The function of Sir2 in promoting longevity in yeast mother cells appears to relate to silencing in the rDNA. The stability of the 100–200 tandem copies of rDNA on chromosome XII requiresSIR2, as the frequency of recombination at that locus increases about 10-fold in sir2 mutants (Gottlieb and Esposito 1989). One of the products of rDNA recombination is extrachromosomal rDNA circles (ERCs) (Fig. 1B), which, once formed, replicate and segregate preferentially to mother cell nuclei (Sinclair and Guarente 1997). ERCs thus accumulate in mother cells as they grow older and ultimately trigger senescence. At least one function of Sir2 in yeast longevity, therefore, is to forestall the appearance of the first rDNA circle in mother cells by creating a silenced chromatin structure.
Silencing requires particular lysines in the extended amino-terminal tail of histones H3 and H4 (Thompson et al. 1994; Hecht et al. 1995;Braunstein et al. 1996). These and other lysines of the tail are acetylated in active chromatin but deacetylated in silenced chromatin (Braunstein et al. 1993, 1996). The deacetylated histones evidently can fold into a more compact, closed nucleosomal structure (Luger et al. 1997). These considerations led to the suggestion that Sir2 could be a histone deacetylase. Further evidence for this claim arose from the global deacetylation of yeast histones observed when Sir2 was overexpressed (Braunstein et al. 1993). However, attempts to demonstrate a histone deacetylase activity by Sir2 in vitro initially met with failure.

Sir2 is a conserved NAD-dependent histone deacetylase

Unlike SIR3 and SIR4, the SIR2 gene is broadly conserved in organisms ranging from bacteria to humans (Brachmann et al. 1995). Studies on the bacterial homolog,cobB, led to the conclusion that this gene could substitute for another bacterial gene, cobT, in the pathway of cobalamin synthesis (Tsang and Escalante-Semerena 1998). cobT was known to encode an enzyme that transferred ribose–phosphate from nicotinic acid mononucleotide to dimethyl benzimidazole. Thus, it seemed possible that Sir2 proteins might be equipped to catalyze a related reaction at the nicotinamide–ribose bond in NMN and perhaps nicotinamide-adenine dinucleotide (NAD), in the latter case resulting in transfer of ADP–ribose. Indeed, it was shown by Frye (1999) that Sir2 proteins from bacteria, yeast, or mammals were able to transfer 32P from NAD to a protein carrier, suggesting that they were ADP–ribosyl transferases. Subsequent work proved that Sir2 could, in fact, transfer ADP–ribose, albeit in a reaction that proceeds only weakly in vitro (Tanny et al. 1999). This latter study led to the proposal that the ADP–ribosyltransferase activity of Sir2 was essential to the in vivo function of silencing.
In studying this ADP–ribosyl transferase reaction, we noticed that peptides of the amino-terminal tails of histone H3 or H4 could accept32P from NAD, but only if the peptides were acetylated. Using acetylated H3, we separated the Sir2-modified product by chromatography and found by mass spectrometry that the molecular weight of the product was actually smaller by 42, indicating that the major modification catalyzed by Sir2 was deacetylation and not ADP ribosylation (Imai et al. 2000). When NAD was omitted, no deacetylation by Sir2 occurred. NADH, NADP, or NADPH could not substitute for NAD in this reaction. The weak ADP–ribosyltransferase reaction did not generate sufficient levels of product to allow detection by this physical method.
Because the deacetylase activity of Sir2 occurs preferentially on histone residues that are essential for silencing, we infer that it is this activity, rather than the ADP–ribosyltransferase, that triggers silencing in vivo. Consistent with this claim, a mutation of Gly-270 of Sir2 to Ala reduces the ADP–ribosyltransferase by 93%, but reduces the deacetylase activity by only 20% and still can function in silencing, repression of rDNA recombination, and extension of life span (Imai et al. 2000). Thus, Sir2 is an NAD-dependent histone deacetylase that may link metabolism and silencing in vivo (Fig.2). The role of the ADP–ribosyltransferase in vivo is still not clear, but this activity is evidently separable from the deacetylase, as a known inhibitor of mono-ADP–ribosyltransferases selectively inhibits the one activity of Sir2 and not the other (Imai et al. 2000). The ADP–ribosyltransferase may turn out to be important to the function of Sir2 in DNA repair, as nuclear mono- and poly-ADP–ribosyltransferases have been associated with DNA repair in mammalian cells (Kreimeyer et al. 1984; Pero et al. 1985; Meyer and Hilz 1986).