Introduction
The epigenetic basis of many developmental, physiological, and metabolic processes is manifest. Epigenetic mechanisms control gene expression by potentially reversible changes in DNA methylation and chromatin structure. The remodeling of chromatin is largely elicited by enzyme-catalyzed posttranslational modifications of the core histone N-terminal tails (Kouzarides, 2007; Li et al., 2007a; Peterson and Laniel, 2004; Strahl and Allis, 2000). These include acetylation, poly(ADP-ribosylation), ubiquitination, methylation, and phosphorylation and represent critical regulatory events of a large array of nuclear responses. Unique combinations of these modifications, for which the “histone code” hypothesis has been formulated (Strahl and Allis, 2000), induce conformational changes of chromatin, rendering it permissive to transcription, silencing, DNA replication, and repair (Cheung et al., 2000a; Kouzarides, 2007; Kurdistani and Grunstein, 2003; Li et al., 2007a; Strahl and Allis, 2000).
Histone acetylation is recognized as one of the most prominent epigenetic marks leading to activation of gene expression (Strahl and Allis, 2000). Acetylation of the 3-amino groups of specific lysine residues in the N termini of core histones is generally associated with transcription activity, as it is thought to induce an open chromatin conformation that allows the transcription machinery access to promoters (Cheung et al., 2000a; Li et al., 2007a; Struhl, 1998). Indeed, acetylation of lysines in histones neutralizes the positive electric charge between the negatively charged DNA backbone and tips the balance toward relaxing chromatin. Deacetylation, on the other hand, would shift the balance back to condensing chromatin and silencing gene expression. The enzymes that elicit these critical transitions are histone acetyltransferases (HAT) and histone deacetylases (HDAC). HDAC-mediated deacetylation of histones correlates with gene silencing (Grunstein, 1997; Struhl, 1998; Wade and Wolffe, 1997; Workman and Kingston, 1998). HDACs have also been implicated in the reversible acetylation of nonhistone proteins, including p53 (Luo et al., 2001; Vaziri et al., 2001), Hsp90 (Kovacs et al., 2005), MyoD (Mal et al., 2001), and E2F1 (Martinez-Balbas et al., 2000). Mammalian HDACs have been classified into four classes based on their structure and regulation (Yang and Seto, 2008). There are seven mammalian enzymes constituting class III; these are homologs of yeast Sir2 (silencing information regulator) and are known as SIRT1 to SIRT7. These proteins are structurally distinct from the other HDACs and have the property of dynamically sensing cellular energy metabolism (Bordone and Guarente, 2005). Indeed, unlike other HDACs, SIRT proteins catalyze a unique reaction that requires the coenzyme NAD+ (nicotinamide adenine dinucleotide). In this reaction, nicotinamide (NAM) is liberated from NAD+ and the acetyl group of the substrate is transferred to cleaved NAD+, generating the metabolite O-acetyl-ADP ribose (Sauve et al., 2006). Due to the NAD+ dependency, SIRTs are thought to constitute one of the functional links between metabolic activity and genome stability and, finally, aging (Bishop and Guarente, 2007).
In yeast, the Sir2 complex mediates transcriptional silencing at telomeres and regulates the pace of aging (Chopra and Mishra, 2005; Oberdoerffer and Sinclair, 2007). Because of the NAD+ requirement for Sir2 deacetylase activity, it is evident that silencing is likely coupled to the metabolic cycle of cells. In C. elegans, one of the Sir2 orthologs, Sir2.1, has been shown to prevent aging (Tissenbaum and Guarente, 2001).
SIRT1, the mammalian ortholog of Sir2, is a nuclear protein that occupies a privileged position in the cell and governs critical metabolic and physiological processes. SIRT1 helps cells to be more resistant to oxidative or radiation-induced stress (Brunet et al., 2004; Luo et al., 2001), promotes mobilization of fat from white adipose tissues, an event that contributes to extending the life span (Picard et al., 2004), and mediates the metabolism of energy sources in metabolically active tissues (Lagouge et al., 2006; Rodgers et al., 2005). At the level of chromatin, SIRT1 enzymatic activity preferentially targets histone H3 at Lys9 and Lys14 and histone H4 at Lys16 (Imai et al., 2000). In addition, a number of nonhistone proteins, including p53 (Luo et al., 2001; Vaziri et al., 2001), FOXO3 (Brunet et al., 2004; Motta et al., 2004), PGC-1a (Nemoto et al., 2005; Rodgers et al., 2005), and LXR (Li et al., 2007a), are regulated by SIRT1-mediated deacetylation, stressing the pivotal function that this regulator plays in cellular control and responses.
A remarkable array of metabolic and physiological processes display daily oscillations (Panda et al., 2002; Storch et al., 2002; Ueda et al., 2002), and an intimate interplay exists between circadian clocks and metabolic rhythms in all organisms (Wijnen and Young, 2006). The discovery that a core element of the circadian clock machinery, the protein CLOCK, is an enzyme with HAT activity (Doi et al., 2006) revealed the crucial role that chromatin remodeling plays in the circadian regulation of gene expression (Hardin and Yu, 2006; Nakahata et al., 2007). More recently, the finding that CLOCK specifically also acetylates nonhistone targets, such as its own partner BMAL1, suggested that it may control a number of physiological cellular functions (Hirayama et al., 2007). The intrinsic nature of SIRT1 as a NAD+-dependent HDAC prompted us to explore the possibility that SIRT1 could participate in circadian control by regulating the HAT function of CLOCK. This would uncover a unique example of control of gene expression by metabolites (Ladurner, 2006).
Here we report that the HDAC activity of SIRT1 is regulated in a circadian manner in cultured cells and in the liver. SIRT1 physically associates with CLOCK and contributes to the acetylated state of CLOCK targets, such as Lys9/Lys14 in the tail of histone H3 and Lys537 in the BMAL1 protein. CLOCK, BMAL1, and SIRT1 colocalize in a chromatin-associated regulatory complex at promoters of clock-controlled genes. Pharmacological inhibition of SIRT1 activity by NAM and the drug splitomicin causes a loss in stringency of circadian gene expression, an effect equally observed in mouse embryo fibroblasts (MEFs) derived from Sirt1 null mice. Importantly, this effect is paralleled by a significant reduction in the oscillation of H3 and BMAL1 acetylation. Finally, using tissue-specific mutant mice, in which the Sirt1 gene is mutated uniquely in the liver, we demonstrate that SIRT1 contributes to circadian regulation in vivo. We propose that SIRT1 functions as an enzymatic rheostat of CLOCK function, thereby transducing signals originated by cellular metabolites to the circadian machinery.
Results
SIRT1 Deacetylase Activity Is Circadian
The CLOCK protein is one of the few core circadian regulators whose levels do not oscillate in most settings (Lee et al., 2001). Thus, we predicted that its HAT activity would oscillate in a circadian manner, thereby explaining the physiological remodeling of chromatin (Doi et al., 2006). An alternative scenario implicates a regulated HDAC, whose activity may function as rheostat of the HAT’s function of CLOCK.
To assess whether SIRT1 may be regulated in a circadian manner we prepared RNA and protein extracts from serum-stimulated cultured MEFs and from mouse liver at various Zeitgeber times (ZT). In both cases, the transcript and protein levels of SIRT1 remained constant, as determined using two anti-SIRT1-specific antibodies (Figure 1A; see also later, Figure 7B) and reverse transcription (RT)-PCR (Figures 1B and 1C, where dbp is shown as a control from the same RNA preparations). We have also determined the levels of SIRT1 in nuclear fractions prepared in various manners. Again SIRT1 protein levels showed either modest or no oscillation (Figure S1 available online). Thus, we turned to determining whether SIRT1 deacetylase activity may oscillate.