Extended heterochromatin domains, which are repressive to transcription and help define centromeres and telomeres, are formed through specific interactions between silencing proteins and nucleosomes. This study reveals that in Saccharomyces cerevisiae, the same nucleosomal surface is critical for the formation of multiple types of heterochromatin, but not for local repression mediated by a related transcriptional repressor. Thus, this region of the nucleosome may be generally important to long-range silencing. In S. cerevisiae, the Sir proteins perform long-range silencing, whereas the Sum1 complex acts locally to repress specific genes. A mutant form of Sum1p, Sum1-1p, achieves silencing in the absence of Sir proteins. A genetic screen identified mutations in histones H3 and H4 that disrupt Sum1-1 silencing and fall in regions of the nucleosome previously known to disrupt Sir silencing and rDNA silencing. In contrast, no mutations were identified that disrupt wild-type Sum1 repression. Mutations that disrupt silencing fall in two regions of the nucleosome, the tip of the H3 tail and a surface of the nucleosomal core (LRS domain) and the adjacent base of the H4 tail. The LRS/H4 tail region interacts with the Sir3p bromo-adjacent homology (BAH) domain to facilitate Sir silencing. By analogy, this study is consistent with the LRS/H4 tail region interacting with Orc1p, a paralog of Sir3p, to facilitate Sum1-1 silencing. Thus, the LRS/H4 tail region of the nucleosome may be relatively accessible and facilitate interactions between silencing proteins and nucleosomes to stabilize long-range silencing.
THE formation of silenced chromatin, or heterochromatin, in eukaryotes is important for proper gene regulation and chromosome stability and helps define centromeres and telomeres. One interesting property of heterochromatin is its capacity to spread along a chromosome to form an extended, repressive domain. This spreading is enabled by specific interactions between silencing proteins and nucleosomes. Consequently, particular surfaces on the nucleosome can be critical to the assembly of heterochromatin. This study reveals that in the budding yeast Saccharomyces cerevisiae, the same nucleosomal surface is critical for the formation of three types of silenced chromatin, mediated by the Sir, RENT, or Sum1-1 complexes, suggesting that this region of the nucleosome may be generally important to long-range silencing.
In S. cerevisiae, silenced chromatin domains associated with Sir proteins are found at the cryptic mating-type loci HMLα and HMRa and subtelomeric regions (Ruscheet al. 2003). A distinct form of silencing occurs in the rDNA repeats and is mediated by the RENT complex. Strains lacking Sir2p, Sir3p, or Sir4p lose silencing at the mating-type loci and consequently express both a and α mating-type information, resulting in the loss of cell-type identity and an inability to mate. Interestingly, silencing and mating can be restored by the SUM1-1 mutation, which was originally identified as a suppressor of a sir2Δ mutation (Klaret al. 1985). The SUM1-1 mutation results from a single-amino-acid change (Chi and Shore 1996), which enables the Sum1 repressor that normally does not spread to form an extended silenced domain (Rusche and Rine 2001; Suttonet al. 2001). The ability of mutant Sum1-1p to spread suggests that Sum1-1p or an associated protein interacts with nucleosomes and raises the question of whether the wild-type, nonspreading Sum1 complex also interacts with nucleosomes. To address these issues, we identified histone residues important for silencing mediated by the mutant Sum1-1 complex and the wild-type Sum1 complex.
Sir silencing at the cryptic mating-type loci is initiated at the flanking E and I silencers, which contain binding sites for the origin recognition complex (ORC) and the transcription factors Rap1p and Abf1p. These factors recruit the Sir proteins. Orc1p, the largest subunit of ORC, interacts directly with Sir1p, stabilizing the Sir proteins at the silencer (Triolo and Sternglanz 1996; Gardneret al. 1999). Sir2p is a NAD+-dependent deacetylase (Imaiet al. 2000; Landryet al. 2000; Smithet al. 2000), and deacetylation of nucleosomes in the vicinity of the silencers facilitates the binding of Sir3p and Sir4p, allowing the Sir complex to propagate along the chromosome through a sequential deacetylation mechanism (Hoppeet al. 2002; Ruscheet al. 2002).
Sir3p is proposed to have two distinct histone-binding domains. A C-terminal histone-binding domain interacts with histone tails (Hechtet al. 1995; Carmenet al. 2002), and the N-terminal bromo-adjacent homology (BAH) domain binds the base of the histone H4 tail and an adjacent surface on the nucleosome core termed the LRS region (Onishiet al. 2007; Buchbergeret al. 2008; Norriset al. 2008; Sampathet al. 2009). Many mutations in histones that disrupt Sir silencing are thought to act by decreasing the affinity of Sir3p for nucleosomes. These include mutations in the histone H4 tail (Johnsonet al. 1990, 1992; Park and Szostak 1990; Altafet al. 2007) and the LRS region (Parket al. 2002; Thompsonet al. 2003; Buchbergeret al. 2008; Norriset al. 2008; Sampathet al. 2009). Mutations that disrupt silencing have also been identified in H2A and H2B (Daiet al. 2010; Kyrisset al. 2010), although the mechanism by which they act remains unknown. In the cell, interactions between Sir proteins and nucleosomes are regulated by post-translational modifications. For example, deacetylation of histones by Sir2p is thought to increase their affinity for Sir3p (Megeeet al. 1990; Johnsonet al. 1992; Carmenet al. 2002; Altafet al. 2007). In addition, the affinity of Sir3p for nucleosomes is modulated by methylation of H3 K4 (Santos-Rosaet al. 2004) and H3 K79, located within the LRS region (Van Leeuwenet al. 2002). Thus, mutations that mimic or block these modifications can also affect silencing.
Unlike the Sir proteins, wild-type Sum1p does not form an extended silenced domain. Sum1p is a DNA-binding protein that recognizes a motif found in the promoters of some midsporulation, NAD+ biosynthesis, and α-specific genes (Xieet al. 1999; Bedalovet al. 2003; Pierceet al. 2003; Zill and Rine 2008). Sum1p works with Hst1p, an NAD+-dependent deacetylase related to Sir2p (Xieet al. 1999; McCordet al. 2003). The mutant form of Sum1p, Sum1-1p, contains a single-amino-acid change (T988I) that redirects Sum1-1p to the cryptic mating-type loci through interactions with ORC. We have proposed that the SUM1-1 mutation acts by changing the binding partners of Sum1p. In particular, loss of T988 reduces the affinity of Sum1p for its DNA consensus sequence, and the presence of isoleucine at position 988 increases the tendency of the protein to self-associate (Safiet al. 2008), which may enable it to spread along chromatin. Additionally, the T988I mutation increases the affinity of Sum1-1p for ORC, leading to the recruitment of Sum1-1p to HMRa and other ORC-binding sites (Rusche and Rine 2001; Suttonet al. 2001; Lynchet al. 2005). Thus, the SUM1-1 mutation changes the location and function of the complex, from a locus-specific repressor to a regional silencer. We predicted that Sum1-1p or another component of the Sum1-1 complex interacts with histones to enable the complex to spread. However, such an interaction has not been documented.
To investigate how the mutant Sum1-1 complex contacts nucleosomes and to determine whether such an interaction is also important for repression mediated by wild-type Sum1p, we conducted genetic screens for mutations in histones H3 and H4 that disrupt Sum1-1 silencing or Sum1 repression. We discovered that the same region of the nucleosome is important for silencing mediated by the mutant Sum1-1, Sir, or RENT complexes but this region is not critical for Sum1 repression. In addition, we present evidence that this region interacts with the Orc1p BAH domain to facilitate Sum1-1 silencing.
S. cerevisiae strains were derived from W303-1b and are described in supporting information, File S1 and Table S1. The SUM1-1, 3myc-SUM1, 7myc-SUM1-1, sir2Δ::HIS3, sir3Δ::LEU2, hst1Δ::KanMX (Rusche and Rine 2001), pPES4-HIS3 (Hickman and Rusche 2007), URA3::(lexAop)8-lacZ, LYS2:(lexAop)4-HIS3 (Hollenberget al. 1995), and VR-ADE2-TEL (Singer and Gottschling 1994) alleles were previously described.
Plasmids were constructed as described in File S1 and Table S2, Table S3, Table S4, and Table S5. pDM9 (HHT1 HHF1), pDM18 (HHT2 HHF2) (Duina and Winston 2004), pJR2292 (3myc-SUM1), pJR2291 (3myc-SUM1-1) (Rusche and Rine 2001), pLR052 (3HA-SUM1), pLR047 (3HA-SUM1-1) (Safiet al. 2008), pTT93 (LexA-ORC5), pGAD424 (GAD), pRH01 (GAD-SUM1), pRH02 (GAD-SUM1-1) (Suttonet al. 2001), pRM430 (HHT2Δ4-30), and pGF29 (HHF2Δ4-28) (Linget al. 1996) were previously described. Point mutations to histones H3 and H4 were obtained from Jef Boeke (Fryet al. 2006) and Michael Grunstein (Kayneet al. 1988; Johnsonet al. 1990) or were constructed as described.
Genetic screen for histone mutations
To generate mutations in histones H3 and H4, the gene of interest was amplified using error-prone PCR [0.2 mM dNTPs, 0.5 μM primers, 1.2 mM MnCl2, 5 ng plasmid pDM18, and 0.75 unit Taq (Roche) (5 units/μl) in 50 μl denatured for 3 min at 94°, followed by 32 cycles of 15 sec at 94°, 15 sec of 55°, and 2 min at 68°]. The HHT2 gene (H3) was amplified from pDM18 using primers (GGCTATGGCTCGGTGTCAAA) and (GCCCCGCAATTATGTCTGTAAA), and the HHF2 gene (H4) was amplified using primers (TACATACGTGTTTGTGCGTAT) and (CCAGGGTTTTCCCAGTCAC). The resultant collection of PCR products was integrated into plasmid pDM18 by homologous recombination in yeast cells (Muhlradet al. 1992). Specifically, pDM18 was digested with BglII and EcoRI to remove HHF2 or with AflII and RsrII to remove HHT2, and yeast (LRY1450 or LRY1849) were simultaneously transformed with the gapped plasmid (100 ng), PCR product (200 ng), and 30 μg salmon sperm DNA (Stratagene, La Jolla, CA). Colonies containing repaired plasmids, many of which contained mutations, were selected on medium lacking tryptophan and replica plated to medium lacking tryptophan with 0.1% 5-fluoroorotic acid (5-FOA) to select for cells that lost pDM9-bearing wild-type H3 and H4 with URA3.
To identify histone mutations that disrupted Sum1-1 silencing, parent strain LRY1450 was used. Colonies were replica plated to a lawn of the opposite mating type (LRY1021 MATa his4) on minimal plates to select for prototrophic diploids. Over 13,000 colonies were screened, and 163 colonies failed to mate. Plasmids were isolated from these yeast, amplified in Escherichia coli, and incorporated back into the reporter strain to confirm the nonmating phenotype. Plasmids that disrupted mating are listed in Table S3.
To identify histone mutations that disrupted Sum1 repression, parent strain LRY1849 was used. Colonies were replica plated to medium lacking histidine and leucine to identify those that failed to repress Sum1p-regulated reporters pPES4-HIS3 and pYGL138C-LEU2. Over 17,000 colonies were screened, and 80 colonies grew on medium lacking histidine and leucine. Plasmids were isolated and retested, and none of these plasmids enabled growth on medium lacking histidine and leucine. Twelve representative plasmids were sequenced and confirmed to have nucleotide and amino acid substitutions in the histone genes.
For the semiquantitative mating assay, one optical density (OD) equivalent of logarithmically growing cells was resuspended in 100 μl yeast minimal medium (YM). Tenfold serial dilutions were prepared, and 3 μl of each dilution was spotted onto rich medium (YPD) or medium containing 5-FOA and lacking tryptophan to monitor growth. For mating, an equal volume of a tester strain of the opposite mating type (LRY1021 MATa his4 or LRY1022 MATα his4) at 10 OD equivalents per ml in YPD was mixed with each sample in the dilution series. Three microliters of this mixture was spotted onto minimal medium to select for the growth of prototrophic diploids. Yeast were grown at 30° for 3 days before imaging.
To assess telomeric silencing, a reporter yeast strain (LRY2423) contained an ADE2 reporter gene integrated at telomere VR (Singer and Gottschling 1994) and lacked chromosomal histone genes. Tenfold serial dilutions were prepared, and 10 μl of the 10,000-fold dilution was spread onto YPD medium. Yeast were grown at 30° for 3 days and placed at 4° for 7 days to allow color development before imaging. Light pink, red, and sectored colonies were scored as (+) for maintaining telomeric silencing, and white colonies were scored as (−) for silencing defective.
To assess Sum1 repression of PES4, the reporter yeast strain was LRY1793 containing reporter pPES4-HIS3. Tenfold serial dilutions were prepared, and 3 μl of each dilution was spotted onto rich medium (YPD) to monitor growth or medium lacking histidine to monitor Sum1 repression.
Chromatin immunoprecipitation (ChIP) was performed as previously described (Rusche and Rine 2001). Two independent cultures of each strain were grown, and 50 OD equivalents of cells were cross-linked with 1% formaldehyde for 30 min. Lysis and subsequent steps were as described using 3 μl of antibody [anti-myc (Millipore 06-549), anti-HA (Millipore H6908), anti-H3 (Upstate 06-755), anti-H3 K4 Me2 (Upstate 07-030), and anti-H3 K4 Me3 (Millipore 04-745)]. Samples were quantified relative to a standard curve prepared from input DNA using real-time PCR with SYBR Green on a Bio-Rad (Hercules, CA) iCycler as described (Safiet al. 2008). Data represent the average of two independent IPs for each strain, each analyzed in two independent PCR reactions. Error bars represent the standard error. Oligonucleotide sequences are listed in Table S6.
For immunoprecipitation of Gal4DBD-myc-SUM1-1 and ORC1-HA (Figure 4), immunoprecipitation was performed as previously described (Hickman and Rusche 2007), using a second cross-linking agent (Kurdistaniet al. 2002). Fifty OD equivalents of cells were cross-linked for 1 hr in 10 ml cold DMA (PBS with 10 mM dimethyl adipimidate, 0.25% DMSO), followed by 1 hr in 1% formaldehyde. Lysis and subsequent steps were as described above. Each sample was analyzed relative to a standard curve prepared from its own input DNA to control for plasmid copy number differences.
Co-immunoprecipitations (co-IPs) were performed as previously described (Rusche and Rine 2001; Safiet al. 2008), using 30 ODs of cells grown in medium lacking leucine and adenine to maintain plasmids. Cleared lysates were incubated with 5 μl rabbit polyclonal IgG antibodies [anti-myc (Upstate 06-549) and anti-HA (Upstate 05-902)] at 4° overnight and then with 60 μl protein A agarose beads (Upstate) for 1 hr. Samples were electrophoretically fractionated on 7.5% polyacrylamide–SDS gels, transferred to membranes, and probed using mouse monoclonal antibodies (Upstate; 05-724 and 05-904).
RNA isolation and quantitative reverse transcriptase PCR
RNA was isolated as described previously (Schmittet al. 1990), and cDNA was synthesized and quantified as previously described (Hickman and Rusche 2007). The standard curve was generated with genomic DNA isolated from wild-type W303-1b. SUM1, HST1, RFM1, and ORC1 transcript levels were calculated relative to NTG1 and then normalized to the wild-type strain. Results represent the average of two independently grown cultures for each strain, each analyzed in two independent PCR reactions. Error bars represent the standard error. Oligonucleotide sequences are listed in Table S6.
Two-hybrid interactions were detected through activation of a LacZ reporter construct as previously described (Safiet al. 2008). To assay for β-galactosidase activity, yeast were grown for 18 hr directly on a Magna Graph Nylon membrane (GE, NJOHY08250) placed on a YPD plate. The filter was flash frozen in liquid nitrogen, placed onto filter paper soaked in developing solution [65 mM sodium phosphate, pH 7.0, 10 mM KCl, 1 mM MgSO4, 0.6% β-mercaptoethanol, and 1% x-gal stock (100 mg/ml in N,N-dimethyl formamide)], and developed for 2 hr at 37° in a closed container.
Genetic screen for histone mutations that disrupt Sum1-1 silencing
To identify mutations in histones H3 and H4 that disrupt Sum1-1 silencing, we developed a genetic screen employing a plasmid shuffling strategy (Figure 1A, left column). A reporter yeast strain contained the SUM1-1 mutation and a sir2 deletion so Sum1-1 silencing could be assessed independently of Sir silencing. In addition, the essential histone H3 and H4 genes were deleted from their chromosomal locations and provided on a plasmid bearing a URA3 marker. Mutations were generated in the histone H3 or H4 genes using error-prone PCR, and these mutated histone genes were incorporated into the reporter strain on plasmids bearing the TRP1 marker. Finally, the wild-type histone genes were shuffled out by selecting against the URA3 marker. The efficacy of Sum1-1 silencing in the presence of these histone mutations was assessed using a mating assay (Figure 1B), as loss of silencing at HMRa in these MATα cells leads to simultaneous expression of a and α mating-type information and sterility. Plasmids from colonies that failed to mate were isolated, amplified in E. coli, and reexamined in the reporter strain to confirm that the mating defect was linked to the plasmid. Over 13,000 colonies were screened: 7500 for H3 and 5700 for H4. Forty-eight plasmids were recovered that conferred a mating defect, each containing one to six mutations (Figure 1C).
To identify those mutations that disrupted Sum1-1 silencing, plasmids containing single-amino-acid mutations were generated and tested in a semiquantitative assay for their ability to disrupt mating. These single mutations were generated by site-directed mutagenesis or were obtained from existing sources (Kayneet al. 1988; Johnsonet al. 1990; Fryet al. 2006; Daiet al. 2008). For each mutation, the fraction of cells that mated in the presence of the mutation was estimated relative to a strain expressing wild-type histones (Figure 1B). Once a disruptive residue was identified on a particular plasmid, the remaining mutations were not investigated. Thirty-one of the 71 histone H3 mutations and 11 of the 18 histone H4 mutations were tested (Table S3). To confirm that the loss of mating resulted from the loss of silencing, as opposed to a defect in the mating pathway, the level of HMRa1 mRNA was assessed in the presence of H3-K4I, H3-E73G, and H4-G28P mutations. In each mutant strain, HMRa1 mRNA was induced compared to a strain with wild-type histones (data not shown), confirming the loss of silencing.
This screen was sensitive enough to identify histone mutations that caused a 10-fold loss of mating, but we focused on mutations conferring ≥100-fold defects in mating. Ten mutations in histone H3 and 4 mutations in histone H4 fell into this category [Figure 2, A (red bars) and B, and Table S4]. These 14 amino acid residues mapped to three distinct regions of the nucleosome; residues 2–4 of the H3 tail, the LRS region in the H3 core, and residues 18–26 of the H4 tail. Interestingly, mutations in two of these regions, the LRS domain and the H4 tail, disrupt Sir silencing (Johnsonet al. 1990; Parket al. 2002). Since the Sir and Sum1-1 complexes both generate extended silenced domains, these regions of the nucleosome may contribute similarly to the formation of both types of silenced chromatin. Therefore, 14 additional mutations in the LRS domain and the H4 tail that were previously demonstrated to disrupt Sir silencing were tested for their effect on Sum1-1 silencing. Seven additional residues that disrupted Sum1-1 silencing were identified in this manner (Figure 2B, light gray boxes, and Table S4). The remaining 7 residues did not have an effect on Sum1-1 silencing (Table S5).
Post-translational modifications of histones modulate the assembly of chromatin structures. Therefore, we specifically examined the importance of 35 modifiable residues by replacing each one with alanine. Only H3-K4A was found to disrupt Sum1-1 silencing (Figure 2B, dark gray box), consistent with our recovery in the original screen of 10 independent plasmids bearing a mutation in H3-K4. After examining 96 individual mutations in H3 and H4, we identified 22 that caused a defect in Sum1-1 silencing, 14 in H3 and 8 in H4 (Figure 2A and Table S4). These mutations cluster in three discrete regions. Interestingly, the H3 LRS region and the H4 tail are adjacent in the nucleosome structure (Figure 2C) and could form a surface that interacts with the Sum1-1 complex or another protein important for silencing.
To determine whether histone mutations that disrupt Sum1-1 silencing are dominant, mating was assessed under conditions that maintain plasmids expressing wild-type and mutant histones. H3-K4A and -K4I resulted in a 10-fold mating defect. These mutations also disrupted mating when expressed in a yeast strain having wild-type chromosomal histone genes. No other mutations were dominant (data not shown). Thus, H3-K4 is particularly critical for Sum1-1 silencing.
Histone mutations do not disrupt Sum1 repression
We reasoned that residues in histones H3 and H4 that are important for Sum1-1 silencing might be important contact points for the Sum1 complex. Therefore, these residues might also play a role in repression mediated by wild-type Sum1p, which differs from Sum1-1p by a single amino acid. A role for nucleosomes in Sum1 repression is also suggested by the requirement for the deacetylase Hst1p (Xie et al. 1999; McCord et al. 2003), which could act on histones. To investigate whether histones contribute to Sum1 repression, a second genetic screen was conducted (Figure 1A, right). The starting strain contained two reporters in which the Sum1p-repressed PES4 and YGL138C promoters drive the expression of HIS3 and LEU2, allowing Sum1 repression to be assessed by growth on medium lacking histidine and leucine. Yeast with wild-type SUM1 and HST1 genes were unable to grow in the absence of histidine and leucine, whereas control strains with sum1Δ or hst1Δ mutations grew well (Figure 1B). The starting strain also had its only copy of the histone H3 and H4 genes on a plasmid bearing the URA3 gene, and a plasmid shuffle strategy was used to screen for mutated histone genes that disrupt Sum1 repression. Over 17,000 colonies were screened, 11,200 for H3 and 6100 for H4, but no plasmids were identified that enabled cells to grow on medium lacking histidine and leucine (Figure 1C). We also specifically examined each histone mutation that was found to disrupt Sum1-1 silencing, and none of these mutations disrupted Sum1 repression. Therefore, Sum1 repression could not be disrupted by single mutations in the H3 or H4 genes.
It is possible that multiple lysine residues must be deacetylated by Hst1p to achieve repression. Therefore, targeted mutations and deletions were tested for their ability to disrupt Sum1 repression (Table 1). In particular, lysines known to be acetylated were replaced by glutamine to mimic the acetylated state, arginine to mimic the deacetylated state, or alanine. The tails of histones H3 and H4 were also deleted to remove the lysine residues entirely. In addition, mutations in H3 that disrupt Sum1-1 silencing were combined with H4-K16Q, as K16 is a critical substrate of Sir2p (Johnsonet al. 1990; Imaiet al. 2000), the paralog of Hst1p. However, none of these alterations to histones H3 and H4 disrupted Sum1 repression of PES4 or YGL138C. Therefore, although particular regions of the nucleosome are critical for silencing mediated by the mutant Sum1-1 complex, no such requirements exist for Sum1 repression.
Histone mutations in the H3 LRS region and H4 tail disrupt Sir silencing
Our ability to identify histone mutations that disrupt silencing mediated by Sum1-1p, which acts over several kilobase pairs, but not Sum1p, which acts over a more restricted region, suggests that nucleosomes are more critical in generating long-range chromatin structures involved in Sum1-1 silencing. Moreover, it is striking that mutations disruptive to Sum1-1 silencing fall in the same regions of the nucleosome important for silencing at the cryptic mating-type loci, telomeres, and rDNA (Johnsonet al. 1992; Parket al. 2002; Thompsonet al. 2003). Therefore, related higher-order structures may be generated by the Sir and Sum1-1 complexes.
To determine whether histone mutations that affect Sum1-1 silencing also affect Sir silencing, a semiquantitative mating assay was used to assess silencing at the HMLα locus and a colorimetric assay was used to score silencing of ADE2 integrated at telomere VR (Singer and Gottschling 1994). Examining both loci enabled us to distinguish mild from more severe mutations, as Sir silencing is more easily disrupted at the telomeres than at HMLα. All residues in the H3 LRS/H4 tail region that disrupt Sum1-1 silencing also disrupted Sir silencing at the telomere (Table 2). Sir silencing of HMLα was less disrupted by histone mutations, and only H3-E73 and residues 17–21 in the H4 tail cause a silencing defect at HMLα. These results are consistent with H3-E73D having the strongest effect on HM silencing of any known H3 mutation (Thompsonet al. 2003). Interestingly, mutations in the H3 tail were not disruptive to Sir silencing, with the exception of H3-K4I and K4R, which disrupted telomeric silencing. Therefore, the H3 tail may serve a specific role in Sum1-1 silencing, whereas the LRS/H4 tail region may have a common function in Sum1-1 and Sir silencing.
Mutations in the H4 tail did not act by interfering with the ISW complex
One possible explanation for silencing defects in the presence of H4 tail mutations is a defect in chromatin remodeling. H4 tail residues R17, H18, and R19 form a binding site for the ATP-dependent chromatin remodeler ISW (Clapieret al. 2002), and mutation of these residues could disrupt Sir and Sum1-1 silencing by inhibiting chromatin remodeling. Indeed, deletion of ISW1 disrupts Sir silencing at HMR (Cuperus and Shore 2002) and HML (Yuet al. 2011). Similarly, we found that isw1Δ decreases Sum1-1 silencing at HMR, resulting in a 100-fold decrease in mating. However, most of the histone H4 mutations we identified disrupted mating at least 1000-fold, inconsistent with these mutations acting solely by disrupting remodeling by the ISW complex.
The BAH domain of Orc1p contributes to Sum1-1 silencing by binding nucleosomes
The LRS/H4 tail domain of the nucleosome, in which many of the mutations fall, interacts with the BAH domain of Sir3p, and this interaction is required for the spreading of the Sir complex (Connellyet al. 2006; Onishiet al. 2007; Buchbergeret al. 2008; Norriset al. 2008; Sampathet al. 2009). Therefore, it is possible that the LRS/H4 tail domain plays a similar role in Sum1-1 silencing and that a BAH domain-containing protein is also critical for Sum1-1 silencing. Sir3p itself is not required for Sum1-1p silencing (Laurenson and Rine 1991), but Orc1p, which is a paralog of Sir3p, might play a similar role. Consistent with this idea, the BAH domain of Orc1p is required for Sum1-1 silencing (Rusche and Rine 2001), and ORC has been implicated in the recruitment of Sum1-1p to chromatin (Rusche and Rine 2001; Suttonet al. 2001; Irlbacheret al. 2005; Lynchet al. 2005). We investigated whether an interaction between the BAH domain of Orc1p and the LRS/H4 tail domain of nucleosomes is important for Sum1-1 silencing. Such an interaction is not anticipated by the previous model that ORC recruits Sum1-1p to silenced loci. Therefore, to determine whether the BAH domain of Orc1p contributes to Sum1-1 silencing by recruiting the Sum1-1 complex or by acting at subsequent steps in silencing, we conducted several experiments.
First, we tested whether a previously described two-hybrid interaction between Sum1-1p and Orc5p, which may not be direct, requires the BAH domain of Orc1p. If so, the primary role of the BAH domain is most likely in recruitment. As in previous assays (Suttonet al. 2001; Safiet al. 2008), the C-terminal portion of Sum1p (786–1062) was fused to the activation domain of Gal4p, and Orc5p was fused to the LexA DNA-binding domain. An interaction between Sum1p and Orc5p resulted in the activation of a LacZ gene driven by a minimal promoter containing LexA-binding sites. As previously described, Orc5p interacted with the mutant Sum1-1p but not the wild-type Sum1p (Figure 3A). Importantly, this interaction still occurred in an orc1Δbah strain (Figure 3A, right), indicating that the BAH domain of Orc1p was not required. Therefore, although ORC recruits the Sum1-1 complex, the BAH domain of Orc1p likely makes a different contribution to Sum1-1 silencing.
To investigate whether the function of the Orc1p BAH domain in Sum1-1 silencing requires its nucleosome-binding capacity, we created mutations homologous to mutations in Sir3p that disrupt its ability to bind nucleosomes (Buchbergeret al. 2008). Such mutations should disrupt Sum1-1 silencing if the BAH domain acts by binding nucleosomes. Indeed, orc1-E84K and orc1-P179L mutations disrupted Sum1-1 silencing, as assessed by mating (Figure 3B). Therefore, the BAH domain likely acts by binding nucleosomes and not the Sum1-1 complex.
The BAH domain of Orc1p contributes to Sum1-1 silencing at a step other than recruitment
To investigate whether the ability of the Orc1p BAH domain to bind nucleosomes is important for a step in Sum1-1 silencing subsequent to recruitment, we created a situation in which Sum1-1p could be recruited to the HMRa locus independently of ORC. A plasmid containing the entire HMRa locus was modified to replace the ORC-binding sites in the E and I silencers with Gal4-binding sites (Figure 4A). Yeast cells that lacked SIR2 and HMRa and expressed both a Gal4 DNA-binding domain (Gal4DBD)-myc-Sum1-1 fusion protein and untagged Sum1-1p were transformed with this HMRa-Gal4 plasmid. These strains also lacked Sir3p, to prevent it from substituting for its paralog Orc1p. If the BAH domain of Orc1p acts at a step subsequent to recruitment, it should still be required to silence this modified HMR locus. Silencing, as assessed by mating, was dependent on the presence of the Gal4DBD-Sum1-1 protein (Figure 4B), indicating that the recruitment of Sum1-1p was occurring through the Gal4DBD rather than ORC. Importantly, mutation or deletion of the BAH domain disrupted mating ∼100-fold (Figure 4B), consistent with the BAH domain providing a function other than recruitment of Sum1-1p.
The requirement for the Orc1p BAH domain in silencing the HMRa-Gal4 locus suggests that Orc1p acts directly at the locus despite the absence of ORC-binding sites in the silencers. To examine this possibility, we performed a chromatin IP assay and found Orc1p enrichment at the E silencer even in the absence of Gal4DBD-Sum1-1p (Figure 4C). Thus, a cryptic ORC-binding site must still be present near the E silencer. Nevertheless, in the presence of Gal4DBD-Sum1-1p, the enrichment of Orc1p was significantly increased at the E silencer and was also observed at other sites within HMRa-Gal4. Therefore, Orc1p is stabilized at the HMRa-Gal4 locus by Sum1-1p. As anticipated, Gal4DBD-Sum1-1p was enriched across the modified HMRa-Gal4 locus (Figure 4D).
Our initial hypothesis was that the interaction of the Orc1p BAH domain with nucleosomes promotes the distribution of Orc1p and Sum1-1p across the silenced locus, as is the case for Sir3p. However, the distribution of Orc1p was more restricted than that of Gal4DBD-Sum1-1p (Figure 4, C and D), suggesting that Orc1p does not spread with Sum1-1p and makes a different contribution to silencing. To determine how the BAH domain of Orc1p promotes silencing at HMRa-Gal4, we performed chromatin IP in strains bearing ORC1 mutations that disrupt nucleosome binding. If the ability of the BAH domain to bind nucleosomes is important for stabilizing the Sum1-1 complex across HMRa-Gal4, these mutations should decrease the associations of Sum1-1p and Orc1p. However, although a significant decrease in enrichment of Orc1p was observed (Figure 4E), these mutations had only a slight effect on the enrichment of Sum1-1p (Figure 4F). Therefore, the BAH domain of Orc1p stabilizes its own association with HMRa-Gal4 and correlates with the maximally silenced state, suggesting that the presence of Orc1p is critical for complete silencing. It is likely that mutations in the LRS/H4 tail region of the nucleosome have similar consequences for Orc1p association with HMRa, but we were unable to directly test this idea due to technical complications of combining the HMRa-Gal4 plasmid with histone mutations.
The BAH domain of Orc1p was not required for the self-association of Sum1-1p
In addition to binding nucleosomes and thereby stabilizing the association of Orc1p with chromatin, the BAH domain of Orc1p could contribute to Sum1-1 silencing in other ways. In particular, the SUM1-1 mutation increases the ability of Sum1-1p to self-associate (Safiet al. 2008), and the BAH domain could stabilize this interaction. We previously demonstrated that two differently tagged alleles of Sum1-1p coprecipitate, whereas similarly tagged wild-type proteins do not coprecipitate (Safiet al. 2008). This coprecipitation was not disrupted by DNaseI and therefore may be independent of association with silenced loci. To determine whether the BAH domain of Orc1p was required for the self-association of Sum1-1p, the coprecipitation experiment was repeated in an orc1Δbah strain. The coprecipitation still occurred (Figure 5, A and B), indicating that the BAH domain was not required for self-association.
Mutations in the N terminus of H3 disrupt Sum1-1 silencing at a step subsequent to recruitment of Sum1-1p
In addition to mutations in the LRS/H4 tail domain, we obtained seven mutations at the N terminus of H3 that disrupted Sum1-1 silencing. These mutations may act through a different mechanism than mutations in the LRS/H4 tail domain as they have distinct properties. In particular, H3-K4A and -K4I were the only mutations with a dominant phenotype, and most H3 tail mutations were not disruptive to Sir silencing. The H3 tail residues important for Sum1-1 silencing cluster around H3-K4, suggesting that the methylation status of H3-K4 could affect silencing. In particular, mutation of H3-K4 would block methylation, and mutations of neighboring residues could interfere with the recognition of H3-K4 as a substrate. Consistent with this notion, Sum1-1 silencing was severely disrupted in a strain lacking Set1p, the methyltransferase specific for H3-K4 (Figure 6A). In contrast, deletion of Dot1p, which methylates H3-K79, located in the LRS domain, resulted in a very modest disruption of Sum1-1 silencing.
Methylation of H3-K4 in silenced domains could be critical for the assembly of Sum1-1p chromatin. Alternatively, methylation of euchromatin may help restrict Sum1-1 silenced chromatin to appropriate domains, as has been proposed for Sir silenced chromatin (Santos-Rosaet al. 2004). To determine whether H3-K4 is methylated or unmethylated at HMRa in the presence of Sum1-1p, chromatin IP was performed using antibodies specific for di- or trimethylated H3-K4 (Figure 6B). In a background where Sir proteins silence HMRa, there was no detectable enrichment of di- or trimethylated H3-K4, as previously reported (Bernsteinet al. 2002; Santos-Rosaet al. 2002). In a sir3Δ strain, in which HMRa1 is expressed, di- and trimethlyation of H3-K4 were enriched, as expected. In a sir3Δ SUM1-1 strain, in which Sum1-1p restores silencing of HMRa, methylation of H3-K4 was not observed, indicating that methylation of H3-K4 within the silenced domain is not necessary for Sum1-1 silencing.
A trivial explanation for the disruption of Sum1-1 silencing by mutations in the H3 tail is that improper methylation of H3-K4 alters the expression of a component of the Sum1-1 complex. To examine this possibility, the steady-state mRNA levels of SUM1-1, HST1, RFM1, and ORC1 were compared by quantitative reverse transcriptase (RT)–PCR in strains expressing wild-type or mutant histones and in a strain lacking Set1p. The levels of these mRNAs did not change significantly (Figure 6C). In addition, immunoblot analysis of Sum1-1p in these strains revealed no significant differences (data not shown). Therefore, a change in the expression of the Sum1-1 complex does not account for the silencing defect in strains expressing H3 tail mutations.
It is possible that mutations in the H3 tail reduce an interaction between this portion of H3 and the Sum1-1 complex, thereby disrupting silencing. To examine this possibility, the association of Sum1-1p with HMRa was assessed in the presence of H3-K4I, the most disruptive mutation. However, in the presence of H3-K4I, the enrichment of Sum1-1p was only slightly reduced compared to a strain with wild-type histones (Figure 6D), suggesting that the Sum1-1 complex assembles properly at HMRa and that a subsequent step in the silencing process must be disrupted. Similarly, mutations in the LRS region (H3-E73G) or the H4 tail (H4-I21V) caused a modest reduction in the association of Sum1-1p with HMRa (Figure 6D), consistent with the modest decrease in Sum1-1p enrichment in the presence of Orc1p BAH mutations (Figure 4E).
In this study, we identified three regions of the nucleosome that are important for long-range silencing mediated by the mutant Sum1-1 complex but not for wild-type Sum1 repression. The H3 LRS and H4 tail regions are in close proximity in the nucleosome structure and may form a potential interaction surface. This LRS/H4 tail region is also important for Sir silencing and is thought to interact with the Sir3p BAH domain. By analogy, we suggest that the BAH domain of Orc1p facilitates Sum1-1 silencing. The extreme N-terminal tail of H3 is more important for Sum1-1 than Sir silencing and disrupts silencing without significantly reducing the association of Sum1-1 with the silenced domain.
The LRS/H4 tail region of the nucleosome is important for multiple types of silencing
It is striking that the same region of the nucleosome, the LRS region and the adjacent base of the H4 tail, plays an important role in silencing mediated by three distinct protein complexes, Sir, RENT, and Sum1-1 (Johnsonet al. 1990, 1992; Park and Szostak 1990; Parket al. 2002; Thompsonet al. 2003; Altafet al. 2007; Buchbergeret al. 2008; Norriset al. 2008; Sampathet al. 2009), suggesting a common mechanism of action. This region could contribute to silencing by promoting proper chromatin compaction, providing a binding site for a common factor, or serving as a multipurpose recruitment site for silencing proteins.
One possibility is that the LRS/H4 tail region is important for multiple types of silencing because it facilitates internucleosome interactions necessary for chromatin compaction. For example, histone H4 basic residues 14–19 and acidic residues in histones H2A and H2B are proposed to interact (Dorigoet al. 2004; Kanet al. 2009). In addition, acetylation of H4 at K16 inhibits 30-nm fiber formation (Shogren-Knaaket al. 2006). However, mutations of H2A and H2B acidic residues did not disrupt Sir silencing (data not shown) and the LRS region is not reported to interact with other regions of the nucleosome. Therefore, assisting in chromatin compaction may not be the primary contribution of the LRS/H4 tail region.
A second possibility is that the LRS/H4 tail region associates with a common factor required for multiple types of silencing. One candidate is the ISW chromatin remodeling complex, which is associated with transcriptional repression and requires H4 tail residues 17–19 for remodeling (Clapieret al. 2002). However, Sum1-1 silencing was more severely affected by mutations in H4 tail residues than by an isw1Δ deletion (data not shown), indicating that the histone tail mutations do not act solely by blocking the activity of ISW.
A third possibility is that the LRS/H4 tail region is relatively accessible and consequently is bound by different silencing proteins, including the BAH domains of Sir3p and its paralog Orc1p. Whether a nucleosome-binding protein is also important for silencing in the rDNA is unclear, as Sir3p and Orc1p are not thought to contribute to rDNA silencing.
Orc1p contributes to Sum1-1p silencing in a capacity other than recruitment
Given the ability of mutations in the LRS/H4 tail region to disrupt Sum1-1 silencing, we suspected that the Orc1p BAH domain, previously shown to be required for Sum1-1 silencing, makes important contacts with nucleosomes. This model is interesting in light of the association of ORC with heterochromatin in a variety of species. For example, in Drosophila and humans, ORC interacts with HP1 (Paket al. 1997; Lidonniciet al. 2004; Prasanthet al. 2004; Authet al. 2006) and is enriched in telomeric and pericentromeric heterochromatin (Denget al. 2007, 2009; Prasanthet al. 2010; Shenet al. 2010). On the basis of the paradigm from S. cerevisiae, ORC is thought to serve as a platform for recruiting heterochromatin proteins. However, ORC could also bind nucleosomes via the BAH domain of Orc1p and thus act like Sir3p, which enables spreading of the Sir complex (Hechtet al. 1995; Carmenet al. 2002; Onishiet al. 2007; Buchbergeret al. 2008). In fact, in the yeast Kluyveromyces lactis, which lacks a distinct Sir3p, Orc1p acts in a Sir3p-like manner (Hickman and Rusche 2010).
Consistent with the “platform” model, ORC is proposed to recruit Sum1-1p to HMRa (Rusche and Rine 2001; Suttonet al. 2001; Lynchet al. 2005). However, the role of the BAH domain in Sum1-1 silencing remained a mystery. Our identification of histone mutations that fall in the LRS/H4 tail region of the nucleosome and disrupt Sum1-1 silencing suggests that an interaction between the Orc1p BAH domain and nucleosomes contributes to silencing. Indeed, mutations in ORC1 predicted to decrease the affinity of the BAH domain for nucleosomes also disrupted silencing, and the BAH domain of Orc1p was still required for silencing when Sum1-1p was recruited to HMRa independently of ORC (Figure 4).
In theory, Orc1p could act in a Sir3p-like manner to promote the spreading of Sum1-1p across HMRa. However, the enrichment of Sum1-1p was only moderately decreased in strains containing mutations in the Orc1p BAH domain (Figure 4) or mutations in the LRS region (Figure 6). Moreover, Orc1p was not uniformly distributed across HMRa, as would be expected if it acted like Sir3p. Thus, the presence of Orc1p at the E silencer may promote transcriptional silencing by altering the conformation of the chromatin fiber or recruiting factors that make the region less permissive for transcription.
It is interesting that the ancestral Sum1p likely acted with Orc1p to achieve silencing, and the SUM1-1 mutation may recapitulate this cooperation. In the yeast K. lactis, the Sum1 complex silences HMLα in conjunction with KlSir4p and KlOrc1p (Hickman and Rusche 2009, 2010). Therefore, the common ancestor of KlSum1p and ScSum1p most likely had a similar function, which was lost in the S. cerevisiae lineage.
The H3 tail is important for silencing
We also identified mutations in the H3 tail that disrupted Sum1-1 silencing but had less impact on Sir silencing. These mutations could act by disrupting specific interactions between the Sum1-1 complex and the H3 tail or reducing the available Sum1-1 by allowing nonspecific binding throughout the genome. However, neither interpretation is consistent with the enrichment of Sum1-1 at HMRa in H3-K4I strains (Figure 6D). Interestingly, similar observations were made for Sir proteins, which remain associated with silenced loci in strains bearing a deletion of the H3 tail that disrupts transcriptional silencing (Sperling and Grunstein 2009). These authors concluded that the H3 tail contributes to chromatin compaction and hence silencing at a step after Sir protein assembly, and the H3 tail may contribute similarly to Sum1-1 silencing.
Histones in heterochromatin formation in other species
Little is known about regions of the nucleosome important for heterochromatin formation in species other than S. cerevisiae. In species outside the budding yeast clade, H3-K9Me and heterochromatin protein 1 (HP1) are hallmarks of heterochromatin, and different portions of the nucleosome might be critical for the formation of this type of heterochromatin. In fact, in Schizosaccharomyces pombe, mutation of histone H3-K9, -S10, or -K14 disrupted pericentromeric heterochromatin formation, but mutation of H4-K8 or -K16 did not (Melloneet al. 2003), consistent with the importance of H3-K9Me.
It is possible that the LRS/H4 tail region of the nucleosome is also important for other types of heterochromatin. For example, the BAH domain-containing protein BAHD1 contributes to heterochromatin in humans (Bierneet al. 2009) and could interact with the LRS/H4 tail region in a manner similar to Sir3p or Orc1p.
Histones do not disrupt promoter-specific repression mediated by Sum1p
A goal of this study was to identify mutations in histones H3 and H4 that disrupt a potential interaction between the Sum1 complex and nucleosomes, which could contribute to wild-type repression. However, no such mutations were isolated in the genetic screen, and candidate approaches did not identify more complex mutations that disrupt repression. Nevertheless, the deacetylase activity of Hst1p is required for Sum1 repression (Hickman and Rusche 2007), and it has been assumed that histones are the relevant substrate of Hst1p. In fact, acetylation of histones H3 and H4 at Sum1-repressed genes increases in the absence of HST1 (Robertet al. 2004; Hickman and Rusche 2007; Weberet al. 2008). However, given that mutations that mimicked the acetylated state did not disrupt repression, the key substrate of Hst1p may not be histone H3 or H4. Instead, Hst1p may deacetylate other proteins such as histones H2A and H2B or the Sum1 complex.
The lack of histone mutations that disrupt Sum1 repression contrasts with other repressors in S. cerevisiae that can be disrupted by histone mutations (Lenfantet al. 1996; Rechtet al. 1996; Sabetet al. 2003; Parraet al. 2006). For example, repression mediated by the Tup1-Ssn6 complex is disrupted by histone mutations that mimic acetylation (Watsonet al. 2000) or diminish the association of the corepressor Tup1p with histone tails (Edmondsonet al. 1996). Unlike Tup1p, the Sum1 complex may not require a stabilizing interaction with histones to achieve repression.
Regional silencing involves the formation of extended domains of heterochromatin, whereas gene-specific repression occurs at individual promoters. This study reveals that in the yeast S. cerevisiae, the same nucleosomal surface is critical for the formation of multiple types of heterochromatin, but not for local repression mediated by a related transcriptional repressor. Thus, this region of the nucleosome may be generally important to long-range silencing. In addition, this region of the nucleosome may act as a binding surface for silencing proteins, including the paralogs Sir3p and Orc1p.
We thank Patrick Lynch, Yuting Fan, Radharani De, and Yang (Sunny) Wu for technical assistance; Fred Winston, Michael Grunstein, Jef Boeke, Junbiao Dai, Daniel Gottschling, Toishio Tsukiyama, Jasper Rine, and Rolf Sternglanz for yeast strains and plasmids; and Junbiao Dai for comments on the manuscript. This work was supported by a grant from the National Institutes of Health (GM073991) (to L.N.R.).
Supporting information is available online at http://www.genetics.org/content/suppl/2011/05/05/genetics.111.129197.DC1.
- Received March 30, 2011.
- Accepted May 2, 2011.
- Copyright © 2011 by the Genetics Society of America