Histone H3 Lysine 36 Methylation Antagonizes Silencing in Saccharomyces cerevisiae Independently of the Rpd3S Histone Deacetylase Complex
Rachel Tompa, Hiten D. Madhani


In yeast, methylation of histone H3 on lysine 36 (H3-K36) is catalyzed by the NSD1 leukemia oncoprotein homolog Set2. The histone deacetylase complex Rpd3S is recruited to chromatin via binding of the chromodomain protein Eaf3 to methylated H3-K36 to prevent erroneous transcription initiation. Here we identify a distinct function for H3-K36 methylation. We used random mutagenesis of histones H3 and H4 followed by a reporter-based screen to identify residues necessary to prevent the ectopic spread of silencing from the silent mating-type locus HMRa into flanking euchromatin. Mutations in H3-K36 or deletion of SET2 caused ectopic silencing of a heterochromatin-adjacent reporter. Transcriptional profiling revealed that telomere-proximal genes are enriched for those that display decreased expression in a set2Δ strain. Deletion of SIR4 rescued the expression defect of 26 of 37 telomere-proximal genes with reduced expression in set2Δ cells, implying that H3-K36 methylation prevents the spread of telomeric silencing. Indeed, Sir3 spreads from heterochromatin into neighboring euchromatin in set2Δ cells. Furthermore, genetic experiments demonstrated that cells lacking the Rpd3S-specific subunits Eaf3 or Rco1 did not display the anti-silencing phenotype of mutations in SET2 or H3-K36. Thus, antagonism of silencing is independent of the only known effector of this conserved histone modification.

IN the budding yeast Saccharomyces cerevisiae, Sir proteins associate with DNA silencer sequences and spread to form silenced chromatin at telomeres and the silent mating-type loci HML and HMR (Jenuwein et al. 2001; Moazed 2001; Rusche et al. 2002). Unlike the initial nucleation event, the spreading of the Sir complex along chromatin appears to be independent of DNA sequence context (for review, see Moazed 2001). The spread of a complex containing Sir2, Sir3, and Sir4 along DNA is dependent on the histone deacetylase activity of Sir2 (Hoppe et al. 2002; Luo et al. 2002; Rusche et al. 2002). Sir3 and Sir4 have greater affinity for H3 and H4 that are hypoacetylated on their N-terminal tails (Carmen et al. 2002). Thus, Sir2-mediated deacetylation of neighboring nucleosomes may promote recruitment of another Sir complex via a protein–protein interaction between Sir4 and Sir2. Due to this mode of self-propagation, euchromatic genes are silenced when integrated near silencer sequences (Schnell and Rine 1986; Huang et al. 1997). While many higher eukaryotes use other components in heterochromatin formation instead of or in addition to Sir proteins, these general principles of sequence-dependent nucleation and sequence-independent propagation are conserved (Moazed 2001). Since all eukaryotic genomes have euchromatic transcriptionally active regions of DNA that abut silenced heterochromatin, cells must possess means to prevent heterochromatin from ectopically spreading beyond normal boundaries into neighboring euchromatin.

In several systems, boundary element (BE) sequences are employed to block heterochromatin spread (reviewed in Bell et al. 2001; Labrador and Corces 2002). In S. cerevisiae, the right boundary element of HMRa has been characterized as a tRNAThr gene (Donze and Kamakaka 2001). Mutations that delete this gene or abrogate its transcriptional activity cause ectopic spread of Sir proteins from HMRa into neighboring euchromatin (Donze and Kamakaka 2001). However, deletion of this boundary element results in only limited repression of the nearest euchromatic gene (Meneghini et al. 2003). Recent work has shown that multiple euchromatin-associated factors prevent heterochromatic spread in a process termed “anti-silencing” (Kimura et al. 2002; van Leeuwen et al. 2002; Suka et al. 2002; Kristjuhan et al. 2003; Meneghini et al. 2003; Santos-Rosa et al. 2004; Jambunathan et al. 2005). An example of one such factor is the histone H2A variant H2A.Z (Meneghini et al. 2003). When the gene coding for H2A.Z (HTZ1) is deleted, Sir proteins spread from telomeres and the silent mating-type loci into neighboring euchromatin to cause ectopic transcriptional silencing of genes in these regions (Meneghini et al. 2003). Similar functions have been observed for several conserved post-translational modifications of core histones, including methylation on lysine 79 of H3 (H3-K79) and acetylation on lysine 16 of H4 (H4-K16) (Kimura et al. 2002; van Leeuwen et al. 2002; Suka et al. 2002).

In this study, we identify a novel anti-silencing function for methylation on lysine 36 of H3 (H3-K36) in S. cerevisiae. This modification is catalyzed by the histone methyltransferase Set2, which has previously been implicated in transcriptional elongation and start site selection (Li et al. 2002, 2003; Strahl et al. 2002; Krogan et al. 2003; Schaft et al. 2003; Xiao et al. 2003; Carrozza et al. 2005; Joshi and Struhl 2005; Keogh et al. 2005). We found that H3-K36 is involved in anti-silencing through a reporter-based screen aimed at identifying residues in H3 and H4 that are necessary to prevent the spread of Sir-mediated silencing from HMRa into neighboring euchromatin. We show by chromatin immunoprecipitation (ChIP) that in the absence of Set2, Sir proteins spread from HMRa and telomeres into neighboring euchromatin. We further demonstrate, using whole-genome transcriptional profiling, that genes near telomeres and HMRa are enriched for those dependent on Set2 for their expression. Methylation of K36 represses erroneous transcriptional initiation by recruiting the Rpd3S histone deacetylase complex via the chromodomain of Eaf3 (Carrozza et al. 2005; Joshi and Struhl 2005; Keogh et al. 2005). We find that cells lacking Eaf3 or Rco1, two subunits of Rpd3S, do not recapitulate the effects of mutations in SET2 or H3-K36. Thus, K36 methylation has at least two independent euchromatic functions, antagonism of silencing and transcriptional start site selection, which are effected by two independent mechanisms.


Yeast strains and screen for histone mutants:

Yeast strains used in this study are described in supplemental Table S1 at http://www.genetics.org/supplemental/. Strain YM2330 was generated from strain JPY16 described in Park et al. (2002) by a plasmid shuffle to replace pDM9 (URA3 CEN HHT1–HHF1) with pJP11 (LYS2 CEN HHT1–HHF1) and integration of a PCR product containing the Candida albicans URA3 gene plus 50 bp of promoter 200 bp to the right of the 3′ end of the tRNA gene at HMRa. SET2, EAF3, RCO1, SIR2, and DOT1 genes were then replaced in this strain with MX markers to generate strains YM2332, YM2333, YM2334, YM2331, and YM2450, respectively. The plasmid BHM225 (TRP1 CEN) was also introduced into these strains, as we noted that Trp− strains showed poor growth on 5′-FOA. To find anti-silencing mutations in histones H3 and H4, a TRP1-marked plasmid containing HHT2 and HHF2 (BHM957) was mutagenized using error-prone PCR on either HHT2 or HHF2 and introduced into the strain, where both wild-type and mutant plasmids were maintained using −Lys-Trp media, and mutants conferring extra growth on 5′-FOA were selected. Trp+ plasmids were rescued from these mutants and reintroduced into the reporter strain to confirm that growth on 5′-FOA was due to the identified mutation. To quantitate the extent of anti-silencing defects, strains containing both wild-type and mutant plasmids were grown to saturation in liquid media and then 10-fold serial dilutions were plated onto selective media containing or lacking 5′-FOA. A similar strategy was followed to assess the anti-silencing defects of set2Δ, eaf3Δ, rco1Δ, and dot1Δ mutants. To assess whether anti-silencing defects due to histone mutants were Sir dependent, mutant plasmids were introduced into YM2331, which is sir2Δ, and plated as above.

Microarray hybridization and analysis:

Microarray hybridization was as described (Meneghini et al. 2003), except that mRNA was selected using the QIAGEN (Valencia, CA) Oligotex mRNA mini kit (catalog no. 70022) as per manufacturer's instructions, and data were uploaded onto a NOMAD database (http://nomad2.ucsf.edu). Genes with significant expression differences between wild-type and set2Δ cells or between wild-type and set2Δsir4Δ cells were identified using the significance analysis of microarrays (SAM) package (Tusher et al. 2001) on four replicate experiments with a 10% false discovery rate. The data for transcriptional profiling experiments of set2Δ or set2Δsir4Δ strains are available at the GEO database (http://www.ncbi.nlm.nih.gov/geo) under series accession no. GSE4934.


ChIP was performed as described (Meneghini et al. 2003) except that quantitative real-time PCR was performed with SYBR green as a label. Three replicates were performed per experiment using independent cultures for each strain. One microliter of anti-Sir3 antibody was used per sample. Polyclonal anti-Sir3 antibody was generated against the C terminus of Sir3.


Identification of histone residues that are necessary to protect euchromatin from ectopic silencing:

Several post-translational modifications have been implicated in anti-silencing. To uncover novel residues in the core histones H3 and H4 that are required for anti-silencing, we devised an unbiased genetic screen. A C. albicans URA3 gene harboring a truncated promoter was integrated so that its ATG was 250 bp to the right of the tRNA gene right boundary element of HMRa (Figure 1a). This reporter construct has been found to be sensitive to mutations that cause the spread of silencing (R. M. Raisner and H. D. Madhani, unpublished results). The strain further contained deletions of the two gene pairs encoding for H3 and H4 (HHT1–HHF1 and HHT2–HHF2) complemented by a LYS2-marked centromeric plasmid containing the HHT1–HHF1 locus and an ADE2 reporter integrated at the chromosome VR telomere (TELVR) (Park et al. 2002). We mutagenized the HHT2 gene by PCR amplification and introduced it into this strain as a TRP1-marked centromeric plasmid containing the HHT2–HHF2 gene pair using a gap repair method. A similar strategy was employed to generate mutations in HHF2. Both wild-type and mutant plasmids were maintained using selective media. Dominant mutants with decreased expression of the URA3 reporter gene were selected on 5′-fluoroorotic acid (5′-FOA), which selects against cells expressing URA3 (Boeke et al. 1987). We sequenced alleles of HHT2 and HHF2 that conferred increased plating efficiency on 5′-FOA and identified eight distinct mutations in H3 and six in H4 that cause anti-silencing defects. These alleles are detailed in Table 1, and examples of anti-silencing defects conferred by these mutants are shown in Figure 1b.

Figure 1.—

A genetic screen identifies several mutations in H3 and H4 that affect anti-silencing adjacent to HMRa. (a) Schematic of the HMRa locus and the surrounding region, indicating the site of integration of the Candida albicans URA3 reporter gene. Ty1 long terminal repeats (LTRs) are shown in light purple, the Ty5 LTR is shown in dark purple, and the tRNA-Thr gene is shown in green. The right boundary element has been defined as the tRNA gene and flanking sequences (Donze and Kamakaka 2001). BE, boundary element. Not to scale. (b) Phenotypes of histone point mutants. To determine the extent of loss of expression from URA3, strains were grown to saturation and diluted in 10-fold steps and plated on media containing or lacking 5′-FOA. Examples of increased plating efficiency on 5′-FOA conferred by several mutations in H3 and H4 and suppression of this effect by deletion of SIR2 are shown. The reporter strain also contains ADE2 integrated near TELV, which confers a pink or red coloring when silenced and white when expressed. All histone mutants that display increased growth on 5′-FOA also show ectopic telomeric silencing as assayed by darker pink coloration. (c) Comparison of phenotypes of strains with mutations in H3-K36 or SET2. Plating of WT and mutant strains on 5′-FOA as described in b is shown.

View this table:

Point mutations identified in histones H3 and H4 that confer dominant anti-silencing defects

To test if these histone mutants caused decreased expression of the URA3 reporter through Sir-mediated silencing, we examined the 5′-FOA growth phenotype for several of these mutants in a strain lacking the Sir2 histone deacetylase. Deletion of SIR2 suppressed the growth phenotype for all histone mutants tested (Figure 1b).

Of the eight residues identified in this screen (Table 1), seven are or are near residues previously shown to be necessary for anti-silencing (indicated by footnote a in Table 1). One identified residue, H3-K36, has not previously been implicated in anti-silencing. We focused on the anti-silencing capacity of H3-K36, as it is the known target of methylation by the histone methyltransferase Set2 (Strahl et al. 2002). The screen identified three different mutations in H3-K36 (Figure 1, b and c, and Table 1). We tested the effect of SIR2 deletion on the expression of the reporter gene of two K36 mutants, K36M and K36E, and found the phenotype of each to be fully suppressed (Figure 1b and data not shown). Deletion of SET2 also conferred increased growth on 5′-FOA in the reporter strain, and the double-mutant set2Δ H3-K36M strain displayed the same phenotype as the single mutants (Figure 1c), consistent with a model in which methylation of H3-K36 prevents ectopic silencing.

H3-K36 methylation protects euchromatin neighboring telomeres and HMRa:

To determine whether H3-K36 methylation protects euchromatin from Sir-mediated silencing in regions other than those flanking HMRa, we determined the transcript profiles of WT and set2Δ strains using whole-genome DNA microarrays (DeRisi et al. 1997). Four replicate experiments were performed, and, using the SAM software package (Tusher et al. 2001), genes were identified whose expression changed significantly in the absence of Set2. We identified 290 genes that decreased significantly in expression in the absence of Set2 and 492 genes with significantly increased mRNA levels. We found that regions within 30–40 kb of telomeres were significantly more likely to contain genes with increased expression in set2Δ cells than other regions (P < 0.001). Additionally, a significant number of genes whose products are involved in oxidoreductase or transporter activity showed increased expression in set2Δ cells (P = 1.87e-14 and 1.7e-9, respectively). Regions within 20 kb of telomeres were significantly more likely to contain genes with decreased expression in set2Δ cells than genes >20 kb from telomeres (Figure 2a and Table 2). The decreased expression of this telomeric cluster of genes was largely due to ectopic Sir-mediated silencing, since the deletion of SIR4 suppressed the expression defect caused by set2Δ for 26 of 37 of these genes (Figure 2b). In contrast, deletion of SIR4 suppressed the expression defects of only 37 of 253 genes >20 kb from telomeres that displayed decreased expression and only 8 of the 492 genes that displayed increased expression in set2Δ cells. Thus the genes that are dependent on Set2 for their expression can be divided into two classes, those that are silenced by Sir proteins in the absence of Set2, which are enriched near telomeres, and those that display reduced or increased expression in a Sir-independent fashion in the absence of Set2, which are found throughout the genome.

Figure 2.—

Set2 prevents ectopic silencing of regions bordering heterochromatin genomewide. (a) Histogram of Set2-dependent genes plotted as a function of distance from telomere. Whole-genome transcriptional profiling of wild-type and set2Δ strains was performed as described (Meneghini et al. 2003), using four replicate experiments from separate cultures. Genes with significant changes in expression between wild type and set2Δ were determined using SAM (Tusher et al. 2001). (b) Suppression of the subtelomeric gene expression defect by deletion of SIR4. Whole-genome transcriptional profiling was performed as in a with WT and set2Δsir4Δ strains. Genes from the plot in a were divided into two classes, those that decreased in both set2Δ and set2Δsir4Δ strains (Sir independent) and those that decreased in a set2Δ strain but not in set2Δsir4Δ strains (Sir dependent). (c) Venn diagram showing overlap between genes that decrease in set2Δ cells and genes that decrease in htz1Δ cells (Meneghini et al. 2003). Subtelomeric: genes within 30 kb of telomeres with decreased expression in htz1Δ or set2Δ. Whole genome: all genes with decreased expression in htz1Δ or set2Δ. Purple numbers indicate overlap between genes with decreased expression in each strain.

View this table:

Genes within 20 kb of telomeres are enriched for those that decrease in expression in set2Δ cells

H2A.Z is among other euchromatin-associated factors that regulate the distribution of Sir proteins (Meneghini et al. 2003). In an htz1Δ strain, genes near telomeres are enriched for those dependent on H2A.Z for their expression (Meneghini et al. 2003). Since set2Δ and htz1Δ strains both exhibit this phenotype, we assessed whether they might function in the same regions of the genome to antagonize silencing. We found that while genes that depend on H2A.Z for their expression do not overlap extensively with those that depend on Set2, those within 30 kb of telomeres do overlap significantly between the two data sets (Figure 2c).

The histone methyltransferase Dot1, which catalyzes the methylation of H3-K79 in yeast, has also been implicated in anti-silencing. Deletion of DOT1 in the anti-silencing reporter strain confers no increased growth on 5′-FOA, and deletion of both DOT1 and SET2 restores growth on 5′-FOA to wild-type levels (Figure 3).

Figure 3.—

Dot1 does not cause spread of silencing to the right of HMRa. Reporter gene assays are shown. SET2, DOT1, or both were deleted in the reporter strain used for the anti-silencing screen, and strains were plated on media containing or lacking 5′-FOA as described in the Figure 1 legend.

Increased Sir protein association in regions flanking silenced chromatin in the absence of K36 methylation:

In euchromatin near telomeres and mating-type loci, as in other euchromatic regions, levels of Sir proteins are normally very low (Meneghini et al. 2003). In mutants such as htz1Δ, the spread of Sir proteins beyond its normal boundaries occurs, resulting in ectopic Sir-mediated silencing of genes in these regions (Meneghini et al. 2003). We sought to test if this was the case in a set2Δ strain. Two regions were selected for analysis. A group of four genes near HMRa showed significant decreases in expression in set2Δ strains (indicated by arrows in Figure 4a). Additionally, in the region adjacent to the right telomere of chromosome XIV, 6 of the 7 telomere-proximal genes decreased significantly in set2Δ cells (Figure 4a). The expression defect of 5 of 10 genes from these two groups was suppressed by deletion of SIR4 (indicated gene names in boldface type in Figure 4a). Expression changes of these genes relative to wild type in set2Δ or set2Δsir4Δ strains are shown in Figure 4b. To determine whether the Sir-dependent repression of these genes in set2Δ mutants was caused by direct ectopic binding of the Sir complex, we used ChIP to compare the occupancy of Sir3 in wild-type vs. in set2Δ strains. We observed an increase over wild-type levels of Sir3 occupancy using PCR probes spanning a 312-bp region starting 302 bp downstream of the 3′ end of the tRNA HMRa BE (Figure 4c, locus D), consistent with a spread of the Sir complex from HMRa beyond normal heterochromatic boundaries. Reproducible increases of Sir3 were found in the region of two other genes that showed decreased expression in the absence of Set2, ADH7 and RDS1 (Figure 4c, loci G and I). The telomere-proximal gene AIF1 showed a large decrease in expression in set2Δ and a corresponding increase in Sir3 levels at the promoter (Figure 4c, locus Z). We also detected increased levels of Sir3 at this locus in an H3 K36A mutant (Figure 4c, right). The relative levels of H3-3meK36 in these regions as assayed by ChIP are shown in Figure 4d (data from Pokholok et al. 2005). Genes that display decreased expression in a set2Δ mutant have high levels of H3-3meK36 with their ORFs, although there is not necessarily a correlation between level of methylation in wild type and degree of silencing in a set2Δ mutant. H3-3meK36 also appears to be enriched at the right boundary element of HMRa, indicating that it may directly influence boundary element function.

Figure 4.—

(a) Schematic of the right telomeres and neighboring regions of chromosomes III and XIV. Locations of primer sets used for ChIP in c are denoted by A–I and Z. Arrows indicate genes that decrease significantly in expression in set2Δ cells as assayed by expression microarray. Genes that are derepressed in set2Δsir4Δ cells are shown in boldface type. This is adapted from Figure 3A in Meneghini et al. (2003). (b) Effect of deletion of SET2 or SET2 and SIR4 on the expression of HMR- and telomere-proximal genes. Shown are the average expression ratios derived from microarray hybridization relative to wild type of the indicated genotypes. Plotted are means and standard error of the mean for four independent experiments performed on each genotype. (c) ChIP using antibodies specific for Sir3. Quantitative PCR was performed using primers as indicated in A. Immunoprecipitation (IP) values were normalized to input values. Plotted are means and standard error of the mean for three independent experiments. (d) Relative enrichments of H3-3meK36 in regions of chromosomes III and XIV studied. Graphs represent data from experiments in Pokholok et al. (2005), where chromatin IPs against H3 or H3-3meK36 were performed and competitively hybridized to whole-genome microarrays.

Set2 protects euchromatin from ectopic silencing independently of Rpd3S:

H3-K36 methylation is a target for binding by the chromodomain-containing protein Eaf3, which is a component of both Rpd3S and the histone acetyltransferase complex NuA4 (Eisen et al. 2001; Carrozza et al. 2005; Joshi and Struhl 2005; Keogh et al. 2005). Recruitment of Rpd3S to chromatin via the interaction between Eaf3 and methylated H3-K36 is necessary to prevent ectopic transcriptional initiation (Carrozza et al. 2005). We therefore assessed whether H3-K36 methylation might also antagonize silencing by recruiting this effector. We tested whether two components of Rpd3S, Eaf3 and Rco1, were required for the anti-silencing function of Set2 by deleting EAF3 or RCO1 in the reporter strain used for the histone genetic screen. Deletion of EAF3 resulted in ∼100-fold enhanced plating efficiency over wild type on 5′-FOA media (Figure 5a). This result was confirmed using three independently derived eaf3Δ strains (data not shown). In contrast, deletion of SET2 conferred ∼10,000-fold enhanced plating efficiency over wild type on 5′-FOA media. Deletion of RCO1 resulted in ∼ ≤10-fold enhanced plating efficiency over wild type on 5′-FOA media (Figure 5a). This result was confirmed using two independently derived rco1Δ strains (data not shown). Thus, deletion of genes encoding Rpd3S-specific subunits essential for transcriptional fidelity failed to recapitulate the anti-silencing phenotype of mutations in SET2 or H3-K36.

Figure 5.—

Set2 antagonizes spread of silencing independently of Rpd3S. (a) Reporter gene assays. SET2, RCO1, or EAF3 were deleted in the reporter strain and plated as in Figure 1. (b) A genetic model that proposes distinct pathways that mediate the roles of H3-K36 methylation in anti-silencing and repression of ectopic initiation sites.


Three modified residues and surrounding residues in histones H3 and H4 are important for antagonizing silencing:

Through an unbiased genetic screen, we identified residues in the core histones H3 and H4 that are necessary to protect euchromatin from ectopic silencing. Mutations of H4-K16 and surrounding residues were identified, consistent with previous reports that acetylation of K16 by Sas2 prevents spread of Sir proteins at telomeres (Kimura et al. 2002; Suka et al. 2002). We further identified three residues in H3 that cluster near H3-K79 (Table 1), which is methylated by Dot1 (Feng et al. 2002; Lacoste et al. 2002; van Leeuwen et al. 2002). Two of these residues, H3-D77 and -D81, were previously identified in a screen for mutants with increased telomeric silencing (Smith et al. 2002). Consistent with these results, loss of methylation at H3-K79 causes increase of Sir2 and Sir3 occupancy at Y′ subtelomeric elements (van Leeuwen et al. 2002). A previous screen for histone mutations identified several residues surrounding H3-K79, including H3-K79 itself, as important for silencing (Park et al. 2002). This study identifies a nonoverlapping set of residues near H3-K79 that are important for anti-silencing. Removal of Dot1, the H3-K79 methyltransferase, does not confer increased growth on 5′-FOA in our anti-silencing reporter strain. Further, deletion of DOT1 suppresses the anti-silencing phenotype of set2Δ in the reporter strain (Figure 3). This may be due to negative regulation of Dot1 by Set2, although it is more likely due to a general redistribution of Sir proteins throughout the genome in the absence of H3-K79 methylation, as has been observed by microscopy in spread mitotic and pachytene nuclei (San-Segundo and Roeder 2000).

A novel function for Set2-mediated methylation of H3-K36:

We identified H3-K36 as necessary for anti-silencing via the reporter-based genetic screen. This residue had not previously been implicated in anti-silencing. We further demonstrated that methylation of this residue by Set2 is necessary to protect euchromatin from ectopic silencing by spread of Sir proteins from heterochromatin. Set2 is recruited to Pol II-transcribed genes via association with the elongating form of Pol II (Li et al. 2002, 2003; Krogan et al. 2003; Schaft et al. 2003; Xiao et al. 2003). Set2 was recently shown to inhibit transcriptional initiation within coding sequences by recruiting Rpd3S and promoting deacetylation of histone tails (Carrozza et al. 2005; Joshi and Struhl 2005; Keogh et al. 2005). In this work, we identify a distinct function for K36 methylation. We found that both mutation of SET2 and several different amino acid replacements of K36 cause Sir-mediated transcriptional silencing of an HMR-adjacent reporter gene. Further, we used genomewide analysis to show that Set2 protects subtelomeric euchromatin from Sir-dependent silencing. In the absence of Set2, genes within 20 kb of telomeres decrease significantly in expression. Concomitantly, Sir3 levels are increased in euchromatic regions in the absence of K36 methylation. Thus we conclude that methylation of H3-K36 antagonizes silencing near telomeres and silent mating-type loci by restricting spread of Sir proteins and further promotes expression of some telomere-distal genes independently of the Sir complex.

Not all genes with decreased expression in a set2Δ strain show increased Sir3 occupancy as assayed by ChIP. For example, GIT1 shows decreased expression in the absence of Set2 by microarray that is dependent on Sir4 (Figure 4, a and b) but no detectable increase in Sir3 in its promoter (Figure 4c, locus F). This discrepancy may indicate that Sir proteins can silence at a distance, or it may simply reflect a difference in detection thresholds between the ChIP and transcriptional profiling technologies. Similarly, the increase in Sir3 occupancy at euchromatic sites adjacent to silenced loci (Figure 4c) leads us to posit a model in which the Sir complex spreads from heterochromatin into adjacent euchromatin in the absence of H3-K36 methylation. However, we can detect more Sir3 at locus G than at locus H (Figure 4c), despite the fact that locus H is more telomere proximal than locus G. This leads us to believe that the spread of silencing proteins does not necessarily proceed in a linear gradient. Alternatively, this discrepancy could be due to noise in the ChIP assay.

H2A.Z was previously characterized in our laboratory as a general anti-silencing factor (Meneghini et al. 2003), similar to the role that we describe here for Set2. We found that a significant fraction of genes within 30 kb of telomeres with decreased expression in the absence of Set2 also show decreased expression in the absence of H2A.Z (28 of 46). However, of all the genes with decreased expression in a set2Δ strain (290), only 60 also show decreased expression in an htz1Δ strain. Consistent with this lack of overlap, htz1Δ and set2Δ strains display synthetic lethality (Krogan et al. 2003), and H2A.Z and methylated H3-K36 are present in distinct regions of genes (Krogan et al. 2003; Guillemette et al. 2005; Raisner et al. 2005; Rao et al. 2005; Pokholok et al. 2005; Zhang et al. 2005). We conclude that H2A.Z and Set2 antagonize silencing in the same regions of the genome. Their synthetic lethality may reflect independent anti-silencing mechanisms, although it is possible that H2A.Z and Set2 are present in two pathways that converge on a common downstream factor or that the synthetic lethality represents redundancy of another function unrelated to anti-silencing.

Histone methylation on K36 has both Rpd3S-dependent and independent functions:

The only known effector for K36 methylation is the Rpd3S complex, which associates with methylated H3-K36 in a manner that requires the chromodomain of the Eaf3 subunit (Carrozza et al. 2005; Joshi and Struhl 2005; Keogh et al. 2005). It has been proposed that the Eaf3 chromodomain specifically recognizes the modified residue in vivo (Carrozza et al. 2005; Joshi and Struhl 2005; Keogh et al. 2005). We observed that the eaf3Δ mutation conferred considerably weaker growth on 5′-FOA of the reporter strain than the set2Δ mutation and that the rco1Δ strain was almost indistinguishable from wild type on 5′-FOA. Therefore, Set2-mediated anti-silencing acts at least in part in a pathway that does not require the recruitment of Rpd3S (Figure 5b). It is possible that the eaf3Δ mutation confers an anti-silencing defect due to the association of Eaf3 with the NuA4 complex, which also antagonizes silencing (Oki et al. 2004), and not Eaf3's association with Rpd3S, since absence of the Rpd3S-specific subunit Rco1 conferred no detectable anti-silencing defect. Taken together, our data imply that methylation of H3-K36 functions through at least two different effector mechanisms.

Links between H3-K36 methylation and cancer:

There are three known homologs of SET2 found in humans, NSD1, NSD2, and NSD3 (Schneider et al. 2002; Rayasam et al. 2003). All three genes have been implicated in cancer. Rearrangements of the genetic sequences of NSD1 and NSD3 that result in truncations are associated with acute myeloid leukemia, and a fusion of NSD2 that is thought to cause overexpression of the gene is associated with multiple myeloma (Schneider et al. 2002). NSD3 is also rearranged in several cancerous cell lines (Schneider et al. 2002). It is possible that ectopic gene silencing occurs in response to defects in K36 methylation in humans as well, which could contribute to cancer initiation and progression.


The authors thank Jef Boeke for strains and plasmids. We are especially grateful to Ryan Raisner for construction of the anti-silencing reporter and assistance with anti-silencing assays. We thank Nguyen Nguyen for technical assistance, Shivkumar Venkatasubrahmanyam for advice on microarray experiments and analysis, and Marc Meneghini for assistance with ChIP experiments. We are grateful to Barbara Panning, Marc Meneghini, and Sigurd Braun for critical reading of this manuscript and to all members of the Madhani lab for valuable discussions and suggestions. This work was supported by a National Science Foundation Graduate Research Fellowship to R.T., a Lymphoma and Leukemia Society Scholar Award to H.D.M., and a grant from the National Institutes of Health to H.D.M. (GM071801).


  • Communicating editor: J. Tamkun

  • Received November 2, 2006.
  • Accepted November 22, 2006.


View Abstract