Using zebrafish in systems toxicology for developmental toxicity testing

Introduction

According to international regulatory guidelines, each drug in development for administration to women of childbearing potential must be tested for developmental toxicity in rodent and non-rodent species (ICH 2005; ICH 2015). Traditional toxicity testing requires data collection on one chemical at a time using common laboratory animal species, such as rat or rabbit (Ema et al. 2012, 2014). Developing a new drug requires a large number of rodents for a teratogenicity safety assessment. This animal-intensive process is not in line with the current trend of applying the 3R principle of humane animal research (replacement, reduction, and refinement) (van der Laan et al. 2012). Substantial efforts have been undertaken to develop alternative methods for the assessment of developmental toxicity, including the mouse embryonic stem cell test, the rat whole embryo culture assay, and zebrafish toxicology testing (de Jong et al. 2011a). None of these assays, however, alone can cover the whole mammalian reproductive cycle, because of its inherent complexity (Kroese et al. 2015). Therefore, recent studies have attempted to combine these alternative methods into a test battery to accurately predict developmental toxicity in humans (Sogorb et al. 2014; Kroese et al. 2015) (reviewed in Sipes et al. 2011). To maximize the power of an integrated systems toxicology approach, each method should be performed using a harmonized procedure.

Zebrafish is currently considered an excellent model organism in various biomedical fields, including developmental toxicology (reviewed in Teraoka et al. 2003; Ota and Kawahara 2013; Nishimura et al. 2015). The development process is highly conserved across vertebrate species, making zebrafish development largely comparable to that of mammalians (reviewed in Sipes et al. 2011; Behl et al. 2015; Nishimura et al. 2015). Zebrafish also allows the evaluation of the complete developmental period of a vertebrate embryo before it becomes fully self-sustainable (approximately 5–6 days postfertilization), and therefore this alternative animal model is not considered a true in vivo system under European law (EU 2010). Given these advantages, Organisation for Economic Co-operation and Development (OECD) developed Test Guideline 236 (TG236) for using zebrafish for acute developmental toxicity testing (OECD 2013). However, TG236 mainly tests for lethality, and only incidentally mentions cataloguing developmental defects. Although several studies have been published on assessing the teratogenicity of chemicals using zebrafish, there remains considerable variation in the testing methods (Beekhuijzen et al. 2015), which can lead to discrepancies among zebrafish toxicity tests using the same chemicals, and reduce the power of integrative analysis.

In this review, we consider zebrafish-based developmental toxicity testing (ZEBDET), while focusing on: (i) exposure conditions; (ii) endpoints measured; (iii) zebrafish housing conditions; and (iv) future directions. Standardizing the testing protocols may be crucial for harmonization of ZEBDET when assessing both lethality and teratogenicity, and may increase the validity of ZEBDET in human risk assessment.

Exposure Time

A major goal of ZEBDET is to complement developmental toxicity testing in mammals. According to ICH S5 (R2), maternal exposure in rat and rabbit begins when the blastocysts undergo gastrulation and ends when the hard palate closes (Thieme et al. 2012). The blastula and gastrula stage of zebrafish at 28°C is equivalent to 2.25–5.25 hours-postfertilization (hpf) and 5.25–10 hpf, respectively (Kimmel et al. 1995). Therefore, zebrafish embryo drug exposure usually begins around 5 hpf, corresponding to late blastula and early gastrula. The closure of the hard palate in rat and rabbit corresponds to the protruding-mouth stage (around 72 hpf) in zebrafish (Kimmel et al. 1995). Most zebrafish organs are fully developed by 96 hpf. The swim bladder is usually inflated by 120 hpf. At this stage, zebrafish can swim freely and feed independently, thus falling into the regulatory frameworks dealing with animal experimentation in Europe (Beekhuijzen et al. 2015; EU 2010). Therefore, ZEBDET is usually performed by 120 hpf.

Selection and Number of Eggs

The quality of zebrafish embryos can be a confounding factor. ZEBDET should be started only with sphere-shaped fertilized eggs without coagulation or bubbles at 4 hpf (Kimmel et al. 1995; Beekhuijzen et al. 2015). Raising healthy female and male zebrafish is also very important for obtaining healthy eggs (discussed in the “Zebrafish housing” section).

OECD TG236 recommends using 20 embryos per concentration to increase intra- and inter-laboratory reproducibility for calculation of LD₅₀ (concentration resulting in mortality of 50%) (OECD 2013). For the assessment of teratogenicity, a large number of embryos per concentration is also preferable to increase the accuracy (Beekhuijzen et al. 2015).

Chorion

A zebrafish embryo's chorion is intact until 2 or 3 days postfertilization (dpf). Pores in the chorion are about 0.5 μm in diameter, and are about 2 μm apart (Rawson et al. 2000; Lee et al. 2007). It is more realistic to think of the chorion as a sieve but not a wall (Schreiber et al. 2009). However, chemical penetration through the chorion can vary, depending on the compound's physicochemical properties (Wiegand et al. 2000; Kim and Tanguay 2014), cationic charge concentration in the aquatic test medium (Cameron and Hunter 1984), and electrostatic attraction between chemicals and the chorion (Burnison et al. 2006). These factors suggest that the chorion can serve as a permeability barrier to chemicals under certain conditions. Panzica-Kelly et al. (2015) demonstrated that the sensitivity to detect teratogens using dechorionated embryos was higher than the sensitivity of chorion-intact embryos (Panzica-Kelly et al. 2015). It has also been demonstrated, however, that the specificity to detect non-teratogens using dechorionated embryos was lower than the specificity obtained from chorion-intact embryos, suggesting that dechorionation at the blastula stage without causing damage to the embryo was challenging (Truong et al. 2014; Panzica-Kelly et al. 2015). Therefore, unless we can develop safe, easy, and accessible methods for dechorionation of zebrafish at the early developmental stages, we think that using dechorionated embryos in ZEBDET may involve trade-offs between sensitivity and specificity. We should also note that there are known teratogens, such as hydroxyurea and ribavirin, whose teratogenicity cannot be detected by ZEBDET, with or without dechorionation (Inoue et al. 2015; Panzica-Kelly et al. 2015). This limitation may be caused by species differences in teratogenic sensitivity, or related to embryo uptake issues associated with zebrafish (Ball et al. 2014; Inoue et al. 2015).

Temperature

Zebrafish can tolerate a temperature range of 24.5 to 28.5°C (Beasley et al. 2012). However, the growth speed of zebrafish embryos varies according to temperature (Kimmel et al. 1995; Vargesson 2007; Howells and Betts 2009) (reviewed in Vargesson 2007; Howells and Betts 2009). When zebrafish embryos are raised at 24.5 to 28.5°C, they reach sphere stage at 4.1 or 3.8 hpf, and shield stage at 5.9 or 5.4 hpf (Beasley et al. 2012). These findings suggest that chemical exposure at a given time-point may have different effects on embryonic development, depending on the ambient temperature. The teratogenic effects of 3,4-dichloroaniline (3,4-DCA) in zebrafish vary between 26 and 28°C (Beekhuijzen et al. 2015). The group found that 53% of zebrafish embryos exposed to 1.5 mg/L 3,4-DCA at 26°C displayed teratogenic effects, compared with only 5% of zebrafish exposed at 28°C (Beekhuijzen et al. 2015). These results suggest that ZEBDET at 26°C may be more sensitive than at 28°C, possibly because a specific developmental stage sensitive to teratogenic effects may take longer at 26°C than 28°C (Beekhuijzen et al. 2015). The other trade-off with different temperatures is the half-life of the chemical; reviewed in (Volz et al. 2011). If the chemical has a long half-life, the rearing temperature may not make as much of a difference, but if it has a short half-life, it may have less developmental time to interact with the developing embryo at the lower temperature. The OECD TG236 (OECD 2013) recommends a temperature of 26 or 26.5°C. Although large-scale ZEBDET at 28 or 28.5°C has been conducted successfully (Gustafson et al. 2012; Ball et al. 2014; Truong et al. 2014), care should be taken when performing comparative analyses of ZEBDET at different temperatures.

Multi-Well Plates and Medium Changes

Because the body length of zebrafish at 5 dpf is around 4 mm, zebrafish larvae can be incubated using a 96-well plate. The advantages of well plates include an increased throughput and reduced amounts of chemicals required in ZEBDET. These advantages depend on the small volume of each well, which may also be a limitation. The medium in each well of a 96-well plate, especially in the outer wells, can evaporate after 4 days of incubation at 28°C (Ali et al. 2011), leading to a change in chemical concentration. Oxygen consumption of zebrafish at 48 hpf is double that the consumption at 24 hpf (Bang et al. 2004). Oxygen consumption of embryos inside the chorion takes place via passive processes, depending on the oxygen gradient (Kranenbarg et al. 2003). Therefore, adequate oxygen concentrations in the surrounding medium are essential for appropriate testing conditions (Lammer et al. 2009). It has been demonstrated that the oxygen concentrations between 2.0 and 3.0 mg/L caused minor developmental retardations and that the oxygen concentrations lower than 0.88 mg/L were 100% lethal (Strecker et al. 2011). OECD TG236 recommends the use of 24-well plates with 2 mL medium per well, because it resolves the question of large concentration changes due to evaporation, while ensuring sufficient oxygen. OECD TG236 also recommends that the medium be changed every day if the chemical degrades by more than 20% after 4 days of static incubation (OECD 2013).

Chemical Concentration

OECD TG236 recommends testing five concentrations of chemicals, ranging from a high concentration that causes 100% lethality to a low concentration that causes no observable effects (OECD 2013). If there are no data available in the literature for the maximal concentration of a chemical in ZEBDET, zebrafish may be exposed to chemical concentration up to the chemical's maximum solubility in the medium. Many chemicals, especially those developed for therapeutic purposes, are hydrophobic and require a solvent (Zuegg and Cooper 2012). Dimethylsulfoxide (DMSO) has been widely used to dissolve these chemicals. Zebrafish exposed to 1% DMSO from 2–4 cell stage to 7 dpf displayed no significant lethality and teratogenicity (Maes et al. 2012), which is consistent with a previous report (Hallare et al. 2006). Uptake of fluorescein into zebrafish embryos with an intact chorion is increased in a medium containing 0.1 and 1% DMSO, compared with 0.01% DMSO (Kais et al. 2013). These results suggest that 0.1 and 1% DMSO may increase the permeability of the zebrafish chorion (Kais et al. 2013). However, the adverse effects of DMSO at low concentrations have also been reported. Zebrafish exposed to 0.01, 0.1 or 1% DMSO from the blastula stage to 7 dpf showed hyperactivity (Chen et al. 2011). The expression of cyp1a mRNA in zebrafish exposed to 0.1% DMSO from 4 to 5 dpf was significantly lower than that of zebrafish exposed to 0.01% DMSO (David et al. 2012). It has also been demonstrated that 0.1% DMSO did not show any effects on the behavior of larval zebrafish (Stewart and Kalueff 2014). Taken together, it is plausible to assume that 0.1% DMSO may be acceptable in ZEBDET. Per OECD TG236 recommendations (OECD 2013); however, DMSO concentrations should be kept to the lowest possible level, especially in longer term tests or when CYP-mediated xenobiotic metabolism is required (David et al. 2012).

In zebrafish, the mouth opens around 3 dpf (Wallace and Pack 2003), and the gills gain functionality around 14 dpf (Rombough 2002; Jonz and Nurse 2005). Therefore, the main route of exposure until 3 dpf is likely to be diffusion through the skin (Diekmann and Hill 2013). Diffusion may be regulated by hydrophobicity of the chemical (Padilla et al. 2012), the chemical class (Padilla et al. 2012), the bioconcentration factor (Scholz et al. 2008), and other physicochemical properties (Diekmann and Hill 2013). For example, the EC₅₀ (concentration which induces a half-maximal response) of two anti-epileptic drugs, valproic acid (VPA) and levetiracetam, for teratogenicity in zebrafish, is 165 μM and 109 mM, respectively, suggesting that VPA may be 660 times more toxic than levetiracetam (Beker van Woudenberg et al. 2014). However, the body burden of VPA and levetiracetam at the EC₅₀ level was 0.087 nM/larva and 1.286 nM/larva, respectively, reducing the difference in teratogenic potency from 660 to 15, possibly because the uptake of levetiracetam in zebrafish was 45 times lower than that of VPA (Beker van Woudenberg et al. 2014). Therefore, if possible, the total body burden of a chemical should be assessed, to correlate the toxic phenotype with the actual absorbed concentration of the chemical.

Ear

The otic placode, from which the ear develops, first becomes visible around 14 hpf on both sides of the hindbrain, approximately midway between the eye and the first somite, then forms the lumen of the otic vesicle (Malicki et al. 1996). Two otoliths appear in the otic vesicle at approximately 19 hpf and increase in size (Malicki et al. 1996). Formation of semicircular canals is initiated around 44 hpf and completed by 64 hpf (Malicki et al. 1996).

Hypothyroidism during the first few post-natal weeks, or prenatal exposure to anti-epileptic drugs, may cause malformation of the ear in mammals (Crofton et al. 2000; Van Houtte et al. 2014) (reviewed in Bartel-Friedrich and Wulke 2007). Thyroid hormones are necessary for normal development of cochlear structures, but ear malformation caused by hypothyroidism can be rescued by T4 replacement therapy (Crofton et al. 2000). Consistent with these reports, anti-thyroid drugs (such as methimazole and propylthiouracil) and anti-epileptic drugs (such as carbamazepine, trimethadione, topiramate, and VPA) have been shown to cause malformation of otoliths in ZEBDET (Weigt et al. 2011; Selderslaghs et al. 2012; Inoue et al. 2015). These findings suggest that prenatal exposure to anti-thyroid drugs may cause ear malformation through hypothyroidism not only in mammals, but also in zebrafish. The mechanism underlying ear malformation induced by anti-epileptic drugs remains to be elucidated.

Eye

When the optic vesicle is developed around 12 hpf, zebrafish eyes become visible as flat structures. When the anterior chamber is formed around 24 hpf, zebrafish eyes become sphere-like. The ganglion cells are differentiated and the axons reach the optic tectum around 40 hpf. The photoreceptor outer segment appears around 55 hpf, and extraocular eye muscles become functional around 70 hpf (reviewed in Neuhauss 2010).

The anthelmintic drug albendazole caused microphthalmia in both zebrafish (Mattsson et al. 2012) and rat (Mantovani et al. 1995; Longo et al. 2013). Sipes et al. (2011) performed an integrative systems toxicology assessment to reveal the molecular mechanism underlying microphthalmia. Evaluating the list of ToxCast phase I chemicals, they identified chemicals that caused microphthalmia in zebrafish, rat and rabbit, and analyzed ToxCast in vivo assays of these teratogens. The study concluded that inflammatory signals, extracellular matrix remodeling, axonal guidance, the transporter ABCB1, and mitotic arrest were significantly correlated in these teratogens. Albendazole binds to β-tubulin, inhibits the polymerization to microtubule, and causes mitotic arrest (Lacey 1990; Feng et al. 2010). These findings suggest that prenatal exposure to albendazole may cause microphthalmia through mitotic arrest in zebrafish and rat.

Jaw

Genetic mechanisms controlling neural crest-derived craniofacial development, including lower jaw formation, are known to be conserved across many vertebrates (reviewed in Schilling 1997). In zebrafish, chondrocytes proliferate in the lower jaw from 2 dpf, and form doublets about midway along the anterior-posterior length of the lower jaw by 6 dpf (Eames et al. 2013).

In humans, anti-epileptic drugs taken during pregnancy can increase the risk of congenital jaw malformation (Koo and Zavras 2013). VPA and topiramate have been shown to cause a hypomorphic lower jaw (Yamashita et al. 2014; Inoue et al. 2015). Many anti-epileptic drugs can also inhibit histone deacetylase (HDAC) (Eyal et al. 2004); the protein HDAC1 regulates neural crest-derived craniofacial development (Ignatius et al. 2013). These findings suggest that prenatal exposure to anti-epileptic drugs may cause neural crest-derived craniofacial malformation through inhibiting HDAC. Amlodipine, a dihydropyridine that blocks L-type calcium channels, caused hypomorphic lower jaw in ZEBDET (Inoue et al. 2015). Although the teratogenic effects of dihydropyridine on craniofacial development have not been reported in mammals, a recent study reported that nisoldipine, another dihydropyridine-class L-type calcium channel blocker, administered prenatally, inhibited the growth of chondrocytes and reduced the size of the lower jaw in mice, suggesting that calcium influx through L-type calcium channels may regulate mandibular development in zebrafish and mouse models (Ramachandran et al. 2013).

Heart

In zebrafish, the heart begins to beat at around 22 hpf, and is partitioned into the sinus venosus, atrium, ventricle, and bulbous arteriosus (Stainier et al. 1993; Hu et al. 2000).

Valproic acid exposure induced cardiac malformations in both human and rat (Binkerd et al. 1988) (reviewed in Alsdorf and Wyszynski 2005). VPA also induced cardiac malformation and pericardial edema in ZEBDET (Pruvot et al. 2012; Inoue et al. 2015). In both human and mouse embryonic stem cells, VPA decreased the expression of nkx2.5, a marker of cardiomyocyte differentiation (de Jong et al. 2011b; Krug et al. 2013), although the mechanism underlying decreased expression of nkx2.5 is still unknown. Diclofenac, a nonsteroidal anti-inflammatory drug, has been shown to cause cardiac malformation and pericardial edema in ZEBDET (Chen et al. 2014; Inoue et al. 2015). The expression of nkx2.5 was also decreased in the zebrafish after diclofenac exposure (Chen et al. 2014). Although diclofenac has been suggested as teratogen in human and rat (Siu et al. 2000; Chan et al. 2001), there are no reports of cardiac teratogenicity associated with diclofenac. This discrepancy remains to be explained.

Somite

The overall process of somite development in zebrafish is similar to that in mammals (reviewed in Stickney et al. 2000). In zebrafish, the first somites appear around 10 hpf. Additional somites are produced at 30-min intervals in a bilaterally symmetric, anterior to posterior wave until a total of about 30 somite pairs bracket the notochord at 24 hpf. Somites give rise to the axial skeleton and the skeletal muscle of the trunk. Muscle fibers begin to form shortly after the somite formation.

Children exposed to alcohol in utero often exhibit deficits such as motor and reflex development (reviewed in de la Monte and Kril 2014). The malformation of skeletal muscle caused by exposure to alcohol during the embryonic period are also observed in zebrafish (Sylvain et al. 2010), chick (Chaudhuri 2004) and rat (Nwaogu and Ihemelandu 1999). It has also been demonstrated that alcohol exposure during the embryonic period disrupts the expression of sonic hedgehog, an important regulator for the somite formation (reviewed in Christ et al. 1998), in zebrafish (Loucks and Ahlgren 2009; Zhang et al. 2014) and chick (Ahlgren et al. 2002). Sonic hedgehog signaling was also identified as one of the key pathways in fetal alcohol syndrome in human (Lombard et al. 2007). These findings suggest that exposure to alcohol during the embryonic period may cause the malformation of skeletal muscle through disruption of sonic hedgehog signaling in zebrafish and mammals.

Feeding

Zebrafish attain sexual maturity within 3–6 months postfertilization (mpf) in laboratory settings, although this may vary considerably with environmental conditions, including temperature, water quality, rearing density, and nutrient factors (reviewed in Westerfield 2007; Lawrence and Mason 2012; Wilson 2012).

The specific nutritional requirements of zebrafish, especially at the first feeding stages of larvae, are yet to be determined (reviewed in Lawrence 2007; Watts et al. 2012). Although artificial diets can be used as a sole food source, the rates of larval survival and growth are markedly reduced when compared with zebrafish fed live diets (such as Paramecium and Artemia) (Goolish et al. 1999). Feeding zebrafish Paramecium alone during the early larval stages can result in smaller larvae and lower survival rates, when compared with an Artemia-only diet (Carvalho et al. 2006). However, Artemia (around 500 μm) are larger than the mouth of first-feeding zebrafish. Therefore, Paramecium (150–200 μm) and/or Brachionus (∼250 μm) are used for first-feeding zebrafish until they are able to feed on Artemia, usually around 10 dpf (reviewed in Lawrence 2007; Wilson 2012). These live diets are visually and chemically attractive to zebrafish, digestible, and distributed more evenly in the water column than artificial diets (Cahu and Zambonino Infante 2001).

Water Quality

Depending on laboratory infrastructure, the water chemistry can vary. Therefore, it is essential to have an accurate understanding of the water's source and chemistry prior to installing the housing system (reviewed in Lawrence and Mason 2012). If the reverse osmosis system, which purifies water removes not only harmful but also beneficial substances such as calcium and magnesium, the purified water must be conditioned by the addition of synthetic sea salts before it is delivered to the fish (reviewed in Lawrence and Mason 2012).

There are two types of zebrafish housing systems (reviewed in Lawrence and Mason 2012). The first is a flow-through system in which clean water is pumped into tanks and the effluent, into which zebrafish excrete wastes, is flushed out. The second is a recirculating system in which the effluent water is pumped into a treatment zone and the wastes removed. The water is then pumped back into the tank. Because the recirculating system reduces water usage and space requirements, many laboratories have adopted this system.

The recirculating system performs solids removal, biological filtration, aeration, and disinfection (reviewed in Lawrence and Mason 2012). Removal of solid wastes is usually performed by passing the water through filter pads, rotating microscreen drum filters, or expandable granular media filters. The water is then filtered to oxidize ammonia nitrogen into nitrite and then nitrate by nitrifying bacteria. Nitrates can be controlled through removal by regular water changes. Because the efficiency of nitrifying bacteria is highly dependent on alkalinity and dissolved oxygen, the alkalinity of the water is maintained by the addition of sodium bicarbonate or calcium carbonate, and the biological filter zone is highly oxygenated. Dissolved oxygen should not be lower than 4 mg/L, and carbon dioxide should be kept at levels less than 20 mg/L.

All recirculating aquaculture systems contain populations of numerous microbes. Although many of these organisms are benign, some may be pathogenic. Ultraviolet sterilizers are most frequently used for disinfection of the recirculating water. Important factors in the effectiveness of ultraviolet sterilizers include intensity, transmittance, and contact time. These factors should be checked regularly to ensure efficient disinfection.

Breeding

No clear sex-determining chromosome has been found in zebrafish to date (reviewed in Nasiadka and Clark 2012). All zebrafish initially have undifferentiated ovary-like gonads, but by 30 dpf, all oocytes disappear from male gonads, and males undergo testicular differentiation (Takahashi 1977). High stocking densities and limited food resources appear to produce more males than females, while lower stocking densities and high food availability produce more females (reviewed in Wilson 2012). Other factors, such as hypoxia and temperature, may also affect sex ratios in zebrafish populations (Uchida et al. 2004; Orban et al. 2009) (reviewed in Orban et al. 2009). Laboratory zebrafish typically attain sexual maturity at 3–4 mpf (reviewed in Nasiadka and Clark 2012). Females have a large, light silver belly that protrudes from the body; their blue stripes alternate with silver stripes, and their dorsal fin displays a stronger yellow hue. Males typically lack the protruding belly and have a streamlined shape, displaying a reddish-gold hue between blue stripes, particularly in the anal and caudal fins (reviewed in Nasiadka and Clark 2012). Although sexually mature zebrafish can spawn at a frequency of two or three times a week (Eaton and Farley 1974), overspawning decreases the quantity and quality of eggs (reviewed in Nasiadka and Clark 2012). An optimal breeding frequency for zebrafish seems to be every 10 days (Niimi and LaHam 1974).

Mating is initiated at the onset of light, and spawning takes place over a short period thereafter (reviewed in Matthews et al. 2002; Nasiadka and Clark 2012). Male gonad pheromones stimulate ovulation in females (van den Hurk and Resink 1992). After ovulation, females release gonad pheromones to attract and stimulate males to perform courtship, which consists of abrupt turns and an elliptical swimming pattern around the female (van den Hurk and Lambert 1983). Both single-pair matings and group crosses can be used (reviewed in Westerfield 2007), but population density and sex ratio in group crosses have a significant impact on reproductive success (Ramsay et al. 2009) (reviewed in Nasiadka and Clark 2012). For example, courtship behavior of males increases in dense populations with a female-biased sex ratio (Spence and Smith 2006). Females tend to spawn smaller clutches at higher densities, possibly because of male territorial behavior, which interferes negatively with female ovipositions. Some pheromones released by females may also inhibit spawning of other females, playing a crucial role in competitive interactions between females (Gerlach 2006). Crossing cages are available in different sizes. However, egg production is decreased in breeding volumes of 200 mL or less (Goolish et al. 1998). Imitation plastic plants or green mesh are often placed in the inner tank of a crossing cage to provide artificial spawning sites (reviewed in Nasiadka and Clark 2012).

Because decreased fertilities and viabilities are often associated with inbreeding depression (Mrakovcic and Haley 1979; Monson and Sadler 2010), sustaining robust zebrafish strains may require maintaining the genetic heterogeneity of these strains, which may interfere with reproducibility of experiments. Using hybrid females of known backgrounds for maintaining the genetic heterogeneity of the strains is therefore recommended (Monson and Sadler 2010) (reviewed in Nasiadka and Clark 2012).

Machine Learning for Objective Morphological Assessment

Jeanray et al. (2015) showed that the scoring of morphology is the main rate- and quality-limiting step in ZEBDET, and may be prone to subjectivity. The large number of substances to be tested and the need for accurate results call for methods allowing automation of data acquisition, as well as identification of defects and classification of the acquired images. Machine learning can be used to classify various developmental toxicities, including curved tail and coagulation in the yolk sac. Other morphological changes, such as a curved trunk and hemostasis, can also be detected by machine learning, albeit less accurately (Jeanray et al. 2015). Because novel algorithms of machine learning are actively developed, it is likely that various morphological abnormalities in ZEBDET could be assessed automatically and objectively using machine learning (Mikut et al. 2013).

Using Transgenic and Genome-Edited Zebrafish to Improve ZEBDET

Developmental toxicity can be caused by multiple mechanisms. However, certain key genes, either marker or causative genes, are known to be involved in the developmental toxicity of various chemicals (reviewed in Piersma et al. 2014). Therefore, transgenic zebrafish, in which fluorescent proteins are expressed under the control of these key genes, can be a powerful tool in ZEBDET to detect developmental toxicities of chemicals with a high degree of sensitivity (reviewed in Dai et al. 2014). This approach also enables us to detect tissue-specific developmental toxicity, such as the induction of green fluorescent protein (GFP) driven by the promoter of hsp70, a marker for oxidative stress response during development, in the eye of transgenic zebrafish embryos, following exposure to tert-butylhydroquinone (Hahn et al. 2014).

Genome-edited zebrafish in which genes involved in developmental toxicity are modified can also be used in ZEBDET. Gene polymorphism in paraoxonase 1 has been shown as a possible determinant of vulnerability to the neurotoxic effects of organophosphorus compounds, including developmental toxicity, in both human and animal studies (reviewed in Costa et al. 2013). New reverse genetic techniques, such as TAL effector nuclease (TALEN) and clustered regularly interspaced short palindromic repeats-Cas9 system (CRISPR/Cas9), can be used to create double-strand DNA breaks at specific sites in the zebrafish genome (reviewed in Ota and Kawahara 2013), and to knock-in human genes into the sites. The genome-edited zebrafish that have sequence variants parallel to humans' may be an ideal model for prediction of various risks in humans, including developmental toxicity of chemicals.

A Systems Toxicology Approach for Human Risk Assessment

A fundamental building block of systems toxicology is the consensus that exposure lead to changes at the molecular level, some of which may induce morphological or functional changes at the cellular and organism level, contributing to toxic outcomes (reviewed in Sturla et al. 2014). Technological developments have enabled us to observe these changes in various species, including zebrafish.

Transcriptomics, or gene-expression profiling, is perhaps the most widely used measurement method in systems toxicology. Transcriptomic analysis is also the best established approach for identifying perturbed biological networks and thereby gaining mechanistic insight into the system's response to an exposure (reviewed in Sturla et al. 2014). Hermsen et al. (2013) performed transcriptome analysis of zebrafish exposed to four developmental toxic compounds – caffeine, carbamazepine, retinoic acid, and VPA. They were able to identify both commonly and selectively expressed genes induced by the four developmental toxicants, suggesting that there may be common and selective modes of action in these toxicants. The use of individual gene expression signatures, as well as pathways enriched in the differentially expressed genes, may be useful starting points for defining gene biomarkers for predicting developmental toxicity (Hermsen et al. 2013).

Image-based high content assays for toxicity screening using transgenic zebrafish embryos with morphology and behavioral endpoints are also gaining popularity. Lantz-McPeak et al. (2015) performed high content screening of zebrafish embryos to examine the developmental toxicity of ethanol, nicotine, ketamine, and caffeine. They used Hb9:GFP transgenic zebrafish embryos in which motor neurons and their axons are labeled with strong neuron-specific expression of GFP under the control of the regulatory elements of the zebrafish Hb9 gene. Using the fluorescence, they quantified the endpoints (body length and fiber length) as measures of teratogenicity (growth retardation), and were able to detect the developmental toxicity of these toxicants. The automated developmental toxicity assay can be performed in a high-throughput manner. Therefore, this approach can significantly expedite the screening of potential teratogens and developmental toxicants, improving the current risk assessment process by decreasing analysis time and required resources (Lantz-McPeak et al. 2015).

Systems toxicology is the integration of traditional and advanced toxicological assays to elucidate the molecular mechanisms and functional changes induced by exposure to a xenobiotic compound (reviewed in Sturla et al. 2014). Systems toxicology develops mathematical models to predict developmental toxicity in humans. Zebrafish is an excellent model in the systems toxicological approach for developmental toxicity testing.

Disclosure

None declared.