Review of Axe’s work by Martin Poenie

This is my response to Doug Axe posted at

Doug Axe and Steve Meyer want us to believe that protein evolution is impossible and that his experiments somehow inform that belief. His experiments, he claims, take the approach of a chemist attempting to make the minimal amount of substitutions needed to transform specificity of one enzyme into that of another similarly-structured enzyme. In my view, the experiments have little to say about evolution and are singularly uninformative. His experiments have also been widely criticized.

As Axe notes, there are 250 out of 381 amino acids that differ between BioF and Kbl such that the two sequences are 34% identical over 381 aligned positions. From an evolutionary perspective, this is a long distance. It is not surprising that his efforts to swap functionality by swapping a few residues failed. Approaches similar to what Axe have done have failed repeatedly. As noted by Harms and Thornton, “The reason studies of this type fall short is that they ignore history” (Harms and Thornton, 2010, Current opinion in structural biology 20: 360-366). The young earth creationist Todd Wood summarized the issue succinctly:

“Instead of ancestral reconstruction, Gauger and Axe focused directly on converting an existing enzyme into another existing enzyme. That left me scratching my head, since no evolutionary biologist would propose that an extant enzyme evolved directly into another extant enzyme. So they’re testing a model that no one would take seriously? Hmmm…” (”)

I quote Wood as a good example of someone whom you would think would be happy to embrace Axe’s conclusions if only they were valid scientifically. But they are not.

It is interesting to note that efforts to change enzyme/binding specificity using ancestral reconstruction actually do work. For example:

Bridgham, J. T., E. A. Ortlund, et al., 2009, “An epistatic ratchet constrains the direction of glucocorticoid receptor evolution.”, Nature 461: 515-519.

Smith, S. D., S. Wang, et al., 2013, “Functional evolution of an anthocyanin pathway enzyme during a flower color transition.” Molecular biology and evolution, 30: 602-612.

I find the Gauger and Axe study uninformative (beyond that which we already knew) because they have no idea what their substitutions actually did to their protein. For example, these mutations could have simply led to an unfolded protein. They reject this idea based on a generalization that “it appears that about 10% or more of the residues in natural proteins need to be changed before the cumulative structural disruption can be expected to cause complete loss of function” (Axe, D. D., 2000, Journal of molecular biology 301: 585-595). Yet in Axe’s own words from that paper, “In both experiments, complete loss of activity demonstrates the importance of sequence context in determining whether substitutions are functionally acceptable”. The fact is, they do not know what they did to their protein other than to change residues and show (unsurprisingly) that it did not work. This type of experimental design would be cause for rejection from any mainstream scientific journal and perhaps an explanation of why he published it in his own journal.

In the wider context of protein evolution, I find this fixation on identifying random point mutations as “Darwinism” and then attempting to disprove Darwinism by showing the odds of this or that very unpersuasive. For one thing, it leaves out a major mechanism for change involving recombination. Recombination can do all the things that Axe thinks are impossible. For example, recombination can generate multiple substitutions in one step and still generate one or two functional proteins as the outcome. Quoting Watson et al. (2011, Evolution, 65:523), “Whereas asexuals must move against selection to escape local optima, sexuals reach higher fitness peaks reliably because they create specific genetic variants that “skip over” fitness valleys, moving from peak to peak in the fitness landscape. This occurs because recombination can supply combinations of mutations in functional composites or ‘modules’, that may include individually deleterious mutations. Thus when a beneficial module is substituted for another less-fit module by sexual recombination it provides a genetic variant that would require either several specific simultaneous mutations in an asexual population or a sequence of individual mutations some of which would be selected against.” Furthermore, as noted by Romero and Arnold, “Laboratory experiments clearly demonstrate the benefits of recombining homologous proteins: intragenic recombination generates new proteins that are functionally diverse while still having a high probability of folding properly and functioning” (2012, PLoS computational biology, 8:e1002713). Indeed, due to its conservative nature, recombination can explore a “functional ridge” between two proteins (Drummond et al., 2005, PNAS USA 102: 5380-5385).

Finally, in regard to ORFans, in my view, Axe’s argument simply backfires. ORFans are found in all genomes; prokaryotes, eukaryotes, bacteriophages and animal viruses. Remarkably, the recently discovered megavirus “Pandoravirus” has 2500 genes, almost all of which are ORFans. The fact that the number of ORFans tends to be constant from one type of organism to the next in prokaryotes and eukaryotes indicates that they not uniquely associated with the Cambrian explosion and that they are likely formed by mundane genetic mechanisms that operate in all organisms. In one particularly informative example, Toll-Riera et al. (2009. Mol. Biol. Evol. 26, 603–612) identified 270 primate-specific ORFans. Of these, 70% contained a transposable element. In other cases, where a particular gene appeared to be ORFan, the same organism had a paralogue that did show homology with other organisms suggesting that the ORFan in question underwent rapid divergence. Contrary to Axe and Meyer, the fact that ORFans could represent new genes generated by genetic mechanisms such as transposition really throws a monkey wrench into their arguments.


13 responses to “Review of Axe’s work by Martin Poenie”

  1. David Snoke

    Jonathan McLatchie has posted a response to this review at

  2. Martin Poenie

    Yes Jonathan responded simply dismissing what he does not like. The question is whether the dismissal carries any weight or is uninformed. For starters, Jonathan writes, “Neither Axe nor Meyer (or, for that matter, anyone else I know of) has argued that ORFan genes are “uniquely associated with the Cambrian explosion”.
    My response was aimed at both Axe and Meyer. So, while they did not directly state that ORFans were uniquely associated with the Cambrian explosion, all you need is to add 2 and 2 to get there. Here is what Axe wrote:
    “Now, since each gene carries the sequence instructions for making a protein, it seems likely that orphan genes tend to encode orphan proteins — proteins that are substantially distinct from any found in other kinds of organisms. And if so, it also seems likely that many of these orphan proteins have distinct structures, or folds, as they are known. Again, we could criticize this claim on the grounds that no one presently knows how to design new protein folds with any proficiency, but this is pointless because reverse engineering has shown that the inference is correct. Proteins with no detectable similarity to any protein of known structure have been found to have unique fold structures in about half of the cases examined. Considering that orphan genes typically account for 10% to 30% of the genes in each sequenced genome, and that multicellular animals have about ten thousand or more genes, this means we can expect to find many dedicated protein folds in each specific kind of animal, right down to the level of species. So while the passage of half a billion years prevents us from actually examining the proteins that were used within the cells that made up the animals that appeared in the Cambrian explosion, the diversity and number of these animal forms leads us to believe that there must have been a corresponding explosion of protein forms. This certainly follows from the facts as we now see them, so Poenie’s assertion is misinformed.”
    So in this quote Axe says that many of these orphan proteins have distinct structures or folds. The diversity of forms leads us to believe that there must have been a corresponding explosion of protein forms. He says that because orphan genes account for 10-30% of the genes in each sequenced genome we can expect to find many dedicated (I infer species specific) dedicated protein folds right down to the species level. As I read this, ORFans are accounting for the explosion of new folds at the Cambrian explosion and represent an explanation for the novelties in different types of animals in the Cambrian explosion. However, the single cell amoeba Dictyostelium has 12,500 predicted proteins (ORFs). Another amoeba “Acanthamoeba catellanii” has 15,455 genes. So my question here is what do all the ORFans mean to these organisms? Based on what Axe wrote, these amoeba should have as many ORFans as in the “explosion of protein forms” in the Cambrian explosion. Now Meyer stated that the minimum amount of new information was a new protein fold. He also thinks there must be many new folds associated with the Cambrian explosion.
    Now getting to the next aspect of Jonathan’s dismissal, I cited the Toll-Riera paper to bring up the point that of 270 primate-specific ORFans 70% contained a transposable element. So, let’s first examine what it means to be an ORF (open reading frame). There are 64 possible codons (groups of 3 bases most of which spell out an amino acid. However there are 2 or 3 stop codons so the odds of randomly generating a stop codon are 1/32 or 1/21 respectively. For each segment of DNA, there are 6 reading frames (3 going in one direction on one strand and 3 going in the reverse direction on the opposite strand). When molecular biologist come across a reading frame where there are hundreds to thousands of codons with no stop codon, they infer that this segment is a gene. However, the Toll-Riera paper also showed that the ORFans they identified were transcribed into RNA. So, there must have been promoters associated with these ORFans and, as we will see, most likely other regulatory elements. Now, one of the problems with ORFans is that they have not received a lot of attention until lately and the identification of actual proteins is rather sparse. However, in the case of human ORFans containing transposons, there is some information.

    As noted by Toll-Riera et al., (2009, Biochemical Society Transactions,37:778) “A surprisingly high number of primate orphan genes, 142, were found to contain TE sequences. The vast majority of these genes were located in primate-specific genomic regions. One example is the F379 retina specific protein family. This family is encoded by four genes that are located in subtelomeric DNA regions. These genes are formed by three coding exons, and the second exon is completely covered by two SINEs (short interspersed transposable elements): an Alu (Flam_C subtype) and an MIR (mammalian-wide interspersed repeat). Another example is Tumor Necrosis Factor gene (p75).

    So with this in mind the reader can evaluate Jonathan’s dismissal.

    “It has yet to be shown that haphazard mechanisms such as transposition of mobile elements are likely to construct functional proteins.”

  3. Martin Poenie

    In part II of my response to Jonathan, I address again the protein experiments of Gauger and Axe. I maintain that these experiments are uninformative and in fact, naïve. Jonathan wants to point to pictures portraying the two proteins in questions so as to say, “look how similar they are!”. This somehow is supposed to justify the proposal to change a few residues to convert one enzyme function to another. Furthermore, with the failure of that approach, they feel entitled to comment freely about the possibility of protein evolution. Nonsense!
    Despite the visual similarity, the two proteins differ by hundreds of amino acids. In fact, from that perspective, they are hardly similar at all. Axe argued that one needs to change about 10 amino acids to get the structure to fall apart – so what about changing hundreds of amino acids? What it means is that many of the interacting amino acids that hold the protein together are quite different between the two proteins. The two proteins look similar but are held together completely differently. Proteins chemists refer to interacting amino acids as epistatic interactions.
    To consider the implication of this drastic difference between the two proteins we have to consider some fundamentals about enzyme catalysis. The foundational principle of enzyme catalysis is based on binding. Enzymes catalyze a reaction by binding to the transition state better than to reactants or products. While an enzyme binds to reactants, it forms stronger bonds with a structure that is in between the reactant and the product i.e. the transition state. So what an enzyme does then is force the substrate into a strained structure. This is driven by the many weak bonds that the enzyme makes with the substrate and even stronger bonds as the substrate is contorted into the transition state. So this is what matters when comparing the two proteins. Showing similarity in skeletal structure might be relevant but it is the interacting amino acids that form the bonds with the substrate that are critical, and that is precisely what may differ when so many amino acids differ in the two proteins.
    In fact, we know nothing from the Gauger and Axe studies as to whether their mutants held together or whether it even bound the substrate at all. The one conclusion they are entitled to make is that all those different amino acids changes between the two proteins really do matter. That is it – period. They are not entitled to make inference about protein evolution because their experiments are completely irrelevant to that. In my view, it is actually naïve to even think about the experiments this way.

  4. David Snoke
    David Snoke

    I’m not going to dismiss the extreme structural similarity of those proteins so easily, Marty. From a probability standpoint, if two designers submitted these structures, they would fail a plagiarism test. There are so many points of similarity, either similar function has forced them to be that way, or you have an impossibly low odds scenario of two different functions having the same structure, with hundreds of structural variables, by chance.

    The other point is that these two structures were chosen because that is as close as one can get. Axe and Gauger would have chosen two that were more similar if there were any more similar out there. This seems to me to point of the issue– even small changes of function require hundreds of mutations, not 2-3. So everything we have to compare is “not evolutionarily near” by your criterion. Either something has just a few mutations and has the same function, or it has a different function and has hundreds to thousands of differences.

    The logic of Axe and Gauger was let’s choose two proteins that look the most like they could have evolved from a common source. Remember, evolution is supposed to be easy. And evolution somewhere along the way has to evolve new functions. It seems to me a very legitimate approach.

  5. Martin Poenie

    Visual closeness is not what will matter. Like I tried to explain, what matters is the right particular amino acid in the right position to form bonds with the substrate. Why didn’t Axe test to see if the opposite substrate even bound at all? Lots of combinations of amino acids will make alpha helix or beta sheets. That is not going to get you anywhere catalytically. Furthermore, at present people who study this estimate that there are a limited number of different folds between 1000 – 10000 with current estimates around 2000. So the same type of folds can be used in lots of different proteins. In any case, I stand by what I said. Just having a similarly folded enzyme is not going to make it likely that the two will be close catalytically. And in fact, it did not work which is exactly what you would expect.

  6. Martin Poenie

    Perhaps it might make sense in these terms. The variable region of antibody molecules have the same structural framework too. But you can make 10^24 different antibodies with different binding specificities.

  7. David Snoke
    David Snoke

    I have been taught that physical homology indicates either common origin or common cause, i.e. convergent evolution. There is a huge physical homology there that begs explanation.

    I would not be surprised if there are a finite number of types of folds that can be mixed and matched, just as there are a finite number of words in the English language. But somewhere along the way, to make evolution go forward, there must be a new fold that happens or is added, or at least a new binding center that is added. It doesn’t seem a stretch to me to say let’s start with something and see if we can get it to add something new, the way that evolution is supposed to.

  8. David Snoke
    David Snoke

    Ann Gauger responds here:

    This debate is oddly taking place across multiple different websites!

  9. Martin Poenie

    Yes – she ignores the bona fide cases that I brought up as if they weren’t there.

  10. Martin Poenie

    Here is a fully characterized new gene. It is only found in one species of yeast thus it might be considered an ORFAN. It originates from the non-coding strand of a gene that is express in many yeast. They showed that it is expressed as a protein which accumulates in the nucleus and it regulates the yeast mating pathway.

    Li, D., Y. Dong, et al. (2010). “A de novo originated gene depresses budding yeast mating pathway and is repressed by the protein encoded by its antisense strand.” Cell research 20(4): 408-420.

  11. David Snoke
    David Snoke

    A couple of times on the comments you have said that people have “ignored” points of yours. I think that is unfair. No one is under obligation to respond to every point that someone else makes. If someone doesn’t respond to a point, there are multiple reasons why they might not: they concede the point, they don’t think the point is relevant or high priority, they think it is obvious to the reader that it is wrong, etc.

  12. Martin Poenie

    Well, ignore in the sense of implying there is no data when there really is.

  13. David Snoke
    David Snoke

    Reply by Doug Axe is here

Leave a Reply