Evolution of evolution

A couple of years ago I learned that Guillaume Beslon and Paulien Hogeweg were collaborating under the umbrella of the EvoEvo project (link). Naturally, that caught my attention. Now the project is finished and I have selected some juicy bits. Enjoy the long version or skip to the take-home messages.

The central idea of EvoEvo is that the evolutionary process is not a static entity. For clarity, since Darwin we take an evolutionary process to be founded on two main components, variation and selection. Reproduction with mutation generates variation, while selection weeds out the individuals that are ill-adapted to the current environment. The key is that these two processes are themselves also the product of evolution. In other words, evolution evolves and EvoEvo studied this evolvability of biological systems.

EvoEvo or evolving coding structure

EvoEvo focused on three topics, woven together in different models of in silico evolution:

the mutational operators leading to genome variability,
the relation between genotypic and phenotypic variability,
environmental influences on genotypic and phenotypic variability.

The notion of coding structure is central to these topics. EvoEvo aptly defines it as genome structure plus the genotype to phenotype mapping. The main topic of EvoEvo may thus be formulated as the evolution of coding structure.

Below I discuss the role of big mutations in shaping genomes, what fitness landscapes evolve under various coding structures and environments, and I present the idea of evolutionary robustness.

The impact of big mutations on (genome) variability

Traditionally point mutations are considered the main source of variation. However, in bacteria (and yeast?) larger scale mutational events are at least equally common. Examples of such events are horizontal gene transfer, gene insertion and deletion, and large chromosomal rearrangement (LCR). And these big mutations do not have the same effects as point mutations.

For a simple reason of interest, I restrict myself to large chromosomal rearrangements. In the models studied by EvoEvo, when populations encounter a new environment, they tend to experience a (severe) drop in fitness. With low fitness LCRs fuel genome expansion, followed by deletion events over a long period of time, named streamlining. Plenty of fitness gain happens during this epoch of genome reduction. Contrary to my intuition, they find that genome expansion need not be a runaway process. If we assume that not only frequency, but also size of LCRs scale with genome size, genome size is limited even without selection. Note: the largest LCR is in fact limited by the largest chromosome.

All in all, EvoEvo finds that a chromosomal rearrangement is a powerful mutational operator. By itself, it is capable of generating high fitness, though adding deletions and point mutations still makes evolution proceed more efficiently. On the other hand, only point mutations are far less efficient!

Playing the devil’s advocate, they give point mutations a major boost by elevating mutation rates. In this scenario, EvoEvo finds that the consequences of an increased point mutation rate depend on the genome structure already present. Virus-like genomes suffer under these high mutation rates, while E.&nbspcoli-like genomes inflate non-coding parts of the genome.

The key insight I gained is that distinct from point mutations, LCRs play with both genome content and mutation rate at the same time. As I wrote above, environmental change induces a pressure to adapt, which leads to genome expansion and thus to higher mutation rates. These allow for an increased sampling of the mutational neighbourhood, thus increasing the probability of finding a favourable phenotype for the new environment.

What fitness landscape do we get?

The fitness landscape is an old metaphor that visualizes evolution as a population of individuals hill-climbing a peak in the landscape. The peak has high fitness, the valleys low fitness, and the geographic coordinates represent genotypes. A focal point of inquiry in biology is the mapping from genotype to fitness, via the phenotype: the genotype-phenotype map, or GP map. In the EvoEvo project, it is concluded that the complexity of the GP map is influenced by both the complexity of the environment and by the variability at the genome level. What does this mean?

First, in the Virtual Cell model a U-shaped fitness landscape evolves, combining the best of both worlds: neutrality and high selection. There are many almost-neutral mutational neighbours and many near-lethal neighbours. Slightly deleterious mutations seldom occur — which runs counter to Muller’s ratchet. This is the idea that slightly deleterious mutations can accumulate in a population without being weeded out by natural selection, leading to a degeneration of the population.

Second, in a scenario of evolving at high mutation rates, a population of in silico RNA replicators ends up in a region of the GP map, where minimal variations from a master genotype correspond to large phenotypic differences. The mutants are found to have new functional roles. They either help the replication of the master sequence, or they hinder reproduction of its competitors. At lower mutation rates, different lineages of RNA replicators evolved. They are treated as species, since they have high genotypic, phenotypic, and functional differentiation. In this case, the GP map of the various species varies from steep to flat.

From these different studies, it becomes clear that variability is a complex trait. EvoEvo finds that multiple measurements are relevant. One needs to distinguish between genotypic and phenotypic variability. Moreover, these should be expressed not only in terms of frequency distributions, but also in terms of phylogenetic depth, mutational distance, and functional differentiation.

Yet another type of robustness: the evolutionary one

Robustness means “Keep calm and carry on”: a biological process is not disturbed by a perturbation. The concept pervades biology and is defined in many ways. It relates to different organizational levels and time scales, as well as relative to a range of disturbances. Here EvoEvo considers robustness at the individual level, the population level, the lineage level, and the ecosystem level.

Evolutionary robustness is the new kid on the block. It captures the persistence of a biological system over evolutionary time, in other words, that it does not evolve itself to extinction. EvoEvo finds that spatial self-organization and multilevel selection are powerful mechanisms to reverse, or at least contain, self-destructive evolutionary trends. Space is known to stabilize and even promote variability in populations and ecosystems. And selection for spatial patterns may be opposite of maximizing an individual’s fitness, which can prevent evolution toward extinction of individual lineages. In other words, multiple levels of selection can automatically emerge through spatial pattern formation. The robustness of the ecosystem resulting from self-organization, allows the evolution of novel genetic mechanisms for robustness.

The key message is that persistence of genes, individuals, populations and spatial patterns are intrinsically linked. They should be considered together.

TL;DR

Finally, EvoEvo summarized in their own words:

Variable genome size and structure enable evolving systems to adjust their evolutionary dynamics to the environmental conditions.
Evolution tunes the ratio of different mutational operators indirectly by evolution of genome size, operon structure, and intergenic regions.
Genomes of variable size can self-regulate their size if we use a per-base mutation rate. Large Chromosomal Rearrangements (LCRs) and InDels are particularly effective owing to their multiplicative effect on genome size.
LCR and InDels provide alternative evolutionary paths and help regulating the mutational neighborhood by linking it to non-coding sequences and genome structure. For contrast, point mutations and crossover only impact the coding sequence.
Individuals can escape local optima by evolving their fitness landscape, hence opening alternative paths toward higher peaks.
Spatial and temporal structure (e.g. evolving on a grid or evolving in cyclic environments) enables stabilization of the population structure, preventing population collapse due to parasites. In other words: evolutionary robustness.
In many situations, evolution of evolvability is linked to evolution of complexity. Indeed, the emergence of new structures adds new levels of selections that interact with the other “lower” ones thereby opening new evolutionary directions.

References

The original EvoEvo reports can be found here. I’ve discussed Work Package 3.