New Strategies for Developmental Toxicity Assessment based on Adverse Outcome Pathways

Home / New Perspectives / Reproductive & Developmental Toxicity / New Strategies for Developmental Toxicity Assessment based on Adverse Outcome Pathways

Reproductive & Developmental Toxicity

New Strategies for Developmental Toxicity Assessment based on Adverse Outcome Pathways

Kamin J. Johnson and Edward W. Carney, Toxicology and Environmental Research and Consulting,
The Dow Chemical Company, Midland, MI

Published: April 10, 2014

About the Author(s)
Kamin Johnson is a Senior Toxicologist for The Dow Chemical Company’s Toxicology & Environmental Research and Consulting function, responsible for the scientific conduct and interpretation of Developmental and Reproductive Toxicology studies.  Kamin has served on National Institutes of Health reproductive toxicology grant review study sections and served as a Counselor for the Reproductive and Developmental Toxicology Specialty Section of the Society of Toxicology.  He has published over 30 articles on the molecular and cellular biology of fetal and postnatal testis function as well as mechanisms of testicular toxicants.

In 2013, Kamin joined Dow after an academic career researching reproductive biology and reproductive toxicology. Kamin received a BS degree in Genetics from the University of Georgia and a PhD in Molecular and Cellular Biology and Biochemistry from Brown University.  After postdoctoral positions at Duke University and Brown University, he was an Assistant Investigator at The Hamner Institutes for Health Sciences.  Just prior to joining Dow, Kamin held adjunct faculty positions at Thomas Jefferson University and the University of Delaware, while conducting research at the Alfred I. duPont Hospital for Children as an Associate Research Scientist.


Edward (Ed) Carney is a Research Fellow and Scientific Director for The Dow Chemical Company’s Toxicology & Environmental Research and Consulting (TERC) function, with responsibility for overall science strategy and advancement of modern approaches to chemical safety assessment.  He also has direct leadership responsibility for TERC’s Systems Toxicology unit.  Ed joined Dow in 1992 as a member of the Developmental & Reproductive Toxicology group conducting regulatory toxicology studies and investigational research, and later went on to positions of increasing leadership across mammalian toxicology.  In 2010 he led the creation of TERC’s Predictive Toxicology Center, which uses in vitro and computational models to develop rapid, low-cost, non-animal safety predictions for a wide range of applications, including new product design and development.  He has published approximately 90 papers to date, in areas such mechanisms of developmental toxicity, integration of toxicokinetics into toxicity testing, endocrine-mediated toxicity, and chemical mixtures assessment.  He also pioneered the development of the rabbit whole embryo technique as a model to study species-specific mechanisms of developmental toxicity.

Dr. Carney represents Dow on US EPA’s Chartered Scientific Advisory Board and Board of Scientific Counselors, the EU’s ECVAM Scientific Advisory Committee, the American Chemistry Council Strategic Science Team, and the Board of Directors of the Hamner Institutes.  He also is an adjunct faculty member in the University of Michigan’s School of Public Health, and lectures for the University of Surrey (UK) Master’s Programme in Toxicology.  Ed previously served as President of the Teratology Society (2012-2013), and as a member of the National Toxicology Program Board of Scientific Counselors (2007-2010).  His educational training includes a BS in Animal Science from Cornell University, an MS in Reproductive Biology from University of Wisconsin-Madison, and a PhD in Reproductive Physiology from Cornell.  Prior to joining Dow, Ed conducted postdoctoral research in molecular developmental biology at Mount Sinai Hospital in Toronto.



For the past several decades, toxicological tests involving pregnant animals (primarily rodents and rabbits) have been the gold standard upon which risk to the developing human embryo/fetus was assessed. The greatest attribute of current animal-based tests is their ability to capture the complexity inherent in developmental biology and observe adverse perturbations to this complex system. Despite these positive attributes, current developmental toxicity regulatory guideline studies have inherent limitations [See Regulatory Testing Overview]. In particular, animal-based tests are resource intensive, low throughput, and provide little mode of action information useful for predicting the human response to chemical exposure. The ideal developmental toxicity test system would possess high human predictive capability, be rapid, include in vivo xenobiotic metabolism, provide mechanistic information to increase confidence in the result, and query a wide exposure range to determine effects at environmentally relevant concentrations (Bhattacharya et al., 2011).  It now seems clear that one single assay may never have all of these attributes.  Instead, novel integrated testing strategies which draw upon knowledge of adverse outcome pathways (AOP) using a combination of in vitro and in silico technologies appear to be the way forward.

Historically, there has been decades of effort devoted to a large number of alternative assays to assess developmental toxicity. The vast majority of these have focused on prediction of teratogenicity, rather than the other manifestations of altered development (i.e., fetal growth retardation, fetal death, functional impairments), and have either relied on direct visualization of altered morphological development (e.g., whole embryo culture) or differentiation along a specific developmental lineage (e.g., limb micromass assay, formation of cardiomyocytes from embryonic stem cells). However, none of these assays has achieved better than approximately 80% predictivity for developmental toxicity in animals.

Recognizing the limitations of these existing alternatives as well as the unsustainable nature of current animal-based testing paradigm, a groundbreaking report was published in 2007 by a National Academy of Sciences committee (NRC, 2007). The vision of this report called for a greater reliance upon newer, human-based in vitro assays and computational systems biology methods to assess human toxicological risk of environmental chemicals, with these assays examining key molecular endpoints within AOPs encompassing the totality of possible toxicological pathways. As far as scientific, regulatory, and industrial stakeholders embracing this 21st Century Toxicology vision, the proverbial train has left the station. The Organization for Economic Cooperation and Development has issued a guidance document on developing an AOP (OECD, 2013), and an initiative among multiple organizations called AOPWiki seeks to develop, catalog, and share AOPs. What remains unknown is the composition of the final landscape of agreed upon AOPs, the suite of in vitro assays to examine toxicity pathway perturbations, the human risk predictability of the developed in vitro assays, and incorporating results from in vitro assays into a risk assessment paradigm (Bhattacharya et al., 2011).

New Approaches for Assessing Developmental Toxicity
‘Omics and Examination of Key Molecular Signaling Nodes

It seems clear that the journey toward a definitive, pathways-based risk assessment supplanting current animal guideline studies may require many years of broad-based scientific effort.  However, there are many interim opportunities to use newer predictive approaches to assess developmental toxicity, which range from in-house evaluation of new product candidates, informing the design of more sustainable alternative chemicals, the assessment of formulations and mixtures, and providing supplemental data for compliance with chemical safety assessment programs such as REACH.  One promising approach to producing a predictive screen is to analyze the response of cells exposed to known developmental toxicants and use these data to develop a biomarker of developmental toxicity using statistical classification techniques. In this approach, high content “omic” screening of embryonic stem cells has been used to identify developmental toxicity biomarkers based upon mRNA (Pennings et al., 2011) or metabolite (Kleinstreuer et al., 2011) expression profiles. The developmental toxicity predictive capability of these biomarkers against a small set of developmental toxicants was reported to be good at approximately 70 – 80% accuracy, which is similar to existing in vitro teratogenicity assays.

Undoubtedly, there is a limited set of toxicological pathways leading to developmental toxicity (actual number is currently unknown), and it may be possible to develop predictive in vitro assays that examine key molecular nodes of mammalian development upon which multiple signaling pathways and differentiation processes converge. For example, Palmer, et al. (2013) profiled the metabolome of human embryonic stem cell cultures following exposure to numerous known developmental toxicants and identified a toxicity biomarker consisting of only two metabolites (ornithine and cystine) that was able to predict teratogenic potential with 75% accuracy. In other cases, the basic biology of mammalian development may guide production of an appropriate developmental toxicity test. Such is the case of a developmental toxicity assay using human induced pluripotent stem cells recently reported by Kameoka, et al. (2014). The transcription factor SOX17 is an established marker of endoderm, which differentiates early during embryonic development and organogenesis. Using SOX17 immunoexpression in human pluripotent stem cells as the only endpoint, Kameoka, et al. reported a high throughput SOX17 imaging assay that was able to predict the in vivo teratogenicity for a set of 86 pharmaceuticals and environmental chemicals with approximately 90% accuracy. These data support the idea that teratogens with disparate molecular targets and phenotypic outcomes initially perturb a subset of molecular pathways converging upon a small number of molecular nodes.

Toxcast Program

A different approach to producing a predictive developmental toxicity model is represented by the ToxCast program at the Environmental Protection Agency. By examining the responses of a large chemical library with known developmental toxicity profiles on a battery of over 600 high-throughput in vitro mechanistic assays, the ToxCast program produced predictive models of rat and rabbit developmental toxicity (Sipes et al., 2011). A different set of assays (i.e. models) produced the best predictivity for rat and rabbit developmental toxicants. While the ToxCast models displayed good predictive accuracy (approximately 70%) within a species, accuracy diminished when the rat model was used to predict rabbit developmental toxicants. This result suggests the importance of using toxicants with known human relevance in the biomarker development process.

Combined Chemical-Biological Approaches

One of the main challenges with both the historical and newer in vitro teratogenicity assays is that they do not cover a broad enough range of developmental stages, embryonic lineages, adverse developmental outcomes or their underlying mechanisms to constitute a robust assessment of safety.  Developing a larger battery of in vitro functional developmental assays is not a pragmatic solution either, as it would likely be more expensive and time consuming than the in vivo studies which they are intended to replace.  Although quantitative structure activity relationship models can identify chemical structural features which might correlate with a broad range of developmental toxicity outcomes, experience to date indicates that they are not by themselves able to sufficiently model all of the biological complexity involved in adverse developmental outcomes. However, recent research suggests that a combination of chemical- and biologically-based methods can be used to generate improved predictions of developmental toxicity (Low et al., 2013). Broad-based, high data content platforms such as cheminformatics, toxicogenomics, and high throughput in vitro mechanistic assays (e.g., ToxCast) are particularly powerful in this regard, and if combined in an integrated, tiered strategy, can do so in an efficient and effective manner.

One example of such a model being developed is rooted in a structure activity relationship decision tree framework for reproductive and developmental toxicity designed to guide identification and evaluation of analogs for use in read-across (Wu et al., 2010; Wu et al., 2013).  In this framework, analogs and substructures for the substance being evaluated are identified and graded for suitability, and toxicity is inferred on the basis of existing data for these analogs.  As the database of chemicals grows (currently nearing 1000), structural rules and alerts for predicting toxicity are developed and incorporated into a decision tree to further guide the evaluation or trigger the need for higher tier experimental data.  To further build upon this framework, transcriptomics and other high content data are being explored as a means of identifying and testing for putative modes of action.

Caveats of Current Predictive Assays

There is no question that the development and ultimately regulatory acceptance of in vitro alternatives for developmental toxicity assessment is fraught with difficult challenges.  Among these are the lack of in vitro metabolizing capability in most in vitro systems, the difficulty in modeling extreme and rapidly changing biological complexity, and the need to ensure safety across an extremely wide range of developmental stages and potential adverse outcomes. From both scientific and product stewardship perspectives, what criteria will need to be achieved before in vitro assays gain acceptance as replacements for animal-based testing? What are acceptable false-negative and false-positive rates?  The answer to many of these questions depends on the purpose for which the assays are being used. Certainly the requirements in terms of assay performance will differ between initial screening, prioritizing a group of chemicals for further testing, providing supplemental data to support a common mode-of-action category for read-across, and higher tier risk assessments upon which safety standards are set.  One thing seems clear is that for many of these more demanding applications, the field needs to move away from single end point (generally teratogenicity) assays and toward high information content approaches which combine knowledge of both chemical structure and biological activity.


At this moment in time, the toxicology community now sees before it an opportunity to reshape developmental toxicology and produce a regulatory testing strategy that potentially generates immediate mechanistic information with limited animal usage in a higher throughput manner. This transition to alternative approaches must be done in a systematic and stepwise manner so that we all have confidence that the new methods adequately protect human health. To build a path toward the ultimate goal of using new methods in risk assessment, it would make sense to first apply these methods for less demanding applications such as prioritization of compounds for subsequent testing, selection of new candidate molecules or sustainable alternatives, and for providing supplemental data to support read-across for programs such as REACH. Along this path, it is prudent to keep our eyes open for potential pitfalls and design inadequacies so that confidence and credibility are maintained. The suite of future predictive strategies must accurately examine adverse toxicological outcomes, be of sufficient sensitivity, and cover the landscape of developmental toxicity, not just teratogenicity.

In the coming years, extensive research involving academic, regulatory, and industrial collaborations will be needed to discover pathways of developmental toxicity and develop appropriate experimental tools to examine these pathways. While much has been accomplished, developing a testing strategy using computers and cultured cells will require several years (decades?) of sustained research activity. Now is the time to put intellectual and monetary resources toward this effort.

©2014 Kamin J. Johnson and Edward W. Carney

Bhattacharya, S., Zhang, Q., Carmichael, P.L., Boekelheide, K., & Andersen, M.E. (2011). Toxicity testing in the 21 century: defining new risk assessment approaches based on perturbation of intracellular toxicity pathways. PloS one, 6, e20887.

Kameoka, S., Babiarz, J., Kolaja, K., & Chiao, E. (2014). A high-throughput screen for teratogens using human pluripotent stem cells. Toxicol Sci, 137, 76-90.

Kleinstreuer, N.C., Smith, A.M., West, P.R., Conard, K.R., Fontaine, B.R., Weir-Hauptman, A.M., Palmer, J.A., Knudsen, T.B., Dix, D.J., Donley, E.L., & Cezar, G.G. (2011). Identifying developmental toxicity pathways for a subset of ToxCast chemicals using human embryonic stem cells and metabolomics. Toxicol Appl Pharmacol, 257, 111-121.

Low, Y., Sedykh, A., Fourches, D., Golbraikh, A., Whelan, M., Rusyn, I., & Tropsha, A. (2013). Integrative chemical-biological read-across approach for chemical hazard classification. Chem Res Toxicol, 26, 1199-1208.

National Research Council (NRC). (2007). Toxicity Testing in the 21st Century: A Vision and A Strategy.  (Washington, DC).

OECD (2013). Guidance Document on Developing and Assessing Adverse Outcomes Pathways. (Series on Testing and Assessment, Number 184.)

Palmer, J.A., Smith, A.M., Egnash, L.A., Conard, K.R., West, P.R., Burrier, R.E., Donley, E.L., & Kirchner, F.R. (2013). Establishment and assessment of a new human embryonic stem cell-based biomarker assay for developmental toxicity screening. Birth Defects Res B Dev Reprod Toxicol, 98, 343-363.

Pennings, J.L., van Dartel, D.A., Robinson, J.F., Pronk, T.E., & Piersma, A.H. (2011). Gene set assembly for quantitative prediction of developmental toxicity in the embryonic stem cell test. Toxicology, 284, 63-71.

Sipes, N.S., Martin, M.T., Reif, D.M., Kleinstreuer, N.C., Judson, R.S., Singh, A.V., Chandler, K.J., Dix, D.J., Kavlock, R.J., & Knudsen, T.B. (2011). Predictive models of prenatal developmental toxicity from ToxCast high-throughput screening data. Toxicol Sci, 124, 109-127.

Wu, S., Blackburn, K., Amburgey, J., Jaworska, J., & Federle, T. (2010). A framework for using structural, reactivity, metabolic and physicochemical similarity to evaluate the suitability of analogs for SAR-based toxicological assessments. Regul Toxicol Pharmacol, 56, 67-81.

Wu, S., Fisher, J., Naciff, J., Laufersweiler, M., Lester, C., Daston, G., & Blackburn, K. (2013). Framework for identifying chemicals with structural features associated with the potential to act as developmental or reproductive toxicants. Chem Res Toxicol, 26, 1840-1861.