Background

The acute fish test (AFT), according to OECD TG 203 [1], is commonly used in the regulatory context to screen for acute toxic effects in aquatic environments [2, 3]. Aquatic toxicity assessment is, therefore, deeply anchored in national and EU legislation. Acute fish data generated with juvenile fish are not only required for the registration of industrial chemicals exceeding production of 10 tons per year [4, 5], for plant protection products [6], veterinary medicines [7], biocides [8] and feed additives [9], but also for environmental monitoring, e.g., in the context of the European water framework directive [10]. As a consequence, the frequent application of OECD TG 203 [1] leads to a very high number of vertebrate experiments, which is not compatible with the 3Rs principle of replacing, reducing and refining animal experiments [11] and the increasing trend towards improvement of animal welfare in animal experimentation [12].

Therefore, the alternative new approach method (NAM) “fish embryo acute toxicity (FET) test” with zebrafish (Danio rerio) embryos according to OECD TG 236 [13] was developed as a substitute to the AFT [14,15,16,17,18,19]. Given that, in legal terms, embryos are not regarded as a protected life stage until they are self-feeding [20, 21], which for zebrafish happens from 120 h post-fertilization (hpf) [12], the FET is classified at least as a refinement, if not as a replacement to the AFT in the sense of Russel and Burch [11]. In various studies, a very good correlation between LC50 values from the FET and AFT has been reported [14, 18, 22,23,24].

However, the FET is still not accepted as a sole alternative to the AFT in most regulatory applications [25], except in waste water testing in Germany according to ISO 15088 [26]. One frequent argument is the assumption that embryos do not possess full metabolization capacities compared to juvenile fish relevant in AFT testing [27, 28]. Limited metabolization capacities of the embryo, however, could lead to an under- or overestimation of toxicity due to biotransformation processes catalyzed by various metabolization enzymes. In particular, one group of phase I metabolism enzymes mainly present in the liver, the cytochrome P450 monooxygenases (CYPs), are most important for bioactivation and detoxification processes as they are capable of catalyzing both oxidation and reduction reactions with broad substrate specificity, whereby the different CYP subfamilies have varying capacities for bioactivation and detoxification [29, 30].

There are various approaches across different levels of biological organization available to study and compare metabolic capacities of zebrafish at different developmental stages. However, due to the differences of applied methods and the still limited number of studies on embryonic and juvenile fish [27, 31], the comparison of metabolic capacities between zebrafish embryos and juveniles remains problematic. One method for the evaluation of metabolism differences is the consideration of the expression, biosynthesis and activity of metabolic enzymes. In a recent review, Loerracher and Braunbeck [31] pointed out that—whenever looked at in detail—all CYP activities could be demonstrated in the zebrafish embryo so far. However, in general, CYP activities [32,33,34,35,36,37] are comparatively low in Danio rerio up to 72/96 hpf compared to later larval developmental stage or in juvenile fish. More specifically, other studies by Verbueken et al. [37] and Otte et al. [38] demonstrated albeit low, but in some cases even constitutive CYP1A activity from the very beginning of zebrafish embryonic development. Another method to differentiate metabolism capacities is to study effects of individual pro-toxicants. In case of allyl alcohol, a drastic discrepancy between AFT and FET data is evident and could clearly be attributed to metabolic enzyme deficiency in zebrafish embryos [39].

To compensate for potential differences in metabolic capacities of zebrafish embryos and juveniles, previous studies included mammalian metabolic systems (rat liver microsomes) into the FET which incubated together with the fish eggs for a limited time of 1–4 h at 32 °C [40,41,42]. Due to the embryotoxicity of the microsomes [43], embryos could only be exposed for a limited time span before transfer to fresh medium for the further test duration, which ended at 48–144 hpf. In contrast, the present study was designed to clearly separate pre-metabolization at optimal mammalian temperature (37 °C) from exposure of the zebrafish embryos for a full exposure duration up to 120 hpf at 26 °C. Thus, from the very beginning of the assay, not only the maternal substances, but also potential metabolites can be tested for embryotoxicity. In addition, incubation of the embryos at temperatures above 26 °C can therefore be avoided, which might represent an additional stressor (besides exposure to the sample and metabolization enzymes).

In contrast to the studies mentioned above, the present study uses S9 homogenates instead of microsomal fractions, since the omission of one centrifugation step during production results in the persistence of cytosolic enzymes in addition to microsomal enzymes [44]. For comparison, two different S9 homogenates are used: A conventional rat liver homogenate chemically induced with phenobarbital/β-Naphthoflavone by ICCR (Roßdorf, Germany) is applied (rat S9), established in several in vitro assays (for example in mutagenicity and genotoxicity testing according to OECD TG 471 [45] and OECD TG 487 [46]). However, conventional rat liver homogenates (microsomes or S9) suffer from various shortcomings: (1) given that they originate from animal experiments, there is a conflict with the idea to establish the FET as an alternative to animal testing in the context of 3R and (2) the composition and activity (and toxicity) of rat liver homogenates may differ considerably even between batches from the same supplier [47]. To overcome the problem of animal experimentation for the production of S9 products, we also test an animal-free, biotechnologically produced S9 (ewoS9R, EWOMIS, Bruchköbel, Germany) produced under-defined conditions from suspensions of rat-derived liver cells cultured in chemically defined medium without animal-derived supplements such as fetal bovine serum. Comparable to the induction of rats for the production of conventional S9 products, rat liver cells are treated in vitro with chemical inducers to stimulate phase I metabolism with special focus on CYP1A activity. The subcellular fraction ewoS9R fraction is obtained by cell harvesting, washing, homogenization and using the supernatant after centrifugation at 9000g.

As model pro-toxicants, allyl alcohol, benzo[a]pyrene (B[a]P), chlorpyrifos (CP), tris(1,3-dichloro-2-propyl) phosphate (TDCPP) and bisphenol A (BPA) are selected to evaluate the effects of rat S9 and ewoS9R, since all of these substances are metabolized via liver enzymes. The industrial chemical allyl alcohol is bioactivated via the cytosolic alcohol dehydrogenase 8a to the more toxic metabolite acrolein [39, 48, 49]. B[a]P, a polycyclic aromatic hydrocarbon, can be oxidized via microsomal CYP1A1/1A2/1B1 enzymes to various metabolites with genotoxic or endocrine disrupting properties [50,51,52,53]. The insecticide CP can be either bioactivated to the neurotoxic chlorpyrifos-oxon or detoxified to 3,5,6-trichloro-2-pyridinol via CYP2B6, 3A4, 2C19 and 1A2 activity [54,55,56,57]. TDCPP, an industrial chemical, is bioactivated by CYP2E1, 2D6, 1A2 and 2C19 present in liver S9 homogenates [58,59,60] to metabolites with elevated mutagenic activity after appropriate bioactivation [61]. The industrial chemical BPA is converted via CYPs 1A2, 2C8, 2C9, 2C18, 2C19, 2D6, 2E1, 3A4, and 3A5 [62,63,64] to various metabolites with differential biological activity: Yoshihara et al. [65] and Jaeg et al. [66] showed that BPA is bioactivated by rat and mouse liver S9 to estrogenic or cytotoxic metabolites, whereas other studies reported lower estrogenic activity after metabolization [67] and lower acute toxicity of BPA metabolites in daphnids [68], if compared with the unmetabolized BPA.

To evaluate the external pre-metabolization, the present study will (a) identify optimal exposure concentrations not affecting embryonic development of both metabolization systems (rat S9 and ewoS9R) including the critical energy carrier β-dihydronicotinamide–adenine dinucleotide phosphate (NADPH), necessary for CYP activity [69] and (b) assess the metabolism capacity of both S9 homogenates exemplary for CYP1A1/1A2/1B1 activity via the fluorescence-based EROD assay [70,71,72,73]. Furthermore, embryotoxic effects of five biotransformable model substances with and without pre-metabolization will be evaluated for toxicity changes and whether external pre-metabolization in the FET increases the comparability of FET to AFT literature data, which could contribute to the acceptance of FET data and, thus, the refinement or even replacement of whole animal testing in regulatory aquatic (eco)toxicology.

Methods

Materials and stock solutions

Details on the origin of chemicals and materials used can be found in the Additional file 1: Tables S1–2, respectively. Stock solutions of benzo[a]pyrene (B[a]P, 0.25–10 mM), bisphenol A (BPA, 5–40 mM), chlorpyrifos (CP, 1–100 mM) and tris(1,3-dichloro-2-propyl)phosphate (TDCPP, 0.1–33 mM) were prepared in dimethyl sulfoxide (DMSO) and stored in brown-glass vials at 4 °C. Allyl alcohol was used as a liquid without solvent [54, 57]. The β-dihydronicotinamide–adenine dinucleotide phosphate (NADPH) stock solution (20 mM) was prepared freshly for each test day in artificial water (294 mg/L CaCl2, 123 mg/L MgSO4, 63 mg/L NaHCO3, 5.5 mg/L KCl, according to ISO 15088 [26]). Artificial water was prepared 1 day before each test day and aerated for at least 1 h; the pH was adjusted to 7.8 ± 0.2.

The rat S9 (protein concentration 33 mg/ml) was obtained from male rats (Wistar, ICCR, Roßdorf, Germany), which were induced by triplicate oral administration of 80 mg/kg body weight phenobarbital and β-naphthoflavone on 3 consecutive days. The rat S9 was stored at -80 °C and thawed maximum twice. The lyophilized ewoS9R was produced from a permanent rat liver cell line cultivated in an animal-reagent-free culture medium, and the cells were induced with CYP1A activators. One vial of lyophilized ewoS9R (stored at room temperature) was dissolved in 469 µl artificial water on the test day to adjust the resulting protein concentration to 4 mg/ml.

Zebrafish maintenance and egg production

Adult zebrafish (wild-type Danio rerio, age of 1–2 years) from the facilities of the Department of Evolutionary Ecology and Environmental Toxicology (Goethe University Frankfurt) were kept in 170 L aquaria with a flow-through system using remineralized osmosis water (26 ± 1 °C, water exchange rate of 30% per week, remineralization with NaHCO3 and sea salt (Aquaforest, Brzesko, Poland). The tank water was purified by UV-light and a biological filter. The zebrafish were fed with dry food (zebrafish SDS, Claus GmbH, Limburgerhof, Germany) twice a day and Artemia sp. (Sanders® Great Salt Lake; Ogden, Utah, USA) once a day. A constant day/night rhythm was maintained at 14/10 h. Plant imitates (weathered inert green filter material) and spawning trays prepared with 1 mm spawning grids were used for egg collection. Mass spawning with 150–200 fish was applied for egg production. Courtship and oviposition began about half an hour after the onset of light, and fertilized eggs were collected at latest 1 h after the onset of spawning.

Fish embryo acute toxicity (FET) test without metabolization

The fish embryo acute toxicity test was performed in a static setup according to OECD TG 236 [13] and Schiwy et al. [74] with the minor modifications mentioned below. Effects and lethality (for endpoints, see Additional file 1: Table S3) were recorded at 48 and shortly before 120 hpf. The test was terminated shortly before 120 hpf, and fish were euthanized by immersion in benzocaine (40 g/L in 96% ethanol).

Test substances were first examined in a standard FET to generate reference data. Therefore, the following assay concentration ranges were used: allyl alcohol 0.5–5,000 µM, B[a]P 0.25–10 µM, BPA 5–40 µM, CP 1–100 µM and TDCPP 0.1–33 µM. Test solutions were prepared by diluting stock solutions with artificial water in glass crystallizing dishes. For B[a]P, BPA, CP and TDCPP, a final DMSO concentration of 0.1% (v/v) was used. From the crystallization dishes, 10 fertilized eggs per test solution were transferred to glass-coated 96-well plates (due to hydrophobicity of the test substances) filled with 200 µl of test solution and sealed with a gas-permeable foil. Due to its liquid, hydrophilic and volatile properties, no solvent was necessary for allyl alcohol; and polystyrene well plates were double-sealed with polyester cover foils to minimize evaporation. Each test was complemented with artificial water as a negative control, 0.1% DMSO as a solvent control and 4 mg/L 3,4-dichloroaniline as a positive control. The embryos were incubated at 26 ± 1 °C until 120 hpf with a light/dark cycle of 14/10 h.

Fish embryo acute toxicity (FET) test with external pre-metabolization

Range-finding experiments were carried out to identify optimal concentrations of rat liver S9, ewoS9R and NADPH for external metabolization, for a detailed description, see Additional file 1: Method M1. For the metabolization with S9 homogenates, exposure solutions were prepared in crystallization dishes (B[a]P, BPA, CP and TDCPP) or in 6- or 24-well plates (allyl alcohol) with the same concentration ranges as mentioned above. In addition, 0.5 mM NADPH in combination with 0.01 mg/ml rat S9 or 0.05 mg/ml ewoS9R were added to the test solutions. The S9 homogenates were kept on ice when not in use. Dishes and well plates were sealed with parafilm® or polyester cover foils, respectively, and incubated for 2 h on a shaking incubator at 37 °C (optimal temperature for mammalian enzymes). Runs with 0.1% DMSO were added as a process control and two S9 homogenate process controls (0.1% DMSO + 0.5 mM NADPH + 0.01 mg/ml rat S9 or 0.05 mg/ml ewoS9R) were included in the metabolization step. After metabolization, the solutions were acclimated to 26 °C before transferring the fertilized eggs for the FET exposure solutions. The tests with pre-metabolization were carried out as described for the tests without pre-metabolization. All experiments were performed in 3–4 independent biological replicates. All data presented originated from experiments, which met the validity criteria set by OECD 236 [13]: with < 10% mortality in negative, solvent and process controls as well as > 30% mortality in the 3,4-dichloroaniline positive control at 120 hpf and ≥ 80% hatching rate in the negative and solvent control at 120 hpf.

For data analysis, concentration–response curves were generated using a non-linear regression model (least squares fit, variable slope, four parameters, constrains of bottom = 0% and top = 100%) in the software GraphPad Prism 9 (GraphPad Software, Boston, USA). The statistical comparison of concentration–response curves was carried out using an extra-sum-of-squares F test with a p value of 0.025 for two pairs of analysis (or 0.0083 in the case of BPA with four pairs of analysis). The statistical significance of different curve parameters (EC50 and hillslope) was calculated to compare the treatments including the metabolization via rat S9 or ewoS9R with the experiments without metabolization. If 50% effect could not be generated in the concentration–effect curves, only hill slopes were compared statistically between the treatments.

Impact of S9-proteins on bioavailability of chemicals

To evaluate whether the addition of S9 homogenates had an impact on the bioavailability of the test compounds, a case study with BPA was conducted. The FET methodology was conducted as described above without the energy source NADPH, but with the S9 homogenates. If similar embryotoxicities between approaches with and without NADPH were observed, this indicated binding of the model substance to the S9 proteins, but no detoxification via the metabolization enzymes.

Biochemical determination of ethoxyresorufin-O-deethylase (EROD) activities

To compare the CYP1A1/1A2/1B1 activities of the S9 homogenates, an EROD assay was conducted according to Donato and Gómez-Lechón [71]. In brief, 10 µM ethoxyresorufin, 100 µM dicumarol, 0.5 mM NADPH and 0.01 mg/ml rat S9 or 0.05 mg/ml ewoS9R were mixed in artificial water in a final volume of 200 µl in a 96-well plate. Assays were conducted under light-reduced settings to account for the light sensitivity of ethoxyresorufin. The S9 homogenates were stored on ice and were added immediately before the start of the measurement. In addition, a negative control (artificial water, 10 µM ethoxyresorufin, 100 µM dicumarol, 0.5 mM NADPH) without S9 homogenates was included for each test. The fluorescence at excitation/emission wavelengths of Em: 588 nm/Ex: 635 nm was measured kinetically for 4.5 h with a Spark multimode plate reader (Tecan, Crailsheim, Germany) pre-heated to 37 °C. The test was carried out in 3 biological replicates and 3 technical repetitions. Mean (calculated from 3 biological replicates) in relative fluorescent units (RFUs) measured with S9 homogenates minus the RFUs of the negative control were plotted against time.

Results

The range-finding experiments for the pre-metabolization step (Fig. 1 and Additional file 1: Fig. S1) revealed that 3.5 h incubation at 37 °C with a combination of 0.5 mM NADPH together with 0.01 mg/ml rat S9 or 0.05 mg/ml ewoS9R did not produce any symptoms of lethal and sublethal toxicity in the FET (≤ 10% effects at 120 hpf).

Fig. 1
figure 1

Identification of optimal pre-metabolization conditions: Effects [%] in the FET with Danio rerio at 120 h post-fertilization (hpf) after incubation at 37 °C with different NADPH-concentrations and different rat S9 (A) oder ewoS9R (B) concentrations for 3.5 h before addition of zebrafish eggs. Dotted lines indicate 10% effect (considered the non-effect threshold according to OECD TG 236). Data given as mean values ± standard deviation from independent biological replicates (n = 3–4). 0 = no effects detected, x = experiment not conducted

Full size image

In these settings, both S9 homogenates clearly showed an EROD activity (Fig. 2 and Additional file 1: Fig. S2). The amount of resorufin formed increases with the addition of both S9 homogenates over the 4.5-h period. Both S9 homogenates show a steeper slope within the first 2 h followed by a gradual levelling off. The increase in EROD activity relative to the negative control was higher with rat S9 (relative fold-increase over controls: 39.1) than with ewoS9R (fold-increase: 3.8).

Fig. 2
figure 2

EROD activity respresented as relative fluorescence units (RFUs) at measuring wavelenghts of 588/635 nm in presence of 0.5 mM NADPH, 100 µM dicoumarol and 0.01 mg/ml rat S9 ( −) or 0.05 mg/ml ewoS9R ( −), respectively. Mean and error bars show relative RFU values corrected for RFUs of the negative control (n = 3) with standard deviation

Full size image

In FET experiments amended with external pre-metabolization, the effects after 48 hpf (Additional file 1: Fig. S5) and 120 hpf (Fig. 3) for the five model substances were recorded with and without metabolization. The calculated EC50 values are shown in Additional file 1: Table S4. At the end of the FET tests (120 hpf), resulting embryotoxicity was significantly different after metabolization for three of five model substances (allyl alcohol, TDCPP, BPA).

Fig. 3
figure 3

Biotransformation potential of the two metabolization systems within the FET using Danio rerio embryos: The effects at 120 h post-fertilization (hpf) of allyl alcohol (A), TDCPP (B), bisphenol A (C), benzo[a]pyrene (D) and chlorpyrifos (E) are shown with or without the addition of 0.01 mg/ml rat S9 and 0.05 mg/ml ewoS9R under addition of 0.5 mM NADPH. Controls contained S9 homogenates and NADPH at the same concentrations w/o chemicals. Data are given as means ± standard deviation (n = 3–4). Model for nonlinear regression (GraphPad Prism 9): [agonist] vs. response, variable slope (four parameters), bottom = 0%, top = 100%. *Significant differences of the best-fit values (hillslope, EC50) with S9 homogenate compared to no addition of an external metabolization system

Full size image

For allyl alcohol in combination with rat S9 and ewoS9R, a significantly higher embryotoxicity was observed compared to allyl alcohol without metabolization at both 48 and 120 hpf, with metabolic activation being stronger with rat S9 than with ewoS9R (Fig. 3 and Additional file 1: Fig. S5). The most frequent effects at 120 hpf without metabolization were coagulation and cardio-vascular effects; in contrast, with metabolization coagulation was most frequent (Additional file 1: Fig. S4).

For TDCPP, the two external metabolization system differed regarding their effects at both time points. At 120 hpf, the addition of rat S9 significantly increased TDCPP toxicity, whereas no significant difference was found for ewoS9R compared to effects of the unmetabolized substance (Fig. 3). However, at 48 hpf (Additional file 1: Fig. S5), the difference was significant with both S9 homogenates, and TDCPP proved more toxic with ewoS9R than with rat S9. The profile of frequent observed effects (see SI, Additional file 1: Fig. S4) was similar comparing the approaches with and without metabolization with coagulation being the most observed effect. However, the percentage of cardio-vascular effects increased with the addition of S9 homogenates.

The implementation of S9 homogenates into the FET with BPA resulted in both a decrease or increase of toxicity depending on time of development. At 120 hpf, addition of both S9-homogenates resulted in significantly lower toxicity of BPA (Fig. 3), with no hatching and cardio-vascular effects being the most frequently observed embryotoxic effects (with and without metabolization, Additional file 1: Fig. S4). In contrast, at 48 hpf (Additional file 1: Fig. S5), a reaction opposite to 120 hpf were seen, with the BPA toxicity being significantly higher only when adding ewoS9R compared to no addition metabolization.

To exclude lower toxicity due to binding to S9 homogenate proteins, S9 homogenates without addition of the energy carrier NADPH were also tested with BPA (Additional file 1: Fig. S3). Without NADPH, the proteins of S9 did not significantly alter the embryotoxicity, as the concentration–response curves were almost identical to the concentration–response relationship of chemical exposure without any S9 addition.

For B[a]P, no significant differences were found between treatments without or with addition of S9 homogenates at both 48 and 120 hpf (Fig. 3 and Additional file 1: Fig. S5). Due to precipitation at concentrations > 10 µM, a full concentration–response relationship could not be established for B[a]P. As the overall number of effects with B[a]P was very low, no frequent effects were reported in Additional file 1: Fig. S4.

Likewise, for CP, no complete concentration–response relationship could be generated due to precipitation at > 100 µM, and no significant differences could be recorded for treatment with and without S9 homogenates at 48 and 120 hpf (Fig. 3 and Additional file 1: Fig. S5). However, a slight increase in toxicity was evident when ewoS9R (48 and 120 hpf) or rat S9 (120 hpf) were added. Frequent effects without pre-metabolization at 120 hpf were cardio-vascular effects, coagulation, no hatching as well as pericardial and yolk sac edemata (Additional file 1: Fig. S4). However, following addition of S9 homogenates, coagulation was the most frequently observed effect.

Discussion

Pre-metabolization for the fish embryo toxicity test (FET)

Inclusion of an external metabolization step prior to initiation of the actual FET to allow continuous exposure of the embryos/larvae to the metabolites bears several challenges:

Temperature

First, the protocol for this study was designed to allow optimal temperatures for each system: 37 °C for incubation with the mammalian S9 homogenates and 26 °C for exposure of zebrafish embryos. For mammalian enzymes, 37 °C is usually the optimal temperature, and a lower temperature results in lower enzyme activity [75]. For zebrafish, however, higher temperatures may cause an acceleration of development [76], which compromises the comparability with FETs performed at 26 °C according to OED TG 236 [13] and may lead to conflicts with animal welfare regulations. Even more important, increased temperature over the entire test period might cause heat stress for the embryos: Although Kimmel et al. [76] did not detect developmental abnormalities at incubation temperatures up to 33 °C in zebrafish, a study by Long et al. [77] demonstrated altered gene expression using an increased temperature of 34 °C. Thus, elevated temperatures in the test create a third stressor in addition to the test compound and the addition of S9 homogenates. By separating a 2 h pre-exposure at 37 °C (without the addition of eggs) from the FET itself (26 °C), such additional stress could be avoided.

Toxicity of S9 homogenates

Both the S9 homogenates and the energy source NADPH showed embryotoxic effects in the FET. Given that effects on zebrafish embryos by liver homogenates have also been documented in literature [43], only short exposure periods of embryos to high concentrations of microsomes (0.7 mg/ml) were performed in previous studies using external metabolism [40,41,42,43]. In range-finding experiments, pre-metabolization with 1 mg/ml S9 homogenates at 37 °C followed by removal of S9 homogenates using (1) a dialysis membrane or (2) heat denaturation and centrifugation to expose the embryos to the metabolites for 120 hpf (details not shown) did not effectively eliminate the embryotoxicity of the S9 homogenates. Consequently, experiments were carried out to optimize protein contents during pre-exposure metabolization: Since the combination of the S9 homogenates (EC10 values: ~ 0.08 mg/ml) and the cofactor NADPH (EC10: 0.47 mM) proved even more toxic than the isolated components (Fig. 1 and Additional file 1: Fig. S1), final concentrations of 0.01 mg/ml (rat S9) or 0.05 mg/ml (ewoS9R) together with 0.5 mM NADPH were in the final experiments.

Species-origin of S9 homogenates

Mammalian metabolization systems have repeatedly been used for the inclusion into the FET [40,41,42]. However, a difference between the metabolization capacities of mammals and fish must be considered: Given that freshly isolated rat hepatocytes have been described to be usually metabolically more competent than fish liver cells [78], rat homogenates were preferred. Hence, the mammalian metabolization within the FET could be described as a conservative and protective approach, but using fish liver homogenates (e.g., according to OECD TG 319B [79]) should also be tested in future studies to strengthen the informative value for ecotoxicology risk assessment (for reference, see [80,81,82]). Given that the FET was designed as an alternative method to conventional animal testing [83] in the context of the 3Rs [11], the primary goal of the present study was the supplementation with a biotechnological product (such as ewoS9R), a product that is derived completely in vitro, and not from animal experiments. To the best of our knowledge, biotechnological approaches using animal-free cultures with fish cells have not been developed so far.

Comparison of FET (with and without metabolism) to AFT data

The comparison of experimental and literature-based FET (without and with pre-metabolization) and AFT data for model substances allyl alcohol, TDCPP, BPA, B[a]P and CP are summarized in Table 1.

Table 1 LC50 values for zebrafish (Danio rerio) exposed to the five model substances (allyl alcohol, tris(1,3-dichloro-2-propyl) phosphate, bisphenol A, benzo[a]pyrene and chlorpyrifos) derived from the fish embryo toxicity test (FET; experimental or literature data) or the acute fish test (AFT, literature data only)
Full size table

If compared to 120 hpf FET EC50 value without pre-metabolization, allyl alcohol proved to be 9–17-fold more toxic in FETs after pre-metabolization with either S9 homogenate. This significant difference can be explained by the presence or absence of alcohol dehydrogenases, which bioactivate allyl alcohol to the more toxic metabolite acrolein [39, 48, 49]. In Danio rerio embryos, alcohol dehydrogenases are present at low levels [39], but are contained in rat S9 homogenates, thus demonstrating the ability to bioactivate allyl alcohol [84, 85]. Elevated concentrations of alcohol dehydrogenases were likely present in the rat S9 homogenate, since despite the lower protein content (0.01 mg/ml), the toxicity of allyl alcohol with rat S9 was higher than after pre-metabolization than with ewoS9R (0.05 mg/ml). This difference might be explained by the origin of the two S9 products: The biotechnologically produced ewoS9R was derived from a rat liver cell line, which had been induced for high CYP1A activity, but has never been optimized for alcohol dehydrogenase contents. In contrast, the rat S9 comprises various cell types induced with a combination of β-naphthoflavone and phenobarbital, which most likely stimulated a broader spectrum of metabolic enzymes including alcohol dehydrogenases.

The strong impact of bioactivation on allyl alcohol and hence the improvement of FET for protective risk assessment becomes clear when comparing FET with AFT data. If the standard FET without metabolization [13] was used as the basis for the risk assessment, the FET LC50 would be 220–1715 times higher than the corresponding AFT LC50. Based on the chemical safety criteria according to ECHA guidance [86], an assessment/safety factor of 1000 would be applied to acute LC50/EC50 fish toxicity data to determine the predicted no-effect concentration (PNEC). Thus, based on FET data without an additional metabolism, the environmental risk for allyl alcohol would clearly be underestimated, if compared to a risk assessment on the basis of the LC50 derived from the AFT (4.8 µM according to Klüver et al. [39]), which would be below or close to the calculated literature (8.2 µM; [39]) or experimental (1.1 µM) PNEC derived from FET data. With a 18–35-fold difference from AFT data, LC50 data from FETs with pre-metabolization provide a much better data base for calculating PNEC values (86.5 nM and 166.6 nM) than conventional FET data.

For TDCPP, a significant bioactivation was found in the FET at 120 hpf after pre-metabolization with rat S9, but not with ewoS9R. TDCPP is metabolized via CYPs 2E1, 2D6, 1A2 and 2C19 [58]. Given that during ewoS9R optimization, the focus was exclusively on CYP1A induction, the formation of the embryotoxic metabolites was likely lower using ewoS9R compared to rat S9, which most likely covers a broader range of enzyme activities. Despite the small differences in the calculated EC50 values, curves differ regarding the hillslope, resulting in larger differences for the EC10 values. Zebrafish embryos with pre-metabolization were slightly more sensitive to TDCPP (LC50 4.8/5.2 µM) than embryos without pre-metabolization (LC50 5.9 µM), which in turn were slightly more sensitive than juvenile zebrafish (LC50 10.6 µM; [87]). Interestingly, after 48 h exposure in the FET, the addition of ewoS9R resulted in a significantly higher than the addition of pre-metabolization with the rat S9; likewise, the hill slopes were different, if compared to the results without metabolization. The stronger effects at 48 hpf might be due to transient increase of toxic metabolites at the beginning of the FET followed by time-dependent degradation during subsequent development [32, 33, 37].

In the case of BPA, significant detoxification was detected following addition of both S9 homogenates. As with TDCPP, the addition of both S9 homogenates led to similar EC50 values compared to the reference without metabolization, but the hillslope of the curve with rat S9 deviated strongly from the reference without metabolization, which led to stronger differences in the EC50 values. A direct comparison to AFT data was not possible, since 50% mortality could not be reached with the concentration range used in this study. However, considering the increase of EC50 values after detoxification, an approximation to zebrafish AFT LC50 can be assumed. In contrast to 120 hpf, significant bioactivation after 48 hpf exposure in the FET was observed after pre-metabolization with ewoS9R (reduced/no pigmentation, edemata, and reduced/no blood flow). Bioactivation in the FET at 48 hpf, but detoxification at 120 hpf might reflect ambiguous data in the literature, which cover both bioactivation [65, 66] and detoxification [67, 68]. As for TDCPP, the transient formation of reactive metabolites followed by detoxification in subsequent stages of development might have triggered such a biphasic reaction of zebrafish metabolites to BPA in the FET. In fact, the presumably rapid development of metabolic capacities in zebrafish embryos from 72/96 hpf should also be considered for BPA metabolization [32, 34,35,36,37, 88]. Furthermore, hatching inhibition as a pronounced effect during later development also reported by Scopel et al. [89] can, of course, only be observed at 120 hpf. Chemical analysis of BPA and potential metabolites might help to clarify the time course of (transient) formation of metabolites via S9 homogenates or the embryos themselves.

To exclude reduced bioavailability due to binding of the test substance to the proteins of the S9 homogenates as a reason for reduced toxicity as previously demonstrated for other substance classes [90], additional experiments were carried out without the energy carrier NADPH, which, however revealed that BPA was detoxified by the S9 homogenates, since the results without the energy carrier did not differ statistically from the reference without S9.

Due to precipitation at concentrations > 10 µM, no concentration-dependent embryotoxicity could be established for B[a]P, which is in line with literature AFT data on B[a]P, which also showed no toxicity. Only Weigt et al. [91] reported a very low LC50 of 5.1 µM B[a]P in the FET. The difference could be related to the usage of 0.5% DMSO as a solvent, which may have improved the chorionic permeability of the compound [88]. B[a]P is bioactivated by the presence of CYP1A1/1A2/1B1, which was detected by the EROD assay in rat S9 and ewoS9R and was also already proven to be bioactivated in in vitro assays by rat S9 and ewoS9R [53].

As for B[a]P, exposure to CP did not result in differences between FETs with and without pre-metabolization, and there was only a minor trend towards bioactivation. This observation, however, is not in line with literature, which report LC50 values of 2.0 µM (AFT, [92]) and 3.7 µM (FET, [93]). In fact, in our experiments, precipitation occurred at concentrations > 100 µM, most likely due to the hydrophobic character of CP (logP of 4.98 [94]). Low bioavailability due to binding of CP to the vessels should have been avoided with the use of glass-coated well plates. In the FET study by Zhang et al. [93], a semi-static test design was applied, and Jeon et al. [92] conducted the AFT with glassware and acetone as solvent. Possibly the slightly different test conditions led to different outcomes.

Conclusions

To summarize, the implementation of an external pre-metabolization step into the FET significantly increased the comparability with the AFT data for allyl alcohol (cf. Table 2). For the other test compounds, this clear conclusion could not be drawn, and observed minor increases/decreases of toxicity as well as putative approximations to AFT data due to the inclusion of pre-metabolization in FET cannot to be rated as biologically relevant. The similarity of AFT and FET data (with or without pre-metabolization) for the biotransformable substances TDCPP, BPA, B[a]P and CP might well be due to comparable biotransformation of the substances by embryonic and juvenile zebrafish, or to a neglectable metabolization rate under the FET/AFT test conditions. Regarding the biological relevance of minor toxicity changes with metabolization, it should be mentioned that not only FET, but—and usually to a greater extent—also AFT data are subject to intra- and inter-laboratory variability [3]. Conventional risk assessment is usually based on AFT data from different species (i.e., not only Danio rerio); thus, inter-species variability must also be considered when comparing and interpreting differences between FET and AFT data [3]. Nevertheless, FET data obtained within one laboratory conducted with and without the implementation of S9 homogenates can indicate if metabolism significantly alters toxicity and, therefore, whether metabolism might be relevant in the context of the tested sample.

Table 2 Overview of toxicity changes via metabolization in the FET with Danio rerio at 120 h post-fertilization (hpf) of the five model substances (allyl alcohol, tris(1,3-dichloro-2-propyl) phosphate, bisphenol A, benzo[a]pyrene and chlorpyrifos) and a possible resulting approximation to AFT data
Full size table

Since a very good correlation between FET and AFT data has repeatedly been documented [14, 18, 22], FETs with pre-metabolization could be applied in the future specifically for selected cases only: (1) the methodology might be applied if existing data or QSAR data indicate relevant biotransformation of the substance. (2) Complex environmental samples could also be tested in FET with and without the addition of S9 homogenates, since for such samples the impact of biotransformation on fish toxicity is extremely difficult to predict.

To further validate the FET with metabolism, other substances with and without metabolism should be tested for which FET and AFT data differ strongly (e.g., albendazole [40]). In addition, other endpoints such as early life stage behavior according to Irons et al. [99] could be investigated with a procedure including pre-metabolization to investigate metabolization in the context of neurotoxic substances. For example, behavioral changes could be investigated with the substance CP as it is metabolized to the neurotoxic metabolite CP-oxone [100].

In addition, to further increase the ecotoxicological relevance of biotransformation processes, (biotechnological) fish S9 should also be examined in the future [79]. In addition, cofactors required for phase II metabolism should be added to the assay in future studies, which could also be relevant for some substances [101]. Here, additional phase II cofactors would also have to be tested for their embryotoxicity, which might affect the concentration of the S9 mixes or NADPH that can be applied. Furthermore, the inclusion of S9 homogenates into the FET could alter the uptake of the substances into the fish embryo cells and the cell excretion via transporters, as the substances properties after metabolization are changed and thus fish embryo cellular transporters could be induced or inhibited via the metabolites [102,103,104,105]. Furthermore, a time-dependent chemical analysis of the metabolites formed by the zebrafish embryos or S9 homogenates might help to gain further insights into the metabolism capacities of the different systems.

Overall, this study showed that the FET with metabolization may be able to compensate for strong differences between FET and AFT data. Using this pre-metabolisation method, the predictive ability of the FET for toxicity to juvenile fish could be improved, if there is evidence of strong biotransformation of the sample. This may reduce or eliminate the need to use juvenile or adult vertebrates for acute ecotoxicological data in future risk assessment.