SIRT3 Functions in the Nucleus in the Control of Stress-Related Gene Expression

Abstract

SIRT3 is a member of the Sir2 family of NAD⁺-dependent protein deacetylases that promotes longevity in many organisms. The processed short form of SIRT3 is a well-established mitochondrial protein whose deacetylase activity regulates various metabolic processes. However, the presence of full-length (FL) SIRT3 in the nucleus and its functional importance remain controversial. Our previous studies demonstrated that nuclear FL SIRT3 functions as a histone deacetylase and is transcriptionally repressive when artificially recruited to a reporter gene. Here, we report that nuclear FL SIRT3 is subjected to rapid degradation under conditions of cellular stress, including oxidative stress and UV irradiation, whereas the mitochondrial processed form is unaffected. FL SIRT3 degradation is mediated by the ubiquitin-proteasome pathway, at least partially through the ubiquitin protein ligase (E3) activity of SKP2. Finally, we show by chromatin immunoprecipitation that some target genes of nuclear SIRT3 are derepressed upon degradation of SIRT3 caused by stress stimuli. Thus, SIRT3 exhibits a previously unappreciated role in the nucleus, modulating the expression of some stress-related and nuclear-encoded mitochondrial genes.

INTRODUCTION

SIRT3 is a member of the sirtuin family of proteins that functions as an NAD⁺-dependent protein deacetylase that targets histone and nonhistone proteins (13, 50). SIRT3 has garnered substantial interest given its apparent role in promoting longevity, a phenomenon associated with caloric restriction, in organisms that span the spectrum from yeast to humans (4, 47). A focus on SIRT3 function has led to a fuller understanding of its role in modulating the adaptation of mitochondria to conditions of low energy input, again reflective of caloric restriction and fasting. Although many studies have argued for an exclusively mitochondrial function of SIRT3 (37, 43), we (41) and others (33, 46) have shown that FL SIRT3 localizes to and functions in the nucleus (see below).

Recent studies have identified SIRT3 targets for deacetylation and demonstrated its role in promoting the production of energy sources that detour from glucose utilization. These SIRT3 targets include acetyl coenzyme A (CoA) synthetase 2 (AceCS2), involved in acetate recycling (18, 42), long-chain acyl-CoA dehydrogenase (LCAD), involved in fatty acid oxidation (20), and ornithine transcarbamoylase (OTC), which increases the metabolic flow of the urea cycle (19). In all cases, the activities of these enzymes are increased by direct SIRT3-mediated deacetylation. These findings highlight the key role of mitochondrial SIRT3 in regearing mitochondrial usage of its alternate fuel capacities, as well as in minimizing the resultant increase in the levels of reactive oxygen species (ROS).

SIRT3 has also been identified as a functional tumor suppressor (23), in large part by targeting proteins that decrease ROS levels that arise as a consequence of increased fatty acid oxidation. SIRT3 overexpression indirectly results in the increased expression of the nuclear genes encoding manganese superoxide dismutase (MnSOD) and catalase through its deacetylation of the transcription factor forkhead box O 3A (FOXO3A), resulting in FOXO3A retention in the nucleus (45). SIRT3 also directly targets superoxide dismutase 2 (SOD2) to promote its antioxidative activity (6, 39). In keeping with its tumor-suppressing role, the absence of SIRT3 correlates with increased glycolysis, a hallmark of tumor cell proliferation. In this regard, a recent study demonstrated that the elevated levels of ROS associated with SIRT3 depletion lead to the stabilization of the hypoxia-inducible factor 1, alpha subunit (HIF-1α), transcription factor that activates the expression of several glycolytic enzymes (11). Consistent with the findings discussed above, these features of SIRT3 function, including its role as a tumor suppressor, are considered to be consequences of its mitochondrial milieu.

Earlier studies have indicated that human SIRT3 exists in a full-length (FL) form that is processed to a distinct short form that localizes specifically to the mitochondria (43). Also, several reports have indicated that it is only the short form of human SIRT3 that is active as an NAD⁺-dependent deacetylase (43). Yet, our previous studies have demonstrated that FL SIRT3 exists in the nucleus (41), and this was later confirmed by others (33, 46). Our findings, using rigorous biochemical assays, demonstrated that both forms of SIRT3 exhibit deacetylase activities that require NAD⁺. Consistent with a role for SIRT3 in the nucleus, we found that SIRT3 is also capable of histone deacetylase (HDAC) activity against acetylated histone H3 lysine 9 (H3K9ac) and H4K16ac in vivo. Such HDAC activity is, in general, believed to be associated with transcriptional repression. Indeed, when artificially recruited to a transgenic reporter, the FL form of SIRT3 gave rise to transcriptional repression that was dependent upon its HDAC activity (41).

We demonstrated that the processing and translocation of SIRT3 to mitochondria is evident upon conditions of cellular stress, including exposure to DNA-damaging agents and even overexpression of the SIRT3 protein itself (41). A subsequent study detected FL murine SIRT3 in the nucleus and cytoplasm of cardiomyocytes, while the short form was detected exclusively in the mitochondria (46). Notably, in that study, the authors found that only the long form of SIRT3 exhibited changes in its levels in a stress-responsive manner. Consistent with a nuclear role for SIRT3, the study showed that SIRT3 functions in a manner similar to SIRT1 to prevent cell death under stress conditions. Both SIRT1 and SIRT3 deacetylate Ku70, which, in turn, sequesters the proapoptotic BAX protein in the nucleus, therefore preventing its translocation to the mitochondria (7, 8, 46).

Based on our previous findings that FL SIRT3 is localized to the nucleus, that FL SIRT3 functions as a specific histone deacetylase and facilitates transcriptional repression when tethered to an artificial promoter, and that SIRT3 is subjected to stress-related processing, we analyzed further the functional role of SIRT3 in the nuclei of human cells. Here, we demonstrate that under conditions of cellular stress, including the presence of DNA-damaging agents, rapid degradation of SIRT3 results, but this degradation is specific to the FL form; the processed (short) form is unaffected. Remarkably, this degradation of nuclear FL SIRT3 results in the derepression of nuclear genes that encode the proteins involved in sensing metabolic stress.

MATERIALS AND METHODS

Plasmids and antibodies.

FLAG-tagged SIRT1, -2, and -3 have been described previously (41, 48, 49), and SIRT4, -5, -6, and -7 were gifts from E. Verdin (UCSF) (34). SIRT3 expression plasmids containing a hemagglutinin (HA) tag at the C terminus and site-directed mutations and deletions of SIRT3 were generated by PCR. The SIRT3 cDNAs were subcloned into the retroviral vector pCX4(IRES)-bsr (a gift from T. Akagi, KAN Institute, Japan) for viral infection. The FLAG-tagged SKP2 construct was kindly provided by M. Pagano (HHMI/NYU Medical Center). The following antibodies and chemicals were purchased: anti-SIRT3 polyclonal antibody (P-19; Santa Cruz Biotechnology, Inc.), anti-SIRT3 monoclonal antibody (for Western blotting after immunoprecipitation [IP]) (C73E3; Cell Signaling Technology, Inc.), anti-HA polyclonal antibody (Sigma), anti-FLAG M2 monoclonal antibody (Sigma), anti-K48-linked polyubiquitin antibody (Genentech), anti-SKP2 monoclonal antibody (Santa Cruz), anti-acetyl-histone H4K16 polyclonal antibody (Millipore), anti-histone H3 polyclonal antibody (ab1791; Abcam), methyl methane sulfonate, sodium arsenite, and MG132 (Sigma), and PYR-41 (UBEI-41; Biogenova).

RNA interference.

To construct the retroviral vectors for the expression of short hairpin RNA (shRNA), oligonucleotides encoding human SIRT3 (5′-gatccccGGTGGAAGAAGGTCCATATCTTgtgtgctgtccAAGATATGGACCTTCTTCCACCtttttggaaa-3′ and 5′-agcttttccaaaaaGGTGGAAGAAGGTCCATATCTTggacagcacacAAGATATGGACCTTCTTCCACCggg-3′) were chemically synthesized. The SIRT3 target sequences, indicated by uppercase letters in the oligonucleotide sequences above, were annealed and ligated into pSUPER.retro (OligoEngine, WA), which was digested with BglII and HindIII. For SKP2 knockdown, double-stranded small interfering RNA (siRNA) (5′-AAGGGAGUGACAAAGACUUUGdTdT-3′) was purchased from Invitrogen and transfected using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions.

Cell culture and retroviral expression.

293T, HeLa, and U2OS cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum. Skp2^+/+ and Skp2^−/− mouse embryonic fibroblasts (MEFs) were kindly provided by K. Nakayama (Tohoku University, Japan). The primary MEFs were immortalized by expressing simian virus 40 (SV40) large T antigen and maintained in a manner similar that for the other cell lines. Ecotropic retroviruses were produced by transient transfection of Platinum-E (Plat-E) cells (31) with the viral vectors. HeLa and U2OS cells expressing the murine ecotropic receptor were infected with retroviruses as described previously (22).

Immunoprecipitation and immunoblotting.

Cells were lysed in ice-cold 1% NP-40 lysis buffer (22). The lysates were immunoprecipitated with the indicated antibodies for 2 h at 4°C and subsequently incubated with protein G-Sepharose (Roche) for 1 h. The precipitated protein complexes were washed three times with lysis buffer without NP-40 and then suspended in Laemmli SDS-PAGE sample buffer. For immunoblotting, the samples were subjected to SDS-PAGE and transferred to a nitrocellulose membrane as described previously (22).

Ubiquitin conjugates in cells.

293T cells were transiently transfected with a cDNA encoding HA-tagged ubiquitin or ubiquitin K48R along with the indicated cDNAs. The cells were lysed in lysis buffer (100 mM Tris-HCl [pH 7.5] and 1% SDS) and boiled quickly for 10 min. The lysates were diluted 10-fold with radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris [pH 7.4] containing 1% NP-40, 0.1% SDS, 1% sodium deoxycholate, 150 mM NaCl, 5 mM EDTA, 50 mM NaF, 2 mM Na₃VO₄, 10 μg/ml each of leupeptin and aprotinin, and 1 mM phenylmethylsulfonyl fluoride [PMSF]), immunoprecipitated with the indicated antibodies for 2 h at 4°C, and subsequently incubated with protein G-Sepharose (Roche) for 1 h. The precipitated proteins were washed three times with RIPA buffer without sodium deoxycholate and SDS, resolved by SDS-PAGE, and visualized after Western blotting using the indicated antibodies.

Subcellular fractionation.

Subcellular protein fractions were prepared using the NE-PER nuclear and cytoplasmic extraction kit (Pierce) according to the manufacturer's protocol.

Chromatin immunoprecipitation and ChIP-seq.

Chromatin immunoprecipitation (ChIP) experiments were performed essentially as described previously (29), except that Dynabeads Protein G (Invitrogen) was used for ChIP sequencing (ChIP-seq) instead of protein G-Sepharose blocked by salmon sperm DNA. ChIP-seq libraries were prepared according to Illumina protocols and sequenced using an Illumina genome analyzer. Sequenced reads (36 bp) from each ChIP-seq experiment were aligned to the human reference genome (assembly hg19) using Bowtie (25) with the -v2 and -m1 parameters to exclude reads that aligned to more than one site in the reference sequence. Duplicate reads were removed with SAMtools (26). Statistically enriched regions (SERs) for each ChIP-seq data set were identified with Qeseq (2) using a 2.5 enrichment cutoff compared to that of the input. The gene targets associated with SIRT3 SERs were obtained with the Genomic Regions Enrichment of Annotations Tool (GREAT) (30) using the default parameters. The ChIP-quantitative PCR (qPCR) primer pairs used were as follows: for hPTK2, 5′-CTTTTCCCCCATCCTTCCCCTCTCC-3′ (forward [Fwd]) and 5′-CCTGCCCTTAGGCAGTCATAGTGCT-3′ (reverse [Rev]); for hGAL3ST1, 5′-ACTGGGCGATTCTGGAATGGGGAAA-3′ (Fwd) and 5′-GACAACTCAGGCCTGTCTGGGGAAC-3′ (Rev); for hOTOG, 5′-TACCCATCCCTTGGGATCCGCTTTC-3′ (Fwd) and 5′-TGCATGGTTCCTGGGTTCTGCTGTA-3′ (Rev); for hZFAT, 5′-CCAATGTGCTAAAACAGAGGGCCACA-3′ (Fwd) and 5′-TGCTTCAGACTCAGGTGGCTTTCCA-3′ (Rev); for hBAZ2A, 5′-AGGCATATGAGGACACCAACATCCT-3′ (Fwd) and 5′-TCACCACTGAAGAGGTCAAGCACCA-3′ (Rev); for hBRIP1, 5′-TCAGAATGTTCTTCTTTGCCTTAACCACA-3′ (Fwd) and 5′-GGAGAATATTTTCTGGGTTCACTTGGAAGA-3′ (Rev); for hTSHZ3, 5′-AGAGGGGTGAGACAACAGCAACGAT-3′ (Fwd) and 5′-CCCTTGGTAGTCAAAGACATTGCCA-3′ (Rev); for hCORIN, 5′-GCTCTAGAGAGTTTTCTGGGGAGCTA-3′ (Fwd) and 5′-GGAGGTAGTGAAGTGCCAACATAAGC-3′ (Rev); for hWAPAL, 5′-CACCACACAAGGGAAACTTGGGTCC-3′ (Fwd) and 5′-AGGTGGCTTCTTAGGATCTGGTTCA-3′ (Rev); and for hGAPDH, 5′-CAATTCCCCATCTCAGTCGT-3′ (Fwd) and 5′-TAGTAGCCGGGCCCTACTTT-3′ (Rev).

RNA isolation, reverse transcription, and qPCR analyses.

Total RNA was isolated with TRIzol reagent (Invitrogen). The RNA was reverse transcribed with SuperScript III (Invitrogen) with oligo(dT) primers. qPCR was performed with SYBR green-based detection on a Roche light cycler. The primer sequences were as follows: for hBRIP1, 5′-AACTCCTTTTCGCCACAGAAACCCC-3′ (Fwd) and 5′-TGGAACCCCTGAATATGCCGTCCTC-3′ (Rev); for hZFAT, 5′-AAGTCAAGCAGGCTAGGTCCCACTC-3′ (Fwd) and 5′-ATAGCTGCACTGGGGGCACTTGTAT-3′ (Rev); for hMSRB2, 5′-GAAAAGGGAACGGAACCGCCTTTCA-3′ (Fwd) and 5′-GACGTACCATGAGCCTCGGAAAACG-3′ (Rev); for hTSHZ3, 5′-CGCAGCAGCCTATGTTTCCGAAGAG-3′ (Fwd) and 5′-TCACTGATGTGTGACTCGCTGTCCA-3′ (Rev); for hBAZ2A, 5′-TCACACAGGGACATCACCCTACAGC-3′ (Fwd) and 5′-TCCATTTCCATGTCAGGCTGGCACA-3′ (Rev); for hBOLA3, 5′-CCACTTCACCATCGGATGTTTGCCA-3′ (Fwd) and 5′-CATCGCCCCACAACCTCCTGAAATG-3′ (Rev); for hGAL3ST1, 5′-CCTGCTGCTGGTGTACTCCTATGCC-3′ (Fwd) and 5′-TCTTCAAGAACACGATGTTGCGCCG-3′ (Rev); for hCORIN, 5′-AGTCTGACGAGGTCAACTGCTCCTG-3′ (Fwd) and 5′-ATCACTCCCATCCTTGCAGTCCTCG-3′ (Rev); for hGNPNAT1, 5′-CGCAGGGCCTCTACGGACCTTACTA-3′ (Fwd) and 5′-CAGGATGTGTTGGGGAAATGGCTGG-3′ (Rev); for hOTOG, 5′-TGCTGCAGGACCTGTAAGGAGGATG-3′ (Fwd) and 5′-ATCGCAGGACACTAGGTTCACAGGG-3′ (Rev); for hPTK2, 5′-TCGGCTTGGCCCTGAGGACATTATT-3′ (Fwd) and 5′-TCTTGCTGGAGGCTGGTCATGACAT-3′ (Rev); for hWAPAL, 5′-TAGTGCTCGGAATCGGCACTGTCTT-3′ (Fwd) and 5′-GCTCTCGCTCAAGGAATAGCTGCAC-3′ (Rev); for hATF4, 5′-CAACAACAGCAAGGAGGATG-3′ (Fwd) and 5′-ATCCAACGTGGTCAGAAGGT-3′ (Rev); for hCOX6B1, 5′-GCTGAAGCCGCTCGCAAGACT-3′ (Fwd) and 5′-GGCTGTCAAAAGGGGCGGTCT-3′ (Rev); for hCOX5B, 5′-TGGCATCTGGAGGTGGTGTT-3′ (Fwd) and 5′-CTTGTTGGAGATGGAGGGGA-3′ (Rev); for hSIRT3, 5′-TGGAAAGCCTAGTGGAGCTTCTGGG-3′ (Fwd) and 5′-TGGGGGCAGCCATCATCCTATTTGT-3′ (Rev); for hHSPA1A, 5′-TGAGCACAAGAGGAAGGAGCTGGAG-3′ (Fwd) and 5′-AAACAGCAATCTTGGAAAGGCCCCT-3′ (Rev); for hDDIT3, 5′-TCACCTCCTGGAAATGAAGAGGAAGAA-3′ (Fwd) and 5′-GTGCTTGTGACCTCTGCTGGTTCTG-3′ (Rev); and for hGAPDH, 5′-GATGACATCAAGAAGGTGGTGAA-3′ (Fwd) and 5′-GTCTTACTCCTTGGAGGCCATGT-3′ (Rev).

Sequencing data accession number.

The sequencing data sets described in this study have been deposited at the National Center for Biotechnology Information Gene Expression Omnibus under accession no. GSE41000.

RESULTS

Stress-induced SIRT3 degradation.

Given our previous findings (41) and those of others (46) demonstrating that cellular stress results in the modulation of SIRT3 protein levels, we began by examining the stability of the SIRT3 protein upon cycloheximide treatment of 293T cells (Fig. 1A, left panel). The specificities of the anti-SIRT3 antibodies used for Western blotting were monitored by samples of SIRT3 knockdown cells (SirT3-siRNA). Endogenous FL SIRT3 (FL) appeared to be quite unstable, exhibiting a half-life of ∼30 min. In addition, treatment with the DNA-damaging agent methyl methanesulfonate (MMS), the oxidative stressor sodium arsenite (NaAsO₂), or UV irradiation led to strong decreases in the levels of FL SIRT3 but not of its processed form (short) or of another sirtuin, SIRT1 (Fig. 1A, middle and right panels, and B).

Fig 1.

Stress-mediated degradation of SIRT3. (A) Left, turnover kinetics of SIRT3. 293T cells were incubated with 50 μg/ml cycloheximide (CHX) for the indicated times. Degradations of the endogenous full-length (FL) and short (Short) forms of SIRT3 were monitored by Western blotting using anti-SIRT3. As a negative control for endogenous SIRT3 signals, the lysate of 293T cells expressing shRNA for SIRT3 (SirT3-siRNA) was loaded. The asterisks indicate nonspecific (ns) bands. Middle and right, 293T cells were treated with 0.01% methyl methanesulfonate (MMS) and 50 μM sodium arsenite (NaAsO₂) for 2 h and UV irradiated (400 J/m²) 2 h prior to harvesting. The cell lysates were analyzed by immunoblotting with anti-SIRT3 and anti-SIRT1 (top right). (B) Stress agents induce degradation of FL SIRT3. Left, 293T cells were treated with 0.01% MMS and 50 μM sodium arsenite (NaAsO₂) for 2 h and lysed with RIPA buffer. Endogenous SIRT3 proteins were concentrated by immunoprecipitation with anti-SIRT3 antibody from equivalent amounts of lysates and blotted with anti-SIRT3. Right, 293T cells were pretreated with 50 μM PYR-41 for 1 h and then treated with 50 μM NaAsO₂. The cells were further incubated for 2 h in the presence of 50 μM PYR-41. Endogenous SIRT3 was immunoprecipitated and detected by Western blotting. (C) Schematic diagrams of FL-SIRT3-HA and the processed mitochondrial form are shown at the top. MTS, mitochondrial targeting signal; DMSO, dimethyl sulfoxide. Left, HeLa cells stably expressing SIRT3-HA were treated with 0.01% MMS and 50 μM NaAsO₂ for 2 h and UV irradiated (400 J/m²) 2 h prior to harvesting. The cell lysates were analyzed by immunoblotting with anti-HA (top left) and antitubulin (bottom left). Right, densitometric quantification of full-length and short forms of the SIRT3-HA protein displayed in the left panel to determine the expression levels of SIRT3 in response to stress.

Similar results were obtained in the case of cells stably expressing C-terminal HA-tagged SIRT3 from a heterologous promoter, suggesting that this effect was posttranscriptional (Fig. 1C). A range of UV doses also gave rise to dosage-dependent decreases in the levels of FL SIRT3 (Fig. 2A). The decrease was evident within 30 min of UV treatment in the case of 70 J/m² and persisted for at least 6 h (Fig. 2B). Given that the levels of SIRT3 mRNA did not decrease upon UV irradiation (Fig. 2D), the observed decrease in the FL SIRT3 protein must have occurred at the posttranslational level. This stress-induced degradation profile was different in the case of NaAsO₂ treatment, in that the reduction in the SIRT3 level appeared to be more transient, declining within 2 h posttreatment and recovering by 4 h (Fig. 2C). These results suggest that FL SIRT3 is subject to degradation, perhaps through differing pathways depending upon the type of stress treatment applied. Indeed, endogenous FL SIRT3 levels were unperturbed after cells were treated with NaAsO₂ in conjunction with PYR-41, an inhibitor specific for the sole ubiquitin-activating enzyme (E1) in humans (51) (Fig. 1B, right panel).

Fig 2.

(A) SIRT3 protein levels as a function of UV dosage. HeLa (left) and U2OS (right) cells stably expressing SIRT3-HA were irradiated with the indicated UV doses 2 h prior to harvesting. (B) SIRT3 degradation kinetics in response to UV irradiation. Left, U2OS cells stably expressing SIRT3-HA were UV irradiated (70 J/m²) as a function of time, as indicated, and SIRT3-HA levels were examined by Western blotting. Right, densitometric quantification of FL and short forms of SIRT3-HA protein represented in the left panel. (C) As described for panel B but using 50 μM NaAsO₂ treatment in lieu of UV irradiation. (D) mRNA levels of SIRT3 after UV irradiation. As described for panel B, but mRNA levels of SIRT3 were analyzed by RT-qPCR. The values shown are the ratios of mRNA levels after irradiation relative to those at zero time. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Stress-induced SIRT3 degradation by the ubiquitin-proteasome system.

That PYR-41 barred cells from exhibiting NaAsO₂-induced decreases in the levels of SIRT3 suggests that the ubiquitin-proteasome system targets SIRT3 for degradation. To ascertain if SIRT3 can be modified by polyubiquitination, we expressed FLAG-tagged versions of all sirtuins (SIRT1 to SIRT7) independently along with a vector encoding HA-tagged ubiquitin in 293T cells. The FLAG-tagged sirtuin candidates were immunoprecipitated with anti-FLAG antibody and analyzed by Western blotting using anti-HA antibody (Fig. 3A). Both SIRT3 and SIRT4 were highly polyubiquitinated. The polyubiquitination of SIRT3 was greatly reduced, however, when cotransfection was performed with an HA-tagged ubiquitin mutant containing a single point mutation on a lysine at position 48; this indicates that SIRT3 was modified mainly by the K48-linked polyubiquitin chain, the target of proteasome-mediated degradation (Fig. 3B). Consistent with this, MG132 treatment of cells transfected with HA-tagged ubiquitin led to the accumulation of polyubiquitinated endogenous SIRT3 (Fig. 3C).

Fig 3.

Stress-induced SIRT3 degradation is dependent on the ubiquitin-proteasome system. (A) Top, schematic diagram of N-terminal FLAG-tagged SIRT3. Middle and bottom, SIRT3 is polyubiquitinated. 293T cells were transiently transfected with the indicated FLAG-tagged sirtuins together with HA-tagged ubiquitin. FLAG-tagged sirtuins were immunoprecipitated (IP) with anti-FLAG, followed by immunoblotting with anti-HA (middle) and reblotting with anti-FLAG (bottom). (B) The ubiquitin K48R (Ub-K48R) mutant partially blocked SIRT3 ubiquitination. FLAG-tagged SIRT3 was immunoprecipitated with anti-SIRT3 from the lysates of 293T cells expressing the indicated combinations of FLAG-tagged SIRT3 and HA-tagged ubiquitin constructs, followed by immunoblotting with anti-HA (top) and reblotting with anti-SIRT3 (bottom). WT, wild type. (C) Polyubiquitination of endogenous SIRT3. 293T cells were transfected with HA-tagged ubiquitin and treated with 2.5 μM MG132 for 4 h. Endogenous SIRT3 was immunoprecipitated with anti-SIRT3 and blotted with anti-HA (top), anti-K48-linked polyubiquitin chain (middle), and anti-SIRT3 (bottom). (D) UV- and NaAsO₂-induced SIRT3 loss was blocked by E1 or proteasome inhibitors. HeLa cells stably expressing SIRT3-HA were pretreated with the indicated dosage of PYR-41, 5 μM MG132, or DMSO for 4 h (left) and then UV irradiated (100 J/m²) (middle) or treated with 50 μM NaAsO₂ (right). The cells were further incubated for 2 h in the presence of the drugs. The cell lysates were analyzed by immunoblotting with anti-HA and antitubulin.

To determine if the ubiquitin-proteasome system is required for SIRT3 degradation, we treated cells that stably expressed HA-tagged SIRT3 with PYR-41 and the proteasome inhibitor MG132. Indeed, PYR-41 treatment gave rise to elevated basal levels of FL SIRT3, similar to treatment with MG132 (Fig. 3D, left panel). Moreover, PYR-41 treatment abrogated FL SIRT3 degradation, similar to the case with MG132 after exposure to UV irradiation or NaAsO₂ (Fig. 3D, middle and right panels, respectively).

F-box protein SKP2 mediates SIRT3 degradation.

The ubiquitination of proteins for proteasome-mediated degradation occurs through a conserved pathway involving a ubiquitin-activating enzyme (E1), a ubiquitin-conjugating enzyme (E2), and a ubiquitin protein ligase (E3) (12). The latter provides specificity through the direct interaction with the targeted protein. The SCF complex comprising cullin 1, SKP1, Rbx/Roc1, and an F-box protein is a major class of E3, and the SKP2 F-box protein is recognized as an SCF ubiquitin ligase. Among its targets, SKP2 functions in the ubiquitination of several tumor suppressor proteins (12). Given that SIRT3 is a tumor suppressor (23), we first tested if SKP2 associates with SIRT3; we found that endogenous SKP2 coimmunoprecipitated with endogenous SIRT3 (Fig. 4A). Also, SIRT3-HA coprecipitated with FLAG-tagged SKP2 from extracts derived from cells expressing both proteins, and importantly, only the FL form of SIRT3 exhibited this interaction (Fig. 4B). Moreover, cellular exposure to MG132 enhanced this association between SIRT3 and SKP2, suggesting that SKP2 might be involved in the degradation of FL SIRT3. This possibility was bolstered when the half-life of FL SIRT3 was found to be considerably lengthened in the case of Skp2^−/−, relative to Skp2^+/+, MEFs (Fig. 4C). Since an overexpressed SIRT3 deletion mutant lacking the N-terminal 142 amino acids, which was not targeted to the mitochondria, retained its ability to interact with SKP2 (data not shown), we believe that the lack of interaction between SKP2 and the short form of SIRT3 (Fig. 4B) is due to compartmentalization: the short form of SIRT3 is found exclusively in the mitochondria, where there is no SKP2.

Fig 4.

SKP2 is involved in SIRT3 degradation. (A) Interaction of endogenous SIRT3 and SKP2. 293T cell lysates were subjected to IP with anti-SIRT3 and control IgG, followed by immunoblotting (IB) for endogenous SKP2. (B) Full-length SIRT3 interacts with SKP2 in vivo. 293T cells were cotransfected with vector expressing SIRT3-HA and either a vector expressing FLAG-tagged SKP2 or an empty vector. Twenty-four hours posttransfection, the cells were treated for 2 h with MG132 before harvesting, as indicated. Whole-cell lysates (WCL) were subjected to immunoprecipitation (IP) with anti-FLAG and immunoblotting for the indicated proteins. (C) SIRT3 is stabilized in Skp2^−/− MEFs. Top, human SIRT3-HA was stably expressed in Skp2^+/+ and Skp2^−/− MEFs, and its half-life was analyzed by cycloheximide (CHX) treatment. Bottom, densitometric quantification of full-length SIRT3-HA protein from three independent experiments represented in the top panel and normalized by tubulin levels. (D) NaAsO₂-induced SIRT3 degradation was prevented in Skp2^−/− MEFs. The same MEFs as in panel C were treated with 50 μM NaAsO₂ for 2 h or were UV irradiated (100 J/m²) 1 or 2 h prior to harvesting, as indicated. The cell lysates were analyzed by immunoblotting with anti-HA and antitubulin. (E) NaAsO₂-induced degradation kinetics of SIRT3 in Skp2^−/− MEFs. Left, the same MEFs as in panel C were treated with 50 μM NaAsO₂ as a function of time, as indicated, and SIRT3-HA levels were examined by Western blotting. Right, densitometric quantification of FL-SIRT3-HA proteins was performed as described for panel C. (F) NaAsO₂-induced SIRT3 degradation is prevented by Skp2 knockdown. 293T cells were transfected with a control siRNA oligonucleotide or an siRNA oligonucleotide directed against Skp2 mRNA. The cells were then treated with 50 μM NaAsO₂ for 2 h as indicated. The degradation of endogenous SIRT3 was monitored by Western blotting using anti-SIRT3.

To test the repercussions of a deficiency in SKP2 on stress-induced SIRT3 degradation, we compared the levels of SIRT3 in wild-type versus Skp2^−/− MEFs as a function of UV or NaAsO₂ exposure (Fig. 4D). While SKP2 appeared to be dispensable in the case of UV treatment, its absence in the case of NaAsO₂ treatment protected FL SIRT3 from degradation. Indeed, the transfection of SKP2 siRNA oligonucleotide gave rise to increased levels of endogenous FL SIRT3 and blocked the NaAsO₂-induced degradation (Fig. 4F). NaAsO₂-induced degradation kinetics revealed that SKP2 abrogation blocked the degradation only partially and temporarily, given that protein levels returned to pretreatment values in 2 h (Fig. 4E). These findings show that SKP2 regulates the levels of FL SIRT3 under basal conditions and, at least in part, during NaAsO₂-induced stress. On the other hand, the targeting of SIRT3 for degradation during UV exposure occurs in a SKP2-independent manner.

SIRT3 degradation in the nuclei.

Although full-length SIRT3 is markedly degraded under stress conditions, the levels of the mitochondrial short form are unaffected (Fig. 1), suggesting that SIRT3 degradation occurs outside of the mitochondria. To test whether SIRT3 degradation depends on its mitochondrial localization, we generated a deletion mutant (D25SIRT3) that lacks the mitochondrial import signal (43). Consistent with this previous report, when the mutant was stably expressed in HeLa cells, the mitochondrial short form was undetectable (Fig. 5A). Yet, NaAsO₂ treatment and UV irradiation induced the degradation of D25SIRT3 in a proteasome-dependent manner (Fig. 5A). Thus, the mitochondrial targeting of SIRT3 is not required for its stress-dependent degradation.

Fig 5.

SIRT3 degradation in the nuclei. (A) Mitochondrial targeting is not required for SIRT3 degradation. HeLa cells stably expressing D25SIRT3-HA, a mutant lacking the mitochondrial target signal (schematic diagrams, top) and pretreated with 5 μM MG132 for 2 h, were then treated with 50 μM NaAsO₂ or UV irradiated with 400 J/m². The cells were further incubated for 2 h in the presence of 5 μM MG132. The cell lysates were analyzed by immunoblotting with anti-HA (top) and antitubulin (bottom). (B) Degradation of full-length SIRT3 in nuclei. HeLa cells stably expressing SIRT3-HA were pretreated with 5 μM MG132 for 2 h and then UV irradiated (400 J/m²). The cells were further incubated for 2 h in the presence of 5 μM MG132. The nuclear (Nucl.) and nonnuclear (Cyto.) fractions of the cells were prepared, and equivalent amounts of protein were analyzed for SIRT3 expression levels by Western blotting using anti-HA (top). The fractions were characterized by the presence of the mitochondrial marker MnSOD (middle) and the nuclear marker histone H1 (bottom). (C) Degradation of endogenous SIRT3 in nuclei. U2OS and 293T cells were UV irradiated (400 J/m²) 2 h prior to harvesting. The nuclear and nonnuclear fractions of U2OS and 293T cells were prepared, and 25 μg (U2OS) or 35 μg (293T) of protein was analyzed for endogenous SIRT3 expression levels by Western blotting using anti-SIRT3. The fractions were characterized by the presence of the mitochondrial marker MnSOD and the nuclear marker histone H3. As a negative control for endogenous SIRT3 signals, fractionated proteins of U2OS or 293T cells expressing shRNA for SIRT3 (SirT3-siRNA) were loaded. Two asterisks indicate nonspecific bands detected in nuclear fractions.

To determine whether the nuclear form of FL SIRT3 is subject to UV-induced degradation, we separated the nuclei of HeLa cells expressing HA-tagged SIRT3 (SIRT3-HA) from cytoplasmic proteins and analyzed the two fractions for the presence of SIRT3-HA. As reported previously (46), the nuclear fraction contained almost exclusively the FL-SIRT3-HA, while both FL-SIRT3-HA and its mitochondrial short form were detected in the cytoplasmic fraction (Fig. 5B, top left panel). Upon UV irradiation, not only the cytoplasmic but also the nuclear SIRT3 was degraded, and this degradation was completely blocked by MG132 treatment (Fig. 5B). Appropriate fractionation was verified by Western blotting for mitochondrial (MnSOD) and nuclear (histone H1) markers (Fig. 5B, bottom panels). Because the nuclear localization of SIRT3 has been questioned (3, 9, 17, 27), we sought to confirm the existence of endogenous SIRT3 in the nucleus and its stress-dependent degradation. Consistent with the results shown in Fig. 5B, endogenous FL SIRT3 was present in the nuclei of both U2OS and 293T cells, and its levels decreased upon UV irradiation (Fig. 5C). The identity of the SIRT3 band in these Western blots was confirmed by SIRT3 knockdown (SirT3-siRNA) in both cell lines. These results support the conclusions that FL SIRT3 is the predominant form of SIRT3 in the nucleus and that it is subject to degradation in response to stress.

SIRT3 localizes to defined chromatin regions and regulates gene expression.

Given that artificial tethering of SIRT3 to the promoter region of a reporter gene leads to silencing (41), and that endogenous FL SIRT3 is present in the nucleus, we sought to determine whether SIRT3 could regulate gene expression by interacting with chromatin. To identify potential genomic targets for SIRT3, we took an unbiased approach and determined the localization of SIRT3 on chromatin genome-wide using chromatin immunoprecipitation (ChIP) coupled to high-throughput sequencing (ChIP-seq). To maximize the chances of detecting SIRT3 targets on chromatin, we cross-linked U2OS cells using both a short-range (formaldehyde) and long-range [ethylene glycol bis(succinimidylsuccinate) (EGS)] cross-linker. Because our findings above show that UV irradiation results in the degradation of FL SIRT3 in the nucleus (Fig. 5C), we included this condition and SIRT3 knockdown as specificity controls, along with the standard IgG control for ChIP experiments.

Utilizing previously described bioinformatic analyses and statistical thresholds (15), we detected more than 150 genomic regions that were occupied by SIRT3 (enriched regions [ERs]) (Fig. 6A to D). These same regions were not enriched in the ChIP experiments, in which we used a nonspecific IgG control (Fig. 6E), suggesting that they are bona fide genomic targets of SIRT3. Moreover, the SIRT3 antibody did not detect any enrichment at these regions when the levels of endogenous SIRT3 were suppressed either by RNA interference (Fig. 6F) or by treating the cells with UV irradiation (Fig. 6G); this, as we show above, results in the degradation of FL SIRT3 in the nucleus. These controls demonstrate that the SIRT3 ERs are indicative of specific coimmunoprecipitation with SIRT3 rather than of nonspecific DNA binding to the antibodies or to the beads. We visually inspected the ChIP-seq intensity profiles in the vicinities of the transcription start sites (TSS), and from among the genes neighboring the SIRT3 ERs, we selected a number of candidates encoding stress-related, metabolism-related, and mitochondrial proteins based on the established link between SIRT3, the stress response, and the functional relevance of SIRT3 to mitochondrial processes.

Fig 6.

Genome-wide chromatin localization of SIRT3. (A to D) ChIP-seq signal profile for endogenous SIRT3 in untreated (NT), UV-irradiated (UV), and SIRT3 knockdown (KD) U2OS cells. The read densities were calculated on tiling windows of 25 bp and are expressed as the normalized number per 10 million reads sequenced. The University of California at Santa Cruz (UCSC) gene annotation is shown at the bottom of each panel. The scale bar at the top of each panel corresponds to 2 kb. The gray boxes indicate the position of the ERs identified by Qeseq. (E to G) Combined read density profiles for SIRT3 and IgG in untreated U2OS cells (E) or SIRT3 before and after SIRT3 knockdown (F) or UV irradiation (G). Normalized (per 10 million reads) ChIP-seq signal densities were calculated in 10-kb windows spanning the summits of 192 SIRT3 ERs identified in the untreated sample by Qeseq. To control for changes in chromatin accessibility, the signal from the input chromatin was subtracted.

To see the effect of inhibition of nonmitochondrial SIRT3 on mRNA levels of these candidate genes, a catalytically inactive point mutant (41) of nonmitochondrial D25SIRT3 (HY) was stably overexpressed in U2OS cells. Consistent with our previous results showing the gene-silencing activity of SIRT3 using artificial recruitment of the wild-type enzyme (41), several of the stress-related and mitochondrion-related nuclear genes analyzed were upregulated upon overexpression of the dominant negative D25HY form of SIRT3 (Fig. 7A). In addition, SIRT3 knockdown (∼50%) also resulted in the transcriptional upregulation of 8 of the 11 genes analyzed (Fig. 7B). Manual ChIP-qPCR confirmed the presence of endogenous SIRT3 at the ERs identified by ChIP-seq near the affected genes (Fig. 7C). Importantly, SIRT3 knockdown and UV irradiation decreased the amount of SIRT3 bound to these ERs, as predicted by our biochemical data that show extensive UV-induced degradation of SIRT3 (Fig. 5B and C) and by genome-wide analysis (Fig. 6E to G). Together, our results indicate that nuclear SIRT3 binds to genomic targets on chromatin and that its recruitment results in the repression of neighboring genes, which depends on the enzymatic activity of SIRT3.

Fig 7.

Transcriptional deregulation of genes bound by SIRT3 upon its inhibition. (A) Overexpression of catalytically inactive D25SIRT3 upregulates SIRT3 target gene expression. The graph depicts qRT-PCR analysis of putative SIRT3 target genes in U2OS cells, expressing catalytically inactive D25SIRT3 (HY). U2OS cells were infected with a retrovirus expressing D25HY or the control virus. After drug selection, steady-state mRNA levels of the indicated genes were analyzed by real-time qPCR. The values shown are the fold induction of the indicated genes relative to mock-infected cells from three independent biological experiments. The P values are based on Student's t test. Two asterisks indicate a P value of <0.01, and one asterisk indicates a P value of <0.05 compared to results from mock-infected cells. (B) SIRT3 knockdown upregulates SIRT3 target gene expression. U2OS cells were infected with a retrovirus expressing shRNA for SIRT3 (KD) or a mock sequence, and the steady-state mRNA levels of the indicated genes were analyzed by real-time qPCR. The values and P values are as described for panel A. (C) SIRT3 binding to chromatin loci. ChIP-qPCR of endogenous SIRT3 in untreated (SIRT3 none), SIRT3 knockdown (SIRT3 KD), and UV-irradiated (SIRT3 UV) U2OS cells. Enrichment for the SIRT3-bound regions was analyzed by qPCR. The values shown are the percentages of input from three independent ChIPs. The P values are based on Student's t test. Two asterisks indicate a P value of <0.01, and one asterisk indicates a P value of <0.05 compared with results from IgG-ChIP.

Our results show that SIRT3 binds closely to the TSS of the PTK2 (52), GAL3ST1, and OTOG genes (Fig. 6 and 7C), which probably exert their effects on cell growth and metabolic pathways. To further assess the biological importance of stress-induced SIRT3 degradation, we tested the effect of nonmitochondrial D25SIRT3 overexpression on stress-induced gene expression in the case of PTK2, GAL3ST1, and OTOG. These three genes were induced by NaAsO₂ treatment of U2OS cells, as evidenced by our real-time (RT)-qPCR results (Fig. 8A, “mock” lines, and middle panels). Strikingly, D25SIRT3 overexpression inhibited their induction at 2 to 4 h after As (NaAsO₂) treatment (Fig. 8A, upper panels, D25). Although D25SIRT3 is also degraded under conditions of cellular stress (Fig. 5A), its levels are higher than those of endogenous FL SIRT3, because D25SIRT3 is incapable of translocating to the mitochondria and accumulates in the nucleus even when cells are treated with NaAsO₂ (Fig. 8B), thus effectively rescuing nuclear SIRT3 levels from the effects of cellular stress. In contrast, D25SIRT3 overexpression did not affect the NaAsO₂-induced expression of genes that were not targets of SIRT3 (as evidenced by a lack of ChIP-seq enrichment), suggesting that the effect of SIRT3 at its target genes is due to its specific recruitment to chromatin (Fig. 8A, lower panels). However, the D25SIRT3 inhibitory effect was dependent on SIRT3 enzymatic activity, since overexpression of the catalytically inactive D25HY mutant did not suppress the induction of the SIRT3 target genes (Fig. 8A, middle panels). Together, these results show that increased amounts of nuclear SIRT3 inhibit the rapid transcriptional activation of its target genes, PTK2, GAL3ST1, and OTOG, induced by stress signals. NaAsO₂-induced degradation of endogenous SIRT3 appears to constitute an additional layer of regulation that contributes to the rapid induction of stress-responsive genes.

Fig 8.

Overexpression of D25SIRT3 deregulates stress-induced gene expressions. (A) D25SIRT3 overexpression inhibits NaAsO₂-induced gene expression. The graphs depict qRT-PCR analysis of stress-induced gene expression in control versus D25SIRT3 (D25)-expressing U2OS cells (top and bottom panels) and the catalytically inactive D25SIRT3 (D25HY)-expressing U2OS cells (middle panels). U2OS cells were infected with a retrovirus expressing D25, D25HY, or a control vector (mock). The cells were then treated with 50 μM NaAsO₂ (As) for the times (hours) indicated. The mRNA levels of the indicated genes were analyzed by real-time qPCR. The values shown represent the induction ratios (log₂) relative to mRNA levels at zero time in each case. One asterisk indicates a P value of <0.05, and the pound sign indicates a P value of <0.1 compared to results from mock cells. (B) Expression levels of D25SIRT3 in NaAsO₂-treated U2OS cells. Mock and D25SIRT3 (D25)-expressing U2OS cells as described for panel A were treated with 50 μM NaAsO₂ (As) for 2 h and lysed with RIPA buffer. Endogenous SIRT3 and D25 proteins were immunoprecipitated with anti-SIRT3 from equivalent amounts of lysates and blotted with anti-SIRT3. (C) SIRT3 binding correlates with deacetylation of H4K16 at promoter regions of PTK2, GAL3ST1, and OTOG genes. The bar graphs depict ChIP-qPCR of H4K16 acetylation at SIRT3 binding regions in untreated U2OS cells and mock-infected (As/mock), D25SIRT3-expressing (As/D25), and D25HY-expressing (As/D25HY) NaAsO₂-treated U2OS cells. The results are normalized to unmodified H3-ChIP data.

To confirm this conclusion, we analyzed the chromatin state of the three SIRT3 ERs near PTK2, GAL3ST1, and OTOG before and after NaAsO₂ treatment by ChIP (Fig. 8C). The levels of acetylated H4K16 at these SIRT3 ERs increased upon NaAsO₂ treatment (Fig. 8C, As/mock), in keeping with the degradation of SIRT3 and the transcriptional activation of these three genes. The overexpression of D25SIRT3 thwarted the NaAsO₂-induced increases in H4K16ac levels and gene activation observed in the endogenous case through its catalytic activity (Fig. 8C and A, respectively). Together, these results suggest that the degradation of nuclear FL SIRT3 removes it from its chromatin targets, thus relieving its repressive effect on target genes, some of which have key functions in the stress response.

DISCUSSION

Given that the bulk of cellular SIRT3 is found in the mitochondria as a processed short form and that FL SIRT3 was thought to be enzymatically inactive, previous studies focused on the 28-kDa short form, mainly in the context of mitochondrial functions. However, we reported previously that FL SIRT3 is enzymatically active, both in vitro and in vivo (41), suggesting that SIRT3 may exhibit a functional role outside mitochondria through its deacetylase activity. In support of this hypothesis, SIRT3 regulates the expression of nuclear genes, such as PGC-1α and MnSOD, and directly modulates the activity of the FOXO3A transcription factor by deacetylation (23, 44, 45).

Based on our previous results demonstrating that unprocessed FL SIRT3 is localized in the nucleus and suggesting that SIRT3 might be involved in gene silencing through its HDAC activity (41), we sought to determine the function of FL SIRT3 in the nucleus and its biochemical regulation. Here, we show that stress stimuli lead to the rapid degradation of FL SIRT3, resulting in the upregulation of SIRT3 target genes.

Nuclear localization of FL SIRT3.

Our observation that FL SIRT3 resides in the nucleus (41) has been challenged by studies in humans (9) and mice (3, 17, 27) that utilized cellular fractionations followed by Western blotting. However, in two cases, only the short form of SIRT3 was observed; this suggests that the detection method was not sensitive enough, as the FL form is much less abundant (17, 27). Cooper and Spelbrink prepared nuclear fractions from HEK293 cells overexpressing human SIRT3 and did not detect the protein in the nucleus (9). However, using a similar method, Bao et al. detected the exogenous mouse form of SIRT3 in the nucleus, although they reported that the endogenous version was undetectable in MEF cells (3). Yet another study reported both mouse and human FL SIRT3 in the nuclear fractions of mouse heart and of HeLa cells (46). Because we clearly detected both exogenous and endogenous FL SIRT3 in the nuclear fractions of different cell lines (Fig. 5B and C), the discrepancy is likely due to different sensitivities and specificities of the antibodies used for detection. Importantly, we verified the specificity of the antibody by siRNA-mediated knockdown.

Ubiquitination and degradation of SIRT3.

As shown here, SIRT3 degradation is dependent on the ubiquitin-proteasome pathway, with the protein being tagged in vivo with a polyubiquitin chain. It should be noted that the stress stimuli led to degradation even in the case of the stably expressed D25SIRT3 protein lacking its N-terminal mitochondrial localization signal; this shows that mitochondrial localization of SIRT3 is not required for its stress-dependent degradation. This result is consistent with the observation that only FL SIRT3 was degraded rapidly by stress agents and suggests that the polyubiquitination and degradation of SIRT3 take place outside the mitochondria. However, the possible involvement of the ubiquitin-proteasome pathway in regulating the mitochondrial short form of SIRT3 cannot be ruled out at this point, because NaAsO₂ led to a decline in the levels of the short form 4 h after treatment, at which time levels of the long form started to recover (Fig. 2C). We found that an F-box protein, SKP2, associates with FL SIRT3 and regulates the degradation induced by NaAsO₂ treatment. As SKP2 recognizes its substrates, such as p27 and FOXO1, through their phosphorylation (12), oxidative stress may induce phosphorylation of FL SIRT3 to provide SKP2 binding sites. Interestingly, SIRT3 degradation after UV irradiation was found to be SKP2 independent. These observations suggest that the physiological turnover of SIRT3 depends on SKP2 but that additional SKP2-independent mechanisms are activated upon UV irradiation.

Protein ubiquitination is not only the marker for proteasomal degradation. The specific type of ubiquitination that a substrate undergoes determines its cellular fate. Unlike K48-linked polyubiquitin chains, which are usually targeted for degradation by the proteasome, the monoubiquitination of proteins is involved in various cellular processes, including subcellular trafficking (21). For example, the MDM2-mediated monoubiquitination of tumor protein 53 (p53) promotes its stress-induced translocation to the mitochondrion and stimulates apoptosis, while MDM2-mediated polyubiquitination regulates p53 stability (28). Interestingly, blocking the ubiquitination and proteasome activity not only increased the basal levels of full-length SIRT3 but also decreased the levels of its mitochondrial short form (Fig. 3D). This result suggests that ubiquitination and proteasome activity may also be involved in SIRT3 translocation to the mitochondria or proteolytic processing in the mitochondria.

SIRT3 function at target genes.

Using ChIP, we have shown that SIRT3 binds to genetic loci encoding several stress-related proteins and mitochondrial proteins (Fig. 6). The overexpression of nonmitochondrial dominant negative D25HY and SIRT3 knockdown led to the upregulation of these targets (Fig. 7A and B). In addition, nonmitochondrial D25SIRT3 overexpression inhibited stress-induced mRNA expression of the targets (Fig. 8A). The overexpression of D25SIRT3 may overwhelm the ubiquitination-proteasome pathway such that a fraction of this protein escapes degradation (Fig. 8B). Nonetheless, the target gene repression observed here is consistent with our previous report that SIRT3 recruitment to a reporter results in its silencing (41). Among the identified SIRT3 target genes, ZFAT and WAPAL were reported to play roles in antiapoptotic function and oncogenesis, respectively (10, 14, 36, 40). Our observation that SIRT3 controls the expression of these genes (Fig. 7A and B) is consistent with recent studies showing proapoptotic and tumor-suppressing functions of SIRT3 in human cancer cells and genetically manipulated mice (1, 23, 53).

Since sirtuins are dependent on NAD⁺ for enzymatic activity, SIRT3-mediated gene repression is probably regulated by changes in metabolic state and NAD⁺/NADH levels during normal cell conditions, and SIRT3 functions in the mitochondria leading to metabolic changes. Given that SIRT3 targets several genes involved in stress response, DNA damage, lipid metabolism, and mitochondrial function, SIRT3 may regulate intracellular energy status both at the transcriptional level in the nucleus and by posttranscriptional mechanisms in mitochondria. Furthermore, our data suggest that the stress-induced degradation of nuclear SIRT3 would cause induction of the PTK2, GAL3ST1, and OTOG genes that are bound by SIRT3 under normal conditions. The effects of SIRT3 degradation on individual target genes may be also determined by other factors, like chromatin-modifying enzymes and transcriptional factors, whose activities are also modulated by different types of stress.

Given the recent studies showing that Ku70, FOXO3A, and LKB1 are also substrates of SIRT3 outside the mitochondria (38, 45, 46), FL SIRT3 degradation may be crucial for stress-induced signaling pathways in addition to transcriptional regulation. In response to stress, FOXO3A is hyperacetylated and stimulates apoptosis (5, 32). Although the deacetylation of FOXO3A by SIRT1 and SIRT3 is important for cell survival (5, 32, 45), the significance of the initial acetylation of FOXO3A is not fully understood, given that stress-induced acetylation of the FOXO protein seems to be required to target FOXO to transcriptionally active nuclear domains (24). Rapid and transient degradation of the SIRT3 deacetylase might be involved in the rapid acetylation of FOXO that would be required for the initial response to stress.

Our findings suggest a biological role for the stress-dependent degradation of SIRT3: its rapid removal from chromatin targets after stress exposure is needed for the rapid induction of genes required for the stress response. Because SIRT3 binding to chromatin is correlated with the repression of neighboring genes, as well as with the deacetylation of H4K16 (Fig. 8C), our data suggest that its loss leads to the disruption of a silencing machinery that maintains chromatin in an inactive deacetylated state, similar to that of SIRT1 (35). However, SIRT1 is not degraded in response to DNA damage but instead dissociates from its chromatin targets, resulting in transcriptional derepression, and relocalizes to DNA breakpoints to promote repair (35). Given these and our results, we propose that the dissociation of sirtuins may be a critical step in changing gene expression profiles in response to stress stimuli.

In light of the studies about sirtuins so far, their main function may be to promote survival and stress resistance in times of adversity (16). In this context, our finding that the levels of SIRT3 and its binding profile to chromatin can be rapidly modulated by stress signals suggests that its regulation may be an essential component of general stress resistance. Further understanding of the effects of SIRT3 degradation on gene expression and signal transduction may help unveil critical mechanisms that regulate stress response and longevity.