Predicting the composition and function of microbial communities in a given habitat is a major aspiration in microbiome biology. To realize this goal, it is critical to identify which features of microbial communities are reproducible and predictable, which are not, and why. We have addressed this question by studying the assembly of hundreds of communities in simple replicate habitats and connecting the experiments with modeling. We have found that microbial community assembly is generally reproducible and convergent at higher levels of taxonomic community organization and reflects an emergent metabolic self-organization between different functional groups whose ratios can be quantitatively explained with simple models. In turn, taxonomic divergence among replicate communities may arise from multi-stability in population dynamics, which can also lead to alternative, dynamically stable functional states.

Microbial communities frequently invade one another as a whole, a phenomenon known as community coalescence. Despite its potential importance for the assembly, dynamics and stability of microbial consortia, as well as its prospective utility for microbiome engineering, our understanding of the processes that govern it is still very limited. Theory has suggested that microbial communities may exhibit cohesiveness in the face of invasions, leading to correlated invasional outcomes where the fate of a given taxon is determined by that of other members of its community. We have performed over a hundred invasion and coalescence experiments with microbial communities of various origins assembled in two different synthetic environments. We show that the dominant members of the primary communities can recruit their rarer partners during coalescence (top-down co-selection) and also be recruited by them (bottom-up co-selection). With the aid of a consumer-resource model, we found that the emergence of top-down or bottom-up cohesiveness is modulated by the structure of the underlying cross-feeding networks that sustain the coalesced communities.

Most organisms grow spatially whether we look at viruses spreading within a host tissue or invasive species colonizing a new continent. Evolution typically selects for higher expansion rates during spatial growth, but it has been suggested that slower expanders can take over under certain conditions. Here, we report an experimental observation of such evolutionary dynamics. Moreover, we find that the slower expander wins even when the existing theory predicts otherwise. Previous attempts to describe spatial competition only accounted for the global competition mediated by the expansion velocities, but neglected the local competition mediated by the growth rates or other factors. We developed a theory that includes both local and global competition and found that it quantitatively describes our experimental data. In particular, the theory predicted that a slower, but more competitive mutant forms a dented, V-shaped sector as it takes over the expansion front. We indeed observed such sectors experimentally and found that their shapes accurately matched our theoretical predictions. The theory also identified how sector shape changes from the traditional bulge to the novel dent as one varies the local and global fitness. We tested these predictions in simulation, which further illustrated how the tradeoff between local and global fitness could emerge from a mechanistic model of growth and dispersal. Taken together, our results establish a new framework to understand evolutionary and ecological dynamics in expanding populations with arbitrary frequency- and density-dependent selection.

Mathematical models for cooperative biological systems have shown a multitude of tipping points involving regime shifts. Such tipping points can be studied through bifurcation theory. In this talk, I will introduce two different mathematical models considering cooperation and parasitism. The first model describes the dynamics of autocatalysis and is governed by a saddle-node bifurcation. I will discuss the phenomenon behind delayed transitions, given by a ghost that involves extremely long delays towards the system's extinction. The second model, which considers an autocatalytic species with an obligate parasite, experiences a regime shift through a global bifurcation involving a quasi-neutral line. This bifurcation also involves extremely long delays (ghost states) towards the survival of the autocatalytic species or co-extinction. The emergence of scaling laws and the role of stochasticity will be discussed.

Bacteria must monitor the dynamics of their environment continuously, in order to adapt to present conditions and anticipate future changes. But anticipation requires processing temporal information, which in turn requires memory. In this talk I will discuss a possible mechanism through which cells can perform such dynamical information processing by leveraging the recurrent architecture of gene regulatory networks. I will also discuss the size requirements that recurrent networks must fulfill under realistic biological conditions, in order to provide such a dynamic representation of the world. Finally, I will argue that recurrence in biological networks enables not only short-term memory, but also the long-term storage of information, through a phenomenon reminiscent of generalized chaos synchronization.

Susceptible–infected–removed (SIR) models and their extensions are widely used to describe the dynamics of infection spreading. Certain generic features of epidemics are well-illustrated by these models, which can be remarkably good at reproducing empirical data through suitably chosen parameters. However, this does not assure a good job anticipating the forthcoming stages of the process. To illustrate this point, we accurately describe the propagation of COVID-19 in Spain using one such model and show that predictions for its subsequent evolution are disparate, even contradictory. The future of ongoing epidemics is so sensitive to parameter values that predictions are only meaningful within a narrow time window and in probabilistic terms, much as what we are used to in weather forecasts (PNAS 117:26190 (2020).

The emergence of drug resistance is a major limitation to the efficacy of anti-cancer targeted therapies. Drug-resistant mutants that are not present when therapy is initiated may derive from sub-populations of drug-tolerant persister cells, which are known to survive treatment for extended period of time. Whether persister cancer cells pre-exist therapeutic stress or are drug-induced is unclear. Additionally, persister cells can transiently compromise fidelity of DNA replication, but it remains unknown if the mutation rate concomitantly and quantitatively increases.

Here, by combining mathematical modeling and experimental characterization, we show that in colorectal cancer, persister cells are induced by, and do not predate, drug treatment. We then establish a two-step fluctuation assay generalizing the test originally developed by Luria and Delbrück, to measure mutation rates of tumor cells.

We find that colorectal cancer persister cells show a 10- to 100-fold increase of their mutation rate when exposed to clinically approved therapies. Taken together, our results provide a quantitative and predictive framework that could be used to design strategies to restrict therapeutic tolerance and resistance in tumors. The mathematical framework we developed could be broadly used to assess how environmental conditions affect mutability in mammalian cells.

Microorganisms such as 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 new genomic approach to systematically identify bacteria that can exploit public goods produced during the degradation of polysaccharides. Focusing on chitin – the second most abundant biopolymer on the planet, we show that public good exploiters are active in natural marine microbial communities that assemble on chitin particles, invading during early stages of colonization and potentially hindering degradation. Unlike in classical studies of social evolution, exploiters and polysaccharide degraders are not isogenic and instead 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.

In this talk, I will review our efforts in establishing a nonequilibrium landscape and flux theory for evolution. We show explicitly the conventional Wright’s gradient adaptive landscape based on the mean fitness is inadequate to describe the general evolutionary dynamics. We illustrate that the intrinsic potential as being the Lyapunov function(monotonically decreasing in time) does exist and can define the adaptive landscape of the general evolution dynamics for studying global stability. The driving force determining the dynamics can be decomposed into the gradient of the potential landscape and the curl steady state probability flux. Non-zero flux causes detailed balance breaking and measures how far the evolution from equilibrium state. The gradient of intrinsic potential and curl flux are perpendicular to each other in zero fluctuation limit resembling electric and magnetic forces on electrons. We quantified intrinsic energy, entropy and free energy of evolution and constructed non-equilibrium thermodynamics. The intrinsic non-equilibrium free energy is a Lyapunov function. Both intrinsic potential and free energy can be used to quantify the global stability and robustness of evolution. We investigated an example of three allele evolutionary dynamics with frequency dependent selection (detailed balance broken). We uncovered the underlying single, triple, and limit cycle attractor landscapes. We found quantitative criterion for stability through landscape topography. We also quantified dominant evolution pathways and found that these paths do not usually follow the potential gradient and are irreversible due to non-zero flux. We generalized the original Fisher’s fundamental theorem to the general (i.e., frequency dependent selection) regime of evolution by linking the adaptive rate with not only genetic variance related to the potential but also the flux. We show that there is an optimum potential where the curl flux resulting from biotic interactions of the individuals within a species or between the species can sustain an endless evolution even if the physical environment is unchanged. We offer a theoretical basis for explaining the corresponding Red Queen hypothesis proposed by Van Valen. Our work provides a theoretical foundation for the evolutionary dynamics.

Understanding and predicting how gut microbiota respond to antibiotics is a key question of both fundamental and applied significance. Using the data from the two identical experiments where mice were treated with antibiotics, we explored how predictable different community characteristics and responses of individual taxa are. We found that total biomass, species richness and diversity all decline and then recover but the temporal dynamics is less predictable. During the antibiotics treatment, the variance across replicates was large. The community composition response was predictable at the coarse level but less so at finer taxonomic levels. We detected consistent patterns in physiological trait responses, with more oxygen-tolerant, non-sporulating and motile taxa increasing during the treatment. Our analyses also suggest that, at the individual taxon level, the initial response to antibiotics may be more predictable than the subsequent recovery, which may be explained by the stronger dependence of recovery on community composition. The results suggest that different community characteristics may differ in their predictability, and, that predictability is scale-dependent.Understanding the fitness costs of mutations is essential for our comprehension of the architecture of biological systems and its evolution. Many studies of this type, however, have traditionally focused on "simple" mutations. By simple, I refer here to mutations in molecular elements with a specific function, e.g., an enzyme catalyzing a particular biochemical reaction or a transcription factor linked to the activation of a given gene. In this talk, I will focus instead on the study of the fitness costs of “complex” pleiotropic mutations.

One encounters three potential problems when characterizing these costs: 1/to define which molecular elements are likely subjects of complex mutations, 2/to recognize which of the molecular features altered by these mutations are driving the costs, and 3/to identify whether some specific target elements (of the molecular agent) can act as a distinctive reporter of such modified features and, in this way, of the costs. I will consider mutations in the RNA polymerase (RNAP) in Escherichia coli as a model of complex mutations to examine all these three problems. I will finish by discussing the connotations of this study for the understanding of robustness in biological systems.

Understanding the fitness costs of mutations is essential for our comprehension of the architecture of biological systems and its evolution. Many studies of this type, however, have traditionally focused on "simple" mutations. By simple, I refer here to mutations in molecular elements with a specific function, e.g., an enzyme catalyzing a particular biochemical reaction or a transcription factor linked to the activation of a given gene. In this talk, I will focus instead on the study of the fitness costs of “complex” pleiotropic mutations.

One encounters three potential problems when characterizing these costs: 1/to define which molecular elements are likely subjects of complex mutations, 2/to recognize which of the molecular features altered by these mutations are driving the costs, and 3/to identify whether some specific target elements (of the molecular agent) can act as a distinctive reporter of such modified features and, in this way, of the costs. I will consider mutations in the RNA polymerase (RNAP) in Escherichia coli as a model of complex mutations to examine all these three problems. I will finish by discussing the connotations of this study for the understanding of robustness in biological systems.

Natural organisms tend to form complex communities with high biodiversity, such as human gut microbes, tropical trees and ocean fishes. A major challenge in ecology research is explaining the coexistence of multispecies in complex communities. Seminal work by Robert May predicts a stability boundary in ecosystems, suggesting that only a limited number of interacting species can stably coexist. While this prediction has highly influenced our understanding of ecological communities, why the high biodiversity of natural communities largely escapes this boundary is still under intense debate. Here, we address the stability and dynamics of a large number of experimental microbial communities within a wide range of initial richness. Bridging theory and microcosms experiment, we show that, beyond May’s bound, multispecies communities universally exhibit persistent abundance fluctuation. When the community size or interaction strength is high, species exhibit abundance fluctuations emerging from complex species interactions. These emergent fluctuations are inherent and collective behaviors of complex communities, which maintain high biodiversity by allowing episodic local blooms of rare species.

Antibiotics are our primary tool to fight bacterial pathogens, although their use also unintentionally perturbs surrounding microbial communities. Antibiotic dosing is typically determined by the drug susceptibility of the target pathogen, but our understanding of how microbes respond to antibiotics in the multispecies arena remains limited. Here, we study microbial community dynamics following antibiotic perturbations in an experimental community containing Corynebacterium ammoniagenes (Ca) and Lactobacillus plantarum (Lp) that displays two alternative stable states as a result of mutual inhibition. Although Ca was more susceptible to chloramphenicol in monocultures, we found that chloramphenicol exposure nonetheless led to a transition from the Lp-dominated to the Ca-dominated community state. A simple theoretical model predicted, and experiments confirmed, that the slower growth rate of Lp made the Lp-dominated community more vulnerable to a wide range of antibiotic perturbations. The Ca-dominated community, on the other hand, suffered community shifts in experimental conditions in which intraspecies cooperation significantly compromised the growth rate of Ca at low cell density. Our results highlight that species susceptibility to antibiotics is often surprisingly uninformative of community resilience, as growth dynamics in the wake of antibiotic exposure can play a dominant role.

Understanding the principles of colonization resistance of the gut microbiome to the pathogen Clostridioides difficile will enable design of effective next-generation defined bacterial therapeutics. We investigate the ecological and molecular principles of community resistance to C. difficile invasion using a diverse synthetic human gut microbiome. Our results show that the synthetic community strongly inhibits the growth of C. difficile and that species richness is a key determinant of C. difficile growth across a wide range of ecological contexts. Using a dynamic computational model, we infer the inter-species interaction network and show that C. difficile receives the largest number and magnitude of incoming negative interactions. We study the factors influencing C. difficile invasion in a set of small multi-species communities and identify mechanisms of inhibition including acidification of the environment and competition over glucose. We also identify that C. difficile's close relative Clostridium hiranonis strongly inhibits C. difficile via a non-pH mediated mechanism. Our results show that while increasing the initial density of C. difficile can increase its abundance in the assembled community, the community context determines the maximum achievable C. difficile abundance. Our work suggests that the C. difficile inhibitory potential of defined bacterial therapeutics can be optimized by designing communities that feature a combination of mechanisms of species richness, environment acidification, and resource competition.

Dynamical systems often generate distinct outputs according to different initial conditions, and one could infer the corresponding input configuration given an output. This property captures the essence of information encoding and decoding. Here, we use Pseudomonas aeruginosa swarming colony as a model system, and demonstrate the use of self-organized patterns, combined with machine learning, to achieve distributed information encoding and decoding. Our approach exploits a critical property of many natural pattern-formation systems: in repeated realizations, each initial configuration generates similar but not identical output patterns due to randomness in the patterning process. However, for sufficiently small randomness, groups of patterns that each corresponds to a unique initial configuration can be distinguished from one another. Modulating the pattern generation and machine learning model training can tune the tradeoff between encoding capacity and security. Our method is applicable for a wide variety of self-organized pattern-formation systems.

The spread of an enteric pathogen in the human gut depends on many interacting factors, including pathogen exposure, diet, host gut environment, and host microbiota, but how these factors jointly influence infection outcomes remains poorly characterized. Here, we develop a model of host-mediated resource-competition between mutualistic and pathogenic taxa in the gut that aims to explain why similar hosts, exposed to the same pathogen, can have such different infection outcomes. Our model successfully reproduces several empirically observed phenomena related to transitions between healthy and infected states, including (1) the nonlinear relationship between pathogen inoculum size and infection persistence, (2) the elevated risk of chronic infection during or after treatment with broad-spectrum antibiotics, (3) the resolution of gut dysbiosis with fecal microbiota transplants, and (4) the potential protection from infection conferred by probiotics. We then use the model to explore how host-mediated interventions, namely shifts in the supply rates of electron donors (e.g., dietary fiber) and respiratory electron acceptors (e.g., oxygen), can potentially be used to direct gut community assembly. Our study demonstrates how resource competition and ecological feedbacks between the host and the gut microbiota can be critical determinants of human health outcomes. We identify several testable model predictions ready for experimental validation.

Branching pattern formation is common in many microbes. Extensive studies have focused on addressing how such patterns emerge from local cell-cell and cell-environment interactions. However, little is known about whether and to what extent these patterns play a physiological role. Here, we consider the colonization of bacteria as an optimization problem to find the colony patterns that maximize colony growth efficiency under different environmental conditions. We demonstrate that Pseudomonas aeruginosa colonies develop branching patterns with characteristics comparable to the prediction of modeling; for example, colonies form thin branches in a nutrient-poor environment. Hence, the formation of branching patterns represents an optimal strategy for the growth of Pseudomonas aeruginosa colonies. The quantitative relationship between colony patterns and growth conditions enables us to develop a coarse-grained model to predict diverse colony patterns under more complex conditions, which we validated experimentally. Our results offer new insights into branching pattern formation as a problem-solving social behavior in microbes and enable fast and accurate predictions of complex spatial patterns in branching colonies.

The still-evolving 2,4-dinitrotoluene (DNT) pathway of soil bacterium Burkholderia cepacia R34 has been studied as a case of emergence of new metabolic capabilities in environmental microorganisms. The dnt route originated from a precursor naphthalene degradation pathway and the first enzyme (DNT dioxygenase) maintains significant activity towards its earlier substrate. Both in vivo reactions and the associated regulatory system mediated by the DntR transcriptional factor indicate that reactive oxygen species (ROS) generated by the faulty (i.e. uncoupled) reaction of the precursor enzymes with DNT elicit genetic diversification. This could in turn ease the solution of the biochemical and physiological problem. These observations provide a view of evolution as a sort of heterotic computing in which the problem is embodied in the physicochemical frame of the cell and the exploration of the solution space is pushed by its endogenous dynamics. On this basis, it is plausible that some members of a given microbial community are prone to innovate their metabolic capacities much faster than others while the rest may benefit from such innovation through horizontal gene transfer.

Most microbes grow diauxically, utilizing the available resources one at a time rather than simultaneously. The properties of communities of microbes growing diauxically remain poorly understood, largely due to a lack of theory and models of such communities. Here, we develop and study a minimal model of diauxic microbial communities assembling in a serially diluted culture. We find that community assembly repeatably and spontaneously leads to communities with complementary resource preferences i.e., communities where species prefer different resources as their top choice. Simulations and theory explain that the emergence of complementarity is driven by the disproportionate contribution of the top choice resource to the growth of a species. Additionally, we develop a geometric approach for analyzing serially diluted communities, with or without diauxie, which intuitively explains several additional emergent community properties, such as the apparent lack of species which grow fastest on a resource other than their most preferred resource. Overall, our work provides testable predictions for the assembly of natural as well as synthetic communities of diauxically shifting microbes.

Horizontal gene transfer (HGT) allows organisms to share genetic material with non-offspring, and is typically considered beneficial for evolving populations. Recent unexplained observations suggest that HGT rates in nature are linked with environmental dynamics, being high in static environments but surprisingly low in fluctuating environments. Here, using a geometric model of adaptation, we show that this trend might arise from evolutionary constraints. During adaptation in our model, a population of phenotype vectors aligns with a potentially fluctuating environmental vector while experiencing mutation, selection, drift and HGT. Simulations and theory reveal that HGT shapes a trade-off between the adaptation speed of populations and their fitness. This trade-off gives rise to an optimal HGT rate which decreases sharply with the rate of environmental fluctuations. Our results are consistent with data from natural populations, and strikingly suggest that HGT may sometimes carry a significant disadvantage for populations.

The extent to which natural communities can be seen as 'superorganisms' whose collective properties can be directly selected is unclear. Methods of high-throughput screening and selection of microbial communities have recently raised the possibility of exerting artificial selection so as to obtain desirable community-level properties. In this talk, I will address the question of how a target collective phenotype comes to be approached in the course of evolution by successive changes in the traits of the composing units. I will use a simple two-species model to illustrate the effects of selection for community composition on community ecology. In this model, each community in a metapopulation undergoes a growth phase, at the end of which it is assessed for its properties. A new generation of collectives is generated by dilution of the best performers, giving rise to a genealogy of communities. Contrary to species interactions, collective generation time scale and bottleneck size are parameters controllable by the experimenter. A deterministic approximation allows to elucidate the role of these parameters in facilitating the emergence, through the adjustment of species interactions, of an evolutionarily stable composition. Such highly heritable composition is realized - analogous to developmental processes - by 'ecological canalization', which buffers against initial fluctuations in newborn community composition. Finally, I will briefly mention some ongoing work on the effect of selection for collective properties of complex communities.

A major question in microbial ecology is how the coexistence of species is affected by the presence of multiple resources. Microbes in multi-resource environments often display long diauxic lags, which occur when a consumed resource runs out and a microbe has a period of little to no growth while it readjusts its metabolism. We demonstrate experimentally that differences in diauxic lag times allow slow growing species to stably coexist with fast growing species. Our primary example involves an Acinetobacter species (Aci2) and Pseudomonas aurantiaca (Pa) growing on alanine and glutamate. When cocultured on only alanine or only glutamate Aci2 competitively excludes Pa, leading to the expectation that Aci2 would also exclude Pa on a mix of alanine and glutamate. Surprisingly, when cocultured on alanine and glutamate the species coexist at a stable fraction. Monoculture experiments reveal Aci2 grows faster than Pa but Pa has shorter diauxic lags. Through theory and additional experiments, we establish the tradeoff between Aci2’s fast growth and Pa’s short lags as the source of coexistence between these species. Across a large set of two-species two-resource competitions, we show that coexistence occurs more than three times as frequently when the slow-grower is the fast-switcher. We also identify examples in which roles are reversed, specifically in which Aci2 is the fast-switcher and in which Pa is the fast-grower. Our work provides theory and experimental demonstration for diauxic lags as a potentially common source of coexistence in multi-resource environments.

Interactions between genes are a major part of evolution, but they are fundamentally difficult to study due to problems of scale. In the context of bacterial evolution, the widespread presence of horizontal gene transfer amplifies the importance of gene-gene interactions. Detecting gene-gene interactions without performing large numbers of assays requires the development of computational techniques that can handle the necessary volume of genomic data. Here we present such a method, focusing on recent evolutionary events. We apply our method to a database of over 40,000 genomes of S. aureus. We demonstrate that our method can suggest genetic interactions for future study and that our method illuminates the current discourse on bacterial genomics.