Killing by Type VI secretion drives genetic phase separation and correlates with increased cooperation

Abstract

By nature of their small size, dense growth and frequent need for extracellular metabolism, microbes face persistent public goods dilemmas. Genetic assortment is the only general solution stabilizing cooperation, but all known mechanisms structuring microbial populations depend on the availability of free space, an often unrealistic constraint. Here we describe a class of self-organization that operates within densely packed bacterial populations. Through mathematical modelling and experiments with Vibrio cholerae, we show how killing adjacent competitors via the Type VI secretion system (T6SS) precipitates phase separation via the 'Model A' universality class of order-disorder transition mediated by killing. We mathematically demonstrate that T6SS-mediated killing should favour the evolution of public goods cooperation, and empirically support this prediction using a phylogenetic comparative analysis. This work illustrates the twin role played by the T6SS, dealing death to local competitors while simultaneously creating conditions potentially favouring the evolution of cooperation with kin.

Figure 1. T6SS-mediated killing drives phase separation in dense bacterial populations.

We modelled the dynamics of phase separation in fully occupied, randomly seeded square lattices (a). Phase separation between red and blue bacteria capable of mutual killing occurred in an individual-based model (scale bar, 50 cells) (b), in a partial differential equation model (c), and in an Ising spin model (scale bar, 50 magnets) (d). No phase separation occurred between red (C6706) and blue (692–79) T6SS^- mutants of Vibrio cholerae (ΔvasK; e), in contrast to T6SS⁺ strains (f). We varied the efficacy of T6SS while still allowing for growth by culturing V. cholerae at a range of temperatures: 17 °C (h), 25 °C (i), and 30 °C (j). T6SS⁻ controls cultured at 25 °C did not phase separate (g). Scale bars denote 100 μm in e,f, and 1 mm in g–j. Images shown in g–j are representative of four replicate competitions.

Figure 2. Structural analysis of models and experiments.

The static structure factor S(q), plotted versus wavenumber q multiplied by cell size L for the individual based model (IBM; a) and for experiments (b). In the latter, the red and black lines depict two separate fields of view of V. cholerae strains C6706 and 692–79, started at an initial ratio of 1:6, while blue indicates a 1:8 inoculation ratio. The brown line depicts T6SS⁻ mutants, and purple indicates mutual killers grown at 17 °C for 24 h (all others grown at 25 °C). (The brown line is obscured by the purple line, which is nearly identical.) Mutual killing drives phase separation, increasing S(q) at smaller values of q. The relationship between S(q_m) and q_m is summarized in c with open orange cirlcles=experimental data (25 °C and a 1:6 inoculation ratio, as in b), black closed squares=IBM, red closed circles=PDE model (d=0.01), and blue closed triangles=Ising model (T=1); all three models and the experiments follow a universal q_m⁻² trend. S(q) curves collapse when S(q)q_m²L² is plotted versus q/q_m (d), indicating that all models and experiments are undergoing the same coarsening process. Colour denotes model timestep, as in a, while symbols indicate type of model or experiment, as in c. We also examine the creation of spatial structure by calculating a biological metric, assortment (r), through time across 6,000 updates of the IBM (e) and after 24 h in experiments (f). Mutual killers were grown at 30 °C (red), 25 °C (blue) and 17 °C (green). Defective killers were grown at 30 °C (purple), 25 °C (teal) and 17 °C (orange). Plotted is the mean assortment of four replicate populations (mutual killers) and three replicate populations (defective killers)±95% confidence intervals.

Figure 3. Phase separation favours the evolution of cooperation.

The dynamics of competition between cooperators and cheats are shown through time for different starting frequencies. In the absence of T6SS-mediated killing, cooperation is not favoured in either a well-mixed environment (a) or a spatially defined environment (b). In a non-spatial environment with killing via T6SS, cooperators can be protected from cheats when common owing to their advantage in antagonistic interactions, but cannot invade from rarity (c). In contrast, the high assortment created by phase separation allows cooperators to invade from rarity and spread to fixation (d). In a–d, line colour denotes initial cooperator frequency. The spatial organization of cooperators (blue) and cheats (red) during competition is shown in e. Panels correspond to the time-points marked by circles in d.

Figure 4. T6SS is associated with investment in other secreted products.

The phylogenetic distribution of T6SS, T6SS effectors and secretome size across 439 genomes from the Proteobacteria and Bacteroidetes (a). Secretome size of a strain (expressed as a percentage of genome size) increases with both its number of T6SSs (b) and T6SS effectors (c). Lines are the fits of univariate Bayesian phylogenetic mixed models (BPMMs) (Supplementary Tables 3 and 4). Posterior distributions of the effects of the numbers of T6SS (d) and T6SS effectors (e) on secretome size from the multivariate BPMM (Supplementary Table 2). Ninety-five per cent credible intervals of the estimates are shaded. Plot of observed against predicted secretome size from the multivariate BPMM (f), including effects of the number of T6SS, number of T6SS effectors and phylogeny. The line represents a 1:1 mapping.