Histone H3K4 demethylation is negatively regulated by histone H3 acetylation in Saccharomyces cerevisiae

Abstract

Histone H3 lysine 4 trimethylation (H3K4me3) is a hallmark of transcription initiation, but how H3K4me3 is demethylated during gene repression is poorly understood. Jhd2, a JmjC domain protein, was recently identified as the major H3K4me3 histone demethylase (HDM) in Saccharomyces cerevisiae. Although JHD2 is required for removal of methylation upon gene repression, deletion of JHD2 does not result in increased levels of H3K4me3 in bulk histones, indicating that this HDM is unable to demethylate histones during steady-state conditions. In this study, we showed that this was due to the negative regulation of Jhd2 activity by histone H3 lysine 14 acetylation (H3K14ac), which colocalizes with H3K4me3 across the yeast genome. We demonstrated that loss of the histone H3-specific acetyltransferases (HATs) resulted in genome-wide depletion of H3K4me3, and this was not due to a transcription defect. Moreover, H3K4me3 levels were reestablished in HAT mutants following loss of JHD2, which suggested that H3-specific HATs and Jhd2 serve opposing functions in regulating H3K4me3 levels. We revealed the molecular basis for this suppression by demonstrating that H3K14ac negatively regulated Jhd2 demethylase activity on an acetylated peptide in vitro. These results revealed the existence of a general mechanism for removal of H3K4me3 following gene repression.

Eukaryotic genomes are packaged as chromatin by both histone and nonhistone proteins. Chromatin plays an important role in the regulation of numerous cellular processes, including transcription, DNA replication, and repair. The basic unit of chromatin is the nucleosome core particle, which comprises two copies each of histones H2A, H2B, H3, and H4, around which is wrapped 147 base pairs of DNA (1). The amino terminal tails of histones are exposed on the surface of the nucleosome and serve as the main sites for histone posttranslational modifications (PTMs). PTMs, such as acetylation and methylation, have been proposed to act as docking sites for various chromatin-modifying complexes and thus play a role in regulating chromatin structure (2).

Histone H3 lysine 4 trimethylation (H3K4me3) is one of the most commonly used hallmarks of transcriptional activity. In Saccharomyces cerevisiae, mono-, di-, and trimethylation of H3K4 is catalyzed by the histone methyltransferase (HMT) COMPASS (complex proteins associated with Set1), with Set1 as the catalytic subunit (3). Several auxiliary proteins within COMPASS are required for Set1 to catalyze the transition from mono- to di- and from di- to trimethylation, suggesting that the diverse states of methylation are differentially regulated (4). Although H3K4me3 is primarily found at the 5′ region of transcriptionally active genes, di- and monomethylation are found in the mid- and 3′ regions of transcribed genes, respectively (5). These patterns of methylation are thought to be a consequence of the interaction of COMPASS with the elongating form of RNA polymerase II (3).

Although H3K4me3 was once thought of as a stable histone PTM, the dynamic nature of lysine methylation became apparent with the discovery of histone demethylases (HDMs). Two classes of lysine-specific HDMs have been identified: amine oxidases, such as LSD1, and the JmjC (Jumonji C) domain–containing demethylases. This latter class of demethylases can be split further into several subfamilies, including the evolutionarily conserved JARID1 family of demethylases, which is characterized not only by a JmjC domain but also by JmjN, AT-rich interactive, C5HC2 zinc finger, and PHD finger domains (6). In the yeast S. cerevisiae, the lone member of the JARID1 family of demethylases that has been identified is Jhd2.

Previous studies have identified Jhd2 as a demethylase capable of demethylating H3K4 in vivo (7–9) and in vitro (10). Although several studies have focused on the identification and substrate specificity of Jhd2, little is known about how it is targeted to specific regions of the genome or how its enzymatic activity is regulated. Repression of the GAL1, SUC2, and INO1 genes is accompanied by loss of H3K4me3, which is dependent on JHD2 (10, 11). Additionally, overexpression of JHD2 results in global loss of H3K4me3 (7–9), however deletion of JHD2 does not result in a major increase of H3K4me3 in bulk histones, indicating that under steady-state conditions, Jhd2 is unable to demethylate the majority of H3K4me3 in the cell (9, 10).

Genome-wide analysis of histone PTMs in yeast showed that in addition to H3K4me3, the 5′ ends of active genes contain elevated levels of histone H3 acetylation (5, 12, 13). In yeast, histone H3 is primarily acetylated by two histone acetyltransferases (HATs), Gcn5 and Sas3, as part of multiprotein complexes (14, 15). The colocalization of H3K4me3 and histone H3 acetylation suggests that there may be “cross-talk” between these marks, whereby one PTM influences the establishment or maintenance of another (16). Consistent with this, mutation of histone H3 lysine 14 (H3K14) results in a loss of H3K4me3 in bulk histones (17), however the molecular basis for this cross-talk is unknown. In this study, we demonstrated that this relationship was due to the negative regulation of Jhd2 activity by histone H3 lysine 14 acetylation (H3K14ac). These results revealed the basis for the genome-wide colocalization of H3K14ac and H3K4me3, and explained why H3K4me3 is susceptible to demethylation only after gene inactivation.

Results

H3K4me3 Was Dependent on Acetylation of the Histone H3 Tail.

Studies examining the genome-wide localization of histone PTMs in S. cerevisiae, Drosophila, and human cells showed that levels of H3K4me3 positively correlate with that of histone H3 acetylation, suggesting that there may be cross-talk between these modifications (5, 12, 18, 19). To test whether H3K4me3 was dependent on H3 acetylation, we used immunoblot analysis of S. cerevisiae whole cell extracts to examine levels of H3K4me3 in a mutant strain lacking the HATs responsible for histone H3 acetylation. In previously published work, we demonstrated that concomitant deletion of ADA2 and SAS3 abolishes the majority of histone H3 acetylation, without disrupting acetylation of the nonhistone substrate Rsc4 (20). ADA2 encodes a noncatalytic component of all Gcn5-dependent HATs, whereas SAS3 codes for the catalytic subunit of the NuA3 HAT complex. It should be noted that there was still residual H3 acetylation in the ada2Δ sas3Δ double mutant (Fig. S1A), but this was not due to Elp3, a putative H3-specific HAT, or ADA2-independent Gcn5 HAT activity, as demonstrated in Fig. S1 B and C. The histone H3 hypoacetylation in an ada2Δ sas3Δ strain allowed us to test the requirement of H3 acetylation for H3K4me3. When we measured the levels of H3K4me3 in an ada2Δ sas3Δ double mutant, we found that this strain exhibited levels of methylation that were 41% of wild type (Fig. 1 A and B), suggesting that histone H3 acetylation was required for the establishment or maintenance of H3K4me3 levels. Deletion of ADA2 or SAS3 alone did not have as severe of an effect, indicating that these genes played redundant roles in the maintenance of wild-type levels of H3K4me3 (Fig. 1 A and B).

Fig. 1.

Histone H3-specific acetyltransferases were required for H3K4me3. (A and C) Whole cell extracts from the indicated strains were subjected to immunoblot analysis for H3K4me3, H3K4me2, H3K4me1, H3K4me0, H3K14ac, or H3. (B and D) Immunoblots shown in A and C were repeated with three independent cultures and the results quantified after normalization for histone H3. Shown is H3K4me3 (red) or H3K14ac (blue) normalized for histone H3. Error bars indicate mean ± the SE. (E) Corrected cumulative count plot (SeqMonk) of H3K4me3 levels relative to the TSS of 4,637 yeast genes in the indicated strains.

In contrast to the loss of H3K4me3, H3K4me2 levels were increased in an ada2Δ sas3Δ mutant (Fig. 1 C and D), which was consistent with the fact that the histones that lost H3K4me3 in an ada2Δ sas3Δ strain were instead dimethylated. Because H3K4me2 is also associated with transcriptionally active genes (5), these latter results indicated that the loss of H3K4me3 seen in the ada2Δ sas3Δ mutant was not due to a transcription defect. Moreover, genome-wide expression profiling showed that despite having significant growth defects (Fig. S2), the ada2Δ sas3Δ mutant exhibited changes in transcript levels at a relatively limited number of genes (Fig. S3A), with 320 genes (5.6%) showing a greater than twofold increase and 306 genes (5.4%) having a greater than twofold decrease in transcript levels in the mutant. Importantly, none of the genes required for regulation of H3K4me3 (i.e., those encoding Jhd2, components of COMPASS, the PAF complex, or components of the H2B ubiquitylation pathway) were found to be mis-regulated, and the protein levels of Set1 and Jhd2 were not affected (Fig. S3B). Thus, although histone H3 acetylation is a generally ubiquitous mark of transcribed genes, it was only important for expression of a subset of genes under the growth conditions used. These results were consistent with a recent study (21) that demonstrated a similar phenomenon for many histone PTMs.

To confirm that genome-wide levels of H3K4me3 were dependent on histone H3 acetylation, protein–DNA complexes containing H3K4me3 were immunoprecipitated from wild-type and ada2Δ sas3Δ strains, and the coprecipitating DNA subjected to Illumina-based sequencing. Read counts were plotted around the transcription start sites (TSSs) of 4,637 genes (22). To correct for the global change in H3K4me3, immunoblot data were used to normalize the total read count as described (13). A cumulative count plot revealed that, on average, genes showed a loss of H3K4me3 following disruption of the H3-specific HATs (Fig. 1E), which was confirmed by chromatin immunoprecipitation (ChIP)–quantitative polymerase chain reaction (qPCR) at specific loci (Fig. S4A).

Loss of Histone H3 Acetylation Promotes H3K4me3 Demethylation by Jhd2.

The induction of heterochromatin at the HMR locus results in a rapid loss of histone acetylation, followed by loss of H3K4me3 with slightly slower kinetics (23). This, when taken together with our data demonstrating that H3K4me3 levels were dependent on the H3-specific HATs, supported an intriguing hypothesis that in order for Jhd2 to demethylate H3K4me3, histone H3 must be deacetylated. If this was true, then deletion of JHD2, the major H3K4me3 demethylase, from an ada2Δ sas3Δ strain should result in rescue of H3K4me3 levels. We tested this using quantitative immunoblot analysis and found that when JHD2 was deleted from an ada2Δ sas3Δ mutant, H3K4me3 returned to near wild-type levels (Fig. 2 A and B). To confirm that these results were not due to increased H3K4me3 levels at aberrant loci, protein–DNA complexes containing H3K4me3 were immunoprecipitated from an ada2Δ sas3Δ jhd2Δ strain, and the coprecipitating DNA sequenced. A cumulative count plot revealed that the average genome-wide localization of H3K4me3 in an ada2Δ sas3Δ jhd2Δ strain was similar to that of wild type (Fig. 2C), indicating that loss of JHD2 restored H3K4me3 levels to regions that showed methylation loss in the ada2Δ sas3Δ strain. These results were confirmed at specific loci by ChIP–qPCR (Fig. S4B). To rule out the possibility that the restoration of H3K4me3 levels upon loss of JHD2 was an indirect effect of suppression of transcription-related defects in an ada2Δ sas3Δ mutant, we used genetic analysis to demonstrate that deletion of JHD2 did not rescue ada2Δ sas3Δ phenotypes such as slow growth, temperature sensitivity, or sensitivity to 6-azauracil (Fig. S2). Additionally, we carried out genome-wide expression profiling and confirmed that deletion of JHD2 did not change the transcription profile of an ada2Δ sas3Δ strain (Fig. S5).

Fig. 2.

Deletion of JHD2 restored H3K4me3 levels in cells lacking histone H3 acetylation. (A) Whole cell extracts from the indicated strains were subjected to immunoblot analysis for H3K4me3 and H3. A yellow color represented equal red and green intensities. (B) The immunoblot in A was repeated with three independent cultures and the results quantified. Shown is H3K4me3 (red) or H3K14ac (blue) normalized for histone H3 and shown relative to wild-type (WT) levels. Error bars indicate mean ± SE. (C) Corrected cumulative count plot of H3K4me3 levels relative to the TSS of 4,637 yeast genes in the indicated strains.

Although the majority of genes showed restored H3K4me3 levels in the ada2Δ sas3Δ mutant upon loss of JHD2, a subset of genes did not. Interestingly, one such gene, PMA1, is commonly used to examine levels of H3K4me3 by ChIP–qPCR. We examined our data for genes that showed similar behavior and identified 1,147 genes that exhibited incomplete rescue of H3K4me3 levels upon deletion of JHD2. On average, these genes had a greater loss of transcript abundance upon loss of ADA2 and SAS3 compared with all genes (P value of 0.0001). Moreover, of the 306 genes that showed a twofold decrease of transcripts in the ada2Δ sas3Δ strain, over half did not show recovery of methylation levels in a ada2Δ sas3Δ jhd2Δ mutant. We interpreted these data to indicate that although the majority of genes lost methylation of H3K4 in the HAT mutant due to increased Jhd2 activity on the hypoacetylated histones, a subset lost methylation due to decreased RNA polymerase activity.

Histone acetylation facilitates transcription through two mechanisms: (i) weakening of histone–DNA contacts and (ii) recruitment of bromodomain-containing factors that remodel chromatin structure. Methylation of histone H3K4 promotes the chromatin interaction of multiple factors, including HATs, HDACs, HMTs, and HDMs (24). Consequently, H3K4me3 is thought to regulate transcription by triggering alterations to chromatin structure. Because H3K4 methylation and H3 acetylation colocalize, it has always been difficult to tease apart their relative contributions to transcription. Our data led us to question whether restoration of H3K4me3 levels could alleviate the effects of histone H3 acetylation loss. To answer this we identified a set of genes (349) that showed reduced transcript abundance (1.5-fold) and reduced H3K4me3 levels in the ada2Δ sas3Δ strain and showed rescue of H3K4me3 levels upon deletion of JHD2. Upon comparing the levels of transcripts from this set of genes in the ada2Δ sas3Δ and ada2Δ sas3Δ jhd2Δ strains, we found that there was not a significant change in gene expression. Therefore, H3K4me3 in the absence of acetylation does not seem to enhance transcription. This was consistent with the work of others demonstrating that loss of SET1, which encodes the sole H3K4 HMT in yeast, only affects expression of a very limited number (55 in total) of genes (21).

H3K14ac Negatively Regulates H3K4 Demethylation by Jhd2 in Vivo and in Vitro.

The above data demonstrated that ADA2 and SAS3 protected the histone H3 tail from demethylation by Jhd2. To confirm that this was indeed due to acetylation of the H3 tail and to identify the residue involved, we performed site-directed mutagenesis of the acetylatable lysines on the H3 tail. We chose to mutate the residues to glutamine as opposed to arginine, as glutamine substitutions did not have an impact on cell growth. Although mutation of lysines 9, 18, or 23 to glutamine did not alter H3K4me3 levels, mutation of H3K14 recapitulated the loss of methylation observed in the ada2Δ sas3Δ strain (Fig. 3A). The requirement of H3K14 for maintenance of H3K4me3 levels has been documented by others (17), but the mechanism behind this dependency was unknown until now.

Fig. 3.

H3K14ac negatively regulates demethylation by Jhd2 in vivo and in vitro. (A) Whole cell extracts were subjected to immunoblot analysis for H3 (red) and H3K4me3 (green) and the images merged. Yellow indicated equal red and green intensities. The strains expressed either WT or mutant (mut) versions of H3 as shown. (B) ChIP-on-chip analysis of genome-wide levels of H3K14ac and H3K4me3 relative to the TSS. All genes with known TSSs were divided into five transcript length classes: very short (XS) ≤ 750 bp, short (S) 750–1,500 bp, medium (M) 1,500–2,250, long (L) 2,250–3,000, and very long (XL) 3,000–3,750. The five groups comprised 542, 1,859, 1,266, 631, and 291 genes, respectively. All genes in each group were partitioned into 150 bp bins, and the average H3K4me3 enrichment values were calculated and plotted. (C) Jhd2, purified from yeast, was subjected to histone demethylase assays using the indicated peptides. Removal of H3K4me3 was analyzed by immunoblot with H3K4me3-specific antibodies.

If H3K4me3 is dependent on H3K14ac, then it was expected that these marks colocalize on the genome. To test this, protein–DNA complexes containing H3K4me3 or H3K14ac were immunoprecipitated, and the coprecipitating DNA used to probe Affymetrix high-resolution tiling microarrays. Enriched regions were identified by comparing signal intensities of the ChIP to input DNA as previously described (25). Genes were binned by length, and the average enrichment scores for the genes in each bin were calculated. The profiles of H3K14ac and H3K4me3 were highly similar (Fig. 3B), and comparison of the MAT scores for each probe revealed Pearson and Spearman rank correlation coefficients of 0.77 and 0.75, respectively, between the two datasets, indicating that H3K4me3 and H3K14ac significantly colocalized on the genome consistent with the requirement for H3K14ac for maintenance of H3K4me3 levels.

There were two possible explanations for why Jhd2 was unable to demethylate H3K4me3 when H3K14 was acetylated. First, H3K14ac may regulate the interaction of Jhd2 with nucleosomes. We considered this unlikely because previous work has shown that the histone H3 tail is dispensable for the interaction of Jhd2 with chromatin (11). Additionally, we were unable to detect a stable interaction of Jhd2 with peptides representing the first 23 amino acids of the H3 tail regardless of the modification state (Fig. S6A). We also performed a chromatin association assay to test if deletion of ADA2 and SAS3 affected the interaction of Jhd2 with chromatin. Proteins from yeast strains expressing HA-tagged Jhd2 were fractionated into a soluble “nonchromatin” fraction (sup) and a pellet, which contained the bulk of the chromatin (chrom) in the cell. Measurement of Jhd2 levels in each fraction by immunoblot analysis demonstrated that loss of ADA2 and SAS3 did not affect the relative levels of Jhd2 bound to chromatin (Fig. S6B).

A second explanation for our observations is that although Jhd2 can bind acetylated histone H3, it is unable to efficiently demethylate it. To test this, Jhd2, purified from yeast, was subjected to an HDM assay using synthetic peptides corresponding to the H3 tail (residues 1–23) that were trimethylated at K4 (H3K4me3), with and without acetylation at lysine 14 (H3K14ac). Immunoblotting for H3K4me3 was used to detect loss of this PTM upon incubation with Jhd2. Fig. 3C shows that Jhd2 could not demethylate a K14ac peptide to the same extent as an unacetylated one, indicating that the ability of Jhd2 to demethylate H3K4me3 was negatively regulated by histone H3K14ac.

H3K14ac Regulates H3K4 Demethylation During Gene Repression.

From this study we can elucidate a mechanism for the loss of H3K4me3 observed upon transcriptional repression. During active transcription, H3K14ac prevents Jhd2 from demethylating H3K4me3 (Fig. 4A). Transcriptional repression is initiated by binding of a repressor protein (TR) to the promoter that mediates recruitment of one or more histone deacetylase complexes (HDAC), and H3 acetylation is lost. This allows Jhd2 to demethylate H3K4me3 (Fig. 4B). In ada2Δ sas3Δ or H3K14Q mutants, gene repression is not required before demethylation of H3K4me3 by Jhd2, and thus a significant amount of methylation is lost in these strains. To test this model, we examined H3K4me3 loss at the GAL1 gene upon repression in dextrose. Induction of GAL1 with galactose triggers significant histone loss, with histones being quickly restored following repression by dextrose (26). Following repression, H3K4 is demethylated over a period of several hours in a JHD2-dependent manner (10). To determine whether loss of H3K14 affected methylation loss, we repeated this analysis using our histone H3K14Q mutant and found that H3K4me3 levels were greatly reduced at the GAL1 5′ ORF compared with wild type, which supports the model that H3K14ac regulates H3K4 demethylation during gene repression.

Fig. 4.

H3K14ac regulates H3K4 demethylation during gene repression. (A and B) Models for coordinated control of histone acetylation and methylation on the H3 tail. (A) In the presence of an activating stimulus, transcriptional activators (TAs) bind DNA sequences and recruit HAT complexes, which acetylate histone H3. Subsequent recruitment of Set1 by RNAP II results in generation of all three methylation states of H3K4. H3K14ac prevents Jhd2 demethylase activity on H3K4me3, stabilizing this mark. (B) After loss of an activating stimulus or in the presence of a repressive signal, a transcriptional repressor (TR) binds to its DNA sequence and recruits histone deacetylase (HDAC) complexes. HDACs deacetylate the H3 tail, allowing Jhd2 to demethylate H3K4me3. (C) Schematic representation of the GAL1 gene. The arrow indicates the TSS and the line below shows the PCR product detected in D. (D) Yeast strains, expressing WT or mutant versions of histone H3 as indicated, were grown in galactose before transfer to dextrose-only media for the indicated times. ChIP–qPCR was used to measure H3K4me3 levels. Error bars indicate mean ± SE from three independent cultures.

Discussion

In this work we uncovered a role for histone H3 acetylation in the maintenance of H3K4me3 levels through the negative regulation of Jhd2 activity. Our findings explained several inconsistencies observed in previous studies. First, despite being the major H3K4me3 demethylase in yeast, Jhd2 seems unable to demethylate the majority of histones under steady-state conditions (8, 10). We now know that because H3K14ac and H3K4me3 are both targeted to the 5′ ends of transcriptionally active genes, K4me3 will often be found on acetylated histone H3 and as a result be protected from the activity of Jhd2. Second, although H3K4me3 is generally confined to the 5′ regions of transcriptionally active genes, multiple laboratories have observed Set1 bound to the 3′ ends of genes (10, 27). Because these regions are normally hypoacetylated and thus susceptible to demethylation by Jhd2, this may partially explain why the 3′ regions of genes do not have detectable levels of H3K4me3, despite the presence of COMPASS. Finally, induction of heterochromatin by regulated expression of the silencing protein, Sir3, results in loss of both histone H3 acetylation and H3K4me3 from the HMR locus (23). However, removal of H3K4me3 occurs with noticeably slower kinetics than histone deacetylation, which is consistent with a requirement for deacetylation before demethylation.

Although these studies were performed in yeast, we expect the results to be directly applicable to mammalian systems for several reasons. First, in human CD4+ T cells, genome-wide H3K14ac and H3K4me3 levels were found to positively correlate at the 5′ ends of genes, as we have shown here for S. cerevisiae (19). Second, the four mammalian H3K4me3-specific demethylases (JARID1A–D) show similar domain architecture to yeast Jhd2 (6, 28). Interestingly, JARID1A (RBP2) is a component of a Sin3 HDAC complex, which coordinately deacetylates and demethylates histones at E2F4 target genes during myogenic differentiation (29). Although Jhd2 does not seem to be stably complexed with any proteins in yeast (7), the idea that it may be recruited during gene repression by transient interaction with an HDAC is an intriguing one. Of final note, the H3K4-specific HDM, LSD1, is also found complexed with an HDAC (HDAC1 or 2), and hyperacetylated histones were found to be less susceptible to demethylation by recombinant LSD1, as we have seen here with Jhd2 (30, 31). It is important to point out, however, that unlike Jhd2, which acts through a Fe²⁺- and α-oxoglutarate-dependent hydrolysis, LSD1 is a flavin-monoamine oxidase, and thus parallels between LSD1 and Jhd2 must be applied with caution.

The fact that H3K4me3 was dependent on histone H3 acetylation is interesting when one considers that both Gcn5 and Sas3 are components of complexes that contain H3K4me3 binding motifs (32–36). Although we did not observe loss of H3K14ac in bulk histones in a set1Δ mutant (Fig. S7), this does not rule out the possibility that H3K14ac and H3K4me3 are mutually dependent on each other at some loci. Such a self-reinforcing loop would cause a continuous increase in both H3K4me3 and H3 acetylation levels during the period when a gene is transcriptionally active. Although it is still a matter of debate whether histone PTMs are heritable through cell division, an intriguing possibility is that any inheritance that does exist may depend more on the “intensity” of a specific PTM rather than the actual modification per se.

Materials and Methods

Yeast Strains and Antibodies.

All strains used in this study are listed in Table S1. Yeast culture, genetic manipulations, and strain verification were performed using standard protocols (37). Genomic deletions were verified by PCR and whole cell extracts were generated as previously described (38). Antibodies used for immunoblot analysis are listed in Table S2.

ChIP-qPCR, ChIP-on-Chip, and Chromatin Association Assays.

ChIP and ChIP-on-chip were performed as previously described (25, 39). The chromatin association assay was performed as outlined by others (40).

ChIP-seq Library Construction, Sequencing, and Data Analysis.

Chromatin immunoprecipitation was performed essentially as described (5) using bead beating instead of zymolyase to obtain the cell lysate. Briefly, 100 mL of S. cerevisiae was grown to midlog phase and cross-linked with 1% (vol/vol) formaldehyde for 15 min before collection. Cells were disrupted by bead beating and digested to mononucleosomes with 500 units of micrococcal nuclease. H3K4me3 nucleosomes were immunoprecipitated with 20 μg of H3K4me3 antibody (Abcam ab1012) and protein G Dynabeads (Invitrogen).

Library construction for the Illumina platform was performed using a custom procedure for paired-end sequencing. Briefly, 2–10 ng of ChIP material was end-repaired and A-tailed before being ligated to TruSeq PE adaptors. In between each reaction, the material was purified using phenol:chloroform:isoamyl alcohol extraction followed by ethanol precipitation. The resulting material was then amplified in the Phusion HF master mix (NEB) using TruSeq PE PCR primer 1.0 and custom indexed multiplexing primers [5′ AAGCAGAAGACGGCATACGAGATNNNNNNGTGACTGGAGTTCAGACGTGTGCTCTTCCGATC 3′, where “NNNNNN” corresponds to unique hexamer barcodes]. PCR amplification was performed as follows: denaturation at 98 °C for 60 s; eight cycles of (98 °C, 30 s; 65 °C, 30 s; 72 °C, 30 s), and a final extension at 72 °C for 5 min. Amplified libraries were purified using 0.8 (vol) Agencourt AMPure XP solid phase reversible immobilization paramagnetic beads and eluted in 10 mM Tris⋅HCl pH 8.5. An aliquot of each library was run on an Agilent High Sensitivity chip to check the size distribution and molarity of the PCR products.

Equimolar amounts of indexed, amplified libraries were pooled, and fragments in the 200–600 bp size range were selected on an 8% (wt/vol) Novex TBE PAGE gel (Invitrogen). An aliquot (1 μL) of the library pool was run on an Agilent High Sensitivity chip to confirm proper size selection and measure DNA concentration. The pooled libraries were diluted to 15 nM and their concentration was confirmed using the Quant-iT dsDNA HS assay kit and Qubit fluorometer (Invitrogen). Libraries were sequenced on the Illumina HiSeq machine at the UBC Biodiversity Research Centre NextGen Sequencing Facility. Clusters were generated on the cBOT (HiSeq2000) and paired-end 100 nucleotide reads generated using v3 sequencing reagents on the HiSeq2000 (SBS) platform. The hexamer barcode was sequenced using the following primer [5′ GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT 3′]. Image analysis, base-calling, and error calibration were performed using Casava 1.8.2 (Illumina).

Reads were aligned to the S. cerevisiae genome using BWA (41) and visualized using SeqMonk (v0.21.0: www.bioinformatics.bbsrc.ac.uk/projects/seqmonk). The average gene profiles were generated by aligning reads relative to transcription start sites (22). Cumulative read counts were adjusted to reflect bulk modification levels as described (13). The list of genes that show reduced H3K4me3 in both the ada2Δ sas3Δ and ada2Δ sas3Δ jhd2Δ strains was generated through SeqMonk by Pearson correlation, with a correlation coefficient of 0.9 to the read density over gene loci.

Expression Microarray Analysis.

RNA was isolated by the method of hot phenol extraction (37). Poly(A⁺) RNA was amplified and fragmented using the Message Amp III kit (Ambion) as per the manufacturer’s instructions. Hybridizations were performed on a GeneChIP Yeast Genome 2.0 (Affymetrix 900553) according to the manufacturer’s instructions. Expression data were extracted using Expression Console Software (Affymetrix) with MAS5.0 Statistical algorithm.

Jhd2 Purification and HDM Assays.

Jhd2 was overexpressed from pBG1805 (Thermo Scientific Open Biosystems Yeast ORF Collection) in a caf1Δ strain by growth in raffinose followed by an overnight induction with galactose. Cell extracts were prepared by bead beating in 50 mM Hepes, pH 7.5, 150 mM NaCl, 0.1% (vol/vol) Tween 20, and Jhd2 was immobilized on IgG-Sepharose (GE Healthcare). The purified protein was liberated by treatment with PreScission Protease (GE Healthcare) and subjected to demethylase assays (10) using synthetic biotinylated peptides corresponding to the first 23 amino acids of the histone H3 tail (Genscript). To detect methylation loss, peptides were run on 15% (wt/vol) tricine-SDS-urea gels (42) and probed with anti-H3K4me3 antibodies (Active Motif, 39159).