Metabolic model-based analysis of the emergence of bacterial cross-feeding via extensive gene loss

Abstract

Background: Metabolic dependencies between microbial species have a significant impact on the assembly and activity of microbial communities. However, the evolutionary origins of such dependencies and the impact of metabolic and genomic architecture on their emergence are not clear.

Results: To address these questions, we developed a novel framework, coupling a reductive evolution model with a multi-species genome-scale metabolic model to simulate the evolution of two-species microbial communities. Simulating thousands of independent evolutionary trajectories, we surprisingly found that under certain environmental and evolutionary settings metabolic dependencies emerged frequently even though our model does not include explicit selection for cooperation. Evolved dependencies involved cross-feeding of a diverse set of metabolites, reflecting constraints imposed by metabolic network architecture. We additionally found metabolic 'missed opportunities', wherein species failed to capitalize on metabolites made available by their partners. Examining the genes deleted in each evolutionary trajectory and the deletion timing further revealed both genome-wide properties and specific metabolic mechanisms associated with species interaction.

Conclusion: Our findings provide insight into the evolution of cooperative interaction among microbial species and a unique view into the way such relationships emerge.

Fig. 1

A framework for modeling the evolution of species interaction. a To model reductive evolution, genes are iteratively chosen at random as candidates for deletion, the fitness effect of their deletion is evaluated (using a co-culture growth model; panel (b)), and if the fitness effect is relatively small, these genes are deleted. b The co-culture growth model simulates the growth of the two species in a shared environment, and is based on a previously introduced dynamic multi-species model [28]. This model iteratively infers the behavior of each species in the shared environment based on an FBA approach (panel (c)). The predicted growth of each species and the predicted rates at which it uptakes and excretes various metabolites are used to update the abundances of species in the co-culture and the concentration of metabolites in the shared environment over time. c An FBA model is used to predict the growth of each species in a given environment based on the set of metabolic reactions and constraints encoded by the species and the concentration of metabolites in its environment

Fig. 2

Metabolic interactions and their emergence over time. a Evolutionary simulations could result in one of four unique outcomes, determined by the ability of evolved species to grow in mono-culture and co-culture. Plotted are examples of each of these four outcomes, illustrating the fitness of each of the two species in mono-culture and in co-culture over evolutionary time. b The changes in interaction type over time for all 16,317 simulation runs. Each horizontal bar represents a single simulation run, and the color corresponds to the interaction type using the same colors as the titles in panel (a). c Example of an evolved mutualistic community. In the ancestral species tyrosine is produced through the shikimate pathway and dTTP is produced from UDP. In this example evolved mutualistic community, deletions in both species have led to obligate cross-feeding of tyrosine and thymidine. The relevant gene deletions and their impact on metabolic fluxes in each species are highlighted

Fig. 3

Frequencies of metabolites’ availability, cross-feeding, and dependence. The frequencies of metabolites’ availability, cross-feeding, and dependence are shown for species of each interaction type and for each metabolite. Each set of nested circles shows the frequency at which the given metabolite is produced by their partner species and hence available for uptake (blue), the frequency at which this metabolite is utilized by the species through cross-feeding (yellow), and the frequency at which this metabolites is depended on (red). The area of the circle scales with the frequency, but for visualization purposes the portions of the circle extending beyond the rectangular box are not shown. Only metabolites that are depended upon at least 10 times or consumed at least 30 times are shown

Fig. 4

Genome size and gene retention frequency in evolved genomes. a Distributions of the genome size of evolved species from each interaction type. (*: P < 10^− 3; **: P < 10^− 9; ***: P < 10^− 30). b The distribution of retention rates of different genes in the model. Each gene is plotted as a vertical bar with height equal to the fraction of species in non-collapsed simulations that retained it, and the genes are sorted in ascending order of this overall retention rate. Each bar is colored by the fraction of species retaining that gene that are in each of the different interaction types

Fig. 5

Genes and pathways associated with species interaction and cross-feeding phenotypes. a The fraction of genes whose deletion is associated with each interaction type that are assigned to each pathway. The overall size of the pie indicates the number of genes associated with that interaction type. Pathways that are significantly enriched among this set of genes are highlight (bold font and asterisk). Only pathways that contain > 4% of the genes are listed (the rest are pooled into the ‘Other’ category) b Associations between providing or depending on specific metabolites and deletion or retention of specific genes. Orange (purple) bars indicate genes that are lost (retained) significantly more often in species that provide or depend on a specific metabolite compared to independent species. Genes are sorted by membership to pathways (see (c)), and only genes present in at least one pathway are shown. c Pathway membership of genes. Black bars indicate that a gene belongs to that pathway. Pathways are sorted by the number of genes assigned to them, and only pathways with five or more assigned genes are shown