Session B51: Focus Session: Evolutionary Systems Biology I - Evolutionary Dynamics and Rugged Fitness Landscapes

Optimal lineage principle for age-structured populations

Populations whose individuals exhibit age-dependent growth have often been studied using temporal dynamics of age distributions. In this talk, I examine the dynamics of age along lineages. We will see that the lineage point-of-view provides fundamental insights into evolutionary pressures on individuals' aging profiles. I will describe a variational principle that enables exact results for lineage statistics, in a variety of models. I will also discuss measurements on continuously dividing bacterial populations growing in microfluidics devices.   

Why Do Complex Systems Age?

Aging can be defined as the increase in probability of death with time. The observation that organisms, colonies, ecosystems, as well as larger social structures age and die in very similar ways suggest that the reasons underlying aging does not depend sensitively on molecular or cellular details. In this work we argue that aging is an inevitable outcome of the neutral co-evolution of non-aging components which with time become increasingly interdependent. Starting from this hypothesis, we construct generic dependency networks and obtain mortality rate as a function of time, as well as mean life expectancy as a function of organism size, complexity and metabolic rate.   

Learning about evolution from sequence data

Recent advances in sequencing and in laboratory evolution experiments have made possible to obtain quantitative data on genetic diversity of populations and on the dynamics of evolution. This dynamics is shaped by the interplay between selection acting on beneficial and deleterious mutations and recombination which reshuffles genotypes. Mounting evidence suggests that natural populations harbor extensive fitness diversity, yet most of the currently available tools for analyzing polymorphism data are based on the neutral theory. Our aim is to develop methods to analyze genomic data for populations in the presence of the above-mentioned factors. We consider different evolutionary regimes - Muller's ratchet, mutation-recombination-selection balance and positive adaption rate - and revisit a number of observables considered in the nearly-neutral theory of evolution. In particular, we examine the coalescent structure in the presence of recombination and calculate quantities such as the distribution of the coalescent times along the genome, the distribution of haplotype block sizes and the correlation between ancestors of different loci along the genome. In addition, we characterize the probability and time of fixation of mutations as a function of their fitness effect.   

Population genetics inside a cell: Mutations and mitochondrial genome maintenance

In realistic ecological and evolutionary systems natural selection acts on multiple levels, i.e. it acts on individuals as well as on collection of individuals. An understanding of evolutionary dynamics of such systems is limited in large part due to the lack of experimental systems that can challenge theoretical models. Mitochondrial genomes (mtDNA) are subjected to selection acting on cellular as well as organelle levels. It is well accepted that mtDNA in yeast Saccharomyces cerevisiae is unstable and can degrade over time scales comparable to yeast cell division time. We utilize a recent technology designed in Gottschling lab to extract DNA from populations of aged yeast cells and deep sequencing to characterize mtDNA variation in a population of young and old cells. In tandem, we developed a stochastic model that includes the essential features of mitochondrial biology that provides a null model for expected mtDNA variation. Overall, we find approximately 2\% of the polymorphic loci that show significant increase in frequency as cells age providing direct evidence for organelle level selection. Such quantitative study of mtDNA dynamics is absolutely essential to understand the propagation of mtDNA mutations linked to a spectrum of age-related diseases in humans.   

Rare beneficial mutations can halt Muller's ratchet

In viral, bacterial, and other asexual populations, the vast majority of non-neutral mutations are deleterious. This motivates the application of models without beneficial mutations. Here we show that the presence of surprisingly few compensatory mutations halts fitness decay in these models. Production of deleterious mutations is balanced by purifying selection, stabilizing the fitness distribution. However, stochastic vanishing of fitness classes can lead to slow fitness decay (i.e. Muller's ratchet). For weakly deleterious mutations, production overwhelms purification, rapidly decreasing population fitness. We show that when beneficial mutations are introduced, a stable steady state emerges in the form of a dynamic mutation-selection balance. We argue this state is generic for all mutation rates and population sizes, and is reached as an end state as genomes become saturated by either beneficial or deleterious mutations. Assuming all mutations have the same magnitude selective effect, we calculate the fraction of beneficial mutations necessary to maintain the dynamic balance. This may explain the unexpected maintenance of asexual genomes, as in mitochondria, in the presence of selection. This will affect in the statistics of genetic diversity in these populations.   

Clustering and Phase Transitions on a Neutral Landscape

The problem of speciation and species aggregation on a neutral landscape, subject to random mutational fluctuations rather than selective drive, has been a focus of research since the seminal work of Kimura on genetic drift. These ideas have received increased attention due to the more recent development of a neutral ecological theory by Hubbell. De Aguiar et al. recently demonstrated, in a computational model, that speciation can occur under neutral conditions; this study bears some comparison with more mathematical studies of clustering on neutral landscapes in the context of branching and annihilating random walks. Here, we show that clustering can occur on a neutral landscape where the dimensions specify the simulated organisms' phenotypes. Unlike the De Aguiar et al. model, we simulate \textit{sympatric} speciation: the organisms cluster phenotypically, but are not spatially separated. Moreover, we find that clustering occurs not only in the case of assortative mating, but also in the case of asexual fission. Clustering is not observed in a control case where organisms can mate randomly. We find that the population size and the number of clusters undergo phase-transition-like behavior as the maximum mutation size is varied.   

Mutational pathways to drug resistance through a maximally-rugged fitness landscape

Recent laboratory evolution experiments have identified surprising properties in the evolution of trimethoprim resistance in \textit{E.coli} through mutation of the drug's target, DHFR: (1) mutations are acquired in a reproducibly ordered manner; (2) multiple resistant endpoints exist; and (3) some pathways include mutation reversion or conversion. Here we investigate how these properties emerge from the fitness landscape of DHFR by characterizing all combinations of observed DHFR mutations. We see that the effects of mutations are so profoundly dependent on other mutations that sign-epistasis is nearly maximised, and the distributions of most mutations' effects are indistinguishable from randomly increasing or decreasing resistance. This almost `maximally-rugged' fitness landscape contains multiple separated peaks in drug resistance, and 20{\%} of favourable mutational steps are the loss or conversion of a previously acquired mutation. Select pathways through the rugged landscape avoid a common tradeoff between growth and resistance. Empirical characterization of this fitness landscape has identified that ordered but sometimes indirect mutational pathways to multiple endpoints arises from near-maximal levels of sign epistasis.   

Hidden Randomness between Fitness Landscapes Limits Reverse Evolution

Natural populations must constantly adapt to the ever-changing environment. A fundamental question in evolutionary biology is whether adaptations can be reversed by returning the population to its ancestral environment. Traditionally, reverse evolution is defined as restoring an ancestral phenotype (physical characteristics such as body size), and the classic Dollo's Law has hypothesized the impossibility of reversing complex adaptations. However, this ``law'' remains ambiguous unless reverse evolution can be studied at the level of genotypes (the underlying genome sequence). We measured the fitness landscapes of a bacterial antibiotic-resistance gene and analyzed the reversibility of evolution as a global, statistical feature of the landscapes. In both experiments and simulations, we find that an adaptation's reversibility declines as the number of mutations it involves increases, suggesting a probabilistic form of Dollo's Law at the molecular level. We also show computationally that slowly switching between environments facilitates reverse evolution in small populations, where clonal interference is negligible or moderate. This is an analogy to thermodynamics, where the reversibility of a physical process is maximized when conditions are modified infinitely slowly.   

Exploring the fitness landscape of poliovirus

RNA viruses are known to display extraordinary adaptation capabilities to different environments, due to high mutation rates. Their very dynamical evolution is captured by the quasispecies concept, according to which the viral population forms a swarm of genetic variants linked through mutation, which cooperatively interact at a functional level and collectively contribute to the characteristics of the population. The description of the viral fitness landscape becomes paramount towards a more thorough understanding of the virus evolution and spread. The high mutation rate, together with the cooperative nature of the quasispecies, makes it particularly challenging to explore its fitness landscape. I will present an investigation of the dynamical properties of poliovirus fitness landscape, through both the adoption of new experimental techniques and theoretical models.   

Scaling laws and universality for the strength of genetic interactions in yeast

Genetic interactions provide a window to the organization of the thousands of biochemical reactions in living cells. If two mutations affect unrelated cellular functions, the fitness effects of their combination can be easily predicted from the two separate fitness effects. However, because of interactions, for some pairs of mutations their combined fitness effect deviates from the naive prediction. We study genetic interactions in yeast cells by analyzing a publicly available database containing experimental growth rates of ~5 million double mutants. We show that the characteristic strength of genetic interactions has a simple power law dependence on the fitness effects of the two interacting mutations and that the probability distribution of genetic interactions is a universal function. We further argue that the strength of genetic interactions depends only on the fitness effects of the interacting mutations and not on their biological origin in terms of single point mutations, entire gene knockouts or even more complicated physiological perturbations. Finally, we discuss the implications of the power law scaling of genetic interactions on the ruggedness of fitness landscapes and the consequent evolutionary dynamics.   

Speeding up Evolutionary Search by Small Fitness Fluctuations

We consider a fixed size population that undergoes an evolutionary adaptation in the weak mutation rate limit, which we model as a biased Langevin process in the genotype space. We show analytically and numerically that, if the fitness landscape has a small highly epistatic (rough) and time-varying component, then the population genotype exhibits a high effective diffusion in the genotype space and is able to escape local fitness minima with a large probability. We argue that our principal finding that even very small time-dependent fluctuations of fitness can substantially speed up evolution is valid for a wide class of models.   

Experimental observation of critical slowing down as an early warning of population collapse

Near tipping points marking population collapse or other critical transitions in complex systems small changes in conditions can result in drastic shifts in the system state. In theoretical models it is known that early warning signals can be used to predict the approach of these tipping points (bifurcations), but little is known about how these signals can be detected in practice. Here we use the budding yeast Saccharomyces cerevisiae to study these early warning signals in controlled experimental populations. We grow yeast in the sugar sucrose, where cooperative feeding dynamics causes a fold bifurcation; falling below a critical population size results in sudden collapse. We demonstrate the experimental observation of an increase in both the size and timescale of the fluctuations of population density near this fold bifurcation. Furthermore, we test the utility of theoretically predicted warning signals by observing them in two different slowly deteriorating environments. These findings suggest that these generic indicators of critical slowing down can be useful in predicting catastrophic changes in population biology.