This year marks the 60^th anniversary of Russell and Burch’s pioneering book, The Principles of Humane Experimental Technique. Their 3Rs framework has helped to inspire humane and scientific progress in experimental technique. However, it is time to update its strategic application. The 21^st century has already seen the development of promising, high-tech non-animal models, such as organs-on-a-chip and computational approaches that, in our view, will replace animals as the default option in biomedical experimentation. How fast this transition will take place will depend on the pace at which these new models are optimized to reflect the biology of humans, rather than that of non-human animals. While the new methods are likely to reshape all areas in which animals are currently used in science, we particularly encourage their application in biomedical research, which accounts for the bulk of animals used. We call for the pursuit of a three-prong strategy that focuses on (1) advancing non-animal methods as replacements of animal experiments, (2) applying them to biomedical research, and (3) improving their relevance to human biology. As academics and scientists, we feel that educational efforts targeted at young scientists in training will be an effective and sustainable way to advance this vision. Our strategy may not promise an imminent end to the use of animals in science, but it will bring us closer to an era in which the 3Rs are increasingly perceived as a solution to a receding problem. Russell and Burch themselves surely would have welcomed these positive changes.

The adverse outcome pathway (AOP) framework is a conceptual construct that mechanistically links molecular initiating events to adverse biological outcomes through a series of causal key events (KEs) that represent the perturbation of the biological system. Quantitative, predictive AOPs are necessary for screening emerging contaminants and potential substitutes to inform their prioritization for testing. In practice, they are not widely used because they can be costly to develop and validate. A modular approach for assembly of quantitative AOPs, based on existing knowledge, would allow for rapid development of biological pathway models to screen contaminants for potential hazards and prioritize them for subsequent testing and modeling. For each pair of KEs, a quantitative KE relationship (KER) can be derived as a response-response function or a conditional probability matrix describing the anticipated change in a KE based on the response of the prior KE. This transfer of response across KERs can be used to assemble a quantitative AOP. Here we demonstrate the use of the proposed approach in two cases: inhibition of cytochrome P450 aromatase leading to reduced fecundity in fathead minnows and ionic glutamate receptor mediated excitotoxicity leading to memory impairment in humans. The models created from these chains have value in characterizing the pathway and the relative level of toxico­logical effect anticipated. This approach to simplistic, modular AOP models has wide applicability for rapid development of biological pathway models.

Since the EU banned animal testing for cosmetic products and ingredients in 2013, many defined approaches (DA) for skin sensitization assessment have been developed. Machine learning models were shown to be effective in DAs, but the predictivity might be affected by data imbalance (i.e., more sensitizers than non-sensitizers) and limited information in the databases. To improve the predictivity of DAs, we attempted to apply data-rebalancing ensemble learning (bagging with support vector machine (SVM)) and a novel and comprehensive Cosmetics Europe database. For predicting human hazard and three-class potency, 12 models were built for each using a training set of 96 sub­stances and a test set of 32 substances from the database. The model that predicted hazard with the highest accuracy (90.63% for the test set and 88.54% for the training set, named hazard-DA) used SVM-bagging with combinations of all variables (V6), while the model that predicted potency with the highest accuracy (68.75% for the test set and 82.29% for the training set, named potency-DA) used SVM alone. Both DAs showed better performance than LLNA and other machine learning-based DAs, and the potency-DA provided more in-depth assessment. These findings indicate that SVM-bagging-based DAs provide enhanced predictivity for hazard assessment by further data rebalancing. Meanwhile, the effect of imbalanced data might be offset by more detailed categorization of sensitizers for potency assessment, thus SVM-based DA without bagging could provide sufficient predictivity. The improved DAs in this study could be promising tools for skin sensitization assessment without animal testing.

Chemical substances that induce an allergic response in skin upon contact are called skin allergens or sensitizers, while chemical substances that elicit an allergic response only in the presence of light are called photoallergens or photo sensitizers. The Direct Peptide Reactivity Assay (DPRA, OECD TG 442C, 2015) and the Amino Acid Derivative Reactivity Assay (ADRA) are in chemico assays used to discriminate between allergens and non-allergens. The DPRA and the ADRA respectively monitor the depletion of model peptides and modified amino acids induced by crosslinking with test chemicals. In the current study, we compared these two assays and analyzed their suitability to predict the skin sensitization potential of several chemical substances. In order to study the combined effect of a chemical compound and UV light, we modified the DPRA (photo-DPRA) as well as the ADRA (photo-ADRA) by introduction of a photo-irradiation parameter. Analysis using photo-DPRA and photo-ADRA correctly distinguished known photoallergens from non-photoallergens. Upon irradiation, photoallergens selectively showed higher depletion of model peptides or modified amino acids. Thus, photo-DPRA and/or photo-ADRA can serve as non-animal in vitro methods for the identification and assessment of photoallergens/photosensitizers.

To more accurately model inhalation toxicity in vitro, we developed a tetra-culture system that combines lung alveolar epithelial cells, endothelial cells, macrophages, and mast cells in a three-dimensional orientation. We characterized the influence of the added complexity using network perturbation analysis and gene expression data. This allowed us to gain insight into the steady-state profile of the assembled, complete three-dimensional model using all four cell types, and of simpler models of one, two, or three cell types. Gene expression data were analyzed using cause-and-effect biological network models, together with a quantitative network-scoring algorithm, to determine the biological impact of co-culturing the various cell types. In the tetra-culture, macrophages appeared to be the largest contributors to overall network perturbations, promoting high basal levels of oxidative stress and inflammation. This finding led to further optimization of the model using rested macrophages, which decreased the basal inflammatory and cell stress status of the co-culture. We compared transcriptional profiles from publicly available datasets of other in vitro models representing the airways and of healthy human lung tissue with those of our model. We found an increasing correlation between airway models and normal human lung tissue as cell types became more physiologically relevant and the complexity of the system increased. This indicates that the combination of multiple lung-relevant cell types in vitro does indeed increase similarity to the physiological counterpart.

The aim of the study was to develop an in vitro model that mimics the alveolar-capillary barrier and that allows assessment of the respiratory sensitizing potential of substances. The 3D in vitro model cultured at the air liquid interface consists of alveolar type II epithelial cells (A549), endothelial cells (EA.hy926), macrophage-like cells (PMA-differentiated THP-1), and dendritic-like cells (non-differentiated THP-1). This alveolar model was exposed apically to nebulized chemical respiratory sensitizers (phthalic anhydride (PA) and trimellitic anhydride (TMA)) or to irritants (methyl salicylate (MeSa) and acrolein (Acr)) at concentrations inducing 25% cytotoxicity. The exposure to respiratory sensitizers induced den­dritic-like cell activation and a specific cytokine release pattern, while exposure to irritants did not. In addition, the cell surface marker OX40L was found to identify dendritic-like cell activation by high molecular weight allergens. With this in vitro model we can postulate a set of promising markers that allow the discrimination of chemical respiratory sensitizers from irritants.

Besides being an energy store, adipose tissue is an endocrine organ closely associated with the vascular system. Human relevant in vitro models are needed to study adipose tissue and related diseases. Vasculature plays a central role in the development and inhibition of adipose tissue-related diseases.
Here, an adipocyte culture was established from hASC (human adipose stromal cells), and a vascularized adipose tissue model was established from hASC and HUVEC (human umbilical cord vein endothelial cells) co-culture, utilizing the same differentiation procedure. Comparing these models allowed analysis of the effect of vascularization on adipocytes. Both models were characterized on gene (adipocyte and vasculature-related), protein (von Willebrand factor, collagen IV, CD140b and CD144, secretion of leptin, adiponectin and FABP4), and functional (triglyceride accumulation, glucose uptake, and lipolysis) levels. Additionally, the vascularized adipose tissue model was exposed to chemicals with known effects on adipogenesis and angiogenesis (rosiglitazone, chlorpyrifos, prochloraz, mancozeb, butylparaben, 15-deoxy- Δ12,14-prostaglandin J2, bisphenol A, bis-(2-ethylhexyl) phthalate, tributyltin chloride) to compare their effects to the literature. The in vitro vascularized adipose tissue model demonstrated the presence of functional adipocytes and an extensive vascular network, displaying relevant gene and protein markers. Insulin induced glucose uptake, inhibited lipolysis, and influenced vasculature-related genes. The presence of vasculature led to a faster lipolysis inhibition by insulin and modulated responses to chemicals. This novel, thoroughly characterized, vascularized adipose tissue model is a promising new tool for studying adipose tissue as well as the effects of chemicals on adipogenesis and angiogenesis in adipose tissue.

The ability of non-animal methods to correctly predict the outcome of in vivo testing in repeated applications is referred to as precision. Due to dichotomizing continuous read-outs into discrete “positive/negative” hazard data, non-animal methods can reveal discordant classifications if results are sufficiently close to a defined classification threshold. This paper explores the impact of precision uncertainty on the predictive accuracy of non-animal methods. Using selected non-animal methods for assessing skin sensitization hazard as case study examples, we explore the impact of precision uncertainty separately and in combination with uncertainty due to varying composition and size of experimental samples. Our results underline that discrete numbers on a non-animal method’s sensitivity, specificity, and concordance are of limited value for evaluation of its predictivity. Instead, information on the variability and the upper and lower limits of accuracy metrics should be provided to ensure a transparent assessment of a testing method’s predictivity, and to allow for a meaningful comparison of the predictivity of a non-animal method with that of an animal test.

The Virtual Cell Based Assay (VCBA) is an in silico model that simulates the biokinetics of chemicals in in vitro test systems. VCBA simulations can indicate the degree to which the bioavailable concentration varies across chemicals and experimental conditions, thereby providing important contextual information for comparing the results of different in vitro toxicity experiments. The simulated results can also be used to support in vitro to in vivo extrapolation of toxicity data, especially when the VCBA is coupled to a physiologically based kinetic model.
In this work, we selected 83 chemicals previously tested for in vitro cytotoxicity with a neutral red uptake (NRU) assay and used the respective in vitro data to optimize a toxicity and effects model simulating the 3T3 BALB/c cell line in a 96-well microplate with 5% serum supplementation. We then used the optimized parameters to simulate alternative experimental conditions. The simulations show the impact of different physicochemical properties on chemical fate of this diverse group of chemicals and how the different partitioning (to protein, lipid, and plastic) and kinetic (evaporation and degradation) events are intrinsically connected. The results of VCBA simulations were interpreted with respect to the applicability domain of the different QSARs incorporated in the model and the underlying assumptions and uncertainties of the VCBA.

Generalized Read-Across (GenRA) is a data driven approach that makes read-across predictions on the basis of a similarity weighted activity of source analogues (nearest neighbors). GenRA has been described in more detail in the literature (Shah et al., 2016; Helman et al., 2018). Here we present its implementation within the EPA’s CompTox Chemicals Dashboard to provide public access to a GenRA module structured as a read-across workflow. GenRA assists researchers in identifying source analogues, evaluating their validity and making predictions of in vivo toxicity effects for a target substance. Predictions are presented as binary outcomes reflecting the presence or absence of toxicity together with quantitative measures of uncertainty. The approach allows users to identify analogues in different ways, quickly assess the availability of relevant in vivo data for those analogues, and visualize these in a data matrix to evaluate the consistency and concordance of the available experimental data for those analogues before making a GenRA prediction. Predic­tions can be exported into a tab-separated value (TSV) or Excel file for additional review and analysis (e.g., doses of analogues associated with production of toxic effects). GenRA offers a new capability of making reproducible read-across predictions in an easy-to-use interface.

Systematic reviews (SRs) hold promise for implementing the 3Rs in animal sciences: they can retrieve available alternative models, help refine experiments, and identify insufficiencies in, or an excess of, scientific knowledge on a particular topic. Unfortunately, SRs can be labor- and time-intensive, especially the reference screening and data extraction phases. Fortunately, several software tools are available that make screening faster and easier. However, it is not always clear which features each tool offers. Therefore, a feature analysis was performed to compare different reference screening tools as objectively as possible. This analysis enables researchers to select the tool that is most appropriate for their needs.
Sixteen different tools were compared: CADIMA, Covidence, DistillerSR, Endnote, Endnote using Bramer’s method, EPPI-Reviewer, EROS, HAWC, Microsoft Excel, Excel using VonVille’s method, Microsoft Word, Rayyan, RevMan, SyRF, SysRev.com, and SWIFT Active Screener. Their support of 21 features categorized as mandatory, desirable, and optional was tested.
DistillerSR, EPPI-Reviewer, Covidence, and SWIFT Active Screener support all mandatory features. These tools are preferred for screening references, but none of them are free. The best scoring free tool is Rayyan, which lacks one mandatory function: distinct title/abstract and full-text phases. The lowest scoring tools were those not specifically designed for SRs, like Microsoft Word and Endnote. Their use can only be advised for small and simple SRs.
A well-informed selection of SR screening tools will benefit review quality and speed, which can contribute to the advancement of the 3Rs in animal studies.

Quantification of fluorescence colocalization and intensity of strongly overlapping cells, e.g., neuronal cultures, is challenging for programs that use image segmentation to identify cells as individual objects. Moreover, learning to use and apply one of the large imaging packages can be very time- and/or resource-demanding. Therefore, we developed the free and highly interactive image analysis program SUIKER (program for SUperImposing KEy Regions) that quantifies colocalization of different proteins or other features over an entire image field. The software allows definition of cellular subareas by subtraction (“punching out”) of structures identified in one channel from structures in a second channel. This allows, e.g., definition of neurites without cell bodies. Moreover, normalization to live or total cell numbers is possible. Providing a detailed manual that contains image analysis examples, we demonstrate how the program uses a combination of colocalization information and fluorescence intensity to quantify carbohydrate-specific stains on neurites. SUIKER can import any multichannel histology or cell culture image, builds on user-guided threshold setting, batch processes large image stacks, and exports all data (including the settings, results and metadata) in flexible formats to be used in Excel.