Higher-fitness yeast genotypes are less robust to deleterious mutations

Soft Modes as a Predictive Framework for Low-Dimensional Biological Systems Across Scales

All biological systems are subject to perturbations arising from thermal fluctuations, external environments, or mutations. Yet, while biological systems consist of thousands of interacting components, recent high-throughput experiments have shown that their response to perturbations is surprisingly low dimensional: confined to only a few stereotyped changes out of the many possible. In this review, we explore a unifying dynamical systems framework-soft modes-to explain and analyze low dimensionality in biology, from molecules to ecosystems. We argue that this soft mode framework makes nontrivial predictions that generalize classic ideas from developmental biology to disparate systems, namely phenocopying, dual buffering, and global epistasis. While some of these predictions have been borne out in experiments, we discuss how soft modes allow for a surprisingly far-reaching and unifying framework in which to analyze data from protein biophysics to microbial ecology.

The evolution of robustness and fragility during long-term bacterial adaptation

Theory predicts that well-adapted populations may evolve mechanisms to counteract the inevitable influx of deleterious mutations. While mutational robustness can be directly selected in the laboratory, evidence for its spontaneous evolution during general adaptation is mixed. Moreover, whether robustness evolves to include pleiotropic effects remains largely unexplored. Here, we studied the effects of point mutations in the RNA polymerase of Escherichia coli over a 15,000-generation adaptive trajectory. Fitness effects of both beneficial and deleterious mutations were attenuated in fitter backgrounds. In contrast, pleiotropic effects became more severe and widespread with greater adaptation. These results show that trade-offs between robustness and fragility can evolve in regulatory networks, regardless of whether driven by adaptive or nonadaptive processes. More broadly, they illustrate how adaptation can generate hidden variability, with unpredictable evolutionary consequences in new environments.

Systematic genetic perturbation reveals principles underpinning robustness of the epigenetic regulatory network

The molecular control of epigenetic information relies on hundreds of proteins of diverse function, which cooperate in defining chromatin structure and DNA methylation landscapes. While many individual pathways have been characterized, how different classes of epigenetic regulators interact to build a resilient epigenetic regulatory network (ERN) remains poorly understood. Here, we show that most individual regulators are dispensable for somatic cell fitness, and that robustness emerges from multiple layers of functional cooperation and degeneracy among network components. By disrupting 200 epigenetic regulator genes, individually or in combination, we generated network-wide maps of functional interactions for representative regulators. We found that paralogues represent only a first layer of functional compensation within the ERN, with intra- or inter-class interactions buffering the effects of perturbation in a gene-specific manner: while CREBBP cooperates with multiple acetyltransferases to form a subnetwork that ensures robust chromatin acetylation, ARID1A interacts with regulators from across all functional classes. When combined with oncogene activation, the accumulated epigenetic disorder exposes a synthetic fragility and broadly sensitizes ARID1A-deficient cells to further perturbation. Our findings reveal homeostatic mechanisms through which the ERN sustains somatic cell fitness and uncover how the network remodels as the epigenome is progressively deregulated in disease.

Saturated Transposon Analysis in Yeast as a one-step method to quantify the fitness effects of gene disruptions on a genome-wide scale

Transposon insertion site sequencing (TIS) is a powerful tool that has significantly advanced our knowledge of functional genomics. For example, TIS has been used to identify essential genes of Saccharomyces cerevisiae, screen for antibiotic resistance genes in Klebsiella pneumoniae and determine the set of genes required for virulence of Mycobacterium tuberculosis. While providing valuable insights, these applications of TIS focus on (conditional) gene essentiality and neglect possibly interesting but subtle differences in the importance of genes for fitness. Notably, it has been demonstrated that data obtained from TIS experiments can be used for fitness quantification and the construction of genetic interaction maps, but this potential is only sporadically exploited. Here, we present a method to quantify the fitness of gene disruption mutants using data obtained from a TIS screen developed for the yeast Saccharomyces cerevisiae called SATAY. We show that the mean read count per transposon insertion site provides a metric for fitness that is robust across biological and technical replicate experiments. Importantly, the ability to resolve differences between gene disruption mutants with low fitness depends crucially on the inclusion of insertion sites that are not observed in the sequencing data to estimate the mean. While our method provides reproducible results between replicate SATAY datasets, the obtained fitness distribution differs substantially from those obtained using other techniques. It is currently unclear whether these inconsistencies are due to biological or technical differences between the methods. We end with suggestions for modifications of the SATAY procedure that could improve the resolution of the fitness estimates. Our analysis indicates that increasing the sequencing depth does very little to reduce the uncertainty in the estimates, while replacing the PCR amplification with methods that avoid or reduce the number of amplification cycles will likely be most effective in reducing noise.

Soft Modes as a Predictive Framework for Low Dimensional Biological Systems across Scales

All biological systems are subject to perturbations: due to thermal fluctuations, external environments, or mutations. Yet, while biological systems are composed of thousands of interacting components, recent high-throughput experiments show that their response to perturbations is surprisingly low-dimensional: confined to only a few stereotyped changes out of the many possible. Here, we explore a unifying dynamical systems framework - soft modes - to explain and analyze low-dimensionality in biology, from molecules to eco-systems. We argue that this one framework of soft modes makes non-trivial predictions that generalize classic ideas from developmental biology to disparate systems, namely: phenocopying, dual buffering, and global epistasis. While some of these predictions have been borne out in experiments, we discuss how soft modes allow for a surprisingly far-reaching and unifying framework in which to analyze data from protein biophysics to microbial ecology.

Evolution of evolvability in rapidly adapting populations

Mutations can alter the short-term fitness of an organism, as well as the rates and benefits of future mutations. While numerous examples of these evolvability modifiers have been observed in rapidly adapting microbial populations, existing theory struggles to predict when they will be favoured by natural selection. Here we develop a mathematical framework for predicting the fates of genetic variants that modify the rates and benefits of future mutations in linked genomic regions. We derive analytical expressions showing how the fixation probabilities of these variants depend on the size of the population and the diversity of competing mutations. We find that competition between linked mutations can dramatically enhance selection for modifiers that increase the benefits of future mutations, even when they impose a strong direct cost on fitness. However, we also find that modest direct benefits can be sufficient to drive evolutionary dead ends to fixation. Our results suggest that subtle differences in evolvability could play an important role in shaping the long-term success of genetic variants in rapidly evolving microbial populations.

Effects of selection stringency on the outcomes of directed evolution

Directed evolution makes mutant lineages compete in climbing complicated sequence-function landscapes. Given this underlying complexity it is unclear how selection stringency, a ubiquitous parameter of directed evolution, impacts the outcome. Here we approach this question in terms of the fitnesses of the candidate variants at each round and the heterogeneity of their distributions of fitness effects. We show that even if the fittest mutant is most likely to yield the fittest mutants in the next round of selection, diversification can improve outcomes by sampling a larger variety of fitness effects. We find that heterogeneity in fitness effects between variants, larger population sizes, and evolution over a greater number of rounds all encourage diversification.

Environment-independent distribution of mutational effects emerges from microscopic epistasis

Predicting how new mutations alter phenotypes is difficult because mutational effects vary across genotypes and environments. Recently discovered global epistasis, in which the fitness effects of mutations scale with the fitness of the background genotype, can improve predictions, but how the environment modulates this scaling is unknown. We measured the fitness effects of ~100 insertion mutations in 42 strains of Saccharomyces cerevisiae in six laboratory environments and found that the global epistasis scaling is nearly invariant across environments. Instead, the environment tunes one global parameter, the background fitness at which most mutations switch sign. As a consequence, the distribution of mutational effects is predictable across genotypes and environments. Our results suggest that the effective dimensionality of genotype-to-phenotype maps across environments is surprisingly low.

Energy landscapes of peptide-MHC binding

Molecules of the Major Histocompatibility Complex (MHC) present short protein fragments on the cell surface, an important step in T cell immune recognition. MHC-I molecules process peptides from intracellular proteins; MHC-II molecules act in antigen-presenting cells and present peptides derived from extracellular proteins. Here we show that the sequence-dependent energy landscapes of MHC-peptide binding encode class-specific nonlinearities (epistasis). MHC-I has a smooth landscape with global epistasis; the binding energy is a simple deformation of an underlying linear trait. This form of epistasis enhances the discrimination between strong-binding peptides. In contrast, MHC-II has a rugged landscape with idiosyncratic epistasis: binding depends on detailed amino acid combinations at multiple positions of the peptide sequence. The form of epistasis affects the learning of energy landscapes from training data. For MHC-I, a low-complexity problem, we derive a simple matrix model of binding energies that outperforms current models trained by machine learning. For MHC-II, higher complexity prevents learning by simple regression methods. Epistasis also affects the energy and fitness effects of mutations in antigen-derived peptides (epitopes). In MHC-I, large-effect mutations occur predominantly in anchor positions of strong-binding epitopes. In MHC-II, large effects depend on the background epitope sequence but are broadly distributed over the epitope, generating a bigger target for escape mutations due to loss of presentation. Together, our analysis shows how an energy landscape of protein-protein binding constrains the target of escape mutations from T cell immunity, linking the complexity of the molecular interactions to the dynamics of adaptive immune response.

Environment-independent distribution of mutational effects emerges from microscopic epistasis

Predicting how new mutations alter phenotypes is difficult because mutational effects vary across genotypes and environments. Recently discovered global epistasis, where the fitness effects of mutations scale with the fitness of the background genotype, can improve predictions, but how the environment modulates this scaling is unknown. We measured the fitness effects of ~100 insertion mutations in 42 strains of Saccharomyces cerevisiae in six laboratory environments and found that the global-epistasis scaling is nearly invariant across environments. Instead, the environment tunes one global parameter, the background fitness at which most mutations switch sign. As a consequence, the distribution of mutational effects is predictable across genotypes and environments. Our results suggest that the effective dimensionality of genotype-to-phenotype maps across environments is surprisingly low.

Effects of selection stringency on the outcomes of directed evolution

Directed evolution makes mutant lineages compete in climbing complicated sequence-function landscapes. Given this underlying complexity it is unclear how selection stringency, a ubiquitous parameter of directed evolution, impacts the outcome. Here we approach this question in terms of the fitnesses of the candidate variants at each round and the heterogeneity of their distributions of fitness effects. We show that even if the fittest mutant is most likely to yield the fittest mutants in the next round of selection, diversification can improve outcomes by sampling a larger variety of fitness effects. We find that heterogeneity in fitness effects between variants, larger population sizes, and evolution over a greater number of rounds all encourage diversification.

Global epistasis and the emergence of function in microbial consortia

The many functions of microbial communities emerge from a complex web of interactions between organisms and their environment. This poses a significant obstacle to engineering microbial consortia, hindering our ability to harness the potential of microorganisms for biotechnological applications. In this study, we demonstrate that the collective effect of ecological interactions between microbes in a community can be captured by simple statistical models that predict how adding a new species to a community will affect its function. These predictive models mirror the patterns of global epistasis reported in genetics, and they can be quantitatively interpreted in terms of pairwise interactions between community members. Our results illuminate an unexplored path to quantitatively predicting the function of microbial consortia from their composition, paving the way to optimizing desirable community properties and bringing the tasks of predicting biological function at the genetic, organismal, and ecological scales under the same quantitative formalism.

Epistasis facilitates functional evolution in an ancient transcription factor

A protein's genetic architecture - the set of causal rules by which its sequence produces its functions - also determines its possible evolutionary trajectories. Prior research has proposed that the genetic architecture of proteins is very complex, with pervasive epistatic interactions that constrain evolution and make function difficult to predict from sequence. Most of this work has analyzed only the direct paths between two proteins of interest - excluding the vast majority of possible genotypes and evolutionary trajectories - and has considered only a single protein function, leaving unaddressed the genetic architecture of functional specificity and its impact on the evolution of new functions. Here, we develop a new method based on ordinal logistic regression to directly characterize the global genetic determinants of multiple protein functions from 20-state combinatorial deep mutational scanning (DMS) experiments. We use it to dissect the genetic architecture and evolution of a transcription factor's specificity for DNA, using data from a combinatorial DMS of an ancient steroid hormone receptor's capacity to activate transcription from two biologically relevant DNA elements. We show that the genetic architecture of DNA recognition consists of a dense set of main and pairwise effects that involve virtually every possible amino acid state in the protein-DNA interface, but higher-order epistasis plays only a tiny role. Pairwise interactions enlarge the set of functional sequences and are the primary determinants of specificity for different DNA elements. They also massively expand the number of opportunities for single-residue mutations to switch specificity from one DNA target to another. By bringing variants with different functions close together in sequence space, pairwise epistasis therefore facilitates rather than constrains the evolution of new functions.

Genome-scale analysis of interactions between genetic perturbations and natural variation

Interactions between genetic perturbations and segregating loci can cause perturbations to show different phenotypic effects across genetically distinct individuals. To study these interactions on a genome scale in many individuals, we used combinatorial DNA barcode sequencing to measure the fitness effects of 8046 CRISPRi perturbations targeting 1721 distinct genes in 169 yeast cross progeny (or segregants). We identified 460 genes whose perturbation has different effects across segregants. Several factors caused perturbations to show variable effects, including baseline segregant fitness, the mean effect of a perturbation across segregants, and interacting loci. We mapped 234 interacting loci and found four hub loci that interact with many different perturbations. Perturbations that interact with a given hub exhibit similar epistatic relationships with the hub and show enrichment for cellular processes that may mediate these interactions. These results suggest that an individual's response to perturbations is shaped by a network of perturbation-locus interactions that cannot be measured by approaches that examine perturbations or natural variation alone.

Pathway Evolution Through a Bottlenecking-Debottlenecking Strategy and Machine Learning-Aided Flux Balancing

The evolution of pathway enzymes enhances the biosynthesis of high-value chemicals, crucial for pharmaceutical, and agrochemical applications. However, unpredictable evolutionary landscapes of pathway genes often hinder successful evolution. Here, the presence of complex epistasis is identifued within the representative naringenin biosynthetic pathway enzymes, hampering straightforward directed evolution. Subsequently, a biofoundry-assisted strategy is developed for pathway bottlenecking and debottlenecking, enabling the parallel evolution of all pathway enzymes along a predictable evolutionary trajectory in six weeks. This study then utilizes a machine learning model, ProEnsemble, to further balance the pathway by optimizing the transcription of individual genes. The broad applicability of this strategy is demonstrated by constructing an Escherichia coli chassis with evolved and balanced pathway genes, resulting in 3.65 g L-1 naringenin. The optimized naringenin chassis also demonstrates enhanced production of other flavonoids. This approach can be readily adapted for any given number of enzymes in the specific metabolic pathway, paving the way for automated chassis construction in contemporary biofoundries.

Global epistasis in plasmid-mediated antimicrobial resistance

Antimicrobial resistance (AMR) in bacteria is a major public health threat and conjugative plasmids play a key role in the dissemination of AMR genes among bacterial pathogens. Interestingly, the association between AMR plasmids and pathogens is not random and certain associations spread successfully at a global scale. The burst of genome sequencing has increased the resolution of epidemiological programs, broadening our understanding of plasmid distribution in bacterial populations. Despite the immense value of these studies, our ability to predict future plasmid-bacteria associations remains limited. Numerous empirical studies have recently reported systematic patterns in genetic interactions that enable predictability, in a phenomenon known as global epistasis. In this perspective, we argue that global epistasis patterns hold the potential to predict interactions between plasmids and bacterial genomes, thereby facilitating the prediction of future successful associations. To assess the validity of this idea, we use previously published data to identify global epistasis patterns in clinically relevant plasmid-bacteria associations. Furthermore, using simple mechanistic models of antibiotic resistance, we illustrate how global epistasis patterns may allow us to generate new hypotheses on the mechanisms associated with successful plasmid-bacteria associations. Collectively, we aim at illustrating the relevance of exploring global epistasis in the context of plasmid biology.

Reducing Epistasis and Pleiotropy Can Avoid the Survival of the Flattest Tragedy

This study investigates whether reducing epistasis and pleiotropy enhances mutational robustness in evolutionary adaptation, utilizing an indirect encoded model within the "survival of the flattest" (SoF) fitness landscape. By simulating genetic variations and their phenotypic consequences, we explore organisms' adaptive mechanisms to maintain positions on higher, narrower evolutionary peaks amidst environmental and genetic pressures. Our results reveal that organisms can indeed sustain their advantageous positions by minimizing the complexity of genetic interactions-specifically, by reducing the levels of epistasis and pleiotropy. This finding suggests a counterintuitive strategy for evolutionary stability: simpler genetic architectures, characterized by fewer gene interactions and multifunctional genes, confer a survival advantage by enhancing mutational robustness. This study contributes to our understanding of the genetic underpinnings of adaptability and robustness, challenging traditional views that equate complexity with fitness in dynamic environments.

Self-replicating artificial neural networks give rise to universal evolutionary dynamics

In evolutionary models, mutations are exogenously introduced by the modeler, rather than endogenously introduced by the replicator itself. We present a new deep-learning based computational model, the self-replicating artificial neural network (SeRANN). We train it to (i) copy its own genotype, like a biological organism, which introduces endogenous spontaneous mutations; and (ii) simultaneously perform a classification task that determines its fertility. Evolving 1,000 SeRANNs for 6,000 generations, we observed various evolutionary phenomena such as adaptation, clonal interference, epistasis, and evolution of both the mutation rate and the distribution of fitness effects of new mutations. Our results demonstrate that universal evolutionary phenomena can naturally emerge in a self-replicator model when both selection and mutation are implicit and endogenous. We therefore suggest that SeRANN can be applied to explore and test various evolutionary dynamics and hypotheses.

Changing fitness effects of mutations through long-term bacterial evolution

The distribution of fitness effects of new mutations shapes evolution, but it is challenging to observe how it changes as organisms adapt. Using Escherichia coli lineages spanning 50,000 generations of evolution, we quantify the fitness effects of insertion mutations in every gene. Macroscopically, the fraction of deleterious mutations changed little over time whereas the beneficial tail declined sharply, approaching an exponential distribution. Microscopically, changes in individual gene essentiality and deleterious effects often occurred in parallel; altered essentiality is only partly explained by structural variation. The identity and effect sizes of beneficial mutations changed rapidly over time, but many targets of selection remained predictable because of the importance of loss-of-function mutations. Taken together, these results reveal the dynamic-but statistically predictable-nature of mutational fitness effects.

A dynamical limit to evolutionary adaptation

Natural selection makes evolutionary adaptation possible even if the overwhelming majority of new mutations are deleterious. However, in rapidly evolving populations where numerous linked mutations occur and segregate simultaneously, clonal interference and genetic hitchhiking can limit the efficiency of selection, allowing deleterious mutations to accumulate over time. This can in principle overwhelm the fitness increases provided by beneficial mutations, leading to an overall fitness decline. Here, we analyze the conditions under which evolution will tend to drive populations to higher versus lower fitness. Our analysis focuses on quantifying the boundary between these two regimes, as a function of parameters such as population size, mutation rates, and selection pressures. This boundary represents a state in which adaptation is precisely balanced by Muller's ratchet, and we show that it can be characterized by rapid molecular evolution without any net fitness change. Finally, we consider the implications of global fitness-mediated epistasis and find that under some circumstances, this can drive populations toward the boundary state, which can thus represent a long-term evolutionary attractor.

Genome-scale analysis of interactions between genetic perturbations and natural variation

Interactions between genetic perturbations and segregating loci can cause perturbations to show different phenotypic effects across genetically distinct individuals. To study these interactions on a genome scale in many individuals, we used combinatorial DNA barcode sequencing to measure the fitness effects of 7,700 CRISPRi perturbations targeting 1,712 distinct genes in 169 yeast cross progeny (or segregants). We identified 460 genes whose perturbation has different effects across segregants. Several factors caused perturbations to show variable effects, including baseline segregant fitness, the mean effect of a perturbation across segregants, and interacting loci. We mapped 234 interacting loci and found four hub loci that interact with many different perturbations. Perturbations that interact with a given hub exhibit similar epistatic relationships with the hub and show enrichment for cellular processes that may mediate these interactions. These results suggest that an individual's response to perturbations is shaped by a network of perturbation-locus interactions that cannot be measured by approaches that examine perturbations or natural variation alone.

Environmental modulation of global epistasis in a drug resistance fitness landscape

Interactions between mutations (epistasis) can add substantial complexity to genotype-phenotype maps, hampering our ability to predict evolution. Yet, recent studies have shown that the fitness effect of a mutation can often be predicted from the fitness of its genetic background using simple, linear relationships. This phenomenon, termed global epistasis, has been leveraged to reconstruct fitness landscapes and infer adaptive trajectories in a wide variety of contexts. However, little attention has been paid to how patterns of global epistasis may be affected by environmental variation, despite this variation frequently being a major driver of evolution. This is particularly relevant for the evolution of drug resistance, where antimicrobial drugs may change the environment faced by pathogens and shape their adaptive trajectories in ways that can be difficult to predict. By analyzing a fitness landscape of four mutations in a gene encoding an essential enzyme of P. falciparum (a parasite cause of malaria), here we show that patterns of global epistasis can be strongly modulated by the concentration of a drug in the environment. Expanding on previous theoretical results, we demonstrate that this modulation can be quantitatively explained by how specific gene-by-gene interactions are modified by drug dose. Importantly, our results highlight the need to incorporate potential environmental variation into the global epistasis framework in order to predict adaptation in dynamic environments.

Mutualism reduces the severity of gene disruptions in predictable ways across microbial communities

Predicting evolution in microbial communities is critical for problems from human health to global nutrient cycling. Understanding how species interactions impact the distribution of fitness effects for a focal population would enhance our ability to predict evolution. Specifically, does the type of ecological interaction, such as mutualism or competition, change the average effect of a mutation (i.e., the mean of the distribution of fitness effects)? Furthermore, how often does increasing community complexity alter the impact of species interactions on mutant fitness? To address these questions, we created a transposon mutant library in Salmonella enterica and measured the fitness of loss of function mutations in 3,550 genes when grown alone versus competitive co-culture or mutualistic co-culture with Escherichia coli and Methylorubrum extorquens. We found that mutualism reduces the average impact of mutations, while competition had no effect. Additionally, mutant fitness in the 3-species communities can be predicted by averaging the fitness in each 2-species community. Finally, we discovered that in the mutualism S. enterica obtained vitamins and more amino acids than previously known. Our results suggest that species interactions can predictably impact fitness effect distributions, in turn suggesting that evolution may ultimately be predictable in multi-species communities.

Evolution of haploid and diploid populations reveals common, strong, and variable pleiotropic effects in non-home environments

Adaptation is driven by the selection for beneficial mutations that provide a fitness advantage in the specific environment in which a population is evolving. However, environments are rarely constant or predictable. When an organism well adapted to one environment finds itself in another, pleiotropic effects of mutations that made it well adapted to its former environment will affect its success. To better understand such pleiotropic effects, we evolved both haploid and diploid barcoded budding yeast populations in multiple environments, isolated adaptive clones, and then determined the fitness effects of adaptive mutations in 'non-home' environments in which they were not selected. We find that pleiotropy is common, with most adaptive evolved lineages showing fitness effects in non-home environments. Consistent with other studies, we find that these pleiotropic effects are unpredictable: they are beneficial in some environments and deleterious in others. However, we do find that lineages with adaptive mutations in the same genes tend to show similar pleiotropic effects. We also find that ploidy influences the observed adaptive mutational spectra in a condition-specific fashion. In some conditions, haploids and diploids are selected with adaptive mutations in identical genes, while in others they accumulate mutations in almost completely disjoint sets of genes.

A dynamical limit to evolutionary adaptation

Natural selection makes evolutionary adaptation possible even if the overwhelming majority of new mutations are deleterious. However, in rapidly evolving populations where numerous linked mutations occur and segregate simultaneously, clonal interference and genetic hitchhiking can limit the efficiency of selection, allowing deleterious mutations to accumulate over time. This can in principle overwhelm the fitness increases provided by beneficial mutations, leading to an overall fitness decline. Here, we analyze the conditions under which evolution will tend to drive populations to higher versus lower fitness. Our analysis focuses on quantifying the boundary between these two regimes, as a function of parameters such as population size, mutation rates, and selection pressures. This boundary represents a state in which adaptation is precisely balanced by Muller's ratchet, and we show that it can be characterized by rapid molecular evolution without any net fitness change. Finally, we consider the implications of global fitness-mediated epistasis, and find that under some circumstances this can drive populations towards the boundary state, which can thus represent a long-term evolutionary attractor.

Extreme Sensitivity of Fitness to Environmental Conditions: Lessons from #1BigBatch

The phrase "survival of the fittest" has become an iconic descriptor of how natural selection works. And yet, precisely measuring fitness, even for single-celled microbial populations growing in controlled laboratory conditions, remains a challenge. While numerous methods exist to perform these measurements, including recently developed methods utilizing DNA barcodes, all methods are limited in their precision to differentiate strains with small fitness differences. In this study, we rule out some major sources of imprecision, but still find that fitness measurements vary substantially from replicate to replicate. Our data suggest that very subtle and difficult to avoid environmental differences between replicates create systematic variation across fitness measurements. We conclude by discussing how fitness measurements should be interpreted given their extreme environment dependence. This work was inspired by the scientific community who followed us and gave us tips as we live tweeted a high-replicate fitness measurement experiment at #1BigBatch.

Best Practices in Designing, Sequencing, and Identifying Random DNA Barcodes

Random DNA barcodes are a versatile tool for tracking cell lineages, with applications ranging from development to cancer to evolution. Here, we review and critically evaluate barcode designs as well as methods of barcode sequencing and initial processing of barcode data. We first demonstrate how various barcode design decisions affect data quality and propose a new design that balances all considerations that we are currently aware of. We then discuss various options for the preparation of barcode sequencing libraries, including inline indices and Unique Molecular Identifiers (UMIs). Finally, we test the performance of several established and new bioinformatic pipelines for the extraction of barcodes from raw sequencing reads and for error correction. We find that both alignment and regular expression-based approaches work well for barcode extraction, and that error-correction pipelines designed specifically for barcode data are superior to generic ones. Overall, this review will help researchers to approach their barcoding experiments in a deliberate and systematic way.

Epistasis and evolution: recent advances and an outlook for prediction

As organisms evolve, the effects of mutations change as a result of epistatic interactions with other mutations accumulated along the line of descent. This can lead to shifts in adaptability or robustness that ultimately shape subsequent evolution. Here, we review recent advances in measuring, modeling, and predicting epistasis along evolutionary trajectories, both in microbial cells and single proteins. We focus on simple patterns of global epistasis that emerge in this data, in which the effects of mutations can be predicted by a small number of variables. The emergence of these patterns offers promise for efforts to model epistasis and predict evolution.

Global epistasis on fitness landscapes

Epistatic interactions between mutations add substantial complexity to adaptive landscapes and are often thought of as detrimental to our ability to predict evolution. Yet, patterns of global epistasis, in which the fitness effect of a mutation is well-predicted by the fitness of its genetic background, may actually be of help in our efforts to reconstruct fitness landscapes and infer adaptive trajectories. Microscopic interactions between mutations, or inherent nonlinearities in the fitness landscape, may cause global epistasis patterns to emerge. In this brief review, we provide a succinct overview of recent work about global epistasis, with an emphasis on building intuition about why it is often observed. To this end, we reconcile simple geometric reasoning with recent mathematical analyses, using these to explain why different mutations in an empirical landscape may exhibit different global epistasis patterns-ranging from diminishing to increasing returns. Finally, we highlight open questions and research directions. This article is part of the theme issue 'Interdisciplinary approaches to predicting evolutionary biology'.

Mutualism reduces the severity of gene disruptions in predictable ways across microbial communities

Predicting evolution in microbial communities is critical for problems from human health to global nutrient cycling. Understanding how species interactions impact the distribution of fitness effects for a focal population would enhance our ability to predict evolution. Specifically, it would be useful to know if the type of ecological interaction, such as mutualism or competition, changes the average effect of a mutation (i.e., the mean of the distribution of fitness effects). Furthermore, how often does increasing community complexity alter the impact of species interactions on mutant fitness? To address these questions, we created a transposon mutant library in Salmonella enterica and measured the fitness of loss of function mutations in 3,550 genes when grown alone versus competitive co-culture or mutualistic co-culture with Escherichia coli and Methylorubrum extorquens. We found that mutualism reduces the average impact of mutations, while competition had no effect. Additionally, mutant fitness in the 3-species communities can be predicted by averaging the fitness in each 2-species community. Finally, the fitness effects of several knockouts in the mutualistic communities were surprising. We discovered that S. enterica is obtaining a different source of carbon and more vitamins and amino acids than we had expected. Our results suggest that species interactions can predictably impact fitness effect distributions, in turn suggesting that evolution may ultimately be predictable in multi-species communities.

Dynamics of synthetic yeast chromosome evolution shaped by hierarchical chromatin organization

Synthetic genome evolution provides a dynamic approach for systematically and straightforwardly exploring evolutionary processes. Synthetic Chromosome Rearrangement and Modification by LoxP-mediated Evolution (SCRaMbLE) is an evolutionary system intrinsic to the synthetic yeast genome that can rapidly drive structural variations. Here, we detect over 260 000 rearrangement events after the SCRaMbLEing of a yeast strain harboring 5.5 synthetic yeast chromosomes (synII, synIII, synV, circular synVI, synIXR and synX). Remarkably, we find that the rearrangement events exhibit a specific landscape of frequency. We further reveal that the landscape is shaped by the combined effects of chromatin accessibility and spatial contact probability. The rearrangements tend to occur in 3D spatially proximal and chromatin-accessible regions. The enormous numbers of rearrangements mediated by SCRaMbLE provide a driving force to potentiate directed genome evolution, and the investigation of the rearrangement landscape offers mechanistic insights into the dynamics of genome evolution.

The Fitness of Beta-Lactamase Mutants Depends Nonlinearly on Resistance Level at Sublethal Antibiotic Concentrations

Adaptive evolutionary processes are constrained by the availability of mutations which cause a fitness benefit and together make up the fitness landscape, which maps genotype space onto fitness under specified conditions. Experimentally derived fitness landscapes have demonstrated a predictability to evolution by identifying limited "mutational routes" that evolution by natural selection may take between low and high-fitness genotypes. However, such studies often utilize indirect measures to determine fitness. We estimated the competitive fitness of mutants relative to all single-mutation neighbors to describe the fitness landscape of three mutations in a β-lactamase enzyme. Fitness assays were performed at sublethal concentrations of the antibiotic cefotaxime in a structured and unstructured environment. In the unstructured environment, the antibiotic selected for higher-resistance types-but with an equivalent fitness for a subset of mutants, despite substantial variation in resistance-resulting in a stratified fitness landscape. In contrast, in a structured environment with a low antibiotic concentration, antibiotic-susceptible genotypes had a relative fitness advantage, which was associated with antibiotic-induced filamentation. These results cast doubt that highly resistant genotypes have a unique selective advantage in environments with subinhibitory concentrations of antibiotics and demonstrate that direct fitness measures are required for meaningful predictions of the accessibility of evolutionary routes. IMPORTANCE The evolution of antibiotic-resistant bacterial populations underpins the ongoing antibiotic resistance crisis. We aim to understand how antibiotic-degrading enzymes can evolve to cause increased resistance, how this process is constrained, and whether it can be predictable. To this end, competition experiments were performed with a combinatorially complete set of mutants of a β-lactamase gene subject to subinhibitory concentrations of the antibiotic cefotaxime. While some mutations confer on their hosts high resistance to cefotaxime, in competition these mutations do not always confer a selective advantage. Specifically, high-resistance mutants had equivalent fitnesses despite different resistance levels and even had selective disadvantages under conditions involving spatial structure. Together, our findings suggest that the relationship between resistance level and fitness at subinhibitory concentrations is complex; predicting the evolution of antibiotic resistance requires knowledge of the conditions that select for resistant genotypes and the selective advantage evolved types have over their predecessors.

Widespread epistasis among beneficial genetic variants revealed by high-throughput genome editing

The phenotypic effect of any genetic variant can be altered by variation at other genomic loci. Known as epistasis, these genetic interactions shape the genotype-phenotype map of every species, yet their origins remain poorly understood. To investigate this, we employed high-throughput genome editing to measure the fitness effects of 1,826 naturally polymorphic variants in four strains of Saccharomyces cerevisiae. About 31% of variants affect fitness, of which 24% have strain-specific fitness effects indicative of epistasis. We found that beneficial variants are more likely to exhibit genetic interactions and that these interactions can be mediated by specific traits such as flocculation ability. This work suggests that adaptive evolution will often involve trade-offs where a variant is only beneficial in some genetic backgrounds, potentially explaining why many beneficial variants remain polymorphic. In sum, we provide a framework to understand the factors influencing epistasis with single-nucleotide resolution, revealing widespread epistasis among beneficial variants.

Quantifying the local adaptive landscape of a nascent bacterial community

The fitness effects of all possible mutations available to an organism largely shape the dynamics of evolutionary adaptation. Yet, whether and how this adaptive landscape changes over evolutionary times, especially upon ecological diversification and changes in community composition, remains poorly understood. We sought to fill this gap by analyzing a stable community of two closely related ecotypes ("L" and "S") shortly after they emerged within the E. coli Long-Term Evolution Experiment (LTEE). We engineered genome-wide barcoded transposon libraries to measure the invasion fitness effects of all possible gene knockouts in the coexisting strains as well as their ancestor, for many different, ecologically relevant conditions. We find consistent statistical patterns of fitness effect variation across both genetic background and community composition, despite the idiosyncratic behavior of individual knockouts. 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 shape the adaptive landscape in a nascent ecological community.

Transcriptional Potential Determines the Adaptability of Escherichia coli Strains with Different Fitness Backgrounds

Adaptation through the fitness landscape may be influenced by the gene pool or expression network. However, genetic factors that determine the contribution of beneficial mutations during adaptive evolution are poorly understood. In this study, we experimentally evolved wild-type Escherichia coli K-12 MG1655 and its isogenic derivative that has two additional replication origins and shows higher background fitness. During the short time of experimental evolution, the fitness gains of the two E. coli strains with different fitness backgrounds converged. Populational genome sequencing revealed various mutations with different allele frequencies in evolved populations. Several mutations occurred in genes affecting transcriptional regulation (e.g., RNA polymerase subunit, RNase, ppGpp synthetase, and transcription termination/antitermination factor genes). When we introduced mutations into the ancestral E. coli strains, beneficial effects tended to be lower in the ancestor with higher initial fitness. Replication rate analysis showed that the various replication indices do not correlate with the growth rate. Transcriptome profiling showed that gene expression and gene ontology are markedly enriched in populations with lower background fitness after experimental evolution. Further, the degree of transcriptional change was proportional to the fitness gain. Thus, the evolutionary trajectories of bacteria with different fitness backgrounds can be complex and counterintuitive. Notably, transcriptional change is a major contributor to adaptability. IMPORTANCE Predicting the adaptive potential of bacterial populations can be difficult due to their complexity and dynamic environmental conditions. Also, epistatic interaction between mutations affects the adaptive trajectory. Nevertheless, next-generation sequencing sheds light on understanding evolutionary dynamics through high-throughput genome and transcriptome information. Experimental evolution of two E. coli strains with different background fitness showed that the trajectories of fitness gain, which slowed down during the later stages of evolution, became convergent. This suggests that the adaptability of bacteria can be counterintuitive and that predicting the evolutionary path of bacteria can be difficult even in a constant environment. In addition, transcriptional change is associated with fitness gain during the evolutionary process. Thus, the adaptability of cells depends on their intrinsic genetic capacity for a given evolutionary period. This should be considered when genetically engineered bacteria are optimized through adaptive evolution.

Conflicting effects of recombination on the evolvability and robustness in neutrally evolving populations

Understanding the benefits and costs of recombination under different scenarios of evolutionary adaptation remains an open problem for theoretical and experimental research. In this study, we focus on finite populations evolving on neutral networks comprising viable and unfit genotypes. We provide a comprehensive overview of the effects of recombination by jointly considering different measures of evolvability and mutational robustness over a broad parameter range, such that many evolutionary regimes are covered. We find that several of these measures vary non-monotonically with the rates of mutation and recombination. Moreover, the presence of unfit genotypes that introduce inhomogeneities in the network of viable states qualitatively alters the effects of recombination. We conclude that conflicting trends induced by recombination can be explained by an emerging trade-off between evolvability on the one hand, and mutational robustness on the other. Finally, we discuss how different implementations of the recombination scheme in theoretical models can affect the observed dependence on recombination rate through a coupling between recombination and genetic drift.

Challenges and potential solutions for studying the genetic and phenotypic architecture of adaptation in microbes

All organisms are defined by the makeup of their DNA. Over billions of years, the structure and information contained in that DNA, often referred to as genetic architecture, have been honed by a multitude of evolutionary processes. Mutations that cause genetic elements to change in a way that results in beneficial phenotypic change are more likely to survive and propagate through the population in a process known as adaptation. Recent work reveals that the genetic targets of adaptation are varied and can change with genetic background. Further, seemingly similar adaptive mutations, even within the same gene, can have diverse and unpredictable effects on phenotype. These challenges represent major obstacles in predicting adaptation and evolution. In this review, we cover these concepts in detail and identify three emerging synergistic solutions: higher-throughput evolution experiments combined with updated genotype-phenotype mapping strategies and physiological models. Our review largely focuses on recent literature in yeast, and the field seems to be on the cusp of a new era with regard to studying the predictability of evolution.

Mutational robustness changes during long-term adaptation in laboratory budding yeast populations

As an adapting population traverses the fitness landscape, its local neighborhood (i.e., the collection of fitness effects of single-step mutations) can change shape because of interactions with mutations acquired during evolution. These changes to the distribution of fitness effects can affect both the rate of adaptation and the accumulation of deleterious mutations. However, while numerous models of fitness landscapes have been proposed in the literature, empirical data on how this distribution changes during evolution remains limited. In this study, we directly measure how the fitness landscape neighborhood changes during laboratory adaptation. Using a barcode-based mutagenesis system, we measure the fitness effects of 91 specific gene disruption mutations in genetic backgrounds spanning 8000–10,000 generations of evolution in two constant environments. We find that the mean of the distribution of fitness effects decreases in one environment, indicating a reduction in mutational robustness, but does not change in the other. We show that these distribution-level patterns result from differences in the relative frequency of certain patterns of epistasis at the level of individual mutations, including fitness-correlated and idiosyncratic epistasis.

Microbial experimental evolution in a massively multiplexed and high-throughput era

Experimental evolution with microbial model systems has transformed our understanding of the basic rules underlying ecology and evolution. Experiments leveraging evolution as a central feature put evolutionary theories to the test, and modern sequencing and engineering tools then characterized the molecular basis of adaptation. As theory and experimentations refined our understanding of evolution, a need to increase throughput and experimental complexity has emerged. Here, we summarize recent technologies that have made high-throughput experiments practical and highlight studies that have capitalized on these tools, defining an exciting new era in microbial experimental evolution. Multiple research directions previously limited by experimental scale are now accessible for study and we believe applying evolutionary lessons from in vitro studies onto these applied settings has the potential for major innovations and discoveries across ecology and medicine.

Shepherd: Accurate Clustering for Correcting DNA Barcode Errors

Motivation

DNA barcodes are short, random nucleotide sequences introduced into cell populations to track the relative counts of hundreds of thousands of individual lineages over time. Lineage tracking is widely applied, e.g. to understand evolutionary dynamics in microbial populations and the progression of breast cancer in humans. Barcode sequences are unknown upon insertion and must be identified using next-generation sequencing technology, which is error prone. In this study, we frame the barcode error correction task as a clustering problem with the aim to identify true barcode sequences from noisy sequencing data. We present Shepherd, a novel clustering method that is based on an indexing system of barcode sequences using k-mers, and a Bayesian statistical test incorporating a substitution error rate to distinguish true from error sequences.

Results

When benchmarking with synthetic data, Shepherd provides barcode count estimates that are significantly more accurate than state-of-the-art methods, producing 10–150 times fewer spurious lineages. For empirical data, Shepherd produces results that are consistent with the improvements seen on synthetic data. These improvements enable higher resolution lineage tracking and more accurate estimates of biologically relevant quantities, e.g. the detection of small effect mutations.

Availability and implementation

A Python implementation of Shepherd is freely available at: https://www.github.com/Nik-Tavakolian/Shepherd.

Supplementary information

Supplementary data are available at Bioinformatics online.

The distribution of fitness effects of spontaneous mutations in Chlamydomonas reinhardtii inferred using frequency changes under experimental evolution

The distribution of fitness effects (DFE) for new mutations is fundamental for many aspects of population and quantitative genetics. In this study, we have inferred the DFE in the single-celled alga Chlamydomonas reinhardtii by estimating changes in the frequencies of 254 spontaneous mutations under experimental evolution and equating the frequency changes of linked mutations with their selection coefficients. We generated seven populations of recombinant haplotypes by crossing seven independently derived mutation accumulation lines carrying an average of 36 mutations in the haploid state to a mutation-free strain of the same genotype. We then allowed the populations to evolve under natural selection in the laboratory by serial transfer in liquid culture. We observed substantial and repeatable changes in the frequencies of many groups of linked mutations, and, surprisingly, as many mutations were observed to increase as decrease in frequency. Mutation frequencies were highly repeatable among replicates, suggesting that selection was the cause of the observed allele frequency changes. We developed a Bayesian Monte Carlo Markov Chain method to infer the DFE. This computes the likelihood of the observed distribution of changes of frequency, and obtains the posterior distribution of the selective effects of individual mutations, while assuming a two-sided gamma distribution of effects. We infer that the DFE is a highly leptokurtic distribution, and that approximately equal proportions of mutations have positive and negative effects on fitness. This result is consistent with what we have observed in previous work on a different C. reinhardtii strain, and suggests that a high fraction of new spontaneously arisen mutations are advantageous in a simple laboratory environment.

Idiosyncratic epistasis leads to global fitness-correlated trends

Epistasis can markedly affect evolutionary trajectories. In recent decades, protein-level fitness landscapes have revealed extensive idiosyncratic epistasis among specific mutations. By contrast, other work has found ubiquitous and apparently nonspecific patterns of global diminishing-returns and increasing-costs epistasis among mutations across the genome. Here, we used a hierarchical CRISPR gene drive system to construct all combinations of 10 missense mutations from across the genome in budding yeast and measured their fitness in six environments. We show that the resulting fitness landscapes exhibit global fitness-correlated trends but that these trends emerge from specific idiosyncratic interactions. We thus provide experimental validation of recent theoretical work arguing that fitness-correlated trends can emerge as the generic consequence of idiosyncratic epistasis.

Rapid intraspecies evolution of fitness effects of yeast genes

Organisms within species have numerous genetic and phenotypic variations. Growing evidences show intraspecies variation of mutant phenotypes may be more complicated than expected. Current studies on intraspecies variations of mutant phenotypes are limited to just a few strains. This study investigated the intraspecies variation of fitness effects of 5,630 gene mutants in ten Saccharomyces cerevisiae strains using CRISPR–Cas9 screening. We found that the variability of fitness effects induced by gene disruptions is very large across different strains. Over 75% of genes affected cell fitness in a strain-specific manner to varying degrees. The strain specificity of the fitness effect of a gene is related to its evolutionary and functional properties. Subsequent analysis revealed that younger genes, especially those newly acquired in S. cerevisiae species, are more likely to be strongly strain-specific. Intriguingly, there seems to exist a ceiling of fitness effect size for strong strain-specific genes, and among them, the newly acquired genes are still evolving and have yet to reach this ceiling. Additionally, for a large proportion of protein complexes, the strain specificity profile is inconsistent among genes encoding the same complex. Taken together, these results offer a genome-wide map of intraspecies variation for fitness effect as a mutant phenotype and provide an updated insight on intraspecies phenotypic evolution.

Overdominant and partially dominant mutations drive clonal adaptation in diploid Saccharomyces cerevisiae

Identification of adaptive targets in experimental evolution typically relies on extensive replication and genetic reconstruction. An alternative approach is to directly assay all mutations in an evolved clone by generating pools of segregants that contain random combinations of evolved mutations. Here, we apply this method to six Saccharomyces cerevisiae clones isolated from four diploid populations that were clonally evolved for 2,000 generations in rich glucose medium. Each clone contains 17-26 mutations relative to the ancestor. We derived intermediate genotypes between the founder and the evolved clones by bulk mating sporulated cultures of the evolved clones to a barcoded haploid version of the ancestor. We competed the resulting barcoded diploids en masse and quantified fitness in the experimental and alternative environments by barcode sequencing. We estimated average fitness effects of evolved mutations using barcode-based fitness assays and whole genome sequencing for a subset of segregants. In contrast to our previous work with haploid evolved clones, we find that diploids carry fewer beneficial mutations, with modest fitness effects (up to 5.4%) in the environment in which they arose. In agreement with theoretical expectations, reconstruction experiments show that all mutations with a detectable fitness effect manifest some degree of dominance over the ancestral allele, and most are overdominant. Genotypes with lower fitness effects in alternative environments allowed us to identify conditions that drive adaptation in our system.

Population genetics of polymorphism and divergence in rapidly evolving populations

In rapidly evolving populations, numerous beneficial and deleterious mutations can arise and segregate within a population at the same time. In this regime, evolutionary dynamics cannot be analyzed using traditional population genetic approaches that assume that sites evolve independently. Instead, the dynamics of many loci must be analyzed simultaneously. Recent work has made progress by first analyzing the fitness variation within a population, and then studying how individual lineages interact with this traveling fitness wave. However, these "traveling wave" models have previously been restricted to extreme cases where selection on individual mutations is either much faster or much slower than the typical coalescent timescale Tc. In this work, we show how the traveling wave framework can be extended to intermediate regimes in which the scaled fitness effects of mutations (Tcs) are neither large nor small compared to one. This enables us to describe the dynamics of populations subject to a wide range of fitness effects, and in particular, in cases where it is not immediately clear which mutations are most important in shaping the dynamics and statistics of genetic diversity. We use this approach to derive new expressions for the fixation probabilities and site frequency spectra of mutations as a function of their scaled fitness effects, along with related results for the coalescent timescale Tc and the rate of adaptation or Muller's ratchet. We find that competition between linked mutations can have a dramatic impact on the proportions of neutral and selected polymorphisms, which is not simply summarized by the scaled selection coefficient Tcs. We conclude by discussing the implications of these results for population genetic inferences.

Robustness: linking strain design to viable bioprocesses

Microbial cell factories are becoming increasingly popular for the sustainable production of various chemicals. Metabolic engineering has led to the design of advanced cell factories; however, their long-term yield, titer, and productivity falter when scaled up and subjected to industrial conditions. This limitation arises from a lack of robustness - the ability to maintain a constant phenotype despite the perturbations of such processes. This review describes predictable and stochastic industrial perturbations as well as state-of-the-art technologies to counter process variability. Moreover, we distinguish robustness from tolerance and discuss the potential of single-cell studies for improving system robustness. Finally, we highlight ways of achieving consistent and comparable quantification of robustness that can guide the selection of strains for industrial bioprocesses.

Genetic basis of a spontaneous mutation’s expressivity

Genetic background often influences the phenotypic consequences of mutations, resulting in variable expressivity. How standing genetic variants collectively cause this phenomenon is not fully understood. Here, we comprehensively identify loci in a budding yeast cross that impact the growth of individuals carrying a spontaneous missense mutation in the nuclear-encoded mitochondrial ribosomal gene MRP20. Initial results suggested that a single large effect locus influences the mutation’s expressivity, with one allele causing inviability in mutants. However, further experiments revealed this simplicity was an illusion. In fact, many additional loci shape the mutation’s expressivity, collectively leading to a wide spectrum of mutational responses. These results exemplify how complex combinations of alleles can produce a diversity of qualitative and quantitative responses to the same mutation.

Heterogeneity of the GFP fitness landscape and data-driven protein design

Studies of protein fitness landscapes reveal biophysical constraints guiding protein evolution and empower prediction of functional proteins. However, generalisation of these findings is limited due to scarceness of systematic data on fitness landscapes of proteins with a defined evolutionary relationship. We characterized the fitness peaks of four orthologous fluorescent proteins with a broad range of sequence divergence. While two of the four studied fitness peaks were sharp, the other two were considerably flatter, being almost entirely free of epistatic interactions. Mutationally robust proteins, characterized by a flat fitness peak, were not optimal templates for machine-learning-driven protein design - instead, predictions were more accurate for fragile proteins with epistatic landscapes. Our work paves insights for practical application of fitness landscape heterogeneity in protein engineering.

Bulk-Fitness Measurements Using Barcode Sequencing Analysis in Yeast

The use of DNA barcodes for determining changes in genotype frequencies has been instrumental to increase the scale at which we can phenotype strain libraries by using next-generation sequencing technologies. Here, we describe the determination of strain fitness for thousands of yeast strains simultaneously in a single assay using recent innovations that increase the precision of these measurements, such as the inclusion of unique-molecular identifiers (UMIs) and purification by solid-phase reverse immobilization (SPRI) beads.

The population genetics of collateral resistance and sensitivity

Resistance mutations against one drug can elicit collateral sensitivity against other drugs. Multi-drug treatments exploiting such trade-offs can help slow down the evolution of resistance. However, if mutations with diverse collateral effects are available, a treated population may evolve either collateral sensitivity or collateral resistance. How to design treatments robust to such uncertainty is unclear. We show that many resistance mutations in Escherichia coli against various antibiotics indeed have diverse collateral effects. We propose to characterize such diversity with a joint distribution of fitness effects (JDFE) and develop a theory for describing and predicting collateral evolution based on simple statistics of the JDFE. We show how to robustly rank drug pairs to minimize the risk of collateral resistance and how to estimate JDFEs. In addition to practical applications, these results have implications for our understanding of evolution in variable environments.

Drugs known as antibiotics are the main treatment for most serious infections caused by bacteria. However, many bacteria are acquiring genetic mutations that make them resistant to the effects of one or more types of antibiotics, making them harder to eliminate.

One way to tackle drug-resistant bacteria is to develop new types of antibiotics; however, in recent years, the rate at which new antibiotics have become available has been dwindling. Using two or more existing drugs, one after another, can also be an effective way to eliminate resistant bacteria. The success of any such ‘multi-drug’ treatment lies in being able to predict whether mutations that make the bacteria resistant to one drug simultaneously make it sensitive to another, a phenomenon known as collateral sensitivity.

Different resistance mutations may have different collateral effects: some may increase the bacteria’s sensitivity to the second drug, while others might make the bacteria more resistant. However, it is currently unclear how to design robust multi-drug treatments that take this diversity of collateral effects into account.

Here, Ardell and Kryazhimskiy used a concept called JDFE (short for the joint distribution of fitness effects) to describe the diversity of collateral effects in a population of bacteria exposed to a single drug. This information was then used to mathematically model how collateral effects evolved in the population over time.

Ardell and Kryazhimskiy showed that this approach can predict how likely a population is to become collaterally sensitive or collaterally resistant to a second antibiotic. Drug pairs can then be ranked according to the risk of collateral resistance emerging, so long as information on the variety of resistance mutations available to the bacteria are included in the model.

Each year, more than 700,000 people die from infections caused by bacteria that are resistant to one or more antibiotics. The findings of Ardell and Kryazhimskiy may eventually help clinicians design multi-drug treatments that effectively eliminate bacterial infections and help to prevent more bacteria from evolving resistance to antibiotics. However, to achieve this goal, more research is needed to fully understand the range collateral effects caused by resistance mutations.