|Home | About | Journals | Submit | Contact Us | Français|
The circadian timing system coordinates many aspects of mammalian physiology and behavior in synchrony with the external light/dark cycle. These rhythms are driven by endogenous molecular clocks present in most body cells. Many clock outputs are transcriptional regulators, suggesting that clock genes primarily control physiology through indirect pathways. Here, we show that Krüppel-like factor 10 (KLF10) displays a robust circadian expression pattern in wild-type mouse liver but not in clock-deficient Bmal1 knockout mice. Consistently, the Klf10 promoter recruited the BMAL1 core clock protein and was transactivated by the CLOCK-BMAL1 heterodimer through a conserved E-box response element. Profiling the liver transcriptome from Klf10−/− mice identified 158 regulated genes with significant enrichment for transcripts involved in lipid and carbohydrate metabolism. Importantly, approximately 56% of these metabolic genes are clock controlled. Male Klf10−/− mice displayed postprandial and fasting hyperglycemia, a phenotype accompanied by a significant time-of-day-dependent upregulation of the gluconeogenic gene Pepck and increased hepatic glucose production. Consistently, functional data showed that the proximal Pepck promoter is repressed directly by KLF10. Klf10−/− females were normoglycemic but displayed higher plasma triglycerides. Correspondingly, rhythmic gene expression of components of the lipogenic pathway, including Srebp1c, Fas, and Elovl6, was altered in females. Collectively, these data establish KLF10 as a required circadian transcriptional regulator that links the molecular clock to energy metabolism in the liver.
In mammals, including humans, many aspects of behavior and physiology display daily oscillations. Twenty-four-hour rhythms of sleep, feeding behavior, core body temperature, hormone secretion, lipid and carbohydrate metabolism, and blood pressure are well-documented examples. These rhythms are driven by self-sustained endogenous clocks located in virtually all cells of the body and forming an integrated system (44, 48). The circadian (~24-h) system is organized hierarchically with, at the top, a master clock located in the suprachiasmatic nuclei (SCN) of the hypothalamus and oscillating with an approximately 24-h period in the absence of external time cues. Every day, this central clock is reset by light through the retinohypothalamic tract to keep the physiology synchronized to the external light/dark (LD) cycle. Circadian clocks present in the periphery are entrained by the SCN through internal synchronizers that have yet to be identified, and although they display self-sustained oscillations at the single-cell level, at the organ and systemic levels they require an intact SCN clock to remain in phase (63). Forward genetics and biochemical approaches have identified a set of core clock genes that interact through complex positive- and negative-feedback loops to form a molecular clock generating ~24-h oscillations (29). In this mechanism, the two bHLH-PAS transcriptional activators CLOCK (or NPAS2 in some extra-SCN and peripheral tissues) and BMAL1 heterodimerize and transactivate the Per1, Per2, Cry1, and Cry2 genes through binding to E-box DNA response elements located in their promoters. PER and CRY proteins then associate and translocate to the nucleus to repress their own genes by inhibiting CLOCK-BMAL1-dependent transcriptional activity (50). This core feedback loop is the subject of an extensive and increasingly complex posttranslational control through which oscillations are critically sustained and adjusted to a period length of ~24 h (12). Notably, PER1 and PER2 are phosphorylated by the casein kinase isoforms CKI and CKIδ, leading to the subsequent recruitment of the ubiquitin ligase adapter F-box protein bTrCP, followed by proteasomal degradation. Additional CKI substrates include CRYs and BMAL1 proteins, while degradation of CRY1 and CRY2 appears to be regulated through the F-box protein FBXL3, that of BMAL1 is dependent upon SUMOylation (9, 13). Further, the circadian function of BMAL1 requires its acetylation by its heterodimerizing partner, CLOCK, a reaction that is reversed by the action of the NAD+-dependent deacetylase SIRT1 (1, 22). Additional stabilizing loops based on transcription factors of the Rev-erb, ROR, PPAR, and Dec families have also been identified recently (29, 56). This molecular clockwork mechanism ultimately generates oscillations in gene transcription of multiple clock-controlled genes (CCGs) in a highly tissue-specific manner, thereby providing timing information for regulating specific physiological outputs.
Emerging evidence suggests a close link between the circadian clock system and metabolic homeostasis. Epidemiological studies indicate that shift workers are more predisposed to elevated triglyceride and HDL cholesterol levels and obesity than day workers (27). Importantly, the incidence of cardiovascular disease appears to be higher at specific times of the day (4, 18). Clock mutant mice show adipocyte hypertrophy, dyslipidemia, and hepatic steatosis, which are hallmarks of the metabolic syndrome (57). Similarly, mice carrying a null mutation for Bmal1, another key clock gene, display reduced glucose tolerance and increased fat mass (32, 49). Conversely, animals fed with a high-fat diet show altered clock and clock-controlled gene expression patterns in the hypothalamus and in the periphery, as well as perturbed rhythms of locomotor activity and photic response (30, 38).
Genome-wide approaches analyzing circadian gene and protein expression in the liver have revealed that up to 15 to 20% of the transcriptome and proteome is under circadian control (39, 43, 47, 59). Notably, a number of transcriptional regulators are clock controlled, suggesting that a significant part of the circadian genetic network is indirectly or not exclusively regulated by the core clock mechanism. The three PARbZIP transcription factors DBP, HLF, and TEF regulate the circadian expression of xenobiotic metabolizing enzymes in liver (11). Circadian oscillation of several liver-specific mRNAs is also controlled by DEC1 (16). Numerous nuclear receptors are also putative candidates for molecular links between the circadian clock and physiological outputs (62). Along the same lines, PGC-1, a key nuclear receptor coactivator regulating energy metabolism, was recently demonstrated to be a clock target controlling the rhythms of body temperature, metabolic rate, and the expression of genes involved in energy metabolism (35). Krüppel-like transcription factors (KLFs) represent another family of regulators that have been implicated in key biological processes, including cell differentiation and proliferation, apoptosis, and glucose metabolism, but not in circadian physiology (45). Here, we show that one member, KLF10 (10) (also known as Tieg1), previously shown to be a regulator of bone physiology (46, 53), is a circadian-clock-controlled transcription factor regulating genes implicated in glucose and lipid metabolism in liver.
Clock-deficient Bmal1−/− mice in the C57BL/6J background (kindly provided by C. A. Bradfield, University of Wisconsin, Madison, WI) were used at 8 to 12 weeks of age, before they developed arthropathy (5, 6). Klf10−/− mice were backcrossed for 10 generations into the C57BL/6j background. Heterozygous animals were crossed to generate knockout and wild-type littermates. We noticed that following the extensive backcrossing into the C57BL/6j background, female and male Klf10−/− animals had reduced body weights at all ages compared to their wild-type littermates (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm) (analysis of variance [ANOVA], P < 0.001). The animals were housed in a 12-h light/12-h dark cycle (light/dark, 12:12) in a temperature- and humidity-controlled environment and fed ad libitum. Zeitgeber time zero (ZT0) referred to lights on. For the constant-darkness experiment, light was kept off at circadian time zero (CT0) for 24 h. For the fasting/refeeding experiments, 6-month-old mice were randomly divided into three groups, which were either fed, fasted, or refed following fasting. The refed group was fasted for 28 h and then refed with a high-carbohydrate/low-fat diet containing 72.2% of its calories from carbohydrates (Safe) for 24 h prior to the study. The starting times for the experiments were staggered so that all mice were sacrificed at ZT4. All experiments were conducted in compliance with the CNRS guidelines for animal ethics.
To generate the KLF10 expression vector, total RNA from mouse liver was retrotranscribed and PCR amplified using the primers 5′-GGATCCCTCAACTTCGGCGCTTCTC-3′ (forward) and 5′-ATTCTGCGTCTCCATCTTCTG-3′ (reverse), and the resulting fragment was cloned into the pcDNA 3.0 expression vector (Invitrogen). The pcDNA3-CLOCK, pcDNA3-BMAL1, and pcDNA-CRY1 expression vectors have been described previously (16). A genomic fragment encompassing the −558 to +46 (relative to the transcription initiation site) region of the proximal mouse Klf10 promoter was amplified with the primers 5′-GGTTCATCCATCCCTTGCTTC-3′ (forward) and 5′-GCACTGAGACACTAGACGTCG-3′ (reverse) and cloned into the pGL3-Basic vector (Promega) to create the Klf10prom::Luc reporter construct. The 3′-deleted Klf10 promoter construct Klf10Δ233/521::Luc was obtained by PCR amplification of the promoter sequence extending from −558 to −332 from Klf10prom::Luc vector with the forward primer mentioned above and the reverse primer 5′-CAAAGCCTAGTCGCCGC-3′. The PCR product was then inserted upstream of the −42 to +46 region of the proximal Klf10 promoter. The Klf10(E)3x::Luc and Klf10(Emut)3x::Luc reporter constructs contained three copies of the wild-type or mutated Klf10 E-box element inserted upstream of a TATA box-containing pGL2 reporter plasmid, respectively. These two reporter constructs were made by annealing the following phosphorylated oligonucleotides: 5′-GATCTCCACCCCCCACGTGGGGCCGGCTCTCCG-3′ (forward) and 5′-GATCCGGAGAGCCGGCCCCACGTGGGGGGTGGA-3′ (reverse) for the wild-type and 5′-GATCTCCACCCCCCGTATTGGGCCGGCTCTCCG-3′ (forward) and 5′-GATCCGGAGAGCCGGCCCAATACGGGGGGTGGA-3′ (reverse) for the mutated E box. A synthetic DNA fragment (Epoch Biolabs) spanning the −622 to +7 region of the mouse Pepck promoter was inserted into the KpnI-XhoI sites of pGL3basic to generate the Pepck::Luc reporter vector. The 5′ and internal promoter deletion constructs PepckΔ253::Luc, PepckΔ42/325::Luc, PepckΔ514::Luc, and PepckΔ282/514::Luc were generated by digesting the Pepck::Luc vector with either KpnI and NdeI, PstI, KpnI and StuI, or StuI, followed by fill-in with Klenow when necessary and religation. All the constructs were verified by sequencing. NIH 3T3 and HepG2 cells were cotransfected for 6 h with reporter (20 ng) and expression (1 to 100 ng) vectors using Lipofectamine 2000 (Invitrogen). The cells were incubated for 48 h in fresh medium, and luciferase activity was determined using the Bright-glo luciferase assay reagent (Promega) and a Centro LB960 luminometer (Berthold). Luciferase activities were normalized to total protein concentrations. All transfection experiments were performed in triplicate and repeated at least three times.
Liver samples were collected at ZT12, cross-linked with 1% formaldehyde for 10 min, lysed in SDS lysis buffer, and fragmented by sonication. Chromatin was then precipitated with anti-BMAL1 (ab3350; Abcam)- or anti-KLF10 (53)-specific antibodies. Immunoprecipitated chromatin samples were amplified by PCR using the following primers: Klf10, 5′-AGTGGTTCATCCATCCCTTG-3′ (forward) and 5′-CGACTAGGCTTTGCGGAGA-3′ (reverse); Pepck, 5′-CAACAGGCAGGGTCAAAGTT-3′ (forward) and 5′-GCACGGTTTGGAACTGACTT-3′ (reverse). PCR products were separated on a 2% agarose gel and stained with ethidium bromide.
Liver nuclei were purified from pools of 3 or 4 mouse livers (4 to 6 g) according to the procedure described by Gorski et al. (14), except that nonfat milk (20%) and a protease inhibitor cocktail (Sigma) were included in the homogenization buffer. The protein concentration was determined by the method of Bradford. KLF10 was detected in 80 μg of liver nuclear extracts using standard immunoblotting procedures and a rabbit anti-KLF10 polyclonal antibody at a dilution of 1:333 (Eurogentec). Control for equal loading was done by probing the membrane with a monoclonal antibody raised against the rat lamin A/C (sc-7293; Santa Cruz Biotechnology Inc.), a nuclear envelope protein whose expression is not regulated by the circadian clock.
Measurements were performed with a Light Cycler 1.5 (Roche Applied Science) using SYBR green I dye detection according to the manufacturer's recommendations. cDNA, synthesized from 2 to 5 μg of total RNA using random primers and Superscript II (Invitrogen), was added to a reaction mixture (Faststart DNA SYBR green I; Roche Diagnostics) with appropriate primers at 0.5 μM each. The relative mRNA abundance was calculated using a standard-curve method. Expression levels were normalized to the levels of the constitutively expressed 36B4 ribosomal protein mRNA. The sequences of the primer pairs used can be found at http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm.
Livers from 10-week-old wild-type and Klf10−/− male mice were removed at ZT15, and three pools of three livers per genotype were used to isolate total RNA with a nucleospin RNA L kit (Macherey-Nagel). cDNA synthesis, biotin labeling of cRNA, and hybridization on GeneChip Murine Genome 430A 2.0 arrays (Affymetrix) were performed according to the manufacturers' recommendations at the genomic core facility of the Research Institute for Biotherapy (Montpellier, France). Data analysis was performed with ChipInspector software (Genomatix), which uses a single probe rather than the classical probe set approach. Differentially expressed transcripts were identified using a false-discovery rate of 1%, a probe coverage of 3, and a minimal change of 1.5-fold. Functional categorization using gene ontology (GO) terms was done using Bibliosphere software (Genomatix).
NIH 3T3 and HepG2 cells were from ATCC and were grown under standard conditions. Primary hepatocytes were isolated by an in situ perfusion procedure using purified Liberase (Roche Diagnostics). The cells were plated at 2.5 × 104 cells/cm2 in collagen-coated 35-mm culture dishes in William's E medium supplemented with 10% (vol/vol) fetal calf serum, 100 IU penicillin, 100 mg/ml streptomycin, and 20 nM bovine insulin at 37°C in 5% CO2. After cell attachment, the medium was renewed without fetal calf serum and supplemented with 10 nM dexamethasone and 20 nM insulin. For experiments testing the effects of glucose, cells were incubated in low (5.5 mM)- or high (25 mM)-glucose Dulbecco's modified Eagle's medium (DMEM) for 24 h. Glucose production was measured 2 and 24 h after incubation of hepatocytes in glucose-free DMEM supplemented with 16 mM lactate and 4 mM pyruvate. Glucose release in the medium was determined by the GOD-POD colorimetric method (Sclavo Diagnostics International) and normalized to the protein concentration.
Food was removed at ZT0, and blood samples were collected at ZT4 in order to obtain postprandial values. Blood samples were collected from the retro-orbital venous plexus, and serum or plasma was separated by centrifugation for 20 min at 3,000 rpm. Glucose was determined using an Accu-Check glucometer (Roche Diagnostics, France). The plasma d-3-hydroxybutyric acid concentration was determined using commercial kits (R-Biopharm). Total cholesterol, triglyceride, and free fatty acid concentrations were determined by enzymatic assays with the use of commercially available reagents. Hepatic glycogen was determined by the amyloglucosidase method (GAHK-20; Sigma) to measure glucose release. The serum insulin concentration was measured with an enzyme-linked immunosorbent assay (ELISA) kit from Linco Research.
Circadian rhythmicity was evaluated by the Cosinor method using Sigma Plot software. Differences of body weight curves between wild-type and Klf10−/− mice were tested using a two-way analysis of variance (ANOVA). Student's t test was used to evaluate differences between experimental groups and to perform comparisons between wild-type and Klf10−/− mice in each experimental group. A P value of <0.05 was considered significant. All data are reported as means plus standard errors of the mean (SEM).
Microarray data are available at EBI ArrayExpress (http://www.ebi.ac.uk/microarray-as/ae/) under accession number E-MEXP-2089.
Genome-wide analyses of circadian gene expression in peripheral tissues have identified multiple transcriptional regulators as clock targets, indicating that indirect mechanisms linking clock genes to rhythmic physiological outputs may be a common strategy for regulating tissue-specific rhythmic transcriptional networks. Several members of the Krüppel-like family of transcription factors, including Klf9, Klf10, and Klf13, were identified as putative clock-regulated genes in gene expression profiling experiments performed in liver (15, 43, 58). Independently, we report here that Klf10 (also known as Tieg1) displays a robust rhythmic gene expression pattern in mouse liver with peak and trough values at ZT12 and ZT20, respectively (Fig. (Fig.11 A) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). A similar profile, but with a lower amplitude, was also observed in long bone, kidney, and skeletal muscle, but not in epididymal fat tissue and heart (Fig. (Fig.1A).1A). The phase of Klf10 was significantly advanced in bone, kidney, and skeletal muscle compared to liver (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Using highly purified liver nuclei, we observed significantly higher KLF10 protein expression at ZT16 than at ZT4, consistent with the oscillation of the mRNA (Fig. (Fig.1B).1B). To determine whether this oscillation was controlled by an endogenous clock, we first analyzed Klf10 mRNA expression in the livers of mice kept in constant darkness. High-amplitude oscillation of the Klf10 mRNA was observed under these conditions, as in the LD cycle (Fig. (Fig.1C)1C) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Next, we analyzed in liver the phase relationship between Klf10 and the clock genes Bmal1 and Cry1, which encode core components of the positive and negative limbs of the molecular clock, respectively. The results show that Bmal1 maximal expression levels occur approximately 6 h before the rise of the Klf10 mRNA, while Cry1 expression increased when the Klf10 mRNA declined, consistent with a model in which Klf10 transcription would be directly regulated by core clock proteins (Fig. (Fig.1D)1D) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). To further test genetically whether Klf10 is controlled by the molecular clock, we analyzed its expression in the livers from Bmal1 knockout mice, a clock-deficient model that exhibits a profound alteration of the circadian system (6). Expression of Klf10 was dramatically reduced and arrhythmic in both male and female Bmal1 knockout mice compared to their wild-type littermates, indicating that BMAL1 is an upstream positive regulator of Klf10 expression (Fig. (Fig.1E)1E) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Interestingly, Klf11 (also known as Tieg2), a Klf10 paralog, was also found to be regulated by the circadian clock in kidney and epididymal fat tissue (Fig. (Fig.1F)1F) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). In accordance with these gene expression data, we identified a conserved E-box response element within the mouse Klf10 promoter (Fig. (Fig.1G).1G). In vivo ChIP experiments showed that BMAL1 interacted physically with the region of the Klf10 promoter containing this site (Fig. (Fig.1H).1H). A luciferase reporter construct driven by a 604-bp promoter fragment containing this element was significantly stimulated by the CLOCK-BMAL1 heterodimer in a cotransfection assay (Fig. (Fig.1I).1I). As expected, the CLOCK-BMAL1-dependent activation of the Klf10 promoter could be totally antagonized by CRY1, a strong repressor of the core loop in the circadian clock mechanism. Notably, KLF10 itself was able to repress the CLOCK-BMAL1-dependent activation, but not the basal activity, of its own promoter. However, this effect was not mediated through the Klf10 E-box element, as a 288-bp 3′ deletion leaving the element intact was still responsive to CLOCK-BMAL1, but not to KLF10 (Fig. (Fig.1I).1I). This is consistent with the presence of a GC-rich element within the deleted region, as this type of sequence has previously been shown to be recognized by KLF10 (25). The Klf10 E-box element cloned in front of a luciferase reporter driven by a minimal promoter was sufficient to confer responsiveness to the CLOCK-BMAL1 heterodimer, even in the presence of KLF10, as expected from the deletion experiment (Fig. (Fig.1J).1J). In contrast, no transcriptional activation was seen with a mutated version of the same element (Fig. (Fig.1J1J).
Taken together, these gene expression and biochemical data demonstrate that the transcriptional regulator KLF10 is a direct circadian clock target and suggest a mechanism through which KLF10 negatively regulates its own transcription.
To get more insights into the physiological role of KLF10 in the liver, we compared the mRNA profiles of wild-type and Klf10−/− mice using high-density Affymetrix GeneChip microarrays interrogating 14,000 genes. Using a single-probe approach combined with SAM (significant analysis of microarray) statistical analysis, we identified 158 transcripts that showed at least a 1.5-fold change between the two genotypes, with a majority being downregulated in the Klf10−/− animals (Fig. (Fig.2A)2A) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). A GO-based statistical analysis of the biological processes represented in this data set revealed significant enrichment for genes controlling lipid metabolism (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Taken together, genes annotated with metabolic-process-related GO terms and those involved specifically in the control of lipid or carbohydrate metabolism represented 36% and 23.4% of the KLF10-regulated genes, respectively (Fig. (Fig.2A).2A). Merging this data set with those that previously investigated circadian gene expression in the liver indicated that 67 out of 158 KLF10-regulated genes are clock-controlled genes (Fig. (Fig.2B)2B) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm) (39, 43, 52, 58, 61). This subset of KLF10-regulated genes was also enriched for genes regulating lipid and carbohydrate metabolism (Fig. (Fig.2B).2B). Analysis of the phase distribution of the KLF10-regulated CCGs showed that 29 genes have their peaks of expression at CT10 to CT18 and are therefore potentially directly regulated positively or negatively by KLF10 (Fig. (Fig.2C2C).
Notably, many genes encoding components of the lipogenic pathway, including the ATP citrate lyase (Acly), acyl-coenzyme A (CoA) synthetase short-chain family member 2 (Acss2), malic enzyme 1 (Me1), fatty acid synthase (Fas), elongation of very long-chain fatty acid-like 5 and 6 (Elovl5 and Elovl6), stearoyl-coenzyme A desaturase 1 (Scd1), and thyroid hormone-responsive SPOT14 (Spot14), were downregulated in Klf10−/− animals (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Similarly, we noticed that genes coding for the glycolytic enzyme pyruvate kinase (L-Pk), as well as glycerol-3-phosphate-metabolizing enzymes (Gpam and Gpd1) involved in the synthesis of triglycerides, were also downregulated. Conversely, the phosphoenolpyruvate carboxykinase (Pepck) gene, which codes for a rate-limiting-step enzyme of gluconeogenesis, was upregulated (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Genes involved in the regulation of fatty acid utilization, including lipoprotein lipase (Lpl), fatty acid transporter (Cd36), and carnitine palmitoyl transferase (Cpt1a), were also altered in Klf10−/− mice (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm).
The results from this global analysis of the liver transcriptome in Klf10−/− mice suggest that a principal function of KLF10 may be to regulate hepatic metabolism through lipogenesis, glycolysis, and gluconeogenesis. The high proportion of clock-controlled genes among KLF10-regulated metabolic genes further suggests that KLF10 contributes to the circadian regulation of these metabolic pathways.
To extend our analysis of the KLF10-deficient mouse model, we compared postprandial values for blood metabolic parameters in 6-month-old wild-type and knockout animals. The results indicate that Klf10−/− male mice had approximately 20% higher blood glucose levels than wild-type animals (Table (Table1).1). Klf10−/− female mice were normoglycemic but exhibited a 20% increase of plasma triglycerides compared to their wild-type littermates (Table (Table1).1). Insulin, free fatty acids, and total cholesterol were not significantly changed in the mutant animals. These data indicate that the loss of KLF10 resulted in discrete and gender-specific metabolic abnormalities, suggesting a modulatory role of KLF10 in metabolic homeostasis.
The elevated postprandial glycemia seen in the Klf10−/− male mice prompted us to examine the fasting response in these animals. After a 28-h fast, plasma glucose was significantly decreased in both males and females, irrespective of the genotype (Fig. (Fig.33 A). However, plasma glucose levels remained significantly higher in fasted Klf10−/− males than in wild-type animals, while no difference was observed between genotypes in females. Determination of β-hydroxybutyrate indicated that the ketogenic response to fasting was normal in the Klf10−/− males (Fig. (Fig.3A)3A) and females (not shown). Similarly, Pepck was equally induced after fasting in Klf10−/− and wild-type males (Fig. (Fig.3B)3B) and females (not shown). Pepck mRNA expression displayed a robust oscillation in wild-type male mice, with a peak at ZT12 (Fig. (Fig.3C,3C, right) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Consistent with the results shown in Fig. Fig.3B,3B, we observed higher expression at ZT8 to ZT16 and lower levels at ZT0 in fed Klf10−/− male mice, resulting in a 2.1-fold-increased amplitude of the oscillation in the mutant animals (3.08 ± 0.39 versus 1.46 ± 0.28) (Fig. (Fig.3C,3C, right) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). Bmal1−/− male mice were arrhythmic, suggesting that BMAL1 regulates Pepck directly or via other factors in addition to KLF10. Compared to males, wild-type females displayed a significantly lower amplitude of the Pepck oscillation. However, as in males, we observed a time-of-day-dependent upregulation of Pepck, resulting in an increased amplitude of the oscillation. No significant circadian rhythm was seen in Bmal1−/− females (Fig. (Fig.3C)3C) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). These observations strongly suggested that the Pepck gene may be a direct KLF10 target. To test this possibility, the response of a proximal 622-bp mouse Pepck promoter fragment and its deleted derivatives to KLF10 was analyzed in cotransfection assays using HepG2 cells. The results indicated that the wild-type promoter was repressed by KLF10 and that this repressive activity was retained upon deletion of the most 5′ 325 bp (PepckΔ253::Luc and PepckΔ42/325::Luc constructs) (Fig. (Fig.3D).3D). However, when the 5′ deletion was extended up to 514 bp (PepckΔ514::Luc construct) or when a 232-bp internal fragment was deleted (PepckΔ282/514::Luc construct), no repression could be observed, indicating that the region containing nucleotides −297 to −108 mediates the KLF10-dependent repression of the Pepck gene. In the sequential deletion analysis of the Pepck promoter, we identified two potential KLF10 binding sites that contain GC-rich sequences resembling Sp1 binding sequences. Similar GC-rich sequences have previously been identified as potential binding sites for KLF10 (25). In vivo ChIP assays corroborated these transfection data by showing that KLF10 was recruited in liver by the Pepck promoter region containing these sites (Fig. (Fig.3E3E).
To further determine whether the male-specific hyperglycemia was of hepatic or extrahepatic origin, we compared the glucose production levels in primary hepatocytes from wild-type and Klf10−/− mice in an in vitro assay. The results indicated that following incubation for 24 h in the presence of pyruvate, Klf10−/− hepatocytes from male mice produced significantly more glucose then their wild-type controls, whereas no difference was observed between female mice from the two genotypes (Fig. (Fig.3F).3F). These results together suggest a causal link between the loss of KLF10, the upregulation of Pepck, and an increased glucose output in males. Although they display a similar phenotype at the molecular level, females most likely maintain glucose homeostasis in the absence of KLF10 through a compensatory mechanism, such as increased lipogenesis. As feedback regulation is a prominent mechanism in metabolic homeostasis, we also investigated whether glucose was able to regulate Klf10 expression in the liver. Klf10 mRNA expression was monitored over time in primary hepatocytes grown in the presence of either low or high glucose. The results showed that Klf10 was transiently induced up to 4- to 5-fold in cells exposed to high (25 mM) glucose, with a maximum after 2 h, and then returned to basal levels by 24 h, irrespective of the gender of the donor animals. Low (5.5 mM) glucose had no effect, in contrast to a previous observation in fibroblasts (23) (Fig. (Fig.3G).3G). These data indicate that Klf10 is a glucose-responsive gene in liver and are consistent with the negative autoregulation of the Klf10 gene suggested by the results from the promoter analysis (Fig. (Fig.1I).1I). These data together indicate that KLF10 integrates circadian and metabolic cues and controls hepatic glucose production in males through the direct regulation of Pepck.
Microarray data indicated that hepatic lipogenesis is one metabolic pathway that was altered in the absence of KLF10. Additionally Klf10−/− female mice exhibited elevated triglyceride levels. We therefore investigated the fasting/refeeding response in Klf10−/− and wild-type female mice using a low-fat/high-carbohydrate diet, as this nutritional challenge is known to induce lipogenesis. The results indicate that key genes controlling lipogenesis and glycolysis, including Acly, Fas, Elovl6, Scd1, Gpam, Me1, and L-Pk, were all normally induced in Klf10−/− animals (Fig. (Fig.44 A). Accordingly, the Pepck gene was repressed to the same extent in Klf10−/− and wild-type animals. In line with these expression data, hepatic glycogen and plasma insulin levels did not differ significantly between the two genotypes upon refeeding (Fig. (Fig.4B).4B). These data indicate that KLF10 is not required for a normal fasting/refeeding response. Importantly, analysis of Klf10 mRNA expression in fasted and refed wild-type animals revealed a 50% decrease upon refeeding (Fig. (Fig.4C).4C). This effect suggests that downregulation of Klf10 is part of the refeeding response and is consistent with the normal response seen in Klf10−/− animals.
To further explore the potential role of KLF10 in the clock regulation of lipid metabolism, as well as the basis for the female-specific increase of plasma triglycerides in Klf10−/− animals, we analyzed around the clock the liver expression of key lipogenic genes in female and male animals from the different genotypes (wild type, Klf10−/−, and Bmal1−/−). We first examined expression of Srebp1c, which is an essential transcriptional regulator of fatty acid synthesis in the liver (10). The results indicate that female mice from all three genotypes exhibited a 24-h rhythm of Srebp1c expression (Fig. (Fig.5)5) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). However, the Klf10−/− and Bmal1−/− profiles were delayed by approximately 4 to 5 h compared to that from wild-type animals (Fig. (Fig.5)5) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). A 12-h rhythm was observed in wild-type males, consistent with the recently reported 8-h ultradian rhythm for this gene using a 1-h sampling resolution (Fig. (Fig.5)5) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm; http://biogps.gnf.org/circadian/). Bmal1−/− males were arrhythmic, but Klf10−/− males, in contrast, exhibited higher levels during the dark phase, resulting in a statistically significant circadian oscillation. As the observed Srebp1c pattern seen in Klf10−/− females resembled that recently described in REV-ERBα-deficient mice (34), we analyzed the expression of Rev-erbα and its direct target Insulin induced gene 2a (Insig2a) in these animals. Unexpectedly, Klf10−/− females displayed an elevation of Rev-erbα expression at ZT8 with a concomitant reduction of Insig2a at ZT0 to ZT4 (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). This result suggests that KLF10 influences Srebp1c expression essentially independently of the REV-ERBα-INSIG2a pathway. To support this, we did not observe any increase of Srebp1c expression at ZT0 as reported in transgenic mice overexpressing REV-ERBα in the liver (34). We next analyzed Fas and Elovl6, which are two established transcriptional targets of Srebp1c (24, 31). Fas displayed a circadian gene expression in wild-type females, a pattern that was abolished in both Klf10−/− and Bmal1−/− animals (Fig. (Fig.5)5) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). A 12-h rhythm of Fas in phase with the Srebp1 rhythm was observed in wild-type males, while Klf10−/− and Bmal1−/− animals were arrhythmic. Elovl6 was rhythmically expressed in both wild-type male and female mice (Fig. (Fig.5)5) (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). A similar pattern was also observed in Klf10−/− animals, but while the female profile was phase delayed by approximately 6 h, the male profile, in contrast, was phase advanced by approximately 3 h compared to the wild-type controls (http://www.unice.fr/ibdc/pub/Guillaumond.MCB2010.supp/index.htm). The Bmal1−/− mutation abolished the Elovl6 rhythm in both males and females. Altogether, these data demonstrate that KLF10 is required for normal rhythmic gene expression of lipogenic genes in the liver and that its deletion has a different and greater impact in females. The gender-specific impact of the Klf10 mutation on metabolic gene expression changes may be responsible, at least in part, for the elevated triglyceride levels that were specifically observed in Klf10−/− females.
Krüppel-like factors form a family of highly related transcription factors that have been involved in key cellular processes, including cell proliferation, apoptosis, and differentiation (26). In this report, we identify one member, KLF10, of this family as a circadian clock transcriptional output in mammalian liver. KLF10 was initially identified as a primary transforming growth factor beta (TGF-β)-inducible early gene in human osteoblasts (54). Its function was further described as a mediator of the antiproliferative effects of TGF-β in the exocrine pancreas and an estrogen target in bone cells (20, 55). We also report here that Klf10 gene expression is acutely regulated by glucose in the liver. This argues for a pleiotropic role of KLF10 in cellular and organismal biology by integrating antiproliferative, endocrine, metabolic, and circadian signals.
The direct regulation of Klf10 by the circadian clock is supported by several independent lines of evidence, including (i) a 24-h-period oscillation of KLF10 expression in the absence of external time cues, (ii) a phase relationship with components of the core clock feedback that is consistent with a direct regulatory mechanism, (iii) a loss of rhythmic expression in the absence of the essential clock transcription factor BMAL1, and (iv) functional data demonstrating the recruitment of the BMAL1 protein and the CLOCK-BMAL1 activation heterodimer by the Klf10 proximal promoter. While the molecular clocks operate in virtually all cell types, genome-wide profiling experiments have established that a hallmark of circadian gene expression in mammals is tissue specificity (43, 52). Conceivably, this control can be achieved, at least in part, through tissue-specific regulators, such as Klf10, whose circadian expression is restricted to a subset of peripheral organs/tissues, with the liver showing the most robust oscillation. Another such example is the nuclear receptor PPARα, which is a clock-controlled metabolic sensor mostly expressed in the liver, where it regulates fatty acid oxidation (7, 10). Interestingly, KLF10 and PPARα influence the expression of common CCGs, suggesting that the clock regulation of physiological outputs may involve the coordinated actions of several tissue-specific circadian regulators (28). The tissue-specific phase of the Klf10 oscillation is another aspect of its circadian regulation in the periphery. Although the core clock networks driving Klf10 expression are qualitatively identical or very similar in these organs, the phase difference that we observed, for instance, between bone and liver may result from tissue-specific quantitative properties of the clock. Indeed, this has been suggested by experiments longitudinally analyzing circadian clock function across organs and may include, for instance, differences in responsiveness to internal synchronizers and/or the involvement of tissue-specific transcriptional or posttranscriptional regulators (63).
We further show that KLF10 controls a specific metabolic gene network in the liver. Notably, the loss of KLF10 changed the expression of genes involved in lipogenesis, gluconeogenesis, and glycolysis. This is the first evidence that a member of the Krüppel-like family acts as a molecular link between core clock genes and metabolic outputs. Preliminary data indicate that KLF11, a closely related KLF10 paralog that plays an important role in insulin secretion and diabetes, may also play a similar role in the kidney and epididymal white adipose tissue (40). Other transcription factors, including nuclear receptors, basic helix-loop-helix, and bZIP/PAR proteins have previously been shown to link core clock genes to specific circadian outputs (11, 16, 17, 60). Many additional putative circadian regulators of clock-controlled gene transcription have also been identified in various profiling experiments (2, 39, 42, 43, 52, 58). These observations collectively reinforce the notion that a large part, if not most, of the transcriptional regulation of mammalian clock outputs is indirect, as suggested by early work in Drosophila (37). Our findings thus extend the repertoire of circadian transcription factors to the KLF family and predict that all transcription factor gene families may have one or several members acting as first-order circadian regulators in specific organs. A majority of KLF10-regulated genes, including known CCGs, were repressed in our microarray experiment, yet KLF10 is mainly known as a transcriptional repressor and thus may regulate these genes indirectly. This indirect positive regulation is unlikely to involve a transcriptional repressor controlled by KLF10, as no such factor was identified in the data set. However, recent evidence suggests that KLF10 can also transactivate some promoters, such as that of the TGF-β1 gene (8). Alternatively, KLF10 may work as a posttranslationally regulated component in transcriptional regulatory complexes leading to either repression or activation of target genes, as recently demonstrated for KLF5 in the context of lipid metabolism in skeletal muscle (41). Finally, the expression of a number of these genes, including CCGs, may also have changed as a homeostatic response to the upregulation of a few direct KLF10 key targets. The latter possibility is supported by the fact that a small proportion of CCGs, including, for instance, Pepck, were upregulated in the mutant animals at the time corresponding to the peak of KLF10 in wild-type animals.
Several other members of the KLF family have been implicated in metabolic processes. However, KLF10 seems to play a prominent role in the regulation of lipogenesis, since many genes in this pathway have their expression affected by the Klf10 mutation and female Klf10−/− mice exhibit higher triglyceride levels than wild-type mice. Fatty acid synthesis has been shown to exhibit a robust diurnal variation, with a peak during the activity phase (21, 33). Recent work investigating the respective contributions of feeding and of the circadian clock to hepatic rhythmic gene expression using clock-deficient mice and the temporally restricted feeding paradigm has shown that oscillations of food-induced transcripts, including those from the lipogenesis pathway, are modulated and consolidated by the clock (59). Reciprocally, the robustness of the clock and its immediate downstream targets is increased by temporally restricted feeding. KLF10 itself is not directly driven by feeding, but it is a clock target that modulates the expression of food-driven metabolic genes, such as Srebp1c, Fas, and Elovl6. Hence, we propose, in line with the observations reported by Vollmers et al., that a principal function of KLF10 may be to temporally modulate the food-entrained daily oscillation of components of specific metabolic pathways, such as lipogenesis (59). The phase-shifting effect of the Klf10 mutation on the expression pattern of the palmitate elongase-encoding gene Elovl6 and its upstream regulator Srebp1c in females supports this hypothesis. The recent observation that Elovl6 disruption in mice causes obesity and hepatic steatosis is consistent with the hypothesis that misalignment of its rhythmic expression, as seen in our Klf10−/− female mice, could impair triglyceride homeostasis (36). Intriguingly, our data indicate that the role of KLF10 in metabolic gene expression is modulated by gender-specific effects, which may explain why only females display higher triglyceride levels. Indeed, clearly, the effect of the mutation was opposite (Elovl6) or significantly more limited (Srebp1c) in males, which did not show this phenotype. Such gender-specific effects have already been reported in the Klf10−/− model, which exhibits osteopenia in females and cardiac hypertrophy in males (3, 19, 46, 53). The similarity of the Klf10 gene expression profile in the livers of males and females argues against a direct effect of sex steroids on Klf10 circadian regulation in this organ and rather suggests that gender-specific cellular contexts modulate KLF10 activity. Additionally, Klf10 was shown to be an estrogen receptor β-specific target in bone, while the liver expresses mostly the estrogen receptor α subtype (20).
Our data indicate that KLF10, like KLF15, is a modulator of carbohydrate metabolism in the liver. KLF15-deficient mice display severe hypoglycemia upon fasting as a result of decreased availability of amino acid-derived substrates necessary for liver gluconeogenesis (15). In contrast, we show here that Klf10−/− males are hyperglycemic. This phenotype was correlated with a time-of-day-dependent upregulation of the gluconeogenic-enzyme-encoding gene Pepck in Klf10−/− animals fed ad libitum, irrespective of gender. Mechanistically, this effect can be explained by the direct repression of the proximal Pepck gene promoter by KLF10, as shown in this report. The Pepck upregulation was correlated with increased hepatic glucose production seen only in males, suggesting that in females, normoglycemia is maintained via a compensatory mechanism that is presumably linked to increased lipogenesis. One role of KLF10 in the liver might be to blunt the induction of the Pepck gene at the end of the light phase in animals fed ad libitum in order to prevent chronic hyperglycemia. Along the same line, the glucose regulation of Klf10 expression suggests that, additionally, it participates in a negative-feedback loop regulating Pepck expression. These observations collectively suggest that KLF10 integrates both circadian and metabolic cues to regulate a crucial gene of hepatic gluconeogenesis and that KLF10 and KLF15 play distinct and probably complementary roles in this pathway.
Remarkably, KLF10 was dispensable for a normal fasting or refeeding response and was even repressed upon high-carbohydrate-dependent induction of hepatic lipogenesis. We propose that one role of KLF10 is to temporally coordinate liver metabolism by contributing to the adequate synchronization of lipogenesis activation with the feeding phase and modulating glucose production at the end of the resting phase. It is therefore conceivable that mechanisms such as glucose-induced autorepression exist to suppress or overcome the KLF10-mediated circadian gating of metabolism when the organism needs to acutely activate lipogenesis (refeeding following fasting) or gluconeogenesis (prolonged fasting) independently of the time of day.
A growing body of epidemiological and experimental evidence indicates that circadian clock system disruption is detrimental to metabolic homeostasis, yet the precise underlying mechanisms involved remain largely unknown (27, 30, 51, 57). In this context, misregulation of key genes of lipogenesis and glucose metabolism, as observed in the Klf10−/− mice, may constitute one factor contributing to metabolic imbalance in individuals chronically exposed to circadian-cycle-disrupting conditions, such as shift workers and cabin crews. In conclusion, we have identified a novel circadian regulator that specifically links clock genes to liver metabolic physiology.
This work was supported by the University of Nice-Sophia-Antipolis, Centre National de la Recherche Scientifique, Ministère de la Recherche, ARC grants 3729 and 1032, and European Commission grants FP6 CRESCENDO LSHM-CT-2005-018652 and TEMPO LSHG-CT-2006-037543 and NIH grant DE14036 (T.C.S. and M.S.). F.G. was the recipient of a fellowship from Association pour la Recherche contre le Cancer.
Published ahead of print on 12 April 2010.