A central hypothesis in my forthcoming book, "Darwinian agriculture: where does Nature's wisdom lie?" is that past natural selection is unlikely to have missed simple, tradeoff-free improvements. This implies (as discussed in a recent post on drought-tolerant wheat) that tradeoff-blind biotechnology is less likely to succeed, relative to crop-improvement methods that consider tradeoffs, as long as biotechnology is limited to simple changes, like increasing the expression of an existing gene.
More complex improvements (those whose evolution would require a series of steps) are another story, however. Just because some hypothetical horse would kick ass, if it did evolve, doesn't guarantee that it will evolve. The problem is that you can't get from genotype A to some very different genotype Z, except through one or more generations of individuals with intermediate genotypes.
It's fairly easy to get from A to Z, provided that B is at least as fit as A, while C is at least as fit as B, and so on. This can be the case, as shown by experiments on the five-step evolution of antibiotic resistance, discussed in a previous post. But is this the only way a population can evolve a superior genotype? Or does evolution sometimes reach new heights (faster-flying birds, scummier pond scum, etc.) through intermediates that are significantly less fit?
Evolution via less-fit intermediates would expand evolution's options, making it even harder for biotechnology folks to come up with something missed by evolution. And that's what this week's paper seems to show.
"Compensatory evolution in mitochondrial tRNAs navigates valleys of low fitness" was recently published in Nature by Margarita Meer and colleagues.
Compensatory mutations are genetic changes that are harmful in isolation, but beneficial in combination with some other mutation. The chances of such mutations occurring simultaneously (the evolution-denier's caricature of how evolution would have to work) are negligible. The only way to get both mutations in the same individual is for a lineage with one of them to survive long enough to acquire the other. But are we talking about barely surviving, or thriving?
It's often hard to know whether a particular mutation would complement another mutation. But transfer RNA molecules are one place where complementation is easy to understand. These molecules are clover-leaf-like loop structures of RNA which help to translate DNA sequence into protein sequence, linking a DNA triplet (attached to the anticodon loop) to its corresponding amino acid (attached to the acceptor stem) during protein synthesis. (If some intelligent designer had wanted to make us immune to animal viruses and also provide inarguable evidence that humans are special, giving us a different DNA-to-protein code would have been an easy way to do it.)
tRNA loops are held together by AU and GC pair bonds, as shown in the drawing above, kindly provided by Fyodor Kondrashov, whose comments have improved this post. (U is the RNA equivalent of a T in DNA.) AU is strong and GC is strong, but GU and especially AC are weak.
The authors were able to infer the evolutionary history of these bonds, by comparing the equivalent tRNA sequences in related species, i.e., species descended from a common ancestor. For example, where humpback whales have a GC pair, Antarctic minke whales have AU (see above). Using sophisticated computerized "ancestral state reconstruction" (a technique previously used to infer ancient temperatures, among many other applications) the authors concluded that the last common ancestor of these whales had GC at that position.
So, somewhere along the lineage from that ancestor to minke whales, G mutated to A and C mutated to U. We don't know which change happened first -- the odds against them both happening simultaneously are huge -- but this evolutionary two-step, with a second compensatory mutation restoring the strong bond weakened by the first mutation, appears to be typical. This conclusion is based on their finding that, in most cases, when one tRNA base was different in an existing species from its inferred ancestral state, its partner base was also different and complementary, requiring two successive changes.
Was the intermediate genotype significantly less fit, as we would expect from the inability of mismatched partners to bond? To answer this question, the researchers made use of the fact that there are many different tRNAs, at least one for each of the 64 (4x4x4) possible DNA triplets. Some are found on GU-rich DNA strands, which somehow favor A=>G conversions, others on AC-rich strands, which favor the reverse.
So, if we know from ancestral-state reconstruction that an AU was converted to a GC, the intermediate was more likely to be a GU for a tRNA on a GU-rich strand, but an AC on an AC-rich strand, as shown above. GU pairs are more stable, conferring higher fitness, so the fitness of the intermediate generations in a two-step AU=>GC conversion would be greater for tRNAs on GU-rich strands. They found, however, that this conversion had occurred about as often on AC-rich strands.
It appears, therefore that this particular evolutionary two-step often passes through an intermediate with significantly lower fitness than either its ancestor or its compensated descendant. As their conclusion notes:
Whether or not such stepping stones can be used to cross wider fitness valleys remains an open question.