|Home | About | Journals | Submit | Contact Us | Français|
The silenced chromatin at the cryptic mating-type loci (HML and HMR) of Saccharomyces cerevisiae requires a cell cycle event between early S phase and G2/M phase to achieve repression. Although DNA replication per se is not essential for silencing, mutations in many of the proteins involved in DNA replication affect silencing. Each of the four silencers, which flank the silenced loci, includes an origin recognition complex (ORC) binding site (ACS). ORC directly interacted with Sir1 and recruits Sir1 to the silencers. This study describes additional roles for ORC in the architecture of silenced chromatin. Using chromatin immunoprecipitation (ChIP) analysis, we found that ORC physically interacts throughout the internal regions of HMR as well as with silencers. This interaction depended on the presence of Sir proteins and, in part, on the HMR-I silencer. ORC remained associated with the internal regions of HMR even when these regions were recombinationally separated from the silencers. Moreover, ORC could be recruited to the silencers lacking an ACS through its Sir1 interaction.
One mechanism for silencing in eukaryotes involves the formation of heterochromatin, which blocks transcription in a region-specific and non-gene-specific fashion. Once established, silencing in such regions is stably maintained and inherited through multiple cell divisions despite the potentially disruptive effects of DNA replication, recombination, and repair. In Saccharomyces cerevisiae, heterochromatin is found in three locations: telomeres (24), ribosomal DNA (rDNA) (54), and the silent mating-type loci HML and HMR (51), which contain functional copies of MATα and MATa genes, respectively. Silencing of HML and HMR is crucial for proper haploid cell identity (26).
Transcriptional silencing of HMLα and HMRa is controlled by the E and I silencers that flank HML and HMR loci (7). Silencers contain binding sites for the origin recognition complex (ORC), Rap1p, and/or Abf1p, which together recruit Sir1, Sir2, Sir3, and Sir4 proteins, which are essential for initiating and spreading heterochromatin (8, 42, 48, 51). Deletion of SIR2, SIR3, or SIR4 completely abolishes silencing; however, deletion of SIR1 results in a population of cells in which HML and HMR are silenced in some cells, but not in others. Both states of HML and HMR are heritable in sir1 mutants. This and other observations led to the view that Sir1 is required primarily for the establishment of transcriptional silencing (44), whereas Sir2 to Sir4 are required for both the establishment and maintenance of silencing (4, 41). At HML and HMR, Sir1 associates mostly with chromatin at the silencers, whereas Sir2, Sir3, and Sir4 associate with chromatin throughout the silenced region (51). Sir2 is the only protein among the Sir proteins with both structural and enzymatic roles in silencing (30, 31, 35). Once recruited to the silencers via multiple interactions between the silencer binding proteins and other Sir proteins, Sir2 deacetylates the N-terminal tails of histones H3 and H4 of the nearby nucleosomes (9). These hypoacetylated N-terminal tails of histones provide new high-affinity binding sites for Sir3 and Sir4, which are in a complex with Sir2 (25). Histone mutants that mimic the hypoacetylated state rescue the binding and spreading of a catalytically inactive Sir2, Sir2-345, at HML and HMR (62). Thus, hypoacetylated histones provide a foundation for silent chromatin assembly.
A classic study (41) revealed an S-phase dependence for the establishment of silencing. Several proteins that have roles in DNA replication, such as ORC (16, 39, 40), PCNA (64), Dna2, Asf1, and Cac1, also contribute to HML and HMR silencing (15, 32, 53, 55). These findings suggested that DNA replication was required for establishment of transcriptional silencing. However, multiple studies show that neither the initiation of replication at origins that are part of silencers nor the passage of a DNA replication fork through the HML and HMR loci is required for establishment of silencing (30, 34, 37). Nevertheless, each silencer flanking the HML and HMR loci has an ORC binding ARS consensus sequence, and mutations in different ORC subunits lead to decreased silencing (16, 39, 60). ORC directly interacts with Sir1, but no such interaction was detected between ORC and other Sir proteins (17, 63). In addition, ORC's role in silencing can be bypassed by tethering Sir1 to the silencer through Gal4 binding sites (17). These findings suggest that ORC's only role in silencing is recruitment of Sir1 to silencers. If that were the limit of ORC's role in silencing, then mutations in ORC should have no impact on silencing if Sir1 were recruited to silencers by other means.
To understand more fully the effects of ORC on the HML and HMR chromatin and to characterize further the role of ORC in silencing, we measured the silencing level in cells with orc mutations and various configurations of silencers. These experiments revealed unanticipated links between ORC, silencing, and the architecture of the silenced chromatin.
Strains used in this study were isogenic to S. cerevisiae W303 unless otherwise indicated (Table (Table1).1). Gene deletions were done using one-step integration of PCR-amplified knockout cassettes (22, 38) and confirmed by PCR and phenotypic validation. Epitope tagging for immunoprecipitations was done similarly using tandem affinity purification (TAP) tag cassettes (47) or constructs described previously (38). Oligonucleotide primer sets used for PCR, tagging, and knockout in this study are shown in Table Table22.
Rich medium (YPD) and minimal medium (YM) are described previously (52). KanMX4 resistance marker was selected on YPD containing 200 mg/liter of G418 (Geneticin). Modified lithium acetate transformation was used as described previously (3).
After each culture was grown to late log phase, it was diluted to a final optical density at 600 nm (OD600) of 1. Threefold serial dilutions of each culture of interest were spotted onto YPD or synthetic complete medium lacking uracil (SC-Ura) plates for growth and on YM plates sprayed with excess MATα mating tester strain (JRY2726) and incubated at 24°C for 3 days.
Total RNA was prepared using an RNeasy minikit from Qiagen. Genomic DNA was digested on the column using RNase-free DNase (Qiagen). Oligo(dT) primer-directed cDNA was synthesized using the SuperScript III first-strand synthesis system for reverse transcriptase PCR (RT-PCR) kit (Invitrogen). Quantitative PCR (QPCR) analysis was done at least in triplicate on 3 or more independent RNA preparations of each strain. QPCR was performed on a MX3000P machine (Stratagene) using SYBR GreenER QPCR super mix (Invitrogen).
Chromatin immunoprecipitation (ChIP) analysis was done as previously described (12) with minor modifications. Two hundred fifty milliliters of mid-log-phase cultures was used to prepare whole-cell extracts, and 400 μl of this extract used for immunoprecipitation (IP) and 40 μl was used for the input sample. A 50-μl slurry of immunoglobulin G-Sepharose (Amersham Biosciences) per IP was used for TAP-tagged proteins. Agarose-conjugated mouse monoclonal anti-hemagglutinin (anti-HA) (Sigma) was used to IP HA-tagged proteins. For immunoprecipitations with rabbit polyclonal ORC antibody (generous gift from Stephen Bell), 2 μl of antibody was coupled to 50 μl of protein A slurry (Upstate) for 2 h. All immunoprecipitations were done overnight at 4°C. Each ChIP experiment was done in triplicate or more for independently prepared whole-cell extracts. For experiments where no enrichment was detected (see Fig. Fig.8B),8B), ChIP was done in duplicate.
The whole-cell extracts prepared for ChIP analysis were used for DNA blot hybridization analysis. The whole-cell extracts were first digested with proteinase K for 2 h at 37°C and then extracted with phenol-chloroform. After isopropanol precipitation and a 70% ethanol wash, each pellet was resuspended in 50 μl of water. About 15 μg of each sample was electrophoretically separated in a 2% agarose gel and then transferred to a Hybond N membrane. Probes of interest were prepared by PCR and then radiolabeled using [α-32P]dCTP with Amersham RediPrime random prime labeling system (GE Healthcare). DNA blot analysis was done as previously described (56). Blots were analyzed with a Typhoon scanner and ImageQuant software. The most frequently occurring signal value, the mode, was determined for each lane. To evaluate the extent of shearing among different DNA samples, the signal average was determined for each 0.2 mm of each lane on the gel by multiplying the number of signal counts for each fragment by the corresponding size of the fragment and then dividing this value by the total number of counts. The final averages were compared to determine the average shearing difference.
Yeast whole-cell extracts were precipitated using 20% trichloroacetic acid (TCA) and solubilized in SDS loading buffer. Anti-Flag antibody from rabbit (Sigma) was used to detect TAP-tagged proteins, and anti-HA antibody from mouse (Sigma) was used to detect HA-tagged Sir1 (Sir1-HA). The ORC was detected using polyclonal ORC antibody from rabbit. TAP-tagged Sir2 (Sir2-TAP) immunoprecipitations were as previously described (2).
Previous work identified ORC's role in silencing as recruitment of Sir1 protein to the silencers. Indeed, the first 235 amino acids of Orc1, the largest subunit of the ORC complex, contain a bromo-adjacent homology (BAH) domain that interacts with a specific domain in Sir1, known as the ORC interaction region (OIR), but not with other Sir proteins (6, 27, 63). This study revealed additional roles for ORC in silencing and in the architecture of the silenced chromatin.
Two ORC mutations, orc5-1 and orc2-1, cause loss of silencing at HML and HMR (16, 18, 39, 58), particularly in strains in which silencing has already been compromised or in strains with synthetic silencers. The silencing defect of these orc mutations can be bypassed by tethering a Gal4 DNA binding domain-Sir1 fusion protein (Gal4-Sir1) to the silencer through multiple Gal4 binding sites in place of the ORC binding site (ACS) in the HMR-E silencer (17).
If ORC's role in silencing were exclusively to recruit Sir1, then in the absence of Sir1, orc mutations should have no further effect on silencing. The partial defect of sir1Δ mutants allowed us to test whether orc5-1 and orc2-1 mutations had any effects on silencing in sir1Δ orc double mutants. Unexpectedly, the orc5-1 mutation in combination with sir1Δ had a more pronounced silencing defect as measured by mating efficiency than did the sir1Δ mutant or individual orc mutants, indicating that ORC had a role(s) in silencing HMR beyond Sir1 recruitment (Fig. (Fig.1A1A).
To provide a molecular assessment of the effects of orc mutations on residual silencing in sir1Δ mutant cells, we measured the a1 transcript level from HMRa1 by quantitative reverse transcription-PCR (QRT-PCR) in these mutants. As expected, orc2-1 and orc5-1 mutations by themselves had no detectable effect on silencing in cells with wild-type silencers grown at the permissive temperature (Fig. (Fig.1B).1B). However, when combined with sir1Δ in cells incubated at the semipermissive temperature (30°C) for 3 h, a pronounced effect on silencing was observed in both orc mutants, while not affecting the cell viability (data not shown). In agreement with the mating assay, the orc5-1 mutation enhanced the silencing defect of sir1Δ. On the other hand, orc2-1 mutation did not show a significant effect on silencing in the absence of sir1.
To explore the mechanism by which orc mutations affect silencing independently of Sir1, we measured their effect on the association of Sir proteins at HMR and HML using ChIP followed by QPCR analysis (Fig. (Fig.2).2). We evaluated the levels of Sir1, Sir2, and Sir3 proteins at the HML-E and HMR-E silencers and at positions internal to the silenced domain, corresponding to HMRa1 and HMLα1, which were approximately 1.4 kb from silencers. As reported previously, Sir3 is enriched at both silencers and in internal regions of HML and HMR. Neither the orc2-1 mutation nor the orc5-1 mutation had any notable effect on Sir3 occupancy at these positions (Fig. (Fig.2A2A).
Sir1 data offered a somewhat different perspective. As expected, Sir1 was much more enriched at the HML-E silencer than at HMLα1. Interestingly, the orc5-1 mutation resulted in substantial reduction of Sir1 at the HML-E silencer, with a similar though quantitatively smaller impact caused by orc2-1. At HMR, the Sir1 enrichment at the HMR-E silencer was as expected; however, there was still substantial enrichment of Sir1 at the internal HMRa1. Enrichment of Sir1 at both HMR positions was somewhat sensitive to orc5-1 and less so to orc2-1 (Fig. (Fig.2B2B).
The Sir2 localization at HML and HMR in wild-type cells was similar to that of Sir3. However, in contrast to Sir1 and Sir3, Sir2 exhibited a pronounced sensitivity to orc2-1 and much less so to orc5-1 (Fig. (Fig.2C2C).
Sir2 interactions with the silenced domain in sir1Δ mutant strains were reduced in the orc5-1 sir1Δ double mutant in accordance with the further decreased silencing in this strain. However, orc2-1 sir1Δ had very little effect on silencing relative to sir1Δ itself, but the effect of orc2-1 sir1Δ on Sir2-TAP levels at HMR was significant (about 3-fold less than in the sir1Δ strain) (Fig. (Fig.2D).2D). The synergistic effect of orc2-1 and orc5-1 in combination with sir1Δ on Sir2 ChIP levels was unanticipated by all earlier studies of ORC's role in silencing. ChIP analysis for ORC in these mutations showed that, when polyclonal ORC antibody was used, there was more ORC interaction at HMR-E in the orc2-1 sir1Δ double mutant than in either the sir1Δ or orc5-1 sir1Δ mutant strain (Fig. (Fig.2E).2E). This was surprising because the ORC complex in the orc2-1 strain is unstable, and except for the Orc1 subunit, the levels of other subunits were reduced (Fig. (Fig.2F).2F). In fact, the level of Orc5 as reflected in ChIP analysis was greatly reduced in the orc2-1 strain (data not shown), suggesting that the enhanced ORC ChIP in the orc2-1 mutant was probably contributed by the Orc1 subunit. Taken altogether, the data suggest that although both orc mutations resulted in a drop of Sir2 recruitment at HMR, which would negatively affect silencing, enhanced recruitment of Orc1 in the orc2-1 strain may increase silencing through a Sir2-independent way.
To test whether orc mutations had any adverse effects on the overall level of Sir proteins in cells, we evaluated the level of epitope-tagged Sir proteins. Sir1-HA and Sir3-TAP levels were similar in different wild-type and orc mutant strains (Fig. 2G and H). Due to its low abundance, it was difficult to analyze the levels of Sir2-TAP by immunoblotting whole-cell extracts (data not shown). Therefore, to test whether Sir2-TAP was expressed and intact in all the strains, we immunoprecipitated Sir2 and evaluated its integrity by immunoblotting. In all strains, Sir2 was resolved into 3 species (Fig. (Fig.2I).2I). The major species was the expected size for Sir2-TAP (80 kDa). At least one of the other bands was observed in previous Sir2 immunoprecipitations and was presumed to be a degradation product (23). However, although Sir2 degradation was quite low in wild-type and orc5-1 strains, it increased significantly in orc2-1 strains. Despite a considerable drop in the level of intact Sir2 and in the Sir2 ChIP levels at HMR (Fig. (Fig.2C),2C), the silencing in orc2-1 sir1Δ double mutants was comparable to that in sir1Δ mutants (Fig. (Fig.11).
The synergistic effect of sir1Δ orc2-1 double mutants on Sir2 levels at silencers was unexpected. orc5-1 alone or in combination with sir1Δ had little effect on the level of Sir2 protein in cells (Fig. (Fig.2I),2I), ruling out a trivial explanation for the reduced Sir2 occupancy in Fig. Fig.2D.2D. The cause of reduced Sir2 levels in orc2-1 mutants remains unexplained.
ORC is expected to bind at the silencers because each silencer includes at least one ACS sequence, but ORC's binding to both silencers and silenced DNA has not been extensively evaluated before. Prompted by the Sir1 association at HMLα1 and HMRa1 (Fig. (Fig.2B),2B), we speculated that ORC might have a similar distribution due to its interaction with Sir1. To test the association of ORC throughout the silenced domain at HMR, HA epitope-tagged Orc5 (Orc5-HA) and Orc1 (Orc1-HA) subunits, as well as the whole ORC complex, were immunoprecipitated using anti-HA antibodies and polyclonal ORC antibodies, and the immunoprecipitates were evaluated by ChIP analysis. All of these assays showed similar and unexpected results. The ORC was found at HMR-E and HMR-I, as expected, both of which are bona fide chromosomal origins of replication (29, 49). However, ORC was also associated throughout all 3 kb of the silenced domain (Fig. 3B and C). Qualitatively similar results were obtained at HML at which neither silencer is a bona fide origin of replication (Fig. (Fig.3E3E).
Although DNase I footprinting studies show a discrete binding of ORC to the ACS of origins (36), the ORC ChIP signal throughout HML and HMR raised the possibility that ORC might have unexpected binding properties as measured by ChIP analysis at other sites, such as origins of replication not related to silencing. Hence we evaluated the same immunoprecipitates for ORC binding to regions flanking a bona fide origin on the same chromosome as HML and HMR (ARS305). As at the silencers, ORC bound strongly with ARS305. However, in contrast to HML and HMR, ORC did not detectably interact with sequences only 1 kb from ARS305 (Fig. 3D and E). Therefore, the chromatin association of ORC at HMR was distinctly unlike that observed for ORC at a bona fide origin of replication.
The silenced chromatin domain extends beyond the silencers and ends at the boundary elements on either side of HMR (13, 14). ORC association followed a similar pattern (Fig. 3B and C), suggesting the possibility of an expanded role for ORC in HMR silencing. We tested whether Sir proteins and hence silencing were required for ORC to ChIP internal regions of HMR in strains lacking SIR1, SIR2, or both. Deletion of SIR2 completely abolishes silencing, whereas deletion of SIR1 produces cells in one of two mitotically stable phenotypes as described above. Individual deletions of SIR1 or SIR2 decreased ChIP levels of ORC both at HMR-E and at the internal HMRa1 site (Fig. (Fig.4A).4A). Deletion of SIR3 and SIR4 gave effects similar to that of SIR2 deletion and had modest effects on ORC levels at HMR-E but resulted in a 2- to 3-fold drop at HMRa1 (data not shown). Deletion of the Sir proteins did not decrease ORC recruitment at ARS305 (Fig. (Fig.4B)4B) or ARS1 (data not shown). Thus, silencing and/or Sir proteins were required for ORC association throughout HMR. Surprisingly, the combination of sir1Δ and sir2Δ showed a synergistic effect and led to further decrease in ORC ChIP levels relative to sir2Δ alone. Hence, Sir1 contributed to ORC's association with internal regions of HMR even in cells completely defective in silencing HMR. Due to the presence of ORC binding sites at the silencers, ORC could be immunoprecipitated at silencers even in sir1Δ sir2Δ double mutant cells; however, no ORC was detected at the internal regions of HMR in the double mutants.
Recent work suggests that an interaction between the HMR-E and HMR-I silencers enable Rap1-bound HMR-E to interact with the internal regions of HMR as well as HMR-I (61). To test whether the ORC association throughout HMR was dependent upon this silencer interaction as well, we analyzed ORC distribution throughout HMR in cells lacking the HMR-I silencer (Fig. (Fig.4C).4C). Although the ChIP levels of ORC for the internal regions decreased approximately 2-fold (compare HMRa1 levels in Fig. Fig.4A4A to enrichment at locus F in Fig. Fig.4C),4C), there was still 4-fold more ORC at 2 kb from HMR-E than there was at 1 kb from ARS305. Thus, although HMR-I contributed to ORC interaction at HMRa1, significant association remained in the absence of HMR-I.
The unexpected interaction of ORC throughout silenced chromatin led us to consider various alternative explanations other than direct association of ORC with the heterochromatin at these positions. During ChIP analysis, before the immunoprecipitation step, the formaldehyde cross-linked chromatin is sheared, resulting in DNA fragments of 0.2 to 1 kb in length. It has been widely assumed that all structures of chromatin are equally vulnerable to physical shearing by sonication. If the condensed nature of heterochromatic regions were sufficiently refractory to physical shearing, resulting in larger DNA fragments being recovered by immunoprecipitation, then proteins bound to specific sites within heterochromatin might give the illusion of binding across a wider region. Therefore, we tested whether ORC's apparent association with sequences far from its binding sites was due solely to variation in the ease with which different structures of chromatin are sheared by sonication. To address this possibility, whole-cell extracts of wild-type and HMR-IΔ, and sir1Δ sir2Δ mutant cells were prepared as for ChIP analysis. The resulting sheared genomic DNA was electrophoretically separated, blotted onto a membrane, and hybridized with radiolabeled probes from euchromatic (ACT1) and heterochromatic (HMRa1) regions (Fig. (Fig.5).5). The overall shearing of the DNA was similar in the wild-type strain and sir1Δ sir2Δ double mutant and was slightly more extensive in DNA from HMR-IΔ, as judged from the gel image (Fig. (Fig.5A).5A). ACT1 probe hybridization showed a pattern similar to the overall shearing pattern, suggesting that shearing of chromatin containing the ACT1 gene was not affected by these genotypes (Fig. (Fig.5C).5C). In contrast, the a1 probe hybridized to a slower-migrating smear in DNA from wild-type cells compared to the DNA from silencing-deficient cells (Fig. (Fig.5B).5B). The data in this image were quantified by analyzing the signal counts for each fragment size (in 0.2-mm bins) in the autoradiograms (Fig. 5D to F). These graphs clearly indicated the similarity in bulk DNA shearing from these strains and some shearing resistance at HMRa1 in wild-type versus silencing-defective strains. We used the data in these graphs to calculate the modes of each sample. Although ACT1 showed some variation in the mode due to sample-to-sample shearing variations, the wild-type strain showed consistently higher modes for the HMRa1 locus than the sir1Δ sir2Δ double mutant strain did (Fig. (Fig.5G5G).
To determine whether this extent of shearing resistance was adequate to account for the data about ORC distributions, we calculated the total number of fragments that were large enough to include HMRa1 and one or both of the silencers (those longer than 1,440 bp) after normalizing for shearing differences as explained in Materials and Methods. Such fragments would be immunoprecipitated by the ORC complex bound exclusively at the silencers and would be detected by QPCR analysis for HMRa1. The wild-type strain had only 2-fold-more (40%) of these fragments than the sir1Δ sir2Δ double mutant did (22%), and the difference was much less for HMR-IΔ (29%) (Fig. (Fig.5G).5G). This modest difference in shearing propensity was inadequate to explain the 8-fold ORC enrichment at HMRa1 in wild-type strains. These data indicated that ORC actually interacted with internal regions of HML and HMR and that interaction could not be explained by reduced shearing efficiency in heterochromatin. The data so far were compatible with interaction resulting from either ORC binding along the Sir proteins or with long-range interactions facilitated by higher-order structures of the heterochromatin as described in the next section.
The interactions of Sir proteins throughout silenced chromatin as determined by ChIP analysis are commonly interpreted as spreading these proteins across the chromatin. However, it is also possible that some higher-order chromatin structure brings other sequences close to proteins that bind only at their recognition site, thereby enabling them to physically interact and be cross-linked. Indeed, interactions between the two silencers at HMR allow Rap1 bound to HMR-E to interact through out the silenced region (61). The critical test of whether ORC is actually associated with the internal sequences at HMR is to separate the internal silenced chromatin from the silencers themselves prior to cross-linking and then determine whether ORC is still bound to the silenced chromatin at those internal sequences. For this experiment, we used strains with two recombination sites (RS) that allow excision of HMR as an episome, leaving HMR-E and HMR-I silencers in the chromosome (Fig. (Fig.6A).6A). Upon galactose induction of a heterologous recombinase, site-specific recombination at the recombination sites occurs in most cells within one cell cycle. The silencing on this episome is unstable and lost within 2 h of galactose induction (10; data not shown). Therefore, we analyzed ORC interaction at 60 min past induction, during which time silencing is still robust, even though the “loop out” was only 70% complete at this stage. To clearly distinguish HMR sequences that had been looped out from those in the 30% of chromosomes in which recombination had not yet occurred, we used a primer set that specifically amplifies the excised circle (the circle-specific primer set CSP) to differentiate ORC binding at the episome from the internal sequences left at the chromatin. For a positive control, Sir3 bound both the silencers and episome efficiently, and this interaction was lost in the sir4Δ mutant strain (Fig. (Fig.6B).6B). ORC enrichment at internal regions of this construct while still resident in chromosome was easily detected, though 2-fold lower than observed at wild-type HMR (Fig. (Fig.2).2). The twofold reduction was perhaps due to the RS sites, which introduce an additional 750-bp sequence between the silencers, potentially resulting in less efficient ORC interaction at HMRa1 (Fig. (Fig.6C).6C). Nevertheless, upon loop out, the episome was still enriched in the ORC immunoprecipitate, and this interaction was lost in the sir4Δ mutant strain. Despite an overall decrease in the ChIP signal at HMRa1 of the episome relative to the chromosome, ORC still showed significant interactions with internal regions of HMR on the episome, suggesting that ORC was bound to internal regions in a Sir protein-dependent manner.
One argument for the ORC interaction with the episome (Fig. (Fig.6)6) could be that the excised episome, though covalently separated from silencers, might be restricted in its mobility in the nucleus such that it could still cross-link with the silencer-bound ORC. We tested this possibility by using a synthetic silencer that was previously designed to bypass ORC involvement in silencing. This synthetic silencer has four Gal4 binding sites replacing the ACS and has no HMR-I silencer (HMR-GalSS). In cells with normal Sir1 protein, this silencer is unable to silence HMRa1. However, expression of a Gal4 DNA binding domain fused to Sir1 restores HMR silencing even in cells with mutant ORC genes (17). Although the mechanism for how this “supersilencer” bypassed the functional requirement for ORC is not well understood, it offered an opportunity to address the “association by proximity” hypothesis, testing whether ORC recruited to one position in silenced chromatin could ChIP with another. We first determined whether the synthetic silencer bound by Gal4-Sir1 actually recruited ORC to the HMR-GalSS in the absence of an ARS consensus sequence (Fig. (Fig.7A).7A). As expected, no ORC association was detected in the absence of Gal4-Sir1. In contrast, ORC was efficiently recruited to HMR-E in cells expressing Gal4-Sir1, and the interaction was dependent on the BAH domain in the N-terminal region of Orc1 known to interact with Sir1. This result established the ability of tethered Sir1 to recruit ORC to HMR in the absence of ORC's binding site. However, in this context, ChIP analysis revealed that ORC interacted only at the silencer and not at the internal HMRa1 position.
A potential explanation for the absence of ORC's internal interaction in this synthetic construct could be the lack of near-match ACS sequences present next to the core HMR-E silencer. It was possible that, even though these sequences are not sufficient for ORC's long-range interactions at HMR in cells lacking HMR-I (as suggested by the results shown in Fig. Fig.4A),4A), they may still contribute to binding of ORC at the internal regions of HMR. Therefore, we analyzed ORC ChIP levels at a silencer which includes only the 138-bp region of HMR-E silencer that has the ARS and Rap1 and Abf1 binding sites but lacks all the other near-match ACS sequences. Thus, this core silencer was trimmed of all the flanking sequences and near matches to the ARS consensus sequence and was directly comparable in structure to the synthetic silencer used in Fig. Fig.7.7. This trimmed core silencer showed similar ORC enrichment at HMRa1 to that of its parent strain, which had the wild-type silencer (data not shown). Thus, the ORC binding site in the silencer itself was sufficient to mediate associations with internal sequences of HMR without benefit of near matches or other ORC binding sites in its vicinity. Although the mechanism behind the ORC independence of silencing in this context (Fig. (Fig.7B)7B) is unclear, the abundant ORC recruitment to the silencer through tethered Sir1 did not lead to significant interaction between ORC and the internal sequence HMRa1. Therefore, ORC enrichment with the episome in Fig. Fig.6C6C was most simply explained by ORC's physical presence on the episome rather than the physical proximity of the episome to ORC bound at silencers.
Since ORC enrichment at internal regions of HMR relied on silencing, we wondered whether ORC could be recruited to silenced domains per se, even in the absence of an ACS. Subtelomeric sequences and polymerase II (PolII)-transcribed genes in rDNA are silenced in yeast; however, both of these regions harbor at least one ACS that can recruit ORC, precluding a direct test of ACS-independent recruitment of ORC. However, an array of a simple repetitive DNA, C1-3A, which lacks an ACS, is apparently sufficient to assemble silent chromatin in yeast (57). These internal tracts should allow a distinction between ACS-dependent and Sir-dependent ORC recruitment. Moreover, silencing of a URA3 reporter gene by internal C1-3A repeats is Sir1 independent. We performed ChIP analysis for Orc5-HA enrichment at 2 different sequences (A and B) that were proximal to the C1-3A repeats and the reporter gene (URA3) that was silenced by these tracts (Fig. (Fig.8A).8A). Neither sequence enriched in Orc5-HA immunoprecipitation. Hence, at least as defined by this assay, it was possible to assemble Sir-dependent silenced chromatin without the recruitment of ORC. However, as discussed below, the strength of silencing mediated by this C1-3A array is considerably weaker than that evident at HML and HMR.
Previous work and the results in Fig. Fig.7A7A suggested a strong interaction between Sir1 and Orc1 (6, 17, 27, 63 ) and raised the possibility that ORC would recruit Sir1 to chromosomal origins of replication. However, ChIP analysis for Sir1-HA showed no enrichment for any of the six ARS sequences tested (Fig. (Fig.8B)8B) in accordance with earlier data of Gardner and Fox (19).
The work presented here offered an expanded view of the interaction between ORC and silenced chromatin. Previously, the only known role for ORC in silencing was by binding to ARS sequences in silencers, thereby recruiting Sir1 to silencers through binding between Sir1 and the BAH domain of Orc1 (27, 63). Indeed, previous studies also concluded that the requirement for ORC was bypassed when Gal4-Sir1 was tethered to a synthetic silencer because under such conditions, the effects of orc mutations on silencing are suppressed (17). These findings favored the idea that Sir1 may act only at the silencers and, in contrast to Sir2, Sir3, and Sir4, not spread throughout the silenced domain. In fact, in such strains with Gal4-Sir1 bound to the HMR-E silencer, wild-type Sir1 protein in the same cells was not detectable at the silencer or within the silenced domains (50). The work reported here revealed that results with tethered Sir1 protein obscured unexpected complexity regarding the interactions of Sir1 and ORC in silenced chromatin.
The critical observation that led to this study was that the orc5-1 mutation reduced silencing in sir1Δ cells (Fig. 1A and B) and reduced recruitment of Sir2 at HMR and HML (Fig. 2C and D). At face value, that result required that ORC have additional contributions to silencing. Moreover, we found that tethered Gal4-Sir1 fusion protein could recruit ORC to chromatin lacking an ACS (Fig. (Fig.7A).7A). Our efforts to understand these and other phenotypes led to several surprising conclusions.
ChIP assays of Sir1 in wild-type cells revealed that Sir1 was associated with the silencers, as expected, but also with internal regions of HML and HMR. Unlike Sir2, Sir3, and Sir4, which interact as a complex, Sir1 interacts only with ORC and Sir4 (6). Nevertheless, the Sir1 association we detected throughout HML and HMR was similar to that of the other Sir proteins. Sir1's physical association throughout the HML and HMR loci, combined with its ability to recruit ORC, led to our discovery that epitope-tagged Orc5 and Orc1 subunits, and presumably the entire ORC complex, associated with chromatin throughout the HML and HMR loci (Fig. (Fig.33).
Because all previous studies of ORC revealed it to bind only to ARS sequences, its association throughout the silenced domain far from the ARS sequences of silencers led us to explore several explanations for how ORC could appear to interact throughout silenced domains yet actually be bound only to ARS sequences. In brief, we considered whether ORC had peculiar properties in ChIP analyses when bound to ARS sequences, whether silenced chromatin is intrinsically resistant to shearing so as to confound ChIP assays, and whether higher-order chromatin structures at HMR could poise internal sequences of HMR within “cross-linkable distance” of ORC bound to silencers. Our data excluded all three possibilities as an explanation for how ORC associated with sequences throughout the silenced domain: ORC bound to ARS305 behaved in ChIP analyses in a manner expected for a site-specific DNA binding protein, in contrast to the regional associations detected at HML and HMR. We found evidence for a slight shearing resistance of silenced chromatin, but our quantitative analysis indicated that such resistance made, at best, only a modest contribution to the interactions of ORC throughout silenced chromatin (Fig. (Fig.5).5). Finally, we found that the interaction between the E and I silencers revealed by chromosome configuration capture (3C) analysis could not account for the interaction of ORC with HMR sequences recombinationally severed from the silencers. (In principle, we cannot exclude the possibility that long-range interactions accounted for a portion of the signal in our studies due to small quantitative differences in the signal produced from different experimental designs. Indeed, the effect caused by the deletion of the HMR-I silencer may well have affected the stability of the interactions of ORC throughout the silenced domain independently of E-I interactions.)
In support of our results, an independent set of experiments using the ChIP-Seq method to evaluate ORC distribution genome-wide with its higher resolution compared to traditional ChIP methods has also detected ORC signal across HML and HMR, including sequences not associated with the ARS elements at silencers (David MacAlpine and Stephen Bell, personal communication). Recent work from our lab establishes that heterochromatic regions are prone to less enrichment in ChIP-Seq analysis due to shearing resistance by sonication and subsequent elimination at the size selection step (59). Despite this bias against the heterochromatic regions, ORC interaction was detected across the entire HML and HMR loci.
Having exhausted alternative explanations for ORC's association throughout the silenced domain, we tested directly whether ORC molecules associated with sequences within the silenced domains when those sequences were no longer physically linked to the silencers themselves. Using the site-specific recombinase method to excise the internal sequences of HMR as silenced chromatin on an episome, we found that, like the Sir proteins, ORC was clearly associated with these episomal sequences (Fig. (Fig.6).6). Nevertheless, there was a quantitative difference between the extent of association with these excised sequences and with the same sequences left in the chromosome. However, a similar difference was also observed for Sir3 association. The reason for the reduced associations in this experimental design could be because silencing is somewhat labile on the excised episome, with some loss of silencing evident within 60 min (data not shown), or due to an influence on the RS sites added to allow for excision.
An alternative argument for the interaction between ORC and the episome was that the silencer-bound ORC on the chromatin could still interact with the excised episome because of physical proximity. This argument was discounted with an experiment with a synthetic silencer, whose silencing function is independent of ORC. The ORC was efficiently recruited to this silencer through Orc1 and Sir1 interactions, but ORC recruited in this way did not show any interactions with the internal regions (Fig. (Fig.7A).7A). Therefore, the ORC interaction we detected with the episome was most simply explained by ORC's physical presence on the episome.
Formally, it was possible that, unlike the DNA sequence flanking conventional ARS elements, there was some special property of the sequences that make up HML and HMR that allowed ORC to bind these sequences in the absence of recognizable binding sites. Indeed, ORC from larger eukaryotes typically lacks sequence-specific binding sites as origins (21). However, ORC's association with sequences throughout HML and HMR was dependent upon silencing per se (Fig. (Fig.4A)4A) (data not shown) and not simply upon the DNA sequence of the silenced domain. In fact, our analysis showed that the deletion of an almost 700-bp sequence that flanks the HMR-E core silencer results in an internal ORC enrichment that is similar to the wild-type strain (data not shown). In silencing-defective cells, ORC still bound the silencers, presumably through the ARS sequences of the silencers. Thus, it seemed inescapable that ORC itself and Sir1 were structural components of silenced chromatin. The data also suggested that the combination of an ACS in the silencer and Sir1 contributed to the extent of ORC association with the silencers (Fig. (Fig.4A4A).
At present, we do not know how the interactions between ORC and Sir1 are mediated throughout the silenced domain, except that it was probably not through direct DNA binding. Presumably the BAH domain of Orc1 associates with Sir1 throughout HMR, just as it does at silencers. Because Orc1 and Sir3 are paralogs (5), perhaps Orc1 enjoys interactions within silenced chromatin similar to those used by Sir3. Indeed, perhaps ORC binds H3 and H4 histone tails deacetylated at those positions that are the targets of Sir2. It would be interesting to evaluate whether ORC has a different distribution profile in cells lacking the Sas2 acetyltransferase, whose acetylation of histone tails blocks Sir3 binding (33).
The results presented here have forced a reevaluation of the previous idea that ORC mutations affect silencing primarily through their effect on recruiting Sir1 to the silencer. That view was inspired by two independent studies (11, 17) showing that a Gal4-Sir1 fusion protein tethered to the silencer through Gal4 binding sites could suppress the silencing defect of ORC mutations. The fundamental question in such experiments is whether the constructs used to reveal some of the functions of a regulatory site have the capacity to obscure part of the story. Particularly for this synthetic silencer, where Gal4-Sir1 is tethered through four Gal4 binding sites, the silencing may be more robustly established than with a wild-type silencer and hence mask any silencing defect caused by the absence of ORC recruitment.
Our ChIP analysis showed that the whole ORC complex could be recruited to regions lacking a DNA binding site for ORC in a Gal4-Sir1-dependent way and that recruitment was lost in strains lacking the Sir1 interaction domain of Orc1 (N235Δorc1) (Fig. (Fig.7).7). Loss of the ORC recruitment in N235Δorc1 did not affect the silencing adversely as measured by a1 transcript levels, arguing that in the context of this synthetic construct, the absence of ORC had no demonstrable effect on silencing. Nevertheless, the additional loss of silencing caused by orc5-1 in sir1Δ strains (Fig. (Fig.1)1) was free of any complications caused by use of a silencer with multiple Gal4-Sir1 binding sites and is unequivocal in identifying a Sir1-independent role of ORC in silencing. Although the precise nature of this role was incompletely resolved, ORC's absence from internal regions of HMR in cells with Gal4-Sir1 tethered to the silencer indicated that the occupancy of internal sequences by ORC depends upon ORC binding sites at the HMR-E silencer.
The control experiments in this study uncovered a 2-fold relative shearing resistance of silenced chromatin (Fig. (Fig.5).5). A similar resistance is reported for nucleoli in human and mouse cell lines, presumably due to its extensive decoration with proteins and RNA (46). Although shearing resistance could not account for the presence of ORC at HMRa1 in the ChIP experiments, this resistance could nevertheless prove useful in analyses of genomes. The power of deep sequencing enabled by the recent generation of DNA sequencers suggests that variation in the shearing of different chromatin structures will be useful for detecting interesting domains of chromatin structure that will be recognizable by underrepresentation (and perhaps overrepresentation) of regions of the genome in whole-genome sequence reads. Indeed, we and others have recently exploited such impacts on shearing differences to evaluate deep-sequencing strategies for detecting structural features on the chromosomes of yeast and humans (1, 59).
If the complementary surfaces on Sir1 and Orc1 responsible for the physical interaction of these two proteins (4, 20, 63) were sufficient to explain their interaction, then the presence of one protein would predict the presence of the other. One half of this prediction was confirmed. A tethered Gal4-Sir1 protein recruited ORC to a silencer completely lacking an ACS site. However, as previously reported elsewhere (19), we were unable to detect Sir1 at any of the five different bona fide origins of replication (Fig. (Fig.8B).8B). In principle, these data could be explained if something else occupies the BAH domain of Orc1 at origins. If so, such an interaction would probably not be necessary for origin function, since HMR-E and HMR-I are bona fide origins in cells. Alternatively, perhaps something else at silencers, such as Rap1, contributes to forming a complex between ORC and Sir1. At face value, the inability of the C1-3A repeats, which are, in effect, a series of Rap1 binding sites to recruit either ORC or Sir1 would challenge that model. However, the rather weak silencing achieved by these repeats leaves room to consider this possibility. The bottom line is that the source of specificity for ORC-Sir1 interactions remains an interesting and unexplained problem.
The loss of silencing caused by mutations in ORC subunits is clear, but the magnitude of these effects has, in all cases, been small except in strains sensitized to silencing defects. Moreover, a sufficient number of tethered Gal4-Sir1 fusion proteins can recruit ORC to a silencer, but not to internal regions of HMR, yet silencing seems robust in the absence of ORC under these circumstances. So what is the role(s) of ORC at internal regions of HML and HMR?
The most obvious possibility would be some role in mating-type interconversion. Interconversion occurs by a gene conversion mechanism, which requires DNA synthesis, but does not require an origin of replication. Perhaps ORC's presence at HML and HMR and ORC's interaction with other proteins, such as the minichromosome maintenance (MCM) proteins, contribute to the remarkable efficiency of interconversion. Alternatively, perhaps ORC's integration of cell cycle signals, such as those from Cdc28, helps sharpen temporal control of the HO endonuclease to the small temporal window between start and replication of HML and HMR. The tethered Gal4-Sir1 protein in cells lacking the N terminus of Orc1 would provide a useful context for testing these ideas.
In other organisms, ORC also interacts with heterochromatic domains. For example, Drosophila melanogaster Orc2 colocalizes with the heterochromatin protein HP1 in heterochromatin (43), and mutations of orc2 cause a defect in recruiting HP1 to heterochromatin, establishing a role for ORC in heterochromatin structure at a global scale (28). In human cell lines, Orc2 is tightly bound to the heterochromatin and to HP1α and HP1β (45), and this localization is specific to G1-phase and early S-phase cells. Although ORC in S. cerevisiae occupies origins at all stages of the cell cycle, posttranslational regulation of its activity offers potential insight into the cell cycle regulation of events at HML and HMR that have so far eluded detection.
We thank Marc Gartenberg, Virginia Zakian, and Stephen Bell for strains and reagents critical to the execution of these experiments. We thank the members of our lab and Michael Botchan for helpful discussions and comments on the manuscript. We also thank Stephen Bell and David MacAlpine for sharing their unpublished data.
This work was supported by a grant from the National Institute of Health (NIH GMS 31105).
Published ahead of print on 30 November 2009.