Not genes and not environment

Share on FacebookShare on Google+Email this to someoneTweet about this on Twitter

On many measures, identical (monozygotic, MZ) twins are not in fact “identical”, despite the fact that they share essentially identical DNA and highly similar family environments. Indeed, for some traits, such as personality, all non-genetic effects appear to be of the kind that makes siblings different than one another. Peer socialization is one plausible source of this non-shared environment, but stochastic biological events probably play a role as well.

These stochastic effects are seen prominently in studies of aging in the worm C. elegans. Even when genes and environment are held constant[1], there is considerable variation in lifespan (time of death) within a population. (Almost as much variation in relative lifespan as the human population of the US.) A paper published this week in Nature Genetics (ironic) reports that chance variation in the level of induction of a stress-induced reporter predicts (to some extent) variation in lifespan.

When both genotype and environment are held constant, ‘chance’ variation in the lifespan of individuals in a population is still quite large. Using isogenic populations of the nematode Caenorhabditis elegans, we show that, on the first day of adult life, chance variation in the level of induction of a green fluorescent protein (GFP) reporter coupled to a promoter from the gene hsp-16.2 predicts as much as a fourfold variation in subsequent survival. The same reporter is also a predictor of ability to withstand a subsequent lethal thermal stress. The level of induction of GFP is not heritable, and GFP expression levels in other reporter constructs are not associated with differences in longevity. HSP-16.2 itself is probably not responsible for the observed differences in survival but instead probably reflects a hidden, heterogeneous, but now quantifiable, physiological state that dictates the ability of an organism to deal with the rigors of living.

The astonishing implication is that similar reporters could be found that predict the chance variation in human lifespan (or other dimensions of human biodiversity).

How did they do this research?

First, they created a strain of worm which fluoresces in response to exposure to high temperatures (“heat shock”) (panel “a” below). The “heat-shock gene” hsp-16.2 is expressed when worms are heat shocked. The promoter of hsp-16.2 was joined with the protein coding sequence of green fluorescent protein (GFP) and this construct was integrated into the worm’s genome. The intensity of GFP expression (measured by fluorescent intensity) is variable within an isogenic population (panel e). For their experiments, young adult worms were exposed to a heat shock (panel b) and then some time later they the worms were sorted into high, medium and low GFP subpopulations (panels f-h). The level of GFP expression becomes more variable at later times (panels c-d). The GFP expression level is not heritable: the progeny of worms from the high and low groups are indistinguishable when tested for GFP induction.

After sorting into three subpopulations, worms were tested for lifespan (panels a-c) or thermotolerance (i.e., lifespan at high temperature; panels d-f).

Numerous controls follow. The biggest problem for their study is that heat shock is known to cause increased lifespan on its own. They don’t claim to have overcome this confounding effect, and so it seems to me that differential GFP expression may actually be a marker for this effect.

1 – C. elegans has two sexes: male and hermaphrodite. Hermaphrodites are self-fertilizing. Selfing allows for the production of genetically homogenous populations (except random mutations). Worms are grown on solid media in Petri dishes (“plates”) or in liquid cultures. Worms living on the same plate are essentially experiencing the same environment, but undetected variation may exist between different plates or within different regions of the same plate.


  1. my impression is that much of the “penetrance” effect of many deleterious genotypes (that is, don’t get sick sometimes, do other times) have to due with epigenetic probabilities. so, my ? is how much do we really know about epigenetics in humans? are there seminal review articles? i’ve seen stuff on a family with a particular disease and weird penetrance patterns and different patterns of methylation, but it seems all over the place.

  2. Epigenetic differences arise during the lifetime of monozygotic twins. 
    check it out if you haven’t seen the full text

  3. tx. that’s what i was looking for!

  4. some others: 

  5. So, for the layman – could this help explain (partially) those sometimes huge “this should be environmental influence, but we can’t pinpoint what it is” factors from twin studies, etc?

  6. Fanatastic! 
    really enjoyed this – and learnt something new 

  7. I’d be skeptical. Worms are much smaller than humans, and much more R in their selection strategy, suggesting that much more noise is likely. Organisms that follow a K selection strategy, and which have enough cells, even in a blastocyte, to statistically reduce noise to managable levels, probably do so. 
    By removing noise from an organism’s development, they would remove noise from its selective fitness, enabling more efficient natural selection, which could be faster, conveying a long-term selective advantage. Even if they don’t do this, they probably want to minimize noise because in general random deviation from a template is harmful.

  8. some things to note: 
    Worms are small – less than 1000 somatic cells; meaning the per-cell frequency of errors may be higher while the total error load may not be 
    Their development is tightly regulated – their cell lineage is invariant between individuals; at the level of cell division and differentiation, they are physically identical 
    They are short lived – 2-3 weeks, depending on conditions; yet the variability in their lifespan is nearly as great as lifespan variation in humans in the US

  9. What really sparks my curiosity is the idea that chance variation may be structured, so that seemingly disparate kinds of errors are correlated. With enough biomarkers, it may be possible to extract one or more statistical factors that predict variation in health, longevity, etc. This, I think, was completely unexpected. Instead we would have thought that variation was just the noise of mostly unrelated stochastic events. It’s also interested that these variations exist despite the tight control over development. So, I think it’s possible that these correlated variations are in some way regulated; or to put it in teleological terms: a design feature rather than a flaw.

  10. Among humans, identical twins often develop different personalities precisely because they are in constant contact with each other — e.g., one becomes the leader and the other the follower because it’s easier to get things done that way, and it’s not all that exploitative a relationship because the follower would probably make similar decisions if he was the leader. Heinlein’s “Time for the Stars” has an insightful portrait of identical twins.

  11. A confounding factor with MZ twins might be how the egg splits. Did the original egg divide evenly and were the same growth factors evenly distributed. Even the axis along which the egg divides could make a difference. Did one twin arise from a single cell after the egg had already divided several times? Some difference could be due to DNA methylation while others arise due to different concentrations of regulatory factors. (Or so it seems to me from my general reading.)

  12. I think the interesting question is what proportion of twin differences are due to a few (relatively) random events that crystalize into large scale differences versus a constant stream of stochastic events nudging them apart.

  13. If one twin received more organelles in the “split”, or more blood supply from the placenta, and grew minutely larger, stronger, and smarter in the womb–almost all the post-natal differences would logically follow. 
    The stronger, smarter one would be the leader and dictate the terms of their co-evolutionary environment.

  14. what proportion of twin differences are due to a few (relatively) random events that crystalize into large scale differences versus a constant stream of stochastic events nudging them apart.  
    yes, quantifying the dimensionality of the underlying stochastic process.  
    it will be good to take a second look at these reporters transplanted into hsp mutants — the ones which have reduced buffering of developmental noise…