The passage of genes from parents to offspring is a fundamental rule of heredity. However, bacteria violate this rule of strict vertical inheritance by shuttling DNA between cells through horizontal gene transfer (HGT). Common vehicles for HGT are conjugative plasmids, extrachromosomal pieces of DNA encoding the machinery for their own transfer. In addition to standard vertical transmission, genes on such plasmids can move between different strains or even different species of bacteria. This partial “uncoupling” of the evolutionary trajectory of plasmid-borne genes from the evolution of their host has interesting consequences for the ecology and evolution of mobile genes, which will be the focus of this presentation. These consequences revolve around the feature of horizontal transfer itself. I will discuss the impact that cross-species HGT has on protein evolution, focusing on a gene encoding an enzyme that degrades a common class of antibiotics, revisiting Wright’s fitness landscape metaphor in the process. The degree of topographical alignment of landscapes across different bacterial hosts can influence the nature of genetic evolution in the presence of HGT. Furthermore, such alignment affects the potential for adaptation to be distributed across different community members. I will conclude by sketching out a conceptual framework to highlight factors for consideration in the theoretical and experimental study of the evolution of mobile genes in microbial communities.
In their natural habitats, bacteria live in close contact with other species. Genomic studies reveal plentiful evidence of horizontal gene transfer across different species. However, little is known about its rates and fitness effects. What are limiting factors of cross-species gene transfer? How does gene transfer affect bacterial fitness? While gene transfer can benefit bacteria during adaptation to new niches, it also bears the potential of reducing fitness by introducing maladapted genes. We study these trade-offs by combining experimental evolution with genomics, transcriptomics, and molecular biology. We measured the rate of gene transfer between closely related Bacillus species and found that the rate of orthologous replacement is remarkably high. We found that gene transfer confers a net fitness increase and that selection has a broad genetic basis. Currently, we are using high-throughput technology for quantifying the distribution of fitness effects (DFE) of horizontal gene transfer between Bacillus species. The DFE shows both small and large effect beneficial transfers, indicating strong potential for fast adaptive evolution. In different growth conditions, strong effect transfers show pleiotropic effects including a fitness trade-off. With increasing complexity of nutrients and growth conditions, transformation becomes increasingly beneficial. We conclude that transformation has a strong potential for speeding up adaptation to varying environments by profiting form a gene-pool shared between closely related species.
While independently replicating sequences, such as transposons, are common in bacterial genomes, they usually do not persist for long periods of time. To be maintained in the gene pool bacterial mobile genetic elements require to jump hosts. In contrast, short sequence repeats known as REPINs – whose replication is dependent on a non-jumping RAYT transposase – persist for millions of years in bacterial genomes, in the absence of horizontal transfer. Extremely long persistence times and confinement to a single genome makes REPIN populations unique in biology. REPINs duplicate magnitudes less frequently than transposons, which means duplication events cannot be observed in the laboratory. Yet, across different bacterial strains REPIN population biology can be studied. For example, population size fluctuations correlate with available genome space as expected for organisms growing to carrying capacity. Other analyses of REPIN populations within species demonstrate signs of conflict between host and REPINs and how these conflicts are settled. Finally, a search for similar kinds of biological entities suggests that the REPIN-RAYT system may not be unique in bacteria.
Microbes living in natural communities develop antimicrobial resistance (AMR) through complex evolutionary trajectories. Fundamental features of this process emerge only in natural settings and therefore remain poorly understood. How are real-world AMR-associated genetic traits distributed between genes, intergenic regulatory regions, and mobile elements? Following an antibiotic exposure, are AMR-associated genetic changes transient or do they persist after the end of the disturbance? In this talk we will present a new tool that tracks evolutionary dynamics in natural microbial communities evolving under a short period of selective pressure. The tool recovers microbial genomes and their genetic variants from temporal DNA shotgun sequencing data. We overcome sequencing errors and assembly complications to identify true strain-level genetic variants within genomes, including single point polymorphisms (SNPs), local insertions/deletions and traces of large-scale genome rearrangements. By focusing on variants that dramatically increase in their frequency over time the tool pinpoints genetic changes that are associated with a fitness advantage during the disturbance. This approach allows to map out the genomic fitness landscape in natural settings. We will present unpublished data generated from ~50 healthy human subjects that were exposed to antibiotics. The analysis identified thousands of AMR-associated genetic variants that rose in frequency during the disturbance. We will show how most of these events are transient and subside after the end of the antibiotic treatment, and some persist until the end of the sampling period. While a fraction of the variants involves well-studied AMR genes, the majority is poorly characterized and potentially harbors novel mechanisms of resistance. To the best of our knowledge this is the most comprehensive assessment of evolution towards AMR in natural settings to date. This work paves the road to study short-term evolutionary dynamics in natural microbial communities.
While we often track evolution for single traits in microbes, understanding and projecting their ecological functions often requires considering their integrated multitrait phenotypes. For example, understanding how the responses of phytoplankton to environmental change translate into changes in ocean primary production changes depends not only on the direct responses to selection (e.g. warming), but also on correlations between traits such as cell size and composition, nutrient acquisition and population growth rates. In addition, phytoplankton evolve in a highly dynamic environment that can produce migration and demographic events with large effects on evolution. We used experimental evolution to explore how the model diatom genus Thalassiosira moves about multitrait space under relaxed and directional selection in standard and warmed environments. I will discuss how rapid evolution in trait correlations that are often modelled as fundamental or fixed tradeoffs constrains phenotypic shifts in Thalassiosira, and how dynamic ocean environments could facilitate these rapid evolution events. I will also discuss how transient shifts in multitrait correlations may impact the relationship between gene or lineage diversity, which is relatively easy to survey in the surface ocean, and trait values or diversity in phytoplankton.
No two cells are identical, even when sharing the same genetic code. This variability among phenotypes can be found in cell populations regardless of the complexity of the organism — from mammalian neural tissues to bacterial colonies. In the latter, genetically and morphologically identical bacteria often exhibit a myriad of growth states, resulting in drastic fitness variability. Most of this variance is attributed to stochastic processes. However, we showed that deterministic processes are an essential source of phenotypic heterogeneity. One such process is the asymmetric partitioning of intracellular components that happens upon cell division. When rod-shaped bacteria divide, one daughter inherits a conserved cell pole carrying more damaged components, whereas its sibling inherits a newly synthesized pole. Through single-cell microscopy and microfluidic techniques, we showed that this asymmetry drives a variance in fitness — as expressed by individual growth rates — that trickles down along cell lineages. For instance, by following lineages that inherited either new or conserved poles consecutively, we found that they reach distinct states of physiological equilibrium. Although the system was highly stochastic, this deterministic structure of the population was maintained as long as the population experienced low levels of extrinsic damage. However, under lethal levels of oxidation, asymmetry led to differential survival: lineages inheriting conserved poles arrested division, while those receiving new poles remained in equilibrium. Finally, we showed that these results extend to antibiotic responses, with asymmetry driving the fate of bacterial lineages upon drug exposure. Thus, bacterial asymmetry represents a deterministic source of phenotypic heterogeneity, driving survival in the face of environmental pressures and antibiotic treatments.
The fitness effects of all possible mutations available to an organism largely shapes the dynamics of evolutionary adaptation. Tremendous progress has been made in quantifying the strength and abundance of selected mutations available to single microbial species in simple environments, lacking strong ecological interactions. However, the adaptive potential of strains that are part of multi-strain communities remains largely unclear. We sought to fill this gap for a stable community of two closely related ecotypes (“L” and “S”) shortly after they emerged within the E. coli Long-Term Evolution Experiment (LTEE). To this end, we engineered genome-wide barcoded transposon libraries and developed a computational inference pipeline to measure the fitness effects of all possible gene knockouts in the coexisting strains as well as their ancestor, for many different conditions. We found that the fitness effect of most gene knockouts sensitively depends on the genetic background and the ecological conditions, as set by environmental perturbations and the relative frequency of both ecotypes. Despite the idiosyncratic behavior of individual knockouts, we still see consistent statistical patterns of fitness effect variation across both genetic background and community composition. The background dependence of mutational effects appears to reflect widespread changes in which gene functions are important for determining fitness, for all but the most strongly interacting genes. Additionally, fitness effects are correlated with evolutionary outcomes for a number of conditions, possibly revealing shifting patterns of adaptation. Together, our results reveal how ecological and epistatic effects combine to drive adaptive potential in recently diverged, coexisting ecotypes.
Biological functions of many cellular assemblages, ranging from multicellular organisms to microbial communities, rely on diversity among the composing units. Such division of labour is often seen through the lens of game theory, where the accent is posed on the success of different strategies in short-term competition (e.g. one cell type grows faster than another - like in cancer). Such conceptual model has consequences on our expectations for the evolutionary outcomes: costly collective functions should be wiped out by natural selection acting on individual cells. In this talk, I will discuss how viewing reproduction rates as the result of a game sidesteps several important features of the life cycles underpinning collective functions: the coexistence of multiple spatiotemporal scales, the context-dependence of interactions, and selection acting at the collective level. Population-level averages mask relevant processes whereby collective structures repeatedly self-organize. I will illustrate these concepts with models and observations on the eco-evolutionary dynamics of the social amoeba Dictyostelium discoideum, where the effects of interactions are time-dependent. I will end discussing how an externally imposed collective life cycle can drive the evolution of both mutualism and increased diversity.
Bacterial genes can either reside on the chromosome or on plasmids, extrachromosomal genetic structures that can be transferred from cell to cell. The distribution of genes between plasmid and chromosome is not random: certain types of genes are particularly likely to be plasmid-associated. This includes a number of clinically important traits, such as antibiotic resistance and virulence factors. The evolutionary mechanisms that give rise to this pattern are not well understood. Plasmids are occasionally lost during cell replication and thus less reliably inherited than the chromosome, and genes are free to transition between plasmid and chromosome: so what keeps genes on plasmids? We address this question through mathematical modelling. The key insight from our model is that the relative fitness of chromosomal and plasmid-borne genes depends on their relative frequencies (positive frequency-dependent selection). In other words, the fitness of a plasmid-borne gene will be higher in a population in which the chromosomal gene is rare (and vice versa). This positive frequency dependence can keep moderately beneficial genes on plasmids, despite occasional plasmid loss. This leads to a priority effect: whichever form of the gene (i.e., plasmid-borne or chromosomal) is acquired first has time to increase in frequency and thus becomes difficult to displace. Therefore, the relative rate of acquiring the gene on the plasmid versus the chromosome predicts where the gene will be found. Further modelling shows this effect is particularly pronounced when genes are beneficial across a large number of species. All together, the hypothesis that emerges from our work is that plasmid-borne genes are moderately beneficial; functional across a large number of species; and rarely acquired through chromosomal mutation. We suggest traits like antibiotic resistance are often found on plasmids because these genes commonly fulfil these criteria.
The evolution of multicellularity has opened new evolutionary paths to increased diversity and complexity. This transition from single cells to multicellularity involved three processes: cells remained attached to one another and formed groups, cells within these new groups differentiated to perform different tasks, and the emergent groups adapted their life cycles by evolving new reproductive strategies. Recent experiments have provided insights into the selection pressures as well as the genetic toolkit that may have driven the transition from unicellular to multicellular groups and the evolution of cell differentiation. However, how the life cycles of these newly-formed multicellular groups evolved is still unknown. Here, using the budding yeast, Saccharomyces cerevisiae, as a model system, we show that alternating regimes of resource availability can select for regulated multicellular life cycles. Examining a collection of S. cerevisiae wild isolates showed that most strains exist as multicellular clusters during the haploid stage of their life cycles. We also observed that these multicellular states are strongly influenced by their environment. By genetically controlling the size of clusters, we showed that in a low-sucrose environment, previously shown to favor cooperation, cluster size (a proxy for the number of cells in an organism) directly correlated with fitness. Meanwhile, clusters are less fit than single cells when cooperation is not required (glucose) or when the environment is patchy (emulsion). Finally using both wild and engineered yeast strains, we showed that regulated life cycles have a strong advantage over constitutively single-celled or multicellular life cycles when the environment alternates between favoring cooperation and dispersal. Our results suggest that ploidy and environment regulate S. cerevisiae multicellular life cycle and that alternating resource availability may have played a role in the evolution of life cycles. We anticipate that integrating the study of wild and engineered organisms with multicellular life cycles will enhance our understanding of the factors that drove the evolution of life cycles. Specifically, combining QTL mapping of wild isolates and experimental evolution under alternating resource availability may identify the evolutionary drivers of life cycle regulation.
The origin of multicellularity was one of the most significant innovations in the history of life. Our understanding of the evolutionary processes underlying this transition remains limited, however, mainly because extant multicellular lineages are ancient and most transitional forms have been lost to extinction. We bridge this knowledge gap by evolving novel multicellularity in vivo, using the 'snowflake yeast' model system. In this talk, I'll focus on our most Multicellularity Long-Term Evolution Experiment (MuLTEE), in which we've put snowflake yeast through ~5,000 generations of selection. We'll examine how snowflake yeast evolve to be ~20,000x larger, and 10,000x biophysically tougher than their ancestors through a clever change in the way that cells interact within the group. Through a combination of multicellular biophysics and synthetic biology, we'll examine how two key steps in this transition: a multicellular life cycle and heritability of multicellular traits, arise 'for free'. If time permits, we'll examine early steps in the evolution of cellular differentiation. Our approach, which allows for the study of macroevolutionary processes over microevolutionary timescales, demonstrates that multicellularity is less evolutionarily constrained than previously thought.
Bacteria often interact with their environment through extracellular molecules that increase access to limiting resources. These secretions can act as public goods, creating incentives for exploiters to invade and “steal” public goods away from producers. This phenomenon has been studied extensively in vitro, but little is known about the occurrence and impact of public good exploiters in the environment. Here, we develop a genomic approach to systematically identify bacteria that can exploit public goods produced during the degradation of polysaccharides. Focusing on chitin, a highly abundant marine biopolymer, we show that public good exploiters are active in natural chitin degrading microbial communities, invading early during colonization, and potentially hindering degradation. In contrast to in vitro studies, we find that exploiters and degraders belong to distant lineages, facilitating their coexistence. Our approach opens novel avenues to use the wealth of genomic data available to infer ecological roles and interactions among microbes.
The evolutionary battle between viruses and the immune system is likened to a high-stakes arms race. The immune system makes antiviral proteins, called restriction factors, which can stop the virus from replicating. In response, viruses evolve to evade the effects of restriction factors. To counter this, restriction factors evolve too, and the cycle continues, in which both sides rapidly evolve at interaction interfaces to gain or evade immune defense. For example, primate TRIM5α uses its rapidly evolving ‘v1’ loop to bind retroviral capsids whereas the MxA antiviral protein uses its rapidly evolving Loop L4 domain to recognize viruses such as influenza; single mutations in these loops can dramatically improve retroviral restriction. The challenge for the immune system is that mammals do not evolve as fast as viruses. How then, in the face of this disadvantage, can the immune system hope to keep pace with viral evolution? Using deep mutational scanning, we comprehensively measured how single mutations in the TRIM5α v1 loop affect restriction of divergent retroviruses. Unexpectedly, we found that most mutations increase weak antiviral function. Moreover, most random mutations do not disrupt potent viral restriction, even when it is newly acquired via a single adaptive substitution. Our results indicate that TRIM5α’s adaptive landscape is remarkably broad and mutationally resilient, maximizing its chances of success in evolutionary arms races with retroviruses. We also exploit combinatorial mutagenesis at rapidly evolving positions to dissect and enhance the antiviral properties of MxA antiviral proteins, revealing unprecedented capacity for antiviral adaptation and a 'breath versus specificity' tradeoff that constrains their natural evolution.
Large-scale duplications are a highly dynamic class of mutation. They arise and are subsequently lost – often without a trace – at rates far exceeding those typically observed for SNPs. The transient nature of large duplications means that their contribution to evolutionary processes is often overlooked. We are following the dynamics of adaptive, large-scale duplications in evolving populations of the model bacterium Pseudomonas fluorescens SBW25. We have passaged replicate lineages of two slow-growing SBW25 mutants – both lacking one or more tRNA genes – through 100 days (~750 generations) of evolution. We provide evidence that adaptation initially occurs via large-scale duplications, with each duplication fragment containing up to 16% of the wild-type chromosome and affecting the copy number of hundreds of genes (including tRNA genes). As expected, these large-scale duplication fragments are highly unstable and, despite providing a significant growth advantage, are lost at high rates. Our ongoing analyses indicate that, in each evolving population, multiple duplication fragments arise and compete, with progressively shorter (and hence more stable) duplication fragments dominating over time. In one lineage, an alternative, stable SNP (in the promoter of a duplicated tRNA gene) has been detected, and appears to be rising in frequency against the large duplication fragments. Our results demonstrate that large-scale duplications generate an unexpected degree of flexibility in genome content at the population level, with the potential to influence evolutionary outcomes.
In response to infection or vaccination, our immune system creates antibodies that bind strongly to relevant antigens through an evolutionary process called affinity maturation, which involves rounds of somatic hypermutation and selection. A key aspect of this process is the binding affinity landscape, which describes the mapping between antibody sequence and binding affinity to various antigens, and hence plays a role roughly analogous to the fitness landscape (though there are also many other important factors involved). I will describe a combinatorial library assembly and yeast-display system we have developed to systematically measure combinatorially complete binding affinity landscapes relevant for the maturation of broadly neutralizing anti-influenza antibodies (bnAbs). Our results show how widespread epistasis (including high-order epistasis) and pleiotropy may constrain the evolution of these bnAbs.
From phytoplankton producing the planet’s oxygen to wildebeest grazing the Serengeti, each species modifies their ecosystem. These ecological changes can precipitate adaptive evolution, which in turn can lead to further changes in the ecosystem. Previous studies have shown that this coupling between ecological and evolutionary processes is often driven by interactions between species. While there are a number of case-studies of individual demonstrations of eco-evolutionary feedbacks and the role of species interactions in driving these feedbacks, the details of how they work are not well understood. In particular, we do not know how the addition of a species to a community impacts the distribution of adaptations available to other community members, and how these adaptations in turn affect community ecology. We address this gap in an experimental microbial community consisting of the yeast Saccharomyces cerevisiae and the alga Chlamydomonas reinhardtii, which have been previously shown to form an obligatory mutualism in certain laboratory conditions (Hom & Murray, Science 2014). We modified these conditions to make this mutualism facultative, which allowed us to measure (1) how addition or removal of one community member (algae) changes the adaptive mutations available to another member (yeast) and (2) how these mutations in turn alter the ecology of the community. We sampled hundreds of adaptive mutations in each condition to gain a quantitative understanding of how a species interaction affects community eco-evolution. We show that yeast adaptation is highly diverse at both the genetic level and in their effect on community ecology, as some adaptations decrease the yield of one or both species while others increase one or both yields. However, algae algae systematically alter the distribution of community impacts by favoring adaptations that increase the yield of both species. This bias can be explained by understanding the dynamics of rapid adaptation in large asexual populations. As a result of this bias, evolution becomes more repeatable at the community level compared to the evolution of yeast in isolation.
Author: Sotaro Takano, Jean C.C. Vila, Alvaro Sanchez, Djordje Bajić
Microorganisms typically display a wide, and often overlapping, range of metabolic capabilities. In theory, this should favor competitive exclusion, and thus seems at odds with the pervasive coexistence and the diversity observed in natural microbiomes. One form of resource specialization that could partly explain the observed coexistence is the preference hierarchy for different substrates exhibited by most microbes, reflected in well-known diauxic growth patterns. Resource preferences have been shown to impact ecology by allowing microbes with different hierarchies to temporally partition the available resources and coexist. Yet, the extent to which substrate preference hierarchies can evolve and diversify remains largely unexplored. Here, we use a genome-scale metabolic model of a large-scale empirical genotype-phenotype map to show that alterations in the structure of metabolic networks often result in altered substrate preferences. By exploring a universal microbial metabolic genotype space in silico, we show that preference ranks tend to be more evolvable for specific substrates, and for pairs of substrates that are processed by different sets of reactions. We show that the diversification of metabolic preferences strongly depends on the specific topology of metabolism, with key reactions acting as capacitors and potentiators and influencing the evolvability of rank hierarchy. Our analysis sheds light on the evolvability and genetic determinants of microbial resource preference ranks.
There is great deal of interest in exploiting evolutionary tradeoffs to combat drug resistance. Instances of drug resistance have been steadily increasing creating considerable human health and economic impacts. In the United States, the CDC reports that ~35,000 deaths and $55 billion can be attributed to drug resistant infections per year. Collateral sensitivity (CS), where developing resistance to drug A results in sensitivity to drug B, is a type of evolutionary tradeoff that can be exploited to counter drug resistance. While some studies suggest that CS is common, others, in both cancer cells and bacteria populations, have demonstrated that CS is unpredictable and nonrepeatable. This is a problem if our goal is to predict how cell populations evolve to drug A in order to select the most effective drug B. One reason that previous work has not reached a consensus on the likelihood of evolutionary tradeoffs, like CS, that that most work utilizes small experiments with low replicate numbers and cannot survey the wide array of mutations that can contribute to drug resistance. Here, we use a barcoded S. cerevisiae system to track a large population of yeast strains as they develop resistance to different drugs at varying concentrations (drug A; n=12). This system tracked hundreds of thousands of replicate yeast lineages, such that it has the potential to reveal many different adaptive mutants that protect against each drug. From these evolved populations, ~24,000 strains were then subsequently challenged by a second drug (drug B; n=12) and mutants with interesting CS profiles were sequenced. These experiments provide a quantitative understanding of the likelihood of CS, how this likelihood changes depending on drug concentration, and whether patterns of CS are predictable across different mutants. Finally, we are developing a new high-throughput single cell DNA sequencing method that allows us to characterize the genetic basis of drug resistance and CS directly from pooled samples without having to physically isolate individual mutants. We hope this technology will dramatically increase the throughput with which we and others can interrogate the genetic basis of adaptation.
Unbiased and comprehensive genetic screens that are easily scalable to diverse bacterial pathogens would be valuable for obtaining a detailed understanding of antibiotic cross-resistance profiles, phage infection pathways and co-evolutionary landscapes of phage- and antibiotic resistance phenotypes. Recently, we have developed high-throughput barcoded genetic screening technologies that are scalable to different hosts; enable fast and effective genome-wide screens for gene function in competitive fitness assay format. We have applied these complementary loss-of-function and gain-of-function technologies to dozens of phages, antibiotics, metals, bacteriocins, and other antimicrobials targeting E coli, Salmonella and Pseudomonas strains. Our results accurately recapitulate known biology; identify hundreds of novel gene hits that play a role in bacterial fitness in stressful conditions; discover diverse modes of phage resistance and enable high-throughput mapping of genotype-phenotype relationships in relevant environmental conditions. By combining fitness datasets for phages, antibiotics, and antimicrobials or phage-antibiotic combination therapies, such screens could provide an avenue for performing systematic search for genetic trade-offs or ‘evolutionary traps’ and provide a much-needed solution to overcome the antibiotic-resistance pandemic.
Gene regulatory networks are essential to organism survival as they allow rapid adaptation through altering gene expression profiles. These regulatory networks can be key sites of evolutionary change and they provide important insights into the adaptability of various organisms to environmental shifts such as climate change. Mutations drive their evolution, but mutation biases can drive adaptation down one particular route and may limit their ability to explore alternative (and potentially fitter) evolutionary trajectories. One cause of mutational biases are DNA secondary structures such as DNA hairpins that can cause the replication machinery to “trip” and generate a mutation. If this secondary structure is stable, it can create a mutational hotspot. Environmental signals can increase expression of certain genes, which in turn increases the chance of mutation events. This research looks at the evolutionary consequences of these interplaying factors on the evolution of gene regulatory networks by employing a model system that re-evolves motility in different nutrient environments across two bacterial strains of Pseudomonas fluorescens: one with a mutational hotspot in a gene regulatory network (AR2), and one without (Pf0-2x). We focus on the regulation of the motility phenotype of these bacteria, which is transient, only being essential when bacteria are searching for food or escaping toxins or predators and can therefore be easily selected for. By disrupting fleQ, the master regulator for the flagellum, we render them immotile. Under strong selection through starvation, the bacteria reliably re-evolve motility after just a few days, by co-opting a response regulator from another regulatory network (associated with nitrogen regulation) to take over the function of the missing FleQ. Co-option is achieved through mutations in the DNA of the nitrogen response regulator and associated regulatory genes within the network. AR2 has a mutational hotspot in one of these regulatory genes that funnels evolution down the same route nearly every time, with motility restored almost exclusively by a single repeatable SNP. The hotspot is predicted to be caused by a hairpin in the DNA, which can be abolished via the introduction of synonymous changes. Restoration of motility in Pf0-2x is via recruitment of the same nitrogen response regulator as AR2 but it does not contain a mutational hotspot, as such mutations conferring motility are observed across multiple loci but within the same regulatory network. When evolved in complex and simple nutrient environments we see that Pf0-2x has environmentally sensitive biases towards mutations in specific genes, whilst the mutation spectrum in AR2 remains constant across nutrient environments, suggesting the hotspot is a stronger force in determining adaptive outcomes within AR2. The types of mutations we see have varying pleiotropic costs despite restoring the same motility phenotype. The most common mutations typically are not the fittest, highlighting the important role that randomness and mutation bias can play in evolution, rather than selection and persistence of genotypes based purely on fitness.
Tremendous progress has been made in quantifying fitness landscapes and elucidating how the effects of available mutations affect the dynamics of single microbial species in simple environments lacking strong ecological interactions. However, it remains largely unclear how natural selection depends and feeds back onto the spatial and community structure characteristic of most natural microbiomes. I discuss two of our attempts to begin filling this gap. First, I describe microfluidic experiments revealing that eco-evolutionary dynamics can be extremely sensitive to the spatial niche available to a population. The second project is focussed on the community structure of the population, aiming at quantifying the evolutionary potential of a nascent bacterial community. By temporally tracking barcoded transposon libraries derived from a long-term evolution experiment in E. coli, we explore how ecological and epistatic effects combine to drive adaptive evolution in a nascent community of closely related, coexisting ecotypes. We find a pronounced frequency-dependence in the distributions of fitness effects, which could promote long-term co-existence in the face of rampant adaptive evolution.
Populations are fundamental units of ecology and evolution, and delineating ecologically meaningful populations among microbes is important for identifying how they adapt to and interact with their local environment. Here, we develop a method to assign closely related isolates to populations by inferring their gene flow information through a tri-partitioning of SNPs distributed across the genome. By applying this method to the whole genomes of over 16,000 publicly available human gut microbiome isolates, we found that approximately 50% of the microbial taxa in the human gut microbiome are highly recombinogenic, while the other 50% consist of one or more “clonal” populations. Especially representative of the “clonal” taxa are Bacteroides: they can be classified into many “clonal” populations, where isolates in each population share the same unique clonal frame, but harbor different genomic islands. Comparative genomics between the different clonal populations revealed their clonal frames mostly differed by genes that were involved in capsule biosynthesis and vitamin B12 metabolism, as well as carbohydrate active enzymes and flagella related proteins. Often isolates from the same Bacteroides population could be found in biogeographically different human populations, indicating the frequent dispersal and recolonization of human gut Bacteroides populations. We are thus currently investigating the distribution of the clonal Bacteroides populations in humans with different health states, and their links to the population “clonal frame” specific genes. We hope that this work will eventually allow us to develop a general “reverse ecology” framework that uses genomic information to find disease associated microbial populations and adaptations in the human gut microbiome.
Whether evolution is predictable has become an outstanding question in the field of evolutionary biology and requires knowledge of the complex genotype-fitness map. Experimental evolution studies have begun to shed light on this, but it has not yet been determined if predictions can be extended between different species. Here, we use the Pseudomonas fluorescens SBW25 wrinkly spreader (WS) system to predict the outcome of experimental evolution in the closely related P. protegens Pf-5. When cultured in static microcosms, populations of SBW25 rapidly evolve to form a biofilm at the air-liquid interface, with the most adaptive phenotype (the WS) overproducing a cellulosic polymer, ultimately promoting cell-cell adhesion, resulting in a wrinkly colony morphology. The genotype-to-phenotype map of this trait is well characterized, making this system ideal to test the efficacy of predictability in related species. Our results show that predictions based on previous experiments and models in SBW25 were sufficient to predict evolution in Pf-5 across several biological levels. Although Pf-5 lacks the main structural component of SBW25 biofilms, cellulose, we successfully predicted the four main pathways to the adaptive WS genotype, but the phenotypic basis of biofilm formation could only be partially predicted. Types of mutations were also successfully predicted, with loss-of-function mutations in negative regulators being the most common. Mutated regions in genes could also be predicted, but there was little parallelism at the nucleotide level. Finally, we will present novel results experimentally testing predictions of two additional species, P. savastanoi and P. syringae, that encode diverse biofilm genes.
Random mutations and demographic events make evolution inherently stochastic, despite the deterministic force of natural selection. In order to better understand the genetic and ecological factors that drive evolutionary predictability, we use the evolution of resistance to beta-lactam antibiotics in Escherichia coli as experimental model. I will present recent work on the joint effect of mutation and selection bias in bacterial populations of different size, which shows that different-sized populations use similar numbers, but different types of mutations with distinct consequences for the level of resistance. For example, large populations more often use gain-of-function point mutations, including mutations activating a beta-lactamase, leading to high-level resistance, whereas small populations fix more often large deletions and duplications, including the deletion of the same inactive beta-lactamase, leading to 10 times lower resistance levels. We find that these distinct mutation choices can be largely predicted based solely on the rates and fitness effects of different mutation classes. However, some mutation choices require more subtle information about dose-dependent fitness effects of different mutations. If time allows, I may also present recent work on how social interactions within and between bacterial genotypes affect the emergence and selection of resistant mutants.
Fitness landscapes map genotypes to fitness, visualizing possible evolutionary paths. These landscapes are studied both at the conceptual level and made explicit by measuring the fitness of nearby genotypes to create empirical fitness landscapes. Since the mapping of the genotype to fitness depends on the environment, several approaches have been taken to include the environment. One of them is using seascapes, landscapes that change over time, resembling a changing environment. Another aspect is that genotypes themselves might change the environment, leading to the deformability of a fitness landscape. All these methods assume a new genotype removes the old phenotype from the population. However, certain changes of the environment, such as the production of cross-feeding components or the removal of toxins from the environment, allow strains to coexist. Recent results in the sequencing of long-term cultures have shown that coexistence is common and can last at evolutionary timescales. To study the effect of coexistence on evolutionary paths, we introduce 'eco-fitness landscapes', which include ecological mechanisms of coexistence. Coexistence states greatly increase the connectivity of a fitness landscape, and could lead to accessibility of fitness peaks that are otherwise unreachable. Coexistence states can also be incorporated in empirical fitness landscapes. As an example we use a previously characterized empirical fitness landscape of increased cefotaxime resistance in E. coli with a TEM1 gene coding for a beta-lactamase, which can break down the antibiotic cefotaxime. The breakdown results in lower cefotaxime concentrations, which can allow for growth of more susceptible strains, a mechanism called cross-protection. Cross-protection is more pronounced at higher cell densities, because the extracellular concentrations are stronger affected when there are more cells. We measured the properties affecting these dynamics, such as growth rates and antibiotic clearance rates, of eight different mutants, with mutations in the TEM1 gene, and show experimentally that at some densities coexistence is possible. Using this experimental data in a mathematical model we can create an empirical 'eco-fitness landscape'. Then we can predict how evolutionary trajectories are affected by coexistence due to cross-protection, depending on the density of the culture. Including ecology in fitness landscapes may lead to more realistic prediction of evolutionary trajectories and outcomes when processes that promote coexistence are involved. Moreover, including coexistence states in conceptual studies of fitness landscapes can lead to a different view on the reachability of fitness optima.