NAMPT-dependent NAD ⁺ biosynthesis controls circadian metabolism in a tissue-specific manner

To explore the role of NAMPT in regulating the clock in vivo, we used fat-specific Nampt knockout mice (FANKO) (29) and inducible skeletal muscle-specific Nampt knockout mice (iSMNKO) (Fig. 1A and SI Appendix, Fig. S1A↗). Both mouse lines exhibited similar bodyweights and body compositions (SI Appendix, Fig. S1B↗) (29), eating behaviors (SI Appendix, Fig. S1 C and D↗), and locomotor activities (SI Appendix, Fig. S1E↗) compared with littermate controls. Thus, FANKO and iSMNKO mice are ideal models to compare tissue-specific clock function in vivo. In BAT and epididymal WAT (eWAT) from wild-type (WT) mice, Nampt expression oscillated in a circadian manner and peaked during the dark phase (Fig. 1 B and C) consistent with earlier observations (36–38). This pattern of Nampt regulation resulted in significant NAD⁺ oscillation in BAT and a trend toward rhythmicity in eWAT (Fig. 1 D and E). In FANKO adipose depots, the markedly reduced NAD⁺ levels no longer exhibited any oscillatory pattern, supporting the foundational discovery in cell systems that Nampt is responsible for a circadian fluctuation in NAD⁺ levels (22, 23). However, even though NAD⁺ levels were reduced to comparable absolute levels in FANKO BAT and eWAT (12 and 9 pmol/mg tissue, respectively), the subsequent impact on the core clock machinery was different. In BAT, we observed pronounced dampening of both Bmal1 (a.k.a. Arntl), and Reverb-α (a.k.a. Nr1d1) expression (Fig. 1 F–H). Conversely, Arntl and Nr1d1 levels in FANKO eWAT were largely unaffected by Nampt deletion (Fig. 1 I–K). Clock genes in nonadipose tissues from FANKO mice were unchanged (SI Appendix, Fig. S1 F–K↗).

The tissue-specific nature of Nampt circadian regulation was even more evident when comparing the FANKO adipose clocks to the skeletal muscle clock in iSMNKO mice. Consistent with previous work (36, 38, 39), we observed no oscillation in skeletal muscle Nampt levels (Fig. 1L). Yet, even without Nampt oscillation, skeletal muscle NAD⁺ levels displayed a significant, albeit subtle, biorhythm (Fig. 1M), which may be due to diurnality of NAD⁺-consuming reactions. In iSMNKO muscle, Nampt was reduced to approximately the same level as in FANKO BAT and eWAT, and the relative reduction in NAD⁺ levels was 67% compared with 96% and 73% in FANKO BAT and eWAT, respectively. Like adipose tissues, NAD⁺ oscillation in skeletal muscle was totally abrogated by Nampt deletion (Fig. 1M). Nevertheless, the absolute level of NAD⁺ was still relatively high in iSMNKO muscles. Despite loss of Nampt and lower NAD⁺, the expression of Arntl, Nr1d1, and Cry1 in iSMNKO skeletal muscle oscillated similarly compared with WT mice (Fig. 1 N–P). Thus, our findings reveal that the rhythmicity of Nampt and NAD⁺ and the NAMPT-dependence of core clock genes are differentially regulated across tissues.

To investigate the implications of Nampt ablation on a broader spectrum of circadian genes, we performed RNAseq analyses on BAT and eWAT harvested at four time points. We selected the midpoints (i.e., ZT 6 and ZT 18) in each phase as well as time points immediately preceding the transition between light/dark phases (i.e., ZT 10 and ZT 22). Notably, these transitional periods just prior to animals waking up and going to sleep have been shown to be transcriptional “rush hours” with the highest number of peaking circadian genes (36). Rhythmicity was analyzed using the R package LimoRhyde as it allows for assessments of oscillation and amplitude of nonsymmetrical data (40). Nampt ablation markedly diminished the amplitude of clock gene expression in FANKO BAT, particularly attenuating oscillation of Clock, Arntl, Npas2, Cry1, Nr1d2, Rorc, and Dbp (Fig. 1Q). Circadian amplitude was also blunted in FANKO eWAT and was most pronounced for Clock and Dbp expression (Fig. 1Q). In contrast to reduced amplitudes, the times at which clock genes peaked in both depots remained generally unaffected by Nampt deficiency (SI Appendix, Fig. S1L↗). Previously, knockout of Nampt in the liver was shown not to affect the rhythmicity of Arntl, Clock and Per2 (35), whereas the amplitude of Arntl and Per2 were decreased in U2-OS and NIH 3T3 cells in response to SIRT1 knockdown (41). In another study, pharmacological inhibition of NAMPT by FK866 was shown to enhance the amplitude of Per2 in mouse embryonic fibroblasts (23). The amplitude of Clock and Arntl was decreased in BAT, and Per2 trended to be decreased (SI Appendix, Fig. S1M↗). In eWAT, only the amplitude of Clock was decreased (SI Appendix, Fig. S1M↗). In the skeletal muscle of iSMNKO mice, the rhythms of Per2, Drp1, and Clock expression were unchanged (SI Appendix, Fig. S2 A–C↗). This might be due to the relatively high level of NAD⁺ remaining in the skeletal muscle of iSMNKOs. To determine this, we measured core clock gene expression at ZT 4 in constitutive skeletal muscle Nampt knockout (SMNKO) mice and control littermates. In the skeletal muscle of SMNKO mice, NAD⁺ levels were depleted by 91% (30), but Arntl, Nr1d1, Cry1, Per2, Dbp, and Clock expression were still not impacted (SI Appendix, Fig. S2D↗). Together, our findings highlight the cell type–specific nature of NAMPT circadian influence and support that NAMPT acts to selectively potentiate the rhythmicity of the brown and white adipose molecular clocks in vivo.

To determine whether boosting NAD⁺ levels could restore clock gene expression in Nampt-depleted adipose tissue, mice were supplemented with the NAD⁺ precursor nicotinamide riboside (NR). Achieving constant high levels of NAD⁺ in the adipose tissues is challenging with oral or injection protocols, given that NAD⁺ has a high turnover rate in adipose tissues (42–44). Thus, we used a constant 4-d intravenous infusion of NR through a jugular catheter at a dose of 800 mg/kg/d (Fig. 1R) that led to a twofold and a fourfold increase in NAD⁺ levels in WT BAT and eWAT, respectively (Fig. 1S and T). To assess the capacity of NR supplementation to rescue clock function in our NAMPT KO model, adipose tissues were harvested at ZT 6, where we had previously observed a difference in clock gene expression between FANKO mice and control littermates (SI Appendix, Fig. S1M↗). NR rescued NAD⁺ levels in FANKO BAT substantially and in FANKO eWAT to even higher levels than those found basally in WT mice (Fig. 1 S and T). NR led to a complete rescue of Cry1 in both BAT and eWAT of FANKO mice (Fig. 1 U and V), whereas restoration of clock genes, Cry2 and Nr1d1, was depot specific (Fig. 1 W and X and SI Appendix, Fig. S2 E and F↗). Additionally, NR decreased expression of Arntl and Npas2 expression in both genotypes and adipose depots (SI Appendix, Fig. S2 G–J↗), in stark contrast to the effect observed in response to Nampt deletion. Thus, acutely elevating NAD⁺ levels by constant infusion of NR partially restores long-term disruption of the circadian clock and underscores the tight regulatory connection between NAD⁺ levels and the core clock.

We next explored how NAMPT regulation of the core clock affected global transcriptional programs. Consistent with earlier studies (36), we found a higher number of genes to be circadian in BAT (2,821) than in eWAT (1,885) (Fig. 2A and Dataset S1↗), of which only 590 were shared between the depots (Fig. 2A). Nampt deletion substantially reduced the BAT circadian transcriptome by 2,500 genes (Fig. 2B and SI Appendix, Fig. S2K↗). FANKO eWAT lost rhythmicity of 1,460 genes relative to WT tissue (Fig. 2C and SI Appendix, Fig. S2L↗), indicating that NAMPT exerts marked control over general transcriptional rhythmicity in both brown and white adipose depots despite the difference in regulation of the core clock. In line with our findings on how NAMPT potentiates the oscillations of the core clock genes, the amplitudes of oscillating genes in BAT and eWAT were significantly reduced in FANKO mice (Fig. 2D) compared with WT littermates, whereas peak time were largely unchanged (Fig. 2E). Beyond its specific role in imparting transcriptional rhythmicity, Nampt deletion changed the overall expression level of 13,333 transcripts (6,801 down/6,532 up) in BAT and 10,063 transcripts (4,562 down/5,501 up) in eWAT. This impact on transcription is consistent with previous work by Yamaguchi and colleagues (28, 45) and highlights a broader requirement for NAD⁺ as an essential transcriptional regulator in addition to the molecular clock.

Gene ontology analysis of the circadian genes in the two adipose depots showed clear similarities. Terms connected to circadian rhythmicity and protein folding were highly enriched in both adipose tissue depots independent of genotype (Fig. 2F) but with a higher number of genes regulated in WT compared with FANKO mice. The unique intersection between the four datasets was visualized by an UpSet plot (Fig. 2G). A total of 315 genes, over half of the rhythmic genes common between the two adipose tissues (590 genes, Fig. 2A), lost rhythmicity in both depots in response to Nampt deletion (SI Appendix, Fig. S2M↗). The highest ranking ontology terms for these shared genes included biosynthesis of triglyceride, cholesterol and phosphatidic acid, lipid droplet organization, and cardiolipin metabolism (Dataset S2↗). These data underscore the importance of NAMPT in the regulation of circadian lipid metabolism in both tissues. As indicated in the UpSet plot, 46 genes gained rhythmicity in response to knockout of Nampt in both tissues. However, these appeared to be related to an immune response (Dataset S2↗).

There were also notable differences in the tissue-specific ontologies of circadian genes. In eWAT of the FANKO mice, genes involved in lipid and cholesterol homeostasis lost rhythmicity, whereas genes related to inflammation gained rhythmicity (Fig. 2F and Dataset S3↗). The increase in rhythmic inflammatory-related genes in the eWAT of FANKO mice may be due to infiltration of immune cells. In line with this, high-fat diet (HFD) substantially increased inflammation in WAT of the FANKO mice (29). The highest ranking terms between the genes gaining rhythmicity in response to Nampt deletion in BAT were immune response, and positive regulation of transcription by RNA polymerase II (Dataset S3↗). Looking at the ontology for genes losing rhythmicity in each of the two depots separately (Dataset S3↗), we found remarkable tissue-specific differences. In BAT, more than 100 genes related to each of the following terms lost rhythmicity: positive regulation of protein phosphorylation, positive regulation of transcription, cytoskeleton organization, positive regulation of cellular component organization, and organic acid metabolic process (Dataset S3↗). This result points to an effect of NAMPT in circadian regulation of general transcription and cytoskeleton organization in BAT. In eWAT, loss of rhythmicity was seen in a high number of genes related to cell cycle regulation and metabolism, explicitly the terms: regulation of cell cycle, organonitrogen compound metabolic process, cellular lipid metabolic process, phosphate-containing compound metabolic process, monocarboxylic acid metabolic process (Dataset S3↗). An analysis of the common rhythmic genes for the two adipose tissues revealed that more transcripts lost rhythmicity specifically in BAT compared with eWAT, 146 vs. 59, respectively (SI Appendix, Fig. S2M↗). Among the BAT-specific transcripts were again genes related to organic acid metabolic process, and positive regulation of transcription (SI Appendix, Table S2↗). These tissue-specific differences, due to Nampt deletion, in the number of oscillating transcripts and on their gene ontology highlight that the effect of NAD⁺ depletion on circadian rhythmicity largely depends on the cellular context, such as gene accessibility, or the presence of specific transcriptional regulators.

Similar to circadian oscillations in gene expression, metabolite levels show a diurnal pattern in adipose tissue (46). The circadian rhythms of metabolite levels in BAT are profoundly affected by disruption of the molecular clock (46). Furthermore, depletion of NAD⁺ affects the metabolite profile in other model systems (31, 47, 48). To describe the effect of Nampt ablation on the diurnal metabolite rhythmicity of BAT, we performed metabolomics analyses on WT and FANKO mice tissues at the midpoint of the light and dark phases, ZT6 and ZT18, respectively. Any metabolite with concentration significantly changed between these two time points was characterized as rhythmic. Out of 760 metabolic features detected in BAT, 322 were rhythmic (Fig. 3A) and 171 of these oscillating metabolites were dependent on NAD⁺. Nampt deletion affected an additional 223 metabolites that were not classified as rhythmic. The critical importance of BAT NAD⁺ biosynthesis for metabolic rhythmicity was further underscored by principal component analysis (PCA). The WT metabolite profiles were clearly distinct at the two times of day, whereas loss of Nampt abolishes this distinction (SI Appendix, Fig. S3A↗, PC2). As many as 93 features showed an interaction between genotype and time of day (Fig. 3A). Among these specific features having different diurnal patterns in response to Nampt ablation were metabolites related to the KEGG pathways Arginine biosynthesis, TCA cycle, and Pyruvate metabolism (Fig. 3B). Nampt ablation completely abolished the diurnal rhythm of lactate and the three TCA cycle intermediates: α-ketoglutarate, malate, and fumarate (Fig. 3C). We observed similar patterns for N-acetylornithine, an intermediate in arginine biosynthesis and the urea cycle (Fig. 3D). These data indicate that NAMPT is a central regulator of circadian energy metabolism in BAT.

We next determined whether NAMPT-specific circadian changes in BAT metabolites were reflected in the blood. We measured blood glucose and lactate at the midpoint of each period (ZT6 and ZT18). Blood glucose was significantly higher in the FANKO mice at both times-of-day compared with controls, and both genotypes exhibited increased blood glucose during the day at ZT 6 compared with ZT 18 (Fig. 3E). Thus, the diurnal pattern of blood glucose levels was unaffected by Nampt deletion. Despite a clear effect in BAT, blood lactate levels did not differ between genotypes or time of day (Fig. 3F). To further investigate whether the rhythmicity of glucose metabolism in other tissues were affected by adipose Nampt deletion, we measured liver glycogen every 4 h. Liver glycogen levels were highly rhythmic in both genotypes, but the rhythmicity was unaffected by adipose Nampt deletion (Fig. 3G). Overall, these findings are consistent with earlier reports that adipose NAMPT plays a role in whole-body glucose metabolism (29, 49), yet the effect of Nampt deletion on metabolite rhythmicity appears to be more restricted locally within BAT.

Integration of the metabolomic analysis with the transcriptomic analysis allowed us to evaluate the overall impact of Nampt ablation on BAT circadian metabolism. Fig. 3H lists the most distinct KEGG pathways in BAT from the combined analysis of metabolites and genes with changed rhythmicity in response to Nampt ablation. The changes in TCA cycle and pyruvate metabolism were only seen on the metabolite level and were therefore likely driven directly by the decrease in NAD⁺ levels in the FANKO mice, given the role of NAD⁺ as a co-factor in these pathways (Fig. 3C). Three pathways, including Aspartate metabolism, De novo pyrimidine biosynthesis, and Starch and sucrose metabolism, were among the distinct KEGG pathways in the combined analysis that demonstrated a change in rhythmicity in response to Nampt ablation at both the metabolite and the transcriptional levels (Fig. 3H). In each case, the circadian expression of relevant enzymes were altered upstream of the changes in metabolite rhythmicity (SI Appendix, Fig. S3 B–D↗). The diurnal rhythms of several other metabolic pathways were only regulated at the transcriptional level. Among these were the Phosphatidylinositol signaling system, Steroid biosynthesis, and Nicotinate and nicotinamide metabolism. One reason for this discrepancy could be due to the higher resolution of the transcriptomic analysis. Alternatively, many genes cycle at the transcriptomic level, without having a diurnal pattern at the protein level. Concerning NAD⁺ metabolism, both nicotinamide and NAD⁺ levels were decreased in the FANKO mice, and the levels showed a time-of-day effect (SI Appendix↗, Fig. S3 E and F), although there was no difference in their diurnal patterns in response to Nampt ablation.

We next investigated NAD⁺-dependent metabolite oscillation on eWAT from the same WT and FANKO cohort used for BAT metabolomics. Of the 746 metabolic features detected in eWAT, 144 were altered by Nampt ablation and only 22 of these were rhythmic metabolites (Fig. 4A). This relatively nominal impact on eWAT metabolites was globally reflected in the PCA plot where there was no clear separation between genotypes or time-of-day (SI Appendix, Fig. S4A↗). While there was no interaction between the genotype and time-of-day effect in eWAT, fumarate, malate, and sorbose levels were significantly affected by both parameters (Fig. 4B). Collectively, these data illustrate the large difference in the diurnal pattern of metabolism between brown and white adipose depots, with 42% of the detected features cycling in BAT and only 3% cycling in eWAT. Moreover, our metabolomic findings are consistent with the transcriptional data emphasizing that NAMPT-dependent NAD⁺ biosynthesis plays a pronounced tissue-specific role in BAT rhythmicity.

NAMPT is essential for the function of adipose tissue (29, 49, 50), particularly apparent in the context of diet-induced obesity, where mice deficient in adipose Nampt fail to expand their adipose tissue to accommodate the lipid burden (29). Obesity is itself a major effector on the clock; in WT mice, high-fat diet (HFD) alters the clock in peripheral tissues including adipose and increases activity level and food intake of the animals during the light phase, thus attenuating the diurnal pattern (51–54). Furthermore, Nampt diurnal oscillation in adipose tissue is dampened in a genetic model of obesity (37), and the circadian metabolite profile of BAT is heavily influenced by HFD (55). Therefore, we sought to describe the effect of Nampt deletion on the diurnal metabolomic profile of BAT in the context of a HFD challenge. We performed metabolomic analysis on mice fed either chow or HFD and collected tissues at ZT 6 and ZT 18. We detected 1,028 metabolic features, 472 were affected by genotype, 251 by time-of-day, and 283 by diet (Fig. 4C). PCA revealed a clear separation between genotypes on both diets, but only mice fed a chow diet separated with time-of-day (SI Appendix, Fig. S4B↗). There were no significant FDR-corrected 3-way interactions, but 43 metabolite features were affected by the three factors: genotype, time-of-day, and diet, individually. Among these was the TCA intermediate cis-aconitate (Fig. 4D), which was rhythmic in WT BAT under both chow and HFD conditions but lost rhythmicity upon Nampt depletion. For two other TCA intermediates, fumarate and malate, loss of Nampt flattened rhythmicity on chow diet, mirroring the circadian-disrupting effects of HFD (Fig. 4 E and F). Comparing the genotype and time-of-day effects separately in the cohorts on chow and HFD gave a striking result. In the chow-fed cohort, 262 metabolic features were affected by genotype, 243 were affected by time-of-day, and 21 showed an interaction between genotype and time-of-day (Fig. 4G). This is a slightly lower number of regulated features compared with the earlier study (Fig. 3A). By contrast, in mice fed a HFD, 314 features exhibited a genotype effect, but no metabolic feature showed any significant diurnal rhythm (Fig. 4H). A strong effect of HFD on metabolite rhythmicity in BAT has been reported (55), but not to this extent. This emphasizes that on top of tissue specificity, diet adds another layer of complexity to the circadian regulation of metabolism.

We next investigated the physiological consequences of the striking alterations in both circadian transcription and metabolism in Nampt-deficient adipose tissue. The main function of murine BAT, the tissue where we observed the larges changes in transcription and metabolism, is thermogenesis and body temperature defense (56). To investigate the effect of Nampt depletion on temperature regulation, we surgically implanted telemetric temperature probes in the intraperitoneal cavity of WT and FANKO mice. This enabled us to continuously monitor body temperature. The rhythmicity of body temperature was not changed in the FANKO mice at room temperature (SI Appendix, Fig. S4C↗). We then challenged mice with 6-h bouts of cold exposure either during the day (at ZT 4) or during the night (at ZT 16) (Fig. 4I). Activity and food intake during the cold exposure were similar between the two time points and genotypes (SI Appendix, Fig. S4 D and E↗). Both genotypes exhibited differential sensitivity to cold at ZT 4 compared with ZT 16 (Fig. 4 J–L). Yet surprisingly, under these conditions, the FANKO mice were significantly less sensitive to cold at both times-of-day compared with control littermates (Fig. 4 J–L). There was no interaction between genotype and time-of-day, indicating that the diurnal rhythmicity of cold sensitivity was independent of adipose NAMPT.

The tissue-specific Nampt-deficient mouse models were based on the Cre-lox system with exon 3 of Nampt (Nampt^tm1Jtree) being flanked by loxP sites (75). To generate the fat-specific Nampt knockout (FANKO), the Nampt^tm1Jtree line was crossed with the Adiponectin-Cre mice (76) (010803, Jackson Labs), kindly provided by Prof. Karsten Kristiansen, as previously described (29). Inducible skeletal muscle-specific Nampt knockout mice (iSMNKO) were generated by crossing the Nampt^tm1Jtree line with the HSA-MCM line, which expresses a Mer-Cre-Mer (mutated estrogen receptor) chimeric protein (77) (025750, Jackson Labs). iSMNKO mice received tamoxifen (2 mg orally in 100 μL corn oil for three consecutive days) 7 wk prior to tissue harvest to induce Nampt deletion in skeletal muscle. Cre^– littermates were used as controls. Only male mice were used in the experiments.

Mice were housed at room temperature in an enriched environment with ad libitum access to water and chow (1310, Altromin) or high-fat diet (D12492, Research Diets). Lights in the housing facility were switched on at 6am and off at 6pm.

All animal experiments were approved by the Danish Animal Experiments Inspectorate under license no. 2014-15-0201-00181, 2018-15-0201-01441 (FANKO), and 2015-15-0201-00792 (iSMNKO).

The assay was performed as previously described (78). In short, tissue was lysed in 400 μL 0.6 M perchloric acid. The supernatant was diluted in 100 mM Na₂HPO₄ (final pH 8.0), and NAD⁺ was measured on the diluted sample. A reaction mix [2% ethanol, 90 U/mL alcohol dehydrogenase, 130 mU/mL diaphorase, 10 μM resazurin, 10 μM flavin mononucleotide, 10 mM nicotinamide in phosphate buffer (100 mM Na₂HPO₄), pH 8.0] was added, and continuous resorufin accumulation was measured for 15 to 30 min by fluorescence excitation at 544 nm and emission at 580 nm. Absolute quantification was achieved through a standard curve made by serial dilution of a standard with a known concentration.

Total RNA was extracted from tissues using TRI reagent (T9424, Sigma-Aldrich) and isolated using the RNeasy Mini Kit (74106, Qiagen). Reverse transcription was performed with the High Capacity cDNA Reverse Transcription kit (4368814, Applied Biosystems). Gene expression was determined by real time quantitative PCR using SYBR green (PP00259, Primerdesign). Expression levels were normalized to ssDNA input measured by a Qubit fluorometer (Thermo Fisher Scientific). Primer sequences are included in SI Appendix, Supplemental Materials and Methods↗.

RNA sequencing libraries were prepared as described (79). Reads were subjected to 38-bp paired-end sequencing on a NextSeq500 (Illumina) and aligned using STAR v. 2.5.3a (80), against the GRCm38.p5 primary assembly and GENCODE comprehensive gene annotations (81) version 15. Details on the bioinformatics analysis are included in SI Appendix, Supplemental Materials and Methods↗.

Sample processing and profiling was performed as described in SI Appendix, Supplemental Materials and Methods↗.

At ZT 6 and ZT 18 blood glucose and lactate levels were measured directly with Contour Next (Bayer) and Lactate Plus (Nova Biochemical) strips, respectively.

Glycogen was extracted from mouse liver samples by acidic extraction using 1 M HCl at 95 °C for 2 h. After a quick spin, tissue extracts were neutralized using 1 M NaOH followed by centrifugation at 16,000 g for 20 min at 4 °C. The supernatant was transferred to new tubes and a reaction mix consisting of 200 mM Tris-HCl, 500 mM MgCl₂, 5.2 mM ATP, 2.8 mM NADP, and 6 μg/mL hexokinase and glucose-6-phosphate dehydrogenase mixture (10737275001, Roche Diagnostics) were added. After 15 min of incubation at RT, the glucose generated from the hydrolyzed glycogen was measured spectrophotometrically at 340 nm (Hidex sense, Hidex). All samples were measured in duplicates along with glucose standards.

Indirect calorimetry was performed using the Phenomaster Home Cage System (TSE Systems). Animals were acclimated in the system for more than 5 d prior to the measurement. All data were recorded in 6-min intervals. Data collection was integrated into the TSE software. Basal food intake is an average over five constitutive days.

Body temperature and gross motor activity was measured using G2 E-Mitters surgically implanted into the intraperitoneal cavity. Mice were anasthetized with 2% isoflurane during surgery and were allowed to recover for 2 wk before the measurements. Signals from G2 E-Mitters were detected by ER400 Energizer/Receivers (STARR Life Sciences Corp.), and data collection was integrated into the TSE software. Basal physical activity is an average over five constitutive days. For the acute cold tolerance test, mice were transferred from room temperature to new cages in the Phenomaster, which were cooled down to 4 °C. Cold challenges were performed both at ZT 4 and ZT 16, and the mice were kept in the Phenomaster at 4 °C for 6 h.

To obtain continuous intravenous (IV) infusions, a catheter was inserted in the right external jugular vein. The catheter was externalized through a dorsal incision in the interscapular region using a single-channel Vascular Access Button (VABM1B/25; Instech Laboratories). Mice were anasthetized with 2% isoflurane during surgery and were allowed to recover for 1 wk before the beginning of the infusions. Based on a pilot experiment, a NR (Elysium Health) dose of 800 mg/kg bodyweight/d was chosen. The infusate had an NR concentration of 22.4 g/L, and a pH of 7.2. After 4 d of constant infusions with NR or saline in their local environment, the mice were taken down at ZT 6.