Circulating blood eNAMPT drives the circadian rhythms in locomotor activity and energy expenditure

We first tested if the hypothalamic NAD⁺ levels display circadian oscillations. For this purpose, we collected the mediobasal hypothalamus (MBH), which includes the SCN, hypothalamic arcuate nucleus (ARC), and ventromedial nucleus (VMN), from lean C57BL6 mice every 4 h throughout two consecutive 12 h-light/dark cycles and measured the NAD⁺ levels in the MBH. We found a marked circadian oscillation in the MBH NAD⁺ levels, which had a major peak at zeitgeber time (ZT) 18 (mid−dark phase) and a minor peak at ZT2 (early-light phase) and reached the lowest levels at ZT6−ZT10 (mid-late light phase) (Fig. 1a). This rhythmic fluctuation persisted after 2 weeks-exposure to continuous darkness (Fig. 1b), thus confirming a circadian rhythm of hypothalamic NAD⁺ levels. NAD⁺ assay in ARC punch biopsy samples also demonstrated a circadian oscillation with a nighttime elevation (Supplementary Fig. 1). Cosinor analysis revealed the rhythm parameters (mesor, amplitude, and period) of hypothalamic NAD⁺ oscillations (Supplementary Table 1).

In mammalian cells, cellular NAD⁺ levels largely depend on the iNAMPT activity¹² and also can be affected by blood eNAMPT in the cells having low iNAMPT expression levels such as neurons and pancreatic β-cells¹⁴. We thus examined whether the circadian rhythm of hypothalamic NAD⁺ levels is in sync with the levels of iNAMPT in the MBH or eNAMPT in the plasma. NAMPT immunoblotting revealed distinct but opposite patterns of circadian fluctuations in the MBH iNAMPT and plasma eNAMPT expression levels (Fig. 1c, d). The MBH iNAMPT expression peaked in the early light phase (ZT2) and showed a trough in the early dark phase (ZT14) (Fig. 1c and Supplementary Table 1). In contrast, the plasma eNAMPT expression elevated during the dark period and decreased during the light period (Fig. 1d and Supplementary Table 1). These findings led us to speculate that a rise in MBH NAD⁺ levels during the dark period may be related to the nocturnal rise in plasma eNAMPT.

Blood eNAMPT can be derived from adipocytes, hepatocytes, and leukocytes^13,15,30,31. We thus examined the circadian fluctuations in the iNAMPT expression in epididymal white adipose tissues, the liver and leukocytes. Consistent with the previous report²⁷, hepatic iNAMPT protein expression showed circadian oscillations with a major elevation during the dark phase (Fig. 1e). This pattern of oscillation matched that of plasma eNAMPT (Fig. 1d, e). In contrast, circadian oscillation of iNAMPT expression in the epididymal adipose tissue showed the opposite pattern of circadian oscillation to those of blood eNAMPT and liver iNAMPT (Fig. 1d−f and Supplementary Table 1). In addition, a previous study has shown that eNAMPT secretion from adipocytes increases under fasting condition¹⁵, arguing against the possible contribution of adipose tissue-derived eNAMPT to the elevation of blood eNAMPT levels during the physiological feeding period. As leukocytes secrete eNAMPT and may contribute to circulating eNAMPT concentrations³⁰, we isolated blood leukocytes to study circadian rhythms in leukocyte iNAMPT expression. We assayed Nampt mRNA expression levels in leukocytes due to little protein amount. Notably, leukocyte Nampt expression levels had a dramatic circadian oscillation with a peak at ZT10 (2 h before the beginning of the dark period) (Fig. 1g and Supplementary Table 1). Taken together, the nocturnal elevation of blood eNAMPT may relate to iNAMPT fluctuations in the liver or leukocytes.

The circadian rhythms of hormonal and metabolic parameters are strongly entrained by the light-dark cycle when food is plentiful, but can shift to match food availability when food supply is restricted³². We therefore tested whether food availability and light-dark cycle entrain the circadian oscillations in plasma eNAMPT level. Mice were either maintained on a reverse light/dark cycle and ad libitum-feeding (light-shift, light-dark entrainment) or allowed access to food only during the light period (ZT0−ZT12) under a normal light/dark cycle (food-shift, food-entrainment) for 14 days (Fig. 1h). The circadian fluctuations in blood eNAMPT expression were blunted in the food-shift group but adjusted to shifted light-dark cycle (Fig. 1i). We also investigated whether altered food availability and light-dark cycle affected the circadian oscillations in hypothalamic NAD⁺/iNAMPT levels. Circadian fluctuations in hypothalamic NAD⁺ levels showed similar changes, although the nocturnal elevation in the light-shift group did not reach a statistical significance (Fig. 1j). The circadian rhythm of MBH iNAMPT expression was blunted by both manipulations, whereas hepatic iNAMPT expression showed a reverse in circadian rhythm by food shift and an adjustment to the shifted light-dark cycle (Supplementary Fig. 2). Therefore, light and food may act as important zeitgeber in the circadian fluctuations of hypothalamic NAD⁺, plasma eNAMPT, and MBH/liver iNAMPT levels.

Metabolic parameters such as LMA, EE, and food intake (FI) are regulated by the hypothalamus and display marked circadian rhythms. Considering circadian oscillations in hypothalamic NAD⁺ and plasma eNAMPT (Fig. 1a, b, d), we hypothesized that plasma eNAMPT may modulate the circadian rhythms of LMA, EE, or FI by affecting hypothalamic NAD⁺ biosynthesis¹⁵. To test this hypothesis, we examined the effect of systemic eNAMPT inhibition on LMA, EE, and FI. eNAMPT inhibition was induced by continuous infusion of FK866, a chemical inhibitor of NAMPT, for 3 days via a peritoneum-implanted osmotic pump (Fig. 2a). Intraperitoneal (i.p.) infusion of FK866 (3.3 and 8 mg/kg) markedly blunted the nighttime increases in LMA and EE in a dose-dependent manner (Fig. 2b, c), whereas it did not block elevations in food intake during the dark period (Supplementary Fig. 3). Interestingly, nocturnal FI was enhanced by FK866 treatment, suggesting that nighttime elevation in systemic eNAMPT activity might relate to satiety formation.

FK866 infusion tended to lower the NAMPT enzyme activity in the plasma, liver, and leukocytes especially during the dark period, and lowered liver NAD⁺ levels (Supplementary Fig. 4a–c and Fig. 2d). FK866 did not alter the MBH NAMPT activity throughout the circadian cycle (Supplementary Fig. 4d) but blocked a nocturnal rise in the MBH NAD⁺ levels (Fig. 2d). This data suggests that systemic inhibition of NAMPT activity can disrupt normal circadian biosynthesis of NAD⁺ in the hypothalamus.

We further tested the regulatory roles of eNAMPT on the circadian rhythms of LMA and EE with systemic administration of small inhibitory RNA targeting murine Nampt (siNampt) (Fig. 2e). Like FK866, injection of siNampt (33 nmol/kg) into the tail vein abolished the nocturnal rises in LMA and EE and increased FI especially in the dark period (Fig. 2f, g and Supplementary Fig. 5). Therefore, siNampt-induced depletion of blood eNAMPT levels recapitulated the effects of chemical eNAMPT inhibition on LMA and EE.

Systemic siNampt treatment reduced plasma eNAMPT levels by ~40% and MBH NAD⁺ levels by 80% compared with injection of non-targeting control siRNA (siControl) when measured at ZT18 (Fig. 2h, i). The liver NAD⁺ levels were also mildly decreased while the adipose NAD⁺ levels were not significantly affected by siNampt injection (Fig. 2i). Systemic administration of siNampt reduced the expression of liver iNAMPT while not significantly altering the iNAMPT protein expression levels in the MBH and epididymal fat (Fig. 2j). The leukocyte Nampt mRNA levels also tended to decrease in siNampt-injected mice (Fig. 2k). These findings suggest the possibility that liver- and/or leukocyte-derived NAMPT may contribute to the dark-period elevation in plasma eNAMPT levels.

We inversely tested the effects of artificial elevation of systemic eNAMPT activity on LMA, EE, and FI. To this end, we injected recombinant NAMPT protein intraperitoneally to normal mice at ZT5 when the endogenous plasma eNAMPT levels reach a trough (Figs. 1d, 3a). Before study, we confirmed the enzyme activity of NAMPT protein (Supplementary Fig. 6) and that i.p. injection of NAMPT (0.3 mg/kg) increased plasma eNAMPT levels at ZT6 to the level of endogenous eNAMPT at ZT18 (Supplementary Fig. 7). For comparison, we also injected NMN (120 mg/kg), a product of NAMPT enzymatic reaction.

During the light period, systemic administration of NAMPT and NMN significantly increased MBH NAD⁺ levels at 1 h post-injection by 40% and 70%, respectively (Fig. 3b). NMN injection also elevated the liver NAD⁺ levels (Fig. 3b). In contrast, NAMPT injection neither increased the hepatic NAD⁺ levels nor the hepatic NAMPT activity and protein expression levels (Fig. 3b and Supplementary Fig. 8). We also administered the same doses of NAMPT protein and NMN during the dark period and examined the changes in the MBH and liver NAD⁺ levels. The dark-period injection of NAMPT tended to increase MBH NAD⁺ levels to a lesser degree compared with light-period injections (Supplementary Fig. 9). Similarly, NMN injection during the dark period modestly elevated the NAD⁺ levels only in the liver (Supplementary Fig. 9). Therefore, the nighttime administration of NAMPT and NMN was less effective in terms of increasing the MBH or liver NAD⁺ levels compared with daytime injections.

In contrast to FK866 and siNampt, i.p. injection of NAMPT and NMN stimulated LMA and EE during the light period (Fig. 3c, d). In addition, intracerebroventricular (i.c.v.) administration of NAMPT at a more than 100 times-lower dose (100 ng) increased LMA and EE as well as hypothalamic NAD⁺ levels (Fig. 3e−h). These results suggest that increased eNAMPT activity can stimulate LMA and EE through a central mechanism. Artificial elevation in hypothalamic NAD⁺ levels by eNAMPT and NMN administration in the light phase led to increased LMA and EE, which mimicked the physiological changes in LMA and EE during the dark period. Thus, these findings support the hypothesis that circadian fluctuation in blood eNAMPT and hypothalamic NAD⁺ levels may contribute to the day-night rhythms of LMA and EE.

On the other hand, i.p.-injected NAMPT/NMN and i.c.v.-administered NAMPT suppressed fasting-induced hyperphagia (Supplementary Fig. 10). These data indicated that NAMPT/NMN-induced elevation in hypothalamic NAD⁺ levels could lead to a negative energy balance through the stimulation of LMA/EE and suppression of FI. Overall, NAMPT and NMN administration induced the opposite changes to those of FK866 and siNampt.

We studied the mechanism by which eNAMPT regulates LMA and EE in a circadian manner. We first tested the possibility that eNAMPT regulates the SCN clock activity. The SCN clock activity was assessed by bioluminescence monitoring of the PER2:luciferase activity in SCN-containing brain slices that were obtained from PER2:LUC mice (Fig. 4a). Treatment of recombinant NAMPT protein (100 nM) for 4 days did not significantly alter the period and amplitude of circadian oscillations of PER2:LUC activity when compared with those of vehicle-treated controls (Fig. 4a). In addition, i.c.v. administration of NAMPT did not significantly alter the mRNA expression of SCN clock genes (Bmal1, Clock, Period 2, and Cryptochrome 1) (Fig. 4b). These results collectively suggest that systemic eNAMPT activity does not significantly affect the SCN clock activity. In contrast, our preliminary study showed the clock activity in SCN slices was profoundly suppressed when cultured in an FK866-containing medium (Supplementary Fig. 11), supporting the importance of hypothalamic neuron iNAMPT in the SCN clock activity as previously suggested³³.

We next examined the expression of LMA/EE-regulating neuropeptides (proopiomelanocortin [Pomc], Agouti-related peptide [Agrp], neuropeptide Y [Npy]) in the ARC. The i.c.v. administration of NAMPT increased the Pomc expression but decreased the Agrp and Npy expressions in the ARC (Fig. 4b). Consistently, FK866 co-administration blocked the i.c.v. NAMPT-induced changes in Pomc, Npy, and Agrp transcript levels (Fig. 4c). These data imply that eNAMPT can modulate the hypothalamic Pomc, Npy, and Agrp expression via the mechanism dependent on the enzymatic activity of NAMPT. In contrast, i.c.v. NAMPT did not significantly alter the expression levels of SIRT1 target genes, Nkx2.1 and orexin receptor 2 (Ox2r)³⁴ in the hypothalamic dorsomedial nucleus (DMN) (Fig. 4b).

We also examined the effects of systemic eNAMPT modulation on the POMC and NPY/AgRP neuronal activity. The POMC neuronal activity was assessed by c-Fos expression in tdTomato-labeled POMC neurons at midday (ZT6) and midnight (ZT18) (Fig. 4d). The percentages of c-Fos⁺ POMC neurons were significantly different between ZT6 and ZT18 (13% vs. 28%, P = 0.04), indicating a significant circadian variation in the POMC neuronal activity according to the circadian fasting (ZT6)−feeding (ZT18) cycle (Fig. 4d). The low POMC neuronal activity at ZT6 was increased by i.p. injection of NAMPT (1 mg/kg), while the high POMC neuronal activity at ZT18 was suppressed by FK866 (10 mg/kg) injection (Fig. 4d). These changes in POMC neuronal activity were consistent with the changes in hypothalamic NAD⁺ levels (i.e., decrease by FK866 and increase by NAMPT) (Figs. 2d, 3b).

The NPY/AgRP neuronal activity, which was assessed through c-Fos expression in GFP-labeled NPY neurons, showed opposite changes to those of POMC neurons during circadian cycles and following NAMPT and FK866 injections (Fig. 4e). These data demonstrate that systemic eNAMPT activity can drive the circadian changes in POMC and NPY/AgRP neuronal activity.

We further tested the role of POMC and NPY/AgRP neurons in the eNAMPT-mediated regulation of LMA and EE. As POMC end product α-melanocyte-stimulating hormone (α-MSH) and AgRP are each endogenous agonist and antagonist of melanocortin 3 and 4 receptors (MC3/4R)³⁵, POMC/AgRP neurons are thought to regulate LMA and EE via these receptors^36–38. We thus tested if the blockade of MC3/4 R with SHU9119 can inhibit the stimulatory effects of NAMPT on LMA and EE. Prior i.c.v. injection with SHU9119 prevented the NAMPT-induced changes in those metabolic parameters (Fig. 4f). These findings imply that eNAMPT controls LMA and EE via the central melanocortin system (i.e., POMC and AgRP neurons and the MC3/4R).

NAD⁺-dependent deacetylase SIRT1 is essential for the normal functions of POMC and NPY/AgRP neurons^21,22,39. Mice lacking SIRT1 in POMC neurons are vulnerable to diet-induced obesity due to reduced EE²², and SIRT1 deletion in NPY/AgRP neurons leads to reduced food intake and lower body mass²¹. We therefore examined the circadian oscillations of hypothalamic SIRT1 expression, and immunoblotting of MBH tissue blocks revealed a marked circadian rhythm in SIRT1 expression that peaked at ZT6−10 and decreased at ZT18 (Fig. 5a and Supplementary Table 2). The opposite circadian oscillation patterns of MBH SIRT1 expression and blood eNAMPT (Figs. 1d, 5a and Supplementary Tables 1, 2) led us to search for another sirtuin as a downstream mediator of eNAMPT, especially during the dark period. In contrast to SIRT1 expression, MBH SIRT2 expression remained stable throughout the day-night cycles (Fig. 5a and Supplementary Table 2).

Double staining of Sirt2/POMC (β-endorphin) or Sirt2/NPY-GFP confirmed Sirt2 expression in both types of neurons (Fig. 5b), and in vitro promoter assay revealed that NAMPT and NMN treatment increased the POMC transcriptional activity but repressed the AGRP transcriptional activity in SH-SY5Y cells (Fig. 5c). Those effects were completely blunted by SIRT2 depletion (Fig. 5c and Supplementary Fig. 12a), thus demonstrating the importance of SIRT2 in the regulation of POMC and AGRP transcription by eNAMPT.

We further tested the role of SIRT2 in vivo by inhibiting hypothalamic SIRT2 through bilateral intra-MBH administration of siSirt2, which successfully depleted MBH Sirt2 in mice (Fig. 5d and Supplementary Fig. 12b). The depletion of Sirt2 in the MBH significantly blunted the NAMPT-induced changes in LMA, EE, and FI as well as those in the Pomc, Npy, and Agrp transcript levels (Fig. 5e, f and Supplementary Fig. 13). These findings suggested that hypothalamic SIRT2 acts as a downstream mediator for the role of eNAMPT in the modulation of LMA, EE, FI, and the expression of Pomc, Npy, and Agrp.

Given an importance of SIRT1 in normal functions of POMC and NPY/AgRP neurons^21,22,39, we additionally tested the involvement of SIRT1 in the eNAMPT regulation of POMC and AGRP gene transcription. In vitro promoter assay showed that SIRT1 knockdown markedly blocked the effects of NAMPT treatment on POMC and AGRP transcriptional activity (Supplementary Fig. 14). These findings suggest that in addition to SIRT2, SIRT1 may also mediate eNAMPT actions in the hypothalamus, and this possibility should be further tested in future studies.

We next sought to delineate the molecular mechanism by which SIRT2 regulates the transcription of POMC and AGRP. Sirtuins regulate transcription by deacetylating and regulating the activity of transcriptional factors such as FOXO1, p53, and nuclear factor of activated T-cells (NFAT)^18,39–41. Among them, FOXO1 is well known as the regulator of both POMC and AGRP transcription^42,43, and previous studies have shown a direct interaction between SIRT2 and FOXO1 in adipocytes^44,45. Sequential immunoprecipitation-immunoblotting of MBH protein extracts demonstrated the molecular interaction between FOXO1 and SIRT2 in the hypothalamus (Fig. 6a). Moreover, in N1 hypothalamic neuronal cells, Sirt2 overexpression decreased the acetylated form of FOXO1 (Fig. 6b). These data show that FOXO1 is a target of SIRT2 deacetylation in the hypothalamic neurons.

Next, we examined the functional interaction of SIRT2 and FOXO1 in the regulation of POMC and AGRP transcription. In line with previous reports⁴³, FOXO1 overexpression suppressed POMC transcription and stimulated AGRP transcription (Fig. 6c). SIRT2 alone expression increased POMC promoter activity but did not significantly alter AGRP promotor activity (Fig. 6c). Notably, SIRT2 coexpression inhibited the FOXO1’s regulation of POMC and AGRP transcription (Fig. 6c), indicating antagonizing effects of SIRT2 on FOXO1 actions. SIRT1 overexpression recapitulated the effect of SIRT2 on FOXO1 regulation of POMC and AGRP transcription (Supplementary Fig. 15) and thus, both SIRT1 and SIRT2 may inhibit the transcriptional activity of FOXO1 in POMC and NPY/AgRP neurons.

The acetylation state of FOXO1 is known to affect its DNA binding, transcriptional activity, and ubiquitin-mediated degradation^46–48. We thus tested if FOXO1 deacetylation alters the ability of FOXO1 to regulate the POMC and AGRP transcription using FOXO1 deacetylation mutant (6KR). In this mutant, six lysine residues corresponding to the proposed FOXO1 acetylation sites (K242, K245, K259, K262, K271, and K291) were replaced with arginine to prevent acetylation⁴⁶. We confirmed a reduction in FOXO1 acetylation in cells expressing FOXO1-6KR (Supplementary Fig. 16a), and the expression of FOXO1-6KR, that mimicked SIRT2-induced FOXO1 deacetylation⁴⁴, failed to regulate AGRP and POMC transcriptional activity (Fig. 6d). Moreover, in vitro expression of FOXO1-6KR led to decreased FOXO1 protein levels (Supplementary Fig. 16b), indicating that FOXO1 deacetylation promotes the degradation of FOXO1, as shown previously^46,48. Collectively, these findings suggest that SIRT2-mediated FOXO1 deacetylation represses the expression and functions of FOXO1 in hypothalamic neuron cells.

We further examined the importance of FOXO1 in the hypothalamic actions of eNAMPT in mice. I.c.v. NAMPT administration significantly reduced the total FOXO1 levels in the MBH (Fig. 6e). Prior i.c.v. injection of proteasome inhibitor MG132 attenuated the NAMPT-induced reduction in hypothalamic FOXO1 levels (Supplementary Fig. 17). Furthermore, the circadian changes in hypothalamic FOXO1 displayed an inverse relationship with circulating eNAMPT levels in normal mice (Figs. 1d, 6f). Thus, this in vivo evidence suggest that eNAMPT may promote hypothalamic FOXO1 degradation, which may occur through NAD⁺/SIRT-mediated FOXO1 deacetylation^46,48. Bilateral intra-ARC injection of siSirt2 significantly blunted the i.c.v. NAMPT-induced decrease in hypothalamic FOXO1 levels at ZT6 (Fig. 6g). Together, these results support an important role of SIRT2 in eNAMPT-induced FOXO1 degradation.

We finally tested whether hypothalamic Foxo1 overexpression can block eNAMPT actions by injecting either Foxo1-GFP-expressing adenovirus (Foxo1-Ad) or GFP alone-expressing control virus (GFP-Ad) into bilateral ARC prior to i.c.v. NAMPT injection (Fig. 6h). We confirmed successful virus injection in the ARC and increased MBH Foxo1 expression (Supplementary Fig. 18). Overexpression of Foxo1 alone in the ARC decreased EE, and hypothalamic Pomc expression levels and tended to lower LMA, but increased Agrp expression levels (Supplementary Fig. 19). Moreover, Foxo1 overexpression blunted NAMPT-induced changes in LMA and EE as well as in Pomc and Agrp expression (Fig. 6h). These data indicate that suppression of hypothalamic FOXO1 is critical for the actions of eNAMPT.

Obesity is strongly associated with disruptions in central and peripheral molecular clockworks⁴⁹. We thus examined whether circadian oscillations in liver iNAMPT−plasma eNAMPT−hypothalamic NAD⁺ axis may be altered by obesity in mice. Obesity was induced by the consumption of HFD for 20 weeks (Fig. 7a). Mice with diet-induced obesity (DIO) had disrupted circadian oscillation in MBH NAD⁺ levels, plasma eNAMPT, and liver Nampt expression levels (Fig. 7b and Supplementary Table 3). The circadian fluctuation in hypothalamic FOXO1 expression levels was reversed and that of LMA was blunted in obese mice as well (Fig. 7c, Supplementary Fig. 20 and Supplementary Table 3).

A previous human study reported a distinct circadian rhythm in plasma eNAMPT/visfatin with a peak in the afternoon, which corresponds to midnight in rodents, and a dip in the early morning⁵⁰. We thus tested whether obesity may alter the eNAMPT circadian rhythm in humans. We examined the 24-h profiles of blood eNAMPT levels in 14 healthy young (≤ 30 years-old) volunteers (7 males, 7 females) with no medical history of chronic illness or malignancy, of whom 7 were lean subjects (BMI < 23 kg/m²) and 7 were individuals with obesity (BMI ≥ 25 kg/m²). The characteristics of the subjects are shown in Supplementary Table 4. Blood samples were collected every 30 min from 6 a.m. to 10 p.m., and then hourly until 6 a.m. in the next morning (Fig. 7d). During the study, three meals were provided at 8 a.m., 12:30 p.m., and 6 p.m., and lights were turned on at 6 a.m. and turned off at 10 p.m. Frequent blood sampling showed a pronounced oscillation pattern of serum eNAMPT concentrations in lean young subjects that had 3 to 7 pulsatile oscillations during the light-on period but no definite oscillations during the light-off period (Fig. 7e). Although a distinct difference in the eNAMPT oscillation between the light-on period and light-off period, the timing of daytime peaks varied among individuals. Notably, 3 subjects (2 males and 1 female) with a BMI > 30 kg/m² showed significant oscillations during the light-off period, whereas subjects with mild obesity (BMI 25–29.9 kg/m²) had no oscillation during this period (Fig. 7e). The area under the curve (AUC) value of serum eNAMPT during the dark period was significantly higher in the subjects with obesity (BMI > 30) than in the lean subjects, although the AUC of serum eNAMPT during the light period also showed a tendency to increase in individuals with obesity (Fig. 7f). Although our data was obtained from a small number of young healthy subjects, these human data suggest that day-night difference in blood eNAMPT oscillations can be altered under chronic overnutrition conditions and this phenomenon might relate to altered day-night rhythms of physical activity in individuals with obesity^51,52.

Recombinant human NAMPT protein was obtained from Adipogen International (AG-40A-0018). We purchased FK866 from Enzo Life Sciences (ALX270-501), SHU9119 from Phoenix Pharmaceuticals (043-24) and NMN from Sigma (1094-61-7).

Male C57BL/6 mice (8–10-weeks old) were obtained from Orient Bio. Animals were housed under temperature-controlled conditions (22 ± 1 °C), humidity (55 ± 5%), and subjected to a 12-h light/dark cycle with lights on from 8 a.m. with ad libitum access to food to 8 p.m. unless indicated otherwise. Thus, 8 a.m. is zeitgeber time 0 (ZT0). Mice were fed a standard diet (Agripurina) with free access to food and water unless indicated otherwise. To induce diet-induced obesity, mice were fed a high-fat diet (60% fat; Research Diets, D12492). Mice with tdTomato-labeled POMC neurons were generated by mating POMC-Cre mice (provided by Dr. Joel K. Elmquist, University of Texas Southwestern Medical Center) with mice carrying the reporter Rosa26-loxP-STOP-loxP-tdTomato allele (Jackson Laboratory, 007909). NPY-hrGFP mice (Jackson Laboratory, 006417) were used for the purpose of AGRP neuron labeling. Dr. Joshep S. Takahashi (University of Texas Southwestern Medical Center) provided PER2:LUC knock-in mice used in the SCN clock study. All procedures were approved by the Institutional Animal Care and Use Committee of the Asan Institute for Life Sciences under the project license 2012-11-024 (Seoul, Korea).

SH-SY5Y human neuroblastoma cells (ATCC, CRL-2266) and murine hypothalamic N1 cells (Cedarlane, CLU101) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum and penicillin and streptomycin (100 units/ml each).

Plasma, liver, epididymal adipose tissue and the MBH were collected from ad libitum-fed young mice every 4 h (starting at 10 a.m.) throughout two consecutive circadian cycles. We dissected the MBH in the anterior border of the optic chiasm, posterior border of the mammillary body, ventral third from the upper border of the anterior commissure, and medial third from the lateral border of the lateral sulcus in the ventral side of the brain. Leukocytes were isolated from pooled blood samples (three mice to one sample) using the Ficoll-Paque PLUS Cytiva (Sigma, 17-1440-02). In food-shift studies, food was available only during the light period (8 a.m. to 8 p.m.) for 14 days before sacrifice. In light-shift studies, mice were subjected to a reverse light/dark cycle (lights on from 8 p.m. to 8 a.m.) with free access to food for 2 weeks. Mice were sacrificed either at 2 a.m. or 2 p.m. to collect plasma and tissues. We repeated these experiments at least twice.

Blocks of MBH, liver, and fat tissues were weighed, lysed with 150 μl of 1 M HClO₄, and neutralized by adding 50 μl of 3 M K₂CO_2, as previously described¹⁴. Small tissue blocks from 4 liver lobes were pooled to measure the liver NAD⁺ levels. After centrifugation for 15 min (4 °C, 13,000 × g), 100 μl of supernatant was mixed with 30 μl of 50 mM K₂PO₄/50 mM KHPO₄ (1:1 v/v, pH 7.0) and loaded onto the column (AHima HPC 18AQ 5 μM, 15 × 4.6 cm). HPLC was performed at a flow rate of 1 ml/min. NAD⁺ was eluted as a sharp peak at 10 min. NAD⁺ was quantitated based on the peak area compared with a standard curve and normalized to tissue wet weight. The tissue NAD⁺ levels were also measured using a NAD assay kit (Bioassay systems, E2ND-100). Homogenized tissue samples were lysed with NAD⁺ extraction buffer included in the kit and then subjected to heat extraction at 60 °C for 5 min. After centrifugation of samples, 40 μl of supernatant was used for NAD⁺ analysis.

Plasma and serum were collected from mice and humans, respectively, and stored at −70 °C until use. Mouse blood eNAMPT levels were measured using an ELISA kit (Abnova, KA3659) according to the manufacturer’s instructions. A volume of 10−20 μl of mouse plasma was diluted (1:5 or 1:10) in lysis buffer immediately before the assay. Human blood eNAMPT levels were measured using a sandwich ELISA kit (Adipogen, AG-45A-0008). A volume of 100 μl of human serum was assayed without dilution. The lower and upper limits of detection were 0.12 and 8 ng/ml, respectively. The intra- and inter-assay coefficients of variation were 5.6% and 5.9%, respectively. Measurement of NAMPT enzyme activity was performed using a NAMPT activity assay kit (Abcam, 221819) and commercial protocol. Enzyme activity was normalized by the protein amount. All samples were assayed in duplicates.

We measured circadian rhythms of LMA and EE using a Comprehensive Lab Animal Monitoring System (CLAMS, Columbus Instruments). This system measured oxygen consumption (VO₂) and carbon dioxide production (VCO₂) using an indirect calorimeter and we calculated EE by using the manufacturer’s formula. Light and feeding conditions were identical to the home cage conditions unless indicated otherwise. Mice were adapted to the CLAMS cages for 48 h before studies.

We infused FK866 to mice using an osmotic mini-pump (Alzet, 1003D) placed in the peritoneum. FK866 (3.3 and 8 mg/kg/day) was dissolved in a mixture of DMSO-saline (1:10 v/v) and infused at a flow rate of 0.5 μl/h for 3 days. Control animals received the same volume of vehicle. During infusion, mice were placed in CLAMS chambers to monitor LMA, EE, and FI. On the final day of the study, animals were sacrificed every 4 h throughout a circadian cycle to collect plasma, liver, MBH and white blood cells.

We randomly assigned animals to receive either control or Nampt small inhibitory RNA (siRNA) one day before the study to match the average body weights between the two groups. Mice received siRNA targeting murine Nampt (33 nmol/kg, resuspended in 100 μl of RNase-free PBS; Dharmacon; E-040272-00-0005) via tail vein injections at ~10 a.m. The control group received an equal amount of non-targeting, scrambled, control siRNA (Dharmacon; D-001910-10-20). After siRNA injections, we monitored LMA, EE, and FI until sacrifice at 2 a.m. (ZT18) on the 3rd day. We collected samples from white blood cells, plasma, hypothalamus, liver, and epididymal fat. Knockdown was considered successful if there was a >50% decrease in plasma eNAMPT concentrations compared with the average levels of controls at the end of the study. Only the animals demonstrating successful knockdown were included in the data analysis.

We administered NAMPT protein and NMN at indicated doses i.p. (100 μl volume) or i.c.v. (2 μl volume) to C57BL/6J mice in the mid-light phase (ZT6–ZT8). Control mice received the same volume of vehicle. After a 48-h acclimatization period, we monitored LMA, EE, and FI for 4 h following injections using CLAMS. For the i.c.v. injection study, we implanted permanent 26-gauge stainless steel cannulae into the third ventricle (1.8 mm caudal to bregma, 5.0 mm ventral to the sagittal sinus) by using a stereotaxic surgery. One week after surgery, we confirmed positioning of the cannulae by a positive dipsogenic response to angiotensin II (50 ng, i.c.v.) and then daily handled mice for 1 week to minimize stress response. We used only the mice with correctly positioned cannulae for the study. We randomly assigned mice to the control or experimental groups 1 day before the study to match the average body weights between groups. We dissolved NAMPT, SHU9119, and MG132 in 0.9% saline prior to injection and administered at a volume of 2 μl.

Preparation of organotypic SCN cultures and bioluminescence monitoring were carried out as previously reported with minor modifications⁶⁰. Neonatal (5–7 days old) PER2:LUC knock-in mice were sacrificed, and the brains were quickly obtained. The isolated brains were chilled in ice-cold Gey’s balanced salt solution (GBSS) supplemented with 0.01 M HEPES and 36 mM D-glucose, bubbled with 5% CO₂ and 95% O₂, and then coronally sectioned into 400-μm-thick slices with a vibratome (Campden Instruments). The slices were maintained on a culture insert membrane (Millicell-CM, Millipore) and placed in a culture medium (50% minimum essential medium, 25% GBSS, 25% horse serum, 36 mM D-glucose, and 1x Antibiotic-Antimycotic solution) at 37 °C. The SCN explants were cultivated for more than 2 weeks before being used in experiments. All culture media and supplements were purchased from Thermo-Fisher Scientific Co. unless specified otherwise. For bioluminescence recording, SCN cultures from the PER2:LUC mice were maintained in sealed 35-mm petri dishes with 1 ml of culture medium containing 0.3 mM D-luciferin (Promega) at 36 °C. Light emission was measured and integrated for 1 min at 10-min intervals with a dish-type wheeled luminometer (AB-2550 Kronos-Dio, ATTO). The background-subtracted bioluminescence profiles were analyzed using the Cosinor analysis software (http://www.circadian.org↗).

POMC-tdTomato mice and NPY-hrGFP mice received injections of either NAMPT (1 mg/kg, i.p.) at 1 p.m. (ZT5) or FK866 (10 mg/kg, i.p.) at 1 a.m. (ZT17). One hour after injection, mice were transcardially perfused with 4% paraformaldehyde (PFA) for 15 min using a peristaltic pump. Whole brains were collected, post-fixed with 4% PFA overnight, and kept in 30% sucrose for 48 h. Brain sections (20 μm) were obtained and incubated overnight at 4 °C with a rabbit anti-c-Fos antibody (1:1000, Synaptic Systems, #226 003). Tissues were rinsed with PBS and then incubated with Alexa Fluor 488-conjugated anti-rabbit IgG (1:500, Invitrogen) at room temperature for 1 h. After washing with PBS, the slices were mounted using Vectashield mounting media (Vector Laboratories) and viewed using an LSM 700 confocal microscope (Carl Zeiss). Three matched brain sections, which included the ARC, were analyzed in each animal. The numbers of POMC and NPY/AgRP neurons with nuclear c-Fos immunoreactivity were counted and expressed as a percentage of the total number of POMC or NPY/AgRP neurons. c-Fos positive neurons were counted using ZEN microscope software (version 2.1 blue edition, Carl Zeiss).

Adenoviruses expressing FOXO1-GFP fusing gene (Foxo1-Ad, 1 × 10¹¹ plaque-forming unit) were microinjected bilaterally into the mediobasal hypothalamus (5.7 mm depth, 1.8 mm caudal to the bregma, 0.3 mm lateral from the sagittal suture) via a syringe pump (Harvard Apparatus) at a rate of 100 nl/min for 5 min (0.5 μl/injection site)⁴². Animals were randomly assigned to experimental groups according to their body weight. Control animals received the same amount of GFP-Ad. On the third day of adenovirus injection, we monitored LMA, EE, and FI following i.c.v. administration of NAMPT or saline. Appropriate injection of the adenovirus was verified by the presence of green fluorescence in the MBH under a confocal fluorescence microscope at the end of the study. (Supplementary Fig. 18a). Results from animals that were shown to have received appropriate injections were included in the analyses. As a separate experiment, mice injected with GFP-Ad or Foxo1-Ad were sacrificed 2 h after i.c.v. injection of saline or NAMPT (100 ng) on the third day after adenovirus injection to determine the hypothalamic neuropeptide expression. In this experiment, we assessed successful viral injection with increased Foxo1 expression (Supplementary Fig. 18b).

siRNAs specific to murine/human Sirt1 and Sirt2 were purchased from Dharmacon (M-061727-01 for mouse Sirt2; M-004826-02 for human SIRT2, E-003540-00 for human SIRT1). For the cell study, SH-SY-5Y cells were transfected with siControl, siSIRT1 or siSIRT2 (50 nM each). 48 h later, total RNA and protein were extracted to check for successful knockdown of the target gene (Supplementary Figs. 12a and 14b). For the animal study, murine siSirt2 was resuspended in RNase-free water and mixed with lipofectamine (9:1 v/v, Invitrogen) to a final concentration of 1 mM. Sirt2 siRNA was injected bilaterally (0.5 μl each side) into the ARC over a 5 min period as previously described⁴². The same amount of non-targeting scrambled control siRNA (Dharmacon; D-001910-10-20) was administered to the control group. On the second day of siRNA injection, NAMPT (100 ng) or saline was administered via the i.c.v.-implanted cannulae at ZT6 for the LMA, EE, and FI study. On the 3rd day, mice were sacrificed at ZT6 (1 h after i.c.v. saline or NAMPT injections) to collect the MBH. Animals showing a successful gene knockdown (below 50% of the average levels of controls) were included in the data analysis. The experiments were repeated at least twice.

I.c.v. administration of NAMPT (100 ng) was carried out at ZT16. MBH tissue blocks were collected 2 h after NAMPT injection, and total RNA was extracted from the MBH using Trizol reagent (Invitrogen) and was quantified by spectrophotometry. To analyze hypothalamic region-specific gene expression, SCN, ARC and DMN tissues were microdissected from 10 μm-thick hypothalamic slices using a laser microdissection system (Leica LMD6500) and RNA was extracted using QIAGEN RNeasy Micro Kit (#74004). The mRNA levels of genes were determined using real-time PCR analysis (Perkin-Elmer) using the appropriate primers (Supplementary Table). The expression level of each mRNA was normalized to that of glyceraldehyde 3-phosphate dehydrogenase (Gapdh) or ß-actin mRNA. 5

Mice were sacrificed by decapitation under the indicated conditions. Plasma and tissues were collected, immediately frozen in liquid nitrogen, and stored at −70 °C until use. Tissue lysates (50 μg protein) were separated by 8% SDS-PAGE and transferred to a PVDF membrane (GE Healthcare). Following incubation in blocking buffer, membranes were incubated overnight at 4 °C with antibodies against NAMPT (1:1500, mouse, Adipogen, AG-20A-0034), FOXO1 (1:1000, rabbit, Cell Signaling, 2880S), SIRT1 (1:1000, mouse, Santa Cruz, sc-74465), and SIRT2 (1:1000, rabbit, Cell Signaling, 12650). A 2-μl volume of plasma was diluted in 18 μl of lysis buffer and then subjected to immunoblotting for eNAMPT. Plasma protein loading was normalized using transferrin (1:1000, rabbit, Abcam, ab82411). For the immunoprecipitation assay, we extracted proteins from mouse hypothalamus or from cells transfected with Flag-tagged FOXO1- and SIRT2-expressing plasmids (1 μg each). We then incubated protein extracts (500 μg) with 1 μg of primary antibodies against FOXO1 (Cell Signaling, 2880 S), Flag (Sigma, F3165) or an equivalent amount of rabbit immunoglobulin-γ (IgG) overnight before adding 100 μl agarose bead (Roche). Subsequently, we performed immunoblotting using antibodies specific to SIRT2, FOXO1, and acetylated lysine (1:1000, rabbit, Cell Signaling, 9441). Blots were developed using a densitometer (VersaDoc Multi Imaging Analyzer System, Bio-Rad) and normalized using β-actin immunoblotting (1:1000, mouse, Santa Cruz, sc-47778). We quantified fluorescence intensity using Image J (version 1.52p, NIH).

Whole brains were collected from cardiac-perfused mice with 50 ml saline followed by 50 ml 4% PFA in DEPC water via the left ventricle of the heart under anesthesia induced with 5 mg/kg Rompun and 40 mg/kg Zoletil. Brains were post-fixed with 4% PFA at 4 °C overnight, and kept in 30% sucrose solution until the tissues sank to the bottom of the container. Brain sections (14-μm thick) including the hypothalamus were generated using a cryostat (Leica). Brain slices were incubated with antigen retrieval and protease agent, and hybridized with Sirt2 probes (ACDBio, 1161441-C1) for 2 h at 40 °C, followed by step-wise amplification according to the manufacturer protocol. Following the completion of RNA in situ hybridization, brain sections were subjected to immunofluorescence staining by incubation with 3% BSA in PBS for 1 h and then with diluted POMC (ß-endorphin) antibody (1:1000, rabbit, Phoenix, H02233) in 3% BSA in PBS overnight at 4 °C. After washing, slides were incubated with the appropriate Alexa-Flour 488-conjugated secondary antibodies (1:1000) at room temperature for 2 h. Immunofluorescence images were obtained using a confocal microscope (Carl Zeiss 710, Germany). For NPY and Sirt2 double staining, brains were obtained from NPY-hrGFP mouse and subjected to Sirt2 in situ hybridization and followed by immunostaining using GFP antibody (1:1000, goat, Abcam, ab6673) as described above.

The promoter regions of the human AGRP gene (nucleotide positions –1000 to –1) or POMC gene (nucleotide positions –877 to +320) were cloned and ligated into a luciferase reporter as previously described⁴². SH-SY5Y cells were cultured in 24-well plates and were transfected with pAGRP-luc or pPOMC-luc reporters (50 ng) with or without human SIRT2 (100 ng), human wild-type FOXO1 and FOXO1-6KR (100 ng each, gift from Dr. Tadahiro Kitamura at Gumma University), siControl, siSIRT1, siSIRT2 (50 nM each), or CMV-β-gal (25 ng) using lipofectamine. At 48 h after transfection, SH-SY5Y cells were treated with NAMPT protein (1, 10, and 100 nM) for 2 h following 2 h-serum starvation. At the end of the study, cells were harvested to measure luciferase activities, which were normalized to those of β-galactosidase. Transfections in each condition were performed in at least triplicates and the experiments were repeated three times. Data are shown as fold changes compared with mock vector-transfected controls.

For human studies, 14 healthy Koreans (10 males, 4 females) aged 18–30 years were recruited via local website advertising. Seven subjects (5 males, 2 females) were lean (BMI < 23 kg/m²) and 7 subjects (5 males, 2 females) were individuals with obesity (BMI ≥ 25 kg/m²), according to the definition of obesity from the International Diabetes Federation–Western Pacific Region. All subjects signed an informed consent form. This study was approved by the Institutional Review Board of Asan Medical Center (Seoul, Korea) (IRB approval number: 2008-0314, approved date: August 11, 2008) and carried out in accordance with the principles of the Declaration of Helsinki. Subjects were advised to sleep between 10 p.m. and 6 a.m. for 1 week before the study. All subjects were admitted to the Asan Clinical Research Center at 5 p.m. on the day before the study day and were provided with dinner at 6 p.m. An in-dwelling intravenous catheter was inserted into the forearm vein for blood sampling after dinner. On the study day, blood was drawn into serum-separating tubes containing aprotinin (250 kallikrein inhibitor units; Bayer) at 30-min intervals from 6 a.m. to 10 p.m. Thereafter, samples were drawn hourly until 6 a.m. on the next morning. Subjects had breakfast at 8 a.m., lunch at 12:30 p.m., and dinner at 6 p.m., which consisted of 60% carbohydrates, 25% fat, and 15% protein. Lights were off from 10 p.m. to 6 a.m.; the lights were on at all other hours. Subjects were allowed a rest period except when eating and using the lavatory.

Investigators were not blinded in conducting animal experiments or collecting samples for analysis. Where possible, researchers were blinded during data analysis.

Statistical analysis was performed using SPSS version 24 (IBM Analytics) and Graphpad Prism version 7. Statistical significance between groups was determined using one- or two-way analysis of variance, followed by a post hoc Fisher’s least significant difference test and an unpaired Student’s t-test, as appropriate. JTK-CYCLE algorithm (‘MetaCycle’ package of R software, version 4.1.1) was used to detect circadian rhythmicity and circadian rhythm parameters were assessed using the cosinor package of R software (version 4.1.3; R Foundation for Statistical Computing, Vienna, Austria). Wald test was used to compare amplitude and acrophase between lean and obese mice while Student’s t-test was used for the comparison of mesor. Statistical significance was defined at P < 0.05. Data are presented as mean ± standard error of the mean (SEM).

Further information on research design is available in thelinked to this article. Nature Portfolio Reporting Summary

Nature Communications thanks Ricardo Orozco-Solis, Yuanzhong Xu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

All data generated or analyzed during this study are provided with this paper in the source data files.are provided with this paper. Source data