Community composition and the environment modulate the population dynamics of type VI secretion in human gut bacteria

Abstract

Understanding the relationship between the composition of the human gut microbiota and the ecological forces shaping it is of great importance; however, knowledge of the biogeographical and ecological relationships between physically interacting taxa is limited. Interbacterial antagonism may play an important role in gut community dynamics, yet the conditions under which antagonistic behaviour is favoured or disfavoured by selection in the gut are not well understood. Here, using genomics, we show that a species-specific type VI secretion system (T6SS) repeatedly acquires inactivating mutations in Bacteroides fragilis in the human gut. This result implies a fitness cost to the T6SS, but we could not identify laboratory conditions under which such a cost manifests. Strikingly, experiments in mice illustrate that the T6SS can be favoured or disfavoured in the gut depending on the strains and species in the surrounding community and their susceptibility to T6SS antagonism. We use ecological modelling to explore the conditions that could underlie these results and find that community spatial structure modulates interaction patterns among bacteria, thereby modulating the costs and benefits of T6SS activity. Our findings point towards new integrative models for interrogating the evolutionary dynamics of type VI secretion and other modes of antagonistic interaction in microbiomes.

Extended Data Fig. 1.. Abrogation of T6SS activity does not provide a fitness benefit to B. fragilis under laboratory conditions growing on mono and di-saccharide as carbon source.

Growth curves and solid growth in mono and co-culture for the Producer and Resistant strains subjected to growth on different sole carbon sources (mono and di-saccharide) in defined minimal broth or solid agar media, including D-fructose (a), D-galactose (b), D-xylose (c), and sucrose (d). For solid agar growth co-culture, producer and resistant strains were mixed at a 1:1 ratio by OD₆₀₀, inoculated in high-density spots before being harvested for enumeration of colony forming units by plating on BHIS supplemented with either erythromycin or tetracycline. Mann-Whitney tests at each timepoint indicated no significant difference between strains. For liquid experiments, the results are from three biological replicates that each represent the mean of eight technical replicates. For solid growth, results are from six biological replicates from two independent experiments. Data shown represent mean values ±SD.

Extended Data Fig. 2.. Abrogation of T6SS activity does not provide a fitness benefit to B. fragilis under laboratory conditions growing on polysaccharides as carbon source

The growth of the Producer and Resistant strains in defined minimal media with different carbon sources (polysaccharide) is shown in liquid and solid (mono and co-culture) growth in the different panels. For solid growth, B. fragilis Producer strain with Erm^R and Resistant strain with Tet^R were grown on defined minimal media with different carbon sources separately (monoculture) and together in a 1:1 ratio (co-culture). Each strain was quantified via calculating colony-forming units on BHIS agar plates supplemented with erythromycin or tetracycline. Carbon sources used in defined minimal media included potato starch (a), glycogen (b), and mucin (c). Mann-Whitney test showed no difference between the different strains at each timepoint. For liquid experiments, the results are from three biological replicates that each represent the mean of eight technical replicates. For solid growth, results are from six biological replicates from two independent experiments. Data shown represent mean values ±SD.

Fig. 1.. The GA3 T6SS is associated with altered composition of the human gut microbiota.

(a) Summary of frequency of metagenomic detection of the GA3 T6SS in diverse infant (TEDDY N=56, DIABIMMUNE N=187) and adult (HMP N=108, MetaHIT N=73, Yachida et al N=219) human gut microbiome samples containing B. fragilis. (b-c) Comparison of abundance of B. fragilis-specific GA3 structural genes with B. fragilis marker genes in microbiome samples from (b) 6-to-18 month-old infants (DIABIMMUNE, N=187) and (c) adults (HMP, N=108). Scatter plots depict the relationship between the log10 abundance of B. fragilis and the log10 abundance of GA3 for samples in which at least 100 reads map to B. fragilis marker genes. GA3 abundance in each sample was adjusted with a pseudo count of 1e-7 and thus samples without detectable GA3 are present at −7. The percentage of B. fragilis-containing samples lacking GA3 is indicated at lower right in each plot. (d-e) PcoA plots depicting B. fragilis-positive shotgun metagenomic samples from representative cohorts, (d) 6-to-18 month old infants from the DIABIMMUNE cohort and (e) the adult HMP cohort. Samples are colored by the presence (red) or absence (blue) of the GA3 T6SS. Distances between samples were calculated using unweighted unifrac with genus-level relative abundance values and a phylogenetic tree of bacterial genera. A single loading vector is shown for Bacteroides and the separation between GA3+ and GA3- was assessed in each plot using PERMANOVA tests (DIABIMMUNE p > 0.05, HMP p < 0.01). (f) A phylogenetic tree of Bacteroidaceae species that highlights significant PhILR nodes in red identified from analysis of HMP microbiomes, with Bonferroni correction for multiple tests.

Fig. 2.. Recurrent evolutionary loss of the GA3 T6SS from B. fragilis.

(a) Schematic comparing intact and degenerated GA3 loci (red genes) from syntenic regions of select B. fragilis genomes. Homology is indicated by shaded regions linking genes between genomes of different strains. (b) Maximum likelihood phylogenetic tree of B. fragilis strains based on single nucleotide polymorphisms in aligned B. fragilis MetaPhlAn3 marker genes. GA3 presence and absence are indicated by closed and open circles (respectively) at branch tips. Inferred unique loss (red) and gain (green) events during B. fragilis diversification are indicated by stars, based on ancestral state reconstruction.

Fig. 3.. Mouse colonization reveals a fitness cost of the B. fragilis GA3 T6SS.

(a) Schematic representing the B. fragilis NCTC 9343 GA3 T6SS operon, comparing the “Producer” T6SS-active strain (magenta) to the “Resistant” T6SS-inactive strain (blue). The in-frame 3-gene chromosomal deletion mutation (ΔtssBC-clpV) in the Resistant strain is indicated by lack of shading. (b-c) Liquid and solid growth assays for monoculture and co-culture in in defined minimal media for two different sole carbon sources (b) D-glucose and (c) Synergy inulin. For liquid assays, data represent the mean ±SD of three biological replicates and eight technical replicates. For solid co-culture assays, Producer and Resistant strains were mixed at 1:1 ratio by OD₆₀₀ at high density and the growth was measured by calculating the CFU of each strain on BHIS agar plates containing antibiotics from six biological replicates on two independent experiments. (d) Producer and Resistant strains were inoculated via oral gavage of a 1:1 mixture into antibiotic-treated SPF mice. Strain relative abundance (normalized to 100%) was tracked via barcode-specific quantitative PCR performed on DNA isolated from fecal pellets at indicated days post gavage. (e) Quantification of the percent relative abundance of Producer strain (n=8 mice). Box plots indicate the median and interquartile range. One-way ANOVA test was used to compare abundance between sampling at day 1 (one day after gavage) and all subsequent sampling (***P<0.001, ****P<0.00001). (f) Percent relative abundance of Producer strain in each of 8 mice in comparison to the total quantity of gDNA detected for both Producer and Resistant strain. Mouse from which bacteria were isolated for whole genome sequencing is indicated by a star.

Fig. 4.. Co-colonization with a Sensitive strain alters Producer-Resistant strain dynamics in mice.

(a) Schematic representing the B. fragilis NCTC 9343 GA3 T6SS operon from the Sensitive strain. The in-frame deletion mutations (ΔE-IΔtssBC-clpV) are indicated by lack of shading. Producer, Resistant, and Sensitive strains were inoculated in a 1:1:1 ratio and the gDNA quantity of each strain was tracked for 88 days via qPCR on genomic DNA isolated from fecal pellets. Percentages of Producer (b), Resistant (c), and Sensitive (d) strains are shown as the median of eight mice for each sampling from days 1 to 88. Box plots indicate the median and interquartile ranges and bars indicate the minimum and maximum values. Significance between mean detected abundances for each strain comparing between day 1 after gavage and all subsequent samplings was assessed by Kruskal-Wallis test (*P<0.05 **P<0.01 ***P<0.001 ****P<0.0001). The percentage of detected Producer (e), Resistant (f), and Sensitive (g) gDNA is shown for each mouse individually over time. Results are representative of two independent colonization experiments, each with N = 8.

Fig. 5.. Reaction-diffusion model reveals local and spatial dynamics of T6SS Producer, Resistant and Sensitive genotypes.

(a-c) Ternary plots showing vector fields and sample trajectories for different values of the density parameter $g$ (low density: $g=0.1$, moderate density: $g=0.2$, and high density: $g=0.5$). Each point in the triangle represents a different community composition, with the distance to each edge representing the proportion of each genotype in the population. Open and closed circles at the vertices denote the unstable and stable steady states, respectively. (d-f) Progression of community composition for the trajectories shown in (a-c). At moderate and high densities, there is alternation of the dominant genotype. (g-j) Simulation of the system including spatial structure, where cells can move throughout a 2D surface. For moderate densities, in (h-i), the dominant genotype at the end of the simulation depends heavily on the initial distribution of the genotypes across the surface.

Fig. 6.. Individual-based biofilm simulations suggest dispersal-recolonization regimes can dictate selective loss of T6SS.

(a) An example of the biofilm simulation space containing cell groups of T6SS Producer (red), Sensitive (yellow/orange), and Resistant cells (blue). Below the full simulation space, partial frames are shown for the recolonization regime in which cells dispersed from the biofilm surface upstream can re-attach to the biofilm surface downstream, promoting admixture between different strains/species. We also implemented a sloughing regime in which the full biofilm system was periodically cleared and the surface instantaneously recolonized with a sub-sample of the population at the time point just prior to disturbance. (b) In the recolonization regime, simulations tended toward mixed regimes containing mostly T6SS Resistant cells, which most closely resembled the mouse experimental outcomes. We note that the pure-Resistant population state is predicted to be invadable in the long term by T6SS Sensitive cells if the Producer strain has been eliminated. (c) The sloughing regime, in which biofilm clearance was stronger, tended to generate isolated clonal clusters of cells of each strain along the basal surface. This regime tended more strongly to favor T6SS Sensitive cells, because they remained out of contact with T6SS Producers and had the highest growth rate of the three strain types.