The long-term evolution experiment


I’m attending the 2nd ASM Conference on Experimental Microbial Evolution (#ASMEME) in Washington, DC. The meeting opened last night with a keynote address by Rich Lenski on the long-term evolution experiment (LTEE). If you’re not familiar with it, the LTEE involves twelve populations of E. coli bacteria that have been transferred every damn day for the last 28 years. That’s right, twelve transfers every day since Ronald Reagan was President.

Since E. coli undergoes about 6.6 doublings per day under the experimental conditions, that means that the bacteria in this experiment have been evolving for over 65,000 generations. In that time, it has produced a wealth of information about evolutionary processes and spun out countless related experiments. The LTEE is so iconic that you usually don’t have to explain, at least to evolutionary biologists, which long-term evolution experiment you’re talking about. It has also played a role in some controversies, not least the “Lenski affair.”

So much has been learned from the LTEE that it would be impossible to cover all of it in a 45-minute talk; rather, Dr. Lenski’s keynote summarized just a few of the highlights. One of the recent findings, based on the first 50,000 generations, was that fitness (relative to the ancestral strain) is increasing more slowly than it did near the beginning of the experiment, but without any apparent upper bound. In other words, the rate at which fitness is increasing has slowed, but it shows no sign of stopping. This is remarkable when you consider that the environment in this experiment is, as nearly as possible, constant.

Figure 2 from Wiser et al. 2013. (A) Hyperbolic (red) and power-law (blue) models fit to the set of mean fitness values (black symbols) from all 12 populations. (B) Fit of hyperbolic (solid red) and power-law (solid blue) models to data from first 20,000 generations only (filled symbols), with model predictions (dashed lines) and later data (open symbols). Error bars are 95% confidence limits based on the replicate populations.

Figure 2 from Wiser et al. 2013. (A) Hyperbolic (red) and power-law (blue) models fit to the set of mean fitness values (black symbols) from all 12 populations. The hyperbolic curve would imply that the rate of increase in fitness approaches an asymptote, but the power-law curve is a better fit to the observed data. (B) Fit of hyperbolic (solid red) and power-law (solid blue) models to data from first 20,000 generations only (filled symbols), with model predictions (dashed lines) and later data (open symbols). Error bars are 95% confidence limits based on the replicate populations.

An even more recent paper fully sequenced the genomes of two clones from each of the twelve populations at 500, 1,000, 1,500, 2,000, 5,000, 10,000, 15,000, 20,000, 30,000, 40,000, and 50,000 generations. Six of the twelve populations have been taken over by hypermutators, mutants that are defective in DNA repair pathways and so experience mutations at up to 100 times the normal rate. Interestingly, subsequent mutations dampened the hypermutator effect, resulting in mutation rates that were still elevated, but lower than when the strains first became hypermutators.

Genomic evolution was primarily driven by selection, based on the overrepresentation of nonsynonymous mutations (which affect the amino acid sequences of proteins) and intergenic point mutations (which potentially affect gene expression) relative to synonymous mutations (which don’t affect amino acid sequences and are assumed to be neutral) in the evolving populations:

Figure 4 from Tenaillon et al. 2016. a, Synonymous mutations, scaled so that the mean of five non-mutator populations (excluding point mutation and IS150 hypermutators) is unity at 50,000 generations. b, Nonsynonymous mutations, scaled using the same rate as synonymous mutations after adjusting for sites at risk for both classes. c, Intergenic point mutations, scaled using the same rate as synonymous mutations after adjusting for sites at risk. Each symbol shows the mean for sequenced genomes from a non-mutator or premutator lineage. Colours are as in Fig. 1. Note the discontinuous scale; populations with zero mutations are plotted below. Black lines connect grand means; shading shows standard errors calculated from replicate populations.

Figure 4 from Tenaillon et al. 2016. a, Synonymous mutations, scaled so that the mean is 1 at 50,000 generations. b, Nonsynonymous mutations, scaled using the same rate as synonymous mutations after adjusting for sites at risk for both classes. c, Intergenic point mutations, scaled using the same rate as synonymous mutations after adjusting for sites at risk.

Parallelism, in the sense of the same gene containing mutations across different populations, was high. In the non-mutator populations, 57 genes, representing only 2.1% of the protein-coding portion of the genome, contained over half of the observed nonsynonymous mutations. The authors conclude that

…adaptation can remain a major driver of molecular evolution long after an environmental shift. Our experimental results thus support a selectionist view of molecular evolution, complementing indirect evidence based on comparative genomics in bacteria, Drosophila and humans.

The first part of this is hard to argue with. Selection is clearly still an important factor even after 50,000 generations (that’s more generations than separate modern humans from Neanderthals!). The ‘selectionist view’ is also justified, at least for this experiment, though the much smaller effective population sizes of many natural populations are likely to increase the relative contribution of genetic drift.

Parallelism across populations is often taken as evidence of predictability in evolution, and I think it’s fair to say that the degree of similarity among the outcomes of the twelve populations is surprising. There are also, though, some clear examples of contingency. The best known of these is the appearance of citrate consumers in one of the twelve populations. This innovation was contingent not only in the sense that it only appeared in one population (and only after 30,000 generations). It also apparently required a ‘potentiating’ mutation, which does not confer the ability to consume citrate, to be present at high frequency before the ‘actuating’ mutation would be beneficial.

Despite creationist objections to the contrary, this is a clear example of the evolution of a novel trait. Were the bacteria previously able to consume citrate in the presence of oxygen? No. Everyone agrees on this; it is so universally true of E. coli  that it is considered diagnostic. Can they do so now? Yes, this too is undisputed. If being able to do something that you couldn’t previously do doesn’t qualify as an innovation, I would love to hear your definition of innovation.

 

Stable links:

Blount ZD, Borland CZ, Lenski RE (2008) Historical contingency and the evolution of a key innovation in an experimental population of Escherichia coli. PNAS, 105, 7899–7906.

Quandt EM, Gollihar J, Blount ZD et al. (2015) Fine-tuning citrate synthase flux potentiates and refines metabolic innovation in the Lenski evolution experiment. eLife, 4, 1–22.

Tenaillon O, Barrick JE, Ribeck N et al. (2016) Tempo and mode of genome evolution in a 50,000‐generation experiment. Nature (online early), 1–21.

Wiser MJ, Ribeck N, Lenski RE (2013) Long-term dynamics of adaptation in asexual populations. Science, 342, 1364–1367.

 

Comments

  1. Pierce R. Butler says

    DonDueed @ # 1: … has any of the twelve polulations turned into a crocoduck?

    C’mon, how do you think they measure “fitness”?

  2. blf says

    No no no, it isn’t populations which turn into crocoducks, an individual bug has give birth to a crocoduck…

  3. Menyambal says

    I like the emphasis on the length of generations. Some folks think us humans are more evolved, but they don’t realize that with our 20-year generations, we are very slow evolvers. (So to speak, and with several caveats.)

Leave a Reply

Your email address will not be published. Required fields are marked *