E. Coli with a more diverse palate

Today wasn't a good day in the lab for me. There weren't any explosions or failed cultures, just an impossible NOESY spectrum. It's tough to determine your proline isomer when your spectrum doesn't have the characteristic peaks for the cis or trans state (note: there are no other states). At times like this, it helps to be reminded of the value of persistence, which brings me to today's paper, involving an experiment that's been going on for 20 years.

A few weeks back I mentioned Carl Zimmer's excellent book Microcosm: E. coli and the new science of life, which details the study of one of the world's best-understood organisms and the insights it has given us. E. coli has loads of useful properties that make it a great research tool, and one of these is its doubling time, which in the lab typically ranges from a few hours to as few as 20 minutes depending on conditions. This has obvious benefits for researchers who just want to use the bacteria as a means of producing protein or DNA, but it also means that scientists interested in evolution can realistically expect to use E. coli to answer questions that would require tracking a population over tens of thousands of generations. In a recent paper in PNAS, Richard Lenski's long-term evolution experiment (LTEE) has accomplished just that.

One can, for instance, ask whether evolution is random or deterministic. That is, given some specific context, is a particular outcome (or kind of outcome) inevitable, or will the evolutionary history of an organism preclude some possibilities and make others more likely? Stephen Jay Gould maintained the latter—that preceding states would provide such a significant part of the context for future mutations that rewinding time to the beginning and doing the whole thing over again might lead to a completely different world of life. This position has intuitive appeal, but it could also be the case that natural selection is so powerful that certain maximally-adapted states would be achieved regardless of evolutionary history. Obviously, we can't rewind time to test these possibilities directly, but perhaps some model system could help us.

Enter E. coli and the LTEE, an experiment that procedurally sounds very simple. What Lenski did was to found 12 populations of E. coli from two clones. Once the cultures were started, his team took part of each culture every morning and diluted it into some fresh DM25 medium, a minimal medium containing 139 µM glucose and 1.7 mM citrate. Now, E. coli loves glucose, but one of the defining characteristics of the species is that it cannot take up citrate in an aerobic environment. There is citrate inside a bacterium (as an intermediate in a metabolic pathway), and there is citrate outside the bacterium, but what is outside cannot come in. And that is how things stayed for more than 30,000 generations. A number of differences in appearance and other properties have been noted, but for a very long time the ability to eat citrate did not evolve.

Eventually, shortly after the 33,000th generation, random mutations in one of the populations gave its bacteria the ability to transport the more-plentiful citrate across their membranes. Over a very few generations, the maximum density of the cultures exploded as the bacteria gained access to the more-plentiful food source. By itself, that's pretty cool, as it represents laboratory observation of a mutation that changes a defining characteristic of a species. But this new ability also made it possible to test the importance of the previous mutations.

You see, Lenski's lab had taken samples of their E. coli every 500 generations and frozen them in glycerol at -80 °C. This sounds harsh, but this kind of deep freeze allows them to resurrect these populations for future study. And that's just what Lenski's researchers did. They pulled the samples out of deep freeze and replayed evolution for ~3700 generations for each of them, testing to see whether the ability to transport citrate under aerobic conditions evolved again.

If pre-existing context is not particularly important, one would expect that there would be some low probability of evolving citrate transport that was equal for all previous generations. On the other hand, if Gould is right, then there would be essentially no chance of evolving citrate transport for early populations, and then after some potentiating mutation occurred, a higher chance for later populations. Blount et al. show that the latter is the case. In their replays, citrate transport never evolved in populations earlier than the 20,000th generation, and only appeared regularly in replay experiments after the 30,000th generation. This would suggest that some potentiating mutation occurred before generation 20,000 that provided a genetic context in which subsequent mutations could produce citrate transport.

Obviously, one suspect for this mutation would be a change in the DNA reproduction machinery that made subsequent mutations more likely in general. It's true that some of the other populations in the LTEE had enhanced mutation rates, but not this one. A subsequent experiment tracking mutations in another gene showed no difference in mutation rate between the ancestors and the potentiated clones. So the answer is more complex. Unfortunately, the authors do not yet know exactly what mutation occurred in this time frame to potentiate citrate transport. However, our ability to examine the genes of bacteria has increased significantly since the LTEE began. Full-genome sequencing of these bacteria (now underway) should give us some powerful insights into the evolutionary history of this population.

So, would evolution play out the same way if we rewound and started again? These results suggest that it would not, and historical contingency is likely to be the case for many systems. However, it may also be true that some particular characteristics are so constrained, or confer such an enormous selective advantage, that life will take these avenues no matter what. In the case of citrate transport, the existing genetic context appears to be very significant, but generalizing this observation to other systems may not be valid. Nonetheless, Gould's point is proved even if contingency is a property of some systems. The Lenski experiment shows him to be in the right, even if it took 20 years to do it.

Carl Zimmer has a post on this article as well, complete with a few intelligent questions and some crazy rantings.

1. Blount, Z.D., Borland, C.Z., Lenski, R.E. (2008). Inaugural Article: Historical contingency and the evolution of a key innovation in an experimental population of Escherichia coli. Proceedings of the National Academy of Sciences 105(23) p. 7899-7906 DOI: 10.1073/pnas.0803151105