Inflammation and bacteriophages affect DNA inversion states and functionality of the gut microbiota

Abstract

Reversible genomic DNA inversions control the expression of numerous gut bacterial molecules, but how this impacts disease remains uncertain. By analyzing metagenomic samples from inflammatory bowel disease (IBD) cohorts, we identified multiple invertible regions where a particular orientation correlated with disease. These include the promoter of polysaccharide A (PSA) of Bacteroides fragilis, which induces regulatory T cells (Tregs) and ameliorates experimental colitis. The PSA promoter was mostly oriented "OFF" in IBD patients, which correlated with increased B. fragilis-associated bacteriophages. Similarly, in mice colonized with a healthy human microbiota and B. fragilis, induction of colitis caused a decline of PSA in the "ON" orientation that reversed as inflammation resolved. Monocolonization of mice with B. fragilis revealed that bacteriophage infection increased the frequency of PSA in the "OFF" orientation, causing reduced PSA expression and decreased Treg cells. Altogether, we reveal dynamic bacterial phase variations driven by bacteriophages and host inflammation, signifying bacterial functional plasticity during disease.

Keywords: Bacteroides; Crohn’s disease; DNA inversions; bacteriophages; functional plasticity; gut microbiome; immunomodulation; inflammatory bowel diseases; phase variation; ulcerative colitis.

Graphical abstract

Figure 1

Bacteroides species exhibit differential orientations of invertible regions during health and disease (A) Selected significantly differentially oriented invertible DNA regions (inverted repeats (IR) segments, see also Table S1) (Wilcoxon rank-sum test, adjusted p value < 0.05) in at least one comparison between healthy, CD (Crohn’s disease), and UC (ulcerative colitis). Red indicates the forward orientation, and blue represents the reverse orientation in comparison with the reference genome. (B) Prevalence of functional genes in proximity to invertible DNA regions significantly different between healthy individuals and IBD patients. (C) Differentially oriented invertible DNA regions in B. fragilis NCTC 9343. PSA, polysaccharide A; PSH, polysaccharide H. (D) Differentially oriented invertible DNA regions in B. thetaiotaomicron VPI-5482. CPS, capsular polysaccharide. (E) Differentially oriented invertible DNA regions in Phocaeicola dorei MGYG-HGUT-02478. In (C), (D), and (E), data represent the median (line in box), IQR (box), and minimum/maximum (whiskers). (Wilcoxon rank-sum test, ^∗p < 0.05; ^∗∗p < 0.01).

Figure 2

Relative orientation of the PSA promoter of B. fragilis is affected by inflammation (A) Illustration of murine model of inflammation. “Humanized” mice harboring human microbiota spiked with B. fragilis NCTC 9343, and mice monocolonized with B. fragilis NCTC 9343 were exposed to 3% DSS (day = 0 in the illustration) at 4 weeks of age. (B) Calprotectin levels (ng/mL) measured in different days of the experiment. Lines represent the standard deviations. Green: control group; red: DSS-treated mice. (C) The body weight change of mice measured on different days of the experiment. Lines represent the standard deviations. Green: control group; red: DSS-treated mice. (D) Ratio of B. fragilis PSA’s promoter “ON” orientation measured by qPCR on different days of the experiment (n = 8–12 in each time point). Data represent the mean (cross), median (line in box), IQR (box), and minimum/maximum (whiskers). (Wilcoxon rank-sum test, ^∗p < 0.05; ^∗∗p < 0.01.) Green: control group; red: DSS-treated mice. (E) Ratio of B. fragilis PSA’s promoter “ON” orientation measured by the PhaseFinder tool, in different days of the experiment. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (Wilcoxon rank-sum test, ^∗p < 0.05; ^∗∗p < 0.01.) Green: control group; red: DSS-treated mice. (F) Ratio of B. thetaiotaomicron CPS3’s promoter “ON” orientation measured by the PhaseFinder tool on different days of the experiment. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (Wilcoxon rank-sum test, ^∗p < 0.05; ^∗∗p < 0.01.) Green: control group; Red: DSS-treated mice. (G) Alpha-diversity (observed species) between groups and timepoints. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (Wilcoxon rank-sum test, ^∗p < 0.05; ^∗∗p < 0.01, ^∗∗∗p < 0.001.) Green: control group; red: DSS-treated mice. (H) Alpha-diversity (Shannon index) between groups and timepoints. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (Wilcoxon rank-sum test, ^∗p < 0.05; ^∗∗p < 0.01, ^∗∗∗p < 0.001.) Green: control group; red: DSS-treated mice. (I) Beta diversity (Aitchison distance) between groups and time points. Principal coordinates analysis (PCoA) of Aitchison distances between bacterial communities of different groups and time points. Each point represents a single sample, colored according to group and timepoints: light green: control at day 0; green: control at day 6; light red: DSS-treated at day 0; red: DSS-treated at day 6. The mean (centroid) of samples in each group is indicated with a blank circle. Ellipses represent 0.95 confidence intervals of each group. (J) Beta-dispersion values (Aitchison distances from the centroid) between groups and timepoints. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (betadisper, p > 0.05.) (K) Differential bacterial abundance between DSS-treated mice on day 0 and day 6 detected by the Maaslin2 algorithm. Red dots indicate differentially abundant bacteria that were determined by adjusted p value < 0.05 and log₂ fold change >1 and <−1, respectively. Blue dots indicate B. fragilis and B. thetaiotaomicron. (L) Ratio of B. fragilis PSA’s promoter “ON” orientation measured by qPCR on different days in gnotobiotic mice monocolonized with B. fragilis NCTC 9343. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (Wilcoxon rank-sum test, ^∗p < 0.05; ^∗∗p < 0.01.) Green: control group; red: DSS-treated mice. See also Figure S1 and Table S2.

Figure 3

Evidence of phage correlation with B. fragilis polysaccharide A promoter orientation (A) Experimental design of culturing B. fragilis in patients' fecal filtrates. (B) Ratio of the “ON” orientation of the PSA promoter of B. fragilis, measured by qPCR, after ex vivo exposure to fecal filtrates of IBD patients before and after treatment with anti-TNF (4 patients on infliximab [HR] and 7 patients on Humira [HuR]). Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (one-sided Wilcoxon rank-sum test, ^∗p < 0.05.) Green: B. fragilis grown in a 1:1 ratio of M9 and PBS not exposed to fecal filtrates; orange: B. fragilis exposed to fecal filtrates of patients before anti-TNF treatment; yellow: B. fragilis exposed to fecal filtrates of patients after anti-TNF treatment. Each dot represents 3 individual experiments; lines connect experiments from the same patient; shapes are determined by the patients’ treatments: circle, HR; triangle, HuR. (C) Calprotectin levels (μg/g) measured in patients’ feces. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) (one-sided Wilcoxon rank-sum test, ^∗p < 0.05.) Dots represent samples; lines connect samples from the same patient; shapes are determined by the patients’ treatments: circle, HR; triangle, HuR. (D) Differential viral taxonomic units’ abundances, (from the IBDMDB cohort, count table from Nishiyama et al.³⁴), between samples with low “ON” orientation of the PSA promoter (<40%) and high “ON” orientation (>60%). Differentially abundant viral taxonomic units were detected by the DeSeq2 algorithm (Wald test, p < 0.01). Red dots indicate differentially abundant bacteria that were determined by p value < 0.01 and fold change >1.5 and <−1.5, respectively. (E) Phage-to-host abundances ratios of viral OTUs predicted to infect B. fragilis and detected in (D). Data represent the median (line in box), IQR (box), minimum/maximum (whiskers), dots represent individual samples. Dashed line represents the mean phage-to-host ratio of all phages predicted to infect B. fragilis in all samples (mean = 0.009). (F) Differential viral taxonomic units' abundances, (from the IBDMDB cohort, count table from Nishiyama et al.³⁴), between samples with low “ON” orientation of the CPS3 promoter (<40%) and high “ON” orientation (>60%). Differentially abundant viral taxonomic units were detected by the DeSeq2 algorithm (Wald test, p < 0.01). Red dots indicate differentially abundant bacteria that were determined by p value < 0.01 and fold change >1.5 and <−1.5, respectively. (G) Phage-to-host abundances ratios of viral OTUs predicted to infect B. thetaiotaomicron and detected in (F). Data represent the median (line in box), IQR (box), minimum/maximum (whiskers), dots represent individual samples. Dashed line represents the mean phage-to-host ratio of all phages predicted to infect B. thetaiotaomicron in all samples (mean = 0.03). See also Figure S2 and Table S3.

Figure 4

Phage exposure alters the expression of the PSA locus and surface PSA (A) Phylogenetic tree based on the whole genome of viral OTUs identified as bacteriophages against B. fragilis as well as Bacteroides bacteriophages Barc2635, B40-8, B124-14, crAss002, and Enterobacteria phage lambda. MAFFT was used to perform multiple sequence alignment, and the average-linkage method was used to construct the phylogenetic tree. Colors denote the association between the viral OTUs abundances in Nishiyama et al., IBDMDB cohort, with active Crohn’s disease (light red) or with both active Ulcerative colitis and active Crohn’s disease (red). See also Figure S3. (B) Experimental design of in vivo experiments. (C) Ratio of B. fragilis PSA’s promoter “ON” orientation measured on day 10 by qPCR in fecal samples of gnotobiotic mice monocolonized with B. fragilis with and without the Barc2635 bacteriophage. Horizontal lines represent the means. Blue: control group; yellow: Barc2635-treated mice. Each dot represents a mouse. (Mann-Whitney test, ^∗∗∗p < 0.001.) (D) Expression levels of upaY in ceca of gnotobiotic mice monocolonized with B. fragilis with and without the Barc2635 bacteriophage at day 10. Levels are shown as 2ˆ(-ΔCT) with rpsL as a reference gene. Each dot represents a mouse, and error bars represent the standard deviations. (Mann-Whitney test, ^∗∗∗∗p < 0.0001.) (E) PSA presence on the surface of B. fragilis exposed to the Barc2635 bacteriophage, detected by anti-PSA antibodies on day 10. Bacteria were analyzed by flow cytometry for the expression of PSA, using Rabbit anti-PSA antibodies. Horizontal lines represent the means. Each dot represents a mouse.| (F) Experimental design of in vivo competition experiments. (G) Fold change of Δpsa\ΔmpiM44 at day 10 and day 0 in the Barc2635 bacteriophage treated group (pink) compared with Δpsa\ΔmpiM44 at day 10 and day 0 in the control group (black). Error bars represent the standard deviations. Each dot represents a mouse (Mann-Whitney test, p > 0.05.) See also Figure S4.

Figure 5

Bacteriophage-driven phase variation in B. fragilis results in reduction of Tregs (A) Experimental design of in vivo experiments. (B) CD4⁺ to CD8⁺ ratio out of CD45⁺TCRβ⁺ live cells in colon. Single cells were isolated from colon lamina propria. Immune cells were analyzed by flow cytometry. Horizontal lines represent the means. Each dot represents a mouse. Green, GF mice; blue, mice monocolonized with B.fragilis; yellow, mice monocolonized with B.fragilis and Barc2635. (C) RORγt⁺ helios^- percentages out of FOXP3⁺ cells in colon. Single cells were isolated from colon lamina propria. Immune cells were analyzed by flow cytometry. Horizontal lines represent the means. Each dot represents a mouse. Green, GF mice; blue, mice monocolonized with B.fragilis; yellow, mice monocolonized with B.fragilis and Barc2635. ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001, one-way analysis of variance (ANOVA). See also Figure S5.