Experiments with "fitness landscapes" explain evolution of interacting genes
A reader asked an interesting question about the difficulty of coordinated evolution of groups of genes. Although I welcome comments and questions, I won't usually have time for detailed responses. and I'd already discussed one paper this week. But then Huxley brought in a recent issue of Nature he'd been chewing on, and there it was: "Empirical fitness landscapes reveal accessible evolutionary paths" (Nature 445: 383-386). So I guess I should take this dog-given opportunity to talk about the evolution of multiple interacting genes. The Nature paper is a review article with no original data, so isn't eligible for my regular weekly paper discussion, but maybe it's OK as a bonus paper, especially since the most interesting papers it discusses were published within the last year and they do contain original data.
The exciting thing about these papers is that people are starting to use molecular methods in experiments that solve "you can't get there from here" problems in evolutionary biology.
Let's start with a one-gene example, the paper by Weinreich et al. (Science 312:111-114). They looked at the evolution of antibiotic resistance involving a beta-lactamase gene in certain bacteria. Even with only one gene, evolution of resistance required five mutations. The odds against all of these mutations happening to any one bacterium are really small, even though there are lots of bacteria. So evolution of resistance probably required five successive steps. The problem is, what if one of the intermediate genotypes (step 3, say) were less fit than earlier genotypes in this evolutionary path? The intermediates might die out before any of them mutated to the next step. They didn't know what order the mutations occurred in; there are 32 possible intermediate genotypes (2 to the fifth power) and 120 possible 5-step paths from susceptibility to resistance (5 factorial). So Weinreich et al. made all 32 genotypes and compared their antibiotic resistance. Then they were able to calculate which of the 120 possible pathways involved a decrease in antibiotic resistance at one or more steps, and which pathways, if any, had increasing resistance at each intermediate step. 102 of the possible pathways involved intermediates with lower resistance. Therefore, evolution by any of these pathways would be unlikely, if the only mechanism was natural selection driven by the presence of the antibiotic. The 18 remaining pathways didn't involve intermediates with decreased resistance, but some were still considered more probable than others, for various reasons. Note that there's no doubt about whether this gene evolved; both "before" and "after" versions are out there. Rather, these experiments let us assign relative probabilities to different pathways by which the evolution occurred.
You might think that evolution of interacting genes would be much harder. As the review put it, "if the lock is modified first, the intermediate is not viable because the old key does not fit, and vice versa." Bridgham et al. (Science 312:97-101) looked at the problem of how one hormone and one receptor could evolve into two hormones and two receptors. In this case, evolution involved duplication of genes, a common event also responsible for innovations like color vision. Once there were two copies each of hormone and receptor gene, one copy of each could evolve a new function without losing the original function. Before molecular methods were widely available, we might have speculated about this mechanism, but now it can be tested experimentally. Bridgham et al. used DNA sequences of existing animals to infer ancestral sequences, synthesized genes with the inferred sequence, and tested hormone-receptor interactions in cultured cells containing the synthesized sequences, etc., confirming "recruitment of an older molecule, previously constrained for a different role, into a new functional complex."
There are some other nice examples in the Nature review, including experiments by my colleague Tony Dean and collaborators that let them draw inferences about complex evolutionary events (different enzyme substrate and different cofactor) over a billion years ago. But that's enough for now. I expect we'll be seeing more of this sort of study, further clarifying the evolution of interacting groups of genes.
The same issue of Nature has a neat paper about a microbe that eats chloroplast-containing prey. Rather than digest the chloroplasts, it keeps them, thereby benefiting from ongoing photosynthesis. Chloroplast function requires some genes in the prey nucleus, but fortunately the predatory microbe keeps the nuclei, too, for up to 30 days. It would be fun to come back in a million years and see how this interaction evolves. Or maybe try a little experimental evolution, setting up conditions that select for longer retention of the prey genes?
Thanks for all the comments, especially the chimp movies and suggestion for a paper to discuss!
I made up the part about Huxley.