APPROVED: 23 November 2016 AMENDED: 7 April EFSA Scientific Colloquium 22. Epigenetics and Risk Assessment: Where do we stand?

Transcription

1 EVENT REPORT APPROVED: 23 November 2016 AMENDED: 7 April 2017 doi: /sp.efsa.2016.en-1129 EFSA Scientific Colloquium 22 Epigenetics and Risk Assessment: Where do we stand? European Food Safety Authority and Tina Bahadori 1, David Bell 2, Sandra Ceccatelli 3, Raffaella Corvi 4, Christer Hogstrand 5, Sharon Munn 4, Eric Nilsson 6, David Spurgeon 7, Jochen Vom Brocke 2, Diane Wray-Cahen 8, Matt Wright 9, Marco Binaglia 10, Jean-Lou Dorne 10, Nikolaos Georgiadis 10, Andrea Germini 10, George Kass 10, Tobin Robinson 10, Annamaria Rossi 10, Reinhilde Schoonjans 10, Andrea Terron 10, and Hubert Noteborn 11 Abstract The issue of epigenetic changes and their impact on human health and life span was prominently discussed at EFSA s second scientific conference Shaping the future of food safety, together in Milan. Epigenetic changes are molecular changes mainly in chromatin, such as DNA methylation, histone modifications, that modulate gene expression directly or indirectly through the expression of noncoding RNAs. There is increasing evidence to suggest that individual lifestyles, nutrition and environmental stressors can affect epigenetic processes and as a result, alter phenotypes, longevity, health and disease both within generations (from embryogenesis to adulthood) and in a transgenerational manner. In response to the interest in this issue, EFSA has selected epigenetics as the subject of its 22nd scientific colloquium, which was held on 14 and 15 June 2016 in Valencia, Spain. About 100 scientists, risk managers and policymakers discussed where we stand regarding our knowledge of epigenetic mechanisms. The overall objective of the discussions was to identify the potential role of epigenetics in food safety risk assessment. The colloquium was organised around four discussion groups looking at the following themes: incorporating epigenetics data in mode of action analysis; epigenetics and chemical risk assessment in humans; epigenetics in risk assessment of farmed animals for food production; epigenetics and environmental risk assessment. The main takehome message from the colloquium was to ask and seek answers to those questions that will increase our understanding of epigenetics. What do epigenetic modifications mean for safety assessment? How do we study them? What is the size of such modifications that we need worry about? Cooperation and collaboration between the various scientific disciplines and with the clinical side of epidemiology was identified as a necessary strategic element to improve scientific risk assessment. Key words: epigenetics, scientific colloquium, risk assessment, DNA methylation, histone acetylation, non-coding RNA Question number: EFSA-Q Correspondence: any enquires related to this output should be addressed to scer@efsa.europa.eu 1 Environmental Protection Agency, USA 2 European Chemicals Agency EFSA Supporting publication 2016: EN-1129

2 3 Karolinska Institute, SE 4 Joint Research Centre, European Commission 5 University College London, UK 6 Washington State University, USA 7 Centre for Ecology and Hydrology, UK 8 Foreign Agricultural Service of United States Department of Agriculture, USA 9 Newcastle University, UK 10 European Food Safety Authority 11 Netherlands Food and Consumer Product Safety Authority, NL European Food Safety Authority, 2016 Disclaimer: The views or positions expressed in this publication do not necessarily represent in legal terms the official position of the European Food Safety Authority (EFSA). EFSA assumes no responsibility or liability for any errors or inaccuracies that may appear. Amendment: Editorial corrections were carried out on page 1-2 (authors list), 22 (references) and (county of origin of participants) that does not materially affect the contents or outcome of this scientific output. To avoid confusion the older version has been removed from the EFSA Journal, but is available on request, as is the version showing all the changes made. Suggested citation: EFSA (European Food Safety Authority) and Bahadori T, Bell D, Ceccatelli S, Corvi R, Hogstrand C, Munn S, Nilsson E, Spurgeon D, Vom Brocke J, Wray-Cahen D, Wright M, Binaglia M, Dorne JL, Georgiadis N, Germini A, Kass G, Robinson T, Rossi A, Schoonjans R, Terron A and Noteborn H, EFSA Scientific Colloquium 22 - Epigenetics and Risk Assessment: Where do we stand? EFSA Supporting publication 2016: EN pp. ISSN: European Food Safety Authority, 2016 Reproduction is authorised provided the source is acknowledged 2 EFSA Supporting publication 2016: EN-1129

3 Table of contents Abstract Introduction Abstracts of speakers in the opening plenary session Summary of Discussion Groups results Discussion Group 1 - Incorporating epigenetic data in mode of action and adverse outcome pathway frameworks Discussion Group 2 - Epigenetics and chemical risk assessment in humans Discussion Group 3 - Epigenetics in risk assessment of farmed animals for food production: what do we need to consider? Discussion Group 4 - Epigenetics and environmental risk assessment: mechanisms, testing and data gaps Final plenary discussion Overall conclusions and recommendations References Abbreviations Annex A Programme of the Colloquium Annex B Participants at the Colloquium EFSA Supporting publication 2016: EN-1129

4 1. Introduction The 22 nd meeting in the EFSA Scientific Colloquium Series was held in Valencia, Spain on June 2016 and addressed the status of knowledge around epigenetics and its implications for the risk assessment performed by the EFSA. Epigenetic changes are molecular changes in chromatin, such as DNA methylation and histone modifications, that modulate gene expression without affecting the encoded gene sequence. Expression of non-coding RNAs is also sometimes considered an epigenetic mechanism of gene regulation. There is increasing evidence to suggest that individual lifestyles, nutrition and environmental stressors can affect epigenetic processes and as a result modify gene expression which leads to changes in phenotypes, longevity, health and disease both within generations (from embryogenesis to adulthood) and in a trans-generational manner. Considerable research in this area has focused on vertebrates (including human), although the number of studies published on invertebrate species, plants and even prokaryotes is rapidly increasing including studies on classical biological model species, as well as non-model organisms. This 22 nd EFSA Scientific Colloquium brought together experts dealing with epigenetics and human health, animal health and environmental risk assessment of chemicals. The objective of the colloquium was to hold an open scientific debate on where we stand regarding knowledge of epigenetic mechanisms and the potential use and integration of epigenetic data in chemical risk assessment. The colloquium was run as a plenary session with keynote presentations followed by a discussion session organised in four breakout groups focussing on the following topics: Discussion Group 1 discussed the incorporation of epigenetic data in mode of action and adverse outcome pathway frameworks; Discussion Group 2 focused on the implications for chemical risk assessment in humans; Discussion Group 3 looked at the risk assessment needs for farmed animals for food production; Discussion Group 4 focused on the implications for environmental risk assessment. The Colloquium was attended by approximately 90 experts from Europe, Canada and the USA; the present publication reports the abstracts of speakers in the opening plenary session and summarizes the discussions and conclusions of the Colloquium. 2. Abstracts of speakers in the opening plenary session Environmental and nutritional effects on epigenetic regulation Robert Feil, Montpellier Institute of Molecular Genetics (IGMM), CNRS, France Epigenetic phenomena in animals and plants are mediated by DNA methylation and by other covalent chromatin modifications. Such epigenetic modifications are often stably maintained, but they are also reversible. There is a rapidly growing interest in whether the environment can modulate the establishment and maintenance of epigenetic modifications, and could thereby influence gene expression and phenotype. Studies on experimental animal models have indeed provided evidence that chemical pollution, dietary components, temperature changes and other external stresses can have long-lasting effects on development, metabolism and health, sometimes even in subsequent generations. For many of these examples, however, it remains unclear how the environmental cues trigger observed epigenetic changes and whether these are causally involved in the aberrant phenotypes (Feil & Fraga, Nature Reviews Genetics 13, 97, 2012). Nevertheless, genetically controlled studies in mice have shown clear epigenetic effects of nutrition and chemical components, particularly of endocrine disruptors, and highlight that certain genes are highly susceptible. For different reasons, environmental epigenetic effects have been more complicated to explore in humans, and it has been 4 EFSA Supporting publication 2016: EN-1129

5 generally difficult to draw firm conclusions. Most reported epigenetic changes are minor and it has been difficult to link these to the observed phenotypic effects. In future research it should be important to address how precisely chemical components and nutrition lead to epigenetic changes at specific genes. It should be important also to explore which regions in the human genome are most highly susceptible to epigenetic changes and whether these could be useful as biomarkers to assess environmental and nutritional effects. Molecular mechanisms of chemical injury: Input from epigenomics Jos Kleinjans, Department of Toxicogenomics, Maastricht University, The Netherlands The epigenome may be defined as a record of the chemical changes to the DNA and histone proteins of an organism which can be passed down to an organism's offspring. A particular feature of the epigenome is that unlike the underlying genome which is largely static, the epigenome can be dynamically altered by environmental conditions, e.g. by toxic chemicals. However, investigations on the inducibility of the epigenome by toxicants, in particular in cellular assays in vitro, are still rare. We therefore undertook to analyze the dynamics of the epigenome of the human A459 lung cell line during a 2 week exposure to arsenic (As), which due to its detoxification through methyl binding, presents a prototypical toxicant affecting DNA methylation patterns. We indeed found dose- and timedependent modifications of the epigenome, to some extent complemented by concordant changes in gene expression levels, in particular centered around the p53-mediated tumor suppressor network, thus subscribing to the suggested lung carcinogenicity of As. We thereupon questioned whether also the epigenome of human primary cells is inducible, and if so, whether such alterations are persistent, e.g. after chemical exposure has stopped. Within the context of the EU FP7 project DETECTIVE, we therefore explored primary human hepatocytes (PHHs), treated daily with prototypical liver toxicants known to induce specific endpoints of liver injury, e.g. aflatoxin B1 (hepatocarcinogenicity), cyclosporine A (cholestasis) and valproic acid (steatosis), for 5 days, followed by a 3 days wash out period. Overall, we concluded that the epigenome of PHHs indeed appears inducible by these hepatotoxicants. Epigenomic responses appear highly dynamic, but also to some extent persistent. Also novel response patterns have been disclosed which to some extent can be translated to gene signatures from diseased livers, thus identifying promising biomarkers for repeated dose toxicity in vitro. The role of epigenetics in toxicological risk assessment of furan in food Dan Doerge, National Center for Toxicological Research, FDA, Jefferson, USA Furan is an important thermal processing contaminant found in many common cooked foods. Metabolic activation of furan by CYP 2E1 occurs primarily in the liver, where the formation of a reactive electrophilic species, cis-2-butene-1,4-dial (BDA) is the proximate cause of tissue damage. At sufficient dosing, furan is hepatotoxic and hepatocarcinogenic in rodent bioassays. Risk assessment of furan in foods entails significant uncertainties, including extrapolations of toxic effects produced at doses orders of magnitude above upper-bound human dietary intake, inter-species differences in responses between rodents and humans, and carcinogenic mechanisms. Furan carcinogenicity testing was undertaken in rats to address uncertainty regarding dose-response and mode of action, and included evaluation of toxicokinetics and metabolism; in vivo genotoxicity; subchronic and chronic toxicology using consensus histopathology; and mechanistic evaluations, with a focus on epigenetic endpoints. After oral administration, furan partitioned into the liver and was rapidly eliminated. Despite extensive time- and dose-dependent hepatotoxicity, the major BDA-DNA adduct was not detected in the liver after single or repeated dosing. Similarly, furan dosing produced negative results for several in vivo genotoxicity endpoints in Big Blue rats (e.g., micronucleus, cii, Hprt, Pig-a). However, furan treatment modified in a dose-dependent manner the methylation of liver DNA and acetylation/methylation of associated histones, which were linked with gene-specific expression changes. In some cases, the epigenetic changes persisted after dosing had stopped. A strategy of combining transcriptomic and DNA methylation analyses was used to identify high priority furan treatment-specific genes affected in the liver. All critical toxic effects in rats treated chronically with furan, including neoplastic and non-neoplastic, were evaluated for mode of action and dose-response. 5 EFSA Supporting publication 2016: EN-1129

6 Epigenetic endpoints proved useful in expanding the mode of action analysis for furan, by identifying reversible and irreversible changes in specific genes associated with stem cell differentiation and cancer-related pathways in rat liver. Epigenetics in risk assessment of farmed animals for food production Heiner Niemann, Institute of Farm Animal Genetics, FLI Mariensee, Germany Epigenetics entails the study of changes in gene function that are mitotically or meiotically inherited, but are not based on a change in DNA sequence. Epigenetic changes, predominantly DNA methylation and histone modifications, play a crucial role in defining the temporal and tissue specific gene expression profile. Assisted reproductive technologies (ARTs) are well developed in domestic cattle and include artificial insemination (AI), embryo transfer (ET), in vitro embryo production (IVP), and somatic cell nuclear transfer (SCNT). ARTs have been used to shorten the generational interval, to propagate valuable genetic material from breeding populations, and for biomedical and reproductive research. Moreover, bovine oocyte/embryo development is increasingly being used as an important model for studies in humans. Albeit ARTs are useful tools for improving reproduction in the cattle industry, some of the procedures involved could potentially affect gametes and embryos by causing aberrant epigenetic marks which in turn would lead to aberrant gene expression profiles. Increasing evidence indicates that ARTs may be associated with changes of the epigenetic profile and an elevated percentage of imprinting disorders. Despite the widespread application of ARTs under commercial conditions, the exact mechanisms leading to epigenetic disorders and aberrant gene expression are not yet fully understood not only in the bovine species, but also in the mouse model and in humans. To improve the results of ARTs, further studies are necessary to understand how epigenetic regulation is affected by ART in gametes, early embryos and post-implantation. A battery of diagnostic tests to identify, prevent and/or reduce epigenetic disorders and changes in gene expression after use of bovine assisted reproductive technologies could be beneficial in this respect. However, current evidence suggests that risks with regard to food safety, derived from epigenetics in livestock are minimal only and are likely covered by the current regulatory system. Impact of epigenetics in environmental risk assessment Kevin Chipman, School of Biosciences, The University of Birmingham, UK Epigenetic modifications form part of the mechanism whereby organisms respond to their environment including in response to their diet and various stressors including chemical pollutants. These modifications can be beneficial or protective but, particularly when occurring at a critical stage in development, can lead to delayed onset of disorders including cancer. There is also the possibility of transgenerational effects depending on the species and the level and timing of the exposure to the stressor. Modifications to the epigenome (both DNA and histone changes) can be detected and monitored in non-model species taken from the environment and this will be demonstrated via a study of DNA methylation in flatfish taken from impacted environments that show a high incidence of liver tumours. Based on studies in twins, that show epigenetic changes throughout a lifetime independent of genetics, we propose the use of epigenetic footprints to inform on prior stress in organisms. Clonal daphnia provide a useful model for such studies including the use of resurrection ecology. The time has arrived for epigenetics to be used as part of environmental monitoring programmes but for routine risk assessment it is premature due to deficiencies in knowledge particularly around beneficial versus adverse responses, specificity of stress induced changes, reversibility and the basis of potential transgenerational effects. 6 EFSA Supporting publication 2016: EN-1129

7 3. Summary of Discussion Groups results 3.1. Discussion Group 1 - Incorporating epigenetic data in mode of action and adverse outcome pathway frameworks Chair: David Bell, European Chemicals Agency Rapporteur: Matt Wright, Newcastle University, UK Epigenetic changes are molecular changes in the chromatin that modulate the expression of genes. The changes include DNA methylation, histone modifications and the expression of non-coding RNAs. There is increasing evidence to suggest that life style and environmental stressors can affect epigenetic processes and as a result alter phenotypes, longevity, health and disease both within generations (from embryogenesis to adulthood) and in a trans-generational manner (Kaelin & McKnight, 2013; Lalevee & Feil, 2015; Qureshi & Mehler, 2013). Mode of action (MOA) and adverse outcome pathway (AOP) are frameworks that provide a basis for identifying sequential chains of causally linked and empirically observable events at different levels of biological organisation that lead to an adverse health or ecotoxicological effect. These frameworks have been developed over the past 10 years as central elements that provide a mechanistic basis to support chemical risk assessment (Edwards et al., 2016; Meek et al., 2014; Simon et al., 2014). While the MOA and AOP frameworks have commonalities with regards to the identification of measurable key events (biological perturbations) that are necessary for toxicity, a key difference is the inclusion of toxicokinetic processes in the MOA framework whereas the AOP framework was initially developed to consider relevant chemical-biological interactions in a toxic outcome. Recent attempts have been made to use OMICs data, and in particular transcriptomics data, in MOA frameworks through the identification of functional pathways and their incorporation as key events (Moffat et al., 2015). The aim of this discussion group was to critically discuss whether epigenetic events that will have an impact on gene expression can be identified and integrated in MOA and AOP frameworks as key events. In addition, the discussion group aimed at identifying data gaps and research needs Discuss key epigenetic mechanisms and their potential relevance to key events under MOA/AOP analysis. With regard to epigenetic changes and their potential relevance to key events in an AOP, the discussion group considered an epigenetic event could be a molecular initiating event or a key event. Whilst it was considered that an adverse outcome may not necessarily be dependent on an epigenetic change (i.e. that an epigenetic change might not be a key event in an AOP), identifying the key events may in some cases require knowledge of an epigenetic change(s) which is required for the adverse effect. It was therefore considered that it was necessary to incorporate understanding of epigenetic changes in a study in order to have a full understanding of mechanism(s) of action which rely upon epigenetic key events. Given the current state of knowledge and that the primary function of epigenetic alterations is in modulating gene expression, the discussion group considered that epigenetic changes which do not give rise to a biologically significant change in gene expression are unlikely to cause toxicologically relevant effects. At the present time, there is limited data to indicate that specific epigenetic changes (e.g. methylation changes at a particular locus in the genome) lead to predictable adverse outcomes. However, it is necessary to be able to link specific epigenetic changes to specific adverse outcomes as a prerequisite for establishing causality, i.e. that the specific epigenetic change causes the specific adverse outcome. The demonstration that the specific epigenetic change is required for the specific adverse outcome is also a prerequisite for incorporating the epigenetic change into an AOP approach as a key event or molecular initiating event. The discussion group considered that chemically-induced epigenetic changes may be key events in a transgenerational or delayed onset mode of action. The discussion group emphasised that contextualising epigenetic datasets in an MOA/AOP framework is helpful to understand their relevance for chemical toxicity What in vitro, in vivo assays and non-testing strategies are available to generate relevant data on epigenetic changes for incorporation into MOA analysis and AOP development? Discuss study design (choice of tissues, 7 EFSA Supporting publication 2016: EN-1129

8 experimental time points, target organ, cellular model, in silico tool) and biological relevance. Specific techniques were not discussed in detail. It was mentioned however that two publications on in vitro and in vivo testing strategies for epigenetic endpoints in the context of chemical risk assessment (Grealy and Jacobs, 2013; Marczylo et al., 2016) and a recent ECETOC report on epigenetics and risk assessment that includes methodological aspects 1 are available. In vitro methods were considered to have well described limitations which may complicate interpretation if used to support a toxicological investigation. For example, the in vitro cell system may have an aberrant epigenome. If proliferation of cells is required for an epigenetic key event, this may influence the results of an assessment performed in the in vitro cell system that does not support proliferation. Likewise, if metabolic activation is required for the toxic action of a substance, then the metabolic capability of the in vitro cell system may be an experimental limitation too. However, the discussion group generally agreed that it is currently possible to detect (quantitative) epigenetic changes such as DNA methylation changes in cells or tissues. Despite technological advances in associated technologies (such as DNA sequencing) and reduced costs, the analyses of data remains a niche ITbased skill and the analyses are relatively expensive. Furthermore, if it is desired to identify a point of departure for a key epigenetic event in a toxicological study, the experimental design will require at the very least - an appropriately powered analysis of changes over time and with dose, which will amplify costs. At the present time, predicting adverse effects from epigenetic changes was not considered possible by the Discussion Group; main challenges are specificity (that a specific epigenetic change causes a specific adverse effect) and issues surrounding validation (e.g. is the effect robust, does it happen in all model systems/ experimental setups, does the specific epigenetic change always cause the specific adverse effect, etc.). In the specific case of adverse outcome(s) which are linked to trans-generational epigenetic changes, the group concluded that, although human data suggest a causative effect (that the epigenetic changes cause the transgenerational adverse outcome), it still remains necessary to demonstrate causality for this hypothesis in experimental models. It was emphasized that if in the future a model system is developed to screen chemicals for their ability to induce epigenetic alterations, this model must be relevant and well characterised for the adverse outcome endpoint How can epigenetic data be integrated with other OMICs data for MOA analysis and/or AOP development? The discussion group considered that linking epigenetic data with other OMICs data will enrich the dataset and may even be essential to identify key events in a MOA/AOP approach (e.g. confirming that specific epigenetic changes lead to specific changes in gene expression). Such combined data illustrates that these are complementary approaches, and together may give clues to other key events within an MOA. The group stressed the importance of using a structured framework, such as MOA or AOP, to evaluate the relevance of new information such as epigenetic or genomic data What are the data gaps and research needs? The group discussed these issues at some length since it was generally agreed that epigenetics could in principle play a role in all non-communicable diseases. The group identified a number of critical data gaps for the future use of epigenetic data in MOA and AOP frameworks. These include: to what extent are epigenetic changes reversible?; to what extent does the human epigenome vary without leading to phenotypic differences or adverse effects?; what is the dose-response relationship between chemically induced epigenetic effects and the subsequent adverse effect? (and therefore whether there might be a threshold below which epigenetics changes do not lead to an adverse outcome); to what extent species differ in their sensitivities to epigenetic changes? (and which experimental species is best predictive of man?); to what extent are in vitro models relevant to in vivo responses? The discussion group therefore foresees significant research need. However, some relevant data may already be available from research in other disciplines such as epigenome signatures from clinically diseased organs or studies focused on the developmental origins of health and chronic disease (e.g. foetal origins of disease). This underscores future need for interdisciplinary collaboration and a 1 WR25: Workshop on Omics and Risk Assessment Science 3 October 2013 Workshop Report no.25. Omics and risk assessment science 25-26) 8 EFSA Supporting publication 2016: EN-1129

9 targeted exploitation of available human epidemiological and clinical data. A specific research priority could be identifying chemically-induced specific epigenetic changes that cause predictable adverse outcomes (i.e. epigenetic key events), and clarifying whether these epigenetic key events are specific to particular classes of chemicals or have broader relevance to chemical toxicology. Further, it was considered essential to know which epigenetic changes are biomarkers of exposure and which are predictive of adverse outcomes. Hence, the identification of robust epigenetic biomarkers is also an important scientific goal. Agreeing that some specific epigenetic changes may be key events and predictive of adverse outcomes, the discussion group noted that in the future it may be possible to identify a chemical as an epigenotoxin. Key research needs include a greater understanding of chemically induced epigenetic change and its function, and the development of validated in vitro and in vivo screening assay systems for the identification and characterisation of epigenetic alterations in a toxicologically-relevant manner Discussion Group 2 - Epigenetics and chemical risk assessment in humans Chair: Tina Bahadori, Environmental Protection Agency, USA Rapporteur: Raffaella Corvi, Joint Research Centre, European Commission Human risk assessment of chemicals is traditionally based on the identification of the hazard posed by the chemical of interest, the characterisation of the hazard through a battery of in vivo and in vitro experimental tests and the derivation of health-based guidance values that are compared with exposure data. In vivo studies used to identify and characterise hazards typically include among others, body weight, organ weight, gross and histopathology, serum chemistry, haematology, reproductive function. Using the NOAELs reported from these in vivo studies or following benchmark dose (BMD) modelling of the experimental data, predictive cancer-based and non-cancer-based reference points (RP) (points of departure) are then identified and used to establish appropriate health-based guidance values. In addition to the biochemical or pathology-based types of experimental data, molecular data derived from OMICs technologies (transcriptomics, proteomics, metabolomics) have also recently received attention for their applicability in hazard characterisation (EFSA, 2014). Indeed, in vivo toxicogenomics and toxicoproteomics data have been successfully applied to BMD modelling and predictive cancerbased and non-cancer-based RP identification and have been reported to complement other toxicity data (Moffat et al., 2015; Thomas et al., 2011; Thomas et al., 2012; Webster et al., 2015; Chepelev et al., 2014). Thus, OMICs technologies have the potential to provide important data in support of hazard identification and hazard characterisation. Epigenetic changes, through DNA methylation, histone modifications and the expression of non-coding RNAs, modulate the expression of genes, and hence contribute not only to changes in cellular mrna and protein levels but may also generate novel biomarkers. The aim of this discussion group was to critically discuss whether the epigenetic data can be integrated in chemical risk assessment by supporting the identification of RP for hazard characterisation. In addition the discussion group aimed at identifying data gaps and research needs Do toxicity studies with their current design allow to measure adverse effects resulting from epigenetic changes? Give examples. There is to date accumulating evidence that epigenetic changes, which may control gene expression, occur in the genome upon exposure to various environmental, chemical and biological stressors. Technological advances now enable the measurement of epigenetic modifications of various natures. It is therefore envisaged that toxicity studies with their current design may in principle allow to monitor changes and to measure adverse effects resulting from epigenetic modifications. However, the current knowledge does not allow agreement on which endpoints to capture and whether these lead to an adverse outcome or just a step within a molecular pathway. 9 EFSA Supporting publication 2016: EN-1129

10 In this context, a more general question was raised, namely whether our current risk assessment paradigm adequately captures biological complexity or whether it should be revisited to better reflect the extent to which complexity is being evaluated in understanding biology. Several limitations of the current paradigm were identified such as its limited ability to assess multiple exposures, the timing (window) of exposure and multigenerational and transgenerational effects. Biology is clearly more complex than our current toxicological models. One may ask whether the present paradigm is ready for the emerging science or if the science is ready for a new paradigm. There was some concern that adding epigenetics-related toxicity parameters would increase the burden to toxicological testing and could generate higher complexity as well as lead to a detection of too many spurious changes resulting in an increase of uncertainty. Nevertheless, it was recognised that epigenetic data have the potential to support risk assessment in several ways; for instance, by revealing AOPs and identifying individual key events. They also may lead to the identification of early markers of effects, which may be more sensitive than those picked up using standard tests. Such developments may positively impact on testing costs. However, such markers need to be evaluated with caution to avoid that a too high sensitivity makes the markers irrelevant for assessing safety. Another contribution of epigenetic data in risk assessment is their potential to explain inter-individual variability. A major obstacle for the integration of epigenetics in risk assessment is the lack of knowledge on the baseline and the contribution of background variability. It is at the moment not clear if epigenetic modifications represent main regulators of gene expression or just effect modifiers. Epigenetics is an important and natural part of biology. Therefore, there is a need to understand normal and background biology in order to appreciate epigenetic effects. We might ask whether we are expected to move more and more towards a personalised or customised risk assessment. The importance of communicating with the public in a simple and understandable manner was highlighted. However, there was no full agreement within the group on how this would be best achieved. On one hand it was mentioned that a more complex risk assessment is not necessarily beneficial for conveying a simple message. On the other hand, it was considered crucial not to compromise on the science and make the effort to transmit a simple message despite the complex underlying science Discuss key issues to integrate in vivo data reporting epigenetic effects, including non-coding RNA-based biomarkers such as mir-122, in BMD modelling or identification of RP for hazard characterisation? This question assumed more aspirational readiness for the science than it was revealed by the state of art. Even if those data would be available, in addition to standard requirements, they would not necessarily be integrated into risk assessment. Data availability is not the only bottleneck to data integration. Sometimes good quality data are not considered for risk assessment for different reasons that may include policy reasons, lack of experience or difficulty in interpretation. For instance, epidemiological or human evidence are at times not included in chemical risk assessments, because of, for example, the lack of experience among traditional risk assessors to reconcile and integrate data that is not derived from studies or tests conducted using standardised protocols. While using epigenetic effects for hazard characterisation was considered premature, understanding dependency between epigenetic changes and adversity was regarded as a priority. In the following discussion the definitions of adverse effect and biological relevance used by EFSA were considered. Adverse effect: An effect is considered adverse when leading to a change in the morphology, physiology, growth, development, reproduction or life span of an organism, system or (sub)population that results in an impairment of functional capacity to compensate for additional stress or an increase in susceptibility to other influences (WHO/IPCS EHC 240). Biologically Relevant Effect: Biologically relevant effect is defined as an effect considered by expert judgement as important and meaningful enough for human, animal, plant or environmental health. It 10 EFSA Supporting publication 2016: EN-1129

11 implies a change that may alter how decisions for a specific problem are taken (EFSA Scientific Committee, 2011). Based on these definitions and on the existing science, it was clearly considered not possible, at this time, to predict an adverse effect from epigenetic changes. It was considered necessary to initially define an adverse effect in relation to the different endpoints and to the modulation of the underlying biological variability. Once this is defined, available analytical methods can be identified which can be validated in the future to ensure harmonisation of the test methods. Finally, the generation of sufficient data on epigenetic changes and the consequent understanding of the sequence of events from early changes to adverse effects may contribute to hazard characterisation. It might be necessary to reconsider the measure of adverse effects in the current risk assessment paradigm. Can animal models mimic human effects appropriately? Relying on efficient epidemiological studies may help to identify more human relevant and global endpoints of adversity. It was also asked whether the current risk assessment framework can accommodate all the emerging science. Fundamental science is often the driver in bringing hot topics into applied science, but with new endpoints, caution is needed not to repeat errors done in the past. Are we satisfied with a pragmatic approach to risk assessment or do we need a more aspirational risk assessment? It was recognised that every study is a compromise between design and choice What in vivo, in vitro assays and non-testing strategies are available to generate relevant data on epigenetic changes for incorporation into risk assessment frameworks? Discuss study design (choice of tissues, experimental time points, target organ, cellular model, in silico tools) and biological relevance. In silico models could be developed when data on more chemicals become available. For instance, it could be envisaged that based on the growing number of studies mapping methylation patterns of exposed cells or tissues, appropriate QSARs could be developed. These models could be used as part of a holistic approach to priority setting and for selection and design of additional studies. Well-designed epidemiological studies that measure epigenetic modifications may also help in generating relevant data. At present, data from a few older epidemiological studies are usually cited to demonstrate impacts of epigenetic changes in humans. From a risk assessment point of view, ultimately the goal is to establish a Reference Point (RP, also known as point of departure). Even if a RP were derived from epigenetic data, and if this RP were different from the existing ones, would this represent a meaningful contribution to risk assessment? Considering that the risk assessment framework is an overall assessment as embraced by the 21st century approach that points to integration of information, it is improbable that epigenetic based RP would contribute on its own to risk assessment without a thorough understanding of the underlying mechanisms, anchoring it to a certain endpoint and considering other information. It was also suggested that current assays may already capture some epigenetic effects within the risk assessment; at least this seems to be the case for cancer endpoints. There seems to be some expectation to better address the topic of transgenerational effects in the future, as it is questionable if the current risk assessment model captures them appropriately (whether or not these are epigenetic or not). Some studies show a mix of transgenerational and evolutionary changes; this needs further clarification What are the data gaps and research needs? It was considered fundamental to initially pose the right basic questions. Given that epigenetic modifications do occur, how do these modifications modulate gene expression? If the consequences are adverse effects, is it possible to understand the extent to which these effects are adverse? And if the phenomena are understood, can they be integrated in risk assessment? And finally what are the endpoints of specific importance that should be regarded? 11 EFSA Supporting publication 2016: EN-1129

12 If we are expecting to consider epigenetic effects we need to invest in more clever study designs that would allow us to meaningfully demonstrate and probe these effects and understand if and how they should be incorporated into decision making. It is evident that, before we will be able to explain how to use epigenetics there is a need to better understand the baseline biology and variability, including inter-individual variability and inter-species differences. The ongoing exposome EU projects (HELIX, EXPOsOMICS, HEALS, HERCULES) might help in this. There is also a need for mechanistic understanding of epigenetic changes and their link to gene expression. Determining the sequence of events and the association between early events and adverse outcome would be important. In the future, when more is known, would it help to add an endpoint that is not necessarily adverse or would it create more uncertainty? This is an interesting question, as epigenetic changes that do not have any phenotypic consequence may nevertheless become bigger and be easier to monitor than tiny epigenetic effects of exposure that exert a known and relevant effect (e.g. on a specific gene). Most of the publications to date focus on a few chemicals at high exposure and a few biological parameters. There is a need to extend our knowledge and generalise what is known up to now with information on more chemicals to identify characteristic patterns for specific endpoints at lower and more relevant dose exposures. There is a need for epidemiological studies with relevant biomarkers of exposure and epigenetic effects that can be integrated with data from appropriate toxicological studies (e.g. exposure, endpoints) and mechanistic information. A multi-disciplinary approach is necessary to take advantage of the progress made in other fields and discuss how this knowledge may advance risk assessment. At present there seems to be a disconnection between risk assessment and public health. In particular, more interaction between toxicologists, epidemiologists and clinicians is required. Translational science needs to be further exploited to identify the most appropriate models in a more holistic approach Discussion Group 3 - Epigenetics in risk assessment of farmed animals for food production: what do we need to consider? Chair: Diane Wray-Cahen, USDA Foreign Agricultural Service, USA Rapporteur: Eric Nilsson, Washington State University, USA Risk assessment of animals used for food production focuses on the identification of potential hazards, hazard characterization and quantitative estimates of risk. Scientific advances in the understanding of biology can improve the hazard identification and characterization processes. Potential new tools need to be assessed for their usefulness, including whether the science is sufficiently advanced for the tool to improve hazard identification and be incorporated into the risk assessment process. Changes in an animal s epigenome can result in phenotypic changes, without changing the primary sequence of that animal s DNA. An animal s epigenome is dynamic and subject to change over time. It can be influenced by a variety of factors, including: stage of development, nutrition, environment, disease, or social factors. Epigenetic changes are a normal mechanism by which organisms respond to their environment by changing gene expression. These changes can result in a positive or negative outcome for the animal (e.g. alter normal developmental processes or disease susceptibility). Some of these outcomes can result in alterations of the phenotype that are transient, while others may have permanent impacts on animal health, including disease susceptibility and longevity, from embryogenesis to adulthood and potentially in a trans-generational manner. These phenomena have been studied in farmed animals (Freeney et al. 2014), but only to a relatively limited extent. Animal breeding is focused on animal improvement and the heritability of certain phenotypes; epigenetic variability can influence these phenotypes, including production traits. Currently breeders employ modern biotechnologies such as genomic selection, semen sorting, in vitro fertilisation, and 12 EFSA Supporting publication 2016: EN-1129

13 cloning. Some of these techniques have been evaluated for their effects on the epigenetic status of the adult animals (e.g. for cloning, see US-FDA (2008), EFSA (2008), Japan FSC (2009), for in vitro embryo production (Smith et al., 2015)). In the near future, gene editing and gene drive technologies may be added to the list; these have the potential to more precisely control genomic selection and manipulation, and may also allow for targeted epigenome editing (Qi et al., 2013). The aim of this discussion group was to critically discuss the current state of the field of epigenetics, including whether any subset of marks and changes to those marks can be causally associated with phenotypic changes and whether sufficient information exists to use these to characterize potential hazards and make inferences regarding animal health or food consumption risks. Several different aspects of risk assessment were discussed including whether epigenetic data could or should potentially inform risk assessment relative to: Food safety risk to humans: Could epigenetic data help identify potential hazards for human consumers of meat or dairy products from farmed animals? Risk of disease or poor welfare for the farmed animals: Could epigenetic data help predict whether farmed animals are likely to be more or less susceptible to disease in the future? Assessment of risk to food supply (food security): Could epigenetic data help with projections of health in farm animal populations, and whether that impacts the human food supply? The discussion group aimed at identifying data needs, data gaps, and potential research needs Do changes in epigenetic marks lead to predictable phenotype changes? This question speaks to the idea of whether epigenetic changes can be used as biomarkers to identify potential hazards and lead to improved assessments of risk of adverse outcomes or poor welfare for farmed animals. The answer from this discussion group was sometimes yes, sometimes no. There is very limited data at this point in farmed species. There may be important and relevant marks that have yet to be discovered, and the cell type within which an epigenetic change takes place may also be important. There is evidence of correlation between epigenetic changes and phenotypic changes in some cases, but there is often no evidence of causal links. That is to say, it is often not clear why a particular set of epigenetic changes would lead to a phenotypic change or a future adverse or positive outcome. This link was considered necessary for risk assessment purposes, as causal links are traditionally an important part of risk assessment. It was also noted that not all epigenetic changes represent a hazard What are key epigenetic marks? Are changes in these marks biologically significant? (How is normal defined, biologically and statistically?) There may not be specific key epigenetic marks, for which changes in individual marks could be used as predictive biomarkers of potential hazards, such as an increased risk to the animals health or food safety. There is a need for baseline data so as to define normal epigenetic variation and to determine which changes may be biologically significant. Every cell type has its own epigenome, and there may also be considerable individual variation. Suites of epigenetic changes may be more useful in this respect, with a certain proportion of a defined set of possible epigenetic changes indicating a potential hazard or increased risk. However, what those specific sets of epigenetic changes are has not yet been determined for any farmed animal or health risk. Epigenetic changes are a normal part of biological adaptation and can vary over the lifetime of an animal, as well as due to nutritional and environmental stressors. Assisted reproductive technologies (ARTs) can also trigger epigenetic changes. There is also the potential to intentionally edit the epigenome to induce phenotypic changes without changing the genome. ARTs, such as cloning, involve embryo culture technologies that can cause epigenetic changes in animals, as might be expected following any environmental exposure or change. However, improvement of culture conditions and other technical refinements have resulted in improved outcomes. From a regulatory or risk assessment perspective, perhaps normal would be better defined as the absence of certain 13 EFSA Supporting publication 2016: EN-1129

14 defined hazards or negative outcomes, rather than trying to capture or describe the breadth of expected phenotypic variation within a population of farmed animals. Are the food products from farmed animals with epigenetic changes safe for humans to eat? Likely yes. For example, food from animals produced using ARTs, including cloning, have been determined to be as safe to consume as food from other animals. Epigenetic data is not expected to replace current food safety inspection practices. Epigenetic data may help predict whether an animal will be more or less susceptible to disease in the future, but an animal that is currently unfit for human consumption would be removed from the food chain by current food safety inspection practices. At this time, there is not sufficient evidence to indicate whether or not epigenetic tools could improve food safety risk assessments. However, future research may be able to associate certain epigenetic changes with food nutrient content or altered shelf life. Are there food supply security concerns associated with farmed animals having epigenetic changes? With further research, it is possible that particular epigenetic changes may be identified that are correlated with increases in adverse (or positive) outcomes in farm animals, or perhaps with alterations in fertility. In such situations epigenetic data could potentially add value as part of risk assessment for the human food supply. However, at this point there is insufficient evidence to indicate whether or not this would be the case. It is possible that a suite of epigenetic changes in food animals could represent no hazard for humans when products are consumed as food, but may indicate the presence of a hazard for the animal in the farm environment. These data might be useful for those doing environmental risk assessments What are the available methods for assaying epigenetic marks? What would be necessary to validate them? The best-developed assays of the epigenome are for identifying sites of DNA methylation within the genome. New techniques for determining histone modifications, changes to non-coding RNA expression, changes in chromatin structure, etc. are being developed for farmed animal species. Validation of assays of epigenetic marks can refer to different aspects of validation. One aspect is validation of epigenetic changes as being in a causal pathway leading to a phenotypic change for the animal. For purposes of risk assessment this may be in the category of information that scientifically could be nice to know, but it may not reach the threshold of something that one needs to know in order to conduct a risk assessment or improve on the current risk assessment process. Opinions differed on this point within the discussion group. Other considerations for validation of assays for epigenetic marks: Is there a causal link to phenotype? Which changes are biologically relevant? Are the assays robust/repeatable? Do the effects of certain epigenetic changes hold across tissues, species, breeds and gender, etc.? Considerable research is needed to determine this for the many species and breeds of farmed animals. One must isolate pure cell types to do epigenetic analyses, as the underlying epigenome differs markedly between different types of cells. The development and sharing of isolation techniques for different species and cell types is needed What are the knowledge gaps and how they can be addressed? There are many knowledge gaps about the mechanisms of how epigenetic changes lead to changes in gene expression and, ultimately, to changes in phenotype. (Some of these are mentioned above.) This is true even in research model species, and is much more the case for farmed animal species. Therefore, descriptive research is needed for each group of farmed animals (each breed) in order to characterize baseline epigenetic information relevant to these groups. For example, when during chicken development do the primordial germ cells undergo their normal programmed DNA demethylation and re-methylation? It is during this developmental window that the animals are potentially most sensitive to the effects of environmental toxicants to induce epigenetic changes. Such research needs to be done for all farmed animals. Discussion among colloquium participants included the mitochondrial genome and indicated that the potential for the existence of a mitochondrial epigenome remains controversial. Thus far little research attention has been devoted to exploring this 14 EFSA Supporting publication 2016: EN-1129