Hypothesis-driven approach to developmental toxicity assessment: Using mechanistic information to inform testing.

Developmental toxicity assessment relies on standardized guideline protocols in which animals (usually rats and/or rabbits) are exposed to the test(s) agent(s) and pregnancy outcomes are assessed at an organismal level. Increasing information about mechanisms of toxicity now allows improved selection of in vivo and in vitro models for assessing developmental toxicity and prediction of developmental outcome by investigating the mode of action (MoA) of the test agent, allowing for a more flexible resource-efficient approach. Read-across, already widely used for chemical assessment, relies on a combination of cheminformatics to select suitable analogs and any of a variety of methods to prove biological similarity and/or a common metabolic pathway. Some of these methods include high-throughput test batteries (e.g., ToxCast) and transcriptomics linked to large databases of gene expression profiles. These can be used to both generate and test hypotheses about MoA of novel compounds. Increasing availability of induced pluripotent stem cells provides greater range of biological models that closely mimic the human biology relevant for addressing a specific hypothesis. Examples are given of how (1) understanding mode of action can be used to identify activity cliffs in a series of analogous chemicals, (2) the use of metabolism data in an example demonstrating that closely related analogs do not all have to be tested in developmental toxicity protocols, and (3) how analysis of gene expression can be used to identify divergent pharmacology in similar chemicals. It is possible using the approaches described to design more flexible, hypothesis-driven approaches to assess developmental toxicity.