It is widely maintained that biology is undergoing an epigenetic revolution. According to this narrative, the gene is being dethroned from its privileged explanatory and investigation-guiding roles. In its place, scientists are focusing on various epigenetic factors—equally significant to genes in their casual and information-bearing functions, or so it is argued—that have long been neglected in the study of development and evolution.
The study of human disease is one of the fields that epigenetics is expected to transform. Biomedical interest in epigenetics traces back to the discovery that widespread loss of DNA methylation is associated with cancer [1]. At the time, it was a significant discovery that cancer could be triggered not only by mutation in gene sequence, but also by the removal of methylation marks. During the 2000s, biomedical work on epigenetics explored the tendency for cells to acquire an elevated vulnerability to stress [2]. This phenomenon was associated with alterations to DNA methylation triggered by environmental factors, such as a reduction in quality of diet [3], that are potentially transmitted to offspring in utero [4]. More recently, we are seeing the rise of large-scale research consortia such as the Encyclopedia of DNA Elements (ENCODE), which seeks to identify all functional elements in the human genome by focusing in particular on “regions of transcription, transcription factor association, chromatin structure and histone modification” [5, p. 57]. ENCODE’s most controversial and widely publicized result states that over 80% of the human genome is associated with some biochemical function [5]. From a gene-centric perspective, this claim would be surprising, since protein-coding regions constitute a mere 4% of the human genome [6]. Detractors object that ENCODE’s finding relies on an overly permissive definition of function, that their study used unjustifiably weak criteria for identifying genetic candidates as functional, and that their framework cannot explain differences among species’ genome sizes [7,8,9,10]. In defense of ENCODE, some authors interpret their controversial statement as an estimate of the proportion of genomic regions that are of potential biomedical interest [11]. Generally speaking, it is clear that epigenetics has motivated considerable research within the biomedical sciences, challenging conventional notions of biological function and expanding the range of entities thought to be functionally relevant to human disease.
It is tempting to follow authors like Eva Jablonka and Marion Lamb, who claim that epigenetics involves a paradigm shift in biology [12], or Russell Bonduriansky and Troy Day, who suggest that epigenetics constitutes a “new understanding of inheritance and evolution” [13]. This is a seductive picture, especially to philosophers. Conceptual change in science is an established field of philosophical research. The study of gene concepts has been one of the most fecund topics within the philosophy of biology. This work reveals that scientific conceptions of the gene and genetic disease are in an ongoing historical dialogue with technological advances in biology [14,15,16]. To many philosophers, it would be unsurprising if further technological developments led to additional modifications to scientific conceptions of heredity. Gene concepts have proven to be fluid, the thinking goes. Why should gene centrism itself not be up for grabs?
Some authors challenge the suggestion that there is an epigenetic revolution afoot. It is possible to distinguish three general objections. The first takes issue with the claim that epigenetic insights qualify as revolutionary. Peter Godfrey-Smith notes that over the course of its historical development, molecular biology has become gradually less doctrinaire [17]. Theoretical principles that were central to this discipline in its early stages, such as G.W. Beadle’s one gene–one enzyme hypothesis, have become less important as molecular details have been filled in. Epigenetic phenomena might have posed a serious challenge to the principles on which molecular biology was founded. However, in Godfrey-Smith’s view, these phenomena are less threatening now that principles have been supplanted with mechanistic details.
A second objection focuses on the various meanings of ‘epigenetic’ [18,19,20]. Some instances of epigenetic regulation merely involve the (gene-mediated) influence of an environmental factor on some phenotype. Gene centrists have always allowed that environmental factors influence gene expression. Such examples of epigenetic phenomena are therefore not unorthodox. At the same time, the term epigenetic sometimes refers to the open-ended transmission of a phenotypic change that involves no change in gene sequence. This phenomenon is thought to be rare in eukaryotes [21], but would indeed call for a radical shift in biological thinking if it were common. Conflating familiar epigenetic effects with rarer or more controversial phenomena potentially gives a distorted impression of what the study of epigenetics is about.
A third objection concerns the functional interpretation of certain epigenetic phenomena. Epigenetic revolutionaries point to examples of phenotypic mutation that are induced by some environmental change, appear to be adaptive for the organism, and involve no change in DNA sequence, but are transmitted in sexual lineages across generations. Such examples are interpreted as evidence for a switch-like mechanism that rapidly adapts the phenotype to environmental change. This mechanism is allegedly less visible from a research program focused on genes. Also, if adaptive epigenetic inheritance is common, this challenges the neo-Darwinian idea that phenotypic adaptation typically involves random genetic variation and selection.
Our first aim in this paper is to explore an alternative explanation of epigenetic inheritance that views it not as an adaptive epigenetic switch, but rather as the byproduct of transposon dynamics. This explanation has long been available but is rarely considered, raising the question of whether transposon dynamics generally tend to be neglected in discussions about epigenetics. Our second aim is to address this question using a quantitative analysis of papers sampled from the Web of Science. In this way, we examine the popularity of epigenetics versus transposons across different disciplines over the past five decades. Finally, using a qualitative analysis comparing different conceptions of epigenetics across disciplines over the past twenty-five years, we compare the varied disciplinary views of epigenetics researchers on the topics of heritability and function.
Epigenetic switches and the significance of transposons
One of the most widely discussed examples of epigenetic inheritance involves the transmission of coat coloration in lab mice. The agouti gene is expressed in mouse hair follicles and normally produces a dark brown coat. However, in some mice there is a change in the expression of this gene, producing a coat that appears sometimes yellow or on other occasions variegated. All strains of mice share an identical agouti gene with no variation in nucleotide sequence. Differences in coat color are instead produced by variation in methylation patterns upstream of the pigment gene. An interesting feature of this example is that color pattern is maternally inherited for up to three generations, indicating that parents transmit methylation patterns to their offspring.
The agouti gene has become a model system for epigenetics. For instance, a study by Dana Dolinoy et al. exposed female mice to bisphenol A (BPA) and noticed a shift toward yellow in the coat color distribution of their offspring [22]. Again, variation in coat color was caused not by a DNA mutation but rather by a change in methylation. Moreover, the effect was counteracted when female mice were fed a diet supplemented with methyl donors.
Such examples have been interpreted as evidence for an epigenetic inheritance mechanism, or switch, that rapidly adapts organisms to their environment. In discussing agouti gene expression in mice, Jablonka and Lamb propose:
Because it provides an additional source of variation, evolution can occur through the epigenetic dimension of heredity even if nothing is happening in the genetic dimension. But it means more than this. Epigenetic variations are generated at a higher rate than genetic ones, especially in changed environmental conditions, and several epigenetic variations may occur at the same time. Furthermore, they may not be blind to function, because changes in epigenetic marks probably occur preferentially on genes that are induced to be active by new conditions. [12, p. 144]
Likewise, Bonduriasnki and Day claim that the agouti mouse example “shows how such epigenetic traits could contribute to adaptive evolution” [13, p. 58]. There are three basic components to this interpretation. First, there is the proposal that phenotypic changes are induced by the environment. Second, there is the claim that those changes involve a modification to methylation or some other epigenetic mark, but no change in gene sequence. Finally, there is often the suggestion that epigenetic changes are biased toward adaptive phenotypic responses. The conjunction of these three propositions is what we refer to as the epigenetic switch hypothesis.
Others have raised doubts about the existence of epigenetic switches because the relevant effects persist for no more than three generations. To be of evolutionary interest, it is argued, an epi-mutation would have to persist for much longer. A recent review by Alfredo Sánchez-Tójar et al. found little evidence for such transgenerational epigenetic effects. However, this remains a topic for further research [21].
Perhaps a more philosophically interesting objection concerns the fact that the agouti mutation involves the suppression of a transposable element, located upstream of the agouti gene. Jablonka and Lamb mention in passing that “there was a small extra bit of DNA (originating from a transposon) in the regulatory region of a coat color gene” [12, p. 142]; but they overlook the theoretical significance of this point. As we explain in the next few paragraphs, the fact that epigenetic mutations are often transacted by transposable elements suggests an alternative to the epigenetic switch hypothesis.
Transposable elements (TEs) are mobile strands of DNA capable of jumping into new chromosomal locations. The act of transposition (jumping) often involves the creation of additional TE copies. Hence, individual TEs can replicate multiple times per generation in a process akin to meiotic drive. It is well known that TE insertion can interfere with protein synthesis or cause various sorts of harmful mutation. Organisms have thus evolved a variety of mechanisms for deactivating, suppressing, or removing TEs from the genome. These mechanisms, in turn, impose a selection pressure on TEs to evolve ways to overcome the host organism’s defenses. Over millions of years, these coevolutionary dynamics have given rise to eukaryotic genomes replete with TEs—with 40–60% of the nuclear DNA in humans descending from TEs—most of which are temporarily silenced or permanently deactivated [23].
There are several reasons TEs may appear to have organism-beneficial functions when they are in fact deleterious. One way for a TE lineage to potentially avoid deactivation or deletion is by inserting copies very close to a protein-coding gene [24]. These sites are safe havens, so to speak, because the host cannot easily methylate TEs at these locations without altering the expression of its own genes. It is therefore no surprise that many TEs preferentially insert close to protein-coding genes [25].
It is easy to mistake these stealthy TEs for organism-beneficial insertions [10]. Genomics researchers identify the strands of DNA located adjacent to genes as regulatory regions because they contain transcription factor binding sites. The occurrence of TEs within regulatory regions has led some genomics researchers to implicate them in gene regulation, disregarding the possibility that the TEs might simply be hiding in a safe location. This interpretation is further supported by the fact that TEs contain their own binding sites which are normally used to harness the host’s replication machinery for their own benefit. Hence, TEs are especially effective mimics of genuine regulatory regions.
Another deceptive feature of TEs is that they are activated by stress. When an organism is exposed to chemical, thermal, or other forms of stress, there is sometimes a burst of TE activity [25]. Barbara McClintock has interpreted TE bursts as evidence for a switch-like mechanism that facilitates rapid phenotypic adaptation by elevating mutation rate [26]. Once again, however, the situation looks different from the perspective of TE–host coevolution. Organisms employ various strategies to protect genes from TE insertion. Some suppression strategies occur at the level of the DNA strand, where methyl groups are inserted on top of transposon binding sites to prevent them from being recognized by the host’s transcription factors and replicated. In fact, it is now thought that DNA methylation originated as a system for TE suppression, with gene regulation a secondary (exapted) function [27]. Important for our argument is that suppression mechanisms are themselves compromised by stress. Just as a parasite can get the upper hand on a patient with a compromised immune system, so can TEs flourish in a genome with weakened suppression. Thus, what appears to be switch-like behavior in response to environmental change might in fact be a breakdown in TE suppression machinery.
These considerations cast new light on the agouti mouse example. Recall that variability in coat color is caused by variable methylation patterns surrounding TE insertion upstream of the pigment gene. It is quite plausible that different color morphs represent different levels of TE suppression, with more heavily methylated strains being a step ahead in the coevolutionary arms race. Were this TE to degrade or be removed, the site would presumably cease to become hyper-methylated and the yellow phenotype would disappear. Moreover, if this example is typical, and epigenetic effects typically involve an effort to suppress TEs, then it is unlikely that epigenetic mutations will have adaptive effects. It is essentially up to the transposon to determine where it wants to insert. Selection acting among TE lineages (within the organism) will favor transposons that avoid detection and deletion. This might involve stealthy insertions close to genes in some cases or in other cases the avoidance of genic regions altogether, but there is no reason to expect an insertion preference for regions that will benefit the host.
David Haig argues that it is often in the evolutionary interest of both the organism and the transposon for TE insertions to be silenced in somatic tissues (as opposed to the germ line) [28]. This allows the host organism to survive and reproduce, passing along its complement of TEs to the next generation. Evolutionary interests conflict more directly in the germ line. If a TE insertion kills the host, then the TE will be removed from the population. This imposes a downward selection pressure moderating the rate of TE replication. However, it has long been recognized that in sexual species it is difficult for selection to entirely purge the genome of determinantal TEs [29]. Eukaryotic organisms are stuck with these genetic parasites and, again, there is no reason to expect that TEs will preferentially insert into regions that are likely to benefit the host. Nor does the methylation of those insertions occur with some directed beneficial effect on the organism, other than to mitigate the negative effects of a TE insertion on normal host function. These considerations cast doubt on the idea that epigenetic responses to environmental change will tend to be adaptive, at least, not insofar as they are associated with the suppression of TEs.
If epigenetic differences are typically driven by responses to TE insertion, this also has implications for the persistence of epi-mutations. Organisms are engaged in a constant effort to detect and suppress TEs. Eventually, active TE insertions will degrade and no longer attract methylation. As a result, any TE-mediated switch will have a limited life span because processes within the organism are actively degrading it.
What about the suggestion that epigenetic switches respond to specific environmental cues? From a coevolutionary perspective, not just any environmental factor can be “hooked up” to the epigenetic machinery. If the loss of methylation is typically caused by a breakdown in TE suppression, then only harmful environmental factors will induce this type of epigenetic change. Relatedly, after the stressful conditions have subsided, the TE suppression machinery ought to resume its job of methylating TE insertions. Hence, unless the organism is exposed to a continual regime of stress, persisting over many generations, one would expect TE-based epigenetic mutations to be short lived.
The topic of TE–host dynamics is a fascinating area of research that would take us beyond the objectives of this paper to describe in detail. We hope to have said enough to at least raise questions about the ways that examples of epigenetic inheritance are interpreted by some proponents of the epigenetic switch hypothesis. At the very least, one might expect that considerations about TE dynamics would be raised as an alternative explanation for examples such as agouti gene expression in mice. Instead of being viewed as an epigenetic switch, the environmental induction and epigenetic transmission of the colored phenotype might simply be the byproduct of TE suppression. Why has this alternative been largely ignored by authors working on epigenetic inheritance?
It has been suggested that the fields of molecular biology and genomics are simply out of touch with recent trends in evolutionary biology [30]. This could be due to insufficient evolutionary training in those fields. Another potentially relevant factor is the high prevalence of adaptationist thinking within molecular biology and genomics. A number of authors have noted that adaptationist hypotheses are unjustifiably popular in these disciplines [7,8,9, 31, 32]. Another, non-exclusive possibility concerns the influence of large research consortia like ENCODE and the economic incentives driving these projects. Garnering large sums of public funding sometimes involves interpreting results in ways that sound exciting, revolutionary, or relevant to human disease. Describing examples like the agouti mouse coat coloration as an epigenetic switch sounds more exciting than the alternative possibility, that this phenomenon is the fleeting, stress-induced byproduct of a genetic parasite.
We have suggested that information about TE–organism coevolution recommends an explanation of certain epigenetic phenomena that rivals the epigenetic switch hypothesis. This raises the question of whether, given the ballooning popularity of epigenetics research, those coevolutionary dynamics are generally being overlooked or downplayed. This question can be explored by comparing the relative popularity of epigenetic versus transposon research over time and across disciplines. We expect that researchers working in the field of evolution, who are familiar with genome-level coevolutionary dynamics, are less enthusiastic about epigenetics compared to researchers working in proximal biological sciences, where evolutionary thinking is less common. Likewise, if the attraction to epigenetics is influenced in part by large research consortia like ENCODE, then one might expect epigenetics to be more popular in biomedical biology and genomics compared to other disciplines.
A related set of questions concerns the ways that different disciplines conceptualize epigenetics. It is possible that researchers in biomedical fields rarely embrace the epigenetic switch hypothesis and use epigenetic to refer to different phenomena than researchers working in other disciplines, for instance. The remainder of this paper describes two bibliometric studies attempting to shed light on these questions.