The Bur1 Cyclin-Dependent Protein Kinase Is Required for the Normal Pattern of Histone Methylation by Set2

Strains and media.

The S. cerevisiae strains used in this study are listed in Table 1. All media, including yeast-peptone-dextrose (YPD), synthetic complete drop-out medium (e.g., SC−Ura), and minimal and sporulation media were made as described previously (44). 5-Fluoroorotic acid (5-FOA) sensitivity was assayed on media that contained 1 g/liter 5-FOA, and medium containing 2% formamide was used to assay formamide sensitivity. Standard genetic methods for mating, sporulation, transformation, and tetrad analysis were used throughout this study (44).

Genetic methods.

A yeast genomic library containing mTn-LEU2/lacZ insertions, provided by P. Ross-Macdonald and M. Snyder (Yale University, New Haven, CT), was used to mutagenize SY54 (bur1Δ1 [pGP59 = CEN URA3 BUR1]) (see Table 1) essentially as described (3). Approximately 51,000 Leu⁺ transformants were obtained and replica plated to SC−Leu-5-FOA plates. 5-FOA-resistant colonies were selected after 3 to 4 days of growth and used as templates in PCRs using primers diagnostic for the BUR1 open reading frame. These primers generate an 880-bp PCR product indicative of BUR1⁺ and a 1,937-bp PCR product indicative of bur1Δ1::TRP1. 5-FOA-resistant candidates that generated only the 1,937-bp band were pursued further. Candidates were crossed with GY480 (BUR1⁺) and tetrads were dissected to determine if viability was linked to the transposon-based LEU2 marker. Two transposon-linked suppressors tentatively designated mut3 and mut59 were obtained. The transposon insertion site was identified by transformation with pRSQ2-URA3 (3), digestion of the genomic DNA with EcoRI, ligation under dilute conditions, recovery in bacteria, and DNA sequencing of the transposon-insert junction.

Gene disruptions were carried out using a kanamycin resistance cassette generated from plasmid pRS400-KanMX4 as the template by a PCR-assisted method (55). PCR products were purified using a PCR purification kit (QIAGEN), and transformed into the diploid yeast strain GY763. Following transformation, yeast cells were grown in liquid YPD for 4 h and spread onto YPD plates containing G418 (200 μg/ml; BRL) to select transformants that had the gene disrupted by the kanamycin cassette via homologous recombination.

Plasmids.

A SacI-XhoI fragment from pBM3352 (22) containing SET2 was subcloned into pRS416, generating pCYY23 (CEN URA3 SET2). The SacI-XhoI fragment containing set2C82YD83Q from pBM3354 (22) was subcloned into pRS416 to generate pCYY25 (CEN URA3 set2C82YD83Q).

Western blots.

Yeast crude extracts were prepared by growing 30 ml of cells to a concentration of 2 × 10⁷ to 3 × 10⁷ cells per ml. Cells were harvested by centrifugation at 2,000 rpm for 5 min and resuspended in 500 μl of breaking buffer (50 mM Tris [pH 7.5], 10% glycerol, 10 mM MgCl₂, 1 mM EDTA, 100 mM NaCl, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, one tablet of protease cocktail [Roche]). Cells were disrupted by vortexing with glass beads, and extracts were clarified by centrifugation at 16,000 × g for 10 min (59). Proteins were separated by 10 to 20% gradient polyacrylamide gel electrophoresis (PAGE) gel (Bio-Rad) and transferred to Immuno-Blot polyvinylidene difluoride membranes using a Bio-Rad Transblot SD transfer cell. Membranes were blocked in 3% bovine serum albumin (or 1% bovine serum albumin for monomethyl K36 antibody) in PBS-Tween for 1 h and incubated overnight at room temperature with primary antibody followed by peroxidase-conjugated anti-rabbit immunoglobulin G secondary antibody.

The antibody complexes were detected using enhanced chemiluminescence reagents (Pierce) as specified by the manufacturer. Antibodies that recognize total histone H3 (ab1791), monomethylated histone H3 K36 (ab9048), and trimethylated histone H3 K36 (ab9050) were obtained from Abcam, and the antibody that recognizes dimethylated histone H3 K36 was obtained from Upstate (07-369). Western analysis for the antibody specificity test was performed as described above except that peptides were spotted onto polyvinylidene difluoride membranes and allowed to dry overnight. The membranes were incubated with primary antibody for 1 hour, followed by a 1 -h incubation with secondary antibody. For bur1, bur2Δ, and ctk1Δ strains the amount of extract was increased 20 to 50% to adjust for the smaller amount of total histone H3 typically observed in those extracts.

Chromatin immunoprecipitations.

Chromatin immunoprecipitations were performed essentially as described (27). Chromatin was sheared using an Ultrasonics Inc. sonicator at level three with 12 pulses of 20 seconds. Immunoprecipitations (IPs) were performed using 2.5 μl antibody to monomethyl H3 K36, 1.0 μl antibody to trimethyl H3 K36, and 1.0 μl anti-H3 antibody. Inputs and IP samples were analyzed by standard PCRs to which 1 μCi of ³²P-labeled dATP had been added. Products were resolved on 1.8% agarose gels and quantified using a Storm 840 scanner and ImageQuant software (Molecular Dynamics). Results are normalized to inputs and expressed as ratios of methylated histone H3 K36 to total histone H3. The results shown are representative of several independent experiments that gave similar results.

Primers for HMRa and the ACT1, PMA1, and GAL10 open reading frames were as follows: HMRa, 5′CAGTTTCCCCGAAAGAACAA and 5′CCATCCGCCGATTTATTT; ACT1, 5′CCAATTGCTCGAGAGATTTC and 5′CATGATACCTTGGTGTCTTG; PMA1, 5′AAGTCGTCCCAGGTGATATTTTGCA and 5′AACGAAAGTGTTGTCACCGGTAGC; and GAL10, 5′GATCTTCCATACAAAGTTACGGand 5′GGCCTCGACACCCC.

Genomic suppressors of bur1Δ.

Deletion of BUR1 was reported to cause lethality or extremely poor growth (13, 17, 42). In our strain background, bur1Δ strains grew extremely slowly, but revertants that grew significantly better arose readily, prohibiting propagation of pure bur1Δ strains. We anticipated that these viable bur1Δ revertants contained genomic suppressor mutations that would be informative for understanding the roles of BUR1 in vivo. Although revertants of bur1Δ arose spontaneously, we utilized a transposon mutagenesis strategy to identify suppressors of bur1Δ and to facilitate rapid identification of the mutant genes. A bur1Δ strain, covered by a BUR1⁺ URA3 CEN plasmid to maintain viability, was transformed with an mTn-lacZ-LEU2 integrating-plasmid library (3), selecting for transposon insertions that allowed loss of the BUR1 URA3 plasmid.

The entire screening involved three steps. First, ∼51,000 Leu⁺ transformants were screened for 5-FOA-resistant colonies, which could include the desired viable bur1Δ strains, plasmid-borne ura3 mutations, or gene conversion of bur1Δ to BUR1⁺ (Fig. 1A). To distinguish between these possibilities, the 5-FOA-resistant colonies were screened by PCR amplification using primers that distinguish the presence of bur1Δ and BUR1⁺ (Fig. 1B). Finally, a genetic cross was utilized to determine whether the mutation responsible for any viable bur1Δ strains was linked to the LEU2⁺-marked transposon. Through this process we obtained two bur1Δ deletion strains in which there was no detectable source of BUR1⁺ and survival of which was strongly linked to the LEU2-marked transposon.

The insertion site was identified for both suppressors by rescuing the transposon and DNA sequencing of the transposon-insertion junction. Interestingly, both suppressors were due to insertions into SET2, which encodes a histone methylase that targets lysine 36 of histone H3 (22, 52). One insertion occurred after amino acid 136 in the SET domain that is required for methylase activity, while the other insertion was after amino acid 495 in the WW domain, suggesting that suppression was due to loss of SET2 function. This conclusion was confirmed by showing that a set2 deletion suppressed bur1Δ as efficiently as the original transposon insertion alleles (Fig. 2A) and that transformation with a SET2⁺ CEN plasmid reversed the suppression phenotype (Fig. 2B).

The identification of set2 mutations as suppressing the severe growth defect of bur1Δ was important because, despite the widely inferred importance of histone methylation, set2 deletion strains are healthy and display very few phenotypes; set2Δ has been reported to cause both 6-azauracil-sensitive and 6-azauracil-resistant phenotypes (18, 20, 25, 26, 47), and set2Δ increases transcription from a gal4ΔUAS allele (22). Although suppression of gal4ΔUAS by set2Δ appears analogous to our Bur⁻ selection, which uses suc2ΔUAS as a reporter, set2Δ did not cause Bur⁻ or Spt⁻ phenotypes, which are typically associated with bur1 and bur2 mutations, and bur1-8 did not reverse the Gal⁻ phenotype of gal4ΔUAS (Fig. 3C and data not shown). These results suggest that although SET2 and BUR1 have independently been implicated in elongation, their roles are nonoverlapping, and instead are at least partially antagonistic.

Genetic analysis of the Bur1-Set2 relationship.

A critical test for any suppressor is the specificity of the suppression phenotype (41), and the suppression of bur1Δ inviability by set2Δ was quite specific by several criteria. Two other histone methylases have been identified in yeast; Set1 methylates histone H3 on lysine 4 (2, 19, 32, 43) and Dot1 methylates H3 on K79 (34, 54). Deletion of either SET1 or DOT1 did not suppress bur1Δ, either singly or in combination, and set1Δ and dot1Δ also did not strengthen the suppression by set2Δ (Fig. 3A and data not shown). This demonstrates the specificity of the interaction and the lack of redundancy between the histone methylases, at least with respect to BUR1.

The specificity of suppression was tested in other ways. To test for gene specificity, double mutants were created with set2Δ and deletions of other genes that are involved in elongation. set2Δ reversed the formamide-sensitive and cold-sensitive phenotypes of bur2Δ mutants (Fig. 3B), had a modest effect on the growth of bur2Δ strains (Fig. 3C), and was unable to suppress the inviability caused by deletion of SPT5, SPT6, or SPT16 (data not shown), each of which is phenotypically related to BUR1 and has a role in elongation. Although we cannot rule out the possibility that set2Δ might suppress subtler mutations in these or other genes, suppression by set2Δ was specific for the Bur1-Bur2 complex and did not extend to any other elongation or transcription factors tested to date. Finally, set2Δ did not suppress all bur1 or bur2 mutations; set2Δ suppressed the temperature-sensitive phenotype of the bur1-8 missense mutant (data not shown), the inviability of bur1Δ mutants, and the poor growth, formamide-sensitive, and Cs⁻ phenotype of bur2Δ mutants, while other bur1Δ and bur2Δ phenotypes and healthier bur1 or bur2 alleles were not suppressed. Since set2Δ suppresses bur1Δ and bur2Δ null alleles, this example of allele specificity is not indicative of any physical interaction between the two proteins, but likely suggests that severe defects in the kinase complex are required to observe the relationship with SET2. Combined, these tests demonstrate the specificity of the relationship between the Bur1-Bur2 complex and SET2.

The identification of set2Δ as a suppressor of the bur1Δ growth defect functionally linked BUR1 to histone methylation. But was suppression due to defective methylation of histone H3, or could suppression be due to some other unsuspected role of SET2? Two results clearly indicated that suppression was due to loss of SET2-dependent methylation of K36. First, a set2 C82Y missense mutation that causes defective methylation activity in vitro (22) suppressed bur1Δ as effectively as set2Δ (Fig. 4A), suggesting that loss of methylation activity was responsible for suppression. Second, if suppression of bur1Δ and bur2Δ was due to loss of methylation of histone H3 K36, then mutation of K36 should also suppress bur1Δ. As predicted, a histone H3 K36R mutation suppressed bur1Δ as effectively as set2Δ, strongly indicating that suppression of bur1Δ was caused by loss of K36 methylation (Fig. 4B).

Molecular analysis of the Bur1-Set2 relationship.

To gain a better understanding of the relationship between BUR1 and SET2, we investigated the effect of bur1 mutations on the methylation status of histone H3 in vivo using antibodies that specifically recognize monomethylated, dimethylated, and trimethylated lysine 36 of histone H3. The specificity of the antibodies was first tested by assaying for whether the antibodies cross-reacted with unmodified or methylated histone H3 peptides, and as shown in Fig. 5A, each antibody was highly specific for the appropriate methylated peptide.

Using these antibodies in Western blots of crude cell extracts, we found that a combination of monomethylated, dimethylated, and trimethylated forms of lysine 36 was detected in wild-type strains, and as expected, all methylation was SET2 dependent (Fig. 5B). In a temperature-sensitive bur1-8 mutant strain, the levels of trimethylated K36 were significantly reduced, while monomethylated and dimethylated K36 persisted longer than in the BUR1⁺ strain, with the monomethylated K36 levels decreasing after prolonged incubation at the nonpermissive temperature (Fig. 5B and C). This specific reduction of trimethylated K36 could be detected at the permissive temperature in the bur1-8 strain and was exacerbated at 37^oC, while methylation of K36 was unaffected in a wild-type strain at either temperature. A similar reduction of trimethylated K36 was also observed in a bur2Δ strain (Fig. 5B). These results suggest that the Bur1-Bur2 kinase is directly or indirectly required for the transition from a dimethylated to trimethylated state of histone H3 at K36. This is analogous to the recently proposed requirement of Bur1 for the transition of K4 methylation from the dimethylated to the trimethylated state (23).

The Western blots shown in Fig. 5 assay the methylation status of total histone H3 in crude cell extracts. To analyze the K36 methylation state at specific genes, chromatin immunoprecipitation assays were performed using the monomethylated K36- and trimethylated K36-specific antibodies. Previous chromatin immunoprecipitation assays demonstrated that the K36 dimethylation pattern is unaffected by bur2Δ (57). At two constitutively highly expressed genes, ACT1 and PMA1, the level of monomethylated K36 was essentially unchanged in BUR2⁺ versus bur2Δ strains, while the level of trimethylated K36 was significantly reduced, consistent with the results obtained in the crude extracts (Fig. 6A).

At two genes that are transcriptionally inactive under the growth conditions used, HMR and GAL10, the levels of trimethylated K36 were low and largely unaffected by the bur2Δ mutation. When GAL10 was induced on galactose-containing medium, the level of trimethylated K36 increased significantly in a BUR2⁺ strain, but not in a bur2Δ mutant (Fig. 6B). The slight increase in trimethylated K36 observed in the bur2Δ strain is likely due to low levels of Bur2-independent Bur1 kinase activity or BUR-independent methylation by Set2. Combined, these results reveal that both at individual genes and in total cell extracts, the Bur1-Bur2 protein kinase is required for the transcription-associated increase in histone H3 K36 trimethylation, but not for K36 monomethylation or dimethylation.

To determine whether the reduction of trimethylated K36 was observed in other mutant backgrounds, we began by surveying a panel of Spt⁻ mutants, which are phenotypically most closely related to BUR1 mutants (Fig. 7A). A variety of patterns were observed; mutations in SPT4 and SPT5 (DSIF subunits) had no effect on K36 methylation, an spt16 mutation selectively reduced trimethylation, although not to the extent observed in bur1 or bur2 strains, and spt6 mutants showed allele-specific effects. An spt6-140 strain displayed wild-type methylation in these crude extracts, while only the monomethylated form of K36 was observed in an spt6-1004 strain. The reason for this allelic difference is currently unknown, but phenotypic differences between these two alleles have been observed previously; both strains display strong Ts⁻, Bur⁻, and Spt⁻ phenotypes, but only the spt6-1004 mutation increases initiation from cryptic promoters (15).

Strains containing mutations in genes related to SET2 were examined next (Fig. 7B). Three patterns were observed again; a deletion of CTK1, which encodes a protein kinase previously reported to be required for K36 dimethylation (58), completely abolished all methylation of K36, similar to set2Δ, while deletion of SET1, DOT1, or RAD6, which are responsible for methylation of K4 and K79 of histone H3 and ubiquitylation of histone H2B, respectively, had no effect on K36 methylation. Finally, three truncations of the Rpb1 CTD (Fig. 7B) selectively reduced K36 trimethylation, as in bur1 and bur2 mutants, with the CTD truncations having a more severe effect than the bur mutations. Combined, these results suggest that the methylation pattern of K36 is affected by many factors in a complex manner, with the selective loss of K36 trimethylation being observed in bur1, bur2, spt16, and rpb1 CTD truncation strains. More extensive methylation defects that differ from the pattern observed in bur1 mutants occur in spt6 and ctk1 mutants.

Posttranslational modifications of histones have been proposed to affect nucleosome function by two broad mechanisms, either by directly affecting interactions with DNA and the other histones, or by affecting the recruitment of modification-specific histone-binding proteins (10). In particular, methylation marks on histone H3 on lysine 4, lysine 9, and lysine 27 stimulate direct binding of Chd1, HP1, and polycomb (PC), respectively, via recognition of the methylated residues by the Chd1, HP1 and PC chromodomains (1, 5, 9, 21, 38). Based on these precedents, we expected that methylation of K36 recruits a chromodomain-containing protein.

Three chromodomain proteins, Chd1, Eaf3, and Esa1, have been identified in yeast (38), and all three have chromatin-mediated functions in transcription. As CHD1 and EAF3 are nonessential genes, we tested whether deletion of either of these genes could also suppress bur1Δ, and indeed, chd1Δ and eaf3Δ individually suppressed the bur1Δ growth defect (Fig. 8A). Unlike set2Δ-mediated suppression, which reduced all K36 methylation, the methylation pattern in bur1Δ eaf3Δ and bur1Δ chd1Δ strains was unchanged compared to that of bur1 mutants, with normal levels of mono- and dimethylated K36 but greatly reduced trimethylation (Fig. 8B). We conclude that eaf3Δ and chd1Δ suppress bur1Δ differently than set2Δ does; while set2Δ results in loss of methylated K36 marks, eaf3Δ and chd1Δ abolish downstream effects of K36 mono- or dimethylation. In agreement with these results, a recent report (16) demonstrated that bur1Δ is suppressed by eaf3Δ and chd1Δ.