Are Developmental Mistakes Essential to Evolution?

BQO - MaselFlickr Charles Clegg (CC)

In biology, as in life, mistakes happen — and most of them generally make things worse, not better. For example, “mistakes” can occur when cells are dividing and reproducing, as the DNA in one of your cells is copied and split between two cells. These mistakes accumulate as we age, making the cells of an old person more sluggish than the cells of a young person — or worse, cancerous. What’s striking is that such developmental mistakes nevertheless turn out to be essential to the process of evolution — in a good way.

To take a different example, consider the many opportunities for error in gene expression — the process by which a gene gives rise to a protein. Gene expression involves “transcribing” a particular sequence of DNA into a new RNA molecule, which is then “translated” to produce a protein. This is a very complex process. A typical protein is composed of about 500 amino acids, each of which is encoded through a triplet of nucleotides — meaning that 1,500 nucleotides need to be correctly transcribed from DNA to RNA, which must then be correctly translated into a protein. After that, there is even more room for error: this protein then has to be “folded” properly so that it can perform its particular biological functions.

It is remarkable that any proteins come out the right way at all. Errors during the transcription process or the translation process can result in a protein that has the wrong sequence of amino acids. And even if they have the correct sequence, proteins can easily misfold into shapes that not only fail to function, but are toxic to the cell. Indeed, a significant fraction of the activity of every cell is dedicated to what is, in effect, an elaborate garbage-disposal system that cleans up the mess left behind by the large number of flawed proteins that get produced.

These developmental errors during gene expression are distinct from the more familiar germline mutations, in which offspring wind up with a new version of a gene, one that is different from the versions of either parent. Germline mutations are particularly dangerous errors because they don’t just affect single proteins or cells that could be destroyed and recycled; instead, germline mutations permanently affect the whole organism as well as its future children and grandchildren.

Perhaps unsurprisingly, mutations — errors in copying germline DNA — occur much less often than do errors in processes such as transcription or translation. Preventing errors is costly, biologically speaking, and so our bodies invest only as much as they need to in the task. Because the damage from a mutation in the germline is greater than the damage from making a single faulty protein molecule, our costly safeguards to prevent mutations in eggs or sperm are more meticulous than those that ensure quality control during the production of a protein.

While the magnitude of the consequences of germline mutations is so much greater than that of gene-expression errors, their specific effects can be remarkably similar at the molecular level. For example, DNA (deoxyribonucleic acid) molecules, which use an “alphabet” of four nucleotides, abbreviated as G, A, T, and C, are transcribed into RNA (ribonucleic acid) molecules that use a similar alphabet of G, A, U, and C. The similarity between the chemistry of the DNA alphabet and the chemistry of the RNA alphabet means that the effects of a permanent mutation that changes a G to an A in the DNA are similar to the effects of a mistake in transcription that also changes G in the DNA to A in the RNA.

This similarity at the molecular level between germline mutation and errors in gene expression opens up a remarkable opportunity for mutations to get “prescreened” before they even occur —meaning that natural selection can occur prior to mutation. If you have, say, 1011 copies of a protein in your body, and a particular error happens at a rate of 10-5, then 1 million of your copies of the protein, on average, have that same error. So although your genotype encodes what a particular protein is “supposed” to be like, by making errors your body contains not only the “correct” version of the protein coded for by a given gene, but potentially a whole range of other similar — though “incorrect” — versions of that gene’s protein.

With the error rate in this example, that could be enough for natural selection to take notice. To see how, imagine you have a permanent, germline mutation that doesn’t affect how well your protein works in normal cases, when it’s transcribed and translated correctly. But the mutation does change the fraction of error-containing variants that work properly, say from 40 percent to 42 percent. That means slightly less work for your cells’ garbage-disposal system and more fitness for you — making you healthier and more likely to survive and reproduce. In other words, this mutation benefits you, evolutionarily speaking. Natural selection doesn’t just judge how well a gene works when its proteins are produced correctly, but also how well all the error-containing versions of the gene’s proteins work.

Because the same protein variants can be produced either by mutation or by gene expression error, evolution under constant pressure from developmental errors leads to improvements not only in developmental robustness, but also in what biologists call “evolvability” — the capacity to produce genetic variants that help adaptation. For evolution by natural selection to work, a reasonable fraction of genetic mutations has to be not just harmless but also helpful. This seems like a tall order. Aren’t random changes to any highly complex thing likely to break it, rather than improve it? Yes, but they are much less likely to break it, and hence more likely to improve it, if those mutations have already been prescreened.

One of the “incorrect” protein variants your body produces in error today could be the new “normal” tomorrow, if the right permanent germline mutation comes along. So the more resilient your proteins are to the presence of these “mistakes,” the higher the number of viable options that evolution has to work with, and the easier it is to pick out the good ones. Developmental errors provide a kind of playground where future genetic options can be tried out on a small scale, before they are risked on the larger, evolutionary stage.

The evolution of evolvability may seem like a problematic idea at first, because evolution has no foresight. How can evolution know what will be adaptive? On one level, this is true, of course: evolution does not know what will turn out to be useful. But evolution does know what won’t be useful. Genetic variants that increase the likelihood of death or sterility generally fit the bill of uselessness. Breaking essential housekeeping genes that deal with basic cellular functions is also almost always a bad idea, evolutionarily speaking. But the effects of other, less drastic variants are harder to predict, and may depend on the circumstances. Mutations that fit into this latter category tend to be ones that are mostly harmless, mimics of the kind of developmental errors that evolution tolerates. Problematic mutations, by contrast, especially catastrophic ones, resemble at the molecular level the kind of developmental errors that evolution moves far away from.

Developmental errors give evolution a glimpse of the future, so to speak — not enough to tell it where to go, but enough to help it avoid some of the more obviously bad paths. By a process of elimination, whatever is left over after this “prescreening” process can only be better than the original set of choices.

By contrast, if there were no developmental errors, there would be no effective prescreening for robustness and the evolvability that comes with it. Evolution — if it worked at all — might end up taking other, less productive paths, producing proteins so fragile that almost any mutation would destroy them.

Without errors, in other words, evolution’s creativity would be stifled. If we were already perfect biological organisms, evolution would have nowhere to go — no diversity to explore, no fountain of creativity. The takeaway? Embrace the waste, the mess, and the errors — embrace the imperfection. Without it, the diverse wonders of the natural world could not have come to be.

Discussion Questions:

  1. Could some species be less error-prone than others, and hence less evolvable?
  1. Have more evolvable genotypes with higher error rates been favored by natural selection, leading to the evolution of evolvability?
  1. If we allowed computer hardware to make more errors, could software become as evolvable as biological systems?
  1. How often do small mistakes in our personal and business lives inadvertently give us a preview of some better option — leading to a better outcome than if everything were always done “correctly”? Are these mistakes more like permanent “mutations”, or like “gene expression errors” that provide a miniature preview of what a future mutation might be like?

Discussion Summary

Two themes that came up in the discussion about developmental mistakes and evolution were: (1) agency/intent, and (2) the challenge of finding the right language to express scientific ideas.

The two are related. A major challenge for the language in which we give popular accounts of evolutionary biology is to avoid making claims about agency that are not warranted by the science. This is partly an issue of the popularization of science: In the original scientific research, many of the ideas are expressed in the language of mathematics, which is naturally far more precise and has a smaller range of connotations.

But this is not exclusively an issue of popularization. Professional scientists don’t think exclusively in mathematical terms; we all are guided by metaphors and narratives to some degree. Moreover, some technical scientific jargon is also drawn from ordinary English words, which can no doubt be confusing to a popular audience. But although their ordinary meaning may be supplanted during the long training scientists go through, this process is bound to be incomplete, with the broader range of connotations retaining some salience for scientists too, not always at a conscious level.

So the discussion points to some new big questions not only about how we express scientific ideas in the public domain, but also about how language and metaphor might influence scientific inquiry itself.

New Big Questions:

  1. How do we discuss scientific topics in the public domain for which full definitions require mathematics and/or highly technical jargon and when partial definitions can seem to imply things that are not meant?
  1. If scientists’ insider language is difficult for others to follow, who can best act as a useful outsider, scrutinizing the choice of metaphors and narratives that might subconsciously influence their work?

13 Responses

  1. Elliot G. says:

    If developmental mistakes or errors are essential to evolution, aren’t they, in some sense, not mistakes or errors at all? Would it be better to call them developmental “experiments” or something like that, something less judgmental? What, exactly, do “mistake” and “error” really mean in this context?

    • Ventus says:

      What or who is responsible for the development of such mutational occurrences in nature?

      • Joanna Masel Joanna Masel says:

        This article isn’t primarily about mutation (inherited errors), but about errors in gene expression and other developmental processes. I would turn the question around: Noise and mess are the default, so who or what is responsible for avoiding errors?

    • Joanna Masel Joanna Masel says:

      In this context, when one version is made 99.9% of the time, and 10 others are each made 0.01% of the time, those 10 alternatives are called “errors.” Another reason we call a class of variants “errors” might be that they tend to happen more when selection is weaker and less when selection is strong. These words are not meant to imply judgment in the ordinary, human sense.

  2. Jon Perry says:

    Great article! I never thought of or heard of this little positive feedback loop before.

    Questions:

    1. From this I would expect new genes to produce higher numbers of toxic error proteins, and old genes (of similar length) to produce higher numbers of neutral error proteins. Has this been demonstrated?

    2. Even more basic of a question, has anyone done work on a specific protein to find a ratio of toxic vs neutral variations? I suppose a solid method for this would need to be worked out before we could compare old and new genes.

    • Joanna Masel Joanna Masel says:

      Both excellent suggestions. We have some unpublished work that suggests something close to #1, namely that genes that have more errors (often simply because they are expressed more often) have errors that are less likely to be toxic. The bottleneck in proving this was to find a sufficiently reliable proxy for which polypeptides are likely to be toxic. We are also working on genes of different ages, but haven’t put the two together. Expression level is a very strong confounding factor in this sort of study, and would be hard to control for.

      #2 is best answered by a recent set of studies that go by the name “deep mutational scanning,” which applies to individual proteins. We are also working on a less precise, but also less labor-intensive whole-genome approach to this question, but are not yet sure that it is going to work.

  3. Kimberly Kesselbaum says:

    Thank you, Professor Masel, for this interesting article, and thank you also to BQO for this article. I hope to use it in some of my homeschooling classes.

    A question for you: Evolution through developmental mistakes seems like a pretty inefficient way to evolve. Does this model of evolution go against or complicate the idea that nature tends to “choose” the most parsimonious means to bring about its ends?

    • Joanna Masel Joanna Masel says:

      Evolution is a “satisficing” process, which is a fancy way of saying that its solutions only have to be good enough, and no better. Evolution by natural selection doesn’t necessarily go towards any particular end, just towards something that is at least somewhat better than what went before, given the current circumstances. This process need not be particularly efficient.

  4. wduch says:

    Nature has learned to prescreen beneficial mutations, so the mutation landscape is biased towards beneficial mutations. Probability of mutations in different parts of DNA is certainly non-random, and there are robust mechanism that are essential to support basic mechanisms of living organisms, resistant to mutations. The rate of mutations of some genes is regulated by other genes. In research on creativity Campbell (1960) idea of Blind Variation, Selective Retention (BVSR) has been quite successful. Although random factors may exert evolutionary pressure genetic variation is not quite random, and probably only partially blind, biased towards selective retention of some traits. Epigenetics and genetic assimilation are interesting candidates for such learning mechanisms. Anyway, as a guiding principle search for bias in mutations is worth pursuing.

    • Joanna Masel Joanna Masel says:

      I wouldn’t say that Nature has “learned” to do this: It simply happens. The assertion that “mutation is random” has many different meanings, most of them problematic. In reality, yes, the biases in mutation are important to evolution.

  5. Brendo says:

    This is a really interesting article!

    I was wondering, you write that some transcription errors are similar to germline mutations because they both basically change the amino acid sequence of proteins. What about other kinds of developmental errors that result in post-translation defects in the protein, such as protein misfolding or having the protein sent to the wrong part of the cell or having the wrong kinds of extra molecules like sugars added on to the finished protein?

    Do any of these kinds of errors in protein synthesis that don’t affect the amino acid sequence have any influence on the evolutionary process?

    • Joanna Masel Joanna Masel says:

      Absolutely! I talked about change to amino acid sequences because it’s particularly easy in that case to see the relationship between expression errors and mutation errors. But the argument I was making also applies to all the other developmental errors that you mention, too.

  6. karu says:

    This is a very interesting article. What I want to point out, and what I think is very worthy of critical discussion, is that you (and the whole of evolutionary biology and genetics) are using a highly metaphorical language here. You wisely indicate that by using quotation marks when you are invoking terms like “mistake,” “error,” “transcription,” “translation” for the first time. I would say that it would be better to always use quotation marks around words like those, at least to remind the reader of the fact that this is metaphorical speech.

    Metaphoric descriptions like those are used all the time in genetics and biology. I think there lie some serious epistemological problems here. What we get here are highly anthropomorphic descriptions of natural processes. Some questions that I find interesting for a discussion:

    It is not enough to say, “hey, the description I give here is metaphoric, I don’t really mean this to be taken literally.” This would not be science. The question is: Can we do without the metaphors in our scientific accounts of those genetic processes? And if so, what would we be left with then? Could we still say the same things, make the same extrapolations? Would some questions or problems turn out to be only pseudo-scientific, because they couldn’t even be articulated in a strictly scientific terminology?

    One example: We couldn’t even speak of “mistake” or “error” if we didn’t presuppose that what we are talking about is a process of “copying” or “translation.” But what actually makes us claim that? I understand what is meant if it is said that a human person is copying or translating something. I understand even what is meant by saying that a (human built) machine is copying something. In both cases what is presupposed is that there is an intention involved. Intention presupposes subjectivity. There is no such thing as a “subject” that has intentions here – “nature” or “evolution” are not subjects that intend (and sometimes fail) to do something.

Leave a Reply