Big Question to Developing Solutions: A Decade of Progress in the Development of Aquatic New Approach Methodologies from 2012 to 2022

Abstract

In 2012, 20 key questions related to hazard and exposure assessment, and environmental and health risks of pharmaceuticals and personal care products (PPCPs) in the natural environment were identified. A decade later, this article examines the current level of knowledge around one of the lowest ranking questions at that time, number 19: “Can non–animal testing methods be developed that will provide equivalent or better hazard data compared with current in vivo methods?”. The inclusion of alternative methods that replace, reduce, or refine animal testing (the 3Rs) within the regulatory context of risk and hazard assessment of chemicals generally faces many hurdles, although this varies both by organism (human centric versus other), sector and geographical region or country. Focusing on the past 10 years, only works that might reasonably be considered to contribute to advancements in the field of aquatic environmental risk assessment, are highlighted. Particular attention is paid to methods of contemporary interest and importance, representing progress in a) the development of methods which provide equivalent or better data compared with current in vivo methods such as bioaccumulation, b) weight of evidence, or c) -omic based applications. Evolution and convergence of these risk assessment areas offer the basis for fundamental frame shifts in how data is collated and used for the protection of taxa across the breadth of the aquatic environment. Looking to the future, we are at a tipping point, with a need for a global and inclusive approach to establish consensus. Bringing together these methods (both new and old) for regulatory assessment and decision making will require a concerted effort and orchestration.

INTRODUCTION

Increased awareness for sustainability needs

In 2012, twenty key questions related to hazard and exposure assessment, and environmental and health risks of pharmaceuticals and personal care products (PPCPs) in the natural environment were identified (Boxall et al., 2012). A decade later, this article examines the current level of knowledge around one of the lower ranking questions at that time, number 19 (out of 20): “Can non-animal testing methods be developed that will provide equivalent or better hazard data compared with current in vivo methods?”. The increase in production and diversification of synthetic chemicals poses a global challenge due to complex human and environment exposure scenarios, coupled with a lack of toxicity data for the majority of chemicals in the environment. While ranked low at the time, today the need for non-animal-based methods is widely recognized as being essential for all three pillars of sustainability (economic, environmental, and social). Non-animal-methods (NAMs) offer potential economic and ethical opportunities for ‘greener’ chemicals and new assessment tools with high throughput testing services, allowing for a safer environment with broad societal support. This promise has heralded shifts at the top levels of government and industry to reduce reliance on in vivo animal testing for risk and safety assessment purposes represented by an increasing number of groups supporting and working on non-animal-based methods (Table S1).

Human and environmental risk assessment share similar challenges

Early on, efforts to reduce reliance on in vivo animal testing were largely focused on developing new methods to replace individual tests or at least to find ways to refine or reduce the number of animals used in a laboratory study. However, in the past 10 years the language around non-animal testing methods has changed to a broader multi-method-based approach now collectively referred to as a New Approach Methodologies (NAMs), encompassing the overall process rather than individual non-animal methods. Chemical risk assessment (RA) and management was established as a scientific field over 40 years ago, with principles and methods developed on how to conceptualize, assess, and manage risk. Historically, human risk assessment (HRA) and environmental risk assessment (ERA) developed independently within specific regulatory chemical sectors (chemicals, biocides, pharmaceuticals etc.), resulting in the use of different terminology, separate databases, and varying regional requirements. Traditionally, HRA includes an assessment of possible exposure pathways, kinetics including the potential for bioaccumulation within the organism, sensitive organs, mode of actions and no effect levels. Conceptually similar, ERA encompasses an assessment of exposure pathways and the fate (i.e. kinetics) of a substance within the environment (surface water, sewage treatment plants, soil, sediment, and groundwater), including its persistence and bioaccumulation in addition to its impact on numerous organisms. However, unlike HRA, toxicity to multiple organisms (aquatic, terrestrial and microorganisms) is examined and no-effect concentrations in the more sensitive organisms are identified. Both HRA and ERA, require extrapolation of results from one or a few experimental species within their specific artificial environments to define protection levels for other relevant organisms, be it humans or various environmental. Notably, HRA aims to provide specific information on a multitude of human organ systems, mode of action and their interaction for humans with their variable genetics and lifestyle, while ERA aims to provide specific information on a multitude of organisms and their interactions within their variable environment (Figure 1). Extrapolation can be carried out pragmatically using standard assessment factors or adaptations thereof, or more scientifically, engaging probabilistic data-based extrapolation models. There is also growing recognition that comprehensive and reliable identification of similarities and differences between organisms will enhance cross-species extrapolation of potential adverse toxicological effects (Rivetti et al., 2020), with international consortiums established to start to address the challenges in extrapolating knowledge across classes (e.g. International Consortium for Advancing Cross-Species Extrapolation in Regulation [ICACSER]).

Figure 1:

Outline of conceptual similarities between human health risk assessment (HRA) and ecological risk assessment (ERA): Risk assessment require tasks which are conceptually similar for HRA and ERA. Current regulatory approaches are based on a battery of animal tests for HRA and ERA, which are slowly being replaced by new approach methodologies (NAMs). Challenges for extrapolation from the testing models to reality relate for HRA to the level of detail for the multitude of modes of action, organs, and interactions (all used for globally harmonized system of classification and labeling of chemicals), as well as human variability. For ERA, target organs of environmental organisms are not of immediate interest for regulators with the extrapolation challenges relating to the identification of an overall low-/no-observed- effect concentration for the multitude of organisms, populations, their interactions, and environmental variability. Importantly, mode of action* information in ERA currently is critical only for the identification of endocrine disruption. Despite these conceptual similarities, availability of NAMs and related guidance is much limited for ERA compared with HRA. OECD = Organization for Economic Co-operation and Development.

The use of animals in HRA and ERA is still increasing, despite growing awareness of the need to reduce reliance on animal testing, in line with the “reduction” and “replacement” aspects of the 3Rs principles (with the third R being refinement, centered on minimizing the pain, suffering, distress or lasting harm that research animals might experience) first proposed over 60 years ago. In Europe, the total number of animals used in testing in REACH testing (https://bit.ly/3xF4O4A) has essentially doubled in four years, from 1.1 million as reported in 2016 to 2.4 million animals (https://bit.ly/3xsXxUH). This trend is likely to continue for REACH given the imminent addition of testing requirements for endocrine disruptor identification and similar changes are expected in other regions which may further increase animal use.

A major challenge in terms of the use of laboratory animals arises from the need to evaluate bioaccumulation potential, which may lead to chronic outcomes not necessarily revealed in current regulatory toxicity testing. Bioaccumulation generally relies on determining the bioconcentration factor (BCF) as the sole decisive metric, with the recognition that slow metabolism can result in potentially higher bioaccumulation, with implications for both the environment and human health. It is important to highlight that there are more complex metrics which can also be calculated including bioconcentration factor (BCF; aqueous exposure routes), the biomagnification factor (BMF; dietary exposure route), bioaccumulation factor (BAF; all possible exposure routes) or trophic magnification factors (TMF) derived from mesocosm studies. While standard test protocols are available for the determination of BCF and BMF’s under well-defined laboratory conditions, a BCF is typically required in risk assessment to estimate concentrations in prey for the investigation of risks from secondary poisoning.

For example, the current European Medicines Agency (EMA) guideline for environmental assessment of human medicinal products requires a fish bioconcentration test (OECD TG 305) in Phase I for Persistence, Bioaccumulation and Toxicity (PBT) screening of drug substances with a log Kow > 4.5 and in Phase II for those with a log Kow > 3. It is important to note this guideline is currently under revision, however, as per the log Kow criteria in the guideline currently in effect, an animal study is required. While the test can use hundreds of fish per chemical, efforts to reduce organisms are reflected in the current guidelines. Specifically, the option to use a ‘minimized’ test requiring fewer organisms to estimate kinetic BCF using fewer sampling time-points, provided uptake and depuration are expected to follow first-order kinetics. In addition, a single test concentration may be used in the full or minimized test design when it is likely that the BCF is independent of the test concentration (Burden et al., 2017).

Integration of NAMs in HRA and ERA, at disparate pace

Significant efforts to reduce the use of animals via technological, computational, and scientific advances have given rise to new approach methodologies or non-animal methods, both of which have been used interchangeably in the literature (NAMs) and specifically reference any non-animal technology, methodology, approach, or combination thereof that can be used to provide information on chemical hazard and RA that avoids the use of intact animals (https://bit.ly/3HnMY9L). NAMs information includes various predictive in silico methods and models (e.g., quantitative structure-activity relationships [QSARs], physiologically-based toxicokinetic modeling [PBTK]), in vitro testing (e.g., cell-based, cell-free, biochemical assays) and embryo testing1¹ (e.g., whole animal exposure prior to independent feeding such as the fish embryo toxicity [FET] test). Furthermore, they may also include a variety of state-of-the-art methods, such as “high-throughput screening” and “high-content methods”, as well as some of the more conventional methods that aim to improve understanding of toxic effects using toxicokinetic-toxicodynamic (TK-TD) knowledge. Further information can be found in the supplemental materials. Providing information on chemical hazard and RA that avoids the use of intact animals, the OECD has standardized and internationally approved test guidelines for several NAMs for HRA used to evaluate dermal absorption, dermal irritation, eye irritation/corrosion and dermal sensitization potential and genotoxicity. Further, in progress at OECD level is HRA NAM validation for carcinogenicity and developmental neurotoxicity. In contrast, OECD standardized ERA NAMs are available for acute aquatic toxicity, aquatic bioconcentration/clearance and some tests for endocrine mechanisms.

Originally suggested as an alternative to animal studies, NAMs may also be used as a complement to animal testing, increasing our understanding of internal concentrations of compounds and how they relate to mechanisms of toxicity, as well as to answering scientific questions which cannot be well addressed by current in vivo regulatory testing. A series of reviews (https://bit.ly/3ts3vDO) highlight the non-animal models that are being used for basic and applied biomedical research such as on neurodegenerative diseases and immune oncology. Indeed, several of the NAMs established for HRA or biomedical research are also relevant for ERA, although have not been applied in that context yet. While progress towards change in the field of HRA draws on numerous articles, experiences and recommendations resulting in an explosion of scientific and policy initiatives, a similar level of engagement and change has so far not been observed for ERA (Figure 1).

The (need for) actions towards the use of NAMs

Catalyzed by the announcement of the US EPA to reduce animal testing and funding by 30% by 2025 and eliminate by 2035 (https://bit.ly/3tDKUVI), numerous new EPA policies and new guidance for the use of NAMs has been issued for HRA. Likewise, direct European Union (EU) funded research projects in combination with research partnerships have facilitated the development and use of NAMs in combination with the establishment of new assessment frameworks for regulatory toxicology (e.g., EU-ToxRisk², ASPIS cluster³ (Ontox, PrecisionTox, Risk-Hunt3R), PARC⁴)).

Some countries are outpacing others, at least in HRA, and appear to be acting as global catalysts for change. Ongoing discussions with relevant stakeholders across the globe have resulted in some movement in terms of adoption of some NAMs, often reinforced by changes to legislation, examples of which are summarized in Table S1 with further examples for specific working groups and associated legislations also briefly outlined. Yet further action within numerous individual overlapping sectors but especially at the governmental level will be key to ERA specific change, with some suggestions outlined in Figure 2. These actions by various sectors create the necessary change to support the adoption of NAMs in aquatic ERA and specifically the priority research questions outlined (Text box).

Figure 2:

Suggested activities by various stakeholders which could support the necessary change towards the increased use of NAMs.

Text Box.

Priority research and implementation questions for the next 5–10 years to support adoption of NAMs in aquatic ERA. The order of the questions does not indicate their relative priority.

1. What is necessary to evolve the discussion and mutual understanding between developers and end-users (industry/regulatory agencies) on when and how to further methods/approaches as new science or regulatory change emerges?

2. How can potential divergence of environmental NOEC be better assessed? Can variability between taxa and due to multiple environmental modifiers, like chemical mixture effects, abiotic stressors including climate change and biotic stressors like variable food-webs be incorporated?

3. Increase fundamental research on:

   • TK-TD divergence in sensitivity between a wider selection of organisms
   • Kinetic IVIVE models which protect numerous organisms’ population demographics
   • Better use/reuse and interoperability of -omics based RA toxicology data

4. (How) can knowledge about the uncertainties of traditional tests for environmental protection be better used to define benchmark criteria for the scientific and regulatory acceptance of NAMs data?

The increasing availability of NAMs, combined with political change, have now created an opportunity to redefine how we carry out RA, providing a rare opportunity to enhance knowledge of associated hazard and exposure. While the technological and methodological landscape has evolved rapidly in support of these changes, regulatory acceptance of alternatives to in vivo testing methods, capacity, and training in them have not kept pace, with various reasons cited (Mondou et al., 2020). One of the important reasons relates to the fact that the regulatory use of NAMs needs consensus of hundreds of experts and stakeholders, which is far beyond any usual scientific review process. However, even in the consumer products sector, where animal testing has been phased out in some global regions by law, conflicting requirements mean that traditional toxicity tests continue to be conducted in addition to NAMs (Fentem et al., 2021). More optimistically though, there are recent examples of regulatory agencies actively encouraging discussion with registrants regarding the use and submission of data from innovative technologies including NAMs (e.g., EMA Innovation Task Force (https://bit.ly/3aNzurV)). Further action by regulatory agencies appears essential for the evolution towards a NAM based regulation. Moreover, for success and a truly sustainable transformation, parallel initiatives by academic and industry actors and collaboration between all sectors are necessary and we indicate some ideas towards these goals in Figure 2.

The aim of this paper is to focus on the status quo and potential evolutions for the use of NAMs in ERA, in particular fish or other aquatic organisms. A non-exhaustive list of articles of interest which benefited developments in ERA are supplied in supplemental materials. Focusing on the past 10 years, only works that might reasonably be considered to contribute to advancements in the field or methods of particular special contemporary interest and importance, are highlighted.

CURRENT STATE-OF-THE SCIENCE FOR AQUATIC SPECIES

ERA of chemicals is largely based on aquatic ecotoxicology and faces several scientific challenges including the large number of species that are potentially affected in addition to the large number of chemicals emitted into the environment, various life stages, potentially chronic exposures, and the need to assess impacts at a population level, with a vast array of abiotic and biotic modifiers (Text Box). Several of the aforementioned variables can be modeled using interspecies correlation estimates (ICE; https://www3.epa.gov/webice/), species sensitivity distributions (SSDs), chemical toxicity distributions (CTDs) and the ecological threshold of concern (EcoTTC). Building on available experimental data, they can in principle be used with input data from in vivo methods or NAMs, allowing the prediction of acute fish toxicity, for example. Further, work is refining the available knowledge to what extent other plants or invertebrates are more sensitive than fish for the evaluation of acute toxicity of many compounds (e.g., (Rawlings et al., 2019)), since this could provide better returns for the protection of ecology, while also benefiting efforts to reduce animal testing. Moreover, scientifically improved predictions for environmentally safe concentrations may be generated by large multidisciplinary studies which incorporate both the development and use of NAMs which provide mechanistic data as well as the compilation of systematic knowledge about evolutionary conservation of (eco)toxicological mechanisms among species. Several online resources for comparative and predictive toxicology supporting this goal are available.

Yet, some assessments (e.g., bioaccumulation) are being carried out in fish, not just as an important component of ERA for ecosystems but also for human health protection (i.e., ingestion of contaminated fish via the diet). Thus, the fish as a model organism is pragmatically being employed in various fields including toxicology, pharmacology, and etiology of human disorders. Due to its small size, rapid growth, and freely accessible embryonic stages it may provide some practical advantages compared to mammalian species. Based on its evolutionary relationship with mammals it may also offer opportunities for molecular mechanistic studies and cross-species extrapolation. Furthermore, if testing is limited to the embryonic stages, then a higher throughput, and a reduction in laboratory animal usage is achievable, whilst still protecting the environment. However, to be fully implemented, more fundamental knowledge concerning mechanisms of toxicological outcomes in non-model or target organisms needs to be generated. Nevertheless, and despite the slow pace, there have been some recent achievements (not exhaustive) in aquatic non-animal alternatives, in terms of regulatory accepted methods but also emerging technologies which offer glimpses of a non-animal-based framework for aquatic ERA (see below).

Aquatic Embryo Testing and Weight of Evidence Assessment

Assessment of acute fish toxicity is an integral part of environmental hazard and RA regulations and is classically carried out using the acute fish toxicity test (AFT) which is conducted according to OECD TG 203 or similar guidelines, although other ecotoxicological endpoints are also used depending on specific need. AFT is the most frequently used vertebrate ecotoxicology assay because it is required in nearly all global regulatory schemes for the purposes of RA in addition to classification and labeling of chemicals (Burden et al., 2020). While this test has a number of recognized limitations, such as being low throughput, lacking in mechanistic information and reports of significant uncertainties, in addition to the severe suffering involved due to the nature of the test, there is currently a lack of consensus by regulators on an alternative approach. Though, studies are emerging which demonstrate that many of these issues can be addressed using alternative methods (e.g., (Paparella et al., 2021)).

Currently, two experimental methods are standardized and approved as OECD test guidelines as alternatives to the in vivo fish toxicity test: The fish gill cell line acute toxicity test using rainbow trout (Oncorhynchus mykiss) RTgill-W1 cell line (ISO 2019; OECD TG 249) and the fish embryo acute toxicity test [FET] (OECD TG 236), although so far their use has been limited due to a continued preference for traditionally accepted approaches and perceived difficulties in interpreting and combining new data types. While the FET test has not been accepted as a standalone replacement for regulatory purposes such as under the REACH regime (Sobanska et al., 2018), it can provide significantly more information about test compounds than originally envisioned during the TG guideline development (e.g., von Hellfeld et al., (2022)). Due in part to the versatility of the protocol, it has been beneficial to the development of numerous decision-making tools, some of which will be discussed in later sections. Furthermore, other alternative regulatory assessment assays are emerging, with the OECD recently releasing test guidelines utilizing transgenic Xenopus laevis, Danio rerio and Oryzias latipes embryos for evaluation of potential endocrine activity (OECD TGs 248: Xenopus Eleutheroembryonic Thyroid Assay (XETA); 250: EASZY assay - Detection of Endocrine Active Substances, acting through estrogen receptors, using transgenic tg(cyp19a1b:GFP) Zebrafish embrYos; and 251: Rapid Androgen Disruption Activity Reporter (RADAR) assay). These are in draft form at various levels using O. latipes (Rapid Estrogen Activity In Vivo (REACTIV) assay), andDaphnia magna (Short-term Juvenile Hormone Activity Screening Assay using Daphnia magna (JHASA)s. Notably the latter tests currently cannot contribute to any reduction or replacement of adult animal tests that definitively assess endocrine disruption, since they are considered to inform on endocrine mode of action only, and not on potential adversity as required to meet the current definition of an endocrine disruptor (IPCS 2002). However, guidance outlining the specific conditions for the use of the XETA as a mechanistic assay to detect thyroid active substances as an alternative to the in vivo amphibian metamorphosis assay (AMA, TG 231) for plant protection products have recently been published by the European Chemicals Agency (ECHA) and the European Food Safety Authority (Andersson et al., 2018). For a broader up-to-date review of the current and potential use for NAMs in the assessment of endocrine activity and disruption, please refer to Mitchell et al. (2023).

Weight of Evidence (WoE) assessment is frequently cited as necessary for a wide variety of decision-making needs due to the complexity of environmental data (Hall et al., 2017). It is generally understood as a method for decision-making which relies on multiple sources of information and lines of evidence and is expected to be the game changer for the regulatory acceptance of NAMs for acute fish toxicity. One such example of this can be found in the CEFIC LRI ECO51 project SWiFT (http://swift.hugin.com), which has developed a comprehensive online toxicity assessment system with built in examples to facilitate the acceptance of NAM data to routinely fill the regulatory requirements currently provided by the AFT. This Bayesian model integrates FET data with numerous lines of evidence including toxicity data from algae, daphnids, RTgill-W1 cell line (OECD TG 249) and information on fish-neurotoxicity and biotransformation in addition to QSARs and diverse ecotoxicological and physico-chemical datasets (Moe et al., 2020). Combined, there is over 87% agreement with AFT outcomes (when the aim is to predict if the LC50 is above or below 1 mg/L), which considering the uncertainties of AFT data represents a practically perfect correlation. While the latter project demonstrated how a quantitative WoE approach can successfully lead to animal test replacement for AFT, it has become increasingly clear that the integration of numerous lines of evidence will also be important in progressing the development and acceptance of non-animal methods in ERA, beyond this specific endpoint.

Omics Technologies, Complex Data, and Computational Approaches

Emerging and existing technologies have a significant impact on toxicological investigations and regulatory science alike. Increased access to mechanistic information via -omic approaches, can directly inform adverse outcome pathway (AOPs) frameworks, assisting in identification of mode of action. Perspectives on how high content -omic data sets can support ERA through the AOP framework was recently summarized (Brockmeier et al., 2017). Technical guidance for the use of AOPs in developing Integrated Approaches to Testing and Assessment (IATA) was harmonized at OECD level (https://bit.ly/3tCZVqB). However, while these methods (e.g., proteomics, lipidomics, metabolomics and transcriptomics) have been reviewed in numerous contexts including in the development of AOPs, chemical RA and the prospects and challenges of multi-omics data integration in toxicological research, these technologies have had limited acceptance for regulatory purposes (Text Box; Viant et al., 2019). Their absence in decision making has been attributed to the lack of best practice, standardization, and reporting guidance, all of which build confidence in methodological results. Different multi-stakeholder groups, including the OECD and specific advisory groups (https://bit.ly/39c6byL) have collaborated to improve the adoption of these approaches in ERA through the development of guidance documents and frameworks (Harrill et al., 2021; Viant et al., 2019). But, without a clear strategy to evaluate emerging technologies which are both rapid and appropriate, their full potential will remain largely unrecognized and unused (Anklam et al., 2022). Yet, notable change is emerging. In April 2022, the first transcriptomics-based NAM, the Genomic Allergen Rapid Detection Test Method for Skin Sensitization (GARD^™skin), was approved at OECD level, representing a breakthrough for regulatory acceptance of such technology. Furthermore, this sets a new standard for new OECD test guideline development, which will be beneficial to the whole field.

In the adoption of -omic technologies, an enormous volume of complex data will be generated, which when combined with experimental databases, should provide sufficient data to enable in silico modelling of the ecotoxicity of existing and new chemicals. In parallel, this may also increase confidence in –omics data and computational approaches via their mutual support. While numerous examples of developed and validated computational approaches for hazard assessment are available from academia, their application outside this sector is confined (e.g., (Luechtefeld, Marsh, et al., 2018)), with limited regulatory acceptance of in silico modeling. To facilitate the adoption of such tools, the OECD has developed a QSAR Toolbox (www.qsartoolbox.org) to improve regulatory acceptance for computational approaches which can be used for both the prediction of simple toxicity and fate properties and for more complex endpoints such as reproductive or repeated dose toxicity. Especially for the latter, the combined use of (Q)SARs (e.g., (Burden et al., 2016)) and experimental in vitro data may be useful to support an expert-based chemical category formation and read across of traditional in vivo data between chemicals within the same chemical category. Other (Q)SAR software tools like the free VEGA⁵ platform include the use of read across via a user-friendly interface which enhances transparency for the uncertainty of the proposed model versus experimental data for the structurally nearest neighbor. Furthermore, advances in technology are enabling complex algorithms in the form of machine learning (ML) and artificial intelligence (AI) to be applied in this context, underpinned by curated compound libraries (e.g., Tox21) specifically designed for the purpose of gaining better understanding of the chemical basis of toxicology [chemoinformatics]. The use of AI approaches (alongside other computational models) are also driven by the increasing availability of data whereby in vivo or in vitro databases for exposure or effect endpoints can be leveraged to increase generalizability of predictive models. ML algorithms are reported to be more powerful than traditional (Q)SAR approaches in terms of predictivity, (e.g., (Chen et al., 2022; Huang et al., 2016)). In addition, they may enable inter-species extrapolation for chemical safety and the prediction of chemical hazard across fish taxa for prioritization purposes (Wu et al., 2022), all without additional animal testing. While cases have demonstrated that these ML models could generate equivalent or better hazard data (Luechtefeld, Rowlands, et al., 2018), significant barriers remain to AI/ML adoption in the regulatory setting (Miller et al., 2018). Addressing these issues will be directly beneficial in reducing animal use.

Additionally, while acute fish toxicity appears to correlate well with the in vitro rainbow trout gill cell line assay (RTgill-W1), other situations where NAM data could be used to predict specifically systemic toxicity (e.g., toxicity to fish liver cells) require further information. Specifically, in linking the observed effects related to the internal concentrations of a chemical at the target site (e.g., in the blood or in a target tissue), rather than concentrations in the water, information on the absorption and interaction of chemicals within the living fish is critical. In silico techniques such as physiologically based pharmacokinetic (PBPK) models, or ADME-PK are increasingly being adopted to link exposure to physiological and health outcomes. One such example lies in prior in vivo work with fish demonstrating that pharmaceuticals with comparable pharmacological in vitro activity can have highly different in vivo risk due to their different uptake and PK profile (Margiotta-Casaluci et al., 2016). Moreover, it was previously demonstrated that the explicit consideration of internal exposure parameters (i.e., blood concentrations) can dramatically improve the accuracy of toxicity prediction for pharmaceuticals and facilitate the extrapolation of clinical and pre-clinical data to fish species (Margiotta-Casaluci et al., 2014). The systematic implementation of fish-specific PK considerations in the ERA process would allow the development of In Vitro-In Vivo Extrapolation (IVIVE) approaches to interpret the relevance of in vitro data for the specific fish-toxicity, in line with significant efforts ongoing in HRA (Punt et al., 2020). To overcome the challenge of species specificity and availability of chemical-specific parameters, (Wang et al., 2022) recently reported on a generalized fish PBK model that can be applied to a broader range of fish species and chemicals. A recent perspective article highlighting the application of PBTK model coverage combined with external exposure modeling provided better support for protective decisions allowing a shift towards new technologies that allow holistic evaluation of chemicals (Text Box; Cohen Hubal et al., 2019). In this, specific priorities requiring further work to build sufficient confidence were identified in a joint EPAA-EURL ECVAM ADME workshop (Bessems et al,, 2014). As with most computational approaches, barriers to their use have been recognized as a multipronged issue that relates to the availability and reliability of training data, the extrapolation of outcomes beyond the model’s domain of applicability and the lack of computational literacy among relevant stakeholders (Miller et al., 2018).

Towards Increasingly Complex In Vitro Culture Systems

Bioaccumulation potential of chemicals is traditionally assessed in terms of a reductionist BCF, though not necessarily ecologically relevant for hydrophobic chemicals because dietary exposure, and hence the potential for biomagnification, is not included. The complexity of environmental food-webs, possibly assessed via TMFs using mesocosm studies, may affect bioaccumulation and biomagnification in real environments via a wealth of biological modifiers. Nevertheless, on purpose, regulators usually do not consider any of these biological modifiers and prefer to regulate by reductionist BCF or BMF values, since these allow between chemical comparisons of intrinsic bioaccumulation or biomagnification potential without complex biologic modifiers. The latter may be relevant for one mesocosm or environment but not for another and they practically cannot be assessed comprehensively. This scientifically well defensible preference for reductionist approaches focusing on relative effect sizes between chemicals rather than on absolute effect sizes in real environments may be thought-provoking for the utility and acceptability of NAMs in general. Specifically, to what extent may we reduce complexity within the test systems, recognizing that regulatory toxicology can only assess comparative toxicity between chemicals?

Long-term research has resulted in the development of NAMs to assess the bioaccumulation potential of chemicals resulting the development of in silico and in vitro methods to estimate bioaccumulation potential comparable to in vivo (e.g., Kropf et al., (2020)), with accompanying information on reliability through an international ring trial (Nichols et al., 2018). Acceptance of the in vitro biotransformation assays with rainbow trout [fish] primary hepatocytes and S9 fractions by the OECD (TG 319A & 319B) as well as OECD guidance referencing the uncertainties for the computational in vitro data integration for BCF prediction as well as the uncertainties of the experimental BCF value (OECD GD 280) represents a significant step forward in this field, although discussions on their harmonized regulatory application for HRA (using human or rat hepatocytes) is still ongoing (Louisse et al., 2020). Likewise, the recent the recent acceptance of the fish gill cell line acute toxicity test as a predictor for acute fish toxicity is the result of decades of work (Fischer et al., 2019). Further OECD validation work is ongoing for the use of the freshwater amphipod Hyalella azteca (HYBIT) for bioaccumulation testing. These recent steps towards the ethically and scientifically desired regulatory acceptance and use of these protocols represent solid foundations from which we can expand.

Regulatory acceptance of NAM data in this field may be considered ‘low hanging fruit’ given the broad use of pragmatic logKow or QSAR-based regulatory decision tools for environmental assessments as well as the established regulatory acceptance of in vitro mammalian data in the drug development and approval process. But, to realize its full potential, regulatory acceptance of IVIVE would be beneficial, and this requires a concerted effort. For example, by generating fish in vitro hepatocyte and liver S9 data for various compounds, the application of the IVIVE approach for the estimation of bioaccumulation potential can be established, providing there is the necessary information for case study development which builds confidence. Such work may profit from consensus growing in the use of IVIVE for human safety and efficacy assessments (Bell et al., 2018).

For decades, two-dimensional (2D) cell cultures have been primarily used as in vitro screening tools to evaluate toxicity and predict drug impact in humans and more recently fish. Several NAMs have been developed and deployed to generate fish specific ADME parameters in different compartments, including both in vitro systems for relevant organs (i.e., gills, liver, intestine) and computational approaches integrating multi-organ or system level data. Examples of these include the Fish Gill Cell Culture Systems (FIGCS) used to predict chemical bioconcentration factors (BCF) and uptake/excretion dynamics (Chang et al., 2021). A more sophisticated metabolically competent three-dimensional (3D) in vitro system based on spheroidal aggregate cultures (spheroids) was first applied to humans, and later successfully applied using rainbow trout (Oncorhynchus mykiss) liver to study chemical metabolism with promising results (Baron et al., 2017, Hultman et al., 2019, Lammel et al., 2019).

The intestine also represents a major site of chemical interaction and toxicity, but until recently data on uptake through food chains was almost nonexistent, although this is changing slowly. Several groups have demonstrated that rainbow trout primary intestinal cells can be maintained in vitro in both 2D and 3D and used to investigate chemical metabolism in this important but often overlooked compartment (Langan et al., 2018). Limitations associated with the use of primary cells has been overcome with the generation of an immortalized rainbow trout intestinal cell line [RtgutGC] (Kawano et al., 2011), that has been successfully used to understand chemical transfer (Schug et al., 2019) and improved further using coculture with an intestinal fibroblast cell line (RTgutF) (Drieschner, Vo, et al., 2019). In keeping with the WoE approach, the development of a tiered testing strategy which integrates these in vitro systems could increase the chance that regulators accept risk assessments without data from the in vivo OECD 305 test to determine bioaccumulation in fish. Furthermore, the data generated with such NAMs could be used to accelerate the growing field of fish-specific PBPK models, which is currently limited (Wang et al., 2022), with a recent review highlighting how in vitro toxicity data can be used in risk assessment and decision making in the European Union (X. Chang et al., 2022).

In line with human organ-on-chip development, it is possible to foresee the application of microfluidic devices to enhance the biological relevance of the in vitro models mentioned above or to even integrate multiple organs in a single chip as a complex model recapitulating the complexity of an intact living organism. Since the early 1990’s, microfluidics has been increasingly used in chemical and biological research due to their potential numerous benefits, including improved physiological complexity and emulation of systemic effects in vitro. Organ-on-chip (OOC) devices, also known as microphysiological systems (MPS), have seen dramatic advances in the sophistication of biology and engineering over the past decade, facilitated by the convergence of multiple previously disparate technologies. Although progress has been primarily driven by human studies, with the common use of 3D human liver-on-a-chip (e.g., (Moradi et al., 2020)), the application of OOCs in aquatic toxicity testing is limited. Despite early developments in toxicity testing on flow-through RTgill-W1 cultures (Glawdel et al., 2009), and the development of a two-compartment intestinal barrier model with similar properties to salmonids intestines (Drieschner, Könemann, et al., 2019), little progress has been made in other fish organs. Furthermore, despite the availability of primary and immortalized static 3D fish hepatic cultures, progress on adaptation of this technique to the development of a fish specific liver-on-a-chip in vitro model to evaluate bioaccumulation potential has been limited.

The key drivers of (human) OOC development, including improved and longer-term phenotypic maturity such as expression and activity of xenobiotic metabolizing enzymes, are expected to improve the IVIVE of fish and other organism hepatocyte models. Yet, there are certain fundamental characteristics of microfluidics, which still render these assays technically demanding, especially for metabolic clearance predictions. For example, in flow-through systems, the chemical’s residence time within the cell culture chamber [~10’s mm-long] is defined by the linear flow rate and is thus inherently very short. Consequently, the majority of the OOC experiments focus on rapid measurement of pharmacological/toxicological endpoints or transepithelial transport and (human) disease modeling in vitro. To achieve hours-long chemical exposure times on microfluidic devices, often necessary in metabolic clearance determinations, dedicated recirculation systems need to be established both for mammalian and aquatic organisms alike. Furthermore, technological advances for measurement of the associated tiny volumes (μL) and increases in the limit of detection of analytical methodologies for more environmentally relevant concentrations are necessary for wider adoption. Although this technology is relatively new, its potential impact and implementation in the context of risk assessment of chemicals is already underway (Nitsche et al., 2022). To realize its potential, further work must address the lack of standardization of applicable materials (culture platforms) and protocols (e.g., shear force), which presently poses major challenges to interlaboratory comparisons and regulatory acceptance of microfluidics-derived data (Allwardt et al., 2020). In this regard, CEN-CENELEC⁶ established in 2021, the results of an EURL ECVAM survey (Batista Leite et al., 2021) and the respective activities by the Standards Coordinating Body in the US are foreseen to accelerate wider adaptation of the OOC concept to in vitro assessment of human and fish pharmacokinetics alike.

Integrated Approach to Testing and Assessment

The methods outlined provide some of the raw data or information about the hazard and/or exposure to a chemical, but still require a framework and workflow to be used for regulation. Thus, any regulatory testing and assessment would require the initial definition of the regulatory purpose and preferably a societal agreement about suitable goals and tools used for the evaluation. A conceptual example may build on OECD guidance for HRA (OECD GD No. 275) and is outlined here for environmental protection:

Step 1: establish if conservative estimates for environmental protection levels are sufficient for the specific regulatory situation. Such estimates may be based on computational EcoTTC values or refinements (already being applied in HRA), including a very broad set of in vitro bioactivity data and kinetic modeling (termed Next-Generation-Risk-Assessment, NGRA; (Friedman et al., 2020)) intending to protect, but not to predict any specific adverse effects at organism or population level. For ERA, such an approach may include besides NAMs, a relevant number of plants and invertebrates to provide a useful environmental No-Adverse-Effect-Concentration (NOAEC).

Step 2: where more information on possible vertebrate level effects appears necessary, NAM derived data and kinetic information may be used to derive a mode-of-action hypothesis, which can be tested with increasingly more complex in vitro methods (e.g., OOC/microphysiological systems) with a step wise improvement in kinetic modeling with increasing information. If relevant and available, additional in vivo vertebrate data from similar chemicals may be integrated by read across.

Importantly, the approach referenced here builds on exposure considerations and integrates these with hazard data. Exposure information is legally required for pesticides, biocides, and chemicals. Especially in Europe, regulation is also based on GHS classification, such that a potential hazard may have severe regulatory consequences independent from any exposure considerations, e.g., endocrine disrupting properties or chronic aquatic toxicity in combination with persistence- and bioaccumulation criteria. Nevertheless, exposure information may lead to adaptations of regulatory hazard information requirements on a case-by-case basis and future regulations could implement new default approaches, if this could increase the sustainability of regulations (as recently discussed e.g., for HRA in Ball et al. 2022).

However, where to stop a tiered assessment, as indicated above, would depend on available resources, acceptable uncertainties, and societal values. Such an approach could be positioned as part of an IATA and would logically include the interplay of TK-TD, environmental/interspecies extrapolation, and an appropriate and transparent level of uncertainty (Laroche et al., 2018). IATAs were intended to be flexible, but some elements can be standardized, which are referred to as Defined Approaches (DA) consisting of testing strategy and a fixed data interpretation procedure. Support for such approaches has resulted in guidance documents by the OECD in addition to case studies demonstrating proof of concept for regulatory fit (https://bit.ly/3I1P4ww). Moreover, such IATAs may also include the integration of NAM data to support the read-across of traditional animal test data between chemically and biologically similar chemicals.

FUTURE RESEARCH PRIORITIES AND RECOMMENDATIONS

Importance of collaboration

Numerous methods have been developed and implemented at various stages of the RA process; however, direct efforts are needed which allow for the integration and connection of these different approaches (Text Box). A significant advancement of scientific confidence in the practical application integrating NAM data into read-across approaches for risk assessment was achieved during the HORIZON 2020 EU-ToxRisk project (https://bit.ly/3QjN9Hp). This resulted in a unified strategy for the development of case studies established in partnership with regulatory agencies and contextualizing NAM data in a scientifically defensible way (Krebs et al., 2020), related workshops on how to make this approach global (Rovida et al., 2020) and ultimately recognized within the OECD Mutual Acceptance of Data (MAD) system. Building on prior research projects, the HORIZON 2020 RISK-HUNT3R project builds on prior outcomes to establish an overall human centric NGRA framework for chemicals which are designed to promote a combination of computational toxicology, in vitro toxicology, and systems biology, assuming this approach will lead to faster and more risk accurate procedures (Pallocca et al., 2022). Such an approach may also be undertaken in other organisms, with precedent already established.

Importantly, such activity alongside the 4Ćs principle (communication, cooperation, commitment, and coordination), may start to overcome some of the previously reported cross sectional barriers to the adoption of NAMs, which include uncertainty about the value of the new models, the lack of harmonization of regulatory requirements and acceptance criteria and the high levels of risk aversion (Punt et al., 2017). IATAs building on NAMs will provide different types of data and this will require a new/expanded scientific understanding of risk and uncertainty and a multi-stakeholder agreement on the regulatory use of these different data. Therefore, collaboration between government, regulators, academia, industry, and NGOs is key to success and we outline some options for respective and practical collaborative actions in Figure 2.

Importance of standardization

To improve confidence in the validity of NAMs, reproducibility and (re)usability among end users, transparent and comprehensive reporting may be furthered with an approach similar to the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines (Percie du Sert et al., 2020), which were developed to promote robust and reproducible animal research. Additionally, an important component of reducing animal usage is the enhancement of the reusability of data, for which the FAIR (findable, accessible, interoperable reusable) principles for scientific data management were developed (Wilkinson et al., 2016), and which should be applied to optimize knowledge growth. These guidelines set out the minimum information that should be included in any publication reporting the use of animals; endorsed by over a thousand journals and highly cited. However, while standardized guidelines for reporting of animal studies are available, standardized reporting for NAM methods is limited. Improved reporting standards is not new in the ecotoxicology field, although more effort should now be made to focus on transparency within NAM studies going forward. Reproducible science requires reproducible reporting, building confidence and trust in the process. In this respect, having a unified strategy on reporting facilitates adequate interpretation of the data to ensure overall scientific and toxicological validity. In line with this need for harmonized and comprehensive reporting, the EU-ToxRisk project built on earlier work for the standardization of non-guideline methods, reporting on the results of a case study for the regulatory use of 23 NAMs, which involved regulators reviewing the case studies and reporting, establishing method documentation, data processing and chemical testing pipelines (Krebs et al., 2020). The method documentation readily incorporated well established guidance documents on good in vitro methods practice (GIVIMP) and good cell culture practice (GCCP), both of which apply to all in vitro testing irrespective of organism. It should further be noted that these documents are not static, and with increasing technological advances in addition to increasingly complex culture systems, proposed strategies need space to incorporate these complementary recommendations for increased reproducibility and transparency, such as the latest version of the GCPP guidelines (Pamies et al., 2020). Likewise, the NC3Rs⁷ has initiated the development of reporting guidelines for in vitro research (RIVER – Reporting In Vitro Experiments Responsibly). As with the established reporting guidelines, once NAM-specific guidelines are developed, there needs to be mechanisms employed to ensure their uptake, including endorsement by internationally renowned journals. This need is also echoed in Figure 2.

Importance of the recognition of uncertainties

Following decades of research, the reality is that routine toxicity testing cannot fill the large gaps that experimental scientists and assessors/regulators regularly identify. The shift away from studying whole organisms, sometimes in advance of legislative change, to increased data availability and predictive power will assist in strengthening our confidence in the establishment of cause-effect relationships – a basic tenet of RA. For pragmatic reasons, regulatory toxicology is based on data from relatively few methods, which were internationally standardized. Such a practically manageable, well standardized set of methods should allow a minimum safety standard and a harmonized regulation of chemicals with minimal trade barriers from differential data availability. However, an additional scientific estimate for the toxicity of many chemicals might be provided using systematic reviews of all available data going forward, including scientific literature. Within such a new assessment, the uncertainties and inconsistencies could be spelled out in a scientifically correct way.

Traditionally the approach for NAM validation was to assess NAM data relative to data from the existing regulatory standards, rather than all the available scientific evidence. Yet, we do not know empirically if these standard tests actually represent the best science, or how predictive they are of environmental outcomes. This is especially true for some of the more complex animal tests (e.g., OECD TG 240 Medaka extended one generation reproduction test (MEOGRT) or OECD TG 241 Larval amphibian growth and development assay (LAGDA) which could not be validated during their development due cost, So, how can one scientifically demonstrate that any new regulatory approach is at least as useful as the current one? As a first step, this requires full transparency of practical limitations and scientific uncertainties of the current animal-based approaches (Text box). Scientists, together with regulators, must start to routinely report and discuss these in their daily work, which may provide the necessary common understanding for any next steps (Figure 2). Consequently, applying a systematic approach to characterize methods, current and new approaches can be qualitatively or semi-quantitatively compared for practicalities and scientific uncertainties. This has been demonstrated for human developmental neurotoxicity (Paparella et al., 2020) and fish toxicity (Paparella et al., 2021), with the same principle also applied to quantify the variability of rodent repeat dose studies, recognizing that any new method cannot predict data from the traditional method more precisely than the traditional method can predict itself via replication (Pham et al., 2020).

CONCLUSIONS

Regulatory toxicology must be recognized as a natural and social science, allowing for the regular reexamination of its basic concepts, by asking questions such as: What type of testing for which type of chemicals is required from an economic, societal, and ethical perspective? Besides cosmetics, are there other chemicals for which animal testing may become unacceptable, such as biocides, and which can be replaced by non-chemical alternatives? When do we need to predict adverse effect types and when are estimates for non-adverse effect concentrations sufficient? What level of uncertainty is acceptable for decision making and what level of pragmatic precaution can be taken to compensate for uncertainties? Inclusive societal forums may need to be established to discuss and answer such questions.

The challenge to change from animal testing to NAMs is different between sectors. Personal education, moral values, professional experience, opportunities, and hierarchies as well as peer group forces are influential in the development, use and selection of scientific methods. However, in principle academic researchers are free to formulate their research questions in a way that they may be directly addressed with NAMs. Here, innovation and exploration of new methods is encouraged, enabling “big data” generation while avoiding animal use. Such activity may significantly contribute to building evidence, confidence and potentially case studies to support regulatory needs. Indeed, this is the de facto case for many regions where animals in science are increasingly strictly regulated by government or institution. So, what needs to be done to effect transformation? We outline some suggested actions for the various stakeholders in Figure 2. As highlighted in the EPAA Blue sky workshop, “disruptive thinking” is required to reconsider chemical legislation, validation of NAMs and to embrace the opportunities to move away from reliance on animal tests (Mahony et al., 2020).

Therefore, the answer to the initial question of “Can non-animal testing methods be developed that will provide equivalent or better hazard data compared with current in vivo methods?” is nuanced, but clearly yes for some of the available approaches, while also recognizing that NAMs may be used in a protective standalone approach without predicting any specific in vivo method, applying to HRA and ERA alike. A divergence in views occurs with identifying which regulations can exclusively be based on NAMs and further when this will occur. The answer to such questions heavily depends on resources invested in regulatory evolution, policy change and the readiness by all for substantive changes versus minor adaptations of regulatory practices. Regardless, to realize the full potential of NAMs, more work is needed, much of which overlaps with various publications following HRA workshops outlined in supplemental additional reading, in addition to progress in other sectors and countries paving the way for the adoption of NGRA (e.g., (Bhuller et al., 2021; Escher et al., 2022; Friedman et al., 2020)). To accomplish the goal of providing equivalent or better hazard protection compared with current in vivo methods, the authors acknowledge that NAMs and their combinations in a standardized WoE approach have been developed which could provide equivalent or better hazard protection, while also acknowledging that research is always required to move ahead. In light of this review, the authors recommend prioritizing the following research questions outlined in the Text box. Furthermore, we have also outlined some policy-research questions (Table S2) required to increase pace and diversity in addition to future proof the area of toxicology and RA. To answer all outlined questions will require funding, leadership, guidance, and active endorsement. Looking to the future, we are at a tipping point, with a need for a global and inclusive approach to establish consensus. Bringing together all of this work for regulatory assessment and decision making will require a concerted effort and orchestration.

Data Availability statement

No data has been generated in the writing of this article. However, all articles which have been cited or influenced the summarization outlined are provided in supplementary material on a section-by-section basis.