New Concepts of Old Epigenetic Phenomena and Their Implications for Selecting Specific Cell Populations for Epigenomic Research

KEYWORDS::

• cancer stem cells
• cell subpopulations
• CpG islands
• methylome
• next-generation sequencing

While all cells in any given organism share an identical genome, different cell types are characterized by distinct epigenomes. The variabilty of the epigenome among different cells in tissues can confound the findings and force researchers to average results that can be of no or limited value. If we are to understand the biological and biomedical significance of epigenetic phenomena, it is important to map the epigenome in detail. For example, the biological role of 5-methylcytosine in different contexts needs to be considered. With approximately 200 cell types in the human body, the number of methylomes to be analyzed seems high. Furthermore, in addition to being highly variable between cell types, 5-methylcytosine fluctuation in time, even within a single cell has been reported [1,2]. Inherent to this divergence of DNA methylation patterns between cell types is the question of whether those differences in methylation are the cause or the consequence of changes arising during differentiation.

With the advent of epigenomics, high-throughput and genome-wide profiling technologies have emerged making it possible to study epigenetic profiles at both the genome-wide scale and at high resolution. However, despite their genome-wide coverage, high resolution and cost–effectiveness, most of these methods are not designed to be compatible with the analysis of minute amounts of biological material, and thus, are not capable of resolving the problem of mixed populations of cells. This is particularly true for tumors analysis, where the presence of normal (‘contaminating‘) cells and epigenetic heterogeneity is common.

Cytosine methylation of the human genome has been known for over 30 years, and is probably the most studied nongenetic modification. As a consequence of that intensive research, we have learned that 5-methylcytosine in mammals occurs mostly during CpG configuration, a dinucleotide under-represented in the genome but enriched in regions called CpG islands (CGI), commonly found in the promoters of genes (reviewed in [3]). From a functional perspective, cytosine methylation has been considered a mechanism to differentially shut down gene expression, and keep control of unscheduled transcription from repetitive regions of the genome. However, the view of methylation as a means of stable transcriptional silencing stems from detailed analyses of defined loci. Genome-wide epigenomic tools have begun to provide global data to refine, and sometimes to redefine, established concepts. The basic strategies of methyl antibody (or protein) affinity immunoprecipitation, enzymatic (methyl-sensitive) digestion and bisulfite modification, have been taken to the genome-wide level through the combination with microarrays (e.g., promoter arrays, tiling arrays and bead arrays); (reviewed in [4]). In addition, with the advent of massive parallel sequencing technology, we will extend our view of the distribution and dynamics of DNA methylation. Especially, massive parallel sequencing on bisulfite-converted DNA, with its single-base resolution and genome-wide coverage, promises to provide definitive maps of the human methylome [5]. These new tools have validated some of the previous paradigms of DNA methylation but, at the same time, new concepts are emerging. For example, almost a quarter of all DNA methylation found in embryonic stem cells occurs in a non-CpG context [6]. In addition, 5-methylcytosine can be converted into 5-hydroxymethylcytosine by the 2-oxoglutarate-and Fe(II)-dependent oxygenases TET1, TET2 and TET3 [7–9], opening a whole new field of epigenetic study. Finally, one important question regarding methylome analyses is where in the genome are the most informative changes found. Recent findings suggest that extensive DNA methylation changes caused by differentiation (and frequently deregulated in cancer) take place at CpG island ‘shores‘, regions of comparatively low CpG density close to CpG islands [10]. In addition, differential methylation during development has been shown at imprinted loci [11,12], gene-body cytosines [5,13], and intergenic CGIs [14]. Moreover, bisulfite data supports the view that differentially methylated regions are over-represented within the non-CGI category of promoters [15]. Unraveling the significance of non-CpG, non-CGI, gene body, intergenic methylation and hydroxy-methylation, with respect to gene expression and development is an evident priority. Therefore, it will be necessary to switch from our promoter CGI-centered view of DNA methylation, and gain insight into the distribution and function of these ‘nonclassical‘ modifications.

The answer to the questions posed above will partially depend on a proper identification of the corresponding ‘cellular states‘ of differentiation. Detailed kinetic experiments will define the temporal association of a specific DNA methylation profile and a switch in phenotype. An illustration of this was given by the recent study of the transition from human blood monocytes to dendritic cells, two easily distinguished cell types [16]. That study defined reproducible changes in DNA methylation, not confined to the promoters. Moreover, because of the lack of cell division during this process, the authors concluded that the changes in 5-methylcytosine were part of an active mechanism [16]. The immune system offers multiple similar examples where well-defined cell subpopulations switch to a different cell type in response to environmental stimuli. Because of its relevance and the availability of a detailed cell characterization and tools for cell separation (e.g., magnetic cell separation or fluorescence-activated cell sorting), based on surface markers and function, this system represents a precious field in which to understand the role of the methylome. A closer understanding of this role will derive from falsifiable hypotheses, for example, after blocking the activity of DNA methyltransferases during the process of cell-type switching. Under these conditions, DNA methylation may be shown to be necessary for cell differentiation or, alternatively, it may be important in stabilizing a defined phenotype. In the latter situation, cells may be prone to return to their previous stage of differentiation in the absence of 5-methylcytosine exchange. In either case, methylome patterns could be used as markers of differentiation or cell type. Furthermore, there is evidence that methylation at specific loci may define transcriptional potential rather than active transcription at a given time point. For example, the definition of T helper (Th) cell subtypes is based on the cytokine response to an external stimulus [17]. Methylation of the promoters of those cytokines may predict the direction of the response, and therefore the Th subtype, in the absence of a phenotypic change [18,19].

Unfortunately, apart from the hematopoietic system, few other tissues permit a proper identification and purification of cell types for epigenetic analyses. This is illustrated by the common finding of partial levels of methylation in studies intended to be tissue-specific. As a binary type of data (a given cytosine can be methylated or not, as the only options) intermediate levels of methylation in a cell can only be obtained in cases of differential allele or strand methylation. However, we have become used to intermediate levels of methylation (between 0 and 100%) that are likely to reflect heterogeneity within the tissue. Cell discrimination becomes even more complex in pathological conditions, and especially cancer, where redistribution of cell subpopulations may be inherent to the disease. Another important situation is given by the cells with the highest tumorigenic potential within tumors, also known as cancer stem cells. Intense research in the last few years has been devoted to their characterization, because of their potential role in resistance to therapy, clinical relapse and metastasis. In a way analogous to normal stem cells, the epigenome of cancer stem cells may define their main biological features, including pluripotency and the ability to self renew [20]. However, current strategies to isolate cancer stem cells are based on a few cell surface markers (e.g., CD44, CD133, ALDH or EpCAM) or dye exclusion techniques that only allow partial enrichment (reviewed in [21]). Future research on this direction should parallel the developments in characterization of normal progenitor and stem cells and how their differentiation potential correlates with a specific 5-methycytosine signature.

Biomedical interest in DNA methylation centers on the possibility that epigenetic variation between individuals can have repercussions for health. In addition to cancer, other human diseases have been hypothetically linked to abnormalities in DNA methylation, although causality is notoriously difficult to establish. However, emerging evidence supports a role for epigenetic modifications as a mechanism through which the external environment (maternal, social andchemical) and internal environment (fluctuating hormones) interact in order to produce cellular effects. Evidence in this sense suggests that DNA methylation is more dynamic than previously thought. This is illustrated at the single-locus level in response to signaling activation [1,2], cell differentiation [3] and at the genome-wide level during induced cellular reprogramming [22]. Bisulfitome sequencing data is likely to enrich the list of biological situations where cytosine methylation represents a dynamic process and where its deregulation is linked to human disease. A well-controlled experimental design in specific cellular subtypes is a fundamental factor to consider. Therefore, improving the tools for better selecting target cells and making future-generation sequencing technology suitable for the single-cell epigenomics should overcome the current problems associated with the analysis of heterogeneous cell populations. Coming back to the original question of the present discussion, although further research in characterizing cell types is required in some contexts, our current knowledge of normal cell subpopulations should be used as a reference for decoding the language of the methylome in health and disease.

Financial & competing interests disclosure

Neither the authors nor the authors‘ institutions have a financial or other relationship with other people or organizations that may inappropriately influence the authors‘ work or this editorial. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.