In Saccharomyces cerevisiae, genes located at the telomeres and the HM loci are subject to transcriptional silencing. Here, we report results of screening a Gal4 DNA-binding domain hybrid library for proteins that cause silencing when targeted to a silencer-defective HMR locus.
TRANSCRIPTIONAL silencing in Saccharomyces cerevisiae occurs through a specialized chromatin structure at the telomeres and the HM loci, HML and HMR (reviewed in Ruscheet al. 2003). The HM loci consist of mating-type genes that are kept transcriptionally silent by cis-acting elements, termed the E and I silencers (Brandet al. 1985; Mahoneyet al. 1991). The HMR-E silencer is composed of A, E, and B sites that are bound by the origin recognition complex, Rap1, and Abf1, respectively (Shore and Nasmyth 1987; Bellet al. 1993). These factors, in turn, recruit the silent information regulator proteins, Sir1–4, which leads to silencing of nearby genes (Ruscheet al. 2003). If two or more silencer elements at HMR-E are deleted and replaced by Gal4-binding sites, silencing is lost. Silencing can be restored by expression of Gal4 DNA-binding domain (GBD) hybrids fusing GBD to known silencing proteins (Chienet al. 1993; Buck and Shore 1995; Lustiget al. 1996; Triolo and Sternglanz 1996); we refer to this as targeted silencing.
AGBD library was screened to identify proteins capable of targeted silencing at HMR. Several known silencing proteins and Sir-binding proteins were identified. Novel proteins were also identified and named Esc because they establish silent chromatin when targeted to DNA. A few other previously characterized proteins, with no known role in silencing, were also found to give SIRdependent targeted silencing.
Targeted silencing screens: Strains with either two or three silencer elements at HMR-E deleted and replaced by binding sites for Gal4 and containing an hmr::URA3 reporter (designated Aeb::G and aeb::G) were used for separate screens (Figure 1). The proteins identified in these screens are listed (Table 1) and targeted silencing by some of them is shown (Figure 2).
In the screen with the Aeb::G silencer, the known silencing factor Sir1 and Sir-interacting proteins Rad7, Rif1, Ris1, and Ubp10 were identified (Table 1; Hardyet al. 1992; Paetkauet al. 1994; Zhang and Buchman 1997; Singeret al. 1998). Presumably, these gave targeted silencing by binding a Sir protein directly, leading to the recruitment of the Sir protein complex. We also identified Vac8, a component of junctions formed between the nuclear envelope and the vacuole (Panet al. 2000). Perhaps Vac8 brings the derepressed HMR locus to the nuclear periphery, where silencing proteins are concentrated (Palladinoet al. 1993; Gottaet al. 1996; Andruliset al. 1998). Two novel proteins with silencing activity were identified and named Esc1 and Esc2. Using two-hybrid assays, we showed that Esc1 interacts with Sir4 (Andruliset al. 2002) and another group showed that Esc2 interacts with Sir2 (Cuperus and Shore 2002); this is probably why they gave targeted silencing. A serine-rich peptide was also found; this peptide is not derived from an open reading frame (ORF). Several proteins predicted to have at least one transmembrane domain (Snc1, Nyv1, Src1, and Gtt3) were also identified (data not shown). We have previously shown that membrane proteins can give targeted silencing, presumably because they bring the HMR locus to the nuclear periphery where there is a higher concentration of Sir proteins (Andruliset al. 1998).
In the screen with the aeb::G silencer, many proteins were found (Table 1). Sir1, Rif1, Esc2, and Ris1, identified in the first screen, were isolated again. The Sir1-binding protein, Orc1, and the Sir2 homolog, Hst1, were identified. The Sir2-interacting Net1 protein was also found, as were Hir1, a repressor of histone gene transcription (Sherwoodet al. 1993), and Rpb4, an RNA polymerase II subunit. A previously uncharacterized protein was also identified and named Esc4. The serine-rich peptide found in the first screen was identified twice in this screen. Finally, a GBD hybrid to a 17-amino-acid peptide with the sequence IFLRLVKRPWP GQNFAP gave silencing. This peptide, like the serinerich peptide, is not derived from an ORF. It is possible that these peptides may mimic a binding site for a silencing protein.
A targeted silencing screen also was undertaken with a strain that had both HMR silencers deleted and replaced with GBD-binding sites. Two proteins, Esc1 and Rif2, a Rap1-interacting factor (Wotton and Shore 1997), were identified (data not shown).
SIR-dependent targeted silencing: As mentioned, many of the proteins identified in these screens bind to Sir proteins. To test the SIR dependence of targeted silencing by the GBD hybrids, each of the hybrid proteins was introduced into a targeted silencing reporter strain deleted for the SIR2, SIR3, or SIR4 gene. None of the hybrid proteins gave targeted silencing when tested in these sir mutant strains (data not shown). Thus, it is very likely that the silencing observed was due to the endogenous silencing machinery. It is not clear why Hir1 or Rpb4 gave SIR-dependent targeted silencing. Perhaps Hir1 binds appropriately modified histones and thus seeds the formation of the Sir2–4 complex. At first sight, SIR-dependent targeted silencing by Hst1 does not seem surprising because Hst1 is a fairly close homolog of Sir2 and overexpression of Hst1 can partially suppress the silencing defect of a sir2 mutant at HMR (Brachmannet al. 1995). However, a novel form of silencing mediated by the SUM1-1 protein requires Hst1 but is independent of Sir2, Sir3, and Sir4 (Suttonet al. 2001).
Characterization of Esc2: We decided to focus our attention on one of the three previously uncharacterized proteins identified in these screens, Esc2. Esc1 has already been described (Andruliset al. 2002), and Esc4 will be described elsewhere. Targeted silencing by Esc2 was most efficient in the presence of one remaining HMR-E silencer element, but also was seen when all three were absent (Figure 3A). Silencing by Esc2 required the Gal4-binding sites and endogenous Sir2, Sir3, and Sir4 proteins, but was only partially dependent on Sir1 (Figure 3A). The C-terminal region of Esc2 protein shows a significant similarity to the small ubiquitin-like protein, SUMO (58% identical or similar residues for amino acids 393–452 of Esc2). To address the role of the Esc2 C-terminal SUMO-like domain in silencing, we constructed several GBD hybrids containing Esc2 fragments and tested them for targeted silencing. Removal of the Esc2 C-terminal SUMO-like domain did not affect the ability of Esc2 to give targeted silencing (Figure 3B). Also, this domain alone was not sufficient for silencing the reporter gene. These deletion studies suggest that the targeted silencing domain of Esc2 does not include the SUMO domain and lies between amino acid residues 115 and 260 (Figure 3B). While this article was in preparation, two other groups identified Esc2 by use of genetic screens for silencing proteins (Dhillon and Kamakaka 2000; Cuperus and Shore 2002).
Future directions: These targeted silencing screens allowed us to identify proteins implicated in the formation of silent chromatin. The screens have not been saturated, as we have identified only one clone for most of the proteins. It is likely that other factors that can establish silencing at HMR remain to be identified. The silencing peptides discovered here may provide insight into critical domains to which silencing proteins bind.
This one-hybrid silencing approach can be applied to identify factors that are locus specific by performing targeted silencing screens at derepressed HML or ribosomal DNA loci with Gal4-binding sites. Future studies using this system should enhance our understanding of the general mechanisms of transcriptional silencing.
C.E. thanks Stan Fields for support and advice. This work was supported by National Institutes of Health grant GM-28220.
Communicating editor: F. Winston
- Received August 7, 2003.
- Accepted October 7, 2003.
- Copyright © 2004 by the Genetics Society of America