Zebrafish larvae as a New Alternative Methodology (NAM) for the assessment of compounds’ teratogenicity

Combining a zebrafish-based high throughput platform with artificial intelligence for human developmental toxicity assessment.

By Carles Cornet, Head of Toxicology at ZeClinics.

Carles Cornet is an expert in zebrafish biology and drug discovery. He did his master’s degree in biomedicine straddling between the University of Barcelona and the German Cancer Research Center in Heidelberg. Carles joined ZeClinics team in 2015 as a PhD student, where he has optimized the use of the zebrafish model as a novel tool to assess drug safety and antitumoral efficacy. Since January 2020, year in which he obtained his PhD in Biomedicine from the Pompeu Fabra University in Barcelona, He is currently in charge of the toxicology area as well as of the CNS behavioral efficacy area.

We have previously introduced in our blog several global initiatives for zebrafish model validation in which ZeClinics is participating, such as the DNT, the IATA, and the SEAZIT. These initiatives have arisen in response to:

  1. The problem of the low acceptance rate of drugs entering clinical trials, which suggests the information gathered during preclinical phases, specifically that obtained from animal models, provides a lower-than-expected prediction of toxicity and therapeutic effects in human patients.  
  2. The pressing need from regulatory agencies to reduce mammal experimentation by 2025 and complete abolishment by 2035.

With this in mind, at ZeClinics we are using the zebrafish larvae model also for the assessment of compounds' teratogenicity [1].

DevTox - Developmental Toxicology

3Rs-aligned method for high throughput teratology assessment in vivo.

The importance of early teratogenic risk identification

Teratogenesis is the process that disrupts the normal development of an embryo or fetus by causing permanent structural and functional abnormalities, growth retardation, or even lifelong mental disabilities [2]. It can be caused by exposure to certain chemicals (teratogens), found also naturally, which interfere with embryonic development in multiple ways. As humans are in contact with a huge variety of chemicals already from the zygote stage, continuing the exposure during embryo and fetus development, this threat is of crucial importance. Hence, to avoid the health risks associated with teratogens, their identification in the early phases of pharmaceutical drug and industrial chemical development has become a priority.

Highlighting the importance given to chemical hazard assessment is the implementation in 2007 of the REACH (Registration, Evaluation, Authorization and Restriction of Chemicals) regulatory framework in the European Union. This regulation is aimed at the protection of human health and the environment from risks associated with chemical exposure [3]. In addition, and in line with the objective to reduce animal experimentation, one important aspect of this action is the use of new approach methodologies (NAMS).

A paradigm shift in teratogenicity assessment

Traditionally, in vivo teratogenicity studies have relied on mammalian models such as adult rodents and rabbits. These models suffer from elevated experimental costs and low throughput and normally require invasive methods. Their replacement by biomarkers, in vitro models such as cell and tissue cultures, and whole organisms models such as zebrafish larvae or chick embryos can increase throughputs while decreasing costs. However, there are still few biomarkers for teratogenicity properly validated.

On the other hand, in vitro teratogenicity studies have shown significant improvements in terms of predictivity, reaching accuracies of around 90 % for some validation studies [4,5]. But its use is still challenging due to its reductionist nature, which obviates complex biological processes present in intact organisms, potentially reducing predictivity compared to whole-organism models.

Teratogenicity studies involving whole organisms have relied on intact invertebrates such as drosophila [6] or planarians [11], as well as lower vertebrate embryos including amphibians [7], fish [1,8], chick embryos [9], and cultured mammalian embryos [10]. Among them, the zebrafish embryo model has emerged as a particularly promising choice as, from a translatability point of view, they have shown to be more sensitive, accurate, and representative of human biology [1,11].

What makes zebrafish an interesting model for developmental toxicity?

Zebrafish has been extensively studied, described, and used as model organism in ecotoxicology to assess the effects of chemicals and their risk to the environment. One of the zebrafish characteristics, accelerated development, allows the study of molecular mechanisms resembling those in mammals much earlier. For instance, molecular mechanisms controlling organogenesis can be assessed already at 120 hours post fertilization (hpf), when this process is completed.

In fact, all digestive and metabolism-related organs, including the liver, are functional from 72 hpf and physiologically mature by 120 hpf.  The onset of glomerular filtration is estimated at 48 hpf. The thyroid gland is fully functional between 70–80 hpf. Key neuronal transcripts are present as early as 36 hpf, and neurogenesis occurs from 36 to 120 hpf, when embryos are already capable of displaying free swimming, rudimentary learning and memory features, and response to a variety of stimuli [1].

In addition to its rapid development, the zebrafish embryo’s optical transparency allows for the easy evaluation of developmental morphological endpoints. This rapid development and optical transparency, together with their small size, external fertilization, and large progenies, allows for the use of zebrafish embryos, instead of adults, for the quick screening of large batteries of molecules. Finally, it is important to highlight that zebrafish at larval stages is considered a non-animal method. Hence, experiments performed at this developmental stage are aligned with the 3Rs principles.

Leveraging zebrafish and artificial intelligence combined potential for teratogenicity screenings

As already introduced at the beginning of this blog, at ZeClinics we are leveraging the unique biological features zebrafish larvae possess to develop a NAM for the early identification of compounds teratogenicity. This NAM harbors promising applicability during both pharmaceutical and industrial chemical development.

Specifically, we have generated an automated high-throughput screening platform for teratogenic drugs. The platform was validated by testing on zebrafish larvae a library of 31 compounds previously categorized as teratogens and non-teratogens in mammals, and by assessing 16 phenotypical parameters related to embryo development.

Classification of compounds according to the results obtained in this platform was compared with existing scientific evidence of teratogenicity in rodents and humans, in order to calculate the teratogenic predictivity potential of zebrafish larvae.

As shown in the following figure, results obtained using this platform show a high correlation with results obtained in murine models, with very high sensitivity, specificity, and accuracy. More importantly, zebrafish results show a high correlation with humans, even increasing the prediction level reported in rodents for the same compounds (higher sensitivity, specificity, and accuracy) [1].

Since then, we have used the information obtained using this validated platform to train a deep learning algorithm, which is currently able to discriminate all the defined phenotypic parameters, sort larvae as positive or negative for qualitative phenotypes, and extract values from quantitative ones. Therefore, it demonstrates that combining the experimental advantages of the zebrafish larval model with artificial intelligence allows for high-throughput, fully automated detection of compound teratogenicity.

Our hope is that the usage of this or similar platforms could help in the establishment of a faster and more reliable human teratogenic risk assessment based on NAMs. Furthermore, such type of initiatives demonstrates that zebrafish can complement if not replace mammalian testing while being equally protective of human health. Finally, it reinforces the evidence for the zebrafish to be present not only within the regulated safety studies in the chemical industry but also in the pharmaceutical preclinical regulatory phases.


  • The zebrafish embryos serve as a system alternative to animal testing which, however, preserves the complex biology of a whole organism.
  • The predictive teratogenic potential of zebrafish is similar, when not superior, to that obtained using traditional rodent models.
  • The use in combination of NAMs such as zebrafish larvae and artificial intelligence can help in increasing the prediction of toxicity and therapeutic effects in human patients while reducing experimental times, costs, subjectivity, and animal experimentation.


[1] Jarque S, Rubio-Brotons M, Ibarra J, Ordoñez V, Dyballa S, Miñana R, Terriente J. Morphometric analysis of developing zebrafish embryos allows predicting teratogenicity modes of action in higher vertebrates. Reprod Toxicol. 2020 Sep;96:337-348. doi: 10.1016/j.reprotox.2020.08.004.

[2] Gilbert-Barness E. Teratogenic causes of malformations. Ann Clin Lab Sci. 2010 Spring;40(2):99-114.

[3] OJL396, Regulation (EC) No 1907/2006 of the European Parliament and of the Council of 18 December 2006 Concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH), Establishing a European Chemicals Agency, Amending Directive 1999/4, 2006. PMID: 20421621

[4] Kameoka S, Babiarz J, Kolaja K, Chiao E. A high-throughput screen for teratogens using human pluripotent stem cells. Toxicol Sci. 2014 Jan;137(1):76-90. doi: 10.1093/toxsci/kft239.

[5] Paradis FH, Huang C, Hales BF. The murine limb bud in culture as an in vitro teratogenicity test system. Methods Mol Biol. 2012;889:197-213. doi: 10.1007/978-1-61779-867-2_12. 

[6] Schuler RL, Hardin BD, Niemeier RW. Drosophila as a tool for the rapid assessment of chemicals for teratogenicity. Teratog Carcinog Mutagen. 1982;2(3-4):293-301. doi: 10.1002/1520-6866(1990)2:3/4<293::aid-tcm1770020310>3.0.co;2-w.

[7] Dawson DA, Fort DJ, Newell DL, Bantle JA. Developmental toxicity testing with FETAX: evaluation of five compounds. Drug Chem Toxicol. 1989 Mar;12(1):67-75. doi: 10.3109/01480548908999144.

[8] Orrego R, Guchardi J, Beyger L, Krause R, Holdway D. Comparative embryotoxicity of pulp mill extracts in rainbow trout (Oncorhynchus mykiss), American flagfish (Jordanella floridae) and Japanese medaka (Oryzias latipes). Aquat Toxicol. 2011 Aug;104(3-4):299-307. doi: 10.1016/j.aquatox.2011.04.015.

[9] Kucera P, Burnand MB. Routine teratogenicity test that uses chick embryos in vitro. Teratog Carcinog Mutagen. 1987;7(5):427-47. doi: 10.1002/tcm.1770070502.

[10] Kitchin KT, Schmid BP, Sanyal MK. Rodent whole-embryo culture as a teratogen screening method. Methods Find Exp Clin Pharmacol. 1986 May;8(5):291-301. PMID: 3724303.

[11] Hagstrom D, Truong L, Zhang S, Tanguay R, Collins ES. Comparative Analysis of Zebrafish and Planarian Model Systems for Developmental Neurotoxicity Screens Using an 87-Compound Library. Toxicol Sci. 2019 Jan 1;167(1):15-25. doi: 10.1093/toxsci/kfy180. 

3Rsalternative modelsanimal experimentationanimals in scientific researchartificial intelligenceChemical toxicitydevelopmental toxicologyecotoxicologyenvironmental chemicalsenvironmental toxicologyHigh Throughput ScreeningNAMspredictive toxicologypredictivityREACHregulatory complianceteratogenicityvalidityZebrafishzebrafish embryos