The arc of evolutionary genetics may be irreversible

One of the banes of modern life is the stack of papers in one's "to-read" list. I guess that goes to show how cushy modern life is, as what sort of complaint is that? In any case, I began to consider this after reading Joe Thornton's magisterial response to Michael Behe's giddy excitement over his most recent paper, An epistatic ratchet constrains the direction of glucocorticoid receptor evolution. Thornton dispatches Behe's muddled misconceptions with economy and precision, but after reading the paper, as opposed to cogent summaries such as Carl Zimmer's in The New York Times I'm even more at a loss as to how Behe arrived at the conclusions he did as to the paper's significance (please read the paper, available on Thornton's lab website, and then try and make sense of what Behe is asserting) . But at least Michael Behe prompted me to push the paper to the top of my stack.

To understand why this paper is important, one has to know a bit of the history of evolutionary theory (ironic in that this paper points to the importance of contingency, that is, historical sequence of events, in evolution). The arguments about whether evolutionary process is contingent or not go back a century. One can see glimmers of it in the debates between R. A. Fisher and Sewall Wright, the two preeminent figures of 20th century population genetics (Fisher arguably invented the field). Fisher and Wright are relevant to the discussion around this paper because of their views about the genetic architecture of evolutionary process. The former tended to elaborate a simple perspective whereby the focus was on genetic variants of independent and additive effect against an averaged genetic background. As an example, presumably the thousands of genes which effect normal variation in height, and are selected en mass when a shift in height away from the mean value results in higher reproductive fitness. In a Fisherian world the adaptive landscape is smooth and without discontinuity, with a simple symmetrical peak around which the species' total population flows. The world may not be flat, but its geometry is elegant. In contrast, Sewall Wright conceived of the adaptive landscape as far more rugged, subject to the complex interlocking effects of epistasis as well as stochastic random walks, and fragmented into numerous populations with some barriers to gene flow. Wright's ideas were not entirely coherent, as his biographer Will Provine lays out in Sewall Wright and Evolutionary Biology, but the metaphors and generally more empirical methodology left an imprint (while Fisher's contributions tended to be theoretical, Wright was an experimental geneticist with a more intuitive feel for the shape of reality). Though Stephen Jay Gould gave a hearty nod to Sewall Wright as a primary influence (though he also argued that Wright watered-down his own theories to conform to Neo-Darwinian orthodoxy), Gould's nemesis William D. Hamilton declared that Wright was the strongest influence upon his later work in his collection of papers. The combination of Wright's extremely long career (he was 1 year older than Fisher, but outlasted him by 26 years) and his occasional tendency toward lack of clarity may go some way to explaining how and why such disparate intellectuals could look to him as an influence.

The tension between Fisher's elegant formalism which led one toward an ahistorical sensibility, and Wright's more sloppy empirically informed conjectures, has to a great extent echoed down the decades. In the 1960s Ernst Mayr attacked "beanbag genetics," in other words, Fisher's genetics. Fisher's British colleague, J. B. S. Haldane, rose to the challenge (Fisher had died by this point), publishing A Defense of Beanbag Genetics. In more recent years Stephen Jay Gould laid out his vision of historical contingency in Structure of Evolutionary Theory. As the inevitable counterpoint Richard Dawkins took Simon Conway Morris' side in The Ancestor's Tale, arguing for the inevitability of particular morphological types due to the deterministic character of natural selection. And so it goes.

Thornton's paper does not resolve this general argument, but it does illustrate starkly the constraints which natural selection faces in a universe of finite time:

The extent to which evolution is reversible has long fascinated biologists...Most previous work on the reversibility of morphological and life-history evolution...has been indecisive, because of uncertainty and bias in the methods used to infer ancestral states for such characters...Further, despite theoretical work on the factors that could contribute to irreversibility...there is little empirical evidence on its causes, because sufficient understanding of the mechanistic basis for the evolution of new or ancestral phenotypes is seldom available...By studying the reversibility of evolutionary changes in protein structure and function, these limitations can be overcome. Here we show, using the evolution of hormone specificity in the vertebrate glucocorticoid receptor as a case-study, that the evolutionary path by which this protein acquired its new function soon became inaccessible to reverse exploration. Using ancestral gene reconstruction, protein engineering and X-ray crystallography, we demonstrate that five subsequent 'restrictive' mutations, which optimized the new specificity of the glucocorticoid receptor, also destabilized elements of the protein structure that were required to support the ancestral conformation. Unless these ratchet-like epistatic substitutions are restored to their ancestral states, reversing the key function-switching mutations yields a non-functional protein. Reversing the restrictive substitutions first, however, does nothing to enhance the ancestral function. Our findings indicate that even if selection for the ancestral function were imposed, direct reversal would be extremely unlikely, suggesting an important role for historical contingency in protein evolution.

To be terse about it, you have ancestral phenotype A and derived phenotype D. And you have ancestral set of amino acids and a derived set of amino acids. In a "beanbag" conception you simply focus on the amino acids which directly effect A & D, and assume that the "background" can be ignored as a first approximation. But it turns out that it doesn't work like that, in their experiments the researchers found that there are other amino acids which had changed in the interval between A & D and blocked potential evolutionary reversal by simply changing the core set of amino acids back (labeled "XYZ" in the paper).

There are always supportive biological process around many mechanisms which you take as "givens," the formless background. As you are focusing on the genes and traits of interest these givens may also change, and no longer be givens which can be ignored. It turns out that there were other amino acids which changed between the ancestral and derived states (labeled "W" in the paper) which made a reversion back to the ancestral genotype result in loss of function. The ancestral genotype was out of sync with its background singers, so to speak.

In the paper they used some ingenious methods to map from the amino acids to the biophysical structure of the proteins generated by a particular acid sequence. After all, it is the nature of these structures which determines the function of the DNA-binding transcription factor. If you are curious about the details of the mechanism in terms of how it physically manifests, I recommend you follow the link I provided earlier. The conformation of proteins is not my cup of tea, rather, I will continue to speak of more abstract constructs in the paper, the cluster of amino acids XYZ and W. Not only does the state of W result in the lack of functionality of xyz (the ancestral state), but reversing W back its ancestral state first often reduces functioning as well! And there you have the ratchet. Stepwise changes result in fall into the fitness valley, and there lay extinction. Rather, there must be an unlikely leap across to directly reverse path. There are likely other ways in which one could reconstruct the ancestral phenotype through other sets of amino acids which would neutralize W first, and so allow for the mutation of XYZ back to xyz without penalty, but as noted in the paper the longer the path the more unlikely the probability of wending back toward the original location.

So why did w mutate to W in any case? They speculate that the change was either as a modifier of XYZ, or, that it was neutral. If the change was neutral we are now adding an element of stochasticity, that is, the ratchet is somewhat random. It's as if some of the time when you crossed a bridge it collapsed behind you, and some of the time it remained sturdy. Imagine you're making your way through a network of such bridges which collapse stochastically after you've crossed. Now assume that the goal is to go from point A to point Z, and then reverse back to point A. If all the paths are initially open, but some of them close up behind you by chance then every single reversal from Z to A may have to take a different path. These are the random acts of contingent history.

Here is the conclusion of the paper:

Our observations suggest that history and contingency during glucocorticoid receptor evolution strongly limited the pathways that could be deterministically followed under selection.The 'adaptive peak' represented by the promiscuous AncGR1 is a relatively close neighbour in sequence space to themore specific AncGR2. This peak was occupied in the ancestor of jawed vertebrates--indicating that no intrinsic constraints prevent its realization--but it became far more difficult to access just 40 million years later because of intervening epistatic mutations. Selection is an extraordinarily powerful evolutionary force...nevertheless, our observations suggest that, because of the complexity of glucocorticoid receptor architecture, low-probability permissive substitutions were required to open some mutational trajectories to exploration under selection...whereas restrictive substitutions closed other potential paths. Under selection, some kind of adaptation will always occur...but the specific adaptive forms that are realized depend on the historical trajectory that precedes them. The conditions that once facilitated evolution of the glucocorticoid receptor's ancestors were destroyed during the realization of its present form...The past is difficult to recover because it was built on the foundation of its own history, one irrevocably different from that of the present and its many possible futures.

There's a fair amount of hedging in the text, as there should be in science. How relevant are these sorts of insights on the biomolecular level to gross morphology? I suspect this is exactly the sort of thing that Richard Dawkins would reply with; specifically, that his support for Simon Conway Morris' thesis was on the scale of gross higher-level phenotypes, not details of biomolecular function so closely coupled to the state of amino acids. As a reductionist it seems that one would be behooved to focus on the real and concrete manifestation of the genetic substrate which has such a direct connection to structural constraints. And yet I wonder if perhaps this opens up an avenue to rescue those who would argue that history does not matter on a larger scale, because inevitability is a function of the laws of physics. As an example, the persistent reduction in appendages across taxa characterized by larger body size may be due to physical biomechanical efficiencies which natural selection will hone in upon given enough time.

Neither the partisans of contingent history or those of the march of inevitable history in this battle really espouse an either/or model. Rather, they're fighting on the margins as to what is, or isn't, significant, what should, or shouldn't, be emphasized. After all, selection itself to a great extent is stochastic, maximizing fitness with the parts on hand because of the limits of time. It seems that the most forceful implication of this type of research is that biological cartographers need to fan out and map the lay of the land, one simply can't be complacent with a simple heuristic assuming idealized topographies.

Below I've reformatted Figure 4. If the prose above was unintelligible, look at hard....

i-0bf7a958413317bd77de833cfecdeacb-thornton4.png

Citation: An epistatic ratchet constrains the direction of glucocorticoid receptor evolution, Jamie T. Bridgham, Eric A. Ortlund, Joseph W. Thornton, Nature 461, 515-519 (24 September 2009) doi:10.1038/nature08249

Categories

More like this

Great post and excellent summary of a convoluted history. I'm with not sure how to evaluate the generality of the evidence in this paper, though, on the same ground you attribute to Dawkins. They argue that reverting to certain genotypes or endophenotypes (e.g. a particular protein's structure) might be next to impossible. But given that many genotypes might give equivalent (in terms of selection) phenotypes, organisms would still be relatively unconstrained in phenotypic space for many traits. For example, selection for "tall" would still be easily accommodated even if one particular tall-promoting locus became constrained.

Now I'm all about evolutionary constraints and think they are massively underappreciated, especially by fields like evo-psych that seem to assume an inordinate lability and selection on mental function in recent hominid history. But I think the important constraints are in developmental processes, not so much in DNA/protein sequence evolution.

Although Iâm sure the creationists will misconstrue the conclusion (as Behe does), itâs important to note that the authorsâ conclusion is limited to âdirect reversalâ of the (general) evolutionary sequence that actually occurred. That is, they are saying that going back to the ancestral function via a reversal of the general path taken to get the current function is what is âextremely unlikelyâ.

Behe continues to make the same fundamental error he made in the âEdge of Evolutionâ: assuming that the low probability of the *specific path* taken to get from point A to point B is evidence that getting from point A to point B (or back again) by *any route* is similarly unlikely.

A beautiful analysis of the details of this case. But the general phenomenon is (?used to be) well known. I showed it experimentally in Drosophila in 1964 (Heredity 19:734-739), and Lewontin has emphasized it several times.