1,2 beacon center for the Study of Evolution in Action, Michigan State University, East Lansing, mi, usa

FIGURES

Figure 1. The E. coli Long-Term Evolution Experiment (LTEE).

The LTEE involves a simple serial transfer regime (top). 1% of each population is transferred to a fresh volume of growth medium DM25 every 24 hours, after which the populations are incubated for another 24 hours. Each population experiences approximately 6.6 generations per day. The experiment includes twelve, initially identical populations that have been evolved in parallel, with viable samples of each frozen every 500 generations (bottom).

Figure 2. Expansion of the Ara–3 Population.

As Cit⁺ variants rose to high frequency, the size of the Ara–3 population expanded several-fold due to the high concentration of citrate in the DM25 growth medium. The expansion caused the population to become much more turbid after twenty-four hours of incubation. (Figure adapted from Blount et al. 2012)

Figure 3. Phylogeny of the Ara–3 Population.

Population phylogeny based on sequenced clonal genomes. Colored lines and shaded areas correspond to major identified clades. (Figure adapted from Blount et al. 2012)

Figure 4. Cit⁺ Actualization.

In the Cit^– ancestor (top), the citT gene is silent when oxygen is present. In Cit⁺ genomes (bottom), a tandem duplication places a new copy of citT adjacent to and under the control of the promoter that normally controls the aerobically expressed rnk gene. The new rnk-citT module supports the expression of the CitT citrate-succinate transporter protein when oxygen is present, providing access to the citrate resource. (Figure adapted from Blount et al. 2012.)

1^ In microbiology, a clone is a population grown from a single founding cell. Because of asexual reproduction, this population will be more or less genetically identical, meaning that we can study its properties to understand the properties of the founding cell. Due to this genetic homogeneity, we also tend to use “clone” interchangeably with “genotype.”

2^ In essence, parallel replay experiments look at whether or not populations starting from the same initial evolutionary state then take parallel evolutionary paths when evolving under the same conditions. Another class of experiments, “historical difference experiments,” is designed to examine whether or not populations that start from different evolutionary states follow convergent paths to a common evolutionary endpoint during evolution in the same conditions (Travisano et al. 1995a, Blount in press; Spor et al. 2014).

3^ Citrate was not added to DM25 to provide an ecological opportunity, but has always been an ingredient of DM-type media (Davis and Mingioli 1950). DM was originally formulated by the bacteriologist Bernard Davis when he was developing a method of isolating mutants of E. coli that exploits the fact that penicillin only kills actively growing cells (Davis 1949). When penicillin is added to a culture of bacteria growing in a minimal medium the only cells that are not killed are those that are not actively growing because they have mutations that prevent them from synthesizing all nutrients not provided by the medium, and therefore are not actively growing. Davis found citrate increased the killing efficiency of penicillin, so he included it in his standard recipe. Davis’s penicillin method of mutant isolation came into wide use in the 1950s, which in turn led to DM becoming a widely used minimal growth medium for other applications.

We now know that the citrate in DM improves the availability of iron for E. coli. At neutral pH with oxygen present, iron occurs primarily as an insoluble ferric ion. Citrate binds to ferric ions to form a stable, soluble complex called ferric dicitrate. E. coli has a transport system that specifically allows it to bind ferric dicitrate, and take up the iron, though not the citrate, from it (Cox et al. 1970; Frost and Rosenberg 1973; Hussein, Hantke, and Braun 1981; Härle et al. 1995).

E. coli requires a tiny amount of citrate to take up iron, as little as 10 μM. DM25 contains 1700 μM. Why the excess? First, DM was designed specifically by Davis (1949) for his penicillin method of isolating mutants, which he knew worked better with citrate. Because E. coli was not known to use citrate at all, Davis likely did not think it was terribly important to find the lowest effective citrate level. Second, citrate’s role in iron acquisition was not discovered until the 1970s (Frost and Rosenberg 1973). In short, DM was developed for a particular experimental method long before the biological role of citrate in the medium had been discovered and the concentrations strictly necessary for it to fulfill that role had been determined, and its recipe reflects this. The high concentration of citrate in DM25 and the large ecological opportunity it presents in the LTEE was therefore contingent upon its unique history combined with conservative lab practice.

4^ This is to say that there may be alternate mutations that carry the same adaptive benefit, but that do not simultaneously exist on the pathway leading to the historically contingent trait.

5^ The Cit⁺ mutants recovered during the replay experiments show that there were actually many possible actualizing mutations. All of them have mutations that involve citT, and most clearly place copies of citT under the control of new promoters, but the actual mutations vary considerably. Many are tandem duplications similar to that found in the original Cit⁺ variant, and more of these form are variants of the rnk-citT module, but several place citT under the control of other promoters. Could these alternate actualizing mutations have changed what evolutionary paths Cit⁺ evolution might have taken? Good question. I don’t know. (Yet.)

6^ Or perhaps it has experienced a small one. The Cit^– lineage that persisted in the Ara-3 population after Cit⁺ became dominant went extinct between 43,500 and 44,000 generations. Careful study by Turner et al. (2015) showed that the extinction was not a deterministic matter of competitive exclusion of Cit^– subsequent to niche invasion by a Cit⁺ subpopulation. Cit^– clones from just prior to extinction were able to reinvade the population even thousands of generations later, showing that the Cit^– niche was still open. Moreover, when replicates of the population started from samples taken prior to Cit^– extinction were evolved for five hundred generations, no instances of extinction were observed. These findings led to the conclusion that the extinction was caused by an unknown environmental perturbation to which the Cit^– subpopulation was more sensitive, the LTEE version of a chance cataclysm. Not only does this represent another instance in which contingency has played a role in the Cit⁺ story, its impact is ongoing. Cit⁺ evolution in Ara-3 after the extinction takes place in an ecosystem substantially altered by loss of Cit^–. How contingent might its subsequent evolutionary trajectory be upon the extinction event? This is a question we will be able to assess in the future. One of the extinction replay experiments, founded from a population sample frozen at 43,000 generations, was chosen to be continued indefinitely as the thirteenth LTEE population, Ara-7. It will permit comparison of how Cit⁺ has evolved under conditions in which the Cit^– extinction did or did not take place. Stay tuned.

7^ What if the Cit⁺ function were to evolve in another of the LTEE populations? Would this mean that Cit⁺ is not a contingent trait? I don’t think so. My work pretty well demonstrates that the trait’s evolution was dependent upon the occurrence of the potentiating mutations, which in turn means that it was contingent upon the occurrence of an evolutionary history that included such mutations. The recurrence of such a history in another population would not negate the trait’s contingency. Speaking in terms of path dependence, such a recurrence would still argue that Cit⁺ is path dependent, but that there was not a single path upon which it was dependent. This is an important point—evolutionary path dependence, and thereby historical contingency, exists on a continuum ranging from those outcomes that may be reached via one or a few paths, to those that are path-insensitive and guaranteed.

8^ This concept is similar that of the “genetic neighborhood,” which deals with a genotype and the local region of genotypic space that contains genotypes one or two steps removed from it (Lewontin 1974; Wilke et al. 2001; Covert et al. 2013). The two concepts are related, but my idea with variation space is that is that all points in the genotypic landscape are reachable at a certain probability determined by the frequency of the necessary mutation or mutations. Variation space thus deals with how the accessibility of different points in genotypic space varies.