The microbiome interacts with the circadian clock and dietary composition to regulate metabolite cycling in the Drosophila gut

The w¹¹⁸ iso31, per¹ fly lines present in the study were maintained on standard cornmeal/molasses medium at 25°C under 12:12 LD conditions. The germ-free flies used in this study are the same as in the previous report (Feltzin et al., 2019; Zhang et al., 2023). The Newborn fly embryos within 12 hours were rinsed in 100% ethanol, dechorionated in 10% bleach for 2 minutes, then immediately rinsed three times in germ-free PBS. The germ-free embryos were transferred to autoclaved standard germ-free molasses-cornmeal-yeast medium containing 1 mM kanamycin, 650 μM ampicillin (61-238-RH, MediaTech), and 650 μM doxycycline (D9891, Sigma-Aldrich). Germ-free flies were maintained on germ-free mediums with three antibiotics for four generations and then maintained on the same medium without antibiotics for the next four generations. Bacterial contamination of flies was monitored by homogenizing larvae in germ-free PBS. Aspirate supernatant after brief centrifugation and monitoring the bacterial growth on DeMan, Rogosa, and Sharpe-agar plates (Lekova, 2017).

Diets preparation followed the protocols below. Normal chow diets: 64.67 g of corn meal, 27.1 g of dry yeast, 8 g of agar, and 61.6 mL of molasses in 1 L ddH₂O. For high protein diets: 64.67 g of corn meal, 150 g of dry yeast, 8 g of agar, and 61.6 mL of molasses in 1 L of ddH₂O. For high sugar diets: 64.67 g of corn meal, 27.1 g of dry yeast, 8 g of agar, 61.6 mL of molasses, and 243 g of sucrose in 1 L of ddH₂O (Bedont et al., 2021; Morris et al., 2012). After autoclaving the diets at high temperature, temperature was allowed to drop to 40°C and then 10.17 mL of 20% Tegosept was added to each 1 L diet and mixed thoroughly. Germ-free conditions were maintained throughout. For all treatments, 5–7-day-old adults laid eggs on their respective diets (is031 and per¹ flies were fed normal chow diets, while is031 flies were fed high protein or high sugar diets). The subsequent generation of newly hatched adults continued to feed on their respective diets for 3 days. Then, newly hatched female adult flies aged 5–7 days were separated by gender and subjected to the feeding paradigm in the presence of light-dark cycles beginning on day 5. In the timed feeding experiment, both ad libitum (AF) flies and those subjected to timed feeding (TF) were transferred to germ-free medium vials at Zeitgeber time 0 (ZT0), which corresponds to the actual time of 9:00 am. They were then relocated into either a new germ-free medium or a 1.1% agar vial at ZT10 (7:00 p.m.) for a 14-hour fast. In the ad lib feeding experiment, per¹ flies were fed normal chow diets, while iso31 flies received high protein and high sugar diets. AF flies were maintained in germ-free medium vials at all times, with germ-free fresh diets replaced according to the timed feeding schedule. Fly female guts were dissected on the fifth day at ZT0, ZT6, ZT12, and ZT18 within a 12:12 light-dark (LD) cycle. This dissection took place after the flies had been continuously fed for 4 days under ad lib or timed feeding conditions. For each condition, four biological repeat samples were collected both from microbiome-containing and germ-free female flies (>180 guts per sample, which corresponds to at least 10 mg for each sample). All experimental processes were carried out on ice, and samples quickly transferred to dry ice for storage after each sample was collected.

For 24-hour feeding, we used the Ex-Q assay. Fifteen females were pre-fed on the three experimental diets described above for 3 days. On day 4, flies were transferred to the same food supplemented with 2.5 mg/mL of FD&C blue FCF brilliant blue dye for 48 hours. Flies were switched back to the specific diet without the dye for an extra 24 hours. Excrement was suspended into 150 uL of MilliQ water. Absorbance was measured at 620 nm. Raw data was divided by the number of flies and further normalized to the minimum value of the entire dataset. For time courses, females were pre-fed on the three experimental diets described above for 3 days. On day 4, groups of three flies were transferred to the same food supplemented with 2.5 mg/mL of FD&C blue FCF brilliant blue dye for 4 hours (i.e., for the ZT2 collection, flies were transferred to the blue food at ZT22 and collected at ZT2). Flies were collected and homogenized using a Tissue Lyser in 200 ul of PBS. After centrifugation to remove debris, absorbance was measured at 620 nm. Data was divided by the number of flies and further normalized to the minimum of each dataset, high protein and high sugar separately.

After acquisition, primary metabolomics data is converted from LECO peg files to NetCDF files for processing. Files are submitted to in-house data processor BinBase for annotation and a report is exported into an Excel file. In Excel, data is curated by blank reduction with a fold change of 3 or lower by using the max peak height from samples/average blank peak height. After reduction, peak heights are normalized by sum normalization. This is done using the un-normalized peak height of the feature divided by the mTIC (sum of the annotated metabolites) of the respective sample multiplied by the average mTIC of the samples. Normalized peak heights are then reported in an Excel document. Chromatograms first undergo a quality control check in which internal standards are examined for consistency of peak height and retention time. Raw data files are then processed using an updated version of MS-DIAL software, which identifies and aligns peaks and then annotates peaks using both an in-house mzRT library and MS/MS spectral matching with NIST/MoNA libraries. All MS/MS annotations are then manually curated by a lab member to ensure that only high-quality compound identifications are included in the final report. Enrichment analysis was conducted through the online software MetaboAnalyst 5.0.

Statistical details of experiments can be found in the figure legends. Circadian statistical analysis was performed in R using JTK_CYCLEv3.1. JTK_p-value <0.05 and JTK BH.Q <0.2 deem as cycling metabolites (Pang et al., 2022).

We showed previously that the Drosophila gut microbiome does not exhibit obvious oscillations in a 12:12 hours LD cycle, but it significantly reduces the number and strength of host transcript cycling (Zhang et al., 2023). To investigate the role of the microbiome in gut metabolite cycling, we assayed metabolite cycling at different times of day in microbiome-containing (AM) and germ-free (AS) fed ad libitum (Figure 1A). Metabolic profiling was performed separately to characterize primary metabolites, lipids, and biogenic amines (MetaboLights datasets MTBLS8819 contain the raw data). Principal component analysis (PCA) of the total expressed primary metabolites did not show significant separation between AM and AS flies at any time of day, indicating that this class of molecules is not changed by loss of the microbiome. However, a clear separation between the two groups was observed for lipids and biogenic amines (Figure 1—figure supplement 1A–C).

Given our transcriptomic results, we were expecting that the gut microbiome would reduce the number of cycling metabolites; however, we observed the opposite. JTK cycling analysis of the three different types of metabolites revealed that the likelihood of oscillations increased in AM flies compared to AS flies. 26.9% (227 of 843) primary metabolites were rhythmic in AM flies compared to 20.3% (171 of 843) in AS flies. A further reduction in AS was observed for lipids whose rhythms dropped from 29.5% (1329 of 4510) cycling in AM to 14.4% (648 of 4510) in AS. Similarly, the proportion of biogenic amines cycling decreased from 39.0% (1689 of 4330) to 24.7% (1068 of 4330) (JTK_p-value <0.05) (Figure 1B, D, F, H, J and L). In addition, only about half of the cycling metabolites in germ-free flies are shared with microbiome-containing flies (Figure 1—figure supplement 2). Consistent with the role of the microbiome in favoring metabolic rhythms, lipids and biogenic amines that cycle in both AM and AS flies have a higher amplitude in AM flies (Figure 1—figure supplement 3A–C). Although the microbiome did not significantly change the overall distribution of phases for the cycling primary metabolites and lipids, it had a substantial effect on the phase of biogenic amine oscillations that shifted from peaking in the middle of the day (ZT6) in the AM flies to a peak in the early morning (ZT0) in AS flies (Figure 1B, D, F, H, J and L). Morning peaks were also noted for lipids under almost all conditions, which may be driven by the morning peak of feeding.

Analysis of the abundance of cycling metabolites in AM, compared to their non-cycling counterparts in AS, also revealed that lipids and biogenic amines had higher expression levels in AM flies compared to AS flies. However, there was no significant difference between AM and AS flies in abundance of primary metabolites (Figure 1—figure supplement 4A, E and I). Likewise, the overall abundance, regardless of cycling, of lipids and biogenic amines was higher in AM flies but primary metabolites were equivalent in AS and AM (Figure 1—figure supplement 5A–C).

Overall, this indicates that the microbiome increases the cycling of all metabolites in the gut but has more dramatic effects on lipids and biogenic amines.

Timed feeding affects host metabolism and benefits host health, according to several studies (Chaix et al., 2019; Gill et al., 2015; Lundell et al., 2020). In this study, we aimed to determine how microbiome-mediated metabolite cycling interacts with a timed feeding paradigm in which flies are provided with food only from ZT0 to ZT10. Similarly to ad libitum conditions, PCA of the total expressed primary metabolites showed only minor separation between the four groups— AM, AS, microbiome-containing flies subjected to timed feeding (TM), and germ-free flies subjected to timed feeding (TS)—but a clear separation was observed for lipids and biogenic amines just as seen for AM and AS flies (Figure 1—figure supplement 1A–C). Consistent with the results of our previous study on transcript cycling, we found that timed feeding significantly increased the number of metabolites cycling in germ-free and microbiome-containing flies (Figure 1B–M, Figure 1—figure supplements 2B,C,E,F, H, I and 6A–D).

Time-restricted feeding coordinated the phase of all three-metabolite classes with TM and TS flies displaying very similar phase distributions. Interestingly, primary metabolites peaked around a single time of day whilst lipids and biogenic amines displayed two distinct phases in timed feeding conditions (Figure 1B–M). Moreover, the observation that TM and TS flies display similar metabolite phases differs from what happens in ad lib conditions in which the phase of biogenic amines and lipids changes in the absence of a microbiome. In particular, metabolites peaking at ZT9 increased with TF in all metabolite groups. This could suggest an anticipation of sleep at the end of the feeding window. Interestingly, a smaller proportion of cyclers is specific for the ad libitum condition compared to the TF in both sterile and microbiome-containing flies (Figure 1—figure supplement 2).

Analysis of abundance of cycling metabolites under ad lib feeding and timed feeding conditions revealed that, in general, rhythmic metabolites tend to have higher overall expression levels compared to non-rhythmic metabolites, and this phenomenon is more prominent in the case of cycling biogenic amines (Figure 1—figure supplement 4A–L). Additionally, we also looked at the amplitude of metabolites that cycle in both microbiome-containing and microbiome-free flies under the two feeding paradigms. In this case, the three metabolite groups did not always show a larger amplitude in the timed fed groups, which slightly contradicts the potential benefits of this type of intervention (Figure 1—figure supplement 3D–I).

To further determine whether timed feeding modulates the effect of the microbiome, we compared the number of metabolites cycling with TF in AS and AM flies and found that loss of the microbiome had a larger effect in ad lib fed flies. This was reflected in the ratios depicting metabolite cycling in microbiome-containing versus germ-free flies, such that the AM/AS ratio was higher than the TM/TS for lipids and biogenic amines, although not for primary metabolites (Figure 1N–P). Moreover, the TS/AS ratio for lipids and biogenic amines was also higher than TM/AM (Figure 1—figure supplement 1D–F). These data suggest that loss of the microbiome has less of an effect on timed fed flies than on ad lib fed, and, conversely, timed feeding has less effect on microbiome-containing flies. Thus, time-restricted feeding improves cycling and may compensate for the lack of a microbiome.

To determine whether circadian clocks play a role in the enhanced metabolite cycling in microbiome-containing flies, we next subjected per¹ flies with (per¹-AM) and without (per¹-AS) a microbiome to ad lib feeding conditions and examined the cycling of metabolites in their guts (Figure 2A). The results showed that the PCA for the three metabolite classes differentiated per¹-AM and per¹-AS flies (Figure 2—figure supplement 1A–C). Surprisingly, loss of the microbiome increased metabolite cycling in per¹ flies, not for primary metabolites, but increasing the proportion of lipids cycling from 17.8% to 22.2% and almost duplicating the proportion of biogenic amines cycling from 18% in per¹-AM to 30.5% cycling features in per¹-AS flies (Figure 2B, D and F). Moreover, all cycling metabolites exhibited completely opposite phases in per¹-AM relative to per¹-AS flies (Figure 2C, E and G). Phases also changed for cycling metabolites shared between AM and AS (Figure 2—figure supplement 2A–C). Along these lines, we found an even lower overlap of cycling metabolites between AM and AS in per¹ flies (Figure 1—figure supplement 2), indicating that the clock modulates not just the phase but also the identity of cyclers.

Analysis of the abundance of cycling metabolites in per¹ flies revealed that, in general, cycling lipids and biogenic amines have higher expression in AM compared with their non-cycling equivalents in AS, whereas primary metabolites show minor difference between the two groups (Figure 2—figure supplement 3A–C). This result is consistent with the comparison of abundance in microbiome-containing and germ-free iso31 flies.

These datasets suggest a complex interaction between the microbiome and the host clock. It was noticeable that lack of a host clock resulted in reduced metabolite rhythms in microbiome-containing flies (Figure 1—figure supplements 2A, D, G and 6A, E) but enhanced cycling in microbiome-free flies (Figure 1—figure supplements 2J, K, L and 6B, F). This also caused profound changes in phase (Figure 2). Overall, it appears that the host clock stabilizes metabolite rhythms against changes in the microbiome.

Changes in diet composition can significantly affect gut microbiome rhythms (Flint, 2012; Frazier et al., 2022). Therefore, we also explored the cyclic metabolome of flies with or without a microbiome and fed on two different diets; a high sugar and a high protein diet (Figures 3A and 4A). From the PCA , we noticed that high sugar diets trigger metabolic differences between flies that contain a microbiome and microbiome-free flies; however, high protein diets do not prompt such difference (Figure 2—figure supplement 1 D–F and G–I).

On high protein diets, the lack of a microbiome reduced the number of rhythmic metabolites (Figure 3B, D and F). This was consistent with the role of the microbiome in promoting metabolic rhythms. However, the effects of a high sugar diet diverged substantially from those of a high protein diet. With the high sugar diet, rhythmic lipids and biogenic amines increased in germ-free flies, similar to what we observed in a clock mutant background (Figure 4B, D and F). Indeed, when the three metabolite groups were analyzed together, flies on a high sugar diet showed enhanced cycling of the metabolome when the microbiome was absent (Figure 1—figure supplement 6I and J).

We also looked at phases of cycling metabolites. In flies fed a high protein diet, a predominant peak around ZT15 was observed for all metabolites regardless of the presence or absence of a microbiome (Figure 3C, E and G). This suggests that protein metabolism imposes a specific metabolic rhythm that is not dependent on the gut microbiome. In case of the high sugar diet, we noticed that the absence of microbes changed the phase distribution of both lipid and biogenic amines but not of primary metabolites (Figure 4C, E and G).

We also compared the dataset generated from flies fed a standard diet versus these nutrient-specific paradigms. For flies with a normal microbiome, the fraction of cycling metabolites decreased from 33.5% on standard diet to 22.2% and 24.5% with a high protein and high sugar diet, respectively (Figure 1—figure supplement 6A, G and I). Thus, overall metabolite cycling decreases on high protein and high sugar diets. But while high protein potentiates the effect of the microbiome in promoting cycling, high sugar suppresses it (Figure 1—figure supplements 2A, D, G, M–R and 6G–J).

Overall, the results suggest that nutrient-specific diets interact in different ways with the microbiome to regulate the cycling of gut metabolites.

We further explored how the phase of cycling is affected by the endogenous clock, diet, and microbiome. The phase of cycling in per¹ flies was different from that of wild-type flies. The average phase of primary metabolites in wild-type flies falls around ZT6 while per¹ flies mainly have a peak phase near ZT3. Lipids lose a morning peak in the clock mutant and most dramatically, biogenic amines go from peaking in the middle of the day to the middle of the night (Figure 5A and B, left panels). This effect of the host clock on the phase of cycling metabolites was even stronger in microbiome-free flies (Figure 5A and B, right panels). Phases were also different on the different diets in microbiome-containing flies, with just primary metabolites conserving a peak around ZT6 in both the standard and high sugar diet; in high protein, these metabolites showed a peak around ZT15. Lipids exhibit a cycling peak primarily around ZT0 and ZT6 in standard diets. However, the phase shifts to around ZT6 and ZT15 for high protein diets, and to around ZT3 and ZT15 for high sugar diets. Biogenic amines have a cycling phase that falls around ZT15 in both high protein and high sugar diets, while in standard diets, the peak occurs around ZT6 (Figure 5A, C and D left panels). Germ-free flies also show significant effects of diet composition on cycling phase (Figure 5).

To determine whether metabolite cycling is affected by altered feeding, we investigated food intake at different times of day in different genotypes and under different conditions and diets. Given that much has already been published on timed feeding, and the fact that it does not significantly affect food consumption (Gill et al., 2015; Villanueva et al., 2019), we chose to investigate feeding rhythms of AM and AS iso31 and per¹ flies on different diets. The results showed that, compared to normal chow diets, high protein diets decreased food consumption, while high sugar diets increased food consumption (Figure 4—figure supplement 1A). Flies fed high protein diets displayed two feeding peaks, one in the morning and one in the evening, as we previously demonstrated with standard diets. The feeding rhythm showed no significant difference between AM and AS flies, in both iso31 and per¹ strains (Figure 4—figure supplement 1B). However, high sugar diets reduce the amplitude of feeding rhythms and eliminate the evening feeding peak. Indeed, the feeding rhythm nearly disappears in AS flies, regardless of genotype, on a high sugar diet (Figure 4—figure supplement 1C).

The extensive metabolic profiling performed here indicates that the gut microbiome is a central piece in generating daily cycles of metabolites and that depending on the host circadian clock and the composition of the diet, the extent and directionality of that modulation varies.

To get a further understanding of the biological implications of this cycling metabolic datasets, we performed pathway enrichment analysis. First, we explored the ad libitum dataset in which we found that absence of a microbiome reduced metabolic rhythms (Figure 1). To our surprise, despite the reduction in cycling metabolite number in flies lacking a microbiome, the pathway enrichment identified numerous pathways as significantly enriched in the AS-only cycling group. These included biosynthesis of branched (valine, leucine, and isoleucine) and aromatic (phenylalanine, tyrosine, and tryptophan) amino acids as well as the aminoacyl-tRNA biosynthesis pathway that is relevant for protein synthesis (Figure 6A). Another amino acid pathway, involving alanine, aspartate, and glutamate, was found significant in AS and AM cycling sets, with different specific metabolites cycling in these datasets. No pathway related to amino acids or protein metabolism was specific to the AM cycling metabolites, suggesting that, in general, the microbiome abrogates daily oscillations of protein anabolism.

We found that similar pathways were significant in the cycling dataset from microbiome-containing flies subjected to a daytime-restricted feeding paradigm versus ad libitum feeding (Figure 6B). Additionally, we determined whether pathways affected by timed feeding were different in flies with and without a microbiome and found that amino acid pathways cycle significantly in both groups (Figure 6D). Overall, the data suggest that amino acid and protein biosynthesis cycles are dampened by the gut microbiome but potentiated by time-restricted feeding and the latter has a larger effect in microbiome-containing flies, indicating a microbiome and TF interaction. An interesting point here is how timed feeding influences the same amino acid pathways in microbiome-containing and germ-free flies (e.g., valine, leucine, and isoleucine biosynthesis) but sometimes through distinct cycling metabolites. For instance, threonine, leucine, isoleucine, 4-methyl-2-oxovaleric acid, and valine in TM (Figure 6—figure supplement 1 and Supplementary file 1). In TS, cycling metabolites in this pathway are isoleucine and valine. Likewise, specific metabolites of a pathway showed differential regulation in AS and AM flies. In AS on a high sugar diet, L-threonine and leucine cycle. Another amino acid, 4-aminobutyric acid, cycles in both AM and AS flies (Figure 6—figure supplement 1 and Supplementary file 1). An acetylated derivative of a natural essential amino acid, N-acetyl-l-leucine, uniquely cycles in AS flies but not in AM flies. On the other hand, the TCA cycle appeared as a shared cycling category in all comparisons (AM vs AS, Figure 6A; AM vs TM, Figure 6B, AM vs per¹ AM, Figure 6E), suggesting consistency in the rhythmicity of this pathway. Lastly, we aimed to identify pathways whose cycles were exclusive to microbiome-containing flies. We found that alpha-linoleic metabolism cycled significantly in AM flies compared to AS flies (Figure 6A). This pathway cycled significantly in AM and TM flies (Figure 6B), indicating that its cycling is not dramatically influenced by timed feeding and may be largely dependent on the microbiome. However, in the absence of a microbiome, timed feeding seems to drive cycling of this pathway (see comparison of AS and TS in Figure 6C). In addition, cycling of this pathway requires a functional circadian clock as it is lost in per¹-AM flies. Thus, cycling of alpha-linoleic metabolism depends upon the microbiome and the clock and can also be driven by timed feeding.

Consistent with the results from flies fed a normal chow diet, we observed that germ-free flies fed a high sugar diet showed an enrichment of amino acid-related metabolism in the absence of a microbiome (Figure 6F), which indicates that sugar content did not alter the functional category of cycling metabolites by KEGG analysis. Pathway analysis of cycling metabolites under high protein conditions was not durable due to the reduced number of features.