Revolutionizing Developmental Neurotoxicity Testing – a Journey from Animal Models to Advanced In Vitro Systems

,


Fig. 1: Illustration of the knowledge gap of DNT in the chemical universe
Based on Grandjean andLandrigan, 2006, 2014;Aschner et al., 2017 1 Introduction Our 2014 article in this series, "Developmental neurotoxicity -Challenges in the 21 st century and in vitro opportunities" analyzed the state of the art and emerging opportunities (Smirnova et al., 2014).Much progress has been made since.Here, we will summarize this, emphasizing the strategic development from having no alternative to animal testing to the recent establishment of the Organisation for Economic Co-operation and Development (OECD) "Initial Recommendations on Evaluation of Data from the Developmental Neurotoxicity (DNT) In Vitro Testing Battery" (OECD, 2023) over 18 years.
Developmental neurotoxicity (DNT) is a big public health concern in the context of autism spectrum disorder (ASD), attention deficit hyperactivity disorder (ADHD), and other neurodevelopmental delays (Tab.1).The developing brain is more vulnerable to environmental perturbations than the mature brain due to pharmacokinetic factors, reduced defense mechanisms, and the complex processes of neurodevelopment (e.g., proliferation, migration, and differentiation) (Rice and Barone, 2000;Rodier, 1995).These processes take place in strictly controlled timeframes over several years and create different windows of vulnerability.As we summarized in our 2014 article (Smirnova et al., 2014), interference with these developmental processes by, e.g., chemical exposure, can contribute to neurodevelopmental disorders such as ADHD, intellectual disabilities, and ASD (Kuehn, 2010;Sagiv et al., 2010;Grandjean and Landrigan, 2004;Landrigan, 2010, Rossignol et al., 2014).However, very few substances have been identified as developmental neurotoxicants due to the general lack of toxicity information (Grandjean andLandrigan, 2006, 2014) (Fig. 1).by epidemiological studies in establishing causal relationships between environmental exposures and neurodevelopmental disorders, due to difficulties in study design, biosampling, and exposure metrics.
In response to these challenges, the article explored the movement towards alternative DNT assessment methods, initiated through international workshops and conferences by expert groups and institutions.This effort aimed to identify promising alternative approaches, such as the use of adverse outcome pathways (AOPs) (Leist et al., 2017) and integrated testing strategies (ITS) (Hartung et al., 2013a;Rovida et al., 2015), to enhance the predictivity and relevance of DNT testing.The article emphasized the need for innovative biotechnological and computational methods to overcome the limitations of current testing paradigms.Despite the progress made, it acknowledged the ongoing challenges in validating and integrating these new approaches into regulatory frameworks.The article called for continued innovation, collaboration, and strategic evaluation to improve the quality and efficiency of DNT testing, ultimately aiming for better protection of human health without relying on animal testing.

Limitations of DNT animal testing
The case of DNT is a great illustration of how difficult testing for complex hazards in animal models actually is.Due to the potential impact of chemicals on human brain development, the development and adoption of OECD Test Guideline (TG) 426 (OECD, 2007) and the extended one-generation reproductive toxicity test guideline 443 (OECD, 2018) are significant milestones in the evolution of developmental neurotoxicity (DNT) testing.OECD TG 426, established in 2007, provides a framework for conducting in vivo DNT studies.The guideline is designed to identify chemicals that may affect the nervous system during development, characterizing any chemical-induced alterations and estimating dose levels for regulatory uses.It includes specific endpoints to evaluate functional, behavioral, and morphological effects on the nervous system, with additional testing of offspring exposed in utero and during early lactation (Makris et al., 2009).However, it cannot realistically measure typical human adverse outcomes, such as reduced language capacities, autism spectrum behavior or reduced IQ scores.Instead, it measures some endpoints that are not typical characteristics of human pathologies like altered grooming behavior, altered light-dark preferences, whisker reflexes, delayed eye opening, etc.
TG 443, also known as the extended one-generation reproductive toxicity study, expands upon previous guidelines by including assessments of reproductive and developmental effects within a single study (Moore et al., 2016).It allows for the examination of multiple generations and includes optional modules to assess developmental neurotoxicity and immunotoxicity.This guideline aims to provide a more comprehensive understanding of a substance's potential reproductive and developmental hazards while using fewer animals than traditional two-generation studies.In actual practice, the reduced number of animals needed does not really hold (see below).Both TG 426 and TG 443 have been subject to critical evaluation and refinement to ensure that they reflect the best available science for assessing DNT potential in human health risk assessment using animals.The guidelines emphasize the importance of considering the complex nature of neurodevelopment and the need for a variety of test methods to address different aspects of DNT (Makris et al., 2009;Arts et al., 2023).
Traditional animal-based developmental neurotoxicity (DNT) testing faces significant challenges and limitations.OECD TG 426 relies entirely on in vivo animal experiments, which are designed to detect gross neurological and behavioral abnormalities.However, these tests are not formally validated in ring trials and are both costly and time-consuming.For a single DNT study, approximately 1,200 rat pups are used, and the experimental part of the study lasts about three months.Data evaluation (e.g., pathology reading) takes about two years, with overall costs reaching around $1.4 million per substance.Ethical concerns arise from the use of animals in such large numbers, especially considering the sensitivity of the tests may not be sufficient to detect subtle neurotoxic effects that could be relevant to human health.Also, the scientific relevance to human health effects is questionable (Smirnova et al., 2014): Interspecies differences: One of the primary limitations is the interspecies differences between humans and the animals used in DNT testing such as rats or mice.These differences can affect the extrapolation of data from animals to humans, as the mechanisms of neural development and the responses to toxic substances can vary significantly between species.The human brain is our most complex organ, and replicating its development in animal models is inherently challenging.
Genetic variability: Another limitation is the inability of animal models to reflect the inter-individual genetic and epigenetic differences found in human populations.These differences can influence susceptibility to neurotoxicants, and animal models typically lack this variability.For example, the use of inbred animal strains reduces genetic diversity, which can limit the predictivity of the tests for human health effects.
Behavioral and histological interpretation: Interpreting behavioral effects in animals is also difficult.Behavioral tests in animals may not accurately represent the complex human behaviors or cognitive functions that could be affected by neurotoxicants.Additionally, potential artifacts in morphometric neuropathological measures can arise, and criteria for observation measures are not always clear, leading to uncertainty in the evaluation of histological data.
Ethical and practical concerns: Ethical concerns regarding the use of animals in research are significant and have led to calls for the reduction, refinement, and replacement of animal testing.Moreover, animal-based DNT tests are expensive, timeconsuming, and require a large number of animals, which is prohibitive for routine chemical screening.This results in many potential DNT chemicals remaining unidentified due to the impracticality of testing them all using current animal-based guidelines.
In response to these limitations, there is a push for the development of new approach methodologies (NAMs), such as human in vitro tests and microphysiological systems (MPS) of the brain.These systems aim to mimic the cellular composition, regional architecture, and functionality of the human brain more closely than animal models (Hogberg and Smirnova, 2022).They offer the potential for higher throughput and better predictivity of human health.In summary, while animal-based DNT testing has provided critical data for regulatory decisions, the limitations regarding interspecies differences, genetic variability, interpretation of behavioral and histological data, ethical concerns, and practicality highlight the need for alternative testing strategies that are more predictive of human health effects.Advances in human-based in vitro models and MPS are promising developments in this field.
Economical aspects (Meigs et al., 2018) also need to be considered: Some DNT concerns are covered also in the twogeneration study (TG 416), considered the most definitive study for developmental and reproductive toxicity (DART).The extended one-generation reproductive toxicity test guideline (TG 443), which has a DNT module that can be triggered or be part of routine depending on regulatory requirements, was expected to reduce the costs of TG 416 studies (Schiffelers et al., 2015).However, an analysis commissioned by ECHA2 showed that the worldwide average price for TG 416 is €285,842 (European average is €318,295), while for TG 443 the average price for the basic study (without second generation and extra cohorts) is €414,273, and with second generation €469,778, i.e., 60% more than TG 416.With extra cohorts, the price increases to €507,444 for DNT cohort or €440,414 for the immunotoxicity cohort; both cohorts and the second generation costs €655,195.The expectation to save drastically on animal numbers has also been disproven (Knight et al., 2023;Rovida et al., 2023)

The path of alternatives to in vivo DNT testing up to the DNT-4 meeting in 2014
The first article on DNT in this series (Smirnova et al., 2014) was written in conjunction with the DNT-4 meeting in Philadelphia.The following summary of how we got to Philadelphia shall illustrate the unique strategic development to replace animal testing starting from zero.It all started with Alan Goldberg, at the time director of CAAT, approaching Thomas Hartung, at the time head of the European Center for Validation of Alternative Methods (ECVAM), in 2004, saying, "Thomas, we have to do something about this."With an ever-growing team of collaborators, they carried out 9 workshops and 4 international DNT conferences.Alan Goldberg stated, "I chose DNT as the topic for TestSmart as there were no specific requirements for DNT, so we did not have to match against an in vivo approach."The most important outcome of this was the engagement of regulatory agencies: The US Environmental Protection Agency (US EPA) from the very beginning, and later European Food Safety Authority (EFSA) and the OECD, and ultimately guidance was developed (Fig. 2).
The discussion started in Aspen with Pamela Lein, Alan Goldberg, and Kevin Crofton talking about the need for development and neurotoxicity testing around 2003 (Lein et al., 2005).ECVAM had started working on this topic around the same time with Sandra Coecke, Anna Bal-Price, Chantra Eskes and Helena Hogberg (Hogberg et al., 2010(Hogberg et al., , 2011;;Bal-Price et al., 2010).During this time, we were continuously developing assays, with a "family" element of researchers joining such as Bill Mundy, Ellen Fritsche, Marcel Leist and many others.With Thomas Hartung, Erwin van Vliet, and Helena Hogberg moving to CAAT in 2009, laboratories focusing on DNT were created and workshops were held.The series of international conferences co-organized by CAAT, ECVAM, and US EPA will be summarized in the following.

The first DNT workshop at ECVAM 2005
At ECVAM, we organized the first workshop as a collaboration between CAAT, ECVAM, and CEFIC3 (the chemical industry in Europe) on April 19-21, 2005, in Ispra, Italy.It did a lot of important things.The article "Incorporating in vitro alternative methods for developmental neurotoxicity into international hazard and risk assessment strategy" by Coecke et al. (2007) reports on the first workshop aimed at integrating in vitro alternatives for developmental neurotoxicity (DNT) testing into global risk assessment frameworks.The workshop's goal was to identify and catalog in vitro methods that could predict and identify DNT hazards.Discussions focused on the science of DNT, including models capable of capturing critical mechanisms and processes, and policy and strategy for integrating alternative methods into regulatory frameworks.The report outlines recommendations and priorities for future work, emphasizing the need for high-throughput screening, in vitro DNT models, and their regulatory validation.It highlights the importance of identifying and validating in vitro tests that can effectively predict DNT hazards, the critical role of mechanistic understanding in developing these tests, and the necessity of a collaborative approach among scientists, regulators, and stakeholders to integrate these methods into risk assessment strategies.Already in the 2007 paper on the 2005 workshop, we discussed that embryonic stem cells might be a possibility for humanizing this.This was a year before Yamanaka developed induced pluripotent stem cells and still ethically a very problematic aspect.The question was what are the endpoints and aspects of neural development that need to be modeled in these cultures?The focus was made on a certain number of key events, which we felt should be measured in order to model or assess developmental neurotoxicity.The second question we posed already then was "which cell model?"This implies a tradeoff between increasing complexity and throughput.All of these models from simple cell lines to complex Microphysiological Systems (MPS) and lower organisms and even emerging computational methods have their own rights in this.At ECVAM and later transferring this to CAAT, we started to focus on 3D models -aggregates -at the beginning.We felt this is a good trade-off between complexity and throughput.Honegger and his group in 1979 published in Nature the model of reaggregating rat fetal brain cells, which was successfully adopted by our group (Honegger et al., 1979).It led us to the first paper describing the use electrophysiological recording in DNT (van Vliet et al., 2007), and to the first paper, where DNT was assessed using metabolomics (van Vliet et al., 2008).It allowed us to address more recently developmental neurotoxicity of flame retardants (Hogberg et al., 2021).Additionally, the 2007 paper outlines strategic recommendations for advancing the use of in vitro methods for DNT, emphasizing the need for a concerted effort to address scientific and regulatory challenges.For many, this sounded like "Mission Impossible" at the time, but it started to break down the Herculean challenge into doable steps.This constituted the roadmap, which ultimately led to the OECD guidance in 2023.
In summary, the key messages from the 2005 workshop (Coecke et al., 2007) focus on the urgent need to develop and incorporate in vitro alternative methods for developmental neurotoxicity (DNT) into international regulatory frameworks.It highlights the importance of identifying and validating in vitro tests that can effectively predict DNT hazards, the critical role of mechanistic understanding in developing these tests, and the necessity of a collaborative approach among scientists, regulators, and stakeholders to integrate these methods into risk assessment strategies.Additionally, the paper outlines strategic recommendations for advancing the use of in vitro methods for DNT, emphasizing the need for a concerted effort to address scientific and regulatory challenges.

The first international DNT conference DNT-1 in 2006
The first international DNT meeting, TestSmart DNT-1, took place in Reston, VA, in 2006 and its primary objective was to bring together stakeholders (e.g., industry, academia and regulators) to identify concerns relating to science and policy and how to develop alternative testing methodologies (Lein et al., 2007).The paper discusses the imperative need for developing alternative methodologies to traditional animal-based DNT testing.The work emphasizes the collaboration between CAAT, the U.S. EPA, and the National Toxicology Program (NTP) towards creating TestSmart DNT -aimed at fostering alternative DNT models.These initiatives focus on identifying and prioritizing chemicals posing DNT risks, integrating alternative DNT methods into regulatory decision-making, and exploring possibilities for reducing, refining, or replacing animal use in DNT testing.
The progress from the 2005 workshop to the 2006 conference reflected a significant advancement in the field of DNT testing: While Coecke et al. (2007) focused on the initial discussions and the conceptual framework for incorporating in vitro methods into regulatory practices, Lein et al. (2007) highlighted active collaborations and specific initiatives like TestSmart DNT, aimed at practically implementing these alternative methods.This transition marks a movement from theoretical and strategic planning to actionable steps towards validation and regulatory acceptance of in vitro DNT testing methods, demonstrating an increased commitment to reducing animal testing through scientifically robust alternatives.

The second international DNT conference DNT-2 in 2008
In 2008, TestSmart DNT-2 assessed the progress made in developing DNT alternatives, reassessed the priorities and recommendations founded at DNT-1, and established ways to use in vitro data in decision making.This meeting produced a document with recommendations on how to develop alternative DNT methods for screening and prioritization of chemicals (Crofton et al., 2011), which included: 1. Test methods should incorporate one or more endpoints that model key aspects of human neurodevelopment.The test method models the biological process, the test system employs the appropriate cells/organisms, and the endpoints measure relevant developmental features.2. The ability to correctly and accurately measure the intended DNT endpoint must be demonstrated by the use of a set of compounds termed "endpoint-specific controls" or "tool compounds".The measured endpoint should reflect the intended neurodevelopmental process.3. The dynamic range of the DNT endpoint should be characterized to determine the measurable extent of change from control values.4. Concentration-response relationships should be characterized, ideally testing at least five concentrations over a wide range.This is critical for comparing the sensitivity of different methods.5. Positive and negative control chemicals should be tested that are known to reliably affect or not affect the measured DNT endpoint by known mechanisms.6.Initial training sets of chemicals should be tested, including those known to elicit or not elicit DNT responses based on in vitro data.This evaluates the method's ability to screen moderate numbers of chemicals.7. Larger testing sets of chemicals should then be screened, including those known to cause or not cause DNT effects in vivo.
This demonstrates the method's ability to test larger chemical numbers.8.The metho's sensitivity, specificity and predictivity in identifying chemical' DNT potential should be analyzed based on these reference chemical sets.In summary, the recommendations focused on demonstrating the relevance, reliability, sensitivity, and throughput of alternative DNT methods using well-characterized endpoint assays and reference chemical testing sets, as a framework for developing screening-level alternative DNT tests.The goal was establishing the method' fitness for screening and prioritizing large chemical numbers.
The progress from the initial workshop and DNT-1 to DNT-2 recommendations built upon these foundations to provide more specific, actionable recommendations for developing alternative DNT methods amenable to screening and prioritizing large numbers of chemicals.In summary, Crofton et al. (2011) provided a more detailed framework and guidelines for developing alternative DNT assays, with an emphasis on their suitability for screening/prioritization purposes.The DNT-2 recommendations aimed to facilitate a transition from conceptual discussions to practical assay development and use for regulatory purposes.These considerations and experience from the DNT field had an important impact on the development of broader concepts for in vitro test method development and for novel validation approaches of such assays (Leist et al., 2010(Leist et al., , 2012;;van Thriel et al., 2012).

3.4
The third international DNT conference DNT-3 in 2011 in Varese, Italy DNT-3 was held in Varese, Italy, in 2011 and concluded substantial progress since DNT-2 in applying alternative DNT tests (Bal-Price, 2012).DNT-3 built upon the recommendations from Crofton et al. (2011) and showed progress in several key areas: 1.There was general consensus on the urgent need to develop alternative DNT testing strategies that are faster, more costefficient, and predictive of human outcomes.High-throughput in vitro DNT models are needed to test the large numbers of chemicals for which there is little to no DNT data.2. Significant progress was reported in applying in vitro and non-mammalian test systems to DNT, including human stem/progenitor cell-based assays, though more work is needed to validate these alternative models against human DNT outcomes.3. Generating data across multiple alternative models using a common set of test chemicals was identified as a critical need to facilitate comparisons and determine which models/endpoints are most predictive.4. Establishing a reference set of positive and negative control chemicals for DNT was also deemed important for evaluating alternative models.5. Cell-based assays covering key neurodevelopmental processes like proliferation, migration, differentiation, synaptogenesis, and network formation should be applied as functional DNT endpoints.6. Computational modeling and bioinformatics approaches should be utilized to evaluate the predictive capacity of alternative DNT models/batteries.7. Despite progress, in vitro models are not yet able to replicate the complexity of the developing nervous system.Their relevance to human outcomes must still be cautiously evaluated.In summary, DNT-3 highlighted advancements in developing higher-throughput alternative models based on the DNT-2 framework, while identifying crucial data gaps, such as the need for more cross-model comparisons using standardized reference chemicals.The emphasis shifted to practical application of DNT alternatives for screening/prioritization, beyond the initial proof-of-concept stage discussed in 2008.

The first Food for Thought … article on DNT
Building upon the seminal works discussed above, our 2014 paper (Smirnova et al., 2014) presents several key conceptual advances in the development of alternative approaches for DNT testing.These advances reflect the rapid progress made in the field since the publication of the earlier papers and provide a forward-looking perspective on the future of DNT testing.
First and foremost, we emphasize the pressing need for more efficient, cost-effective, and mechanistically informative methods to address the paucity of DNT data for the vast majority of chemicals in commerce.This builds upon the discussions in the previous papers, which highlighted the limitations of current animal-based DNT testing paradigms, including their high cost, low throughput, and questionable relevance to human health outcomes.We argue that the development of cheaper, faster, and more predictive alternatives is not only a scientific imperative but also an ethical one, given the large numbers of animals used in regulatory DNT studies.
Secondly, the paper discusses the concept of adverse outcome pathways (AOPs) and their central role in guiding the development of integrated testing strategies (ITS) for DNT.AOPs provide a structured framework for linking molecular initiating events (MIEs) to adverse outcomes at the organism or population level, via a series of key events at the cellular, tissue, and organ levels.By mapping the landscape of known or putative AOPs relevant to DNT, researchers can identify critical pathways and processes that should be targeted by alternative test methods.This advances the mechanistic approach advocated by Crofton et al. (2011) and provides a rational basis for designing and interpreting in vitro and in silico assays in terms of their relevance to in vivo outcomes.
Thirdly, we highlight the transformative impact of recent technological advances on the development of alternative DNT models.These include the use of human stem cell-derived neural models, which offer a more physiologically relevant and speciesspecific platform for studying neurodevelopmental processes and their perturbation by chemicals.Advances in 3D cell culture techniques, such as brain organoids, allow the recreation of complex tissue-level interactions in vitro, bridging the gap between traditional 2D models and intact organisms.High-throughput screening (HTS) technologies, such as automated imaging and multielectrode arrays, enable the rapid and quantitative assessment of key neurodevelopmental events, such as neurite outgrowth, synaptogenesis, and network formation.These technological breakthroughs, which were still in their infancy when the earlier papers were published, have greatly expanded the toolkit available for alternative DNT testing.
Building on these advances, we proposed a conceptual framework for linking in vitro test systems to an endophenotype of disturbed functional or structural connectivity in the brain.It was postulated and assumed that all DNT effects eventually resulted from the disturbance of at least one key neurodevelopmental process (KNDP), such as cell proliferation, differentiation, migration, and myelination.This pragmatic approach acknowledges that our current understanding of the complete chain of events linking MIEs to adverse outcomes in the developing brain is still limited.It also anticipated currently discussed test strategies that focus more on determining the highest non-toxic (no toxicity endophenotype) concentration/dose, than on exactly defining the type of adverse outcome (e.g., attention deficit vs language disturbance).By focusing on a core set of well-characterized neurodevelopmental processes, which are known to be sensitive to chemical perturbation and critical for proper brain development, researchers can develop a battery of fit-for-purpose assays that cover the main pathways of DNT.This framework provides a roadmap for assay development and validation in the absence of complete AOPs.
Another important issue tackled in 2014 is the challenge of defining adversity in the context of alternative DNT models.While apical endpoints, such as changes in behavior or cognitive function, are typically used to define adverse effects in regulatory animal studies, these complex outcomes can be difficult to recapitulate in vitro.We discuss the need to establish clear criteria for distinguishing between adaptive and adverse responses at the cellular and molecular level, taking into account factors such as the magnitude, duration, and reversibility of the effects.This discussion was very much driven by our concomitant discussions on Pathways of Toxicity (Kleensang et al., 2014) versus Pathways of Defense (Hartung et al., 2012) in the context of resilience of biological systems (Smirnova et al., 2014).We also highlight the importance of considering the biological relevance of the observed changes, rather than relying solely on statistical significance.This nuanced discussion of adversity builds upon the earlier works (Blaauboer et al., 2012) and underscores the need for a careful and context-dependent interpretation of alternative DNT assay results.
Finally, we emphasized several key factors that should be considered when designing and interpreting alternative DNT models.These include potential interspecies differences in neurodevelopmental processes and chemical susceptibility, which underscore the need for human-relevant models.We also discuss the importance of considering indirect mechanisms of DNT, such as alterations in thyroid hormone signaling or placental function, which may not be captured by models focused solely on direct effects on neural cells.Additionally, we highlight the critical role of exposure timing, given the dynamic nature of neurodevelopment and the existence of critical windows of vulnerability.These considerations, while not entirely new, are given greater prominence in the paper, reflecting a more nuanced and biologically informed approach to alternative DNT testing.
In conclusion, the 2014 paper summarized a significant conceptual advance in the field of alternative DNT testing, building upon the groundwork laid by earlier publications.By integrating the latest scientific understanding of neurodevelopment with technological advances in in vitro and in silico modeling, we provided a comprehensive and forward-looking perspective on the challenges and opportunities in this field.We articulated a vision for a new paradigm of predictive toxicology, based on a mechanistic understanding of the pathways underlying DNT and their perturbation by chemicals.While acknowledging the limitations of current alternative models, we offered a roadmap for future research and development, emphasizing the need for a multi-disciplinary and collaborative approach.As such, the paper served as a valuable synthesis of the state-of-the-art in 2014 and a guidepost for the future direction of alternative DNT testing.

DNT-4 -Toward adverse outcome pathways and fit for purpose assays for DNT
DNT-4 -Toward Adverse Outcome Pathways and Fit-for-Purpose Assays for DNT, was held in Philadelphia, PA, in 2014.The meeting brought together diverse stakeholders (academia, industry, and regulatory bodies) from around the globe (Asia, Canada, Europe, and US) that discussed the subsequent steps required to move the in vitro DNT tests forward.Based on experiences from previous DNT meetings, the format of this meeting contained scientific presentations followed by smaller breakout groups where specific topics were discussed.The following is the report of what was discussed during DNT-4.The summary does not necessary reflect all opinions of the participants and the following statements are our own reflections of the meeting.As similar topics were discussed in several breakout groups, the report has been organized based on the initial meeting program.The AOP framework is a tool to combine existing knowledge concerning the linkage between a molecular initiating event (MIE) and an adverse outcome (AO) at the individual or population level (Ankley et al., 2010).The AOP covers the whole pathway including chemical properties, MIE, cellular response, organ response, organism response and finally the population response4 .It is clear that one test or study alone will not be able to capture this whole pathway.However, combining data from several DNT tests, including different models (e.g., cells and non-mammalian species) and endpoints for different developmental processes (e.g., proliferation, differentiation, and myelination) with existing in vivo data and data from epidemiological studies can help to reduce the uncertainty and give a better toxicity prediction.It should be emphasized that AOPs are not chemical specific, better explained is that any chemical that perturbs the identified MIE with sufficient potency and duration will have an effect on following chain of key events (KE) identified in the AOP.The advantage with this approach is that not all element in the AOP needs to be identified before the concept can be useful (Bal-Price and Meek, 2017).The AOPs will give us confidence that our cellular changes can lead to DNT effects in humans.In addition, by assembling all existing information, data gaps can be identified and assist in focusing on generating the right data, methods development, and data research needs.The AOPs will increase the certainty in the in vitro test for risk assessors and regulators to make decisions.

New concepts and test strategies
Currently, only a few AOPs for DNT have been developed according to the OECD guidance document (Bal-Price et al., 2015a).Once more data is obtained additional AOPs will be built.However, the AOP framework partakes especial challenges for neurotoxicity and DNT assessment.Firstly, there is a lack of basic knowledge about the pathophysiology of neurological diseases, e.g., autism and the understanding of the underlying MIEs and kEs.Moreover, there is a low understanding on compensatory processes in the nervous system, such as homeostasis and resilience.It was recognized that it takes time to build AOPs, however, the incompleteness in an AOP should not impede its use.The complexity of the development of the central nervous system will likely construct a network of AOPs instead of the suggested linear AOPs.
To make it even more complex to generate AOPs, it is likely that gene and environmental interactions play a crucial rule in the case of many neurodevelopmental diseases.This increase of ASD prevalence is partly due to increased diagnostic criteria.However, it cannot be fully attributed to diagnostic substitution (Hertz-Picciotto and Delwiche, 2009) and genetic causes; an increasing number of studies suggest that environmental factors contribute to this increase.Several genes involved in ASD and other neurodevelopmental disorders have shown to be susceptible to environmental perturbation (Pessah and Lein, 2008).In vitro models are preferable to evaluate these gene and environmental interactions and how it might affect neurodevelopment at the cellular level.The adverse effect on the cellular level can then be linked to adverse effects in patients with developmental disorders to build confidence in the AOP.For example, genes involved in synapse formation and elimination have shown to be disturbed by environmental exposure and epidemiological studies have shown that many children with ASD have increased connectivity in local circuits of the cortex (Keown et al., 2013).Such an approach to associate cellular effects with disease outcome will likely be crucial for regulators to make use of in vitro data for risk assessment.
A breakout group "Science of DNT Models and AOPs" elaborated further on this topic.The previous DNT meetings had already identified several promising models for DNT studies, e.g., cell lines, primary cultures, 3D aggregating cultures, zebrafish, C. elegans and drosophila (Coecke et al., 2007;Lein et al., 2007).At DNT-4 this breakout group deliberated further advantages and limitations of these models with a focus on what window of neurodevelopment they cover and if/how genetic variation can influence DNT endpoints in vitro.
It was concluded that different alternative models are suitable for various stages of neurodevelopment.Suggested as the most fitting cell models for earlier stages were precursor cell lines and stem cells (e.g., embryonic and induced pluripotent) together with non-mammalian species such as C. elegans and zebrafish.Some of these models, like induced pluripotent stem cells and the non-mammalian species can also cover the later stages of development while primary cultures are mainly relevant for the later periods.It was further discussed that different species have divergent developmental time frames, both in vitro and in vivo, e.g., rodent cells develop faster than human cells and the zebrafish have shorter developing time than mice.Shorter development can be an advantage as this would speed up the experimental time and make it more suitable as a screening model for DNT effects.However, it is unclear how species might differ in sensitivity to environmental perturbations.Several cell-based studies have reported species differences in response to various compounds.Though, there were concerns among the participants that these effects could be artifacts as it is difficult to compare species with such different developing pattern.Do these studies reflect actual species differences or just develop trajectory?Especially in the early stage of development the group felt that differences could be large while the later stages more easily can be compared between models.It was remarked that these challenges do not only concern different species but as well cells from different donors.Development from donors can differ depending on what time point the cells are taken as well as the genetic background.
It is well known that the genetics play a role in sensitivity in exposure to chemicals and drugs and it will clearly influence DNT endpoints in vitro.As the human population is heterogenetic and consists of vulnerable subgroups this can as well be an advantage.Genetic variations have long been explored in simpler organisms such as the nematode C. elegans and zebrafish (Nishimura et al., 2015).However, with the recent use of induced pluripotent stem cells, this is increasingly studied also in human cells (Pamies et al., 2017).It was specified that human cells in DNT studies would be preferred, as it would avoid species differences.However, there is a tremendous gap of relevant data on human exposure and DNT effects.Therefore, as we have more rodent in vivo data, rodent cells can be used as an important bridge between species.Correlating human cells to rodent cells and the rodent in vivo data can enhance extrapolation to the human situation (Maass et al., 2023;Algharably et al., 2023).Nevertheless, to be applicable the models need to be characterized based on the maturation as a function of days in vitro (DIV) and associated to the developmental stage in vivo.Is 7 DIV comparable to 7 days in vivo?It was discussed that measurements of specific markers could contribute to identifying the developmental stage of the model, e.g., specific miRNA, mRNAs and/or proteins.
The conclusion of this breakout group was that alternative models need to be well characterized to be useful in any toxicity study, including DNT.Often models lack information about metabolic capacity, understanding of different cell types, and association to the in vivo situation.However, many models have shown great competence and are useful for measurement of several developmental processes such as differentiation, proliferation, migration, and survival.The in vivo (rodent) predictability for a subset of chemicals seems to be good, however, many more chemicals need to be tested to fully understand the real potential (see below for deeper discussion).To better understand the relevance towards the human situation, the DNT community could benefit by interacting with the clinical community to correlate toxicity to, e.g., exposure scenarios, existing biomarkers, and pathological mechanisms.

Mechanistic and omics tools, as well as functional endpoints, to increase assay information content
In this session, different approaches to evaluate DNT using ~omics tools were presented.Dr Jennifer Freeman, Purdue University, gave some examples using the zebrafish, a non-mammalian model, and Dr Lena Smirnova described the use of 3D in vitro brain models.Dr Milou Dingemans, Utrecht University, informed how to integrate multiple data from the European Commission funded project DENAMIC5 , one of the bigger investments in Europe to tackle the DNT problem.
The benefit of using non-mammalian species, such as zebrafish and C. elegans for DNT assessment was already discussed in DNT-1 and -2.There are several advantages with using these models, e.g., they are inexpensive, fast, amenable to genetic modifications and allows whole organism and behavior studies.In the presentation of Dr Jennifer Freeman, Purdue University, the zebrafish was combined with transcriptomic and epigenomic approaches to identify DNT mechanisms for chemicals (Lee and Freeman, 2014).
The DENAMIC project Ides both in vitro, Zebrafish, in vivo models, and human cohorts and applies transcriptomics, proteomics, neurotransmitter profiles and miRNA to identify biomarkers for neurodevelopmental disorders.The focus is on mixtures and the data is integrated with the aim to understand mechanisms of the interactions of low doses and neurodevelopmental disorders.
The benefit of using 3D in vitro models, such as increased cell-cell interactions and better representation of the complexity of the in vivo brain.Different models human, rodents, single or multi-cellular types were evaluated for their advantages and disadvantages.The models are combined with transcriptomics, metabolomics, and miRNAs (Smirnova et al., 2015a) profiles to identify pathways of toxicity after chemical exposure that may lead to adverse effects on the developing brain.
The use of omics approaches will generate high-content information to better understand the mechanisms of chemicals as in the vision of the Toxicology in the 21 st Century (Krewski et al., 2020).However, the generation of massive data might lead to new challenges and was discussed in the breakout group on techniques.
The keynote by Marcel Leist from Konstanz University described the use of human ESCs combined with transcriptomics endpoints.It was demonstrated that different drugs affected different markers and biological processes, and that statistical prediction models could be developed that identified signature transcriptome changes for subgroups of DNT toxicants.Moreover, it was highlighted that short toxicant exposures may be more useful to identify modes-of-action, while longer exposures of differentiating stem cell models lead to epigenetic changes that affected the differentiation track of the cells and had permanent effects even after drug washout.The conclusion was that transcriptome patterns, recorded at non-cytotoxic chemical exposures, can allow for chemical classification and therefore hazard prediction.

How to accelerate testing for DNT
One of the major problems in the field of DNT is the lack of methods to assess hazard of the high amount of chemicals on the market.In this session speakers from the US government were describing current screening approaches for DNT at NIEHS by Dr Mamta Behl and at US EPA by Dr William Mundy and Dr Timothy Shafer.The number of chemicals that lack DNT data are too many to test with traditional in vivo guidelines.Therefore, the National Toxicology Program at NIEHS are working on strategies to screen and prioritize compounds for in vivo DNT testing.A list of 80 compounds have been developed that are tested by collaborators using various cell systems such as primary, ESC, iPSCs, and cell lines containing both different neurons and astrocytes as well as non-neuronal cells e.g., cardiomyocytes.Alternative models such as nematode and Zebrafish are also included.The prioritization is based on various data analysis models.
The high throughput testing program at UI includes ToxCast (700 assays and over 2000 chemicals tested) and Tox21 (in partnership with NTP, FDA and NCATS, screening more than 10,000 chemicals).Moreover, NHEERL/ToxCast are developing assays for more complex endpoints including DNT.The EPA assays for DNT are developed after the AOP framework with high throughput molecular assays e.g., ion channels, receptors, enzymes, for MIEs, proliferation, differentiation, neurite outgrowth, synaptogenesis, migration, and apoptosis assay and multi electrode array (MEA) for KEs and Zebrafish behavior assay for AO.A list of positive and negative reference chemicals has been developed.

The use of AOPs and alternative assays for safety assessment
For the alternative assays to be useful they need to be applied in a regulatory framework.However, different regulatory bodies have different requirements and needs.In this session Dr William Slikker, from NCTR, FDA, and Dr Anna Lowit from Health Effects Division, US EPA, gave two regulatory agencies perspective on the use of alternative methods for DNT assessment.
In the case of drug regulation at FDA, it was described that the challenge with only using mechanistic data is that drugs are not entirely pharmacologically identical.In this case, the data may suggest similar risk but the "safety margin" may not be the same.However, mechanistic data can still be helpful and together with in vivo experiments provide a better understanding of the DNT effects.
The regulation at US EPA is more flexible and the use of in vitro data when there exists knowledge from an AOP can be used to support read across for similar compounds.Furthermore, in vitro data can be used in a weight of evidence evaluation to determine data needs or to review a waiver justification.The US EPA is supporting the 3Rs in their regulatory program.In the case of DNT, US EPA granted 8 waivers and requested 1 study from December 8, 2011, to April 11, 2014.Several in vitro study design considerations to satisfy regulatory use were discussed including cell types, presence/absence of serum, parent chemical vs. metabolite and test concentrations.Furthermore, there are different fit for purpose depending on the intended use e.g., screening and prioritization vs. replacement of in vivo guideline or single assay vs. batteries.
Thomas Hartung discussed the use of integrating testing strategies (ITS) (Hartung et al., 2013a) for DNT assessment.An integrated test strategy is an algorithm to combine (different) test result(s) and, possibly, non-test information (existing data, in silico extrapolations from existing data or modeling) to give a combined test result.They often will have interim decision points at which further building blocks may be considered.Several aspects should be considered, for example the flexibility in combining components of the ITS, the applicability domain, and the efficiency in terms of cost, time, and technical difficulties.

Frontiers in DNT testing
The last session of the meeting described new cutting-edge technologies for DNT presented by stakeholders from government, and academia.Dr Andrea Seiler from the German Federal Institute for Risk Assessment (BfR) described the frame of a joint project funded by the German Ministry for Research and Education (BMBF) with the goal to develop standardized predictive cell-based in vitro assays for DNT testing.From US EPA, Dr Nisha Sipes used literature mining to identify MIEs in an AOP framework for cleft palate.Such informatic data can further be applied to develop multicellular virtual-tissue models (VTMs) e.g., neural tube closure in the virtual embryo, that are based on cell-level models driven by biological networks and rules that can be used for predictive toxicology.From the academia the extensive characterization of human iPSC cultured in 3D were presented by Dr David Pamies from Johns Hopkins University.Human iPSCs gives the opportunity to study patients with different genetic backgrounds and can explore if various genes give sensitivity to chemical exposure.ESCs and iPSCs in 3D were also applied at the University of Applied Sciences Geneva presented by Dr Luc Stoppini, but with focus on their functionality using various MEAs.Functional effects after exposure to chemicals were further evaluated with gene expression for various receptors and metabolomics for neurotransmitters to develop an in vitro assay for toxicology and drug screenings.The 3D models were further enhanced by incorporation of monocytes to study neuroinflammation and combined with other organ cultures in a chip platform and blood brain barrier.Finally, Dr Keith Cheng from Penn State College of Medicine discussed a novel imaging technology in whole Zebrafish.The high-throughput, technique is capable of detecting changes in any cell type at cell resolution (voxel dimensions of ~1 micron) and makes it possible to identify and characterize virtually every organ, tissue, and cell type contributing to soft tissue architecture including specific structures such as nerve tracts and vessels.

Conclusions from DNT-4
"Where are we at DNT-4?"The meeting ended with a panel and plenary discussion of the steering committee moderated by Dr Alan Goldberg.Several challenges and limitations with current approaches were identified.
The relatively new AOP framework is still not mature enough to use for regulatory decisions but will be important to understand how to link cellular events to adverse outcomes.The major challenges with the AOPs are to understand the relationships between the events and to get quantitative data.To achieve this, different models and technologies are needed.Some of the "new" human models (e.g., iPSCs) and technologies (e.g., omics and miRNA) might be useful tools, particularly when linked to epidemiological studies.However, the models need to be very well characterized.
There is a need to accelerate chemical testing in the assays that have been developed, make sense of the data from highcontent assays and link cellular effects to adverse outcomes.
Prediction and interpretation cannot be determined without screening more compounds.The major limitation of current DNT tests is still the same as at DNT-2 and -3, lack of data generation.
For this reference chemicals are needed, which is challenging as there is limited DNT in vivo data for most chemicals.After identification and testing of reference chemicals there should be a selection of tests to build a battery and eventually an ITS.To collect generated and existing data in a common database will be essential.Furthermore, a strong policy program that can influence funding organizations (both in EU and US) to support assay development and chemical screening is needed.Establishment of a DNT Secretariat including different stakeholders from various parts of the world was suggested as a path forward.
There seems to be a change in mindset of regulators, with a first step to make use of in vitro tests combined with revised in vivo test for risk assessment.However, it should be kept in mind that different regulatory bodies have different needs.It will be crucial to have them involved during test development.To demonstrate that in vitro tests are useful we need to generate more data, validate the tests and link the generated data to adverse outcomes in humans.

Methodological advances as a basis for a new approach for DNT
The Biotech Revolution has clearly fueled DNT testing: Significant advancements have revolutionized DNT testing, such as the development of organotypic cultures, the use of stem cells, and the application of high-content methods.These technologies have enabled the creation of more physiologically relevant in vitro models that can mimic key aspects of human neurodevelopment.Organotypic cultures, stem cells, and high-content methods have significantly contributed to the development of alternative testing methods for DNT by providing more human-relevant and mechanistically informative systems that can reduce reliance on animal testing (Groot et al., 2013).

Stem cells
The availability of human stem cells, particularly human-induced pluripotent stem cells (iPSCs), has revolutionized the field of developmental neurotoxicity (DNT) testing by providing new methodologies that are more relevant to human biology.iPSCs can be differentiated into various cell types of the central and peripheral nervous systems, enabling the modeling of different stages of brain development and the assessment of chemical toxicity during these stages (Yamada et al., 2019;Kobolak et al., 2020).
Gene-environment studies using patient cells -iPSCs facilitate the study of gene-environment interactions by using cells from patients with developmental disorders (Ilieva et al., 2018;Russo et al., 2019;Wegscheid et al., 2021;Villa et al., 2021;Santos et al., 2023;Kilpatrick et al., 2023).This approach provides insights into whether certain genetic backgrounds confer increased sensitivity to environmental stressors.This kind of research is crucial for understanding the complex interplay between genetic predispositions and environmental factors in the development of neurological disorders.
Gene engineering with risk genes: The advent of gene-editing technologies like CRISPR/Cas9 has allowed to introduce or correct mutations in iPSCs.This enables the study of specific risk genes in a controlled environment and the observation of their effects on neurodevelopment: For instance, our earlier study demonstrated a potential synergy between a mutation in the high-risk autism gene CHD8 and exposure to the organophosphate pesticide chlorpyrifos in an iPSC-derived 3D brain model (Modafferi et al., 2021).By engineering iPSCs with known risk genes for neurological conditions, scientists can dissect the pathways through which these genes contribute to disease and identify potential therapeutic targets.
"Living biopsy" and disease modeling: iPSCs can be considered a "living biopsy" of a patient's condition, as they capture the patient's genetic makeup and can be differentiated into disease-relevant cell types, providing a platform for studying disease mechanisms and potential treatments (Smirnova and Hartung, 2024).
Personalized medicine and toxicology: The use of iPSCs in DNT testing paves the way for personalized medicine and toxicology.By generating iPSCs from individual patients, it is possible to create personalized models of disease and predict individual responses to drugs and environmental toxicants (Fritsche et al., 2018).This approach could lead to more tailored and effective treatments with fewer side effects, as well as safer and more targeted drug development.
Challenges and future directions: Despite these advances, challenges remain in the use of human stem cells for DNT testing.These include the need for improved methods to differentiate iPSCs into fully mature and functional neurons and glia, the development of standardized protocols for toxicity testing, and the integration of these new methods into regulatory frameworks (Fritsche et al., 2018;Yamada et al., 2019).As the technology progresses, it is expected that human stem cell-based models will become increasingly important tools for assessing the safety of chemicals and drugs, ultimately reducing the reliance on animal testing and improving human health outcomes.

Organotypic cultures and microphysiological systems
Organotypic cultures are complex in vitro models that maintain or reconstruct the architecture and multi-cellular complexity of the original tissue.We see a revolutionary change (Hartung and Tsatsakis, 2021) at this moment in achieving relevant human cell culture.These cultures can, in our case, mimic key aspects of human neurodevelopment and are used to study basic biological processes such as differentiation, proliferation, migration, and neurite growth, which are fundamental to understanding DNT.By testing the disturbance of these biological activities by chemicals, we can identify potential neurotoxicants.The availability of stem cells and the bioengineering of so-called microphysiological systems (MPS) is an enormously important task.
We have organized three large workshops with opinion leaders from all over the world (Marx et al., 2016, 2020, andin preparation).The second one led to this Science paper on human MPS -microphysiological systems for drug development (Roth et al., 2021).Out of this, also, developed a series of conferences, the MPS World Summits, and the international MPS society.In 2022 in New Orleans, out of 655 registrants, 65 came from the FDA, showing the enormous resonance of this topic with the agency.430 people met there in person, which was tripled to 1300 people in Berlin and similar numbers are expected for Seattle in 2024.
Our own work led us to humanize brain organoids (Hogberg et al., 2013) published shortly after the first iPSC-derived model (Lancaster et al., 2013) of the first brain organoids.In 2016, at AAAS we finally showed how we can mass produce them (Pamies et al., 2017).Since then, we have been using this for a number of disease models.In respect to DNT, these models are spontaneously, electrophysiologically active and they include most of the brain cells which we know except microglia.All types of neurons we were looking for, astrocytes, oligodendrocytes, even in reasonable proportions, the standard model has about 40% glial cells and we just developed protocols to up them to physiological levels of 50% (Morales Pantoja et al., 2023b).We can also add microglia to the systems.One of the key features is myelination, with about 40% of the axons being myelinated Pamies et al., 2017;Chesnut et al., 2021a,b;Romero et al., 2023).One can see the beautiful structures of oligodendrocytes wrapping themselves around the axons, which is a quite unique feature because very few human models show myelination.We used this model for DNT, assessing pesticides such as rotenone (Pamies et al., 2018) and chlorpyrifos (Modafferi et al., 2021), but also the antidepressant Paroxetine (Zhong et al., 2020), which has been under debate for about 20 years as possibly contributing to neurodevelopmental disorders.The discussion was never closed, but we could show that clinically relevant concentrations disturb brain development in our experimental system.
One of the key ideas is that these types of models show all of the mechanisms which we have identified before as critical for DNT.Is this, perhaps, a possibility to replace the long battery of tests with something that combines all of these assays?Because we observe aspects of neurodifferentiation, myelination, neurite outgrowth, synaptogenesis, glial migration and gliosis, and also the neural network through electrophysiology in a single model, it appears possible to multiplex the different assays.The brain organoids are ideally suited for doing exactly this because many of the processes we are interested in are happening there.And they are undergoing at least a critical phase of development reflecting currently about five months of embryo development, possibly even more we only drive them towards this.So, we can probably observe in a single set-up a lot of these relevant mechanisms.This is more promising than developing a lot of assays, which no lab in the world can have all at the same time for the same substance.This is the starting point for a project which was funded by the EPA6 , announced when the EPA in 2019 decided to move out of animal testing by 20357 .It is trying to develop a mini "brainbow" by fluorescent reporter genes, introduced, engineered into these brain organoids so that, noninvasively, all of these mechanisms can be studied, at the same time (Romero et al., 2023), we were developing 3D electrophysiology around them, and this led to shell electrodes embracing brain organoids (Huang et al., 2022).
Another line of work is to use this for gene-environment interactions (Butera et al., 2023;Suciu et al., 2023a).Our hypothesis is that it is a susceptible genome that meets exposure at vulnerable times, because we cannot explain the enormous increase in autism just by genomics.There must be an environmental component.In Modafferi et al. (2021), we showed for the first time, such a pair: CHD8, a known risk gene for autism and chlorpyrifos, a substance with some liabilities at least at high concentrations and doses, synergize.And this led now to the creation of an NIH Autism Center of Excellence8 , where we are following in total in 18 partner centers 175,000 children in order to look for such gene-environment interactions and we are verifying experimentally as a second ongoing part these pairs in the brain organoid systems.
Challenges and future directions: Though more complex models have the potential to enhance DNT testing there are some challenges to consider before use in regulatory applications (Hogberg and Smirnova, 2022).The major limitations are lack of standardization of protocols and lower reproducibility.Moreover, the throughput of such models is still low compared to the simple methods previously developed.The culturing time is often extended and can be costly making such models, as of today, more useful as a follow up method for a small set of prioritized compounds identified in an initial screening approach.
In conclusion, organotypic cultures and MPS as complementary or orthogonal assays for the current screening battery of assays for DNT can support the translation of in vitro mechanistic effects to in vivo DNT outcomes.Once these models can demonstrate that they follow the readiness criteria for regulatory application (Bal-Price et al., 2018a) there is time to refine the current DNT battery.

Advantages of high-content screening in DNT testing
High-content screening (HCS) methods (van Vliet et al., 2014) have significantly transformed DNT testing by introducing a more nuanced, efficient, and comprehensive approach to evaluating the effects of chemicals on neural development.These methods leverage automated imaging and analysis to measure multiple biological parameters within cells, offering a profound leap in the ability to assess functional endpoints critical for DNT testing, such as neurite outgrowth and electrophysiology.
Enhanced efficiency and throughput: HCS methods enable the rapid and efficient processing of moderate to large numbers of chemicals, essential for the development of high-throughput DNT testing strategies.This capability is crucial given the vast number of chemicals in the environment that have not been tested for neurotoxicity due to the traditionally time-consuming and resource-intensive nature of DNT testing.
Multiparametric analysis: One of the key strengths of HCS is its ability to simultaneously measure multiple parameters within the same assay, providing a more comprehensive understanding of neurotoxic effects.This multiplexing capability allows for the assessment of various aspects of neuronal health, including cell viability, neurite outgrowth, synaptic function, and cellular signaling pathways, within a single experiment (Li and Xia, 2019).
Improved sensitivity and specificity: The automated nature of HCS, combined with advanced image analysis algorithms, enhances the sensitivity and specificity of DNT testing.High-content methods can detect subtle changes in neuronal morphology and function that may be indicative of neurotoxicity, even at low doses of chemicals (Schmuck et al., 2017;Persson and Homberg, 2016).This sensitivity is critical for identifying potential neurotoxicants that may not produce overt toxicity but could still have significant impacts on neural development and function.
Application to complex models: HCS methods are compatible with complex in vitro models, including 3D organoid cultures and human-induced pluripotent stem cell (iPSC)-derived neuronal models (Fritsche et al., 2018).These advanced models more closely mimic human neural development and disease, enhancing the relevance of DNT testing to human health.High-content methods allow for the detailed analysis of these complex models, providing insights into the mechanisms of neurotoxicity and the potential for developmental disorders (Schmuck et al., 2017).
Challenges and future directions: Despite the advantages of HCS in DNT testing, challenges remain.These include the need for standardized protocols and validation of HCS methods for regulatory acceptance.Additionally, the complexity of data generated by high-content methods requires sophisticated bioinformatics tools for analysis and interpretation (Li and Xia, 2019).The integration of HCS with emerging technologies, such as machine learning and artificial intelligence, holds promise for addressing these challenges.These technologies can enhance the analysis of complex datasets, improve the predictive power of DNT testing, and facilitate the identification of novel neurotoxicants (Schmuck et al., 2017).
In conclusion, high-content screening methods have revolutionized DNT testing by providing a more efficient, sensitive, and comprehensive approach to evaluating the neurotoxic potential of chemicals.As these methods continue to evolve and integrate with advanced computational tools, they will play an increasingly important role in protecting human health from the adverse effects of environmental chemicals.

Key characteristics of developmental neurotoxicants
The concept of "key characteristics"9 , properties of chemicals and other agents that confer potential hazard, was first developed for carcinogens and was based on properties of known human carcinogens as classified by the International Agency for Research on Cancer (IARC).These key characteristics of carcinogens were applied in the evaluation of diverse carcinogens and are now used as the basis for the evaluation of mechanistic data at IARC.Recently the key characteristics of male and female reproductive toxicants (Arzuaga et al., 2019;Luderer et al., 2019) and of endocrine disrupting chemicals (La Merrill et al., 2020;Cediel-Ulloa et al., 2022) have been described and those for other toxicant areas are in development.A (developmental) neurotoxicity working group10 met at UC Davis on September 17-18, 2019.Pamela Lein, UC Davis, and Thomas Hartung, Johns Hopkins, co-chaired the meeting which was hosted by Martyn Smith (UC Berkeley) and Lauren Zeise (OEHHA, CalEPA).The group examined the literature and developed 12 key characteristics (KCs), considering neurotoxicity both during development and in later life.They are refining the KCs in a series of follow-up teleconference calls and preparing a manuscript for publication.

Advances in artificial intelligence supporting DNT testing
Artificial Intelligence (AI) has significantly transformed developmental neurotoxicity (DNT) testing, offering innovative approaches to understanding and predicting the neurotoxic effects of chemicals and drugs.This transformation is evident in several key areas, including the enhancement of predictive models, the integration with complex biological systems, and the improvement of data analysis and interpretation.
Enhanced predictive models: AI, particularly machine learning (ML), has been instrumental in developing predictive models for DNT testing in the context of the ongoing ONTOX project (Vinken et al., 2021, see below).These models can analyze vast datasets, identifying patterns and relationships that may not be apparent through traditional analysis methods.For instance, a study demonstrated the use of machine learning to predict in vitro neurotoxicity induced by nanoparticles, highlighting the potential of non-testing approaches in hazard assessment (Furxhi and Murphy, 2020).By leveraging features such as exposure dose, duration, and cell type, AI models can provide a more nuanced understanding of neurotoxicity risks.
Integration with complex biological systems: AI's ability to handle complex, high-dimensional data makes it particularly suited to integrating with advanced biological systems used in DNT testing, such as human pluripotent stem cell-derived neural constructs (Schwartz et al., 2015).These systems can model human neurodevelopment more accurately than traditional animal models, but they generate large amounts of data that can be challenging to analyze.AI can process and interpret this data, identifying key indicators of neurotoxicity and enhancing the relevance of DNT testing to human health.We have discussed earlier the opportunities of modeling MPS by computational approaches (Smirnova et al., 2018).
Improvement of data analysis and interpretation: High-content screening methods, which generate large volumes of data on cellular responses to chemicals, have become a cornerstone of modern DNT testing (Shafer, 2019).AI algorithms can analyze these data efficiently, extracting meaningful insights on neurotoxic effects.For example, AI has been used to map drug-induced neuropathy through in-situ motor protein tracking, combining imaging data with machine learning for a more accurate assessment of neurotoxicity (Yi et al., 2021).
Challenges and future directions: Despite these advancements, challenges remain in the application of AI to DNT testing.These include the need for large, high-quality datasets for training AI models, the interpretation of AI-generated predictions, and the integration of AI approaches into regulatory frameworks (Fritsche et al., 2017).Addressing these challenges will require continued collaboration between toxicologists, data scientists, and regulatory bodies.
AI has the potential to revolutionize DNT testing by enhancing the predictive accuracy of neurotoxicity assessments, enabling the integration of complex biological data, and improving the efficiency of data analysis.As AI technologies continue to evolve, they will play an increasingly important role in identifying neurotoxic risks and protecting human health.
The EU ONTOX project (Vinken et al., 2021), funded by Horizon 2020, is at the forefront of leveraging AI for developmental neurotoxicity (DNT) testing (beside liver and kidney toxicity) and broader chemical risk assessments without the use of animal testing.By focusing on the development of non-animal new approach methodologies (NAMs) and probabilistic risk assessment (PRA), ONTOX aims to align with 21 st -century toxicity testing principles.A key aspect of the project involves addressing the acceptance and validation of AI in risk assessment, as highlighted during the first ONTOX Stakeholder Network Meeting held in March 2023 (Diemar et al., 2024).This meeting brought together various stakeholders, including regulatory authorities, companies, academia, and non-governmental organizations, to discuss the challenges and opportunities associated with implementing AI-driven NAMs and PRA.The discussions underscored the need for capacity building, sustainability, and regulatory acceptance of AI technologies in the context of ensuring consumer safety and advancing chemical risk assessment methodologies.The ON'OX project's efforts to integrate AI into DNT testing and chemical risk assessment represent a significant step towards reducing reliance on animal testing while enhancing the accuracy and efficiency of toxicity evaluations (Diemar et al., 2024).
In summary, these advanced biotechnological tools have enabled the creation of more relevant and efficient in vitro models for DNT testing.They have the potential to provide mechanistic insights into how chemicals affect neurodevelopment, which is crucial for the development of integrated testing strategies that can ultimately reduce the need for animal testing.

The rise of integrated testing strategies (ITS) aka IATA & defined approaches
The early discussions and recommendations for animal-free systemic toxicity testing laid the groundwork for the development of Integrated Testing Strategies (ITS).These strategies were envisioned to provide a more effective approximation of regulatory information needs than standalone assays.The consensus report on the future of animal-free systemic toxicity testing, which emerged from expert workshops convened by CAAT-Europe, outlined a general strategy for animal-free test approaches (Leist et al., 2014).This strategy was informed by the US National Research Council's vision for Toxicity Testing in the 21 st Century, published in 2007, which advocated for a shift towards more human-relevant, non-animal methods.
ITS are designed to integrate various information sources, including in vitro assays, in silico models, and human biomonitoring data, to predict the toxicity of substances.The integration of these diverse data sources is facilitated by computational tools such as Bayesian networks and machine learning, which can handle complex datasets and uncover patterns that may not be evident through traditional analysis.
Bayesian networks are probabilistic models that can combine data from different sources and account for uncertainties, providing a structured approach to integrating evidence and making predictions.Machine learning techniques, on the other hand, can analyze large datasets to identify features that are predictive of toxicity outcomes, improving the accuracy and efficiency of toxicity predictions.
The promise of ITS lies in their ability to approximate the information that would traditionally be obtained from animal testing, but in a more human-relevant and ethical manner.By leveraging emerging tools for data integration, ITS can provide a more comprehensive assessment of potential toxicants, taking into account various factors such as toxicokinetics, hazard testing, and the mapping of information along adverse outcome pathways (Leist et al., 2014).
The development of ITS represents a significant advancement in the field of toxicology, aiming to reduce the reliance on animal testing while still meeting the regulatory requirements for safety assessment.As these strategies continue to evolve, they are expected to become an integral part of the regulatory landscape, providing a more efficient and ethically responsible approach to toxicity testing.

The role of adverse outcome pathways (AOPs)
The concept of adverse outcome pathways (AOPs) (Willet, 2018) has become a cornerstone in the rational design of integrated testing strategies (ITS) for developmental neurotoxicity (DNT) testing.This framework is instrumental in identifying specific targets and mechanisms that are critical in the development of neurotoxic effects, thereby facilitating the development of testing strategies that are both more mechanistically informed and predictive of human health outcomes.
Importance of AOPs in ITS design for DNT testing: AOPs offer a structured approach to understanding the complex mechanisms underlying DNT, enabling researchers to pinpoint specific biological processes that can be targeted for testing.By mapping out the sequence of events that lead to adverse neurodevelopmental outcomes, AOPs help in identifying relevant biomarkers and endpoints that can be incorporated into ITS (Hernández-Jerez yet al., 2021;Willet, 2018).This mechanistic understanding is crucial for developing assays that are not only sensitive to specific neurotoxic effects but also relevant to human health.
Facilitating mechanistically informed testing strategies: The identification of key events and targets within DNT AOPs has significantly contributed to the development of more mechanistically informed testing strategies.For instance, the EFSA Panel on Plant Protection Products and their Residues developed AOP-informed integrated approaches to testing and assessment (IATA) case studies for the DNT hazard identification of pesticides like deltamethrin and flufenacet (Hernández-Jerez et al., 2021).By focusing on specific key events identified in the AOPs, such as alterations in neural proliferation, differentiation, and synaptogenesis, these strategies can more accurately predict the potential for chemicals to cause DNT.
Enhancing predictivity with emerging tools for data integration: The integration of emerging tools for data analysis, such as Bayesian networks and machine learning, has further enhanced the predictivity of ITS for DNT testing.Bayesian networks, for example, allow for the probabilistic quantification of the weight of evidence (WoE) across different data sources, including in vitro assays and in silico models, within the AOP framework.Machine learning algorithms can analyze complex datasets from highthroughput screening assays to identify patterns and predict outcomes based on identified key events (Zhang et al., 2023).These computational tools enable the integration of diverse data types, improving the ability of ITS to approximate regulatory information needs effectively.

Conclusion:
The development and application of AOPs in the design of ITS for DNT testing represent a significant advancement in the field of toxicology.The currently available AOPs for DNT, however, are relatively few: Data retrieved from AOP-Wiki11 on 24 Mar 2024 list eight AOPs (#6, #12, #13, #17, #31, #54, #499, #500) of which five are endorsed by OECD's WPHA/WNT12 .It will take a community effort to expand this, to combine it to a network (Pistollato et al., 2020;Sachana et al., 2021b;Spînu et al., 2022;Pitzer et al., 2023), and to make it the basis of a testing strategy.By providing a structured framework to understand the mechanistic basis of neurotoxicity, AOPs can facilitate the development of testing strategies that are not only more informed by the underlying biology but also more predictive of human health outcomes.The integration of advanced computational tools for data analysis further enhances the capacity of ITS to provide comprehensive and reliable assessments of potential neurotoxicants, moving towards more effective and human-relevant toxicological evaluations.

The development of new approaches for DNT as an example for the strategic development of alternatives to animal testing
The 2014 conference represented a turning point toward regulatory engagement.We have to highlight the involvement of regulatory agencies (EPA, EFSA, OECD) and the development of guidance documents that emerged from these collaborations.In parallel to embracing the technological developments, the concepts for tackling DNT developed further with this discussion moving more toward implementation and regulatory use (Bal-Price et al., 2018b).

Conceptual workshops further advancing DNT testing
The International Stakeholder NETwork, which took place in 2014 in Zurich, Switzerland, which was helping to define now how do we get all of these technical developments into some type of regulatory context.The workshop was sponsored by EPAA (The European Partnership for Alternatives to Animal Testing), CAAT and SCAHT (Swiss Centre for Applied Human Toxicology).The resulting article "International STakeholder NETwork (ISTNET): creating a developmental neurotoxicity (DNT) testing road map for regulatory purposes" by Bal-Price et al. discusses the challenges and advancements in the field of DNT testing, particularly the lack of toxicological hazard information for most compounds.The paper emphasizes the need for new approaches to generate experimental data that can inform regulatory decisions (Bal-Price et al., 2015b).
The ISTNET meeting focused on the concept of Adverse Outcome Pathways (AOPs) as a promising tool to promote the development of test systems aligned with regulatory needs.AOPs are considered crucial for assembling predictive ITS for DNT.The paper outlines a stepwise approach to AOP-based DNT testing, starting with incomplete AOPs for compound grouping and focusing on key events of neurodevelopment.The next steps include applying the AOP concept in regulatory DNT testing, using AOP intersections for economic development of screening assays, and transitioning from qualitative descriptions to quantitative network modeling.
The report also highlights the importance of communication and discussions between stakeholders -regulators, industry, and academia -to define a regulatory need-driven road map for an ITS for DNT.It acknowledges the limitations of current animalbased test methods due to high costs and the use of large numbers of animals, advocating for in vitro/in silico modeling approaches to provide value-added data for regulatory purposes.These approaches are expected to reduce animal testing, lower costs, and increase testing efficiency using high-throughput systems (HTS) to estimate environmental hazards to human health.
The ISTNET meeting concluded that alternative approaches such as in vitro test methods, quantitative structure-activity relationships (QSARs), read-across, and the application of the concepts of KNDP, AOP and toxicity endophenotypes could meet regulatory requirements for DNT testing.The main focus of the meeting was increasing the use of alternative data sources in DNT risk assessment and risk management decisions.
In summary, the article by Bal-Price et al. (2015b) advanced the concepts of DNT testing by promoting the use of AOPs to guide the development of ITSs, advocating for the integration of in vitro and in silico methods into regulatory frameworks, and emphasizing the need for a collaborative approach among stakeholders to address the challenges in DNT testing.The recommendations and discussions from the ISTNET meeting serve as a roadmap towards more efficient, predictive, and animalfree DNT testing strategies that are aligned with regulatory needs.
The workshop played a pivotal role in integrating EFSA and the OECD into the development of animal-free systemic toxicity testing strategies, including developmental neurotoxicity (DNT) testing.These discussions laid the groundwork for future activities aimed at integrating AOP-based ITSs into regulatory decision-making processes for chemical safety assessment.The contributions of EFSA and OECD to these activities have been significant, providing scientific expertise and regulatory perspectives that are essential for the acceptance and implementation of new, animal-free testing strategies.Their subsequent activities have helped to advance the field of DNT testing by promoting the development of more predictive, efficient, and ethically responsible testing strategies that better reflect human health outcomes: EFSA and the OECD have been actively involved in advancing the field of DNT testing, such as the development of non-animal test methods and integrated approaches to testing and assessment (IATA) for DNT.The OECD has coordinated international efforts to enhance DNT testing, acknowledging the limited historical use of in vivo DNT test guidelines.Workshops over the past decade have led to a consensus among stakeholders on the need for a DNT testing battery based on in vitro endpoints and alternative species assays.The OECD has initiated specific activities to facilitate this project, including collating available DNT in vitro methods, forming a DNT testing battery, generating a reference set of chemicals for testing, and developing an OECD guidance document including IATA cases studies (Sachana et al., 2019).These case studies aimed to assess the applicability of the DNT in vitro testing battery (IVB) (Blum et al., 2023) in regulatory risk assessment of pesticides.The approach included systematic literature reviews, expert knowledge elicitation, and Bayesian network analysis to integrate evidence within the AOP framework.
These activities by EFSA and OECD represent significant steps towards improving DNT testing by promoting the use of alternative methods and IATA frameworks in regulatory decision-making.The efforts aim to address the challenges of lacking DNT data for numerous chemicals and the difficulty in interpreting results from traditional animal-based tests.The goal is to develop more predictive, efficient, and ethically responsible DNT testing strategies that better reflect human health outcomes (Sachana et al., 2019;Hernández-Jerez et al., 2021).
Then, EFSA and the OECD came into the game, really bringing in the big cannons to move these things forward.The OECD/EFSA workshop on developmental neurotoxicity (DNT) in October 2016 (Fritsche et al., 2017) made significant progress towards advancing the use of alternative, non-animal test methods for regulatory purposes.Key points of consensus reached by the diverse group of international stakeholders included: 1.There is an urgent need for a DNT testing strategy using in vitro methods and alternative models to begin screening and prioritizing the large number of untested chemicals for their potential effects on the developing nervous system.2. A battery of currently available in vitro DNT assays, based on critical neurodevelopmental processes, is ready for use now for screening and prioritization purposes.Further development and standardization of the testing battery can enable its use for hazard assessment and to support risk management decisions in the future.3. A roadmap should be established to define procedures and milestones for implementing this new approach to DNT testing.
Priorities include chemical testing to build confidence in the testing battery, establishing performance standards, and developing an OECD guidance document on an integrated approach to DNT testing and assessment.The OECD project was summarized by Sachana et al. (2019Sachana et al. ( , 2021a)).Key steps in this project include: 1) Generating a reference set of chemicals to test a battery of in vitro DNT assays spanning key neurodevelopmental processes; 2) Selecting the battery of in vitro DNT assays based on readiness criteria; 3) Testing the reference chemicals in the battery to build confidence in the alternative approaches; 4) Developing integrated approach to testing and assessment (IATA) case studies using DNT in vitro battery data for different regulatory applications; and 5) Drafting an OECD guidance document on the use of alternative DNT testing methods within an IATA framework.The ultimate goal is to accelerate the development, standardization, and regulatory acceptance of alternative testing strategies that can more rapidly and cost-effectively screen chemicals for their potential to cause DNT.
The review by Schmidt et al. (2016) presents a comprehensive overview of the progress and technical possibilities in in vitro neurotoxicity and developmental neurotoxicity (DNT) screening, emphasizing the urgent need for alternative testing methods due to challenges in extrapolating animal data to humans and the limited capacity of animal testing to cover all substances requiring evaluation.The paper discusses various cellular platforms used in neurotoxicity testing, ranging from animal and human cell lines to advanced human-induced pluripotent stem cells (hiPSC) and organ-on-a-chip models, highlighting their respective advantages and limitations.It covers common endpoints of neurotoxicity testing, including cell viability, neurite outgrowth, synaptic function, and electrophysiological properties, among others.Furthermore, it explores analytical methods such as high-content imaging and electrophysiological screens to assess these endpoints.The review underscores the importance of integrating multiple cellular models and endpoints to capture the complexity of neurotoxic effects and advocates for the development of integrated testing strategies and multi-omics approaches to improve prediction models for assessing chemical hazard potential.This work contributes significantly to the field by outlining the current state, challenges, and future directions for in vitro neurotoxicity and DNT screening, promoting the shift towards more human-relevant, efficient, and ethical testing methods.Recognizing that future testing approaches will likely involve a battery of alternative and complementary tests, the paper focuses on the first generation of alternative DNT tests that target fundamental neurodevelopmental processes, such as neuronal differentiation, precursor cell migration, or neuronal network formation.These processes are crucial as they capture toxicants with diverse targets and modes of action and can be linked to toxicity endophenotypes, which are alterations in neural connectivity leading to neurofunctional deficits in humans.The workshop that led to this review defined criteria for selecting positive/negative controls, prepared recommendations for their use, and initiated the setup of a directory of reference chemicals.The workshop convened to define criteria for selecting positive/negative controls and to initiate a directory of reference chemicals.Over 50 endpoint-specific control compounds were identified for initial technical optimization of tests, and an additional set of 33 chemicals, considered direct DNT toxicants, was proposed for further test development.The paper emphasizes the importance of AOPs in regulatory DNT testing and the use of over 100 compiled compounds to address specificity, adversity, and the use of alternative test systems.This structured approach aims to enhance the development of high-throughput screening methods and improve the predictivity of DNT testing methodologies (Aschner et al., 2017).
We next assessed test readiness in a subsequent workshop and developed a list of reference compounds: Bal-Price et al. (2018a) presents a major conceptual advance in providing a framework to evaluate the readiness of new approach methods (NAMs) for developmental neurotoxicity (DNT) testing for regulatory purposes.The authors propose a set of readiness criteria and a scoring system to assess the readiness of individual DNT in vitro assays as well as the overall DNT in vitro battery (DNT IVB) for various regulatory applications.This allows a quantitative assessment of the current status of assay development and what further work is needed to increase regulatory confidence in these alternative methods.The approach outlined here provides a role model for evaluating alternative methods in other fields of toxicology.Key elements include: 1) defining readiness criteria based on the regulatory need and context of use, 2) quantitative scoring of assay readiness based on multiple defined criteria, 3) evaluating the battery of assays as a whole in terms of biological coverage, predictive performance, and overall readiness for different regulatory uses, and 4) using case studies to demonstrate how the alternative methods can be applied in an integrated approach to testing and assessment (IATA).This framework allows a transparent and objective evaluation of NAMs to facilitate their regulatory acceptance and use.While more work is needed to fully validate the DNT IVB (Juberg et al., 2023), the strategy presented here is an important step forward in advancing the use of alternative methods for safety testing.
The US EPA held a public meeting of the FIFRA Scientific Advisory Panel on 15-17 Sep 2020 to review the use of new approach methodologies to derive extrapolation factors and evaluate developmental neurotoxicity for human health risk assessment 13 .Key points include14 : 1.The Panel reviewed several in vitro NAM assays developed by EPA and European researchers to evaluate important neurodevelopmental processes that may be disrupted by chemical exposure.While finding strengths in the assays, the Panel noted limitations such as the lack of important cell types, absence of functional/mechanistic assessments, and difficulties extrapolating to in vivo effects.2. The Panel considered the use of in vitro to in vivo extrapolation (IVIVE) approaches to compare NAM assay results to doses causing acetylcholinesterase inhibition in animals for organophosphate pesticides.They found the IVIVE approach reasonable but raised concerns about model assumptions and performance.3. The Panel reviewed analyses deriving interspecies and intraspecies extrapolation factors from in vitro acetylcholinesterase inhibition data in rat and human tissues.Limitations were noted in the analysis methods, sample representativeness, and sample sizes to characterize human variability.4. Overall, the Panel saw value in the NAM approaches while providing numerous recommendations to address limitations and uncertainties in using the data for human health risk assessment.This evaluation represents an important starting point for the regulatory implementation of in vitro DNT approaches for the most important use case of agrochemicals.Their specific concerns 14 represent an important agenda for future developments: a.The absence of hormonal factors (sex hormones, thyroid, stress hormones) b.The influence of neurotransmitter signaling c.The influence of chemical-induced systemic changes (e.g., inflammation, oxygen levels and distribution) d.The influence of maternal factors (maternal infection, hormonal, organ system dysfunction, placenta integrity) In addition, the in vitro assays: a.Will be limited in their ability to detect adaptive or compensatory processes b.Do not account for critical cell-cell interactions required during neurodevelopment c.Have difficulty distinguishing between neuroactive and neurotoxic compounds d.Do not reflect human genetic diversity when using human cell lines from one human

The impact of OECD guidelines and the guidance document
An OECD expert group for in vitro DNT testing was created in 2017.The outcome of this was the proposal to develop, in essence, an integrated approach to testing and assessment, IATA.The OECD calls it an integrated approach to testing and assessment -we recently called it invivitrosiin vitro, in vivo, and in silico combined (Caloni et al., 2022).The development and adoption of OECD Test Guideline 426 and the extended one-generation reproductive toxicity test guideline (TG 443) were important steps in the evolution of DNT testing.They have incorporated improvements and recommendations from expert consultations, addressing critical issues in DNT testing and paving the way for the inclusion of alternative methods.
An important contribution was a report commissioned by EFSA: Crofton and Mundy (2021) provide guidance on the interpretation and use of data generated from assays currently included in the Developmental Neurotoxicity In Vitro Battery (DNT IVB) for regulatory purposes.Their report "External Scientific Report on the Interpretation of Data from the Developmental Neurotoxicity In Vitro Testing Assays for Use in Integrated Approaches for Testing and Assessment" has the following key points: 1.The DNT IVB includes a battery of in vitro assays that cover key neurodevelopmental processes such as proliferation, migration, differentiation, neurite outgrowth, synaptogenesis and neural network formation.2. These assays are considered key events at the cellular level that are plausibly related to modes of action of developmental neurotoxicants in vivo.3. When interpreting data from individual assays, it's important to evaluate the biological relevance of the test system, assay quality and reproducibility, testing with a training set of chemicals, data analysis methods, and use of a decision model to classify chemicals.4. To evaluate the DNT IVB as a whole, factors to consider include the predictive power compared to in vivo data, consistency of results across the battery, comparison of potency to other in vitro endpoints, and mapping to outcome pathways.5. Use of DNT IVB data should be guided by the consistency of the in vitro data, biological plausibility, incorporation of in vitro to in vivo extrapolation models, and consideration of uncertainties in the context of the regulatory need.6.While not a full replacement for animal studies, DNT IVB data is already being used to inform screening, prioritization, and weight of evidence approaches in different regulatory applications.Further work is needed to develop standardized reference chemical lists, data analysis pipelines, and tiered testing strategies.

From screening hits to potential DNT toxicants
Screening takes an increasingly larger role in DNT, driven, e.g., by large screening and prioritization programs by the US EPA or the NTP/NIEHS, and by the availability of more, better, and more robust high throughput assays in cells, organoids and simple organisms.Screening is a scientific discipline developed in drug discovery, and a lot of experience has been collected that was incorporated in a screening culture and best practices within the discipline of pharmacology.In toxicology, the technology has been adopted and adapted, but the culture is lacking behind.Often, the distinction between a screen hit and a potential toxicant is not clear and needs further clarification for efficient use of this approach for DNT testing.The pharmacological terminology is as follows: screens produce "preliminary positives" in the respective assay.After a confirmation under more controlled conditions, they can be called "confirmed positives".They need to be filtered, e.g., for technical artefacts, or secondary effects due to cytotoxicity.In many pharmacological screens this eliminates 90% of the screen, and only few "true positives" remain.These survivors are usually further tested in secondary assays.If they survive this step, then one can talk of potentially interesting assay hits and one can setup the hypothesis that these are potential toxicants.(see Fig. 3).This hypothesis then needs to be further evaluated to call compounds "relevant toxicological hits".The evaluation is a process that either disproves the hypothesis or gathers evidence that increases the plausibility of the hypothesis to be true.Important factors in this are similarity to other well-known toxicants or a known mechanism that feeds into an AOP, in addition to toxicokinetic behavior likely to lead to target occupancy at relevant exposures.In the absence of a strong structural similarity argument or mechanistic plausibility, the relevance is often hard to evaluate and requires extremely relevant and predictive DNT assays.Exemplary publications of this process are based on a neural crest migration assay and polychlorinated biphenyls as preliminary hits (Nyffeler et al., 2018) or a neurite outgrowth assay (Krug et al., 2013b) and berberine as hit compound (Suciu et al., 2023b).A very recent example is a screen of 1800 ToxCast compounds for toxicity to developing oligodendrocytes.The extensive hit characterization and toxicological follow-up was necessary to profile, e.g., quaternary amine based amphiphilic compounds as new class of potential DNT toxicants (Cohn et al., 2024).

Challenges and future directions
One of the primary challenges is the need for further validation of alternative DNT methods.While a growing number of in vitro and in silico models have been developed and shown to be promising predictors of DNT, most of these methods have not yet undergone the rigorous validation process required for regulatory acceptance.Traditional validation involves demonstrating the reproducibility, predictivity and relevance of a test method for its intended purpose, typically through a series of inter-laboratory studies using a standardized protocol and a diverse sets of reference chemicals.This process can be time-consuming and resourceintensive, requiring collaboration among multiple stakeholders, including test method developers, validation bodies, and regulatory agencies.This concept of validation is outdated, as it was developed for simple endpoints and for methods that by themselves (i.e., 1:1) predict such endpoints.For methods that are part of a test battery, the principles of relevance and predictivity require a redefinition, and the use of reference sets of chemicals is only of limited use.For DNT, the overall number of reference chemicals is too small for a classical relevance approach with a statistical prediction model (Mundy et al., 2015).More importantly, also the few reference chemicals with known DNT effects in humans or animals are mostly not well characterized for their mechanism and their mode of action.They can therefore hardly be used for mechanistic assays or KNDP assays.
To facilitate the validation of alternative DNT methods, there is a need for a clear and consensus-based framework that outlines the key steps and criteria for establishing the scientific and regulatory validity of these methods.This framework should take into account the specific challenges and opportunities associated with DNT testing, such as the complexity of neurodevelopmental processes, the importance of considering species differences (Baumann et al., 2016) and exposure timing, and Fig. 3: How to move from screening results to toxicological hits Screens, if run by high throughput technology, often generate "preliminary positives" that need to be further qualified and characterized.A first step involves repetition of the test, possibly with a more stringent prediction model, more stringent laboratory procedure controls (e.g. according to the OECD good in vitro laboratory practice (GIVIMP) guidance document) and with new, well-controlled compound stock solutions.The true positives (TP) are then obtained after elimination of technical artefacts (e.g.fluorescent compounds, or compounds sticking unspecifically to proteins), control for cytotoxicity and cheminformatics filtering for known pan-assay interfering compounds (PAINS).Certainty on the bioactivity of such compounds is then obtained classically in secondary and tertiary assays for similar target pathways or structures.These are ideally orthogonal in the sense, that they use other test systems and/or readout technologies.In a second phase, such "convincing" assay hits are evaluated for their toxicological relevance.Strategy I tests the relevance by establishing the similarity of the assay hit to known, well-characterized toxicants.The definition of similarity relies on (i) structural similarity, (ii) similar ADME properties, and (iii) a similar bioactivity profile/mode-of-action (MoA).Strategy II tests the relevance based on the expected MoA or the activation of a reliable adverse outcome pathway (AOP).Essential issues are (i) whether the concentrations triggering an AOP molecular initiation event (MIE) or key event (KE) are realistically reached, whether it is likely that the triggered AOP also continues to the adverse outcome.A more comprehensive investigation would use quantitative in vitro-to-in vivo (IVIVE) extrapolation models and it would consider whether key event relationships (KER) are affected by modifiers or counter-regulations.
the need for a battery of complementary assays that cover different modes of action.The recently updated document by the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM), "Validation, Qualification, and Regulatory Acceptance of New Approach Methodologies" 15 describes a more flexible, fit-for-purpose validation strategy with emphasis on integrating results from multiple NAMs rather than one-to-one replacement.DNT is described as one endpoint example where such a strategy will be important and hopefully speed up the process.Efforts to implement such a framework are ongoing, as exemplified by the work of the OECD's DNT Expert Group and the collaborative workshops organized by CAAT, ICCVAM, ECVAM, and other stakeholders.Possibly, the concept needs to be even more radically revised for some assays, i.e., not relying at all on classical reference chemicals but rather on mechanistic validation concepts.
Another challenge is the integration of new technologies, such as high-throughput screening (HTS), -omics approaches, and computational modeling, into regulatory DNT testing strategies.While these technologies offer the potential for more rapid, efficient, and mechanistically informative testing, their application in a regulatory context raises several issues related to data quality, interpretability, and relevance.For example, HTS assays may generate large amounts of data on the effects of chemicals on specific molecular or cellular endpoints, but the relationship between these effects and adverse outcomes in the developing brain may not always be clear.Similarly, -omics approaches, such as transcriptomics (Krug et al., 2013a;Rempel et al., 2015) and metabolomics (Modafferi et al., 2021), can provide a wealth of information on the biological pathways and processes perturbed by chemical exposure, but translating this information into actionable regulatory decisions may require additional steps, such as the development of quantitative adverse outcome pathways (AOPs) and the establishment of thresholds for adverse effects.
To address these challenges, there is a need for ongoing dialogue and collaboration between the developers of new technologies, the regulatory agencies responsible for implementing DNT testing requirements, and the broader scientific community.This dialogue should aim to identify the key data needs and decision-making criteria for regulatory DNT testing, and to develop a roadmap for the progressive integration of alternative methods and new technologies into regulatory frameworks.This may involve the development of tiered testing strategies that combine in vitro, in silico, and in vivo approaches in a way that balances the need for efficiency and predictivity with the requirements for regulatory certainty and human health protection.
Looking ahead, there are also significant opportunities for improving the predictivity and efficiency of DNT testing through advancements in data integration and machine learning.One promising approach is the development of integrated testing strategies (ITS) that combine multiple types of data, such as in vitro assay results, physicochemical properties, and in silico predictions, to provide a more comprehensive and reliable assessment of DNT potential.By leveraging the strengths of different data streams and taking into account the uncertainties and limitations associated with each, ITS can potentially provide a more accurate and efficient means of prioritizing chemicals for further testing and informing regulatory decisions.
The development of ITS for DNT testing can be facilitated by advances in machine learning and artificial intelligence (Kleinstreuer and Hartung, 2024), which offer powerful tools for integrating and analyzing large and complex datasets.Machine learning algorithms, such as random forests, support vector machines, and deep neural networks, can be trained on existing DNT data to identify patterns and relationships that may not be apparent from traditional statistical analyses.These models can then be used to predict the DNT potential of new chemicals based on their structural and biological similarity to known toxicants, or to identify the most informative combinations of assays and endpoints for a given regulatory context.
As more data become available from alternative DNT testing methods and other sources, such as high-throughput screening programs and academic research, there will be increasing opportunities to refine and optimize ITS using machine learning approaches.This could involve the continuous updating of predictive models as new data are generated, the identification of novel biomarkers and endpoints that are more predictive of adverse outcomes, and the development of more sophisticated decision-making frameworks that take into account multiple lines of evidence and sources of uncertainty.
To fully realize these opportunities, however, there is a need for greater standardization and harmonization of DNT testing methods and data reporting practices.This includes the development of common ontologies and data formats for describing neurodevelopmental processes and endpoints, the establishment of minimum reporting standards for in vitro and in silico DNT studies, and the creation of centralized and publicly accessible databases for storing and sharing DNT data.Efforts to promote such standardization and harmonization are already underway, as exemplified by the work of the OECD's Extended Advisory Group on Molecular Screening and Toxicogenomics (EAGMST) 16 and the collaborative projects funded by the European Union's Horizon 2020 research and innovation program.
In conclusion, while significant progress has been made in developing alternative approaches for DNT testing, there remain important challenges to be addressed in terms of method validation, regulatory acceptance, and data integration.At the same time, there are exciting opportunities for leveraging advances in machine learning and artificial intelligence to improve the predictivity and efficiency of DNT testing, and to support the development of more effective and scientifically sound regulatory strategies for protecting human health.By working together to address these challenges and opportunities, the scientific and regulatory communities can help to ensure that the promise of alternative DNT testing is fully realized, and that the safety of chemicals is assessed in a way that is both rigorous and efficient, and that reflects the best available science.

Implementation
Implementing advanced testing strategies for DNT in regulatory and research contexts presents several challenges.While ITS offer a promising approach to approximate regulatory information needs more effectively, their adoption is hindered by various factors.One of the main challenges is the complexity of DNT itself, with a multitude of key events and targets that need to be covered by the building blocks of an ITS.The design of ITS is ideally based on established mechanisms of health effects, such as adverse outcome pathways (AOPs), but the small number of studies leading to regulatory decisions in DNT limits this approach.
Moreover, the integration of emerging tools for data integration, including Bayesian networks, machine learning tools, and sensitivity analysis, is essential for the continuous optimization of ITS.However, the practical application of these tools in a regulatory context is still in its infancy, and there is a need for further development and validation of these methods.Additionally, the current regulatory frameworks may not be fully equipped to accommodate the novel approaches that ITS and AOPs represent, requiring updates and adaptations to existing guidelines and practices.
The challenges extend to the quality assurance of non-animal tests, where Good Cell Culture Practices (GCCP) (Pamies et al., 2022) and the concept of "mechanistic validation" (Hartung et al., 2013b) are critical to ensure the reliability of the results.However, establishing these quality standards and gaining widespread acceptance can be a slow and complex process.Furthermore, the interpretation of DNT study results requires a substantial amount of expertise, and there is considerable flexibility in the study design, which introduces potential sources of variability.
In summary, while the strategic development of pathway-based approaches to DNT testing is underway, the implementation of these advanced strategies faces challenges related to the complexity of DNT, the need for further development and validation of data integration tools, the adaptation of regulatory frameworks, and the establishment of quality standards for non-animal tests.number of chemicals in commerce that have not been adequately tested for DNT potential, and the growing recognition of the importance of protecting the developing brain from environmental chemical exposures.By enabling the prioritization of chemicals for further testing and providing a more comprehensive and reliable assessment of DNT potential, alternative methods can help to ensure that regulatory decisions are based on the best available science and are protective of human health.
The development of alternative methods for DNT testing has been a collaborative effort, involving contributions from researchers in academia, government, and industry, as well as from regulatory agencies and other stakeholders.This collaboration has been essential for ensuring that the methods developed are not only scientifically valid and relevant, but also practical and implementable within a regulatory context.The establishment of networks and forums for the exchange of knowledge and ideas, such as the DNT Workgroup and the series of DNT conferences organized by CAAT and its partners, has been crucial for fostering this collaboration and driving the field forward.
Looking back over the past 18 years, the field of DNT testing has made remarkable progress.In 2006, when the first DNT conference was held, the field was heavily reliant on animal-based testing methods, with few alternatives available.The discussions at that conference focused on the need for more rapid and efficient testing methods, and the potential for in vitro and non-mammalian models to fill this need.Over the subsequent years, a range of alternative methods were developed and evaluated, culminating in the publication of a comprehensive review by Bal-Price et al. in 2012 that identified a number of promising in vitro and alternative animal models for DNT testing.
The period from 2014 to the present has seen a rapid acceleration in the development and application of alternative methods for DNT testing, driven in part by advances in stem cell biology, genome editing, high-throughput screening, and computational modeling.The establishment of the OECD DNT Expert Group in 2017 and the publication of the OECD's "Initial Recommendations on Evaluation of Data from the Developmental Neurotoxicity (DNT) In Vitro Testing Battery" 17 in 2023 marked important milestones in the regulatory acceptance of alternative methods.
Despite these advances, however, much work remains to be done to fully realize the potential of alternative methods for DNT testing.Further research is needed to optimize and validate these methods, to demonstrate their reliability and relevance for regulatory decision-making, and to integrate them into testing strategies that provide a comprehensive and efficient assessment of DNT potential.Continued collaboration among researchers, regulators, and industry stakeholders will be essential for addressing these challenges and advancing the field towards a more predictive and ethical testing paradigm.
In conclusion, the strategic development of alternatives to animal testing for DNT over the past 18 years represents a major advancement in the field of toxicology, one that has been driven by scientific innovation, regulatory need, and a commitment to more human-relevant and ethical testing approaches.While significant challenges remain, the progress made to date is a testament to the dedication and collaboration of the many stakeholders involved in this effort.By continuing to work together to refine and validate alternative methods for DNT testing, we can move towards a future in which the safety of chemicals is assessed in a way that is both scientifically rigorous and ethically sound, and that truly protects the health of our most vulnerable populations.What we have seen over the last two decades is a scientific revolution.For the first time, an example of a strategic development by a community led to the development of testing strategy which is completely novel in its approach.It is a testing strategy composed of different components.It is mechanism based, it has been a consensus process internationally, these are all things we should apply to more areas of animal experimentation.The combination of these disruptive technologies will lead us to far more than just replacing animal tests.We can do things we did not dream of when we started this 20 years ago.The challenge now is implementation!

Fig. 2 :
Fig. 2: Timeline of international DNT meetings : While the TG 416 in REACH registrations used on average 2,590 animals (3,098 with dose range finding studies (DRF)), TG 443 used 1,318 (1,826 with DRF) and with second generation 2,226 (2,734 with DRF).Test duration of TG 443 compared to TG 416 is certainly an advantage as long as the second generation is not triggered.However, the overall feasibility of animal testing as a strategy for DNT remains a challenge.What is lacking is a gap analysis showing what the in vivo tests are not getting.What are the substances we are not picking up and what are the mechanisms in humans not reflected in the animal model?This information can be used as a point of comparison to assess what alternate tests can replace or improve.The challenges posed by the lack of data for almost all chemicals and the prohibitive costs and limitations of animal-based testing prompted a strategic development to find new approaches.However, the initial absence of alternatives to animal testing and the urgent need for a paradigm shift towards more efficient, cost-effective, and human-relevant methods posed an enormous challenge.
The first session explained the adverse outcome pathway (AOP) framework and gave examples on how it can be used for DNT assessment.The presentation of the keynote speaker Dr Kevin Crofton (US EPA, presented by Dr William Mundy, US EPA) described the critical importance of the adverse outcome pathway (AOP) framework to link DNT research to regulatory needs.Dr Anna Price, European Commission, further explained the concept of the AOPs applied to DNT evaluation by reporting on a AOPs workshop organized by the European Commission.Next two presenters Dr Ellen Fritsche, IUF -Leibniz Research Institute of Environmental Medicine, and Dr Pamela Lein, University of California Davis, gave some practical examples how in vitro and epidemiological data can inform AOPs.
The workshop and resulting paper "Reference Compounds for Alternative Test Methods to Indicate Developmental Neurotoxicity (DNT) Potential of Chemicals: Example Lists and Criteria for their Selection and Use" by Aschner et al. addresses the critical gap in information regarding the developmental neurotoxicity hazard posed by industrial and environmental chemicals.