Field of Science

Stochastic cell differentiation my ass

Jean-Jacques Kupiec est un scientifique français. He has a new book out by the title The Origin of Individuals (2009), and Eric Werner reviews it in the latest issue of Nature.

Kupiec's idea is that there is a stochastic element to development which dominates over determinism:
In the standard view, development is controlled by the binding of protein transcription factors to promoters that activate genes in the DNA. These genes in turn generate proteins, including other transcription factors and signalling molecules that activate yet more genes. A cascade of gene activation results, leading to the proliferation and differentiation of cells that ultimately generates the organism. Assuming that molecular interactions and gene activation are predictable, the development process should be deterministic. [My emphasis.]
Kupiec disagrees. Rather, the inherent stochasticity at the level of molecules implies, to him, that cell differentiation (= cellular development) is also stochastic:
Kupiec argues that this picture is wrong. Gene activation is inherently stochastic, he says, and, therefore, cell differentiation must also be stochastic. Transcription factors attach with certain probabilities to many binding sites in gene promoters, implying that chance plays a dominant role in gene activation and expression. Similarly, cell signalling pathways, and thereby cell interactions, are stochastic, as proteins may bind promiscuously to many partners with various odds. Many interactions and pathways are possible. [My emphasis.]
What he is saying here is that since the processes that activate the expression of genes - which in turn make proteins for the different parts of the cell - have a random element (imagine a soup with different molecules bouncing around in a stochastic fashion, turning expression on and off when they come into contact with promoters and enhancers), then the outcome of that cell must also be random. And this is fundamentally wrong.

The big lesson from thermodynamics is that when you put a lot of molecules together, each will be controlled by random (i.e. non-deterministic) effects, but the properties of the gas or liquid are deterministic. There is a transition between stochasticity and determinism when random critters are joined to form a mob. This will surprise no one who has taken high school physics (and listened).

In fact, if we take Kupiecs argument seriously, then all higher level processes are also therefore stochastic. When you play baseball, whether you hit the ball or not must also a random process. Whether you get to have offspring or not must be a random process. This is clearly absurd to the nth degree.

Thus, Kupiec bases his whole book (at least according to Werner's review) on a totally wrong assertion. Had he been better informed from the beginning, so much time could have been saved. And if you think that a year or two to write a book with a provocative idea isn't that much, then consider that Kupiec has had this idea for at least 26 years:
Kupiec Jean-Jacques, A probabilist theory for cell differentiation, embryonic mortality and DNA C-value paradox, Specul Sci Technol, 1983;6:471-478.

A second minor quip with the review (I'm not sure if this is due to Werner or Kupiec) is a very common mischaracterization of what constitutes evolution:
Put simply, evolution requires two processes — variation and selection. An organism's offspring each varies slightly; natural selection picks out those that survive to generate more such organisms, again with their own subtle variations.
That wouldn't be how to put it simply. That would be how to put it wrong. Evolution requires variation and heritability. Selection is not required. Random effects will lead to evolution, too, by changing the frequencies of traits within the population. Selection is sort of an added bonus, really, which makes things a lot more interesting. But it isn't required.

13 comments:

  1. I'm pretty clueless about the science behind this, so let me sum up what I think is being said here. Kupiec is saying that the process by which DNA turns into a complete organism is greatly influenced by randomness. That, in effect, you could have two meaningfully different phenotypes gestate out of identical gene sequences.

    This does seem kind of crazy to me, because of one very obvious counter-argument: identical twins. I have cousins who are identical twins. They were physically identical at birth and as young children. When they began to differentiate themselves from one another psychologically, they made different choices which activated different potentials in the gene set they both share. So now, as women in their twenties, they look different, as much as they looked identical at age seven.

    The randomness comes in when one out of many, many sperm happens to be the one to fertilize the egg which happened to descend from the ovaries that month! That's plenty enough randomness to ensure every individual (or set of twins or triplets) has unique genetic potentials to explore.

    And if cell differentiation were really all that random, not only would identical twins not look identical--the vast majority of pregnancies would end in stillbirths.

    Tres gauche, Kupiec. And nice catch Bjørn.

    ReplyDelete
  2. And if cell differentiation were really all that random, not only would identical twins not look identical--the vast majority of pregnancies would end in stillbirths.

    Exactly. According to the review (which btw isn't too unfavorable, to my astonishment) the is some directing from the environment, but it isn't explained too well there. I'll still pass on reading the book, though.

    ReplyDelete
  3. Are you aware that stochastic effects dominate on some system sizes and disappear when the system becomes infinitely large? Stochastic effects at the cell level do not imply a macroscopic stochasticity to life in general. This review is ridiculous. You mount your attack on his basic misunderstanding of physics yet you have mischaracterized the most basic aspect of stochastic processes and non-deterministic physics.

    ReplyDelete
  4. Are you aware that stochastic effects dominate on some system sizes and disappear when the system becomes infinitely large?

    What do you think?

    Stochastic effects at the cell level do not imply a macroscopic stochasticity to life in general.

    Kupiec apparently argues that it does imply macroscopic stochasticity, though. I, too, ague it is not true.

    This review is ridiculous. You mount your attack on his basic misunderstanding of physics yet you have mischaracterized the most basic aspect of stochastic processes and non-deterministic physics.

    Who is misunderstanding what?! I do indeed attack Kupiec for his (career-long) misunderstanding of physics. Please do tell me how I how I have mischaracterized anything. Perhaps it is you who have misunderstood what I wrote.

    If you care to explain yourself, then maybe we could resolve this question.

    P.S. Who are you?

    ReplyDelete
  5. Dear Bjorn, you should read the book before attacking it!
    Arthur

    ReplyDelete
  6. Dear Bjorn, you should read the book before attacking it!
    Arthur


    Arthur, I am not attacking the book. Just the idea of stochastic cell differentiation.

    I generally do not think whole books need to be read before a cogent critique can be made against a single point. Rather, if the author can't explain that single point in less than a whole book, then he needs a course in writing.

    ReplyDelete
  7. What is so wild about the idea of stochastic cell differentiation. The latest understanding of stem cell state switching is that it is stochastic. http://web.mit.edu/biophysics/papers/NATURE2009C.pdf

    I'm confused. Are you saying you don't think any cell differentiation is stochastic?

    ReplyDelete
  8. I'm confused, are you saying that no cell differentiation is stochastic or that the idea that it can influence macroscopic phenotypes is wrong? The latest understanding of stem cell differentiation is that it is stochastic. http://web.mit.edu/biophysics/papers/NATURE2009C.pdf What is your argument against stochasticity in differentiation?

    ReplyDelete
  9. Bubbloy/Josh,

    What is your argument against stochasticity in differentiation?

    Here's what I wrote in the main post, in case you missed it:

    What he is saying here is that since the processes that activate the expression of genes - which in turn make proteins for the different parts of the cell - have a random element (imagine a soup with different molecules bouncing around in a stochastic fashion, turning expression on and off when they come into contact with promoters and enhancers), then the outcome of that cell must also be random. And this is fundamentally wrong.

    The big lesson from thermodynamics is that when you put a lot of molecules together, each will be controlled by random (i.e. non-deterministic) effects, but the properties of the gas or liquid are deterministic. There is a transition between stochasticity and determinism when random critters are joined to form a mob. This will surprise no one who has taken high school physics (and listened).

    In fact, if we take Kupiecs argument seriously, then all higher level processes are also therefore stochastic. When you play baseball, whether you hit the ball or not must also a random process. Whether you get to have offspring or not must be a random process. This is clearly absurd to the nth degree.


    Am I saying that "no cell differentiation is stochastic or that the idea that it can influence macroscopic phenotypes is wrong?"

    No, I am not saying that. I am saying that when Kupiec is saying that "Gene activation is inherently stochastic, he says, and, therefore, cell differentiation must also be stochastic", then he is clearly wrong (that all cell differentiation is stochastic), because we know very well that it isn't: if it was you couldn't make organs that worked.

    Could it be that some cell differentiation is stochastic? I am not saying that it could not. i am only saying that the argument presented above in the main post quoted from the review of Kupiec's book is false, and that cell differentiation is clearly not always stochastic, but mostly deterministic.

    ReplyDelete
  10. Sorry for the double post. I don't see why what you say implies most cell differentiation is deterministic. This may just be definitions, but a deterministic event should take place at a set time whereas a stochastic event has some characteristics time scale and distribution. An event that will definitely result in one outcome is still stochastic. If the event always happens after the same exact amount of time then I would call it deterministic. Can you say more about what Kupiec's scandalous claims are? So far I'm not seeing anything to blow my mind about his sanity.

    ReplyDelete
  11. For those who do not have access to Nature, here is the whole book review:

    Evolutionary embryos

    Eric Werner1
    BOOK REVIEWED-The Origin of Individuals

    by Jean-Jacques Kupiec

    World Scientific Publishing: 2009. 276 pp. $72, £54

    A central question in biology is how multicellular organisms develop from a single cell and how development is controlled. The standard view is that the process is deterministic, following directives governed by information located in the genome. Molecular biologist Jean-Jacques Kupiec contradicts this picture. In the fascinating The Origin of Individuals he argues that there is no plan, pre-pattern or program encoded in the genome. Instead, cell differentiation and development include a random element.
    Evolutionary embryos

    N. BROMHALL/PHOTOLIBRARY.COM

    In the standard view, development is controlled by the binding of protein transcription factors to promoters that activate genes in the DNA. These genes in turn generate proteins, including other transcription factors and signalling molecules that activate yet more genes. A cascade of gene activation results, leading to the proliferation and differentiation of cells that ultimately generates the organism. Assuming that molecular interactions and gene activation are predictable, the development process should be deterministic.

    Kupiec argues that this picture is wrong. Gene activation is inherently stochastic, he says, and, therefore, cell differentiation must also be stochastic. Transcription factors attach with certain probabilities to many binding sites in gene promoters, implying that chance plays a dominant role in gene activation and expression. Similarly, cell signalling pathways, and thereby cell interactions, are stochastic, as proteins may bind promiscuously to many partners with various odds. Many interactions and pathways are possible.

    As a result of this underlying unpredictability, Kupiec claims, stochastic cellular actions such as cell growth, cell differentiation and cell death must be constrained somehow to ensure that the correct sequence of development occurs. Otherwise, a fertilized egg could grow into any organism.

    The problem that ordered biological structures are rarer than the many possible random states led the physicist Erwin Schrödinger in his 1944 book What is Life? to contrast the science of life with physics: in statistical thermodynamics, macroscopic order is generated from disorder, whereas for life to develop, order must be generated from order. Schrödinger introduced the notion of a code script — analogous to a program — contained in the chromosomes, which acts as both plan and operative factor to prevent disorder by guiding the development process.

    Kupiec disagrees with the idea of programs. Because of the stochastic nature of protein interaction and gene expression, he says, there can be no Aristotelian form or program to give order to life and ward off entropic chaos and death.

    But without a code to follow, how can a particular organism develop from a single cell? Kupiec's radical solution is to apply Darwin's theory of evolution. Put simply, evolution requires two processes — variation and selection. An organism's offspring each varies slightly; natural selection picks out those that survive to generate more such organisms, again with their own subtle variations. In the development of an organism the stochastic nature of gene activation and protein interactions permits a vast array of possible developmental outcomes. Darwinian selection, Kupiec argues, constrains development so as to consistently form a particular organism. The local environment of the cell is the selecting agent, choosing which cells survive, differentiate, divide or die. He demonstrates his concept through computer simulations that generate simple patterns of two cell types. Each cell reacts with some probability to its local conditions to determine its next state.

    ReplyDelete
  12. In setting up local environment and metabolic interactions, cells must, however, use signalling protocols or programs that specify how the cells react — even if they involve probabilities. Yet Kupiec claims that because of stochastic protein interactions such programs cannot exist. But even if we allow simple reactive protocols to control any cell's reaction to its local environment, such strategies can produce only simple patterns. They cannot achieve the complex structures and functions generated in many multicellular organisms. Such reactive strategies can, at best, pass on information to determine the cell's next state.

    Thus, in Kupiec's proposal, the local environment must host the constraining information necessary to form an organism. But it is not clear how this information could be stored and conveyed. In his view, the environment functions like a complex, external 'homunculus', magically controlling embryonic development at every step. Kupiec also fails to explain why differentiated cells remain stable if gene activation is stochastic, or why cellular control strategies and protocols exist at all.

    Kupiec's version of a Darwinian-like cell-selection process needs to be robust and invariant. It must be more restrictive than typical Darwinian selection, which permits the formation of a diverse array of organisms and species to form. It must explain why a particular embryo forms, not just any embryo. It must account for the similarity of identical twins; the precision with which the left side mirrors the right in bilaterally symmetrical organisms; and why a mouse differs from a horse or a potato. A further issue is that even if there is local molecular randomness, it need not be passed on to the cell or to the developmental control architecture of the organism. Organisms consistently pass through the same stages during development, irrespective of minor variations in their local and maternal environments.

    By treating the genome only as a generator of proteins, Kupiec adopts an implicitly reductionist view of development. But organisms of many species have virtually identical protein structures, yet their control architecture is vastly different, just as a house and a skyscraper can be made of the same parts. Every complex structure needs specific control information to develop, and the only reasonable source of that information is the genome, not some blind local evolutionary selection process. The genome and cell cooperate by means of an epigenetic interpretation system by which control information in the genome is interpreted and executed by the cell. Thus the genome encodes more than protein building-blocks — it contains a hidden control code. Such a feature could explain the vast non-coding regions in the genome; Kupiec prefers to think of these regions as mere space fillers determining gene-activation probabilities.

    Kupiec's model also fails to account for global and temporal relationships. Local information is not powerful enough to generate global relationships in an organism — all the more so if it is probabilistic. Because the growth process of an embryo is ordered in time, directives from the genome must be linked to form control networks. The architecture of an organism is complex both spatially and temporally.

    Kupiec is a very successful writer, deservedly so. I enthusiastically recommend this courageous book with its iconoclastic viewpoint. The Origin of Individuals is a pleasure to read, presenting complex ideas clearly and effectively. Whether one agrees with him or not, Kupiec's is an inspiring work, a thought-provoking rollercoaster ride through the history of ideas about the origins of ontogeny.

    1. Eric Werner is in the Department of Physiology, Anatomy and Genetics, and the Computing Laboratory, University of Oxford, South Parks Road, Oxford OX1 3QX, UK.
    Email: eric.werner@dpag.ox.ac.uk

    ReplyDelete
  13. An event that will definitely result in one outcome is still stochastic.

    Okay, we may just be playing with words here, while we actually agree on the biology. I don't agree with your understanding of stochastic, and it seems to me that both Kupiec and the reviewer of his book, Werner, thinks of stochasticity is something to do with randomness, like I do.

    Gene activation is inherently stochastic, he says, and, therefore, cell differentiation must also be stochastic.

    Kupiec apparently thinks that the stochasticity that we can agree exist at the molecular level translate into stochasticity at the level of cells (this is btw what I think is insane). I surely do not. Do you?

    ReplyDelete

Markup Key:
- <b>bold</b> = bold
- <i>italic</i> = italic
- <a href="http://www.fieldofscience.com/">FoS</a> = FoS