Remembrance of things past--in your genes? Part I: Is epigenetics 'Lamarckian'?

Marcel Proust's epic novel, In Remembrance of Things Past, was a 20th century masterpiece of thoughtful reflections.  It is about small things that trigger recollections of events that happened earlier in life but that were otherwise lost to memory (an alternative title translation from the French is In Search of Lost Times).  For Proust it was the madeleine, a small cake he dipped in tea (the most famous example), or any number of other unexpected nostalgic triggers.  These evoke times past without one explicitly trying to dig into one's memory: they just come up.  We all have that experience, I assume. An odor, a piece of music, a food trigger a particular birthday or Christmas or girlfriend's name, or event at the beach.  In some instances, Proust's protagonist remembered things about his parents or other relatives.  But wistful as such memories were, they are things in the past that can be retrieved in memory but not in reality.  Or can they?

Forget the parents?
I'm not talking about, say reincarnation or strangely eerie experiences like deja vu. Instead, I'm thinking of reports over the years, and increasingly these days, of genomes that 'remember' events that happened in their, and their ancestors' past.

The consensus idea about genome evolution, supported with overwhelming evidence of all sorts, is that genetic mutations, that is, changes in DNA sequence, occur through various chemical process that are random with respect to any functional effect they may have.  It is chance and various forms of selection (see our series of posts on these, starting here), that determine which changes proliferate over the generations.

Before the understanding of the nature of genetic inheritance, going back from Darwin's wild guesses about gemmules, to Lamarck's famously dissed ideas about the inheritance of acquired characteristics*, even to Hippocrates, was the reasonable idea that your traits were somehow the result of elements (now we would call them 'molecules') that traveled to your gonads to be transmitted to offspring.  They were molecular images, so to speak, of who you were.  It made sense, but Mendel's work and much else showed clearly that that is not how inheritance works, or evolves.  Lamarckian thinking of that sort is out, and not because the Mendelians are bullies, but because there isn't any evidence for it!  Or is there?

Lamarck redux, or trying too hard?
There are many incentives for scientists to try hard to be the next Darwin, and to press their hopeful ideas to the public.  But most such claims are hopeful monsters, quickly shown not to be true. For example, ten years ago two Purdue plant geneticists published a paper in Nature that reported that the plant Arabidopsis, related to mustard, had a self-correcting mechanism that for generations could restore 'good' gene versions, replacing 'bad' mutants.  This if true would be a kind of Lamarckism, in that the organism could remember what was good and impose it (as I recall, the authors were not claiming to be resuscitating Lamarck, but that this was a new or different kind of inheritance).

It quickly turned out that the results were due to some experimental artifact, and pollen contamination and/or other problems undermined these results and nobody, probably not even the original authors, believes them any longer (the authors have tried to suggest that the artifacts didn't explain everything, but this hasn't been convincing and not even the authors seem to be following it up).  Nice try, but no cigar.

At the same time, over recent years, there have been findings that suggest that experiences acquired during life, that involve gene expression, could be transmitted to offspring, but were not encoded in DNA sequence.  Instead, they were epigenetic, that is, they involved modifying the DNA sequence in a way that affected which genes were being used in given contexts.  Clear examples involved coat color genetics in mice and some physiological responses related to obesity and associated traits.  The idea is that a fetus could acquire the change in gene usage while in utero, which would affect its traits, and then 'remember' the gene-usage setting and transmit it to their offspring.

There have long been suggestions that offspring can, even when adults, have physiological traits that resemble their parents, not through inheriting genetic variation but inheriting physiological states themselves.  Examples included traits like blood pressure, where mothers' state during pregnancy set their children on a related path, such as to having high blood pressure.  Terms like 'set point' were used to describe how the infant's body was 'set' to respond to its life experience in a way that resembled how the mother's state was, but that was not because of her specific genotype.  These results could not be related to known genes at the time, but technology has improved and there are several examples and some of the genetic basis seems to be becoming known.

Epigenetic changes are real
In short, what we know has to do with gene usage, not gene sequence.  Gene usage is affected by very well-known mechanisms that modify the chemical state of a given DNA region in ways that enable a nearby gene to be used (or, depending on the mark, prevent usage).  This is known as epigenetic marking because it is not due to mutations in DNA sequence.  The difference is basically that between Lamarckian and modern inheritance ideas in that epigenetic changes can be directed--that is, set or removed--based on experience.  Generally, the idea is that epigenetic changes are responses that relevant cells 'know' how to make in a given environmental context.

These effects thus seem 'Lamarckian' in a restricted sense, but that has been thought to be very restricted.  Your body's cells, say muscle or heart cells,  may respond to their environment (e.g., what's passing by in the blood stream) by context-specific mechanisms that use epigenetic marking to turn some genes on or shut others down.  This may be inherited by the body's cells, when they divide, unless changing circumstances lead them to change the genes they're using.  But that would not necessarily be inherited, because all you transmit to your offspring is a sperm or egg cell.  There is no reason to think that, for example, nutritional components that affect how insulin is used or how fat cells store energy would affect the use of energy-storing genes in sperm or egg cells.

This might happen, however, if the body cells, or even all the cells of a fetus including the germ line, also sense the maternal environment, and that in turn induces similar changes in the fetal cells.  The physiological settings of the fetus would reflect the mother's experience, not by being transmitted in her egg cell but by the effects of her blood constituents via the placenta.

A fine review of the state of knowledge as of 2014 is by Heard and Martienssen ("Transgenerational Epigenetic Inheritance: Myths and Mechanisms," Cell, 3/27/14, p99). This post today and Monday's post take some selective bits from that, but if you are interested in this subject, it seems to be an excellent source to read carefully.

This figure shows how exposure in utero can transmit to the offspring (F1 generation) or grandchildren (F2), but only if it also appears in the next generation in unexposed individuals is it really now incorporated into the genome for future generations.  The figure labels stress and nutrition as possible causes of gene-expression change that could get into the germline.

Figure from Heard and Martienssen, 2014, Cell

If altered epigenetic settings affected all the fetus's cells, then the settings could be inherited by its offspring, that is, the mother's grandchildren.  Even with no further environmentally induced effects, this could indeed be transmitted for multiple generations, and be more truly 'Lamarckian' inheritance into posterity. Unless the pattern really shows up in the 3d generation, it will not be considered truly transgenerational.  But even then, or with the examples, there are at least three problems to consider.

First, we know that these set-points are generally changeable during life.  Circumstances set and un-set them, as cells respond to their environment.  Indeed, epigenetic changes are in large part responsible for how tissues differentiate--into stomach, lung, brain, skin, etc.--during embryonic development.  That's how you become a differentiated organism.  Having gene-usage too rigidly programmed during an adult's life (the parent), could prevent the offspring from even becoming an offspring.  So there is likely to be a re-set mechanism so a fertilized egg can start life anew.  Why would your Mom's experience override your own responses?

Secondly, the evidence to date suggests, at least, that even the persistent marking that's been observed fades or disappears eventually.  It is not as permanent as changes in the DNA sequence itself (the genes proper).  The problem that raises is that it won't be part of long-term evolution.  Unless--unless a phenomenon called 'genetic assimilation' occurs.  That's when something not engraved in DNA persists for whatever favoring reason and eventually some 'real' mutations--DNA sequence changes--with similar effect arise.  In that case, the actual hard-wired changes can persist, with the 'good' trait being produced even when the epigenetic marking has long gone.  But that then is a form of ordinary rather than Lamarckian selection, and how often would such things arise in the world? This has been debated since CH Waddington's advocacy of the idea as an important evolutionary process, in the mid-20th Century, and indeed the idea was proposed in the late 1800s before any actual genes were known, in a somewhat different context and known as the Baldwin effect.

Thirdly, and at least as important, too much epigenetic change could prevent new mutations--'real' evolutionarily relevant change--from having effects, if gene usage patterns, which are an important part of evolution, were too rigidly entrenched by DNA marking.

Fourthly, sperm and egg cells are developed in particular cell lineages in a fetus, initially called primordial germ cells (PGCs), cell lineages that are isolated from cell lineages that form the rest of the body, and vice versa.  So how is it that something specifically affecting gene usage in a particular organ, like a blood vessel, kidney or eye, would also be 'set' in the gonadal cells?  How exotic would such an information-passing mechanism have to be, and if it exists would we have to re-think our skepticism about Lamarckian inheritance--because if such specific mechanisms exist, they would really be the transmission in the patrimony, of things acquired by experience during life?

In fact, one recent paper has suggested that this may be occurring.  Male rats exposed to a particular odor that is known to activate one particular odor-recipient ('olfactory receptor', or OR) gene and at the same time exposed to a mild fear-inducing stress, were conditioned to activate that OR when exposed to the stress. These males were mated to unconditioned females, and the investigators report that the offspring males respond strongly to that odorant (via the specific OR gene).  But when such offspring males, who had not been exposed to the conditioning fear stress, were mated with unconditioned females, the next, grand-rat, generation also showed the preferential usage of this OR gene.

This latter study is remarkable if you think of the fact that a rat has about 1000 different OR genes, two instances of each (one inherited from each parent), and they are scattered in sets of varying numbers across most of its chromosomes.  So how is it that the rat's body 'knows' how to just mark that one particular OR gene, not just in its nose cells but also in its sperm cells, so the marking can be inherited?

If this experiment is to be believed, and it is remarkable enough that it must be carefully tested, and the effect shown to last for further generations, then we have to wonder what mechanism might underlie it. Is it really Lamarckian?  There is a sort of precedent, in the worm C. elegans, in which olfactory gene switching can be inherited for multiple generations and may have adaptive function.  But if this is true in mammals, is it general enough to challenge the axiom about inheritance on which current evolutionary theory rests?

Adaptive function would seemingly be something a species would 'want' to be permanent and not easily erasable or, put in a more mechanistic way, the freeze-in-place mechanism would eventually work against adaptation if circumstances changed.  And how would such decision-making characters work?  Wouldn't genetic assimilation remove this from epigenetic control?

Some new papers in the June 4th issue of the journal Cell raise questions, or perhaps raise serious problems, about how to interpret these various results on gene-usage effects and their inheritance, and we'll talk about them tomorrow.
-------------------------------------

*Lamarckian inheritance has been badly misunderstood, as we'll discuss in part 3 of this series later this week.