Adipose tissue (AT) is a key metabolic organ which functions are rhythmically regulated by an endogenous circadian clock. Feeding is a “zeitgeber” aligning the clock in AT with the external time, but mechanisms of this regulation remain largely unclear. We tested the hypothesis that postprandial changes of the hormone insulin directly entrain circadian clocks in AT and investigated a transcriptional-dependent mechanism of this regulation. We analyzed gene expression in subcutaneous AT (SAT) of obese subjects collected before and after the hyperinsulinemic-euglycemic clamp or control saline infusion (SC). The expressions of core clock genes PER2, PER3, and NR1D1 in SAT were differentially changed upon insulin and saline infusion, suggesting insulin-dependent clock regulation. In human stem cell–derived adipocytes, mouse 3T3-L1 cells, and AT explants from mPer2Luc knockin mice, insulin induced a transient increase of the Per2 mRNA and protein expression, leading to the phase shift of circadian oscillations, with similar effects for Per1. Insulin effects were dependent on the region between −64 and −43 in the Per2 promoter but not on CRE and E-box elements. Our results demonstrate that insulin directly regulates circadian clocks in AT and isolated adipocytes, thus representing a primary mechanism of feeding-induced AT clock entrainment.
Introduction
Adipose tissue (AT) is a key energy storage strongly contributing to the regulation of carbohydrate and lipid metabolism. AT also acts as an endocrine organ, secreting adipokines and cytokines that are able to modulate systemic insulin sensitivity, systemic inflammation, and appetite regulation and therefore plays an essential role in the pathogenesis of metabolic diseases (1). Many metabolic and secretory AT functions are time of day dependent. They are controlled by an endogenous circadian clock that generates local 24-h metabolic and secretory AT rhythms to coordinate AT physiology with day-night changes and corresponding rhythms of food intake (2,3). During the active phase (physiological day) food is consumed and nutrients are transported to white AT (WAT) where they are stored in the form of triglycerides. In contrast, during the inactive phase (physiological night), triglycerides from adipose stores are released as free fatty acids to serve as energy substrates for other organs (2). Correspondingly, a range of key regulators of glucose uptake, lipogenesis, lipolysis, and insulin signaling in adipocytes show circadian rhythms (4–6). Both in humans and in rodents, up to 25% of the WAT transcripts are expressed in a time of day–dependent manner (4,7).
The molecular basis governing these 24-h rhythms in AT metabolism (as well as other physiological and cellular processes) is an endogenous circadian clock comprised of a network of core clock genes (BMAL1, CLOCK, PER1/2/3, CRY1/2, etc.) organized in transcriptional, translational, and posttranslational feedback loops (8). The circadian system in mammals consists of a central clock in the suprachiasmatic nucleus (SCN) of the hypothalamus and peripheral clocks controlled by the SCN clock and in turn rhythmically regulating a number of clock-controlled genes within various signaling pathways (9). The important role of the AT circadian clock for metabolic regulation was confirmed in animal models with genetic clock disruption demonstrating various adipose phenotypes with or without weight gain, disturbances of adipogenesis, and dyslipidemia (10–12).
Circadian clocks are self-sustained oscillators, which, however, need to be adjusted by external cues (in German, “zeitgebers”) to synchronize (entrain) endogenous rhythms with environmental cycles (8). Although light is a dominant zeitgeber for the SCN clock, the main external cue for peripheral clocks including AT is food intake (7,13). The proper entrainment of endogenous rhythms to the environment is a prerequisite for metabolic health. Dysregulation of circadian rhythms due to conflicting feeding and light stimuli, e.g., during jet lag, shift work, or nighttime feeding, disturbs metabolic homeostasis, which is associated with increased risk of obesity, type 2 diabetes, metabolic syndrome, and cardiovascular diseases (2,14). Notably, even a 3-day circadian misalignment in humans results in impaired glucose tolerance and decreased insulin sensitivity similar to those observed in subjects with prediabetes (14,15). In particular, nighttime feeding inverts clock rhythms in the mouse AT, uncoupling them from the SCN (7). Even a 5-h delay in meal timing shifts the clock phase in human AT (16).
Not only feeding time but also other feeding-associated factors affect the body’s internal clock. For example, caloric restriction influences the circadian clock in human AT and mouse liver (17,18). In addition, even isocaloric changes of food composition, i.e., dietary fat and carbohydrate content, alter clock rhythms in human AT (19).
While feeding-induced clock entrainment occurs in all organs except the SCN (20) and can be observed also in animals with SCN lesion (21) or genetic clock disruption (22), the mechanisms of feeding-induced clock entrainment are still not well understood. Elucidation of mechanisms of how food consumption entrains circadian clocks will help to develop strategies to maintain metabolic health and to counteract metabolic disturbances.
Recent rodent studies and in vitro experiments provided evidence that the postprandial increase of insulin secretion may be one of the most important systemic feeding signals entraining the circadian clock in the whole body (23–27). Insulin is a hormone secreted by the pancreas in response to the increase in feeding-induced blood glucose level promoting the glucose uptake from the blood into liver, fat, and skeletal muscle cells. Insulin is a key regulator of carbohydrate, fat, and protein metabolism, generally enhancing anabolic and inhibiting catabolic cellular processes (28). Amazingly, insulin is also able to entrain circadian clock oscillations, as was shown in mouse liver, hepatocytes, fibroblast culture, and even hair follicles (23–27). However, whether insulin can modulate the circadian clock in human AT and, in particular, which mechanisms are involved are still unknown. Here, we show that insulin directly entrains circadian clocks in human and mouse AT and isolated adipocytes and uncover a transcriptional-dependent mechanism of this regulation.
Research Design and Methods
Human Study
A total of 17 middle-aged obese subjects with normal glucose tolerance (clinical trial reg. no. NCT00774488, ClinicalTrials.gov [Supplementary Table 1]) underwent at least one of the following procedures for 4 h after an overnight fast: 1) control experiments (0.9% saline infusion [SC], n = 11) and 2) the hyperinsulinemic-euglycemic clamp (EC) with continuous infusion of 40 mU · m2 body surface · min−1 human insulin (Actrapid; Novo Nordisk) at a steady-state capillary plasma glucose concentration of 4.4 mmol/L (n = 10) as previously described (29). Infusions started at 8.00 a.m. and finished at 1.00 p.m. Blood glucose concentrations were determined using a glucose oxidase method, and serum insulin was measured by ELISA assay (Mercodia AB). Subcutaneous AT (SAT) biopsies were taken before (−40 min) and 240 min after start of infusions, from distinct sites, at the level of the umbilicus. The study was approved by the University of Potsdam Institutional Review Board, and all volunteers gave written informed consent.
Transcriptomic Analysis of Human Adipose Tissue Samples
Total RNA was extracted from SAT samples with the RNeasy Lipid Tissue Mini Kit (QIAGEN), and 300 ng RNA was hybridized to the Agilent 60-mer whole human genome (4 × 44 K) single-color DNA microarrays (Agilent Single Color 12391) (Agilent Technologies). Data were analyzed with Agilent GeneSpring GX software.
Cell Culture
Human mesenchymal stem cells (hMSCs) were isolated from SAT samples of three obese subjects and differentiated as previously described (30) After 14 days, cells were transferred into DMEM/F12, supplemented with 3% FCS (HyClone; Thermo Fisher Scientific), 8 µg/mL biotin, and 15 mmol/L d-pantothenate for 48 h, which was changed to the serum-free medium 2 h before insulin stimulation.
3T3-L1 cells were maintained in DMEM (Invitrogen) containing 4.5 g/L glucose, 10% FBS, and 1% penicillin/streptomycin. For differentiation, maintenance medium was supplemented with 1.7 μmol/L insulin, 500 μmol/L 3-Isobotyl-1-Methylxanthine, and 1 μmol/L dexamethasone for 48 h and with 1.7 μmol/L insulin only for another 48 h.
Promoter Analysis
The vectors pGL3basic (pGV-B2)-mPer2-Luc (31), pGL3ppp-Per1Prom, and pGL4.23(luc2/minP)7×CRE (32) were kindly provided by other groups. For construction of the vectors pL6-mPer2-Luc and pL6-mPer1-Luc, 3.4 kb long fragment of the murine Per2 promoter (spanning from −3,309 to 105 connected to the coding region of firefly luciferase) and 2.9 kb segment of the mPer1 promoter (spanning from −2,938 to 42) were cloned into a lentiviral vector, pLenti6/V5-DEST Gateway (Thermo Fisher Scientific). The cloning of truncated versions of the Per2 promoter (200 base pairs [bp], 170 bp, 149 bp, 126 bp, 100 bp) in a lentiviral vector was performed using gene synthesis. The CRE as well as the NFY and SP1 binding sites in the Per2 promoter were mutated by site-directed mutagenesis using a QuikChange Site–directed mutagenesis kit (Agilent Technologies) and mutagenesis primers (Supplementary Table 2). In addition, a 7×CRE reporter construct containing seven repeats of the CRE element was cloned into the lentiviral vector. For the annotations of the mPer2 and mPer1 promoter regions, the descriptions by Yoo et al. (33) and Travnickova-Bendova et al. (34) and MatInspector software were used.
Lentivirus Transduction
Lentiviruses were produced with HEKT293T cells as previously described (35). Lentiviral transduction of 3T3-L1 fibroblasts was performed for 1 day, and cell selection was made over 7 consecutive days with 10 µg/mL blasticidin starting 48 h after infection. Transduction of hMSC-derived adipocytes was performed by incubation of the cells with lentiviruses for 24 h.
Animals and Ex Vivo Tissue and Cell Culture
Heterozygous male mPer2Luc knockin mice (33) expressing an mPERIOD2::LUCIFERASE (mPER2::LUC) fusion protein and male C57Bl6/J mice were maintained in a 12-h light/12-h dark cycle. All procedures were authorized by a local animal ethics committee and performed in accordance with German animal protection law (Deutsches Tierschutzgesetz). Animals were killed by cervical dislocation at zeitgeber time 6 (ZT6) for the isolation of AT, adipocytes, stromal vascular fraction (SVF), and peritoneal macrophages. For bioluminescence assays, epididymal AT was minced into 1 × 1 × 2 mm3 pieces and synchronized in 1 μmol/L dexamethasone for 30 min at 37°C. Adipocytes and the SVF were isolated from AT using collagenase A (Sigma-Aldrich) digestion in KRBH buffer for 1 h at 37°C and centrifugation for 2 min at 400g. Peritoneal macrophages were received by peritoneal lavage and cultured in DMEM with 4.5 g/L glucose and 10% FBS.
Bioluminescence Recordings
Bioluminescence recordings of tissues and cells were performed at 37°C in a luminometer (TopCount; PerkinElmer), LumiCycle (Actimetrics), or light-tight boxes using a single photomultiplier tube (HC135-11MOD; Hamamatsu Photonics) in 5-min bins for at least 7 days and in phenol-free DMEM or DMEM/F-12 with 1% penicillin/streptomycin and 250 μmol/L luciferin (PJK GmbH). Insulin (Sigma-Aldrich) or mock treatment was performed after the first bioluminescence peak during the drop phase unless stated otherwise. Bioluminescence time series were analyzed using ChronoStar 3 (Stephan Lorenzen, Institute for Theoretical Biology, Humboldt-Universität zu Berlin). The analysis of insulin-induced phase shifts (phase response curve [PRC]) and insulin-mediated induction of bioluminescence (area under the induction curve) was developed by Dr. Sarah Lück and Dr. Nicole Wittenbrick, respectively, using R software (https://www.r-project.org).
Quantitative Real-time PCR
Total RNA was extracted by the PureLink RNA Mini Kit, TRIzol (Life Technologies) or RNeasy Lipid Tissue Mini Kit (QIAGEN). cDNA was synthesized with a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) or M-MLV Reverse Transcriptase (Thermo Fisher Scientific). Quantitative real-time PCR (qPCR) was performed using specific primers (Supplementary Table 2).
ChIP-Seq Data Analysis
For the identification of possible binding regions for several transcription factors, we analyzed ChiP-Seq data. Already aligned files were downloaded and converted to BED files using bedtools (https://github.com/arq5x/bedtools2/releases/) and then filtered for chromosomes in linux. BED files were loaded into R (version 4.0.3) with use of Gviz (version 1.32.0), rtracklayer (version 1.48.0), and chipseq (version 1.38.0) packages. ChIP-Seq scores were calculated within a size of 10 bp and plotted with the plotTracks function.
Statistical Analyses
Statistical analyses were performed with R software, SPSS 20.0 (SPSS) and GraphPad Prism 5 software. Data are presented as means ± SD unless otherwise stated. A P value <0.05 was assumed as significant.
Data and Resource Availability
Microarray data are available online in the Gene Expression Omnibus (GEO) database (GSE40589). ChIP data used for analysis are publicly available in ENCOCE (https://www.encodeproject.org/) and GEO (https://www.ncbi.nlm.nih.gov/geo/) databases. Other data sets generated or analyzed during the current study are available from the corresponding author upon reasonable request.
Results
Insulin Changes Clock Gene Expression in Human AT In Vivo
To investigate whether insulin impacts the circadian clock in human AT in vivo, we conducted a randomized single-blind crossover trial involving 17 overweight male volunteers with normal glucose tolerance (Supplementary Table 1). Subjects underwent a euglycemic-hyperinsulinemic clamp (EC) (n = 10) and/or a normal saline infusion, used as a control experiment (SC) (n = 11) (Fig. 1A). In the steady state of the EC, circulating insulin levels strongly increased in the comparison with baseline and were higher than in SC experiments (P < 0.001), in the presence of similar blood glucose concentrations in both tests (Fig. 1B).
SAT transcriptomes were analyzed before (−40 min) and 240 min after start of insulin or saline infusions. While in control experiments (SC), the expression of 427 transcripts (1.1%) in SAT was altered, 689 transcripts (1.8%) were differentially expressed upon insulin infusion (EC) (Supplementary Fig. 1) and 136 transcripts overlapped in SC and EC. As expected, in both control and insulin infusion experiments, functional annotation revealed that altered gene expression over the course of 240 min is associated with the circadian rhythm pathway (Fig. 1C and Supplementary Table 3). Importantly, however, insulin infusion in EC additionally induced changes in genes associated with circadian entrainment, AMPK, phosphatidylinositol, and mTOR pathways (Fig. 1C and Supplementary Table 3) which were not found in SC. Analysis by qPCR confirmed a time-dependent decrease in the expression of core clock genes PER1, PER2, PER3, NR1D1, and NR1D2 expression at 240 min of SC and/or EC experiments compared with the −40 min expression level. Notably, PER2 and PER3 expression significantly decreased after 4 h of the SC experiment but not upon the insulin infusion (Fig. 1D). In contrast, NR1D2 expression was not affected upon saline infusion but was downregulated upon insulin infusion (Fig. 1D). Other clock genes (CLOCK, BMAL1, CRY1, CRY2, RORA, DBP, and TEF) did not show saline- or insulin-dependent mRNA expression changes (Fig. 1E). Consistent with previous reports that insulin affects clock gene rhythms, in particular PER2 oscillations, in mouse liver (23–26), our data suggest that postprandial insulin directly or indirectly affects the AT clock in humans.
Insulin Affects PER2 Oscillations in AT and Adipocytes
To test whether insulin has an impact on PER2 oscillations in AT, we treated hMSC-derived adipocytes with insulin in vitro and assessed PER2 levels. Insulin treatment (100 nmol/L) for 4 h resulted in an induction of PER2 mRNA expression (Fig. 2A). In hMSC-derived adipocytes transduced with a 0.5-kb mPer2 promoter luciferase reporter construct (36), insulin treatment induced transient increase of the mPer2 promoter activity and corresponding phase delays of Per2 bioluminescence rhythms (Fig. 2B). This suggests that insulin acutely regulates Per2 mRNA expression.
We further analyzed direct insulin effects using tissues/cells from the mPer2Luc reporter mouse, which enables the real-time imaging of the mPER2::LUC protein expression. We treated epididymal AT explants ex vivo with 100 nmol/L insulin and monitored mPER2::LUC bioluminescence rhythms for several days. Acute increase of mPER2::LUC expression upon insulin treatment and corresponding phase shifts of circadian rhythms were detected (Fig. 2C and Supplementary Fig. 2A and B). Similarly, insulin led to an increase in mPER2::LUC activity both in isolated adipocytes (Fig. 2C and Supplementary Fig. 2C and D) and the SVF (Supplementary Fig. 2E). In contrast, this effect was not found in peritoneal macrophages (Fig. 2D). Thus, we demonstrate direct insulin effects on the circadian oscillator in AT and in isolated adipocytes.
Insulin Phase Shifts the Circadian Clock of AT in a Time-Dependent Manner
To test whether insulin has a time of day–dependent effect on the oscillator in AT, fat pads of the mPer2Luc mouse were treated ex vivo with insulin or solvent at different times during the circadian cycle. Fig. 3A shows that insulin and solvent had distinct effects on circadian rhythms of mPER2. For example, stimulation with insulin 7 h after the mPER2 peak led to a phase advance of ∼9 h, while solvent treatment led to a phase delay of ∼3 h. The insulin phase response curve (PRC [Fig. 3B]) shows a continuous decline from large phase advances to phase delays between approximately circadian time (CT) 0 and CT16. A “break point” is observed approximately at CT18, after which phase advances occurred again. In contrast, the majority of phase advances of the solvent PRC were between approximately CT6 and CT16.
Insulin Transcriptionally Regulates the Expression of Clock Genes in AT
We further tested, in a mouse model, whether the acute induction of mPER2 protein expression by insulin results from an increased level of mPer2 mRNA. To this end, epididymal fat pads of male C57BL/6J mice were stimulated with either 100 nmol/L insulin or solvent. mPer1 as well as mPer2 mRNA was significantly increased 0.5–1.0 h and 1.0–3.0 h after stimulation, respectively, with a magnitude of approximately twofold (Fig. 4A and B). In contrast, we found the strong and rapid but transient induction of the immediate early gene (IEG) c-fos, known to be activated rapidly in response to a wide variety of stimuli correlating with the resetting of circadian clocks (37) (Fig. 4C). mNr1d1 mRNA was decreased 3 h after insulin treatment, and mCry1 slightly increased after 1 h of stimulation (Fig. 4E and F). The expression level of mBmal1 and mCry2, however, was not altered (Fig. 4D and B).
Insulin treatment of differentiated (Supplementary Fig. 3) and nondifferentiated (data not shown) 3T3-L1 cells in vitro showed similar results. mPer2, mPer1, and mc-fos mRNA were induced upon insulin treatment with kinetics similar and to an extent similar to those in ex vivo AT explants (Supplementary Fig. 3A, C, and D). In addition, the induction of mPer2 was insulin dose dependent (Supplementary Fig. 3B). mBmal1 transcript levels increased slightly after 3 h, and mClock levels stayed unaffected (data not shown), whereas mNr1d1 transcripts was slightly reduced. mCry1 and mCry2 showed opposing trends, but the effects were rather small (Supplementary Fig. 3E–H).
To test whether the changes in transcript levels upon insulin treatment were dependent on active transcription, we treated 3T3-L1 cells with the RNA polymerase II blocker triptolide (38) simultaneously with insulin treatment and found attenuated the induction of clock gene expression after 1 h (mPer1 and mc-fos) or 2 h (mPer2 and mBmal1) (Supplementary Fig. 4), suggesting indeed a direct transcriptional regulation.
Promotor Analysis of the mPer2 Gene in 3T3-L1 Cells
We next investigated the insulin-mediated transcriptional regulation from the mPer2 promotor and compared it with mPer1 promotor regulation. To this end, we used reporter 3T3-L1 cells harboring a 3.4-kb fragment and a 2.9-kb fragment of the mPer2 and mPer1 promoter, respectively (Fig. 5A). The oscillatory amplitude of the mPer1 promoter was lower than that of mPer2 (Fig. 5B). Both promoters strongly responded to the treatment with insulin, although the mPer2 promoter showed two times higher inducibility (Fig. 5C–F) and its maximal activity was reached later in comparison with mPer1 (2.67 ± 0.2 h for mPer2 and 1.87 ± 0.08 h for mPer1) (Fig. 5G). In addition to the acute induction of the promoter activity, the perturbation of the oscillator with insulin led to phase delays in the circadian rhythmicity of the mPer2 promoter with an average of −8.59 ± 2.09 h. In contrast, a slight phase advance of 0.32 ± 0.8 h was determined for solvent (Fig. 5C).
To find out more about the transcriptional regulation of the mPer2 promoter, we tested 1) whether CRE elements—known to be required for mediating light effects in the SCN (39)—are sufficient for insulin-mediated upregulation and 2) whether the CRE element (−1,607) (Supplementary Fig. 5A) is necessary for its insulin-induced activation. For this, we used a 7×CRE reporter construct containing seven repeats of the CRE element upstream of a minimal promoter fused to a luciferase. Insulin did not lead to a substantial induction of CRE elements within this construct (Supplementary Fig. 5A–C). Further, a mutation of the CRE binding site within the Per2 promoter did not attenuate the induction of the promoter upon insulin treatment (Supplementary Fig. 5D and E), indicating that insulin-inducible expression from the mPer2 promoter is CRE independent.
We tested, further, whether the activation of the promoter upon insulin stimulation might require activation of the E-box located 20 bp upstream of the transcription start site (TSS) (33) (Supplementary Fig. 6A). The mutation of the E-box led to the loss of the promoter rhythmicity (data not shown) but did not attenuate the induction of the promoter activity upon insulin stimulation (Supplementary Fig. 6).
Next, we performed a series of mPer2 promoter truncations to identify insulin-responsive regions. To this end, we generated five truncated versions of the mPer2 promoter ranging from −94, −64, −43, −20, and 6 to 105 relative to the TSS, resulting in constructs comprising 200 bp, 170 bp, 149 bp, 126 bp, and 100 bp, respectively (Fig. 6A and B). All five mPer2 constructs responded to the insulin treatment with an acute and transient increase in bioluminescence (Fig. 6C and D). Three promoter constructs, mPer2-3.4 kb, -200 bp, and -170 bp, responded with a similar magnitude to the insulin treatment (assessed as a mean difference of the area under the induction curve between insulin and solvent treatment), whereas the constructs mPer2-149 bp, -126 bp, and -100 bp responded with a substantially smaller magnitude (Fig. 6E). This indicates a potentially important insulin-responsive region between the −64 and −43 positions upstream of the TSS. Within this region, we found a potential binding site (CCAAT) for the insulin signaling–associated transcription factor nuclear transcription factor Y (NFY) (40), located in the −58 region (Fig. 6A and Supplementary Fig. 7). CCAAT is also a potential binding site for CCAAT/enhancer binding proteins (C/EBPs) contributing to the regulation of adipocyte differentiation and metabolism (41). Eleven bp upstream of the CCAAT box, we found a GC box (GGGCGG), which co-occur with CCAAT boxes (42) and might represent a binding site for the transcription factor SP1 (Fig. 6A and Supplementary Fig. 7) also playing a critical role in the regulation of gene expression in response to insulin (43). Similarly, potential NFY/C/EBP and SP1 binding sites were found within the mPer1 promoter within the region near the TSS (Supplementary Fig. 8). ChIP-Seq data analysis confirmed the presence of binding peaks for these transcription factors in Per1 and Per2 promoters of mouse 3T3-L1 and embryonic stem cells (ESC) and human HepG2 cells near the TSS (Supplementary Fig. 9). Notably, double mutation of the above-mentioned NFY and SP1 binding sites within the Per2 promoter, but not the mutation of each binding site separately, attenuated the insulin induction of the promoter (Supplementary Fig. 10), although SP1 mutation seems to contribute more to this effect. This suggests that the binding of both transcription factors is required for this regulation.
Discussion
Both rodent and human AT are well known to harbor circadian clocks, which generate 24-h rhythms in AT functions and which phase can be shifted or even uncoupled from the SCN by feeding (7,44). In the current study, we demonstrate that the feeding-associated hormone insulin directly regulates circadian clocks in human and mouse AT and isolated adipocytes affecting the phase of PER oscillations and thus representing a primary mechanism of feeding-induced clock entrainment in AT.
Mechanisms by which feeding entrains the clock in peripheral tissues were intensively discussed in previous works. Circadian clock gene expression responds to feeding rapidly—within 30 min after refeeding in the liver of fasted rats and within 1 h in mice (45). Because this phenomenon is also observed in animals with SCN lesions, a hypothesis of a “food entrainable oscillator” located outside of the SCN was suggested some decades ago, but its existence remained controversial (46). Further, a range of metabolic cues were suggested for this phenomenon, especially in the liver. They include postprandial changes of nutrients—glucose, lipids, and amino acids—as well as their combinations and their plasma metabolites (47–51). Further, rodent studies clearly demonstrated a role of feeding-associated hormones in clock entrainment including glucagon (52), glucagon-like peptide 1 (53), leptin (54), ghrelin (55), and oxyntomodulin (56).
The role of the key metabolic hormone insulin (and related hormone IGF-I) as a link between feeding and circadian clock was recently shown in mouse liver, hepatocytes, fibroblasts, and hair follicles (23–27). Several studies have also mentioned that insulin can affect circadian clocks in mouse WAT (23,25), but not in human AT, and underlying mechanisms were only partly elucidated (23). In our study, we investigated insulin effects on circadian clocks in vivo in human SAT, ex vivo in mouse AT explants, and in vitro in the culture of hMSC-derived adipocytes and mouse 3T3-L1 adipocytes.
We first conducted transcriptome analysis of human SAT biopsies from obese subjects during insulin infusion and control saline infusion. As expected, genes of the circadian rhythm pathway were affected within 4 h of both saline and insulin infusion. Saline-dependent changes of clock gene expression highly likely represent the time-dependent changes in this time frame because similar time-dependent clock expression changes within 4 h ante meridiem were observed in human blood cells and ATs by our and other groups (4,16,19,57). Importantly, only insulin infusion induced changes in the circadian entrainment pathway as well as AMPK, phosphatidylinositol, and mTOR pathways, which are suggested to be involved in the insulin-dependent clock entrainment in other tissues and cells (23,24,26). Notably, the PER2 and PER3 mRNA expression significantly decreased after 4 h of the SC experiment but not upon the insulin infusion, while NR1D2 expression was downregulated in the EC experiment but not upon saline infusion, supporting the role of NR1D2 in the metabolism-clock interaction (58). Based on this human data, we hypothesized that changes of circulating insulin levels acutely regulate the circadian clock in AT modulating the oscillation of multiple genes within a clock machinery. Supporting our hypothesis, previous mouse studies demonstrated that insulin injection rapidly increases mPER2 protein abundance in vivo in freely moving mPER2::LUC animals as well as mPer2 mRNA expression and protein levels in various organs (25,26).
We also found that insulin treatment led to an increase in mPER2 protein levels and corresponding phase shift of PER2 rhythms in AT explants and in isolated adipocytes and preadipocytes, as well as the SVF. In turn, changes in PER protein levels and oscillations are sufficient to reset the phase of the whole molecular clock machinery (59). Notably, both adipocytes and different cell types within the SVF (preadipocytes, endothelial precursor cells, immune cells, etc.) might contribute to the observed insulin-dependent shifts of the PER2 oscillations. However, we did not find insulin effects on the clock rhythmicity in macrophages; thus, these cells are possibly not involved in the observed effect. Interestingly, the direction of the phase shifts in AT varied by the time of insulin treatment. In our experiments, insulin PRC resembles a type-0 PRC including large phase delays and phase advances, which are interrupted by a “breakpoint” (60). Type-0 PRCs are characteristic for strong zeitgebers, as shown by the data of Crosby et al. (26) in mouse fibroblasts. The time of day–dependent restricted responsiveness of the clock to zeitgebers is known to be modulated via a phase-dependent gating (61). Human AT indeed demonstrates a time of day–dependent insulin sensitivity (5); however, the underlying mechanism is still not known. Sato et al. (23) showed that the bidirectional response to insulin treatment could not be explained by the expression level of the insulin receptor; nevertheless, they demonstrated that mPer2 induction is essential for phase shifts upon insulin stimulation.
Interestingly, the AT oscillator also showed the phase-shifting effect of solvent, which was different from insulin treatment. This phenomenon could be explained by 1) temperature changes during handling of the organ cultures, 2) the mechanic perturbation of the recording system, or 3) the supplementation of the solvent solution changing nutrient concentration within the medium. Our observation shows that the AT oscillator is quite susceptible to environmental changes, which could also explain large phase shifts in response to insulin treatment (62).
Our results show that insulin regulates the expression of clock genes via transcriptional changes. Indeed, both in human and mouse AT explants, insulin acutely induces Per2 transcript and protein levels. The induction of the Per2 mRNA and the mPER2 protein in AT explants showed roughly similar kinetics and magnitude, suggesting a correlated regulation of their expression. In contrast, in mouse fibroblasts, Crosby et al. (26) showed that insulin regulates mPER2 due to increased translation, whereas pharmacological inhibitors of transcription and proteasomal degradation do not affect the insulin-induced regulation. This discrepancy might be explained by cell- and tissue-specific mechanisms and could manifest an additional mechanism for the increase of the protein as well.
In mouse AT explants and in 3T3-L1 cells, we further detected a rapid and transient induction of mPer1 and mc-fos. Together with the mPer2 gene, these IEGs are implicated in the resetting of peripheral clocks (63). In 3T3-L1 adipocytes the induction of IEGs was attenuated via RNA polymerase II blocker, confirming their de novo gene synthesis via transcription. Additionally, a nonacute Bmal1 transcriptional activation could be involved in the entrainment of the clock (64).
Along this line, we performed a promoter analysis of the mPer1/2 genes. An acute elevation of the promoter activity of both genes as seen in AT and adipocytes as well as differentiated hMSCs again indicated a direct transcriptional regulation. In addition, we excluded the involvement of CRE elements and E-boxes for insulin-mediated mPer2 activation. Instead, we identified a potentially important regulatory region for NFY and SP1, highly conserved transcriptional factors, likely contributing to the insulin-dependent metabolic regulation. In particular, NFY and SP1 are implicated in the insulin-induced activation of the sterol regulatory element binding protein 1c (SREBP1c) (40), a key gene of lipogenesis and adipocyte differentiation. NFY binds to CCAAT boxes (65), located in the −58 region of the mPer2 promoter. The transcription factor SP1 binds to GC-rich elements that co-occur with CCAAT boxes (42), and such a GC box (GGGCGG) was present upstream of the CCAAT box in the mPer2 promoter. Notably, NFY and SP1 binding sites are also present in the mPer1 promoter and might also play a role in mPer1 activation, although the activation of the mPer1 gene via insulin occurred earlier than that of mPer2. Further, transcription factor C/EBP, a key regulator of differentiation and energy metabolism in adipocytes (41), might also bind to the CCAAT site. We therefore hypothesized that the activation of NFY (C/EBP) or SP1 binding sites, or both, in the region between the −64 and −43 positions upstream of the TSS might be required for the activation of mPer2 via insulin (Fig. 7). Moreover, other transcription factors, such as HNF4 and serum response factor (SRF), might contribute to this regulation interacting with NFY and SP1 (32,66,67) (Fig. 7).
To test this, we conducted ChIP-peak calling and an experiment with mutations of NFY and SP1 binding sites within this promoter region. The ChIP-Seq analysis confirmed a binding of all three transcription factors (NFY, SP1, and C/EBP) in the mouse and human Per2 promoter, and double NFY and SP1 binding site mutations attenuated the insulin induction, whereas the mutation of each binding site separately did not show this effect. This suggests that the binding of both transcription factors is required for the insulin-dependent clock regulation, but further studies are needed to decipher an exact mechanism.
Other molecular pathways, involved in the insulin-dependent clock regulation in AT and not addressed in the current study, might be also tissue specific and therefore differ from other tissues. In the liver, both PI3K and MAPK pathways are required for this regulation (23,24), whereas in AT, Sato et al. (23) could confirm only the involvement of the MAPK pathways. In fibroblasts, an activation of the mechanistic target of rapamycin (mTOR), increased phosphoinositide signaling, and miRNA downregulation are interrelated (26). Consistent with these findings, we also observed the alterations of AMPK, phosphatidylinositol, and mTOR pathways in human SAT upon the insulin infusion (but not upon the saline infusion), which suggest their involvement in the insulin effects on clock. For instance, the insulin-induced activation on E4BP4 via the mTOR pathway and its binding to the D-box (−150 bp) within the Per2 promoter (Supplementary Fig. 7) might contribute to the resetting of circadian clock (68) and explain the rapid insulin-dependent induction of the longest promoter used in our experiments (Fig. 6C). Thus, as already mentioned above, careful investigations in future studies is needed of the role of distinct signaling pathways and selected transcriptional factors involved in the insulin-induced clock regulation in AT.
In conclusion, our results demonstrate that insulin directly regulates circadian clocks in AT and isolated adipocytes. Thus, insulin not only controls metabolic functions in AT but also acts as a time cue (Zeitgeber) for local clocks in this tissue. Hence, postprandial transient insulin increase might represent a primary signal underlying the entrainment of the AT clock by feeding. In experiments with AT explants and isolated adipocytes, we provided evidence that insulin directly regulates AT clock, although the regulation via insulin-induced systemic or local metabolite changes may also contribute to the observed phenomenon. Most likely, insulin provides the AT clock entrainment together with nutrients and other postprandial regulated hormones such as glucagon and GLP-1. In particular, combined insulin and glucose injections induced a larger and longer increase of PER2 expression compared with the insulin or glucose administration alone (26). Moreover, both action and secretion of insulin are subjected to circadian control (44,69,70), indicating complex control mechanisms. This regulation might be disturbed upon insulin resistance, which might contribute to the alterations of peripheral clocks observed in subjects with type 2 diabetes and severe obesity (71,72). The insulin-induced clock alterations might also represent an undesirable effect of the insulin treatment in patients with diabetes, which possibly depends on the insulin types (e.g., short- or long-acting insulins) and dosage/timing and requires future investigation. In other words, every time when we eat (or inject insulin), we reset our AT clock synchronizing metabolic processes in AT (and other organs) to fasting-feeding cycles and corresponding organism needs (Fig. 7). This also explains why mistimed eating disturbs circadian rhythms and induces metabolic problems such as obesity and associated diseases (2,14).
N.T., O.P.-R., A.F.H.P., and A.K. contributed equally to the manuscript (N.T. and O.P.-R. share first authorship; A.F.H.P. and A.K. share last authorship).
Clinical trial reg. no. NCT00774488, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.14791830.
Article Information
Acknowledgments. The authors thank all study participants for their cooperation. The authors thanks A. Wagner, S. Grosch, A. Borchert, K. Sprengel, and F. Werner (German Institute of Human Nutrition Potsdam-Rehbruecke, Germany) for technical assistance. The authors appreciate the group of Ohdo Shigehiro (Kyushu University, Fukuoka, Japan), Bert Maier (Charité, Berlin, Germany), and Alan Gerber from the Schibler laboratory (University of Geneva, Switzerland) for sharing the pGL3basic (pGV-B2)-mPer2-Luc, pGL3ppp-Per1Prom, and pGL4.23(luc2/minP)7×CRE vectors. The authors also thank Stephan Lorenzen for developing the ChronoStar software, Bert Maier and Raik Paulat for help in designing and building the temperature-adjustable light-tight boxes, Sabrina Klemz and Jeannine Mazuch (Charité, Berlin, Germany) for help in macrophage experiments, Nicole Wittenbrick for her contribution to the bioluminescence data analysis, Joseph Takahashi for providing PER2::LUC mice, and also Ute Abraham and Silke Reischl (Charité, Berlin, Germany) for the support of animal and cloning experiments.
Funding. This study was supported by a grant from the German Research Foundation (DFG grants Pf164/021002 NR [to Ö.G. and A.F.H.P.], KFO218 PF164/16-1 [O.P.-R., A.F.H.P., A.K.] and RA 3340/3-1 [to O.P.-R.]) and by the Morgagni Prize of the European Association for the Study of Diabetes 2020 (O.P.R.). M.S. was supported by the German Research foundation (DFG grant SCHU 2546/4-1).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. N.T., O.P.-R., V.M., S.L., M.K., V.J.N., P.G., C.S., N.G., N.R., and A.K. analyzed the data. N.T., O.P.-R., V.M., A.G., A.-C.O., Ö.G., and C.S. conducted the experiments. N.T., O.P.-R., A.G., M.K., V.J.N., M.O., C.S., and N.R. acquired the data. N.T., O.P.-R., M.S., A.S., N.R., A.F.H.P., and A.K. designed the experiments. N.T. and O.P.-R. wrote the manuscript. N.T. and O.P.-R. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.