Schedule for: 19w5150 - Out-of-Equilibrium Processes in Evolution and Ecology

Microbial populations undergo multiple stages of growth, including a lag phase, an exponential growth phase, and a stationary phase. Mutations can therefore improve the frequency of a genotype not only by increasing its exponential growth rate, but also by decreasing the lag time or adjusting the yield (resource efficiency). However, recent experiments in yeast and E. coli have shown that these traits can also vary across genetically-identical single cells. Both the mean and variance of these distributions can evolve under mutation, selection, and genetic drift. To interpret this data in an evolutionary context, we develop a framework for population genetics with single-cell variation across multiple phases of microbial growth. We show how variation in single-cell lag times creates large fluctuations in lineage dynamics, allowing a mutation to fix more rapidly than would otherwise be expected. We also quantify how selection acts on this variability, which we use to predict patterns of coevolution for these growth traits.

The rapid global increase of multidrug-resistant organisms presents a major global health threat that will dramatically reduce the efficacy of antibiotics and thus constrain the number of effective treatments available to patients. As opposed to analogous efforts in viral epidemiology, accurate reconstruction of the pandemic spread of antibiotic resistance remains intractable for reasonable sample sizes due, in large part, to the high rate of homologous recombination and horizontal gene transfer that prevents the application of traditional phylogenetic approaches. Lastly, complete assemblies are a prerequisite to such quantitative study. In this talk, I will present a novel computational framework for bacterial evolution that generalizes the traditional linear reference genome to a pan-genomic non-planar graph. Using this framework, I will analyze 120 antibiotic resistant genomes collected in Basel, Switzerland over the course of a decade, sequenced with both ONT and Illumina data. We show that the evolution of antibiotic resistance exhibits a nested doll structure in which genetic transposition, homologous recombination, and clonal expansion occur at similar time-scales.

Many bacteria are capable to sense and actively follow chemical gradients. For Escherichia coli, much is known about the swimming machinery and the molecular signaling involved in chemotaxis. However, much less is known about its physiological role. Chemotaxis has been suggested to be a strategy of foraging in starvation conditions with swimming being triggered as an emergency response to find new nutrient sources. However, this hypothesis has never been probed rigorously. I will thus present a systematic investigation of chemotaxis and its dependence on growth conditions for E.coli. In contrast to the proposed starvation response, cells are chemotactic active in good growth conditions but stop swimming soon after encountering starvation. Considering the collective motion of cells along self-generated gradients I show how this up-regulation of chemotaxis with good growth conditions can serve as a foresighted strategy to occupy new habitats and optimize population growth. This navigated form of range expansion outcompetes the Fisher-Kolmogorov dynamics and might be a very general principle shaping the evolution of expanding populations.

Most populations are spread over spatial ranges much bigger than any one individual will explore in its lifetime. How does the simple fact of this spatial structure affect adaptive evolution and genetic diversity? We quantify how space interacts with recombination (sex) to slow down or speed up adaptation by changing the amount of clonal interference among beneficial mutations, and to change how adaptation reshapes neutral diversity through genetic hitchhiking. We also show how the spatial and temporal variation in genetic diversity revealed by sequencing can be used as a lens to watch how organisms move, capturing rare jumps that have large effects on populations' evolution.

Microbes exist in complex, multi-species communities with diverse interactions that play an essential role in both human health as well as the health of the planet. Over the last decade tremendous progress has been made in characterizing these communities, but the lack of experimentally tractable model systems has made it difficult to discern the rules governing microbial community assembly and function. In this talk I will describe our recent experimental efforts to develop a bottom-up approach to understanding the dynamics of these communities.

Spatially coexisting within- species bacterial diversity has now been observed in a variety of contexts. Why doesn’t survival of the fittest destroy such diversity? Most work on multiple interacting types — usually species — explicitly or implicitly assumes that the interactions with siblings are substantially stronger than with other types: essentially assuming niche-like ecology. Instead, we consider general Lotka-Volterra models in which the interactions with self and other are statistically similar. If there are antisymmetric correlations in the interactions — most naturally for one bacterial and one phage species each with micro diversity but no specificity — no stable communities exist and the ensuing chaos drives most types extinct. But with spatial structure of a number of identical islands and low migration between them, some types go globally extinct but a majority survive for times that are exponentially long in the number of islands. We then show that if the initial diversity is substantial, even “adiabatically” slow evolution increases the diversity, without bound in some models, and to a saturation level with evolution continuing forever without slowing down, if there is no upper bound to “generalist” fitness.

Microbes live in complex communities which play a central role in global nutrient cycles. The flow of nutrients around these cycles are defined by and define microbial community structure and metabolic function. As such, a primary concern of microbial ecology is understanding the eco-evolutionary process by which functional microbial communities arise. Towards this end, we present work interrogating the flow of nutrients through microbial communities in two contexts. Our first study uses denitrification as a model process whereby bacterial communities respire organic carbon in the absence of oxygen using nitrogen compounds as electron acceptors. Using natural isolates, sequencing and metabolite measurements we develop a statistical approach to mapping community level denitrification rates to genomic composition. In a second study we exploit closed microbial communities - self-sustaining microbial biospheres which persist indefinitely when supplied with only light. Closed ecosystem persistence relies on the sustained cycling of nutrients driven by photosynthesis. We present a new, high-precision, measurement of carbon cycling in closed ecosystems and go on to show how the composition of the community impacts its carbon cycling capabilities and long-term survival.

The dynamics of some systems are directional, meaning that the state of a system can be characterized by a function that increases in time. This includes, for example, the growth of entropy in physical systems, or the growth of fitness in certain formulations of Darwinian evolution. Yet even when directionality holds, it might be sensitive to perturbations, such as interactions in evolutionary dynamics. I'll describe a model of ecological community dynamics which features a phase-transition, from a regime where directionality is sensitive to a regime where it is robust. This has far reaching consequences: In the latter phase, the system admits many alternative community states, that are able to expand in space, forming (exact or approximate) copies of themselves. This leads to community-level selection, in analogy with Darwinian selection, with the increasing function acting as a fitness.


Microbes are everywhere. For every human cell of our body, there is at least one bacterial cell living inside us. Due to the advancement of sequencing techniques, there has been an explosion of data that track the composition of microbial communities. In a traditionally data-poor discipline as ecology, this novel richness of data represents a unique opportunity to understand quantitatively how different ecological forces shape diversity. In this talk, I will present three independent, emergent statistical patterns of distribution of abundances across species and communities, which are conserved across different ecosystems. I will then discuss how these "laws" can inform us about the fundamental mechanisms that are shaping the composition of these communities.

Machine learning algorithms can be fooled by small well-designed adversarial perturbations. This is reminiscent of cellular decision-making where ligands (called antagonists) prevent correct signalling, like in early immune recognition. We draw a formal analogy between neural networks used in machine learning and models of cellular decision-making relevant for immunology. We apply attacks from machine learning to simple decision-making models, and show explicitly the correspondence to antagonism by weakly bound ligands. We then apply a gradient-descent approach from machine learning, similar to an evolving agent trying to fool the cellular decision system. We reveal the existence of two regimes characterized by the presence or absence of a critical point for the gradient. This causes the strongest antagonists to lie close to the decision boundary. Our work connects evolved cellular decision-making to machine learning, and motivates the design of a general theory of adversarial perturbations, both for in vivo and in silico systems.

Ecological species can spread their extinction risk in an uncertain environment by adopting a bet-hedging strategy, i.e, by diversifying individual phenotypes. I will present a theory of bet-hedging for populations colonizing an unknown environment that fluctuates either in space or time. We find that diversification is more favorable for range expansion than in the well-mixed case, supporting the view that range expansions promote diversification. For slow rates of variation, spatial fluctuations open more opportunities for bet-hedging than temporal variations. Opportunities for bet-hedging reduce in the limit of frequent environmental variations. These conclusions are robust against demographic stochasticity induced by finite population sizes. 

Ref. P. Villa-Martin, M.A. Muñoz, S. Pigolotti, Plos Comp. Biol. 15(4): e1006529 (2019).


A recently proposed mechanism suggests that transient compartmentalization could have preceded cell division in prebiotic scenarios. Here, we study various classes of transient compartmentalization dynamics. We show that two regimes are possible: In a diffusion-limited regime (e.g. simple autocatalysis), a large noise is generated at the population level due to asynchronous growth. In contrast, in a replication-limited regime with many steps (e.g. polymerization), a low noise is generated at the population level. Since strong noise will yield many unviable population compositions, polymerization can present a strong fitness advantage. For deterministic growth dynamics, we introduce mutations that turn functional replicators into parasites. This can either lead to coexistence or parasite dominance, and we derive the phase boundary separating these two phases as a function of relative growth, inoculation size and mutation rate. We show that transient compartmentalization allows coexistence beyond the classical error threshold.

Biological evolution can be described as a population climbing a fitness landscape, and has inspired a variety of derivative-free optimization algorithms. Here we describe how phenotype evolution has sophisticated optimization properties. In particular, natural selection approximates second order gradient descent (Newton's method), and recombination is efficient in generating diversity. We use these insights to design a new type of derivative-free optimization algorithm for continuous problems.

Living systems evolve one mutation at a time, but a single mutation can alter the effect of subsequent mutations. The underlying mechanistic determinants of such epistasis are unclear. Here, we argue that the physical properties of a biological system can generically and easily constrain its epistasis. We analyze the interaction between mutations in generic models of proteins and biochemical networks. In each case, a slow, collective physical mode is actuated upon mutation, reducing the dimensionality of mutational effects and thus the rank of the epistatic matrix. This, in turn, reduces the ruggedness of the sequence-to-function map. By providing a mechanistic basis for experimentally observed global epistasis, these results suggest that slow collective physical modes can make biological systems more evolvable. 

Frequencies of synonymous codons are typically non-uniform, despite the fact that such codons correspond to the same amino acid in the genetic code. This phenomenon, known as codon bias, is broadly believed to be due to a combination of factors including genetic drift, mutational effects, and selection for speed and accuracy of codon translation; however, quantitative modeling of codon bias has been somewhat elusive. I will present a biophysical model which explains genome-wide codon frequencies observed across 20 organisms. Our model implements detailed codon-level treatment of mutations and includes two contributions to codon fitness which describe codon translation speed and accuracy. We find that the observed patterns of genome-wide codon usage are consistent with a strong selective penalty for mistranslated amino acids, while the dependence of codon fitness on translation speed is much weaker on average. Treating the translation process explicitly in the context of a finite ribosomal pool has allowed us to highlight the biophysical underpinnings of codon-level selective pressures. Overall, our approach offers a unified biophysical and population genetics framework for understanding the origin of codon bias.

Species level data show that gut microbiota can be remarkably resilient to brief environmental perturbations like antibiotics. However, little is currently known about how this ecological robustness is implemented at the strain level. In this talk, I will describe our recent efforts to address this question, by analyzing longitudinally sampled metagenomes from a single antibiotic-treated individual over a six-month period. We have developed new statistical approaches to simultaneously measure the ecological and evolutionary dynamics across multiple species in this community during the course of antibiotic treatment, revealing dramatic shifts in the genetic composition of individual species. I will show how these data can help distinguish between possible mechanisms of resilience, and what they can teach us about the population genetic forces that act within this community.

Cancer is the result of a stochastic evolutionary process characterized by the accumulation of mutations that are responsible for tumor growth, immune escape, and drug resistance, as well as mutations with no effect on the phenotype. Stochastic modeling can be used to describe the dynamics of tumor cell populations and to obtain insights into the hidden evolutionary processes leading to cancer. I will present recent approaches that use branching process models of cancer evolution to quantify intra-tumor heterogeneity and the development of drug resistance, and their implications for interpretation of cancer sequencing data and the design of optimal treatment strategies. 

Proteins can fold spontaneously into well-defined three-dimensional structures and can carry out complex biochemical reactions such as binding, catalysis, and long-range information transfer. The precision required for these properties is achieved while also preserving evolvability – the capacity to adapt in response to fluctuating selection pressures in the environment. What is the basic design of proteins that supports all of these properties? Recent work suggests that rather than direct physical analysis, statistical analysis of genome sequences provides a powerful and general approach to this problem. Using different methodologies, this approach has revealed both direct structural contacts as well as collective functional modes within protein structures. In this talk, I will present the current state of these approaches and the possibility of unifying them into a single theoretical framework for representing the evolutionary design of proteins.

The advent of a new modeling paradigm known as ‘differentiable programming’ makes possible bespoke machine-learned models of biological phenomena that are partly learned from data and partly informed by human-derived biophysical knowledge. In this talk I will describe three instantiations of this new approach for (i) de novo protein structure prediction, (ii) elucidation of the combinatorial grammar underlying metazoan signaling networks, and (iii) design of new protein function. In all cases qualitative improvements in model accuracy or speed, or both, are achieved using differentiable programming, enabling new scientific insights into biological macromolecules and the networks they comprise.

Models of evolutionary dynamics often focus on trajectories of variation due to a specific condition of selection, but the natural process often involves multiple, potentially opposing selection events. In such complex adaptive landscapes, when multiple opposing selection factors are simultaneously present or rapidly fluctuating in an environment, evolution is expected to be a genetic search on a convoluted fitness landscape with several pitfalls due to incompatibilities between genetic changes. I will present our recent studies addressing how evolutionary constraints on multiple fitness conditions are arranged in a bacterial efflux protein and how the rate and mechanism of protein evolution can vary under different (opposing or overlapping) selection conditions.


Antibodies mediate immunity to viruses and other pathogens by binding to specific antigenic targets. Our bodies deploy an evolvable arsenal of naïve antibodies capable of responding to diverse antigens. Upon exposure to a new antigen, potent and mature antibodies emerge from this arsenal through the Darwinian process of affinity maturation, consisting of iterative rounds of mutation and selection for improved binding to the antigen. This work seeks to understand the underlying relationship between sequence and function that guides this somatic evolutionary process by combining bioinformatic analysis of antibody repertoire sequencing data and high-throughput mutational scanning approaches. This framework is being applied to the broadly neutralizing HIV antibody VRC01, a representative of an important class of antibodies that are the subject of ongoing efforts in HIV vaccine design and antiviral therapy. By combining phylogenetic reconstruction of the VRC01 lineage with high-throughput ligand-binding assays, we are uncovering the relationship between sequence and function that guides VRC01 affinity maturation, revealing how epistasis, stochasticity, and historical contingency contribute to the complex maturation trajectory of this antibody. The results will highlight the landscape over which this broad class of therapeutically-important antibodies develop, with key implications for antibody engineering and HIV vaccine design.

Over the last ten years high-throughput sequencing has enabled increasingly quantitative measurements of the diversity of lymphocyte receptor repertoires. A striking finding of these sequencing efforts has been that the clone sizes of cells sharing the same receptor are heavy-tail distributed. Here, we present a simple neutral birth-death model for immune repertoire formation in which all cells compete for a global resource. Homeostatic control of proliferation leads to a founder effect, in which large clones emerge early when there is less competition. We show that this mechanism produces a transient but long-lived regime of power-law scaling of clone sizes. A reanalysis of a cohort study shows that indeed early founded T cell clones are over-represented among the most abundant clones. We use data about how the founder effect diminishes over time to constrain how much peripheral selection impacts immune repertoire dynamics. Overall, our work suggests that dynamical processes early in life have a strong and long-lasting influence on the structure of the immune repertoire.

Features of the CRISPR-Cas system, in which bacteria integrate small segments of phage genome (spacers) into their DNA to neutralize future attacks, suggest that its effect is not limited to individual bacteria but may control the fate and structure of whole populations [1]. In our model, we find that early dynamics of large phage clones is largely independent of bacterial dynamics but crucially depends on the burst-size of phage infections. In contrast, the fates of early phage mutants are strongly influenced by the feedback from bacterial population that creates a time-dependent fitness landscape for that phage type. Taken together, we quantify the role of population parameters in maintaining phage and bacterial diversity where CRISPR-cas is in the play. 
[1] https://www.pnas.org/content/115/32/E7462)

Recent advances in immunotherapy have brought to the fore questions regarding the conditions under which the adaptive immune system can recognize and eliminate cancer cells. Here we introduce a simple model of at the competition between detection and invasion and apply it to recent data from the TRACERX experiments. Our model produces interesting hypotheses regarding the age=dependence of cancer incidence and the ratio of cancer incidence "attempts" versus actual progression to detectable tumors.