I think it’s fair to say that uncovering the distribution of mutational effects is one of the central problems (if not the central problem) in molecular evolution and genetics today. I mean “mutational effects” in a very broad sense, including the direct effects on the genome itself (e.g., point mutations, indels, larger structural rearrangements), effects on genes and their products (e.g., protein abundances, catalytic activity), and of course effects on whole-organism traits like fitness. Our ultimate goal is to determine the rate at which mutations with these various effects occur, which we could then use to determine which types of mutations tend to dominate in populations over time.
I think this is where controlled, laboratory evolution of microbes like E. coli and yeast is making its biggest impact. The group of Maitreya Dunham (University of Washington) recently posted a preprint to bioRxiv with some interesting contributions to this topic, focusing on both haploid and diploid yeast in different environmental conditions:
Empirical determinants of adaptive mutations in yeast experimental evolution Celia Payen, Anna B Sunshine, Giang T Ong, Jamie L Pogachar, Wei Zhao, Maitreya J Dunham https://dx.doi.org/10.1101/014068
The study has two main parts. In the first, they assemble a large library of mutant strains to approximate the distribution of all “single-step” mutations. That is, they create strains with each gene deleted (to represent knockout mutations) and with each gene on low- and high-copy plasmids (to represent mutations that increase expression). They measure the relative fitness of all these strains (in glucose-, sulfate-, and phosphate-limited environments) using barcode sequencing.
Most of the results here are straightforward in my opinion: the vast majority of mutants are effectively neutral; deletions tend to be slightly deleterious, and more deleterious in haploids than in diploids. But a few observations stood out to me, such as that there are more beneficial deletions in haploids than in diploids (at least in sulfate- and phosphate-limited environments), and that increasing copy number can be just as deleterious as beneficial.
Assuming this is a reasonable approximation for all single-step mutations, the authors sought to check whether it was consistent with the collection of mutations that were actually observed to spontaneously arise and establish in evolution experiments. I thought this was the most fascinating part of the paper: they performed a meta-analysis of 106 previously-published evolution experiments in yeast, yielding 1167 mutations in 1088 genes. They also performed new experiments that resulting in 150 additional mutations.
Some really interesting trends appeared here. Haploids tended to adapt by losing functions (via premature stops, typically near the start codon), while diploids tended to adapt by amplification (via intergenic and 5’ upstream mutations). Presumably this is because deletions tend to be recessive in heterozygotes and therefore not so beneficial in diploids, although some deletions were found to be dominant. However, these distinct haploid/diploid trends only held in glucose-limited environments, and not really in phosphate- or sulfate-limited. Another important caveat is that copy-number variants are believed to be more important in diploid evolution, but most of the experiments surveyed here didn’t report them. The other fascinating aspect of the meta-analysis was that a large number of the observed mutations were concentrated in a small number of genes. Moreover, these recurrently-mutated genes were enriched in “high-impact” mutations (frameshifts, gain or loss of a start or stop codon).
Finally, they wanted to reconcile the mutations observed in all these experiments with their new systematic survey. In general, it seemed that the recurrently-observed mutations aren’t well explained by the survey: 41 of 70 mutated genes were not associated with beneficial mutations in the survey, and in fact 24 observed mutations were associated with (low-impact) deleterious effects in the survey. However, mutations that were observed are enriched for high-fitness effects.
I think the main limitation of this work is that deletions and amplifications are only a small part of the mutational repertoire — certainly mutations can have more subtle effects such as fine-tuning protein stability or binding interactions. So their systematic survey leaves out a wide range of mutational possibilities, which is perhaps why it’s not surprising that it doesn’t fully account for the observed mutations in experimental evolution. Of course, a systematic exploration of those more subtle effects is a daunting task, although some recent studies have begun to construct exhaustive single- and double-mutant libraries for individual proteins. Nevertheless, the growing emphasis on high-throughput mutational scans is clearly a very positive step toward acquiring the comprehensive quantitative picture that we need to address major evolutionary questions.