The intestinal clock drives the microbiome to maintain gastrointestinal homeostasis

Diurnal rhythms in microbiota composition and function have been demonstrated in animal models and in humans^9–12. However, it has not been demonstrated whether these rhythms are a response to rhythmic external cues (Zeitgebers), such as the light-dark cycle, or are generated by endogenous clocks²² and, thus, persist when the organism is placed in an environment devoid of timing cues. To address this question, we compared fecal microbiota rhythms of the same wild-type mice kept in a rhythmic 12-hour light/12-hour dark (LD) cycle and constant darkness (DD) for two weeks (Fig. 1A). Host-driven rhythmic factors, which might influence microbiota composition, such as locomotor activity, food intake as well as total GITT, did not differ between light conditions (Supplementary Fig. 1A-D). 16 S rRNA profiling of fecal samples revealed clustering based on sampling time points rather than light condition (Fig. 1B, E). Generalized UniFrac distances (GUniFrac) quantification to ZT1 identified rhythmicity in both light conditions (Fig. 1C). Importantly, 24-hour rhythms of species richness and Shannon effective number of species found in LD persisted in DD, supporting their circadian origin (Fig. 1D). Rhythms were also preserved in DD on phylum level with the two most dominant phyla Bacteroidetes and Firmicutes, oscillating in antiphase (Fig. 1E, F). Although microbiota composition is commonly analyzed by relative abundance, rhythmicity of highly abundant taxa may mask oscillations of small microbial communities, as previously demonstrated in fecal samples collected in LD¹³. Therefore, using synthetic DNA spikes, we performed relative quantification of the copy number of 16 S rRNA genes according to Tourlousee et al.²³, from here on referred to as ‘quantitative analysis’. With both approaches, highly abundant phyla and families, including Lachnospiraceae and Muribaculaceae, showed comparable circadian rhythmicity in LD and DD (Fig. 1F; Supplementary Fig. 1E). Few families, such as Prevotellaceae showed significant diurnal (LD), but no circadian (DD) rhythmicity, suggesting that their rhythms are regulated by the environmental LD cycle (Supplementary Fig. 1E). After removal of low-abundant taxa (mean relative abundance > 0.1%; prevalence > 10%), the remaining 580 zero-radius OTUs (zOTUs) displayed robust and comparable 24-hour oscillations in both light conditions, independent of the analysis (Fig. 1G, H; Supplementary Fig. 1F–I). The wide distribution of peak abundances, e.g., bacteria peaking during the day (mainly belonging to Muribaculaceae) and during the night (Lachnospiraceae) (Fig. 1G, H; Supplementary Fig. 1F–H), suggests that different microbial taxa dominate different daytimes. Importantly, more than 60% of all identified zOTUs were found to significantly oscillate in LD and in DD (Fig. 1G, H, Supplementary Fig. 1F–H, Supplementary data 1–3), suggesting that their rhythmicity is generated by the circadian system. Notably, comparable amounts of rhythmic zOTUs belong to the major phyla Firmicutes and Bacteriodetes (Supplementary data 2). Previous literature identified lower percentages of microbial rhythmicity^9,10. To assess this discrepancy we compared our results with the previously published data set from Thaiss and colleagues²⁴. Although the different methods to assign taxonomic units, namely OTUs (97% similarity strain assignment) in contrast to zOTUS (100% similarity strain assignment) caused disparity between amounts of rhythmic OTUs/zOTUs, the total abundance of rhythmic microbiota was comparable between both studies (Heddes et al.: 74%, Thaiss et al.: 83%, Supplementary Fig. 1L. Supplementary data 4). Rhythmicity of the majority of zOTUs (e.g. the genera Alistipes, Oscillibacter and Fusimonas) identified in our study by relative analysis was further validated by quantitative analysis (73% in LD, 58% in DD) (Supplementary Fig. 1I, J). The diurnal rhythmicity found in ¾ of all zOTUs examined in LD using JTK_CYCLE²⁵ is generated by the endogenous circadian system, since DODR analysis^26,27 confirmed that more than 95% of zOTUs show similar rhythmicity in DD (e.g. the genera Eubacterium and members of the family Lachnospiraceae) (Fig. 1G; Supplementary Fig. 1K, Supplementary data 1, Supplementary data 3).

Light and food are the main Zeitgeber for the central and peripheral clocks, respectively^2,28. In our study, external light conditions only mildly effected microbial rhythms (Fig. 1) and previous research showed that rhythmic feeding time influences the phasing of microbiota rhythmicity^10,11. This prompted us to determine the dependency of bacterial oscillation on rhythmic food intake behavior by investigating microbiota rhythmicity in the absence of both Zeitgeber food and light. Therefore, we collected feces from starved mice during the 2^nd day in DD (Fig. 2A). Although starvation altered microbial composition, clustering according to collection time points was significant in both feeding conditions (Fig. 2B, D). Microbiota diversity became arrhythmic in the absence of food (Fig. 2C). However, rhythms persisted in the two dominant phyla, highly abundant families, including Muribaculaceae and Lachnospiraceae and the majority of zOTUs (Fig. 2E–H, Supplementary data 1). Of note, circadian food intake behavior enhanced the amount of rhythmic zOTUs by 16% according to DODR analysis on quantitative data (Supplementary data 3). Nevertheless, circadian rhythms of highly abundant microbial taxa belonging to the families Muribaculaceae and Ruminococaceae maintained under starving conditions in both relative and quantitative analysis (Fig. 2E–I).

Together, these data show that although rhythms in gut bacteria are slightly influenced by environmental light and food conditions, intrinsic factors are the dominant drivers of circadian microbiota fluctuations.

An important role of the host clock on diurnal microbiota fluctuation has previously been suggested based on experiments on mice with clock dysfunction in all tissues under LD conditions^10,13,14. The GI tract represents an important interface for cross-talk between IECs, bacteria and their metabolites²⁹ and thus might likely be involved in controlling microbiota composition. To identify the circadian origin of microbial rhythmicity, we used mice with IEC-specific deletion of the essential core clock gene Bmal1 (Bmal1^IEC-/-). The rhythmic activity, food intake behavior as well as GITT and stool weight of Bmal1^IEC-/- mice did not differ from control littermates, reflecting a functional central pacemaker (Supplementary Fig. 2A–E). Arrhythmic core clock gene expression in the jejunum, cecum and proximal colon, in contrast to rhythms in the liver, confirmed intestine-specific clock ablation in Bmal1^IEC-/- mice (Supplementary Fig. 2F). Rhythmic genes relevant for host-microbe crosstalk, in particular Tlr2, Muc2, Nfil3 and Hdac3^30–35 lost rhythmicity in Bmal1^IEC-/- mice (Supplementary Fig. 2G). In line with this, microbiota composition in DD and LD differed significantly between Bmal1^IEC-/- mice and controls (p = 0.037 DD, p = 0.018 LD) and circadian rhythmicity in community diversity (species richness) observed in control mice was abolished in Bmal1^IEC-/- mice (Fig. 3A, B, Fig. 4A). Quantitative analysis revealed loss of rhythmicity of Firmicutes in Bmal1^IEC-/- mice in LD and DD (LD p = 0.514, DD p = 0.313), although, oscillation in relative abundance of the major phyla was found in both genotypes (Figs. 3C, Fig. 4C). Alterations in quantitative compared to relative rhythmicity analyses are not due to differences in 16 s copy number over time, as both genotypes show similar patterns in LD and DD conditions (Supplementary Fig. 2H, Fig. 4B). In accordance to results obtained from whole-body clock-deficient mice¹³, IEC-specific loss of Bmal1 decreased the quantitative abundance of both phyla, which was not visible in relative data sets (Fig. 3C, Fig. 4C). Similarly, rhythmicity observed in the relative abundance of families, including Lachnospiraceae and Ruminococcaceae, was abolished performing quantitative analysis (Supplementary Fig. 2I). Furthermore, dramatically disrupted circadian oscillations of Bmal1^IEC-/- mice in both relative and quantitative analyses was observed on the level of all 580 identified zOTUs illustrated by heatmaps (Fig. 3D, E). Of note, in Bmal1^IEC-/- mice the phase of remaining microbial oscillations was delayed using quantitative analysis (Fig. 3D, E). Based on both analyses more than 60% of fecal zOTUs, which account for more than two-third of bacterial abundance, underwent circadian oscillation in controls (Fig. 3D, E, Supplementary data 1). Importantly, the amount of rhythmic zOTUs was dramatically reduced by 2/3 in mice with IEC-specific Bmal1 deficiency in both analyses, independent of the light conditions (Fig. 3D, E, Fig. 4D, I, Supplementary data 1–3). Circadian zOTUs in control mice, which were arrhythmic in Bmal1^IEC-/- mice in both relative and quantitative analysis, belong predominantly to the families Lachnospiraceae and Ruminococcaceae of the phylum Firmicutes (Supplementary Fig. 3J, Supplementary data 2). The genera Lactobacillus, Ruminococcus, Anaerotignum, Odoribacter and Alistipes are among the taxa with differential rhythmicity comparing Bmal1^IEC-/- mice with their controls based on DODR analysis and in addition significantly differed in their abundance between genotypes (Fig. 3F, G). Although the majority of zOTUs was under intestinal clock control, rhythmicity of around 20% of zOTUs (e.g. Lactobacillus animalis and Pseudoflavonifractor) persisted in Bmal1^IEC-/- mice (Fig. 3D, E, Supplementary data 1). To address whether these remaining rhythms were driven by the rhythmic food intake behavior observed in Bmal1^IEC-/- mice kept under ad libitum conditions in DD (Supplementary Fig. 2D), microbiota rhythmicity in Bmal1^IEC-/- mice was analyzed in feces from starved mice during the 2^nd day in DD (Fig. 2A). Although beta diversity was significantly altered in the absence of rhythmic food intake, profiles of major phyla were undistinguishable between feeding conditions (Fig. 4E, F). Only a small amount of the remaining rhythmic zOTUs in Bmal1^IEC-/- mice significantly lost rhythmicity upon starvation based on DODR analysis (Supplementary data 3), such as bacteria belonging to Turicimonas muris and Muribaculaceae (Fig. 4G-I, Supplementary data 1).

Taken together, rhythmicity analyses of zOTU in Bmal1^IEC-/- mice and controls under different light and food conditions (Fig. 3, Fig. 4, Supplementary Fig. 2) highlight that the IEC clock represent the dominant driver of circadian gut microbiota oscillation.

To address the potential physiological relevance of microbial oscillations, PICRUST 2.0 analysis was performed on intestinal-clock controlled zOTUs. Loss of microbial rhythmicity in Bmal1^IEC-/- mice was reflected in assigned pathways involved in sugar- and amino acid metabolism, vitamin biosynthesis and relevant for fatty acid (FA) metabolism such as β-oxidation, FA elongation and short-chain FA (SCFA) fermentation (Fig. 5A, Supplementary Fig. 3I). This prompted us to test the functional connection between intestinal clock-driven microbial rhythmicity and intestinal homeostasis (Fig. 5B–H). Indeed, Procrustes analyses (PA) identified an association between intestinal clock-controlled zOTUs and SCFA concentrations measured by targeted metabolomics (p = 0.001) (Fig. 5B). The level of valeric acid (p = 0.04, Mann-Whitney U test) and low-abundant branched-chain fatty acids (BCFAs) (p = 0.01, Mann-Whitney U test), including isovaleric, isobutyric, 2-methylbutyric differed between genotypes, although the amount of total and highly abundant SCFAs were comparable (Fig. 5C, Supplementary Fig. 3A). Nevertheless, concentrations of SCFAs negatively correlated with the relative abundance of intestinal clock-driven taxa, mainly belonging to the family Muribaculaceae of the phylum Bacteroidetes (Fig. 5D, Supplementary Fig. 3B). Positive significant correlations with multiple SCFAs, in particular BCFAs, where found with zOTUs belonging to the phylum Firmicutes, including the SCFA producers Lachnospiraceae and Ruminococcaceae (Supplementary Fig. 3B)³⁶. Similar to the results observed for SCFAs, associations were found between intestinal clock-controlled zOTUs and BAs levels (p = 0.04, Fig. 5G). Positive correlations with primary BAs and negative correlations with secondary BAs were observed with several intestinal clock-controlled taxonomic members mainly belonging to Firmicutes (Fig. 5D, Supplementary Fig. 3D). For example, DHLCA and 7-sulfo-CA negatively correlated with Oscillibacter and Eubacterium (Fig. 5D, Supplementary Fig. 3D). In addition, almost half of the measured BAs differed among genotypes and, BAs, such as the conjugated primary BA, TCDA, and secondary BA, 7-sulfo-CA, lost rhythmicity in Bmal1^IEC-/- mice (Fig. 5E-F). Interestingly, most alterations were observed for secondary BAs (Fig. 5H, Supplementary Fig. 3C), which were linked to various GI diseases (reviewed by³⁷). In summary, these results indicate that deletion of intestinal clock function causes loss of rhythmicity in the majority of bacterial taxa and alters microbial functionality.

To further test the physiological relevance of intestinal clock-driven rhythmic microbial composition and its functionality on host physiology, GF C57BL/6 mice were colonized with cecal content from Bmal1^IEC-/- or control mice (n = 4) (Fig. 6A). To confirm the importance of the gut clock in cecal microbial rhythmicity and composition, we performed 16 S sequencing as well as targeted metabolomics on cecal content of the donor mice. In line with results obtained from feces, 50% of the rhythmic zOTUs in the controls lost rhythmicity within cecal content of Bmal1^IEC-/- mice (Supplementary Fig. 4A-B). Moreover, several SCFA, including isovalerate and 2-methylbutyrate lost rhythmicity in Bmal1^IEC-/- mice cecal content (Supplementary Fig. 4C).

Importantly, ~76% of bacteria from the control and Bmal1^IEC-/- donor, were transferred into the recipients (Supplementary Fig. 3E). Reduced richness was observed in mice after microbiota transfer (Supplementary Fig. 3F), in line with previous reports³⁸. Interestingly, microbiota composition of recipient wild-type hosts significantly differed depending on the genotype of the donor (Fig. 6B; Supplementary Fig. 3G, H). Most of the transferred taxa belonged to Firmicutes, whose rhythmicity strongly depends on a functional intestinal clock, whereas abundances in Bacteroidetes were highly suppressed (Fig. 6C; Supplementary Fig. 3G). Importantly, lack of rhythmicity was transferred to mice receiving microbiota from Bmal1^IEC-/- mice (Fig. 6D, E). In particular rhythmicity of zOTUs belonging to e.g. Alistipes, shown to be driven by the gut clock (Fig. 3F) and maintained circadian rhythmicity in the host after transfer from control donors, lack rhythmicity in mice receiving microbiota from Bmal1^IEC-/- mice (Fig. 6E). Interestingly, after transfer of Bmal1^IEC-/- -associated arrhythmic microbiota several bacterial-derived metabolites, including BCFA as well as the secondary bile ketolitocholic acids were significantly upregulated and the primary BA β-muricholic acid (b-MCA) was significantly downregulated (Fig. 6F), similar to observations in made in donor mice (Fig. 5C, F, Supplementary Fig. 3A-D). Moreover, altered expression of intestinal clock genes in jejunum and colon was observed after transfer of arrhythmic microbiota in comparison to recipient mice receiving rhythmic microbiota (Supplementary Fig. 5A). In addition, the expression of genes involved in host-microbe interaction, including Arg2 and Tlr4, significantly differed depending on the genotype of the donor (Fig. 6G). Bmal1^IEC-/- -associated microbiota transfer induced the expression of Il33 and NfkB both known to be involved in intestinal inflammatory responses^39,40 and reduced the expression of Ang4 and Hdac3 (Fig. 6G), which are known to integrate microbial and circadian cues relevant for inflammatory and metabolic intestinal functions^{31,33,35,41,42}. Although mesenteric lymph nodes and colon weights of mice associated with arrhythmic microbiota were undistinguishable from controls and histology scores were unaffected, we noticed an increase in spleen and jejunum weight (Fig. 6H, Supplementary Fig. 5B, C), indicating a role of intestinal clock-controlled microbial on host intestinal homeostasis. Indeed, transfer of arrhythmic microbiota altered immune cell recruitment to the lamina propria. For example, an increase in dendritic cells (CD11c +) was detected in the small intestine whereas recruitment of T cells (CD3 + CD4 + , CD3 + CD8 +) and dendritic cells (CD11c +) to the lamina propria of the colon was decreased (Fig. 6I, Supplementary Fig. 5D). Similar trends in immune cell recruitment and intestinal gene expression were observed in Bmal1^IEC-/- mice in SPF conditions (Supplementary Fig. 5E, F), suggesting that intestinal-clock controlled microbes influence gut immune homeostasis. Importantly, alterations found following transfer of Bmal1^IEC-/- -associated arrhythmic microbiota were not observed after transfer of microbiota from control mice kept in either LD, DD or starvation conditions, highlighting the physiological relevance of gut-clock-controlled microbiota (Supplementary Fig. 5G-I). Distinct gut microbial functionality controlled through food, light or the gut clock were further conformed through PICRUST analysis (Supplementary Fig. 6). In particular, a completely different metabolic signature was obtained in mice lacking a functional gut clock. For example, although loss of pathways in nucleotide metabolism was shared between conditions, loss of the gut clock additionally disrupts rhythmicity of pathways related to the TCA cycle and sugar metabolism (Supplementary Fig. 6). The profiles of LD and food regulated pathways were instead dominated by short chain fatty acid (SCFAs) metabolism, Amino acid metabolism and Vitamin, carrier and cofactor biosynthesis. In conclusion, our results support a different impact of the gut clock on microbial functionality compared to light and food.

Together, these results indicate that deletion of intestinal clock function caused loss of rhythmicity in the majority of bacterial taxa, alters microbial composition and functionality and resets the host’s intestinal homeostasis.

Experiments were conducted at Technical University of Munich in accordance with Bavarian Animal Care and Use Committee (TVA ROB-55.2Vet-2532.Vet_02-18-14).

All male mice, unless stated otherwise, were individually housed in cages with running wheels. Handling and activity measurements during experiments were performed as described⁷⁹. Wheel-running activity was analyzed using ClockLab software (v6.0.52, Actimetrics). The last 10–14 days of each condition were used to determine the period (tau, calculated using a X² periodogram and confirmed by fitting a line to the onsets of activity), the duration of the active period (alpha), the amount of activity and the subjective day/night activity ratio (where the subjective day under DD conditions is the inactive period between the offset of activity and the onset of activity and the subjective night is the active period between the onset of activity and the offset of activity).

Male Bmal1^IEC-/- and their control littermates Bmal1^IECfl/fl (n = 6/genotype) were maintained under LD cycle for 2 weeks (age 8–10 weeks), switched to a DD cycle for 2 more weeks (age 10–12 weeks), kept in constant light (LL) for an additional 2 weeks (age 12-14w) and finally returned back to LD till the age of 18–20 weeks before sacrifice. Average daily food intake was measured over 5 consecutive days in the second week of the above indicated light conditions. Of note, Fecal sample collection in darkness was performed after adjusting for each mouse’s individual free-running period (Supplementary Fig. 2A). For example, the activity period varied by ~0.5 hours per day within mice of the same genotype and consequently accumulates to a ~7-hour phase difference between individual mice after 2 weeks in DD. Fecal collection in LD was performed according to normal Zeitgeber time. Samples were collected every 3 hours over the course of a 24h-day (Fig. 1A, Supplementary Fig. 2A).

Eleven-week-old Bmal1^IEC-/- and control male mice (n = 11) with ad libitum food intake were used for fecal sample collection. At the age of 12-13 weeks, Bmal1^IEC-/- mice (n = 12) and their control (n = 11) littermates were starved beginning at CT1 on the 1^st day in DD. Fecal samples were collected as detailed above starting at CT13 on the 2^nd day in DD (after 13 hours of starvation) till CT10 (after 34 hours of starvation) at the indicated time points (Fig. 2A).

The diurnal food intake pattern of individually housed mice was recorded using an automated monitoring system (TSE LabMaster Home Cage Activity, Bad Homburg, Germany) at the age of 10–12 weeks. Mice were habilitated for 3 days. Then data were collected for a full 24-hours profile at the 4^th day. Cumulative food intake was recorded through high precision balances connected to the food baskets. Food basket weight were summed up in 1 min intervals. Total food intake was calculated as 1^st derivative of cumulative food intake and consumption was summed up at intervals of 1 hour. One-minute intervals where the weight loss was directly followed by a similar weight gain in food baskets within a period of 3 minutes were excluded. Circadian food intake of individually housed mice was recorded in the second day of darkness by weighing the food every 3 h.

Complete GI transit time (GITT) was measured by administering natural carmine red (6%, Sigma Aldrich) dissolved in 0.5% methylcellulose (Sigma-Aldrich) by gavage (100ul) according to⁸⁰. 16–18 week-old male Bmal1^IEC-/- and Bmal1^fl/fl mice were starved 6 h prior to gavage. Time of gavage was considered T0. In case of presence, fecal pellets were taken out of the cage every 10 minutes and checked for red color. GITT was registered as the time between T0 and the time of presence of carmine red in the fecal pellets.

All animals were sacrificed by cervical dislocation at the age of 18–20 weeks in the second day of darkness at the indicated circadian times (CT) or in LD12:12 conditions, in 4 h intervals starting at 1 hour after lights off (ZT1). In constant darkness samples were collected at the indicated time points used for control mice in LD. Eyes were removed prior to tissue dissection. Tissues were harvested and snap frozen on dry ice and stored in −80 degrees.

RNA was extracted from snap frozen tissue samples with Trizol reagent. cDNA was synthesized from 1000 ng RNA using cDNA synthesis kit Multiscribe RT (Thermofischer Scientific). qPCR was performed in a Light Cylcer 480 system (Roche Diagnostiscs, Mannheim, Germany) using Universal Probe Library system according to manufacturer’s instructions. For genes expression the following primers and probes were used: Brain and Muscle ARNT-Like 1 (Bmal1) F 5′-ATTCCAGGGGGAACCAGA-’ R 5′-GGCGATGACCCTCTTATCC-3′ Probe 15, Period 2 (Per2) F 5′-TCCGAGTATATCGTGAAGAACG-3′ R 5′- CAGGATCTTCCCAGAAACCA-3′ probe 5, Nuclear receptor subfamily 1 group D member 1 (Reverbα) F 5′- AGGAGCTGGGCCTATTCAC-3′ R 5′-CGGTTCTTCAGCACCAGAG-3′ probe 1, Toll-like receptor 2 (Tlr2) F- 5′-GGGGCTTCACTTCTCTGCTT-3′ R 5′-AGCA TCCTCTGAGATTTGACG-3′ probe 50, Angiogenin 4 (Ang4) F 5′-CCCCAGTTGGAGGAAAGC-3′ R 5′-CGTAGGAATTTTTCGTACCTTTCA-3′ probe 106, Mucin 2 (Muc2) F 5′- GGCAGTACAAG AACCGGAGt-3′ R 5′-GGTCTGGCAGTCCTCGAA-3′ probe 66, Histone Deacetylase 3 (Hdac3) F 5′- GAGAGGTC CCGAGGAGAAC-3′ R 5′-CGCCATCATAGAACTCATTGG-3′ probe 40, Tumor necrosis factor alpha (Tnfa) F 5′- TGCCTATGTCTCAGCCTCTTC-3′ R 5′- GAGGCCATTTGGGAACTTCT-3′ probe 49, Leucine Rich Repeat Containing G Protein-Coupled Receptor 5 (Lgr5) F 5′- CTTCACTCGGTGCAGTGCT-3′ R 5′- CAGCCAGCTACCAAATAGGTG-3′ probe 60, Toll like receptor 4 (Tlr4) F 5′-GGACTCTGATCATGGCACTG −3′ R 5′-CTGATCCATGCATTGGTAGGT-3′ probe 2, Nuclear factor kappa-light-chain-enhancer of activated B cells (Nfkb) F 5′- CCCAGACCGCAGTATCCAT −3′ R 5′- GCTCCAGGTCTCGCTTCTT-3′ probe 47. RNA abundance was normalized to the housekeeping gene Elongation factor 1-alpha (Ef1a) F 5′- GCCAAT TTCTGGTTGGAATG-3′ R 5′-GGTGACTTTCCATCCCTTGA-3′ probe 67. For Interleukin33 (Il33) gene was used sybr green to run the gene using the following primer F 5′- GAACATGAGTCCCATCAAAG −3′ R 5′- CAGCTGGTTATCTTTTACTCC −3′ and RNA abundance was normalized to the housekeeping gene Ef1a F 5′- GCCAAT TTCTGGTTGGAATG-3′ R 5′-GGTGACTTTCCATCCCTTGA-3′.

Genomic DNA was isolated from snap-frozen fecal pellets according to a modified protocol of Godon and colleagues⁸¹, as previously described¹². DNA NucleoSpin gDNA columns (Machery-Nagel, No. 740230.250) were used for DNA purification. In a two-step PCR the V3-V4 region (using the primers 341F-ovh and 785r-ov) of the 16 S rRNA gene was amplified from 24 ng DNA. After pooling, the multiplexed samples were sequenced on an Illumina HiSeq in paired-end mode (2×250 bp) using the Rapid v2 chemistry, in accordance with¹². Two negative controls, consisting of DNA stabilizer without stool, were used for every 45 samples to control for artifacts and insure reproducibility. High-Quality sequences of read counts > 5000 were used for 16 s rRNA data analysis. Reads FASTQ files were consequently processed using an in-house developed NGSToolkit (Version Toolkit 3.5.2_64) based on USEARCH 11⁸². A trim score of 5 was used on the 5′ end and 3′end for the R1 and R2 read followed by chimera removal⁸³ using the FASTQ mergepair script of USEARCH⁸². Quality filtered reads were merged, deduplicated, clustered and a denoised clustering approach was applied to generate zero-radius operational taxonomic units (zOTUs)^82,84. Using zOTUs analysis provide the best possible resolution considering the 16 s rRNA sequencing, by correcting the sequencing error (denoising step) and then identifying each unique sequence (100% similarity) as a different microbial strain. On the other hand, previously OTU analysis was used, which is based on merging different strains which has more than 97% similarity into one single OTU, and assign it to one microbial strain⁸⁵. Taxonomic assignment was performed with the EZBiocloud database⁸⁶. Notably, For Supplementary Fig. 1L Taxa were assigned to Greengenes database in order to be able to compare our results with the data obtained in the study by Thaiss and colleagues¹⁰. Data was further analyzed with the R-based pipeline RHEA⁸⁷. Phylogenetic trees are generated by a maximum likelihood approach, which was performed on an alignment generated by MUSCLE with the software MegaX⁸⁸. Trees were visualized and annotated with the use of the online tool EvolView (http://www.evolgenius.info/evolview/↗)⁸⁹. Spike-in of 12 artificial DNA standards mimicking 16 S rRNA genes (a surrogate for bacterial numbers) were used to determine the quantitative copy numbers of rRNA genes per gram of fecal sample between samples. Briefly, the same amount of artificial DNA (6 ng) was added to each weighted fecal sample before DNA extraction. After sequencing as described above, FASTQ files were mapped against the spike FASTA sequences (using bowtie2), removing the spike reads and generating a new FASTQ file. By comparing the spike sequencing reads to the fecal bacterial reads; we calculate the quantitative number of 16 S rRNA gene copies per gram of sample. The copy number of 16 S rRNA gene is proportional to the number of bacteria present in a sample. Thus, this approach enables estimation of microbial abundances relative between samples, suitable for comparative analysis according to Tourlousse et al. 2017²³.

Mice were gavaged with cecal microbiota at ZT13 from either cecal content (CT/ZT13) of Bmal1^IEC-/- mice (n = 4, mixture), or control mice kept in LD (n = 5, mixture), DD (n = 4) or starvation (n = 5) into germ-free wild type C57BL6 recipient mice at the age of 10 weeks. 100 µl of 7 × 10⁶ bacteria/µl were used for gavaging each mouse. Mice were weekly monitored for bodyweight changes and feces was sampled at week 5 after gavage in DNA stabilizer for 16 s analyses. Native fecal samples for targeted metabolomics were taken at ZT4 and ZT16. Afterwards, mice were sacrificed at the 2nd day in DD at CT13 at the age of 16 weeks. All mice were kept in the gnotobiology facility in isolators equipped with HEPA-filters at 22 ± 1 °C with a 12-h light/dark cycle (lights on 5 am till 5 pm). Mice were single housed and had ad libitum access to autoclaved chow (V1124-300, Sniff Diets, Soest, Germany) and autoclaved water.

Immune cells were isolated from freshly isolated jejunum and colon. Intestinal tissues were flipped, washed out and cut into 1 cm pieces. To remove epithelial cells, pieces were incubated in DMEM with 20 µL of 1 M DTT. After shaking for 15 minutes, tissues were incubated PBS with 200 µL of 150 mM EDTA at 37 °C with shaking. After 3 times harsh shaking in Hanks buffer, jejunum tissue was then digested at 37 °C for approximately 15 min in Thermoshake (200 rpm) with 0.6 mg/ml type VIII collagenase (Sigma-Aldrich). Colonic tissue was simultaneously digested under the same condition for approximately 25 min but with the mixture of 0.85 mg/ml type V collagenase (Sigma-Aldrich), 1.25 mg/ml collagenase D (Sigma-Aldrich), 10 µl/ml Amphotericin (100x) 1 mg/mL Dispase II, and 30U/ml Deoxyribonuclease I (Sigma). Following digestion, intestinal cells were passed through a 40 µm straine. Consequently, cells were fixed with 2% PFA, washed, and stored in RPMI at 4 °C until further processing. r

For intracellular stainings, cells were permeabilized with saponin 0.5% and stained with anti CD8 PE-, anti CD3 PerCP/Cy5.5, anti CD4 FITC-, anti-IL-17a PE/cy7-, anti Infy- APC conjugated antibodies at dilution 1/100-1/50 for 30 min.

Surface stainings were performed using anti CD11c PE-, anti CD11b APC Cy7-, anti F4/80 PE/cy7-, anti Ly6G APC conjugated antibodies at dilution 1/50 for 30 minutes. Cells were analyzed on a LSR-II (BD Biosciences) flow cytometer and analysis was performed using FlowJo software (FlowJo, LLC, version 10.7.2). Gating strategies are illustrated in Supplementary Fig. 7

In formalin fixated tissues in paraffine were cut into 5μm thick slices and consequently stained according to the following steps: xylene/ 5 min, xylene/ 5 min, Ethanol 100%/ 5 min, Ethanol 100%/ 5 min, Ethanol 96%/ 2 min, Ethanol 96%/ 2 min, Ethanol 70%/ 2 min, Ethanol 70%/ 2 min, Water/ 30 s, hematoxylin/ 2 min, tap water/ 15 s, Scotts Tap Water/ 30 s, Water/ 30 s, Ethanol 96%/ 30 s, Eosin/ 30 s, Ethanol 96%/ 30 s, Ethanol 96%/ 30 s, Ethanol 100%/ 30 s, Ethanol 100%/ 30 s, Xylene/ 90 s, Xylene/ 90 s (Leica ST5020 multistainer). DPX new mounting media (Merck) was added to preserve the tissues. Stained slides were scanned and further analyzed for histological scoring. Histological scores were assessed blindly based on the degree of immune cell infiltration of all colonic wall layers (mucosa, submucosa and muscularis), crypt hyperplasia, goblet cell depletion and mucosal damage, resulting in a score from 0 (not inflamed) to 12 (severely inflamed) according to Katakura method⁹³.

Antigens retrieval of the de-paraffinized and re-hydrated sections was performed by heating the slices in citrate buffer for 34 minutes. After multiple washing steps and tissues were blocked by donkey blocking buffer (1 h at room temperature and 1 hour at 4 °C). Tissue sections were incubated overnight with Anti E-cadherin (1:300, mouse, Abcam) anti-CD3 (1:400, rabbit, Sigma). Tissue were washed with PBS and followed by incubation with secondary antibodies (donkey antirabbit 546, donkey antimouse 647, both from Invitrogen with a dilution 1:200) for 1 h at room temperature. Finally, DAPI (Sigma) were used to stain the nuclei and then tissues were mounted using Aquatex. Sections were visualized with Fluoview FV10i microscope (Olympus, Shinjuku, Japan).

For prediction of functional of metagenomic functionality. Sequence of the gut controlled zOTUs, captured as described above, were used to construct the metagenome using PICRUST2.0⁹⁴. Corrected zOTU 16 s rRNA gene copy number is multiplied by the predicted functionality to predicted the metagenome. Resulted enzymatic genes classified according to Enzyme Commission (EC) numbers were mapped to Metacyc pathways. Superclasses were removed and Metacyc pathways abundance was used for statistical analysis using STAMP (2.1.3). Statistical differences were calculated based on White’s non-parametric t-test and Benjamini Hochberg dales discovery rate to adjusted for multiple testing.

Statistical analyses were preformed using GraphPad Prism, version 9.0.0 (GraphPad Software), JTK_cycle v3.1.R (²⁵) or R. With the use of the pipeline Rhea (Lagkouvardos) between-sample microbiota diversity is calculated by generalized UniFrac using GUniFrac v1.1. distances. Quantification of GUniFrac distances was always done in comparison to ZT1 or CT1. Different in microbiota composition was calcularted by PERMANOVA test on generalized Unifrac dissimilarity matrix and illustrated by MDS, The bar next to the graph indicate the dissimilarity for the distance of the bar⁸⁷. Circadian and diurnal patterns of individual 24 h period graphs, as well as phase calculation were analysed by fitting a cosine-wave equation: y=baseline + (amplitude∙cos(2 ∙ π ∙ ((x-[phase shift)/24))), with a fixed 24-h period for phyla, richness and single zOTU or by using the non-parametric algorithm JTK_cycle for overall rhythmicity of all zOTUs. Connected straight lines in of individual 24 h period graphs within the figures indicate significant rhythmicity based on cosine analyses whereas dashed lines indicate non-significant cosine fit. Comparison of rhythm between data sets were preformed using an adjusted version of the Compare rhythm script²⁶, which performs JTK_cycle followed by DODR⁹⁵ available at (DOI: 10.5281/zenodo.7049842). Some CCGs when indicated were analysed with a harmonic - regression: y= baseline + (amplitude A * cos (2 ∙ π ∙ ((x - [phaseshift A) / 24))) + (amplitude B * cos (4 ∙ π ∙ ((x - [phase shift B) / 24))). Amplitude calculations depicted in the manhattan plots are based on the output of JTK_cycle and the phase was calculated by cosine-wave regression. Analysis between two groups was performed using the non-parametric Mann-Whitney test (two-sided). A p value ≤0.05 was assumed as statistically significant.

Heatmaps were generated using the online tool “heatmapper.ca”⁹⁶. Heatmaps were sorted based on peak phase of controls. Abundance plots were generated using SIAMCAT package (v1.10.0) in R using the “check.associations() function”⁹⁷. Visualisation of zOTUs and samples trees were conducted using the online platform “evolgenius.info”⁸⁹.

Further information on research design is available in thelinked to this article. Nature Research Reporting Summary

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work.

The Microbiota sequencing data and metabolite data generated for this study have been deposited in Sequence Read Archive (SRA, submission ID: SUB10646422/Bioproject ID: PRJNA779969) and Zenodo (10.5281/zenodo.6962318). The processed data generated in this study are provided in theand as a Source Data file. Cosine wave regression P values are provided in Supplementary data.are provided with this paper. Supplementary Information 5 Source data

The modified version of the compare rhythm script is availabe (10.5281/zenodo.7049842).