In the June 2011 issue of PLOS Genetics the laboratory of University of Oregon evolutionary biologist Joseph Thornton published “Mechanisms for the Evolution of a Derived Function in the Ancestral Glucocorticoid Receptor,” the latest in their series of papers concerning the evolution of proteins that bind steroid hormones (Carroll et al., 2011). In earlier laboratory work they had concluded that a particular protein, which they argued had descended from an ancestral, duplicated gene, would very likely be unable to evolve back to the original ancestral protein, even if selection favored it (Bridgham et al., 2009). The reason is that the descendant protein had acquired a number of mutations which would have to be reversed, mutations which, the authors deduced, would confer no benefit on the intermediate protein. They used these results to argue for a molecular version of “Dollo’s Law,” which says roughly that a given forward evolutionary pathway is very unlikely to be exactly reversed.
In my previous comments here at ENV on this interesting work, I noted that there is nothing time-asymmetric about random mutation/natural selection, so that the problem they saw in reversing the steroid hormone receptor evolution did not have to be in the past — it could just as easily have been in the future. The reason is that natural selection hones a protein to its present job, with regard to neither future use nor past function. Thus, based on Thornton’s work, one would not in general expect a protein that had been selected for one function to be easily modified by RM/NS to another function. I have decided to call this the Time-Symmetric Dollo’s Law, or “TSDL.”
But if there is such a thing as a TSDL, did the forward evolution of the steroid-hormone protein-receptor manage to avoid it? That question had not yet been addressed. Was the protein lucky this time, encountering no obstacles to its evolution from the ancestral state to the modern state? If so, then maybe TSDL is occasionally an obstacle, but not so often as to rule out modest Darwinian evolution of proteins (as I had thought before reading Thornton’s earlier work).
Well, thanks to the Thornton group’s new work, we can now see that there are indeed obstacles to the forward evolution of the ancestral protein. The group was interested in which of the many sequence changes between the ancestral and derived-modern protein were important to its change in activity, which consisted mostly of a considerable weakening of the protein’s ability to bind its steroid ligands. They narrowed the candidates down to two amino acid positions, residues 43 and 116. Each of the changes at those sites decreased binding by over a hundred-fold. However, when the researchers combined both mutations into a single protein, as occurs in the modern protein, binding was not only decreased — it was for all intents and purposes abolished. Upon further research the group showed that a third mutation, at position 71, was necessary to ameliorate the effects of the combination of the other two mutations, bringing them back to hundreds-fold loss of function rather than essentially complete loss of function.
Carroll et al. (2011) conjecture that the mutation at position 71 occurred before the other two mutations, but that it had no effect on the activity of the ancestral protein. So let us count the ways, then, in which “fortune” favored the evolution of the modern protein. First, an ancestral gene duplicated, which would usually be considered a neutral event. Thus it would not have the assistance of natural selection to help it spread in the population. Next, it avoided hundreds of possible mutations which would have rendered the duplicated gene inactive. Third, it acquired a neutral mutation at position 71. Thus, again, this mutation would have to spread by drift, without the aid of natural selection. Once more, the still-neutral gene manages to avoid all of the possible mutations that would have inactivated it. Next, it acquires the correct mutation (either at position 43 or 116) which finally differentiates it from its parent gene — by reducing its activity a hundred-fold! Finally, somehow the wimpy, mutated gene (putatively) confers upon the lucky organism some likely-quite-weak selective advantage.
The need to pass through multiple neutral steps while avoiding multiple likely-deleterious steps to produce a protein that has lost 99% of its activity is not a ringing example of the power of Darwinian processes. Rather, as mentioned above, it shows the strength of TSDL. Darwinian selection will fit a protein to its current task as tightly as it can. In the process, it makes it extremely difficult to adapt to a new task or revert to an old task by random mutation plus selection.
Dollo’s law holds going forward as well as backward. We can state the experimentally based law simply: “Any evolutionary pathway from one functional state to another is unlikely to be traversed by random mutation and natural selection. The more the functional states differ, the much-less likely that a traversable pathway exists.”
(1) Carroll, S. M., E. A. Ortlund, and J. W. Thornton, 2011 “Mechanisms for the evolution of a derived function in the ancestral glucocorticoid receptor.” PloS. Genet. 7: e1002117.
(2) Bridgham, J. T., E. A. Ortlund, and J. W. Thornton, 2009 “An epistatic ratchet constrains the direction of glucocorticoid receptor evolution.” Nature 461: 515-519.
Photo credit: Conor Ogle, Wikimedia Commons