Oxyntomodulin regulates resetting of the liver circadian clock by food

We screened a commercially available metabolic peptide library (Obesity Peptide Library, Phoenix Europe GmbH; DE) for factors capable of resetting luciferase activity rhythms in organotypic liver slice cultures from Per2::LUC circadian reporter mice (Yoo et al., 2004). Interestingly, out of 200 peptides applied during the descending phase (∼180°, corresponding to the early morning) of the luciferase activity rhythm, only a few produced marked phase shifts, including three proglucagon-derived peptide (PGDP) hormones: exendin-4, OXM, and GCG (Figure 1—source data 1). Exendin-4 has been isolated from the salivary gland of the Gila monster, with no analogue in rodents or humans. To compare the effectiveness of mammalian PGDPs in liver clock resetting, we treated slices with increasing doses of OXM, GCG, and the three other commercially available PGDPs, glicentin-related pancreatic polypeptide (GRPP), glucagon-like peptide-1 (GLP-1), and glucagon-like peptide-2 (GLP-2) (Figure 1A). GLP-1, GLP-2, and GRPP (0.5–450 nM) had no significant resetting effects on PER2::LUC phase compared to PBS-treatment (Figure 1B–D). GCG resulted in phase delays of up to 3 hr, but only at relatively high concentrations (Figure 1E). In contrast, OXM reset PER2::LUC rhythms in liver slices at much lower doses, resulting in phase delays of up to 8 hr at higher concentrations (Figure 1F). To act as a true Zeitgeber signal one would expect differential OXM effects depending on treatment time, that is, a circadian gating effect. We tested this by applying OXM at different phases of the PER2::LUC rhythm. To validate the setup, slices were treated at different PER2::LUC phases with 100 μM of the glucocorticoid analog dexamethasone (DEX), which was previously shown to reset hepatocyte clocks in vivo (Balsalobre et al., 2000). In a phase-dependent manner DEX treatment reset PER2::LUC activity rhythms in slices (Figure 1—figure supplement 1). Very similar to what had been observed after DEX treatment in animals (Balsalobre et al., 2000), application in the first quarter of the PER2::LUC activity rhythm (0–90°) resulted in phase delays, while later treatments produced phase advances (100–180°) or had no marked effect (around 270°). Likewise, OXM effects were phase dependent. Delays were predominantly observed at 90–210° of the PER2::LUC cycle with a maximum around 180°, while only modest phase shifts were seen at 270–360° (Figure 1G). Though GCG also showed potential in resetting liver clock rhythms, OXM emerged as the most potent liver clock synchronizer from our screen. Moreover, contrary to GCG, OXM secretion is directly induced by food consumption in humans (Le Quellec et al., 1992), making it an attractive candidate for linking meal timing and clock function. Therefore, we focused on OXM for further analyses.

So far no OXM-specific receptor has been identified; however, OXM can bind to and activate both GCG and GLP-1 receptors (Jorgensen et al., 2007). Gcgr transcripts are strongly expressed in the murine liver (Sinclair et al., 2008). In contrast, the majority of previous studies failed to detect a full-length Glp1r mRNA in murine hepatocytes (Campos et al., 1994; Dunphy et al., 1998; Panjwani et al., 2013). We performed RT-PCR analyses for all annotated coding Glp1r exons on cDNA preparations from wild-type mouse livers with pancreas as positive control. Glp1r transcripts were present in pancreas, but undetectable in liver samples (Figure 2—figure supplement 1), in line with the absence of significant liver clock resetting effects of GLP-1 (Figure 1B) and potent resetting of PER2::LUC rhythms by OXM treatment in slices from PER2::LUC x Glp1r^−/− mice (Figure 2A). On the other hand, blocking glucagon receptors (GCGR) signaling by co-treatment with 2-(4-Pyridyl)-5-(4-chlorophenyl)-3-(5-bromo-2-propyloxyphenyl)pyrrol (Calbiochem Glucagon Receptor Inhibitor II; GRI-2) potently inhibited GCG- and OXM-induced clock resetting in Per2::LUC slices (Figure 2B).

GCGR is a G protein-coupled receptor that, via protein kinase A, leads to phosphorylation and activation of the transcription factor cyclic adenosine monophosphate (cAMP) response element-binding protein (CREB) (Gonzalez and Montminy, 1989; Dalle et al., 2004). This pathway is reminiscent of the SCN, where nocturnal light exposure induces Per gene transcription via cAMP signaling and CREB activation downstream of the N-methyl-D-aspartate receptor (Welsh et al., 2010). To investigate if OXM would impinge on the hepatic clock machinery in a similar way, we treated liver explants with OXM and performed chromatin immunoprecipitation (ChIP) analysis to measure CREB binding to cAMP response elements (CRE) in the Per1 gene promoter. 30 min after OXM treatment, CREB binding was significantly increased at the Per1 CRE, but not at downstream sequences (Figure 2C). In addition, we analyzed clock gene expression in liver slices at different intervals after treatment with OXM at 180°. Per1 expression was transiently induced 60 min after addition of OXM to the medium, returning back to normal levels after 120 min (Figure 3A). Similarly, Per2 was induced by OXM after 60 min, but mRNA levels remained high even after 120 min (Figure 3B). No significant effect was seen on Bmal1 expression at all time points (Figure 3C). In line with the absence of OXM-induced phase shifts at 360° (Figure 1G), Per1/2 and Bmal1 mRNA levels were unaffected by OXM treatment at this phase (Figure 3D–F). Induction of Per1 and Per2 expression was preserved in Glp1r^−/− slices, suggesting that hepatic OXM effects are independent of GLP-1R signaling (Figure 3G,H).

To test if the effect of OXM on liver clock gene activity is preserved in vivo, we analyzed hepatic Per1/2 transcription after OXM treatment in wild-type mice. Analogous to what we observed in slices, robust Per1 and Per2 induction was observed after i.v. injection of OXM at Zeitgeber time (ZT) 3 (Figure 4A). When animals were treated with OXM at the opposite phase of the LD cycle (ZT15)—a time when nocturnal mice usually eat and, thus, no food-mediated clock shifts would be expected—no induction of Per gene expression was observed (Figure 4B). Of note, in situ hybridization (ISH) showed no acute effect of OXM treatment on Per expression in the SCN at ZT3 (Figure 4C), indicating that OXM acts primarily on the liver clock and in line with the observed phase stability of the SCN clock under time-restricted feeding conditions (Damiola et al., 2000).

To assess OXM effects on liver clock phase, we next treated wild-type mice with either PBS or OXM at the beginning of their rest phase on the first day in constant darkness (DD) and under fasting conditions, thus excluding potential confounding effects of light exposure or food intake. Per2 and Dbp expression were determined from liver cDNA preparations at different time points using qPCR. We detected phase delays of Per2 and Dbp mRNA rhythms in livers of OXM-treated mice relative to those in PBS-injected control animals (Figure 4D), in line with the phase-delaying effects of OXM administration in slices at this time (Figure 1G). In addition, Per2, but not Dbp, rhythms appeared dampened after OXM injection.

Tissue clocks regulate local physiology via coordination of transcriptional programs. To test if OXM treatment would impinge on hepatic energy metabolism, we analyzed the expression of important metabolic transcripts after OXM treatment. Similar to what we observed for clock gene activity, transcript profiles of genes involved in liver carbohydrate metabolism were found either phase delayed (Foxo1 and Pdk4; Figure 5A,B) and/or dampened (Foxo1 and Pklr; Figure 5A,C). Of note Pepck, which was previously described as a clock output gene in liver (Lamia et al., 2008) was not rhythmic under these conditions, and OXM had no further effect on Pepck mRNA levels (Figure 5D). Expression levels of the glucose transporter Slc2a2 (Glut2) and the pyruvate transporter Slc16a7 were also dampened or phase-delayed, respectively (Figure 5E,F).

In summary, our data so far show that OXM treatment resets liver circadian mRNA rhythms in a phase- and dose-dependent manner, indicating that it may be involved in food-induced resetting of the liver circadian clock and metabolic machinery.

In humans, OXM levels in the blood rise in response to food intake (Le Quellec et al., 1992). To test whether this effect is conserved in mice, we determined diurnal plasma oxyntomodulin-like immunoreactivity (OLI) profiles in mice with ad libitum food access and in fasted animals. In fed mice, OLI levels were elevated during the active dark phase, while under fasting conditions OLI concentrations were constant and consistently low (Figure 6A). In line with this, non-rhythmic Gcg mRNA expression in the gut was observed under fasting conditions (data not shown). These data suggest a link between food intake and OXM secretion. To test this more directly, we used a 12-hr fasting-refeeding paradigm. Mean fasting OLI levels in the early morning (ZT1) were ∼3.5 ng/ml, but showed high inter-individual variation (Figure 6B). Upon food intake, a rapid increase (relative to individual fasting levels) was observed after 20 min. This effect persisted for more than 1 hr before returning to baseline levels (Figure 6C). Of note, OLI induction after OXM injections were about twofold–threefold higher than what was observed after refeeding (Figure 6—figure supplement 1). To test if postprandial OLI induction is sufficient to affect liver clock gene expression, we analyzed Per1/2 mRNA levels in livers of wild-type mice after fasting-refeeding. Parallel to the rise in plasma OLI, we observed a transient postprandial increase of hepatic Per1 expression. Per2 expression showed a delayed, but a more persistent induction (Figure 6D). Food-mediated Per activation was partly inhibited by treatment with purified anti-OXM IgG to neutralize the effects of endogenous OXM in wild-type (Figure 6E) and in Glp1r^−/− mice (Figure 6F). Importantly, the effects of refeeding on insulin, GLP-1, and GCG plasma levels were not affected by treatment with anti-OXM IgG, suggesting that these peptides are not involved in the activation of food-induced hepatic Per expression (Figure 6G–I).

We next tested if OXM signaling regulates food-mediated phase resetting of the liver clock using Per2::LUC slice preparations. Per2::LUC mice were starved overnight (ZT12-24/0) and treated with either IgG or anti-OXM antibodies prior to re-feeding or extended starving. Animals were sacrificed at ZT4 and liver slices were prepared to determine luciferase rhythm phases. After a 12-hr fast, food intake caused a significant phase delay of PER2::LUC activity in liver slices and this effect was attenuated by OXM neutralization (Figure 6J,K). No effect on PER2::LUC phase was seen after anti-OXM treatment alone (Figure 6K).

Together, these data suggest that elevated OXM levels in response to food intake affect hepatic clock gene expression. Neutralization of endogenous OXM signaling inhibits food-induced clock gene induction and rhythm shifts, suggesting that OXM may act as a metabolic synchronizer of hepatic clocks.

All animal experiments were ethically assessed and licensed by the Office of Consumer Protection and Food Safety of the State of Lower Saxony and in accordance with the German Law of Animal Welfare (license nos. V312-7224.122-4 and 33.12-42502-04-12/0893). For all experiments adult wild-type mice male C57BL/6J (8–24 weeks old) were used. If not stated otherwise, mice were provided with food and water ad libitum. To investigate the effect of OXM on gene expression by qPCR or ISH, mice were peripherally treated with OXM (i.v. 4 µg/mouse; i.p. 25 µg/mouse) or vehicle (PBS) at ZT3, ZT15, or 12 hr after light-off. For luminescence measurements adult heterozygous males Per2::LUC (Yoo et al., 2004) and Per2::LUC x Glp1r^−/− were used. Glp1r^−/− mice were maintained on a C57B/6J background (Scrocchi et al., 1996). All mice were exposed to a 12-hr: 12-hr light–dark cycle with 100 lux in the light phase (LD12:12). Animals were sacrificed at indicated time points by cervical dislocation. Animals euthanized during the dark phase were handled under red light and eyes were removed before dissection. All tissue samples were collected and immediately snap-frozen on dry ice or liquid nitrogen. For long-term storage tissues were kept at −80°C.

In order to test Per induction and phase shifts after refeeding, mice were fasted for one night (13 hr, or from ZT12—ZT1) and either food deprived until decapitation or refed at ZT1. The refed mice received either anti-OXM rabbit IgG (i.p. 50 µg/mouse; T-4800; Bachem, Bubendorf, Switzerland) or control IgG from unimmunized rabbit serum (I5006; Sigma-Aldrich, Seelze, Germany) at ZT1 when food was returned.

Luminescence was measured from cultured liver slices of heterozygous Per2::LUC mice as described previously (Yoo et al., 2004) modified to include the use of culture plate inserts (Millipore, Billerica, MA). Briefly, the median lobe of the liver was isolated and 300-µm slices were prepared using a vibratome (Campden Instruments, Loughborough, UK). The slices were immediately placed onto a culture plate insert in 35-mm petri dishes filled with 1-ml culture medium (D-MEM with high glucose, w/o L-glutamine and phenol red; Life Technologies, Darmstadt, Germany) supplemented with 3 mM sodium carbonate (Sigma–Aldrich), 10 mM HEPES buffer, 2 mM L-glutamine, 2% B-27 supplement, 25 U/ml penicillin/streptomycin and 0.1 mM D-luciferin (all Life Technologies). Luminescence was measured in a luminometer (Actimetrics, Evanston, IL) at 32.5°C with 5% CO₂. Analyses were performed using the LumiCycle analysis (Actimetrics) and Prism software packages (GraphPad, La Jolla, CA). PER2::LUC activity in slices closely follows a sine wave shape. The intersection of the ascending cross-section of the sine wave with the x-axis was defined as 0°/360°, the peak as 90°. Degrees at the time point of treatment were calculated as follows: T_p[°] = ((T_p[hsm] − P_bt[hsm]): P_bt) × 360 + 90 with T_p = treatment phase; ° = degree; hsm = hours after start of measurement; P_bt = peak before treatment. If the result was >360°, the value was subtracted by 360. Raw data were baseline subtracted with running averages of 24 hr. Peaks were defined as middle time point between two troughs of the sine wave. Period was determined as the time between peaks averaged over 2–3 consecutive cycles. For the duration of treatment, samples were maintained at 32.5°C to avoid resetting of clock gene expression rhythms due to temperature changes. Phase shifts were determined by comparing extrapolated peak times from sine wave fits before and after treatment. Unless otherwise stated, peptides used for experiments were dissolved in culture medium and administered at a final concentration of 450 pM.

Quantitative real-time PCR (qPCR) was performed with a CFX96 thermocycler system (Bio-Rad, Munich, Germany) with iQ-SYBR Green SuperMix (Bio-Rad). Relative quantification of expression levels by a modified ΔΔCT calculation was performed as described (Pfaffl, 2001). ß-Actin was used as a reference gene. Statistical analyses were performed using GraphPad Prism software. Circadian profiles of clock gene expression were normalized against the average values over all time points. Induction analyses were normalized against untreated conditions (0 min). PCR primer sequences are listed in Figure 6—source data 1.

The Per1 probe corresponds to nucleotides 1 to 619 (GenBank accession number AF022992↗) and Per2 corresponds to nucleotides 229 to 768 of GenBank AF036893↗. PCR products had been cloned into pCR II TOPO vector using TOPO TA Cloning Kit (Life Technologies) (Oster et al., 2002). Linearization of vectors for in vitro transcription was done with EcoRI. ³⁵S-UTP (PerkinElmer, Waltham, MA) labeled RNA probes were prepared using RNA Transcription Kit (Maxi Script Labeling Kit, Life Technologies) with T7 or T3 RNA polymerases according to the manufacturer's protocol. 10-µm cryosections were cut using a Leica CM3050 cryostat. Cryosections were fixed in 4% paraformaldehyde, acetylated in acetic anhydride and dehydrated with ethanol. Hybridization was performed over night at 55–58°C. Autoradiographs were analyzed by densitometry (Bio-Rad GS-800) using QuantityOne software (Bio-Rad). Three sections per brain were used and background values were calculated from adjacent tissue areas on the same slide for each section. Measurements from different animals/experiments were compared for statistical analysis using GraphPad Prism (GraphPad).

Liver slices were homogenized and immediately cross linked with 1% formaldehyde. Chromatin was sonicated for 15-s on/20-s off cycles for 22 min using a Bioruptor sonicator (Diagenode, Denville, NJ). Samples were incubated overnight at 4°C with CREB antibody (ab31387, Abcam, Cambridge, UK). After clearing, samples were incubated with A/G agarose beads (Thermo Scientific, Braunschweig, Germany) for 1 hr at 4°C followed by repetitive washings. After boiling for 10 min in 10% Chelex (Bio-Rad) with Proteinase K (150 mg/ml), samples were spun down and DNA-containing supernatant was collected for PCR. qPCR was performed as described above, and values were normalized to percentage of input. Primer sequences were: 5′-CAGCTGCCTCGCCCCGCCTC-3′/5′-CCCAAGCAGCCATTGCTCGC-3′ (Per1 CRE) and 5′-CCCCGCAGTCCTACGGTGCTG-3′/5′-AAGCCCCCAAACAACTGAAGGT-3′ (500 bp downstream sequence).

Blood collection for radioimmunoassay (RIA) was performed at ZT1 after 12 hr fast. Mice were allowed to recover for 3 days, then fasted again followed by treatments (with or without refeeding). Blood was collected from the tail vein at 0 min, 20 min, 60 min, and 120 min after treatment. Plasma concentrations of OLI were determined by RIA (Phoenix Pharmaceutics, Karlsruhe, DE) according to manufacturer's protocol modified to use a 50% reaction volume. GLP-1 (EZGLP-1T-36K, Millipore, Darmstadt, Germany), insulin (Catalog# 90080, CrystalChem, Downers Grove, IL), and GCG plasma levels (EZGLU-30K, Millipore) were determined by ELISA according to the manufacturers' protocols.

Data were analyzed with GraphPad Prism (GraphPad). Mann–Whitney tests were used for simple comparisons. For dose responses one-way ANOVAs and for two-factor comparisons two-way ANOVAs with Bonferroni post-tests were used. A p-value of less than 0.05 was considered significant.