Traditionally, evolutionary biology has mostly taken a retrospective view, looking backwards in time to infer past evolutionary dynamics. Over the past 30 years, evolution experiments in the laboratory have become a valuable complementary technique to study evolution in real time. Microbial populations in shaken flasks are an ideal model system to do this, because their short generation times and easy reproducibility allow for the study of dozens to hundreds of replicates. Our understanding of microbial evolution in these simple laboratory environments has dramatically improved in recent years.

Microbial populations in the wild face vastly more complex conditions: they grow as spatially structured communities called microbial biofilms, often consisting of interacting mixtures of different species fulfilling different purposes, subject to various, potentially self-generated, biophysicochemical gradients of, e.g., oxygen or nutrients, which are in turn altered by the physical structure of the community. In short, natural population are subject to a vast variety of ecological interactions, and it has remained unclear how much can be learned from well-mixed liquid culture experiments about how ecology affects evolution in more complex scenarios.

In this dissertation, I approach this question using one of the simplest possible ecological aspects: the fact that most populations grow in spatially structured communities. Using microbial colonies as an experimental model system, I examine the effect of spatial structure on evolutionary dynamics in a variety of ways. First, Chapters 2-4 investigate the fates of neutral mutations and the dynamics of beneficial mutations in microbial colonies to find that both the neutral diversity resulting from spontaneous mutations and the strength of adaptation is increased in colonies compared to microbial populations grown in shaken flasks. The second half of the thesis is concerned with the effects of environmental heterogeneity on evolutionary dynamics. In Chapter 5, randomly disordered environments are used to examine the competition of selection and extrinsic noise in a model system for spontaneous beneficial and deleterious mutations. In these experiments, extrinsic noise can almost entirely overpower selection such that beneficial variants cannot leverage their advantage to further their evolutionary success. Chapter 6 discusses the effects of gradients on the emergence of antibiotic resistance and how convective flow can shape the trade-off between selection for resistance and the efficacy of treatment.

Overall, the results presented in this thesis suggest that spatial structure can have a momentous influence on the evolutionary dynamics of many dense cellular populations like biofilms and tumors: not only do the dynamics of adaptation change quantitatively in spatially structured populations, but qualitatively different patterns of evolutionary dynamics emerge that cannot arise in well-mixed population. Environmental heterogeneity can also have a strong influence on the speed and the direction of adaptation: whereas random heterogeneity in the environment prevents the spread of beneficial variants, the presence of antibiotic gradients can facilitate the rapid emergence of resistance. This work thus offers a glimpse into the profound and complex ways in which ecology can impact evolution even in simple model systems.

Living systems offer a richness of behaviors that are of broad interest to many fields of science. For instance, cells that are dwelling in their natural environment are mostly subject not only to the scarcity of energy resources, but also the space in which they can grow and live. This space limitation eventually leads to the emergence of contact forces (mechanical stress) between neighboring cells or the cells and their confining environment. These emergent forces may further have a crucial impact on the cells' biology, the dynamics of the whole population, or even the integrity of confinement (resulting in the remodeling of the environment). Even though the general importance of these forces has been widely recognized, technical difficulties and the complexity of the emergent phenomena prevented much progress in this direction. In this thesis, using mostly computer simulations, I make steps towards overcoming these barriers. In the first part of this thesis, I describe, on the coarse level, how the geometric properties of micro-confinement entails clogging of the microbial populations of Saccharomyces cerevisiae. These clogged populations are found to be quite disordered, with intermittent growth dynamics and heterogeneous mechanical stresses - properties much more like those of granular materials than a continuum. Thus, granular materials are an appealing framework to describe dense microbial populations. However, using a simple 2D model, I numerically show that the coupling between cellular growth rate and mechanical stress gives rise to deviation from the expected behavior of inanimate granular materials, and it increases the complexity of the emergent phenomena. The simple and coarse model used in the first part of the thesis is sufficient for relatively low density systems, but it is not adequate for strongly compacted systems. Thus, in the second part of this thesis, I employ the Finite Element Method to study in detail the structure and mechanics of the disordered packings of elastic shells - our proxy model for dense cellular packings. Therein, I discuss how deformations resulting from large compressive forces couple the structural and mechanical properties of the compact packings. Finally, using Lattice-Boltzmann simulations, I investigate the fluid transport in such compacted packings of deformable shells. I show that a relatively simple model proposed by Kozeny & Carman, combined with a percolation theory, can capture the fluid transport in porous materials. This result is of interest not only in dense biological systems, but also in a broader class of granular porous materials.

Microbes proliferate and migrate to take over territories. The dynamics of population growth canbe classified by the rate of proliferation and migration. When proliferation is predominant, the population density tends to saturate, and cells form expanding clusters. On the other hand, when migration is more active than the timescale of proliferation, the populations get almost well-mixed like a liquid culture in a shaken test tube. Both extreme cases have been extensively studied in previous research. However, the intermediate regime, where proliferation and migration consort with each other, has not been understood well. With this thesis, I attempt to provide an overarching insight into the interplay of proliferation and migration using microfluidic experiments and computer simulations. In Chapter 2, we show that the balance of proliferation and diffusion results in the sharp transition between two density states, gaseous and jammed states, using a newly-developed microfluidic device named microfluidic panflute. The density dependence of the diffusivity is shown to be fundamental to producing bifurcating behaviors with hysteresis. We further discuss the ecological impact of the density transition on invasion resistance. Chapter 3 characterizes the clone size distribution of jammed populations by fluctuation tests and lineage tracing with microfluidics. We show the characteristic power-law decay of the site frequency spectrum. We further discussed applying our results to cancer research: the site frequency spectrum can be reconstructed by sampling tumors spatially. In chapters 4 and 5, characteristic behaviors of jammed and gaseous populations are discussed. Chapter 4 shows the impact of the shape of physical boundaries on the population dynamics in a jammed state. In chapter 5, gaseous populations in various types of flow are investigated. This thesis contributes to the understanding of microbial population dynamics in spatial structure. Also, experimental techniques developed in this thesis, especially microfluidic systems, have the potential to be a platform for microbial experiments to investigate the ecological and evolutionary dynamics under spatial constraints.

With the advent of many new and revolutionary technologies (e.g. cheap high-throughput sequencing, precise gene editing, and more), biology has become a much more quantitative science over the past few decades. With these technologies in hand, we find ourselves able to probe some of the “fundamental” questions and assumptions that undergird the fields of evolution and ecology, both in controlled laboratory environments and in natural settings. However, with these new, quantitative observations, it is clear that some of the frameworks we use to think about biological populations are insufficient to describe relatively simple scenarios.

One of the key assumptions shared by population genetics and theoretical ecology is that these two fields are distinct. In other words, it has been taken for granted that ecological and evolutionary processes act separately and on disparate timescales. However, this may not necessarily be the case, where even in controlled laboratory evolution experiments, ecological structure frequently evolves on human observable times. This leads to an interesting (and broad) set of questions – what are the new timescales to consider in a joint eco-evolutionary process? How does eco-evolutionary feedback affect known observables? What new observables might be relevant? And, importantly, what should we find surprising in such a setting?

The contents of this dissertation hope to start to answer some of these questions by proposing and analyzing relatively simple models of eco-evolutionary dynamics. Since the task involves combining models that fall under the distinct classes of population genetics and ecology, the resulting joint models are necessarily more complex. However, by taking cues from statistical physics, I study these models in an explicitly high-dimensional setting, finding some forms of simplification.First, inspired by experimental observations of diversification resulting from the evolution of novel resource preferences, I (in joint work with Benjamin Good and Oskar Hallatschek) propose a minimal model of evolution in the setting of resource consumption with trade-offs. This model combines aspects of niche construction theory from the realm of ecology, with directional selection from the realm of population genetics. We study the low and high dimensional behavior of the model and describe its relatively simple steady state behavior which is dominated by resource generalists.

Second, I extend this model to include epistasis, or ‘rugged’ trade-offs. I show that the simple behavior of the non-epistatic model yields to a richer phase diagram when there is even weak epistasis. I show that in the many resource limit, the resource generalist state becomes ‘fragile’ to small epistatic fitness differences. This results in a transition in which the steady state gives way to a state of ‘punctuated equilibrium’ in which the ecosystem spends long times waiting for fitness mutations which bring about rapid rearrangement of the resource strategies of resident strains. This can be understood in light of the form of the Lyapunov function, which naturally separates into fitness specific and ecology specific components.

Finally, I propose a simple model of predator-prey co-evolution in a high dimensional setting. Using a combination of stochastic and deterministic simulations and theory inspired by the physics of disordered spin systems, I show that co-evolution stabilizes such populations for sufficiently variable interactions and for sufficiently high mutation rates, which stands in sharp contrast to expectations from ecological models alone. I also derive the phase boundary between a stable eco-evolutionary phase and an extinct phase, showing the dependence on relevant parameter combinations.

Deciphering the processes that govern microbial evolution allows us to make predictions of systems ranging from pathogen evolution to climate-driven ecosystem shifts. One fundamental evolutionary process is genetic drift, which is the stochastic change in population composition due to the randomness of birth and death processes. Genetic drift can lead to the loss of genetic diversity and weaken the efficacy of natural selection; thus, inferring its strength and understanding how it arises is important for understanding evolutionary dynamics. Despite decades of research on genetic drift, we still have limited empirical tools to infer the strength of genetic drift from data. Additionally, the strength of genetic drift is often considered to be a static property of a population on short timescales rather than an evolvable trait. In this thesis, we develop new methods for empirically measuring the strength of genetic drift and test hypotheses about the mechanisms that set the strength of genetic drift.

In Chapter 2, we test the hypothesis that mutations can change the strength of genetic drift, the first requirement needed for a trait to be evolvable. We focus on microbial colonies, which are a model system for range expansions. To test this hypothesis, we develop a new experimental method to measure the strength of genetic drift in high throughput using fluorescence microscopy. We find that mutations significantly affect the strength of genetic drift by causing changes in the self-organized spatial structure of the colony. These changes to genetic drift substantially affect the probability that new beneficial mutations escape stochastic extinction, providing evidence that genetic drift may be an evolvable property of a population.

In Chapter 3, we investigate the strength of genetic drift in SARS-CoV-2 evolution at the host population level. We develop a statistical inference framework for inferring the strength of genetic drift simultaneously with measurement noise from lineage frequency time series data. Applying this method to genomic data from England, we find that the strength of genetic drift is consistently, throughout time, higher than expected given the number of people infected with COVID-19 at the community level, even after correcting for measurement noise and epidemiological dynamics. We also find evidence for spatial structure in SARS-CoV-2 transmission at the regional level. The levels of genetic drift that we observe are higher than the estimated levels of superspreading found by modeling studies that incorporate data of contact statistics in England. We discuss how even in the absence of superspreading, high levels of genetic drift can be generated via jackpot events in a deme model.

The new experimental and computational methods developed in this thesis allow us to infer the strength of genetic drift, and thus gain a better understanding of evolutionary dynamics, in a larger range of laboratory and natural settings. We find that random single mutations can change the strength of genetic drift and affect downstream evolution, suggesting that genetic drift can be an evolvable trait of a population. Finally, we find that multiple mechanisms can alter the strength of genetic drift at the population level.