DNA Replication-Independent Silencing in S. cerevisiae

See allHide authors and affiliations

Science  26 Jan 2001:
Vol. 291, Issue 5504, pp. 646-650
DOI: 10.1126/science.291.5504.646


In Saccharomyces cerevisiae, the silent mating loci are repressed by their assembly into heterochromatin. The formation of this heterochromatin requires a cell cycle event that occurs between early S phase and G2/M phase, which has been widely assumed to be DNA replication. To determine whether DNA replication through a silent mating-type locus, HMRa , is required for silencing to be established, we monitored heterochromatin formation atHMRa on a chromosome and on a nonreplicating extrachromosomal cassette as cells passed through S phase. Cells that passed through S phase established silencing at both the chromosomalHMRa locus and the extrachromosomalHMRa locus with equal efficiency. Thus, in contrast to the prevailing view, the establishment of silencing occurred in the absence of passage of the DNA replication fork through or near theHMR locus, but retained a cell cycle dependence.

Heritable states of gene expression are central to the development of life. Gene repression and activation play pivotal roles in the differentiation of totipotent cells into different cell types, each of which selectively and stably expresses only a subset of the genes in the genome. DNA replication can play a role in changing patterns of gene expression (1–3) and thus is a possible mechanism for disrupting chromatin states before their reprogramming and for the de novo establishment of those states. There are also clear examples of changes in gene expression and differentiation that occur independently of DNA replication (4).

For years, one of the strongest suggestions of a role for DNA replication in establishing heritable transcriptional states came from studies of yeast mating types. In S. cerevisiae, mating competence requires heritable repression at the silent mating-type loci, HML and HMR. The formation of heterochromatin at HML and HMR requires regulatory sites called silencers, which flank these silent loci. Silencers contain binding sites for the origin recognition complex (ORC), Rap1p and Abf1p [(5) and references therein]. In addition, the four Silent Information Regulator proteins, Sir1p, Sir2p, Sir3p, and Sir4p, are structural components of yeast heterochromatin (3, 5, 6). Sir2p is an enzyme with histone deacetylase activity (7). The principal role of Sir1p is the establishment rather than maintenance of silencing, presumably by enhancing a limiting step in heterochromatin formation (5). The recruitment of Sir1p to the silencer through interactions with ORC is thought to lead to silencing, resulting in the recruitment of Sir2p, Sir3p, and Sir4p (8,9).

Passage of yeast cells through S phase is required to establish silencing at HMR (10), and DNA replication has been the leading candidate for the requisite S phase event. The hypothesized link between DNA replication and silencing is supported by indirect evidence. For example, two of the four silencers are chromosomal origins of replication (11). Similarly, ORC has roles in both replication and silencing (12–14). In addition, mutations in genes encoding an essential DNA helicase, DNA2, or the proliferating cell nuclear antigen (PCNA) loading factor,RFC1, disrupt silencing at the telomeres and ribosomal DNA (rDNA), respectively (15, 16). Alterations of two replication-coupled chromatin assembly factors, CAF-1 and ASF1, also disrupt silencing (15–18). Finally, mutant forms of PCNA defective in CAF-1 interaction are defective in establishing silencing (19, 20). These data suggest that CAF1 and PCNA link DNA replication to chromatin assembly and silencing.

However, some data raise doubt about this model. For example, DNA replication initiation at silencers is not required for silencing (21). Also, the roles of ORC in replication and silencing are genetically separable (22–24). Tethering Sir1p to a silencer bypasses a need for ORC in silencing that locus (9), but does not bypass the requirement for S phase passage (21). All observations could be explained if the link between replication and silencing were the passage of a replication fork though HMR, rather than a requirement for replication initiation at a silencer. Here we present a critical test of whether DNA replication is required to establish the silenced state.

To test directly whether DNA replication through HMRwas the S phase event requirement for silencing, we excised a replication-defective HMR cassette from its chromosomal locus during G1. Cells with the excised HMRcassette were induced to express Gal4-Sir1p, which is required to establish silencing, then were allowed to pass through S phase and subsequently rearrested at the G2/M boundary. The establishment of chromosomal and extrachromosomal silencing was monitored by RNA blots (Fig. 1).

Figure 1

Experimental Design. A genomicHMR locus consisting of a synthetic silencer containing four Gal4p binding sites, a Rap1p binding site and an Abf1p binding site integrated at the HMR-E silencer and the genes encoding a1 and a2 (black or gray double-headed arrow indicating expression or repression, respectively) was flanked by two FRT sites (white arrowhead) oriented to allow excision by FLP recombinase. The excised HMR locus lacked any origin of replication (11, 21). The establishment of silencing at HMRa was regulated by controlling expression of the chimeric Gal4-Sir1p (gray circle) via theMET3 promoter (21). FLP recombinase expression was regulated with the GAL10 promoter (29). Cells grown with Gal4-Sir1p off, resulting in HMRa being on, were arrested in G1 with α factor. Induction of FLP with 2% galactose led to excision of theHMR locus (light gray area) from the chromosome (black area) in G1. Expression of Gal4-Sir1p was then induced, and FLP recombinase was repressed in medium lacking methionine and containing 2% raffinose. Cells expressing Gal4-Sir1p were then released from G1, were allowed to proceed to G2/M, and were rearrested with benomyl plus nocodazole. The probe (medium gray area) and restriction sites used (Figs. 2B and 3A and Table 1) are noted.

Both the HMR-E and HMR-Isilencers contain origins of replication (11), which were incompatible with this experimental design that evaluated silencing in the absence of HMR replication. Thus, we used a strain with a replication-defective silencer containing a Rap1p binding site, an Abf1p binding site, and four Gal4p binding sites in place of the ORC binding site (4xGal4-Rap1-Abf1) flanking HMR. TheHMR-I silencer was deleted and the entire locus was flanked by FLP1 recombination target (FRT) sites at which Flip recombinase (FLP) catalyzes site-specific recombination (25). The orientation of the sites allowed excision of HMR from the chromosome. The FLP enzyme, an Int-like recombinase, leaves no free broken DNA ends as intermediates or products of recombination (26). In these cells, silencing at HMR is dependent on the chimeric protein, Gal4-Sir1p, which binds Gal4 sites in the synthetic silencer, bypassing the requirement for ORC (8, 21). In addition, silencing at HMRin this strain was dependent on SIR2, SIR3, andSIR4, indicating that silencing mediated by tethered Sir1p or a wild-type silencer was mechanistically similar (27). Moreover, the tethered Sir1p form of silencing causes an altered superhelical density and enrichment for deacetylated histones expected of ORC-dependent silencing (28).

Cells were arrested in G1 by the pheromone α factor, and the culture was split in two. HMRa in these cells was de-repressed, expressing a1 mRNA due to the absence of Gal4-Sir1p. In one culture, the HMR cassette was excised from the chromosome by inducing FLP recombinase from theGAL10 promoter (29). The other culture was held in G1 without inducing FLP1, leavingHMR in the chromosome. In G1, FLP1expression was subsequently repressed in the first culture, and Gal4-Sir1p was then expressed for 1 hour via the MET3promoter in both cultures. Both cultures were then released from G1, allowed to pass through S phase, and rearrested two hours later at the G2/M boundary using microtubule inhibitors. Thus, in one culture, HMRa was replicated in its chromosomal context, and, in the second culture, the excisedHMR locus passed through S phase without being replicated. Parallel control experiments were performed without expression of Gal4-Sir1p, and silencing was not established in these cells. Cell cycle progression and arrest were monitored by microscopy.

After passage through S phase, the level of a1 mRNA from HMR was monitored. The half-life of a1 mRNA is less than 3 min, so changes in transcription initiation are rapidly reflected in the levels of a1 mRNA. In cells at G2/M, a1 mRNA from the chromosomal locus was reduced to 11% of the level in G1 cells. Thus, silencing was efficiently established at the chromosomal HMR locus (P = 0.018, n = 3) [Fig. 2A, lanes 8 and 9, and Web table 1 in (30)]. Similarly, after passage through S phase, the level of a1 mRNA from the excised HMR cassette in cells at G2/M was reduced to 14% of the level in G1cells (P = 0.018, n = 3) (Fig. 2A, lanes 5 and 6, and Web table 1). The efficiency of silencing ofHMR in both contexts was quantitatively similar (P = 0.51, n = 3). Silencing was not established at HMR in either context without Gal4-Sir1p (Fig. 2A, lanes 15 and 16, or lanes 12 and 13, respectively, and Web table 1). In addition, the level of a1 mRNA from the chromosomal HMR locus or from the excised HMRcassette in G1-arrested cells was similar, with or without Gal4-Sir1p (Fig. 2A, lanes 3 and 5, and lanes 10 and 12; or lanes 5 and 8, and lanes 12 and 15). Thus, excision of HMR did not promote or inhibit transcription of a1 from HMR. Also, a1 mRNA levels in G1-arrested cells from either the chromosomal HMR locus or excised HMRcassette was unaffected by expression of Gal4p-Sir1p for at least 1 hour (Fig. 2A, lanes 4 and 5, and lanes 7 and 8). Thus, silencing was established as efficiently on the excised HMR cassette as on the chromosomal HMR locus, and silencing in either context required both the expression of Gal4-Sir1p and passage through the cell cycle. These results implied that passage from G1 to G2/M was required for silencing, in support of earlier findings (10, 21). However, the cell cycle requirement was not passage of the DNA replication fork throughHMR.

Figure 2

Establishment of silencing in the absence of DNA replication. Cells were treated as described in Fig. 1 and text. (A) RNA analysis of the establishment of silencing. The blot was probed for a1 mRNA (upper band) and subsequently stripped and reprobed for the SCR1 loading control (lower band) (30). (B) Analysis of the excision efficiency of HMR and the nonreplication of the excisedHMR. Cells were treated as described in text and in Fig. 1. Total genomic DNA was harvested, digested with Sna BI and Eco NI, and separated on a 0.7% agarose gel before analysis by DNA blots using a 1697–base pair (bp) probe that hybridized to both the HMRcassette and the flanking chromosomal DNA (30). This probe detected a 4.6-Kbp fragment from chromosomal HMR or a 2.6-Kbp fragment from the excised HMR cassette, and a 6.7-Kbp fragment, which harbored the chromosomal locus flanking the excised HMR (Fig. 1). Data were quantified using a PhosphorImager (Molecular Dynamics). Expression of Gal4-Sir1p is indicated by + or –. L, log phase cells; G1, α-factor arrested cells; G2/M, benomyl and nocodazole arrested cells; Excised, the excised HMR cassette; Chrom.,HMR at the chromosomal locus.

The interpretation of the previous experiment hinged critically on knowing whether HMR was efficiently excised by FLP recombinase and whether the excised HMR had some unanticipated capacity to replicate. Analysis of DNA blots hybridized for HMR and for flanking chromosomal DNA resolved both issues. To determine the efficiency of the excision, we hybridized a probe homologous to both the HMR locus and chromosomal sequences flanking the site of excision to DNA from cells either expressing FLP recombinase or not, and either expressing Gal4-Sir1p or not (Fig. 2B, lanes 5 and 12) (27). The efficiency of excision in G1 was 91 ± 7.7% (n = 3) for cells expressing Gal4-Sir1p and 93 ± 4.7 % (n = 3) for cells lacking Gal4-Sir1p (30). Thus, most cells expressing FLP recombinase had excised HMRfrom the chromosome before S phase.

To monitor whether the excised chromosomal locus was replicated, we compared the intensity of the hybridization signals from the excisedHMR in both G1 and G2/M arrested cells to those from sequences flanking the site of excision. The flanking chromosomal sequences double each S phase. Whether the excisedHMR could replicate was resolved by determining if the relative hybridization intensities changed between G1 and G2/M (31). The ratio from G1-arrested cells was calculated, and the amount of the excised HMR cassette relative to the chromosomal DNA flanking the site of excision was set to 100% to normalize between experiments (Table 1). If the excisedHMR cassette did not replicate, this hybridization ratio should have been reduced to 50% at G2/M. In contrast, if the excised HMR replicated, this hybridization ratio should have remained constant (100%). Indeed, this ratio was reduced to an average of 50% in cells lacking Gal4-Sir1p and to an average of 62% in cells containing Gal4-Sir1p (Fig. 2B, Table 1). These two ratios were not significantly different (P = 0.51,n = 3). Thus, no evidence of replication of either the expressed or silenced excised HMR locus was observed. Further analysis establishing that differential degradation of the excised HMR relative to the chromosomal locus was not responsible for the differences observed is provided (30).

Table 1

The excised HMR cassette was not efficiently replicated. Cells were treated as described in Fig. 1. Noted are the final arrests in α factor (10 μg/ml) (G1) and the subsequent release and rearrest in benomyl (30 μg/ml) and nocodazole (10 μg/ml) (G2/M). The relative level of excised HMR was calculated as the ratio of the PhosphorImager units as follows: [(excised HMR/chromosome ΔHMR in either G1 or G2/M arrest)/(excised HMR/chromosome ΔHMR in G1 arrest)] × 100 ± σ ; n = 3. The ratios during the G1 arrest and the ratios during the G2/M arrest either in the presence of absence of Gal4-Sir1p were compared to determine whether they were similar, and data were analyzed by using the one-sided Wilcoxon rank sum test (P = 0.018). The ratio during the G2/M arrest in the presence of Gal4-Sir1p was compared to that in the absence of Gal4-Sir1p to determine whether they were similar, and data were analyzed by using the two-sided Wilcoxon rank sum test (P = 0.51). The ratio during the G1 arrest has been set to 100% to normalize the data between independent experiments, one of which is shown in Fig. 2B.

View this table:

To measure more sensitively whether the excised HMR could replicate, we monitored its loss in dividing cells (Fig. 3A). In this experiment, the initial G1-arrested culture was split in two and the HMRcassette was excised by inducing Flp1p in both cultures. In one culture, Gal4-Sir1p was expressed while maintaining the G1arrest. However, upon release from G1, the cells were grown logarithmically and samples were harvested hourly for 8 hours. In this experiment, hybridization to the extrachromosomal HMRrelative to the flanking chromosomal sequences was reduced by approximately 50% per generation, reflecting a lack of replication of this locus (Fig. 3A). Together, these results confirmed thatHMR was efficiently excised from the chromosome in G1 and that the extrachromosomal HMR was not efficiently replicated, even after multiple cell divisions.

Figure 3

The excised HMR cassette was not replicated. Cells were treated similarly to the description in Fig. 1, except that upon release from the G1 arrest [0 hours in (A) or 0 min in (B) and Web fig. 1], cells were grown logarithmically and samples were harvested hourly for 8 hours (A) or every 5 min for 60 min and then every 10 min for an additional 30 min (B). Samples were analyzed by DNA blots and quantified using a PhosphorImager. Data represent one experiment from two independent experiments with comparable results. (A) The excised HMR cassette was rapidly lost from cycling cells in either the presence or absence of silencing. Samples were analyzed as described in Fig. 2B. The relative DNA content for each time point was calculated as follows: [(excisedHMR/chromosomeΔHMR at time indicated)/(excisedHMR/chromosomeΔHMR during G1)] or [(chromosomeΔHMR/chromosomeΔHMR at time indicated)/(chromosomeΔHMR/chromosomeΔHMRduring G1)]. The doubling time for all cells was about 90 min. (□), excised HMR in the absence of Gal4-Sir1p; (◊), chromosomeΔHMR in the absence of Gal4-Sir1p; (○), excised HMR in the presence of Gal4-Sir1p; and (▵), chromosomeΔHMR in the presence of Gal4-Sir1p. (B) Monitoring HMR for hemimethylation as a mark for passage of the replication fork using cells that constitutively expressed dam methylase (JRY7144). Total genomic DNA was harvested, and analyzed as described in (30). The relative level of hemimethylation at a1 was calculated as follows: [(full-length restriction fragment at time indicated/fragment generated by sensitivity of Dpn I site closest to a1 at time indicated)/(full-length restriction fragment during G1/fragment generated by sensitivity of Dpn I site closest to a1during G1)]. The signal from the full- length restriction fragment relative to the fragment generated by sensitivity of Dpn I site closest to a1 at 0 min was 13 and 14% for the chromosomal locus and excised HMR cassette, respectively. (□), chromosome containing HMR in the absence of Gal4-Sir1p, (◊), excised HMR in the absence of Gal4-Sir1p. FACS analysis of cells from (B), indicating that both cultures entered and passed through S phase at the same rate, is available in Web fig. 1 (30).

As an independent measure of whether the excised cassette could replicate, we monitored the passage of the replication fork through both a chromosomal and an excised nonsilenced HMR locus (Fig. 3B). In this experiment, Escherichia coli dammethylase, which methylates GATC sites, was expressed from the constitutive integrated URA3 promoter. Passage of the replication fork though a locus during either origin-based or repair-based replication causes transient hemimethylation. Fully methylated sites are sensitive to cleavage by Dpn I, whereas hemimethylated (or nonmethylated) sites are resistant. Thus, passage of a replication fork leads to transient resistance of sites to cleavage by Dpn I. Upon release from G1, transient resistance to digestion by Dpn I was observed for chromosomal HMR but not for excised HMR. Thus, a replication fork did not replicate Dpn I sites near a1 in the excised HMR, providing direct physical evidence for lack of replication.

This study provided the first mechanistic test of whether DNA replication was required to form heterochromatin. The HMRlocus of yeast was silenced efficiently regardless of whether the locus was in the chromosome or excised as a covalently closed circular DNA molecule. This excised HMR was not detectably replicated. Nevertheless, passage of the cells from G1 to G2/M through S phase was required for silencingHMR in either context. It is unlikely that the passage of time itself, rather than cell cycle passage, is required for Gal4-Sir1p to mediate silencing. Expression of Gal4-Sir1p for 4 hours between an arrest in G2/M and rearrest in early S phase of the subsequent cell cycle did not allow establishment of silencing (21), whereas in this study expression for a total of 3 hours between G1 and G2/M did. Conceivably, passage of a replication fork through HMR might enhance the efficiency of silencing, even though these data demonstrated that it was not required to silence all, or virtually all, HMR loci in the population. Moreover, cells containing sixHMR-bearing plasmids have all HMR loci silenced (32), indicating that all necessary silencing components are also likely in excess in this study. Therefore, at this level, silencing components do not seem to be limiting for HMRsilencing. It is unclear what, if any, step would be enhanced by passage of the replication fork.

Two different assays for replication of the excised HMRlocus detected no replication of this DNA molecule. However, we could not exclude the possibility that a low, undetectable level of DNA replication occurred. It is possible that a small subset of excised molecules could have replicated or that a very small amount of replication, such as repair-coupled DNA synthesis, could have occurred on a larger fraction of the population. The first possibility would not affect our conclusions because we observed complete repression ofHMRa expression. Thus, even if a few molecules did replicate, the nonreplicated majority of the molecules were silenced.

A low level of repair-coupled synthesis could conceivably deliver replication-coupled proteins like PCNA or CAF1 to the excised HMRmolecules. Repair synthesis can occur at any time during the cell cycle, even in nondividing cells, and hence would not explain the cell cycle dependence of silencing. Moreover, for repair-coupled synthesis to explain silencing, essentially all of the excised HMRmolecules must have suffered some damage that requires repair. The only event experienced by all excised HMR molecules was FLP-mediated recombination. FLP recombinase catalyzes the entire recombination reaction, requiring no other proteins (26) and leaving no substrates for repair synthesis. Also, no repair-coupled synthesis of the excised HMRmolecules was detected by transient resistance to Dpn I cleavage. Thus, it was unlikely that replication-coupled process contributed to the quantitative silencing observed here. We are unaware of any evidence that mutations or damage occurs more readily on circular plasmids in yeast than on the chromosome.

In addition to offering a new mechanistic insight on silencing, these data place renewed importance on the role of proteins involved in DNA replication, such as PCNA, Rfc1p, Asf1p, Dna2p, and CAF-1, in silencing. An important challenge is to learn how those proteins affect silencing when silencing can be mechanistically divorced from both replication initiation and from the passage of a replication fork. PCNA left behind on a previously replicated template can mark that template as “competent” for CAF-1–dependent chromatin assembly (33). If PCNA from the previous cell cycle remains associated with HMR upon entering the subsequent G1 phase, it may be excised with HMR and therefore available to establish heterochromatin. The efficiency of silencing observed here would require that some feature ofHMR causes retention of PCNA. Alternatively, these proteins may have a role in silencing other than in its establishment (18). For example, once heterochromatin is established at a locus, it must be maintained throughout that cell cycle and duplicated in each subsequent cell cycle. Indeed, recent data underscore the dynamic nature of heterochromatin composition in vivo, even on nonreplicating DNA molecules (3). Perhaps proteins like CAF1 and PCNA have a replication-coupled role in the inheritance of heterochromatin at HMR, or possibly in its maintenance.

The results of this study have reframed the essential outstanding issues in establishing heterochromatin. One goal now is to learn what replication-independent event happens in this cell cycle window that is essential for silencing. The second goal is to uncover how replication proteins play a role in silencing when replication itself is not required. Both questions should provide fundamental insights into how cells assemble specific structures of chromatin in a spatially and temporally organized manner.

  • * To whom correspondence should be addressed. E-mail: jrine{at}


View Abstract

Navigate This Article