An important new paper came out recently that supports claims made by intelligent design scientists. Mike Behe has already addressed this paper here at ENV, but because its implications are so important I want to draw them out further, and make the connections with intelligent-design research explicit.
Doug Axe and I published a paper this year in which we examined how many mutations are required to convert the function of one enzyme to that of another. The two enzymes are structurally very similar, but have different chemistries. This is the kind of conversion that neo-Darwinism says must have happened many times over the course of evolutionary history, yet we found that for the protein pair we studied, conversion would require a minimum of seven mutations. Based on a realistic population genetics model, we calculate that the waiting time for a bacterial population to acquire seven specific mutations in a duplicated gene, none of which provide any functional benefit until all seven are present, is something like 1027 years. That’s a ten with 27 zeros after it. To put this in perspective, the age of the universe is believed to be on the order of 1010 years.
In response to our work, one critique was that we didn’t start with the right ancestral protein. The argument went something like this: unless the right ancestral sequence is used as a starting point, conversion between two structurally similar enzymes is very difficult. The proteins are adapted for different functions, and their folds are stabilized by numerous subtle interactions between amino acids. There is no path between the two enzymes, because epistatic interactions can prevent the desired reconfiguration. Epistatic interaction in this context means that mutating one amino acid position within a protein changes the effect of other mutations. Two mutations that are beneficial individually can be harmful when combined. Therefore, only when the actual evolutionary history is duplicated by using an ancestral starting point will there be a path to the new protein forms.
Yet, for this to be true, life on earth must have been very lucky (or designed) to evolve to where we are today. Assume as a starting point a minimal cell with about 300 genes. For that cell to evolve the full array of modern proteins, it must have had just the right ancestral proteins, proteins that could produce the current diversity without getting stuck in evolutionary blind alleys due to epistatic constraints. This kind of argument makes enzyme evolution harder, not easier to explain. Either conversion to new chemistry is hard, requiring many coordinated changes, and we are very lucky (or designed) to be here, or conversion is easy, requiring only a few changes (two or less, plus a duplication), in which case we should be able to see it happen in the lab.
So which is it? Based on this new research, it increasingly appears that either we are very lucky or we are intelligently designed.
Joseph Thornton and Sean Carroll and colleagues have studied the evolution of the corticosteroid receptors in vertebrates for years. They first used sequence analysis to reconstruct a putative ancestral receptor that had the ability to recognize two kinds of corticosteroid hormone. They then identified critical mutations that apparently had permitted a shift in the hormone receptor’s preference for one kind of hormone over the other. But when they examined the backward path from modern to ancestral form, it would not work due to negative epistatic interactions like the kind described above
Now, in a new study, they report that the three critical mutations required for change to have happened in the forward direction, the path supposedly followed by unguided evolution, also show epistatic interactions. Two mutations are beneficial individually but eliminate activity when combined. The third mutation is essentially a pre-adaptation that must be in place first; when it is present, the other two mutations become jointly beneficial, and lead to the new activity. Yet that pre-adaptation is by itself neutral — that is, of no selective benefit.
Mike Behe pointed out the improbability of this evolutionary path in his article titled “Wheel of Fortune: New Work by Thornton’s Group Supports Time-Symmetric Dollo’s Law.” The first step on the path involves a duplication of the receptor gene. Step two is the acquisition of the specific pre-adaptive mutation in one copy of the duplicated receptor gene.
Now, here’s the problem. In order for the remaining steps in the pathway to occur, the mutant duplicate must retain the ability to be expressed, either constitutively or upon induction by some upstream regulator. Also, no further mutations can occur, except those that are either neutral in effect, or that are the precise two mutations needed to arrive at the new selectable function.
At this point, we must remember that Darwinian evolution has no foresight. It doesn’t know about a hands-off policy for genes that are almost ready for a new function. Mutations will continue to pile up, regardless of their effect on the receptor’s function, and many of these will either turn off expression or inactivate the gene. This is precisely the point missed by many models of evolution that claim easy adaptation to new complex functions.
In bacteria and yeast, broken or useless genes that are expressed are pruned rapidly and overwhelmingly. Even regulated genes that are not being expressed are frequently lost. This is because there is a metabolic cost to maintaining unnecessary DNA, and removing it gives a selective advantage in a competitive environment. Mike Behe described many such examples in his article published in the Quarterly Review of Biology. Ralph Seelke and I also demonstrated this in a paper published last year.
In vertebrates, the strength of purifying selection (selection to remove unneeded DNA) is greatly reduced, due to the overwhelming effects of genetic drift. But mutation is still impartial, and in a target the size of the glucocorticoid receptor, there are many more ways to get it wrong than right. Pre-adapted duplicates with no benefit are highly unlikely to survive unscathed long enough for the next mutation(s) to show up, the one(s) that might actually confer a selective advantage.
It turns out that a complex adaptation requiring duplication and two mutations is right on the edge of what bacterial populations can hope to accomplish, given the cost of maintaining a duplicate. No one knows the limit for higher organisms, but it is likely to be small, as Behe suggested in his book the Edge of Evolution.
So to return to the main point: it would appear that converting a protein from one function to another is not simple or easy, even when the protein in question starts with some level of activity for the target. A simple change in binding preference in a pre-existing hormone receptor (such as described by Thornton’s lab) requires a gene duplication, three mutations in a particular order, plus continued expression, with no inactivating mutations. Random mutation, drift, and selection are highly unlikely to hit the mark. That leaves luck or design as an explanation, but to call it luck would seem to be pushing the probabilities.
And if merely changing binding preferences is hard, even when you start with the right ancestral form, then converting an enzyme to a new function is completely beyond the reach of unguided evolution, no matter where you start.