Two modes of evolution shape bacterial strain diversity in the mammalian gut for thousands of generations

Abstract

How and at what pace bacteria evolve when colonizing healthy hosts remains unclear. Here, by monitoring evolution for more than six thousand generations in the mouse gut, we show that the successful colonization of an invader Escherichia coli depends on the diversity of the existing microbiota and the presence of a closely related strain. Following colonization, two modes of evolution were observed: one in which diversifying selection leads to long-term coexistence of ecotypes and a second in which directional selection propels selective sweeps. These modes can be quantitatively distinguished by the statistics of mutation trajectories. In our experiments, diversifying selection was marked by the emergence of metabolic mutations, and directional selection by acquisition of prophages, which bring their own benefits and costs. In both modes, we observed parallel evolution, with mutation accumulation rates comparable to those typically observed in vitro on similar time scales. Our results show how rapid ecotype formation and phage domestication can be in the mammalian gut.

Fig. 1. Colonization success depends on microbiota diversity and rates of molecular evolution across mice.

a Time series of the abundances of the invader (circles) and resident (squares) E. coli lineages. Each circle represents the mean value and error bars represent 2*Standard Error (2SE) (n = 3 technical replicates, Supplementary Data 1). Time series of the microbiota species diversity (Shannon diversity). The values of microbiota diversity observed are shown as a gray area. b Rate of molecular evolution, per generation, in each host (sweeps, intergenic mutations, Insertion Sequences). dN and dS, and dN/dS ratio. c E. coli mutation accumulation during in vivo and in vitro evolution (data of non-mutator lines from). “Mutations M(t)” correspond to the sum of allele frequencies at each sampling point for the in vitro and in vivo experiments (Supplementary Data 6 and 8 to 14). Blue and red hues represent mice where the invader colonized the gut alone or together with the resident E. coli lineage, respectively. d Level of mutational parallelism in the invader lineage across hosts: the genetic targets of adaptation and the frequency of hosts where it was changed is shown, as well as the changes that likely involve loss of function. Mutation targets also found in Lenski’s in vitro evolution experiment, data from, are highlighted in gray.

Fig. 2. Evolution of invader E. coli in the absence of the resident.

a–d Frequency of the mutations identified in the invader E. coli population across time in mouse D2, B2, E2 and I2 (top panels) and corresponding Muller Plots showing the spread of mutations observed in each mouse (middle panels). Microbiota composition (phylum level) along time (bottom panels), showing a high temporal stability in the long term. *1 fimA, frlR, ykfI, dgoR, dcuA/aspA, qseC, cydA; *2 ycbC, rpoC, frlR. Mutations that reached frequency above 95% are highlighted in color, other frequency trajectories shown in gray (Supplementary Data 5, 9, 10, 11 and 14).

Fig. 3. Evolution of invader E. coli when co-existing with the resident strain.

a–c Frequency of the mutations identified in the invader E. coli population across time in mouse A2, H2 and G2 (top panels), the corresponding Muller Plots (middle panels) and microbiota compositions at the phylum level (bottom panels). The spread of de novo mutations is indicated by the gene targets where they emerged, intertwined with HGT events, mediated by two phages (in light and dark orange) and two plasmids (green circle – small plasmid: ~69Kb; blue circle – big plasmid: ~109Kb) acquired from the resident strain which started to be detected on day 90 in mouse A2 but much earlier in the other mice. The frequency of phage- or plasmid-driven HGT events can be found in Supplementary Data 8, 12, 13, and 19. In mouse H2 invader evolution could be followed along 2475 generations, after which extinction occurred and in mouse G2 evolution could be followed for 1350 generations. Mutations that reached frequency above 95% are highlighted in color, other frequency trajectories are shown in gray (Supplementary Data 5, 8, 12 and 13). d Evidence for prophage domestication, shown by a reduction of lytic induction during gut adaptation. The ancestral lysogen and clones randomly sampled from mouse A2 at days 104 and 493, and mouse H2 at day 165, which bear Nef and KingRac prophages (lysogens), had their prophage induction rate (per hour) measured in a mitomycin C assay. P values were calculated by a two-sided t-test (n = 4 biologically independent replicates per clone tested, Supplementary Data 18). e, Molecular evolution of the resident strain in mouse A2. The loci highlighted in color were targets of evolution in both the resident and the invader E. coli strains. Mutations hitting genes with metabolic functions are indicated with solid lines (i.e. dgoR, glpR, srlR, psuK/fruA and lrp).

Fig. 4. Modes of evolution and fitness tradeoffs.

a–c A $p$-$\tau$ selection test captures the joint statistics of fixation probabilities a, and times b. This test c, identifies dominant diversifying selection in mice D2, B2, and E2 (brown), and directional selection generating clonal interference in mice A2, G2, and I2 (green) or periodic sweeps in mouse H2 (dark green). In mouse D2, where no fixations occur, we use the time to the end of the experiment lower bound for τ. The reference value $\tau \approx 2$ expected for directional selection in the low-mutation regime is shown as dashed line. Short-term trajectories (<2000 generations) are marked by small symbols. The data from Lenski’s in vitro evolution experiment, for the same lines as in Fig. 1, are shown as open circles. d Host specificity of in vivo adaptation. Relative fitness of pools of evolved clones from mice D2, B2, A2 and H2 when competing against the ancestor in new mice (n = 4, 2 male and 2 female per competitive fitness assay of each of the independently evolved populations). The height of each bar represents the value of competitive fitness of the evolved clones in each mouse, measured as the slope of the linear regression on Ln(CFUevolved/CFUancestor) over the first 4 days after gavage. The error bars represent the standard error of the linear regression slope. e In vivo adaptation leads to growth trade-offs in vitro. Growth rate (per hour) measured in LB medium (3–4 replicates) of randomly isolated invader E. coli clones evolved in the gut of mice D2, B2, E2, I2, A2, H2 and G2 during 436, 493, 167, 104, 493, 165 and 90 days, respectively. P values were calculated by a two-sided one-sample t-test (n = 3 to 4 biologically independent replicates per clone tested, Supplementary Data 24).