Epigenetics: the burden of proof vs the folly of dismissal

We haven't written much about epigenetics for a while, in part because it's so trendy that it's impossible to know what much of it means or how it's all going to shake out, and in part because there are so many different interpretations of the word that it's hard to know whether everyone's talking about the same thing.  Tools to detect epigenetic changes in the genome, that is, specific locations that have been chemically modified in ways that affect nearby gene transcription, are now available.  Still, it is clearly a fad in the sense that once tools are there the scientific community seizes them in a bandwagon effect, showing up in study designs and grant applications and so on, in ways that can exceed the reality.  Everybody now simply 'has' to do an epigenetic analysis on their favorite project.  And partly for this reason, not everyone else accepts that epigenetics will prove in the long run to be a significant actor in development and disease.

Epigentics: what is it?
We've posted about epigenetics in the past (e.g., here and here).  Relevant to today's post, the idea of epigenetic changes is that DNA can be chemically modified not by changing its nucleotide sequence itself, but by altering the packaging of the chromosome near genes that are to be expressed (or repressed).  Since different cells in the body are different because of their differential use of genes (cells in your brain and your skull have the same DNA, but turn on different sets of genes to become what they are, as well as to do what they do throughout life), this is simply a statement of one of the mechanisms by which it's achieved.  That's not the controversy.

The controversy goes a bit deeper and reaction to claims of epigenetic changes are emotional and often vehement.  There are two reasons for this.  First, what causes epigenetic marking of specific chromosome regions is the state of the cell at a given time, and that can change in response to the conditions it senses--its environment.  The evidence is, further, that until active mechanisms alter the epigenetic marking of chromosomes, the gene-expression pattern of the cell is inherited when it divides.  Since the inducing mechanism is part of the environmental situation of the cell, this gets uncomfortably close to Lamarckian inheritance.  The cartoon example of Lamarck's pre-Darwinian idea is the giraffe stretching to reach high leaves and, if successful, passing on long-neck genes to its offspring.  This is the antithesis of modern Darwinian theory (though Darwin himself toyed with it in his own theory of inheritance).  That Darwinian theory has randomly arising mutation being screened by natural selection to pick the successful genes: the lucky giraffe that happened to have inherited a long-neck genetic variant ate better and had more girafflets than its shorter-necked peers.  Here, the facts speak for themselves, and nothing known suggests that striving for something can in itself engineer heritable genetic change, specific DNA mutations, to make that something happen.

Masai giraffe; Wikipedia

But the second reason that epigenetics touches raw ideological nerves, especially in regard to humans, is that one school of thought wants to see everything human as written deterministically in our genomes: you (including your behavior) are what your DNA sequences prescribes.  Anyone offering any other suggestion is by this group widely denigrated without inhibition as a soft-headed denier of the importance of heredity.  That's because if environments really do affect your achieved nature, then genetic determinism and all that goes with it are no longer biological universals set in stone.

But what if, beyond environmental effects on gene expression in an individual during its lifetime, those effects were heritable into the next and future generations?  That would suggest that we are not just dealing with a fad made possible by a fancy bit of gear that can help you get a grant, but that there are things about our achieved natures and our evolution that we don't yet really understand.  And a couple of papers, one from last year, and one more recent, struck us as worth writing about, for different reasons.

Epigenetic---and then some?
A lot of attention was paid to a December 2013 paper in Nature Neuroscience ("Parental olfactory experience influences behavior and neural structure in subsequent generations", Dias and Ressler). Their point was first to show that your experiences involving odor detection are specific and can leave a long-lasting chemical and behavioral 'memory'.  Dias and Ressler exposed mice to a particular well-studied single-molecule odor, and coupled that exposure with a shock to the mouse's foot, to condition the animals to fear the odor (a logic resembling Pavlov's famous dog experiments).  This triggered the activation of cells that express a particular odor-detection (olfactory receptor, or OR) gene, out of the repertoire of about 1000 such genes, as well as the conditioned fear response upon smelling the odor, even absent the shock. But the authors reported something much more remarkable, and challenging to understand.

What they found was that the behavioral response to that same odor is activated in at least two future generations that had not been exposed to the fear-conditioning.

We subjected FO mice to odor fear conditioning before conception and found that subsequently conceived F1 and F2 generations had an increased behavioral sensitivity to the FO-conditioned odor, but not to other odors... Bisulfite sequencing of sperm DNA from conditioned F0 males and F1 naive offspring revealed CpG hypomethylation in the Olfr151 gene.  In addition, in vitro fertilization, F2 inheritance and cross-fostering revealed that these transgenerational effects are inherited via parental gametes.  Our findings provide a framework for addressing how environmental information may be inherited transgenerationally at behavioral, neuroanatomical and epigenetic levels.

A very important part of this is that the transmission was by males who had been conditioned, via their sperm, to females who had no such experience....and then to their sons' offspring (that is, the marking was present on the sons' sperm cells, directing hyper-expression of the OR gene in the grandchild-mice).

Reaction to this paper seemed to fall along party lines, with determinists doubting that the results could be real, and others intrigued.  Indeed, this is curious because the experience affects not just the startled males' odor-detecting mechanism in its nose cells where odors are detected, and fear response, but seems to imprint the specific effect on sperm cells.  There is no means known (to us, at least) by which this could occur, unless all cells' OR genes are affected during the F0 males' conditioning, nor are we qualified or patient enough to judge whether the study, or its set-up in some way has led to a misleading result.  To be fair, while the authors didn't demur to send their paper to a Nature journal, nor (in expected fashion) did the Nature journal demur from publishing without requiring such a mechanism to be shown, the authors themselves in fact did not venture a mechanism and recognized the issue in the Discussion.

If this sort of specific epigenetic mechanism does in fact persist across generations, without further conditioning, many questions are raised.  Not only is the targeting mechanism important to know, but since every generation has different experiences, what sort of expression mishmash would new pups have after millions of years of evolution in all sorts of environments?  Nonetheless, mice (and we, and trees, and even bacteria) are differentiating organisms that respond to environmental conditions in ways that certainly include altered gene expression.  So being skeptical may be fully justified, and this is not in any sense an "Aha!" moment for Lamarckians.  But its strangeness to what is currently known is also no reason to dismiss it because it doesn't fit your, say, genomic determinist or selectionist predilections. This is especially so because there is in fact a lot of evidence for environmentally induced changes in gene usage, and hence in the traits, of organisms including humans.  And that brings us to the other, more recent paper.

Do big bodies mean big epigenetic news?
The other paper is a recent report in The Lancet ("DNA methylation and body-mass index: a genome-wide analysis," Dick et al., 2014), which describes the results of a genome-wide analysis of methylation at CpG sites and its association with obesity, measured by BMI.  CpG refers to a C nucleotide being next to a G nucleotide along a DNA strand.  Methylation is a way of chemically attaching a small tag to that CpG in gene regulating areas of a chromosome, that makes it hard for the proteins that are needed to express a nearby gene to bind to the DNA to do their job.  That is, the expression of methylated genes is repressed.

A commentary in the June 7 Lancet applauds the work of Dick et al., and heralds the beginning of the "EWAS [epigenome-wide association study] era."  Of course, one's first reaction might be a sigh of 'here we go again!' in regard to hype far out-performing hope, a new fad for rescuing hopeless non-replicable findings, and journals having to sell copy and holding no standards of circumspection. But how should one react?

The possible significance of epigenetics to disease has not been lost on epidemiologists, and a new field called epigenetic epidemiology is abornin', counting on the importance of non-sequence modifications of DNA, in particular methylation and acetylation patterns, to (finally!) explain patterns of disease.  In that sense EWAS may be important, or may to a cynic just be an E-for-G swap to keep the GWAS funding flowing.

The Dick et al. paper is from this burgeoning field.  Epidemiology had many successes in the last century identifying environmental causes of disease, but when complex chronic diseases overtook infectious diseases as leading causes of death, the field had a much rougher time finding the causes of major diseases, and predicting who would get them.  So, epidemiology turned to genetics, but ran into the same problem genetics itself was up against -- complexity.  But if specific epigenetic changes can now be attributed in a useful way to environmental factors, on say the McFood-O-Meter scale, the claim will be that reductionist science has found the mechanism that shows that new epidemiological studies will have to be funded to focus on the risk factor that causes the epigenetic change.

It's not an entirely new idea.  For some years, George Davey Smith, an epidemiologist at the University of Bristol, has been advocating the use of 'Mendelian randomisation ', a strategy to see whether a variant in a gene whose function relates to processing some environmental factor has the same effect in people not exposed to that factor as those who are.  Maybe someone will cook up other prevention or treatment strategies if epigenetic mechanisms prove important.  

Dick et al. identified five methylation sites in the genome that in their sample were associated with being overweight by the Body Mass Index (BMI) criterion: three were in intron 1 of the HIF3A gene. HIF3A is a gene that regulates response to reduced oxygen levels.  The authors note that "Although the main focus on HIF has been its role in cellular and vascular response to changes in oxygen tension during normal development or pathological processes (eg, cardiovascular disease and cancer), compelling and increasing experimental data suggest that the HIF system also plays a key part in metabolism, energy expenditure, and obesity."

Have Dick et al. found the cause of obesity?  Well, no.  As the Lancet commentary points out, there are numerous difficulties in epigenetic research, a primary one being that a gene won't be modified in every tissue, nor all the time, nor even necessarily in every cell of the appropriate type in a given tissue.  That means that the choice of tissues in which to search for methylation and when to look are crucially important considerations.  Dick et al. did test various tissues, and found that methylation varied.

Another important issue in epigenetic studies is determining the order of events -- which came first, the disorder or the DNA modification?  That is, does the disorder lead to methylation of genes involved, or does methylation of related genes cause the disorder?

Dick and colleagues attempt to address the issue of causality by applying a mendelian randomisation approach to interrogate the causal relation between HIF3A methylation and BMI. This approach uses a genetic proxy for DNA methylation (namely, methylation quantitative trait loci) to identify a causal relation between an exposure or trait and epigenetic variation, assuming that genetic associations are largely immune to residual confounding and reverse causation. Dick and colleagues identified two upstream single nucleotide polymorphisms that were independently associated with DNA methylation at a HIF3A locus in both the discovery and replication cohorts. However, these single nucleotide polymorphisms were not associated with BMI in the study cohorts or the high-powered GIANT consortium dataset, suggesting that hypermethylation at the HIF3A locus is likely to be a result of increased BMI rather than a causal association between increased methylation and BMI.

So, apparently the obesity came first, methylation later, and has nothing necessarily to do with the cause of obesity. Interestingly, The Lancet still describes this as an important study. "Dick and colleagues’ study represents an important advance for both obesity-related research and the specialty of epigenetic epidemiology." Why? "The widespread uptake of instruments such as the Illumina 450K HumanMethylation array means that large collaborative EWAS meta-analyses can be done, building on the success of similar approaches in genetics."

Have instrument, will use it.

The burden of proof vs the folly of dismissal
It's early days yet in the understanding of the role of epigenetics in disease and behavior, and there's a lot left to be learned.  There is now a wealth of experimental literature on cells as well as a variety of laboratory species, demonstrating some of the mechanisms of gene regulation that involve epigenetic changes of DNA.  There are carefully done experimental studies that show multi-generational transmission of such changes. There have also been epidemiological and even experimental studies of intra-uterine or maternal experience affecting things like body weight in offspring.  Thus, even without specific epigenetic data at the genome level we have every reason to expect that life experience at any age could affect even complex traits.  And what would be more likely than some sort of epigenetic mechanism to be responsible?

One should also keep in mind that trans-generational correlation can look very much like regular genetic transmission and make a trait look 'genetic' in the classical sense, rather than in the epigenetic sense.

It clearly befalls those advocating, and those dismissing, epigenetic inheritance to keep their powder dry until we can see more clearly into the whites of the genome's eyes.  In fact, since we are obviously differentiated organisms descended from a single cell, who respond in all sorts of physical and behavioral ways to our internal and external environments, it seems obvious that some such mechanisms are fundamental to genome function, as experience clearly suggests.  But how well complex traits like body shape or odor detection would be transmitted not just across cell divisions in specific types of responding cells, but also across generations, is far from clear.

Keeping our powder dry should be automatic for scientists, as this is a very important question well worthy of careful investigation.  But whether we can keep obfuscation by ideology and equipment salesmen at bay is just as serious a question.