Circadian GLP-1 Secretion in Mice Is Dependent on the Intestinal Microbiome for Maintenance of Diurnal Metabolic Homeostasis

The incretin hormone glucagon-like peptide 1 (GLP-1) lowers glycemia through stimulation of glucose-dependent insulin secretion (1). As a consequence, GLP-1 receptor agonists and inhibitors of endogenous GLP-1 degradation are now both widely used for the treatment of patients with type 2 diabetes (2). GLP-1 is secreted by the intestinal L cell after nutrient ingestion. However, GLP-1 secretion also follows a circadian rhythm, demonstrating greater release in response to identical glucose loads at the onset of the dark/feeding period compared with the light/fasting period in rodents as well as time-of-day–dependent secretory patterns in humans (3–6).

Circadian rhythms are endogenous biological rhythms that follow a 24-h period (7). These rhythms are determined by the molecular clock, which consists of two basic helix-loop- helix transcription factors, BMAL1 (aryl hydrocarbon receptor nuclear translocator-like protein 1 [ARNTL] or brain and muscle ARNT-Like 1) and CLOCK, that bind to E-box promoter elements of the period (PER1/2/3) and cryptochrome (CRY1/2) genes. In turn, PER/CRY-dimers inhibit BMAL1/CLOCK expression, thereby forming an autoregulatory feedback loop. Although originally described in the hypothalamic suprachiasmatic nuclei, the molecular clock is also expressed in metabolic tissues that respond to the diurnal feeding-fasting cycle, including islets, hepatocytes, skeletal muscle, and adipocytes (8–10). Consistent with a physiological role for this peripheral clock, disruption of circadian rhythmicity, as occurs in shift workers, is a risk factor for metabolic disorders, including type 2 diabetes (11). However, circadian disruption is also induced by obesogenic feeding, resulting in both impaired insulin secretion and glucose intolerance (11,12). Importantly, while the β-cell expresses a cell-autonomous circadian clock, insulin secretion in rats follows a circadian pattern in response to oral but not intravenous glucose (13–15). These findings implicate the incretin hormones, glucose-dependent insulinotropic polypeptide (GIP) and/or GLP-1, in β-cell entrainment. Indeed, treatment with GLP-1 synchronizes the molecular clock in isolated β-cells, and greater insulin responses to exogenous GLP-1 are observed at the onset of the feeding compared with the fasting period in rats, suggesting a role for GLP-1 in the circadian secretion of insulin (4,16).

It remains unknown how the circadian release of GLP-1 is entrained in vivo and whether circadian GLP-1 secretion plays a critical role in time-dependent glucose homeostasis. Interestingly, although the intestinal microbiome is not directly exposed to light-dark cycles, oscillations in microbial composition and function have been reported to follow diurnal host feeding rhythms (17,18). Furthermore, intestinal dysbiosis is associated with metabolic defects, including insulin resistance, as well as alterations in both GLP-1 secretion and glucose metabolism (19,20). We therefore hypothesized that diurnal changes in the microbiome play a role in circadian GLP-1 secretion and its subsequent effects on glucose homeostasis.

Male and female animals were used in all experiments. C57Bl6/J mice (5–7 weeks of age; The Jackson Laboratory) were acclimatized to the animal facility for 1 week before experimentation. Proglucagon (Gcg)-Venus C57Bl/6J mice (a gift from Drs. F. Reimann and F.M. Gribble, University of Cambridge [21]) were bred at the University of Toronto and used at 5–7 weeks of age. For the duration of each experiment, mice were housed under a 12/12-h light/dark cycle at constant room temperature with free access to water and were fed regular chow (RC; 18%, 58%, and 24% calories from fat, carbohydrate and protein, respectively; overall caloric density of 3.1 kcal/g [2018 Envigo]) or a high-fat/high-sucrose Western diet (WD; 41%, 43%, and 17% calories from fat, carbohydrate [29% sucrose, 14% other], and protein, respectively; overall caloric density of 4.7 kcal/g [D120798, Research Diets]) for 16 weeks. Antibiotic-induced microbial depletion (AIMD) was induced by oral gavage (twice a day) of an antibiotic/antifungal cocktail (ampicillin, metronidazole, and neomycin [100 mg/kg each], vancomycin [50 mg/kg], and amphotericin B [1 mg/kg]) or water (vehicle) during the final 2 weeks of feeding (20). RC-germ-free C57Bl/6N mice were bred and maintained under standard germ-free conditions before transfer into a sterile 12/12-h light/dark room for 1 week, followed by oral glucose tolerance tests (OGTTs) at 7–9-weeks of age. Feces from RC animals were collected throughout the 24-h day and resuspended in PBS for fecal microbiome transplantation. Aliquots (200 μL; 5 g/mL) were administered to 8-week-old germ-free mice by oral gavage, daily for 2 days, followed by a 3-week rest period and OGTTs. All animal studies were approved by the University of Toronto Animal Care Committee (Toronto, Ontario, Canada) and conformed to Canadian Council on Animal Care guidelines.

Mice were weighed weekly, peripheral fat was assessed by DEXA scan, and visceral fat was weighed after sacrifice. Energy intake was measured every 4 h for 24 h on a separate cohort of animals.

Animals that had been fasted for 4 h were administered OGTTs (5 g glucose/kg body wt [22,23]), intraperitoneal glucose tolerance tests (IPGTTs; 2 g glucose/kg), insulin tolerance tests (ITTs; 0.25 units insulin/kg i.p.), and/or intraperitoneal GLP-1 and GIP injections (70 pmol/kg in combination with an IPGTT [24]). Experiments were conducted at zeitgeber (ZT) 2, 6, 10, 14, 18, and 22 (ZT0 being 0600 h), with dark-period studies performed under a red light. Animals underwent two metabolic tests (only), conducted at two different times (i.e., ZT2 and ZT14), with a 1-week rest period between tests. At t = 0, 10, and 60 min post gavage/injection, tail vein glycemia was measured using a LifeScan glucometer, and 100 μL of blood was collected (i.e., a total of 300 μL/mouse) for analysis of plasma GLP-1, insulin, and glucagon by Meso Scale Discovery Assay, and plasma GIP by Millipore ELISA (22,23).

As previously described (22), distal small intestinal crypts from 16-week RC- and WD-fed mice were plated onto Matrigel for 24 h and then incubated with vehicle or 50 μmol/L forskolin/3-isobutyl-1-methylxanthine (IBMX) for 2 h. Media and cell lysates were collected to assess active GLP-1 concentrations by ELISA (Millipore). Secretion was calculated as the percentage GLP-1 in the medium normalized to the cell + medium content, averaged for four wells per mouse to make n = 1.

Ileal and colonic sections were stained using an anti–GLP-1 primary (Abcam) and Alexa Fluor 555–coupled secondary antibody for L-cell counting (22). Ileal mucosal RNA was extracted using the RNeasy Plus Mini Kit with QIAshredder (Qiagen), reverse-transcribed with 5× All-In-One RT Mastermix (Applied Biological Materials), and analyzed using TaqMan Fast Mix Gene Expression Assay with primers for proglucagon (Mm00801712_m1) and histone 3a (Mm01612808_g1), as previously described (22,23).

Colonic fecal DNA was isolated using the DNeasy PowerSoil Kit, and 16S rRNA gene sequencing was performed by the Centre for the Analysis of Genome Evolution & Function (University of Toronto, Toronto, Ontario, Canada). Cecal feces were processed by the Analytical Facility for Bioactive Molecules at the Hospital for Sick Children (Toronto, Ontario, Canada), and short-chain fatty acids (SCFAs) and bile acids (BAs) were analyzed by gas chromatography–mass spectrometry and liquid chromatography–mass spectrometry (Biocrates), respectively.

Single cells from the distal small intestine of RC and WD mice were sorted using a BD FACSMelody cell sorter (BD Bioscience) at the University of Toronto Flow Cytometry Facility with DAPI staining and side- and forward scatter to remove dead and clustered cells. Approximately 10,000 Venus-expressing cells were purified into LoBind tubes (Eppendorf) with Buffer RLT (Qiagen), as previously described (25,26). Total RNA was extracted using the RNeasyPlus Micro Kit (Qiagen) with genomic DNA removal. Library preparation and sequencing was performed by the Donnelly Sequencing Centre (Toronto, Ontario, Canada).

All data are expressed as mean ± SD. CircWave (www.hutlab.nl) was used to determine 24-h rhythms. All other data were analyzed by ANOVA (one-, two-, or three-way), followed by the Tukey test for three or more groups or by the Student t test for two groups.

Supplementary Tables 1–8 are available are available at the following link: https://hdl.handle.net/1807/101930.

C57Bl/6J WD-fed mice demonstrated obesity and disruption of feeding patterns compared with RC-fed animals (Supplementary Fig. 1A–D). Fasting glycemia, plasma insulin, glucagon, GIP, and GLP- 1, as well as insulin tolerance, displayed fluctuations throughout the day in all mice, with obese animals being more insulin resistant (Supplementary Fig. 1E–K). Therefore, for direct comparison of glycemic and hormonal responses to challenge, all data were analyzed as the change from basal.

In response to both oral and intraperitoneal glucose, RC-fed mice displayed a significant 24-h rhythm in glycemic tolerance, while insulin demonstrated a rhythm only after oral administration (Fig. 1), implicating the incretin hormones in entrainment of diurnal insulin release. WD-fed mice displayed elevated fasting glycemia throughout the 24-h day and were significantly hyperglycemic in response to an IPGTT (Fig. 1A–C and Supplementary Fig. 1E). However, after an OGTT, WD-fed animals maintained normoglycemia, accounted for by a significant elevation in insulin secretion (Fig. 1D–F), suggesting that the incretin hormones are also important for temporal metabolic homeostasis in WD-induced obesity.

GIP and GLP-1 secretion displayed parallel 24-h rhythms in RC mice after an OGTT, with the peaks occurring at the onset of the dark/feeding period, ZT14, and the troughs at the beginning of the light/fasting period, ZT2 (Fig. 2A–C). Similar analyses in WD animals revealed a complete loss of rhythmicity in both hormones. However, whereas modest elevations in GIP were also observed, plasma GLP-1 levels were massively increased throughout the 24-h day, suggesting its more dominant role in driving the elevated insulin secretion under WD conditions (Fig. 2A–C).

To determine the role of the incretins in rhythmic insulin secretion, RC- and WD-fed mice were administered GIP or GLP-1 at their trough and peak secretory time points. In RC mice, both incretins stimulated greater release of insulin at ZT2, thereby lowering glycemia despite greater insulin resistance (Fig. 2D–G and Supplementary Fig. 1K). However, administration of GLP-1, but not GIP, in WD mice increased insulin and maintained normoglycemia at both ZT2 and ZT14, indicating that while both incretins are important for entraining rhythmic insulin secretion in normal mice, the increase in GLP-1 in response to oral glucose is essential for the maintenance of 24-h glucose homeostasis in WD animals.

To examine molecular mechanisms underlying the time- and diet-dependent changes in L-cell secretory capacity, the L-cell transcriptome from RC- and WD-fed Gcg-Venus mice was determined. Importantly, these animals exhibited the same temporal GLP-1 secretory pattern as normal RC and WD mice (Supplementary Fig. 2). Analysis of core clock and clock-controlled gene expression in RC animals established the existence of a primary L-cell clock (Fig. 3A). Bmal1 (Arntl) was significantly increased at ZT2 in L cells from RC mice, whereas Per1/2/3 and Rev-erbα/β (Nr1d1/2) were antiphasic to Arntl, increasing at ZT14 (Fig. 3A). The clock-controlled genes, Dbp and Tef, also exhibited significantly higher expression at ZT14, indicative of a functional L-cell clock (Fig. 3A). Pathway analysis of high-confidence data (i.e., communities with at least three nodes) also revealed profound differences in the L-cell transcriptome at the peak and trough of GLP-1 secretion. The L cell appeared to be more quiescent at the trough of GLP-1 secretion, ZT2, and in contrast, exhibited marked enrichment in genes for nutrient sensing (i.e., carbohydrates, chemical stimuli, and olfaction) at ZT14, the peak of GLP-1 release (Fig. 3B and Supplementary Table 1) Furthermore, SNARE complex gene sets were determined to be significantly enriched to ZT14 (Fig. 3C).

In contrast to the RC transcriptome, WD L cells exhibited no significant differences in clock gene expression between ZT2 and ZT14, indicative of WD-induced clock disruption (Fig. 3A). While WD L cells retained many of the inherent time-of-day variances observed in RC L cells, there were profound differences between the transcriptomes. Unexpectedly, and despite the increased GLP-1 secretion in WD mice, mitochondrial function, which is known to be essential for GLP-1 secretion, was reduced in WD L cells at ZT2 and ZT14. Conversely, WD L cells demonstrated enrichment in genes for fatty acid metabolism, consistent with the high-fat feeding (Fig. 3D and E). Pathways involved in immune responses and synaptic transmission were also upregulated at both time points, the latter of which included genes encoding core SNARE and SNARE accessory proteins (Fig. 3D–G and Supplementary Tables 2 and 3).

In marked contrast to the findings in vivo, ileal and colonic intestinal crypt cultures from WD animals demonstrated normal GLP-1 secretory responses, with no differences in cellular GLP-1 content (Fig. 4A and B). These changes were independent of crypt and villus L-cell numbers, and indeed, WD feeding reduced the L-cell count in ileal crypts and villi (Fig. 4C). Collectively, these findings suggested that circadian GLP-1 secretion and the altered responses in WD-fed animals occur consequent to some factor present only in vivo, such as the intestinal microbiome.

16S rRNA gene sequencing of colonic feces at ZT2 and ZT14 revealed increased Verrucomicrobia at the peak compared with the trough of GLP-1 secretion in RC mice, and this effect of time was lost in association with increased relative abundance in WD mice, despite their overall reduction in diversity (Fig. 5A and Supplementary Fig. 3A and B). These changes were also apparent at the species level, because the major contributor to the Verrucomicrobia phyla, Akkermansia muciniphila, a species that improves glucose tolerance and increases GLP-1 release (27,28), paralleled the time- and diet-dependent patterns in GLP-1 secretion (Fig. 5B and Supplementary Fig. 3C and D).

RC-fed animals exhibited decreased cecal SCFAs at the peak of GLP-1 secretion, while WD-fed mice lacked rhythmicity and had reduced levels (Fig. 5C–E). Conversely, cecal BA levels demonstrated increases in several known L-cell stimulators, including the secondary BA, deoxycholic acid, at the peak of GLP-1 secretion in RC animals, and upregulation in all measured BAs independent of time in WD mice (Fig. 5F).

In silico analysis of the 16S rRNA gene sequences demonstrated that the composition of the microbial community changed between the peak and trough of GLP-1 secretion. Moreover, analysis of the functional genes present in the community identified both time- and diet-dependent changes in metabolic pathways. Analysis of gene function, as represented by Kyoto Encyclopedia of Genes and Genomes (KEGG) orthology (KO), also identified time-dependent changes in the RC mice only, with all KOs decreased at the peak time point of GLP-1 secretion only, including functions related to transport systems and nucleotide metabolism (i.e., adenosine deaminase [29]) (Fig. 5G and Supplementary Table 4). In contrast, WD feeding induced both increases and decreases in KOs compared with RC feeding at both ZT2 and ZT14 (Fig. 5H and I and Supplementary Tables 5 and 6). Metabolic pathway analysis also revealed profound enrichment of biotin biosynthesis in the WD microbiome at both ZT2 and ZT14 (Supplementary Tables 7 and 8). When taken together, the changes in intestinal microbiome composition and function with both time and diet were consistent with the patterns observed in GLP-1 secretion.

Antibiotic-induced microbial dysbiosis in RC animals reduced diversity, depleted all major phyla, and ablated the normal temporal 16S rRNA gene expression patterns, including that for Verrucomicrobia (Fig. 6A). This was accompanied by a massive surge in the relative abundance of Proteobacteria, including the pathogenic Ralstonia (Fig. 6A and Supplementary Fig. 4C). In contrast, the microbiome of WD AIMD animals was colonized almost entirely by Firmicutes and, in particular, the largely beneficial Lactobacillus (Fig. 6A and Supplementary Fig. 4C). Microbial depletion also decreased all BAs in RC mice, at both ZT2 and ZT14, while WD animals demonstrated a decrease only in primary and secondary BAs, maintaining the diet-induced increase in conjugated primary and secondary BAs (Fig. 6B). AIMD also profoundly increased fasting GLP-1 in both the RC and WD animals, at both ZT2 and ZT14 (Fig. 6C). Notwithstanding, fasting levels of insulin and glucose were unaltered in RC AIMD animals (Fig. 6D and E), consistent with the known glucose dependence of the incretin hormones for facilitation of insulin secretion (1). In contrast, fasting WD and AMID-WD animals were hyperglycemic at both ZT2 and ZT14 in association with insufficient increases in insulin levels (Fig. 6D and E).

OGTTs revealed that the microbial dysbiosis in the AIMD mice completely abolished not only the time-dependent GLP-1 secretory pattern in RC animals but also the WD-induced, time-independent elevation in GLP-1 responses (Fig. 6F–I). This elevation in absolute GLP-1 levels in RC AIMD animals (i.e., basal plus stimulated secretion) was associated with improved glucose tolerance at both ZT2 and ZT14, despite normal insulin levels. However, this did not translate to the WD-AIMD mice, because their absolute GLP-1 levels were markedly lower than those of their WD comparators. WD-AIMD mice also appeared to have elevated but insufficient insulin secretion, resulting in an inability to normalize glucose tolerance at both time points. This may be a result of the known effects of the microbiome to modulate both insulin secretion and insulin sensitivity independent of gut-derived factors (17). Thus, in the absence of an intact microbiome, WD animals were unable to maintain diurnal normoglycemia in association with impaired circadian GLP-1 secretion.

Finally, to confirm the centrality of the microbiome for circadian GLP-1 release, OGTTs were conducted throughout the 24-h day in germ-free mice. As in the RC-AMID animals, fasting GLP-1 levels were remarkably elevated compared with normal mice at each time point, and this was paralleled by increased insulin levels and normoglycemia (Fig. 7A and Supplementary Fig. 5A and B vs. Fig. 1B and C and Fig. 2C). Furthermore, while AMID mice maintained L-cell capacity to respond to an OGTT, germ-free animals lacked a GLP-1 circadian rhythm and failed to response positively to the oral glucose load, demonstrating that complete absence of a microbiome impairs the L-cell secretory response (Fig. 7B and C). Consistent with these findings, germ-free mice also lacked 24-h rhythms in both insulin secretion and glycemic tolerance (Supplementary Fig. 5C–E). However, after RC fecal microbiome transplantation, germ-free animals not only regained time-dependent microbial changes, including the increase in Verrucomicrobia and A. muciniphila at ZT14 (Fig. 7D and E), but their fasting levels of GLP-1 were also restored to normal, and the ZT2 versus ZT14 GLP-1 secretory pattern was reestablished (Fig. 7F–H). This was accompanied by parallel changes in insulin and glucose responses (Supplementary Fig. 6). Interestingly, a significant positive correlation (r² = 0.48, P < 0.01) was found between A. muciniphila and the GLP-1 response in RC-recolonized germ-free animals, further supporting the role of A. muciniphila as a positive regulator of GLP-1 secretion.

The hormones GIP and GLP-1 are responsible for ∼50% of nutrient-induced insulin secretion (1,2). Epidemiologic data and findings from clock-gene knockout mice have both indicated an essential role for circadian secretion of insulin in the maintenance of diurnal euglycemia, implicating one or both of the incretins in this patterning (11,14). The results of the current study demonstrate that parallel 24-h rhythms in both GLP-1 and GIP are essential for the temporal regulation of insulin secretion and glucose homeostasis in normal mice, whereas circadian secretion of GLP-1 plays a more important role under conditions of WD feeding. Further, the intestinal microbiome was established to be an integral component of the pathway regulating diurnal GLP-1 release, adding a previously unsuspected level of complexity to the metabolic clock.

As reported for both rats and humans (3–6), GLP-1 and GIP demonstrated time-of-day– dependent secretion in normal mice, whereby peak and trough levels were observed at the onset of the dark/active period (ZT14) and light/inactive period (ZT2), respectively, likely as anticipatory responses to the diurnal changes in food intake. However, although a diurnal pattern in insulin release was observed only in response to oral glucose administration, thereby implicating the incretin hormones in β-cell entrainment, the insulin secretory rhythm did not parallel the pattern in incretin release, with peak insulin levels being observed during the light period. These findings stand in contrast to those in rats, wherein the diurnal patterns in GLP-1 and insulin are parallel (4). Nonetheless, both basal and GLP-1-stimulated insulin secretion were greater in mice at ZT2, consistent with the increased demand posed by greater insulin resistance observed during the rest phase. Expression of the GLP-1 receptor in the β-cell has been reported to be arrhythmic (14). However, several key downstream mediators of β-cell GLP-1 signaling do exhibit diurnal expression, including, most notably, key exocytotic SNARE proteins (14,15), thereby providing a possible mechanism by which insulin secretion is more sensitive to stimulation by GLP-1 at certain times of day.

As in RC animals, WD mice also exhibited diurnal glucose homeostasis in response to oral, but not intraperitoneal, glucose administration. However, normoglycemia in WD animals was associated with arrhythmic elevations in insulinemia as well as with massive increases in GLP-1 and more modest changes in GIP secretion that were accompanied by loss of their normal secretory patterns. Furthermore, normoglycemia in the WD-fed mice was maintained after GLP-1, but not GIP, administration. These findings are consistent with a recent study showing that increased GLP-1 in WD-fed rats drives insulin secretion, albeit at a single time point during the day (30), and extend them to demonstrate that although both incretin hormones entrain rhythmic insulin secretion under normal conditions, GLP-1 appears to play a more important role in maintaining diurnal glucose homeostasis under conditions of WD feeding.

The cellular mechanism(s) underlying circadian rhythms in GLP-1 secretion have, to date, been largely limited to studies using the murine (m) GLUTag L-cell line (4,23,31). The results of the current study have now demonstrated the existence of a clock in the primary L cell, with gene expression patterns in L cells from normal mice being consistent with those of other metabolic tissues, including higher expression of Arntl at ZT2 than at ZT14 (32). However, unexpectedly, the clock gene expression patterns differed between mGLUTag and primary L cells, in that GLP-1 secretion parallels the positive arm of the (i.e., Arntl) in the mGLUTag L cells, whereas GLP-1 release peaked in correlation with increased expression of the negative arm (i.e., Per1/2/3, Rev-erbα/β) in the primary L cells. However, the parallel between Rev-erbα/β expression and the GLP-1 secretory pattern is consistent with the established role of these transcription factors as links between circadian rhythms and metabolism in other tissues (33). Notably, in association with loss of the normal pattern in GLP-1 secretion, these relationships were disrupted in L cells from WD mice, which demonstrated the known effects of obesogenic feeding to dampen rhythmic clock gene expression in metabolic tissues (15).

Specific components of the transcriptome in primary L cells from RC mice were also found to differ between the peak and trough time points of GLP-1 secretion. Most notably, transcripts related to sensory pathways were upregulated at ZT14, consistent with a greater capacity of the L cell to respond to the normal initiation of nutrient ingestion at this time. Furthermore, the enrichment of SNARE complex gene to the peak of GLP-1 nicely paralleled current literature establishing the essential roles of several SNARE proteins in diurnal GLP-1 secretion (22,23,31,34). Conversely, WD feeding increased the expression of transcripts related to exocytosis at both time points, coinciding with the profound increases in GLP-1 release observed in these animals. Similar observations have been made in human β- and α-cells, wherein decreased clock gene expression is associated with reduced insulin and glucagon release, respectively, in association with disrupted secretory granule docking and exocytotic pathways (16).

The enrichment of mitochondrial function in RC primary L cells was somewhat unexpected, given both the known requirement for ATP in GLP-1 release (35) and the finding that palmitate, a major component of the WD diet, suppresses ATP production and GLP-1 secretion in mGLUTag L cells (36,37). Furthermore, ileal and colonic intestinal crypt cultures from WD animals did not demonstrate enhanced GLP-1 secretory responses as had been anticipated based on the in vivo findings. Notwithstanding, the findings of both compositional and functional changes in the microbiome and in A. muciniphila, in particular, strongly implicated the microbiome as a key factor regulating the parallel temporal- and diet-dependent changes that were observed in GLP-1 secretion, as was further supported by the restoration of these patterns after RC-fecal microbiome transfer into germ-free mice. The intestinal microbiome has been established to regulate intestinal epithelial gene expression in a circadian manner, and importantly, is thought to control, at least in part, expression of Rev-erbα (33,38,39). Consistent with these findings, we have shown that Rev-erbα expression in primary L cells parallels the GLP-1 secretion patterns, which may therefore be, at least potentially, driven by time- and diet-dependent differences in the microbiome. Furthermore, increases in Akkermansia are also known to be associated with enhanced secretion of GLP-1 as well as with improved metabolic control in both mice and humans (27,28).

Use of both dysbiotic and germ-free mice demonstrated the necessity of an intact microbiome for regulation of circadian GLP-1 release. Fasting GLP-1 levels in RC AIMD mice significantly exceeded those of the WD AMID animals, possibly as a consequence of a proinflammatory intestinal environment induced by Proteobacteria, leading to the release of inflammatory cytokines known to stimulate GLP-1 secretion (40). In contrast, WD AIMD mice had increased Lactobacillus, which are known to synthesize GLP-1 secretagogues and may, therefore, contribute to their more physiological, yet elevated, basal levels of GLP-1 (41). Interestingly, elevated GLP-1 secretion in germ-free animals has also been associated with SNARE regulatory pathways (19). Finally, further supporting the notion that GLP-1 is essential for entrainment of insulin release, lack of the GLP-1 rhythm in germ-free mice was associated with loss of the normal 24-h patterns in both insulin and glycemia. Additionally, although no significant changes in food intake patterns were noted in the WD mice, given the known role of food intake in the regulation of GLP-1 secretion, a limitation of the current study is the lack of determination of this parameter in the microbiome-depletion and germ-free mouse studies. Nonetheless, the combination of the loss-of-function approaches (i.e., WD, AIMD, germ-free mice) with the gain-of-function study (i.e., RC-fecal microbiome transfer into germ-free mice) strongly implicates the microbiome as an essential determinant of both basal and diurnal GLP-1 secretion.

How the microbiome modulates GLP-1 release remains to be established. The observed changes in secondary BAs, which paralleled the patterns in GLP-1 secretion, are consistent with the alterations in the microbial community over time as well as with the known increase in BAs as a result of obesogenic feeding (42). Furthermore, BAs are known to stimulate GLP-1 release by the L cell through activation of the G protein–coupled BA receptor TGR5 and/or the nuclear BA receptor FXR (42). In contrast, despite evidence supporting the notion of SCFAs as stimulators of GLP-1 release through GPR41 and/or GPR43 (43), the temporal and diet-dependent rhythms observed for SCFAs in the current study were antiphasic to GLP-1 secretion. However, cecal SCFA concentrations are determined not only by their initial production but also by the efficiency of their absorption from the intestine (44). As such, it is possible that the decrease in measured cecal SCFAs is a result of increased SCFA absorption at the time of peak GLP-1 secretion and under WD conditions, thereby enhancing SCFA signaling through L-cell GPR41 and/or GPR43. Finally, it remains possible that functional changes in the microbiome, such as in adenosine deaminase activity, may be of relevance to the L-cell response to an OGTT, because extracellular adenosine levels and adenosine receptor signaling can affect cellular metabolism through enhanced glucose utilization (29). Further studies will be required to determine which, if any, of these different pathways are involved in the regulation of circadian GLP-1 secretion by the microbiome.

Collectively, the findings of the current study position the incretin hormone GLP-1 as a key mediator of diurnal metabolic homeostasis. Although a normal intestinal microbiome is essential to establishing rhythmic GLP-1 release, pursuing a better understanding of the mechanism(s) by which the microbial community or specific commensal organisms entrain the intestinal L cell may provide novel therapeutic approaches for the treatment of metabolic disease, particularly in the setting of circadian disruption.

Acknowledgments. The authors are extremely grateful to Drs. F. Reimann and F. Gribble (University of Cambridge) for the generous gift of Gcg-Venus mice, Dr. P. Larraufie (ABIES Doctoral School) for assistance with the FACS and RNA extraction methods, M. Clemenzi (University of Toronto) for assistance with the IPGTTs and ITTs, A. Biancolin (University of Toronto) for assistance with plotting using the R package limma, A. Cao and L. Kent (University of Toronto) for maintenance of the germ-free facility, D. White for assistance with FACS (Flow Cytometry Facility, University of Toronto), and the Centre for the Analysis of Genome Evolution & Function and Analytical Facility for Bioactive Molecules (Hospital for Sick Children, Toronto, Ontario, Canada) for technical assistance.

Funding. S.E.M. was supported by graduate awards from the Ontario Graduate Scholarship program and the Banting and Best Diabetes Centre (University of Toronto), A.M. by graduate awards from the Canadian Institutes of Health Research, the Ontario Graduate Scholarship, the Banting and Best Diabetes Centre, and the Biological Rhythms Training Program (University of Toronto), B.J.C. by a Tier II Canada Research Chair, K.N. by the John P. Mitchell Cancer Award from the Undergraduate Research Opportunity Program, University of Toronto, P.G. and M.E.S. by summer studentships from the Banting and Best Diabetes Centre, and P.L.B. by a Tier I Canada Research Chair. This work was supported by the Banting and Best Diabetes Centre (Pilot and Feasibility Grant, to D.J.P. and P.L.B.) and an operating grant from the Canadian Institutes of Health Research (PJT-15308 to P.L.B.). Some of the equipment was supported by the 3D (Diet, Digestive Tract and Disease) Centre funded by the Canadian Foundation for Innovation and Ontario Research Fund (project numbers 19442 and 30961).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.E.M. researched data, analyzed data, and wrote the manuscript. A.M. researched data, analyzed data, and wrote the manuscript. B.J.C. analyzed data and reviewed and edited the manuscript. K.N. analyzed data and reviewed and edited the manuscript. A.W. analyzed data and reviewed and edited the manuscript. P.G. researched data and reviewed and edited the manuscript. M.E.S. researched data and reviewed and edited the manuscript. D.J.P. provided funding and reviewed and edited the manuscript. P.L.B. provided funding, contributed to discussion, and wrote the manuscript. P.L.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Portions of this study were presented in abstract form at the 55th Annual Meeting of the European Association for the Study of Diabetes, Barcelona, Spain, 16–20 September 2019.