|Home | About | Journals | Submit | Contact Us | Français|
Previous work has shown that the N terminus of the Saccharomyces cerevisiae Sir3 protein is crucial for the function of Sir3 in transcriptional silencing. Here, we show that overexpression of N-terminal fragments of Sir3 in strains lacking the full-length protein can lead to some silencing of HML and HMR. Sir3 contains a BAH (bromo-adjacent homology) domain at its N terminus. Overexpression of this domain alone can lead to silencing as long as Sir1 is overexpressed and Sir2 and Sir4 are present. Overexpression of the closely related Orc1 BAH domain can also silence in the absence of any Sir3 protein. A previously characterized hypermorphic sir3 mutation, D205N, greatly improves silencing by the Sir3 BAH domain and allows it to bind to DNA and oligonucleosomes in vitro. A previously uncharacterized region in the Sir1 N terminus is required for silencing by both the Sir3 and Orc1 BAH domains. The structure of the Sir3 BAH domain has been determined. In the crystal, the molecule multimerizes in the form of a left-handed superhelix. This superhelix may be relevant to the function of the BAH domain of Sir3 in silencing.
Epigenetic silencing is a term used to describe the heritable transmission of a transcriptionally inactive state. The silent mating type loci HML and HMR and telomeres of the budding yeast Saccharomyces cerevisiae are examples of loci that undergo this type of transcriptional silencing and have served as a paradigm for studying this process.
HML and HMR harbor copies of the mating type information genes, α and a, respectively. They are involved in mating type interconversion with the actively transcribed MAT locus. Transcriptional silencing at these loci relies on the existence of cis-acting DNA regulatory elements, termed silencers (E and I), which flank both loci. These elements recruit the DNA binding proteins Rap1, Abf1, and ORC, which then serve to recruit the silent information regulators (Sir) 1, 2, 3, and 4 (11, 13, 33). The widely accepted view of silencing at these loci (and at telomeres) is that histone tails are deacetylated through the action of Sir2, a NAD-dependent histone deacetylase, creating a binding surface on nucleosomes for the binding of Sir3 and Sir4. Multiple rounds of deacetylation lead to the formation of a Sir2/3/4 polymer that spreads on the nucleosomes of the silent region, altering the chromatin and making it unavailable for transcription. The detailed structure of silent chromatin is not known, and exactly how transcription is prevented is a matter of dispute (6, 34).
Sir3 is essential for the establishment and maintenance of the silent state at the HM loci and telomeres. Genetic, two-hybrid, and biochemical studies have identified interactions of Sir3 with histones H3 and H4, Sir4, Rap1, Abf1, and Sir3 itself (reviewed in references 11, 13, and 33). Interestingly, all these interactions are within the C-terminal two-thirds of the Sir3 protein. Nevertheless, expression of a Sir3 construct lacking the N-terminal region (hereafter referred to as the N terminus) is not sufficient to promote silencing at HML or telomeres in place of the full-length protein (17). Moreover, coexpression of the same C-terminal construct with the N terminus of Sir3 does lead to silencing (17, 23). Furthermore, point mutations in the N-terminal region can weaken or abolish silencing (35, 39). These data suggest that the N terminus of Sir3 plays an important role in transcriptional silencing and that it has the ability to function when separated from the rest of the protein (1).
The N-terminal region of Sir3 contains a BAH (bromo-adjacent homology) domain within it. The BAH domain has been identified in several chromatin-associated proteins, including Rsc1 and Rsc2, components of the RSC chromatin remodeling complex in S. cerevisiae; Dnmt1, the major DNA methyltransferase in mouse and human cells; Orc1, the largest subunit of the origin recognition complex; and others (4). The majority of what is known about the BAH domain comes from work with Orc1 in S. cerevisiae. Interestingly, this region of Orc1 is not required for its role in replication but is essential for its role in silencing. Our laboratory demonstrated previously that the Orc1 BAH domain interacts with Sir1 and brings it to the silent locus (38). Subsequently, the regions essential for this interaction were narrowed down to the Orc1 H domain within its BAH domain and a region in the C-terminal portion of Sir1 termed the Orc1-interacting region (OIR) (2, 12, 40). This has led to the idea that the BAH domain is a protein-protein interaction module that serves to recruit proteins involved in transcriptional regulation.
The BAH domain of Orc1 (Orc1BAH) is highly similar in sequence to the BAH domain of Sir3 (50% identity within the first 214 amino acids [aa]), and this region of Orc1 can functionally replace the N terminus of Sir3 in silencing (1). Even though the Orc1 and Sir3 BAH domains are so similar and Orc1BAH binds to Sir1, all attempts to detect an interaction between Sir1 and the Sir3 BAH domain (Sir3BAH) have been unsuccessful. Zhang et al. previously showed that replacing the H domain in Sir3 with the H domain of Orc1 was sufficient to promote an interaction between the C terminus of Sir1 and full-length Sir3 (40). These data suggest that the H domain, which resides within the BAH domain (see Fig. Fig.7),7), determines the specific protein-protein interaction of the BAH domain. Very recently, the cocrystal structure of Orc1BAH with the Sir1 OIR has been solved; it provided detailed knowledge of the Orc1BAH-Sir1 interaction (18, 19). Sir3 is only 35% identical toOrc1 within its H domain, and a high degree of variability within BAH domain-containing proteins occurs within this region. This may explain why the Sir1 OIR does not bind to Sir3 and suggests that there may be a difference in the way that the Orc1 and Sir3 BAH domains exert their silencing effects.
Here, we describe a study of Sir3BAH and Orc1BAH undertaken to characterize their roles in silencing. We show that these domains have the ability to silence the HM loci, at least partially, in the complete absence of the rest of the Sir3 protein. This silencing requires overexpression of SIR1. A gain-of-function point mutation within Sir3BAH allows this module to silence even without extra SIR1. We identify two novel properties of Sir3BAH: first, an ability to bind to oligonucleosomes and DNA and, second, a genetic interaction with the Sir1 N terminus. Finally, we present the crystal structure of Sir3BAH and show that it multimerizes to form a novel helical structure in the crystal, which may reflect its mode of binding to chromatin.
Strains used in this study are listed in Table Table1.1. Deletions of SIR1, SIR2, SIR3, SIR4, and ARD1 were made by gene replacement using the S. pombe his5+ gene or the Escherichia coli kanMX6 gene (27).
SIR3 or ORC1 fragments were cloned into the vector pFBL23 (TRP1, 2μm), a gift from Jacques Camonis, and expressed from the ADH1 promoter. These fragments, generated by PCR with BamHI and SalI ends, were cloned such that a Sir3-LexA or Orc1-LexA hybrid protein was expressed, with LexA at the C terminus of each protein. The following plasmids express C-terminal LexA hybrid proteins, all in pFBL23: pJC35 expresses Sir3 aa 1 to 380, pJC61 expresses Sir3 aa 1 to 380 with the D205N mutation, pJC14 expresses Sir3 aa 1 to 214, and pJC116 expresses Orc1 aa 1 to 214. pJC60 expresses SIR31-214 D205N-LexA in pFBL23; the PCR fragment was generated from pBTM-SIR3N205 (32). pJC59 contains the coding region for aa 1 to 380 of the SIR3 open reading frame followed by a stop codon, cloned as a BamHI-SalI fragment into pFBL23; therefore, this Sir3 fragment is not expressed as a LexA hybrid. pPY15, expressing full-length SIR3-LexA, was constructed by first destroying the PstI site in pFBL23 using Klenow polymerase to create pPY14 and then inserting a PCR-generated BamHI-SalI fragment into pPY14. pJC52 contains the entire SIR3 open reading frame, expressed from the ADH1 promoter, inserted into the vector pADH424 (TRP1 2μm) as a BamHI-SalI PCR fragment (29).
For recombinant Sir3 protein expression and purification, each SIR3 fragment was generated by PCR as an NcoI-SalI fragment and cloned at these sites into pET28b (Novagen), leading to the following plasmids expressing proteins with a C-terminal His6 tag: pJC65, containing the coding region for the first 380 aa of SIR3; pJC69, containing the coding region coding for the first 214 aa of SIR3; and pJC70, containing an NcoI-SalI fragment amplified from pBTM-SIR3N205 (32) and expressing SIR31-214 D205N-His6.
The SIR1 overexpression plasmid pES13B and the sir1ΔN plasmid pES13 have been described previously (36). The SIR1 mutant plasmids are derivatives of pES13B and were constructed by site-directed mutagenesis performed according to Stratagene's QuikChange site-directed mutagenesis protocol. First, a BamHI-BamHI SIR1 fragment from pES13B was cloned into pBluescript KS(+) to create pJC93. The corresponding mutation was made in pJC93, and the BamHI fragment with the mutation was then used to replace the equivalent fragment of pES13B to create pJC103 (sir1R31G), pJC118 (sir1L39P), or pJC119 (sir1D41N).
For patch mating assays, cells transformed with the indicated plasmids were patched onto the appropriate synthetic complete selective medium. Patches were grown for 1 day and transferred onto yeast-peptone-dextrose (YPD) by replica plating along with the appropriate mating tester strain (DC16 or DC17). After 1 day, these cells were transferred by replica plating onto synthetic dextrose plates to select for diploids. Diploids were allowed to grow for 2 days.
For quantitative mating assays, cells transformed with the indicated plasmids were grown in the appropriate liquid selective medium to exponential phase. Dilutions were made in YPD and plated onto YPD to count the total number of cells. Appropriate dilutions were also plated onto synthetic dextrose after mixing with 107 cells of an exponentially growing mating tester strain (DC16 or DC17). Mating efficiency was calculated as the fraction of cells that mated. In all cases, data were normalized relative to a value of 1 for wild-type SIR3 on a plasmid. All values are the means of data from three independent assays.
Sir31-214-His6, Sir31-214 D205N-His6, and Sir31-380-His6 were expressed from pJC69, pJC70, and pJC65, respectively, in BL21-CodenPlus (Stratagene). Expression was induced by the addition of 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) and 3% ethanol for 5 or 6 h. The proteins were purified using Novagen His Bind Affinity resin according to the manufacturer's instructions. Purified proteins were dialyzed against a buffer containing 50 mM HEPES, pH 7.5, 25 mM KCl, 1 mM EDTA, 1 mM EGTA, and 5 mM magnesium acetate. This buffer was used previously for assays to detect full-length Sir3/nucleosome binding (15). Protein concentrations were estimated by comparing Coomassie blue staining of samples to bovine serum albumin standards and were quantitated using a Bio-Rad Bradford assay.
Yeast nucleosomes were isolated with a protocol similar to that used for mammalian cells (8). Yeast nuclei were prepared as described previously (10). They were washed with MSB (20 mM HEPES, pH 7.5, 400 mM NaCl, 1 mM EDTA, 1 mM β-mercaptoethanol, 0.5 mM phenylmethylsulfonyl fluoride [PMSF], 5% glycerol, 2 mg/ml leupeptin, 2 mg/ml pepstatin, 5 mg/ml aprotinin, 5 mM sodium butyrate, 5 mM β-glycerophosphate, 0.5 mM spermidine, 0.15 mM spermine) and extracted with HSB (20 mM HEPES, pH 7.5, 650 mM NaCl, 1 mM EDTA, 340 mM sucrose, 1 mM β-mercaptoethanol, 0.5 mM PMSF, 2 mg/ml leupeptin, 2 mg/ml pepstatin, 5 mg/ml aprotinin, 5 mM sodium butyrate, 5 mM β-glycerophosphate, 0.5 mM spermidine, 0.15 mM spermine). The extract was dialyzed against LSB (same as HSB except with 100 mM NaCl). Calcium chloride was added to a 3 mM final concentration, and the chromatin was digested with 0.1 U/ml micrococcal nuclease for 5 min at 37°C. The reaction was stopped by the addition of 50 mM EDTA and NaCl at a final concentration of 0.6 M. The digested chromatin was concentrated to 2 ml on a Centriprep 30 column (Amicon) and injected onto a Superose 6 gel filtration column (Pharmacia). The column was equilibrated with HSB without sucrose prior to sample loading. Fractions (0.5 ml) were collected and analyzed for purity and oligonucleosome size by agarose gel electrophoresis. Appropriately sized oligonucleosome fractions were pooled, concentrated, and dialyzed against a solution containing 20 mM HEPES, pH 7.5, 1 mM EDTA, 1 mM β-mercaptoethanol, and protease inhibitors. Samples were stored at −80°C.
Four micrograms of yeast oligonucleosomes or 2 μg of 146-bp DNA was incubated with or without Sir3 protein in a solution containing 20 mM Tris, pH 8.0, 4 mM EDTA, pH 8.0, 0.5 mM dithiothreitol, and 1 mM PMSF for 1 h at room temperature. The reactions were analyzed by electrophoresis at room temperature on a 1.8% agarose gel (containing ethidium bromide) for 30 min at 100 V.
Chromatin immunoprecipitation was performed on strain JCY4 expressing either LexA, SIR31-380-LexA, or SIR31-380 D205N-LexA, as described previously (24). Extracts were sonicated 18 times for 10 s at level 2 with an Ultrasonics Inc. sonicator (model W220-F). Immunoprecipitations were performed using 10 μl antibody to LexA (D-19; Santa Cruz). Input and immunoprecipitation samples were analyzed by standard PCR mixtures to which 1 μCi of 32P[dATP] had been added. Products were resolved on 1.8% agarose gels and quantified using a Storm840 scanner and ImageQuant software (Molecular Dynamics). Results are normalized to inputs and expressed as enrichment relative to LexA alone. The results shown are representative ofseveral independent experiments. Primers used are as follows: for HMRa, 5′-CAGTTTCCCCGAAAGAACAA and 5′-CCATCCGCCGATTTATTT; for HMR-E, 5′-ACCAGGAGTACCTGCGCTTA and 5′-TGCAAAAACCCATCAACCTGG.
Sir3BAH (aa 1 to 219) was expressed in E. coli as a polyhistidine-tagged protein using a pET15b vector. Four amino acids, Met-His-Met-Thr, were designed to follow the thrombin cleavage site in the fusion protein before the first methionine of Sir3. The extra amino acids were added because they were used for our previous successful crystallization of Orc1BAH (40). The recombinant protein was first purified on a Ni-chelating column followed by in-column digestion of the polyhistidine tag and then ion-exchange column chromatography. The purified Sir3BAH protein was concentrated to 15 to 20 mg/ml for crystallization by the hanging-drop vapor diffusion method.
We obtained Sir3BAH crystals in several conditions containing polyethylene glycol 20,000 or sodium formate. The best-diffracting crystals were obtained with 4 M sodium formate. Thus, we used crystals from this condition for all the crystallographic studies here. Native X-ray diffraction data were collected at the X26C beamline of National Synchrotron Light Source, Brookhaven Nation Laboratory. The data were processed using the HKL suite of programs (31). The structure was solved by molecular replacement with the AMoRe program (30) using the Orc1BAH structure as the search model. Interactive cycles of model building and refinement were carried out using the programs O and CNS (3, 21). Statistics of crystallographic analysis are shown in Table S1 of the supplemental material. The figures were prepared using the Pymol program (9).
Atomic coordinates of the Sir3BAH structure have been deposited in the Protein Data Bank under accession number 2FYU.
In an attempt to understand the role of the Sir3 N terminus in silencing, initially, we expressed the first 380 amino acids of the protein and assessed its ability to silence HML and HMR in the absence of full-length Sir3 (i.e., in sir3Δ strains). This Sir3 fragment was expressed from a plasmid as a LexA hybrid, with LexA at the C terminus of the hybrid protein. Silencing was assessed both qualitatively by patch mating (Fig. (Fig.1A)1A) and quantitatively (Fig. (Fig.1B).1B). To our surprise, this protein was able to silence both HML and HMR in a significant fraction of cells (Fig. (Fig.1).1). When Sir1 was overexpressed, silencing was almost as good as that seen with full-length Sir3 (Fig. (Fig.1).1). The Sir3 N-terminal fragment was also expressed without the LexA tag. In that case, a smaller fraction of the cells was silenced than with the LexA hybrid, but again, it was improved by Sir1 overexpression (Fig. (Fig.11).
In view of these results, we next assessed the ability of Sir3BAH and Orc1BAH constructs, coding for the first 214 amino acids of the respective proteins and each fused to LexA, to silence the HM loci. Sir3BAH and Orc1BAH could silence HML in a significant fraction of cells but only if Sir1 was overexpressed (Fig. (Fig.2).2). It should be emphasized that silencing by Orc1BAH occurred in the absence of any Sir3 protein whatsoever. HMR could also be partially silenced by Sir3BAH, but not by Orc1BAH, again only when Sir1 was overexpressed (Fig. (Fig.22).
Expression of either BAH fragment without LexA did not restore silencing in these assays. It is likely that the enhanced silencing provided by the LexA tag was due to the ability of LexA to dimerize. In support of this, Sir3BAH-glutathione-S-transferase (GST) silenced just as well as the LexA-tagged version of Sir3BAH, and GST can also dimerize. Moreover, these tags had to be C-terminally fused, as an N-terminal LexA fusion to Sir3BAH gave no silencing (data not shown). This was not surprising, as we and others have reported previously that tagging Sir3 at its N terminus greatly weakened Sir3 function (7, 16, 17). This loss of function is due to the need for the N-terminal Ala residue of Sir3 to be both exposed and acetylated (39).
A hypermorphic allele of SIR3, sir3D205N, is known to suppress mutations in histones H4 and H3 that lead to a loss of silencing at HML (20, 37). This same sir3 allele suppressed the telomeric silencing defect of several rap1 alleles (26, 32). In each case, the sir3D205N allele improved silencing, in some cases to almost wild-type levels. Since residue D205 resides within Sir3BAH, we tested the effect of the D205N mutation on silencing by this protein. We found that Sir3BAH D205N-LexA greatly improved silencing at HML and HMR compared to wild-type Sir3BAH and even gave some silencing without extra Sir1 (Fig. (Fig.2).2). The D205N mutation was also introduced into the two Sir31-380 constructs shown in Fig. Fig.1.1. In those cases, the mutation also improved silencing, and in the case of HMR, silencing was improved almost to the level found with full-length Sir3 (data not shown).
Having identified a unique silencing function for the BAH domains of Sir3 and Orc1, we sought to determine if this silencing was dependent on the other Sir proteins and on NatA, the enzyme that acetylates the N-terminal Ala residues of Sir3 and Orc1 (14, 39). Plasmids expressing Sir3BAH, Orc1BAH, or Sir3BAH D205N, all C-terminally tagged with LexA, were cotransformed with a plasmid that overexpressed SIR1 into MATa strains with a deletion of SIR3 and a deletion of SIR1, SIR2, SIR4, or ARD1 (ARD1 codes for the catalytic subunit of NatA). We assessed silencing at HML by patch mating. Deletion of SIR2 or SIR4 abolished silencing by all three BAH proteins, as did deletion of SIR1 in the absence of the Sir1-overexpressing plasmid (Fig. (Fig.3).3). The same result was seen at HMR with Sir3BAH and Sir3BAH D205N and with the Sir31-380 constructs (data not shown). We conclude that silencing by each BAH domain depends on the usual silencing proteins (Sir1, Sir2, and Sir4). Figure Figure33 also shows that Sir3BAH and Orc1BAH could not silence in an ard1Δ strain, and Sir3BAH D205N silencing was diminished by the ard1Δ mutation. Thus, acetylation of the N termini of Sir3BAH and Orc1BAH is required for them to act in silencing. This is expected, given that NatA is known to acetylate full-length Sir3 and Orc1 and that Sir3 must be N-terminally acetylated in order to function fully (14, 39).
Finally, we checked whether the N terminus of H4, known to bind to a Sir3 C-terminal region (5, 25), was required for Sir3BAH or Sir31-380 silencing. Plasmids encoding these proteins were introduced into sir3Δ derivatives of strains with deletions of the H4 N terminus or deletions of either amino acids 4 to 14 or amino acids 4 to 23 (22). Silencing by the Sir3 N-terminal fragments was greatly weakened in the deletion of aa 4 to 14 and abolished in the deletion of aa 4 to 23 (data not shown).
Since the Sir3BAH and Sir31-380 proteins gave Sir-dependent silencing, it seemed likely that they would spread from the silencers to the silenced loci, just as is seen with Sir2, Sir3, and Sir4 in a wild-type cell, when all the full-length Sir proteins are present. To monitor the presence of Sir31-380-LexA both at the HMR-E silencer and in the HMR-silenced region, we used chromatin immunoprecipitation with an antibody to the LexA tag to immunoprecipitate Sir31-380-LexA or Sir31-380 D205N-LexA. To increase the fraction of cells silenced, Sir1 was also overexpressed. As seen in Fig. Fig.4,4, Sir31-380 and Sir31-380 D205N were greatly enriched at both the HMR-E silencer and the HMR locus itself. The presence of Sir31-380-LexA at the HMR locus was also examined in sir1 and sir4 mutants. As expected, no enrichment over the background was seen in these mutants (data not shown).
In order to test whether Sir3BAH bound to oligonucleosomes or DNA, we performed in vitro experiments using recombinant Sir3BAH purified from E. coli and oligonucleosomes from yeast. To obtain oligonucleosomes, chromatin was isolated and partially digested with micrococcal nuclease, and oligonucleosomes were size fractionated on a gel filtration column (Fig. (Fig.5A).5A). Two pooled fractions were used for binding assays (Fig. (Fig.5A,5A, labeled A and B): fraction A had oligonucleosomes with approximately 2 to 5 nucleosomes, and fraction B had mono- and dinucleosomes. Binding was assessed by a standard gel shift assay. Briefly, the Sir3BAH fragment was incubated with oligonucleosomes, and the reaction mixtures were electrophoresed on a 1.8% agarose gel containing ethidium bromide to assess the mobility of the oligonucleosomes. The results of this experiment are shown in Fig. Fig.5B.5B. The addition of Sir3BAH did not affect oligonucleosome mobility, whereas the addition of Sir3BAH D205N led to a decrease in mobility. C-terminal GST-tagged Sir3BAH, Sir3BAH D205N, and Orc11-214 were also tested in this assay. As with the His-tagged proteins, only the Sir3BAH D205N protein had the ability to retard the mobility of the oligonucleosomes (data not shown). These results show that the Sir3 D205N mutation greatly increased the ability of Sir3BAH to bind to oligonucleosomes.
Georgel et al. previously reported that full-length Sir3 bound to DNA (15). We reasoned that the association of Sir3BAH D205N with nucleosomes could be occurring through an interaction with DNA. We tested Sir3BAH and Sir3BAH D205N binding to DNA using a gel shift assay with a 146-bp DNA fragment. Just as with oligonucleosomes, Sir3BAH did not shift DNA, whereas Sir3BAH D205N bound DNA and retarded its mobility at protein concentrations equivalent to those seen for nucleosomes (Fig. (Fig.5C).5C). These data show that the D205N gain-of-function mutation found within the BAH domain of Sir3 allowed the BAH domain of Sir3 to associate with nucleosomes through an interaction with DNA. Interestingly, the Sir31-380 protein bound to both oligonucleosomes and DNA even without the D205N mutation (Fig. (Fig.5C5C and data not shown).
As mentioned above, a C-terminal region of Sir1 interacts with Orc1BAH but does not interact with Sir3BAH. We reasoned that a protein with similarity to this region of Sir1 might exist and be the relevant Sir3BAH partner. We conducted a BLAST search using Sir1 amino acids 346 to 678 as the query and, to our surprise, found significant homology with a region in the Nterminus of Sir1. Amino acids 471 to 609 of Sir1 are 28%identical and 45% similar to amino acids 13 to 149 of Sir1 (Fig. (Fig.6A).6A). Bose and colleagues have previously shown that the minimal region of Sir1 necessary to support an interaction with Orc1BAH in the two-hybrid system maps to amino acids 490 to 611 of Sir1, and those authors defined this domain as the OIR (2). This is almost exactly the same region of Sir1 that appears to be duplicated in the Sir1 N terminus (Fig. (Fig.6A).6A). Mutations within the OIR of Sir1 that disrupt the Sir1-Orc1BAH interaction and lead to a loss of establishment of silencing have been identified (2). There are 12 such mutations in the region between residuesY489 and Y607 of Sir1 (Fig. (Fig.6A,6A, boxed in gray). Eight of these 12 residues are conserved between these two regions of Sir1, which we will refer to as Sir1C and Sir1N; moreover, 6 of these 8 residues are identical.
Interestingly, the Sir1N region is not needed for complementation of sir1Δ mutants. For example, that region of Sir1 was missing in the original SIR1-complementing clone (36) (see below). For simplicity, we will refer to this truncation allele as sir1ΔN and the protein as Sir1ΔN. Although plasmids overexpressing Sir1ΔN complement sir1 mutations as well as do plasmids overexpressing the full-length protein, only overexpression of full-length Sir1 restores silencing to mutations that weaken silencing, such as a leaky sir3 or sir4 allele or nat1/ard1 mutations (36). Interestingly, the region of Sir1 missing in Sir1ΔN encompasses the SirN region with similarity to Sir1C (Fig. (Fig.6A).6A). We sought to determine if this duplicated region in the N terminus of Sir1 was necessary for the silencing exhibited by the various Sir3 and Orc1 BAH fragments that required Sir1 overexpression (Fig. (Fig.11 and and2).2). We compared BAH silencing in the presence of a plasmid expressing Sir1ΔN to that of a plasmid overexpressing full-length Sir1. As is shown in Fig. Fig.6C,6C, full-length Sir1 is required for silencing by these proteins; the Sir1ΔN protein did not support BAH silencing. It did complement a sir1Δ strain, however (Fig. (Fig.6B),6B), as has been shown previously (36).
The above-described experiments demonstrated that the N terminus of Sir1 is required for silencing by Sir3BAH and Orc1BAH. Since this is the region of Sir1 that is duplicated in Sir1C, and since residues within Sir1C that affect the Sir1C/Orc1BAH interaction are conserved in the Sir1 N terminus, we engineered three of these mutations, R31G, L39P, and D41N (Fig. (Fig.6A,6A, highlighted in red), into the N terminus of Sir1. We tested each mutant's ability to complement the silencing defect of a sir1 strain and to support BAH silencing.
For the complementation test, plasmids expressing full-length Sir1, Sir1ΔN, Sir1R31G, Sir1L39P, Sir1D41N, and a vector were transformed into a strain that harbors a TRP1 reporter at HMR and is deleted for SIR1. Complementation of sir1Δ was assessed by the inability to grow on medium lacking tryptophan (Fig. (Fig.6B).6B). As expected, a deletion of SIR1 abolished silencing of the TRP1 reporter in this strain, and a plasmid expressing Sir1 complemented this deletion in that it silenced the TRP1 gene. The plasmid expressing the truncated version of Sir1 (Sir1ΔN) also complemented sir1Δ (this confirms the previously reported data), and each of the SIR1 N-terminal point mutants also complemented sir1Δ (Fig. (Fig.6B).6B). A similar assay was performed using a reporter at HML, and the results were the same as those for HMR (data not shown). These data suggest that each mutant is expressed and is able to complement in the context of normal silencing.
In order to determine the effect of these N-terminal SIR1 mutants on BAH silencing, each of the mutants was tested for its ability to enhance Sir3BAH and Orc1BAH silencing at HML. The various Sir1 plasmids were cotransformed with plasmids expressing Sir3BAH or Orc1BAH, and silencing at HML was assessed (Fig. (Fig.6C).6C). For both BAH proteins, Sir1L39P and Sir1D41N led to an almost complete loss of silencing at HML, whereas the Sir1R31G mutant had no effect. These data demonstrate that the Sir1 N terminus plays an essential role in BAH silencing. Given the sequence similarity between Sir1N and the Sir1C OIR (which binds Orc1BAH), we wanted to test if Sir1N directly interacted with Sir3BAH. Unfortunately, we were unable to isolate soluble fragments of Sir1N expressed in E. coli, so we could not do this experiment with purified proteins.
As described above, Sir3BAH shares significant sequence homology with Orc1BAH, but they have distinct functions in transcriptional silencing. To understand the structural basis for their functional differences, we have determined the crystal structure of Sir3BAH and compared it with the Orc1BAH structure determined previously (40).
The Sir3BAH structure was determined by molecular replacement using the Orc1BAH structure as the search model. The structure has been refined to 1.9-Å resolution. There are two Sir3BAH molecules in one crystallographic asymmetric unit. Similar to Orc1BAH, Sir3BAH contains a predominantly β-sheet BAH domain and a ~40-residue helical domain, termed the H domain (Fig. 7A and B). The H domain of Orc1BAH has been shown to be important for interacting with the C-terminal domain of Sir1. The structures of Sir3BAH and Orc1BAH can be superimposed with a root-mean-squared deviation of 1.1 Å using the Cα positions of the BAH core for alignment (Fig. (Fig.7C).7C). Significant conformational differences are found in three regions. (i) Residues 1 to 5 of Sir3BAH form an antiparallel β-strand with residues 11 to 15. In the Orc1BAH structure, the N-terminal tail projects out of the core domain and interacts with a neighboring molecule. (ii) Residues 77 to 82, which form a β-hairpin in the Orc1BAH structure, are disordered in the Sir3BAH structure. Curiously, these residues are identical between the two proteins. (iii) Residues 161 to 164, which are spatially adjacent to the H domain, are in a conformation further away from the H domain, by as much as 8.8 Å, than they are in the Orc1BAH structure. Amino acids connecting β1 and β2 (aa 18 to 37 in Sir3BAH) are disordered in both the Orc1BAH and Sir3BAH structures.
It is interesting that the N-terminal tails of Sir3BAH and Orc1BAH have significantly different conformations, despite 100% sequence identity of the first eight residues. In the Orc1BAH structure, the N-terminal tail is involved in interacting with neighboring molecules (40). This region of Sir3BAH is important for silencing (39), and in the structure, it is involved in protein-protein interactions between the two molecules within the same asymmetric unit. Three additional stretches of amino acids are also involved in the protein-protein interactions. They are residues 42 to 48, 148 to 151, and 162 to 170. Gln157, located between the last two regions, also contacts the other molecule.
In the crystal lattice, the Sir3BAH dimer in the asymmetric unit packs with symmetry-related molecules to form a contiguous left-handed superhelix. Figure Figure8A8A shows a view of the crystal packing along the superhelical axis. Three “strands” of parallel protein dimers wrap around each other to form a tubular architecture. The inner diameter of the superhelix is approximately 80Å. The pitch of the superhelix is approximately 197 Å, with the closest and furthest separations being 130 Å and 250 Å, respectively (Fig. (Fig.8B).8B). Obviously, the organization of the superhelix in the crystal lattice is under a rather artificial condition of crystal packing. Nevertheless, in many cases, the crystal packing can stabilize weak and transient interactions that might be difficult to detect in solution. It is interesting that the minimal diameter of a nucleosome core particle is about 85 to 90 Å (28). Considering the physiological function of Sir3 in spreading along the chromatin in the silent mating type loci, it is possible that Sir3 will coat the nucleosomes within the silent chromatin domain in a fashion similar to the one observed in the crystal packing. It is also of interest that the inner surface of the superhelix is highly negatively charged, thus making it energetically unfavorable for interaction with the backbone of DNA directly (Fig. (Fig.8B).8B). However, the charge distribution in Sir3BAH is consistent with the inner surface contacting the highly positively charged N-terminal tails of histones. As seen Fig. Fig.8C,8C, the D205 residues are on the outside surface of the superhelix, in a position where they could interact with DNA, as is discussed below.
We have shown that short N-terminal fragments of Sir3, lacking any of the known protein-interacting domains of full-length Sir3, give Sir1-, Sir2-, and Sir4-dependent silencing in a measurable fraction of cells. This silencing is greatly increased, and in some cases totally dependent, on overexpression of full-length Sir1 or upon introduction of the hypermorphic sir3 D205N mutation. Similarly, the Orc1BAH fragment can also give silencing, at least at HML, as long as Sir1 is overexpressed. For both Sir3 and Orc1, the fragments encompassing amino acids 1 to 380 silence a larger fraction of cells than do the corresponding BAH proteins (Fig. (Fig.11 and and22 and data not shown).
Previous studies have demonstrated that mutations in the N-terminal region of Sir3 interfered with its silencing function (35, 39). Our observation that the BAH domain can silence in the absence of full-length Sir3 suggests that the BAH domain possesses the essential silencing features of the full-length protein, including (i) the ability to be recruited to the silent chromatin, (ii) interaction with the nucleosome, and (iii) spreading along the silenced region. Below, we discuss possible mechanisms by which the Sir3 BAH domain functions in these processes.
Full-length Sir3 has been shown to interact with Rap1, Sir4, and histone H3 and histone H4 N-terminal tails, and these interactions are thought to be responsible for the recruitment of Sir3 to the silent HM loci. Interestingly, the Sir3 regions involved in these interactions are located outside the BAH domain (also beyond the first 380 residues). A natural question to ask is what protein-protein interactions recruit the Sir3 BAH domain to the silent chromatin. Our circumstantial evidence suggests that Sir3BAH interacts with the N-terminal domain of Sir1 and that this interaction is important for the recruitment of Sir3BAH (Fig. (Fig.66).
The previously characterized role of Sir1 in silencing was to bind to Orc1BAH and to attract Sir4 to the HM silencers through this interaction. These interactions involved a C-terminal domain of Sir1 called the OIR. Here, we have shown that there is a region near the N terminus of Sir1 that shows significant similarity to the OIR (Fig. (Fig.6A).6A). Moreover, this region is required for silencing by Orc1BAH and Sir3BAH (Fig. (Fig.6C).6C). Interestingly, the N-terminal region of Sir1 is not required for complementation of sir1Δ mutants (Fig. (Fig.6B)6B) but is required for the suppression of various silencing mutations such as nat1, ard1, sir3-8, and sir4-9 (36). It is noteworthy that the sir3-8 allele changes residue E131 within the BAH domain to K and that nat1/ard1 affects the acetylation of residue A2 (14, 35, 39). We hypothesize that the N-terminal region of Sir1 interacts with the H domain of Sir3BAH just as the OIR of Sir1 interacts with H domain of Orc1BAH (Fig. (Fig.7).7). We have been unable to demonstrate this directly, however, because of the insolubility of the Sir1 N terminus when expressed in E. coli.
Whereas Sir3BAH, Orc1BAH, and Sir31-380 could partially silence the HM loci (Fig. (Fig.11 and and2),2), absolutely no silencing by these proteins was observed at telomeres in a sir3Δ strain (data not shown). We think that this is because there are no ORC binding sites near the telomere reporter gene, and therefore, Sir1 cannot be attracted to that region. Since Sir1 is absolutely required for silencing by these Sir3 and Orc1 fragments, no telomeric silencing is observed in the absence of Sir3. This fits with the results described previously by Gotta et al. that demonstrated that the enhancement of telomeric silencing by Sir31-503 could only be seen in the presence of full-length Sir3 (17). No full-length Sir3 was present in our experiments.
Our in vitro binding data show that the silencing capacity of the Sir3 N-terminal domain correlates with the DNA/oligonucleosome binding ability. Sir3BAH with the D205N mutation bound DNA better than Sir3BAH, and it silenced better (Fig. (Fig.22 and and5).5). The same was the case with Sir31-380 (data not shown). The binding results are consistent with a previous observation that full-length Sir3 has the ability to associate with the nucleosome and DNA (15). The better DNA binding ability exhibited by the D205N mutant provides an explanation for the long-standing puzzle that this mutation suppresses silencing defects that occur in several places on the nucleosome and on Rap1 (20, 26, 32, 37). We believe that the mode of suppression could be due to a better association with the DNA wrapped around the nucleosome. This better DNA association would then compensate for a weaker association between Sir3 and another region of the nucleosome or with a mutated Rap1.
The structure of Sir3 BAH shows that the protein surface is quite negatively charged (calculated pI of 5.3 for residues 1 to 219 and pI of 6.0 for full-length Sir3). The negatively charged surface favors an interaction with positively charged histone tails but disfavors interactions with DNA. The D205N mutation lessens the negative charge in a particular region of the protein surface, which is likely to promote a tighter association with DNA. The structure of Sir3 BAH does not resemble any sequence-specific DNA binding protein modules, which is probably to be expected, because Sir3 spreads a long distance along the silent chromatin. It is interesting to compare Sir3 with the nucleosome binding properties of HP1, a protein essential for epigenetic silencing in higher eukaryotes that has the ability, once targeted, to spread great distances, much like Sir3. In addition to the well-documented interaction with methylated histone H3, HP1 also exhibits nonspecific DNA association, and it has been suggested that this interaction could explain how it can spread (41). We hypothesize that the Sir3/DNA association could explain how Sir3 has the ability to spread, once targeted.
Full-length Sir3 can interact with itself, Sir4, and the nucleosome, which form the molecular basis for its ability to spread along the silenced chromatin domain. As we mentioned earlier, no such properties of Sir3 BAH or the longer Sir3 (aa 1 to 380) have been documented, and yet they are capable of silencing. We have discussed above that the Sir3 N-terminal domain appears to have an intrinsic DNA binding activity and that this binding activity is enhanced in the presence of the D205N mutation. Here, we will examine the self-interaction ability of Sir3 BAH, a hallmark property exhibited by silencing proteins such HP1 and full-length Sir3.
While we could not detect self-interaction of purified Sir3BAH produced in bacteria, the crystal structure provides indirect evidence to suggest that Sir3 BAH may self-associate. Sir3 BAH is crystallized with two molecules in one asymmetric unit, and crystal packing organizes the noncrystallographic dimer into a contiguous superhelical structure. We are intrigued by the Sir3 BAH quaternary structure in the crystal because it reveals a favorable property for Sir3 BAH spreading. It is worth noting that two Sir3 BAH regions important for silencing, the N terminus and a region encompassing D205, are both involved in intermolecular interactions that lead to the formation of the superhelical structure. We suggest that Sir3 BAH can self-interact and that this property is important for the role of full-length Sir3 in silencing.
We thank Xiaorong Wang and Ann Sutton for advice, M. Grunstein for antibodies, J. Camonis for the LexA plasmid, and Annie Heroux and Dieter Schneider for help during data collection at the X26C beamline of the National Synchrotron Light Source of Brookhaven National Laboratory.
The work was supported in part by the W.M. Keck Foundation (R.-M.X.) and NIH grants GM63716 to R.-M.X. and GM28220 to R.S.
†Supplemental material for this article may be found at http://mcb.asm.org/.