CONCLUSIONS AND RECOMMENDATIONS

The aim at the outset of the workshop was four-fold; here the conclusions and recommendations arising from this workshop will be discussed under each of the four relevant headings:
Define Epigenetics and understand its potential value for reproductive toxicology
During the workshop, speakers defined Epigenetics as follows:
“Heritable modifications, superimposed on DNA base sequence, that regulate gene expression.” (Jessica LaRocca).
“Heritable information governing a cell state unrelated to DNA sequence variability, or information that can be inherited from a parent cell that is not encoded in the DNA sequence.” (John Greally).
“Chemical modifications of DNA that control expression of genes.” (Daniele Fallin).
“Chemical modifications of chromatin (histone PTMs, ncRNAs) which affect gene expression and may be heritable, and play a role in reproductive toxicology.” (Peter Alestrøm).
All four definitions seem concordant. The potential value for reproductive toxicology is that the study of epigenetics may elucidate important mechanisms that address why and how early life exposures can result in adverse health outcomes later in life (i.e. epigenetic analysis may contribute to understanding mechanism/mode of action). However, since the linkage of specific epigenetic alterations to adverse apical outcomes has not been established, it is too early to causally link epigenetic changes to altered health outcomes, and thus, too early to apply routine epigenetic assessments to regulatory applications.
Understand the relationship between epigenetic change and adverse endpoints
Proof of principle has not yet been fully established; however methods to help achieve this were described during Day 1. Whilst there have been numerous studies attempting to test how toxicant exposures during pregnancy affect the epigenome of offspring, these studies are poorly interpretable and poorly reproducible (OECD, 2012). In order to move forward with our understanding of the relationship between epigenetic change and adverse endpoints, future studies should be carefully designed to yield high confidence, high interpretability and high reproducibility. As such, guidelines of best practice should be developed and disseminated. The next step in furthering our understanding of the relationship between epigenetic change and adverse effects is to identify strong apical endpoints for use in models to investigate mechanisms of toxicity and causal linkage between epigenetic and apical endpoint changes. Research proposals and model compounds to achieve this were identified during the two-day workshop.
The workshop also identified several areas to be included in future epigenetic considerations regarding epigenetics:
Agreement on definitions and semantics.
Decide and justify what to measure: DNA methylation (including DNA hydroxymethylation) and histone post-transcriptional modifications of genes and miRNAs.
Decide and justify what model systems are relevant to human health (rats, zebrafish, in vitro assays, etc).
Use a consortium approach: multi-disciplinary engagement in the design of epigenetic studies is a requirement. Necessary disciplines may include: toxicology, regulatory toxicology, epidemiology, molecular epigenetics, statistics, bioinformatics, and developmental / reproductive biology.
Develop a Roadmap for the practical use of epigenetic studies in regulatory applications
Regulators stated that they need more data on apical and epigenetic endpoints for chemicals of high concern and that this could be extracted from augmented TGs. However, proof of principal is needed before epigenetics can be incorporated into regulatory applications: models with strong apical endpoints are required to investigate epigenetic mechanisms of toxicity and validate a robust functional linkage between epigenetic and apical (adverse) endpoints. The following elements will be required to achieve this:
Model compounds should be selected on the basis of a strong understanding of known phenotypic (apical endpoint) effects that are relevant to the hypothesis (e.g. if the endpoint is male infertility, then this is not an appropriate model to test for effects in F2 or F3 offspring because no offspring will be produced). Possible model compounds were suggested:

The organism of choice: choices must be based upon a thorough understanding of their advantages and limitations with regard to risk assessment in humans. Biological relevance to the human and mechanistic understanding to underpin regulatory utility is the primary driver. Cost, time and throughput criteria should also be considered. For example, zebrafish are evolutionarily more distant from humans compared to mammalian models, and their eggs are pre-treated with chemical preparations. However, they have the advantage that F2 is sufficient to study transgenerational inheritance as compared to F3 in mammals (however, workshop participants agreed that investigating transgenerational effects would not add value at this time). Some of the commonly used rat strain reference genomes are poorly annotated and may require upgrading with additional genome sequencing to maximise interpretation of experimental data and make them more useful as research models. However, their ability to gestate offspring makes the rat model more relevant for some applications when extrapolating to humans.
Consistent and standardised data management and transparency of experimental design: Critical to the practical use of epigenetic studies is that they be of high confidence, high interpretability and high reproducibility – which have not been the case to date. Therefore, participants propose the need for developing and disseminating standard operating procedures and data interpretation principles.
A consortium approach is required in both the design and implementation stage of epigenetic projects. Bioinformatics expertise will be necessary to ensure proper analysis of high content data, but also required will be experts in toxicology, regulatory toxicology, epidemiology, molecular epigenetics, histopathology, and developmental/reproductive biology.
Defining normality for the epigenetic endpoint(s): including normality at the time of analysis and normality within the system (tissue, cell, etc).
In vitro studies are needed to complement in vivo studies as they will help elucidate and validate mechanistic understanding, including secondary mechanisms (Kanno et al, 2013).
Additional elements such as dose-response and No-Effect-Level determination, exposure route and stability vs. transience must also be addressed. Epigenetic study must add value over and above what is already available in terms of mechanistic insight or predictive capacity.

Generate a prioritised research agenda
A consensus should be reached on the study type, species and strain. The studies should be performed under standardised conditions as required for regulatory studies (e.g. OECD TG 421, 2015). It might also be useful to ask CROs or companies’ experimental facilities to provide control tissue out of current studies. This would save animal usage as well as costs and would guarantee defined conditions. Three possible research proposals were outlined as shown below:
Develop in vivo exposure models that will provide reproducible apical and epigenetic endpoints that can be used for correlative studies involving molecular and cellular assays. Complement with in vitro studies to further elucidate and validate mechanistic understanding and markers. Include the identification of sensitive and predictive early epigenetic markers for latent adverse outcomes following early life exposure. Evaluate feasibility of these epigenetic markers for relevance in human health risk assessment.
Define epigenetic normality across different laboratories, species and tissues.
_{Develop an “Enabling Resources in a Data Analysis and Coordination Centre” for data management and analysis standardisation.}