Executive Summary of the Epigenetic Mechanisms in Cancer Think Tank

Executive Summary of the Epigenetic Mechanisms in Cancer Think Tank

The purpose of the Epigenetics Think Tank was to bring together experts from basic and translational sciences to discuss the potential of epigenetic research to inform cancer biology, diagnostics and therapeutics, and to make specific recommendations to NCI for action in this field. The think tank began with brief presentations by each participant on the first evening. On the second day, three discussion sessions addressed cancer epigenetics, basic mechanisms in chromatin structure, and epigenetic technology and epigenomics. A special presentation was included on the European Epigenomic Network of Excellence. On the final morning, the panel summarized the discussions and prioritized recommendations.

The Think Tank highlighted a clear link between epigenetic alterations and cancer. Epigenetic changes such as DNA methylation and imprinting are targets of environmental carcinogens and diet, and probably are responsible for much of the population variation relevant to cancer risk. The role of hypomethylation in gene activation, aging and chromosomal instability is an emerging significant issue. The panel was particularly excited by the apparent role of epigenetic changes in the earliest stages of carcinogenesis, because these are potentially reversible with treatment. This provides an epigenetics-based strategy for cancer prevention. Epigenetic changes also clearly play an important role in tumor progression, tumor cell heterogeneity and possibly metastasis. A key need is to map these early changes through the analysis of methylation and other epigenetic marks in a well-defined set of reference human normal and tumor tissues. In addition, it is important to determine the temporal relationship of epigenetic alterations, as well as their relationship to underlying DNA sequence variation in the population.

This is an exciting time in the understanding of the mechanisms of epigenetic transcriptional regulation and chromatin structure. Protein complexes that manipulate nucleosomes, organize larger chromatin domains and set boundaries of chromatin structure have been discovered in recent years; now research must focus on discovering how they work in such cellular functions as replication, transcription and differentiation. Crucial next steps are to understand the role of key histone modifications, cis-acting elements, and regulatory proteins in setting, maintaining, and reprogramming epigenetic memory. Other key unknowns are the determinants of higher-order chromatin structure, the function of micro RNAs or siRNA in mammalian gene regulation and genome stability, and the mechanism of transmission of epigenetic marks through cell replication.

The Think Tank concluded that the highest priority should be development of a U.S. Human Epigenome Project of analogous scope to the Genome Project. The immediate establishment of a Working Group to define the criteria and plan the approach to the Epigenome Project is timely and highly recommended. There was strong consensus that we are now at the stage at which we can "genomicize" epigenetic information. Most of the necessary technologies are already available, although some tools for analysis of epigenetic alterations at the genome-wide level are still needed. New quantitative and computational methods may be required, including integrating epigenetic information with existing databases. Mapping epigenetic information to the human genome will maximize its usefulness and allow us to determine the relationship between genetic and epigenetic variation in the population, and its relationship to cancer. Such a project would be immediately worthwhile for its basic science insights as well as for enabling the use of epigenetic information in prevention or therapeutic strategies.

Introduction

The think tank began on the first evening with brief presentations of scientific advances, questions and barriers by each participant. The following day, three discussion sessions addressed Human Cancer Epigenetics; Basic Mechanisms in Chromatin Structure and Epigenomics; and Technology and Translation. There was also a presentation on the European Epigenome Program. On the final morning, the discussions were summarized, and the participants created a prioritized list of major and minor recommendations to NCI which will best advance the science.

In each session, the discussion was organized around a set of major questions. The panel began with these questions, and proceeded to discuss the issues and most critical next steps, and concluded each discussion with the next crucial questions to be answered and needs for the field to progress.

Human Cancer Epigenetics

Mechanisms of cancer initiation and progression by epigenetic events

Epigenetic changes are modifications to the genome that are heritable during cell division but do not involve change in DNA sequence. Expression of individual genes, of gene regions or a whole transcriptional program are regulated not by the DNA sequence, which is the same in every cell, but by the epigenetic marking and packaging that regulates chromatin structure. DNA methylation, histone variants and post-translational modifications, nucleosome positioning factors, boundary setting elements, and chromatin loop and domain organizer complexes are all elements of the epigenome.

Alterations in methylation, imprinting and chromatin are ubiquitous in cancer. A clear link between epigenetic changes and cancer, as well as other diseases, has in many case been established. And beyond these specific instances, epigenetics has the potential to explain many aspects of cancer: these marks can be modified by environmental factors; the age dependance of cancer corresponds to epigenetic changes in aging; and the quantitative nature of epigenetic changes corresponds to changes seen in cancer. The frequency of epigenetic alterations is orders of magnitude greater than that of genetic mutation, but unlike mutation, they are reversible.

The most frequently identified epigenetic alterations in cancer are changes in DNA methylation, although others have been identified as well. DNA methylation at CpG sequences (frequently in stretches called “islands”) is a basic marker of the genome that appears to have several functions. Methylation of gene regulatory regions has long been associated with transcriptional inactivation, and differences in methylation pattern in the promoter regions of tumor suppressor genes are documented in some cancers (e.g., p16/INK4a in pancreatic cancer). Hypomethylation is thought to either permit gene expression where it should not occur, or to encourage chromosome instability, especially when it occurs in the centrosome region. In general, global hypomethylation with local hypermethylation of CpG sequences is associated with cancer, although this concept may eventually be revised.

Two well described instances of epigenetic changes involved in cancer were used to illustrate different epigenetic mechanisms of cancer initiation throughout the meeting. Both involve loss of imprinting (LOI), which is DNA methylation that silences specific genes on either the maternal or paternal allele lineage. In Wilms’ tumor, approximately 95% of cases do not follow the two-hit model as it has been understood, with an inherited mutation in one allele of a tumor suppressor followed by a mutation in the other allele; instead, the second hit usually is LOI that activates the IGF2 gene. In colon cancer, IGF2 is also frequently activated, but another mechanism is involved, in which genetic and epigenetic mechanisms appear to be complementary. The LOI seen in these tumors also exists in the surrounding normal tissue and appears to predate the tumor; the theory is that this “field effect” renders the mucosa highly susceptible to subsequent insults, such that the overexpression of IGF2 (or other growth factors) due to LOI renders mutations such as APC mitogenic rather than apoptotic.

While the link between epigenetic changes and cancer is clear, the mechanisms behind the abnormal changes – indeed, the mechanisms by which normal patterns are established and maintained – at this time are largely unknown. The relationship between cause and effect is also controversial; while histone acetylation and DNA methylation are permissive for transcription, there is conflicting evidence as to whether these marks are the signals for gene expression, or if they follow the passage of the polymerase and serve to reinforce the “active” gene structure. DNA methylation differences at specific loci exist in Wilms’ tumor and Beckwith-Wiedemann syndrome, but a mosaic of methylation states is observed in the population of tumor cells. This raises the question of the degree to which the methylation pattern is a cause of the malignant growth. Because normal methylation patterns are not well defined, it is difficult to determine what is abnormal in cancer. In spite of the widely documented link between methylation changes and cancer, none of the known methyltransferase genes are known oncogenes or tumor suppressors. Thus, several crucial questions remain to be answered for methylation to be optimally used in cancer diagnosis or treatment.

A fascinating current issue is whether there are sequence variations in the human genome that create susceptibility differences by predisposing the site to epigenetic changes. Noncoding DNA such as repetitive elements and transposons may play a role in genome-wide methylation patterns. These elements are the main sites of CpG methylation in the genome. The example discussed by one think tank participant was that of a transposed element that altered the methylation pattern at a gene locus in a mouse model in a manner that responded to dietary factors. In the offspring, the transposon could be differently methylated depending on the maternal diet, with a higher methylation state leading to an obesity phenotype, demonstrating a means by which disease states could have an underlying genetic/epigenetic basis that responds to environmental factors. In this example, a genetic event led to the possibility of stochastic epigenetic changes. There is currently no known human locus where methylation variation leads to trait or susceptibility variation, or where a transposable element creates a site of potential expression variation; however, the likelihood was thought to be high that such examples will be found.

Histone modification, especially acetylation, has long been recognized to influence transcriptional capacity, but we must now begin to understand how DNA methylation and histone modifications have coordinated effects. A self-reinforcing cycle between these two epigenetic marks appears to exist, such that transcriptionally active or inactive regions of chromatin (i.e., euchromatin or heterochromatin) are self perpetuating through replication. The most important next step will be to understand the degree to which changes in these epigenetic marks are the causative regulatory factor in chromatin function changes, or whether other complexes play the regulatory role and these marks play a mainly reinforcing or maintenance role. This issue was the focus of some debate, since different results have shown these marks to play both an active and a passive role. Histone acetylation, for example, has been shown to cause transcriptional activity and also to result from the passage of the polymerase complex, thus being self-reinforcing, while loss of single specific histone acetylations can cause the spread of heterochromatin over a gene region. The recently proposed histone “code” of modifications was challenged by some panel members because such questions of cause and effect cannot yet be answered.

While DNA methylation of specific genes or general pattern changes are the most widely reported epigenetic signature in cancer, a cancer connection with chromatin organization is also well documented. SNF5, a subunit of the SWI/SNF chromatin remodeling complex family, is deleted in the aggressive pediatric cancer rhabdosarcoma. It acts as a tumor suppressor, activating expression of p16/INK4a and repressing several cyclins and CD44, a cell surface glycoprotein implicated in metastasis. Another example is that of the tissue-specific BUR (base unwinding region)-binding protein SATB1, which appears to organize chromatin loop domains and act as a positioning factor that attaches genes to Matrix Attachment Regions (MARs), giving them the proper nuclear localization for transcriptional activation. SATB1 specifically binds to the major break region in the characteristic chromosomal translocation of some lymphomas, creating the MAR which is thought to be a susceptible site. It will be interesting to search for similar proteins specific to other cell types and determine their expression or activity in cancer cells, since SATB1 has the ability to cause global changes in gene expression pattern.

Several specific needs were identified to advance our understanding of the epigenetic mechanisms functioning in cancer. It will be important to define genetic-epigenetic relationships; e.g., to identify mutations that lead to epigenetic changes. It will also be necessary to determine the relationships among such transacting modifiers of the epigenome as methyltransferases, transcription factors and genome organizers, as well as identify the factors and/or pathways that are activated in conjunction with these relationships. Another significant area in need of investigation is the effects and dynamics of epigenetic modifications of repetitive elements, transposons, pseudogenes, and novel transcripts. New insight into the organization and functioning of transcriptional networks and pathways may be gained from understanding principles of chromatin organization. New experimental model systems (e.g., to study RNAi; global CpG methylation and hypomethylation in cancer; pseudogenes in heterochromatin origins; and histone acetylation/deacetylation) are needed to study these features. Finally, it will be important to identify and develop applications for reagents for site-specific modifications of histones, such as acetylation-site-specific antibodies, and identify ways to decrease cost and increase availability of these reagents.

Interplay of epigenetics and cancer B where does epigenetics fit in the big picture of initiation and progression?

Despite the cancer associations highlighted above, the question remains of how frequently epigenetic gene activation or silencing is a “first cause” event that initiates dysplasia or creates susceptibility. Depending on the cancer, the first hit may be a mutation followed by epigenetic changes leading to a functional second hit, or abnormal or age-related epigenetic changes may create conditions that enhance environmental assaults or release suppression of tumor progression. Once the primary tumor is under way, the many steps by which a motile, metastatic phenotype emerges are likely to include epigenetic changes in gene regulation. Tumor cells generally must obtain a motile, spindle phenotype while traveling to the new site, then change phenotype again as they become established as a metastatic lesion. Regarded in this light, the reversibility and changeability of epigenetic states, as opposed to mutations, is a thought provoking idea. The reexpression of genes in metastatic cells, and the regulation of some genes involved in the angiogenic switch, are thought potentially to have an epigenetic etiology. In all of these questions, the lack of experimental models is a major barrier. Specifically lacking are models of very early epigenetic changes or very early stages of disease.

One promising focus for research is the question of epigenetic methylation changes as a timing event in the path from stem cells to differentiated cells to tumor cells (i.e., DNA methylation as a developmental timer). Another area of investigation might involve examination of epigenetic changes in “normal” cells and cells predisposed to cancer, to distinguish between normal epigenetic variation or changes with age, and potentially dysregulating epigenetic changes leading to disease. The epigenetic status of normal cells surrounding tumors, including angiogenic, immune, and stromal cells, may also play a significant role, and is virtually unknown. It will also be necessary to identify the relationship of epigenetic changes to chromosome instability and recombination in cancer.

Comprehensiveness of epigenetic studies of cancer B how much information do we need?

It will be important to conduct a genome-scale analysis of domains of epigenetic change at the local level, in gene groups or clusters, and globally. A genome-wide screen for DNA methylation changes would avoid the bias from candidate gene studies; on the other hand, the data set from a single region would allow us to study all epigenetic features, and determine how they relate to each other. There were conflicting opinions on which approach would be most valuable at this time. It will be necessary to investigate the functional significance of specific epigenetic marks in order to better understand how to resolve this question. It is important to determine the relationship between hypomethylation and gene activation, as there are several examples of loss of methylation linked to new gene expression and cancer. Before demethylating agents become widely used to treat cancers, this question must be addressed. The relationship of epigenetic changes to histone modifications is also important, and may involve the results of clinical applications, such as the effects of drugs on gene expression.

It will be highly informative to add epigenetic information to existing databases of genetic information. Gene regions could be layered with information on methylation and other marks that relate to transcriptional activity. However, as discussed in the final section on Epigenomics, a judicious approach to selecting experimental material and identifying the regions of chromatin to analyze must be used. To date, drugs that target DNA methylation and histone acetylation have been used with some degree of success in the clinic, albeit the more that is learned about the function of these epigenetic marks the less clear it becomes how the drugs work.

Basic Mechanisms in Chromatin Structure

Key Players in Chromatin Organization

A number of molecules with critical roles in chromatin organization, remodeling and epigenetic modification have been identified in recent years. However, our understanding of which ones are the key regulators and how they accomplish their roles in development, in the identity of different tissues, or during cellular functions such as replication is far from complete. Further basic research to better understand the mechanisms of action and determine the regulatory pathways of the known enzymes and complexes is crucial and timely. At the same time, it is important to try to broaden the spectrum of known chromatin modifying enzymes, if we wish to form a comprehensive picture. To understand how chromatin structure marked by specific epigenetic modification is established, it is necessary to identify additional players at the levels of DNA, RNA and proteins, and then go on to identify the dynamic relationships between them. The role of non-coding RNAs is addressed in a later section.

There is evidence that DNA methylation/demethylation affects gene transcription by signaling chromatin structure changes, rather than simply “blocking” access by polymerases or transcription factors. Thus, since DNA methylation status is the most widely and readily measured epigenetic mark, it will be important to determine its role in chromatin organization. At the same time, it is important to search for long-term epigenetic marks other than DNA methylation and histone modification.

A major challenge in the near future will be to identify key determinants of locus-specific chromatin structures; that is, key histone modification sites and key cis-acting DNA elements (e.g., BURs, which are recognized by global gene regulators) which have dominant roles in the formation of specific chromatin structure, as well as the protein complexes which interact with these sites. To understand how epigenetic states are established, it will be necessary to identify new interacting proteins that are involved in chromatin targeting and remodeling. These partner proteins currently include PCNA, RNA polymerase, nucleosome assembly factors, CTCF, and loop organizers. Other factors must be presumed to exist which have yet to be identified; however, it is not too soon to begin to ask which part of the genome is regulated by which of the known factors.

The known molecules that regulate accessibility of the DNA to regulatory factors include DNA methyltransferases, histone variants, and enzymes that covalently modify histones through acetylation, phosphorylation, methylation and other chemical moieties. Another class of chromatin-modifying complex includes those that move nucleosome positions to reveal or occlude DNA sequence. Three families of these complexes have been identified – SWI/SNF, ISW1, and NuRD – all of which consist of a core subunit with an ATPase domain, plus associated subunits presumed to be regulatory. Other chromatin modifiers include the Methyl-CpG-Binding Domain (MBD) protein family, Position-effect variegation (PEV)-related factors, and chromatin loop organizers. Although presumed to exist based on the observed dynamics, neither DNA demethylases nor histone demethylases have yet been identified. Discovery of such crucial enzymes will have high impact on understanding of chromatin structure maintenance and remodeling. Additionally, any of these enzymes, loop organizers or boundary element-binding proteins may be subject to post-translational modifications that regulate their functions, although very little is known at all about such regulation of these molecules. Many of these proteins can be studied in yeast; for example, yeast studies have demonstrated that Sir2 histone deacetylase induces the spread of heterochromatin, while Sas2 acetyltransferase disrupts heterochromatin, and together, they appear to establish chromatin boundaries. Efforts are now being directed at understanding the roles of known higher eukaryote analogs of these enzymes.

Mechanisms for higher-order folding of chromatin

To understand chromatin structure and function, it is essential to expand our knowledge of how chromatin is packaged beyond the nucleosome structure. Which proteins are responsible for higher-order packaging (e.g., loop domain organization) and the biological significance of such chromatin folding are fundamental questions that have only recently been formally addressed. It will be an important step to identify more loop organizers in addition to SATB1, in different tissues and at different developmental stages, and to distinguish between general and specific proteins. This will require improvement in the highly variable loop capture assay, and the development of new technologies to analyze higher-order chromatin structures (e.g., in mega-base domains).

It is necessary to determine the role of loop structures in delineating functional chromatin domains that are marked by specific epigenetic modifications. Such an approach, which introduces the third dimension into epigenetic analysis, will reveal the contribution of higher-order folding to epigenetic modification and overall genome organization. It will be important to identify the mechanisms that regulate functions of loop organizers such as SATB1, and to discover additional analogous protein complexes. Finally, it will also be important to analyze the roles of loop structure and loop organizers in cancer and explore their diagnostic utility. Some observations suggest that the organizer protein BORIS competes with the related protein CTCF to disrupt boundary function, and may alter transcription pattern on a large scale when abnormally expressed in cancer. Additionally, the global organizer SATB1 is aberrantly expressed in some cancers. It will be significant to explore whether these proteins could be effectively targeted in cancer therapy.

The role of RNA in gene regulation and genomic organization

Non-coding RNA appears to play several roles in chromatin regulation. Emerging evidence suggests that micro RNAs (miRNA) or siRNAs can regulate over very large chromatin territories. These small RNAs, produced by defined processing pathways, are known to silence genes at the translational and perhaps also the transcriptional level; however, a role in maintaining heterochromatic regions is currently only postulated. It will be important to determine the generality of RNAi control in the genome beyond repetitive elements, and the function of RNAi in genome organization. Currently it is unknown how many miRNAs are present in our genome. It is still to be determined how they can be identified, and how their targets can be validated. Then, moving beyond quantification of miRNAs, we will have to determine the role of miRNA in gene expression, where it acts, and what link if any it has to cancer. Another intriguing question is whether viruses and other pathogens modify the epigenetic profile of the genome in ways that involve siRNAs and/or miRNA, leading to dysregulation and potentially diseases such as cancer.

Intergenic transcription is a novel mechanism by which heterochromatin may be excluded from transcriptionally active regions. It will be necessary to determine the link between intergenic transcription and RNAi. An important question is whether siRNA-based regulatory networks exist, and what the role and mechanisms of large noncoding RNA transcripts (e.g., Xist) are in chromatin function.

Epigenetic inheritance in somatic cells

This area of chromatin epigenetics research in particular is still in its early stages and warrants further, extensive investigation. DNA methylation is initiated de novo by DNA methyltransferases and maintained by Dnmt1, and this provides a mechanism of epigenetic inheritance. However, there are epigenetic phenomena in organisms such as Drosophila that lack any significant DNA methylation. Epigenetic states may be transmitted through specific chromatin structures, such as histone modifications, but how this is achieved is largely an enigma. Lysine methylation on histones 3 and 4 provides relatively stable marks on histones. But, again, how such marks are maintained through the cell cycle or through development is not known, given that there are mechanisms for active loss of histone methylation, in addition to passive loss by dilution at each round of replication. Replication timing may be a crucial mechanism in setting or preserving the epigenetic marks, since euchromatin replicates early in the cell cycle, while heterochromatin replication is delayed. The histone modification Acode@ has not yet been demonstrated to be heritable. The true function of histone modification patterns remains to be determined, and there was disagreement as to whether an active role in determining chromatin functions has been convincingly demonstrated, as opposed to a more passive role as a result of replication or transcription.

The question of what epigenetic information is preserved through replication, as well as the related question of what modifications are causative of chromatin organization as opposed to simply the more passive result of replication, were the topics of extensive discussion. The roles of DNA methylation and histone modifications were discussed in this light.

A critical next step is to determine the mechanisms of the replication of specific chromatin structure, identifying the DNA sequences and the targeting of chromatin remodeling factors involved in setting up specific chromatin structures during or after DNA replication. Part of the same effort will be to test the role of histone modifications in epigenetic inheritance through the cell cycle, and to determine the dynamics of histone modifications during DNA replication and the cell cycle. We must deepen our understanding of how DNA methylation relates to gene expression and chromatin structure.

Epigenetic remodeling and genome stability

It will be important to determine the mechanisms of regulating or reversing the more stable epigenetic marks. For example, DNA methylation can be either an active or a passive process, as discussed above, and once methylated DNA demethylation may be via an active or passive mechanism. Also, when histones are replaced with variants that have different functions or modification sites, are these mechanisms active regulatory mechanisms or secondary results of other functions? In any case, it remains unknown if they have relevance to cancer.

Chromosomal rearrangements are a common feature of a wide variety of neoplastic lesions and are thought to have a causal role in tumorigenesis. Genome instability, leading to carcinogenesis, may occur due to the genome-wide DNA hypomethylation observed in aging. Global hypomethylation is frequently observed in tumor cells, and has been associated with abnormal chromosomal structures, as observed in cells from patients with ICF (Immunodeficiency, Centromeric instability and Facial abnormalities) syndrome. However, the mechanistic link between genomic instability and hypomethylated genomic DNA remains to be discovered.

It will be very important to determine the mechanism for the creation of neocentromeres. Centromeres are key players in genome integrity, and defects in their function results in mis-segregation of chromosomes and aneuploidy. Neocentromeres are ectopic centromeres that originate occasionally despite the complete absence of normal centromeric alpha-satellite DNA in the regions. No alteration of DNA sequence occurs, and it remains unknown how they are created, but they are still able to form a primary constriction and assemble a functional kinetochore. Neocentromeres have been detected in certain cancers, and are thought to be generally detrimental. However, no systematic screening for these structures has been conducted.

Finally, it will be important to investigate the role of chromosome territories, and factors that organize these territories, in recombination and genome instability. Chromatin must be viewed in the context of whole chromosome territories, organized into specific nuclear zones. These territories occupy non-random positions in the interphase nucleus. The mechanisms confining each chromosome in a defined territory remains unknown, however, as well as the question of whether changes in the position of the territories occur during differentiation or carcinogenesis. The specific location of genes within nuclei is important for proper gene expression, and nuclear architecture contributes to the localization of genes. Thus, improper positioning of genes may contribute to their improper regulation, potentially leading to cancer. Cancer is associated with disruption of gene expression patterns and global disorganization of chromosome organization within the nucleus. The roles of nuclear architecture, including chromosome territories, in genome stability and proper gene regulation must be studied in depth.

Epigenomics, Technology, and Translational Opportunities

We are presently approaching the stage where we can begin to “genomicize” epigenetic information. At the think tank, much of the discussion of needed technology and translational opportunities took place in the context of developing a Human Epigenome Project, analogous to the Genome Project. The science is clearly developed to a point that it is timely to plan a strategy for such an effort. The key question is what information the efforts should focus on obtaining.

There are two decisions that need to be made before launching into a large-scale effort. First, what epigenetic information should be gathered and incorporated into the Epigenome? DNA methylation is the first consideration, and most of the investigators want to make genome-wide DNA methylation profiling using microarrays the top priority. After that, other important elements that should and could be incorporated include unequal allelic gene expression (including imprinting); histone variants and modifications; chromatin organizer proteins; and cis-acting marks for epigenetic signals. It must be remembered that different types of epigenetic information present different challenges. For example, the two fundamental epigenetic marks, histone modifications and DNA methylation, differ in stability; certain histone modifications can change quickly (acetylation reversed by deacetylases, methylation reversed by histone replacement, etc.), while CpG methylation of DNA is more stable.

Second, what tissue samples should be used? Unlike DNA sequence, epigenetic marks and states certainly differ between tissues and change with the age of the donor. A discussion is needed about what we know and what we need to know, to permit these choices to be made. It would be useful to ask what relationship exists between diversity in the population and diversity in epigenetic marks, and what new epigenetic population-based and family-based algorithms need to be developed.

Key Issues and Potential Pitfalls

New technologies are needed to study some questions of higher order chromatin organization and function, and broader training is needed in current state-of-the-art procedures such as the loop capture assay or the ChIP-on-chip assay. Nonetheless, current methodology is sufficient for conducting genome-wide screens of much of the epigenetic information. Integration of epigenetic information into existing genome databases would be optimal. A computer screen that incorporates all levels of epigenetic information for each stretch of DNA sequence in the genome was envisioned by some of the panel. In principle, data from genome-wide or chromosome-wide bisulfite conversion/sequencing, histone modifications, and mRNA expression profiling, can be overlaid on the existing human genome sequence, to provide an integrated picture of the genome and epigenome. This information, with associated annotations, could be displayed at a resolution ranging from the whole chromosome to the individual nucleotide in a clickable format, simply by adding a series of new Ainformation fields@ to available, widely used genomic databases. With this coordination of data, the project would begin to provide answers about the potential for epigenetic therapy and how the epigenome can be targeted for chemotherapy or chemoprevention.

Discussion of precisely what epigenetic marks and what tissues would be included first did not produce a consensus at this meeting. All agreed that future organizational meetings would be required to plan an Epigenome Project. This project should take an approach somewhat like the Human Genome Project, with a sequencing effort that provided baseline information on which further genetic variation could be identified. A comprehensive analysis of DNA methylation is needed, including quantitative variation in the density and breadth (i.e. numbers of modified residues) of methylation marks. In addition, novel bioinformatic approaches must be developed.

In contrast to the DNA sequence, the epigenome varies among different normal cell types and, perhaps more dramatically, between normal and cancer cells. So, selection of the Areference cells@ and Areference tissues@ for analysis in the Epigenome must be done with extreme care. To deal with the problem of cell heterogeneity, microdissection will be essential in preparing the samples. An even more fundamental issue concerns the identification of the closest normal counterpart for any give human cancer type, i.e. intelligent and meaningful selection of the Acontrol@ samples for comparing to the cancer samples. Significant progress is being made in this area, largely through gene expression profiling and immunophenotyping, and this information needs to be taken into account in making this critical decision on sample selection. It would be a disastrous waste of resources to analyze the epigenome in tissue samples that subsequently prove to be poorly defined and heterogeneous.

Genome Imprinting and Allelic Asymmetry

Interindividual Variation in Epigenetic Marks

Mechanisms of Epigenetic Marking

Environmental Effects on the Epigenome

Cancer Detection using Epigenomic Targets

Epigenomic Drug Targets in Cancer Treatment

Bioinformatic Approaches to Epigenome Analysis

Benefits of a Human Epigenome Project

Epigenetic marks, defined as modifications of the genome that can be faithfully propagated through cell divisions, and that affect central biological processes like gene expression and genome stability, have been amply shown to influence both cancer predisposition and tumor progression. But in contrast to the DNA sequence, our understanding of these epigenetic marks remains fragmentary, with most if not all of the available data coming from highly selected regions of the genome, examined by non-exhaustive methods. It is likely that a comprehensive analysis of the epigenome, carried out in carefully chosen and well defined normal and cancerous cells/tissues, will produce some truly surprising and paradigm-shifting results, and may well yield fundamental insights into cancer pathogenesis, detection and treatment. This integrated epigenomic information will allow us to begin answering the question of what is the basis for epigenetic therapy and how can the epigenome be targeted for chemotherapy or chemoprevention.

Following some discussion of the European Epigenomic Network of Excellence (see below), Think Tank participants discussed pursuing a similar effort in the United States that would comprise a significant, dedicated investment whose management and direction would be formulated by an established network or committee. The justification for pursuing such a project included:

$ The series of discoveries in the last 10 years showing that epigenetics is central to understanding of the development and genetics of cancer.

$ Moving forward with an Epigenome Project is timely in light of the completion and findings of the Human Genome Project.

$ To gain a better understanding of human disease, including cancer, investigators need to take a genome-wide perspective of epigenetics.

$ It will facilitate the standardization and dissemination of information across the research community.

$ It will increase visibility of the research community as a coordinated group.

$ Improved cost effectiveness, decreased redundancy, and better comparisons are expected outcomes with standardization across research teams, labs, and groups.

$ Keeping the project in the public domain would help not only to maintain but to encourage openness of communication, development, and discoveries, including technological developments and advances.

$ This strategy would provide an infrastructure for reviewers to improve genome study sections through increased awareness, expertise, and innovation.

$ It will provide a much-needed, coordinated opportunity for quantitative/statistical epigenetics and the development of new tests, tools, and measures.

$ It will provide a missing component of integrative biology.

Special Presentation: European Union 6^th Framework ProgrammeCThe Epigenome

Dr. Patrick Varga-Weisz summarized the EU=s Epigenomic Research Program, which provides a platform and structure to facilitate the coordination of epigenetics research beyond local efforts. The program takes advantage of established research groups and facilitates the development of new research talent. A Acore consortium@ of 25 leading scientists in epigenetic research receives funding through a Joint Program of Activities (JPA) for the first phase of the Epigenomic Network of Excellence (NoE), which will run from 2004 through 2009 with an annual budget of approximately 12.5 million Euros.

The purpose of the Epigenome Program is to integrate European epigenetics and epigenome research; to support new researchers; and to facilitate dialogue among investigators. More specifically, the three key goals are to coordinate and support:

● European epigenetic research: To unravel epigenetic mechanisms in the post-genomic era, the NoE will prioritize research into molecular mechanisms of epigenetic control. To achieve this goal, the NoE will mount an internationally competitive research program through a Joint Program of Activities (JPA).

● The next generation of European scientists: A funding program for 22 independent investigators will distribute monies in two rounds over 5 years to integrate the most promising newly established teams (NET program) in Europe.

● An interactive Website: Scientists and the public will be provided with up-to-date knowledge of the field through a Web site maintained by the NoE. This Web site [for the preliminary Web site, go to http://www.epigenome.imp.ac.at/index.html] will develop into a major resource for the scientific community, and it will seek to establish a dialogue with the public by providing first hand knowledge in an appropriate form for a non-specialist audience.

The NOE research program is organized into eight distinct subprograms focusing on different aspects of epigenetic control: