Readers will know that from time to time (e.g., here and here) I weigh in on the persistent and loud claim that epigenetic inheritance (that is, the transmission from parent to offspring of traits that are not coded for in the DNA) will have huge effects on the current paradigm of neo-Darwinian evolution. There’s a segment of the evolution community who sees this form of nonstandard inheritance as a revolution in our field. That’s because, perhaps, some environmental modifications of an organism—changes induced by climate, diet, or the like—might become inherited, forming a type of “Lamarckian” inheritance. (Lamarck [1744-1829] was a French zoologist who proposed that evolution occurred by the inheritance of acquired traits.)
Now it’s unlikely that, say, a change in diet or habits alone will cause changes in an organism that can be passed on to its offspring. Athletes don’t tend to produce muscular children, nor amputees legless ones. But there is one type of “acquired” trait that can be inherited, at least for a few generations: the attachment of methyl groups to DNA. Certain components of the DNA, most particularly the cytosine residues (“C”s: one of the four DNA bases), have a tendency to be methylated: a one-carbon methyl group attaches itself to the 5 position of the ring in that DNA base. Those changes, which can have important effects on gene regulation, can themselves be coded for in the DNA (that is, there is some gene which gives instructions for another gene to become methylated). But some methylations seem to occur spontaneously, without any genetic instruction, and these are in effect environmentally induced changes that can be inherited from parent to offspring. It is those changes that many evolutionists point to as the “nongenetic” inheritance that could revolutionize our view of evolution.
Whether such changes can indeed be important material for evolution is the subject of a new paper in Nature by Becker et al. The authors examined DNA methylation in the plant Arabidopsis thaliana, a small (and largely self pollinated) plant known as the “Drosophila of plants” because its ease of culture and short generation time (six weeks from one generation to the next!) makes it suitable for breeding experiments.
The point of the experiment was to start with a single homogeneous inbred line of the plant and then subculture it into ten lines, all of which were genetically identical at the start of the experiment. Each of the ten lines was then propagated for thirty generations through inbreeding, and then subjected to DNA sequencing. That DNA sequencing detects which bases are methylated, and then the methylation patterns of each line could be compared to see two things: 1) how much the initially identical lines differed in methylation patterns after 30 generations; that is, how much the environment had created different methylation patterns in the different lines; and 2) how “heritable” those changes were among generations. That could be seen by comparing the generation-30 lines with some generation-3 lines that were ancestral. If stable methylation patterns arose in the different lines that could be transmitted over 27 generations, that means that, at least in the short term, acquired changes in the DNA could be inherited.
Here are the important findings:
- The lines rapidly accumulated differences in methylation patterns of DNA, and some of these could be stably inherited over at least 27 generations (they had only two lines to do this 27-generation comparison, so the results are tentative). But many changes also disappear over time, and thus are not stably inherited.
- Despite this high mutation rate, the number of methylated sites does not accumulate linearly with time. The authors consider this their most important finding because it indicates “that many [epigenetic changes] are not stably inherited over the long term.”
- The “mutation rate” to methylation changes of cytosine bases was very high: roughly a thousand times higher than the “real” mutation rate at which one DNA bases changes to another (the latter are what has been considered the main heritable basis of evolutionary change).
- Methylation occurred at certain “privileged” regions of the genome: in the coding parts of genes (“exons”) rather than noncoding parts, and in regions closer to “transposable elements” (bits of the DNA known to jump around in the genome).
- Methylation could affect gene expression. Analyzing the three genes with the highest differences in expression levels among lines, the authors found that the less-methylated genes were expressed more strongly. Methylation tends to reduce gene expression, then, but of course reduced gene expression can also be of evolutionary significance.
- The authors don’t know how this differential methylation of DNA bases occurs, though they hypothesize that “siRNAs” (small interfering RNA molecules, which affect gene expression themselves) play a role.
What’s the upshot? First, that environmental changes in DNA not mediated by genes did occur: the different sublines of the plant, though genetically identical, accumulated different methylation patterns in their DNA. These environmental “mutations” occur very rapidly and some of them are inherited over several generations. Some of them could affect gene expression, too. Putting it all together, the experiment shows that it’s theoretically possible for environmental influences to produce inherited changes that could affect evolutionarily important traits (in this case the level of gene expression). In other words, it’s possible for evolution to occur in a Lamarckian way.
Does this, then vindicate Lamarckian inheritance and presage a revolution in evolution? I don’t think that’s likely. As I’ve written before, every evolutionarily important change that has shown to be inherited, and has been mapped to specific positions in the genome, shows that real genetic mutations—not methylated changes in DNA bases—are responsible. While it’s possible that some adaptations or evolutionary changes could rest on epigenetic modifications not involving substitutions of one DNA base for another or an interruption of DNA sequenes by the interposition of other sequences, we haven’t yet found any.
At the end of the paper, the authors raise another problem with touting these environmental changes as important sources of evolution change: the modifications don’t appear stable in the long term, and so couldn’t be the basis of adaptations that arise and are stable over thousands or millions of years:
Perhaps our most important finding is that the number of epimutations does not increase linearly with time, indicating that many are not stably inherited over the long term. In addition to DMPs [dfferentially methylated positions] and DMRs [differentially methylated regions] that arose apparently independently in several strains, we even discovered a DMR that had become demethylated after 31 generations, but was re-methylated in the following generation. This suggests that DNA methylation in specific regions of the genome can fluctuate over relatively short timescales. Such sites can be considered as going through recurrent cycles of forward and reverse epimutation, which is very different from what is found at the level of the genome sequence, where reverse mutations are exceedingly rare. Importantly, reversion rates directly determine the ability of any type of allele to be subject to Darwinian selection. This needs to be taken into account when considering the potential of epialleles as a factor in evolution.
Translation: These epigenetic changes in the DNA aren’t all that stable, folks, so we need to be really careful before touting them as important aspects of evolution.
My conclusion: Though the results are intriguing—especially the observation of a high rate of epigenetic “mutation”—there’s still no reason to see this type of heritable change as presaging the overturning or drastic revision of the current neo-Darwinian view of evolution.
___________________
Becker, C. et al. 2011. Spontaneous epigenetic variation in the Arabidopsis thaliana methylome. Nature, in press. Published online, doi:10.1038/nature10555
Were the ten lines exposed to very different environments? If not, why would one expect that methylation changes would be stable? In other words, if the specific methylations didn’t have an opportunity to confer selective advantage to a line, we shouldn’t necessarily expect them to persist in a line, should we?
If they were stable, we would expect them to persist because the lines were inbred – there wasn’t any outcrossing, so no new variation was introduced. The fact that some methylation changes didn’t persist shows that they were sometimes spontaneously lost between one generation and the next.
Good point – “Plants were grown on soil under long-day conditions (23 °C, 16 h light, 8 h dark) after seeds had been stratified in 150 nM GA-supplemented water at 4 °C for 6 days. Siblings were grown independently at different time points. Positions of the pots were randomized.” But I see no mention of differing conditions for different lines on a quick look. However is the point of methylation that it is a change that does not have to produce a selective advantage or disadvantage? If one did it would surely wipe the line out or make it dominate – while methylation lasted?
My point was that if the selective environment isn’t different among the lines, then there is no pressure for the methylation to differ among lines, or to become stable over generations. Consider if this study were merely looking at changes in genetics, rather than epigenetics — surely we would say that a lack of stable genetic changes tells us nothing if the lines have the same selective environment?
Perhaps I’m missing something here, but if the question is whether stable heritable differences can arise in different lines (due to whatever mechanism), this doesn’t seem to me to be the optimum methodology to address that.
I think you are missing something – the key point is that there is no variation within the lines, and they’re being inbred. So the only way for methylation to change is if it is spontaneously lost (or gained) from one generation to the next. You’re quite right that if there were variation within the lines we would expect some methylation variants to be lost just by chance.
Also –
“Consider if this study were merely looking at changes in genetics, rather than epigenetics — surely we would say that a lack of stable genetic changes tells us nothing if the lines have the same selective environment?”
The equivalent genetic study would be looking at genetic variants in a clonal organism over the generations. There’s no initial variation, and no sexual outcrossing, so we would expect these genetic variants to persist (barring new mutations) simply because they are heritable.
Right, but that clonal study would essentially be looking at genetic drift, right? And we wouldn’t argue that genetic drift would necessarily produce as stable changes in a line as would a genetic change that is selected for, correct?
I really do understand the setup, but my point is that we shouldn’t expect any significant heritable differences between the lines, or for any differences to be terribly stable, when there is no selection differences among the lines. And that is true whether we are looking at heritable methylation or heritable genetic changes.
Look at it this way — if we repeated the study looking for stable genetic differences among the lines, you wouldn’t expect to find any, correct? So why use this approach for looking at epigenetic differences?
I hope I’m not missing your point here, but if we repeated it looking at genetic differences, we wouldn’t expect to find any at all (barring rare new mutations) because all the lines were derived from a single individual. But if we *did* find genetic differences, we would expect them to be stable because there’s no variation within the lines. Since there’s no variation, there’s nothing for selection or drift to act upon – the way for a genetic variant to be lost would be through a new mutation.
^ sorry, meant to write “…the only way”
Right, so the reported setup is a terrible way to generate any potential heritable differences — in other words, this isn’t the best way to answer “how much the initially identical lines differed in methylation patterns after 30 generations; that is, how much the environment had created different methylation patterns in the different lines”.
Right, and that set up would answer how stable we should expect those effects to be in the absence of any selective pressure — we might expect the changes to remain except in the case of additional random mutations to those changes. But if the organisms were in an environment where those changes were extremely beneficial, we might expect them to be far more stable in the population, as they are preserved via selection pressure against any further random mutation. My point here is that “heritability” as measured by this study isn’t a very good measure of how likely a beneficial change will spread and endure in a population. What the study tells us is that without such positive selection pressure, such methylation “mutations” tend to be lost — it doesn’t tell us whether positive selection pressure would be unable to maintain such changes over time.
Maybe I’m missing something fundamental here, but the study seems to be trying to address whether methylation can be an engine for natural selection by using a methodology that explicitly eliminates any selection forces. And yes, in such a case methylation acts differently than genetic changes. But that doesn’t seem to really answer the question.
Tulse, in order for something to be fodder for natural selection, you first have to show that it can be reliably replicated, and that’s what this study set out to do. Measuring selection presssure wasn’t the goal; measuring stable inheritance was. From that point of view, holding the environment constant across all generations of all lines was the right approach, since varying the enivronment would only have muddied the waters with respect to the goal of measuring heritability. How would you know how much variation was due to environmental factors rather than epigenetic instability? If you observed a persistent pattern in a particular line, how would you know whether it was inherited or created de novo by the environment in each generation? The only way to answer such questions is to eliminate the environment as a variable.
But that presumes that inheritance has to be stable to an arbitrary degree in order to spread through a population that is under selection, and I don’t think that is true. Wouldn’t a methylation change that conferred perfect fitness to its possessor over those without (i.e., all who develop it survive and reproduce, all who don’t die) spread and maintain throughout a population even if the change were not 100% inheritable?
The study saw methylation changes “die out”, and thus concluded they aren’t stable. My argument is that methylation changes that actually impact fitness may not “die out” as readily, precisely because they impact fitness and thus would spread through the population.
I am a bit out of my depth here, but consider this:
(1) Methylation changes occur rapidly in response to environment,
(2) Methylation changes can result in gene expression changes (doesn’t matter if adaptive), and
(3) They are stable for tens of generations
It seems to me, then, that methylation can pave the way for “normal” adaptive evolution by quickly changing the gene expression context in which the new, “normal” mutation arrives. In other words, methylation pushes the system off its evolutionary gene-expression equilibrium in anticipation of incoming mutations and, at least naively, there are reasons to expect that in this equilibrium the new mutation would be disfavoured. If I am not mistaken, this is similar to what Mary Jane West-Eberhard has been writing about.
If something like this occurred regularly, it would be a somewhat novel model of evolution (still, not a revolution of course). I’d be interested what the practicioners think of such a scenario.
Ugh, bad typo: novel *mode*, not *model*.
Thanks for commenting on this really interesting result. Here’s a question: is it necessarily true that environmentally-induced epialleles must be inherited over the long term in order to be evolutionarily important? Studies of maternal environmental effects (which may rely on epigenetics as a mechanism of transmission) have shown that environmentally-induced phenotypes that persist for just a single generation can have remarkable impacts on individual fitness and population growth. I agree that it’s unlikely that epialleles will form the basis of adaptations that persist over thousands or millions of years, but that seems like a fairly narrow view of what is evolutionarily important.
The proportion of methylation occurring in exons compared to introns or even UTRs was interesting. One of the current epigenetic models is that inherited methylation changes may have an effect on evolution due to methylated bases having a higher rate of DNA mutation – and thus the epigenetic change may increase the chances of permanent genetic change in the same genomic location. I would guess, however, that this is still going to be a rare event, albeit much more common than a non methylated base undergoing the same change. One would need to sequence very large numbers of individual cases in order to pick up this type of DNA mutation at the epigenetic site. In this current report they did not sequence enough plants to really test this hypothesis.
Most methylation in humans occurs in promoter regions where most cpg islands are found; these islands are generally unmethylated or differentially methylated. If a cpg in a promoter region is methylated then the gene is almost always under expressed compared to a non-methylated version.
This has been known for some time as have the observations that many disorders have their own specific methylation patterns: tumor suppressor genes in cancer patients being methylated, oncogenes in healthy patients being methylated, increased global methylation in brain tissue of people with dementia, increased global methylation as a function of age, etc.
So like you said DNA methylation effects individual fitness, and in some cases methylation patterns are inherited across generations. The old guard like Coyne and Moran are being a little too eager to evoke officer Bar Brady from South Park: “Nothing to see here, move along.”
There is a very good book about Thale cress & how one single plant influenced the (Norwich) botanist Nicholas Harberd – Seed to Seed: The secret life of plants, http://tinyurl.com/6dn7s7j
Thanks for this distillation of the article. It was fairly opaque to me until I read your analysis. I wonder, is it a different kettle of fish with beasties that are larger & have multiple breeding seasons rather than only having one shot at reproducing? Does a particular methylation in a male gamete remain the same from the early period of fertility to say the middle or end?
Another recent study showed that methylation rates and nucleotide substitutions (i.e. “real” mutations) are correlated.
http://www.ncbi.nlm.nih.gov/pubmed/21696599
Wouldn’t this set up a potential dynamic where a population can try out new variations very rapidly through methylation, and if certain variants are favorable (rapidly increasing in the population), then the abundance of methylated variants would increase the odds of a mutation arising at that locus, potentially making the change permanent?
So it seems to me quite clear here that epigenetic changes are “important to evolution” only in the sense that any other type of environmental or phenotypic effect is. It can affect evolution, but it is not the substrate for accumulated changes — if methylation struggles for stability after only 30 generations, it essentially doesn’t count at all after tens of thousands!
It still could be “important” in the sense that genes could use methylation to improve their inclusive fitness, e.g. a gene that triggers a particular methylation position under certain environmental conditions might improve its inclusive fitness if that methylation can be passed down through a few generations.
And random methylations could potentially have swayed the path of evolution: In the same way that a random event such as a sub-population finding themselves in a different type of environment for a number of generations could influence what mutations are kept, the same thing could happen from a random methylation I suppose.
But it just seems obvious from this data (as well as from common sense) that there’s no way that the accumulation of epigenetic mutations could play any role in descent with modification.
I think a start would be to look at the methylation patterns between species, such as human and chimp, which have been done and have been found to be highly conserved across their genomes. This has also been done in human and mouse where we see a predictably less conserved pattern. We have even looked at cpg substitutions between genomes for species variation and rates of change. How would you interpret conserved cpg substitions between species as well as conserved global methylation patterns?
I would look at it funny, considering that they ought to go away, see above.
Are we sure they aren’t locked in, say by histones for non-coding DNA?
They don’t go away because CpGs are nucleotides and their location in the genome is conserved across lineages. The question I think being asked is whether methylation (which effects gene expression) of CpGs is selected for.
And even then conserved methylation patterns are not themselves evidence of selection. This is because a CpG dinucleotide undergoes slow deamination if unmethylated and rapidly deaminates if methylated. This leads to stable CpG concentrations across related genomes.
To suggest selection you would have to show evidence of an islands increase or decrease GC content…a destablilizing event inferring either loss or gain of methylation. For that you need at least two points of comparison, say human and chimp.
This has been done.
Primate CpG islands are maintained by heterogeneous evolutionary regimes involving minimal selection.
Cell, 2011
My question was if histones can preserve methylation (or perhaps prevents it), as AFAIK it binds tightly on places were there is little transcription.
I’m not sure. Any DNA binding protein can alter histone binding and ultimately chromatin structure, right, so methylation specific binding proteins would fall under that rubric.
Also, I would reverse the cause and affect of the last clause: tight binding reduces transcription…we think.
So what would change if Lamarck had been right? I mean – you could imagine a system in which the germ line could be directly affected by changes in the soma. The only difference with neo-Darwinism would be that evolution would be more rapid than we expect. Natural selection would still work, but it would be operating not just on characters but also on plasticity. No? Epigenetics is fascinating not so much because it undermines neo-Darwinism as because it shows that all those analogies we use – the genome as ‘a blueprint’ or ‘a programme’ are just that – limited analogies, with very simplistic consequences. Living matter is more complicated.
With such plasticity species would not be stable entities would they, as every individual would vary more easily so the only brake on evolution would be the increased incompatibility of individuals…?
I would like to make few points. First of all, calling these mechanisms as Lamarckian is wrong since he never accepted the direct influence of the environment on organisms, but were their habits and actions (not purposeful desire per se) that could have some effect on their evolution. Perhaps this narrow definition of epigenetic mechanisms and inheritance could be called Neo-Lamarckian.
Now, I’ve always considered that due to the spontaneous, reversible and ¨environmentally¨ induced epimutations(chromatin structure changes), these cannot play an important role in evolution.
Second, we must see these epigenetic inheritance systems (EISs) or mechanisms as an adaptation resulting from a more Darwinian evolutionary process. It is doubtless the important role that these mechanisms play in the short term evolution of species (very few generations perhaps), and on their continuous adaptation to short term environmental fluctuation, but due to the nature of these epigenetic changes they cannot be important in the long term evolution.
The evolution of the epigenetic mechanisms is the issue that must interest us, since they allow species to have higher phenotypic plasticity, robustness and perhaps evolvability, allowing them to respond quickly and reversibly to the environment in few or just one generation. These mechanisms most probably evolved through a Darwinian process, later allowing a kind of ¨Lamarckian evolution¨.
Thanks for all the interesting comments. I also seem way over my head here on the technical details, but it seems to me that we are blurring the meaning of “environmentally induced” change. After all, cosmic ray induced mutation of a base is also “environmentally induced” in one sense. If methylation is just something that has a random chance of happening (Jerry used the word trend) minus a specific causal tie to a particular (macro?) environmental condition or ‘habit’ of an organism, then even if methylation changes did accumulate and were stable, we would still be a very long way off from a “Lamarckian” paradigm.
Tulse and I appear to have exceeded the reply limit on the first comment. I wanted to add that I see his point now, which (as I understand it) is that in a real population selection might be strong enough to allow advantageous methylation variants to spread, even if they aren’t stable over many generations.
My guess is that selection would have to be pretty strong for this to happen, since these things seem to be persist only for a few generations. Anyone know of any studies looking specifically at the spread of methylation variants through a population?
That kind of study hasn’t been done yet, but it’s exactly the kind of thing that needs to happen.
Has anyone grokked why well nourished parents give rise to taller children? Why the effect is cumulative over several generations (vide the gradual increase in height among the Japanese)?
Possible counterexample: I ran into a guy at the grocery store, originally from South Sudan, probably Dinka, w-a-y over 6′ tall (in fact, probably the tallest individual I’ve ever met in person), but skinny as all get out. Bet he wasn’t in an overnourished environment!
No need for height to be forced by the same causes. I can well imagine that in the Sudan, the height of the grass is an incentive to be able to see over it when hunting, or to be able to see predators that might take an interest in native individuals for their lunch.
Just idle speculations…
I would say sexual selection is more likely…?
Because the grass is never that high!
Are there actual researchers making this claim or just some know-nothing journalists?
So, methylation is a heritable method of muting certain genes. Most genes are quite long so presumably methylation is not an either/or situation, you can have degrees of methylation. Is this right? But all methylation can do is dial down gene expression. I’m sure this can be very useful, in time of dearth muting genes for growth would probably be advantageous, but it’s hardly going to produce enormous changes in an organism, you can’t evolve feathers via methylation however stable it might be. Also it clearly isn’t very stable so it’s purely for short term adaptation. I wonder if an animal like a horseshoe crab which has changed little in millions of years, has a different methylation regimen than other organisms. Or animals that live in very stable environments like caves where there is little change, might not need methylation as much as other animals.
(subscribing)