Transcriptional silencing in Saccharomyces cerevisiae occurs at the silent mating-type loci HML and HMR, at telomeres, and at the ribosomal DNA (rDNA) locus RDN1. Silencing in the rDNA occurs by a novel mechanism that depends on a single Silent Information Regulator (SIR) gene, SIR2. SIR4, essential for other silenced loci, paradoxically inhibits rDNA silencing. In this study, we elucidate a regulatory mechanism for rDNA silencing based on the finding that rDNA silencing strength directly correlates with cellular Sir2 protein levels. The endogenous level of Sir2p was shown to be limiting for rDNA silencing. Furthermore, small changes in Sir2p levels altered rDNA silencing strength. In rDNA silencing phenotypes, sir2 mutations were shown to be epistatic to sir4 mutations, indicating that SIR4 inhibition of rDNA silencing is mediated through SIR2. Furthermore, rDNA silencing is insensitive to SIR3 overexpression, but is severely reduced by overexpression of full-length Sir4p or a fragment of Sir4p that interacts with Sir2p. This negative effect of SIR4 overexpression was overridden by co-overexpression of SIR2, suggesting that SIR4 directly inhibits the rDNA silencing function of SIR2. Finally, genetic manipulations of SIR4 previously shown to promote extended life span also resulted in enhanced rDNA silencing. We propose a simple model in which telomeres act as regulators of rDNA silencing by competing for limiting amounts of Sir2 protein.
CLASSICAL transcriptional silencing in the budding yeast Saccharomyces cerevisiae was originally identified at the silent mating-type loci HML and HMR, which are located on chromosome III (for review see Loo and Rine 1995). A related form of silencing also occurs at telomeres and is known as telomere position effect (TPE; Gottschlinget al. 1990). Foreign genes inserted in these regions are repressed in a reversible manner that resembles the position effect variegation observed in the heterochromatin of more complex eukaryotes (Pillus and Rine 1989; Gottschlinget al. 1990; Weiler and Wakimoto 1995; Dobieet al. 1997). Both of these chromosomal regions display a general inaccessibility to DNA-modifying enzymes in vitro and in vivo (Strathernet al. 1982; Kostrikenet al. 1983; Gottschling 1992; Singh and Klar 1992; Loo and Rine 1994). Genes transcribed by either RNA polymerase II or III (Pol II or Pol III) are generally repressed (Schnell and Rine 1986; Gottschlinget al. 1990; Huanget al. 1997), although at telomeres, the degree of silencing depends on the intrinsic features of each individual promoter (Renauldet al. 1993). These properties, combined with similarities to classically defined eukaryotic heterochromatin, such as DNA replication late in S phase, location near the nuclear envelope, and physical compaction, have led to proposals that the silent mating-type loci and telomeres in yeast are functionally equivalent to heterochromatin (Thompsonet al. 1994; Braunsteinet al. 1996).
A novel form of silencing was described recently in the ribosomal DNA (rDNA) of S. cerevisiae (Bryket al. 1997; Smith and Boeke 1997). The rDNA consists of 100–200 copies of a 9.1-kb unit organized into a tandem array on chromosome XII (Petes and Botstein 1977; Philippsenet al. 1978). Each 9.1-kb unit contains the gene for 35S ribosomal RNA precursor, transcribed by RNA Pol I, and the gene for 5S rRNA, transcribed by RNA Pol III (Figure 1). These two transcription units are separated by a nontranscribed spacer (NTS) that contains an origin of DNA replication (Szostak and Wu 1979; Skryabinet al. 1984). We originally demonstrated that the expression of several different Pol II–transcribed marker genes integrated into various regions of the rDNA was repressed (Smith and Boeke 1997). Of the Silent Information Regulator (SIR) genes, rDNA silencing depends only on SIR2 (Smith and Boeke 1997). In contrast, all four SIR genes contribute to silencing the HM loci, whereas TPE only requires SIR2, SIR3, and SIR4 (Haber and George 1979; Klaret al. 1979; Rineet al. 1979; Rine and Herskowitz 1987; Aparicioet al. 1991). TPE and HM silencing also depend on other factors, including Rap1p (reviewed in Shore 1994), as well as histones H3 and H4 (reviewed in Loo and Rine 1995). The role of non-SIR silencing factors and cis-acting elements in mediating rDNA silencing is just beginning to be investigated. The UBC2 (RAD6) gene has been recently implicated in both TPE (Huanget al. 1997) and rDNA silencing (Bryket al. 1997). Topoisomerase I (TOP1), not previously implicated in silencing, is also required for silencing Ty1 elements in rDNA, as well as for suppression of mitotic rDNA recombination (Christmanet al. 1988; Bryket al. 1997).
We previously demonstrated that deletion of SIR2 increases the accessibility of rDNA to psoralen cross-linking in vivo (Smith and Boeke 1997), suggesting that SIR2 is required for the formation or maintenance of a specialized rDNA chromatin structure that is repressive to RNA Pol II. Using in vivo formaldehyde cross-linking, Sir2p has been found to be associated with rDNA chromatin, mostly within the NTS region (Gottaet al. 1997). In addition, loss of SIR2 increases the frequency of mitotic and meiotic recombination within the rDNA (Gottlieb and Esposito 1989). Taken together, these findings indicate that SIR2 plays a major role in rDNA function, and they are consistent with the nucleolar localization of Sir2 (Gottaet al. 1997).
Although SIR4 is not required for rDNA silencing, its deletion paradoxically enhances the strength of rDNA silencing (Smith and Boeke 1997), suggesting that SIR4 may regulate such silencing. TPE and HM silencing regulation can occur at the level of Sir protein dosage. For example, increased SIR3 gene dosage enhances the strength of TPE, resulting in the spreading of repressive telomeric chromatin toward the centromere (Renauldet al. 1993). Sir3p is physically associated with this extended telomeric chromatin (Hechtet al. 1996). Another example is that diploid strains containing one copy of SIR4 display unstable repression at HMR (Susselet al. 1993), and overexpression of either SIR4 or its C-terminal fragment greatly reduces silencing at both HM loci and telomeres (known as the “anti-SIR” effect; Marshallet al. 1987; Renauldet al. 1993; Cockellet al. 1995). Sir1p levels are limiting for establishment of HM silencing (Stoneet al. 1991). So far, there have been no reports that SIR2 dosage affects TPE (Renauldet al. 1993) or HM silencing, although high-copy SIR2 expression does enhance rDNA silencing in the rDNA (Fritzeet al. 1997). SIR2 overexpression leads to a decrease in histone acetylation at the HM loci (Braunsteinet al. 1993), and high levels of SIR2 overexpression are toxic to cells, provoking chromosome instability (Holmeset al. 1997).
It has been suggested that the local concentration of Sir proteins is critical in determining silencing efficiency, even at internal chromosomal locations (Stavenhagen and Zakian 1994; Lustiget al. 1996; Mailletet al. 1996; Marcandet al. 1996). This hypothesis is supported by other studies showing that telomeres, the Sir proteins, and Rap1p colocalize into a limited number of discrete intranuclear foci, predominantly near the nuclear membrane (Palladinoet al. 1993; Cockellet al. 1995; Gottaet al. 1996). Competition for these limiting amounts of silencing factors has been shown to occur between telomeres and the HM loci. The rap1s mutation, rap1-12, simultaneously reduces HMR-specific silencing and enhances telomeric silencing (Buck and Shore 1995). This shift in the balance of silencing to telomeres was proposed to result from sequestration of Sir4p at telomeres, thus reducing the amount of Sir4p available for HMR silencing (Buck and Shore 1995; Marcandet al. 1996).
We demonstrate in this article that SIR4 regulates rDNA silencing strength by controlling the level of Sir2p that is available to the rDNA. We show that rDNA silencing is highly sensitive to even small changes in Sir2 levels, and that there are normally limiting amounts of Sir2p available for rDNA silencing. We also show that SIR4 negatively affects rDNA silencing. SIR4 overexpression severely reduced the strength of rDNA silencing, and sir4Δ strengthened rDNA silencing through a SIR2-dependent, SIR3-independent mechanism. Finally, manipulations of SIR4, which were previously shown to redistribute Sir proteins (including Sir2p) to the nucleolus and promote longevity, also cause an increase in rDNA silencing strength. We propose a Sir2p competition model in which Sir4p regulates the shuttling of Sir2p between competing rDNA and telomeric compartments.
MATERIALS AND METHODS
Media and plasmids: Unless stated otherwise, media were used as described previously (Roseet al. 1990; Smith and Boeke 1997). Ty1 transposition induction medium (YNB/casamino acids + Trp galactose) consisted of yeast nutrient broth (YNB; Difco, Detroit) supplemented with 2% casamino acids, 2% galactose, 160 μm adenine, and 800 μm tryptophan. SC (FOA) medium contained 5-FOA (PCR Incorporated) at 2 mg/ml and glucose at 2%. Pb2+-containing medium (modified lead acetate, MLA) consisted of 0.3% peptone, 0.5% yeast extract, 4% glucose, 0.02% (w/v) ammonium acetate, 0.1% Pb(NO3)2, and 2% agar.
The mURA3-marked GAL-Ty1 overexpression plasmid pJSS36-6 and the CEN, SIR2 plasmid pCAR237 were described previously (Smith and Boeke 1997). pLP304 (2μ, LEU2 SIR3) and pLP305 (2μ, LEU2 SIR4) were also described previously (Stone and Pillus 1996). pLP347 consists of a HST1 SacI fragment ligated into the SacI site of YEp351. pSIR2μ was constructed by inserting the HindIII genomic SIR2 fragment from pCAR237 into the HindIII site of pRS425 (Christiansonet al. 1992). pJSS71-13 (2μ, HIS3, SIR2) was constructed by ligating a XhoI-NotI fragment from pCAR237, which contains SIR2, into the XhoI-NotI sites of pRS423. pLP754 (2μ LEU2 SIR4-42) consists of a NotI-SalI SIR4-42 fragment that was blunted by Klenow treatment and then ligated into the SmaI site of YEp351 (Hillet al. 1986). pLP793 (integrating TRP1 sir4-42) was constructed by ligating the SacI-SalI sir4-42 fragment from pLP754 into the SacI-SalI sites of pRS304. pJH3A and pJH5.1A (provided by James Broach) were previously described (Ivyet al. 1986; Marshallet al. 1987). pJSS73-5 was constructed by ligating an EcoRI-SacII fragment containing the GAL1 promoter and a C-terminal fragment of SIR4 from pRO135 into the same sites of pRS313. pRO135 was a plasmid derived from a galactose-inducible overexpression library (Liuet al. 1992) and was provided by Rohinton Kamakaka. The kanMX4 gene of pRS400 (Brachmannet al. 1998) was derived from pFA6a-kanMX4, provided by Achim Wach and Peter Philippsen (Wachet al. 1994).
Yeast strains: All strains used in this study are derived from JB740 or its derivatives (Table 1; Smith and Boeke 1997). JS50-1 is a MATα/α diploid isolated during transformation of JB740 with the galactose-inducible Ty1-mURA3 plasmid pJSS36-6 (Smith and Boeke 1997), presumably via polyethylene glycol (PEG)-mediated fusion (J. S. Smith and J. D. Boeke, unpublished data). JS249 was derived from JS50-1 by loss of pJSS36-6. To produce JS101 and JS109, transposition was induced in JS50-1 to produce Ty1-mURA3 integration events into the rDNA or non-rDNA locations, as described previously (Smith and Boeke 1997). SIR gene deletions are complete ORF deletions and were performed by PCR-mediated gene disruption as described elsewhere (Baudinet al. 1993; Lorenzet al. 1995). The HIS3, LEU2, or kan MX4 genes were PCR amplified from pRS403, pRS405, or pRS400, respectively (Sikorski and Heiter 1989; Brachmannet al. 1998), using oligodeoxynucleotide primers containing 5′ flanking sequences complementary to the SIR genes (Smith and Boeke 1997). One SIR2 allele in the MATα/α diploid JS101 was replaced with HIS3 (sir2Δ::HIS3) to produce heterozygote JS106, which can mate. The second SIR2 allele of this heterozygote was replaced with LEU2 (sir2Δ::LEU2) to produce a homozygous SIR2 knockout that was a nonmater.
sir3Δ::kanMX4 was similarly introduced into JS308 (MATa) producing JS316. The MATα sir4Δ::HIS3 deletion strain JS219 has been described previously (Smith and Boeke 1997). JS316 was transformed with pLP304 (2μ LEU2 SIR3), JS219 was transformed with pLP305 (2μ LEU2 SIR4), and the resulting transformants were crossed. After loss of the LEU2 plasmids, tetrads of the resulting diploid XJS3 were dissected to generate congenic SIR+(JS333), sir4Δ (JS337), sir3Δ (JS335), and sir3Δ sir4Δ (JS339) spores. Diploid XJS5 was similarly sporulated, and tetrads were dissected to generate the congenic sir2Δ and sir2Δ sir4Δ spores. To produce a sir4-42 strain, pLP793 was linearized by SacII cleavage within TRP1 and then integrated into the trp1Δ63 locus of the sir4Δ strain JS337, resulting in strain JS347.
Tetrad dissection: Diploid strains XJS3 and XJS5 were grown as patches on YPD medium for ~20 hr. Cells from these patches were transferred to 2 ml of liquid sporulation medium (1% potassium acetate, 0.002% zinc acetate) and incubated 1 day at 25°, followed by 4 days at 30°. Tetrads were dissected on a YPD plate and incubated for 3 days at 30° to allow spores to grow up as colonies. Dissection plates were replica plated to SC-His, YPD+G418 (200 μg/ml), and MLA to determine the SIR genotype and rDNA silencing strength.
Sir2p-specific antibody: A Sir2p-specific antiserum was raised using a synthetic peptide (CGVYVVTSDEHPKTL) comprising the 14 C-terminal amino acids plus a nonencoded cysteine residue. The cysteine residue was added to facilitate m-maleimidobenzoyl-N-hydroxysuccinimide ester (Pierce, Rockford, IL) conjugation of the peptide to carrier keyhole limpet hemocyanin. Rabbits were immunized, and serum was collected using standard protocols (Harlow and Lane 1988). Serum was diluted and used without further purification, as described below.
Protein extraction and immunoblotting: Strains were grown to saturation in 10 ml of YPD or SC-Leu medium, diluted to an A600 of 0.2 in YPD or SC-Leu, and grown to an A600 of ~1.0. Pelleted cells were resuspended in 100 μl lysis buffer (50 mm Tris-HCl, pH 7.5, 1% SDS, 5 mm EDTA, 14.3 mm 2-mercaptoethanol, 2 mm PMSF, 1 μg/ml leupeptin, 2 μg/ml antipain, 10 μg/ml benzamidine, 1 μg/ml chymostatin, 1 μg/ml pepstatin A, and 10 U/ml aprotinin). Glass beads (0.45–0.5 mm diameter) were added to the meniscus, and the tubes were vortexed at full speed for five 30-sec pulses and placed on ice in between pulses. Another 100 μl of lysis buffer was added, followed by boiling for 3 min and centrifugation (14,000 × g for 20 min at 4°). An equal volume of 2× loading buffer was added to the supernatant and then boiled again before loading.
Some 300 A280 units of each extract were separated on a 10% SDS polyacrylamide gel and then electroblotted to an Immobilon-P filter (Millipore, Bedford, MA) at 300 mA for 1 hr using a mini-Protean II apparatus (Bio-Rad, Richmond, CA). Filters were blocked for 45 min with 10 ml of blocking solution (1× PBS/0.05% Tween 20/5% milk), then incubated with rabbit polyclonal α-Sir2 antibody 2916 (1:5000 dilution in blocking solution) or control mouse monoclonal α-tubulin antibody B-5-1-2, 6.1 mg/ml (1:5000 dilution; Sigma, St. Louis, MO) for 1 hr at room temperature. The appropriate secondary antibody (anti-rabbit HRP conjugated or anti-mouse HRP conjugated; Amersham Corp., Arlington Heights, IL) was used at a 1:10,000 dilution for 1 hr in blocking solution. Filters were developed on Kodak XAR5 film using the ECL detection system (Amersham).
RNA blot analysis: Saturated YPD cultures were diluted to an A600 of 0.2 in YPD and grown at 30° to midlog phase (A600 of 1.4–1.6). Forty micrograms total RNA, isolated as described by Chapman and Boeke (1991), was separated on a formaldehyde-containing 1.2% agarose gel and transferred to Gene-screen Plus (NEN-Dupont, Boston, MA) in 10× SSC. The filter was probed with an antisense URA3 RNA probe labeled with [α-32P]UTP (800 Ci/mmol). The URA3 RNA probe was transcribed by T7 RNA polymerase from pBS-URA3 linearized with StuI, producing a 363-nt antisense probe. The filter was prehybridized in 10 ml of 5× SSPE, 5× Denhardt's solution, 1% SDS, 50% formamide (w/v), and 100 μg/ml boiled herring sperm DNA at 60° for 3 h, followed by hybridization in the above solution without DNA at 60° for 16 h. The filter was washed twice at room temperature 2× SSC/0.1% SDS, two times at room temperature in 0.2× SSC/0.1% SDS, and twice at 42° in 0.2× SSC/0.1% SDS. The filter was stripped and reprobed with a 1.1-kb BamHI-HindIII fragment of actin, which was derived from pΔ10-AHX3 (Chapman and Boeke 1991), as a loading control. The filter was imaged on a PhosphorImager (Molecular Dynamics, Sunnyvale, CA) and quantitated using ImageQuant software.
Ribosomal DNA silencing is highly sensitive to changes in SIR2 dosage: The silencing reporter strains used in this study all carry mURA3- or MET15-marked Ty1 elements integrated within the NTS of rDNA (Figure 1). Both of these markers are partially silenced when integrated within the rDNA tandem array, but are fully expressed when located outside the rDNA or within a single rDNA gene copy that is artificially positioned outside the array (Smith and Boeke 1997). mURA3 expression is measured by growth on medium lacking uracil, whereas MET15 expression is measured by growth on medium lacking methionine or by colony color on Pb2+ containing medium (Cost and Boeke 1996; Smith and Boeke 1997).
We previously demonstrated that silencing of marker genes in the rDNA is completely dependent on the SIR2 gene. Furthermore, deletion of SIR2 increased the accessibility of rDNA to psoralen cross-linking (Smith and Boeke 1997). These results are consistent with a role for SIR2 in the formation or maintenance of a specialized rDNA chromatin structure. Because the rDNA tandem array makes up 5–10% of yeast chromatin, there are likely to be limiting amounts of Sir2p available to act on the rDNA. To test this hypothesis, we first constructed the diploid reporter strain JS101 (D39), which contains a single Ty1-mURA3 element integrated at the rDNA (Figure 1; see materials and methods). SIR2 was then deleted one allele at a time, producing heterozygous and homozygous sir2 deletion strains. The strength of rDNA silencing of the mURA3 reporter, as measured by growth on -Ura medium, was tested using a serial dilution growth assay for each of the resulting strains. Interestingly, when a single copy of SIR2 was deleted (JS106), growth on SC-Ura plates increased approximately fivefold over that of the SIR2/SIR2 (+/+) parent (Figure 2A). At the same time, its sensitivity to FOA increased 5- to 25-fold. This result indicates that a single gene dose of SIR2 is insufficient to fully silence mURA3 within a diploid cell. When the second SIR2 allele was deleted from the heterozygote, creating a homozygous knockout strain (JS108), growth on SC-Ura increased another fivefold to the level of positive control non-rDNA insertion JS109 (Figure 2A). This latter result is identical to that previously observed when SIR2 was deleted from a haploid strain (Bryket al. 1997; Smith and Boeke 1997). The FOA-resistant colonies that grow in the double knockout are completely Ura− and represent mitotic loss of mURA3 (data not shown) because of an increased rate of rDNA recombination events leading to complete loss of the mURA3-marked repeat (Gottlieb and Esposito 1989). The SIR2 deletions did not affect overall growth on nonselective media. The steady-state mURA3 RNA level in these strains was determined by quantitative RNA blot analysis using a URA3-specific riboprobe and is shown schematically in Figure 2B. The relative mURA3 RNA levels corresponded to the relative amount of Ura+ growth for each strain, indicating that growth assays are a reliable method for determining silencing efficiency in the rDNA.
For independent confirmation of the intermediate SIR2 dosage effect, Ty1-mURA3 expression driven from the Ty1 promoter was quantitated from the same URA3-probed filter (Figure 2C). Again, the level of the Ty1-mURA3 6.7-kb message directly correlated with the dosage of SIR2 (Figure 2C), indicating that the dosage effect was not specific to the mURA3 promoter, as measured in Figure 2, A and B, but it also influenced the Ty1 promoter. The level of the mURA3 message in a sir2 homozygote was equivalent to that in the non-rDNA control JS109 (Figure 2B), but the level of the larger Ty1-mURA3 message was significantly greater when transcribed from a locus outside the rDNA than within the rDNA, even in the sir2/sir2 homozygote (JS108). We attribute this large difference to the sensitivity of the Ty1 promoter to chromosomal position effects (Curcio and Garfinkel 1991). In summary, by both a growth assay and by directly examining transcription, the SIR2/sir2 heterozygote is phenotypically intermediate between the sir2/sir2 and SIR2/SIR2 homozygotes.
To confirm that Sir2 protein levels had actually changed in these cells, the relative amounts of Sir2p were analyzed by immunoblotting using an α-Sir2 antibody (Figure 2D); the steady-state Sir2p levels corresponded to SIR2 gene dosage. These results indicate that silencing of a single Ty1-mURA3 element inserted at the rDNA of a diploid cell is exquisitely sensitive to Sir2p levels. Whereas rDNA silencing was highly sensitive to SIR2 dosage, HMRa silencing, as measured by a sensitive mating assay, was not affected by the twofold difference in SIR2 dosage (data not shown).
Increased dosage of SIR3 was previously shown to enhance telomeric silencing, but increased SIR2 dosage had no effect (Renauldet al. 1993). To determine whether increased SIR2 or SIR3 dosage could strengthen the rDNA silencing observed in a haploid isolate, CEN (low-copy) and 2μ (high-copy) plasmids containing SIR2 were transformed into the SIR2 haploid Ty1-mURA3 isolate JS124 (S2). This haploid silencing reporter strain contains a single Ty1-mURA3 insertion (within the NTS2) upstream of the 5S rRNA gene (Figure 1; Smith and Boeke 1997). The level of silencing is shown for each strain in Figure 3A; CEN SIR2 caused a small (less than fivefold) increase in repression compared to the empty CEN vector. Furthermore, the 2μ SIR2 plasmid caused between a 5- and 25-fold increase in repression compared to the empty 2μ vector. These results indicate that there are limiting amounts of cellular Sir2p for rDNA silencing, and that silencing can be enhanced beyond wild-type levels by providing additional Sir2p. In contrast, 2μ plasmids containing SIR3 or the closely related SIR2 homolog HST1 (Brachmannet al. 1995; Derbyshireet al. 1996) had no effect on rDNA silencing strength. In addition, high-copy SIR3 did not effect the enhanced silencing caused by high-copy SIR2 when the two genes were coexpressed in the same strain (data not shown). Another SIR2 homolog, HST3, also had no effect on rDNA silencing when overexpressed from the GAL1 promoter (data not shown). Unexpectedly, the level of Ura+ growth was modestly greater for the CEN vector set than for the 2μ vector set, including the empty vectors. The reason for this difference is currently unclear, but could be caused by the known effects of CEN plasmids on the cell cycle (Spencer and Hieter 1992).
The relative steady-state amounts of Sir2p produced by CEN SIR2 or 2μ SIR2 plasmids in strain S2 are shown in Figure 3B. The modest increase (approximately two-fold) in Sir2p caused by CEN SIR2 that is associated with enhanced silencing further emphasizes the extreme sensitivity of rDNA silencing to SIR2 dosage. 2μ SIR2 boosted Sir2p protein levels approximately twofold more than CEN SIR2, concomitant with a greater increase in rDNA silencing strength. SIR2 overexpression is toxic to cells (Holmeset al. 1997), which could explain the relatively small increase in Sir2p levels produced by the 2μ SIR2 plasmid. Cells expressing very high levels of Sir2p would be selected against.
SIR4 regulation of rDNA silencing is dependent on SIR2, but not on SIR3: We previously demonstrated that sir4Δ causes enhanced silencing of both mURA3 and MET15 genes inserted in the rDNA (Smith and Boeke 1997). This result suggested that SIR4 negatively regulates rDNA silencing. We were interested in understanding the nature of this regulation. The enhanced silencing phenotype of a sir4Δ mutant resembled the enhanced silencing caused by increased SIR2 expression. Therefore, the simplest explanation for enhanced silencing would be a sir4Δ-induced increase in cellular Sir2p levels. This initial hypothesis was incorrect because Sir2p levels were unaffected by sir4Δ (data not shown). In wild-type cells, Sir4p, Sir3p, Rap1p, telomeric DNA, and some Sir2p are normally colocalized within discrete foci at the nuclear periphery (Gottaet al. 1996). The bulk of Sir2p is nucleolar (Gottaet al. 1997). When SIR4 is deleted, all detectable Sir2p and Sir3p redistributes to the nucleolus (Gottaet al. 1997). The redistribution of Sir3p is especially dramatic as it is undetectable in wild-type nucleoli (Gottaet al. 1997). Another possible explanation for the enhanced rDNA silencing in a sir4Δ strain could therefore be the nucleolar redistribution of Sir3p, Sir2p, or both. To test these possibilities, congenic SIR, sir2, sir3, sir4, sir2 sir4, and sir3 sir4 strains were qualitatively tested for rDNA silencing strength using a MET15-based colony color assay in an epistasis analysis. In this color assay, Met+ cells produce white colonies on Pb2+ containing medium (MLA), and Met− cells produce dark brown colonies (Cost and Boeke 1996). Strains containing Ty1-MET15 within the rDNA produce colonies with an intermediate tan color, representing partial silencing of MET15 (Smith and Boeke 1997). Using this assay, enhanced rDNA silencing is visualized as a darker colony color. As shown in Figure 4, the colony colors of a sir4Δ strain and a sir3Δ sir4Δ strain were equivalent at all stages of growth, indicating that SIR3 was not required for the enhanced rDNA silencing phenotype caused by sir4Δ. The sir2Δ sir4Δ strain produced highly sectored white colonies that were phenotypically indistinguishable from the sir2Δ strain and indicative of loss of rDNA silencing (Smith and Boeke 1997). Therefore, SIR2 is necessary for the sir4Δ effect of enhanced rDNA silencing, but SIR3 is not. The specificity of these effects suggests that the enhanced rDNA silencing caused by deletion of SIR4 does not require relocalization of Sir3p to the nucleolus but, rather, occurs by a SIR2-dependent mechanism.
High-copy SIR4 reduces rDNA silencing: Whereas deletion of SIR4 enhances rDNA silencing, it drastically reduces HM and telomeric silencing. Overexpression of SIR4 or its C terminus reduces telomeric and HM silencing in a dominant negative fashion, which has been termed the “anti-SIR” effect (Marshallet al. 1987; Renauldet al. 1993; Cockellet al. 1995). The C terminus of Sir4p interacts with the Sir3 and Rap1 proteins, and these interactions are proposed to help recruit a Sir2p/Sir3p/Sir4p complex to telomeres and the silencer elements of the HM loci (Morettiet al. 1994; Cockellet al. 1995; Hechtet al. 1995; Moazedet al. 1997). The dominant negative activity of excess SIR4 toward TPE and HM silencing is proposed to be caused by antagonism of these interactions (Marshallet al. 1987; Morettiet al. 1994; Cockellet al. 1995). For example, SIR4 overexpression could result in the formation of excess incomplete silencing complexes at the expense of complete ones.
To examine the role of SIR4 dosage on rDNA silencing, a high-copy plasmid containing full-length SIR4 was introduced into the haploid Ty1-mURA3 rDNA insertion isolate S6, and the resulting strains were tested for rDNA silencing strength in a serial dilution growth assay on -Ura medium (Figure 5). Empty vectors had no effect on the repression of Ura+ growth. However, the SIR4 plasmid increased the Ura+ growth of S6 by at least 25-fold, indicating that high-copy SIR4 severely reduces rDNA silencing strength (Figure 5), very similar to the “anti-SIR” effect that occurs at telomeres and HM loci. Silencing was partially restored by co-overexpression of SIR2 with SIR4, suggesting that SIR4 negatively affects rDNA silencing by interfering with SIR2 function. Highcopy SIR4 also reduced silencing in sir4 and sir3 mutant versions of S6 (data not shown), indicating that SIR3 was not required for the high-copy SIR4–dependent loss of silencing. The strength of rDNA silencing therefore correlates inversely with SIR4 dosage.
To determine which domains of SIR4 antagonize rDNA silencing, 2μ plasmids expressing various domains of SIR4 were tested for their effect on rDNA silencing using the Ura+ growth assay. The results are shown in Figure 6A and tabulated in Figure 6B. Overexpression either of the SIR4-42 allele, which truncates the extreme C terminus, or of the C-terminal 40% of SIR4 (pJH5.1A) had a dominant negative effect on rDNA silencing. pJH5.1A was also moderately toxic to these strains for unknown reasons. Surprisingly, the extreme C terminus of SIR4 actually caused an increase in rDNA silencing strength (pJH3A) similar in amplitude to the effect of a 2μ SIR2 plasmid. Increased rDNA silencing was also observed with pJSS73-5, which overexpresses the C-terminal 23% of Sir4p (Figure 6B). The region in common between the constructs that cause derepression (shaded area) overlaps with the region of Sir4p previously shown to interact with Sir2p, but not with Sir3p (see black bar in Figure 6B; Moazedet al. 1997). Based on these results, we propose that the inhibitory effect of Sir4p on rDNA silencing requires its physical interaction with Sir2p.
How could overexpression of the extreme C terminus of Sir4p cause increased rDNA silencing? In addition to the SIR genes, UTH4 and YGL014W are also required for a long life span in yeast (Kennedyet al. 1997). Overexpression of the extreme C-terminal Sir4p domain restores a long life span to specific uth4 mutant strains, a phenotype shared by the same uth4 strains bearing the SIR4-42 mutation (Kennedy et al. 1995, 1997). Interestingly, the SIR4-42 allele in single copy also results in redistribution of the Sir complex from telomere foci to the nucleolus, regardless of UTH4 genotype (Kennedyet al. 1997). The enhanced rDNA silencing produced by overexpression of the SIR4 C terminus (pJH3A) in Figure 6 could therefore be caused by redistribution of telomeric Sir2p to the nucleolus (rDNA). If this is true, then an integrated SIR4-42 allele would be predicted to also cause enhanced rDNA silencing. To test this hypothesis, we integrated the SIR4-42 allele into a strain deleted for SIR4 and tested rDNA silencing of MET15 (Figure 7). In this case, SIR4-42 caused enhanced rDNA silencing (dark colony color) similar to the sir4Δ strain (Figure 7). A similar phenotype was also produced by introducing a stop codon after amino acid 1237 of the endogenous SIR4 gene (data not shown). These results suggest that mutations which cause redistribution of the Sir complex to the nucleolus, especially redistribution of Sir2p, also result in enhanced rDNA silencing. Remarkably, pJH3A further strengthened the enhanced rDNA silencing phenotype of a sir4Δ strain (data not shown), suggesting that overexpression of the SIR4 C terminus may release even more Sir2p for rDNA silencing than does sir4Δ. Alternatively, the SIR4 C terminus could release an additional separate factor that acts at the rDNA.
The importance of SIR2 to cellular function is emphasized by the identification of a family of SIR2 homologs that exist in many diverse organisms, including bacteria, yeast, plants, and mammals (Brachmannet al. 1995; Derbyshireet al. 1996). We and others previously showed that SIR2 was required for rDNA silencing in S. cerevisiae (Bryket al. 1997; Smith and Boeke 1997). The data presented here demonstrate that rDNA silencing is highly sensitive to the dosage of SIR2 and SIR4, but is independent of SIR3 dosage. The strength of rDNA silencing is proportional to the amount of Sir2p produced, suggesting that modulation of cellular Sir2p levels or the redistribution of the normal Sir2p pool could regulate rDNA silencing levels. Such regulation of silencing by modulation of Sir2p has not been previously observed for the other known forms of silencing and may be specific to the regulation of rDNA silencing.
Model of rDNA silencing regulation by SIR2 and SIR4: Immunolocalization studies have demonstrated that the Sir2, Sir3, Sir4, and Rap1 proteins colocalize along with telomeric DNA to several subnuclear foci that normally associate with the nuclear periphery (Palladinoet al. 1993; Gottaet al. 1996). This accumulation of Sir proteins at the telomeric foci may act to produce a critical concentration necessary for efficient silencing and chromosomal integrity (Palladinoet al. 1993; Cockellet al. 1995; Gottaet al. 1996). The silencing complex is also required for silencing at the HM loci, resulting in competition between telomeres and the HM loci for limiting amounts of Sir3 and Sir4 proteins (Buck and Shore 1995; Marcandet al. 1996). Mutations that lengthen telomeres, such as rap1s, cause the balance of silencing to be tipped toward strengthened TPE (Buck and Shore 1995). This is proposed to result from titration of Sir4p away from the HM loci. Furthermore, it has been proposed that the sequestration of Sir3p and Sir4p by Rap1p at telomeres controls silencing at artificially created, internal, nontelomeric silencing sites (Mailletet al. 1996; Marcandet al. 1996).
Whereas some Sir2p is localized to the telomeric foci, the bulk is in the nucleolus (Gottaet al. 1997). We have now demonstrated that the wild-type level of Sir2p is limiting for rDNA silencing and that silencing strength is proportional to SIR2 dosage. In a recent independent study, high-copy SIR2 was also shown to decrease the expression of ADE2 and CAN1 markers inserted in the rDNA (Fritzeet al. 1997). In contrast to the effect of SIR2 dosage, the strength of rDNA silencing is inversely proportional to SIR4 dosage. We propose a model for the regulation of rDNA silencing in which the amount of Sir2p available in the nucleolus for silencing is controlled by the sequestration of Sir2p at telomeres through an interaction with Sir4p (Figure 8). Several lines of evidence support this model.
When SIR4 is deleted, rDNA silencing becomes enhanced (Smith and Boeke 1997). This is the opposite effect of what happens at telomeres and HM loci, where SIR4 is absolutely required for silencing (Ivyet al. 1986; Rine and Herskowitz 1987; Aparicioet al. 1991). Deletion of SIR4 also results in the redistribution of the telomere-localized Sir3p and Sir2p to the nucleolus (Gottaet al. 1997), which is the subnuclear location of the rDNA. Epistasis analysis revealed that SIR2, but not SIR3, is required for the enhanced rDNA silencing caused by sir4Δ. Therefore, the enhancement in silencing cannot be caused simply by relocalization of Sir3p to the nucleolus. It is more likely that the redistribution of telomeric Sir2p to the nucleolus increases the local concentration of Sir2p at the rDNA, resulting in stronger rDNA silencing (see Figure 8, model). In the case of sir3Δ, Sir4p displays a diffuse nuclear localization pattern. Sir2p and Sir4p have been shown to interact independently of Sir3p when not associated with chromatin (Moazedet al. 1997). Therefore, this diffuse nuclear Sir4p would prevent most of the telomeric pool of Sir2p from redistributing to the rDNA (Figure 8, model), thus causing only the minor enhancement in rDNA silencing that we observe.
Overexpression of SIR4 results in loss of rDNA silencing (Figure 5) similar to the dominant negative “anti-Sir” effect of SIR4 overexpression on TPE and HM silencing (Marshallet al. 1987; Renauldet al. 1993; Cockellet al. 1995). The negative effect of SIR4 on HM silencing can be compensated for by co-overexpression of SIR3 (Marshallet al. 1987). Strikingly, co-overexpression of SIR2 partially alleviates the overexpression effect of SIR4, resulting in restoration of rDNA silencing (Figure 6A). These results suggest that Sir4p disrupts the normal nucleolar function of Sir2p in a titratable manner.
We found that a specific domain of Sir4p is responsible for the negative effect on rDNA silencing. This region of Sir4p (amino acids 731–1049) overlaps with a domain previously shown to physically interact with Sir2p (Moazed and Johnson 1996; Moazedet al. 1997; Strahl-Bolsingeret al. 1997). SIR4 overexpression results in delocalization of the telomeric Sir4p staining to a diffuse nuclear pattern (Mailletet al. 1996). This excess nuclear Sir4p could potentially titrate Sir2p away from the rDNA, therefore causing derepression of rDNA silencing (Figure 8, model).
A specific mutant allele of SIR4, called SIR4-42, results in redistribution of the Sir silencing complex from telomeres to the nucleolus and suppresses the allele-specific decrease in life span caused by a specific mutation in UTH4, a gene required for long life span (Kennedy et al. 1995, 1997). Overexpression of the extreme C terminus of SIR4 also increases life span, presumably by redistributing the Sir complex to the nucleolus (Kennedyet al. 1995). Nucleolar relocalization of the Sir complex would increase the intranucleolar Sir2p concentration. Strikingly, both of these SIR4 manipulations cause an increase in rDNA silencing, as predicted by our model. In summary, release of rDNA silencing-competent Sir2p from telomeres should enhance rDNA silencing, whereas increased sequestration of Sir2p at telomeres should weaken rDNA silencing. Consistent with this model, mutant alleles of cdc17, rfc1, and rif1, which increase telomere length in yeast (Hardyet al. 1992; Adams and Holm 1996), also result in weakened rDNA silencing (J. S. Smith and J. D. Boeke, unpublished data).
Deregulation of rDNA silencing could have profound metabolic implications for the yeast cell. For example, loss of the repressive SIR2-dependent rDNA chromatin structure causes an increase in the percentage of actively transcribed rDNA gene copies (Smith and Boeke 1997). Increased rDNA silencing is therefore predicted to cause a reduction in the percentage of actively transcribed rDNA gene copies.
Structural differences between rDNA and other forms of silencing: Sir3p is a proposed structural component of yeast telomeric heterochromatin (Hecht et al. 1995, 1996). Therefore, the lack of SIR3 involvement in rDNA silencing emphasizes the fundamental difference in silencing complexes that act at the rDNA vs. the telomeres/HM loci. Sir3p exclusion from the putative rDNA silencing complex may allow for efficient chromatin remodeling that may be necessary in a genomic region, such as the rDNA, which is efficiently transcribed by Pol I and Pol III, but may need to be constrained in terms of other transactions, such as rDNA recombination (Christmanet al. 1988; Gottlieb and Esposito 1989; Bryket al. 1997; Smith and Boeke 1997).
What is the role of SIR2 in rDNA silencing? Not only is Sir2p localized in the nucleolus (Gottaet al. 1997), but Sir2p preferentially associates with rDNA (Gottaet al. 1997). We have previously demonstrated that in sir2Δ strains, transcriptionally inactive rDNA regions become more accessible to psoralen cross-linking in vivo, indicating that Sir2p is involved in either establishment or maintenance of a specialized repressive chromatin structure in the rDNA (Smith and Boeke 1997). Furthermore, a recent study showed that SIR2 modulates the accessibility of rDNA chromatin to micrococcal nuclease digestion in vitro and dam methyltransferase activity in vivo in a dosage-dependent manner (Fritzeet al. 1997). These results, taken together, strongly support the hypothesis that Sir2p contributes to the formation of chromatin structure in the rDNA. Therefore, the modulation of rDNA silencing strength by changes in SIR2 dosage is likely to result from direct changes in the Pol II-repressive chromatin in the rDNA. It is still unclear whether Sir2p is a direct structural component of the rDNA chromatin, or whether it is present in a regulatory capacity. Sir2p does not have a demonstrated DNA-binding activity (Buchmanet al. 1988), suggesting that it might associate with rDNA through interactions with other proteins.
The transcriptional activity of eukaryotic genes correlates with the level of nucleosome acetylation of lysine residues on the N terminus of the core histones (for review see Sternglanz 1996). Accordingly, histones of the HM loci-associated chromatin have been shown to be hypoacetylated compared to other nonsilenced chromosomal regions (Braunstein et al. 1993, 1996). Overexpression of SIR2 correlates with decreased histone acetylation levels in S. cerevisiae (Braunsteinet al. 1993). It is therefore possible that inactive rDNA chromatin is associated with hypoacetylated histones. The rDNA histone acetylation level might be similarly influenced by SIR2. Interestingly, it has been suggested that deacetylated histones are associated with transcriptionally inactive rRNA genes in rat tumor cells (Mutskovet al. 1996). However, this has not yet been directly examined in yeast. SIR2 could be directly or indirectly involved in histone deacetylation, or it could prevent acetylation by blocking the access or activity of acetyltransferase on rDNA histones. Whether the histone acetylation state of rDNA histones controls rDNA silencing remains to be determined.
A large family of genes homologous to SIR2 have been identified in other species ranging from bacteria to human (Brachmannet al. 1995; Derbyshireet al. 1996). Included in this family are four homologs in S. cerevisiae known as HST1, HST2, HST3, and HST4. HST1 overexpression has previously been shown to partially complement the mating defect of a sir2Δ mutant (Brachmannet al. 1995). However, overexpression of HST1 or HST3 had no effect on rDNA silencing strength. These results suggest that individual members of the HST family of genes do not have identical functions. It will be interesting to test the effects of other hst mutants on rDNA silencing. So far, the only SIR2 homolog in other species that has been characterized is from the yeast Kluyveromyces lactis (Chen and Clark-Walker 1994). Mutation of this SIR2 homolog causes an increase in ethidium bromide sensitivity and was reported to cause an increase in rDNA recombination (Chen and Clark-Walker 1994). These results suggest that SIR2 homologs may function broadly in regulation of chromatin, including rDNA chromatin.
rDNA silencing and aging: The life span of S. cerevisiae cells is defined by the relatively fixed number of asymmetric cell divisions that a mother cell can achieve (Mortimer and Johnston 1959). A link between silencing and life span determination was identified through the isolation of a dominant gain of function mutant of SIR4 (SIR4-42), which restored a long life span to a short-lived uth4-14c strain (Kennedyet al. 1995). Because the Sir protein complex is redistributed from telomeric foci to the nucleolus in SIR4-42 cells, the rDNA has been proposed to be the site of this SIR4-42 gain of function and perhaps to be associated with aging (Kennedyet al. 1997). Interestingly, phenotypes associated with the SIR4-42 mutation, including sterility and Sir protein redistribution to the nucleolus, also occur naturally in old mother cells (Smealet al. 1996; Kennedyet al. 1997), suggesting that the nucleolar redistribution of the Sir protein complex may somehow compensate for or protect against the cumulative effects of aging on the nucleolus (Guarente 1997; Kennedyet al. 1997).
We have demonstrated that the SIR4-42 mutation or overexpression of the extreme C terminus of SIR4causes an increase in the strength of rDNA silencing. Does this mean that increased rDNA silencing causes increased life span? As stated above, both of these SIR4 manipulations can restore longevity to specific, short-lived uth4 mutant strains in a SIR3-dependent manner (Kennedyet al. 1995). However, deletion of SIR4 modestly shortens life span (Kennedyet al. 1995) yet enhances rDNA silencing in a SIR2-dependent, SIR3-independent manner. Moreover, SIR3 and SIR4 are required for longevity (Kennedyet al. 1995) but not for rDNA silencing (Smith and Boeke 1997). Taken together, these findings indicate that increased rDNA silencing is not sufficient to lengthen life span.
Enhanced rDNA silencing may, however, be an important component of the cellular response to aging. Our results suggest that increased rDNA silencing strength is a consequence of mutations or conditions that cause redistribution of the Sir complex (especially Sir2p) to the nucleolus, which could be a phenocopy of what happens in old cells. Deletion of the SGS1 gene was recently demonstrated to accelerate the nucleolar fragmentation phenotype that occurs normally in very old wild-type yeast cells (Sinclairet al. 1997). SGS1 is a member of the RecQ helicase family, which includes the Werner's syndrome gene WRN that is implicated in premature aging in humans (Yuet al. 1996). Sgs1p is localized in the nucleolus, and deletion of the gene shortens life span (Sinclairet al. 1997) and increases the frequency of rDNA recombination (Gangloffet al. 1994). It was therefore proposed that nucleolar fragmentation represents a cause of yeast aging, and that redistribution of the Sir silencing complex to the nucleolus delays these changes (Guarente 1997; Sinclairet al. 1997). More recently, this nucleolar fragmentation has been shown to result from accumulation of extrachromosomal rDNA circles, which form multiple nucleoli (Sinclair and Guarente 1997). These circles also accumulate in strains with a subthreshold level of DNA topoisomerases I and II (Kim and Wang 1989). Given that TOP1 and SIR2 are both required for rDNA silencing (Bryket al. 1997; Smith and Boeke 1997), it is possible that the Sir2p-dependent chromatin structure associated with rDNA silencing could contribute to the counteraction of aging through suppression of rDNA circle excision.
We thank James Broach, Leonard Guarente, Rohinton Kamakaka, Joanna Lowell, Peter Philippsen, and Achim Wach for supplying plasmids, Danesh Moazed and Siyuan Le for comments on the manuscript, and members of the Boeke lab for helpful advice and discussions. L.P. thanks Jasper Rine for his early support (GM54778) of this work, which has been continued with National Insititues of Health (NIH) grant GM54778 (L.P.) J.S.S. is a postdoctoral fellow of the Leukemia Society of America. This work was supported in part by NIH grant CA16519 to J.D.B.
Communicating editor: F. Winston
- Received December 24, 1997.
- Accepted March 23, 1998.
- Copyright © 1998 by the Genetics Society of America