Epigenetics is the study of heritable traits that are not dependent on the primary sequence of DNA. That’s a short, simple definition, and it’s also largely unsatisfactory. For one, the inclusion of the word “heritable” excludes some significant players — the differentiation of neurons requires major epigenetic shaping, but these cells have undergone a terminal division and will never divide again — but at the same time, the heritability of traits that aren’t defined by the primary sequence is probably the first thing that comes to mind in any discussion of epigenetics. Another problem is the vague, open-endedness of the definition: it basically includes everything. Gene regulation, physiological adaptation, disease responses…they all fall into the catch-all of epigenetics.
Here’s another definition, cited by Mary Jane West-Eberhard in Developmental Plasticity and Evolution. Epigenetic factors are defined as:
…those heritable causal interactions between genes and their products during development that arise externally to a particular cell or group of cellswithin the same individual, and condition the expression of a cell’s intrinsic genetic factors (i.e., genome) in an extrinsic manner. In other words, epigenetic factors are the contributions to a cell’s environment by genes in other cells of the same individual.
Just to confuse matters even more, I surveyed the long line of developmental biology textbooks on my office shelf, and most don’t even mention epigenetics. Not because it isn’t important, of course, but because developmental biology basically takes epigenetics entirely for granted — development is epigenetics in action! Compare an epidermal keratinocyte and a pancreatic acinar cell, and you will discover that they have exactly the same genome, and that their profound morphological, physiological, and biochemical differences are entirely the product of epigenetic modification. Development is a hierarchical process, with progressive epigenetic restriction of the fates of cells in a lineage — a dividing population of cells proceeds from totipotency to pluripotency to multipotency to a commitment to a specific cell type by heritable changes in gene expression; those cases where there is modification of the DNA, as in the immune system, are the exception.
In part, the root of the problem here is that we’re falling into an artificial dichotomy, that there is the gene as an enumerable, distinct character that can be plucked out and mapped as a fixed sequence of bits in a computer database, and there are all these messy cellular processes that affect what the gene does in the cell, and we try too hard to categorize these as separate. It’s a lot like the nature-nurture controversy, where the real problem is that biology doesn’t fall into these simple conceptual pigeonholes and we strain too hard to distinguish the indistinguishable. Grok the whole, people! You are the product of genes and cellular and environmental interactions.
With our current state of knowledge, though, we can at least separate the two operationally. We can go into a cell, or into the online databases, and pull out DNA sequences, like this little snippet from human chromosome 15:
gaattctact aatgtttaaa aaattaatac caataaagtc ttacaaaaat atagaagtag
We can also see how mutations that change that sequence affect the organism, and we can also see that sequence being passed on from parent to child. Those are genetic traits; they are characterized by an overall stability, with deviations being the fascinating and important exception.
Epigenetics is messier and more fluid, and therefore harder to pin down. The genome is actually not a simple sequence of letters, it is a more complex chemical structure that is bound up with proteins called histones forming a complex called chromatin. This is what cells are actually working with when they execute a ‘genetic program':
Strands of DNA (in blue) are wrapped around spools of histones forming a unit called the nucleosome; these units are folded and wrapped into a great tangled loops and whorls, the chromatin. This is what is modified by epigenetic processes. We can break these processes down into several different categories. The two big ones are methylation of DNA and modification of the histones themselves.
DNA modification. Stretches of DNA can be inactivated by covalently attaching methyl groups, which can interfere with the binding of transcriptional enzymes, and can also be signals to recruit enzymes that modify associated histones. Cells have enzymes called methyltransferases that bind to specific dinucleotides (a cytosine adjacent to a guanine) and attach a methyl group to the cytosine. Methylated DNA is silent DNA.
Histone modification. Those roughly spherical histone complexes also have dangling N-terminal tails that can also be covalently modified by acetylation, phosphorylation, ubiquitination, or methylation. These changes affect how tightly packed the chromatin will be: in loosely packed chromatin, called euchromatin, the DNA is more accessible and more active, while in tightly packed chromatin (heterochromatin), the DNA is more inactive.
Histone variants. All histones are not alike! Some variants are more permissive of transcription, while others facilitate tighter packing. Activity in a region of DNA can be modulated by the kinds of histones used.
Chromosomal arrangement. There is growing evidence that at least some aspects of the 3-dimensional arrangement of DNA in the nucleus is non-random — that is, the DNA isn’t a willy-nilly tangle of spaghetti, but folded in some specific ways that bring widely separated regions into association. One of the prettiest examples of this is the control of olfactory receptor expression in mice: by unknown mechanisms, a single specific receptor gene is activated in an olfactory cell by the association of a distant enhancer element with a single receptor gene.
Every other mechanism of gene regulation. Heritable modifications of DNA are easily seen as epigenetic factors, but similarly, just about every known regulatory mechanism is in some sense also heritable. The concentration of transcription factors and RNA in the cytoplasm, for instance, affects levels of gene activity…and those factors are passed on in mitosis as well. Especially by the West-Eberhard definition above, epigenetics opens up into a vast catalog of everything that modifies gene expression.
How about some examples?
One clear example of a long-term epigenetic modification is X chromosome inactivation. Mammalian females have two X chromosomes, while males have only one, which could create problems of differences in dosage — left unregulated, females would have twice the concentration of the gene products found on the X chromosome, and we know that many genetic effects are sensitive to concentration differences. The mammalian strategy is not to make genes on the X in males work twice as hard to compensate, but to instead shut down one whole X chromosome in females. This is accomplished by, largely, extensive methylation of histones on the inactivated X, and by recruitment of repressive histone variants. The chromosome is heterochromatized to shut it down.
Which X chromosome is silenced is heritable. If a cell in a female embryo happens to shut down the X chromosome it inherited from the mother, leaving the paternal X active, all of its subsequent daughter cells will also shut down that same X. This is fixed; all future progeny of that cell, from that early embryonic state until the female grows old and dies, will be using the same single X. This is definitely a long term commitment.
One other interesting phenomenon occurs in eutherian embryos. Initially, the paternal X chromosome (that is, the one carried in the sperm) is always inactivated, and the earliest steps of development are carried out using only the maternal X chromosome. The extraembryonic tissues persist in this pattern, but the embryo itself later briefly activates all X chromosomes, and then in females, randomly shuts down one of them. This leads to the interesting situation that mammalian females tend to be mosaic, with an invisible (except in cases like calico cats) mottling of cells that have arbitrarily shut down one or the other X chromosome. It’s this later choice that is locked in for the rest of the individual’s life.
Another instance is genomic imprinting. It’s not just the X chromosome that is differentially activated or inactivated depending on whether it is maternal or paternal; other genes on non-sex chromosomes also show differences. The best known example is a bank of genes on human chromosome 15. In males, some of these genes are silenced; in females, a different set of genes in the same area are shut down. The pattern of inactivation is perpetuated in the sperm and egg, so sperm cells always carry a chromosome 15 with those genes methylated, while egg cells similarly have the female pattern of inactivation. Now normally, this has no detectable consequences to the embryo. It has one paternal and one maternal copy of chromosome 15, so it still has one copy of each gene that is entirely functional. All is well.
However, there are instances where the embryo must rely on just one of the chromosomes, and then things can go wrong. What if the sperm cell carries a chromosome 15 that has a defective allele, or one that is completely deleted? Then the embryo must use the maternal chromosome 15 copy, but what if that allele is maternally inactivated, or imprinted? Then it effectively has no copies of that gene product to use in development.
Another situation is called uniparental disomy. Sometimes there are errors in mitosis or meiosis called nondisjunction, in which a cell inherits an extra copy of a chromosome (a well known example is Down syndrome, where individuals have an extra copy of chromosome 21, with serious effects). Being trisomic, or having an extra copy, for chromosome 15 is lethal to the embryo, so that no such individuals make it to term. However, sometimes they can be spontaneously rescued by a second mistake, a loss of one of the extra chromosomes, reducing them back down to two copies of chromosome 15. Here’s the catch, though: which one is lost is random. If the individual has two copies of the maternal chromosome 15 and one copy of the paternal chromosome 15, and sheds the paternal chromosome, it’s no longer trisomic, but it does bear two chromosomes with only the maternal pattern of imprinting. This can also have serious consequences.
Individuals that develop with only maternally imprinted copies of these gene on chromosome 15 have something called Prader-Willi syndrome, a disorder characterized by mental retardation, obesity, and short stature; if instead the individual has only the paternally imprinted copies of chromosome 15, they have Angelman syndrome, a different disorder with severe mental retardation, and characteristic changes in facial features and movement (the original descriptions called them “puppet children” for their howdy-doodyesque appearance and jerky limb movements). These individuals may have identical genetic factors, and the only difference is in the epigenetic modifications of their chromosomes.
Another concern is the role of epigenetics in disease. Some chronic diseases, such as cirrhosis of the liver, are more than just an acute reaction to an environmental insult — they represent long-term changes in the pattern of gene expression in the cell lineages of the organ. Our cells are responsive, and they can be changed epigenetically in our lifetimes.
Some cancers seem to be facilitated by what are called epimutations — changes, not in the DNA itself, but in the pattern of methylation such that genes that play a role in our defenses against cancer are inactivated. Epigenetic silencing of the gene MLH1, for instance, is associated with some colorectal cancers.
One of the ways viruses can affect us is that they insert their genomes into ours — they induce a dramatic genetic change, which can be deleterious and which can be passed on by dividing cells. Epigenetic processes are defenses against the propagation of viral infections. Methyltransferases can sweep through and silence viral insertions, preventing them from promoting viral proliferation.
I began this with a couple of definitions of epigenetics. Perhaps a simpler, non-technical way to think of it all is that it represents a kind of cellular memory that can be passed down to daughter cells. It’s not as specific as the sequence of DNA, but it is sufficient to reconstitute the state of gene activity between generations. It’s central to understanding development as well as how organisms interact with their environment, and is intertwined and inseparable from our understanding of the gene.
Cavalli G (2006) Chromatin and epigenetics in
development: blending cellular memory with cell
fate plasticity. Development 133:2089-2094.
Egger G, Liang G, Aparicio A, Jones PA (2004) Epigenetics in human disease and prospects for epigenetic therapy. Nature 429:457-463.
Kiefer JC (2007) Epigenetics in development. Dev Dyn 236:1144-1156.