REV-ERBα regulates brain NAD⁺ levels and tauopathy via an NFIL3–CD38 axis

Abstract

Nicotinamide adenine dinucleotide (NAD⁺) is a critical metabolic co-enzyme implicated in brain aging, and augmenting NAD⁺ levels in the aging brain is an attractive therapeutic strategy for neurodegeneration. However, the molecular mechanisms of brain NAD⁺ regulation are incompletely understood. In cardiac tissue, the circadian nuclear receptor REV-ERBα has been shown to regulate NAD⁺ via control of the NAD⁺-producing enzyme NAMPT. Here we show that REV-ERBα controls brain NAD⁺ levels through a distinct pathway involving NFIL3-dependent suppression of the NAD⁺-consuming enzyme CD38, particularly in astrocytes. REV-ERBα deletion does not affect NAMPT expression in the brain and has an opposite effect on NAD⁺ levels as in the heart. Astrocytic REV-ERBα deletion augments brain NAD⁺ and prevents tauopathy in P301S mice. Our data reveal that REV-ERBα regulates NAD⁺ in a tissue-specific manner via opposing regulation of NAMPT versus CD38 and define an astrocyte REV-ERBα–NFIL3–CD38 pathway controlling brain NAD⁺ metabolism and neurodegeneration.

Lee et al. show that the circadian clock protein REV-ERBα controls brain NAD⁺ levels by regulating the NAD⁺-consuming enzyme CD38. Global or astrocytic REV-ERBα deletion or pharmacologic REV-ERB inhibition protects against tau pathology in mice.

Main

NAD⁺ is an abundant metabolite required for multiple cellular functions, including redox homeostasis, DNA repair, metabolism and histone/protein deacetylation via control of sirtuins (SIRTs)^1–3. NAD⁺ homeostasis is dependent on a balance between production and consumption, with NAMPT being the rate-limiting enzyme for production, whereas several proteins, including CD38, poly-ADP-ribose polymerases (PARPs) and SIRTs, are consumptive^1–3. Interestingly, NAD⁺ is intertwined with circadian clock function, as NAMPT expression is under circadian control in the liver, whereas both NAD⁺ itself and NAD⁺-dependent deacetylase SIRT1 regulate core clock function directly^4,5. Because NAD⁺ levels decline with aging in most tissues, including the brain³, augmentation of NAD⁺ levels has been proposed as a therapeutic strategy to counteract aging and prevent neurodegenerative diseases^2,3,6.

REV-ERBα (encoded by Nr1d1) is a circadian clock protein, nuclear receptor and transcriptional repressor, which serves critical functions in the control of metabolism and inflammation^7–11. REV-ERBα expression is dampened in patients with Alzheimer’s disease (AD), who often suffer from sleep fluctuations and disrupted circadian function¹², and global REV-ERBα deletion can prevent amyloid plaque pathology in an AD mouse model by enhancing microglial phagocytic activity¹³. Conversely, microglia-specific REV-ERBα deletion can exacerbate tau pathology in male mice¹², suggesting that REV-ERBα may exert complex cell-type-specific effects on AD pathology. In cardiac tissue, REV-ERBα is required to maintain NAD⁺ levels via rhythmic expression of NAMPT¹⁴. In that setting, REV-ERBα deletion de-represses NFIL3/E4BP4, leading to suppression of NAMPT, loss of NAD⁺ levels and cardiomyopathy. However, it is unknown if a similar pathway is present in the brain or how such a REV-ERBα–NAD⁺ axis might influence neurodegeneration. As REV-ERBα-targeted therapeutics are being developed, understanding the overall and cell-type-specific effects of this pathway on brain NAD⁺ and neurodegeneration has clear translational importance^15,16. Thus, we investigated REV-ERBα effects on brain NAD⁺ levels and neurodegeneration under basal conditions and in the setting of tau pathology, which plays a critical role in AD and other age-related neurodegenerative conditions. We observed that REV-ERBα regulates NAD⁺ levels in the brain through an NFIL3–CD38 pathway in astrocytes and that global or astrocyte-specific REV-ERBα deletion augments brain NAD⁺ levels and protects mice from tauopathy. Here we provide initial findings that pharmacological antagonism of REV-ERBα can mitigate tau pathology in PS19 mice, suggesting that the protective effects of REV-ERBα inhibition predominate in the brain and that REV-ERBα inhibitors may hold promise for AD therapy.

Results

REV-ERBα deletion suppresses CD38 expression and induces NAD⁺ levels in the brain

To investigate the effects of REV-ERBα on gene expression in the brain, we generated global, postnatal REV-ERBα knockout (RKO) mice using the tamoxifen-sensitive CAG::CreER^T2 line and Nr1d1^fl/fl mice, which produce full loss of REV-ERBα expression¹⁷. REV-ERBα deletion was induced by tamoxifen at 2 months of age, and hippocampal transcriptomes were analyzed by bulk RNA sequencing (RNA-seq) at 10 months of age (Fig. 1a). We observed abundant expression of Cre in the hippocampus of the Cre⁺ group and found more than approximately 50% Nr1d1 reduction by tamoxifen treatment (Fig. 1b). We performed bulk RNA-seq and identified 470 differentially expressed genes (DEGs) in Cre⁺ versus Cre⁻ hippocampus (P < 0.05, |fold change (FC)| > 50%; 262 higher, 208 lower). Elevated genes in Cre⁺ mice included Tex11, Plvap, Bmal1(Arntl), C4b, Bag3, Plin4 and Cxcl5, and downregulated transcripts included Apoa1, Apoa2, Lcat and Ccr7 (Fig. 1c). Gene Ontology term pathway analysis using these DEGs identified ‘Cation transport’ and ‘Steroid metabolic process’ as top processes altered by postnatal REV-ERBα deletion in the brain (Fig. 1d,e). Interestingly, we observed that the transcription factor Nfil3, a known target of REV-ERB-mediated transcriptional repression¹⁴, was upregulated in Cre⁺ mice, whereas the critical NAD⁺-consuming enzyme Cd38 was strongly downregulated after REV-ERBα deletion (Fig. 1e,f)^18,19. As NAD⁺ levels can vary with time of day in liver^4,5, we prepared hippocampal lysates from Cre⁻ and Cre⁺ mice euthanized every 6 hours across the day under standard 12-hour:12-hour lighting, to exclude time-of-day effects. We saw upregulation of the REV-ERB target Fabp7 at all timepoints, indicating strong REV-ERBα deletion. Cd38 transcript did not vary by time of day in Cre⁻ mice and was consistently suppressed across timepoints in Cre⁺ mice (although the 12:00 timepoint did not meet significance), suggesting that REV-ERBα regulation of Cd38 is not time-of-day dependent (Extended Data Fig. 1a). Moreover, CD38 protein level was also downregulated in RKO hippocampus (Extended Data Fig. 1b). We next measured the NAD⁺ concentration using brain cortex samples and found that it was significantly increased in RKO brain (Fig. 1g) compared to wild-type (WT) (Cre⁻) controls, in sharp contrast to a previous report in cardiac tissue¹⁴. NAD⁺ metabolism is generally driven by NAD⁺-consuming enzymes such as CD38, SIRTs and PARPs^1,20. Among these, only Cd38 was significantly downregulated by REV-ERBα deletion in hippocampus (Fig. 1h), suggesting that REV-ERBα could affect brain NAD⁺ level through CD38 modulation.

Fig. 1. REV-ERBα deletion suppresses the NAD⁺-consuming enzyme CD38 and enhances brain NAD⁺ levels.

a, Schematic showing experiments with inducible global REV-ERBα KO mice (CAG::Cre^ERT2;Nr1d1^fl/fl mice, termed RKO). TAM, tamoxifen. b, Cre expression (WT, n = 13 mice; RKO, n = 17 mice) and Nr1d1 (REV-ERBα) deletion efficiency (WT, n = 14 mice; RKO, n = 18 mice) in RKO (Cre⁺) compared to WT (Cre⁻) mouse brain. c, Volcano plot showing differential gene expression from WT and RKO hippocampus (log₂FC cutoff = 0.5, −log₁₀ (P value) cutoff = 1.3, corresponds to P value of 0.05 using limma-voom). d, Top five upregulated or downregulated biological processes identified for DEGs in REV-ERBα KO brain from bulk RNA-seq. e, Heatmap representing genes from ‘Cation transport; upregulation’ and ‘Steroid metabolic process; downregulation’. Nfil3 and Cd38 are noted with red asterisks. f, Expression of Nfil3 and Cd38 from WT and RKO hippocampus by qPCR (WT, n = 12 mice; RKO, n = 9 mice). g, Increased NAD⁺ levels in RKO cerebral cortex compared to WT (WT, n = 12; RKO, n = 13). h, Transcripts of three major NAD⁺-consuming enzymes, Sirt1, Parp1 and Cd38, in WT and RKO hippocampus (from RNA-seq data in c). i, Nampt expression in hippocampal tissue is unchanged in RKO group compared to WT group (WT, n = 12 mice; RKO, n = 9 mice, qPCR). j, Expression of Nfil3 transcript in Arc nucleus from control (n = 4 mice) and hypothalamic REV-ERB α/β KO brain (H-DKO, n = 5 mice) and cardiomyocyte-specific REV-ERBα/β KO heart (CM-DKO, n = 3 mice per group). k, Cd38 and Nampt transcript expression in Arc nucleus tissue from control (n = 4 mice) and hypothalamic REV-ERBα/β KO (H-DKO, n = 5 mice). l, Cd38 and Nampt transcript expression in heart tissue from control and cardiomyocyte REV-ERBα/β KO (CM-DKO) mice (n = 3 mice per group). m, Knockdown of Nfil3 with siNfil3 siRNA causes increased expression of Cd38 in primary mouse astrocyte cultures. Control siRNA is labeled as ‘siCon’. n, Diagram depicting proposed indirect regulation of CD38 and NAD⁺ by REV-ERBα via NFIL3 inhibition. Note that the NFIL3–NAMPT interaction is minimal in brain but predominates in heart. **P < 0.01, ***P < 0.005, ****P < 0.001; ‘NS’ is non-significant by two-tailed t-test. Error bars represent mean ± s.e.m. ptn, protein.

Source data

Extended Data Fig. 1. Dampened Cd38 mRNA and protein expression on Rev-erbα knock-out (KO) mouse hippocampus across the day.

(a) Tamoxifen treated global REV-ERBα KO mice (CAG::CreER^T2; Nr1d1^fl/fl; RKO) and Cre- controls (WT) were sacrificed every 6 hours over a 24-hour period under standard 12 h:12 h light:dark conditions and gene expression was assayed using hippocampal tissue. Cd38 transcript did not vary by time of day, but was decreased at all timepoints (except 12 pm). A known target of REV-ERBα mediated repression, Fabp7, was highly induced compared to WT (Cre-) at all timepoints. N = 3 mice/genotype/ in each time point. (b) Lower level of CD38 protein in RKO hippocampus than WT at 12 pm (WT, n = 5; RKO, n = 4 mice). *p < 0.05, **p < 0.01, and ****p < 0.001 by 2-tailed T-test. Error bars represent mean ± SEM.

Source data

REV-ERBα regulates NAD⁺ level via NFIL3–CD38 axis not through Nampt in the brain

In cardiac tissue, REV-ERBα directly represses Nfil3, which is required for expression of Nampt, an enzyme that catalyzes the first reversible step in NAD⁺ biosynthesis and nicotinamide (NAM) salvage¹⁴. REV-ERBα deletion in heart, therefore, reduces Nampt expression and causes NAD⁺ depletion¹⁴, the opposite of what we observed in brain (Fig. 1g). Moreover, we found that REV-ERBα KO had no effect on Nampt expression in hippocampus (Fig. 1i), suggesting divergent regulatory pathways in heart versus brain. To address this further, we examined transcriptomic data from previous studies of hypothalamic and heart-specific REV-ERBα/β deletion^14,21. As expected, Nfil3 expression was increased in both the arcuate (Arc) nucleus of the brain and heart tissue after REV-ERB deletion (Fig. 1j). REV-ERB deletion significantly reduced Arc Cd38 levels without any effect on Nampt, similar to our data in hippocampus (Fig. 1k). However, myocardial REV-ERB deletion had no effect on Cd38 expression but strongly suppressed Nampt, which was previously shown to cause NAD⁺ depletion¹⁴ (Fig. 1l). As Cd38 is primarily expressed in astrocytes in mouse brain according to a brain transcriptomic atlas (https://brainrnaseq.org/)²², we suspected that astrocytes may be the main site of the REV-ERBα effects on NAD⁺. We hypothesized that astrocyte REV-ERBα regulates Cd38 expression via NFIL3. To address this, we knocked down Nfil3 using small interfering RNA (siRNA) in primary cultured astrocytes and observed a decrease in Cd38 levels (Fig. 1m). Although REV-ERBβ (Nr1d2) can have overlapping functions with REV-ERBα (refs. ^8,23), we observed only weak induction of Nfil3 and no significant Cd38 reduction after siRNA-mediated REV-ERBβ knockdown in cultured astrocytes, suggesting that REV-ERBβ has a minimal impact (Extended Data Fig. 2).

Extended Data Fig. 2. REV-ERBβ (Nr1d2) knockdown does not alter Cd38 expression in primary astrocytes.

Knockdown of Nr1d2 (REV-ERBβ) by siRNA transfection in cultured primary astrocytes induces Bmal1, has a small effect on Nfil3 transcript levels, but Cd38 does not significantly change (siControl, n = 7; siNr1d2, n = 7 plates from two mice). **p < 0.01, ****p < 0.001, and ns is non-significant by 2-tailed T-test. Error bars represent mean ± SEM.

Source data

To further support our hypothesis that NFIL3 regulates Cd38 transcription directly, we analyzed an existing mouse single-cell assay for transposase-accessible chromatin using sequencing (ATAC–seq) dataset, which probes accessible chromatin in more than 800,000 individual nuclei from 45 regions that span multiple brain regions and maps the state of 491,818 candidate cis-regulatory DNA elements in 160 distinct cell types, for NFIL3 binding motifs in the Cd38 promoter²⁴. We retrieved astrocyte-specific candidate cis-regulatory elements (cCREs) and identified a peak (cCREs353987) that was specific to astrocytes from the mouse brain white and gray matter and located at the 5′ transcription start site (TSS) of the Cd38 genomic region (Extended Data Fig. 3). Furthermore, multiple lines of evidence from three additional databases support the functionality of this genomic region, including the Mouse ATAC–seq Atlas database, DNase-seq data from the Cistrome database as well as the ENCODE cCREs database from the ENCODE project^25–27. Using the position weight matrix (PWM) of NFIL3 (motif M01819 from the Catalog of Inferred Sequence Binding Preferences (CIS-BP) database), the Patser program^28,29 identified 12 putative low-affinity binding sites of NFIL3 within the peak cCREs353987. The PWM derived from the 12 NFIL3 low-affinity sites was similar to the NFIL3 motif from the CIS-BP database (Extended Data Fig. 3). These results suggest that NFIL3 may directly regulate Cd38 expression through this astrocyte-specific enhancer. Thus, in the brain, REV-ERBα deletion leads to de-repression of NFIL3, which, in turn, represses Cd38, leading to decreased CD38-dependent NAD⁺ degradation and ultimately causing increased NAD⁺ levels. These results suggest that REV-ERBα may use distinct molecular pathways in brain versus peripheral tissues to regulate NAD⁺ level and finetune tissue NAD⁺ levels via opposing regulation of Nampt versus Cd38, with Nampt suppression predominating in heart and Cd38 suppression predominating in brain (Fig. 1n).

Extended Data Fig. 3. Identification of 12 binding motifs for NFIL3 in Cd38 genomic region.

(a) scATAC-seq peak cCREs353987 was specific to astrocytes from the mouse brain white and grey matter as well as cerebral nuclei (or basal ganglia). It is located at the 5’ TSS of Cd38 genomic region. Multiple putative low-affinity binding sites of the transcription factor NFIL3 (motif M01819 from CIS-BP database) were detected within the peak cCREs353987. Furthermore, multiple lines of evidence from three additional databases support the functionality of this genomic region including the Mouse ATAC-seq Atlas database, DNase-seq data from the Cistrome database as well as the ENCODE cCREs database from ENCODE project. (b) The score represents the binding affinity of NFIL3 to the putative NFIL3 finding sites. (c) The sequence alignment logo of the 12 NFIL3 binding sites.

REV-ERBα deletion induces Nmnat3 and alters NAD⁺-related metabolites

To further explore the effects of REV-ERBα on NAD⁺ metabolism, we examined expression of transcripts encoding several enzymes involved with NAD⁺ metabolic pathway^1,2 in Cre⁻ WT and Cre⁺ RKO hippocampus, including Nnmt (nicotinamide N-methyltransferase), Sarm1 (sterile alpha and TIR motif containing 1), Nmnat1 (nicotinamide mononucleotide adenylyl transferases 1), Nmnat2 and Nmnat3 as well as Nampt. Surprisingly, we found significant induction of Nmnat3 transcript in RKO hippocampus compared to WT (Extended Data Fig. 4a), whereas no other genes were altered. Again, hippocampal NAD⁺ levels were increased in RKO mice (Extended Data Fig. 4b). Interestingly, Nmnat3 did not differ in heart tissue from RKO mice compared to WT mice, whereas Nampt transcript was significantly reduced (Extended Data Fig. 4c). NMNAT3 is involved in the synthesis of NAD⁺ in mitochondria, which can exert protective effects in models of axonal degeneration and Parkinsonʼs disease^30–32. Thus, REV-ERBα could regulate brain NAD⁺ level by Nmnat3 induction instead of Nampt modulation.

Extended Data Fig. 4. Nmnat3 is induced by REV-ERBα deletion in the hippocampus, but not in the heart tissue.

(a) Comparing mRNA expression of NAD⁺ metabolic enzymes and (b) NAD⁺ level between WT (Cre-) and RKO (Cre + ) hippocampus (WT, n = 16; RKO, n = 16 mice). (c) Nmnat3 is not altered in heart tissue in global RKO mice, while both Nampt and Sarm1 are decreased (WT, n = 4; RKO, n = 4 mice). *p < 0.05 and ns is non-significant by 2-tailed T-test. Error bars represent mean ± SEM.

Source data

Next, we performed untargeted metabolome analysis using WT (Cre⁻) and RKO (Cre⁺) cortex tissues to verify the effect of REV-ERBα on brain metabolites related to NAD⁺ metabolism. We identified a total of 1,467 features (Extended Data Fig. 5a), with principal component analysis (PCA) identifying 185 differentially abundant metabolites (Student’s t-test, P < 0.1) separating the RKO group (blue dots and ellipses) from the WT group (yellow dots and ellipses) with nominal significance (Extended Data Fig. 5b). Among 185 metabolites, 12 features corresponded to the unique NAD metabolites¹⁹. Of these NADP⁺, ADP-D-ribose, nudifloramide, NAM and glutamine were increased in RKO brain, whereas cyclic ADP-ribose (cADPR), guanosine 5-triphosphate (GTP), adenosine diphosphate (ADP), arachidonic acid, L-(−)-malic acid, flavin adenine dinucleotide (FAD) and 2-oxogultaric acid were decreased (Extended Data Fig. 5c), indicating that REV-ERBα deletion causes altered NAD⁺ metabolism in the brain.

Extended Data Fig. 5. Principal component analysis (PCA) and Volcano plots representing deregulated metabolites between WT (Cre-) and RKO (Cre + ) cortex.

(a) PCA plot of all the metabolite features (left) and differentially abundant metabolite features (right) between ‘Cre-’ and ‘Cre + ’ groups: Each point represents a sample, colored by ‘Cre-’ (Yellow) and ‘Cre + ’ (Blue). Ellipses represent 95% confidence intervals (95% CI) for each group. (b) 185 metabolite features are differentially abundant between the groups by p-value < 0.1. (c) Volcano plot of differentially abundant features: Each Point represents a metabolite feature in the analysis, with the x-axis showing the log₂ fold change (Log₂FC) between ‘Cre + ’ versus ‘Cre-’ and the y-axis representing the –log10 p-value. Points are colored in blue based on the NAD derivatives or related metabolites. Red horizontal lines indicate the significance threshold (Student T-test, p < 0.1), and the points above the line are considered marginally significant. Text labels are provided for the name of the 12 marginally significant NAD related metabolites.

Global REV-ERBα deletion increases brain NAD⁺ levels and improves synaptic signaling and neuroinflammatory gene expression in PS19 tauopathy mice

Considering its effect on neuroprotective NAD⁺ levels in the brain, we next sought to determine the impact of REV-ERBα deletion on neurodegeneration. Because REV-ERBα deletion suppressed amyloid-β pathology in a previous study¹³, we examined its potential effect in a mouse model of tauopathy, which is another major contributor to AD pathogenesis³³. We crossed the human mutant P301S tau mouse line PS19 (ref. ³⁴) with our CAG::CreER^T2;Nr1d1^fl/fl line, generating four genotypes (Cre⁻PS19⁻ (termed WT), Cre⁺PS19⁻ (termed RKO), Cre⁻PS19⁺ (termed PS19) and Cre⁺PS19⁺ (termed PS19;RKO)). All mice were treated with tamoxifen at 2 months and euthanized at 10 months, when moderate tau pathology and neurodegeneration are present¹². We first analyzed the hippocampal transcriptomic signature of WT versus PS19 mice by bulk RNA-seq to determine the effects of tau pathology alone. In PS19⁺ mice, 1,931 DEGs were upregulated, the majority of which are involved in ‘Cytokine production’, ‘Cell activation’ and ‘Inflammatory responses’, whereas the 915 downregulated DEGs were related to ‘Synaptic transmission’, ‘Ion transport activity’ and ‘Behavior’ (Extended Data Fig. 6a,b), demonstrating expected changes in synaptic function and inflammation in PS19 mice. Brain NAD⁺ levels decline in human brain with AD progression²¹, and we also observed lower levels of NAD⁺ in 10-month-old PS19 mouse brain (Extended Data Fig. 6c).

Extended Data Fig. 6. PS19 mice exhibit increased inflammatory gene expression and decreased brain NAD⁺ levels.

(a) Volcano plot showing differentially gene expression (DEGs) from 10-months old WT vs. PS19 brains (Log₂ fold changes cutoff = 0.5, −Log₁₀ P-value cutoff = 1.3, corresponds to p-value of 0.05 using Limma-voom). (b) Top 5 upregulated or downregulated biological processes identified for DEGs from PS19 mouse brain from bulk RNA-sequencing. (c) NAD⁺ levels in PS19 mouse brain (n = 11 mice) compared to WT (n = 6 mice). **p < 0.01 by 2-tailed T-test. Error bars represent mean ± SEM.

Source data

We next compared PS19 and PS19;RKO mice to observe the transcriptomic changes caused by REV-ERBα deletion in PS19 mice brain (Fig. 2a). Interestingly, many inflammatory genes were more strongly expressed in PS19 mice (565 genes), such as microglial activation markers (Aif1, Cd68, Lpl and Cd36, left side of plot), whereas 434 genes, including circadian transcripts (Cry2, Npas2, Per2and Arntl, right side of plot), were upregulated in PS19;RKO mice (Fig. 2b). As shown in Fig. 2c, 1,344 DEGs overlapped between WT versus RKO group and PS19 versus PS19;RKO, which we categorized into four different clusters based on their expression pattern: cluster ‘a’: down in PS19/up in PS19;RKO; cluster ‘b’: up in both PS19 and PS19;RKO; cluster ‘c’: down in both PS19 and PS19;RKO; and cluster ‘d’: up in PS19/down in PS19;RKO (Fig. 2d). The populations of groups ‘a’ and ‘d’ were dominant, with 53.27% and 43.53% of DEGs, respectively, among the four groups (Fig. 2e). Expression of most genes in cluster ‘a’ (716 genes) were involved in ‘Synaptic signaling’ and ‘Regulation of ion transport’, suggesting that REV-ERBα deletion efficiently mitigated dysfunction of synaptic and ion transport activity in PS19 mice. In contrast, cluster ‘d’ (585 genes) was enriched for ‘Cytokine production’ and ‘Cell activation’, showing that these processes in PS19 mice were alleviated by REV-ERBα deletion. In addition, ‘Defense response to virus’ and ‘Protein catabolic process’ were more dominant in PS19;RKO compared to the PS19 group (cluster ‘b’, 41 genes), but we found only two genes (Amnand Gdf10) in cluster ‘c’ (Fig. 2f). We also observed low Cd38 and high Nfil3 expression (Fig. 2g) with induction of brain NAD⁺ levels in PS19;RKO mice (Fig. 2h). Altogether, REV-ERBα deletion prevented tau-mediated downregulation of synaptic genes as well as upregulation of inflammatory genes, suggesting a protective effect.

Fig. 2. REV-ERBα deletion activates the NFIL3–CD68 axis to increase brain NAD⁺ levels and improves inflammatory and synaptic gene expression in PS19 tauopathy mice.

a, Schematic showing 10-month-old Cre⁻;P301S tau⁺ (PS19) and Cre⁺;P301S⁺ RKO (PS19;RKO) mice. b, Volcano plot showing differential gene expression from PS19 and PS19;RKO brains (log₂FC cutoff = 0.5, −log₁₀ (P value) cutoff = 1.3, corresponds to P value of 0.05 using limma-voom). Left side indicates genes increased in PS19; right side indicates genes increased in PS19;RKO. c, Venn diagram analysis showing DEGs that were differentially expressed in PS19/WT and PS19;RKO/PS19, indicating regulation by both tau and REV-ERBα. d, Division of the 1,344 DEGs from c into four groups: ‘a’, increased by RKO but not tau; ‘b’, increased by tau and further by RKO; ‘c’, decreased by RKO and by tau; ‘d’, increased by tau and decreased by RKO. e, Pie chart showing proportion of the 1,344 DEGs in each group. f, Top three Gene Ontology functional enrichment results for each group of DEGs except group ‘c’. RKO rescued tau-mediated downregulation of synaptic genes, enhanced tau-mediated expression of protein catabolism genes and reduced tau-mediated inflammatory gene expression. g, REV-ERBα KO increased Nfil3 (n = 10 mice) and suppressed Cd38 expression (n = 8 mice) in PS19 mouse brain. h, Increased NAD⁺ levels in PS19;RKO mice brain compared to PS19 (PS19, n = 15 mice; PS19;RKO, n = 16 mice). **P < 0.01, ****P < 0.001; ‘NS’ is non-significant by two-tailed t-test or two-way ANOVA with Sidakʼs multiple comparisons test. Error bars represent mean ± s.e.m. ptn, protein.

Source data

Global REV-ERBα deletion reduces tau accumulation and glia activation in PS19 mice

Tau pathology is characterized by accumulation of hyperphosphorylated tau (pTau;AT8⁺), aggregated, misfolded tau (MC1⁺), neuroinflammation and brain atrophy^12,34. We next examined tissue from PS19 and PS19;RKO mice treated as above at 10 months for pTau (AT8) and misfolded pathological tau (MC1) by immunohistochemistry. REV-ERBα deletion significantly reduced both AT8⁺ and MC1⁺ tau in PS19;RKO mouse brain in the hippocampus and entorhinal cortex (Fig. 3a) compared to PS19 mice. We confirmed by western blotting that pTau was significantly decreased in the PS19;RKO group (Extended Data Fig. 7a).

Fig. 3. Tau deposition and glial activation in PS19 mouse brain are ameliorated by REV-ERBα deletion.

a, pTau and aggregated/misfolded tau were stained by AT8 and MC1 antibodies, respectively, in hippocampus (HIP) and entorhinal cortex (EC). Percent area positive for each staining was quantified (PS19, n = 9 mice; PS19;RKO, n = 11 mice). Scale bars, 500 μm. b,c, Representative images of activated astrocytes (GFAP, cyan), total microglia (IBA1, red) and activated microglia (CD68, green) in PS19 and PS19;RKO mouse CA3 region (b) and EC (c) with quantification (PS19, n = 9 mice; PS19;RKO, n = 12 mice). Scale bars, 500 μm. d, Graphs depicting expression of eight genes in PS19 versus PS19;RKO HIP, including a REV-ERBα target (Fabp7) as well as genes involved in glial activation (Gfap, Aif1and Cd68), inflammation (Il1b and Tnf) and lipid metabolism (Apoe and Plin2). *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001; ‘NS’ is non-significant by two-tailed t-test. Error bars represent mean ± s.e.m.

Source data

Extended Data Fig. 7. REV-ERBα KO does not affect levels of acetylated Tau (ac-Tau) in PS19 mice.

(a) Western blot analysis of pTau (AT8) protein in cortex samples from PS19 (n = 5 mice) and PS19;RKO (n = 5 mice) mouse brain samples. Each intensity value is normalized to total hTau (HT7) (n = 10 mice/group). (b) Representative western blot images showing the effect of REV-ERBα on total SIRT1 protein expression under both basal (WT, n = 10; RKO, n = 13 mice) and tau expressing condition (PS19, n = 13; PS19;RKO, n = 12 mice) in vivo. β-tubulin was used as a loading control. (n = 10-13 mice/group). (c) Western blot analysis of ac-Tau (K174, K274) protein in PS19 and PS19; RKO mouse brain samples. Each intensity value is normalized with total hTau (HT7) expression (n = 10-12). *p < 0.05, ****p < 0.001, and ns is non-significant by 2-tailed T-test. Error bars represent mean ± SEM.

Source data

Acetylated forms of tau (ac-Tau) can promote tau aggregation³⁵, and tauopathy progression can be delayed by treatment with a monoclonal antibody targeting ac-Tau³⁶. SIRT1 is an NAD⁺-dependent deacetylase³⁷, which can de-acetylate tau to reduce tauopathy-mediated damage^38,39. Although Sirt1 transcript was unchanged in REV-ERBα KO mouse brain (Fig. 1h), SIRT1 protein was significantly increased in the RKO group, although it did not change in PS19 mice (Extended Data Fig. 7b). Nevertheless, we hypothesized that increased NAD⁺ could drive SIRT1-mediated tau de-acetylation and reduce ac-Tau in PS19;RKO mice. To assess this, we measured ac-Tau expression using tau K174 and K274 antibodies^38,39 by western blot and normalized this to total human tau expression by HT7 antibody. However, no significant changes were observed in either K174 or K274 ac-Tau protein (Extended Data Fig. 7c), suggesting that REV-ERBα deletion does not reduce tauopathy via increasing SIRT1-mediated tau de-acetylation.

We next measured gliosis and observed that astrocyte (GFAP) and microglial (IBA1 and CD68) reactivity were significantly decreased in parallel with tau pathology reduction in both hippocampus (Fig. 3b and Extended Data Fig. 8a) and entorhinal cortex (Fig. 3c and Extended Data Fig. 8b) of PS19;RKO mice compared to PS19 mice, whereas microgliosis was not severely changed in RKO hippocampus in the absence of tau pathology (Extended Data Fig. 9a,b). We noted a small decrease in GFAP⁺ astrocytes in RKO hippocampus (Extended Data Fig. 9a,b), although staining with S100β, a pan-astrocyte marker, was unchanged, suggesting that the changes are due to GFAP expression, not astrocyte number (Extended Data Fig. 9c,d). We further observed downregulation of several neuroinflammation-related genes such as pro-inflammatory cytokines (Tnf, Il1band Il6) and glial markers (Gfap, Aif1 and Cd68) in PS19;RKO compared to PS19 (Fig. 3d). Collectively, global, postnatal REV-ERBα deletion protected against tau deposition and tau-mediated neuroinflammation.

Extended Data Fig. 8. Glial markers in hippocampus and entorhinal cortex of PS19 and PS19/RKO.

(a) Representative images of activated astrocytes (GFAP: cyan), total microglia (IBA1: red), and activated microglia (CD68: green) in PS19 and PS19; RKO mouse whole hippocampus and (b) the entorhinal cortex (EC). (PS19, n = 9; PS19;RKO, n = 12 mice) Scale bars, 500 μm.

Extended Data Fig. 9. Global post-natal REV-ERBα deletion does not induce microglial activation but slightly decreases GFAP+ astrocytes.

(a) Representative images showing activated astrocytes (GFAP; cyan) and toal microglia (IBA1; red), and activated microglia (CD68; green) in WT and RKO mouse hippocampus and (b) quantification of % area for each marker, normalized to WT (WT, n = 6; RKO, n = 8 mice). (c) Immunofluorescent co-staining depicting S100β (red) and GFAP (green) in WT and RKO hippocampus. (d) Quantification of images in (c) (WT, n = 11; RKO, n = 12 mice). Scale bars, 500 μm (whole hippocampus); Scale bars, 50 µm (inset image). *p < 0.05, ***p < 0.005 and ns is non-significant by 2-tailed T-test. Error bars represent mean ± SEM.

Source data

REV-ERBα deletion improves hippocampal volume and CA1 thickness in PS19 mice

To determine the effect of REV-ERBα deletion on tau-mediated neurodegeneration, we examined the volumes of hippocampal regions CA1, CA3 and dentate gyrus (Fig. 4a,b). In the absence of tau, REV-ERBα deletion has no impact on hippocampal volumes. On the other hand, both CA1 and dentate gyrus volumes were reduced in PS19 mice compared to WT control, and this atrophy was partially rescued by REV-ERBα deletion (Fig. 4c). We next stained for synaptoporin (SPO), a presynaptic CA3 mossy fiber marker⁴⁰. REV-ERBα deletion in the absence of tau mildly reduced SPO staining in the CA3 region compared to WT, consistent with synapse loss that is milder than we previously reported in germline REV-ERBα KO mice⁴¹. SPO levels were more severely reduced in PS19 mice, and this was not rescued in PS19;RKO mice (Fig. 4d,e). Finally, CA1 neuronal layer thickness with NeuN staining was significantly reduced in PS19, and this was rescued in male PS19;RKO mice but not female mice, which generally have milder pathology (Fig. 4f,g). Previous studies suggested that tau expression may alter expression of the immediate early gene c-fos, which reflects increased neuronal activity in the brain^42,43. To examine a possible effect of REV-ERBα on neuronal activity in PS19 mice, we performed c-fos staining, which was not different between genotypes (Extended Data Fig. 10). Together, these results demonstrate that, although REV-ERBα deletion may mildly reduce SPO⁺ synapses at baseline, it also prevents tau-mediated hippocampal neurodegeneration.

Fig. 4. Hippocampal volumes and CA1 thickness in PS19 mouse brain are improved by REV-ERBα deletion.

a, Schematic showing compartmentation of mouse hippocampus region into the CA1, CA3 and dentate gyrus (DG) based on DAPI staining. b, Representative images of DAPI staining from four different groups (WT, RKO, PS19 and PS19;RKO). Scale bars, 500 μm. c, Quantification of DAPI percent area in CA1, CA3 and DG compartments in all four genotypes of mice showing loss in PS19 mice, which is rescued in PS19;RKO mice in two regions (WT, n = 8; RKO, n = 14; PS19; n = 10, PS19;RKO, n = 12 mice). d,e, Representative images showing SPO (cyan) staining in all four groups (WT, RKO, PS19 and PS19;RKO) (d) and quantification showing mild synapse loss in RKO and more severe in PS19 without rescue by REV-ERBα deletion (WT, n = 8; RKO, n = 14; PS19; n = 10, PS19;RKO, n = 12 mice) (e). Scale bars, 500 μm. f,g, NeuN staining of four groups to measure CA1 thickness (f) and quantification showing CA1 thinning in PS19 mice that is rescued by REV-ERBα deletion specifically in males (WT, n = 8; RKO, n = 14; PS19; n = 10, PS19;RKO, n = 12 mice) (g). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001; ‘NS’ is non-significant by two-way ANOVA with Sidakʼs multiple comparisons test. Error bars represent mean ± s.e.m.

Source data

Extended Data Fig. 10. Hippocampal c-fos staining is not altered by REV-ERBα deletion in regular or PS19 mice.

(a) Representative images showing immuno-labelling for c-fos (green) and NeuN (red) in four different groups; WT, RKO, PS19, PS19;RKO with (b) quantification (% area normalized to Cre- control group, n = 12-17 mice per group). Scale bars, 500 μm (whole hippocampus); Scale bars, 50 µm (inset image). ns is non-significant by 2-tailed T-test. Error bars represent mean ± SEM.

Source data

REV-ERBα modulates Cd38 expression specifically in astrocytes

We suspected that astrocytes may be the main cell type to drive the effect of REV-ERBα on tauopathy through CD38-mediated NAD⁺ regulation. Thus, we generated mice with conditioned deletion of REV-ERBα in specifically in astrocytes in the brain, using the Aldh1l1-Cre^ERT2 mouse line⁴⁴. We termed these mice ARKO (Aldh1l1-Cre^ERT2;Nr1d1^fl/fl) (Fig. 5a) and treated Cre⁻ and Cre⁺ littermates with tamoxifen at 2 months of age. Deletion efficiency of REV-ERBα was confirmed by examining representative REV-ERBα downstream genes such as Fabp7 and Bmal1, both of which increased in expression due to loss of REV-ERBα-mediated repression (Fig. 5b). To investigate whether Cd38 expression changes more in response to REV-ERBα KO in astrocytes than microglia, we compared hippocampal Cd38 expression in three different cell-type-specific, postnatal REV-ERBα KO mouse lines, including our global REV-ERBα KO (CAG::CreER^T2;Nr1d1^fl/fl; Fig. 1), astrocyte-specific REV-ERBα KO ARKO and microglia-specific REV-ERBα KO MRKO (Cx3cr1::CreER^T2;Nr1d1^fl/fl), which were previously described¹². For all of these mouse lines, Cre⁻ littermates were used as control, and all were treated with tamoxifen at 2 months and harvested at 4 months. Cd38 downregulation was observed in global and astrocytic REV-ERBα KO (ARKO) but not in microglial REV-ERBα KO (MRKO) (Fig. 5c). Moreover, astrocytic REV-ERBα KO caused increased brain NAD⁺ levels similar to global KO mice (Fig. 5d), indicating that astrocytes could be the main cell type for the action of REV-ERBα on CD38 function.

Fig. 5. REV-ERBα–NFIL3–CD38 axis controls astrocyte NAD⁺ levels.

a, Mouse model of astrocyte-specific REV-ERBα KO (ARKO, Aldh1l1-CreER^T2;Nr1d1^fl/fl). b, Astrocyte-specific REV-ERBα KO mice (ARKO) show increased expression of known downstream REV-ERBα target transcripts, including Fabp7 and Bmal1 in the cortex after tamoxifen treatment, compared to Cre⁻ littermates (WT) (n = 6 mice per group). c, Comparing hippocampal Cd38 expression across three different tissue-specific REV-ERBα KO mouse lines: global (RKO; CAG-Cre^ERT2, n = 9 mice; paired WT, n = 12 mice), astrocyte (ARKO; Aldh1l1-Cre^ERT2, n = 6; paired WT mice, n = 6 mice) and microglia (MRKO; Cx3cr1-Cre^ERT2, n = 10 mice; paired WT, n = 12 mice). d, Increased brain NAD⁺ levels in ARKO mice (n = 9; paired WT, n = 8). e, AT8⁺ pTau pathology in hippocampus (HIP) and entorhinal cortex (EC) of PS19 and PS19;ARKO mice at 9 months. Scale bars, 500 μm. f, Hippocampal AT8 percent area from d (PS19, n = 11 mice; PS19;RKO, n = 8 mice). g, Expression of transcripts related to astrocyte (Gfap) and microglial (Iba1/Aif1 and Cd68) activation in HIP of WT, ARKO, PS19 and PS19;ARKO mice (WT, n = 8; ARKO, n = 9; PS19, n = 7; PS19;ARKO, n = 7). h, Cd38 mRNA levels in all four groups (n = 6 mice per genotype). i, Similar brain NAD⁺ levels between PS19 and PS19;ARKO mice at 9 months (PS19, n = 10; PS19;ARKO, n = 8). *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001; ‘NS’ is non-significant by two-tailed t-test or two-way ANOVA with Sidakʼs multiple comparisons test. Error bars represent mean ± s.e.m. EC, entorhinal cortex; HIP, hippocampus; ptn, protein.

Source data

Astrocytic REV-ERBα deletion augments brain NAD⁺ levels and mitigates tau pathology in PS19 mice

We next sought to determine if astrocyte-specific deletion of REV-ERBα in the brain is sufficient to recapitulate the effects on tau pathology that we observed in global PS19;RKO mice. We crossed ARKO mice with P301S^+/−;Nr1d1^fl/fl mice to generate PS19;ARKO (Cre⁺) and PS19 (Cre⁻) control littermates (all with two floxed Nr1d1 alleles), all of which were treated with tamoxifen at 2 months and then aged to 9 months to assess pathology. We observed a significant reduction of hippocampal AT8⁺ pTau pathology in ARKO;PS19 mice compared to Cre⁻;PS19 controls and saw a trend toward reduction of pTau in the entorhinal cortex region (Fig. 5e,f). Transcripts related to glial activation, including Gfap and Iba1, were also reduced in ARKO;PS19 mice (Fig. 5g), along with downregulation of Cd38 expression (Fig. 5h), indicating that astrocytic REV-ERBα deletion itself can prevent tauopathy. NAD⁺ levels in ARKO;PS19 brain showed a trend toward induction without statistical significance compared to PS19 alone, perhaps due to mixed effects of tauopathy and REV-ERBα deletion on NAD⁺ (Fig. 5i).

Astrocytic Cd38 reduction enhances NAD⁺ levels and tau phagocytosis in vitro

To clarify a molecular mechanism by which astrocyte REV-ERBα might impact tau pathology through CD38 reduction, we next assessed the effect of CD38 knockdown on astrocytic tau phagocytosis in vitro. We transfected primary murine astrocyte cultures with siRNA targeting REV-ERBα (siNr1d1) or a non-targeting siRNA (siCon), achieving Nr1d1 knockdown with more than 70% efficiency (Fig. 6a). As expected, Cd38 was reduced (Fig. 6a), and NAD⁺ levels were also significantly increased in siNr1d1-transfected cells compared to control (Fig. 6b). To test whether CD38 affects NAD⁺ levels, we knocked down Cd38 using siRNA (siCd38) in astrocyte cultures (Fig. 6c) and observed that Cd38 knockdown led to a significant increase in NAD⁺ (Fig. 6d). We next explored the possibility that diminished Cd38 expression might affect the phagocytic activity of astrocytes. To address this, we used dye-quenched bovine serum albumin (DQ BSA), a reagent that fluoresces only upon lysosomal cleavage and is used to monitor lysosomal proteolytic activity^45,46. In the presence of pretreatment with cytochalasin D (Cyto D), an inhibitor of actin polymerization and phagocytosis (Fig. 6e), proteolysis of DQ BSA was dramatically inhibited (Fig. 6f), indicating that this system reflects the phagocytic ability of the cells. We observed an increase in DQ BSA signal in astrocytes after Cd38 knockdown (Fig. 6g), suggesting that Cd38 depletion elevates general protein degradation through the endolysosomal pathway. Next, we examined the impact of Cd38 knockdown on astrocyte uptake of tau aggregates. We knocked down Cd38 in primary astrocytes and then treated with FITC-conjugated tau aggregates, as we previously described¹². Engulfment of FITC⁺ tau was increased by Cd38 depletion (Fig. 6h), suggesting that CD38 deficiency, as is seen in REV-ERBα KO, enhances astrocyte tau uptake and lysosomal proteolytic activity. These findings provide a plausible mechanistic link among astrocytic REV-ERBα deletion, Cd38 suppression and decreased tau pathology.

Fig. 6. Cd38 knockdown in astrocytes enhances tau phagocytosis and lysosomal activity.

a, Effect of siRNA treatment with control (siCon) or Nr1d1 (siNr1d1) siRNA in primary astrocyte cultures on Nr1d1 and Cd38 transcript levels (n = 14 from biological replicates). b, NAD⁺ levels in siCon- or siNr1d1-treated cultured primary astrocytes (n = 6 from biological replicates). c, Knockdown efficiency of siCd38 siRNA on Cd38 mRNA in cultured primary astrocytes (n = 6 from biological replicates). d, Increased NAD⁺ levels in siCd38-transfected cultured primary astrocytes (n = 6 biological replicates). e, Graphic showing experimental strategy for DQ-Red BSA assay using Cyto D, an actin polymerization inhibitor that blocks phagocytosis. f, Cyto D treatment (20 μM) significantly reduced the DQ BSA⁺ astrocyte population (non-stain, n = 6; DQ BSA, n = 9; DQ BSA + Cyto D, n = 9 from biological triplicates). g, Knockdown of Cd38 in primary astrocyte cultures increased the DQ BSA⁺ population compared to control group (n = 9 plates from three mice). Non-treat group did not receive DQ-BSA. h, Engulfment of FITC-tau aggregates was increased in siCd38-treated primary astrocytes (non-treat, n = 4; siCon, n = 7; siCD38, n = 7 plates from three mice). Typical flow cytometry output as well as both mean fluorescence intensity (MFI) and %FITC⁺ astrocytes are shown. Non-treat group did not receive FITC-tau. **P < 0.01, ***P < 0.005 and ****P < 0.001 by two-tailed t-test or two-way ANOVA with Sidakʼs multiple comparisons test. Error bars represent mean ± s.e.m. ptn, protein.

Source data

Administration of REV-ERBα antagonist drug SR8278 relieves the tau pathology in PS19 mice

Based on the observed effects of REV-ERBα deletion on tauopathy, we next performed therapeutic intervention in PS19 mice using a brain-permeant small-molecule REV-ERB antagonist, SR8278 (ref. ¹⁶). Previous studies showed beneficial effects of SR8278 in amyloid-β pathology¹³ and Parkinson disease models⁴⁷. We performed subcutaneous injection of SR8278 (50 mg kg⁻¹ daily at Zeitgeber time 6 (ZT6)) for 2 weeks in PS19 mice starting at 8.5 months, and then we euthanized mice at 9 months. This is a critical point in the disease progression when tau aggregation and inflammation accelerate. We analyzed tau pathology by AT8 (pTau) and MC1 (misfolded tau) staining and observed that both were significantly decreased by SR8278 administration (Fig. 7a,b). We further measured transcripts related to circadian clock (Bmal1, Per2, Cry2and Nr1d1), microglial activation (Cd68and Trem2), astrocyte activation (Gfap, Emp1and Aqp4), complement components (C3, C1q and C4b), inflammation (Il1b and Tnfa) and lysosomal dysfunction (Nlrp3 and Lamp1) in vehicle (VEH)-treated or SR8278-treated PS19 mouse brain (Fig. 7c). SR8278 efficiently induced Bmal1 expression by inhibiting repressor activity of REV-ERBα and led to reduction of neuroinflammatory responses, including decreases in Gfap, Tnfa and C1q (Fig. 7d). Glial activation markers GFAP and IBA1 were also significantly downregulated as measured by immunohistochemistry (Fig. 7e). As genetic REV-ERBα deletion in PS19 led to high brain NAD⁺ levels with pathological changes, we assessed the effect of SR8278 on brain NAD⁺ in PS19. However, SR8278 did not significantly increase NAD⁺ levels, although there was a trend toward induction (Fig. 7f). In summary, an acute regimen of REV-ERBα antagonist SR8278 improved tau pathology, suggesting that it may be a promising strategy for AD therapy.

Fig. 7. Inhibition of REV-ERBα function with SR8278 improves tau pathology.

a,b, Hippocampal staining of pTau (AT8) (VEH, n = 6 mice; SR8278, n = 6 mice) (a) or aggregated tau (MC1) in piriform cortex (VEH, n = 3 mice; SR8278, n = 4 mice) (b) in PS19 mice treated daily from age 8.5 months to 9.0 months with VEH or SR8278. Scale bars, 500 μm. Dotted lines in b indicate borders of piriform cortex. c, Heatmap showing gene expression changes in PS19 mouse brain from VEH-treated and SR8278-treated mice. d, Individually plotted genes showing diminished neuroinflammation as well as the increased Bmal1 expression after SR8278 treatment compared to VEH group (n = 3 mice per group). e, Representative images showing reduced microglial activation (IBA1, red) and astrogliosis (GFAP, green) in cortex of VEH versus SR8278 mice, with quantification (VEH, n = 3 mice; SR8278, n = 4 mice). Scale bars, 500 μm. f, NAD⁺ levels in cortex from VEH-treated or SR8278-treated mice (VEH, n = 6 mice; SR8278, n = 7 mice). *P < 0.05, **P < 0.01; ‘NS’ is non-significant by two-tailed t-test. Error bars represent mean ± s.e.m.

Source data

Discussion

Our data demonstrate a REV-ERBα–NFIL3–CD38 axis that regulates brain NAD⁺ levels. This pathway is active in astrocytes, where REV-ERBα inhibition augments NAD⁺ and enhances astrocyte tau uptake and lysosomal function, ultimately protecting against tau-mediated neurodegeneration in vivo. In the heart, the REV-ERBα–NFIL3 axis does not alter Cd38 expression but, instead, suppresses Nampt, with a previous report showing that REV-ERBα/β deletion in cardiac tissue reduces NAD⁺ and causes cardiomyopathy¹⁴. We observed that REV-ERBα has an opposite effect and a unique mechanism for controlling NAD⁺ levels in the brain through CD38 regulation, not through Nampt, suggesting that REV-ERBα uses opposing regulation of Nampt versus Cd38, both downstream of Nfil3, to finetune NAD⁺ levels in a tissue-specific manner. Considering the immense interest in augmenting NAD⁺ levels to prevent aging and neurodegenerative diseases^3,20, our results illustrate the tissue-specific role of REV-ERBα in NAD⁺ regulation and reveal a neuroprotective function for REV-ERBα inhibition in tauopathy.

Accumulating evidence demonstrates that circadian proteins can influence neurodegenerative disease pathogenesis through modulation of glial activation and proteostasis^12,13,48,49. Here we show, in an AD-relevant tauopathy model, that inhibition of the circadian nuclear receptor REV-ERBα, either genetically or pharmacologically, enhanced NAD⁺ levels and reduced tau pathology and neurodegeneration. Moreover, our results show that REV-ERBα–NFIL3–CD38–NAD⁺ axis regulates astrocyte phagocytosis of tau, as well as lysosomal proteolytic activity, providing a mechanism to explain the protective effect on tau accumulation observed with REV-ERBα deletion in PS19 mice in vivo. These findings are in keeping with our previous work showing that germline REV-ERBα deletion enhances microglial amyloid-β phagocytosis and reduces amyloid plaque deposition in 5×FAD mice¹³. Our studies highlight the potential importance of CD38 in the brain as a molecular target of REV-ERBα for NAD⁺ regulation. CD38 has been suggested as a possible therapeutic target for diverse age-dependent diseases, including AD⁵⁰, as genetic deletion of CD38 reduces pathology in an AD model⁵¹. Pharmacological inhibition of CD38 can reduce neuroinflammation in response to lipopolysaccharide (LPS)⁵², prevent astrogliosis and demyelination in a lysolecithin model⁵³ and improve metabolic function and lifespan in aged mice⁵⁴. Our findings could help to illuminate promising therapeutic approaches targeting CD38 to prevent tau pathology in AD and other diseases.

Although we have focused on NAD⁺ regulation here, it is likely that REV-ERBα exerts multiple effects on neurodegenerative pathology, which may work in parallel with the NAD⁺ induction pathways that we have described. For example, REV-ERBα deletion also induces Bag3 (Fig. 2d), a chaperone protein that can mediate tau degradation^55,56, which we previously showed to be expressed in astrocytes downstream of the core clock protein BMAL1 (ref. ⁴⁹). As REV-ERBα levels are suppressed in Bmal1 KO mouse brain⁵⁷, our findings also suggest that upregulation of Bag3 by loss of REV-ERBα may contribute to the protective effect observed here in PS19 mice. REV-ERBs also influence mitochondrial function and other pathways, which we have not addressed here.

The effects of REV-ERBα in the brain under basal conditions and in the setting of neurodegeneration are complex, but our findings here suggest that the overriding impact of REV-ERBα deletion is protective in tauopathy. Our previous studies showed that germline REV-ERBα KO can stimulate microglial activation, inflammation and synaptic loss under basal conditions^11,41 but had a protective effect on amyloid-β plaque deposition in 5×FAD mice¹³. As we did not observe spontaneous neuroinflammation and microglial activation in inducible postnatal REV-ERBα KO mouse brain as we did in germline REV-ERBα KO mice¹¹, we suspect that these severe neuroinflammatory responses in germline REV-ERBα KO mice may be due to complete gene deletion (as our inducible systems used here achieved only 70–85% deletion) and/or developmental effects. It is feasible that total loss of REV-ERB function in the brain may have some deleterious effects, whereas strong inhibition of REV-ERBs (60–80%), as is seen with tamoxifen-inducible postnatal deletion or with pharmacological inhibitors, may avoid these negative effects and promote protective metabolic and proteostatic changes. Further study is needed to understand optimal REV-ERB manipulation to prevent neurodegenerative pathology.

As further evidence of the complexities of REV-ERBα in the brain, we previously demonstrated that microglia-specific deletion of REV-ERBα promotes inflammation and tau aggregation specifically in male PS19 mice¹², whereas here we show that global and astrocyte-specific REV-ERBα deletion can mitigate tau pathology and augment NAD⁺ levels without sex dependency. This indicates that microglial REV-ERBα has a unique role in regulating tau pathology but that this effect is overwhelmed in our global REV-ERBα KO mice by protective effects, likely mediated by REV-ERBα in astrocytes. Similarly, global pharmacological REV-ERB inhibition with SR8278 has a net protective effect on tau pathology in vivo, suggesting that the microglia-specific phenotype is not dominant. We postulate that REV-ERBα inhibition could be therapeutically appropriate to prevent toxic protein aggregation in tau-related diseases such as AD. However, other studies show that pharmacological REV-ERBα activation can also have protective effects in the brain, as treatment of an accelerated aging mouse model (SAMP8) with the REV-ERB agonist drug SR9009 improved cognitive function and reduced soluble amyloid-β levels⁵⁸. It is possible that REV-ERBα inhibition could be optimal in certain disease stages or states, such as in early stage of AD when tau aggregation is beginning, whereas REV-ERBα activation might be preferable in other circumstances, such as disease states without prominent protein aggregates in which impairment of mitochondrial function predominates. Our studies shed light on crucial neuroprotective mechanisms mediated by REV-ERBα in astrocytes and should help guide ongoing drug development efforts to target REV-ERBα function in AD and other neurodegenerative diseases.

Methods

Mice

All mouse experiments were performed according to protocols approved by the Washington University Institutional Animal Care and Use Committee (IACUC), Office of Laboratory Animal Welfare Assurance D1600245 (USDA registration no. 43-R-008) and under supervision of the Department of Comparative Medicine. Mice were group housed in all experiments and were bred in our facility at Washington University in St. Louis. Nr1d1^fl/fl mice¹⁷ were used for generating global RKO mice. CAG::Cre^ERT2+ and Aldh1l1::Cre^ERT2 mice, both on the C57BL6/j background, and MAPT P301S mice (mixed background, but bred to C57BL6/j) were obtained from The Jackson Laboratory. All mice were housed under a 12-hour light/dark cycle with free access to water and food. Mice were given tamoxifen (Sigma-Aldrich, 2 mg per day for 5 days) by oral gavage at 2 months of age to induce REV-ERBα deletion. Cre⁻ littermates were also treated with tamoxifen. All experimental mice were deeply anesthetized by intraperitoneal injection with Fatal-Plus pentobarbital and subjected to thoracotomy and transcardiac perfusion with cold DPBS containing 0.3% heparin in the afternoon, between 13:00 and 18:00, with most between 13:00 and 15:00. Mice with unnecessary genotypes from each cohort were euthanized by CO₂ according to the approved protocol of the Washington University IACUC.

Immunohistochemistry and imaging

Brain tissue was then either dissected and flash frozen in liquid nitrogen for NAD⁺, protein and RNA studies or drop-fixed in 4% paraformaldehyde solution for 24 hours for sectioning and immunohistochemistry. Fixed hemispheres were then incubated with 30% sucrose for 24 hours at 4 °C on a shaker, and then 40-µm-thick frozen sections were prepared on a Leica sliding microtome. All sections were stored in cryoprotectant solution at −20 °C prior to use. For fluorescence staining, sections were blocked and permeabilized with 3% donkey serum in PBS containing 0.4% Triton X-100 (PBSX) and stained with the following antibodies: mouse anti-NeuN (Sigma-Aldrich, MAB377, 1:500), mouse anti-GFAP (Novus Biologicals, 5C10, 1:2,000), rabbit anti-IBA1 (Wako, 019-19741, 1:1,000), mouse anti-SPO (Abcam, 8049, 1:1,000), rabbit anti-c-fos (Cell Signaling Technology, 9F6, 1:1,000), rabbit anti-S100β (Abcam, ab52642, 1:500) and MC1 (tau aggregate, gift of Peter Davies, AB_2314773). Sections were incubated with primary antibodies in PBSX with 0.4% Triton X-100 containing 1% donkey serum at 4 °C overnight and were then washed three times with PBS and incubated with secondary antibodies (1:400) in PBS with 0.25% Triton X-100 for 1 hour at room temperature. Sections were washed three times with PBS and stained with DAPI (1:1,000) for 15 minutes. Additional washes were performed three times, and then sections were mounted on slides with ProLong Gold mounting media. All fluorescent imaging was performed on a Keyence BZ-X810 microscope. Light intensity and exposure times were determined from each cohort of samples after a survey of the tissue in order to select appropriate parameters that could then be held constant for all slides in that imaging session. These values varied by antibody, but all sections in a given cohort were imaged under identical conditions at the same magnification.

pTau (AT8) staining

Brain sections were washed three times with Tris-buffered saline (TBS) for 5 minutes and incubated with 0.3% hydrogen peroxide (H₂O₂) in TBS for 10 minutes at room temperature. Sections were then washed three times with TBS for 5 minutes and blocked with 3% skim milk in 0.25% Triton X-100 (TBSX) for 30 minutes and then incubated with biotinylated AT8 antibody (Thermo Fisher Scientific, MN1020, 1:1,000) in 0.25% TBSX containing 1% skim milk at 4 °C overnight. After 3× washes with TBS for 5 minutes, the sections were incubated with ABC Elite (Vector Laboratories, PK-6100, 1:400) in TBS for 1 hour and were visualized using 3,3′-diaminobenzidine tetrahydrochloride (DAB) (Chromogen).

SR8278 drug treatment

Aged P301S mice were kept under standard 12-hour light/dark conditions. Pure SR8278 gel was obtained from the Burris laboratory and stored under nitrogen gas at −20 °C. SR8278 was then diluted in a VEH solution of 10% DMSO, 10% Tween 80 and 90% PBS. Fresh stock was made every 3–4 days and kept under nitrogen at 4 °C. The VEH alone without SR8278 was injected as a control. All injections were subcutaneous and given at ZT6 plus or minus 1 hour.

Primary astrocyte and cell culture

Astrocyte-enriched cultures were prepared from mixed glial cultures that were obtained from postnatal days 1–3 (P1–3) mouse pups. Dissected cortical regions were isolated and the meninges removed. Tissue was then trypsinized with 0.05% trypsin-EDTA for 15 minutes at 37 °C. Cells were neutralized with 2 volumes of complete media (DMEM + 10% FBS; FBS + 100 U ml⁻¹ penicillin–streptomycin) and centrifuged at 500g for 5 minutes. After discarding of the supernatant, cells were resuspended with full volumes of complete media, plated on poly-d-lysine (PDL)-coated T75 flasks and grown in complete media. Media were changed 2 days later and kept for 7 days, after which microglia were removed by shaking at 180 r.p.m. for 2 hours in a heated shaker. This leads to approximately 94% astrocyte-enriched culture, although some microglia and oligodendrocyte precursor cells do remain. Astrocytes were collected and seeded on PDL-coated plates for experiments. Conditional REV-ERBα-deleted astrocyte cultures were prepared from the CAG::CreER^T2;Nr1d1^fl/fl pups and Cre⁻;Nr1d1^fl/fl control littermate pups and were treated with 4-hydroxytamoxifen (1.5 μM) for 24 hours after plating to induce gene deletion. To perform LPS studies, we treated cells with 50 ng ml⁻¹ LPS (Sigma-Aldrich, L6529) for 24 hours before harvest. All cells were maintained at 37 °C in a humidified atmosphere with 5% CO₂.

Immunoblotting

Dissected brain tissues were typically sonicated (60% amplitude, 1 minute with 1-second pulse on/off) in RIPA lysis buffer (Nalgene, R0278), and protein concentration was determined by BCA assay (Thermo Fisher Scientific, BCA Protein Assay Kit, 23225). Each sample was separated by electrophoresis on 4–12% SDS-PAGE gels and then transferred electrophoretically to PVDF membranes. The membranes were blocked with 5% skim milk and then washed three times with 0.1% Tween 20 containing TBS (TBST). The membranes were then gently incubated at 4 °C overnight with the following primary antibodies in TBST with 3% BSA: HT7 (Thermo Fisher Scientific, MN1000, 1:500), AT8 (Thermo Fisher Scientific, MN1020, 1:1,000), SIRT1 (Cell Signaling Technology, 2028S, 1:1,000), CD38 (Proteintech, 60006-1-Ig, 1:1,000) and β-tubulin loading control (Invitrogen, MA5-16308, 1:1,000). K174 and K274 were kindly provided by Li Gan^20,21. The following day, the membranes were washed and then incubated with HRP-labeled anti-mouse or anti-rabbit secondary antibodies for 1 hour at room temperature. Subsequently, membrane-bound HRP-labeled antibodies were detected using an enhanced chemiluminescence detection system including Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific). Quantification of the bands was performed with ImageJ 1.44 (Image Processing and Analysis in Java).

NAD⁺ measurement

NAD⁺ was extracted from frozen brain/liver tissues or cultured cells with 0.5% dodecyl trimethyl ammonium bromide (DTAB; Sigma-Aldrich, D8638) with sonication (60% amplitude, 1 minute with 1-second pulse on/off), and the supernatant was harvested. Supernatant was incubated with 0.5 volumes of 0.4 N HCl for acidification for 15 minutes at 65 °C and was equilibrated for an additional 10 minutes at room temperature. For neutralization of the samples, 0.5 M Trizma base was added with an equal volume of NAD/NADH-Glo Detection Reagent (Promega, G9071). Assay plate was incubated for 60 minutes at room temperature with shaking, and luminescence was observed every 20 minutes during incubation using a Cytation 5 plate reader (Agilent, BioTek Cytation 5 Cell Imaging Multimode Reader).

DQ-Red BSA protein cleavage assays

siScramble or siCD38 astrocytes were treated with 1 µg ml⁻¹ DQ-Red BSA for 3 hours. Cells were washed with DPBS twice and trypsinized with TryplE and then collected in astrocyte growth medium (10% FBS, 1% penicillin–streptomycin in DMEM). Collected cells were centrifuged at 1,000g for 5 minutes and resuspended in flow cytometry buffer (1% FBS, 1 mM EDTA in DPBS). Cells were then analyzed on a Beckman Coulter CytoFLEX S flow cytometer. DQ BSA was determined by FL8 Yellow, after pregating for live, single cells. Data were analyzed in FlowJo version 10.10.0.

RNA preparation and quantitative polymerase chain reaction analysis

Tissues were bead-homogenized in 500 µl of TRIzol in a bullet blender for 3 minutes. Cultured cells were homogenized by adding 500 µl of TRIzol per well and were collected after incubation for 10 minutes with shaking at room temperature. TRIzol samples were then subjected to chloroform extraction (1:6 chloroform:TRIzol) and centrifugated at 12,500g for 15 minutes at 4 °C. RNA was then extracted from the supernatant using a PureLink RNA Mini Kit (Invitrogen, 12183018A) according to the manufacturer’s instructions. The RNA concentration was determined using a NanoDrop ND 1000 spectrophotometer. cDNA was then synthesized using approximately 1 µg of RNA using a High-Capacity cDNA Reverse Transcription Kit (Life Technologies, 4368814) according to the manufacturer’s instructions. Real-time quantitative polymerase chain reaction (qPCR) was performed using ABI TaqMan primers and ABI PCR Master Mix buffer on ABI StepOnePlus or QuantStudio 3 thermocyclers (Applied Biosystems). TaqMan qPCR primers were purchased from Thermo Fisher Scientific: Cre, Mm00635245_cn; Cd68, Mm03047343_m1; Gfap, Mm01253033_m1; Aif1, Mm00479862_g1; Il1b, Mm00434228_m1; Il6, Mm00446190_m1; Tnfa, Mm00443258_m1; Nr1d1, Mm00520711_m1; Nampt, Mm00451938_m1; Plin2, Mm00475794_m1; C1qa, Mm00432142_m1; Cd38, Mm00483143_m1; Apoe, Mm01307192_m1; Nfil3, Mm00600292_s1; Fabp7, Mm00445225_m1; and Actn, Mm02619580_g1.

RNA-seq and analysis

RNA-seq and analysis were performed by the Genome Technology Access Center at Washington University using their standard methods, which are summarized here. RNA was extracted as above. Sample RNA integrity was determined using a TapeStation, and library preparation was performed with 10 ng of total RNA for samples with a Bioanalyzer RNA integrity number (RIN) score greater than 8.0. Double-stranded cDNA was prepared using a SMARTer Ultra Low RNA Kit for Illumina sequencing (Takara-Clontech) per the manufacturer’s protocol. cDNA was then fragmented using a Covaris E220 sonicator using peak incident power 18, duty factor 20% and cycles per burst 50 for 120 seconds. The cDNA was blunt ended, had an A-base added to the 3′ ends and then had Illumina sequencing adapters ligated to the ends. Ligated fragments were then amplified for 12–15 cycles using primers incorporating unique dual index tags. The fragments for each sample were then pooled in an equimolar ratio and sequenced on an Illumina NovaSeq 6000 using 150-bp paired-end reads. Basecalls and demultiplexing were performed with Illumina’s RTA 1.9 software, and the reads were aligned to the Mus musculus Ensembl release 76 GRCm38 primary assembly with STAR version 2.5.1a. Gene counts were quantitated with Subread:featureCount version 1.4.6-p5. All gene counts were then imported into the R/Bioconductor package edgeR, and trimmed mean of M values (TMM) normalization size factors were calculated to adjust the samples for differences in library size. Ribosomal genes were removed, and only genes expressed greater than one count per million (CPM) in at least four samples were kept for further analysis. The adjusted TMM size factors and the matrix of counts were then imported into the R/Bioconductor package limma. Weighted likelihoods were then calculated for all samples, and the count matrix was transformed to moderated log₂CPM with limma-voom with quality weights. Differential expression analysis was then performed to analyze for differences between conditions, and the results were filtered for only those genes with Benjamini–Hochberg false discovery rate (FDR)-adjusted P values less than or equal to 0.05.

Untargeted metabolite analysis by mass spectrometry

For each sample, 20 mg of brain tissue was homogenized with 200 μl of pre-chilled 80% methanol using a pulsed tip sonicator on dry ice. The metabolite extraction was performed by adding pre-chilled 80% methanol up to 1 ml and storing at −80 °C overnight. After centrifugation, the supernatant fraction was dried using a SpeedVac vacuum concentrator (Thermo Fisher Scientific) with no heat. The dried pellet was reconstituted in 60 µl of 50% MeOH and transferred to the autosampler vial and injected. The extraction was run using the two liquid chromatography–mass spectrometry (LC–MS) systems on Vanquish Horizon coupled with an Orbitrap IDX Tribrid MS (Thermo Fisher Scientific). The detailed LC–MS methods are as follows. (1) For the C18 column system, 5 µl of extraction was injected on an ACQUITY UPLC HSS T3 column (2.1 × 150 mm, 1.7 µm; Waters). The column was isocratically equilibrated at a flowrate of 0.4 ml min⁻¹ and 30 °C oven temperature with 98% of mobile phase A (0.1% formic acid in water) for 2 minutes, followed by a linear gradient to 27% of mobile phase B (0.1% formic acid in 90% acetonitrile) over 7 minutes, a linear gradient to 98% of mobile phase B over 7 minutes and then 2 minutes at 98% of mobile phase B. The MS1 full scan was performed over 100–600 m/z at 60,000 resolution, using heated electrospray ionization in the positive and negative ion mode, separately. (2) For the hydrophilic interaction chromatography (HILIC) column system, 5 µl of extraction was injected on a BEH Amide column (2.1 × 150 mm, 1.7 µm; Waters). The column was isocratically equilibrated at a flowrate of 0.5 ml min⁻¹ and 50 °C oven temperature with 95% of mobile phase B (acetonitrile, pH 9) for 3 minutes, followed by a linear gradient to 80% of mobile phase B over 7 minutes, a linear gradient to 50% of mobile phase A (20 mM ammonium acetate in water, pH 9) over 5 minutes and then 4 minutes at 50% of mobile phase A. The MS1 full scan was performed over 70–800 m/z at 60,000 resolution, using heated electrospray ionization in the positive and negative ion mode, separately. The ‘Deep Scan’ mode of AcquireX MS2 data acquisition was applied between every batches using the samples pooled by batches with generating one inclusion list followed by three times of exclusion override (inclusion list peak fragmentation threshold, 50%; exclusion duration, 10 seconds). Additional MS settings are as follows: capillary voltage in the positive mode 3.4 kV, negative mode 2.5 kV, ion transfer tube temperature 325 °C, vaporizer temperature 350 °C and normalized collision energy 35%. For monitoring the instrument and column performance, QReSS standard mixture (Cambridge Isotope Laboratories) was run at the starting point and between every batch.

Metabolomics data postprocessing

We used Compound Discoverer 3.3 (Thermo Fisher Scientific) to process raw LC–MS data for spectra selection, retention time alignment between samples, grouping all the detected features by molecular weight and retention time, calculating area under the curve feature abundances and assigning putative class and/or metabolite annotation by elemental composition prediction and spectral library matching. The following parameters were used for the feature detection and grouping: mass tolerance 5 ppm, minimum peak intensity 1 × 10⁴, retention time tolerance 0.2 minutes, signal-to-noise threshold 3 and peak rating threshold 5 in at least 10% of samples. The gaps for missing peaks were filled with redetected peak, matching ion, spectrum noise and trace area. The value was considered as full gap if the gap was filled with spectrum noise or trace area, and the features without full gap in at least 20% of samples were used for further processing. The features lower than 2 of sample-to-blank ratio were marked as background and excluded for further analysis. The parameters for predicting elemental compositions are as follows: mass tolerance 5 ppm, maximum ring-and-double-bonds-equivalent 40, minimum hydrogen-to-carbon ratio 0.1, maximum hydrogen-to-carbon ratio 3.5, intensity tolerance 30%, signal-to-noise threshold 3, minimum spectral fit 30% and minimum pattern coverage 90%. For the library search, the databases from mzCloud (https://www.mzcloud.org/), MassBank of North America (http://mona.fiehnlab.ucdavis.edu), Global Natural Product Social Molecular Networking⁵⁹ and National Institute of Standards and Technology (NIST23) were used. The following parameters were used for library matching: precursor and fragment ion mass tolerance 10 ppm, minimum matching score 0.5 (cosine algorithm for mzCloud and HighChem HighRes algorithm for other databases) and ion activation energy tolerance 20. Any library matches were considered as level two according to the metabolomics standards consortium guidelines⁶⁰. The mass lists of the NAD-related metabolites were generated according to the related pathways and Giner et al.⁶¹, to identify the precursor ions corresponding to NAD metabolites by 5-ppm mass tolerance and 0.5-minute retention time tolerance.

Quantile normalization was performed using the ‘preprocessCore’ package in R. To merge the two batches of the data, the batch effect correction was performed by the ComBat function in the ‘sva’ package in R. PCA and Student’s t-test were performed and visualized using the base R package. The statistics data were visualized using the ‘ggplot2’, ‘FactoMineR’ and ‘factoextra’ packages in R. The pathway enrichment analysis was performed using the ‘RAMP’ package (version 2.0.2) in R.

Nfil3 binding site analysis

The Cd38 locus in the mouse genome (mm10 assembly) was identified on chromosome 5 (chr5:43,843,469–43,978,857, forward strand), located between the 3′ end of Bst1 and the 5′ start of Fgfbp1 genes. Intersection between the astrocyte-specific candidate cCREs derived from brain single-cell ATAC–seq data²⁴ and the Cd38 locus identified astrocyte-specific cCREs within the Cd38locus. Putative regulatory sequences from multiple epigenomic resources, including the Mouse ATAC–seq Atlas²⁷, DNase-seq data from the Cistrome database²⁵ and ENCODE cCREs from the ENCODE project²⁶, were downloaded from corresponding websites. The PWM for Nfil3 (motif ID: Nfil3_M01819_2.0) was retrieved from the CIS-BP database⁶². The Patser program was used to scan genomic sequences to identify putative Nfil3 binding sites in the mouse genome^28,29. Intersection between putative Nfil3 binding sites and peak cCREs353987 identified putative Nfil3 binding sites within the peak cCREs353987. The Integrative Genomics Viewer (IGV) motif tool was used to visualize the results. The sequence logo of the 12 putative Nfil3 binding site sequences that overlapped with the single-nucleus ATAC–seq cCRE peak cCREs353987 was created using WebLogo 3-Create (https://weblogo.threeplusone.com/create.cgi).

Statistics and reproducibility

Sample sizes for P301S mouse experiments were based on sample size calculations using data on tau pathology from previous studies in our laboratory, with a goal of detecting a 50% difference in AT8 percent area with 80% power and α = 0.05. Sample sizes for other experiments were estimated based on previous experience with similar studies. Outliers were identified by the Grubbs test and excluded. Experimenters were blinded to genotype during data collection. Experiments with genetic perturbations were naturally randomized by genotype within the teach cohort, whereas SR8278 experiments were randomized within each litter of mice such that an equal number of mice received drug or VEH in each litter. For the statistical analysis in all figures, graphs depict the mean + s.e.m., and n generally indicates the number of animals and cells, unless otherwise noted in the figure legend. Student’s t-tests or one-way or two-way ANOVAs with Tukeyʼs post hoc tests were performed using GraphPad Prism software version 8. An F-test was first performed for datasets with a single dependent variable and two groups, to determine if variances were significantly different. If not, two-tailed unpaired t-test was performed. For two-way ANOVAs, multiple comparisons testing was performed only if the main effect was significant at P < 0.05. Statistical tests were performed with GraphPad Prism software version 10. P < 0.05 was considered significant, and asterisks indicate the P value: *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001.

Reporting summary

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

Author contributions

J.L. and E.S.M. planned the study and designed the experiments. J.L. performed and analyzed all experiments, except for those indicated below. M.A.L. and L.N.W. provided transgenic mice and intellectual input on data interpretation. J.M.D. and E.I.Q. contributed to developing mouse lines, and R.K. and S.P. assisted with experiments. J.R.V. and T.P.B. synthesized and provided SR8278 as well as intellectual input on its use. L.G. provided in-house antibodies. B.V.L. assisted with NAD⁺ analysis studies, and metabolomic approaches were performed and analyzed by M.S. and Y.-A.G. I.O.S. and G.Z. supported genomic data analysis. J.L. and E.S.M. initially wrote the paper and revised it.

Peer review

Peer review information

Nature Aging thanks Zheng Chen and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

The authors declare that all the data supporting the findings of this study are available from the corresponding authors upon reasonable request. Source data, including raw western blots, are provided with this paper. No new code was generated. Raw RNA-seq files and metadata are publicly available on the Gene Expression Omnibus website under accession number GSE255975.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s43587-025-00950-x.

Supplementary information

The online version contains supplementary material available at 10.1038/s43587-025-00950-x.