Advertisement

Direct evidence for SIR2 modulation of chromatin structure in yeast rDNA

Christian E. Fritze, Kristin Verschueren, Randy Strich, Rochelle Easton Esposito

Author Affiliations

  1. Christian E. Fritze2,
  2. Kristin Verschueren3,
  3. Randy Strich4 and
  4. Rochelle Easton Esposito1
  1. 1 Department of Molecular Genetics and Cell Biology, The University of Chicago, 920 E. 58th Street, Chicago, IL, 60637, USA
  2. 2 Department of Cell Biology, The Scripps Research Institute, 10550 N. Torrey Pines Rd, La Jolla, CA, 92037, USA
  3. 3 Celgen, Herestraat 49, 3000, Leuven, Belgium
  4. 4 Fox Chase Cancer Center, 7701 Burholme Ave, Philadelphia, PA, 19111, USA

Abstract

The yeast SIR2 gene maintains inactive chromatin domains required for transcriptional repression at the silent mating‐type loci and telomeres. We previously demonstrated that SIR2 also acts to repress mitotic and meiotic recombination between the tandem ribosomal RNA gene array (rDNA). Here we address whether rDNA chromatin structure is altered by loss of SIR2 function by in vitro and in vivo assays of sensitivity to micrococcal nuclease and dam methyltransferase, respectively, and present the first chromatin study that maps sites of SIR2 action within the rDNA locus. Control studies at the MATα locus also revealed a previously undetected MNase‐sensitive site at the a1‐α2 divergent promoter which is protected in sir2 mutant cells by the derepressed a1‐α2 regulator. In rDNA, SIR2 is required for a more closed chromatin structure in two regions: SRR1, the major SIRResponsive Region in the non‐transcribed spacer, and SRR2, in the 18S rRNA coding region. None of the changes in rDNA detected in sir2 mutants are due to the presence of the a1‐α2 repressor. Reduced recombination in the rDNA correlates with a small, reproducible transcriptional silencing position effect. Deletion and overexpression studies demonstrate that SIR2, but not SIR1, SIR3 or SIR4, is required for this rDNA position effect. Significantly, rDNA transcriptional silencing and rDNA chromatin accessibility respond to SIR2 dosage, indicating that SIR2 is a limiting component required for chromatin modeling in rDNA.

Introduction

Eukaryotic organisms maintain stable genomes despite the presence of repeated DNA sequences and efficient homologous recombination. The absence of frequent deletions and translocations due to exchange between repeats, particularly in meiosis where recombination is elevated, suggests the existence of mechanisms to suppress such recombination. Indeed, the amount of meiotic recombination per unit length DNA falls dramatically as the size and repetitive DNA content of eukaryotic genomes increases. For example, although human cells contain ∼150 times more DNA per haploid genome than yeast, both species exhibit similar numbers of reciprocal exchanges per meiosis (Lewin, 1990). The basis for this decline in recombination frequencies is not well understood, but may be explained by such diverse mechanisms as changes in chromosome compaction and altered distribution and/or reduced frequency of recombination initiation sites in the genome.

Our laboratory has investigated the pathways of recombination in diverse regions of the genome. Of particular interest is the phenomenon of recombination suppression in yeast, where reciprocal exchange between the ∼120 tandem repeats of ribosomal RNA genes (rDNA) is ∼100‐fold less frequent in meiosis than exchange between unique genes (Petes and Botstein, 1977; Petes, 1979; Zamb and Petes, 1982). Paradoxically, rDNA sequences have been isolated (HOT1) that when removed from the context of the repeated array actually stimulate recombination between artificial duplications, suggesting that reduced recombination in the rDNA array is not due to a lack of recombination initiation sites (Keil and Roeder, 1984; Voelkel‐Meiman et al., 1987). Moreover, many properties of recombination appear to be similar to exchange in non‐rDNA (Ozenberger and Roeder, 1991). Mutations selected for defects in HOT1‐mediated recombination outside the rDNA display reduced rDNA recombination, strongly suggesting that HOT1 enhances recombination in its natural context (Lin and Keil, 1991). In contrast to the cis‐acting HOT1 element, several trans‐acting factors have been identified that are specifically required to maintain low levels of exchange in these regions (Christman et al., 1988; Aguilera and Klein, 1990; Keil and Lin, 1991; Lin and Keil, 1991; Keil and McWilliams, 1993; Huang and Keil, 1995). For example, inactivation of either Topoisomerase I or II dramatically increases the frequency of mitotic recombination in the array, but not elsewhere in the genome (Christman et al., 1988). Lack of Topoisomerase III function has a similar but broader effect, increasing exchange between rDNA genes and Ty delta sequences (Wallis et al., 1989; Gangloff et al., 1994, 1996). Our studies have demonstrated that one of the genes involved in transcriptional silencing (SIR2) suppresses mitotic and meiotic rDNA recombination 10‐ to 15‐fold, while the other SIR genes (SIR1, SIR3 or SIR4) have no effect on exchange in the array (Gottlieb, 1989; Gottlieb and Esposito, 1989). SIR2 has been shown to be dispensable for HOT1's ability to stimulate artificial repeat recombination (Gottlieb, 1989), but it is not clear whether SIR2 influences HOT1‐mediated recombination within the rDNA array.

The phenomenon of silencing in yeast is characterized by the repression of gene expression in a region‐specific but promoter‐non‐specific manner. Silencing has been observed in yeast at the cryptic mating‐type loci (reviewed in Laurenson and Rine, 1992) and at telomeres (where it is designated telomeric position effect; Gottschling et al., 1990). In both cases, repression is associated with specialized chromatin structures whose integrity depends upon a complex combination of cis‐acting sites, several shared trans‐acting factors, and histones. The SIR1, SIR2, SIR3 and SIR4 loci were the first genes identified as required for silencing (Haber and George, 1979; Klar et al., 1979; Ivy et al., 1986; Rine and Herskowitz, 1987). These genes prevent the expression of mating information contained at the cryptic HM loci and regulate the directionality of gene conversion events from the silent loci to the expressed MAT locus. The products of the SIR2, SIR3 and SIR4 loci are now known to interact with one another as well as histones H3 and H4 to form productive ‘silencing complexes’ at the HM loci and telomeres (Moretti et al., 1994; Hecht et al., 1995, 1996; Strahl et al., 1997). Transcriptional silencing at each of the HM loci also depends upon a pair of cis‐acting elements, designated silencers, which are themselves made up of partially redundant regulatory sites (Brand et al., 1985, 1987) and contain an ARS element, RAP1‐binding site (Shore et al., 1987) and ABF1‐binding site (Diffley and Stillman, 1988). Telomeres also contain ARS elements and a RAP1‐binding site (Brand et al., 1987; Buchman et al., 1988), although the ARS elements are contained in the Y′ regions which are dispensable for telomeric silencing.

The idea that expression of silenced loci is regulated by chromatin structure stems primarily from findings that: (i) mutations which abolish silencing produce more open chromatin configurations at the derepressed loci (Nasmyth, 1982; Gottschling, 1992; Singh and Klar, 1992); and (ii) loss of silencing can be generated by histone mutations (Kayne et al., 1988; Thompson et al., 1994b) or mutations which affect histone modifications (Braunstein et al., 1993). Studies by Nasmyth first demonstrated that SIR function is required for the formation of a specialized chromatin structure at the HM loci which is not observed at MAT (Nasmyth, 1982). Since then, expression in yeast of Escherichia coli dam methyltransferase has shown that repression at the cryptic mating‐type loci and telomeres is tightly correlated with the presence of a methyltransferase‐resistant chromatin structure (Gottschling, 1992; Singh and Klar, 1992). These findings demonstrate that chromatin structure at repressed loci can restrict access by dam methyltransferase and, by implication other important proteins, to the DNA.

In this study we have investigated the basis of SIR2‘s suppression of rDNA exchange. We previously proposed that SIR2 limits the accessibility of the rDNA array to the general recombination machinery, based on the finding that RAD50 and RAD52, two genes not normally required for rDNA recombination, are needed for the elevated levels exhibited in the sir2 mutant background (Gottlieb et al., 1989). SIR2's previously defined role in silencing at the cryptic mating type loci further suggested that this limited accessibility might be dependent upon a specialized chromatin structure dependent upon Sir2p which excludes the general recombination system in mitosis and meiosis. Here we report direct evidence that rDNA chromatin structure is responsive to the presence of SIR2. Furthermore, we demonstrate that rDNA chromatin structure and an associated rDNA position effect are responsive to SIR2 dosage. Thus, Sir2p is a limiting component for silencing and recombination in the rDNA, unlike the other known contexts of SIR activity. These findings complement several very recent reports of other rDNA position effects, including Ty transposition into the non‐transcribed spacer region, effects on transcription, and localization of Sir2p to the nucleolus (Bryk et al., 1997; Gotta et al., 1997; Smith and Boeke, 1997) (see Discussion). Together, these findings indicate that SIR2 reduces the accessibility of rDNA sequences by regulating chromatin structure in the rDNA, presumably by altering the proportion of rDNA repeats bound by nucleosomes or regulating an aspect of higher‐order chromatin configuration. The data strongly support the idea that SIR2‐dependent alterations in chromatin structure are responsible for both suppression of recombination and transcriptional silencing in the rDNA.

Results

Identification of a position effect in the rDNA regulated by SIR2

To address whether SIR2 regulates gene expression in the rDNA, a double ADE2‐CAN1 marker was constructed allowing selection for the presence of the marker or counter selection for its excision from rDNA. The marker was integrated into the rDNA array (Figure 1) of strains containing disruptions of the chromosomal ADE2 and CAN1 loci. Expression of ADE2 allows growth on adenine omission medium (−ADE) and gives white colonies on either −ADE or COM (complete) medium. Loss or reduced expression of the ADE2 gene is detected by the appearance of red or pink colonies, respectively, on COM medium due to the accumulation of an adenine biosynthetic intermediate when ADE2 is blocked (Roman, 1956). CAN1 confers dominant sensitivity to the arginine analog canavanine (CanS) (Grenson et al., 1966; Hoffmann, 1985). Loss of CAN1 function allows resistance to canavanine (Canr). As with all unique markers inserted into the rDNA, a fraction of cells in an unselected culture lose the integrated marker via genetic recombination between rDNA repeat units flanking the insertion. Thus, plating cells containing the rDNA::ADE2‐CAN1 marker onto media containing canavanine (CAN media) should only permit growth of red Canr colonies which have lost the rDNA marker by recombination. These red Canr recombinants should not grow on media with canavanine but lacking adenine, since they are adenine auxotrophs.

Figure 1.

(A) Integration of the ADE2‐CAN1 and ADE2 markers into the rDNA array. The ADE2‐CAN1 marker was integrated at the KpnI site in the rDNA of CFY558 and related strains. The unrelated strain REE1285 contains only a single ADE2 insert at the HindIII site near the 5′ end of the 18S coding region. Arrows within the ADE2 and CAN1 genes indicate transcriptional orientation. ADE2 and CAN1 genes not to scale. Hash marks and gray shaded regions at the top of the figure mark the positions of the major SIR2Responsive Regions (SRR1 and SRR2) detected by MNase (see Figures 6 and 8B). E, enhancer; RFB, replication fork barrier; P, promoter; ARS, matches to the ARS consensus sequence; TOPI, REB1, ABF1, RAP1, Topoisomerase I, REB1, ABF1 and RAP1 consensus binding sites. The HOT1 element is composed of the enhancer and promoter regions (E + P). (B) Strategy for mapping rDNA chromatin with MNase. The restriction fragments analyzed by micrococcal nuclease assays are indicated (thin lines) together with the position of probes used for indirect end‐labeling (gray boxes). Restriction sites are AvaI (AI), AvaII (AII), EcoRI (R) and PvuII (Pv). Not all restriction sites for these enzymes are shown.

During the course of our studies with the ADE2‐CAN1 marker we discovered allelic variants of a single locus, independent of CAN1, affecting the resistance of cells to canavanine. The dominant allele, which we called CSS1, confers resistance to canavanine, while the recessive allele, css1, results in super‐sensitivity. The CSS1 gene was cloned and shown to be an essential gene; its detailed characterization will be presented elsewhere (C.Fritze and R.Esposito, in preparation). Recent work by others indicates that it encodes a ubiquitin ligase, suggesting that super‐sensitivity of the mutant is due to a failure of protein turnover and the accumulation of canavanyl‐containing proteins. The present study took advantage of the behavior of each of these alleles to demonstrate in different assays that SIR2 reduces gene expression in the rDNA. For example, as described further below, the combination of CSS1 and SIR2 together led us initially to detect silencing of the CAN1 marker in rDNA, since the level of canavanine resistance resulting from the presence of both wild‐type alleles permits sufficient growth on canavanine‐containing medium to detect normal colonies even in the presence of the wild‐type CAN1 gene encoding sensitivity. The css1 allele, on the other hand, proved very useful in examining SIR2 dosage effects. Strains containing this allele, which are exquisitely sensitive to CAN‐containing medium, facilitate the detection of increased CAN1 gene silencing in rDNA, resulting in canavanine resistance beyond the already high level seen in CSS1 SIR2 strains.

Figure 2 summarizes the phenotypes of each of the CSS1 alleles with respect to the ADE2‐CAN marker in rDNA. Strains carrying css1 and the ADE2‐CAN1 marker which are super‐sensitive to CAN, give no background colony growth on medium containing canavanine. Only rare red recombinants which have excised the marker and are now can1 ade2 grow on this medium. Strains containing CSS1, on the other hand (including W303), exhibit a slow‐growing white Canr phenotype with occasional red Canr papillae on CAN media. In different CSSI backgrounds, 25–100% of colony forming units (cfu) are capable of growth on −ADE+CAN media despite the presence of ADE2‐CAN1 in the rDNA (see also Figure 3, strain 2). Red colonies generate only stable red colonies when replated, while white Canr colonies generate predominantly white colonies. Loss of ADE2‐CAN1 from rDNA in red colonies and its presence in white colonies was confirmed by Southern analysis (unpublished data). These findings initially suggested to us that the CAN1 gene in the rDNA::ADE2‐CAN1 marker might be silenced. However, the CAN1 gene in the rDNA marker could not be fully silenced, since wild‐type cells were still responsive to canavanine with respect to viability, i.e. viability decreased as the level of canavanine in the plating media was increased (unpublished data). As reported in more detail below, this effect on CAN1 expression is only observed when the marker is integrated in rDNA, and not in non‐rDNA. Thus, the phenotype seen in CSS1 strains is not simply due to the presence of an independent suppressor of canavanine sensitivity in this strain, but reflects properties of rDNA and the presence of other factors (e.g. SIR2) on expression.

Figure 2.

Diagrammatic summary of CAN plating phenotypes of rDNA::ADE2‐CAN1 populations. Open and filled circles represent white and red colonies, respectively.

Figure 3.

Effect of SIR2 on expression on the ADE2‐CAN1 marker in the rDNA. Log phase cultures in liquid medium were 10‐fold serially diluted and plated in parallel on the indicated media. Strains 1–6 carry a single copy of ADE2‐CAN1 at the unique KpnI site in the rDNA. Strains 7 and 8 carry a single ADE2 insertion at the HindIII site in the 35S rRNA coding region. Strains 9 and 10 contain no rDNA insertions, but carry a centromeric plasmid with a single copy of rDNA and the ADE2‐CAN1 marker. Strains 11–13 carry no rDNA insertions and have the ADE2‐CAN1 marker integrated at the Δade2 locus. Media is −URA for samples 1–8 and −LYS −URA for samples 9–13 to select for plasmid retention. Strain 1: CFY559 [YCp‐URA3], strain 2: CFY559 [YCp‐URA3‐CSS1], strain 3: CFY559Δsir2 [YCp‐URA3], strain 4: CFY559Δsir2 [YCp‐URA3‐CSS1], strain 5: CFY559Δsir2 [YEp‐URA3‐SIR2], strain 6: CFY520Δsir4 [YCp‐URA3‐CSS1], strain 7: REE1285 [YEp‐URA3], strain 8: REE1285 [YEp‐URA3‐SIR2], strain 9: CFY558 [YCp‐URA3‐rDNA‐ADE2‐CAN1, YCp‐LYS2‐CSS1], strain 10: CFY558Δsir2 [YCp‐URA3‐rDNA‐ADE2‐CAN1, YCp‐LYS2‐CSS1]. Strain 11: CFY570 [YEp‐URA3, YCp‐LYS2], strain 12: CFY570 [YEp‐URA3, YCp‐LYS2‐CSS1], strain 13: CFY570 [YEp‐URA3‐SIR2, YCp‐LYS2‐CSS1].

To determine whether any of the SIR genes are required for the reduced expression of the CAN1 marker in the rDNA, each of the four SIR genes was individually deleted or expressed in high copy in the parental strain containing ADE2‐CAN1 inserted in the rDNA. In a wild‐type strain nearly all cells form colonies on media containing canavanine, and the majority of colonies on CAN media are white, indicating the presence of the rDNA::ADE2‐CAN1 marker (Figure 3, strain 2). However, in the Δsir2 strain, white Canr colonies indicative of reduced CAN1 expression are reduced 10‐ to 100‐fold (Figure 3, strain 4). This phenotype is more clearly visible when SIR2 and Δsir2 strains are compared on −ADE+CAN medium which excludes the growth of red Canr recombinants caused by increased rDNA recombination in the Δsir2 strain (Gottlieb and Esposito, 1989). These results demonstrate that SIR2 function is required for reduced expression of the CAN1 marker in rDNA. In contrast to SIR2, loss of SIR4 (Figure 3, strain 6), SIR1 or SIR3 (unpublished data) produced no detectable difference in expression of ADE2 or CAN1 as judged by colony color and plating efficiency on CAN media.

Dosage‐dependent reduction of gene expression by SIR2

Since a SIR2 strain displayed reduced expression of CAN1 in rDNA::ADE2‐CAN1 as compared with an isogenic Δsir2 strain, we asked whether increased SIR2 dosage further lowers CAN1 expression. In order to detect any potential increase in CAN1 repression by high‐copy expression of SIR2, we employed an isogenic strain which lacks the CSS1 episome to increase the canavanine‐sensitivity of this background. Less than 1×10−4 cfu of the css1‐1 strain form colonies on −ADE+CAN media. (Figure 3, strain 1). Significantly, introduction of a multicopy SIR2 plasmid into this strain restored 100% plating efficiency on −ADE+CAN media, indicating that 2μSIR2 reduces CAN1 expression below the level observed with wild‐type SIR2. Additionally, high‐copy expression of SIR2 resulted in pink colonies, indicating that the ADE2 marker is now also poorly expressed under conditions of high SIR2 dosage (Figure 3, strain 5). Repression by SIR2 is not specific to the particular ADE2‐CAN1 construction, since pink colonies, indicating lowered expression, also result from transformation of 2μSIR2 into strain REE1285 (Figure 3, strain 8), which bears a different ADE2 insert (at the HindIII site in the 35S gene; Figure 1). The absence of colony color differences between SIR2 and Δsir2 genotypes, in light of observable changes in canavanine‐resistance in the same strains, indicates that the threshold of expression for detecting a phenotypic difference is not the same for the ADE2 and CAN1 markers.

Since overexpression of SIR2 represses ADE2 and further reduces CAN1 expression in the rDNA, we determined if this hyper‐repressed phenotype requires the other SIR genes by overexpressing SIR2 in the original panel of SIR deletion strains. The absence of the other SIR products does not alter the pink colony color indicative of reduced ADE2 expression in the rDNA::ADE2‐CAN1 marker, nor does it alter the CAN plating phenotype of 2μSIR2 strains (Fritze, 1994). SIR1, SIR3 and SIR4 are thus dispensable for the reduced expression of the rDNA::ADE2‐CAN1 marker in both wild‐type and 2μSIR2 strains. To confirm that SIR2 repression is specific to the rDNA, we showed that colony color or canavanine sensitivity is unaffected by increased dosage of SIR2 when the ADE2‐CAN1 marker is maintained on a plasmid or integrated elsewhere in the genome (Figure 3, strains 9–13). These studies thus demonstrate that: (i) SIR2 represses both ADE2 and CAN1 in the rDNA; (ii) repression of CAN1 and ADE2 responds to SIR2 gene dosage; and (iii) SIR2‐dependent repression is independent of the SIR1, SIR3 or SIR4 functions. Although a correlation between increased SIR2 dosage and hypoacetylation of histone pools has been shown (Braunstein et al., 1993), these results constitute the first demonstration in yeast of an association between SIR2 overexpression and reduced gene expression.

SIR2 silencing in the rDNA is associated with a small reduction in ADE2 and CAN1 RNA levels

We confirmed that repression of rDNA::ADE2‐CAN1 correlated with a reduction in transcription using S1 nuclease protection assays to quantitate precisely steady‐state levels of CAN1 RNA in SIR2 and Δsir2 cells. Unexpectedly, Pol I read‐through transcripts were detected which originated in the rDNA from the upstream 35S rRNA promoter, extending through the ADE2 and CAN1 markers (see Materials and methods). We determined that SIR2‐dependent silencing of the genes inserted in the rDNA is not dependent on the anomalous transcripts by examining silencing in strains in which ADE2 was integrated into the rDNA in the opposite orientation from the 35S transcript to eliminate formation of read‐through transcripts from the upstream 35S promoter (see Figure 1). Only RNAs with the wild‐type RNA start site were transcribed in the single marker rDNA::ADE2 integrant, and high‐copy SIR2 still reduced expression of the ADE2 marker (Figure 3, strain 8). Therefore, our analysis of transcript levels focused on the wild‐type ADE2 and CAN1 messages. The presence of a single copy of SIR2 results in a small but measurable average decrease in ADE2 and CAN1 RNA levels, which was observed in several independent trials of this experiment (Figure 4B). High‐copy SIR2 expression further reduces the level of CAN1 and ADE2 mRNAs ∼40% as compared with wild‐type (Figure 4C). The same trend was discerned under conditions of increased SIR2 dosage with ADE2 RNA levels in the integrant containing the independent ADE2 insertion in opposite orientation to the 35S transcript which produces only wild‐type transcripts (Figure 4D). No SIR2‐dependent changes in ADE2‐CAN1 RNA levels are seen when ADE2‐CAN1 is located outside the rDNA at the ADE2 locus (Figure 4A.) These data indicate that the amount of ADE2‐CAN1 marker transcription in the rDNA varies inversely with increasing SIR2 dosage, in parallel with SIR2's phenotypic effects on gene expression.

Figure 4.

Transcription of rDNA markers in the presence or absence of SIR2. Equal amounts of total RNA from duplicate samples prepared from independent colony isolates were analyzed with CAN1, ADE2 and HIS3 probes by S1 nuclease protection analysis as described. Arrows labeled p, r and wt indicate positions of residual probe, read‐through transcript and wild‐type transcript, respectively. RNA amounts were quantitated by phosphorimagery as described in Materials and methods. To control for loading variation, the CAN1 and ADE2 RNAs were normalized to the cluster of bands as shown above, which were quantitated as a unit, representing different transcription termination sites detected by the 3′‐HIS3 probe. Each bar graph represents quantitation for the S1 experiment shown below it. Error bars represent the standard deviation of the data. (A) Transcription of the ADE2‐CAN1 marker integrated outside the rDNA at the Δade2 locus. Δsir2 = CFY570 [YEp‐URA3], 2 μSIR2 = CFY570 [YEp‐URA3‐SIR2]. (B) Transcription of the rDNA::ADE2‐CAN1 marker in SIR2 versus Δsir2 strains. Δsir2 = CFY559Δsir2 [YCp‐URA3‐CSS1], SIR2 = CFY559[YCp‐URA3‐CSS1]. (C) Transcription of rDNA::ADE2‐CAN1 in SIR2 versus 2μSIR2 strains. SIR2 = CFY559 [YEp‐URA3], 2μSIR2 = CFY559[YEp‐URA3‐SIR2]. (D) Transcription of an ADE2 insertion which produces no read‐through RNA. In lanes 1–3, SIR2 = CX100‐2C[YEp‐URA3], 2μSIR2 = CX100‐2C[YEp‐URA3‐SIR2]. In lanes 4–7, SIR2 = CX100‐6D[YEp‐URA3], 2μSIR2 = CX100‐6D[YEp‐URA3‐SIR2].

Loss of SIR2 function alters chromatin structure at both HMR and MAT

SIR2‐specific features in rDNA chromatin were monitored by MNase cleavage and indirect end‐labeling (Wu, 1980; see also Materials and methods). As a control for our studies of MNase sensitivity in rDNA, the same chromatin preparations used for the analysis of rDNA were used to visualize chromatin features at HMR and MAT. It has previously been demonstrated (Nasmyth, 1982) that loss of SIR2 function causes significant changes in the MNase cleavage pattern of silent mating‐type chromatin (HML and HMR), but not of chromatin at the actively transcribed MAT locus. Here, in agreement with the earlier studies of Nasmyth (1982), the MNase cleavage profiles of HMRa chromatin prepared from SIR2 and Δsir2 cells are significantly different across the X region and near the promoter required for a1 and a2 transcription (Figure 5A), but remain unchanged at the MATα HO cleavage site at the Y/Z junction (Figure 5B). Unexpectedly, however, our data revealed a previously undetected SIR2‐dependent change at MAT, specifically in the promoter region of MATα. This site is cleaved by MNase in the SIR2 but not sir2 strain (Figure 5B, lanes 1–10).

Figure 5.

(A) Micrococcal nuclease cleavage of HMRa chromatin in SIR2 and Δsir2. Chromatin from SIR2 (strain YK18, lanes 1–5), and Δsir2 (strain YK4, lanes 6–10) strains, or naked DNA (lanes 11–12) was partially digested with micrococcal nuclease, treated with proteinase K and phenol:chloroform to remove protein, digested with HindIII, transferred to membrane and hybridized to a BamHI–PstI HMRa probe as described in Materials and methods. ● indicates positions at which MNase cleavage differs between SIR2 and Δsir2. X and Z represent sequence elements at HMRa, arrowheads show the a1 and a2 transcripts. Numbers to the right of the figure correspond to the size in bp of markers co‐electrophoresed alongside the sample lanes. (B) MNase cleavage of MATα chromatin. MNase digestion and other treatments as described above, but with hybridization to a HindIII–NruI probe to MATα. Lanes 1–5 are nuclei from SIR2 (YK18), lanes 6–10 from Δsir2 (YK4) and lanes 11–15 from Δsir2 in which HMRa has been deleted/disrupted (YK52). ● at left edge of photograph indicates the position of a band which appears in chromatin from SIR2 and Δsir2 hmra::URA3 nuclei but not Δsir2 nuclei. W, X and Z represent sequence elements at MATα; arrowheads show the α1 and α2 transcripts. Naked DNA profiles show MNase cleavage sites at the same positions as lanes 1–5 and 11–15 (Nasmyth, 1982; also unpublished data).

Transcription at MATα1, in contrast to MATa, is repressed by the a1‐α2 repressor in diploid cells via binding of the repressor to an operator site within the MATα1 promoter (reviewed in Herskowitz, 1990). Because silent mating‐cassette derepression in sir2 cells permits formation of the a1‐α2 repressor, we tested whether the change in chromatin structure in the promoter region of MATa depends on the presence of this repressor. Disruption of HMRa to eliminate the a1‐α2 repressor showed that the promoter site at MATα is still cleaved in a sir2 strain lacking the a1‐α2 repressor (Figure 5B, lanes 11–15). Therefore, the protection of a MNase cleavage site in MATα sir2 chromatin is indeed dependent upon the presence of the a1‐α2 repressor, perhaps due to binding of the repressor to its target sequence in the MATα1 promoter.

SIR2 reduces accessibility of rDNA chromatin to nuclease digestion in the non‐transcribed spacer

A physical map of the rDNA repeat unit with relevant restriction sites and sequence elements is shown in Figure 1. Each rDNA repeat consists of 9080 bp of DNA and gives rise to two primary transcripts: a 5S rRNA (Pol III) and a 35S precursor RNA (Pol I) which is processed into 5.8S, 18S and 25S rRNA species. Adjacent 35S coding regions are separated by a non‐transcribed spacer (NTS) which is sub‐divided by the centrally located 5S gene into two ∼1 kb regions, NTS1 and NTS2 (reviewed in Planta and Raue, 1988). NTS1 contains an enhancer for Pol I transcription (E), while NTS2 contains a promoter (P), an ARS element that has been shown to function as an origin of replication in vivo (Skryabin et al., 1984; Linskens and Huberman, 1988), and a binding site for Topoisomerase I (TOPI) (Bonven et al., 1985). A consensus binding site for the transcription factor RAP1 is present in the 25S rRNA coding sequence (Buchman et al., 1988). The HOT1 recombination initiator is composed of the enhancer and promoter for 35S transcription. The NTS1 element also contains a replication fork barrier near the end of the 35S transcription unit which impedes the progress of replication forks which travel in the opposite direction than 35S transcription (Brewer and Fangman, 1988).

Our analysis of MNase sensitivity in the rDNA NTS is presented in Figure 6. The MNase cleavage pattern of wild‐type (SIR2) chromatin differed from naked DNA at eight distinct positions within the NTS1–5S–NTS2 region. Six of these positions are cleaved with reduced efficiency in chromatin as opposed to naked DNA and represent sites of chromatin‐specific protection from MNase (Figure 6, closed circles). The remaining two positions constitute cleavage sites in chromatin which are not present in naked DNA and thus define chromatin‐specific enhanced cleavages (Figure 6, open circles).

Figure 6.

MNase cleavage sites in chromatin of the rDNA non‐transcribed spacer (NTS). MNase digestion and other treatments as described in Figure 5, but digestion with EcoRI and hybridization to probe P6. MNase concentration increases from left to right within each series; lanes 1 and 5 were not treated with MNase. The line diagram at left depicts features of the rDNA gene in this region. ● and ○ indicate MNase cleavages identified as sites of chromatin‐specific protection and enhancement, respectively (see text). These sites constitute SRR1. E, enhancer; gray box and arrowhead labeled 5S represent the positions of coding region and transcript for the 5S rRNA gene. Circled numbers 1 and 2 denote DpnI restriction sites referred to later in the text.

Chromatin prepared from the Δsir2 strain is more sensitive to MNase cleavage at all six sites of chromatin‐specific protection, indicating that loss of SIR2 function reduces the extent of protected chromatin in this region of the rDNA. The differences between Δsir2 and SIR2 chromatin are restricted to a 0.9 kb region starting 300 bp downstream of the 5′ EcoRI site at the border of the enhancer and encompass the 5S rDNA coding region. This region represents the major location of SIR2‐dependent alterations we detected in the entire rDNA, and is here designated SRR1 (SIR2 Responsive Region 1). The same SIR2‐dependent alterations representing SRR1 are also detected by mapping with respect to the EcoRI restriction site at the 5′ end of NTS1 (unpublished data). Loss of SIR2 function does not lead to the elimination of all MNase‐detectable chromatin structure in this region however, since the two sites of chromatin‐specific enhanced cleavage (Figure 6, open circles), are not affected by the SIR2 genotype. These data demonstrate that the loss of SIR2 function either alters the structure of chromatin in SRR1 to allow increased MNase cleavage, or results in a decrease in the proportion of rDNA repeat units which are organized into a chromatin structure that protects the rDNA NTS from MNase cleavage.

Because NTS1 contains the Pol I enhancer and promoter that comprise the HOT1 element as well as a number of other cis‐acting sequences, chromatin features in these regions were examined more closely. No reproducible differences in MNase sensitivity were detected between SIR2 and Δsir2 chromatin in the enhancer or promoter (Figure 7).

Figure 7.

(A) MNase sensitivity of chromatin in the rDNA enhancer region. MNase digestion and other treatments as described in the legend to Figure 5, but with AvaII digestion and hybridization to probe P3. Lanes 1–3 are nuclei from SIR2, lanes 4–5 from Δsir2 and lanes 6–7 are naked DNA. ○ indicate MNase cleavages identified as sites of chromatin‐specific enhancement. The increased intensity of the band in lanes 4 and 5 near the edge of the enhancer (E) was not observed in two subsequent repetitions of this experiment. E, enhancer; gray box and arrowhead denote the 3′‐end of the 25S coding region and 35S transcript, respectively. (B) MNase sensitivity of chromatin in the 35S rRNA promoter. MNase digestion and other treatments as described in Figure 5, but with PvuII digestion and hybridization to probe P3. ● and ○ indicate MNase cleavages identified as sites of chromatin‐specific protection and enhancement, respectively (see text). ARS, ARS consensus sequences; TOP, Topoisomerase I binding site; P, promoter; the heavy black line indicates the 5′ end of the 35S transcript.

The 35S transcription unit contains few chromatin features as measured by MNase

The MNase cleavage profile of the chromatin preparations in the remainder of the rDNA repeat unit, including the 35S coding region, is similar to that of the naked DNA samples (Figure 8). In the 25S coding region, only two sites of chromatin‐enhanced MNase cleavage are detected, and no changes between the MNase cleavage profile of SIR2 and Δsir2 chromatin are visible (Figure 8A). In the 18S coding region a number of chromatin‐specific features are detected (Figure 8B); however, these features are considerably weaker than the differences identified in the NTS. These included two sites of chromatin‐enhanced MNase cleavage (Figure 8B, open circles) whose intensity increases by loss of SIR2 function and a site at which chromatin is more readily cleaved in Δsir2 chromatin compared with both wild‐type chromatin and naked DNA (Figure 8B, filled diamond). A series of complex sir2‐dependent changes can also be seen which we could not unambiguously ascribe to chromatin‐specific enhancement or protection (Figure 8B, filled square).

Figure 8.

(A) MNase sensitivity of the 25S rRNA coding region. MNase digestion and other treatments as described in Figure 5, but with AvaI digestion and hybridization to probe P2. Lanes 1–3 are nuclei from SIR2, lanes 4–5 from Δsir2 and lanes 6–7 are naked DNA. ○ indicate MNase cleavages identified as sites of chromatin‐specific enhancement (see text). RAP1, RAP1 binding site; gray box, 25S rDNA coding region; arrowhead, 35S transcript. (B) MNase sensitivity of chromatin in the 18S and 5.8S rRNA coding regions. MNase digestion and other treatments as described in Figure 5, but with AvaI digestion and hybridization to probe P6. Lanes 1–3 are nuclei from SIR2, lanes 4–5 from sir2 and lanes 6–7 are naked DNA. ○ indicate MNase cleavages identified as sites of chromatin‐specific enhancement. ♦ indicates the position of a site whose cleavage is enhanced in Δsir2 chromatin as compared with both wild‐type chromatin and naked DNA. The filled bar indicates the position of unclassified sir2‐specific features. These sites constitute SRR2. Gray boxes, 18S and 5.8S rRNA coding regions; black line, 35S transcript.

Overall, our analysis of the 35S rRNA region by MNase treatment revealed large stretches of the rDNA which are devoid of positioned nucleosomes, consistent with the high level of transcription in this region (Dammann et al., 1993). Several chromatin‐specific features occur in the 18S coding region, and at a subset of these sites chromatin prepared from Δsir2 cells is more readily cleaved by MNase than chromatin from SIR2 cells. The SIR2‐dependent chromatin alterations in the 18S region are designated SRR2 (SIR2 responsive region 2).

sir2‐specific changes in rDNA chromatin are independent of the a1‐α2 repressor

As indicated earlier, derepression of the silent mating cassettes in a sir2 background permits expression of the a1‐α2 regulatory molecule. To determine whether the sir2‐specific changes in rDNA NTS chromatin were a primary defect of Δsir2 or an indirect consequence of the presence of the a1‐α2 repressor (as was found at MATα), we eliminated the a1‐α2 repressor from the Δsir2 strain by disrupting HMRa as described earlier. Chromatin from the Δhmra strain was visualized by MNase treatment as before. The sir2‐dependent changes in NTS chromatin are still visible in the Δhmra Δsir2 strain (unpublished data), indicating that they do not result indirectly from a1‐α2 expression.

SIR2 restricts accessibility of the rDNA NTS to modifications in vivo

The in vitro analysis of purified nuclei described above demonstrates that loss of SIR2 function increases the accessibility of MNase to chromosomal DNA in the rDNA NTS1/5S region, and to a lesser extent, in the 18S rRNA coding region. In order to determine whether these results reflect changes in rDNA chromatin accessibility in vivo, we examined whether sequences in the rDNA NTS1/5S region of SIR2 and Δsir2 strains are differentially susceptible to methylation by heterologously expressed E.coli dam methyltransferase. Modification of adenosine residues in the sequence GATC by dam methyltransferase constitutes a silent, stable in vivo modification which can be easily detected by Southern analysis with restriction enzymes that discriminate between methylated and unmethylated GATC sequences (Gottschling, 1992; Singh and Klar, 1992). This assay measures directly the extent to which chromatin structure restricts in vivo access to the DNA template and thus provides a convenient method to extend and confirm the observations made with MNase in vitro.

Two GATC sites in the NTS1 were chosen for this analysis. The first GATC site (Position 1) lies near a site of chromatin‐specific protection from MNase that is more efficiently cleaved in a Δsir2 background than in wild‐type (see Figure 6). The second GATC site (Position 2) is in a region of NTS1 that is efficiently cleaved by MNase regardless of SIR2 genotype (Figure 6). If SIR2 regulates the accessibility of rDNA chromatin near Position 1, it should be more efficiently methylated in a Δsir2 strain than in wild‐type and be detected by the presence of a 1331 bp EcoRI–DpnI restriction fragment. Lack of methylation gives rise to a 1519 bp EcoRI–PvuII fragment. The exclusive presence of the 1519 bp EcoRI–PvuII fragment (Figure 9B, lane 4) and the 1331 bp EcoRI–Sau3A fragment (Figure 9B, lane 3) in DNA from a control strain lacking the dam integration confirms the presence of the restriction sites and demonstrates the lack of significant DpnI cleavage in the absence of dam expression. Methylation of Position 2 gives rise to a 670 bp EcoRI–DpnI fragment. Its appearance is predicted to be unaffected by SIR2 function. Analysis of DNA samples in the control dam background once again confirms the restriction pattern and lack of background methylation/DpnI cleavage (Figure 9B, lanes 3 and 4).

Figure 9.

Accessibility of rDNA chromatin to heterologously expressed dam methyltransferase. (A) Schematic map of the rDNA NTS. R, P and D mark the positions of EcoRI, PvuII and DpnI–Sau3AI restriction sites, with corresponding restriction fragment sizes indicated underneath. Circled numerals 1 and 2 refer to the DpnI sites monitored in this experiment (also marked in Figure 6). SRR1, SIR2 responsive region 1; E, enhancer; P, promoter; A, ARS consensus sequences. Boxes labeled P2 and P6 represent probe sequences. (B) Analysis of dam methylation in SIR2, Δsir1, Δsir2 and Δsir4 strains. Genomic DNA was prepared from mid‐log cultures in YPDA medium and digested with either EcoRI, PvuII (R,P, lane 1), EcoRI, PvuII and Sau3A (R,P,S, lanes 2, 3) or EcoRI, PvuII and DpnI (R,P,D, lanes 4–12). Methylation at Position 1 was detected with probe P6; methylation at Position 2 was detected after stripping the membrane and re‐hybridization with probe P2. M, size standards; SIR2, DNA prepared from two independent isolates of UCC1021; SIR2 (dam‐), UCC1001; Δsir1, UCC1044; Δsir2, UCC1042; Δsir4, UCC1045. Lanes 1–5 and 6–12 correspond to different exposures from two regions of the same membrane. (C) Analysis of dam methylation in SIR2, Δsir2 and 2μSIR2 strains. Duplicate SIR2 and Δsir2 samples represent DNA samples purified from independent cultures on separate days. In lanes 9–11, DNA was prepared from mid‐log cells grown in −URA medium and digested with EcoRI, PvuII and DpnI. SIR2, UCC1023 transformed with pRS426 (YEp‐URA3); 2μSIR2, independent isolates of UCC1023 transformed with pJR68 (YEp‐URA3‐SIR2). (D) Quantitation of results for Position 1 presented in (C). Band intensities were measured by scanning and densitometry as described. Values for isolates with the same genotype were expressed as the proportion of methylated sites to total and averaged.

The ability of the dam methyltransferase to modify Position 1 and Position 2 in response to the SIR genotype is shown in Figure 9B (lanes 5–12) and Figure 9C. Less than 40% of GATC sites at Position 1 are methylated in a wild‐type background, indicating restricted access to the sequence by dam methyltransferase. Loss of SIR1 or SIR4 function does not alter the pattern of DpnI cleavage at Position 1 as compared with wild‐type. However, when Position 1 is assayed in a Δsir2 background, >80% of sites were methylated. As predicted from the MNase studies, loss of SIR2 function increases accessibility of this site to methyltransferase. The increased accessibility is due specifically to loss of SIR2 function and not to derepression of the silent mating‐cassettes since increased accessibility is not observed when the SIR1 or SIR4 gene is deleted. In contrast to Position 1, ∼90% of sites at Position 2 are methylated regardless of SIR2 genotype, as expected. Once again, loss of SIR1 or SIR4 does not significantly alter accessibility at Position 2. These results demonstrate that mutations in SIR2 increase accessibility to the formerly protected Position 1 site in vivo as suggested by the MNase studies. Intriguingly, deletion of SIR4 leads to a slight increase in protection of Position 1 in this experiment. This can be explained by the recent finding that delocalization of Sir2p from telomeres in a sir4 background leads to increased accumulation of Sir2p in the nucleolus (Gotta et al., 1997). In summary, SIR2 function (but not SIR1 or SIR4 function) is required in vivo for the maintenance of a more closed rDNA chromatin structure. This finding parallels the results of previous genetic data that SIR2 (but not SIR1, SIR3 or SIR4) suppresses rDNA recombination (Gottlieb and Esposito, 1989) and agrees with our finding that SIR2 is required for reducing expression of marker genes integrated into the rDNA array.

rDNA accessibility responds to increased dosage of SIR2

Since SIR2 chromatin has a more closed configuration than Δsir2 chromatin, we examined whether increased SIR2 dosage further reduces rDNA accessibility. The dam methyltransferase protection assay was thus repeated with a strain bearing a high‐copy SIR2 plasmid (2μSIR2) (Figure 9C, lanes 9–11). In the SIR2 background, the DpnI site Position 1 in the NTS is methylated at 44% of GATC sites, in close agreement with the results from the previous experiment (38%). When SIR2 dosage is increased via expression from a 2μ plasmid, only 25% of sites are methylated. Therefore, high‐copy SIR2 expression results in greater protection of rDNA near Position 1 as compared with a wild‐type background. No effect is observed at Position 2. These data show that rDNA accessibility in the rDNA NTS decreases with increasing SIR2 dosage, indicating that Sir2p is a limiting component or regulatory function required for closed chromatin structure in this region.

Discussion

A SIR2‐dependent position effect in the rDNA

At the silenced HM mating loci and telomeres, SIR2, the other three SIR genes and additional factors are all required for the maintenance of specialized chromatin structures and inhibit transcription in a promoter‐non‐specific and position‐dependent manner. As such, the SIR genes define a set of region‐specific regulators that inhibit gene expression of a chromosomal ‘neighborhood’ as contrasted with repression of individual genes (reviewed in Laurenson and Rine, 1992).

In this study we have demonstrated that SIR2 also regulates region‐specific gene expression in the rDNA repeated array, as has recently been found by (Bryk et al., 1997; Smith and Boeke, 1997). We also present the novel finding that SIR2 function regulates chromatin structure by direct assays of MNase sensitivity and in vivo dam methyltransferase accessibility. At least two regions of the rDNA, designated SRR1 and SRR2, are affected by the presence of Sir2p, resulting in a more closed chromatin structure. In all nine positions where SIR2‐dependent changes could be clearly identified, loss of SIR2 function led to increased cleavage by MNase. The increased susceptibility of these regions to MNase in vitro reflects a change in chromatin structure at these sites which may result from one of a number of diverse mechanisms, including changes in nucleosome positioning or occupancy, DNA unwinding, association of DNA‐binding proteins, or higher‐order chromatin folding. To determine whether increased cleavage by MNase in vitro reflected increased exposure of the underlying DNA template, we assayed the accessibility of a putative SIR2‐protected site in SRR1 to dam methyltransferase. The results of this analysis indicated that at this site SIR2 protects the rDNA sequence from attack by methyltransferase, and by implication from access to other proteins which require interaction with the DNA.

Repression of gene expression by SIR2 was not observed when marker genes were located outside the rDNA array (even when adjacent to a single rDNA repeat), indicating that repression was governed by positional context. Additionally, unlike other known contexts of SIR activity, chromatin structure and repression in the rDNA specifically responded to SIR2 gene dosage, indicating that the SIR2 gene product is a limiting component or regulatory factor which regulates these phenomena in the rDNA. The close association between SIR2 dosage, rDNA chromatin changes and gene expression suggests that SIR2's effect on gene expression in the rDNA is mediated by changes in chromatin structure, specifically via the formation of a more closed chromatin configuration which may exclude transcription factors. The effect of SIR2 in rDNA is to be contrasted with the situation at yeast telomeres, where position effect associated with repression and closed chromatin is specifically responsive to SIR3 dosage (Renauld et al., 1993).

Our result that SIR2 causes pronounced chromatin changes at SRR1 is in good agreement with the recent findings that Sir2p localizes to the nucleolus and that rDNA NTS/5S sequences are enriched in immunocomplexes precipitated by anti‐Sir2p antibodies (Gotta et al., 1997). Taken together with our results, these data build a compelling case that SIR2's effects on chromatin structure, gene expression and recombination result from direct or indirect interaction primarily with the NTS1–5S–NTS2 region of the rDNA repeat unit.

SIR2‐dependent chromatin changes and recombination in the rDNA

We had previously proposed that SIR2‐mediated changes in rDNA chromatin structure may protect rDNA from general recombination functions, since crossing over in this 1 Mb region occurs 50‐ to 100‐fold less frequently than expected based on physical distance (Petes and Botstein, 1977; Petes, 1979; Zamb and Petes, 1982). This was based on the finding that two genes dispensable for the low level of rDNA recombination in wild‐type cells (RAD50 and RAD52) are essential for the elevated recombination observed in a Δsir2 background (Gottlieb, 1989). Recently, a compelling body of evidence has developed demonstrating that DNA accessibility throughout the genome is critical in the initiation of general and site‐specific recombination events. Recombination‐initiating double‐strand breaks (DSBs) have been found during meiosis, predominantly in the upstream regions of yeast genes (Wu and Lichten, 1994). During both mitosis and early meiosis, the DNA at these sites is more exposed than surrounding regions as judged by MNase and DNase I treatment. Events which increase or decrease the accessibility of the DSB sites lead to parallel increases or decreases in recombination initiation (Ohta et al., 1994; Wu and Lichten, 1994). These results suggest that DNA accessibility determines the frequency with which the recombination machinery selects specific sites for recombination initiation. In site‐specific systems, recombination initiation can also be regulated via changes in chromatin accessibility, as in yeast where SIR functions protect the silent mating‐type cassettes from HO cleavage, and in mammals where chromatin structure plays a role in the activation of immunoglobulin gene rearrangements. In a similar manner, the ability of SIR2 to regulate rDNA recombination may stem from its ability to regulate the accessibility of SRR1 and/or SRR2. SIR2‐mediated changes in rDNA chromatin structure may also exclude recombination factors like RAD50 and RAD52, thereby explaining why the products of both genes are dispensable for rDNA recombination in wild‐type cells but are essential for the increased recombination observed in a Δsir2 background (Gottlieb, 1989). The identification of SIR2‐responsive sites in rDNA chromatin now makes possible the detailed analysis of cis‐acting sequences required for specialized chromatin structure and recombination suppression in the rDNA, focusing particularly on the sequences underlying SRR1 and SRR2. It is tempting to speculate that SIR2 regulates chromatin structure in the rDNA by altering histone acetylation (as at the silent HM loci and telomeres; Braunstein et al., 1993), but the effects of histone modifications on chromatin structure in the rDNA remain to be determined.

Does the position effect in the rDNA involve switching between heritable expression states?

Silencing by SIR2 results in intermediate expression of ADE2 and CAN1 genes located in the rDNA array. Even in the presence of a multicopy SIR2 plasmid, reduced expression of ADE2 yields pink colonies (rather than red) and a 40% reduction in steady‐state ADE2 RNA levels as compared with wild‐type. Repression of CAN1 generates a similar decrease in CAN1 RNA and leads to slow growth in the presence of canavanine. Expression of both markers fails to be completely extinguished by SIR2 function. This result is perhaps not surprising, considering that the rDNA array is a region of vigorous and crucial transcriptional activity where stringent silencing would be deleterious. The colony color, CAN growth and RNA level data are compatible with a mode of silencing by SIR2 which is of intermediate strength and uniform in the population. We considered whether this behavior may reflect mixtures of cells with high and low expression states. Transcriptional silencing of telomeric loci is typified by semi‐stable, heritable switching between active and inactive expression states (Gottschling et al., 1990), a characteristic which is shared with the HM loci under some conditions (Pillus and Rine, 1989; Mahoney et al., 1991; Sussel et al., 1993). When the ADE2 locus is placed near a telomere, red/white sectors are formed. The same behavior is observed with an HMR::ADE2 marker in the presence of HMR ARS mutations and certain alleles of rap1 (Sussel et al., 1993). However, this result contrasts with our findings in the rDNA, since rDNA::ADE2 strains display uniform color regardless of SIR2 genotype. Furthermore, while rapid switching between fully ‘on’ and ‘off’ CAN1 expression states could account for slow growth in CAN media, one would expect that the repressed Canr state would coincide with repression of the adjacent ADE2 locus. Since Canr colonies are white rather than red or pink, this description does not appear correct. Rather, we favor the view that rDNA silencing results either from a constant, intermediate level of silencing or rapid on/off switching, a model shared with others (Smith and Boeke, 1997).

The rDNA position effect may require fewer components than other silenced regions

Silencing in yeast has been shown to depend upon a large number of cis‐ and trans‐acting factors, many of which are shared between silencing at the HM loci and telomeres. The differential requirements for individual cis‐ and trans‐acting factors at HML, HMR and telomeres suggest that different combinations of multiple silencing components contribute to silencing at these regions (Laurenson and Rine, 1992). Thus far, the finding that nucleosomes at silenced loci are hypoacetylated in a SIR‐dependent manner (Braunstein et al., 1993) constitutes the only candidate biochemical role of the SIR proteins. While SIR2 leads to reduced acetylation of cellular histone pools, SIR2, SIR3 and SIR4 are all required for maintenance of the hypoacetylated state. In contrast, we have shown here that of the four SIR genes, only SIR2 is required for the three rDNA‐specific phenomena we have described: suppression of recombination, suppression of gene expression, and regulation of chromatin structure. The SIR1, SIR3 and SIR4 functions are dispensable for each of these phenomena in the context of the rDNA. It is not known whether the recently identified SIR2‐related protein family (Brachmann, 1995, #639; Derbyshire, 1996 #619) plays a role in rDNA silencing. At the HM loci and telomeres, a complete understanding of silencing has been complicated by the partially redundant nature of the silencing apparatus (Stone et al., 1991; Sussel and Shore, 1991; Chien et al., 1993; Thompson et al., 1994a). The rDNA may present an alternative, simplified context for dissecting silencing, since silencing is potentially accomplished with fewer functions.

Additional biological consequences of SIR2–dependent chromatin structure in the rDNA

In addition to the suppression of recombination and gene expression, SIR2's ability to regulate rDNA chromatin structure may have additional effects in the rDNA. Recent studies have shown that rRNA gene expression in Saccharomyces cerevisiae is regulated in part by modulating the proportion of rDNA units which are transcriptionally active at any one time (Dammann et al., 1993). Using DNA and RNA cross‐linking techniques in an elegant series of experiments, Dammann et al. demonstrated that active and inactive repeat units are found in the same cell, and the ratio between them varies according to growth phase. Inactive units were shown to be packaged along their length in consistently spaced nucleosomal arrays typical of bulk chromatin. Active units, on the other hand, are devoid of nucleosomes along their coding regions and display a modified nucleosomal structure in the NTS which is thought to reflect the binding of trans‐acting proteins.

The correspondence between the involvement of SIR2 in the regulation of rDNA chromatin structure and the finding that rDNA units are organized into functionally and structurally distinct chromatin configurations suggests the intriguing possibility that SIR2 may also play a role in the regulation of the proportion of active and inactive rDNA transcription units. Both the MNase and methyltransferase procedures generate a cleavage profile for the mass average of rDNA in the assay. Moreover, because the rDNA unit is a repeated structure (∼100 copies/haploid cell), this raises the possibility that the repeated array within a single cell contains a mixture of units with distinct cleavage profiles (and therefore distinct chromatin configurations). It is not clear whether SIR2 function results in a more closed chromatin configuration in all or a subset of rDNA repeat units within one cell. Because of the relationship between the activity of rDNA transcription units and rDNA ARS function (Walmsley et al., 1984; Saffer and Miller, 1986), SIR2 may also play a role in the initiation of rDNA replication analogous to the established correspondence between ARS activity/timing and silencing at telomeres and the HM loci (Miller et al., 1984; Saffer and Miller, 1986; Mahoney et al., 1991; McNally and Rine, 1991; Ferguson and Fangman, 1992; Aparicio and Gottschling, 1994; Lucchini and Sogo, 1994).

Finally, recent evidence indicates that, under certain specialized circumstances, SIR3p and SIR4p are also localized to the nucleolus (Gotta et al., 1997; Kennedy et al., 1997), but this appears to be due primarily by virtue of their potential to interact with SIR2p. Intriguingly, these studies suggest that major redistribution of silencing factors to the rDNA also plays a significant role in senescence (Kennedy et al., 1997). Whether Sir2p plays an additional role in this process other than to aid in the redistribution of silencing factors is not yet clear.

Materials and methods

Strains and media

A complete list of yeast strains used in this study is presented in Table I. Bacterial plasmids were maintained in E.coli DH5αF (Woodcock et al., 1989). Incubations at 30°C and 37°C were used to support growth of yeast and E.coli, respectively. Transformations of E.coli (Dower et al., 1988) and S.cerevisiae (Schiestl and Gietz, 1989) were performed as described. Liquid and solid YPDA, SD, sporulation, dissection media and media containing canavanine were as described (Sherman et al., 1974). Selective media (supplemented minimal media, SMM) were made by supplementing minimal medium with liquid amino acid stock solutions as required to complement strain auxotrophies. Experiments requiring yeast strains with episomal plasmids were conducted with fresh transformants which were propagated under selective growth conditions except were stated. All gene disruptions and integrations were confirmed by Southern analysis (Southern, 1975), including the verification of single‐copy insertions in the rDNA array.

View this table:
Table 1. Strains used in this study

Strain CFY543 was constructed by two‐step gene replacement from a segregant of CFY520 which has lost its rDNA insert. This starting strain was transformed with pURACAN to uracil prototrophy. Integrants whose structure was confirmed by Southern analysis were plated on 5–FOA medium to select for loss of the URA3 marker. Survivors were screened by Southern analysis and for CanS phenotype. CFY556 was generated by disruption of the CAN1 gene in strain CFY543 by a two‐step process (Alani et al., 1987) to generate CFY556. CAN1 was disrupted by transformation with a linear fragment from pΔCAN1 (O.Aparicio) in which the central CAN1 HindIII fragment has been replaced by the hisG‐URA3‐hisG cassette. The ADE2 locus of strain CFY556 was disrupted with the plasmid pΔADE2 (Aparicio et al., 1991) by the same method to form CFY558. The intermediate containing the integration of the hisG‐URA3‐hisG cassette at the deleted ADE2 locus was retained as strain CFY557. The ADE2‐CAN1 cassette was integrated into the rDNA of CFY558 to form CFY559 by linearizing pCAR1 at the unique SacII site in the rDNA and transforming CFY558 to Ade+. CFY559Δsir2 was generated form CFY559 by disrupting the SIR2 gene with pJH103.1 (Ivy et al., 1986). Strain CFY558Δsir2 is a mitotic segregant of CFY559Δsir2 that has lost the ADE2‐CAN1 insert from the rDNA. CFY570 was constructed by transformation of CFY558Δsir2 to Ade prototrophy with circular pADECAN plasmid. Southern analysis was used to screen transformants for isolates which had integrated the plasmid at the ADE2 locus to form the structure depicted in Figure 1. CX200‐2C and CX200‐6D are independent Ura+ Ade+ meiotic segregants from a cross of CFY557 and REE1303. Both were isolated from tetrads with four normally segregating spores and were subsequently plated on medium containing 5‐FOA to select for loss of the hisG‐URA3‐hisG cassette in the ADE2 locus. REE1285 contains an integration of the 3.6 kb ADE2 fragment at the HindIII site within the 35S coding region of the rDNA (originally constructed by R.Keil). The transcriptional orientation of ADE2 is opposite to the 35S transcription unit.

The strains YK18 and YK4 are mitotic segregants of SG25‐36D and G200, respectively, that have lost the URA3 insert in the rDNA (Gottlieb and Esposito, 1989). YK52 is isogenic to YK4, but contains a disruption of HMRa (hmra::URA3). HMRa was disrupted by transforming the strain YK4 to uracil prototrophy with a HindIII fragment derived from pCH4 that contains HMRa sequences in which the coding and flanking regions were deleted and replaced by the URA3 gene. Disruption of the HMRa locus was confirmed by Southern analysis in transformants which mated as a cells (Southern, 1975).

Plasmids

A complete list of yeast strains used in this study is presented in Table II. The plasmid pCAR1 was employed to introduce the ADE2‐CAN1 marker into the rDNA. pCAR1 was constructed by subcloning a 9.1 kb rDNA KpnI fragment from pKK1 (Gottlieb, 1989) into the KpnI site of pUC19 (Yanisch et al., 1985) to form pCF15. Two KpnI sites are found in the published yeast rDNA sequences; the KK1 subclone contains only the KpnI site near the 5′ end of the 26S coding region. A 3.6 kb BamHI fragment containing ADE2 from plasmid L909 (gift from R.Keil) was cloned into the BamHI site of pCF15 to generate pCF16. Finally, a 4.1 kb XhoI CAN1 fragment from plasmid pCAN1‐2‐1 (gift from J.Shuster) was ligated into the SalI site of pCF16 to generate pCAR1. The 35S rRNA, ADE2 and CAN1 genes are all in the same transcriptional orientation. The CAN1 XhoI fragment contains ∼2 kb of upstream sequence and the entire CAN1 coding region. The downstream XhoI site is not in the published CAN1 sequence (Hoffmann, 1985); we have mapped it to position 2202 where the published sequence indicates a 5/6 match for the XhoI recognition sequence (CTCAG as compared with CTCGAG).

View this table:
Table 2. Yeast plasmids used in this study

The centromeric plasmid pCARCEN was constructed by co‐transforming a ura3 yeast strain with the large PvuII fragment of pRS316 and pCAR1. Ura+ transformants were rescued to E.coli and plasmid was purified to confirm that the small PvuII fragment of pRS316 had been replaced with the PvuII fragment of pCAR1 containing the rDNA, ADE2 and CAN1 inserts. pCACEN was generated by partial KpnI digestion of pCARCEN and religation to remove the 9.1 kb rDNA repeat unit.

Linear fragments from plasmids pJI23.2, pJH103.1, pJH107.1 and pDM610.23 (Ivy et al., 1986) were used as described to disrupt the SIR1, SIR2, SIR3 and SIR4 genes respectively. The 2μ‐based plasmids pJH50, pJH423, pKAN63 and pKAN59 (Ivy et al., 1986) contained the SIR1, SIR2, SIR3 and SIR4 genes, respectively. These disruption and 2μ plasmids are marked with LEU2. The plasmid pJR68 was also used for multi‐copy expression of SIR2 (gift from J.Rine).

pCSS1(16) contains a 5.1 kb SalI fragment encoding the CSS1 gene in the SalI site of the URA3 marked centromeric vector pRS316. pCSS1(17) contains the same CSS1‐complementing fragment as pCSS1(16) cloned into the LYS2‐marked centromeric vector pRS317. pBE288 is the HindIII fragment containing the 5′ end of the ADE2 coding region in the HindIII site of pVZ1 (Henikoff and Eghtedarzadeh, 1987).

pCH1 contains a 2458 bp EcoRI fragment containing the 5S rDNA gene and surrounding non‐transcribed spacer sequences cloned into the EcoRI site of the pUC19 polylinker. pCH2 and pCH6 are derived from pCH1 and contain a 200 bp EcoRI–HindIII and a 250 bp SmaI fragment of pCH1, respectively, cloned into the polylinker of pUC19.

RNA analysis

Total yeast RNA was prepared by harvesting ∼4×108 cells from mid‐log cultures. Cell pellets were disrupted for 2.5 min on a vortex mixer in the presence of 0.4 g acid‐washed glass beads, 0.2 ml YRLB (0.5 M NaCl, 0.2 M Tris, pH 7.5, 0.01 M EDTA, 1.0% SDS) and 0.3 ml phenol:chloroform. After addition of 0.3 ml phenol:chloroform and 0.3 ml YRLB, vortexing was continued for 1.5 min and phases were separated. The aqueous phase was re‐extracted with 0.3 ml phenol:chloroform by vortexing for 1.5 min. The aqueous phase was precipitated in RNase‐free tubes with 2 vols 100% EtOH containing 0.05% DEPC for >30 min at −20°C. RNA pellets were washed with DEPC‐treated 70% EtOH, resuspended in DEPC‐treated water, and stored at −80°C. Yield was judged by UV absorbance at 260 nm.

S1 nuclease protection assays were performed as described (Berk and Sharp, 1977) using 20 μg total RNA per sample and 20 000, 30 000 and 50 000 c.p.m. of CAN1, ADE2 and HIS3 probe, respectively. Samples were fractionated on 5% acrylamide (19:1 acrylamide–bisacrylamide)–8.3 M urea gels. Control experiments were performed to confirm that the S1 nuclease protection procedure can detect differences in RNA levels under the conditions used. A constant amount of labeled anti‐sense probe was added to samples containing increasing amounts of target RNA. Signals detected with probes to CAN1, ADE2 and HIS3 were quantitated as described below and responded linearly over a 10‐fold range of input RNA encompassing the experimental 20 μg per lane, indicating sufficient probe excess for quantitative analysis.

Unexpectedly large transcripts in addition to the expected wild‐type messages were detected in strains carrying the rDNA::ADE2‐CAN1 insert using probes homologous to the 5′ ends of the CAN1 and ADE2 genes. In contrast, both probes detected only the expected wild‐type species in strains with wild‐type ADE2 and CAN1 loci and no rDNA inserts. Subsequent Northern and S1 analyses demonstrated that the novel transcripts represent large, unpolyadenylated messages with a size expected for Pol I read‐through transcripts from the nearest upstream 35S rRNA promoter. Production of the novel transcripts depended upon the presence of rDNA sequences upstream of the ADE2‐CAN1 insert since they were found in episomal constructs with an upstream rDNA unit but not in related constructs lacking rDNA sequences (unpublished data). A similar pattern of transcription was noted in another instance where markers were placed downstream of an rDNA transcription region (Stewart and Roeder, 1989).

Probes

DNA probes were gel‐purified and labeled via the manufacturer's instructions with the GENIUS™ kit (Boehringer‐Mannheim). The RNA probe for CAN1 was generated by digesting p5′CAN1 with BsmAI followed by in vitro transcription with T7 RNA polymerase in the presence of [α‐32P]dUTP. A 623 bp RNA probe is produced which contains 50 bp of plasmid sequence at its 5′ end, followed by 573 bp of CAN1 sequence beginning at the Afl2 site in the coding region and ending at the BsmAI site in the CAN1 upstream region. The RNA probe for ADE2 was generated by digesting pBE288 with DraI and in vitro transcription with T3 RNA polymerase in the presence of label. A 945 bp RNA probe is produced which contains 20 bp of plasmid sequence at the 5′ end followed by 925 bp of ADE2 spanning the HindIII site at the 5′ end of the ADE2 coding region through the DraI site in the ADE2 upstream region. Both probes were shown to be specific by their failure to detect messages in RNA prepared from a Δade2 Δcan1 strain (CFY558, not shown). HIS3 RNA probe has been described previously (Surosky and Esposito, 1992).

For the MNase studies, DNA probes p2, p3 and p6 were prepared directly from the rDNA inserts of plasmids pCH2, pCH3 and pCH6, respectively. HMRa and MATα MNase profiles were analyzed by digesting the DNA with HindIII and hybridization with a unique DNA probe that adjoins the distal HindIII site for each locus: a HindIII–NruI fragment for MATα and a BamHI–PstI fragment for HMRa. These fragments were purified by gel electrophoresis and labeled to a specific activity of 1×109 c.p.m./μg using the Stratagene random priming kit. In all other studies, DNA probes were labeled via the manufacturer's instructions with the GENIUS™ kit (Boehringer‐Mannheim).

Image analysis and figure production

S1 gels were quantitated using a Molecular Dynamics PhosphorImager and ImageQuant software. In other cases as indicated, autoradiograms were scanned using an Apple Scanner and AppleScan software in grayscale mode at 300 d.p.i. Scanned images were analyzed by densitometry with NIH Image v1.54. Band intensities were generated by manually identifying bands and determining the background density (BD) for each lane. The software calculated average density (AD) and area (AR) for each selected band; these were combined to yield total band density (TD) as TD = AR(AD–BD). Images of scanned autoradiograms and phosophorimagery files were imported to Adobe's Photoshop v 2.5 software. Pertinent sections of the images were selected and arranged for presentation; no image manipulations were performed and all data presented in any figure is the result of the same experimental trial. Output was generated on a Kodak 7700 dye‐sublimation printer using Kodak paper.

Nuclear isolation

Nuclei were isolated following the Percoll gradient procedure of (Ide and Saunders, 1981), with some modifications in the pretreatment, spheroplasting and lysis steps according to (Huigbregtse et al., 1987): 4–8 g of cells (wet weight) grown to late log or stationary phase in YPDA (2% bactodextrose, 2% bactopeptone, 1% yeast extract, 0.001% adenine sulfate) were pretreated by washing in 28 ml/g wet weight 40 mM EDTA, pH 8.5, 90 mM β‐mercaptoethanol. Spheroplasting was done in 1 M sorbitol, 1 mM EDTA, pH 8.5, 3 mM DTT with 6 mg Zymolyase 20.000T/g wet weight at 30°C until 75% of the cells appeared swollen and without buds when suspended in digestion buffer (50 mM Tris, pH 7.5, 75 mM KCl, 0.5 mM CaCl2, 1 mM PMSF). Spheroplasts were gently resuspended in 5 ml/g wet weight ice‐cold digestion buffer. The suspension was kept on ice for 3 min to allow lysis of the spheroplasts, and the lysates were loaded on 30 ml prespun 30% Percoll (Pharmacia) gradients. Nuclei were withdrawn from the gradient, washed with 150 ml ice‐cold digestion buffer and resuspended in 2 ml of the same buffer.

MNase digestion

Aliquots of 0.6 ml nuclear suspension were digested step‐wise with 0 (untreated) to 18 units of micrococcal nuclease (MNase‐Worthington) for 5 min at 25°C. The reactions were stopped by the addition of 1.8 ml GuHCl solution (6 M GuHCl, 133 mM EDTA, 200 mM NaCl, 0.066% Sarkosyl, pH 8). Purification of the DNA was performed according to Holm et al. (1986).

For the naked DNA controls, 100 μg of total yeast DNA was resuspended in 1 ml 25 mM Tris, pH 7.5, 5 mM CaCl2 and digested at 25°C with 1.8 units of MNase. Aliquots of 300 μl were removed from the reaction after 3, 5 and 8 min and brought to a final concentration of 20 mM EGTA, 0.1% SDS to stop the reaction. The DNA was incubated at 65°C for 30 min with 0.1 mg/ml Proteinase K (Boehringer‐Mannheim) and then extracted thoroughly with phenol/chloroform.

Southern hybridizations

For the MNase studies, digestion of DNA with restriction enzymes was carried out according to specifications of the manufacturer (Boehringer‐Mannheim). Restricted DNA was run on agarose gels and blotted overnight onto nitrocellulose or Nytran nylon membranes (Schleicher & Schuell) in 20× SSC according to (Southern, 1975). Hybridizations were done at 65°C overnight in 5× SSC, 0.5% Sarkosyl, 1× Denhardt's, 0.1 mg/ml sonicated salmon sperm DNA, 10% dextran sulfate. Membranes were washed for 2×10 min in 2× SSC, 0.1% SDS at room temperature and subsequently, when necessary, in 0.2× SSC, 0.1% SDS at 55°C for 30 min. Membranes were exposed to Kodak XAR films overnight at room temperature.

For the analysis of GATC methylation, genomic DNA was prepared from 5 ml logarithmic cultures by the method of Hoffman and Winston (1987) with the addition of a second organic extraction after RNase treatment. Approximately 10 μg DNA was digested overnight in 100 μl with a minimum of 6 units enzyme. Digests were precipitated and run on 1% agarose gels. Transfer to Magna nylon membranes (Micron Separations) was accomplished by the method of Kempter et al. (1991) with transfers of 45 min to 1.5 h. Filters were incubated to dryness at 80°C. Hybridization was carried out in roller bottles at 65°C in a solution of 5× SSC, 0.5% blocking reagent (Boehringer‐Mannheim), 0.1% sarcosyl and 0.02% SDS for a minimum of 4 h. Filters were subsequently washed twice for 15 min at 65°C with 0.1% SDS, 0.5× SSC. Development by chemiluminescent detection proceeded as per instructions in the Boehringer‐Mannheim GENIUS™ kit, with the modification that Wash #2 contains 2% blocking reagent. When needed, blots were stripped for re‐probing by incubation in 0.2 M NaOH, 0.1% SDS at 37°C for 20 min.

Acknowledgements

The authors would like to thank J.Ivy, J.Rine, O.Aparicio, H.Renauld and D.Gottschling for the gift of materials used in this study and S.Gasser for her critical comments on the manuscript. This work was supported by an NRSA training grant to C.F. (GM07183) and grant NIH‐G29182 to R.E.E.

References