Silencing of the cryptic mating-type loci HMR and HML requires the recognition of DNA sequence elements called silencers by the Sir1p, one of four proteins dedicated to the assembly of silenced chromatin in Saccharomyces cerevisiae. The Sir1p is thought to recognize silencers indirectly through interactions with proteins that bind the silencer DNA directly, such as the origin recognition complex (ORC). Eight recessive alleles of SIR1 were discovered that encode mutant Sir1 proteins specifically defective in their ability to recognize the HMR-E silencer. The eight missense mutations all map within a 17-amino-acid segment of Sir1p, and this segment was also required for Sir1p's interaction with Orc1p. The mutant Sir1 proteins could function in silencing if tethered to a silencer directly through a heterologous DNA-binding domain. Thus the amino acids identified are required for Sir1 protein's recognition of the HMR-E silencer and interaction with Orc1p, but not for its ability to function in silencing per se. The approach used to find these mutations may be applicable to defining interaction surfaces on proteins involved in other processes that require the assembly of macromolecular complexes.
EUKARYOTIC chromosomes are organized into structural domains that can regulate their replication, segregation, and expression (Wolffe and Pruss 1996). For example, the chromosomal domains of centromeres direct the assembly of the kinetochore that participates in the segregation of chromosomes (Hyman and Sorger 1995; Ekwallet al. 1996; Saitohet al. 1997). Other domains can significantly influence the expression of genes that reside within them (Loo and Rine 1995; Edmondson and Roth 1996; Elgin 1996; Elgin and Jackson 1997). Although molecules required for assembly of specialized chromosomal domains have been identified, little is known about how the assembly of specialized domains is restricted to some chromosomal regions and not others. Elucidation of the mechanisms that target certain chromosomal regions for assembly into specialized domains should reveal fundamental principles relevant to the overall functional organization of chromosomes.
Transcriptional silencing of the silent mating-type loci, HML and HMR, in the yeast Saccharomyces cerevisiae provides a well-documented example of specialized chromosomal domains that block transcription of genes that reside within them (Herskowitzet al. 1992). In yeast, mating type is controlled by the alleles present at the MAT locus; the MATa allele confers the a-mating phenotype whereas the MATα allele confers the α-mating phenotype. In most yeast strains, a silenced copy of the MATa allele resides at HMR and a silenced copy of the MATα allele resides at HML. Transcriptional silencing of HML and HMR requires the combined action of regulatory sites called silencers, several multifunctional nuclear proteins that bind the silencers directly, and silencing-specific proteins that are thought to recognize the silencer-binding proteins and contribute to the assembly of silenced domains of chromatin (Loo and Rine 1995).
Yeast silencers contain binding sites for the origin recognition complex (ORC), the Rap1p, and/or Abf1p (Brandet al. 1987; Kimmerlyet al. 1988; Bellet al. 1993). Molecular and genetic evidence indicates that these proteins function in silencing by binding their DNA target sites within the silencers (Loo and Rine 1995). Significantly, these proteins have other functions as well. ORC functions as the DNA replication initiator at each of the hundreds of replication origins in the yeast genome (Bell and Stillman 1992; Foxet al. 1995; Lianget al. 1995). Rap1p binds telomeres, where it functions in telomeric silencing and length regulation (Shore 1994). Rap1p also binds to many promoters, where it contributes to transcriptional activation (Shore 1994). Abf1p binds some replication origins, where it contributes to replication initiation (Marahrens and Stillman 1992). Abf1p also binds to many promoters, where it contributes to transcriptional activation (Kanget al. 1995; Gailus-Durneret al. 1996; Rolfeset al. 1997). Not surprisingly, the genes that encode the ORC subunits, Rap1p, and Abf1p, are essential for viability (Loo and Rine 1995). Thus all these proteins have multiple roles: in essential functions such as DNA replication and transcriptional activation and in the nonessential function of transcriptional silencing.
The silencer-binding proteins probably function in silencing by helping to recruit the four Sir proteins to HML and HMR. The Sir proteins, unlike the silencerbinding proteins, are nonessential for life, yet essential for transcriptional silencing (Loo and Rine 1995). Thus, the question of how the silent loci are targeted for assembly into specialized chromatin boils down to how the silencer-binding proteins, ORC, Rap1p, and Abf1p, recruit the Sir proteins to HML and HMR.
The Sir1p is likely to be a key participant in the recruitment of the other Sir proteins because its principal function appears to be in the establishment of the silenced state at HMR and HML (Pillus and Rine 1989; Aparicioet al. 1991; Foxet al. 1997). Sir1p probably recognizes a silencer, at least in part, through interactions with ORC because the requirement for ORC in silencing HMR can be bypassed by a Gal4-Sir1p fusion protein tethered to the HMR silencer (Triolo and Sternglanz 1996; Foxet al. 1997). This Gal4-Sir1p-mediated silencing requires the other Sir proteins (Chienet al. 1993). Moreover, two-hybrid experiments indicate that a portion of Orc1p can physically interact with a portion of Sir1p (Triolo and Sternglanz 1996). Together, these data focus the early steps in silencing on the question of how Sir1p is recruited specifically to the silencers of HMR and HML and not to the many other sites at which individual silencer-binding proteins are found. A simple view would be that the ensemble of silencer-binding proteins creates a composite surface to which Sir1p binds. This model predicts that Sir1p might have a surface dedicated to interacting with the silencer-binding proteins.
To begin addressing the mechanism by which Sir1p recognizes a silencer, we performed a genetic screen to identify alleles of SIR1 that produce mutant Sir1 proteins that are defective in the ability to recognize the HMR silencer yet retain their ability to silence per se. The resulting alleles revealed that the ability of Sir1p to recognize the HMR silencer was genetically separable from the ability of Sir1p to function in the assembly of silenced chromatin. A small discrete region of the Sir1p was identified as being necessary for recognizing the HMR silencer and this same region was required for interacting with Orc1p, the largest subunit of ORC. Thus this region of Sir1p may define a Sir1p-ORC interaction motif.
MATERIALS AND METHODS
The genotypes of the yeast strains and the plasmids used in this study are in Tables 1 and 2. Yeast rich medium (YPD), minimal medium (mM), amino acid and base supplements, and standard yeast genetic methods were as described (Guthrie and Fink 1991). Recombinant DNA methods were as described (Sambrooket al. 1989).
Strain construction: All strains were isogenic to W303 except as noted. The strain used for the α-mating lawn (CFY617, Figures 1 and 4) was constructed by switching the mating type of JRY19 from a to α with a plasmid encoding the HO endonuclease. To construct the strains used for the experiments presented in Table 3 the following crosses were performed: A MATα HMR-SSa sir1Δ::LEU2IVY strain (CFY114) was constructed by crossing a MATα HMR-SSa strain (JRY4492) to a sir1Δ::LEU2IVY strain (JRY2335) and selecting a MATα sir1Δ::LEU2IVY segregant with HMR-SSa. A MATα HMR-SSa (acs– GAL4 ABF1) strain (CFY238) was constructed by using one-step gene replacement to integrate an HMR-SSa (acs– GAL4 ABF1) DNA fragment at the HMR locus in a MATα hmrΔ::URA3 strain (JRY3933). The sir1Δ::LEU2IVY mutant in these strains was described previously (Ivyet al. 1986) and will be designated with IVY in superscript to distinguish it from the sir1Δ::LEU2 deletion mutant described below. Data indicated that the two different SIR1 mutant alleles behaved identically. SS refers to a synthetic version of the HMR-E silencer that consists solely of an autonomously replicating sequence (ARS) consensus sequence (ACS), a binding site for Rap1p (RAP1), and a binding site for Abf1p (ABF1) (McNally and Rine 1991). Mutant derivatives of the silencer are represented by lowercase letters for the mutant sites. GAL4 represents a Gal4p-binding site incorporated in place of the Rap1p-binding site of the synthetic silencer.
To construct the diploid strain used to identify the silencer-recognition-defective SIR1 alleles (CFY456, Figure 1) a hmlaΔp mataΔp HMR-SSα sir1Δ::LEU2IVY strain (CFY444) was mated to a hmlaΔp mataΔp HMR-SSa (acs– GAL4 ABF1) sir1Δ::LEU2IVY strain (CFY436).
To examine the effect integrated GAL4-SIR1 alleles had on silencing the synthetic silencer, a plasmid for directing integration at the TRP1 locus (pRS304) was used to insert the appropriate GAL4-SIR1 allele at the TRP1 genomic locus of a MATα HMR-SSa sir1Δ::LEU2 strain (CFY761) via a unique XbaI site. This procedure allowed the construction of isogenic MATα HMR-SSa sir1Δ::LEU2 strains that contained either GAL4-SIR1::TRP1 (CFY792), GAL4-sir1-100::TRP1 (CFY789), or GAL4-sir1-101::TRP1 (CFY788). The sir1Δ::LEU2 allele in these strains lacked the entire coding region of Sir1p and contained a complete LEU2 gene in its place. To examine the effect the integrated alleles had in silencing a HMR locus flanked by a synthetic silencer containing a Gal4p-binding site in place of the Rap1p-binding site [HMR-SSa (ACS GAL4 ABF1)], the appropriate GAL4-SIR1 allele was integrated at the TRP1 locus of a MATα HMR-SSa (ACS GAL4 ABF1) sir1Δ::LEU2 strain (CFY770). This integration produced isogenic MATα HMR-SSa (ACS GAL4 ABF1) sir1Δ::LEU2 strains that were GAL4-SIR1::TRP1 (CFY795), GAL4-sir1-100::TRP1 (CFY793), or GAL4-sir1-101::TRP1 (CFY794).
To construct the isogenic MATα HMR-SSa strains containing different mutant SIR1 alleles (Table 6), the appropriate mutant SIR1 alleles in pRS406 were exchanged with the chromosomal SIR1 gene by two-step gene replacement into a MATα HMR-SSa strain (JRY4492). Integration was directed to the BglII site of SIR1. Ura+ transformants were transferred to 5-FOA to select for recombinants that regenerated an intact SIR1 gene. Those recombinants that received a mutant SIR1 allele were determined by mating assays with a MATa lawn (JRY19). The weak α-maters contained mutant SIR1 alleles and the integrity of the SIR1 locus in all strains was determined by DNA-blot hybridization.
To construct the diploids used to test whether the mutant SIR1 alleles were recessive (Figure 5), a hmlaΔp mataΔp HMR-SSα strain (ROY1) was mated to a hmlaΔp mataΔp HMR-SSα strain containing either a sir1Δ::LEU2 (CFY444) allele or a sir1-101 allele (CFY732). Six independent diploids from each cross were examined and all yielded the same results. The analogous diploids were constructed with both sir1-100 and sir1-102 alleles and yielded the same results as those shown in Figure 5 (C. Fox, unpublished results).
Plasmid constructions: To construct HMR-SSα for integration at HMR (pCF51), the Eco0109I-BsaAI fragment containing the a1 gene from HMR was replaced with the Eco0109I-BsaAI fragment containing the α1 gene from HML within an EcoRI-HindIII HMR-SS fragment in pUC18. To place SIR1 under the control of the ADH1 promoter (pCF128), a HindIII site was inserted 10 bp upstream of the initiation codon of SIR1 and a HindIII fragment containing the entire SIR1 coding region was cloned into the HindIII site of a plasmid (pRS316) that contained a 1.5-kb BamHI/HindIII ADH1 promoter-containing fragment (from pMA424). To place GAL4-SIR1 and GAL4-sir1-101 under the control of an ADH1 promoter on a 2-μm plasmid for the overexpression experiment (Figure 6), a NotI/KpnI fragment containing either GAL4-SIR1 under the control of the ADH1 promoter (from pCF117) or an analogous fragment containing mutant GAL4-sir1-101 was cloned into pRS426 (pCF413 and pCF415, respectively). The same strategy was used to engineer the same contructs into a pRS416 CEN plasmid (pCF409 and pCF411).
PCR mutagenesis: The region encoding the C-terminal half of SIR1 was mutagenized using the following pair of primers that annealed to the template plasmid pJR909 (SIR1 in pRS316): CAGCTATGACCATGATTACGC, which annealed to pRS316 vector sequences that flanked the 3′ end of SIR1, and GGT TTATTGTCAGGGAAAAGCC, which annealed within the SIR1 coding region 930 bp 3′ of the SIR1 initiation codon in the same orientation as the SIR1 reading frame. The N-terminal half of SIR1 was mutagenized using the following pair of primers that annealed to the template plasmid pCF117 (GAL4-SIR1 in pRS316): CCTCGAGAAGACCTTGACATG, which annealed to the Gal4p coding region, and GGAAATTTCGAAAT TTGATCTAGG, which annealed within the SIR1 coding region 1130 bp 3′ of the SIR1 initiation codon in the opposite orientation of the SIR1 reading frame. A typical PCR reaction contained 30 ng of template plasmid, 30 pmol of each primer, between 5 and 6 mm MgCl2, 0 to 0.15 mm MnCl2, either 200 μm dATP or dGTP, and 1 mm each of the nonlimiting dNTPs. The PCR cycle conditions were as described (Muhlradet al. 1992).
The screen: To identify mutations within the coding region for the C-terminal half of Sir1p that caused a specific silencer-recognition defect, the detector diploid was cotransformed with a linearized plasmid (pCF117), which contained the GAL4-SIR1 fusion-lacking sequences between the BglII and KpnI sites, and PCR-mutagenized DNA that overlapped the deletion and could recombine with the gapped plasmid. To identify mutations within the coding region for the N-terminal half of Sir1p, the detector diploid was cotransformed with a linearized plasmid (pCF117), which contained the GAL4-SIR1 fusion-lacking sequences between the SmaI and BglII sites, and PCR-mutagenized DNA that overlapped the deletion and could recombine with the gapped plasmid. On average, 100 bp of overlapping homology existed between the ends of the gapped pCF117 plasmid and the ends of the PCR fragments allowing for efficient recombination between the plasmid and the PCR fragments and regeneration of an intact plasmid (Muhlradet al. 1992). Transformants were selected on medium lacking uracil. After incubating for 3 days at 30°, the transformants were replica plated to selective medium containing a MATa lawn (JRY19). Colonies exhibiting an α-mating phenotype were recovered from the master plates for further analysis. Two separate pools of PCR-mutagenized fragments were used for screening the C-terminal coding region of SIR1. In separate transformation experiments, 45% of recombinant plasmids failed to complement the sir1Δ::LEU2 mutation, indicating that the level of mutagenesis was high. However, only eight independent silencer-recognition-defective SIR1 mutant alleles were recovered from an examination of ∼18,000 transformants of the detector diploid, suggesting that mutations that caused a specific silencer-recognition defect in Sir1p were relatively rare. Three separate pools of PCR-mutagenized fragments were used for screening the N-terminal coding region of SIR1. On the basis of separate transformation experiments, it was determined that between 25 and 30% of the recombinant plasmids failed to complement a sir1Δ::LEU2 mutation. However, no silencer-recognition-defective SIR1 mutant alleles were recovered from an examination of ∼6000 transformants of the detector diploid. The mutant plasmids were recovered, retested as described in results, and the mutagenized portion of SIR1 was sequenced.
Quantitative mating assays: The quantitative mating assays were performed as described except that the mating lawn used was JRY19 (Ehrenhofer-Murrayet al. 1996). In addition, the mating efficiency of strains that harbored plasmids was determined under conditions that selected for the presence of the plasmid in the resulting diploids.
Immunoblot analysis of wild-type and mutant versions of Gal4-Sir1p: The level of Gal4-Sir1p in crude yeast extracts was determined as described (Ehrenhofer-Murrayet al. 1996). Both 0.044 OD cell equivalents (1×) and 0.132 OD (3×) cell equivalents of crude yeast extracts were examined for the presence of Gal4-Sir1p by immunoblot analysis. The primary antibody was a rabbit polyclonal made against the Gal4p DNA-binding domain (Upstate Biotechnology, Lake Placid, NY).
RNA blot analysis of a1 mRNA and SCR1 RNA: Total yeast RNA was prepared and RNA-blot hybridization was performed with a probe for a1 and, as a loading control, with a probe for SCR1, as described previously (Fox et al. 1995, 1997).
Previous studies established that a fusion protein consisting of the DNA-binding domain of Gal4p fused to the complete coding sequence of the Sir1p complements the silencing defect of sir1Δ strains and silences the HMR locus when tethered directly to the HMR silencer via a Gal4p-binding site (Chienet al. 1993; see also Table 3). Furthermore, Gal4-Sir1p-dependent silencing bypasses the requirement for both the ORC-binding site (ACS) and the Rap1p-binding site in silencing HMR (Triolo and Sternglanz 1996; Foxet al. 1997; see also Table 3). These data set the stage for a screen designed to identify alleles of SIR1 that produce mutant Sir1p molecules specifically defective in their ability to recognize the HMR-E silencer (Figure 1A). In principle, such alleles would identify a region of Sir1p that contacts a feature of the silencer-binding proteins that uniquely identifies silencers.
A screen for SIR1 alleles defective in HMR-E silencer recognition: We used a specialized diploid strain to identify alleles of SIR1 that produced forms of Sir1p that were defective in recognition of the HMR-E silencer yet could still promote silencing of HMR when brought to HMR by other means (Figure 1B). The mutant SIR1 alleles that were silencer-recognition-defective were referred to collectively as sir1srd alleles. The specialized diploid (CFY456), called the detector diploid, was homozygous for a sir1Δ::LEU2 mutation and thus did not produce Sir1p. In addition, both HML loci and both MAT loci contained a-mating-type genes that harbored a deletion in the promoter for the a genes (aΔp). Thus, no mating-type information was expressed from either HML or MAT in this strain. These mutant loci were designated hmlaΔp and mataΔp, respectively. The detector diploid was heterozygous at the HMR locus. One HMR locus contained the α-mating-type genes under the control of the synthetic HMR silencer (HMR-SSα). The synthetic silencer is a simplified version of the natural HMR-E silencer (McNally and Rine 1991) that is completely dependent on SIR1 function (Table 3) and therefore provided the sensitivity to SIR1 function required by the screen. Thus, in the absence of wild-type SIR1, the α genes were expressed from this HMR locus. The second HMR locus contained the a-mating-type genes under the control of a synthetic HMR silencer in which the Rap1p-binding site was replaced with a Gal4-binding site and the ARS consensus sequence was rendered nonfunctional [HMR-SS(GAL4)a; Foxet al. 1997]. In the absence of functional Gal4-Sir1p, the a genes were expressed from this HMR locus (Table 3).
The detector diploid could mate as an a or α cell or exhibit a nonmating phenotype depending upon the SIR1 genotype (Figure 1B). The detector diploid's ability to exhibit all three mating phenotypes of yeast was exploited to isolate silencer-recognition-defective SIR1 alleles (sir1srd). Phenotypically wild-type Gal4-Sir1p, produced by plasmids that harbored no mutations or phenotypically neutral mutations within the SIR1 portion of GAL4-SIR1, would produce a colony with an a-mating phenotype because both HMR loci would be silenced and, in the absence of any mating-type information, yeast cells would exhibit the default a-mating phenotype (reviewed in Herskowitzet al. 1992). In contrast, a mutation that produced a nonfunctional Gal4-Sir1p would cause a nonmating phenotype because both HMR loci would be expressed and, in the presence of both a- and α-mating-type information, yeast cells would exhibit the nonmating phenotype. However, mutations that produced Gal4-Sir1p molecules with a specific defect in silencer recognition would cause the α-mating phenotype because the HMR-SS(GAL4)a silencer would mediate repression of the a-mating-type genes, but the HMR-SSα silencer would be unable to silence the α-mating-type genes. Thus, α mating in the detector diploid was the phenotype caused by a GAL4-sir1srd mutant allele that was specifically defective in its ability to recognize the silencer. To isolate sir1srd alleles, a plasmid encoding Gal4-Sir1p was digested with restriction enzymes such that a region encoding the Sir1p portion was deleted. This gapped plasmid was cotransformed into the detector diploid together with linear DNA consisting of the SIR1 gene that had been mutagenized during its PCR amplification. The ends of the PCR fragments were homologous with the ends of the gapped plasmid and recombined with the gapped plasmid to produce full-length Gal4-Sir1p (Muhlradet al. 1992). Sixteen plasmid-dependent sir1srd mutants were isolated by screening 24,000 transformants for the ability to mate as α cells. Examples of the different mating phenotypes exhibited by the detector diploid are shown (Figure 1C).
To provide a second criterion for testing candidate mutants, two additional haploid strains were transformed with the candidate mutant plasmids isolated from the detector diploid described above. In the first strain (CFY114), the MAT locus contained α genes and the HMR locus contained a genes under the control of the synthetic HMR silencer. This strain also contained a sir1Δ::LEU2 mutation and had a nonmating phenotype (Figure 1D). A plasmid encoding either SIR1 or GAL4-SIR1 transformed into this strain conferred the α-mating phenotype because both Sir1p and Gal4-Sir1p can promote silencing at HMR (Figure 1D). However, plasmids that encoded mutant versions of Gal4-Sir1p that were defective in silencer recognition would fail to cause this haploid strain to acquire an α-mating phenotype. In contrast, these same plasmids would confer the α-mating phenotype to the second strain (JRY5278) that was MATα and contained the a genes under the control of a synthetic HMR silencer in which the Rap1p-binding site was replaced with a Gal4p-binding site. The mutant plasmids recovered from the detector diploid failed to complement the sir1Δ::LEU2 mutation for silencing HMRa (in CFY114), although some of the weaker mutants were capable of supplying some SIR1 function to this strain. However, all the mutant plasmids conferred the α-mating phenotype to the second strain (JRY5278), indicating that the mutant Gal4-Sir1ps were capable of functioning in silencing per se, although they were not capable of recognizing a silencer. Thus all 16 plasmids encoded a silencer-recognition-defective mutant form of the GAL4-SIR1 fusion.
Ten independent plasmids contained missense mutations in the Sir1 protein coding region: To identify the mutations within the Sir1p coding region responsible for the silencer-recognition defect, the relevant SIR1 portion of the 16 plasmids recovered from the screen was sequenced. Ten of these plasmids, which arose from two independent PCR mutagenesis reactions, contained distinct mutations within the coding region of Sir1p (Table 4). Interestingly, two independent plasmids derived from separate PCR reactions each contained the same nucleotide change responsible for an alanine to threonine amino acid substitution at position 505 of Sir1p (Table 4, mutants 11 and 12). In addition, several independent mutations changed the same amino acids within Sir1p (Table 4). For example, mutants 1 and 3 contained different nucleotide changes but both resulted in substitutions of the tyrosine at codon 489. Similarly, four independent nucleotide changes caused substitutions of alanine at codon 505. Each plasmid contained at least one missense mutation in the SIR1 coding region, several contained nucleotide changes that caused no change at the amino acid level, and no plasmid contained a nonsense mutation. Reengineered plasmids containing any one of the single amino acid changes underlined in Table 4 in an otherwise wild-type GAL4-SIR1 gene conferred a silencer-recognition defect. Thus a missense mutation at any one of the five codon positions 489, 490, 491, 493, or 505 within Sir1p was sufficient to create a silencer-recognition-defective mutant form of GAL4-SIR1.
Mutant Gal4-Sir1p proteins defective in HMR-E silencer recognition were expressed at levels similar to that of wild-type Gal4-Sir1p: One possible explanation for the silencer-recognition defect of the mutant Gal4-Sir1psrds would be that the steady-state level of the mutant forms was lower than that of wild-type Gal4-Sir1p and that the Gal4p-binding domain provided a higher affinity for the HMR-SS(GAL4) silencer than Sir1p itself had for the HMR-SS silencer. Two observations argued against this possibility. First, silencing mediated by Gal4-Sir1p tethered to a silencer via a Gal4p-binding site was actually more sensitive to reductions in Gal4-Sir1p concentration than was silencing by the normal mechanisms (C. Fox, unpublished results). Therefore, mutations that caused a decrease in Gal4-Sir1p levels should have failed to silence via a Gal4p-binding site in the detector diploid and would not be identified in the screen. Second, we directly compared the levels of two different mutant Gal4-Sir1ps with wild-type Gal4-Sir1p (Figure 2A). The levels of wild-type Gal4-Sir1p were similar to the levels of the mutant Gal4-Sir1ps even though wild-type Gal4-Sir1p could silence HMR-SSa but the mutant forms could not (Figure 2 and Table 5). Thus the silencer-recognition defect was not due to decreased levels of the Gal4-Sir1p mutant proteins.
A mutant GAL4-sir1srd allele was defective in silencing natural HMR: The preceding data indicated that the mutant GAL4-SIR1 alleles isolated in the screen described above were not capable of recognizing the synthetic HMR-E silencer. If these mutant alleles were defective in a specific protein-protein contact fundamental to Sir1p function, then they should also cause a silencing defect at HMR controlled by the natural, wild-type HMR-E silencer. A sir1Δ mutation causes only a partial silencing defect at natural HMR (Rine and Herskowitz 1987; Pillus and Rine 1989) that is most reliably detected at the level of a1 mRNA expression from HMR by RNA-blot hybridization (Figure 3). Therefore, the level of a1 mRNA was measured in a MATα sir1Δ HMRa strain that expressed a plasmid copy of a GAL4-sir1srd allele (GAL4-sir1-101) by RNA-blot hybridization (Figure 3).
Either SIR1 or wild-type GAL4-SIR1 expressed from a plasmid completely silenced natural HMRa as measured by the lack of detectable a1 mRNA (Figure 3, compare lane 1 to lanes 2 and 3). In contrast, the GAL4-sir1-101 allele failed to provide sufficient SIR1 function to silence natural HMRa as measured by the expression of a1 mRNA (Figure 3, lane 4). Notably, GAL4-sir1-101 provided partial SIR1 function at natural HMR as measured by a1 mRNA levels that were slightly reduced compared to the plasmid control (Figure 3, compare lanes 1 and 2), consistent with the hypomorphic behavior of the sir1srd alleles at the synthetic HMR-E silencer (Table 6). Regardless, these data indicated that the amino acids in Sir1p required for the Sir1p's recognition of the synthetic silencer were also required for efficient Sir1p function at the natural HMR-E silencer.
The mutations in SIR1 that caused a silencer-recognition defect in the context of the Gal4-Sir1p fusion protein also blocked the function of Sir1p: The silencer-recognition defect was defined in the context of Gal4-Sir1p (Figure 1). To determine if the lesions in Sir1p conferring this defect also caused a silencing defect in the absence of the Gal4p DNA-binding domain, individual mutations responsible for the silencer-recognition defect in Gal4-Sir1p were placed within the chromosomal SIR1 gene and silencing in these strains was measured in quantitative mating experiments (Table 6). The most severe mutants exhibited silencing defects that were similar to, although less severe than, those caused by a deletion of the SIR1 gene, supporting the hypothesis that these mutations caused defects in natural Sir1p function (Table 6). It is worth noting that within the context of chromosomal SIR1, the mutations responsible for a silencer-recognition defect did not reduce silencing as severely as they did within the GAL4-SIR1 context, suggesting that the Gal4p domain perturbed or sensitized Sir1p function somewhat (compare Tables 5 and 6). Nevertheless, the mutations causing the most severe phenotype within the context of Gal4-Sir1p exhibited the most severe silencing defects in the context of chromosomal SIR1. The weaker mutations behaved as hypomorphic sir1 alleles within both contexts. In fact, the weakest allele [sir1-107 (A 505 S)] isolated from the detector diploid actually caused a bimating phenotype in this diploid because the α genes at HMR-SSα were only partially silenced. In the context of chromosomal SIR1, the mutation responsible for this weak silencer-recognition defect caused no observable defect in silencing in a strain that was MATα HMR-SSa. However, this same mutation in a mataΔp HMR-SSα strain caused an α-mating phenotype, indicating that even this extremely weak allele caused a measurable silencing defect in the context of the chromosomal SIR1 gene (C. Fox, unpublished results). These data established that the amino acids identified by the mutations that caused a silencer-recognition defect in the context of Gal4-Sir1p were also required for normal Sir1p function.
The amino acids required for Sir1p's recognition of the HMR-E silencer were also required for Sir1p's interaction with Orc1p: A previous study reported an interaction between the carboxy-terminal portion of Sir1p and the aminoterminal portion of Orc1p in a two-hybrid interaction assay (Triolo and Sternglanz 1996). If this interaction were relevant to Sir1p's recognition of the HMR-E silencer, then the amino acids in Sir1p identified by the mutant Sir1psrd proteins might also be required for Sir1p's interaction with Orc1p. Therefore, we tested whether Orc1p interacted with a mutant Sir1psrd protein in the two-hybrid interaction assay.
A Gal4p DNA-binding domain fused to the full-length Sir1p (Gal4db-Sir1p) interacted with a fusion protein that consisted of the Gal4p activation domain and the aminoterminal portion (amino acids 5-228 or 5-268) of Orc1p (Gal4ad-Orc1p) as measured by growth on selective media of the two-hybrid tester strain (PJ69-4A) expressing both fusion proteins (Figure 4A). The Gal4db-Sir1p fusion did not interact with the Gal4p activation domain itself (Figure 4A). Thus, the two-hybrid assay reported here detected an interaction between a fulllength, functional form of Sir1p and the aminoterminal portion of Orc1p. In striking contrast, a mutant Gal4db-Sir1psrd fusion, encoded by a plasmid copy of GAL4-sir1-101, failed to interact with Gal4ad-Orc1pN (Figure 4A). The tester strain grew well with all relevant plasmids on media that did not demand a two-hybrid interaction (Figure 4B), and the levels of Gal4db-Sir1psrd and wild-type Gal4db-Sir1p were equivalent by protein immunoblot analysis (see Figure 6B). In addition, several other Gal4db-Sir1psrd mutant proteins also failed to interact with Gal4ad-Orc1pN in this assay (K. Gardner, unpublished results). Therefore, the amino acids required for Sir1p's recognition of the HMR-E silencer were also required for Sir1p's physical interaction with Orc1p.
The sir1srd alleles were recessive: The mutant SIR1 alleles described above were defective in Sir1p's recognition of the HMR-E silencer and interaction with Orc1p but not in Sir1p's ability to silence once tethered to the silencer. If Sir1p functioned in silencing by recruiting other silencing proteins to the silencer, one might expect that a sir1srd allele would encode a mutant protein that would still associate with other silencing proteins yet be unable to bring them to the silencer. Thus, it was possible that the Sir1psrd mutant proteins would compete with wild-type Sir1p for association with the other silencing proteins and prevent wild-type Sir1p from functioning in silencing. Thus a mutant sir1srd allele might show a dominant silencing defect. Therefore, the mating phenotypes of diploids homozygous for mataΔp and HMR-SSα and heterozygous sir1Δ/SIR1 or sir1-101/SIR1 were determined (Figure 5). The sir1Δ/SIR1 diploid mated with a strong a mating type, indicating that the single copy of SIR1 in this strain silenced the α genes effectively. The slight amount of α-mating phenotype exhibited in this diploid may be due to limiting amounts of wild-type Sir1p. The sir1-101/SIR1 diploid exhibited a mating phenotype indistinguishable from that of the sir1Δ/SIR1 diploid, indicating that the sir1srd alleles were recessive. In fact, even expressing relatively high amounts of a Gal4-Sir1psrd mutant protein from a multicopy plasmid did not cause an observable defect in silencing in a wild-type SIR1 strain as measured by mating of a MATα HMR-SSa strain (Figure 6A, row 3 and Figure 6B). Even in the mataΔp HMR-SSα strain, which can detect slight reductions in silencing as an increased number of cells with an α-mating phenotype, high levels of the mutant Gal4p-Sir1psrd caused only a slight increase in this phenotype over that caused by wild-type Gal4-Sir1p (Figure 6A, bottom row). The levels of both wild-type and mutant forms of Gal4-Sir1p expressed in strains carrying the multicopy plasmids were significantly higher than necessary for Gal4-Sir1p to function through either an HMR-SS silencer or a HMR-SS(GAL4) silencer (Figure 6A, rows 1 and 2 compared to Figure 1D; Figure 6B). Therefore the Sir1srd mutant proteins did not interfere significantly with wild-type Sir1p's function in silencing.
The work presented here was based on the prediction that Sir1p would have a region that was devoted to recognition of a silencer. This silencer-recognition region would function to bring Sir1p to HMR-E, but not be required to assemble silent chromatin per se. In this article, eight new SIR1 alleles were described, referred to collectively as sir1srd alleles, all of which were defective in silencer-recognition but not in silent chromatin assembly. The alleles were designated sir1-100 through sir1-107 in order of decreasing severity of their silencing defect, and the individual missense mutations responsible for the mutant phenotype clustered within a 17-amino-acid region of Sir1p (Figure 7).
A region of Sir1p dedicated to silencer recognition was also required for interaction with Orc1p: The clustering of missense mutations within a small portion of Sir1p, as well as the number of independent mutations giving rise to changes in the same amino acids, suggested that only a small number of amino acid substitutions were tolerated by the constraints set by the silencer-recognition mutant phenotype. However, it is worth noting that other regions within Sir1p may also be required for silencer recognition. These regions would not have been identified by the screen described here if they also played other roles in Sir1p function, such as recruiting other silencing proteins to HMR.
There are at least two possible explanations for how the region of Sir1p defined by the sir1srd alleles could contribute to Sir1p's recognition of the HMR-E silencer. The simplest view is that the amino acids dedicated to Sir1p's silencer-recognition function may represent sites of direct contacts between Sir1p and a protein, such as ORC, that binds silencer DNA directly. In support of this view, a Sir1p-Orc1p interaction was abolished by a single amino acid substitution in Sir1p that created a mutant Sir1psrd protein incapable of recognizing HMR-E. However, we cannot exclude the possibility that the region of Sir1p identified by the Sir1psrd mutations functioned indirectly, by influencing the proper folding of Sir1p, such that another region of Sir1p was made accessible for the appropriate Orc1p-Sir1p contacts required for Sir1p's recognition of HMR-E. If so, then another region of Sir1p must directly contact Orc1p that was not identified in the silencer-recognition screen described here. Regardless of the exact mechanism, the data presented here argue for an interaction between Sir1p-Orc1p that is dedicated to Sir1p's function in silencer recognition but dispensable for Sir1p's function in the assembly of silent chromatin. The simplest explanation for our data was that Sir1p's recognition of a silencer was mediated by a few specific amino acids in Sir1p that directly contact the Orc1p. The issue of how this Sir1p-Orc1p interaction could be used to uniquely distinguish generic origins of replication from silencers was not addressed by the experiments reported here. Perhaps the region of Orc1p that interacts with Sir1p is accessible only when ORC binds a silencer.
Recessive sir1srd alleles: The sir1srd alleles were recessive, indicating that the inability of one type of Sir1p to recognize a silencer and interact with ORC did not interfere with the ability of wild-type Sir1p to function in silencing. At one level, it was suprising that the GAL4-sir1srd did not interfere significantly with wild-type Sir1p in silencing HMR. After all, these mutant forms of Gal4-Sir1p must still interact with some silencing-specific components because they can silence HMR when tethered to the silencer via a Gal4p-binding site. Why did the mutant forms of Gal4-Sir1p fail to compete with wild-type Sir1p for these interactions? By immunoblot analysis, the Gal4-Sir1p proteins were reasonably abundant and, even when overproduced beyond the level needed for function, did not interfere significantly with wild-type Sir1p. An appealing model involves an order-of-addition explanation. Perhaps Sir1p's ability to interact with other silencing components, such as the other Sir proteins, requires that it first bind the silencer either through interactions with ORC or through direct interactions with silencer DNA containing a Gal4p-binding site when expressed as a Gal4-Sir1p fusion. The synthetic silencer at which Gal4-Sir1p was functional lacked an ORC or Rap1p-binding site but contained an Abf1p-binding site and the HMR-I element. Perhaps the juxtaposition of Gal4-Sir1p and the Abf1p, for example, is required to form a surface that can efficiently recruit the other Sir proteins. Thus, in the absence of a region required for silencer recognition, the mutant forms of Sir1p would be recessive to wild type. Implicit in this model is the notion that Sir1p, on its own, is insufficient to recruit other silencing proteins to a locus. Indeed, a LexA-Sir1p fusion tethered to a LexA-binding site on a plasmid upstream of a reporter gene was unable to silence the reporter gene (S. Okamura, personal communication), although this same fusion could silence HMR in the presence of a LexA-binding site at the silencer (C. Fox, unpublished observations).
The amino acids in Sir1p identified by the silencer-recognition defect of the mutant Sir1psrd proteins play an important role in Sir1p's ability to recognize a silencer and interact with Orc1p without dramatically affecting Sir1p's other role(s) in assembling the silent state. Further systematic analysis of such precisely defined Sir1p defects should help unravel the multitude of specific protein-protein contacts that regulate the assembly of multiprotein complexes at defined chromosomal positions. Moreover, the conceptual approach used to identify the sir1srd alleles should be generally applicable to dissecting individual protein-protein interactions that contribute to the stability and function of other multiprotein complexes.
We thank Rolf Sternglanz for providing the ORC1 plasmids used for the experiment reported in Figure 4 and Melissa R. Mielke for technical assistance with the experiment reported in Figure 3. We are grateful to Elizabeth Craig for comments on the manuscript. C.A.F. thanks Michael Sheets for his patience and support and Andrew Dillin and Sara Okamura for listening to convoluted arguments. This work was supported by postdoctoral fellowships from the Leukemia Society of America and the California Division of the American Cancer Society (to C.A.F.) and by grants from the National Institutes of Health (GM31105 to J.R. and GM56890 to C.A.F.). K.A.G. is supported by a Biotechnology Training Grant to the University of Wisconsin-Madison (NIH 5T32 GM08349). C.A.F. is a recipient of a Burroughs Wellcome Fund Career Award in the Biomedical Sciences. Additional support was provided by a National Institute of Environmental Health Sciences Mutagenesis Center grant.
Communicating editor: F. Winston
- Received April 8, 1998.
- Accepted September 23, 1998.
- Copyright © 1999 by the Genetics Society of America