Introduction

Retinol binding protein 4 (RBP4), a specific carrier for retinol in the circulation [1], is produced and released mainly by hepatocytes (80%) [2] and adipocytes (20%) [3]. A previous study has reported that elevation of adipocyte-secreted RBP4 contributes to the development of insulin resistance in mice [4]. Clinical studies have shown that circulating RBP4 levels are increased in individuals with impaired glucose tolerance and type 2 diabetes, and correlate inversely with insulin sensitivity [4–7]. However, these associations are not supported by all studies in human, as several studies found no correlation between circulating RBP4 levels and insulin resistance, impaired glucose tolerance or type 2 diabetes [8–12]. This discrepancy suggests a limited understanding of the relationship between RBP4 and insulin resistance. Although liver is the principal source of circulating RBP4 [2], it remains unknown whether hepatic RBP4 plays a role in modulating glucose homeostasis.

The circadian clock is an intrinsic time-keeping system that enables organisms to adapt their physiological and behavioural rhythms in response to the external environment [13]. While the light-responsive central clock resides in the suprachiasmatic nucleus, individual clocks are present in the peripheral tissues [14–17]. The current molecular model for the circadian clock is based on a transcriptional feedback loop in which aryl hydrocarbon receptor nuclear translocator-like (BMAL1)/clock circadian regulator (CLOCK) heterodimers drive the transcription of cryptochrome 1 and 2 (Cry1 and Cry2) and period circadian clock 1, 2 and 3 (Per1, Per2 and Per3) by binding to the E-boxes on their promoters. In turn, the cryptochrome (CER)/period (PER) complex represses the transcriptional activity of BMAL1/CLOCK, which generates intrinsically oscillatory signals [18–20]. This core clock machinery regulates the rhythmic expression of downstream genes, largely by coordinating with other transcription factors, including their direct targets (e.g. D site-binding protein (Dbp), Pparα [also known as Ppara], and Pparγ [also known as Parg]) [19]. However, little is known about how these clock components influence RBP4 in the liver and plasma.

While recent studies strongly suggest an important role of circadian clocks in glucose homeostasis [21, 22], the underlying mechanism remains to be elucidated. Thus, we hypothesised that the levels of Rbp4 mRNA or RBP4 protein would display diurnal oscillations that would be regulated by BMAL1, and that RBP4 may be required for the regulation of glucose metabolism by the circadian clock. With this in mind, we studied liver-specific-Bmal-knockout mice, a circadian-clock-disruption mouse model that has an insulin-sensitive phenotype. In addition, we took advantage of adenoviruses that selectively inhibit the expression of Dbp or Rbp4 in the mouse liver.

Methods

Animal experiments

Bmal1 lox (also known as Arntl tm1Weit/J) were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Bmal1 lox mice were mated with the C57BL/6 J strain expressing a Cre-recombinase to generate mice with liver-specific deletion of Bmal1 (L-Bmal1 −/−), as described by Lamia et al [21]. RBP4−/−mice were generated on a C57BL/6 J-genetic background by SIDANSAI Biotechnology (Shanghai, China) using transcription activator-like effector nuclease (TALEN)-mediated gene targeting, as described by Chen et al [23]. The 8- to 10-week-old male C57BL/6 mice were purchased from Shanghai Laboratory Animal Centre (Shanghai, China). All mice were housed and bred in the animal facility at Shanghai Institutes for Biological Sciences (SIBS) at 22 ± 2°C, maintained on a 12 h/12 h light/dark cycle and fed ad libitum for at least 2 weeks before the study. Recombinant adenovirus was delivered by systemic tail-vein injection to mice. At 10–25 days after viral infection, mice were subjected to glucose and insulin tolerance tests. See electronic supplementary material (ESM) Methods for further details. All animal care and use procedures were in accordance with guidelines of the Institutional Animal Care and Use Committee of the Institute for Nutritional Sciences (INS), SIBS and the Chinese Academy of Sciences (CAS).

Plasmids

Full-length mouse cDNAs, Rbp4-luciferase (Rbp4-Luc) and shRNA plasmid constructions were cloned into the vectors indicated using standard molecular biology techniques. See ESM Methods for details. For cloning primers and oligonucleotides, see ESM Tables 1 and 2.

Adenoviruses and adeno-associated viruses

Adenoviruses encoding RBP4, retinol binding protein 4 without signal peptide (RBP4-Δ18), control green fluorescent protein (GFP), DBP, BMAL1, with Rbp4 RNA interference (RNAi), Dbp RNAi, Bmal1 RNAi and control unspecific RNAi adenoviruses were generated through homologous recombination between a linearised transfer vector pAD-Track and the adenoviral backbone vector pAD-Easy, as described by He et al [24]. Viruses were purified by the CsCl method and dialysed against PBS buffer containing 10% glycerol as described by Becker et al [25]. The Rbp4 RNAi adeno-associated virus was generated by Obio Technology (Shanghai, China).

Culture of primary hepatocytes

Primary hepatocytes were prepared from mice by collagenase perfusion techniques. See ESM Methods for further details.

Quantitative RT-PCR

Quantitative RT-PCR was used to determine the relative expression levels of mRNAs. See ESM Methods and ESM Table 3 for further details.

Antibodies and western blot

Anti-BMAL1 (ab3350) and anti-DBP (ab22824) were obtained from Abcam (Cambridge, UK). Cellular RBP4 (2947-1) was purchased from Epitomics (Burlingame, CA, USA) and plasma RBP4 antibody was a gift from Y. Liu (Institutes for Nutritional Sciences, Shanghai, China). Tissue and cell proteins were lysed in hypotonic buffer (1% Nonidet P-40, 50 mmol/l Tris-HCl, 0.1 mmol/l EDTA, 150 mmol/l NaCl, pH 7.4). See ESM Methods and ESM Table 4 for further details.

Chromatin immunoprecipitation

Chromatin immunoprecipitation assays were done as described by Sun et al [26]. See ESM Methods for further details. The primer sequences are shown in ESM Table 5.

Generation of reporter cell lines

Reporter cells and clonal lines were generated as previously described [27]. See ESM Methods for further details.

Bioluminescence recording and data analysis

A LumiCycle luminometer (LumiCycle 32, Actimetrics, Wilmette, IL, USA) was used for bioluminescence recording. The LumiCycle Analysis program version 2.53 (Actimetrics, Wilmette, IL, USA) was used to determine circadian variables, as described elsewhere [28].

Luciferase assay

The luciferase reporter assay for HEK293T cells has been described previously [29].

Statistical analysis

Results are shown as mean ± SEM. Statistical analysis was performed using one-way or two-way ANOVA or two-tailed unpaired Student’s t test where appropriate. Differences were considered statistically significant at p < 0.05. All experiments were performed on at least two independent occasions.

Results

Hepatic BMAL1 deficiency dampens Rbp4 expression

To explore the daily changes in Rbp4 expression, we analysed the mRNA levels of Rbp4 and Bmal1 in the livers of mice killed at 4 h intervals around the clock (Fig. 1a, b). The accumulation of Rbp4 mRNA and RBP4 protein displayed robust diurnal rhythms in the livers of wild-type (WT) mice, peaking around the onset of night (zeitgeber time [ZT] 12, Fig. 1a, c). Interestingly, plasma RBP4 levels also oscillated across the 24 h light/dark cycle, but with an 8 h phase advance compared with its hepatic expression (Fig. 1c). Furthermore, liver-specific knockout of Bmal1 significantly attenuated the amplitude of Rbp4 mRNA and protein rhythms in the liver and plasma without altering their phases (Fig. 1a–c). In contrast, the oscillation of Rbp4 mRNA in white adipose tissue was little affected in L-Bmal1 −/− mice (ESM Fig 1a, b). Moreover, the damped oscillation of RBP4 in L-Bmal1 −/− mice was restored to full-blown rhythmicity in the liver and plasma by ectopic expression of hepatic BMAL1 (Fig. 1d–f). Consistently, Bmal1 RNAi-mediated knockdown markedly attenuated the amplitude of the oscillation of Rbp4-Luc in NIH3T3 cells, with little effect on its period as monitored by real-time bioluminescence imaging (Fig. 1g). However, to our surprise, the results of the chromatin immunoprecipitation (ChIP) assay showed no presence of BMAL1 on the Rbp4 promoter (ESM Fig. 1c, d).

Fig. 1
figure 1

BMAL1 regulates the rhythmic oscillation of Rbp4 in the liver. L-Bmal1 −/− and WT littermate mice were killed at 4 h intervals across the 24 h light/dark cycle. Quantitative PCR analysis of relative mRNA levels of (a) Rbp4 and (b) Bmal1 in the liver. Values were normalised to Gapdh and are presented relative to the lowest WT control, set at 1. Black diamonds, WT; white circles, L-Bmal1 −/−. Grey shading and dark bars indicate dark periods (n = 3–4). *p < 0.05 and **p < 0.01 with two-way ANOVA. (c) Immunoblotting, with the antibodies indicated, of liver homogenates or plasma at the ZTs indicated. WT (injected with Ad-gfp), L-Bmal1 −/− (injected with Ad-gfp) and L-Bmal1 −/− (injected with Ad-Bmal1) mice were killed at 4 h intervals across the 24 h light/dark cycle. (d) Immunoblotting, with the antibodies indicated, of liver homogenates or plasma at the ZTs indicated. (n = 3–4). Quantitative PCR analysis of relative (e) Rbp4 and (f) Dbp mRNA levels in the livers. Black diamonds, WT (GFP); white circles, L-Bmal1 −/− (GFP); black circles, L-Bmal1 −/− (BMAL1). (g) Bioluminescence rhythm of Rbp4-Luc collected from NIH3T3 cells transfected with pGE1 (GeneErase shRNA expression vector, see ESM Methods for further details) or Bmal1 RNAi. Period and relative amplitude are presented. n = 3; black lines, pGE1; grey lines, Bmal1 RNAi . **p < 0.01 with Student’s t test. Data are presented as mean ± SEM

Full size image

DBP transactivates Rbp4 in the liver

Further analysis of the Rbp4 promoter sequence showed homology of the nucleotide sequence located from −132 to −809 in the Rbp4 gene with the consensus DBP-binding site (DBS; ESM Fig. 2a) required for the recruitment of transcription factor DBP; hepatic expression of the DBP was abolished in L-Bmal1 −/− mice (ESM Fig. 2b). We then tested the ability of DBP to drive Rbp4 transcription through these putative DBSs. Overexpression of DBP significantly stimulated Rbp4-Luc activity in a dose-dependent manner (Fig. 2a), whereas deletion of the putative DBSs abolished DBP-induced activity (Fig. 2b). ChIP analysis confirmed the presence of DBP on these DBSs in the Rbp4 promoter (Fig. 2c). Consistently, we observed that the amplitude of the oscillation of Rbp4-Luc was greatly attenuated by Dbp RNAi (ESM Fig. 2c) in NIH3T3 cells by using real-time bioluminescence imaging. Moreover, adenoviral expression of Dbp (Ad-Dbp) significantly enhanced the expression of RBP4 in primary hepatocytes. Conversely, Dbp RNAi adenovirus (Ad-Dbpi) decreased Rbp4 mRNA levels, which led to a reduction in levels of both expression and secretion of RBP4 protein in primary hepatocytes (ESM Fig. 2d-f). In vivo investigation revealed that Ad-Dbpi significantly decreased the amplitude of RBP4 protein and mRNA oscillation in mouse liver (Fig. 2d–f). Furthermore, adenoviral overexpression of Bmal1 or Dbp was sufficient to restore both the levels and secretion of RBP4 in Bmal −/− primary hepatocytes, suggesting that DBP is involved in the mechanism by which BMAL1 regulates RBP4 (Fig. 2g).

Fig. 2
figure 2

Rbp4 transcription is regulated by DBP in the liver. (a) Relative luciferase activity of Rbp4-Luc in the presence of 30 ng and 100 ng doses of DBP in 293 T cells. Values were normalised to β-gal and are presented relative to control, set at 1; **p < 0.01 with Student’s t test. (b) Relative luciferase activity of Rbp4-promoter mutants in the presence of vector expressing DBP in 293 T cells. WT, wild-type Rbp4-promoter; ΔDBS1 and ΔDBS2, Rbp4 promoter with a deletion of the putative DBP-binding site 1 (ATTATGTTTT) and 2 (GTTACGAAAG); respectively. White bars, plasmid cytomegalovirus promoter (pc)DNA; black bars, DBP; **p < 0.01 with Student’s t test. (c) ChIP analysis showing the relative recruitment of DBP to DBS1 and DBS2 on the promoter of Rbp4 in mouse primary hepatocytes synchronised by dexamethasone (1 μmol/l) for 2 h. White bars, anti-IgG; black bars, anti-DBP; **p < 0.01 with Student’s t test. Mice injected with Ad-USi or Ad-Dbpi adenovirus were killed at 4 h intervals across the 24 h light/dark cycle. Quantitative PCR analysis of relative mRNA levels of (d) Rbp4 and (e) Dbp in liver. Black diamonds, USi; white circles, Dbpi. Grey shading and dark bars indicate dark periods (n = 3–4); *p < 0.05 and **p < 0.01 with two-way ANOVA. (f) Immunoblotting, with the antibodies indicated, of liver homogenates or plasma at the ZTs indicated. (g) Immunoblotting analysis of RBP4 proteins in WT or Bmal1-null primary hepatocytes infected with adenovirus expressing GFP, BMAL1 and DBP. Hepatocytes were synchronised as before. Relative protein levels of RBP4 are presented. Values are normalised to GAPDH or loading control and are presented relative to control, set at 1; *p < 0.05 and **p < 0.01 with Student’s t test. Data are presented as mean ± SEM. GAPDH, glyceraldehyde 3-phosphate dehydrogenase

Full size image

Hepatic RBP4 modulates glucose metabolism

To determine the effects of hepatocyte-derived RBP4 in regulating glucose homeostasis, we prepared an Rbp4 RNAi adenovirus (Ad-Rbp4i), which specifically reduced Rbp4 expression levels in the liver but not in white adipose tissue (ESM Fig. 3a,b). Compared with mice expressing unspecific RNAi (Ad-USi), levels of resting glucose (the blood glucose level in mice fed ad libitum) were significantly lower in RBP4-knockdown mice around most time points (ZT0, 4, 8, 20) except ZT12 and 16 (Fig. 3a). Consistently, Rbp4i significantly increased insulin sensitivity in mice at ZT4, whereas these effects were not observed at ZT16 (Fig. 3b, c). These two sets of data indicate that RBP4 modulates glucose metabolism mainly during resting/fasting periods (daytime for mice) when hepatic gluconeogenesis is stimulated. Indeed, analysis of gluconeogenic genes demonstrated that mRNA levels of Pck1 and glucose 6-phosphatase (G6pase, also known as G6pc) as well as protein levels of phosphorylated cAMP responsive element binding protein 1 (CREB) and phosphoenolpyruvate carboxykinase 1, cytosolic (PCK1) were decreased by RBP4 knockdown around ZT4, but not ZT16 (Fig. 3d–g). We further confirmed these results by delivering Rbp4i via an adeno-associated virus (Rb4i), which caused very mild staining on the immunoblot (ESM Fig. 4).

Fig. 3
figure 3

Temporal effects of hepatic RBP4 knockdown on glucose metabolism. WT mice were injected with adenovirus expressing Rbp4i or USi control. (a) Resting glucose levels were measured around indicated ZTs (n = 12–14). (b) GTT and (c) ITT were performed around ZT4 and ZT16 (n = 12–14). Mice were fasted for 16 h before GTT and 2 h before ITT. AUCs were measured. Black diamonds and bars, USi; white circles and bars, Rbp4i; *p < 0.05 and **p < 0.01 with one-way ANOVA. (d–f) Quantitative PCR analysis of relative G6pase, Pck1 and Rbp4 mRNA levels in the liver. White bars, ad-libitum-fed USi; hatched bars, ad-libitum-fed Rbp4i mice; grey bars, fasted USi; black bars, fasted RBP4i mice; *p < 0.05 and **p < 0.01 with Student’s t test. Data are presented as mean ± SEM. (g) Immunoblotting of liver homogenates with the antibodies indicated. Mice were killed at ZT4 or ZT16, fed or fasted for 16 h (n = 3–4). Values were normalised to Gapdh and are presented relative to fed control, set (ZT4) at 1. Ad-lib, ad-libitum-fed

Full size image

To explore the possibility that intracellular hepatic RBP4 may be implicated in the regulation of glucose homeostasis, we prepared an adenovirus expressing Rbp4 with the deletion of secretory signal peptide (Ad-Rbp4-Δ18). Injection of this virus significantly increased resting blood glucose levels in mice around ZT0 and ZT4, but had little effect during the rest of the day (ESM Fig. 5a). Consistently, glucose tolerance was impaired in mice expressing hepatic Ad-Rbp4-Δ18, compared with those injected with control GFP adenovirus, at around ZT0 (ESM Fig. 5b). In parallel with these results, the mRNA levels of gluconeogenic genes Pck1, Pgc1α (also known as Ppargc1a) and Fbp1 were also significantly increased in the livers of Ad-Rbp4-Δ18-injected mice at ZT0 (ESM Fig. 5c–f). Meanwhile, protein levels of PCK1 and phosphorylated CREB were increased by Ad-Rbp4-Δ18 in the liver without affecting the total level of CREB protein at ZT0 (ESM Fig. 5g). Importantly, injection of Ad-Rbp4-Δ18 led to significant accumulation of Rbp4-Δ18 within the liver, but had little effect on plasma RBP4 levels (ESM Fig. 5g).

DBP knockdown in the liver improves whole-body insulin sensitivity

As hepatic RBP4 participates in the regulation of insulin sensitivity, we expected DBP to also be involved in glucose metabolism. In line with the effects of Rbp4i (Fig. 3), the injection of DBP-knockdown adenovirus (Dbpi) not only lowered blood glucose levels during the daytime in mice fed ad libitum but also ameliorated insulin sensitivity around ZT4 but not ZT16 (Fig. 4a–c). Consistently, Dbpi significantly reduced the protein and mRNA levels of gluconeogenic genes in livers from 16 h fasted mice and mice fed ad libitum collected around ZT4 but, in contrast, had little effects in those around ZT16 (Fig. 4d–h).

Fig. 4
figure 4

Temporal effects of hepatic DBP knockdown on glucose metabolism. WT mice were injected with adenovirus expressing Dbp RNAi (Dbpi) or USi control. (a) Resting glucose levels were measured around the ZTs indicated (n = 10–12). (b) GTT and (c) ITT were performed around ZT4 and ZT16 (n = 8–10). Mice were fasted for 16 h before GTT and 2 h before ITT. AUCs were measured. (a–c) Black diamonds and bars, USi; white circles and bars, Dbpi; *p < 0.05 and **p < 0.01; one-way ANOVA. (d–g) Quantitative PCR analysis of relative G6pase, Pck1 Rbp4 and Dbp mRNA levels in the liver. White bars, ad-libitum-fed USi; hatched bars, ad-libitum-fed Dbpi; grey bars, fasted USi; black bars, fasted Dbpi; *p < 0.05 and **p < 0.01 with Student’s t test. (h) Immunoblotting of the liver homogenates or plasma with the antibodies indicated. Mice were killed at ZT4 or ZT16, fed or fasted for 16 h (n = 3). Values were normalised to Gapdh and are presented relative to fed control, set (ZT4) at 1. Data are presented as mean ± SEM. Ad-lib, ad-libitum-fed

Full size image

The involvement of RBP4 in BMAL1/DBP regulation of glucose homeostasis

Based on our findings, we speculate that RBP4 is involved in DBP-regulated glucose metabolism. Indeed, transduction of DBP adenovirus significantly increased the amounts of PCK1 protein in primary hepatocytes derived from WT mice, but not in those from Rbp4-knockout mice (Fig. 5a). In addition, adenoviral expression of Rbp4 restored the reduction in PCK1 protein by Ad-Dbpi in primary hepatocytes (Fig. 5b). These observations prompted us to examine the role of RBP4 in meditating the effects of BMAL1 on glucose metabolism. Consistent with our previous study [21], L-Bmal1 −/− mice had enhanced glucose tolerance compared with WT littermates during the daytime, together with decreased protein levels of G6Pase, PCK1 and phosphorylated CREB. However, hepatic expression of Rbp4 by adenoviral delivery reversed such beneficial effects of L-Bmal1 deficiency in mice (Fig. 5c–e).

Fig. 5
figure 5

Hepatic RBP4 is involved in BMAL1/DBP regulation of glucose metabolism. Immunoblotting, with the antibodies indicated, of: (a) WT or Rbp4-null mice (white bars, GFP; black bars, DBP); (b) WT mouse primary hepatocytes infected with indicated adenoviruses (white bars, USi; black bars, Dbpi). Hepatocytes were synchronised by dexamethasone (1 μmol/l) for 2 h. Relative protein levels of PCK1 are presented. Values were normalised to HSP90 or β-actin and are presented relative to control, set at 1; *p < 0.05 with Student’s t test. Glucose tolerance of WT and L-Bmal1 −/− mice before (c) and after (d) the injection of Ad-Rbp4 (n = 7–8) at ZT8. AUCs were measured. Black diamonds and bars, WT; white circles and bars, L-Bmal1 −/−; *p < 0.05 and **p < 0.01 with one-way ANOVA. (e) Immunoblotting analysis in the liver homogenates or plasma of WT and L-Bmal1 −/− mice, with or without injection of Ad-Rbp4 (n = 3). Mice were fed ad libitum and were killed at ZT8. (f) Schematic diagram depicting the circadian regulation of RBP4 by BMAL1 through DBP, which contributes to insulin resistance and glucose intolerance in mice. Data are presented as mean ± SEM

Full size image

Discussion

Previous studies have revealed that a substantial fraction (∼10%) of all liver mRNA is expressed in a rhythmic fashion, and many of the proteins are involved in glucose metabolism [30]. In this study, we found that RBP4 oscillates in a circadian manner throughout 24 h light/dark cycles, not only in the liver and plasma, but also in the white adipose tissue. We also found that this diurnal variation is dampened in liver-specific Bmal1-knockout mice. These results indicate a critical role of BMAL1 in the regulation of hepatic Rbp4 transcription, which determines its rhythm in the plasma. By using ChIP analysis and adenovirus-mediated RNA interference, we provide a mechanism through which BMAL1 drives the circadian expression of hepatic RBP4 via DBP. The liver-enriched transcription factor DBP oscillates in a daily manner under the control of BMAL1 [31, 32]. In line with previous studies [32, 33], mRNA levels of Dbp peak around ZT8 (ESM Fig. 2b), which is compatible with the accumulation of Rbp4 transcripts (Fig. 1a). Our results demonstrate that DBP is a novel transcriptional activator of RBP4 in the liver.

Interestingly, our data show that the rhythm of plasma RBP4 displays an 8 h phase advance compared with its hepatic expression (Fig. 1c). Based on the important role of diet in retinol supply, we speculate that this phase advance could be due to dietary retinol uptake control by feeding cycle. In contrast to the role of RBP4 in the transport of hepatic retinol to plasma during resting/fasting periods (daytime for mice), postprandial retinol is mainly mobilised to peripheral tissues via chylomicrons. Therefore, the oscillation of serum RBP4 peaks in the daytime and reaches a nadir after eating (night-time), whereas hepatic RBP4 oscillates with a reversed phase, regulated by the circadian clock, which leads to the apparent 8 h phase lag between these two rhythms.

Although circulating RBP4 levels are consistently correlated with insulin resistance in mice [4, 34], clinical studies in human have not reached a consensus on this issue [8–12]. We have demonstrated that RBP4 displays an oscillatory pattern in the plasma (Fig. 1c), which provides an alternative interpretation for the proposed correlation. As the amount of RBP4 in the blood changes within a couple of hours, if the sampling time interval is too long or the sampling time points are different, the results could be conflicting, which is further supported by the fact that resting glucose levels are not altered in RBP4-knockdown mice around ZT12 and ZT16 (Fig. 3a). Meanwhile, insulin sensitivity is increased in RBP4-knockdown mice at ZT4, but remains unchanged at ZT16, which strongly implies an underlying circadian process in the regulation of glucose metabolism by RBP4 (Fig. 3b, c). Together, our studies emphasise the need to apply strict control to the timing of blood sampling and to consider circadian effects when interpreting results.

Compared with adipose tissue, the liver is the major site for the synthesis and secretion of RBP4 [1], as one hepatocyte expresses about five times as much Rbp4 mRNA as an adipocyte under either lean [35] or obese [36] conditions. Meanwhile, several studies have found no correlation between adipose Rbp4 mRNA expression and serum protein level [9, 11, 37, 38]. Moreover, Rbp4 mRNA amounts are selectively decreased in the liver, but not in the white adipose tissue, of insulin-sensitive L-Bmal1 −/− mice (Fig. 1a and ESM Fig. 1a). Indeed, accumulation of non-secretory RBP4 in the liver impairs glucose tolerance by inducing gluconeogenic genes (ESM Fig. 5b–g), which suggests that the metabolic actions of endogenous RBP4 within the liver also contribute to the salutary phenotype of RBP4-knockdown mice. Together, our findings highlight the importance of the hepatic RBP4 for systemic glucose metabolism.

Previous studies have reported that mice exhibit diurnal variations in insulin sensitivity: the highest insulin sensitivity of mice was found during the night and the lowest was found during the daytime [39, 40]. Mice with an L-Bmal1 −/− genotype exhibit a time-of-day-dependent insulin-sensitive phenotype [21]. While it has been shown that Glut2 (also known as Slc2a2) has lost its circadian expression in L-Bmal1 −/− mice, the mechanism by which BMAL1 exerts its regulatory function remains largely unknown. Our data support the hypothesis that the insulin hypersensitivity of L-Bmal1 −/− mice could be due to a deficit of RBP4 (Fig. 5c–e). Although DBP has been reported to be potentially involved in the regulation of cholesterol metabolism [41], its role in glucose homeostasis has not been investigated. In our study, we demonstrate that both DBP and RBP4 are required for the normal circadian changes in insulin sensitivity (Figs 3 and 4). The results of decreased hepatic and circulating RBP4 expression in DBP-knockdown mice, along with the similarity in phenotypes of RBP4- and DBP-deficient mice, suggest that DBP may exert its metabolic action through an RBP4-dependent mechanism. Indeed, the ability of DBP to induce PCK1 expression is compromised by RBP4 deficiency and the effect of DBP knockdown on PCK1 protein is abolished by RBP4 adenoviruses (Fig. 5a, b). Furthermore, hepatic overexpression of RBP4 reverses the insulin-sensitive phenotype of L-Bmal1 −/− mice and restores the expression of gluconeogenic genes (Fig. 5c–e), which strongly implicates a key role of RBP4 in the mechanism by which the circadian clock maintains glucose homeostasis (Fig. 5f).

Taken together, our results identify Rbp4 as a novel gene under circadian clock control and provide mechanistic insight into its circadian regulation. Furthermore, our results support the prominent role of hepatic RBP4 in the maintenance of metabolic control, and confirm a time-dependent association between RBP4 and insulin resistance in mice. Importantly, the time-of-day-dependent effect of RBP4 on insulin sensitivity is of potential importance in the clinical prognosis of individuals with type 2 diabetes.