New Approach Methods in Genetic Toxicology

New approach methodologies (NAMs) are human‐relevant, computational, chemical, or cellular‐based methods that can replace, reduce, or refine the use of animals in research and testing. Practically this means replacing standard preclinical endpoints with in vitro/silico tools (or a combination thereof) to derive useful points of departure for human health risk assessments. In 2023 at the 54th annual meeting of the Environmental Mutagenesis and Genomics Society, a workshop sponsored by the Applied Genetic Toxicology Special Interest Group was held on “Incorporating New Approach Methodologies (NAMs) into Modernized Approaches for Genetic Toxicity Assessment.” Based on the discussions there, we compiled this special issue to highlight NAMs that are becoming more widely used in the field. Before we get to a brief review of those technologies, it's worth acknowledging an important aspect of NAMS that wasn't strongly represented there or in the collected manuscripts—validation.

Although the terminology “NAMs” is new, they are a familiar concept in the field of genetic toxicology. In fact, the spirit of NAMs is what ignited the field. Consider the Ames test, a bacterial reverse mutation assay, that was engineered and created to more rapidly detect rodent carcinogens with a mutagenic mode of action (MOA) (Ames et al. 1975). At the time this was novel, provided a rapid (< 1 week) timeline, moved away from animal use, and chemicals could now be screened and prioritized for selective in vivo testing. The Ames test helped to address a growing public concern on chemical‐induced carcinogenesis in the early 1970s (National Cancer Act of 1971 1971; US EPA 1992). While this pioneered our love for counting dots, it also instilled the biggest limitation facing NAM development in genetic toxicology—they were, and generally are, benchmarked to published test results in non‐clinical species. This practice continues as results from in vivo studies are considered “truth” that are used to define sensitivity and specificity of cell‐based assays (Kirkland et al. 2016).

While current genetic toxicity NAMs have addressed the general lack of specificity that current cell‐based assays have, the challenge of benchmarking to human (not rodent) outcomes remains. It is a conceptual catch 22. On one hand, we strive to compare NAM performance to expected human toxicities, yet human data rarely exist as preclinical models have identified mutagenic hazards which limited (or prevented) human exposure. How then should the field adequately validate these new models? This fundamental question has delayed the intended paradigm shift that NAMs offer to toxicological sciences. Progress is being made globally, and in the USA, legislation is in place to adopt NAMs for regulatory use (US Congress 2016, 2022). Yet, only recently have the necessary scientific efforts emerged, across public and private sectors, to build requisite NAM qualification networks. Two examples are the efforts by NAMWISE (https://cordis.europa.eu/project/id/101191595) in the European Union and the Foundation for the National Institutes of Health (https://fnih.org/our‐programs/validation‐qualification‐network‐design‐phase/) in the United States. Once qualification networks are fully vetted, NAMs that complete the process should be considered a suitable alternative to regulatory‐driven animal endpoints.

In the meantime, animal data comparisons may still be a component of NAM performance evaluations since existing models have a plethora of historical data and practical experience across many laboratories. One example is the interlaboratory ring trial performed to support the OECD guideline creation for ToxTracker, which was accepted by OECD July of this year (Hendriks et al. 2024; OECD Environmental and Molecular Mutagenesis 2026). A critical component was benchmarking results to in vivo genotoxicity outcomes, specifically, the rodent erythrocyte micronucleus assay. Comparisons like this are useful when considering replacement of animal tests, as “false positives” could halt product development, while “false negatives” may bring a potential hazard forward. This past dogma assumes that animal models are correctly predicting human health hazards. However, LOAELs from rodent studies are not protective of human health unless non‐empirically derived fit‐for‐purpose uncertainty factors are applied (Weitekamp et al. 2025). Conversely, points of departure derived from NAMs yield more conservative, protective, hazard estimates (Beal et al. 2023; Kuo et al. 2022; Thienpont et al. 2025; Weitekamp et al. 2025; Wills et al. 2021). While this approach is arguably an improvement for assessing potential human health risk, it may shrink safety margins used to advance product development, which could be a deterrent for industry acceptance. Additionally, comparison to human and rodent carcinogenicity test results may skew the performance assessment of NAMs unless MOA is considered.

For example, in 2006 the GreenScreen assay was one of the first recognized NAMs in genetic toxicology. As stated above, it wasn't the first “NAM” per se (see Figure 1 for a chronology of genotoxicity tests) but was rapid and utilized human‐derived lymphoblastoids (TK6 cells) containing a GFP reporter (Hastwell et al. 2006b). This was unique, as instead of a cytogenetic endpoint (i.e., visible DNA structural damage) it visualized transcriptomic changes to inform on genotoxic hazard. Centered around a single gene (growth arrest and DNA damage‐inducible, alpha; GADD45a), it detected activation of the DNA damage response pathway, which gave the endpoint high specificity when compared to regulatory in vitro assays characterized by misleading positive test results driven by cytotoxicity (Birrell et al. 2010; Hastwell et al. 2006a; Kirkland et al. 2016). While transformative, GreenScreen lacked endogenous metabolism, utilized a low top concentration (100 μM), and didn't inform on MOA—which are key aspects of defining genotoxicity in vitro. This was highlighted when a binary comparison of GreenScreen versus human and rodent carcinogenicity showed low sensitivity of this NAM (Olaharski et al. 2009). Even though this comparison did not consider carcinogenic MOA, and was challenged (Walmsley and Billinton 2009), the assay lost momentum in gaining widespread acceptance. This underscored that direct benchmarking of NAMs to human carcinogenicity, when such data exist, hasn't necessarily clarified their utility and in some cases slowed adoption.

FIGURE 1.

Chronology of biological assay development in genetic toxicology. Top: Animal based tests. Bottom: Cell based in vitro assays, with regulatory assays (pre‐1990) isolated in the dashed box. Generally, similar endpoints were first available in animal models prior to being adapted to cell culture (e.g., micronucleus and TGR endpoints), with the exception of HPRT. Abbreviations: HPRT = hypoxanthine phosphoribosyltransferase, CEGA = chicken egg genotoxicity assay, TGX‐DDI = transcriptomic DNA damage inducing biomarkers, Pig‐a = phosphatidylinositol glycan anchor biosynthesis class A. (Bender and Gooch 1962; Bryce et al. 2014; Clive and Spector 1975; Countryman and Heddle 1976; Dertinger et al. 2007; Evans 1970; Gocke and Müller 1988; Gossen et al. 1989; Heddle 1973; Hendriks et al. 2012; Kennedy et al. 2014; Li et al. 2019; Randerath et al. 2001; Singh et al. 1988; Thompson and Rubnitz 1971; White et al. 2003; Williams et al. 2014; Wood et al. 2010; Yamagiwa and Ichikawa 1918)

As validation/qualification networks are developed for NAMs, we should consider existing animal‐free testing paradigms in the field. Why recreate the wheel? Over 15 years ago, the European Union enacted a restriction on animal testing when evaluating the safety of cosmetic ingredients (European Parliament and Council 2009). This necessitated expert guidance on establishing an in vitro‐only testing strategy in the genetox field (Scientific Committee on Consumer Safety 2021). The strategy aligns with the recommendation for fragrance testing (Thakkar et al. 2023), sharing a focus on detecting both predominant DNA reactive MOAs—mutagenicity and clastogenicity. However, the latter includes in silico tools and NAMs as follow up tests to clarify potentially misleading positive test results. Similar workflows are being established for botanical extracts, where NAMs are recommended as the first tier of evaluation (Witt et al. 2025). Learnings from these frameworks should lay the foundation for an in vitro only hazard identification strategy for chemicals and pharmaceutical ingredients.

The 2023 EMGS Workshop focused on current uses of NAMs in the genetic toxicology space with an emphasis on the added value that MOA adds when interpreting data. Herein we showcase this along with other new tools that are at various stages of development, levels of maturity, and areas of application. One of the widely established NAMs, MultiFlow, kinetically evaluates multiple biomarkers, including nuclear gH2AX and pH 3, in TK6 cells to categorize compounds as clastogens, aneugens, or non genotoxicants (Bryce et al. 2016). A new wash out procedure in conjunction with the presence or absence of DNA repair inhibitors or ROS scavengers further categorizes clastogens based on DNA reactivity and other important mechanisms (Bryce et al. 2025). Other labs have begun refining analysis pipelines for this assay, in open‐source software, while others are utilizing ToxPi to efficiently visualize data from this multiplexed assay (Hussien et al. 2025; Trairatphisan et al. 2025). Even with this established tool, basic research continues. Avlasevich and colleagues assess potential well‐to‐well contamination due to volatility, which could confound data interpretation (Avlasevich et al. 2025). Lagunas and colleagues go one step further, interrogating questionable MultiFlow and micronucleus results by exploring the transcriptome, using the nearly‐qualified TGx‐DDI assay (Lagunas Jr et al. 2025). While MOA is the backbone of MultiFlow, combining NAMs together strengthens weight of evidence as it provides orthogonal substantiation for the MOA. Examples of other methods used to discriminate aneugens from clastogens/mutagens are described by Sun and colleagues (Sun et al. 2025).

At face value, the “human relevance” component of NAMs would suggest that test systems derived from humans are requisite. In addition to MultiFlow conducted in TK6 cells, more complex organ‐mimicking cell‐based models can be used. For example, HepaRG cells, derived from human liver, can form 2D or 3D cultures with endogenous metabolic capacity circumventing the need to add exogenous liver homogenate (S9). Herein, Engleward and colleagues have used 2D HepaRG cultures to investigate transferability of CometChip technology across 4 laboratories and highlight the detoxification abilities of this model (Recio et al. 2025). Additional work compared micronucleus induction across 2D HepaRG cultures and TK6 cells for 28 chemicals, suggesting that overt phase 1 metabolism from exogenous S9 may artificially inflate MN response (Allemang and Pfuhler 2026). An alternate model (organotypic large airway human tissues) was used to assess formaldehyde‐induced genotoxicity after subchronic exposures via CometChip and error‐corrected next generation sequencing (Le et al. 2025). Results were negative, which supports that rodent physiology and anatomy contribute to the observed genotoxicity further questioning the translatability of this hazard to humans.

While human‐derived test systems may be the de facto NAM standard, models derived from other species can be utilized provided data are relevant for assessing human health hazards. Since mutation and other genetic toxicities are not species or physiology dependent, the field has historically relied on cells from hamster or murine origin, largely due to their rapid cell cycle. However, tumor‐derived test systems are not genetically wild‐type and may display abnormalities in gene regulation. To circumvent this, primary or primordial cell types can be used, such as in ToxTracker, which utilizes DNA repair‐proficient mouse embryonic stem cells that report on direct and indirect (e.g., oxidative stress) MOAs. This assay has been used to refine a mammalian cell‐based exposure protocol that can better detect nitrosamine induced genotoxicity, a class of concerning mutagens that require specific bioactivation (Geijer et al. 2025). Further enhancements to detecting genotoxic metabolites that are poorly detected (or missed) with exogenous S9 come from in ovo models that have functional hepatic tissue (Williams et al. 2014). Kobets and Williams (2025) present a meta‐analysis comparing turkey and chicken egg genotoxicity data (CEGA) to existing test results in the literature. In CEGA, the method of exposure is further characterized and qualified by showing biodistribution of chemicals post air sac injection (Thakkar et al. 2025). Another example of improving metabolic competency is described by Göpfert et al. (2026) where primary mutamouse hepatocyte cultures are used to assess chemical‐induced mutagenesis without traditional S9. It performed well when compared to traditional Hprt forward mutation results, adding to the growing popularity of test systems with endogenous metabolic capacity.

In silico tools that report on structure activity relationships (SAR) have transformed how we assess genotoxicity. Often, they are a first step in chemical hazard identification and have been included in regulatory workstreams (ICH 2023). Herein, Pradeep and colleagues evaluate a variety of publicly available quantitative SAR tools to prioritize follow up testing for 24 data‐poor tattoo inks (Pradeep et al. 2025). Computational approaches are a practical first step in any hazard assessment strategy, and when combined with cell‐based NAMs, have the potential to replace and reduce the reliance upon traditional animal models. Once SAR identifies a potential hazard, a battery of NAMs can be used to assess biological activity in the context of adverse outcome pathways (AOPs). AOPs provide a framework of chronological biological events that occur from a molecular initiating event to an adverse outcome, in our case, mutation or chromosomal aberrations. To exemplify this, AOP #296 was explored with 4‐nitroquinoline‐1‐oxide, which induces genotoxicity via alkylation and ROS cycling. Using a variety of NAMs, kinetic endpoints in TK6 cells supported that this AOP was operable after 4NQO exposure as key events were demonstrated temporally (Huliganga et al. 2025).

Currently, the field uses NAMs to build weight of evidence for a hypothesized MOA, and/or triage misleading in vitro test results from regulatory endpoints. In the future, frameworks of qualified NAMs will evolve for regulatory use, and with in vitro to in vivo extrapolation, direct assessment of human health risk can occur. While this special issue is not all‐inclusive, it should inform new readers on the current landscape of NAMs in genetic toxicology with the aim of energizing them for the paradigm shift to come.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.