Intestinal Bacteroides modulates inflammation, systemic cytokines, and microbial ecology via propionate in a mouse model of cystic fibrosis

Abstract

Persons with cystic fibrosis (CF), starting in early life, show intestinal microbiome dysbiosis characterized in part by a decreased relative abundance of the genus Bacteroides. Bacteroides is a major producer of the intestinal short chain fatty acid propionate. We demonstrate here that cystic fibrosis transmembrane conductance regulator-defective (CFTR-/-) Caco-2 intestinal epithelial cells are responsive to the anti-inflammatory effects of propionate. Furthermore, Bacteroides isolates inhibit the IL-1β-induced inflammatory response of CFTR-/- Caco-2 intestinal epithelial cells and do so in a propionate-dependent manner. The introduction of Bacteroides-supplemented stool from infants with cystic fibrosis into the gut of Cftr^F508del mice results in higher propionate in the stool as well as the reduction in several systemic pro-inflammatory cytokines. Bacteroides supplementation also reduced the fecal relative abundance of Escherichia coli, indicating a potential interaction between these two microbes, consistent with previous clinical studies. For a Bacteroides propionate mutant in the mouse model, pro-inflammatory cytokine KC is higher in the airway and serum compared with the wild-type (WT) strain, with no significant difference in the absolute abundance of these two strains. Taken together, our data indicate the potential multiple roles of Bacteroides-derived propionate in the modulation of systemic and airway inflammation and mediating the intestinal ecology of infants and children with CF. The roles of Bacteroides and the propionate it produces may help explain the observed gut-lung axis in CF and could guide the development of probiotics to mitigate systemic and airway inflammation for persons with CF.IMPORTANCEThe composition of the gut microbiome in persons with CF is correlated with lung health outcomes, a phenomenon referred to as the gut-lung axis. Here, we demonstrate that the intestinal microbe Bacteroides decreases inflammation through the production of the short-chain fatty acid propionate. Supplementing the levels of Bacteroides in an animal model of CF is associated with reduced systemic inflammation and reduction in the relative abundance of the opportunistically pathogenic group Escherichia/Shigella in the gut. Taken together, these data demonstrate a key role for Bacteroides and microbially produced propionate in modulating inflammation, gut microbial ecology, and the gut-lung axis in cystic fibrosis. These data support the role of Bacteroides as a potential probiotic in CF.

Keywords: SCFA; cystic fibrosis; gut; inflammation; probiotic; propionate.

Fig 1

Bacteroides supernatants downregulate IL-8 production by Caco-2 cells, and IL-8 downregulation is likely not due to a protein. CFTR−/− Caco-2 cells were cultured for 2 weeks in 24-well plates and then exposed to Bacteroides supernatants prepared in sMEM as described in the Materials and Methods. Each point represents the average of 4–6 technical replicates from a single biological replicate. IL-8 was quantified by ELISA after (A) 6 hours and (B) 24 hours of CFTR−/− Caco-2 cell exposure to the bacterial supernatants from the indicated Bacteroides strains. Individual data points are shown, and the median is represented by the horizontal lines, and the dotted line is the median of the sMEM + IL-1β control. Significance was determined by unpaired one-way ANOVA followed by Dunnett’s post-test with sMEM + IL-1β as the reference. *P < 0.05, **P < 0.01, and ****P < 0.0001. (C, D) IL-8 was quantified by ELISA after 24 hours of CFTR−/− Caco-2 cell exposure to the bacterial supernatants from the indicated treatment. IL-8 quantity is displayed as a percentage of the sMEM + IL-1β control in the same biological replicate. Effects of (C) proteinase K treatment and (D) size fractionation on IL-8-reducing activity. Significance was tested by unpaired one-way ANOVA followed by Tukey’s post-test. **P < 0.01. Individual data points and the mean are shown.

Fig 2

CFTR−/− Caco-2 cells are responsive to the anti-inflammatory effects of SCFAs. CFTR−/− Caco-2 cells were cultured for 2 weeks in 24-well plates and then treated with IL-1β alone or with the addition of the indicated concentrations of SCFAs. IL-8 was quantified by ELISA after 24 hours of culture with (A) sodium acetate, (B) sodium propionate, or (C) sodium butyrate at the indicated concentration. Each point indicates the mean of 4–6 technical replicates from a single biological replicate, with 3–4 independent biological replicates performed. Lines connect results from experiments performed on the same day. Significance was tested by unpaired one-way ANOVA followed by Dunnett’s post-test with MEM + IL-1β as the reference. *P < 0.05, **P < 0.01, and ****P < 0.0001.

Fig 3

IL-8 downregulation by B. thetaiotaomicron is propionate dependent. CFTR−/− Caco-2 cells were cultured for 2 weeks in 24-well plates and then incubated with B. thetaiotaomicron supernatants from the indicated strains prepared in sMEM as described in the Materials and Methods. IL-8 was quantified by ELISA after (A) 6 hours and (B) 24 hours of the CFTR−/− Caco-2 cell exposure to the bacterial supernatants from the indicated strains. Each point represents the average of 4–6 technical replicates from a single biological replicate, with three independent biological replicates performed. Significance was determined by unpaired one-way ANOVA followed by Dunnett’s post-test with sMEM + IL-1β as the reference. *P < 0.05, ***P < 0.001 and ****P < 0.0001. The supernatants in these assays were not normalized to cell density, but there is no difference in the growth of the strains in this medium (see Fig. S5A). The SCFAs propionate (panel C) and acetate (panel D) were quantified by gas chromatography - mass spectrometry (GC-MS) (ng/μL) and log2-transformed. The dashed red line indicates the assay limit of detection (LOD). In panel D, the dashed black line indicates the average acetate concentration detected in the medium, as this medium contains acetate above the LOD. Significance was determined by unpaired one-way ANOVA followed by Dunnett’s post-test with B. thetaiotaomicron Δprp as the reference. ns = not significant, ****P < 0.0001. In all panels, individual data points are shown and the median is represented by the horizontal line.

Fig 4

Strain-dependent production of SCFA and IL-8 downregulation. SCFAs were quantified in undiluted, sterile-filtered supernatants by GC-MS as described in the Materials and Methods. (A) Heatmap of log2-transformed concentration (ng/mL) of each detectable SCFA. Pentanoic and 2-methyl pentanoates were not detected above the LOD. (B) Propionate produced by individual isolates. The dashed red line indicates the LOD.

Fig 5

A CftrF508del mouse model reveals that intestinal supplementation of Bacteroides reduces systemic cytokine levels. (A-C) Stool samples were collected from CftrF508del mice either before the onset of antibiotic treatment (baseline) or post-gavage with pooled CF stool ± supplemented Bacteroides. Each point represents data from a single mouse stool in all graphs. (A) Microbial composition of stool was profiled for each mouse by 16S rRNA gene amplicon sequencing to determine Bacteroides relative abundance. “Early” indicates that the stool was collected 2–4 days post-gavage and “Late” indicates collection 12–13 days post-gavage. (B) Propionate was quantified by GC-MS from mouse stool samples during the early or late windows (days 1 & 9, respectively). (A, B) Statistical significance was analyzed by linear model on log2-transformed propionate concentrations, accounting for ±supplemented Bacteroides, experiment, cage, and sex of mouse *P < 0.05, **P < 0.01, and ***P < 0.001. (C) Fecal pellets, then serum, and intestine were collected at the end of the experiment 24 hours after LPS challenge. A complete timeline of the experiment is shown in Figure S10A. Cytokines are displayed as a ratio of the median cytokine concentration in the (−) supplemented Bacteroides condition divided by the median cytokine concentration in the (+) supplemented Bacteroides condition. Cytokines were quantified by Luminex 32-plex panel for the indicated cytokines. Statistical significance was analyzed by linear model on log2-transformed cytokine concentrations, accounting for ±supplemented Bacteroides, experiment, cage, and sex of mouse with P < 0.05 labeled in red, and indicate higher levels for the (−) supplemented Bacteroides condition. (D-G) Serum cytokines quantified by Luminex that were statistically significant between the (+) and (−) supplemented Bacteroides conditions by linear model on log2-transformed cytokine concentrations, accounting for ±supplemented Bacteroides, experiment, cage, and sex of mouse. *P < 0.05 and **P < 0.01. When the high concentration outlier > 10,000 pg/mL in the (−) supplemented Bacteroides G-CSF measurement is removed, this result remains significant with P = 0.0243. After we account for multiple comparisons, by Bonferroni correction, only IL-6 is significant. (D) G-CSF (granulocyte colony stimulating factor), (E) KC (CXCL1; IL-8 homolog), (F) IL-6 (Interleukin 6), and (G) MIP-1β (macrophage inflammatory protein 1β). Horizontal bars in each graph display the median.

Fig 6

Cftr^F508del mouse model reveals that intestinal supplementation of Bacteroides propionate mutant no longer reduces systemic cytokine levels. (A) Heatmap of normalized cytokine levels quantified by Luminex 32-panel in serum (left), lungs (middle), and colon (right). Mice were gavaged with CF stool + supplemented BacteroidesΔtdk or CF stool + supplemented BacteroidesΔprp. Samples were collected at sacrifice 24 hours after LPS challenge. Cytokines associated with a pro-inflammatory response are labeled in green, anti-inflammatory response are labeled in light-blue, and cytokines associated with either response are labeled in pink. Higher z-score levels of respective cytokine are indicated in red, and those lower are indicated in blue. Gray indicates that the cytokine was below detection. (B) Serum (left), lung (middle), and intestinal (right) KC cytokine levels quantified by Luminex 32-panel. Statistical significance was determined by linear model on log2-transformed KC concentrations, accounting for strain and sex of mouse. *P < 0.05, **P < 0.01, and ***P < 0.001. (C) Stool was collected throughout the experiment from mice gavaged with Bacteroides Δtdk (purple) and Bacteroides Δprp (yellow). Stool was serially diluted and plated to count CFU/mL on blood agar + 100 μg/mL gentamicin. Each point indicates stool collected from a single mouse. Upon sacrifice (Sac), intestines were dissected to isolate 3 regions: the cecum, proximal colon, and distal colon. The luminal contents of these three regions were plated for CFUs and make up the CFUs under Sac**, which is a sum of the CFU from the three regions of the intestine assessed (see Fig. S17A). Statistical significance was determined by Kruskal-Wallis with Dunn’s multiple pair-wise comparisons test. (D) Absolute abundance was determined by qPCR using DNA extracted from stool with B. thetaiotaomicron VPI-5482-specific probes to quantify establishment of strains Bacteroides Δtdk (purple) or Bacteroides Δprp (yellow). Statistical significance was determined by linear model for B. thetaiotaomicron DNA (ng/μL), accounting for strain, cage, and sex of mouse. ns, not significant. (E) Absolute abundance was measured by 16S rRNA gene amplicon sequencing from DNA extracted from stool spiked with ZymoBIOMICSSpike-In Control I. Statistical significance was determined by linear model for Bacteroides absolute abundance accounting for strain, cage, and sex of mouse. ns, not significant.