Navigating adaptive landscapes
What is the topography of the adaptive landscapes on which evolution takes place, and to what extent are such landscapes navigable by evolving populations? These are some of the most fundamental questions in evolutionary biology.
Navigare necesse est, livere non necesse
To navigate is necessary; to live is not necessary. According to Plutarch, this is what Pompey the Great shouted to the sailors responsible for bringing grain to Rome when they were frightened to face the waters during a threatening storm. Later, others have taken this epigram to capture the human aspiration to go beyond the bare minimum for survival and to aim higher: To strive for creation, break barriers, and eagerly go into the unknown. To us, this sentence also encapsulates an intrinsic property of living systems: Evolution. Living systems are constantly adapting and innovating, navigating new waters, finding a way on the surface of their adaptive landscapes. What is the topography of these landscapes on which evolution takes place? To what extent are they navigable by evolving populations? These are some of the most fundamental questions in evolutionary biology.
Since Sewall Wright introduced the concept of the adaptive landscape in the early 1930s, such landscapes have received considerable attention from theorists interested in understanding how landscape topography affects evolutionary dynamics. In recent years, attention has shifted from theoretical analyses to empirical characterizations of landscapes constructed from experimental data. This shift has been triggered by advances in high-throughput sequencing and chip-based technologies, which have made it possible to assign phenotypes or fitness values to a large number of genotypes. The study of empirical adaptive landscapes is therefore a burgeoning area of research. However, all of the studies published to date share two limitations. The first is that their landscapes are highly incomplete: They have been constructed from just a small fraction of all possible genotypes. The second limitation is that they focus on just one or a small number of landscapes, making it difficult to generalize their findings. These limitations are completely understandable, because they stem from the hyper-astronomical sizes of most genotype spaces, such as those corresponding to macromolecules like RNA and proteins.
Our goal was to study a large number of complete and empirical adaptive landscapes. To accomplish this, we focused on transcriptional regulation, which drives the development, behavior and physiology of organisms as different as bacteria and humans. More specifically, we focused on the interaction between transcription factors and their DNA binding sites. It is well known that evolutionary change in transcription factor binding sites is both an important means by which gene regulation has evolved and a common culprit in disease. Such change is also involved in countless evolutionary adaptations. Our paper benefits from high-throughput technologies developed in recent years, such as protein binding microarrays and digital footprinting that have facilitated the characterization of thousands of potential transcription factor binding sites in many different organisms, both in vitro and in vivo.
In our paper, we characterize and study 1,137 adaptive landscapes of transcriptional regulation from 129 different eukaryotic species. Each landscape is derived from protein binding microarray data and describes the binding affinity of a transcription factor to all possible binding sites. In these genotype-phenotype landscapes, adaptation is the exploration of sequence space that tries to optimize the capacity of a short DNA sequence to bind a particular transcription factor in the absence of confounding factors such as chromatin context.
We found that these landscapes are highly navigable through single nucleotide changes, indicating that binding affinity – and thereby gene expression – is readily fine-tuned via mutations in transcription factor binding sites. These landscapes comprise few peaks that are made up of dozens to hundreds of sequences, and that vary in their accessibility. These peaks are therefore more similar to a mesa (table mountain) than a narrow peak such as the Matterhorn in the Swiss Alps. These findings, which are based on in vitro data, are supported by three additional analyses that are based on in vivo data. First, high-affinity transcription factor binding sites from rugged landscapes are less prevalent in protein-bound regions of the mouse genome than high-affinity sites from smooth landscapes. Second, gene expression measurements from hundreds of engineered yeast promoters closely reflect landscape topography. And third, the amount of genetic polymorphism in yeast binding sites increases with the number of sequences in a peak. Together, these analyses indicate that landscape topography has partly shaped the portfolio of regulatory DNA in two highly diverged eukaryotic species – yeast and mouse.
The data we study bring the metaphor of the adaptive landscape to life at unprecedented resolution. To our knowledge, this is the first time that multiple complete, empirical adaptive landscapes have been characterized, providing an exceptional opportunity to study their topography and navigability through single nucleotide changes. Affinity-modulating mutations in transcription factor binding sites are important drivers of adaptation, and this paper provides fundamental insight into how such mutations may fine-tune binding affinity. In doing so, it sheds further lights on the evolvability of transcriptional regulation.
We, the Navigators: José Aguilar-Rodríguez, Joshua L. Payne, and Andreas Wagner
Poster Credits: The poster image was created by Stephen J. Reynolds and Julia K. Johnson. It is a contour map of the Poland Junction Quadrangle in Arizona, which is an eroded mesa surrounded by lower, rugged terrain.
The paper can be found here: http://www.nature.com/articles/s41559-016-0045