Negative frequency-dependent selection maintains coexisting genotypes during fluctuating selection

Caroline B. Turner, Sean W. Buskirk, Katrina B. Harris, Vaughn S. Cooper

Preprint posted on August 08, 2019

Article now published in Molecular Ecology at

How do organisms adapt to changing environments? Rather than becoming generalists, a fluctuating environment selects for coexisting sub-populations.

Selected by Defne Surujon


Adaptive laboratory evolution (ALE) experiments have been widely used to study how organisms adapt to particular stresses. A lot of ALE experiments are set up such that a population is continually passaged in culture media, with a gradual increase in some stress factor (e.g. antibiotic concentration). However, even in a simple passaging protocol, the environment experienced by the organisms is subject to fluctuations, for instance in nutrient availability. Natural environments also often have inherent fluctuations, for example there are cycles of temperature in many habitats depending on the time of day or year. Organisms living in these changing habitats should therefore be able to tolerate these changing environmental conditions.

How can an evolving population adapt to tolerate fluctuations in the environment? Either members of the population become generalists, meaning they have high fitness in all the environmental conditions they encounter, or distinct sub-populations emerge and start coexisting, where each sub-population is adapted to a different condition. One main question the work by Turner et al. addresses is whether a population under fluctuating selection evolves to become a single generalist, or a collection of coexisting specialists.

The authors address the question in an evolution experiment using Burkholderia cenocepacia, a Gram-negative bacterium that is found in the soil, but can also infect cystic fibrosis patients and cause life-threatening disease in humans. B. cenocepacia, like many other bacterial pathogens, can exist planktonically (i.e. free-swimming bacterial cells), or attach to a surface and form a sessile biofilm. Biofilm-forming bacteria often cycle between the planktonic and biofilm mode of growth, where planktonic growth enables dispersion, and biofilms provide added protection for the bacteria against environmental insults such as antibiotics. The authors replicate the planktonic-biofilm cycle in their adaptive evolution experiments and compare  the types of mutations observed under fluctuating modes of growth with mutations acquired in the non-fluctuating adaptation experiments (i.e. passaging in either biofilm-only or planktonic-only modes).

Key Findings

Among the three sets of adaptive evolution experiments (biofilm only, planktonic only, biofilm-planktonic fluctuating), the biofilm-adapted and the biofilm-planktonic fluctuating populations had similar outcomes in terms of the genes that acquired adaptive mutation. Specifically, two biofilm-regulating genes, wspA (a surface receptor) and rpfR (a phosphodiesterase/guanylate cyclase),  acquired mutations in all 4 biofilm and all 4 fluctuating populations. The wspA mutations reach high frequency in the biofilm environment very rapidly in both modes of adaptation. In the fluctuating populations, a switch to planktonic growth is followed by another very quick shift in the population – the wspA mutation decreases in frequency, and rpfR mutants start becoming more abundant.

The fact that the frequencies of the two types of mutants cycle in the fluctuating adaptation suggests that the wspA and rpfR mutants are specialists in the biofilm and planktonic modes respectively. When competed against the ancestor, the wspA and rpfR mutants had a fitness advantage in biofilm and planktonic modes of growth respectively. And when pitted against each other, the rpfR could outcompete wspA in the planktonic mode, but wspA did not have a clear advantage over rpfR in biofilms. In fact, the two mutants seem equally fit in the biofilm mode. This sounds more like rpfR being a generalist, and wspA being a biofilm-specialist. How then can wspA have any possibility of reaching prevalence in the population?

The high fitness of rpfR was evaluated by mixing the rpfR and wspA mutants were in equal proportion. This is not what happens in the adaptation experiments, where the population frequencies become less balanced, especially during the biofilm mode. The wspA mutant becomes the majority very quickly, making the rpfR mutant rare. The authors show that when the starting proportions of the two mutants are varied, rpfR has a higher fitness advantage when rare, and is at a disadvantage against wspA when prevalent (this inverse relationship between prevalence and fitness is called negative frequency dependent selection, or NFDS). This can explain why the rpfR mutant is never driven into extinction in these experiments – as wspA starts taking over the population, rpfR is also becoming more likely to be successful.

Interestingly, the populations adapted to grow planktonically acquire mutations in the gene rpoC (which encodes RNA polymerase β) and no mutations in the seemingly advantageous rpfR. While the rpfR mutations do give B. cenocepacia a fitness advantage during planktonic growth, the rpoC mutation provides an even higher fitness advantage in relation to the wildtype ancestor. This is presumably why the rpfR mutant does not reach detectable frequencies during planktonic evolution, as it is outcompeted by the more fit rpoC mutant.

Why I Find this Work Exciting

The replicate populations that were evolved under the same regime (biofilm-only, planktonic-only, biofilm-planktonic alternating) not only share the same mutations, but the population frequency of these mutations over time are also highly similar. This level of parallelism in adaptive evolution experiments solidify the idea that adaptive evolution has deterministic aspects driving it. While the initial appearance of mutations may be more arbitrary, the selection of such mutations is driven by their fitness contributions, and the nature of the selective pressure. While we may not yet have a comprehensive understanding of such factors, the fact that evolution is highly replicable suggests that these deterministic factors play a major role in the outcome of adaptive evolution. And any process that is deterministic can be represented as a mathematical model. This implies that sometime in the near future adaptive evolution may be accurately simulated in silico.

Future Directions and Questions for the Authors

There are multiple mutations on the rpfR gene that appear at low frequency in the biofilm-adapted and fluctuating populations. Are these mutations linked on the same genome, or do they arise independently in distinct sub-populations? Have the investigators done whole genome sequencing on some clonal samples?

In the fluctuating population, the first transition from biofilm to planktonic growth is followed by very drastic changes in mutation frequencies, however in the second cycle the genotype frequencies don’t change as dramatically or quickly. Why might the changes be attenuated over time? And if the adaptation were to be carried out for more cycles of biofilm-planktonic transitions, would the population frequencies keep fluctuating or would the population stabilize at some point?

Would the same coexisting genotypes be observed if the fluctuating evolution experiment is started with the planktonic mode of growth rather than biofilm mode?

When a patient is being treated for an infection or prescribed antibiotics prophylactically, they will likely take doses of the antibiotic on a regular basis, resulting in pulses of the drug. Would the drug pulses also act like a cycling selective pressure that is subject to NFDS? That is, would the selection and short-term adaptation of a bacterial population in a patient behave similarly to these in vitro populations?

Tags: adaptive evolution, biofilm, microbial ecology, nfds, population genetics

Posted on: 18th September 2019


Read preprint (No Ratings Yet)

  • Author's response

    The author team shared

    Thanks for your interest in our work and for your summary highlighting the paper!

    In answer to your first question, we think that the different rpfR mutations occur independently in different sub-populations, though we can’t say definitively that multiple rpfR mutations can never occur in the same background. We did sequence a handful of clones from this experiment and never saw more than one rpfR mutation in a given clone. Another piece of evidence comes from other experiments we’ve done (Traverse et al. 2013; Turner et al. 2018) where rpfR mutations eventually rise to 100% in many populations. In those experiments we see one rpfR mutation rising to 100% and outcompeting any other rpfR mutations, indicating that the different rpfR mutations are on different genetic backgrounds.

    As you noticed, at the first transition from biofilm to planktonic wspA mutations drop dramatically in frequency, whereas the decrease is smaller at the second transition from biofilm to planktonic. We think this may be due to additional mutations being selected in bacteria with the wspA mutation that improve fitness in planktonic conditions. Whether the population fluctuations would be maintained over more switches between environments is something we also are curious about.

    What the differences might be between biofilm-planktonic-biofilm-planktonic evolved populations and planktonic-biofilm-planktonic-biofilm evolved populations is another great question that we also wonder about.

    Your point about pulses of antibiotics is a great one. It would be easy to imagine negative frequency dependent interactions occurring between sensitive and resistant bacteria and that similar dynamics might occur to those we observed in our experiment.


    Traverse, C. C., L. M. Mayo-Smith, S. R. Poltak, and V. S. Cooper. 2013. Tangled bank of experimentally evolved Burkholderia biofilms reflects selection during chronic infections. Proc. Natl. Acad. Sci. USA 110:E250-E259.

    Turner, C. B., C. W. Marshall, and V. S. Cooper. 2018. Parallel genetic adaptation across environments differing in mode of growth or resource availability. Evolution letters 2:355-367.

    Have your say

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

    This site uses Akismet to reduce spam. Learn how your comment data is processed.

    Sign up to customise the site to your preferences and to receive alerts

    Register here