Friday, February 19, 2010

Noisy genes and the limits of genetic determinism

Why are genetically identical monozygotic twins not phenotypically identical?  They are obviously much more similar than people who do not share all their DNA, but even in outward physical appearance are not really identical.  And when it comes to psychological traits or psychiatric disorders, they can be quite divergent (concordance between monozygotic twins for schizophrenia for example is only around 50%).  What is the source of this phenotypic variance?  Why are the effects of a mutation often variable, even across genetically identical organisms?

“Nurture” has been the answer proffered by many, but there is good evidence that environmental or experience-dependent effects can not explain all the extra phenotypic variance and in most cases contribute very little to it.  (See post on “Nature, nurture and noise” on June 24th, 2009 for more on this:

An alternative source of variation is intrinsic to the developmental programme itself.  In particular, small, random fluctuations in the expression of genes at various times during development can have large effects on the phenotypic outcome.  A new study in Nature by Raj and colleagues directly illustrates this point for the first time and highlights several important principles of developmental systems. 

They studied the effects of mutations in components of a genetic network involved in the specification of a small number of intestinal cells in the nematode, Caenorhabditis elegans.  This is the perfect organism for such studies, as the cells in question are individually identifiable and generated in an invariant pattern in wild-type animals.  Mutations in one of the components led to an incompletely penetrant mutant phenotype: some animals made intestinal cells and others did not (even though all had the identical geneotype). 

To determine whether noise in gene expression could explain this diversity the authors directly measured the precise number of messenger RNA molecules being transcribed from the genes encoding other components of this developmental pathway in particular cells of each embryo and correlated these measurements with phenotypic outcome.  They showed that the expression of one of these genes in particular became highly variable in the mutant background.  If, by chance, the level of expression crossed a particular threshold it turned on the master gene responsible for intestinal cell specification and these cells were generated.  If the levels did not cross the threshold then the cells were not generated.  In this way, a bimodal phenotypic distribution can arise from an identical starting genotype. 

This study illustrates several important principles of complex regulatory systems that apply not just to developmental and genetic networks but also to neuronal networks.  First, a certain amount of noise is a normal part of the system – a feature, not a bug – that increases robustness to external variation.  Developmental systems are normally buffered, however, to reduce noise in gene expression and to absorb its effects.  This buffering can be disrupted when individual components of a regulatory system are removed; this is why when genes are mutated, one expects (and always sees) not just a change in phenotype but an increase in phenotypic variability.  The effects of stochastic fluctuations in expression levels of various genes can lead to a continuous distribution of phenotypic outcomes or, as in this case, dramatically different phenotypes.  Interlocking positive and negative feedback loops can generate extremely discrete thresholds, where once a certain level of a component is reached it will reinforce its own expression and shift the network into a different state.  Such bistability is a common feature of complex systems and is sometimes taken advantage of to generate phenotypic diversity or plasticity. 

This study elucidates a molecular mechanism of intrinsic variation in developmental systems and shows that it can have a large effect on the eventual phenotype, even in genetically identical organisms.  No matter how precise the recipe, you can’t bake the same cake twice. 

Raj, A. (2010). Variability in Gene Expression Underlies Incomplete Penetrance in C. Elegans: Using Single Molecules To Study the Development of Single Cells Biophysical Journal, 98 (3), 14-14 DOI: 10.1016/j.bpj.2009.12.087


  1. This information is magnificent as the twins can share psychological problems such as esquisofrenia and other situations, this information helps me to my research, for presentation at the schoolBuy Cialis located in ireland

  2. The noise may be quantum mechanical in origin maybe? Here is another piece by Lubos Motl:

    And from this comments this exchange:

    Does the probabilistic nature of the classical world (as a part of which I would include observed human behavior) emerge from (or have anything to do with) the probabalistic nature of the quantum world? Or should I say the probabilistic nature of our knowledge of what is going to happen next?
    Sunday May 22, 2011, 11:00:51
    – Like – Reply – Delete

    Lubos Motl
    Dear Luke, probably yes, there's probably a very good quantum explanation why the anti-quantum or anti-Copenhagen hysteria had to swell in the recent decades but you would need lots of initial data to calculationally prove that the probability is high.

    The classical world is often thought of as deterministic i.e. non-probabilistic, but as you correctly notice, many things in the actual classical world are actually probabilistic as well. That's why we say that the probability is 1/6 for each number when we throw dice. The chances are determined by the person who threw the dice and random quantum processes in his brain - even random at the quantum level - have probably contributed to the randomness that is ultimately seen in dice.

    So much of the "apparent" randomness in the world of classical phenomena - in the classical approximation - could be removed by knowing the "classical" information about the objects and dices that is more accurate. But it's equally clear that we couldn't remove all of its randomness because the seemingly classical phenomena are ultimately caused by quantum processes. So while I believe that the brain isn't a quantum computer in the sense that the things we want to do with quantum computers are not needed for the calculations, it's still true that many of the responses in the brain - and any system - are random and can only be probabilistically predicted, for example the exact moment when a neuron decides that it has had enough and it will react in some way. This is a similar setup to a radioactive nucleus that decays at a random time, and the quantum randomness makes it impossible to find out when.

    I actually think that your focus on the fact that even the classical world was probabilistic in practice is a conceptually deep one. The only difference in the randomness is that in classical physics, one could imagine that aside from rho(x,p,t) probability distribution on the phase space, there could also exist a "real" x(t), p(t), we just didn't know it exactly. However, the philosophical thesis that a real x(t), p(t) exists independently of our ignorance wasn't ever useful for us to learn anything about the system. But we could still believe that x(t), p(t) existed. According to quantum mechanics, we can no longer believe that a real objective state as a function of (t) exists. But even in the classical world, the existence was totally useless for making predictions, even in principle, so the non-existence in quantum mechanics is analogously harmless.

  3. Thanks Luke for your comment and the link. I am not a huge fan of quantum-level explanations of biological phenomena. While quantum phenomena are certainly likely to contribute to the non-determinacy of neural development, there are a lot of other sources of noise on a classical scale that are likely to have a far larger impact. Noise in gene expression is the most obvious source - exacerbated by the fact that genes tend to be transcribed in bursts, rather than at a steady state. Coupled with the fact that many events in neural development involve small numbers of cells over short time periods, this could significantly increase the variability of the eventual outcome.

  4. There is definitely so much to learn about this. The genes are definitely linked between the two here. So many similarities that go a long with it. Home insurance Stoney Creek