Fascinating new paper in Nature by Joseph Thornton's lab titled "Historical contingency and its biophysical basis in glucocorticoid receptor evolution." In my opinion, Thornton's one of the best scientists out there and this paper shows why. It looks at the history of evolution with a biochemist's eye for detail and also speaks to the big-picture questions. This paper is going to require some repeat reading and digestion, and for now, I'll start at the end, with big-picture thoughts. I find the experiments illuminating and important -- but I disagree with some of the interpretations of the experiments.
The experiments leave no doubt that, for this hormone binding this receptor, there are only a few small roads through the forest of possibilities that led from the original, less-picky receptor to the later, more-picky receptor. Also, there were no signs leading to those roads, that is, no evolutionary pressure that could have helped pick these rare mutations out of the crowd of possible mutations. This is similar to how DNA can change randomly if a water molecule hits it the wrong way. The specific hormone-receptor interaction we have is shaped by randomness and it very well may have a different shape if the tape of life was run again. (The surprise for me is the tight relationship between overall protein structure and the specific hormone structure, which is more intense than I thought it would be -- in other words, I expected proteins to be able to do more with the multitude of shapes they could adopt.)
But, as Simon Conway Morris might say, so what?
The key to the interpretation of this study is that the evolution we're talking about is the specific arrangement of oxygens around a constant, central 4-carbon-ring sterol core. What separates cortisol from the other hormones is the placement of an oxygen over here and not over there. In a dance of co-evolution, this specific arrangement of cortisol's oxygens was chosen out of a welter of possible arrangements and the protein changed alongside, taking a very limited number of roads to do so.
But if those improbable permissive mutations for our cortisol system never happened, then a different arrangement of oxygens on the same carbon core could easily send the same cortisol signal. The specific molecular structures may be unpredictable and contingent, but the fact that some specific arrangment of oxygens will send a specific signal like cortisol can be predicted and repeated. Presumably that's why we have so many signals built around 4-carbon-ring cores decorated with oxygens at different places. The variation in the oxygens isn't as interesting as the constancy in the carbon core, and the fact that all these hormones sending very different signals follow this same, predictable pattern.
Also, the fact that there are only a few roads (2 out of thousands or even millions) that can develop into this specific system doesn't account for the variable speed with which the system can develop to explore those roads. If mutations can be accelerated by stress, could they be accelerated enough to make this improbable path probable? After all, there are only a handful of mutations that are required to change the specificity, and if the rate of mutation for this receptor gene speeds up, the improbable becomes more probable.
Other carbon-based, water-loving organisms on other planets therefore probably have a very different cortisol shape. But although the oxygens may be differently placed in their cortisol, I would predict that there would still be oxygens placed around a carbon-ring core.
The oxygens can flit around the central carbon core and the receptors can mold themselves to those oxygens in many different ways, reshaping the circuitry of the hormone system. But from a more distant perspective, where the exact placement of oxygens can't be seen, the system would adopt much the same shape, using oxygen and carbon in much the same way to send much the same signal. The shape of the foundation is contingent, but the overall style of the house is predictable.
What this paper does is that it clearly places hormone-receptor interactions (within the sterol class of hormones) in the "contingent" category, but I maintain that the chemistry of the signaling system would remain much the same, whichever path the hormone-receptor interactions took. The hormone may be contingent, but the overall shape and chemistry of hormone system is predictable.