The yeast Sir2 (silent information regulator-2) protein functions as an NAD+-dependent histone deacetylase to silence gene expression from the mating-type locus, tolomeres and rDNA and also promotes longevity and genome stability in response to calorie restriction. Homologues of yeast Sir2 have been identified in the three domains of bacteria, archaea and eukaryotes; in mammalian cells, Sir2 proteins also deacetylate non-histone proteins such as the p53 tumour suppressor protein, α-tubulin and forkhead transcription factors to mediate diverse biological processes including metabolism, cell motility and cancer. We have determined the X-ray crystal structure of a Sir2 homologue from yeast Hst2 (yHst2), in various liganded forms, including the yHst2/acetyl-Lys-16 histone H4/NAD+ ternary complex; we have also performed related biochemical studies to address the conserved mode of catalysis by these enzymes as well as the distinguishing features that allow different members of the family to target their respective cognate substrates. These studies have implications for the structure-based design of Sir2-specific small molecule compounds, which might modulate Sir2 function for therapeutic application.
The enzymes that mediate post-translational modification of the N-terminal tails of the histone proteins, which package DNA into chromatin, play key roles in the regulation of gene expression [1–3]. These include histone acetyltransferases  and deacetylases [5,6], whose activities are correlated with gene activation and gene repression or silencing respectively. Since the transmission of genetic information is at the heart of controlling many biological processes including metabolism, cell cycle and cell differentiation, it has come as no surprise that histone modification enzymes play key roles in normal life processes and are aberrantly regulated in several human diseases such as cancer [7–9]. This is particularly true for the deacetylation enzymes.
The ySir2 [yeast Sir2 (silent information regulator-2)] protein deacetylates histones H3 and H4 to silence gene expression from the mating-type locus, tolomeres and rDNA and suppresses genomic instability [10,11]. ySir2 is the prototype of the Class III family of HDAC (histone deacetylase) enzymes that, unlike the Class I and II HDACs, requires the cofactor NAD+ for catalytic activity . Enzymic studies have demonstrated that ySir2 converts NAD+ and acetyl-lysine histone substrates into 2′-O-acetyl-ADP-ribose, nicotinamide and deacetylated histone [13,14]. Sir2 proteins are conserved in the three domains of bacteria, archaea and eukaryotes, and while bacteria and archaea typically contain one or two Sir2 proteins, eukaryotic organisms contain multiple members. For example, budding yeast and humans contain five (Sir2 and Hst1–4) and seven (SirT1–SirT7) Sir2 homologues respectively. Analysis of Sir2 proteins from various species indicates that their protein targets extend beyond histones and that they mediate diverse biological functions . For example, of the seven known human Sir2 proteins, nuclear SIRT1 targets the p53 tumour suppressor protein for deacetylation to suppress the apoptotic programme in response to DNA damage [16,17] and forkhead transcription factors [18,19] to regulate its transcriptional activity; and the cytoplasmic SIRT2 homologue targets α-tubulin for deacetylation to maintain cell integrity .
In both yeast and worms, increased dosage of Sir2 extends lifespan; in yeast, both Sir2 and NAD+ are required for the long-established link between calorie-restricted diets and longevity in many organisms. In addition, since it is well documented that, as cells age, they are more prone to genomic instability, a hallmark of cancer, Sir2 proteins have been implicated as important targets for chemotherapeutic agents. Indeed, a recent study to identify Sir2-activating compounds characterized a family of polyphenol compounds, several of which are currently being used as chemotherapeutic agents . Interestingly, the most potent of these compounds is resveratrol, a natural plant product that is in high abundance in red wine and correlated with increased lifespan and decreased cancer risk in humans .
To understand the detailed mechanism of Sir2 activity, in our laboratory, we determined the structure of a ySir2 homologue, Hst2, in full-length nascent form  as well as the structure of the catalytic domain in various liganded forms [23,24]. Together with associated biochemical studies, we have addressed the following questions: (i) what is the conserved mode of catalysis by Sir2 enzymes? (ii) how do non-conserved regions of the Sir2 enzymes mediate functions that distinguish the different homologues? and (iii) how do different Sir2 proteins target their respective cognate substrates? Implications of these studies for the structure-based design of Sir2-specific small molecule compounds that might modulate Sir2 function for therapeutic application are discussed.
Overall structure of the Sir2 proteins
A sequence alignment of Sir2 proteins reveals that they contain a highly conserved catalytic core domain of approx. 270 residues and more variable N- and C-terminal extensions that are believed to play more protein-specific functions . The structure of the Sir2 catalytic core adopts an elongated shape containing a large and structurally homologous Rossmann fold domain characteristic of NAD+/NADH-binding proteins and a small, structurally more variable domain containing a structural zinc ion [25–29] (Figure 1). A series of loops traverse between the large and small domains, forming a pronounced extended cleft between the two protein domains. The ternary yHst2 complex shows that the two substrates enter the protein through opposite sides of a cleft between the small and large domains of the catalytic core, and the functional groups of both the protein and substrates are buried within a protein tunnel that harbours the region of highest conservation within the Sir2 proteins [23,24] (Figure 1).
Overall structure of the Sir2 protein, Hst2, in ternary complex with acetyl-Lys-16 histone H4 and carba-NAD+
Catalysis by Sir2 proteins
We have obtained ternary complex structures of yHst2, acetyl-Lys-16 histone H4 peptide and one of three NAD+ analogues along the reaction pathway [23,24]: (i) a substrate analogue complex containing carba-NAD+, a non-hydrolysable NAD+ analogue in which a cyclopentane ring replaces the furanose of the nicotinamide-ribonucleotide moiety; an intermediate analogue complex containing ADP-ribose, a close mimic of a reaction intermediate, formed after cleavage of the glycosidic bond between nicotinamide and ADP-ribose (it differs only by the addition of a 1′-OH group); and a product analogue complex containing 2′-O-acetyl-ADP-ribose, the NAD+-derivatized product of the Sir2 reaction. Comparison of these structures reveals that, although nearly 90% of the protein remains structurally invariant, the ribose ring of the cofactor and the highly conserved β1–α2 loop of the protein undergo significant structural rearrangements to facilitate the ordered NAD+ reactions of nicotinamide cleavage and ADP-ribose transfer to acetate (Figure 2).
Proposed catalytic mechanism for Sir2 proteins
The structure of the substrate mimic complex (with carba-NAD+) shows that the carbonyl oxygen of the acetyl-lysine forms hydrogen bonds with the 2′- and 3′-hydroxy groups of the nicotinamide ribose ring (Figure 2). This appears to position what would be the ribose ring oxygen of NAD+ within hydrogen-bonding distance of a water-mediated contact to the side-chain carbonyl of Asn-116. This places the asparagine side-chain carbonyl in position to stabilize the oxocarbenium that has been proposed to form after hydrolysis of the nicotinamide group [13,14,30] and is consistent with solution studies showing that this asparagine is essential for the nicotinamide exchange by Sir2 proteins . The structure of the intermediate analogue complex (with ADP-ribose) shows that the ribose ring of the intermediate complex is rotated by approx. 45° along the ring plane relative to the ribose ring of the substrate analogue complex, to a position that now permits nucleophilic attack of the 1′-carbon of the ribose ring by the carbonyl oxygen of acetyl-lysine. The product analogue complex (with 2′-O-acetyl-ADP-ribose) reveals that the nicotinamide ribose ring is finally rotated by approx. 90° along the ring plane relative to the ribose ring of the substrate analogue complex, to a position that now allows His-135, a residue essential for the deacetylation reaction, to deprotonate the 3′-hydroxy group of the nicotinamide ribose, nucleating the formation of a cyclic acyldioxalane involving the 1′- and 2′-oxygens of the ribose ring. We then propose that a crystallographically well-ordered water molecule, which is held in place by Asn-116, performs the nucleophilic attack of the cyclic acyldioxalane, resulting in the collapse of the cyclic intermediate to the 2′-O-ADP-ribose and lysine reaction products (Figure 2).
Comparison of the three ternary yHst2/NAD+ analogue complexes reveals that the β1–α2 loop also plays important roles in catalysis [23,24]. Specifically, the β1–α2 loop mediates important NAD+ interactions involving several conserved loop residues (Ala-33, Gly-34, Thr-37 and Phe-44). The relatively open conformation of this loop in the product and intermediate analogue complexes also suggests that the loop conformation also facilitates NAD+ substrate access. The more closed conformation of the β1–α2 loop in the product analogue complex suggests that the β1–α2 loop may also play a role in nicotinamide release, since Phe-44 of the β1–α2 loop of the product analogue complex partially occupies the binding site for nicotinamide, and it is also possible that the further burial of the active site further facilitates the acetyl-transfer reaction. Taken together, the nicotinamide ribose ring of the NAD+ and the β1–α2 loop of the Sir2 protein appear to play dynamic roles in NAD+ association and protein catalysis.
Roles of regions N- and C-terminal to the Sir2 catalytic domain
The structure, determined in our laboratory, of the full-length yHst2 protein reveals that regions N- and C-terminal to the catalytic core domain play autoregulatory roles  (Figure 3). Specifically, a C-terminal helix overlaps with the NAD+-binding site and, thus, autoregulates NAD+ binding. In addition, an N-terminal strand sits in the acetyl-lysine-binding site of a symmetry-related subunit in the crystal lattice, mediating the formation of a homotrimer and, thus, autoregulating acetyl-lysine binding. Solution studies in our laboratory correlate with these findings . Specifically, yHst2 constructs in which regions C-terminal to the catalytic core domain have been deleted bind NAD+ approx. 3.5-fold more strongly. Moreover, whereas the full-length yHst2 protein forms a homotrimer in solution with dissociation constant in the low-micromolar range, an N-terminal deletion construct does not form detectable trimers and binds acetyl-lysine substrates 2–3-fold more avidly.
Structure of the autoregulated Hst2 homotrimer
Whether or not this type of enzyme autoregulation occurs with other Sir2 proteins is unknown; however, the sequence divergence within regions N- and C-terminal to the catalytic core domain of the Sir2 proteins suggests that these interactions may differ in other Sir2 proteins, possibly correlating with the substrate-specific roles of different Sir2 proteins.
Substrate-binding specificity by the Sir2 proteins
To date, there is no structure available of a Sir2 protein bound to its cognate acetyl-lysine-containing target, although the ternary yHst2 structures determined in our laboratory [23,24] and the Af2-Sir2/p53 peptide complex reported by Wolberger and co-workers  does provide some insights into substrate recognition by Sir2 proteins (Figure 4). In both structures, the aliphatic arm of the acetyl-lysine residue makes extensive van der Waals interactions with several hydrophobic residues that are highly conserved within the Sir2 proteins. The backbone amino group of acetyl-Lys-16 also makes β-sheet interactions with Sir2-conserved backbone residues of the protein. Taken together, these studies suggest that the observed interactions with acetyl-lysine and the backbone of flanking residues are a conserved feature of the Sir2 proteins.
Interactions between the Sir2 protein, Hst2 and acetyl-Lys-16 histone H4 peptide
Comparison of the Sir2/peptide complexes reveals that, outside of acetyl-lysine and the flanking backbone residues, protein–peptide interactions are very limited and are not conserved between the different complexes. To test the hypothesis that the determinants for substrate-specific binding by Sir2 proteins are mediated, at least in part, by substrate sequences distal to the acetyl-lysine-binding site, the binding properties of cognate and non-cognate peptide substrates to the bacterial Sir2 homologue, CobB, were quantified using isothermal titration calorimetry . Escalante-Semerena and co-workers  had previously shown that CobB deacetylates Lys-609 of Acs (acetyl-CoA synthetase) in vivo to stimulate its enzymic activity . As a cognate peptide, we employed 11- and 15-residue peptides centred on acetyl-Lys-609 of Acs; as non-cognate substrates, we used 11-residue peptides centred on acetyl-lysine targets of other Sir2 proteins, a histone H4 peptide centred on acetyl-Lys-16  and a p53 peptide centred on acetyl-lysine Lys-382 . Analysis of the results reveals that the binding of each of these peptides is exothermic, indicative of hydrogen-bond formation after complexation. However, and quite surprisingly, CobB shows very little discrimination between these substrates, binding each with a dissociation constant between 0.44 and 3.7 μM, with the weakest binding constant for the cognate Acs peptide. Analogous binding studies employing intact Acs proteins specifically acetylated at Lys-609 yield a dissociation constant of 14 μM, but, surprisingly, show an endothermic binding reaction indicative of an entropy-dominant contribution to binding involving a burial of hydrophobic surface and/or structural rearrangement involving CobB, Acs or both proteins. Taken together, these results support the conclusion that substrate specificity determinants of CobB, and probably other Sir2 proteins, derive from regions outside the sequence local to the acetyl-lysine substrate.
Because of the possible association of Sir2 enzymes with human disease and life extension, there is much interest in the design of small molecule compounds that might regulate the activity of Sir2 enzymes [10,15]. Several inhibitors to Sir2 enzymes have been reported, including sirtinol , splitomicin  and nicotinamide , a product of the Sir2 reaction. A shortcoming of these inhibitors, however, is that they have relatively weak binding constants, with IC50 or Km values in the mid-micromolar range. Our structural studies of yHst2 bound to NAD+ analogues suggests that a more fruitful avenue for inhibitor design focus on a mimic of the reaction intermediate, such as ADP-ribose. Indeed, binding studies reveal that ADP-ribose binds to yHst2 with a dissociation constant of 0.4 μM, approx. 100-fold more avidly compared with previously characterized Sir2 inhibitors .
The possibility of generating Sir2-activating compounds that may extend life or promote genomic stability has also generated considerable interest . Our structural studies  suggest a rational approach for developing Sir2-activating compounds that takes advantage of the observation that nicotinamide, a product of the Sir2 reaction, functions as a non-competitive inhibitor of Sir2 by reacting with an ADP-ribosyl-enzyme-acetyl peptide intermediate with regeneration of NAD+ (transglycosidation) [30,36] (Figure 2). This mode of nicotinamide inhibition implies that nicotinamide binds to Sir2 at a site distinct from the nicotinamide group of NAD+, but still in a geometry that promotes the regeneration of NAD+. Interestingly, if we assume that the conformation of ADP-ribose in the intermediate analogue complex mimics the conformation of the NAD+ intermediate that is formed immediately after nicotinamide cleavage, then it is particularly striking that an entering nicotinamide group can be modelled on to the 1′-carbon of the ADP-ribose ring with suitable geometry, without stereochemical clash with the protein, and in a binding site that is distinct from the binding site of the nicotinamide group of NAD+. This suggests that compounds that block the binding site for free nicotinamide, as suggested from our structural modelling, might serve as potent activators of Sir2 with potential therapeutic applications.
Taking together with our structural and biochemical studies employing the yHst2 and CobB Sir2 proteins [23,31], it appears that substrate specificity by Sir2 proteins derives from regions outside the sequence local to the acetyl-lysine substrate. In addition, structural analysis of the intact yHst2 proteins further suggests that regions N- and C-terminal to the conserved catalytic core domain probably also contribute to the protein-specific functions of the Sir2 proteins . Understanding the mechanistic basis for substrate discrimination by Sir2 proteins will require structural analysis of other intact Sir2 proteins as well as their complexes with intact cognate protein substrates. Ultimately, understanding what makes Sir2 proteins different might be at the heart of designing Sir2-specific regulatory compounds with therapeutic value.
Genes: Regulation, Processing and Interference: A Focus Topic at BioScience2004, held at SECC Glasgow, U.K., 18–22 July 2004. Edited by I. McEwan (Aberdeen, U.K.), B. White (Glasgow, U.K.), S. Graham (Glasgow, U.K.), S. Roberts (Manchester, U.K.), A. Sharrocks (Manchester, U.K.), D. Black (Organon, U.K.), S. Newbury (Oxford, U.K.), J. Sayers (Sheffield, U.K.) and A. Lloyd (University College London, U.K.).
I acknowledge K. Zhao, X. Chai, R. Harshaw and A. Clements, who have contributed to this work.