Competitive interactions in Escherichia coli populations: the role of bacteriocins

Abstract

Explaining the coexistence of competing species is a major challenge in community ecology. In bacterial systems, competition is often driven by the production of bacteriocins, which are narrow-spectrum proteinaceous toxins that serve to kill closely related species, providing the producer better access to limited resources. Bacteriocin producers have been shown to competitively exclude sensitive, nonproducing strains. However, the dynamics between bacteriocin producers, each lethal to its competitor, are largely unknown. In this study, we used in vitro, in vivo and in silico models to study competitive interactions between bacteriocin producers. Two Escherichia coli strains were generated, each carrying a DNA-degrading bacteriocin (colicins E2 and E7). Using reporter-gene assays, we showed that each DNase bacteriocin is not only lethal to its opponent but, at lower doses, can also induce the expression of its opponent's toxin. In a well-mixed habitat, the E2 producer outcompeted its adversary; however, in structured environments (on plates or in mice colons), the two producers coexisted in a spatially 'frozen' pattern. Coexistence occurred when the producers were initiated with a clumped spatial distribution. This suggests that a 'clump' of each producer can block invasion of the other producer. Agent-based simulation of bacteriocin-mediated competition further showed that mutual exclusion in a structured environment is a relatively robust result. These models imply that colicin-mediated colicin induction enables producers to successfully compete and defend their niche against invaders. This suggests that localized interactions between producers of DNA-degrading toxins can lead to stable coexistence of heterogeneously distributed strains within the bacterial community and to the maintenance of diversity.

Figure 1

Mutual colicin induction. The proteins of isogenic strains carrying colicin E2 or E7 plasmids and a colicin-free control strain were crudely extracted and used to induce reporter strains carrying ce2a and ce7a promoters fused to the Photorhabdus luminescence luxCDABE reporter operon (Table 1). Colicin E2 crude protein extract was used to induce the pDEW-E7 reporter vector (A; filled circle), whereas colicin E7 extract was used to induce the pDEW-E2 reporter vector (B; filled triangle); the colicin-free strain was tested with the pDEW-E7 (C; open circle) and pDEW-E2 (D; open triangle) reporter vectors.

Figure 2

Community dynamics in an unstructured environment. Flask environment was initiated by introducing E. coli strain ColE2 (closed circles) and ColE7 (open circles) simultaneously into a flask and monitoring their concentrations over time. The dashed line indicates that the abundance of the ColE7 strain has decreased below its detection limit. Data points are the mean of two independent experiments, each performed in duplicate. Bars represent the standard deviation of the average cell concentration.

Figure 3

Community dynamics in a structured environment. A static plate environment was initiated by randomly depositing 24 droplets from pure culture of strains ColE2 and ColE7. The changing spatial pattern of the community is photographed over time (a) showing the spread of the strains droplets (day 3) to lawns bordered by a clearing zone (day 5) that was later colonized by strains resistant to both colicins (day 7). On analysis of the cells' concentration (b), the abundance of E. coli harboring colicin E2 (filled circles) and E7 (open circles) encoding plasmids was shown to remain invariable throughout the experiment. Data points are the mean of two independent experiments, each performed in duplicate. Bars represent the standard deviation of the average cell concentration.

Figure 4

Effect of competition on bacterial population size in mice. Bacterial density (CFU per g fecal matter) monitored over time in mice in control (a) or experimental (b) cages. The control cages hosted mice harboring either E. coli strain BZB1011 bearing pDEW-E2 or mice harboring E. coli strain BZB1011 bearing pDEW-E7. The experimental cages contained one mouse established with the E. coli strain BZB1011 bearing pDEW-E2 and one mouse with E. coli strain BZB1011 bearing pDEW-E7. Each point represents the mean CFU per g feces averaged for strains bearing pDEW-E2 (filled circles) or pDEW-E7 (open circle) recovered from the mice. Bars represent the standard error for each point.

Figure 5

An agent-based simulation of two bacteriocin-producing strains. The interactions between strains P₁ (blue line) and P₂ (red line) with β₁=7.0,β₂=9.0, δ₁=δ₂=1.0, μ₁=μ₂=0.5, π₁=π₂=0.5 and τ₁₂=τ₂₁=0.5 (see Table 2B for a description of these parameters) were simulated. (a) The ‘base case' without bacteriocin-mediated induction (that is, γ₁₂=γ₂₁=0; see Table 2B) showing that the better grower, P₂, invades from low density and displaces its competitor P₁. Strain P₂ was introduced from low density after 50 epochs (marked by the arrows), such that P₁ can first reach its equilibrium. Five replicates are shown. (b) If the better grower P₂ starts at high density, it prevents invasion by P₁ (five replicates are shown, introduction occurs after 50 epochs in each case). (c) The case of bacteriocin-mediated induction (γ₁₂=γ₂₁=5.0) showing that the better grower P₂ is now excluded by a resident population of P₁ (five replicates are shown). (d) When commonly present at higher density, P₂ still excludes P₁ (across five replicates). Thus, bacteriocin-mediated induction can produce a case of mutual exclusion where it would otherwise not be expected. The color reproduction of this figure is available on the html full text version of the paper.

Figure 6

Exploration of parameter space. Here, we use the same parameters as in Figure 5, except that we vary the growth rate of strain 2 (β₂, ranging from 7.5 to 10) and the rate of cross-induction (γ₁₂=γ₂₁, ranging from 0 to 10). For each parameter combination, we run 20 replicates in which strain 2 is introduced at an initial frequency of about 5% after 50 epochs. We record the number of replicates in which strain 2 invades. The ‘black floor' corresponds to runs in which strain 2 experiences uniform extinction. When strain 2 is common, it is able to exclude an invading strain 1 across the entire region of the parameter space shown. Thus, the black floor of this plot corresponds to regions of parameter space in which mutual exclusion occurs. We see that more equitable growth rates and higher rates of cross-induction promote mutual exclusion.