A roadmap towards predicting evolution
Here we lay out a road map for the next 15-20 years for the game-changer Predicting evolution, one of the five game-changers in the Origins Center. This document is primarily aimed at the scientists that work in the fields that contribute to predicting evolution and is meant to position the ongoing research in the Netherlands and to guide our joint efforts to work on the game-changer
Predicting evolution hinges on bridging the multi-scale levels: of biological organisation (from molecule to phenotype), of connecting ecological to evolutionary dynamics (from phenotypes to populations and communities), and of linking large scale abiotic changes to the biological players. Recent progress and developments in characterizing these complex mechanisms and processes bring the goal within reach of predicting the trajectories of evolutionary change. Predicting these trajectories can be done at the DNA sequence level in the form of changes in gene frequencies, or at the level of phenotypes, in the form of changes in the phenotypic trait distribution; ultimately, these levels should be connected.
This roadmap is constructed across six interconnected building blocks that jointly address the key elements that are needed to be able to predict evolution. The logic of these six blocks is to (1) find out how the genetic information is translated to the phenotype – i.e. the genotype-phenotype map – in interaction with the environment, (2) understand why evolution has shaped the genotype-phenotype map the way it is, taking into account past selection pressures and constraints, (3) assess the current and future genetic variation for selection to act on, (4) determine how ecological drivers of selection lead to adaptive phenotypes, also taking into account non-selective processes that determine evolution and eco-evolutionary feedbacks, (5) integrate this information to predict evolutionary change and (6) develop ways in which we can quantify prediction accuracy.,/p>
Combining all these blocks will allow us to determine/quantify the predictability of evolution across levels, systems and time periods. When is evolution predictable, and when not? Although there is some form of hierarchy in these six blocks, there are also ‘feed-backs’. We thus need to work on these interconnected blocks simultaneously.
1. How does genetic information specify the phenotype (i.e. the genotype-phenotype map)? What is the architecture of the collective biological machinery that leads to the integrated phenotype?
This includes genomic features such as gene networks, epigenetic modifications, regulatory elements, as well as protein interactions, metabolic and physiological processes, ultimately leading to morphological, physiological, behavioural and life history traits in the integrated phenotype. Computational evolutionary modeling studies, which allow multilevel G-P maps to evolve, can give insights in generic, and therefore predictable, trends in both short- and long-term evolution.
Different environmental conditions can result in the same genetic code being used in different gene networks, as different parts of the genome may be activated under certain environmental conditions but not under others, for example through epigenetic modifications. This influence of the environment can lead to phenotypic plasticity of traits, or in other cases, to canalized expression of traits as these are buffered against perturbations by environmental conditions. All of these mechanisms have been well-documented in a wide variety of biological systems. Now, we need to develop a comprehensive framework to understand when which strategy is chosen in nature and how this affects the G-P map. Also, we must establish how the relative contribution of these mechanisms affects the nature and speed of evolutionary change – and thereby improve our ability to predict evolution.
We clearly see collaboration here with Game-changer 3 which focuses on the interactions between the different levels of biological organization (Systems biology).
2. How is the genotype-phenotype map shaped by evolution?
Over the course of evolution, properties of the G-P map are expected to evolve. Selection pressures may shape levels of gene regulation, robustness of gene networks, and how these translate into the associated phenotype. Furthermore, physical-chemical and developmental constraints may limit the possible realization of the genotype-phenotype space (the collection of all possible outcomes of the genotype-phenotype map), thereby also shaping the G-P map.
Understanding how the G-P map has evolved under different conditions in historical times may enable the reverse, namely predicting evolution under comparable conditions in the future. Particularly when we zoom out to higher levels of biological organization, convergent evolution of phenotypes is frequently observed, but is it also possible to predict the adaptive G-P map that evolution will select for, based on these insights from convergent evolution?
Symbiotic interactions and hybridization can also drastically alter the G-P map, driving loss and acquirement of genes, genome rearrangements, horizontal gene transfer, etc. How common are key innovations in the evolution of life, do they arise constantly or are they rarely occurring new avenues for evolutionary radiation? Thus, can we predict from the observed patterns of past evolutionary change the response to novel selection pressures?
3. What determines current variation in the genotype-phenotype map?
Heritable variation is essential for evolution, thus for the prediction of evolution a thorough understanding of the extent of existent (epi)genetic and phenotypic variation is needed.
Variation at the genotypic level can take the form of SNPs, copy number variation, gene duplications, inversions, etc. Taken together, this is usually referred to as standing genetic variation. New genetic variation in the genome can be generated by mutations, introgression, genome duplication, and horizontal gene transfer. Are there hotspots of genetic variation in the genome, and if so, do we understand why they exist and in which regions of the genome they are likely to occur? What is the relative contribution of novel genetic variation compared to standing variation along the evolutionary trajectory of species?
At the phenotypic level, we can look at quantitative variation in traits, such as additive genetic variation, permanent environments effects, etc. One of the challenges will be to reconcile this quantitative genetic approach with the molecular genetic and genomic approaches to study variation. This is particularly important to comprehend the effect of the environment on heritable variation at the level of the phenotype (GxE interaction)? How is this linked to standing genetic variation, given that there may be ‘hidden’ variation that is only expressed under certain environmental conditions? Does this affect our ability to predict evolution across different environments?
4. How does selection operate on variation in the genotype-phenotype map?
To predict evolution we need to be able to predict how selection pressures will lead to fitness differences among phenotypes. As phenotypes are part of the G-P map (see 1) and there is variation in the G-P map (see 3), selection on phenotypes will favour the subset of G-P maps sharing the phenotype with the highest fitness under these conditions.
Constructed fitness landscapes can be informative for predicting selective pressures. In empirical work, including experimental evolution studies, simple low dimensional fitness landscapes are often used as it is near impossible to empirically construct high dimensional fitness landscapes.
Multidimensionality may however change the contours of the landscape and may reveal alternative routes to fitness peaks. Theoretical studies should explore the evolution of the dimensionality of the G-P maps, and how this influences the evolutionary dynamics, and its predictability at various timescales.
5. How will ecological drivers of selection lead to evolution?
Ecological drivers of selection can be both abiotic and biotic variables. For the latter category, evolution in one species can lead to a change in selection pressures in other species. To understand the abiotic environment as a driver of selection and how these selection pressures will change in the future, we need to include large-scale environmental changes, such as climate change, as well as extreme events (catastrophes). Multidisciplinary research is key to this part of the road map as predicting the abiotic changes of the Earth requires knowledge from the geosciences, climate sciences and astronomical effects.
Selection is only one of the forces that are involved in evolution, so to predict evolution we also need to take other forces into account. Importantly, other population characteristics including population size, fragmentation, and genetic drift, as well as biological features such as generation time, sexual selection can put constraints on the rate and direction of evolution. Focusing on how constraints on evolution limit the range of possible outcomes, provides new opportunities to improve our ability to predict evolution.
6. How do we quantify prediction accuracy?
Validating evolutionary predictions is difficult particularly at longer timescales. At the shorter timescales, we can test the accuracy of our predictions using empirical work. Quantitative genetics has been relatively successful in predicting changes in phenotypes in response to natural selection, using for instance the multivariate breeders equation, but how this feeds back into changes at the molecular level is unknown. For the medium timescales (10-20 years) we can already rely on stored DNA of long-term field samples and phenotypic trait information or museum specimens or collected by the general public via so-called “citizen science”.
Over longer timescales, paleo-data and whole genome phylogenetic studies can be used to hindcast evolution – in a way forecasting the past- and see if the predicted patterns match observed patterns. Also abiotic conditions need to be hindcast over these timescales, which requires paleoclimatology for example by using fossil pollen or lipid biomarkers from sediment cores.
An open question is where the limitations are in the predictability of evolution. It may well be that evolution is predictable at the very short and long timescales but not at intermediate time scales. So can we predict the outcome of evolution of life on Earth at a 100, a 1000 and 10000 years? The ultimate test would be to predict what life would look like on other places than Earth (linking to Game-changer 4).