Video: Visualizing coevolution in dynamic fitness landscapes

Written by: Bjørn Østman

Primary Source: Pleiotropy

The fitness landscape is the framework for thinking about evolutionary processes the same way the phylogenetic tree is how we think about evolutionary history. It can guide our thinking and even enable us to predict outcomes of evolution.

Fitness landscapes are usually depicted and thought of as static, i.e., not changing in time or space, but in reality they change in response to environmental changes. Populations have different fitness in different environments, so changes in both time and space can influence the fitness landscape. For example, releasing chicken on the moon will drastically change their chances to reproduce.

Many papers have been published about fitness landscapes, but with very few exceptions they investigate static fitness landscapes. Exceptions are landscapes that change between two or three different environmental conditions, such as microbes in salty or acidic conditions.

A consistent criticism of studies that look at evolutionary dynamics – the study of evolving populations – is that the fitness landscape is static, and that this is not realistic. But no one knows to what extent natural fitness landscapes change over time. Both the frequency and magnitude of such changes are completely unknown. On the time-scale of significant evolutionary change, do real fitness landscapes experience changes that make any serious difference to how populations evolve? Do they change qualitatively, with peaks coming in and out of existence? Or are the changes merely quantitative, keeping the rank order of fitnesses the same? The former is a possible solution to the problem of how populations can cross valleys between peaks in the fitness landscape: if a population is stuck on a local peak, just wait until the environment changes and leaves an uphill path to new genotypes and phenotypes. But it could very well be that in most cases most of the time populations are stuck in an approximately static landscape. We really don’t know.

And yet, for all the criticism of studies of static landscapes, not much research has been done on evolution in dynamic fitness landscapes.

One environmental factor that can change the fitness landscape of a population is a population of another species. If one species is in any way dependent on another, then there is a potential for the fitness landscape to depend on the other species.

In the video above we present three such cases of coevolution. (Details of the simulations.)

Moth-orchid coevolution. The moth eats nectar from the bottom of the orchid spur. In order to do that, its proboscis needs to be at least as long the orchid’s spur. In this model, the moth therefore gains some fitness if this is true. The more orchids it can feed on, up to a limit, the more fitness it gains. The orchids have a different agenda. They need to get someone to transport pollen from plant to plant so they can be fertilized. The moths can do this for them: when a moth sucks nectar, it touches the male flower parts and some pollen is deposited on the moth, which it carries to the next orchid, where some pollen is deposited on the female flower parts. However, if the moth’s proboscis is longer than the spur, then the moth can suck nectar without coming into touch with pollen. As a result, orchids gain some fitness if their spurs are longer than some or all of the moth’s proboscises. The orchids therefore affect the fitness landscape of the moths, and the moths affect the fitness landscape of the orchids, driving both of them to have longer and longer proboscises/spurs. We visualize this in a two-dimensional phenotype-fitness landscape, where one axis is the proboscis length in the moth landscape (spur length in the orchid landscape), and the other axis is some arbitrary neutral trait that does not affect fitness.

Rock-paper-scissors. The second dynamic fitness landscape is the familiar rock-paper-scissors system. The phenotypes consist of two arbitrary traits, and the three populations are evolving in sympatry, meaning there is no spatial component in the model. Each of the three populations dominate over one of the other two and is inferior to the third. In this model that means that if an organism has the same phenotype as the some members of the population it dominates, then it gains some fitness. The more individual members it has the same phenotype as, the more fitness it gains (density-dependence). Consequently, if this organism has the same phenotype as a member of the population that it is inferior to, then it loses fitness. This system makes the fitness landscape of each population very dynamic, with peaks and valleys appearing and disappearing over time.
Q: Are there any real systems that work like this?

Host-parasite coevolution. The third dynamic fitness landscape is a system with two populations, where the host loses fitness when it shares a phenotype with parasites, and the parasites gain fitness when their phenotypes are the same. The host organism therefore benefits from being different from the parasite, and the parasite benefits from being similar. This results in a situation where the host population evolves away from the parasite phenotype, and the parasite’s population evolves towards the host phenotype. However, it often happens that the parasite population causes the host population to split into two or more subpopulations centered around dissimilar phenotypes. The parasite population will then evolve to climb only one of those peaks, as is always the case when a population of competing organisms is facing two or more peaks. Climbing that peak will cause the host organisms that make up that peak to die out. As a result, the peak disappears, and the parasite population now finds itself dislocated from the surviving host population. Both the host and the parasite populations now have uniform fitness, and they consequently undergo neutral evolution and drifts about in phenotype space. In order to prevent this situation, we have given the parasite population a per-trait mutation rate that is twice as high as the host population. This makes it much less likely that the hosts can escape, because the parasites can now explore a larger area of phenotype space than the host. They move faster around the fitness landscape.

The last model results in two populations that continue to evolve indefinitely. Given enough time they will explore the whole fitness landscape, obtaining all the possible phenotypes. This is arguably open-ended evolution, in that evolution keeps going and populations do not encounter a stopping point. A definition of open-ended evolution requires that the population never reaches a stable phenotype, which in this case it does not. OEE can also be defined to require that new adaptive traits keep appearing, in which case this coevolving system does not qualify. New traits values keep appearing, but after a while they will not be novel, as they will have been attained and then lost in the past.

Some conclusionary words
While these movies are based on actual simulations of a model with two traits, we haven’t really done any science to speak of. Nothing has been measured and no hypotheses have been tested. However, the visualizations could be used as a tool for hypothesis testing and discovery. We can think of videos just as a modern version of the Cartesian coordinate system that enables us to visualize a temporal component (or another spatial component). When populations are seen evolving right in front of your eyes, we can sometimes observe effects that weren’t apparent by any other means.

More about fitness landscapes
Using fitness landscapes to visualize evolution in action
Evolution 101: Fitness Landscapes
Smooth and rugged fitness landscapes
Crossing valleys in fitness landscapes
BEACON Researchers at Work: Holey Fitness Landscapes – See more at: http://pleiotropy.fieldofscience.com/2014/06/video-visualizing-coevolution-in.html#sthash.OKXWBTnx.dpuf

The following two tabs change content below.
Bjørn Østman
Bjørn Østman is an evolutionary biologist postdoc working in the BEACON Center for the Study of Evolution in Action.
I am interested in many aspects of evolution. I work in computational biology, using various approaches to learn about fundamental processes of evolution. Bioinformatics is good for learning about real genes (data generously supplied by other researchers), and simulations are good for testing the mechanisms of evolution. I am particularly interested in how populations and organisms adapt to changing environments, both at the genetic and phenotypic level. Lately my research has focused on the evolutionary dynamics of populations evolving in rugged fitness landscapes.
Bjørn Østman

Latest posts by Bjørn Østman (see all)