|Home | About | Journals | Submit | Contact Us | Français|
Dot1 methylates histone H3 lysine 79 (H3K79) on the nucleosome core and is involved in Sir protein-mediated silencing. Previous studies suggested that H3K79 methylation within euchromatin prevents nonspecific binding of the Sir proteins, which in turn facilitates binding of the Sir proteins in unmethylated silent chromatin. However, the mechanism by which the Sir protein binding is influenced by this modification is unclear. We performed genome-wide synthetic genetic array (SGA) analysis and identified interactions of DOT1 with SIR1 and POL32. The synthetic growth defects found by SGA analysis were attributed to the loss of mating type identity caused by a synthetic silencing defect. By using epistasis analysis, DOT1, SIR1, and POL32 could be placed in different pathways of silencing. Dot1 shared its silencing phenotypes with the NatA N-terminal acetyltransferase complex and the conserved N-terminal bromo adjacent homology (BAH) domain of Sir3 (a substrate of NatA). We classified all of these as affecting a common silencing process, and we show that mutations in this process lead to nonspecific binding of Sir3 to chromatin. Our results suggest that the BAH domain of Sir3 binds to histone H3K79 and that acetylation of the BAH domain is required for the binding specificity of Sir3 for nucleosomes unmethylated at H3K79.
Gene silencing in Saccharomyces cerevisiae at telomeres and the silent mating type loci is mediated by Sir proteins, which are recruited to DNA elements called silencers by sequence-specific DNA binding proteins (16, 20, 61). Upon the recruitment of Sir2 and Sir4 to silencers, Sir3 can bind, and the silent chromatin structure can subsequently spread in cis by interactions with neighboring nucleosomes (42). Silent chromatin in yeast is characterized by the absence of histone modifications, suggesting that the Sir complex preferentially binds to unmodified histones (16, 61). The NAD-dependent histone deacetylase activity of Sir2 is required for the spread and formation of a repressive Sir2-Sir3-Sir4 (Sir2/3/4) chromatin structure (35, 42, 74), and the binding of Sir3 to histone peptides in vitro has been shown to be negatively affected by the methylation and acetylation of the tails of histone H3 and H4 (8, 42, 63). Binding of Sir3 to histone tails is mediated by the C terminus of Sir3 (20). However, full-length Sir3 can bind to nucleosomes which lack histone tails, suggesting that Sir3 also interacts with other features of the nucleosome (23).
In addition to modifications on the histone tails, silencing is positively affected by the methylation of lysine 79 of histone H3 (H3K79), a residue on the nucleosome core (15, 39, 47, 49, 76). The responsible methyltransferase Dot1 methylates ~90% of histone H3 and does so predominantly in euchromatin (39, 47, 49, 76). In the absence of Dot1, binding of Sir2 and Sir3 at silent chromatin is reduced, and Sir3 becomes redistributed (47, 49, 62, 76). We previously proposed that the methylated H3K79 (H3K79me) in euchromatin prevents nonspecific binding of Sir proteins to euchromatin and thereby enhances targeting of the limiting pool of Sir proteins to silent regions (76). However, how H3K79me affects the Sir protein interactions with chromatin remains unclear. In yeast, Dot1 has also been shown to be required for the efficient activation of DNA damage checkpoints (5, 25, 80), for the activation of the pachytene checkpoint in meiosis (62), and for resistance to radiation (18). In human cells, H3K79 methylation by Dot1 has been implicated in Ras-induced gene silencing (21), as well as leukemic transformation by activation of the Hox genes (51, 52). The mechanism by which H3K79me affects these processes has not been established.
To investigate how the abundant H3K79me affects chromatin structure and function, we employed systematic genome-wide screens for genetic interactions of DOT1, using synthetic genetic array (SGA) analysis (3, 54). Our results have allowed us to develop mechanistic insights into how Dot1 functions in silencing and have uncovered an unappreciated aspect of SGA analysis for studying silencing.
Yeast strains used in this study are listed in Table Table1.1. Yeast media have been described previously (76). Gene deletions were made by PCR-mediated gene replacement (6, 43). The dot1-G401V mutation was introduced by the integration of a genomic copy of dot1-G401V into strains lacking the endogenous DOT1 gene coding sequence by transformation with the plasmid pTW043 digested with MluI. Strains were verified by PCR and immunoblotting analysis. Silencing assays were performed by spot tests using media containing 1 g/liter 5-fluoroorotic acid (5-FOA), and media have been described previously (76). For spot tests at high temperature, all strains were pregrown and subsequently tested at 37°C. Mating assays were performed as previously described (77), with tester strains or strains PT1 and PT2. Mating reactions of the nat1Δ, sir1Δ, and dot1Δ strains were performed with the tester strain BY4709 and plated on yeast complete medium lacking uracil (YC-Ura) (MATa haploid plus diploid) and YC-Ura-Trp (diploid) media. Mating reactions of the sir3Δ strains carrying SIR3 centromere (CEN) plasmids or strains carrying integrated SIR3 alleles were performed with the tester strain BY4734 and plated on YC-Trp (MATa haploids plus diploids) and YC-Trp-Lys (diploids). Plasmid pTW043 was made in two steps. First, the G401V mutation was introduced into pRS315-DOT1 (76) by a three-step PCR protocol. The 5.5-kb BamHI-ClaI DOT1 genomic insert was transferred to pRS306 (6). The single-copy plasmid pHR62-16 (pRS314-SIR3; a gift from H. Renauld) was obtained by cloning a 3.7-kb HpaI fragment of YEp13-SIR3 (36) into the SmaI site of pRS314 (6). The A2T and A2G mutations of SIR3 were introduced by a three-step PCR and verified by sequencing to generate pTW070 and pTW071, respectively. The integration plasmids pFvL277 (pRS304-SIR3N) and pFvL278 (pRS304-SIR3N-A2G) were made by cloning a 530-bp genomic region of SIR3 (from 374 bp upstream to 156 bp downstream of the start of the open reading frame) into pRS304 and digesting them with BclI to integrate the plasmids to truncate endogenous SIR3 and express one copy of the wild-type (WT) SIR3 or sir3-A2G gene. The integration plasmids pFvL279 (pRS304-SIR3) and pFvL280 (pRS304-sir3-A2G) were made by cloning a 3.7-kb genomic region of SIR3 (from 374 bp upstream to 399 bp downstream of the open reading frame) into pRS304 and digesting them with HpaI to integrate the plasmids into the SIR3 promoter region to reintroduce the WT SIR3 or the sir3-A2G mutation into the sir3Δ strains.
Synthetic lethal analysis by SGA was performed twice with a dot1Δ strain (NKI3001) derived from strain Y3656 (75), either manually or by using a BioMek robot (Beckman Coulter) and once with a catalytically inactive dot1-G401V (strain NKI2053) mutant derived from strain Y7092 (75), using a Hamilton ML STAR unit (Hamilton Robotics). The screens were performed as described previously (75), with the following modifications: all media contained 10 mg/liter tetracycline; diploids were grown on presporulation GNA medium (24) prior to being plated on sporulation medium (10 g/liter potassium acetate, 1 g/liter yeast extract, 0.5 g/liter glucose, 14 mg/liter uracil, 14 mg/liter histidine, 71 mg/liter leucine); and haploid selection medium was based on YC medium (77). Positive hits from the first rounds were retested by repeating the SGA protocol or by tetrad analysis. For random spore analysis, cells were plated on haploid selection medium, which contained canavanine and thialysine (S-aminoethyl-cysteine), resulting in the survival of can1Δ lyp1Δ haploids and killing all other haploids and the CAN1/can1Δ LYP1/lyp1Δ heterozygous diploids (75). To specifically select for MATa haploids, histidine was omitted from the medium. CloNat (nourseothricin) was obtained from Werner Bioagents (Germany).
Flow cytometry was performed using SYTOX Green as described previously (28). Sir3 mouse monoclonal antibodies, immunoblotting analysis, and chromatin immunoprecipitation (ChIP) protocols were as described previously (76). The Pgk1 antibody was from Molecular Probes (clone 22C5/A-6457). The primers used to amplify Sir3-bound DNA fragments were ACT1f (CCAATTGCTCGAGAGATTTC), ACT1r (CATGATACCTTGGTGTCTTG), HMLf (TACACTCATATGGCTATACC), HMLr (CTATGCGGGCTTGAAAATGAACAG), TEL-VIRf (CAGGCAGTCCTTTCTATTTC), TEL-VIRr (GCTTGTTAACTCTCCGACAG), HMRf (GAGAATAAGCGCAGGTACTCC), and HMRr (TCTTGAGCGGTGAGCCTCTG). The amplified DNA fragments were separated by 2% agarose gel electrophoresis on gels stained with ethidium bromide and imaged using a GeneFlash unit (Syngene). Data were quantified using TINA software version 2.09 (Raytest). Coimmunoprecipitation of histone H3 with Sir3 was examined by elution of the proteins bound to beads and by reversal of the cross-links by boiling for 30 min in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer, followed by immunoblotting analysis using a polyclonal antibody against the C-terminal peptide QKKDIKLARRLRGER of yeast H3 (F. Frederiks and F. van Leeuwen, unpublished data). For cell fractionation, a whole-cell ChIP extract was split into two fractions. One fraction was sonicated to shear and solubilize chromatin and spun for 5 min at 20,000 × g at 4°C. The supernatant was spun again and then collected as the whole-cell extract (WCE). The other fraction was first spun for 5 min at 20,000 × g at 4°C. The supernatant was spun again and then collected as the soluble fraction. The chromatin pellet was washed three times and finally resuspended in the same volume as the soluble supernatant. This pellet sample was sonicated to shear and solubilize chromatin and spun for 1 min at 20,000 × g at 4°C. The supernatant of this sample was collected as the chromatin fraction. Prior to loading samples on gels, all samples were boiled for 30 min in SDS-PAGE sample buffer to reverse the cross-links.
Quantitative real-time reverse transcription-PCR (qRT-PCR) was performed with total RNA isolated from log-phase cultures, using an RNeasy kit (catalog no. 74104; Qiagen). RNA samples were treated with RNase-free DNase (catalog no. 79254; Qiagen), and cDNA was made by using SuperScript II reverse transcriptase (catalog no. 18064; Invitrogen). Real-time qPCR was performed on an Applied Biosystems 7500 Fast RT-PCR system using SYBR Green-based detection, with the primers ACT1_Qf (TCGTTCCAATTTACGCTGGTT), ACT1_Qr (CGGCCAAATCGATTCTCAA), HMLalpha1_Qf (CTTGTCTTCTCTGCTCGCTGAA), and HMLalpha1_Qr (TCCCATATTCCGTGCTGCAT).
SGA analysis of yeast has been used extensively for synthetic lethal screening and has revealed many hidden relationships or functions of nonessential genes (3, 54). SGA makes use of the collection of nearly all ~5,000 nonessential haploid deletion mutants and involves the mating of a query strain with the entire collection of deletion mutants, followed by meiotic recombination, sporulation, and selection of haploid cells. In the final steps, the growth of single knockouts is compared to the growth of double knockouts to score for phenotypic enhancement or suppression (3, 54). If two genes interact to modulate an essential phenotype, the pair-wise combination of deletion mutants may result in a growth defect or lethality. Although a genetic interaction can reveal information about redundant functions or physical interactions, the primary cause of the phenotype often remains unknown.
To identify genetic interactions of DOT1, SGA analysis was performed with a DOT1 knockout strain (dot1Δ) and repeated with a strain carrying a point mutation in DOT1 (dot1-G401V; dot1V) which renders the protein catalytically inactive (data not shown). In our studies, we have not detected a functional difference between dot1Δ and dot1V. Out of three genome-wide screens, two reproducible interactions were found. The deletion or mutation of DOT1 appeared to be synthetically lethal in combination with the deletion of the genes encoding Sir1, which is bound to silencers and acts by recruiting the Sir2/3/4 complex (4, 19, 60) or Pol32, a nonessential subunit of DNA polymerase δ (Fig. (Fig.1A1A and data not shown). During the course of our studies, genetic interactions between DOT1 and SIR1 and between DOT1 and POL32 were reported, confirming our screen results (9, 57). A closer examination of the growth phenotype by tetrad analysis showed that the dot1V pol32Δ double mutants had a minor growth defect compared to the WT strain or either of the single mutants, while a dot1V sir1Δ strain showed no growth defect (Fig. (Fig.1B).1B). Since Sir1 and Dot1 are both known to be involved in silencing, we reasoned that the “lethality” of the double mutant might have been caused by a silencing defect that uncovers an inherent aspect of the SGA screening process.
Haploid yeasts exist in two mating types, MATa or MATα. The mating type information is encoded by the MAT locus, whereas HMRa and HMLα represent silent copies of mating type information that the cell can use to switch mating types by recombination (16, 61). The final steps of SGA analysis involve the selection for MATa haploid cells by growth on media lacking histidine to select for cells that express the HIS3 gene driven by the MATa-specific MFA1 promoter or the Schizosaccharomyces pombe HIS5 gene driven by the MATa-specific STE2 promoter (3, 75). In MATa cells, the loss of silencing of HMLα results in the presence of a and α information, creating a pseudodiploid cell in which all MATa-specific gene expression is eliminated (61) (Fig. (Fig.2A).2A). Therefore, while silencing is not essential for normal growth, the loss of silencing of HMLα during SGA analysis is expected to result in a lack of growth on MATa-selective media. To determine whether the loss of silencing could explain the observed genetic interactions, spores from the DOT1/dot1V SIR1/sir1Δ heterozygous diploids obtained by SGA were analyzed by random spore analysis. When cells were plated on media selecting for MATa haploids, the dot1V sir1Δ double mutants did not grow, whereas the WT cells and the sir1Δ and dot1V single mutants grew normally (Fig. (Fig.2B2B and data not shown). However, when cells were plated on haploid selection media without selection for histidine prototrophy (allowing for the growth of all haploid cells), the dot1V sir1Δ mutants grew normally (Fig. (Fig.2B).2B). These results showed that the dot1V sir1Δ mutants were viable but that no dot1V sir1Δ double mutants were present that behaved like MATa cells, suggesting that the MATa dot1V sir1Δ double mutants had lost histidine prototrophy because of a mating type defect. To test this suggestion directly, the mating efficiency of spores derived from tetrad analysis was examined. Whereas single mutants of dot1V and sir1Δ showed efficient mating, MATa double mutants of these alleles did not mate, confirming that the cells had lost their MATa mating type identity (Fig. (Fig.2C).2C). Deletion of SIR1 also aggravated the partial HMRa silencing defects of the DOT1 mutants (data not shown) but not enough to cause the complete derepression and loss of mating type identity of MATα cells (Fig. (Fig.2C2C).
To verify that the mating defect of the MATa dot1V sir1Δ strain was caused by the loss of HMLα silencing, DOT1 and SIR1 were knocked out in a strain in which a URA3 reporter gene was integrated at the HMLα locus (Fig. (Fig.2D).2D). Growth assays using this reporter gene are more sensitive than the mating assays or the HIS3 reporter used in SGA (77). Whereas the dot1Δ and sir1Δ single mutants showed a minor decrease in silencing, the double mutant showed a complete loss of silencing of the URA3 reporter gene at HMLα (Fig. (Fig.2D).2D). These results confirmed that the synthetic “lethality” of dot1Δ and sir1Δ in the SGA screens was caused by a genetic interaction that caused the loss of HMLα silencing.
We next investigated the synthetic fitness phenotype of the dot1V pol32Δ mutants (Fig. 1A and B). While the colonies from the dot1V pol32Δ spores after germination of the diploid used in the SGA analysis were somewhat smaller than those of the wild type or strains with either of the mutant alleles alone (Fig. (Fig.1B),1B), in subsequent experiments, the growth rates of these and other independent pol32Δ and dot1Δ/V pol32Δ strains were indistinguishable (see below). In addition, analysis of DNA content by fluorescence-activated cell sorting showed that the pol32Δ mutants accumulate in G2/M phase (31) and that the deletion of DOT1 did not enhance or alter this cell cycle defect (Fig. (Fig.3A3A).
To determine whether synthetic sickness of the pol32Δ dot1V mutants in the SGA screens was enhanced by silencing defects, we performed a random spore plating assay. When random spores from SGA analysis were plated on haploid selection media, the growth defect of the dot1V pol32Δ double mutants was more apparent on MATa selection media than on media lacking selection for the mating type (Fig. (Fig.3B).3B). These results suggested that, similar to the dot1V sir1Δ mutants, the dot1V pol32Δ mutants might have a defect in the silencing of HMLα, although no mating defect could be detected by quantitative mating (data not shown).
Analyses of DOT1 and POL32 deletions in strains carrying a URA3 silencing reporter unambiguously identified a role for POL32 in gene silencing. First, the silencing of URA3 at HMLα was reduced in the pol32Δ strain (Fig. (Fig.3C).3C). When it was combined with the deletion of SIR1 or DOT1, the silencing defect of the pol32Δ knockouts was more severe than expected from the phenotypes of either of the single mutants, suggesting that POL32 acts in a pathway that does not overlap with that involving DOT1 or SIR1. Second, silencing of URA3 integrated at telomere VII-L was greatly reduced in the pol32Δ strain, similar to that in the dot1Δ strain (Fig. (Fig.3D).3D). High temperature has been shown to result in stronger silencing in yeast by unknown mechanisms (2). When telomeric silencing assays were performed at 37°C, the silencing defect of the pol32Δ and dot1Δ single mutants was partially suppressed. In contrast, a high temperature did not restore silencing of the dot1Δ pol32Δ double mutant, confirming that DOT1 and POL32 play nonredundant roles in silencing. Although Sir1 affects silencing at some native telomeres (59), Sir1 is not involved in the recruitment of Sir2/3/4 at truncated telomeres containing reporter genes (16, 20, 61) and did not affect the silencing of the telomeric reporter gene used here (data not shown). Synthetic genetic interactions frequently bridge parallel pathways (82). Therefore, our results suggest that POL32, SIR1, and DOT1 affect silencing via three distinct pathways.
Large-scale genetic interaction screens have helped to identify genes that act in the same process or pathway. Part of the reason for grouping genes together in the same pathway is based on the fact that they share genetic interactions with a common set of other genes (9, 57, 82). Because of our interest in learning more about how DOT1 functions in silencing, we took advantage of this idea and looked for gene mutations that had a synthetic silencing defect when combined with a sir1Δ mutation, just as DOT1 did. We first searched in the literature for mutants that shared with DOT1 a dependence on SIR1 for silencing of HMLα. We then selected mutants that showed additional overlap with dot1Δ in other silencing functions, i.e., (i) a severe reduction in telomeric silencing (67, 76) and (ii) no dependence on SIR1 for HMRa silencing (Fig. (Fig.2C).2C). Two classes of mutants fulfilled these criteria. A telomeric silencing defect and a strong dependence on SIR1 for HMLα (but not for HMRa) silencing have been reported for deletions of NAT1 and ARD1 (71) and for mutations of the N terminus of Sir3 (Sir3-A2G and Sir3-A2T) (22, 70, 78). Nat1 and Ard1 are components of the N-terminal acetyltransferase complex NatA, which acetylates the N-terminal amino acid of a range of proteins, including Orc1 and Sir3 (22, 58, 78). In the sir3-A2G and sir-A2T mutants, the N-terminal alanine of the mature Sir3 protein has been replaced by glycine (A2G) or threonine (A2T) (22, 70, 78). It has been suggested that one way in which NatA affects silencing is by acetylating the N-terminal alanine residue of the mature Sir3 protein (78), which would place NAT1/ARD1 and the N terminus of Sir3 in the same pathway or process.
If genes act in the same pathway, then when null mutations in these genes are combined, it is unlikely that they will have a synthetic interaction (3). To determine whether Dot1, the Sir3 N terminus, and Nat1/Ard1 effect a common silencing process, we examined genetic interactions between these genes with respect to silencing of HMLα, by performing quantitative mating assays of MATa strains. As expected, the dot1Δ and nat1Δ strains showed strong mating defects in combination with sir1Δ. However, mating efficiency of the dot1Δ nat1Δ mutants was no worse than for the single nat1Δ mutant (Fig. (Fig.4A).4A). In addition, we found that the deletion of DOT1 did not affect growth of the nat1Δ or ard1Δ strain on MATa haploid selection media in SGA studies, confirming the fact that HMLα silencing was not aggravated in the double mutants (data not shown).
To study the N terminus of Sir3, the A2G and A2T mutations were introduced into the SIR3 gene on a CEN plasmid and transformed into sir3Δ strains. Plasmid-borne Sir3 and Sir3-A2G were expressed at similar levels as endogenous Sir3 (see below). A WT SIR3 plasmid complemented the loss of mating observed for the sir3Δ and sir3Δ sir1Δ strains, indicating that HMLα silencing was restored (Fig. (Fig.4B).4B). As expected from previous studies (22, 70, 78), the Sir3-A2G mutant could restore mating in the sir3Δ strain but failed to mate in the absence of SIR1 (Fig. (Fig.4B).4B). However, mating of the Sir3-A2G mutant was not affected by deletion of DOT1. The same results were observed with the Sir-A2T mutant (data not shown). Previous studies have shown that a nat1Δ or an ard1Δ mutation combined with sir-A2 mutations have no mating defect (70). However, we could not verify this in our strains because the WT SIR3 plasmid failed to complement the mating defect in the nat1Δ sir3Δ strain, while the nat1Δ single mutants still showed substantial mating (data not shown). We do not know the cause of this defect, but we note that specific silencing defects of plasmid-borne SIR3 have also been reported by others (14). To examine the relationship between Nat1 and the Sir3 N terminus, we constructed SIR3 integration vectors (SIR3i) to replace the endogenous chromosomal copy of SIR3 by WT SIR3 or sir3-A2G. In this context, the nat1Δ SIR3i strains showed mating efficiencies similar to that of the original nat1Δ SIR3 strain (compare Fig. 4A and C). Interestingly, the nat1Δ sir3-A2Gi strains mated as efficiently as the nat1Δ SIR3i strains, while the sir3-A2Gi allele in combination with sir1Δ abolished mating (Fig. (Fig.4C4C).
To confirm that the observed mating defects were caused by the loss of silencing of the HMLα locus, the expression of the α1 gene was determined by qRT-PCR. Although the loss of HMLα silencing leads to the expression of the α1 and α2 genes, the α2 protein together with the a1 protein present in the MATa cells form a heterodimeric transcriptional repressor that represses the expression of α1 and α2 (17, 29). Despite this negative feedback loop, desilencing of HMLα can usually be detected by low levels of α1/α2 mRNA (e.g., see reference 45). Indeed, the WT strain showed low levels of α1 mRNA, whereas strains with no mating or strongly reduced mating (e.g., the sir3Δ, sir1Δ dot1, sir1Δ nat1Δ, and sir1Δ sir3-A2G strains) showed higher levels of α1 mRNA (Fig. (Fig.4D).4D). As expected, the dot1Δ and sir3-A2G single mutants and the dot1Δ sir3-A2G double mutant showed low α1 expression levels, similar to that of the WT. The nat1Δ single mutant, which has reduced mating efficiency (Fig. (Fig.4A),4A), showed intermediate α1 mRNA levels, which were not affected by the additional deletion of DOT1 or the mutation of the Sir3 N terminus (Fig. (Fig.4D).4D). Expression of α1 in the nat1Δ single and double mutant strains was more variable (between duplicates and between independent clones) than in other strains. We do not know the cause of the variability but expect that it is related to the poor growth that we observed for strains lacking NAT1. We conclude that, overall, the mating defects shown in Fig. 4A to C correlate with desilencing of HMLα. The similarity in silencing phenotypes and the lack of genetic interactions between mutants of Dot1, NatA, and the N terminus of Sir3 strongly suggest that the three components affect a common silencing process.
Dot1 methylates histone H3K79 in euchromatin, suggesting that H3K79me affects silencing in trans, possibly by blocking nonspecific binding of Sir proteins to euchromatin, which in turn promotes binding of Sir proteins to silent chromatin (47, 49, 76). Sir1 is bound to silencers and acts by recruiting the Sir2/3/4 complex in cis (4, 19, 60), which explains why DOT1 and SIR1 seem to act in parallel silencing pathways (Fig. (Fig.2).2). How the Sir complex is affected by histone H3K79me is unknown. However, because our genetic studies suggest that Dot1 and the Sir3 N terminus affect the same silencing process, we speculated that the effect of histone H3K79me on silencing might be mediated by a physical interaction between histone H3K79 on the nucleosome core and the Sir3 N terminus.
One prediction of this model is that the Sir3 N terminus shares functions with Dot1 in providing specificity of Sir protein targeting. To test this model, we examined binding of Sir3 and Sir3-A2G to chromatin in vivo. We first analyzed Sir3 expression levels. It has previously been shown that the deletion of NAT1 or the mutation of the Sir3 N terminus does not affect Sir3 protein levels (22). However, in a recent study, the deletion of ARD1 was shown to lead to reduced levels of tagged Sir3 (53). By using immunoblotting analysis, we found that in strains harboring a fixed single integrated copy of SIR3 or sir3-A2G, the Sir3-A2G protein was expressed at lower levels than the WT Sir3 (Fig. (Fig.5A,5A, left panel). In nat1Δ strains, the expression of endogenous Sir3 or that of reintegrated Sir3 (or Sir3-A2G) was even more reduced and undetectable with our antibody (Fig. (Fig.5A).5A). These results suggest that the acetylation of the N-terminal alanine of Sir3 enhances Sir3 stability, but we do not know the mechanism of this stabilization. Interestingly, plasmid-encoded Sir3 and Sir3-A2G were expressed at similar levels (Fig. (Fig.5A,5A, right panel). Thus, the silencing defects of Sir3-A2G were not caused by reduced Sir3 levels. By quantitative PCR we observed that the sir3-A2G CEN plasmid was present at a higher copy number than the WT SIR3 CEN plasmid. Although we do not understand the mechanism by which Sir3 affects plasmid copy number, we suggest that the higher DNA copy number of the sir3-A2G plasmid compensated for a reduction in Sir3-A2G protein levels. We investigated the binding of Sir3 to chromatin in strains expressing plasmid-encoded Sir3 and Sir3-A2G to ensure equal expression levels of the two proteins.
The binding of Sir3 to chromatin was analyzed by ChIP, followed by PCR and quantitation of the PCR products. In the strain expressing Sir3-A2G, Sir3 binding to telomeres was reduced compared to that of WT Sir3 (Fig. (Fig.5B),5B), which is consistent with the observed silencing defect (data not shown). Binding to HMLα was unaffected, which is consistent with the efficient mating of this mutant (Fig. (Fig.4B).4B). In the absence of histone H3K79me (dot1Δ), Sir3-protein binding to telomeres was also reduced, although not as much as in the Sir3-A2G strain (47, 76). Deletion of DOT1 did not enhance any of the effects of the Sir3-A2G mutation on Sir3 binding, offering further evidence that they function in the same process. In accordance with the absence of mating problems of MATα cells, the sir3-A2G mutation or the deletion of DOT1 did not affect binding of Sir3 to the silenced HMRa locus (Fig. (Fig.5B5B).
It has previously been shown that in strains lacking DOT1, Sir3 does not lose its binding to chromatin but becomes redistributed (47, 49, 62, 76). To determine whether Sir-A2G had lost specificity for silent telomeric chromatin and became redistributed, we investigated the overall binding of Sir3 to chromatin by examining the amount of histone H3 coimmunoprecipitated with Sir3. With the WT Sir3, we observed a modest but reproducible coimmunoprecipitation of histone H3 that was higher than the nonspecific pull-down of H3 in sir3Δ strains (Fig. (Fig.5C).5C). The amount of H3 that was bound to Sir3 was similar among the WT, Sir3-A2G, and dot1Δ strains (Fig. (Fig.5C),5C), suggesting that while binding to telomeres was reduced (Fig. (Fig.4B),4B), Sir3-A2G was still bound to chromatin elsewhere. To confirm that in the Sir3-A2G and dot1Δ strains Sir3 was still bound to chromatin in a nonspecific manner, we performed chromatin fractionation experiments. A WCE prepared from cells treated with formaldehyde was split into a soluble fraction containing proteins from the cytosol and the nucleoplasm and an insoluble fraction containing chromatin-bound proteins and other insoluble proteins. The chromatin pellet was subsequently sonicated to shear and solubilize the chromatin fragments. As expected, histone H3 was present in the pellet fraction and absent from the soluble pool (Fig. (Fig.5D),5D), confirming that chromatin was present exclusively in the pellet fraction. In contrast, the glycolytic enzyme Pgk1 was found primarily in the soluble pool with a smaller portion in the chromatin fraction (Fig. (Fig.5D).5D). Sir3 cofractionated with histone H3 in the chromatin-containing pellet fraction in the WT, sir3-A2G, and dot1Δ strains. No Sir3 protein was detected in the soluble pool in any of the strains (Fig. (Fig.5D).5D). These results support those shown in Fig. Fig.5C5C and suggest that in the absence of an acetylated N-terminal alanine of Sir3 or in the absence of histone H3K79me, Sir3 remained bound to chromatin but lost its specificity. Together our results indicate that histone H3K79 methylation by Dot1 affects the targeting of Sir proteins via an interaction between histone H3K79 and the acetylated N terminus of Sir3.
In this study, we have shown that cells lacking DOT1 become dependent on SIR1 for silencing at HMLα and, as a consequence, for maintenance of their mating type identity (Fig. (Fig.2).2). This genetic interaction between SIR1 and DOT1 supports a model in which Sir1 acts in one pathway to enhance the recruitment in cis of Sir proteins to silencers, and Dot1 acts in another pathway to increase specificity and target Sir proteins to the silent loci via a trans effect. While it has been unclear how the Sir complex interacts with the nucleosome core and how histone H3K79me affects these interactions, the results we present here shed some light on this interaction. Previous studies have shown that the C-terminal domain of Sir3 binds to histone tails and that this binding is negatively affected by the methylation of the tail of histone H3 at K4 and acetylation of the tail of histone H4 (8, 42, 63). The N-terminal region of Sir3 contains a conserved BAH domain that is essential for silencing (10, 27, 78). Biochemical studies suggest that the BAH domain of Sir3 is involved in nucleosome binding, although binding was observed only when a hypermorphic mutation was present in Sir3 (10). Thus, the details of the mechanism by which the BAH domain affects silencing and interacts with nucleosomes have remained unclear.
We now present three lines of genetic evidence that are consistent with a model in which the N-terminal part of the BAH domain of Sir3 binds histone H3K79 on the nucleosome core. First, specific point mutations in the Sir3 N terminus, the deletion of NAT1, and the deletion of DOT1 each had very similar silencing phenotypes. All show reduced telomeric silencing and a critical dependence on SIR1 for silencing of HMLα but not for HMRa. Second, no combination of double mutations of the sir-A2G, nat1Δ, and dot1Δ mutants showed a defect in mating that was greater than one of the single mutants (Fig. (Fig.4).4). Furthermore, whereas the sir-A2G and dot1Δ mutants both showed a reduction in Sir protein binding at telomeres, the sir-A2G dot1Δ double mutant showed a binding pattern similar to that of the sir-A2G single mutant (Fig. (Fig.5).5). This is consistent with them all acting in the same silencing process. Third, in the absence of the N-terminal alanine of Sir3 (Fig. (Fig.5),5), as well as in the absence of Dot1 (47, 49, 62, 76), Sir3 lost its binding specificity for silent chromatin at telomeres but was still able to bind to chromatin in general. Based on these results, we propose that the Sir3 N terminus binds the nucleosome core on a surface that includes histone H3K79 and that acetylation of the N-terminal alanine is required for the specificity of Sir3 for unmethylated histone H3K79 (Fig. (Fig.6).6). In the absence of this specificity (mutant or unacetylated Sir3 or no histone H3K79me in euchromatin), Sir3 becomes a promiscuous chromatin-binding factor, which leads to reduced Sir3 binding in silent regions, since Sir3 is in limited supply. Our findings are supported by recent biochemical studies by Onishi et al. (53), who showed that binding of the Sir3 BAH domain to yeast nucleosomes is negatively affected by H3K79me.
Because the sir3-A2G mutants have a stronger silencing defect than the dot1Δ mutant, our results suggest that the loss of H3K79me affects the specificity of Sir3 less than loss of a native acetylated N terminus of Sir3. One possibility is that the BAH domain interacts with additional features of the nucleosome. Indeed, the BAH domain of Sir3 binds to DNA (10) and Onishi et al. showed that the BAH domain of Sir3 binds to nucleosomes via interactions involving histone H3K79 (on the nucleosome core) as well as histone H4K16 (on the N-terminal tail of histone H4) (53). The strong silencing defect of the nat1Δ strains may be caused by a combination of the loss of Sir3 acetylation, the strongly reduced Sir3 protein levels, and the loss of acetylation of other proteins that directly or indirectly affect silencing, such as Orc1 (22, 58, 78). A BAH domain has been identified in several chromatin regulators in yeast (e.g., Yng1, Orc1, and Rsc1) and in higher eukaryotes (e.g., histone methyltransferase Ash1 in flies and Orc1 and DNA methyltransferase Dnmt1 in humans) (7, 26, 50). The BAH domain of Orc1 also binds to yeast nucleosomes (53). It will be interesting to determine whether this is a general property of BAH domains and whether nucleosome binding involves interactions with the nucleosome core. Interestingly, a recent study showed that a C-terminal fragment of Sir3 (not including the BAH domain) binds to short peptides encompassing H3K79 in vitro and that this binding is negatively affected by H3K79me (1). Although it is not clear whether these interactions also occur with nucleosomes in vivo, these results suggest that Sir3 might interact with H3K79 on the nucleosome core via its N-terminal BAH domain as well as its C-terminal domain.
Pol32 is a nonessential subunit of DNA polymerase δ, involved in DNA replication, postreplicational error-prone repair, break-induced replication, and telomerase-independent telomere maintenance (11, 30, 32, 37, 44). Through SGA analysis, we identified POL32 as a regulator of silencing at HM loci and telomeres. Subunits of DNA polymerase δ have not previously been implicated in silencing. Other DNA replication factors, such as PCNA, subunits of DNA polymerase , the clamp loader RF-C, RFC-like proteins, and ORC and MCM proteins, have previously been shown to affect silencing (12, 13, 34, 41, 68, 72, 73). In addition, silencing alterations have been observed with mutants displaying altered replication timing or cell cycle progression (40). Although the mechanism of POL32 function in silencing remains unknown, POL32 deletion aggravated the silencing defect of the DOT1 and SIR1 mutants, indicating that POL32 modulates silencing via a pathway that does not involve SIR1 or DOT1.
The underlying cause of synthetic genetic interactions is often unknown. Our results show that some synthetic fitness phenotypes in SGA analysis can be explained by the loss of mating type identity due to the loss of HMLα silencing. The sir1Δ dot1Δ mutant, which is viable and shows no growth defect, was inviable in the screen due to near complete loss of HMLα silencing. The synthetic fitness phenotype of the pol32Δ dot1Δ mutant that we and others observed in synthetic lethal screens (9, 57) is likely to be the result of enhanced silencing defects as well. To determine whether our observation has implications beyond the interactions described here, we reevaluated the 12 genes, besides DOT1, that showed the strongest interaction with SIR1 in a recent comprehensive chromosome-biology SGA screen (9). Of those 12 genes, SAS2, SAS3, and SAS4 (encoding three members of the SAS-I histone acetyltransferase complex) and ASF1 (encoding a histone chaperone) are indeed known to have a mating defect and show no fitness problems when deleted in the absence of SIR1 (46, 55, 56, 81). The BRE1, LGE1, and RTF1 deletion mutants show reduced levels of histone H3K79me and show telomeric silencing defects and are therefore expected to act like the dot1Δ mutants and have mating problems when SIR1 is deleted (33, 38, 48, 64-66, 79). We reexamined the remaining five mutants and identified a synthetic interaction between SIR1 and ELG1 (Table (Table2).2). A plating assay on haploid selection media with or without selection for mating type MATa (as shown in Fig. Fig.1)1) established that this phenotype was caused by a silencing defect of HMLα (Table (Table2).2). These findings suggest that ELG1, which encodes an RFC-like protein, positively affects silencing at HMLα, which is unexpected because it is known to negatively regulate silencing at telomeres (69). Therefore, besides dot1Δ, 9 of the 12 highest ranked genetic interactions of sir1Δ most likely represent silencing defects. DOT1 and SAS2 play nonoverlapping roles in telomeric silencing (A. W., Faber and F. van Leeuwen, data not shown), suggesting that Dot1 and the SAS-I complex act in different pathways. Whether ASF1 and ELG1 are members of the POL32 or DOT1 pathway or represent yet another pathway remains to be established.
We expect that the many other genetic interactions identified by SGA analyses will provide a valuable resource for the identification of additional genes and processes involved in silencing. The distinction between “true” synthetic lethal interactions and synthetic silencing interactions can easily be addressed by secondary assays, as we have described here.
We thank A. Tong and C. Boone for the SGA strains and advice, H. Renauld for plasmid pHR62-16, E. McIttrick for assistance with the SGA screens, J. M. M. den Haan for Sir3 antibody purification, R. Sternglanz and V. Sampath for helpful discussions and sharing of unpublished data, and M. van Lohuizen and A. Berns for critical reading of the manuscript.
F.V.L. was a special fellow of the Leukemia and Lymphoma Society (Special Fellow no. 3409-04) and was supported by the EU 6th framework program (NOE The Epigenome LSHG-CT-2004-503433).
Published ahead of print on 7 April 2008.