Genetic determinants of in vivo fitness and diet responsiveness in multiple human gut Bacteroides

Abstract

Libraries of tens of thousands of transposon mutants generated from each of four human gut Bacteroides strains, two representing the same species, were introduced simultaneously into gnotobiotic mice together with 11 other wild-type strains to generate a 15-member artificial human gut microbiota. Mice received one of two distinct diets monotonously, or both in different ordered sequences. Quantifying the abundance of mutants in different diet contexts allowed gene-level characterization of fitness determinants, niche, stability, and resilience and yielded a prebiotic (arabinoxylan) that allowed targeted manipulation of the community. The approach described is generalizable and should be useful for defining mechanisms critical for sustaining and/or approaches for deliberately reconfiguring the highly adaptive and durable relationship between the human gut microbiota and host in ways that promote wellness.

Fig. 1. The effect of diet and diet oscillations on the configuration of the 15 member artificial community containing 11 wild-type strains and the four mutant libraries, including the structure of the population of B. cellulosilyticus WH2 mutants

(A) Experimental design. 10 to 12-week-old germ-free male C57BL/6J mice were gavaged with the indicated consortium of 15 strains. Animals were given the LF/HPP and/or the HF/HS diet in the order shown. Fecal samples were collected at the indicated time points for INSeq and microbial RNA-Seq analyses. (B) Principal coordinates analysis (PCoA) of Hellinger distance measurements based on the relative abundance of B. cellulosilyticus WH2 Tn mutants in fecal samples, as defined by multi-taxon INSeq analysis. Each colored circle represents the population of all mutants in B. cellulosilyticus WH2 sampled from an individual mouse in the indicated diet treatment group at a given time point. Each empty circle represents the average location of the mutant population for a given group at 4, 10 and 16 days post gavage. The lines connect time points in a given group. The color of the lines indicates the diet. Dashed lines indicate the diet oscillation groups. (C,D) PCoA of Hellinger distance measurements based on COPRO-Seq data. Each colored circle represents a fecal community sampled from an individually caged mouse belonging to the indicated diet treatment group. These circles were ordinated in the same coordinate space but are being shown as two separate panels for clarity. Vertical dashed lines in panel D indicate the time point when a diet switch occurred.

Fig. 2. HF/HS diet-specific fitness determinants in B. thetaiotaomicron 7330

(A) Circos (26)-generated alignment of the B. thetaiotaomicron VPI-5482 and B. thetaiotaomicron 7330 genomes. Grey lines connect segments of DNA conserved between these two genomes (these regions were identified by alignment with Mauve; 27). The color-coded outer circle denotes the similarity between their proteins: green, >90% identity (based on Blastp alignment); blue, 70%-90% identity; red, <70% identity; black, intergenic regions. GC skew was calculated using a sliding window size of 10kb (yellow, GCskew>0; purple, GCskew<0). (B) COPRO-Seq analysis of the relative abundance of the two B. thetaiotaomicron strains in the fecal microbiota of mice sampled two weeks after gavage while consuming the HF/HS or LF/HPP diets. Mean values ± SEM are shown (n=5 individually caged mice harboring a community consisting of 11 wild-type and the four mutant libraries/treatment group). Note that the summed relative abundance of the two strains remains the same even though the relative representation of the individual strains is significantly different in the two diet contexts (p<0.001, 2-way ANOVA). (C) HF/HS diet-specific fitness determinants in B. thetaiotaomicron 7330 involved in degradation of glycosaminoglycans associated with the intestinal mucosa (genes highlighted with red lines together with their EC annotations). (D) HF/HS diet-associated changes in the z-scores of 7330-strain-specific fitness determinants that are involved in transcription regulation (Btheta7330_04415) or are components of operons encoding transport systems (vertical black lines to the left of the gene ID denote individual operons).

Fig. 3. Arabinoxylan increases the relative abundance of B. cellulosilyticus WH2 in vivo

(A) INSeq analysis reveals that all genes in PUL BcellWH2_04321-4327 have significant fitness indices (z-values) in the HF/HS diet context. BACWH2_04327, encoding a hybrid two-component system (HTCS) regulator, is the only gene in this PUL that has a significant fitness effect on the LF/HPP diet. Functional annotations for genes in the PUL are shown together with the direction of their transcription. Fitness indices for each gene in the different diet contexts (orange, HF/HS; green LF/HPP) are plotted as mean values ± SEM. The horizontal dashed line indicates the cutoff for significance (p<0.05; z-test with FDR correction). Abbreviations: dist_CE, distant relatives of carbohydrate esterase; GH, Glycoside Hydrolase; HTCS, Hybrid Two-Component System. (B) Experimental design. Adult C57BL/6J germ-free mice were gavaged with a consortium containing 11 wild-type strains plus the four Bacteroides INSeq libraries. Animals were fed the HF/HS or LF/HPP diets with or without supplementation of their drinking water with 7.5% arabinoxylan (n=5 individually-caged mice/group). (C) The relative abundance of B. cellulosilyticus and B. ovatus was defined by COPRO-Seq analysis of fecal samples collected at the indicated time points. Mean values ± SEM are plotted. ***, p <0.001 compared to the reference group A at 14 days post-gavage (dpg); +++, p <0.001 for within group comparisons of the 30 dpg versus 14 dpg fecal sample (Student’s t-test after FDR correction). B. ovatus, the only other Bacteroides strain in the community that exhibited significantly increased growth in minimal medium supplemented with arabinoxylan (Fig. S13B) did not manifest a significant change in its relative abundance in vivo when arabinoxylan was added to the drinking water (ns, not significant).