Adaptive Evolution within Gut Microbiomes of Healthy People

Abstract

Natural selection shapes bacterial evolution in all environments. However, the extent to which commensal bacteria diversify and adapt within the human gut remains unclear. Here, we combine culture-based population genomics and metagenomics to investigate the within-microbiome evolution of Bacteroides fragilis. We find that intra-individual B. fragilis populations contain substantial de novo nucleotide and mobile element diversity, preserving years of within-person history. This history reveals multiple signatures of within-person adaptation, including parallel evolution in sixteen genes. Many of these genes are implicated in cell-envelope biosynthesis and polysaccharide utilization. Tracking evolutionary trajectories using near-daily metagenomic sampling, we find evidence for years-long coexistence in one subject despite adaptive dynamics. We used public metagenomes to investigate one adaptive mutation common in our cohort and found that it emerges frequently in Western, but not Chinese, microbiomes. Collectively, these results demonstrate that B. fragilis adapts within individual microbiomes, pointing to factors that promote long-term gut colonization.

Keywords: Bacteroides; adaptive evolution; de novo mutation; horizontal gene transfer; human microbiome; metagenomics; microbial evolution; microbial genomics; whole-genome sequencing.

Figure 1 |. Each subject’s B. fragilis population is dominated by a single lineage.

(A) Phylogenetic reconstruction shows that isolates cluster by subject, with one outlier isolate from Subject 10. Isolates are colored according to subject. (B) Isolates from different subjects generally differ by < 100 single nucleotide differences (SNPs) while isolates from different subjects differ by >10,000 SNPs. Mutational distances between all pairs of isolates. Inset: Intra-subject pairs separated by >18,000 SNPs all involve the outlier isolate from Subject 10.

Figure 2 |. B. fragilis lineages diversify for years in healthy individuals via de novo SNPs and MEDs.

(A) The phylogeny of isolates from L05 is shown as an example, demonstrating both SNP and mobile element differences (MEDs; see also Figures S1–S3). Thin lines connect each isolate to a colored circle, which indicates the timepoint of isolation. Relative coverage (compared to the mean genomewide) across two MEDs is also shown. (B) The number of SNPs and MEDs identified for each lineage. (C-D) Estimate of the B. fragilis molecular clock using two different methods. (C) Each shape represents the average number of new SNPs per isolate from the indicated timepoint not present in the set of SNPs at initial sampling. (D) Estimate of molecular clock using root-to-tip distances for L01 only. (E) Distance and inferred time to most recent common ancestor at initial sampling (dMRCA and tMRCA, respectively). Gray dots represent individual isolates and bars represent averages. For L08, purple dots represent hypermutator isolates, and the average presented excludes these. (F) The spectrum of mutations in the hypermutator sublineage (purple) differs substantially from that of the normal sublineages of L08 (yellow) and 11 other lineages (gray; error bars represent standard deviation). Inset: Phylogeny for L08.

Figure 3 |. Mobile elements are transferred within the microbiome of individual people.

(A-B) The phylogeny of isolates from L04 illustrates the gain of MED04–1 over time. Shading reflects the average relative coverage of the MED (compared to the mean genomewide). (B) Average relative coverage across the length of MED04–1 for different samples. Colors are as indicated in (A). (C-D) Combining isolate whole genomes and metagenomes reveals an inter-species mobile element transfer event. (C) Metagenomic libraries from both time points of L01 show high relative coverage of a putative integrative conjugative element (ICE), while only isolates from the later timepoint have coverage of this ICE. Isolates from one sample show slightly higher relative coverage as these genomic were prepared differently (Methods). (D) A rooted parsimonious phylogeny of the putative ICE across 4 species. Isolates with identical ICE sequences from a same phylogenetic group were merged into a single node (see also Figures S1B–S1C).

Figure 4 |. Genes involved in polysaccharide utilization and cell envelope biosynthesis undergo parallel adaptive evolution within individual subjects.

(A) An example gene under parallel evolution from L02 is shown, demonstrating that observed mutations are of independent origin and occur in distinct isolates. Nodes represent individual isolates and are colored by sampling dates. (B) A total of 16 genes were identified as undergoing parallel evolution in the 12 lineages. These 16 genes are grouped by inferred function (Table 5). Each dot in the table represents an independent mutation event, colored by type of mutation. (C) The number of genes mutated in parallel within at least one lineage deviates significantly from neutral simulations (P<0.001, Methods). (D) A classic signature of selection, dN/dS, indicates adaptive evolution in genes under parallel evolution (P<0.001, Binomial test), but not for other genes mutated within subjects. Mutations across lineages show a significant signature of purifying selection (P<0.001, Binomial test). Error bars represent 95% confidence intervals. (E) Mutations in SusC homologs under selection were enriched at the interface between the proteins (P< 0.001, Methods).

Figure 5 |. Evolutionary dynamics over a 1.5 year sampling period reveals a steady increase in mutational frequencies and a stable coexistence of two sublineages.

(A-C) We combined 206 stool metagenomes and 187 isolate whole genomes to infer evolutionary dynamics within L01. (A) Branches with at least 4 isolates are labeled with colored squares that represent individual SNPs. One SNP was inferred to have happened twice and is indicated in both locations (purple). (B) Frequencies of labeled SNPs were inferred from metagenomes. Circles represent SNP frequencies inferred from isolate genomes. (C) We combined these data types to infer the trajectory of sublineages prior to and during sampling. Sublineages are labeled with names and colored as in (A). The two major sublineages, SL1 and SL2, are separated by dashed lines. Black diamonds represent transient SNPs from genes presented in Figure 4. (D) The identity of SNPs shown in (B-C). SNPs in the 16 genes under positive selection are bolded and transient mutations in these genes are indicated with parentheses. Negative numbers indicate mutations upstream of the start of the gene. (E) All isolates from SL2, but only 13% from SL1 carry putative prophage MED01–2. (F-H) Relative abundances of pairs of isolates during competition assays, over two rounds of passages. Dashed lines represent 1:100 dilution at hour 18. Each line represents the average of 3 technical replicates, and error bars represent standard error of the mean.

Figure 6 |. Comparison to published metagenomes reveals a mutation that emerges independently and frequently in Western, but not Chinese populations

(A) We examined the prevalence of a common amino acid change in available metagenomes. The percentage of metagenome samples with a polymorphism or fixed proline at this position was greater in Western populations than in Chinese populations (n=152, 136 respectively). Error bars represent standard error. (B) A neighbor-joining phylogeny of inferred B. fragilis genotypes within public metagenomes demonstrates that this mutation emerged independently and repeatedly. Phylogeny is shown as a dendrogram to better visualize the independent emergence of Q100P mutations. (C) Between lineages, genes under parallel evolution show significant signatures of purifying selection, as indicated by dN/dS (for 13 genes with inter-lineage mutations, Methods). This analysis represents tens of thousands of years of evolution (Methods), in contrast to Figure 4C. Error bars represent 95% confidence interval. The dashed line represents the average dN/dS for all inter-lineage SNPs. (D) Four models that could account for the discrepancy of natural selection at different timescales.