This will be my last post. I have spent way too much time on this, as valuable and insightful as it has been. I have appreciated our dialog, T_aquaticus, and I hope you have as well.
I am going to take one more stab at explaining the rationale behind our research. When Doug and I were first starting our research I used to argue with him. "No biologist would expect us to be able to convert Kbl into BioF--they are too divergent. There's too much epistasis!" And he would reply that modern proteins should be no different from ancient ones. If ancient proteins could evolve new functions then, they should be able to now. We had chosen the closest relative we could find to test, and even swapped out nearly the whole active site and binding pocket,but it didn't work, whether by rational design or directed evolution.
The complaint immediately was, just as I had said, "but you didn't start with the ancestral form!"
The point of all our research was to demonstrate that only a functional folded protein with some degree of starting activity for the target was amenable to conversion. Furthermore, if the only transitions that can take place are the ones that did take place, then we are extremely lucky to have a functioning metabolism, let alone a mind. Blind evolution has a problem doing this sort of thing. It would appear that epistasis, even at the beginning, constrains the paths that a starting protein can follow, and it continues to do so all along the way.
Are there paths between modern proteins, paths leading to innovation? If there are, they must be constrained as well. We would expect that. Now we get to the problem of cooption, or recruitment. It is our observation that conversion of an enzyme to a new function can work in a reasonable amount of steps, > provided it already has a small amount of activity for the target function (overlap in substrate preference, shared chemistry, etc.) If it has none of the target activity, the enzyme won't be converted without substantial reworking, more reworking than evolutionary processes could produce. All the papers that claim recruitment to a new function, if you look at them closely, describe proteins with some degree of overlap to start.
It's a catch-22. According to received wisdom, you have to retrace history, rewind the clock, to be able to interconvert protein function. This suggests evolutionary paths are locked in by epistasis, and there are only one or a few ways to move forward. The fitness landscape may be so rugged that only one or a few paths exist. That's a very thin needle for unguided evolution to thread, especially if the emerging new function is very weak, and therefore not accessible to selection.
A number of studies have looked at the process from the other end, where the two proteins being compared differ by just a few mutations. The possible pathways from one protein to the other are tested to see if any of them are possible evolutionary trajectories. For beta-lactamase, a 5 mutation difference meant 120 potential pathways but only 18 were evolutionarily feasible. And that's with the two proteins being nearly identical.
What I'm saying is that evolutionary pathways to new enzyme function appear to be incredibly fine-tuned. We shouldn't just say "and this thing was coopted to a new function" without testing whether it's possible. It's not easy at all. Joe Thornton can tell you about it. So can David Baker and Daniel Weinreich. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4313208/
The title of his paper is > Should evolutionary geneticists worry about higher-order epistasis? The answer is yes.
Believe me or don't believe me. Just think about it.
And thanks to all my interlocutors.
P.S. T_aquaticus, I hope you read the whole paper.