Sirtuins are NAD+-dependent enzymes that have been implicated in a wide range of cellular processes, including pathways that affect diabetes, cancer, lifespan and Parkinson's disease. To understand their cellular function in these age-related diseases, identification of sirtuin targets and their subcellular localization is paramount. SIRT3 (sirtuin 3), a human homologue of Sir2 (silent information regulator 2), has been genetically linked to lifespan in the elderly. However, the function and localization of this enzyme has been keenly debated. A number of reports have indicated that SIRT3, upon proteolytic cleavage in the mitochondria, is an active protein deacetylase against a number of mitochondrial targets. In stark contrast, some reports have suggested that full-length SIRT3 exhibits nuclear localization and histone deacetylase activity. Recently, a report comparing SIRT3−/− and SIRT+/+ mice have provided compelling evidence that endogenous SIRT3 is mitochondrial and appears to be responsible for the majority of protein deacetylation in this organelle. In this issue of the Biochemical Journal, Cooper et al. present additional results that address the mitochondrial and nuclear localization of SIRT3. Utilizing fluorescence microscopy and cellular fractionation studies, Cooper et al. have shown that SIRT3 localizes to the mitochondria and is absent in the nucleus. Thus this study provides additional evidence to establish SIRT3 as a proteolytically modified, mitochondrial deacetylase.
- sirtuin 3 (SIRT3)
Class III histone deacetylases, or sirtuins, are an evolutionarily conserved family of enzymes that utilize NAD+ as a substrate to deacetylate the ϵ-amino group of lysine from acetylated proteins, yielding two additional products, nicotinamide and O-acetyl-ADP-ribose . In yeast, Sir2 (silent information regulator 2) transcriptionally silences the mating-type loci, telomeric DNA regions and the ribosomal RNA locus by maintaining histones in the hypoacetylated state . Strikingly, gene-dosage of Sir2 in yeast, flies and worms appears to mediate lifespan extension similar to the effects of caloric restriction, whereas the lack of Sir2 opposes this effect . Additionally, sirtuins have been implicated in a number of cellular processes, including cell survival, cell-cycle regulation and genomic stability, fatty acid synthesis, and glucose and insulin homoeostasis .
The seven members of the mammalian Sirtuin (SIRT1–7) family have various substrates and are compartmentalized to distinct cellular locations. The most studied sirtuin, SIRT1, is both nuclear and cytosolic, and has been demonstrated to deacetylate Foxo1, p53, NF-κβ (nuclear factor κβ), Ku70, PGC-1α (peroxisome-proliferator-activated receptor γ co-activator-1α) and AceCS1 (cytoplasmic acetyl-CoA synthetase 1) . SIRT2 deacetylates α-tubulin in the cytoplasm, whereas SIRT6 and SIRT7 mediate DNA repair and rRNA transcription respectively in the nucleus . In the mitochondria, SIRT4 ADP-ribosylates and inhibits GDH (glutamate dehydrogenase), whereas the functional activities of SIRT5 remain unclear .
The localization and the cellular role of SIRT3 has been the subject of some controversy. Initial studies had suggested that SIRT3 resides in the mitochondria and functions as an NAD+-dependent deacetylase [4,5]. Remarkably, a recent report suggested that upwards of one-fifth of detected mitochondrial proteins are acetylated . Given that SIRT3 is the only reported mitochondrial deacetylase with robust activity, this suggests that SIRT3 is probably the main mitochondrial deacetylase and may, in part, control energy flux via post-translational modification of metabolic proteins . However, several contradictory reports have suggested that SIRT3 resides in the nucleus and exerts epigenetic control by deacetylating histone substrates [8,9].
One source of debate arises from inconsistent reports of post-translational processing of the human and mouse SIRT3 homologues. Initially, hSIRT3 (human SIRT3) was reportedly expressed as a 44 kDa protein and targeted to the mitochondria via an N-terminal localization sequence . Upon entry into the mitochondrial matrix, hSIRT3 is processed at the N-terminus by a mitochondrial matrix peptidase into the activated 29 kDa enzyme . These conclusions are corroborated by the fact that truncations, as well as mutations in the N-terminal mitochondrial localization sequence, prevent the protein from proper mitochondrial localization, as well as efficient proteolytic cleavage . In contrast, mSIRT3 (murine SIRT3) was suggested to lack the N-terminal mitochondrial localization sequence, and instead, is translated beginning at the conserved deacetylase domain [9–11]. Curiously, the same cDNA-encoded mSIRT3 has been shown to be associated with the mitochondrial inner membrane . These studies raise the question of whether human and mice utilize different mechanisms to translocate SIRT3 into the mitochondria. A genomic observation may afford a resolution to this discrepancy. In this issue of the Biochemical Journal, Cooper et al.  present sequence analysis that indicates the initiator methionine of mSIRT3 exists further upstream in the mSIRT3 genomic region, resulting in a protein sequence that encodes a conserved mitochondrial targeting signal similar to hSIRT3. However, experimental evidence for the functionality of this differentially processed version of mouse SIRT3 has yet to be revealed.
Mitochondrial localization of SIRT3 has been demonstrated in three separate fluorescence microscopy studies utilizing a C-terminal GFP (green fluorescent protein)-fusion expression construct [4,5,13]. In addition, cellular fractionation experiments in which nuclear, cytosolic and mitochondrial components were separated by differential centrifugation and probed for SIRT3 with C-terminal specific antibodies demonstrated that SIRT3 is enriched in the mitochondria . Cooper et al.  provide additional support for the lack of nuclear SIRT3 by utilizing stringent nuclear fractionation techniques. Although these studies utilize SIRT3 overexpression, several additional lines of evidence support endogenous SIRT3 mitochondrial residence. Endogenous SIRT3 has been detected in the mitochondria of HEK-293 cells, and decreased protein levels were observed by siRNA (small interfering RNA) knockdown . Additionally, endogenous SIRT3 was clearly distinguishable from Myc-tagged SIRT3 as a faster migrating band in Western blotting experiments using fractionated mitochondria . Collectively, this evidence supports SIRT3 mitochondrial localization regardless of the expression level in cells. A study comparing mSIRT3+/+ and mSIRT3−/− mice has provided the strongest evidence of endogenous mSIRT3 mitochondrial localization; careful isolation and subcellular fractionation of mSIRT3+/+ liver lysates demonstrated that SIRT3 is localized exclusively to the mitochondria, and is absent from the nucleus and cytoplasm . Furthermore, SIRT3−/− mice demonstrate global hypoacetylation of mitochondrial proteins, and adding recombinant SIRT3 to mitochondrial extracts from SIRT3−/− animals reduces protein acetylation. These combined studies strongly support the conclusion that SIRT3 localizes to the mitochondria and acts as the predominant NAD+-dependent deacetylase.
In dramatic contrast, two studies have suggested that full-length SIRT3 is localized to the nucleus. One study proposes that SIRT3 is localized to the nucleus upon co-expression of SIRT5 . Until the mechanism of how SIRT5 influences the biological activities of SIRT3 is demonstrated, the caveats of forced SIRT3 and SIRT5 co-overexpression remain. A separate study reported that SIRT3 localizes to the nucleus and translocates to the mitochondria only upon cellular stress, such as overexpression or an oxidative challenge [8,9]. Although an intriguing possibility, the antibody employed in these studies raises questions as to whether all the endogenous SIRT3 was detected. Because the antibody was raised against the N-terminal domain of SIRT3, the same domain that is proteolytically clipped upon mitochondrial translocation, the use of this antibody in immunofluorescence studies would make it improbable for detection of mitochondrial SIRT3. Seemingly contradictory, the same study reported that an antibody raised against the C-terminal domain of SIRT3 clearly identifies endogenous SIRT3 in the mitochondria fraction of 293F cells . By measuring SIRT3 levels with antibodies against portions of SIRT3 that are not proteolytically processed, Cooper et al.  demonstrated internally consistent data that discerns endogenous and overexpressed SIRT3.
Lastly, SIRT3 has been identified as a direct NAD+-dependent deacetylase for the substrates AceCS2 and GDH [7,15,16]. In vitro assays demonstrated that full-length SIRT3 lacked measurable NAD+-dependent deacetylation, but upon cleavage of the mitochondrial localization sequence, gained deacetylase activity towards histone peptide [4,5]. When amino acids 1–119 were removed, recombinantly expressed SIRT3 displayed high catalytic activity against histone H4 peptide as well as acetylated AceCS2 in vitro . This contrasts with a separate study where unprocessed full-length SIRT3 could exchange nicotinamide in the presence of acetylated histone peptide and NAD+ , an assay that has been used to demonstrate catalytic activity of sirtuins. Although nicotinamide exchange indicates that SIRT3 is mechanistically competent for NAD-cleavage, the caveat of this assay is that it does not measure the full deacetylation reaction. Therefore it is problematic to conclude that full-length SIRT3 is a functioning protein deacetylase.
In this issue of Biochemical Journal, Cooper et al.  re-examine the localization of hSIRT3 via fluorescence localization and subcellular fractionation experiments, as well as providing insight into the genomic nature of mSIRT3. In doing so, additional compelling evidence supports the conclusion that endogenous SIRT3 resides in the mitochondria as a highly active protein deacetylase. Alternative cellular localization of SIRT3 cannot be excluded, and further investigations will be necessary to provide convincing evidence that SIRT3 resides elsewhere under ‘stress’. Among the critical questions to be answered, it is not yet clear how SIRT3 expression activates the nuclear transcription of mitochondria-related genes such as those coding for UCP1 (uncoupling protein 1), PGC-1α and COX (cytochrome oxidase) IV and V . As one possibility, SIRT3 may deacetylate proteins involved in metabolic pathways or signalling cascades that emanate from the mitochondria and signal to the nucleus. These signals could exist as deacetylated proteins or metabolites, such as O-acetyl-ADP-ribose or nicotinamide.
- © The Authors Journal compilation © 2008 Biochemical Society