NAD⁺ Controls Circadian Reprogramming Through PER2 Nuclear Translocation to Counter Aging

Summary

Disrupted sleep-wake and molecular circadian rhythms are a feature of aging associated with metabolic disease and reduced levels of NAD⁺, yet whether changes in nucleotide metabolism control circadian behavioral and genomic rhythms remains unknown. Here we reveal that supplementation with the NAD⁺ precursor nicotinamide riboside (NR) markedly reprograms metabolic and stress-response pathways that decline with aging through inhibition of the clock repressor PER2. NR enhances BMAL1 chromatin binding genome-wide through PER2^K680 deacetylation, which in turn primes PER2 phosphorylation within a domain that controls nuclear transport and stability and which is mutated in human advanced sleep phase syndrome. In old mice, dampened BMAL1 chromatin binding, transcriptional oscillations, mitochondrial respiration rhythms, and late evening activity are restored by NAD⁺ repletion to youthful levels with NR. These results reveal effects of NAD⁺ on metabolism and the circadian system with aging through the spatiotemporal control of the molecular clock.

Graphical Abstract

Introduction

All mammals possess an internal circadian clock that couples sleep-wake and fasting-feeding cycles with the light-dark cycle (Bass and Lazar, 2016; Rey and Reddy, 2013). The core clock is a cell-autonomous transcription-translation feedback loop comprised of heterodimeric activators (CLOCK/BMAL1) that transcribe their own repressors (PERs/CRYs/REV-ERB) in a cycle that repeats itself every ~24 hrs. CLOCK/BMAL1 transcriptional activity during the light period is enabled through sequestration of nascent PER and CRY repressor proteins in the cytoplasm. During the dark period, PER and CRY translocate into the nucleus and associate into a macromolecular complex that inhibits CLOCK/BMAL1 and recruits regulators of rhythmic gene oscillation (Aryal et al., 2017; Hardin et al., 1990; Kume et al., 1999; Meyer et al., 2006; Yagita et al., 2002). Ubiquitin-mediated degradation of PER and CRY ends the repressive phase and permits a new cycle to begin (Eide et al., 2005). While this core clock cycle persists in the absence of external cues, the clock is also responsive to changes in environmental conditions through mechanisms that have not been fully elucidated (Bargiello and Young, 1984; Bargiello et al., 1984; Gallego and Virshup, 2007; Konopka and Benzer, 1971; Lowrey et al., 2000; Meng et al., 2008; Reddy et al., 1984; Toh et al., 2001; Zehring et al., 1984).

Mounting evidence has indicated that the circadian clock is intimately linked with metabolic homeostasis. In mice, genetic disruption of molecular clock leads to diet-induced obesity and metabolic disorders (Hirota et al., 2012; Marcheva et al., 2010; Paschos et al., 2012; Turek et al., 2005). Dysregulation of circadian rhythms in humans is also associated with obesity and increased risk for metabolic syndrome (Bass and Lazar, 2016; Rey and Reddy, 2013). The connection between clock disruption and metabolic disease can also be understood based on the observation that ~10% of the mammalian transcriptome is expressed rhythmically with enrichment in pathways involved in intermediary metabolism (Fang et al., 2014; Koike et al., 2012; Mure et al., 2018; Storch et al., 2002; Zhang et al., 2014). Conversely, changes in nutrient and high fat feeding lead to alterations in rhythms of locomotor activity, clock gene expression, and bioenergetics (Atger et al., 2015; Eckel-Mahan et al., 2013; Guan et al., 2018; Kohsaka et al., 2007; Sato et al., 2017; Tognini et al., 2017). However, how nutrients impact circadian function remains poorly understood.

Several lines of evidence indicate that decline in circadian processes is a hallmark of aging . Aging leads to increased daytime napping and a shift towards earlier hours of waking and sleep-onset in humans (Buysse et al., 1992; Ohayon et al., 2004; Roenneberg et al., 2007; Walch et al., 2016), as well as increased risk for developing obesity, diabetes, and metabolic syndrome (Folsom et al., 1993; Ford et al., 2002; Morley, 2008). At the molecular level, aging is associated with dampened oscillation of clock gene expression (Rosenberg et al., 1991; Sato et al., 2017; Solanas et al., 2017; Yamazaki et al., 2002), and genetic abrogation of Bmal1 in mice leads to a spectrum of premature aging disorders and reduces lifespan (Kondratov et al., 2006). Despite the link between aging and disrupted circadian rhythms, little is known regarding the underlying molecular mechanisms that mediate these phenotypes.

A clue to the mechanism underlying a decline in circadian rhythms and metabolism with age stems from the observation that the longevity-associated deacetylase SIRT1 interacts with the core clock complex, though how SIRT1 impacts the clock remain incompletely understood (Asher et al., 2008; Chang and Guarente, 2013; Nakahata et al., 2008; 2009; Ramsey et al., 2009). SIRT1 activity requires the sentinel metabolite nicotinamide adenine dinucleotide (NAD⁺) (Imai et al., 2000), which declines with age (Braidy et al., 2011; Camacho-Pereira et al., 2016; Massudi et al., 2012; Yoshino et al., 2011). Boosting NAD⁺ in rodents ameliorates age-associated insulin resistance, protects against noise-induced hearing loss, central brain injury, heart failure, diabetic neuropathy, enhances postpartum functions and extends lifespan (Brown et al., 2014; Cantó et al., 2012; Diguet et al., 2018; Ear et al., 2019; Massudi et al., 2012; Trammell et al., 2016; Vaur et al., 2017; Yoshino et al., 2011; Zhang et al., 2016), though whether these NAD⁺-dependent benefits improve circadian programming is unknown. Based on the observation that calorie restriction helps maintain hepatic circadian gene expression and the NAD⁺ metabolome (Sato et al., 2017), we hypothesized that reversing the decline in NAD⁺ with age would restore clock activity. Here we examine the effect of pharmacologic and genetic manipulations of NAD⁺ in the control of genome-wide circadian transcriptional rhythms and identify NAD⁺ as a mediator of PER2 nuclear extrusion and BMAL1 chromatin binding activity. We find that pharmacologic elevation of NAD⁺ failed to reprogram transcription in Bmal1 knockout mice, emphasizing a requirement for the molecular clock in NAD⁺ action. Moreover, provision of nicotinamide riboside (NR), an NAD⁺ precursor (Bieganowski and Brenner, 2004), in drinking water of old mice countered the decline in nighttime locomotor activity rhythms and enhanced molecular oscillations in mitochondrial respiration and transcription, revealing a metabolic feedback pathway to the clock important in health with aging.

Results

Supplementation with NR promotes global remodeling in the rhythmic expression of stress and metabolic gene networks through HSF1

To examine whether NAD⁺ affects genome-wide circadian transcription, we provided NR (~400 mg/kg/day) (Bieganowski and Brenner, 2004) in drinking water ad libitum to 4 month old mice for a period of 4 months, which resulted in a 4-5-fold increase in NAD⁺ levels in liver, soleus muscle, and hypothalamus without changing nicotinamide (Figures 1A, S1A–B). At the age of 8 months, livers were harvested for RNA-sequencing (RNA-seq) across six time points from ad lib-fed animals housed in 12:12 light:dark (LD) conditions. Analysis of rhythmic transcription using eJTK_Cycle revealed that 3,588 hepatic transcripts oscillated in animals provided water only, while NR supplementation reprogrammed ~50% of the oscillating hepatic transcriptome (Figure 1A). We identified three major categories of genes with altered diurnal expression following NR supplementation (Figure 1A), including those exhibiting (i) a loss (825 genes), (ii) a gain (753 genes), or (iii) a phase-shift in oscillation (512 genes) (FDR-adjusted p-values <0.1 or >0.9 for oscillating or non-oscillating genes, respectively) (Figure 1A–B, Table S1). We found that NR supplementation exerted distinct effects on gene expression rhythms at different phases of the light-dark cycle. Of the genes that lost oscillation following NR supplementation, 538 (65%) exhibited a peak in expression in either the early light (ZT2-ZT8) or the early dark (ZT16-ZT18) period in the H₂O condition (Figure 1C). Interestingly, NR supplementation caused de novo oscillation of 505 (67%) of the gain genes within the same early light and early dark periods (Figure 1C). The early dark period also contained the largest group of phase-shifted genes; 120 (23%) genes newly peaked at ZT18 following NR supplementation, of which 79% phase-advanced from a peak at ZT0 in the H₂O condition (Figure 1C). Gene ontology and promoter transcription factor (TF) motif analysis of genes that lost oscillations following NR revealed enrichment in pathways associated with translation, insulin signaling, and ketogenesis and in motifs that were bound by E26 transformation-specific (ETS) and FOXO (Forkhead) family TFs (Figure 2A). Transcripts that gained oscillation encompassed those involved in RNA splicing and nucleotide metabolism and were enriched with homeobox (HOX) and signal transducer and activator of transcription (STAT) motifs (Figure 2A). Phase-shifted genes were primarily involved in the spliceosome and enriched in the heat shock factor (HSF) motif (Figure 2A). We note that analyses in Figures 1–2 are also presented in Figure S1C using FDR-adjusted p-values of 0.05 and 0.95 for oscillating and non-oscillating genes, respectively, to identify genes for promoter motif and gene ontology analyses. Here, we reach the same conclusions as in Figures 1–2 with fewer genes (Figure S1C). These results implicate a broad range of collaborative TFs in mediating reprogramming of circadian transcription of stress and metabolic gene networks in liver following NAD⁺ elevation with NR.

Figure 1. Supplementation with the NAD⁺ precursor nicotinamide riboside elevates tissue NAD⁺ levels and promotes global remodeling of hepatic circadian transcriptome.

(A) Model depicting the NAD⁺ salvage pathway and nicotinamide riboside (NR) conversion to NAD⁺. Hepatic NAD⁺ quantified by HPLC in WT mice provided NR-supplemented or regular water for 4 months (***p≤0.001, ANOVA) (n=3/timepoint/group). eJTK_Cycle analyses following RNA-seq every 4 hrs across the 24 hr day identified transcripts that lost, gained, or phase-shifted >4 hrs oscillations following NR (FDR-adjusted p-value <0.1 and >0.9 for cycling and non-cycling genes, respectively). (B) Normalized counts (DESeq2) plotted for representative transcripts of genes that lost, gained, or phase-shifted (≥4 hrs) oscillations, or were unaffected, following NR. Data are double-plotted for better visualization. (C) For each NR-reprogrammed group, radial histograms show the number of genes whose oscillations peak within each 2-hr window, with the radius corresponding to 150 genes. Dark gray shading indicates phases with the largest number of affected genes. See also Figure S1 and Table S1.

Figure 2. NR drives rhythmic expression of stress and metabolic gene networks through the recruitment of HSF1.

(A) The top enriched promoter-proximal TF motifs (± 2 kb of the transcription start site, JASPAR motif set) and KEGG ontology terms are shown for each NR-reprogrammed group. Each phase is uniquely colored and corresponds to colors in radial histograms in Figure 1C; phases with the largest number of affected genes are indicated by brackets. (B) ChIP-seq for HSF1 in liver (ZT8) following 2 months of NR-supplementation compared to H₂O. Scatter plot depicts average log-transformed tag densities for HSF1 peaks that annotate to NR phase-shifted genes (n=2). UCSC genome browser image of average HSF1 tag density at the Hspb1 locus in liver during H₂O or NR supplementation. Maximum track height is indicated on the y-axis, and gene orientation is indicated. See also Figure S1.

The identification of HSF1 as an NAD⁺-responsive factor highly enriched in phase-shifted genes prompted us to test whether HSF1 chromatin binding was induced following NR treatment. We next performed HSF1 ChIP-sequencing in liver of NR-treated mice at ZT8 and found that NR induced a ~27% increase in HSF1 chromatin binding proximal to genes phase-shifted by NR (p=7.8E-9), which included HSF1 targets such as Hspb1 (Figure 2B) (Mahat et al., 2016). The increase in HSF1 binding occurred independent of changes in body temperature (Figure S1D). Together, these results indicate that HSF1 participates in the phase-shifting response to NR.

NAD⁺ is synthesized de novo from tryptophan and from three vitamin precursors made available by digestion of the NAD⁺ metabolomes in foods (Bogan and Brenner, 2008). Because NAD⁺-consuming enzymes degrade NAD⁺, inhibition of the nicotinamide salvage pathway by Nampt inactivation in combination with fasting is a powerful way to depress NAD⁺-dependent processes (Yang and Sauve, 2016). We therefore next tested whether depletion of NAD⁺ also impacts circadian transcriptional rhythms by performing RNA-seq in 18-hr fasted liver-specific Nampt KO mice (L-Nampt^−/−), which have a 30-60% reduction in hepatic NAD⁺ (Figure 3A). We observed reprogramming of ~50% of the oscillating transcripts in the liver of fasted L-Nampt^−/− mice relative to the fasted control (Figure 3A, Table S2). Importantly, L-Nampt^−/− mice displayed reduced expression of genes that were elevated by NR, and increased expression of those that were down-regulated by NR (binomial test, p=6.387E-11) (Figure 3B). Together with the finding that NR supplementation reprogrammed circadian transcription of stress and metabolic gene networks (Figure 1A), these data indicate that NAD⁺ is necessary and sufficient for circadian transcriptional regulation.

Figure 3. Abrogation of NAD⁺ biosynthesis induces genome-wide disruption of the hepatic circadian transcriptome.

(A) Model depicting the NAD⁺ salvage pathway, highlighting the contribution of NAMPT to NAD⁺ biosynthesis. Hepatic NAD⁺ quantified by HPLC in 4-6 mo old liver-specific Nampt knockout (L-Nampt^−/−) and control (Alb-Cre and Nampt^fx/fx) mice following an 18-hr pre-collection fast (***p<0.001, ANOVA) (n=2-6/timepoint/group). eJTK_Cycle analyses following RNA-seq every 4 hrs across the 24 hr day identified transcripts that lost, gained, or phase-shifted (≥4 hrs) oscillations following Nampt ablation (FDR-adjusted p-value <0.1 and >0.9 for cycling and non-cycling genes, respectively) (n=2-6/timepoint/group). (B) Heat map depicting the maximal diurnal differential transcriptional responses to NR supplementation and NAD⁺ depletion in L-Nampt^−/− mice for the 1500 most up- and down-regulated genes. Log-transformed fold change at each timepoint from NR relative to H₂O (orange) and L-Nampt^−/− relative to control (purple) for representative genes (n=2-6/timepoint/group). See also Figure S1 and Table S2.

Requirement for BMAL1 in chromatin remodeling and HSF1 recruitment following pharmacologic NAD⁺ elevation

Given that increased NAD⁺ by NR supplementation altered the landscape of rhythmic transcription in liver (Figure 1A), we next sought to examine whether BMAL1 was necessary for modulation of rhythmic transcription by NAD⁺ (Figure 4A). We compared 4-month old wild-type (WT) and Bmal1 knockout (Bmal1^−/−) mice following administration of the NAD⁺ precursor nicotinamide mononucleotide (NMN) (Figure S1E). NMN administration allowed us to study the effect of elevation of NAD⁺ at the zenith of BMAL1 DNA binding, which corresponds with the lowest point of NAD⁺ oscillations (ZT6) (Koike et al., 2012). In WT mice, NMN supplementation led to differential expression of 299 genes (FDR-adjusted p-value <0.1), including increased expression of the clock repressors Per1 and Cry2 (Figure 4B, Table S3), indicating that NAD⁺ exerts transcriptional control in part through induction of the clock activators CLOCK/BMAL1. Surprisingly, ~95% of the genes differentially expressed in WT liver were no longer differentially expressed in the Bmal1 mutants, suggesting a primary role for BMAL1 in regulating gene expression in response to NAD⁺ (Figure 4B, Table S3). We next tested whether BMAL1 chromatin binding was affected by NMN supplementation. Indeed, ChIP-seq for BMAL1 showed a ~25% increase in binding density following NMN at ZT6 (p=9.5E-47), including at canonical BMAL1 target genes Per1 and Cry2 (Figure 4C, Table S3). To determine whether BMAL1 was also necessary for the induction of HSF1 during NR supplementation, we performed HSF1 ChIP-seq in both WT and Bmal1^−/− mice supplemented with NR for 2 months at ZT8. Strikingly, we observed that the increase in HSF1 chromatin binding at NR-phase-shifted genes was abrogated in Bmal1 mutants (Figure 4D). Given previous evidence that BMAL1 modulates chromatin structure in liver (Menet et al., 2014), we reasoned that induction of BMAL1 activity from increased NAD⁺ may enhance chromatin accessibility. Indeed, the assay for transposase accessible chromatin-sequencing (ATAC-seq) revealed decreased nucleosome-free signal at HSF1 binding sites at phase-shifted genes in Bmal1^−/− mice, suggesting a role for BMAL1 in establishing HSF1-accessible chromatin binding sites (Figure 4E). Together, these findings show that NAD⁺ drives circadian transcription by stabilizing BMAL1 chromatin binding and facilitating the recruitment of collaborative TFs such as HSF1.

Figure 4. Requirement for BMAL1 in transcriptional reprogramming and HSF1 recruitment following pharmacologic NAD⁺ elevation.

(A) Model to interrogate the role of BMAL1 in NAD⁺-dependent transcriptional oscillations. (B) RNA-seq in liver (ZT6) in WT and Bmal1^−/− mice injected with NMN or PBS at ZT2. (Left) Log-transformed fold-change (NMN/PBS) and FDR-adjusted p-values shown for genes differentially expressed by NMN in WT liver. (Right) Heatmap depicting differentially expressed genes from NMN in WT and Bmal1^−/− liver. RNA-seq data is transformed by row z-score across treatments and genotypes and sorted by fold change from NMN in WT (FDR-adjusted p-value <0.1) (n=2-3). (C) BMAL1 ChIP-seq in liver (ZT6) following NMN or PBS injection at ZT2. Scatter plot of average log-transformed tag-densities for BMAL1 peaks that annotate to genes differentially expressed by NMN in WT liver (n=2). (D) ChIP-seq for HSF1 in WT and Bmal1^−/− liver at ZT8 following 2 months of NR supplementation compared to H₂O. Log-transformed fold-change in average HSF1 tag-density from H₂O-treated WT mice for peaks that annotate to NR phase-shifted genes (n=2) (*** p < 0.001). (E) ATAC-seq in liver (ZT8) of WT and Bmal1^−/− mice. Average nucleosome free reads are quantified at HSF1 peaks that annotate to NR phase-shifted genes. (n=2-4). See also Figure S1 and Table S3.

NAD⁺ and SIRT1 regulate nuclear localization of the clock repressor complex

NAD⁺ is an obligate substrate of the major nuclear class III deacetylase SIRT1 (Imai et al., 2000) which is also found in complex with CLOCK/BMAL1 (Asher et al., 2008; Nakahata et al., 2008); yet the role of SIRT1 in genome-wide transcription during NR supplementation has not been determined. To evaluate whether NAD⁺-induced genome-wide reprogramming requires SIRT1, we performed RNA- and BMAL1 ChIP-seq following elevation of NAD⁺ through the acute administration of NMN in liver-specific Sirt1 KO (L-Sirt1^−/−) and control mice (Figure 5A). NMN was injected at ZT2 and tissues collected at ZT6 in order to elevate NAD⁺ at the time of day when it is lowest (Figure 3A). We detected 277 genes that were significantly differentially expressed by NMN in the control mice (FDR-adjusted p-value <0.1), with 140 genes up-regulated (Log₂FC>0, red dots) and 137 genes down-regulated (Log₂FC<0, blue dots) by NMN (Figure 5A). In L-Sirt1^−/− liver, 233 (84%) of the NMN-responsive genes were no longer significantly differentially expressed and 260 (94%) had a smaller absolute fold change in expression from NMN (binomial test, p=2.2E-16) (Figure 5A), indicating that SIRT1 is necessary for the response to NAD⁺. Similarly, loss of SIRT1 prevented induction of BMAL1 chromatin binding in response to NMN (Figure 5A, Table S3) (p<2.15E-140). Together, these data indicate that NAD⁺ mediates global transcription through both SIRT1 and BMAL1.

Figure 5. Pharmacologic and genetic manipulations of NAD⁺ and SIRT1 mediate nuclear localization of the clock repressor complex through PER2 deacetylation to control transcription.

(A) (Left and middle) RNA-seq in liver (ZT6) in control (Sirt1^fx/fx) and liver-specific Sirt1 knockout (L-Sirt1^−/−) (Alb-Cre;Sirt^−/−) mice following NMN or PBS injection at ZT2. (Left) Data for heatmap was transformed by row z-score across treatments and genotypes and sorted by fold change from NMN in control (n=3-4). (Middle) Log-transformed fold-change (NMN/PBS) in gene expression in control and L-Sirt1^−/− mice shown for genes differentially expressed by NMN in control liver (FDR-adjusted p-value <0.1). Red and blue dots indicate genes that are induced or repressed, respectively, by NMN in control liver (n=3-4). (Right) BMAL1 ChIP-seq in control and L-Sirt1^−/− liver (ZT6) following NMN or PBS injection at ZT2. Heatmap depicts row z-score transformed average tag-densities for BMAL1 peaks that annotate to genes differentially expressed by NMN in WT liver (n=2). (B) Clock repressor protein levels in nuclear fractions from dexamethasone-synchronized Sirt1^fx/fx and Sirt1^−/− MEFs. “n.s.” indicates non-specific band. (C) PER2 half-life (hr) following cycloheximide (CHX) addition to DMSO- or FK866-treated MEFs expressing the PER2::LUCIFERASE protein from the endogenous Per2 locus. (***p<0.01). (D) CRY1 western blotting following PER2-FLAG immunoprecipitation from nuclei of HEK293 cells treated with either FK866 (10 nM, 24 hr) or EX527 (1μM, 24 hr). (E) Acetyl-lysine and FLAG Western blotting of immunoprecipitated FLAG-PER2 following FK866 or EX527 treatment in HEK293 cells. (F) IP-LC-MS/MS analysis of PER2 immunoprecipitated from FK866-treated HEK293 cells. Frequency of unique spectra containing acetyl-lysine modifications shown across 4 replicates and representative spectra for PER2 acetyl K680. (G) Western blotting of CRY1 following PER2-FLAG immunoprecipitation from nuclei of HEK293 cells expressing WT, acetyl-mimetic (K680Q), or de-acetyl-mimetic (K680R) PER2 mutants. The IgG sample was run on the same gel as the other samples shown. (H) Biochemical model demonstrating mechanism by which NAD⁺ alleviates prolonged repression of BMAL1 by nuclear PER2 monomer: SIRT1 uses NAD⁺ to deacetylate PER2 on lysine 680, driving dimerization with CRY1, PER2 phosphorylation at the FASPS site, and ultimately PER2 degradation. See also Figures S2 and S3 and Tables S3 and S4.

While the transcriptional co-activator PGC1α is activated by SIRT1 and has been reported to induce Bmal1 RNA expression (Foteinou et al., 2018; Liu et al., 2007; Rodgers et al., 2005), we did not observe changes in Bmal1 RNA with either NMN or in liver-specific Sirt1 knockout animals (Table S3); however, there were marked increases in the RNA expression of the clock repressors Per1 and Cry2 in response to NAD⁺ (Figures 1A, 4B, Table S3). These observations suggest that NAD⁺ influences BMAL1 transcriptional activity rather than expression of Bmal1 through PGC1α. SIRT1 inactivation has also been reported to increase H3K9/14 acetylation at the BMAL1-binding site at the Dbp locus (Nakahata et al., 2008). However, ChIP-sequencing for histone acetyl marks that are SIRT1-dependent (H3K9Ac) and SIRT1-independent (H3K27Ac) following NR supplementation revealed increased acetylation of both marks at the Dbp locus and genome-wide (Figure S2A). Specifically, NR increased the average H3K9Ac tag density for all peaks identified genome-wide by 31% (p<3.7E-264) and specifically increased tag density for peaks at the Dbp locus by 29% (p<0.00043) (Figure S2A). Similarly, NR increased the average H3K27Ac tag-density at all peaks identified across the genome by 26% (p<1.4E-139) and at peaks at the Dbp locus by 22% (p<0.0015) (Figure S2A). Together, increased histone acetylation and SIRT1-dependent and independent sites following NR supplementation suggests that these effects are not directly dependent on SIRT1. Lastly, we were unable to observe differences in BMAL1 acetylation using an anti-acetyl-BMAL1 antibody (Millipore AB15396) (data not shown).

SIRT1 has been reported to promote the deacetylation and degradation of PER2 (Asher et al., 2008), although the molecular mechanism whereby PER2 acetylation affects PER2 stability remain unknown. Given that NAD⁺ did not appear to directly affect transcription of Bmal1 or acetylation of histones, we sought to examine whether NAD⁺ regulates SIRT1-dependent recruitment of BMAL1 to chromatin through control of the PER/CRY repressor complex. During the light period, cytoplasmic sequestration of clock repressors PER and CRY enables sustained CLOCK/BMAL1 activation of transcription. During the dark period, PER and CRY proteins translocate into the nucleus where they inhibit CLOCK/BMAL1 (Aryal et al., 2017; Hardin et al., 1990; Koike et al., 2012; Kume et al., 1999; Meyer et al., 2006; Michael et al., 2017; Yagita et al., 2002; Ye et al., 2014). Phosphorylation and ubiquitin-mediated degradation of PER and CRY then inhibits the repressive phase (Eide et al., 2005). Since PER and CRY proteins undergo nuclear transport, association, and stability (Gallego and Virshup, 2007; Meng et al., 2008), we first sought to test whether NAD⁺/SIRT1 impacts CRY and PER subcellular distribution across the day. Surprisingly, in Sirt1 knockout cells, we found markedly increased nuclear accumulation of PER2 compared to controls, whereas we observed reduced levels of nuclear PER1, CRY1, and CRY2 (Figures 5B, S2B). Further, PER2 exhibited increased stability upon cycloheximide chase following NAD⁺ depletion with the NAMPT inhibitor FK866 (Figures 5C). Finally, PER2-FLAG displayed reduced binding to CRY1 within the nucleus following either NAMPT inhibition with FK866 or SIRT1 inhibition with EX527 (Figure 5D). These findings indicate that NAD⁺-SIRT1 controls the spatiotemporal distribution, abundance, and oligomerization of core clock repressor proteins.

We next hypothesized that NAD⁺-SIRT1 may control changes in PER2 localization and stability through posttranslational modification (PTM) of PER2. Indeed, NAMPT and SIRT1 inhibition with FK866 and EX527 treatment, respectively, increased acetylation of FLAG-PER2 (Figure 5E), consistent with the prior report that loss of SIRT1 increased PER2 acetylation (Asher et al., 2008). To identify the specific acetylated residues on PER2, we subjected PER2 to LC-MS/MS under conditions of NAD⁺ depletion through FK866 treatment and identified hyperacetylation of PER2 at lysine 680 (Figure 5F, Table S4). Mutation of PER2^K680 to the acetyl-mimetic PER2^K680Q revealed decreased association with the repressor CRY1 within nucleus relative to WT and the deacetyl-mimetic PER2^K680R (Figure 5G), consistent with a role for NAD⁺ and SIRT1 in the association of PER2 with CRY1 (Figure 5D). Moreover, the primary amino acid sequence proximate to PER2^K680 is similar to that of Drosophila PER and is known to mediate intermolecular complex formation with CRY1 (Hennig et al., 2009; Yildiz et al., 2005). These data indicate that acetylation modulates the nucleocytoplasmic transport of the PER2-CRY1 complex.

In addition to acetylation of PER2^K680, we observed regulation of PER2 phosphorylation following NAMPT inhibition with FK866. Indeed, acetylation of PER2^K680 occurs near the orthologous sites of impaired phosphorylation in human (PER2^S659) and Drosophila (PER2^S693) variants that cause circadian and sleep disruption (Bargiello and Young, 1984; Bargiello et al., 1984; Reddy et al., 1984; Toh et al., 2001; Top et al., 2018; Zehring et al., 1984). We identified numerous sites of PER2 phosphorylation in NAD⁺-depleted conditions (Figure S3A–B), including at PER2^S693, which is orthologous to the PER^S site in Drosophila that regulates its stability and activity (Garbe et al., 2013; Toh et al., 2001; Vanselow et al., 2006) and near the β-TRCP binding domain, which is necessary for ubiquitin-mediated degradation of PER2 (Figure S3A–B) (Zhou et al., 2015). We next sought to test whether acetylation of PER2^K680 is critical in priming phosphorylation of PER2^S659, which is mutated in familial advanced sleep-phase syndrome (FASPS) in humans (Toh et al., 2001). Immunoprecipitation of the acetyl-mimetic mutant PER2^K680Q and immunoblotting for phosphorylation at PER2^S659 revealed decreased phosphorylation at this site relative to WT PER2 and the deacetyl-mimetic PER2^K680R mutant (Figure S3C). Thus, NAD⁺/SIRT1 controls PER2 acetylation and coordinates PER2 phosphorylation through a pathway similar to that involved in human sleep/wake regulation (Figure 5H).

NR enhances circadian transcription and evening locomotor activity in old mice

Aging is associated with decreased circadian amplitude and robustness in both mice and humans (Bass and Lazar, 2016; Campisi et al., 2019; Rey and Reddy, 2013). Further, Bmal1 mutant mice display premature aging and have low NAD⁺ levels (Kondratov et al., 2006; Ramsey et al., 2009). NAD⁺ levels decline with age, and provision of both NR and NMN have been shown to ameliorate age-associated insulin resistance, protect against noise-induced hearing loss, central brain injury, heart failure, diabetic neuropathy, enhance postpartum functions and extend lifespan in rodents (Brown et al., 2014; Cantó et al., 2012; Diguet et al., 2018; Ear et al., 2019; Massudi et al., 2012; Trammell et al., 2016; Vaur et al., 2017; Yoshino et al., 2011; Zhang et al., 2016). We therefore sought to test the hypothesis that restoration of NAD⁺ levels in old mice might counteract the age-related decline in circadian rhythms (Figure 6A). We chose to examine “young” mice at 10 months of age and “old” mice at 22 months of age because NAD⁺ declines in mice older than 12 months of age and there is no change in weight between mice at these ages (Figure S4A) (Braidy et al., 2011; Camacho-Pereira et al., 2016; Turturro et al., 1999). We first compared levels of the NAD⁺-SIRT1 target clock protein PER2 in young and old mice, in addition to determining the genome-wide binding of BMAL1 in liver from the same populations of young and old mice (Figure 6B–C). We observed an increase in PER2 in old mice, consistent with enhanced repression (Figure 6B). Additionally, old mice displayed reduced amplitude of BMAL1 binding across the day with significantly reduced binding during the light period (Figures 6C, S4B, Table S5). Remarkably, NR supplementation for 6 months in old mice (Figure S4A) led to a dramatic reduction in PER2 (Figure 6B) and a corresponding induction in BMAL1 DNA binding (p=1.4E-237), particularly at the zenith of BMAL1 activity (ZT8) (Koike et al., 2012) (Figures 6C, S4B, Table S5). Gene ontology analysis of genes proximal to BMAL1 peaks in NR- but not H₂O-treated old mice revealed enrichment in pathways associated with metabolism and longevity and in genes with protein structures containing the NAD⁺-binding Rossman-fold (Figure S4C). Of note, the age-dependent decline in BMAL1 chromatin occupancy was independent of changes in total BMAL1 protein levels (Figure 6B). Together, our findings indicate that decreased NAD⁺ in old mice causes increased PER2 that diminishes circadian rhythms.

Figure 6. NR enhances circadian transcription and evening locomotor activity in old mice.

(A) Model to interrogate the contribution of NAD⁺ to circadian genomic and behavioral rhythmic phenotypes during aging. (B) PER2 and BMAL1 levels in liver (ZT8) of 10 month (“young”) and 22 month (“old”) WT mice treated either with NR-supplemented or regular drinking water for 6 months. (C) (Top) BMAL1 ChIP-seq in liver (ZT8 and ZT20). Z-score normalized average tag-densities for genomic regions corresponding to the center of each BMAL1 peak identified in old mice treated with NR ± 1 kb and sorted by size (n=2) (Bottom) Average BMAL1 tag-density and standard error for both replicates within each group. (right) Average background-subtracted PER2::LUC oscillations from soleus of young and old water- and NR-treated mice transgenic for the Per2::Luciferase reporter monitored ex vivo in a lumicycle (n=3). (D) Oxygen consumption rates (OCR) of mitochondria isolated from young and old, water- and NR-treated liver (ZT8 and ZT20) by Seahorse Bioanalyzer. FCCP-treated respiration with palmitate relative to basal (*p<0.05) (n=4-5). (E) (Left) Representative average 24 hr profile of ad lib wheel running activity of young and old water- and NR-treated WT mice. Arrow denotes ‘late-night activity’. (Right) Quantification of average wheel revolutions/min in the late night (ZT22-0) (n=8-10) (*p<0.05). See also Figure S4 and Table S5.

Since NAD⁺ regulates transcription by activating BMAL1 in young mice, we hypothesized that age-associated decline in BMAL1 chromatin binding and the correspondingly low NAD⁺ was associated with alterations in circadian transcription. To test this, we first measured BMAL1 transcriptional activity using live-cell luminometry recording in old (22 mo) genetic reporter mice expressing the Per2::Luciferase(Luc) transcriptional reporter compared to young (10 mo) mice and under NAD⁺ repletion with NR (Yoo et al., 2005) (Figure 6C). We observed a 90% decrease in amplitude of PER2-LUC oscillations in soleus of old mice (p<0.03) that was restored in old NR-treated mice to within 40% of the amplitude in young mice (Figure 6C). To next determine whether decreased NAD⁺ with age altered circadian transcriptional reprogramming by collaborative TFs, we performed ChIP-seq for HSF1 (Figures 2, 4D–E). Similar to the reduced genome-wide DNA binding by BMAL1 in old mice, HSF1 also displayed reduced chromatin binding during the daytime in old mice which was restored following NR supplementation (Figure S4D). Aging is also associated with decline in mitochondrial function (Petersen et al., 2003), a clock-controlled process (Peek et al., 2013; Turek et al., 2005). Indeed, “metabolism” is a top GO term associated with decreased BMAL1 chromatin binding in old age (Figure S4C). Thus, we analyzed respiration in isolated mitochondria harvested from liver of young and old mice provided chronic supplementation with NR. Intriguingly, we observed reduced oxygen consumption in old mice in the presence of palmitic acid specifically at ZT20, which was rescued by NR supplementation (Figure 6D). Thus, restoration of NAD⁺ levels in old age with NR impacts BMAL1 chromatin binding, stress-responsive transcription and a circadian decline in fat oxidation.

In addition to cell and molecular signatures of reduced circadian function, a behavioral hallmark of aging in humans is disruption of circadian sleep-wake rhythms especially in the evening (Buysse et al., 1992; Ohayon et al., 2004; Roenneberg et al., 2007; Walch et al., 2016). We therefore next assessed the impact of aging and NAD⁺ elevation by NR supplementation on 24-hr rest-activity patterns of old mice. While old mice with low NAD⁺ displayed significantly diminished late-night activity between ZT22 and ZT0 (a period that corresponds to human late afternoon), NR supplementation restored this locomotor activity (denoted by arrow in Figure 6E). Similarly, animals deficient in NAD⁺ (CAG^CreER; Nampt^fx/+) displayed reduced late-night locomotor activity (Figure S4E). Thus, countering the decline in NAD⁺ in old age leads to both restoration of youthful patterns of transcription and circadian physical activity.

Discussion

NAD⁺ coordination of transcription and behavior through the molecular clock

Here we elucidated the mechanism of transcriptional remodeling by NAD⁺ and defined a role for the core molecular clock in this process. Studies in yeast were the first to identify NAD⁺ as a cofactor for histone-modifying enzymes involved in the response to caloric restriction, and subsequent work implicated an intersection between the NAD⁺ pathway and the core clock. Further, in both aging and metabolic stress, there is decline in NAD⁺, yet how changes in NAD⁺ might modulate activity of core clock transcription factors has remained incompletely known. Given that pharmacologic NAD⁺ elevation improves healthspan (Brown et al., 2014; Cantó et al., 2012; Ear et al., 2019; Massudi et al., 2012; Trammell et al., 2016; Vaur et al., 2017; Yoshino et al., 2011; Zhang et al., 2016), the potential remains that some of the benefits of NAD⁺ may involve circadian reprogramming. Results presented here implicate a role for NAD⁺ in remodeling the circadian transcriptome through deacetylation of the core clock repressor PER2, which in turn promotes the activity of CLOCK/BMAL1. Our studies clarify the mechanisms underlying SIRT1-mediated regulation of the clock, for which discrepant mechanisms were previously proposed (Asher et al., 2008; Nakahata et al., 2008). Our identification of acetylation on PER2^K680 by mass spectrometry and implication of this site in nuclear PER2/CRY1 heterodimer formation and CK1-dependent phosphorylation supports findings that proposed a role for SIRT1 in destabilizing PER2 (Asher et al., 2008). Further, our discovery that NR increases acetylation of both H3K9 and H3K27 genome-wide and at the Dbp locus suggests that changes in histone acetylation may be an indirect effect of SIRT1 activity.

Time of day as a variable in the metabolic control of epigenetics

Increasing evidence suggests that metabolic cues provide feedback to regulate epigenetic enzymes important in gene expression. For example, metabolites produced through the TCA cycle, such as alpha-ketoglutarate and 2-hydroxyglutarate, are modulators of the jumonji family of demethylases (Dang et al., 2009; Losman et al., 2013; Tsukada et al., 2006). However, how NAD⁺ affects global transcription has remained an open question. Our observation that NR supplementation exerts broad effects on rhythmic gene transcription in liver and our identification of NAD⁺ as a regulator of SIRT1-mediated deacetylation of PER2 establishes a mechanism by which NAD⁺ regulates the clock. It will be interesting to further probe the effects of NAD⁺ on clock repressor complex formation and nuclear-cytoplasmic shuttling. Supplementation with NR revealed a collaboration with HSF1 in a phase-specific set of remodeled genes, consistent with transcriptomic evidence that HSF1 induces transcription of metabolic genes (Mahat et al., 2016). Additionally, recent genomic analyses in liver following high-fat feeding suggest that macronutrients reprogram transcription through the recruitment of collaborative TFs (SREBP/PPARα) (Guan et al., 2018). In this regard, core clock proteins have been shown to shape the chromatin landscape through direct effects on nucleosome remodeling (Menet et al., 2014), through long-range chromatin interactions (Aguilar-Arnal et al., 2013; Kim et al., 2018; Mermet et al., 2018; Sobel et al., 2017), and through recruitment of regulatory cofactors (Duong et al., 2011). One possibility is that local changes in chromatin accessibility may arise in response to tethering of SIRT1 to the CLOCK/BMAL1/PER2 complex. Such interactions may affect phase-specific recruitment of collaborative TFs, many of which are SIRT1 targets (Purushotham et al., 2009; Walker et al., 2010; Westerheide et al., 2009). Conversely, reciprocal regulation of clock oscillations by temperature may involve similar TF interactions (Buhr et al., 2010; Reinke et al., 2008; Saini et al., 2012). It is tempting to speculate that cells undergoing shifts in NAD⁺ metabolism may exhibit alterations in rhythmic transcription, such as in immune cells with activation and genotoxic stress.

Epigenetic regulation of circadian behavioral and metabolic cycles in aging and disease

Disruption of circadian patterns of sleep/wake activity, as occurs in shiftwork, neurodegenerative disorders, and aging, has been tied to metabolic disorders. Our results show that NAD⁺ acts through SIRT1 to deacetylate PER2 and regulate phosphorylation of a CK1 site on PER2 that is mutated in human familial advanced sleep phase (Toh et al., 2001). Whether NAD₊ repletion might enhance metabolism through the alignment of feeding and sleep time remains to be established. One therapeutic implication may be that targeting SIRT1 activity via NAD⁺ (Ondracek et al., 2017) may ameliorate age-related decline in behavior and metabolism . At the cellular level, modulating circadian function through targeting PER2^K680 may provide a therapeutic avenue in disease contexts including sleep loss associated with neurodegenerative and shiftwork disorders. Enhancing the amplitude of daily rhythms in oxidative metabolism may be one mechanism by which NAD⁺ repletion improves resistance to metabolic stress with age.

STAR METHODS

Resource Availability

Lead Contact

Further information and request for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Joseph Bass (j-bass@northwestern.edu).

Materials Availability

No new materials or reagents were generated in this study.

Data and Code Availability

RNA-seq, ChIP-seq, and ATAC-seq data sets have been deposited in NCBI’s Gene Expression Omnibus database (GSE133989).

Experimental Model and Subject Details

Mice

All animal procedures were in accordance with guidelines of the Institutional Animal Care and Use Committee, and all mice were housed at 23-25°C in the Center for Comparative Medicine at Northwestern University and maintained under 12hr light:12hr dark (LD) cycles with ad lib access to regular chow and water unless otherwise indicated. Mice were maintained on Teklad chow 7912 containing 100 mg niacin / kg diet for the duration of this study in order to provide vitamin B3 in excess of daily nutritional requirements of mice (15 mg niacin / kg diet) (National Research Council (US) Subcommittee on Laboratory Animal Nutrition, 1995). 4 and 16 month old wild-type C57BL/6NCrl male mice were obtained from Charles River or the NIA colony at Charles River, respectively. Bmal1^−/− and Per2::Luciferase mice were a gift from J. Takahashi (UT Southwestern). Namp^fx/fx and Sirt1^fx/fx mice were a gift from S. Imai (Washington University in St. Louis). Alb-Cre and CAG^CRE-ER mice were purchased from Jackson Laboratories. Mice crossed to CAG^CRE-ER received 3 tamoxifen (Sigma) injections (200 μg/g body weight i.p. in corn oil) administered every other day.

Cell Lines

Per2::Luciferase MEFs were a gift from J. Takahashi (UT Southwestern). Sirt1^fx/fx MEFs were isolated from pregnant mice at embryonic day 14 or 15 following trypsin digestion as previously (Peek et al., 2013). MEFs were immortalized by lentivirus transduction of shRNAs targeting p19ARF (Addgene 14090, Northwestern Skin Disease Research Center). Recombination of LoxP sites was achieved through 3 treatments with adenovirus expressing Cre or empty vector (Vector Biolabs, MOI 500). MEFs were synchronized with dexamethasone (100 nM, 30 min) and collected at indicated timepoints. pCMV-SPORT2-PER2 expressing plasmid (Addgene 16204) was subjected to site-directed mutagenesis according to manufacturer instructions (NEB) and transiently transfected into low-passage HEK293 cells (Clontech).

Methods Details

Pharmacologic NAD⁺ supplementation

Freshly prepared, sterile-filtered nicotinamide mononucleotide (NMN, Oriental Yeast Company) was injected (i.p. 500 mg/kg in PBS) 4 hours prior to sample collection. Nicotinamide riboside (NR, Chromadex) was prepared to 3.2 g/L in drinking water, sterile filtered, and provided ad lib to mice in autoclaved, light-protected water bottles for 4 months to examine the impact of NR supplementation on the hepatic circadian transcriptome (Cantó et al., 2012). For HSF1 ChIPsequencing studies, NR was provided for 2 months since Canto et al demonstrated that similar metabolic phenotypes were apparent even after only 1 month of treatment (Cantó et al., 2012). NR was replaced every 3 days after lights off to prevent bacterial growth and minimize behavioral disruptions. Water consumption rates were monitored for 1 month and were not different between H₂O and NR-supplemented mice. Mice drank ~3.75 mL water per day, achieving a dose of ~400 mg/kg in a 30 g mouse.

Nucleotide measurements

NAD⁺ was quantified as previously described (Peek et al., 2013). Briefly, cut and weighed tissue was homogenized in perchloric acid (Sigma) in the TissueLyzer (Qiagen). Following neutralization with K₂CO₃, supernatant was diluted 1:1 in mobile phase and analyzed by HPLC (Shimadzu) on a Supelco LC-18 column (Sigma), UV-Vis detector at 260 nm. Nicotinamide was quantified as previously described (Trammell and Brenner, 2013) by buffered, boiled ethanol extraction followed by liquid-chromatography mass-spectrometry by NRomics at the University of Iowa.

RNA-sequencing and analysis

RNA was isolated from 10-20mg of liver tissue using trizol reagent and purified on the Zymo Direct-Zol RNA miniprep kit. RNA quality was assessed by Bioanalyzer (Agilent). RNAs with a RNA Integrity Number (RIN) greater than 8 were used for library prep. Libraries were constructed from 250 ng of RNA using the NEBNext RNA Ultra Directional library prep kit from NEB. Average library size and concentration were determined by Bioanalyzer and qPCR (NEBNext Library Quant kit), respectively, prior to pooling. Pooled libraries were quantified by qPCR and loaded onto a NextSeq 500/550 v2 high output flow cell from Illumina (FC-404-2002) and 75bp single-end sequencing was performed on a NextSeq 500 sequencer. Libraries were sequenced to an average depth of ~15M or ~40M aligned reads for rhythmic or differential analyses, respectively. Using standard parameters, sequences were aligned to the mm10 genome with STAR (v2.5.2) and assigned to Ensembl features (GRCm38.vM12) with subread:featureCounts (v1.5.1).

Analysis of differential expression at each timepoint was performed with standard parameters in DESeq2 (v1.24.0) after removal of genes with 0 read counts. FDR-adjusted p-value cutoff for significance was set to 0.1. DESeq2-normalized counts were used for analysis of rhythmic expression with eJTK_Cycle (v3.1.R). Period lengths were allowed between 20 and 28 hours to account for sampling frequency (4 hours). Adjusted p-values of 0.1 and 0.9 were chosen for ‘rhythmic’ and ‘non-rhythmic’ genes respectively. Genes were classified as “unaffected” if they were similarly rhythmic in both conditions, as “gain” if they were non-rhythmic in the control condition and rhythmic in the experimental condition, as “loss” if they were rhythmic in the control condition and non-rhythmic in the experimental condition, and “phase-shift” if they were rhythmic in both conditions but were annotated with peak phases that were greater than 4 hours different. Heatmaps were sorted by peak phase for the control condition for loss, unaffected, and phase-shift reprogramming groups, while they were sorted by peak phase of the experimental condition for the gain reprogramming group. Gene ontology and promoter motif analysis were performed on genes from each reprogramming group that were sub-divided by phase. Gene ontology analysis of the KEGG term database was performed with the HOMER findGO command. Enrichment of known motifs +/− 2 kb of the TSS were determined using the HOMER findMotif command. The JASPAR vertebrate DNA binding motif database was used as the known motif set with a threshold set to 6.5.

ChIP-sequencing and analysis

Liver was collected at the indicated time points and processed for ChIP-seq as previously published (Perelis et al., 2015). Briefly, liver was fixed in disuccinimidyl glutarate (2mM in PBS + 1% DMSO, 30 min) and formaldehyde (1% in PBS, 10 min) prior to aliquoting and freezing at −80°C. For BMAL1 (antibody produced in Bass Lab as in (Perelis et al., 2015), Millipore ABE2599) and HSF1 (Cell Signaling 4356) ChIP-Seq, nucleii from a quarter of a liver were isolated in buffer (150 mM NaCl, 5 mM EDTA pH8, 50 mM Tris-HCl pH8, 0.35% NP-40) in the presence of protease inhibitors by needle, then sheared in 1% SDS buffer in a sonicator (diagenode) for 6 cycles 30on/30off at 4°C. An aliquot for input samples was taken after shearing. Chromatin was diluted 1:10 in dilution buffer (.01% SDS, 1.1% Triton X-100, 167 mM NaCl, 1.2 mM EDTA pH8, 1.67 mM Tris-HCl pH8), and immunoprecipitations with 15 μg of antibody were performed overnight at 4°C, rotating. Secondary-conjugated, BSA-blocked paramagnetic beads were used to pull down protein/DNA complexes. Chromatin was de-crosslinked and purified by Qiagen MinElute column prior to library prep with NEBNext Ultra II library prep kit. Inputs were quantified by Qubit Fluorometer (Invitrogen) for each liver, and 2 ng of equimolar pools of the replicates per condition were processed into libraries as above. Libraries were size-selected to 200-600 bp on the Sage Pippin Prep prior to PCR amplification. Libraries were processed and pooled for sequencing as above. Libraries were sequenced to an average depth of ~15M aligned reads.

Bowtie2 (v2.2.4) was used to align sequencing data to the mm10 genome with standard parameters. For each ChIP, peaks that enrich over the appropriate input were called using the HOMER findPeaks command with settings -style factor, -size 275, -fragLength 250. For scatter plots, heatmaps, box and whisker plots, tag-density for individual peaks for each replicate as described was quantified by HOMER annotatePeaks with setting -size given, and averaged. For histograms, the average tag-density across numerous peaks for each replicate was determined with annotatePeaks with the settings -hist 5, size −2000. For gene ontology analyses, sam files for both replicates were merged (samtools1.2 merge command) and peaks that enrich over the input were identified as above and compared across groups with HOMER mergePeaks. Gene ontology analysis of KEGG and Interpro databases was performed with HOMER findGO command using non-redundant genes annotated from identified peaks. For tornado plots, tag-density +/− 1kb from the center of the peak as described was generated from merged sam files using the HOMER annotatePeaks command with settings -ghist 25, and -size 2000. Tornado plots were z-score normalized and sorted from max-min signal according to old, NR-treated mice at ZT8.

ATAC-sequencing and analysis

100 mg of fresh liver was dounced in ice-cold PBS, passed through a 40 uM filter, and 50,000 cells were processed as described previously (Corces et al., 2017). Briefly, samples were resuspended in ATAC-RSB (10 mM Tris-HCl pH 7.4, 10 mM NaCl, and 3 mM MgCl₂ in water) containing 0.1% IGEPAL, 0.1% Tween-20, and 0.01% digitonin and incubated on ice for 3 min prior to 20x dilution in ATAC-RSB containing 0.01% Tween-20. Nuclear pellets were subjected to tagmentation at 37°C for 30 min in Nextera TD buffer (Illumina) containing 33% PBS, 0.01% digitonin, 0.1% Tween-20, and 2.5 units Nextera transposase enzyme (Illumina). Following transposition reaction, DNA was purified using MinElute columns (Qiagen), subjected to 9 amplification cycles, size-selected by Sage Pippin Prep for fragments between 100-1000 bp, and sequenced by 42 bp paired-end sequencing to an average depth of ~50M aligned reads per sample. Nucleosome-free and nucleosome-bound fragments were segregated based on size distribution using the R Package ATACSeqQ (Ou et al., 2018). The HOMER findPeaks command with settings -size 150 and -fragLength 41 was used to find nucleosome free regions/peaks. Scatterplots and histograms were made as described in the ChIP-sequencing method above.

Western blotting

Transfections occurred 24 hours after seeding and media changed at 48 hours after seeding. Cells were collected at indicated time points or 72 hours after seeding. FK866 (10 nM) (Sigma) or EX527 (1μM) (Sigma) was diluted in DMSO and supplemented to cells without changing media 24 hours before collection. For subcellular localization experiments, cytoplasmic fractions were isolated from the supernatant following treatment with CER1/CER2 reagents according to manufacturer instructions (Thermo, NE-PER kit). Nuclear fractions were isolated by resuspending the nuclear pellet in 5 volumes RIPA buffer and shearing (Diagenode) 3 cycles at 30sec on/30sec off. Whole cell lysates were isolated by treating cell pellets with RIPA as above. All lysis buffers contained 1X protease inhibitors (Roche), 1X phosphatase inhibitors (Roche), and 1μM Trichostatin A (Sigma) and 20mM nicotinamide (Sigma) to inhibit deacetylases. Anti-FLAG-PER2 immunoprecipitations were carried out by incubating 0.25 mg protein lysate with 30 μL of 50% slurry of anti-FLAG conjugated paramagnetic beads (Sigma) in PBS + 5% Tween-20 in protein lo-bind tubes (Eppendorf) overnight at 4°C, rotating. Western blots for H3 (CST 9715), β-ACTIN (CST 4970), GAPDH (CST 5174), BMAL1 (Santa Cruz sc-48790), p-PER2-S659 (Thermo PA5-38901), PER2 (gift from C. Lee (Florida State University)), PER1 (Abcam ab136451), CRY2 (Alpha Diagnostic CRY21-A), CRY1 (Bethyl A302-614A), and FLAG-HRP (Sigma). HRP-conjugated secondary antibodies were from Sigma and Rockland (TrueBlot).

Ex vivo PER2::LUC measurements

Soleus was excised from transgenic mice expressing the full-length PER2 fused to LUCIFERASE (PER2::LUC) and cultured ex vivo in a lumicycle (Actimetrics) as previously described (Yoo et al., 2005). Briefly, the full-length soleus muscle was spread flat on a 0.2 μm filter (Millipore) exposed to luciferin-containing media (Dulbecco’s modified Eagle’s medium (DMEM; Gibco, 1.2 ml) containing sodium bicarbonate (352.5 μg/ml), 10 mM HEPES (Gibco), 2 mM L-glutamine, 2% B-27 serum-free supplement (Invitrogen), penicillin (25 U/ml), streptomycin (Gibco, 20 μg/ml), and 0.1 mM luciferin sodium salt (Biosynth AG)) on the basal side. Dishes were sealed with vacuum grease and a round cover slip and maintained in a lumicycle at 37°C. Amplitude was determined by calculating the half-difference in height of the background-subtracted maxima and minima on the 3^rd phase of oscillation (~3 days after start of recordings).

PER2::LUC stability

MEFs expressing full-length PER2 fused to LUCIFERASE (PER2::LUC) were tested for PER2 stability in a lumicycle as described previously (Zhou et al., 2015). Briefly, MEFs were synchronized with dexamethasone, then incubated in a lumicycle (Actimetrics) in media containing luciferin as above. 17 hours post-shock, FK866 (30 nM) or DMSO was added to the dish without changing the media. 22 hours post-shock, cycloheximide (40 μg/mL) was added to the dish without changing the media. Background-subtracted luminescence readings were normalized to signal at 21.5 hours post-shock. Half-life was calculated as the time following cycloheximide addition in which the luminescence signal was decreased by greater than 50%.

LC-MS/MS analysis of post-translational modifications

FLAG-tagged PER2 was overexpressed in HEK293 cells treated with FK866 (Sigma) or DMSO and immunoprecipitated from whole-cell lysate as above. Immunoprecipitate was gel-purified and a ~1 cm x 0.5 cm piece was cut from the gel at the location that corresponds to the predominant FLAG-PER2 band (between 150 kDa and 160 kDa) and transferred to protein lo-bind tubes (Eppendorf). Gel pieces were washed with ammonium bicarbonate (Ambic, Fisher) and acetonitrile (Fisher), reduced with dithiothreitol (Promega), washed with Ambic/iodoacetamide (Sigma), then subjected to in-gel proteolysis with 300 ng trypsin (Promega) at 30°C overnight, and washed with Ambic and formic acid (ThermoFisher). Samples were then submitted to the Northwestern Proteomics Center of Excellence where peptides were analyzed by LC-MS/MS using a using a Dionex UltiMate 3000 Rapid Separation nanoLC and a Q Exactive™ HF Hybrid Quadrupole-Orbitrap™ Mass Spectrometer (Thermo Fisher Scientific Inc, San Jose, CA). Approximately 1 μg of peptide samples was loaded onto the trap column, which was 150 μm x 3 cm in-house packed with 3 um C18 beads. The analytical column was a 75 um x 10.5 cm PicoChip column packed with 3 um C18 beads (New Objective, Inc. Woburn, MA). The flow rate was kept at 300nL/min. Solvent A was 0.1% FA in water and Solvent B was 0.1% FA in ACN. The peptide was separated on a 120-min analytical gradient from 5% ACN/0.1% FA to 40% ACN/0.1% FA. The mass spectrometer was operated in data-dependent mode. The source voltage was 2.10 kV and the capillary temperature was 320 degrees C. MS¹ scans were acquired from 300-2000m/z at 60,000 resolving power and automatic gain control (AGC) set to 3x10⁶. The top 15 most abundant precursor ions in each MS¹ scan were selected for fragmentation. Precursors were selected with an isolation width of 2 Da and fragmented by Higher-energy collisional dissociation (HCD) at 30% normalized collision energy in the HCD cell. Previously selected ions were dynamically excluded from re-selection for 20 seconds. The MS² AGC was set to 1x10⁵. All samples were run in duplicates. Proteins were identified from the tandem mass spectra extracted by Xcalibur version 4.0. MS/MS spectra were searched against the SwissProt Homo Sapiens database contain mouse proteins of interest using Mascot search engine (Matrix Science, London, UK; version 2.5.1). All searches included carbamidomethyl cysteine as a fixed modification and oxidized Met, phosphorylated Ser Thr and Tyr, deamidated Asn and Gln, acetylated N-term, and acetylation on Lys as variable modifications. Three missed tryptic cleavages were allowed. The MS¹ precursor mass tolerance was set to 10 ppm and the MS² tolerance was set to 0.05 Da. A 1% false discovery rate cutoff was applied at the peptide level. Only proteins with a minimum of two peptides above the cutoff were considered for further study. The search result was visualized by Scaffold (version 4.8.3. Proteome Software, INC., Portland, OR). Acetylation of PER2^K680 was manually validated.

Behavioral analysis

Locomotor activity was analyzed in young and old mice treated with NR or H₂O or in CAG^CreER; Nampt^fx/fx mice and controls before and after tamoxifen treatment. Mice were singly housed in standard mouse cages equipped with running wheels in standard 12:12 LD conditions. Average 24-hour wheel running activity from each mouse was determined from a 14-day period for each mouse using ClockLab software (Actimetrics). The average activity between ZT22 and ZT0 was quantified.

Live animal body temperature monitoring

Mice were implanted with a IPTT-300 temperature transponder into the subcutaneous dorsal region. Following one week of recovery, body temperature was assessed every 2 hrs for 24 hrs using a hand-held DAS-8007-IUS reader from outside the cage, enabling remote body temperature measurements with minimal disruption to the animals (Bio Medic Data Systems). Mice had ad lib access to food and water for the duration of the experiment.

Oxygen consumption rates

Liver mitochondrial respiratory rhythms were measured as previously described (Peek et al., 2013). Briefly, mitochondria were isolated from fresh liver in mitochondrial isolation buffer (70 mM sucrose, 210 mM mannitol, 5 mM HEPES, 1 mM EGTA, and 0.5% (w/v) fatty acid-free BSA (pH 7.2)) prior to resuspension in 1xMAS buffer (70 mM sucrose, 220 mM mannitol, 10 mM KH₂PO₄, 5 mM MgCl₂, 2 mM HEPES, 1.0 mM EGTA, and 0.2% (w/v) fatty acid-free BSA (pH 7.2). 10 μg of mitochondria in 1xMAS buffer containing 80 μM palmitoyl-carnitine was loaded into each well of a 96-well Seahorse bioanalyzer (Agilent) and centrifuged for 20 min in 2000g at 4°C to adhere mitochondria. Oxygen consumption rates were taken basally and following each sequential injection of 4mM ADP, 10μM oligomycin, 10μM FCCP, and 10μM antimycin A (Sigma). FCCP-treated, “maximal” OCR was normalized to basal readings to control for collection/plate differences and quantified.

Quantitation and statistical analysis

Statistical analysis was performed with unpaired two-tailed Student’s t test except where otherwise noted in results, figure legends, or methods. Behavioral analysis of CAG^CRE=ER;Nampt^fx/fx mice before and after tamoxifen administration was performed with a paired two-tailed Student’s t test. Where appropriate, data are represented as mean +/− SEM. Differences were considered significant when p<0.05 except where otherwise noted in results, figure legends, or methods. Two-way ANOVA was performed to compare effects of NR-treatment or L-Nampt−/− on NAD⁺ and nicotinamide rhythms. P-value was reported for variation from the drug/genetic intervention.

Supplementary Material

3

Table S1. NR reprograms hepatic diurnal transcriptome. Related to Figure 1.

4

Table S2. Genetic depletion of NAMPT reprograms hepatic diurnal transcriptome. Related to Figure 3.

5

Table S3. Transcriptional response to pharmacologic elevation of NAD⁺ require both BMAL1 and SIRT1. Related to Figures 4–5.

6

Table S4. LC/MS-MS reveals PER2 acetylation at K680. Related to Figure 5.

7

Table S5. NR restores rhythmic BMAL1 chromatin binding. Related to Figure 6.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-BMAL1 (ChIP-Seq)	J. Bass, Millipore	ABE2599 (Perelis et al., 2015)
Anti-BMAL1 (WB)	Santa Cruz	sc-48790
Anti-HSF1	Cell Signaling	4356
Anti-H3	Cell Signaling	9715
Anti-H3K27Ac	Active Motif	39133
Anti-H3K9Ac	Active Motif	39137
Anti-β-ACTIN	Cell Signaling	4970
Anti-GAPDH	Cell Signaling	5174
Anti-p-PER2-S659	Thermo Fisher	PA5-38901
Anti-PER2	C. Lee (Florida State University)
Anti-PER1	Abcam	ab136451
Anti-CRY2	Alpha Diagnostic	CRY21-A
Anti-CRY1	Bethyl	A302-614A
Anti-FLAG-HRP	Sigma	F1804-50UG
Chemicals, Peptides, and Recombinant Proteins
Tamoxifen	Sigma	T5648-1G
Corn Oil	Sigma	C8267-500ML
Trypsin	Fisher Scientific	25300054
Dexamethasone	Sigma	D4902-100MG
Nicotinamide mononucleotide (NMN)	Oriental Yeast Company	44500900-1g
PBS	Mediatech	21-040-CM
Nicotinamide riboside (NR)	Chromadex	N001
Perchloric acid	Sigma	244252-100ML
K2CO3	Sigma	367877-10G
Trizol Reagent	Molecular Research Center	TR118
Disuccinimidyl glutarate	Proteochem	C1104-1 gram
DMSO	Sigma	D8418-100ML
Formaldehyde	Polysciences	18814
Sodium chloride (NaCl)	Fischer Scientific	BP35-10
Ethylenediaminetetraacetic acid (EDTA)	Lonza	51234
Trizma® hydrochloride solution	Millipore-Sigma	T2319-1L
NP-40	Sigma	NP40s
SDS	Sigma	L3771-ikg
Triton X-100	Sigma	X100-100ML
MgCl2	Sigma	M8266
IGEPAL	Sigma	I8896-50ML
Tween-20	Sigma	P1754-500ML
Digitonin	Sigma	D141-100MG
Nextera transposase enzyme	Illumina	15027865
Nextera TD buffer	Illumina	15027865
FK866	Sigma	F8557-25MG
EX527	Tocris	2780
HEPES	Gibco	15630080
NE-PER kit	ThermoFisher	78835
B27 Supplement	Invitrogen	17504-001
Trichostatin A	Sigma	T8552-1MG
Nicotinamide	Sigma	N3501-100MG
Ammonium bicarbonate	ThermoFisher	A643500
Acetonitrile	ThermoFisher	A955500
Dithiothreitol	Thermo Fisher	R0861
Ambic/iodoacetamide	Sigma	I1149-5G
Formic acid	ThermoFisher	P128905
Dulbecco’s modified Eagle’s medium (DMEM)	Gibco	MT10013CV
Sodium bicarbonate	Sigma	S5761-500G
Fetal bovine serum (FBS)	Atlanta Biological	S11550
Penicillin-Streptomycin	Gibco	15140122
Luciferin sodium salt	Biosynth AG	L8220
Cycloheximide	Sigma	239764-1GM
Protease inhibitor cocktail, Complete mini	Roche Applied Science	11836170001
Phosphatase inhibitor cocktail, phoSTOP	Roche Applied Science	04906845001
Anti-FLAG conjugated paramagnetic beads	Sigma	M8823-1ML
L-Glutamine	Thermo Fisher Scientific	25030081
Sucrose	Sigma	S7903-1KG
KH2PO4	Sigma	P3786-100g
Palmitoyl-carnitine	Sigma	P1645-5MG
Oligomycin	Sigma	O4876-5MG
FCCP	Sigma	C2920-10MG
Antimycin A	Sigma	A8674-25MG
Mannitol	Sigma	M9546-250G
EGTA	Bostonbioproducts	BM-151
Fatty acid-free BSA	Sigma	A7511-5G
Critical Commercial Assays
Zymo Direct-Zol RNA miniprep kit	Zymo Research	R2051
NEBNext RNA Ultra Directional Library Preparation Kit	New England Biolabs	E7420S
NEBNext Library Quant Kit	New England Biolabs	E7630S
NEBNext Ultra II Library Prep Kit	New England Biolabs	E7645S
Q5 site directed mutagenesis kit	New England Biolabs	E0552S
MinElute columns	Qiagen	28006
Software and Algorithms
Bowtie (v2.2.4)	http://bowtie-bio.sourceforge.net/index.shtml	(Langmead et al., 2009)
subread:featureCounts (v1.5.1).		(Liao et al., 2014)
STAR (v2.5.2)		(Dobin et al., 2013)
DESeq2 (v1.24.0)		(Love et al., 2014)
eJTK_Cycle (v3.1.R).		(Hutchison et al., 2015)
ClockLab Data Collection and Analysis System (v6.0)	http://actimetrics.com
R Package ATACSeqQC		(Ou et al., 2018)
HOMER (v4.10)	http://homer.ucsd.edu/homer/index.html	(Heinz et al., 2010)
Xcalibur (v4.0)
Mascot search engine (v2.5.1).	Matrix Science, London, UK
MS/MS spectra scaffold (v4.8.3)	Proteome Software, Portland, OR
Deposited Data
RNA-seq, ChIP-seq, ATAC-seq	NCBI GEO	GSE133989
Experimental Models (Cells)
Per2::Luciferase MEFs	J. Takahashi (UT Southwestern)
Sirt1−/− MEFs	J. Bass Lab	Harvested from Sirt1fx/fx mice treated with 3 rounds Adeno-Cre (Vector Biolabs)
HEK293 cells	Clontech
Experimental Models (Mice)
C57BL/6CR (4 mo, 16 mo)	Charles River, NIA Colony
Bmal1−/−	J. Takahashi (UT Southwestern)
Per2::Luciferase	J. Takahashi (UT Southwestern)
Namptfx/fx	S. Imai (Washington University)
Sirt1fx/fx	S. Imai (Washington University)
Alb-Cre	Jackson Laboratories
CAGCre-ER	Jackson Laboratories
Recombinant DNA
shRNA targeting p19ARF	Addgene	14090
pCMV-SPORT2-PER2	Addgene	16204