Short-term fish toxicity testing is one of the oldest and most established components of chemical hazard assessment. The result of a standard acute test, the median lethal concentration (LC₅₀), feeds directly into aquatic hazard classification, environmental risk assessment and regulatory decisions under frameworks such as REACH. For decades it has been treated as an indispensable data point. It also comes at a real cost.

Conventional fish tests raise animal welfare concerns, particularly because the endpoint in the standard acute test is mortality, so the procedure is associated with a high degree of severity. They also require substantial numbers of animals. Knight et al. (2026) estimated that around 66,000 fish would be needed to satisfy completed, ongoing and pending short-term fish toxicity tests under REACH. Behind every one of those numbers is an animal, and behind the total is a regulatory system that keeps generating demand for the same type of data. The question is no longer only whether we can reduce this number, but how fast and on what scientific basis.

Why the gill?

The gill is a rational place to start when looking for an in vitro surrogate for whole fish toxicity. Gills are in constant, direct contact with the surrounding water. They are the main interface through which waterborne chemicals enter the body, and they are also a primary target of toxic damage for many substances. For chemicals that act through general mechanisms such as narcosis (baseline toxicity), membrane disruption or non-specific cytotoxicity, damage to gill cells is a plausible proxy for effects on the whole organism.

The RTgill-W1 assay in brief

The RTgill-W1 fish gill cell line assay was adopted as OECD Test Guideline 249 in 2021. It uses a continuous cell line derived from the gills of rainbow trout (Oncorhynchus mykiss). Cells are exposed to a chemical across a concentration range, and cell viability is then measured. The result is an effect concentration (EC₅₀), which can be used to predict the acute toxicity that would be observed in fish (the LC₅₀) without testing live animals.

The guideline measures viability with several fluorescent indicators that capture different aspects of cell health, including metabolic activity, cell membrane integrity and lysosomal integrity. Using multiple readouts helps because chemicals can damage cells in different ways, and a single indicator may miss an effect that another one detects. The format is also well suited to routine testing. Cells are exposed in multiwell plates over a short period, which makes the assay faster and cheaper than a conventional fish test and amenable to testing many concentrations and chemicals in parallel.

The OECD guideline recognises that the assay has the potential to significantly reduce the number of fish used in chemical hazard assessment, particularly in an area that to date relies on a high degree of severity in the acute fish toxicity test. This is an important statement. It places the method not only as a scientific alternative but also as a contributor to the 3Rs principle of replacement, reduction and refinement.

How reliable is it?

The evidence base for RTgill-W1 is among the strongest of any fish cell line approach. Tanneberger et al. (2013) compared EC₅₀ values from the gill cell assay with in vivo fish LC₅₀ values for a range of chemicals and reported good agreement. That study was an important proof of concept that a simple cell based system could capture much of the variation in acute fish toxicity across different substances.

A method intended for regulatory use needs more than good correlation in a single laboratory, however. It also needs to be shown that different laboratories can run it and obtain consistent results. Fischer et al. (2019) addressed this by evaluating the repeatability and reproducibility of the assay, which is a key part of the validation process that led towards international acceptance. Together, these lines of evidence were the foundation for the adoption of TG 249.

It is worth being precise about what “good agreement” means, though. Predictive performance varies with chemical properties and mode of action. Agreement tends to be strongest for chemicals whose acute toxicity is driven by general cytotoxic mechanisms, and it is more variable for substances with specific modes of action, or whose behaviour in the test system is complicated by their physicochemical properties. A method can be well validated overall and still have a defined domain in which its predictions are more or less trustworthy. Recognising that domain is not a weakness of the assay. It is part of using it responsibly.

What are the limitations?

No single in vitro system reproduces a whole animal, and RTgill-W1 is no exception. The assay does not capture all toxicity mechanisms equally well.

Specific modes of action. Some neurotoxic chemicals act on specific receptors or ion channels, and these targets may not be present, or may not be functionally equivalent, in gill cells. Such chemicals can be far more toxic to a whole fish than to cultured cells, in which case the assay may underpredict acute toxicity. For substances where neurotoxicity is the expected driver of lethality, a cell viability readout alone is unlikely to be sufficient.

Biotransformation. Differences in metabolic capacity between a cell line and an intact fish can also affect predictions. If a chemical needs to be metabolically activated to form a more toxic metabolite, a cell line with limited metabolic competence may not produce that metabolite, and the toxicity of the parent compound may be underestimated. The opposite can occur as well, since in a whole organism detoxification may reduce toxicity relative to what is seen in cells.

Exposure and test system considerations. In any cell based assay, the concentration the cells actually experience may differ from the nominal concentration, especially for volatile, hydrophobic or otherwise difficult substances that adsorb to plastic or are lost during the exposure. This matters for interpretation and is one reason why the guideline and the supporting literature stress careful attention to test chemical properties.

Because of these constraints, the applicability of the assay should always be evaluated in the context of the intended regulatory use. A prediction that is fit for one purpose, such as screening or prioritisation, may need additional supporting evidence for another, such as a definitive classification decision.

Could integration be the answer?

Most likely yes. The most realistic route to reducing live fish testing is not to expect one alternative method to replace a fish test on its own, but to combine information from several sources. This is the logic behind an Integrated Approach to Testing and Assessment (IATA) for acute aquatic toxicity.

Within an IATA, RTgill-W1 results can be considered alongside complementary lines of evidence, including:

• QSAR predictions, which can flag chemicals expected to act through baseline toxicity and those that might not
• Data from tests with algae and daphnia, which add information from other trophic levels and sensitivities
• Fish embryo toxicity (FET) test results, which capture effects in a developing whole organism

These can be combined in a weight of evidence framework, where each method’s strengths compensate for the blind spots of another. For instance, a cell assay that cannot detect a receptor mediated neurotoxic effect may be complemented by embryo data or by a mode of action based alert from QSAR. Conversely, consistent results across several independent methods increase confidence in a prediction considerably.

Structured approaches to evidence integration are an active area of work. Lillicrap et al. (2020), for example, evaluated a Bayesian network to strengthen the weight of evidence for predicting acute fish toxicity from fish embryo toxicity data. Approaches of this kind point towards a future in which integration is more transparent, more quantitative and more reproducible than expert judgement alone, which is exactly what regulators need in order to rely on non animal evidence.

So, are we ready?

For many chemicals, RTgill-W1 is already a credible way to reduce or avoid live fish testing, and its status as an OECD Test Guideline is a genuine milestone for the field. The remaining work is less about proving that the assay works in general and more about three practical tasks:

1. Defining the applicability domain: for which chemical classes and modes of action are predictions reliable, and where are they not?
2. Developing clear, shared rules for how to combine RTgill-W1 data with other evidence in an IATA, so that conclusions are consistent between assessors and between regulatory contexts.
3. Building regulatory confidence through experience, case studies and transparent reporting of when alternative data were sufficient and when they were not.

The most useful question is therefore not “Can we replace the fish test?” but “For which chemicals and for which regulatory purposes can we already do so, and what evidence would we need to extend that list?” Answering that question chemical class by chemical class is how the field will move from reducing fish use to phasing out routine acute fish testing.

References

• Fischer, M., Belanger, S. E., Berckmans, P., Bernhard, M. J., Bláha, L., Coman Schmid, D. E., Dyer, S. D., Haupt, T., Hermens, J. L. M., Hultman, M. T., Laue, H., Lillicrap, A., Mlnaříková, M., Natsch, A., Novák, J., Sinnige, T. L., Tollefsen, K. E., von Niederhäusern, V., Witters, H., . . . Schirmer, K. (2019). Repeatability and reproducibility of the RTgill-W1 cell line assay for predicting fish acute toxicity. Toxicological Sciences, 169(2), 353-364. https://doi.org/10.1093/toxsci/kfz057
• Knight, J., Rovida, C., Willett, K., & Ingram, J. (2026). Fish count, too: The animal toll of REACH aquatic toxicity tests. ALTEX: Alternatives to Animal Experimentation, 43(1), 142-157. https://doi.org/10.14573/altex.2506011
• Lillicrap, A., Moe, S. J., Wolf, R., Connors, K. A., Rawlings, J. M., Landis, W. G., Madsen, A., & Belanger, S. E. (2020). Evaluation of a Bayesian network for strengthening the weight of evidence to predict acute fish toxicity from fish embryo toxicity data. Integrated Environmental Assessment and Management, 16(4), 452-460. https://doi.org/10.1002/ieam.4258
• OECD. (2021). Test No. 249: Fish cell line acute toxicity: The RTgill-W1 cell line assay (OECD Guidelines for the Testing of Chemicals, Section 2). OECD Publishing. https://doi.org/10.1787/c66d5190-en
• Tanneberger, K., Knöbel, M., Busser, F. J. M., Sinnige, T. L., Hermens, J. L. M., & Schirmer, K. (2013). Predicting fish acute toxicity using a fish gill cell line-based toxicity assay. Environmental Science & Technology, 47(2), 1110-1119. https://doi.org/10.1021/es303505z