Historical contingency in the evolution of E. coli

Blogging on Peer-Reviewed Research

While I was traveling last week, an important paper came out on evolution in E. coli, describing the work of Blount, Borland, and Lenski on the appearance of novel traits in an experimental population of bacteria. I thought everyone would have covered this story by the time I got back, but there hasn't been a lot of information in the blogosphere yet. Some of the stories get the emphasis wrong, claiming that this is all about the rapid acquisition of complex traits, while the creationists are making a complete hash of the story. Carl Zimmer gets it right, of course, and he has the advantage of having just published a book(amzn/b&n/abe/pwll) on the subject, with some excellent discussion of Lenski's work.

The key phrase is right there at the beginning of the title: historical contingency. This paper is all about how accidents in the genetics of a population can shape its future evolutionary trajectory. It is describing how a new capability that requires some complex novelties can evolve, and it is saying plainly that in this case it is not by the fortuitous simultaneous appearance of a set of mutations, but is conditional on the genetic background of the population. That is, two populations may be roughly equivalent in fitness and phenotype, but the presence of (probably) neutral mutations in one may enable other changes that predispose it to particular patterns of change.

Here, read the abstract for yourself, paying special attention to the parts I've highlighted.

The role of historical contingency in evolution has been much debated, but rarely tested. Twelve initially identical populations of Escherichia coli were founded in 1988 to investigate this issue. They have since evolved in a glucose-limited medium that also contains citrate, which E. coli cannot use as a carbon source under oxic conditions. No population evolved the capacity to exploit citrate for >30,000 generations, although each population tested billions of mutations. A citrate-using (Cit(+)) variant finally evolved in one population by 31,500 generations, causing an increase in population size and diversity. The long-delayed and unique evolution of this function might indicate the involvement of some extremely rare mutation. Alternately, it may involve an ordinary mutation, but one whose physical occurrence or phenotypic expression is contingent on prior mutations in that population. We tested these hypotheses in experiments that "replayed" evolution from different points in that population's history. We observed no Cit(+) mutants among 8.4 x 1012 ancestral cells, nor among 9 x 1012 cells from 60 clones sampled in the first 15,000 generations. However, we observed a significantly greater tendency for later clones to evolve Cit(+), indicating that some potentiating mutation arose by 20,000 generations. This potentiating change increased the mutation rate to Cit(+) but did not cause generalized hypermutability. Thus, the evolution of this phenotype was contingent on the particular history of that population. More generally, we suggest that historical contingency is especially important when it facilitates the evolution of key innovations that are not easily evolved by gradual, cumulative selection.

What Blount et al. are doing is testing SJ Gould's old claim that if we replayed the tape of life, we would not get the same results each time. Each step in evolution is dependent on prior history — it is contingent — and since many of the steps are driven by chance yet unfiltered by selection, we cannot predict the direction of evolution.

We can't rewind the whole planet, but with careful design, we can set up populations that can be rewound. Lenski has done this by setting aside 12 separate populations of E. coli 20 years ago, each one evolving independently and in its own direction. So far, over 44,000 generations have passed in the flasks in Lenski's lab. This is a long time, and at the typical mutation rates present in these creatures, it means that every nucleotide has been mutated singly multiple times in the population — in other words, there has been ample time to thoroughly explore the single substitution search space. In addition, a sample of each population was taken and frozen every 500 generations, so they can go back in time at will and examine their genome or even restart the line. Imagine what we could learn if some ambiguously benevolent space aliens had visited the earth every 5-10,000 years, snatched up a couple of random hominin/primate tribes, and had them tucked away in cryogenic storage — that's what this experiment is like.

These bacteria have been raised in a constant environment, one which is somewhat less than ideal: they've been fed on small quantities of glucose, and nothing but glucose, in a lean regimen that has encouraged selection for somewhat different properties than you'll find in your gut, one of the normal habitats of E. coli. They have evolved, and even have distinctive morphological characters, and many of their properties are consistent from population to population. There is one property that would be useful for the bacteria, but that has evolved in only one of the 12 populations: the ability to use citrate as a carbon source. There's plenty of citrate in the medium, and it would be a bit of a coup for any bacterium to acquire the ability to take up and metabolize it, but it just hasn't happened as often as might be hoped…except in one of the 12 populations, which around the 33,000th generation, suddenly expanded its stable population size by exploiting citrate in its environment.

How did that happen? As the abstract states, they were testing two alternatives. In one, the new ability is purely the product of an extremely rare mutation, some unlikely combination of events that gave a fortunate individual in this population the ability to take up and use citrate. If this were the case, and we rewound the tape of E. coli history back to before the mutation arose, and allowed it to play forward again, we'd expect no enhanced likelihood of a repeat performance — it's just like the other 11 populations. The other alternative is that the population had some prior enabling characteristic, some quirk in its genome that didn't really affect survival in one way or another, but that, in combination with some other ordinary mutation of ordinary probability, could predispose the population to acquire the useful citrate characteristic. In this case, rewinding the tape of life back to before the appearance of the ability, and re-running it forward, would show an increased frequency of reappearance of the ability. Furthermore, by running the tape back further still, they can identify when the enabling change in the population first arose.

The citrate+ trait was first observed in the population called Ara-3 at roughly generation 33,000. By looking back at the frozen populations, they determined that the initial mutation that enabled growth on citrate actually appeared sometime between generation 31,000 and generation 35,000. These early generations were not as efficient at growing on citrate, so another mutation is thought to have occurred around generation 33,000 that allowed much more rapid growth. E. coli from generations prior to 31,000 had no significant, detectable ability to grow on citrate.

So they pushed it back further, by taking samples from earlier generations and allowing them to replicate again, replaying history. If the citrate mutation was a rare, unique mutation, they wouldn't expect to see the novel trait arise again. What they saw, though, was that the bacteria sampled after the 20,000th generation re-evolved the citrate capability with a greater frequency — there is something that arose around generation 20,000 in the Ara-3 population that did not make them citrate+, but did make it easier for subsequent generations to evolve citrate+, confirming their hypothesis of a historical contingency.

This is the lesson: the likelihood of certain mutations arising is strongly affected by historical contingencies — different populations will have different probabilities of producing a particular trait. There were at least 3 events in the history of this one population of E. coli that enabled growth on citrate. The first was an enabling variation at around generation 20,000; the second was an initial mutation that actually allowed slow citrate uptake at around generation 31,000; and the third was a refinement at generation 33,000 that made the bacteria grow much better on citrate. Note: 3 mutations had to occur to produce the visibly better growing citrate+ population.

The creationists are already leaping all over this result and garbling and twisting it hopelessly. Michael Behe was quick to claim vindication, saying that these results support his interpretation.

I think the results fit a lot more easily into the viewpoint of The Edge of Evolution. One of the major points of the book was that if only one mutation is needed to confer some ability, then Darwinian evolution has little problem finding it. But if more than one is needed, the probability of getting all the right ones grows exponentially worse. "If two mutations have to occur before there is a net beneficial effect — if an intermediate state is harmful, or less fit than the starting state — then there is already a big evolutionary problem." And what if more than two are needed? The task quickly gets out of reach of random mutation.

Wait a minute — has he read the paper? This is an experiment that revealed a trait that required at least three mutations. Yet there it is, produced by natural evolution, with no intelligent design required; and when the experiment is re-run with populations that had the initial enabling variant, they re-evolved the ability multiple times. It seems to me that this work demonstrates that drift, chance, historical contingency, and selection are sufficient to overcome his "big evolutionary problem", and directly refute the premise of his book.

If the development of many of the features of the cell required multiple mutations during the course of evolution, then the cell is beyond Darwinian explanation. I show in The Edge of Evolution that it is very reasonable to conclude they did.

This is simply baffling. Behe claims that he has shown in his book that the result observed by Lenski and colleagues could not occur without intelligent intervention…yet it did. He is trying to argue that an experiment that showed evolution in a test tube did not show evolution in a test tube. Behe's claims are comparable to someone living after the time of Kepler and Newton trying to claim that because Copernican circular orbits don't fit the data cleanly, the earth must be stationary — in response to research that shows the earth is moving. That is how backward Behe's claims are.

Behe is a bad note to end on, so let's look at the paper's conclusion. The answer does not lie in an imaginary designer, but in the reality of historical variation. And this is a lovely discovery.

…our study shows that historical contingency can have a profound and lasting impact under the simplest, and thus most stringent, conditions in which initially identical populations evolve in identical environments. Even from so simple a beginning, small happenstances of history may lead populations along different evolutionary paths. A potentiated cell took the one less traveled by, and that has made all the difference.

(Crossposted to Pharyngula)


Blount ZD, Borland CZ, Lenski RE (2008) Historical contingency and the evolution of a key innovation in an experimental population of Escherichia coli. Proc Natl Acad Sci U S A 105(23):7899-7906.