Wednesday, September 08, 2010

Mutations and Complex Adaptations

Michael Lynch is one of those rare scientists who not only think outside the box but successfully stimulate others to do so. I read all of his papers and I'm always very impressed, even though I don't always agree with everything he says.

I recently attended the 18th Annual Meeting of the Society for Molecular Biology and Evolution in Lyons, France, where I met Michael Lynch for the first time. He gave a talk on Evolution of Mutation Rates, a topic many of us treat with a large degree of skepticism—until we're confronted by Michael Lynch. He makes a convincing case for variable rates of mutation in different species and he challenged us (me) to defend the idea that there was a linkage between DNA replication rates and mutation rates. That's a linkage I've always assumed would constrain mutation rates to a narrow range. Now I'm not so sure.

I've been putting off posts about the many exciting things I heard in Lyon because I'm busy with the 5th edition of my textbook but SteveF provoked me into saying something about Michael Lynch by posting a comment on my blog [Larry, you might find this shiny new paper by Michael Lynch in PNAS interesting]. Damn you, SteveF, and thanks.

We had been discussing how the IDiots view mutation and I mentioned that Michael Behe was mostly, but not entirely, correct when he said that if two mutations are required for a complex adaptation then it is very unlikely to happen [Bated Breath].

In a paper just published in PNAS, Michael Lynch explains why it's "not entirely correct" (Lynch, 2010).
The development of theory in this area is rendered difficult by the multidimensional nature of the problem. One strategy has been to ignore all deleterious mutations and to assume that selection is strong enough and mutation weak enough relative to the power of random genetic drift and recombination that evolution always proceeds by the sequential fixation of single mutations (e.g., refs. 6–11). Such an approach provides a useful entree into the evolutionary dynamics of rare adaptive mutations with large effects. Under these conditions, the expectations are clear—with larger numbers of mutational targets and a reduced power of random genetic drift, the rate of adaptation will increase with population size, although more slowly than expected under the assumption of sequential fixation (12, 13). The motivation for these models, which are specifically focused on total organismal fitness, derives from case studies of adaptations with apparently simple genetic bases, e.g., some aspects of insecticide resistance (14), skin pigmentation (15), and skeletal morphology in vertebrates (16).

Nevertheless, a broad subset of adaptations cannot be accommodated by the sequential model, most notably those in which multiple mutations must be acquired to confer a benefit. Such traits, here referred to as complex adaptations, include the origin of new protein functions involving multiresidue interactions, the emergence of multimeric enzymes, the assembly of molecular machines, the colonization and refinement of introns, and the establishment of interactions between transcription factors and their binding sites, etc. The routes by which such evolutionary novelties can be procured include sojourns through one or more deleterious intermediate states. Because such intermediate haplotypes are expected to be kept at low frequencies by selection, evolutionary progress would be impeded in large populations were sequential fixation the only path to adaptation. However, in all but very small populations, complex adaptations appear to be achieved by the fortuitous appearance of combinations of mutations within single individuals before fixation of any intermediate steps at the population level (e.g., refs. 17–26).
Read the paper. You'll find an interesting discussion of recombination—a discussion that does not assume most of the standard myths about recombination. Lynch points out that when it comes to fixing two independent mutation the effect of recombination is just as likely to break up linkage as enhance it. Recombination cannot make much of a contribution to the fixation of two mutations that are required for a complex adaptation unless the mutations are closely linked (e.g. same gene).

However, there are some circumstances where large population sizes can overcome the problem of fixing multiple mutations even if there's a negative correlation between mutation rate and population size. This is the "scaling" parameter mentioned in the title of Lynch's paper.

This is not unlike what Behe's says in The Edge of Evolution where he points out that in malaria parasites (e.g. Plasmodium falciparum) the probability of a double mutation is significant because there are trillions of organisms. In large mammals, however, the probability is much lower because the population size in much smaller.
Changing multiple amino acids of a protein at the same time requires a population size of an enormous number of organisms. In the case of the malaria parasite, these numbers are available. In the case of larger creatures, they aren't.
Behe concludes that this is the "edge" of evolution. Since these kinds of mutations are required for complex adaptations, it follows that evolution can't account for complex adaptations. You'll have to read Behe's book to find out who can design such complex adaptations.

So far, this is pretty much standard orthodoxy. Given that multiple, independent, mutations might be required simultaneously it's very unlikely that evolution will ever see them in some species. It's one of the reasons why Behe's book is so unexciting. There are other ways to account for the adaptive value of multiple mutations, including the fact that many of the individual mutations may be slightly deletersious but, nevertheless, fixed by random genetic drift.

What Lynch's paper shows is that the standard orthodoxy might be wrong! His models suggest that fixation of multiple mutations in small population may be well within the range of probability required for evolution of complex adaptations.
In summary, the preceding results suggest that some general scaling properties may exist for the rapidity with which various types of adaptations can be assimilated in different populationgenetic contexts. In particular, prokaryotes appear to be much more efficient than eukaryotes at promoting simple to moderately complex molecular adaptations, and substantially so for those involving joint changes at different genetic loci. In contrast, adaptations requiring three or more novel mutations may arise more frequently in small populations, regardless of the level of recombination between selected sites. In the absence of comprehensive information on the molecular basis of adaptation in multiple lineages (i.e., the typical number of sites involved and their degree of epistatic interactions), these general predictions are currently difficult to test. Nevertheless, the ideas presented herein are likely to bear significantly on a number of ongoing controversies regarding the nature of adaptation, including the barriers imposed by adaptive valleys in a fitness landscape (22, 40), the role of compensatory mutation in evolution (41), and the relative rates of incorporation of adaptive and nonadaptive mutations in various lineages (42–44).(my emphasis-LAM)

Lynch, M. (2010) Scaling expectations for the time to establishment of complex adaptations. Proc. Natl. Acad. Sci. (USA) publishe online, Sept. 7, 2010 [doi: 10.1073/pnas.1010836107]


  1. Great commentary Larry. I haven't yet read the Lynch paper, I just stuck it at the top of my list of things to read which, is of course never ending and growing larger by the day.

    Like you I'm a big fan of Lynch's work, and read many of his papers. Like you I'm always impressed, even when I don't agree with everything he says.

    One thing I wonder about in this case, is how many of these complex adaptations are necessarily adaptive in the first place, and not predominantly the product of neutral evolution as in Constructive Neutral Evolution. When all we have is the end result (new protein function, new complex, etc) it becomes difficult to differentiate the two, especially if you look at recent work from Joe Thornton's lab on what ti looks like when you try and reverse the evolutionary path taken. Granted in that case evolution was proceeding by selection, but there is no reason to think that the same isn't true under neutral conditions.

    That isn't to deny that such complex adaptations don't happen, I'm sure they do and this paper seems promising in saying that the probability of occurrence in small populations may be much more reasonable than was thought. But I am willing to bet that initially neutral fixations followed by purifying selection may be almost impossible in many cases to distinguish from complex adaptations.

  2. "That's a linkage I've always assumed would constrain mutation rates to a narrow range. Now I'm not so sure"

    Congratulations. You are a bit more pluralist now!

  3. Thanks for your thoughts Larry, very interesting. I agree about Michael Lynch, he's very thought provoking. I should also point out that for those who are interested in exploring his research, he puts his papers up online for anyone to read:

    (the most recent not up yet)

    He had another recent paper that's worth a read, "The Rate of Establishment of Complex Adaptations".

    Behe even gets a mention in this one.

  4. This continues a trend in some other work I've seen on multiple mutations, which is:

    (1) Interesting modeling of what would/can happen in cases where adaptations would require multiple mutations, but

    (2) Not a heck of a lot of evidence that the actual adaptations that we actually see around us in the world would have actually required multiple mutations to produce.

    E.g. the observation:

    1. "this protein-protein binding site uses 5 key amino acids"

    ...Does Not Equal the statement...

    2. "the evolution of this protein-protein binding site required the lucky combination of 5 different point mutations before selection could favor it"

    The idea that (1) implies (2) is just Behe's bogus IC argument ("The flagellum currently requires 20 proteins, therefore they all had to come into existence at once!") taken down to the point-mutation level.

    It makes no more sense than saying "Humans require heads, hearts, stomachs, and kidneys to live; therefore, humans couldn't have evolved gradually. Furthermore, we can test the plausibility of the evolutionary scenario by cutting off the heads of humans and seeing how well these reconstructed ancestors function."

  5. @NickM: Very true. We know that especially in the case of gene duplication followed by divergence (what I work on), a gradual shift in function is plausible and experimentally verified in many cases.

    The only time I would expect multiple mutations to be absolutely required are in the case of traversal of very deep fitness valleys and given relatively small evolutionary timescales such that a more circuitous route with multiple substitutions at a given relevant site being unlikely. Meaning it is likely very difficult to prove that such a situation was necessary.

    That said I think that it likely happens more frequently than is necessary.

  6. "It's one of the reasons why Behe's book is so unexciting."

    Right. The pattern is always the same. A new idea is put out and it is deemed ridiculous. Then after some length of time it is pretended that it is "unexciting", as if everyone knew that already.
    I am not a creationist. But this pattern is just embarrassing.

  7. Right. The pattern is always the same. A new idea is put out and it is deemed ridiculous. Then after some length of time it is pretended that it is "unexciting", as if everyone knew that already.
    I am not a creationist. But this pattern is just embarrassing.

    Actually, Larry has been saying this right from the start (as well as pointing out that some critiques of EoE were misguided).

  8. You'll find an interesting discussion of recombination—a discussion that does not assume most of the standard myths about recombination. Lynch points out that when it comes to fixing two independent mutation the effect of recombination is just as likely to break up linkage as enhance it.

    Isn't that one of the "standard myths"? Once recombination has brought together two elements, they are not invariably torn asunder in the pop-geneticist's 'well-stirred pot', even without physical linkage. Having happened once suggests that the two elements existed locally in space and time, which (unless it was just 'one of those things') gives the possibility that local concentrations may favour its re-occurrence. Each time it happens, 50% of offspring will retain the linkage, and the rest stand a fair chance of re-establishing it, given the 'local' argument. Recombination breaks as well as forges links, but the heterogeneity of populations alone can prevent that from being a balanced equilibrium, even in neutral situations and separate chromosomes.

  9. "The light shines in the darkness, but the darkness has not understood it."



    hawking is WRONG

    science cannot explain NOTHING!



    add comment moderation to your BS or more people will die with you...


    plush safe he think

    but with recent revelations about James Randi, I think he likes DICKS!





    THE B**BQUAKE - 911

    you are going to pay the price for this….


    there is a lot of sh*t to flush!


    Is America burning yet?

    Maybe we need some more...

    we use the DIVINE against the ESTABLISHMENT... you?

    we do better DEMOLITIONS than you, savage...


    let them know if the MDC continues more people will die...




    Actually it is a ROYAL FLUSH!!!

    Let me show you how ATHEISTS were partially responsible for 911


    You don’t even have SCIENCE on your side…

    You’re a perfect example of when PHILOSOPHY becomes an ENEMY OF LIFE...

    not quite samantha with her *supernatural spit*, eh?

    this isn't one of your little WORD GAMES...

    blasphemy is a DEATH SENTENCE

    you people actually BELIEVE the BS you preach!

    GOD 1 - atheists 0



    Repent and turn to God or be destroyed...


    my interpretation of the STATUE FIRE... it symbolizes the SPIRITUAL DEATH of atheism...



    The B**BQUAKE - 911

    Let me show you the FATE OF TRAITORS...

    THE BOOBQUAKE - 911!

  10. LYNCH doesn't understand biochemistry or biophysics at all.

    He fails to grasp the fact that an accumulation of suboptimal changes will only serve to destablise the translated peptide product.

    A head is useless without a body. If a new motif is produced by chance, it needs to be part of a stable protein that folds properly.

  11. One thing I have not seen emphasized (IANAB) is the effect of increasing population size on rare, possibly deleterious genes. I don't think I'm just talking about founder effect on genetic drift. Just as a rising tide lifts all boats, exponential growth in a new environment or growth in, say seasonal environments will lead to rare suboptimal mutations occasionally being a significant part of the total genetic map.
    I'm thinking of such things as rabbits and cane toads in Australia, lionfish in the Carribbean, malaria parasites infecting a person, or flies that have overwintered.
    Not that all the variation will be the subject of future beneficial mutation, but those scenarios are likely to have both constrained diversity and increased exposure of rare genotypes.


  12. @Reza Anyone who does work on molecular evolution, particularly people who do work on the evolution of protein domains and folds, could tell you that Lynch's statements are a hell of a lot more accurate than yours.

    It is obvious from observed sequences and from experimental studies that proteins can accomodate an incredible amount of sequence diversity while retaining their function and structure.

    One of my biggest pet peeves perpetuated by protein biochemists and enzymologists is categorical statements of mutations that are totally destabilizing or render a particular protein non-functional without acknowledging that caveat that it is only destabilizing or deleterious in the context of THAT particular sequence and organism. Many of those same mutations can be found in perfectly functional proteins in other organisms.

  13. DG:

    Protein sequences are flexible and versatile- up to a point. The problem is that Lynch assumes that random changes accumulated by drift, not selection, can somehow produce a functioning protein. This is fantasy.

    As Lynch, admits, there is a tendency for compensatory mutations to make up for any damage - restoring and not changing protein function.

    In short, Lynch's paper focuses on the thereotical pop genetics and not on some basic biochemical realities.

  14. @Reza: Shifts in function can arise through either drift or selection, or a combination of the two. There are obvious cases of clearly orthologous proteins that have acquired or switched functions over evolutionary time, with the main examples being seen when the orthologs are more distantly related. Bacterial or Archeal sequences compared to their Eukaryotic counterparts for example.

    And of course shifts in function are likely to happen much more frequently after gene duplication events. There is no compelling reason why all of those cases must be the result of positive selection. The reality is that they are likely the result of both neutral and adaptive changes followed by purifying selection.

  15. @DG

    Gene duplication has turned out to be a major disappointment. Gene duplicates tend not to acquire new functions other than those closely related to the ancestral one (eg RNASE1B). Lynch and Qian have shown that "subfunctionalization" - the differential partionining of function and expression among paralogs - is far more realistic.

    Also, gene duplicates do come under purifying selection because they can serve as useful backups or provide double dosage. They do not diverge along brave new evolutionary trajectories.

    Nonetheless, selection is relaxed among duplicates but this only leads to partial degeneration and not innovation.

  16. @Reza No one would deny that subfunctionalization isn't more common, it definitely is. However, neofunctionalization does happen. Even shifts to "closely" related functions are far from disappointing.

    I'm not stating that duplication alone is the recipe for more radical shifts in function. It is obvious that there are a variety of more complex processes also occurring including things like domain swapping, domain re-arrangements, etc.

  17. @DG:

    Sure neofunctionalization can be said to happen. But most of the cases observed do not indicate any real biological innovation.

    For example, Sdic in D melanogaster is a new gene that has evolved from a duplicate of Cdic by way of a gene fusion event that deleted the latter's N-terminus - this contained domains specific to a cytoplasmic dynein - i.e. the ability to interact with dynactin.

    Sdic, therefore "evolved" into an axomemal flagellar dynein almost by default - it couldn't serve as a cytoplasmic dynein.

    So, even in the rare cases of neofunctionalization, what we see is a loss of function as much as any gain of function.

    Btw, exon shuffling and domain swapping does produce hybrids proteins but not those with new motifs and the like.

  18. This comment has been removed by the author.

  19. This comment has been removed by the author.

  20. @Reza The Sdic,Cdic, AnnX example is an interesting case. One could describe it as "neofunctionalization by default through loss of function" if one wished. But the flip side is that it has also gained, relative to Cdic, a new expression pattern. In multicellular organisms, this is a key area for evolutionary innovation. Neofunctionalization is likely to almost always result in loss of ancestral function during the gain of any new function, No one would dispute that because it is a trivial point. But not all cases, perhaps most, are not merely going to be "gains by default" as you illustrate with the Sdic example.

    Whether domain shuffling and re-arrangement produces novel sequence morifs is irrelevant to my point. Can domain re-arrangement and swapping produce proteins with novel functions? Yes. That is clearly the case.

    Getting back to your original comment about the evolution of new motifs I would still say that you are not giving enough credit to the often robust nature of protein folding and function. Yes, it is true that there are some small mutations that can be completely destabilizing, at the same time there are large mutations (particularly large insertions) that do not interfere with a proteins folding and function. I think it is unfair to say that Lynch doesn't understand biochemistry or biophysics. His work hardly contradicts observational evidence of protein sequence, structure, and function when examined in the light of evolutionary diversity.

  21. Apologies for the duplicated posts. I kept getting a comment too large error and some of the attempts apparently did actually go through.

  22. @Reza

    Sure neofunctionalization can be said to happen. But most of the cases observed do not indicate any real biological innovation.

    If the resulting paralogous protein interacts with a new set of protein binding partners, carries out the same type of enzymatic reaction but on a different substrate, is differentially regulated, has altered expression patterns, or anything similar this is certainly biological innovation any way you want to cut it. The vast majority of neofunctionalization events, especially recent ones (which are the easiest to characterize) are going to be towards related functions yes. To characterize these as non innovative though is ridiculous.

  23. I'd recommend not replying to "reza". He's a fairly notorious troll (usually goes by the name of atehistoclast) on various discussion boards, who has a habit of pestering evolutionary biologists, submitting papers to journals (none of which have been accepted yet) and denying the holocaust. There's some debate as to whether he's actually serious, or if it's some very elaborate joke. Either way it's most definitely not worth the time responding.

  24. @SteveF Thanks for the heads up, the name seemed familiar from a few other comment threads on here and the last couple of replies I was getting the head against a brick wall feeling

  25. You see, both Lynch and Carrol are focused on the primary ,and not the tertiary structure, of proteins.

    A head is useless without a body.

    I think Dan Hartl sums it up best on his website:

    He says that "We have put forward a new model of protein evolution based on small, nearly neutral, compensatory amino acid replacements affecting protein folding, stability, and aggregation. Proteins are finicky molecules, marginally stable and prone to aggregation, yet they must function in a crowded cellular environment."

    And that is why their evolution is constrained.

  26. @Reza I Probably shouldn't bother but I am going to respond to this anyway for the benefit of others.

    Nothing Hartl says on his website or in his papers contradicts what I have been saying in this thread. It should be noted that the model he is describing largely pertains to proteins undergoing evolution mediated by point mutations.

    Secondly, no one considers protein evolution to be unconstrained. That would be an incredibly naive position. While I am sure Lynch and many others are not experts on protein biochemistry and biophysics they do have at least some appreciation for the fact that proteins are three dimensional models, not merely sequences.

    Lastly, there are a great many of us within the broad field of molecular evolution who work on these issues of protein evolution and biological innovation who explicitly consider protein function and structure into account. Yes, proteins can be finicky, they can also be robust. These statements do not have to be mutually exclusive when you consider context.

  27. > [Lynch's] models suggest that fixation of multiple mutations in small population may be well within the range of probability required for evolution of complex adaptations.

    But Behe is not modeling evolution, he starts by observing what evolution actually did--and then he extrapolates. If Michael Lynch is modeling, and then extrapolating, then his results should match Behe's!

    Given that his results should match what nature actually does.

  28. @DG

    Yes, proteins can be robust

    But it depends on the context, as you say, and what kind of site mutations affect them.

    The epistasis of multiresidue interactions, which Lynch alludes to, can force compensatory changes to rebalance things somewhat.

    What concerns me about Lynchie, is that he thinks that if you take away selection from the process of gene/protein evolution, you can allow for a greater chance of finding a novel adaptation.

    Random mutations and random drift = random noise, not new information. I'm sorry, but that is the case.