The intestinal clock drives the microbiome to maintain gastrointestinal homeostasis

Abstract

Diurnal (i.e., 24-hour) oscillations of the gut microbiome have been described in various species including mice and humans. However, the driving force behind these rhythms remains less clear. In this study, we differentiate between endogenous and exogenous time cues driving microbial rhythms. Our results demonstrate that fecal microbial oscillations are maintained in mice kept in the absence of light, supporting a role of the host's circadian system rather than representing a diurnal response to environmental changes. Intestinal epithelial cell-specific ablation of the core clock gene Bmal1 disrupts rhythmicity of microbiota. Targeted metabolomics functionally link intestinal clock-controlled bacteria to microbial-derived products, in particular branched-chain fatty acids and secondary bile acids. Microbiota transfer from intestinal clock-deficient mice into germ-free mice altered intestinal gene expression, enhanced lymphoid organ weights and suppressed immune cell recruitment. These results highlight the importance of functional intestinal clocks for microbiota composition and function, which is required to balance the host's gastrointestinal homeostasis.

Fig. 1. Fecal microbial rhythms persist in constant darkness.

A Representative actogram of control (Bmal1IECfl/fl) mice exposed to light-dark (LD) cycle and two weeks of constant darkness (DD). Fecal sampling times are indicated by red arrows. B Beta-diversity illustrated by MDS plots of fecal microbiota based on generalized UniFrac distances (GUniFrac) in LD and DD conditions. Significance was calculated with PERMANOVA with Bonferroni adjustment. C GUniFrac quantification over time relative to ZT1. D Diurnal (LD) and circadian (DD) profiles of alpha-diversity. E Dendrogram of microbiota profiles based on GUniFrac in LD and DD condition taxonomic composition at phylum and family level for each sample is shown as stacked bar plots around the dendrogram. The blue inner circle indicates the sampling time. F Diurnal (LD) and circadian (DD) profiles of the relative (left) and quantitative abundance (right) of phyla. G Heatmap depicting the relative abundance of 580 zOTUs (mean relative abundance>0.1%; prevalence > 10%). Data are normalized to the peak of each zOTU and ordered by the peak phase in LD conditions. Z-score represents relative abundance where yellow means high abundance and blue low abundance. Pie-charts at the right indicate the amount of rhythmic (blue) and arrhythmic (grey) zOTUs identified by JTK_Cycle (Bonferroni adj. p value ≤ 0.05). H Significance (Bonferroni adj. p-value JTK) and amplitude of rhythmic and arrhythmic zOTUs (left) and phase distribution (right) in LD and DD based on relative analysis. Significant rhythms are illustrated with fitted cosine-wave regression using a solid line (significance: p value ≤ 0.05. LD (light-blue) and DD (dark-blue). n = 6 mice/time point/light condition. Data are represented as mean ± SEM. Source data are provided as a source datafile.

Fig. 2. Food intake behavior marginally masks microbiota rhythmicity.

A Schematic illustration of the experiment design. B Beta-diversity illustrated by MDS plot based on generalized UniFrac distances (GUniFrac) distances of fecal microbiota stratified by individual time points in ad libitum (top, black) and starvation (bottom, purple). C Circadian profiles of alpha-diversity. D PCoA plots of fecal microbiota based on GUniFrac distances stratified by feeding condition. E, F Circadian profiles of the relative abundance of the major phyla (E) and family (F) of fecal microbiota. G Heatmap depicting the relative abundance of 511 zOTUs (mean relative abundance > 0.1%; prevalence > 10%). Data are normalized to the peak of each zOTU and ordered by the peak phase in ad libitum condition. Z-score represents relative abundance where yellow means high abundance and blue low abundance. Significance and amplitude (JTK_Cycle analysis) of rhythmic and arrhythmic zOTUs (top) identified by JTK (Bonferroni adj. p value ≤ 0.05) and phase distribution (based on cosine regress ion analysis) (bottom). Dashed line indicates adj. p value=0.05. H, I Circadian profile of relative (H) and quantitative (I) abundance of zOTUs of fecal microbiota. PERMANOVA test with Bonferroni adjustment to determine if the separation between group in B & D is statistically significant. Significant rhythms (cosine-wave regression, p value ≤ 0.05) are illustrated with fitted cosine-wave curves; data points connected by dotted lines indicate no significant cosine fit curves (p value > 0.05) and thus no rhythmicity. zOTUs were further analyzed with the adjusted compare rhythm script based on DODR. n = 83 ad libitum samples from 11 independent animals, 81 starvation samples from 11 independent animals for relative data in B-H (repeated measures). n = 39 Ad libitum samples from 5 independent animals, 37 starvation samples from 5 independent animals for quantitative data in I (repeated measures). Data are represented by mean ± SEM. Source data are provided as a source datafile.

Fig. 3. The intestinal circadian clock drives circadian microbiota composition.

A Beta-diversity illustrated by MDS plot of fecal microbiota based on generalized UniFrac distances (GUniFrac) stratified by genotype. Significance was calculated by PERMANOVA with Bonferroni correction. B Circadian profile of alpha diversity. C Circadian profile of relative (top) and quantitative (bottom) abundance of the major phyla. D, E Heatmap depicting the relative abundance (D) and quantitative abundance (E) of 580 zOTUs (mean relative abundance > 0.1%; prevalence > 10%). Data are normalized to the peak of each zOTU and ordered by the peak phase of control mice. Z-score represents relative abundance where yellow means high abundance and blue low abundance. Pie-charts at the right indicate the amount of rhythmic (colored) and arrhythmic (grey) zOTUs identified by JTK_Cycle (rhythmic = Bonferroni adj. p-value ≤ 0.05) based on relative (D) and quantitative (E) analysis. Significance and amplitude of rhythmic and arrhythmic zOTUs (top) and phase distribution (bottom) in controls and intestinal epithelial specific Bmal1-deficient (Bmal1^IEC-/-) mice is depicted on the right of the heatmaps. Dashed line indicates p-value = 0.05. Bar charts in (F) represent the intestinal controlled zOTUs (Adj. compare rhythm script based on DODR abundance comparison between control and Bmal1^IEC-/- (two-sided Wilcoxon, adj. p-value <0.05). Box and bar plots illustrate the alteration in quant. abundance (adj. p-value ≤ 0.05) and fold change of gut controlled zOTUs in the fecal samples with examples depicted in (G) Significant rhythms are illustrated with fitted cosine-regression solid lines; data points connected by dotted lines indicate no significant cosine fit curves (JTK Bonferroni adj. p-value > 0.05) and thus no rhythmicity. Bmal1^IECfl/fl controls (black) and Bmal1^IEC-/- (red). n = 6(control);5(Bmal1^IEC-/-) mice/time point (B,C,G). n = 48 (control, 6 biological independent mice, repeated measures); 40(Bmal1^IEC-/-, 5 biological independent mice, repeated measures) (A,D,E,F). Data are represented as mean ± SEM. Source data are provided as a source datafile.

Fig. 4. Microbiota profiling of Bmal1^IEC-/- mice in light-dark conditions and under starvation.

A Beta-diversity illustrated by MDS plot based on generalized UniFrac distances (GUniFrac) of fecal microbiota of Bmal1^IEC-/- (brown) and their Bmal1^IECfl/fl (grey) controls in LD conditions. Diurnal profiles of (B) 16 s copy number and (C) major phyla quantitative abundance. D Heatmap illustrating zOTUs quantitative abundance over time of fecal microbiota of Bmal1^IEC-/- and their controls in LD conditions. Data are ordered by the peak phase in LD conditions of the control group. E Beta-diversity principal coordinates analyses plot (PCoA) based on generalized UniFrac distances (GUniFrac) of fecal microbiota stratified by food availability. F Circadian profiles of phyla Firmicutes and Bacteroidetes. G Heatmap illustrating microbiota loosing/changing rhythmicity after starvation in Bmal1^IEC-/- mice according to the adjusted compare rhythm script based on DODR (adj. p value ≤ 0.05). Data are normalized to the peak of each zOTU and ordered by the peak phase in ad-libitum condition. H Circadian profiles of the quantitative abundance of example zOTUs masked by the food-intake behavior (adjusted compare rhythm script based on DODR, adj. p value ≤ 0.05). I Pie charts indicating the percentage of rhythmic/arrhythmic zOTUs in different condition according to JTK cycle analysis (rhythmic = JTK Bonferroni adj. p value ≤ 0.05). Significant rhythms (Cosine regression p-value ≤ 0.05) are illustrated with fitted cosine regression; data points connected by dotted lines indicate no significant cosine fit curves (p-value > 0.05) and thus no rhythmicity (C,F,H). zOTUs were further analyzed with the adjusted compare rhythm script based on DODR. Significance was calculated by PERMANOVA with Bonferroni adjustment (A,E). Ad libitum (black), starvation (purple). n = 6(Control LD); 5(Bmal1^IEC-/- LD)/time point (B,C,D). n = 43 independent Ad libitum samples from 6 independent animals, 44 independent Starvation samples from 6 independent animals (F-H). Data are represented as mean ± SEM. Source data are provided as a source datafile.

Fig. 5. Metabolic functioning of intestinal clock-controlled bacteria.

A Heatmap of MetaCyc Pathways predicted using PICRUST2.0 on intestine clock-controlled zOTUs rhythmic in control (left) and arrhythmic in Bmal1^IEC-/- (right) mice. Pathways are colored according to their subclass. B Procrustes analyses (PA) of fecal microbiota and SCFA levels. The length of the line is proportional to the divergence between the data from the same mouse. C SCFA concentrations in feces. D and Spearman correlation (p value ≤ 0.05 and R ≤ − 0.5; red or R ≥ 0.5; blue) between SCFA and gut controlled bacteria taxa. E Conjugated and secondary (F) fecal BAs levels. G PA as described in (D) with fecal bile-acid (BA) levels. H Circadian profiles of BAs. Significant rhythms are illustrated with fitted cosine-regression (solid line); data points connected by dotted lines indicate no significant cosine fit curves (p value > 0.05) and thus no rhythmicity. n = 5(Bmal1^IEC-/-);6(control)/time point (H). n = 40(Bmal1^IEC-/-, 5 biological independent samples), 48(Control, 6 biological independent samples), repeated measures (C, E, F). Control (black) and Bmal1^IEC-/- (red). Data are represented as mean ± SEM. Significance was calculated with two-sided Mann-Whitney U test (C,E,F). BCFA = branched-chain fatty acids. Cholic acid (CA), a-Muricholic acid (aMCA), b-Muricholic acid (bMCA), Taurocholic acid (TCA), Taurochenodeoxycholic acid (TCDCA), Tauroursodeoxycholic acid (TUDCA),Taurohydrodeoxycholic acid (THDCA), Taurolithocholic acid (TLCA), Taurodeoxycholic acid (TDCA), Tauro-a-Muricholic acid (TaMCA), Glycochenodeoxycholic a c(GidCDCA), Glycocholic acid (GCA), Deoxycholic acid (DCA), Lithocholic acid (LCA), y-Muricholic acid (y-MCA), 12-Dehydrocholic acid (12-DHCA), 12-Ketolithocholic acid (12-keto-LCA), 3-Dehydrocholic acid (3-DHCA), 6-Ketolithocholic acid (6- keto-LCA), 7-Dehydrocholic acid (7-DHCA), 7-Sulfocholic acid (7-sulfo-CA), Allocholic acid (ACA), Cholicacid-7ol-3one (CA-7ol-3one), Ursocholic acid (UCA), Dehydrolithocholic acid (DHLCA), Hyodeoxycholic acid (HDCA), Murideoxycholic acid(MDCA), Ursodeoxycholic acid (UDCA). Source data are provided as a source datafile.

Fig. 6. Intestinal clock-controlled microbiota are essential for intestinal homeostasis.

A Schematic illustration of transfer experiments with mixture of cecal microbiota obtained from Bmal1^IECfl/fl control and Bmal1^IEC-/- donors (n = 4) into 10 weeks old germ-free BL6 wild-type recipient mice till sacrifice at 16 weeks (16w). Fecal profiles were collected 5 weeks after gavage at the age of 15 weeks (15w). B MDS plot of GUniFrac distances of donor (CT13) and recipient mice (n = 6/genotype/time point) after 5 weeks of transfer. Significance was calculated by PERMANOVA with Bonferroni correction. C Diurnal profiles of fecal microbiota at phylum level. D Heatmap depicting the relative abundance of zOTUs ordered by their cosine-regression peak phase according to the recipient controls. On the left the first two columns indicate donor zOTU abundance. E Diurnal profile of relative abundance of example zOTUs. F Fecal SCFA and BAs concentrations. G Relative gene expression of Tlr4, Arg2, Ang4, Tnfa, Il33, Nfkb, Hdac3, Lgr5 in the proximal colon of recipient mice of control (black) or Bmal1^IEC-/- (red) cecal microbiota. H Organ weights of recipient mice after receiving control (black) or Bmal1^IEC-/- (red) cecal microbiota. I Frequency of CD3 + CD4 +, CD3 + CD8 +, CD11c + cells in jejunum rhythms are illustrated with fitted cosine-regression; data points connected by dotted lines indicate no significant cosine fit curves (p value > 0.05) and thus no rhythmicity. n = 6 mice/genotype. Data are represented as mean ± SEM. Significance was calculated by Mann-Whitney U test, two-sided (F, G, H, I). Source data are provided as a source datafile.