or to run customized analyses.

Besides individual chemicals, the software enables data input for chemical mixtures and complex samples. With this, analysis and comparison of individual chemicals with mixtures are easily possible in terms of quality and quantity of effects and effect concentrations. Complex mixtures in environmental samples contain up to several hundred chemicals55,56 that occur in the aquatic environment across 6–7 orders of magnitude from the ng/L to the mg/L range.56−58 With our study, we could show that the range of concentrations causing sublethal to lethal effects in ZFE for a diverse set of chemicals covers 7 orders of magnitude from the low μg/L to high mg/L ranges. With this, our study reveals for the first time the suitability of the test system ZFE to cover a wide range of concentrations that are relevant in terms of real-life exposures.

In this study, we aimed to investigate the potential of phenotypic fingerprints based on structured and machine-readable ZFET data and metadata and decided on a particular way how to integrate the data for the analyses. We developed fingerprints for specific time points and integrated the observations across different concentrations. The database is structured in a way that multidimensional analyses considering at least concentrations and time for each observation and effect endpoint is possible once the experiments are set up in ways to systematically gain this data. Such data-cube-based analysis can be established in the future to inform and improve chemical risk assessment. Furthermore, toxicological data provided in a FAIR way can be integrated into larger assessment schemes and data networks applied to train AI models on our way toward an in silico toxicology.

Acknowledgments

The development of the INTOB software was supported and funded by the DataHub initiative of the Research Field Earth and Environment of the Helmholtz Association and the Federal Ministry of Education and Research (BMBF), by the UFZ via the INTOB “transmarket” project, a technology transfer project within the “transfun” program, and by the BMBF project DiMEP (FKZ: 16LW0475K). R.M. was funded by the NFDI4Bioimage project (Project number 501864659) funded by the German Research Foundation (DFG). Data were provided and generated within different periods of the program-oriented funding (POF) of the Helmholtz Research Program Earth and Environment (POF II to POF IV (Topic 9)). The authors thank Ulrike Schlägel for support in statistical questions, Paula Klein for data import, and Rebekka Lange for support in the preparation of figures. The authors acknowledge the subcontractor, InfAI GmbH Leipzig, for software development and programming.

Data Availability Statement

Raw data and code at Zenodo (10.5281/zenodo.11030299).

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.4c11757.

  • Additional data analyses, experimental metadata information, standardized effect vocabulary (PDF)

  • Data_file_S1_zotero_phenotype_library (TXT)

  • Table_S4_zotero_phenotype_library (TXT)

Author Contributions

⊥ P.M. and N.K. equal contribution.

The authors declare no competing financial interest.

Supplementary Material

es4c11757_si_001.pdf (1.3MB, pdf)
es4c11757_si_002.txt (2.5MB, txt)
es4c11757_si_003.txt (4.6MB, txt)

References

  1. EPA . Toxic Substances Control Act (TSCA). https://ehs.usc.edu/hazmat/ec/tsca/ (accessed Aug 12, 2024).
  2. EU . Regulation (EC) No 1107/2009 Legislation on Plant Protection Products (PPPs). https://food.ec.europa.eu/plants/pesticides/legislation-plant-protection-products-ppps_en (accessed Aug 12, 2024).
  3. EU . Regulation (EC) No 1272/2008 Classification, Labelling and Packaging of Substances and Mixtures. https://eur-lex.europa.eu/eli/reg/2008/1272/oj (accessed Aug 12, 2024).
  4. EU . Regulation (EC) No 1907/2006 Concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH). http://data.europa.eu/eli/reg/2006/1907/2022-12-17 (accessed Aug 12, 2024).
  5. Ankley G. T.; Bennett R. S.; Erickson R. J.; Hoff D. J.; Hornung M. W.; Johnson R. D.; Mount D. R.; Nichols J. W.; Russom C. L.; Schmieder P. K.; Serrrano J. A.; Tietge J. E.; Villeneuve D. L. Adverse Outcome Pathways: A Conceptual Framework to Support Ecotoxicology Research and Risk Assessment. Environ. Toxicol. Chem. 2009, 29 (3), 730–741. 10.1002/etc.34. [DOI] [PubMed] [Google Scholar]
  6. Patlewicz G.; Simon T.; Goyak K.; Phillips R. D.; Rowlands J. C.; Seidel S. D.; Becker R. A. Use and Validation of HT/HC Assays to Support 21st Century Toxicity Evaluations. Regul. Toxicol. Pharmacol. 2013, 65 (2), 259–268. 10.1016/j.yrtph.2012.12.008. [DOI] [PubMed] [Google Scholar]
  7. Gunnarsson L.; Jauhiainen A.; Kristiansson E.; Nerman O.; Larsson D. G. Evolutionary Conservation of Human Drug Targets in Organisms Used for Environmental Risk Assessments. Environ. Sci. Technol. 2008, 42 (15), 5807–5813. 10.1021/es8005173. [DOI] [PubMed] [Google Scholar]
  8. Howe K.; Clark M. D.; Torroja C. F.; Torrance J.; Berthelot C.; Muffato M.; Collins J. E.; Humphray S.; McLaren K.; Matthews L.; McLaren S.; Sealy I.; Caccamo M.; Churcher C.; Scott C.; Barrett J. C.; Koch R.; Rauch G. J.; White S.; Chow W.; Kilian B.; Quintais L. T.; Guerra-Assuncao J. A.; Zhou Y.; Gu Y.; Yen J.; Vogel J. H.; Eyre T.; Redmond S.; Banerjee R.; Chi J.; Fu B.; Langley E.; Maguire S. F.; Laird G. K.; Lloyd D.; Kenyon E.; Donaldson S.; Sehra H.; Almeida-King J.; Loveland J.; Trevanion S.; Jones M.; Quail M.; Willey D.; Hunt A.; Burton J.; Sims S.; McLay K.; Plumb B.; Davis J.; Clee C.; Oliver K.; Clark R.; Riddle C.; Elliot D.; Threadgold G.; Harden G.; Ware D.; Begum S.; Mortimore B.; Kerry G.; Heath P.; Phillimore B.; Tracey A.; Corby N.; Dunn M.; Johnson C.; Wood J.; Clark S.; Pelan S.; Griffiths G.; Smith M.; Glithero R.; Howden P.; Barker N.; Lloyd C.; Stevens C.; Harley J.; Holt K.; Panagiotidis G.; Lovell J.; Beasley H.; Henderson C.; Gordon D.; Auger K.; Wright D.; Collins J.; Raisen C.; Dyer L.; Leung K.; Robertson L.; Ambridge K.; Leongamornlert D.; McGuire S.; Gilderthorp R.; Griffiths C.; Manthravadi D.; Nichol S.; Barker G.; Whitehead S.; Kay M.; Brown J.; Murnane C.; Gray E.; Humphries M.; Sycamore N.; Barker D.; Saunders D.; Wallis J.; Babbage A.; Hammond S.; Mashreghi-Mohammadi M.; Barr L.; Martin S.; Wray P.; Ellington A.; Matthews N.; Ellwood M.; Woodmansey R.; Clark G.; Cooper J.; Tromans A.; Grafham D.; Skuce C.; Pandian R.; Andrews R.; Harrison E.; Kimberley A.; Garnett J.; Fosker N.; Hall R.; Garner P.; Kelly D.; Bird C.; Palmer S.; Gehring I.; Berger A.; Dooley C. M.; Ersan-Urun Z.; Eser C.; Geiger H.; Geisler M.; Karotki L.; Kirn A.; Konantz J.; Konantz M.; Oberlander M.; Rudolph-Geiger S.; Teucke M.; Lanz C.; Raddatz G.; Osoegawa K.; Zhu B.; Rapp A.; Widaa S.; Langford C.; Yang F.; Schuster S. C.; Carter N. P.; Harrow J.; Ning Z.; Herrero J.; Searle S. M.; Enright A.; Geisler R.; Plasterk R. H.; Lee C.; Westerfield M.; de Jong P. J.; Zon L. I.; Postlethwait J. H.; Nusslein-Volhard C.; Hubbard T. J.; Crollius H. R.; Rogers J.; Stemple D. L. The Zebrafish Reference Genome Sequence and Its Relationship to the Human Genome. Nature 2013, 496 (7446), 498–503. 10.1038/nature12111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Knudsen T. B.; Martin M. T.; Chandler K.; Kleinstreuer N.; Judson R. S.; Sipes N. S. Predictive Models and Computational Toxicology. Methods Mol. Biol. 2013, 947, 343. 10.1007/978-1-62703-131-8_26. [DOI] [PubMed] [Google Scholar]
  10. Patton E. E.; Zon L. I.; Langenau D. M. Zebrafish Disease Models in Drug Discovery: From Preclinical Modelling to Clinical Trials. Nat. Rev. Drug Discovery 2021, 20 (8), 611–628. 10.1038/s41573-021-00210-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Peterson R. T.; Macrae C. A. Systematic Approaches to Toxicology in the Zebrafish. Annu. Rev. Pharmacol. Toxicol. 2012, 52, 433–453. 10.1146/annurev-pharmtox-010611-134751. [DOI] [PubMed] [Google Scholar]
  12. Tal T.; Yaghoobi B.; Lein P. J. Translational Toxicology in Zebrafish. Curr. Opin. Toxicol. 2020, 23–24, 56–66. 10.1016/j.cotox.2020.05.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Truong L.; Reif D.; Mary L. S.; Geier M.; Truong H.; Tanguay R. Multidimensional in Vivo Hazard Assessment Using Zebrafish. Toxicol. Sci. 2014, 137, 212–233. 10.1093/toxsci/kft235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. EU Directive 63 . 63/EU of the European Parliament and of the Council of 22 September 2010 on the Protection of Animals Used for Scientific Purposes, Off. J. Eur. Union, 2010, pp 33–79.
  15. Strähle U.; Scholz S.; Geisler R.; Greiner P.; Hollert H.; Rastegar S.; Schumacher A.; Selderslaghs I.; Weiss C.; Witters H.; Braunbeck T. Zebrafish Embryos as an Alternative to Animal Experiments—A Commentary on the Definition of the Onset of Protected Life Stages in Animal Welfare Regulations. Reprod. Toxicol. 2012, 33 (2), 128–132. 10.1016/j.reprotox.2011.06.121. [DOI] [PubMed] [Google Scholar]
  16. De Castelbajac T.; Aiello K.; Arenas C. G.; Svingen T.; Ramhoj L.; Zalko D.; Barouki R.; Vanhaecke T.; Rogiers V.; Audebert M.; Oelgeschlaeger M.; Braeuning A.; Blanc E.; Tal T.; Ruegg J.; Fritsche E.; Marx-Stoelting P.; Riviere G. Innovative Tools and Methods for Toxicity Testing within PARC Work Package 5 on Hazard Assessment. Front. Toxicol. 2023, 5, 1216369 10.3389/ftox.2023.1216369. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Tal T.; Myhre O.; Fritsche E.; Ruegg J.; Craenen K.; Aiello-Holden K.; Agrillo C.; Babin P. J.; Escher B. I.; Dirven H.; Hellsten K.; Dolva K.; Hessel E.; Heusinkveld H. J.; Hadzhiev Y.; Hurem S.; Jagiello K.; Judzinska B.; Kluver N.; Knoll-Gellida A.; Kuhne B. A.; Leist M.; Lislien M.; Lyche J. L.; Muller F.; Colbourne J. K.; Neuhaus W.; Pallocca G.; Seeger B.; Scharkin I.; Scholz S.; Spjuth O.; Torres-Ruiz M.; Bartmann K. New Approach Methods to Assess Developmental and Adult Neurotoxicity for Regulatory Use: A PARC Work Package 5 Project. Front. Toxicol. 2024, 6, 1359507 10.3389/ftox.2024.1359507. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. OECD . Test No. 236: Fish Embryo Acute Toxicity (FET) Test; OECD Publishing, 2013. [Google Scholar]
  19. Scholz S.; Sela E.; Blaha L.; Braunbeck T.; Galay-Burgos M.; García-Franco M.; Guinea J.; Klüver N.; Schirmer K.; Tanneberger K.; Tobor-Kapłon M.; Witters H.; Belanger S.; Benfenati E.; Creton S.; Cronin M. T. D.; Eggen R. I. L.; Embry M.; Ekman D.; Gourmelon A.; Halder M.; Hardy B.; Hartung T.; Hubesch B.; Jungmann D.; Lampi M. A.; Lee L.; Léonard M.; Küster E.; Lillicrap A.; Luckenbach T.; Murk A. J.; Navas J. M.; Peijnenburg W.; Repetto G.; Salinas E.; Schüürmann G.; Spielmann H.; Tollefsen K. E.; Walter-Rohde S.; Whale G.; Wheeler J. R.; Winter M. J. A European Perspective on Alternatives to Animal Testing for Environmental Hazard Identification and Risk Assessment. Regul. Toxicol. Pharmacol. 2013, 67 (3), 506–530. 10.1016/j.yrtph.2013.10.003. [DOI] [PubMed] [Google Scholar]
  20. Scholz S.; Fischer S.; Gundel U.; Kuster E.; Luckenbach T.; Voelker D. The Zebrafish Embryo Model in Environmental Risk Assessment--Applications beyond Acute Toxicity Testing. Environ. Sci. Pollut. Res. 2008, 15 (5), 394–404. 10.1007/s11356-008-0018-z. [DOI] [PubMed] [Google Scholar]
  21. d’Amora M.; Giordani S. The Utility of Zebrafish as a Model for Screening Developmental Neurotoxicity. Front. Neurosci. 2018, 12, 976. 10.3389/fnins.2018.00976. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. McCollum C. W.; Ducharme N. A.; Bondesson M.; Gustafsson J. A. Developmental Toxicity Screening in Zebrafish. Birth Defects Res. C Embryo Today 2011, 93 (2), 67–114. 10.1002/bdrc.20210. [DOI] [PubMed] [Google Scholar]
  23. Kokel D.; Bryan J.; Laggner C.; White R.; Cheung C. Y.; Mateus R.; Healey D.; Kim S.; Werdich A. A.; Haggarty S. J.; Macrae C. A.; Shoichet B.; Peterson R. T. Rapid Behavior-Based Identification of Neuroactive Small Molecules in the Zebrafish. Nat. Chem. Biol. 2010, 6 (3), 231–237. 10.1038/nchembio.307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Schüttler A.; Altenburger R.; Ammar M.; Bader-Blukott M.; Jakobs G.; Knapp J.; Kruger J.; Reiche K.; Wu G. M.; Busch W. Map and Model-Moving from Observation to Prediction in Toxicogenomics. Gigascience 2019, 8 (6), giz057 10.1093/gigascience/giz057. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Teixidó E.; Kieβling T. R.; Klüver N.; Scholz S. Grouping of Chemicals into Mode of Action Classes by Automated Effect Pattern Analysis Using the Zebrafish Embryo Toxicity Test. Arch. Toxicol. 2022, 96 (5), 1353–1369. 10.1007/s00204-022-03253-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Truong L.; Marvel S.; Reif D. M.; Thomas D. G.; Pande P.; Dasgupta S.; Simonich M. T.; Waters K. M.; Tanguay R. L. The Multi-Dimensional Embryonic Zebrafish Platform Predicts Flame Retardant Bioactivity. Reprod. Toxicol. 2020, 96, 359–369. 10.1016/j.reprotox.2020.08.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Thessen A. E.; Marvel S.; Achenbach J. C.; Fischer S.; Haendel M. A.; Hayward K.; Kluver N.; Konemann S.; Legradi J.; Lein P.; Leong C.; Mylroie J. E.; Padilla S.; Perone D.; Planchart A.; Prieto R. M.; Muriana A.; Quevedo C.; Reif D.; Ryan K.; Stinckens E.; Truong L.; Vergauwen L.; Vom Berg C.; Wilbanks M.; Yaghoobi B.; Hamm J. Implementation of Zebrafish Ontologies for Toxicology Screening. Front. Toxicol. 2022, 4, 817999 10.3389/ftox.2022.817999. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Bradford Y. M.; Van Slyke C. E.; Ruzicka L.; Singer A.; Eagle A.; Fashena D.; Howe D. G.; Frazer K.; Martin R.; Paddock H.; Pich C.; Ramachandran S.; Westerfield M. Zebrafish Information Network, the Knowledgebase for Danio rerio Research. Genetics 2022, 220 (4), iyac016 10.1093/genetics/iyac016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Olker J. H.; Elonen C. M.; Pilli A.; Anderson A.; Kinziger B.; Erickson S.; Skopinski M.; Pomplun A.; LaLone C. A.; Russom C. L.; Hoff D. The ECOTOXicology Knowledgebase: A Curated Database of Ecologically Relevant Toxicity Tests to Support Environmental Research and Risk Assessment. Environ. Toxicol. Chem. 2022, 41 (6), 1520–1539. 10.1002/etc.5324. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Williams A. J.; Grulke C. M.; Edwards J.; McEachran A. D.; Mansouri K.; Baker N. C.; Patlewicz G.; Shah I.; Wambaugh J. F.; Judson R. S.; Richard A. M. The CompTox Chemistry Dashboard: A Community Data Resource for Environmental Chemistry. J. Cheminf. 2017, 9 (1), 61. 10.1186/s13321-017-0247-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. R Core Team . R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2021.
  32. Ritz C.; Baty F.; Streibig J. C.; Gerhard D. Dose-Response Analysis Using R. PLoS One 2015, 10 (12), e0146021 10.1371/journal.pone.0146021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Aalders J.; Ali S.; de Jong T. J.; Richardson M. K. Assessing Teratogenicity from the Clustering of Abnormal Phenotypes in Individual Zebrafish Larvae. Zebrafish 2016, 13 (6), 511–522. 10.1089/zeb.2016.1284. [DOI] [PubMed] [Google Scholar]
  34. Selderslaghs I. W. T.; Van Rompay A. R.; De Coen W.; Witters H. E. Development of a Screening Assay to Identify Teratogenic and Embryotoxic Chemicals Using the Zebrafish Embryo. Reprod. Toxicol. 2009, 28 (3), 308–320. 10.1016/j.reprotox.2009.05.004. [DOI] [PubMed] [Google Scholar]
  35. Bittner L.; Teixidó E.; Keddi I.; Escher B. I.; Klüver N. PH-Dependent Uptake and Sublethal Effects of Antihistamines in Zebrafish (Danio rerio) Embryos. Environ. Toxicol. Chem. 2019, 38 (5), 1012–1022. 10.1002/etc.4395. [DOI] [PubMed] [Google Scholar]
  36. Brotzmann K.; Escher S. E.; Walker P.; Braunbeck T. Potential of the Zebrafish (Danio rerio) Embryo Test to Discriminate between Chemicals of Similar Molecular Structure-a Study with Valproic Acid and 14 of Its Analogues. Arch. Toxicol. 2022, 96 (11), 3033–3051. 10.1007/s00204-022-03340-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Brotzmann K.; Wolterbeek A.; Kroese D.; Braunbeck T. Neurotoxic Effects in Zebrafish Embryos by Valproic Acid and Nine of Its Analogues: The Fish-Mouse Connection?. Arch. Toxicol. 2021, 95 (2), 641–657. 10.1007/s00204-020-02928-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Saunders L. M.; Mishra A. K.; Aman A. J.; Lewis V. M.; Toomey M. B.; Packer J. S.; Qiu X.; McFaline-Figueroa J. L.; Corbo J. C.; Trapnell C.; Parichy D. M. Thyroid Hormone Regulates Distinct Paths to Maturation in Pigment Cell Lineages. eLife 2019, 8, e45181 10.7554/eLife.45181. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Walpita C. N.; Crawford A. D.; Janssens E. D. R.; Van der Geyten S.; Darras V. M. Type 2 Iodothyronine Deiodinase Is Essential for Thyroid Hormone-Dependent Embryonic Development and Pigmentation in Zebrafish. Endocrinology 2009, 150 (1), 530–539. 10.1210/en.2008-0457. [DOI] [PubMed] [Google Scholar]
  40. Bohara K.; Timilsina A.; Adhikari K.; Kafle A.; Basyal S.; Joshi P.; Yadav A. K. A Mini Review on 6PPD Quinone: A New Threat to Aquaculture and Fisheries. Environ. Pollut. 2024, 340 (Pt 2), 122828 10.1016/j.envpol.2023.122828. [DOI] [PubMed] [Google Scholar]
  41. Peng W.; Liu C.; Chen D.; Duan X.; Zhong L. Exposure to N-(1,3-Dimethylbutyl)-N’-Phenyl-p-Phenylenediamine (6PPD) Affects the Growth and Development of Zebrafish Embryos/Larvae. Ecotoxicol. Environ. Saf. 2022, 232, 113221 10.1016/j.ecoenv.2022.113221. [DOI] [PubMed] [Google Scholar]
  42. Zhang S. Y.; Gan X.; Shen B.; Jiang J.; Shen H.; Lei Y.; Liang Q.; Bai C.; Huang C.; Wu W.; Guo Y.; Song Y.; Chen J. 6PPD and Its Metabolite 6PPDQ Induce Different Developmental Toxicities and Phenotypes in Embryonic Zebrafish. J. Hazard. Mater. 2023, 455, 131601 10.1016/j.jhazmat.2023.131601. [DOI] [PubMed] [Google Scholar]
  43. David A.; Pancharatna K. Effects of Acetaminophen (Paracetamol) in the Embryonic Development of Zebrafish, Danio rerio. J. Appl. Toxicol. 2009, 29 (7), 597–602. 10.1002/jat.1446. [DOI] [PubMed] [Google Scholar]
  44. Wang H.; Kang G.; Ma C.; Lian H.; Zhao K.; Zhao B.; Feng Y.; Dong W. Inhibitory Effect of Acetaminophen on Ocular Pigmentation and Its Relationship with Thyroxine in Zebrafish Embryos. Bull. Environ. Contam. Toxicol. 2024, 112 (2), 39. 10.1007/s00128-024-03867-0. [DOI] [PubMed] [Google Scholar]
  45. Xu C.; Niu L.; Liu J.; Sun X.; Zhang C.; Ye J.; Liu W. Maternal Exposure to Fipronil Results in Sulfone Metabolite Enrichment and Transgenerational Toxicity in Zebrafish Offspring: Indication for an Overlooked Risk in Maternal Transfer?. Environ. Pollut. 2019, 246, 876–884. 10.1016/j.envpol.2018.12.096. [DOI] [PubMed] [Google Scholar]
  46. Ma T.; An X.; Wu P.; He X.; Luo Y. Effects of Insecticide and Herbicides on Thyroid Disturbances in Zebrafish. Toxics 2024, 12 (8), 570. 10.3390/toxics12080570. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Álvarez F.; Arena M.; Auteri D.; Leite S. B.; Binaglia M.; Castoldi A. F.; Chiusolo A.; Colagiorgi A.; Colas M.; Crivellente F.; De Lentdecker C.; De Magistris I.; Egsmose M.; Fait G.; Ferilli F.; Santonja G. G.; Gouliarmou V.; Halling K.; Nogareda L. H.; Ippolito A.; Istace F.; Jarrah S.; Kardassi D.; Kienzler A.; Lanzoni A.; Lava R.; Leuschner R.; Linguadoca A.; Louisse J.; Lythgo C.; Magrans O.; Mangas I.; Mavriou G.; Mioč A.; Miron I.; Molnar T.; Padovani L.; Padricello V.; Panzarea M.; Parra Morte J. M.; Rizzuto S.; Romac A.; Rortais A.; Santos M.; Serafimova R.; Sharp R.; Szentes C.; Terron A.; Theobald A.; Tiramani M.; Vianello G.; Villamar-Bouza L. Peer Review of the Pesticide Risk Assessment of the Active Substance Lenacil. EFSA J. 2024, 22 (7), e8860 10.2903/j.efsa.2024.8860. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Baumann L.; Ros A.; Rehberger K.; Neuhauss S. C. F.; Segner H. Thyroid Disruption in Zebrafish (Danio rerio) Larvae: Different Molecular Response Patterns Lead to Impaired Eye Development and Visual Functions. Aquat. Toxicol. 2016, 172, 44–55. 10.1016/j.aquatox.2015.12.015. [DOI] [PubMed] [Google Scholar]
  49. Blader P.; Strähle U. Zebrafish Developmental Genetics and Central Nervous System Development. Hum. Mol. Genet. 2000, 9 (6), 945–951. 10.1093/hmg/9.6.945. [DOI] [PubMed] [Google Scholar]
  50. Sobanska M.; Scholz S.; Nyman A. M.; Cesnaitis R.; Gutierrez Alonso S.; Kluver N.; Kuhne R.; Tyle H.; de Knecht J.; Dang Z.; Lundbergh I.; Carlon C.; De Coen W. Applicability of the Fish Embryo Acute Toxicity (FET) Test (OECD 236) in the Regulatory Context of Registration, Evaluation, Authorisation, and Restriction of Chemicals (REACH). Environ. Toxicol. Chem. 2017, 37 (3), 657–670. 10.1002/etc.4055. [DOI] [PubMed] [Google Scholar]
  51. Andrade C. Valproate in Pregnancy: Recent Research and Regulatory Responses. J. Clin. Psychiatry 2018, 79 (3), 22082. 10.4088/JCP.18f12351. [DOI] [PubMed] [Google Scholar]
  52. Ostrea E. M. Jr.; Reyes A.; Villanueva-Uy E.; Pacifico R.; Benitez B.; Ramos E.; Bernardo R.; Bielawski D.; Delaney-Black V.; Chiodo L.; Janisse J.; Ager J. Fetal Exposure to Propoxur and Abnormal Child Neurodevelopment at 2 Years of Age. Neurotoxicology 2012, 33 (4), 669. 10.1016/j.neuro.2011.11.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Nöth J.; Busch W.; Tal T.; Lai C.; Ambekar A.; Kießling T. R.; Scholz S. Analysis of Vascular Disruption in Zebrafish Embryos as an Endpoint to Predict Developmental Toxicity. Arch. Toxicol. 2024, 98 (2), 537–549. 10.1007/s00204-023-03633-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Teixidó E.; Kießling T. R.; Krupp E.; Quevedo C.; Muriana A.; Scholz S. Automated Morphological Feature Assessment for Zebrafish Embryo Developmental Toxicity Screens. Toxicol. Sci. 2019, 167 (2), 438–449. 10.1093/toxsci/kfy250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Finckh S.; Carmona E.; Borchardt D.; Büttner O.; Krauss M.; Schulze T.; Yang S.; Brack W. Mapping Chemical Footprints of Organic Micropollutants in European Streams. Environ. Int. 2024, 183, 108371 10.1016/j.envint.2023.108371. [DOI] [PubMed] [Google Scholar]
  56. Finckh S.; Beckers L.-M.; Busch W.; Carmona E.; Dulio V.; Kramer L.; Krauss M.; Posthuma L.; Schulze T.; Slootweg J.; Von der Ohe P. C.; Brack W. A Risk Based Assessment Approach for Chemical Mixtures from Wastewater Treatment Plant Effluents. Environ. Int. 2022, 164, 107234 10.1016/j.envint.2022.107234. [DOI] [PubMed] [Google Scholar]
  57. Busch W.; Schmidt S.; Kühne R.; Schulze T.; Krauss M.; Altenburger R. Micropollutants in European Rivers: A Mode of Action Survey to Support the Development of Effect-Based Tools for Water Monitoring. Environ. Toxicol. Chem. 2016, 35 (8), 1887–1899. 10.1002/etc.3460. [DOI] [PubMed] [Google Scholar]
  58. Halbach K.; Möder M.; Schrader S.; Liebmann L.; Schäfer R. B.; Schneeweiss A.; Schreiner V. C.; Vormeier P.; Weisner O.; Liess M.; Reemtsma T. Small Streams–Large Concentrations? Pesticide Monitoring in Small Agricultural Streams in Germany during Dry Weather and Rainfall. Water Res. 2021, 203, 117535 10.1016/j.watres.2021.117535. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

es4c11757_si_001.pdf (1.3MB, pdf)
es4c11757_si_002.txt (2.5MB, txt)
es4c11757_si_003.txt (4.6MB, txt)

Data Availability Statement

The data used in this study is available at Zenodo (10.5281/zenodo.11030299). This includes the data retrieved from the INTOB database, the code used for the analysis, and the results of the literature research as a Zotero library.

Raw data and code at Zenodo (10.5281/zenodo.11030299).

Articles from Environmental Science & Technology are provided here courtesy of American Chemical Society

RESOURCES