Ecological interactions emerge spontaneously in an experimental study of bacterial populations cultured for 60,000 generations, and sustain rapid evolution by natural selection. See Letter p.45
At the end of On the Origin of Species, Charles Darwin describes a tangled bank1 — an ecosystem of plants, birds and insects whose evolution is inextricably linked by their dependence on one another for survival. Researchers who study evolution in the wild must contend with the tangled bank. Controlled laboratory experiments provide an alternative approach that may avoid the complexities of ecological interactions. But on page 45, Good et al.2 analyse evolution in laboratory populations of bacteria and make a surprising discovery: ecological interactions emerge spontaneously and sustain evolution by natural selection for tens of thousands of generations.
The archetypal study of evolution in the laboratory was initiated almost 30 years ago, when the biologist Richard Lenski founded 12 identical populations of the fast-growing bacterium Escherichia coli. His ongoing long-term evolution experiment (LTEE) is designed to study adaptation to a constant laboratory environment, removing as many complexities and confounders as possible. Every day, each of the 12 populations is transferred to a fresh flask of nutrients to enable continued cell growth and division. Samples from these populations are periodically frozen, to preserve a 'fossil record' for retrospective analysis.
This reductionist approach has been richly rewarding. The LTEE has produced striking discoveries, including markedly elevated mutation rates in some of the formerly identical populations, and the evolution of a radical metabolic trait in another population3. The frozen fossil record has allowed researchers to resurrect ancestral strains, replaying the tape of life to tease apart the molecular causes of these striking phenomena.
Evolution in Lenski's experiment is complex enough to yield surprises, but simple enough to be comprehensible. So far, researchers have approached the LTEE much as natural historians might — by identifying examples of remarkable phenomena and digging deep to dissect their causes. But do these phenomena reflect the typical course of evolution in LTEE populations? To answer this question, Good et al. adopt a fresh approach to analysing LTEE samples: characterize what is typical, rather than catalogue what is remarkable. Their statistical approach shows that phenomena previously observed as single, remarkable examples actually occur with regularity.
The authors analysed the dynamics of molecular evolution over 60,000 generations in the LTEE. They reconstructed the temporal trajectories of all mutations that reached a threshold frequency, by sequencing samples from each population every 500 generations. Their analysis pipeline exploits correlations across samples taken from the same population at different time points to discriminate true mutations from sequencing errors.
In a crucial advance over previous studies, Good et al. used a statistical technique called a hidden Markov model to estimate when each mutation first entered the population, and when it subsequently either became extinct or swept through the entire population. Such events cannot be observed directly because their signal is swamped by noise from sequencing errors. The upshot is a set of high-resolution frequency trajectories for nearly 35,000 mutations over time.
The trajectories inferred by Good and colleagues add rich detail to our understanding of evolution in experimental populations. The mutations detected are distributed non-uniformly across the genome and, within a gene, non-uniformly in time. Some genes acquired mutations in many of the 12 populations independently — a highly non-random phenomenon called parallelism.
Parallel mutations in a gene often occurred in the early phase of the experiment. This indicates that, as previously reported4, in many genes, mutations quickly provide their maximum possible fitness gain for the cell. After this, further mutations in the gene hold no adaptive benefit. But Good et al. also observed that some genes harboured parallel mutations that arose late in the experiment.
These late-onset genes tend to be mutated in some populations but not others, suggesting historical contingency — the idea that a seemingly innocuous early mutation can open up an avenue towards a later mutation, for instance by making it beneficial rather than harmful. Historical contingency was known3 to have a role in one metabolic innovation in the LTEE, and in other domains of life5. But Good and colleagues' findings suggest that historical contingency is commonplace. As a result, the genetic targets of natural selection change over the course of evolution.
The authors' most profound discovery is the spontaneous emergence of ecological interactions that fuel ongoing evolution (Fig. 1). Persistent subgroups have previously been identified in one of Lenski's populations6, but Good et al. reveal that at least 9 of the 12 populations divide into two separate clades (genetic groups). These clades co-exist for tens of thousands of generations, and so must be maintained by some form of interdependence.
These emergent ecologies sustain ongoing adaptation at a roughly constant pace throughout the entire experiment, even though the benefits of ongoing adaptation to the entire population have been shown to decline over time4. Mutations sweep through a clade at generation 50,000 nearly as frequently and rapidly as they sweep through a population at generation 5,000. Indeed, the characteristic pattern of multiple competing beneficial mutations7 that occurs in the early phase of Lenski's populations continues to occur in each clade throughout the experiment.
For all the statistical regularities that Good et al. reveal, the discovery of pervasive ecological interactions in the LTEE will complicate future work. Evolution and ecology were once considered separate fields, safely studied in isolation. There have been several documented cases of ecological and evolutionary changes occurring on the same timescale8, especially when infectious agents compete against host immune defences9,10. But the LTEE was conceived as an experiment in the simplest possible environment precisely to avoid such complications, and so the spontaneous emergence of ecological niches in this setting is a shocking development.
Not only can ecological and evolutionary processes co-occur, but it now seems impossible that they can be disentangled. Future work must address whether ongoing adaptation in these populations is merely sustained by multiple ecological niches, whether it is accelerated by interactions between clades, or whether evolution in these clades actively reshapes the ecological balance that has emerged.