|Home | About | Journals | Submit | Contact Us | Français|
Posttranslational modifications to histones have been studied extensively, but the requirement for the residues within the tails for different stages of transcription is less clear. Using RNR3 as a model, we found that the residues within the N terminus of H3 are predominantly required for steps after transcription initiation and chromatin remodeling. Specifically, deleting as few as 20 amino acids, or substituting glutamines for lysines in the tail, greatly impaired K36 methylation by Set2. The mutations to the tail described here preserve the residues predicted to fill the active site of Set2, and the deletion mimics the recently described cleavage of the H3 tail that occurs during gene activation. Importantly, maintaining the charge of the unmodified tail by arginine substitutions preserves Set2 function in vivo. The H3 tail is dispensable for Set2 recruitment to genes but is required for the catalytic activity of Set2 in vitro. We propose that Set2 activity is controlled by novel intratail interactions which can be influenced by modifications and changes to the structure of the H3 tail to control the dynamics and localization of methylation during elongation.
Chromatin dynamics play important roles in DNA-directed events such as transcription, repair, replication, and recombination. Compact chromatin structure prevents proteins from accessing DNA and represses these functions, and conversely, nucleosome-free regions are found upstream of promoters, allowing for the binding of regulatory factors. Maintenance and alteration of chromatin structure depends on ATP-dependent chromatin remodeling complexes and covalent histone modifications. The concept of a histone code has emerged due to the plethora of modification types and locations and the interdependence of one modification on another (3, 21). Furthermore, certain modifications are linked to transcription activation, while others are linked to repression.
Understanding how cross talk among histone modifications regulates transcription has been of great interest. Several instances of trans-tail regulation have been observed, where one histone modification is regulated by the interaction between the histone-modifying enzyme and residues in another histone in the nucleosome. For example, Dot1 methylation of H3 lysine 79 requires an interaction between Dot1 and a small basic patch on the H4 tail (18). There also exist more complex relationships in which one histone modification precedes and is required for the second modification. Mono-ubiquitylation of H2B lysine 123 precedes trimethlation of H3 lysine 4 by COMPASS (13, 28) and lysine 79 methylation by Dot1 (43, 60). Histone modifications are regulated by intratail mechanisms also, where residues within the same histone control the modification at another site. Methylation of H3 arginine 2 inhibits COMPASS methylation (19, 24), and similarly, phosphorylation of H3 serine 10 inhibits methylation of lysine 9 (46). Interestingly, serine 10 phosphorylation also stimulates GCN5-mediated acetylation of lysine 14 (7, 35). Generally, examples of intratail regulation described thus far involve interactions between residues in close proximity to the modification site and modifications that are discriminated by the structure of the active site of the enzyme (11).
Methylation of H3 lysine 36 is greatly enriched at the 3′ ends of transcribed genes in Saccharomyces cerevisiae (1, 44, 45), and the trimethlated form (me3) correlates strongly with transcription frequency (2, 26, 32, 39, 62). Set2 interacts with RNA polymerase II (RNAPII) and methylates H3 on K36 in the wake of elongation (25, 32). Lysine 36 methylation serves to recruit the Rpd3 histone deacetylase (HDAC) complex to reset the chromatin, preventing intragenic transcription (6, 22, 30). Set2 methylation, independent of the Rpd3 pathway, has also been shown to play a role in the prevention of heterochromatin spreading in a mechanism that is understood to a lesser degree (56). These observations suggest that lysine 36 methylation serves multiple functions in the writing and interpreting of histone modifications.
Set2 is regulated at several levels, including its recruitment to genes by the phosphorylation of serine 2 on the C-terminal domain (CTD) of RNAPII by Ctk1 (25, 32, 61). Furthermore, Ctk1 regulates the levels of Set2 by preventing its degradation, presumably by sequestering it to RNAPII (63). Another level of regulation is the structure of the nucleosome. Set2 was first characterized as a “nucleosome-selective” histone methyltransferase (54); however, low levels of activity on histone H3 alone, and on H3 peptides, have been observed (14, 42). The nucleosome selectivity is mediated by an interaction between Set2 and a residue within the core domain of H4, lysine 44, allowing it to dock onto nucleosomes (14). This provides another example of trans-tail regulation of histone modifications. Despite recent advances, still very little is known about how the dynamics and localization of Set2-dependent methylation are controlled in vivo.
Here we describe studies showing that residues within the H3 tail predicted to be distant from those occupying the catalytic site of Set2 are required for Set2-directed K36 methylation in vivo. Furthermore, these mutants display all the hallmark phenotypes of a set2Δ strain, including 6-azauracil (6-AU) resistance and production of cryptic transcripts from within open reading frames (ORFs) (6, 26). Lysine substitution mutants suggest the charge of the tail is important for its regulation of K36me. Biochemical analyses using recombinant components suggest that the tail directly stimulates Set2 activity. Together, these results support a model by which Set2 methylation is regulated by intratail interactions within H3, which have the potential to be regulated by modification of residues or by the recently described proteolytic cleavage of the tail (15).
The strains used in this study are listed in Table Table1.1. Gene deletions and tagging were carried out by standard techniques (5, 36). Cells were grown in yeast extract-peptone-dextrose at 30°C. For 6-AU sensitivity, threefold dilutions of cultures transformed with a URA3+ plasmid were spotted onto either synthetic complete (SC)-URA medium or SC-URA plus 100 μg/ml 6-AU and grown at 30°C.
Yeast culture was grown to an optical density at 600 nm (OD600) of 0.7, and 15 ml of cells was harvested for total RNA extraction as previously described (47). Where indicated, cells were treated with 0.02% methyl methanesulfonate (MMS) for 1 or 2.5 h prior to harvesting. Briefly, 20 micrograms of total RNA was separated on formaldehyde-containing agarose gels and transferred to a Hybond-XL membrane (GE Biosciences, Piscataway, NJ) by capillary blotting. Probes for RNR3, Scr1, and STE11 were prepared by PCR. The ScR1 signal of each sample was used to correct for loading of RNA. All data represent at least three independent experiments. Error bars represent the standard errors.
The following antibodies were used in chromatin immunoprecipitation (ChIP): 1 μl anti-TATA-binding protein (anti-TBP; our lab), 2 μl anti-RNAPII (8WG16; Covance, Princeton, NJ), 8 μl anti-Ser2-P (H5; Covance, Emeryville, CA), 1 μl anti-H3 C terminus (ab1791; Abcam, Cambridge, MA), 1 μl anti-H3 K36me3 (ab9050; Abcam, Cambridge, MA), 1 μl anti-H3 K36me2 (07-369; Upstate/Millipore, Billerica, MA), and 1 μl anti-myc (9E10 ascites; Covance, Emeryville, CA). The following antibodies were used in Western blotting: 1:10,000 dilution of anti-H3 C terminus (ab1791; Abcam, Cambridge, MA), 1:1,000 of anti-H3 K36me3 (ab9050; Abcam, Cambridge, MA), 1:5,000 of anti-H3 K36me2 (07-369; Upstate/Millipore, Billerica, MA), 1:10,000 of anti-H3 K79me2 (04-835; Upstate/Millipore, Billerica, MA); K36Ac (07-540; Upstate/Millipore, Billerica, MA), and 1:10,000 of anti-FLAG M2 (F1804; Sigma, St. Louis, MO).
ChIP assays were performed as previously described, with minor changes (53). Yeast cultures (100 ml) were grown in yeast extract-peptone-dextrose medium to an OD600 of 0.6 to 0.8. The cells were cross-linked with 1% (vol/vol) formaldehyde at room temperature for 15 min. The formaldehyde was quenched by the addition of glycine. MMS-induced cells were treated with 0.02% MMS and incubated for 2.5 h prior to formaldehyde treatment. Whole-cell extracts were prepared by glass bead disruption and sheared into fragments averaging 200 to 600 bp in size using a Bioruptor (Diagenode, Philadelphia, PA). Whole-cell extracts were immunoprecipitated (IP) with the antibodies indicated throughout the text. After purification, the precipitated and input DNAs were analyzed by semiquantitative PCR. Primer sequences are available upon request. PCR products were analyzed by electrophoresis and ethidium bromide staining, scanned with the Typhoon system (Molecular Dynamics), and quantified by using ImageQuant. The percent IP represents the following product: (IP signal/input signal) × 100. For histone modification ChIPs, the data are represented by the formula (percent IP modified)/(percent IP of total H3) in order to account for variations in histone levels. All data represent at least three independent experiments from multiple extracts. Error bars represent the standard errors.
Yeast cultures (25 ml) were grown to an OD600 of 0.6 to 0.8. Whole-cell extracts were prepared by glass bead disruption. Extracts were centrifuged, and 1 volume of 3× sodium dodecyl sulfate (SDS) load buffer was added before boiling. Extracts were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE), transferred to nitrocellulose membranes (Whatman), probed with the indicated antibodies, and detected by enhanced chemiluminescence (Pierce). Bulk chromatin was prepared as described previously (16).
A DNA fragment of Set2 encoding amino acids 1 to 261 was cloned into pGEX6P by standard molecular biology techniques. Full-length wild-type and R195G Set2-FLAG were described previously (a gift from Brian Strahl) (55). Constructs were transformed into Escherichia coli BL21(DE3 LysS), and proteins were purified using glutathione-agarose beads (glutathione S-transferase-Set2 1-261) or anti-FLAG resin (Set2-FLAG) according to the manufacturer's recommended conditions (Amersham-Pharmacia and Sigma). Set2 1-261 was released from the glutathione beads with PreScission protease (GE Healthcare). Protein quality and quantity were determined by SDS-PAGE with Coomassie blue staining. Methyltransferase assays were performed as described previously with minor alterations (54). Briefly, various amounts of Set2 1-261 were incubated with 4 μg of recombinant Xenopus laevis nucleosomes and 0.6 μCi of S-[3H]adenosylmethionine ([3H]SAM) for 30 min at 30°C. Recombinant Xenopus nucleosomes were prepared and reconstituted as previously described (37). After reconstitution, nucleosomes were purified using high-performance liquid chromatography, quantified by spectrophotometry, and analyzed by SDS-PAGE and native gel electrophoresis prior to use in methyltransferase and pull-down assays. For nucleosome pull-down assays, 1 μg of the wild-type or R195G Set2-FLAG was incubated with various amounts of recombinant Xenopus nucleosomes in binding buffer (10 mM Tris, pH 7.4, 50 mM NaCl, 1 mM MgCl2, 1% Triton X-100, 1% Ficoll, 0.5 mM phenylmethylsulfonyl fluoride) for 1.5 h at 4°C. Set2-FLAG resin was washed three times in binding buffer, and 20 μl of 3× SDS-load buffer was added before boiling. Nucleosome pull-down assays were then analyzed by Western blotting using anti-H3 C terminus and anti-FLAG antibodies.
In S. cerevisiae, the histone N termini are nonessential, enabling investigations into their roles in transcription regulation (38). Historically, deletion of the H3 tail has been associated with derepression of transcription (57); however, recent gene expression profiling of H3 tail mutants revealed it is required for activation and repression in vivo (49, 50). Whether or not the activation defects were direct and the mechanisms behind the activation defects were not examined. Moreover, uncertainty remains whether the mutations affect chromatin remodeling, initiation, or elongation. The DNA damage-inducible RNR3 gene was used as a model to examine the role of the histone N termini in transcription activation (34, 65). We examined the activation of RNR3 in strains harboring a deletion, or lysine-to-glutamine substitutions, within the N-terminal tail of H3 and compared the mRNA induction levels to their respective wild-type backgrounds (Fig. 1A and B). Despite the long-known role of the H3 tail in repression, neither the H3 Δ1-28 nor the H3 K-4, -9, -14, -18, -23, or -27-Q (H3 K→Q hereafter) mutant displayed strong derepression of RNR3 under the uninduced condition (−MMS); however, both mutants had significantly reduced levels of induced RNR3 transcription (Fig. (Fig.1B).1B). The H3 K→Q mutant exhibited a defect in transcription similar to the H3 Δ1-28 strain, suggesting that the modifiable lysines on the H3 N terminus are of particular importance for high levels of induced transcription.
To determine which steps in the activation of RNR3 were affected by the histone tail mutations, we first examined preinitiation complex (PIC) formation in the mutants by measuring the recruitment of TBP to RNR3. In the wild-type strain there was a four- to fivefold increase in TBP recruitment upon induction (Fig. (Fig.1C).1C). Surprisingly, neither histone tail mutation displayed a significant reduction in TBP recruitment. This was in contrast to the 60 to 70% loss of mRNA under the same conditions (Fig. (Fig.1B).1B). This result suggests that chromatin remodeling is unaffected in these strains, because remodeling is required for TBP recruitment (52). In fact, we previously reported that SWI/SNF recruitment was unaffected by the deletion of the H3 tail (55). We next analyzed the recruitment of RNAPII to the promoter in the histone mutants (Fig. (Fig.1D).1D). RNAPII recruitment to the promoter was ~45 to 50% lower in the H3 Δ1-28 strain than that observed in the wild type; however, the loss was not proportional to the loss of RNR3 expression. The RNAPII recruitment to the promoter was unaffected in the H3 K→Q strain, yet mRNA levels were greatly reduced in this mutant (Fig. (Fig.1B).1B). These results suggest that the transcription defects in this mutant predominantly occur after PIC formation.
Next, we characterized the H3 tail mutants further by examining the density of RNAPII across the gene. Upon activation, the increase in RNAPII recruitment is consistently ~5- to 7-fold above the uninduced level in the wild-type strain, indicating its even distribution across the gene (Fig. (Fig.1E).1E). Deletion of the H3 N terminus results in lower levels of RNAPII recruitment to the RNR3 promoter after induction. However, RNAPII levels within the ORF (at ORF B, ORF C, and ORF D) were equal to that observed in wild-type cells. Even in the uninduced state (−MMS), there is greater RNAPII density in the RNR3 ORF than at the promoter in the H3 Δ1-28 mutant. A smaller H3 deletion (Δ1-20) also resulted in a similar shift in RNAPII density (data not shown). To better illustrate the difference in the distribution of RNAPII between the wild type and the H3 Δ1-28 mutant, we plotted its density across RNR3 in each strain, relative to that recruited to the promoter (Fig. (Fig.1F).1F). The wild-type strain showed little variation in RNAPII density across the RNR3, until it dropped off at the untranslated region (UTR) near the transcription termination site, suggesting polymerase was distributed equally. On the other hand, RNAPII density increased progressively further downstream in H3 Δ1-28 cells, up to 2.5-fold higher than that observed at the promoter. Assuming that RNAPII is loaded onto the promoter, the pattern is consistent with the “piling up” of RNAPII across RNR3, suggesting that RNAPII is having difficulties completing transcription across the gene. The shift in RNAPII density could be explained by a defect in 3′-end formation or the appearance of cryptic transcripts within the gene. The data forthcoming in the manuscript predict that cryptic transcripts might arise in the mutants. However, we failed to detect intragenic transcripts originating from RNR3 in the H3 tail mutants, even when RRP6, a subunit of the exosome, was deleted (data not shown). This suggests that RNR3 does not harbor a cryptic promoter(s) capable of initiating intragenic transcription.
To provide further evidence for elongation defects we examined the sensitivity of the mutants to 6-AU, an inhibitor of transcription elongation (17, 48). The H3 Δ1-28 and H3 K→Q strains were spotted below their respective wild types. Sensitivity to 6-AU is thought to vary in different strain backgrounds, and therefore each mutant must be compared directly to its parent strain. Interestingly, instead of sensitivity, we found that both the H3 Δ1-28 and H3 K→Q strains displayed resistance to 6-AU relative to their cogenic wild-type strain (Fig. (Fig.1G).1G). The resistance to 6-AU was very obvious in the H3 K→Q mutant, which grew well on uracil-deficient medium but grew better than the wild type on medium containing 6-AU. While the growth of the H3 Δ1-28 mutant on medium containing 6-AU was similar to that of the wild-type strain, this strain displayed a significant slowed growth phenotype on medium lacking the drug (Fig. (Fig.1G).1G). Thus, the relative growth rate of the H3 Δ1-28 mutant on medium lacking versus medium containing elongation inhibitors indicates a resistance to 6-AU. set2Δ strains are among the few elongation mutants known to show resistance to 6-AU (4, 23, 25). Therefore, we compared the sensitivity of a set2Δ mutant constructed in the MSY590 genetic background and found it displayed a resistance level to the drug similar to the histone mutants (Fig. (Fig.1G1G).
Because we found that mutations to the H3 tail resulted in 6-AU resistance similar to that in a Δset2 strain, we sought to determine if mutating the N terminus of H3 affected Set2-mediated H3 K36 trimethylation (me3). In wild-type cells, we observed a transcription-dependent increase in H3 K36me3 levels toward the 3′ end of the open reading frame, where the levels peak over the region described as ORF D (Fig. (Fig.2A).2A). A similar pattern has been reported at other actively transcribed genes (1, 26, 44). Furthermore, the increase in H3 K36me3 was dependent upon SET2 (data not shown). Significantly, both the H3 Δ1-28 and H3 K→Q mutations caused a strong reduction in the level of H3 K36me3 across RNR3 (Fig. (Fig.2A).2A). These observations are not trivial, because the mutated residues in the histone tail analyzed here are relatively distant from the region of H3 expected to fill the active site of Set2. In fact, a histone peptide containing only amino acids 29 to 41 is capable of being methylated by Set2 (42). Furthermore, an H3 Δ1-20 mutant also resulted in a significant (~50%) reduction of K36me3 across RNR3 (data not shown) and displayed other phenotypes associated with the loss of K36me (see below). Thus, residues more than 15 amino acids away from K36 play a significant role in methylation by Set2.
H3 K36 dimethylation (me2) is also associated with active genes, although this mark is more broadly distributed across the genome and does not correlate as well with transcriptional frequency as the trimethlated form of K36 (1, 44). We examined K36me2 levels at RNR3 by ChIP. Here, too, we observed a significant reduction in K36me2 in both mutants over the ORF (ORFs B and D) of RNR3 under the induced condition (Fig. (Fig.2B).2B). Interestingly, K36me2 was reduced across the entire RNR3 locus in the H3 Δ1-28 mutant, including the promoter and UTR, but was reduced only in the ORF in the H3 K→Q mutant. The levels of K36me2 were unaffected in the promoter and UTR in the in the H3 K→Q mutant. This suggests that there may be a differential requirement for maintaining K36me2 in ORFs versus adjacent regions. The reason for this is unclear at this time. Nonetheless, the data show that mutating the H3 tail affects K36me2 within the ORFs of genes (also see below). The quality and selectivity of the commercial K36me1 antibodies prevented us from examining this modification.
Acetylation of K36 (H3 K36Ac) is another highly conserved histone modification (40). Mutation of the H3 tail could lead to an increase in K36Ac due to reduced HDAC recruitment, which would indirectly reduce K36me levels. The levels of H3 K36Ac over the promoter and ORF D of RNR3 were examined by ChIP. A DNA damage-dependent increase in this modification was detected over the promoter in wild-type cells (Fig. (Fig.2C).2C). Gcn5 is the histone acetyltransferase predominantly responsible for this modification (40), and SAGA acetylates nucleosomes over the RNR3 promoter (64). Interestingly, the level of K36Ac is constitutively high in untreated cells, and inducing the gene led to its reduction within the ORF (Fig. (Fig.2C).2C). The reduction in acetylation is likely caused by the methylation of K36 by Set2 during transcription (Fig. (Fig.2A).2A). Deleting GCN5 reduces the level K36Ac in the ORF to levels slightly below that observed after activation, indicating it is responsible for writing this mark (not shown). The reduction of K36Ac levels and accompanying increase in methylation suggest that the acetyl mark must be removed prior to Set2 methylation. A likely candidate is Rpd3, although this was not examined. The pattern in the mutants is complicated but supports the notion that increased K36Ac is not the cause of the reduced methylation (Fig. (Fig.2C).2C). Generally, deleting the tail or mutating the lysine residues led to a decrease in the level of acetylation at both the promoter and ORF of the gene in the mutants. We speculate that the reduction in acetylation may be due to reduced SAGA recruitment or activity. SAGA may interact with the tail of H3 through the bromodomains within Gcn5 or Spt7 (20), which would be disrupted in the mutants.
In order to determine if the reduction in K36me3 is specific for RNR3, we examined methylation levels at PMA1 and PYK1, two constitutively expressed genes that have been used extensively as models to study cotranscriptional histone modifications (9, 26). As reported previously, there is a high level of H3 K36me3 within the 3′ ORF regions of both PMA1 and PYK1 in the wild-type strain (Fig. (Fig.3A).3A). Deletion of the H3 N terminus resulted in a significant decrease in H3 K36me3 within the ORFs of PMA1 and PYK1. Thus, the same phenotype is observed at both constitutively transcribed and highly induced genes. This indicates that the methylation defect observed in the H3 mutants may be a global phenomenon. To test if H3 K36me3 is reduced across the genome, we examined its levels in chromatin from wild-type, H3 Δ1-28, H3 K→Q, H4 Δ2-26 (control), and set2Δ strains by Western blotting. The amount of total H3 was measured using an antibody recognizing the C terminus of H3. Consistent with the ChIP results, K36me3 levels were lower in the H3 tail mutants than in wild-type cells (Fig. (Fig.3B).3B). As a control, we analyzed a strain containing a deletion of the H4 tail. The H4 Δ2-26 mutant showed no defect in H3 K36me3, suggesting specificity for the requirement for the N terminus of H3 in Set2-mediated methylation. The levels of K36me2 in bulk chromatin were examined by Western blotting. Surprisingly, we observed a small, but reproducible, reduction in K36me2 in the H3 Δ1-28 mutant, but no detectable decrease in the H3 K→Q mutant, after correcting for the load of histones (Fig. (Fig.3C).3C). The reductions in K36me2 observed in bulk chromatin by Western blotting were smaller than those observed in the ChIP assays at active genes (Fig. (Fig.2B).2B). K36me2 is more widely distributed throughout the genome, and it is possible that the reductions in K36me2 are restricted to the ORFs of active genes. Thus, blotting for bulk histones may underestimate the reduction in methylation over regions where K36me2 are the highest.
We also explored the possibility that mutations to the H3 N terminus would affect other histone methylation marks associated with elongation. Because the mutants lacked K4 or had a glutamine substitution at that residue, this precluded us from analyzing K4 methylation. However, we examined H3 K79 methylation by Western blotting and found that mutations to the N terminus of H3 did not significantly affect K79me2 levels (Fig. (Fig.3D).3D). This result suggests that the H3 tail plays a role specifically in Set2 methylation.
K36 methylation is required to recruit the Rpd3S HDAC complex, which in turn deacetylates nucleosomes within ORFs of genes (6, 22, 23, 30). Deleting SET2, or other effectors of this pathway, results in intragenic transcription within certain genes, including STE11 (6). We have already determined that deleting components of the Rpd3S complex or mutating the H3 tail does not produce cryptic transcripts within RNR3 (reference 53 and data not shown). This is not surprising, because a genome-wide analysis revealed that not all genes are capable of producing cryptic transcripts (8, 31). Such transcripts cannot arise unless there are sequences within the ORF that can serve as a promoter, and it is not expected that every gene will produce a cryptic transcript. Thus, we examined the RNAs produced from STE11 by Northern blotting in the mutants to determine if the reduction in K36 levels has functional consequences on the cell (Fig. (Fig.4A).4A). Deletion of SET2, EAF3 (a chromodomain-containing component of the Rpd3S complex), or CTK1 (RNAPII CTD Ser2 kinase required for Set2 activity) all resulted in short intragenic STE11 transcripts, as expected. This phenotype was also observed in the H3 Δ1-28 and H3 K→Q strains, as well as in a strain containing a smaller deletion of the N terminus, H3 Δ1-20, but to a lesser degree (Fig. (Fig.4A).4A). Deletion of SET2 or CTK1 results in a six- to sevenfold increase in the ratio of short to long transcripts, while each of the H3 tail mutations resulted in an approximately three- to fourfold increase. The intermediate levels of intragenic transcription in the H3 mutants are presumably due to the fact that these mutations do not completely abolish Set2 methylation (Fig. (Fig.2).2). This further emphasizes the role of the H3 N terminus in Set2 methylation and indicates that the loss of methylation in these mutants has functional consequences. Deletion or mutation of the histone H4 does not cause intragenic transcription at STE11, once again suggesting that the effects we observed on K36me are specific to mutation of the H3 N-terminal tail. Furthermore, even though recruitment of Rpd3S to genes by K36me reduces H4 acetylation (6, 30), in the absence of the H4 tail, others tails on the nucleosome can suppress intragenic transcription at STE11.
In order to confirm that the intragenic transcripts arise due to a loss of H3 K36 methylation, we examined the patterns of H3 K36me3 and me2 modification at STE11 (Fig. (Fig.4B,4B, primer schematic). Recently it was shown that H3 K36me2, in the absence of H3 K36me3, is sufficient to recruit the Rpd3S complex and repress intragenic transcription (33, 63), so this mark was also examined. Both H3 mutants displayed less than 50% of the levels of H3 K36me3 and K36me2 at the 3′ ORF compared to wild-type cells (Fig. 4C and D). Furthermore, since K36me2 is sufficient to suppress intragenic transcription, the production of intragenic transcripts in the H3 tail mutants provides another line of evidence that mutating the histone H3 tail affects K36me2.
We next turned to determining the mechanism of how the H3 tail regulates Set2 activity. Set2 is recruited to genes by the phosphorylation of Ser2 (S2) on the CTD of RNAPII by Ctk1, resulting in the cotranscriptional methylation of histone H3 (25, 32, 61). The N-terminal tail of H3 could play a role in Ctk1-mediated phosphorylation of RNAPII, which would then result in reduced recruitment of Set2. This was tested by measuring the levels of S2 phosphorylation of RNAPII at RNR3 (Fig. (Fig.5A).5A). Upon induction with MMS, an increase in S2 phosphorylation was observed at the 3′ end of RNR3 (ORF D and the 3′ end). The peak of S2 modification at ORF D is consistent with reports that it occurs predominantly in the middle and 3′ ends of genes. The low level of S2 phosphorylation observed near the promoter of RNR3 is likely due to cross-reactivity of the antibody with differently modified forms of RNAPII, because the same level of cross-linking was observed in a Δctk1 strain (Fig. (Fig.5A,5A, right panel). In contrast, the increase in the MMS-induced level of S2 phosphorylation at the 3′ end of RNR3 was strongly reduced in the ctk1Δ mutant. Neither the H3 Δ1-28 nor the H3 K→Q mutations reduced the level of S2 phosphorylation (Fig. (Fig.5A,5A, left and middle panels, respectively). Thus, the lack of K36 methylation is not caused by reduced Ser2 phosphorylation of the CTD of RNAPII.
It is possible that the N terminus of H3 plays a role, either direct or indirect (such as the recruitment of the PAF complex), in the recruitment of Set2 to genes. In order to investigate this possibility, we examined the recruitment of Set2 (Set2-myc) to RNR3 and PYK1 (Fig. 5B and C, respectively). In the wild-type strain, Set2-myc recruitment increased almost threefold upon induction of transcription at ORF D, the same location of the peak of K36me3 (compare Fig. Fig.5B5B and and2A).2A). We also observed recruitment of Set2 to ORF B and the 3′ end of RNR3, but to a lesser degree. Importantly, neither the H3 Δ1-28 nor the H3 K→Q mutations reduced Set2 recruitment, and in fact, Set2 cross-linking was actually slightly higher in the H3 mutants than in the wild-type strain. Furthermore, Set2-myc recruitment to PYK1 was equal in wild-type and mutant cells (Fig. (Fig.5C).5C). These data indicate that the N terminus of H3 is not required for Set2 recruitment, suggesting that the tail regulates Set2 activity at a postrecruitment step. The data also suggest that mutating the H3 tail does not affect Set2 levels within the cell.
The PAF complex interacts with the CTD of RNAPII and regulates both H3 K4 and K36 methylation (9, 27). Deletion of PAF complex components PAF1 or CTR9 nearly eliminates H3 K36me3 and significantly reduces K36me2 at genes, presumably by reducing Set2 recruitment (9, 41). While we did not observe a Set2 recruitment defect in the H3 mutants, it is possible that the PAF complex regulates Set2 at a postrecruitment step as well, such as by stimulating its activity. We examined the recruitment of an epitope-tagged version of Paf1 to RNR3 (Fig. (Fig.5D).5D). In the wild-type strain, Paf1-myc recruitment increased ~5-fold at ORF B and ~3-fold at the RNR3 promoter, ORF D, and the 3′ end under the induced condition (+MMS). Interestingly, the peak of Paf1-myc recruitment occurred at ORF B, a region where H3 K4me3 and K36me3 methylation overlap considerably upon induction of RNR3 (Fig. (Fig.5D,5D, ,2A,2A, and unpublished data). Neither the H3 Δ1-28 nor the H3 K→Q mutations affected the recruitment of Paf1-myc to most regions of RNR3 (Fig. (Fig.5D,5D, left and right panels, respectively). At ORF B, Paf1-myc recruitment was ~50% less than the wild type in the H3 Δ1-28 mutant and ~35% less in the H3 K→Q mutant. However, since Paf1 recruitment to ORF D, which is the location of the peak of H3 K36 methylation and Set2 recruitment, was not affected in the H3 mutants, the defect in K36 methylation is not likely to be caused by reduced PAF complex recruitment to genes.
Spt6 is another elongation factor recently identified as a regulator of Set2 activity. Inactivation of the conditional spt6-1004 mutant results in a loss of H3 K36me3 (63). Cellular levels of Set2 are reduced in the spt6-1004 mutant (63); however, restoration of Set2 levels by overexpression does not rescue K36me3 (63). This suggests a second role for Spt6 in regulating Set2 activity, in addition to its ability to regulate its levels in the cell. Thus, it is possible that mutating the H3 tail indirectly affects Set2 activity by reducing Spt6 recruitment to genes. The cross-linking of Spt6-myc to RNR3 was examined before and after induction with MMS, and its recruitment increased across RNR3 upon induction with MMS (Fig. (Fig.5E).5E). The largest increase (~8-fold) was observed at the ORF B region (Fig. (Fig.5E).5E). Neither the H3 Δ1-28 nor H3 K→Q mutations adversely affected Spt6 recruitment to RNR3, and the pattern of Spt6 recruitment in the H3 mutant strains mirrored that of the wild type. Therefore, the reduced levels of K36me in the H3 mutants cannot be explained by impaired Spt6 recruitment. The experiments described in Fig. Fig.55 tested all known factors that regulate Set2 activity in vivo, and we found these parameters to be intact. These observations suggest that the H3 tail regulates Set2 activity through a novel mechanism.
The data presented above suggest that mutating the H3 N terminus affects Set2 activity at a postrecruitment step, possibly by directly modulating its activity. For instance, the N terminus of H3 might be required to stimulate the enzymatic activity of Set2 or for the ability of Set2 to bind or recognize nucleosomes. These possibilities were tested using an in vitro histone methyltransferase assay with a recombinant SET domain of Set2 (Fig. 6A and B) and either wild-type or mutant versions of recombinant Xenopus nucleosomes. Histone methyltransferase activity was measured by the incorporation of [3H]SAM into nucleosomes, and the data show that Set2 was significantly more active on wild-type nucleosomes versus the version lacking the H3 N terminus (Fig. (Fig.6B).6B). Importantly, nucleosomes lacking the H4 tail were at least as good of a substrate for Set2 as wild-type nucleosomes, which correlates well with other data showing that K36 methylation is unaffected by a deletion of the H4 tail in vivo (Fig. (Fig.3B3B and and4A4A and data not shown).
The reduced activity of Set2 on the mutant nucleosomes can be caused by reduced binding of the enzyme to the substrate or by reduced catalytic activity. In the former case, Set2 should interact better with wild-type nucleosomes than H3 mutant nucleosomes. Pull-down assays were performed using immobilized Set2-FLAG protein and recombinant nucleosomes, and the bound fraction was detected by Western blotting. Both wild-type and H3 Δ1-26 nucleosomes were pulled down by Set2-FLAG, but not FLAG resin alone (Fig. (Fig.6C).6C). Interestingly, the results suggest that the H3 Δ1-26 nucleosomes interacted more strongly with Set2-FLAG than the wild-type nucleosomes (Fig. (Fig.6C,6C, compare lanes 3 and 4). To further confirm this observation, the assays were repeated while titrating the amount of nucleosome added to each pull-down mixture. These results also support the conclusion that Set2 pulled down more H3 Δ1-26 nucleosomes than wild-type nucleosomes (Fig. (Fig.6D,6D, compare lanes 1 to 3 with 4 to 6). Thus, the mutation of the H3 N terminus does not affect the binding of Set2 to nucleosomes, and this finding further suggests that mutation of the tail affects the catalytic activity of Set2.
The enhanced binding of the mutant nucleosomes to Set2 was unexpected, and we hypothesized that it resulted from the reduced activity of Set2 on these nucleosomes. Methylation of the tail may weaken the binding of Set2 to the newly modified nucleosome, and because the mutant nucleosomes are modified less well, they would bind better to Set2. Even though exogenous SAM was not added to the binding reaction mixtures, Set7/9 methyltransferase copurifies with SAM when expressed in E. coli (10, 59). Copurification with SAM with rSet2 would cause nucleosome methylation and reduced binding to the substrate. If true, we expect that a catalytically inactive version of Set2 would bind nucleosomes better than wild-type Set2. Furthermore, the enhanced binding of Set2 to the mutant nucleosomes, relative to wild-type nucleosomes, would be lost because the wild-type nucleosomes could not be modified. We repeated the assays using a catalytically inactive (R195G) version of Set2 that had been characterized previously (54). The R195G Set2-FLAG interacts equally well with both wild-type and H3 Δ1-26 nucleosomes (Fig. (Fig.6E).6E). Importantly, the equal binding is not caused by a reduction in binding of Set2 R195G to the mutant nucleosomes, but rather from an increase in the binding of the R195G mutant to the wild-type nucleosomes (Fig. (Fig.6E,6E, top panel, lanes 1 to 6, compared with D, lanes 1 to 6). This suggests that the binding of Set2 to nucleosomes is inversely correlated to its ability to modify K36 and that the turnover of nucleosomes by Set2 may be important for the processivity of the elongation complex during transcription. These results strongly suggest that the H3 N terminus directly stimulates the activity of Set2.
Substituting glutamines for lysines greatly reduces H3 K36me3 in vivo. Glutamine mutations have been used to mimic the charge effects of lysine acetylation and have been shown to have similar effects on chromatin folding in vitro as acetylation (58). This raises the possibility that the charge of the H3 tail may be important for Set2 methylation, or alternatively, mutating six lysine residues causes the nucleosome to be unrecognizable by Set2. These possibilities can be discerned by analyzing the methylation levels in strains containing arginines substituted for the lysine residues. Bulk levels of K36me3 were examined in a strain containing all six lysines replaced by arginines (Fig. (Fig.7A).7A). Unlike the H3 K→Q mutant, the H3 K→R mutant does not result in a noticeable reduction in global H3 K36me3 levels (compared Fig. Fig.3B3B and and7A).7A). In fact, methylation may be slightly enhanced.
Since the effects of the H3 tail mutants on K36me were most apparent within the ORF of genes, we examined the levels of K36me3 at specific genes by ChIP. In this case, we normalized the level of K36me3 to the amount of RNAPII at the gene because the H3 K→R mutant displays somewhat less RNAPII at RNR3, PYK1, and PMA1 (data not shown). The reduction in RNAPII occurs over the promoter, indicating that it is not related to the loss of K36me3 (data not shown). Since Set2 is recruited to genes through an interaction with the RNAPII CTD (25, 32), normalization to RNAPII levels provides a better assessment of the effects of the tail mutants on K36me3. This was repeated for the K→Q in parallel. When H3 K36me3 was plotted relative to RNAPII density, the H3 K→Q mutant clearly exhibited a K36me3 defect, equal to that when not normalized to RNAPII (Fig. 7B, C, and D). However, the H3 K→R mutant results in a mild enhancement of Set2 methylation at RNR3, PYK1, and PMA1 (Fig. 7B, C, and D). To provide additional evidence that the K→R mutant does not cause reductions in K36me3 similar to the glutamine substitution or deletion mutants, we examined the 6-AU sensitivity and the appearance of intragenic transcription in the strain. The results show that although the mutant grows more slowly on all media tested, the relative growth compared to wild-type cells on media lacking or containing 6-AU was the same, indicating that the mutant strain is no more resistant to the drug (Fig. (Fig.7E).7E). Finally, Northern blotting for STE11 showed that the H3 K→R strain does not exhibit intragenic transcription like the H3 K→Q and Δset2 strains (Fig. (Fig.7F).7F). There is a small increase in the larger of the two short intragenic products, but the amount of intragenic transcription is clearly less. The small increase in the shorter transcript could be caused indirectly by mechanisms independent of K36me3. Collectively, the data support the hypothesis that the charge of the tail is important for Set2 activity and, importantly, that the glutamine mutations are not indiscriminately making the tail unrecognizable.
Intratail regulation of histone modifications has been described. The existing paradigms, however, involve the control of one modification by the modification state of another residue located within close proximity and within the active site of the enzyme (19, 35, 42). Here we describe a novel mechanism where H3 K36me3 is regulated by residues of H3 predicted to be located outside of the catalytic site of Set2. The crystal structure of most SET domains with histone peptides reveal no more than 9 amino acids total within the active site of the protein, and contacts with residues on each side of the modification site are observed (12). More relevant to our data, Set2 can methylate a histone peptide containing only amino acids 29 to 41 (42). This suggests that the reduction in Set2 activity observed in the H3 mutants examined here is not merely due to an alteration of the interaction of residues within the active site of Set2. Thus, to our knowledge, the stimulatory effect of the H3 tail on Set2 activity is unique to these other forms of intratail regulation. It is certain that the H3 tail plays multiple roles in transcription activation and they are not restricted to regulating K36me. Deleting SET2 had a small effect on the activation of RNR3, less than that of the tail mutants (not shown). Another function suggested by data presented here is regulation of the recruitment or activity of SAGA. Future studies will be required to understand all of the roles of the H3 tail in the transcription of RNR3.
Our data are fully consistent with the tail affecting Set2 catalytic activity. Deleting the tail had no effect on the interaction between Set2 and nucleosomes, once the differences in the turnover of nucleosomes were corrected for by using the catalytic mutant of Set2 (Fig. (Fig.6E).6E). The binding of the tail to Set2 could affect its activity through an allosteric mechanism, or it could position K36 within the active site of Set2. In the latter case, the H3 N terminus interacts with Set2 to position lysine 36 next to the catalytic residues or restricts movement of the tail within the active site of Set2 (Fig. (Fig.8,8, left panel). Deleting the H3 tail, or mutating the lysines to neutral glutamines, could result in the misalignment of K36 within the active site, reducing the modification of the residue (Fig. (Fig.8,8, middle and right panels, respectively). We favor this scenario, versus the allosteric mechanism, based on the previously described regulation of K36me by the isomerization of an adjacent proline residue, P38, by the prolyl-isomerase Frp4 (42). According to this model, Set2 can methylate lysine 36 only when P38 is in the trans conformation, which orients K36 in the active site. An unresolved question about the proposed prolyl isomerization switch is what favors the active isomer versus the inactive form of P38. Prolyl isomerization is highly reversible and subject to equilibrium conditions. Interactions between the H3 tail and Set2, which could be regulated by modification or cleavage by cellular proteases, could favor the active conformation.
The region within Set2 that interacts with the H3 tail is not yet fully characterized. We observed that the SET domain of Set2 (amino acids 1 to 261) has greater activity on wild-type nucleosomes than those lacking the H3 tail (Fig. (Fig.6B).6B). From this, we conclude that the intratail regulatory interaction occurs within the first 261 amino acids of Set2. Since deleting the tail or mutating the lysines to glutamine results in similar phenotypes, this suggests that the charge of the H3 tail is important for mediating the interaction with Set2. Accordingly, we show that lysine to arginine mutations do not negatively affect K36me3 in vivo. The pI of the SET domain of Set2 (amino acids 1 to 261) is 4.94, and it is possible that an acidic patch on Set2 could act as the interaction partner for the basic H3 tail. Determining the structure of the SET domain of Set2 could provide further evidence for this, and allow for the direct testing of this model, but unfortunately this has not been achieved yet.
An intriguing possibility is that modification of residues within the tail of H3, such as acetylation, can regulate the levels of K36me3. Histone H3 acetylation, found predominantly at promoter regions where H3 K36 methylation is low (29), also neutralizes the charge of H3 lysine residues similar to the glutamine mutations. It is possible that acetylation of H3 helps create a barrier between the promoter and ORF regions by limiting H3 K36 methylation. This might be especially important when the end of one gene is in close proximity to the promoter of another gene. Alternatively, cotranscriptional acetylation within genes may serve to dampen Set2 activity ahead of the elongating polymerase. Histone H3 acetylation and Set2 methylation may negatively regulate the activity of each other to strike the proper balance of histone modifications. We attempted to directly test this model in vivo by deleting GCN5, the histone acetyltransferase responsible for the DNA damage-induced H3 acetylation at RNR3 (49), and examining the change in H3K36me3 patterns and levels at RNR3. While we did observe an increase in H3 K36me3 relative to RNAPII density in the gcn5Δ mutant at RNR3, we also observed a loss of H3 K36Ac as well (data not shown). Thus, we could not exclude the possibility that the increase in H3 K36me3 is the result of more “methylatable” K36 residues due to the loss of K36Ac.
The histone tail deletion mutants used here mimic a naturally occurring phenomenon in eukaryotic cells. The programmed “clipping” of the H3 tail at residue 21 occurs in vivo in S. cerevisiae within nucleosomes at the promoters of activated sporulation control genes (51). While it has not been described yet, tail cleavage could occur within other regions of the genome. Our data suggest a consequence of the removal of the H3 tail, reduced Set2 methylation. Clipping the H3 tail will inhibit Set2-dependent K36me, a mark whose function is to reset and repress chromatin by the recruitment of HDACs. Reducing K36me by tail cleavage could increase the expression of certain genes. This is not occurring at RNR3, as expression of this gene is impaired in the tail mutant. As mentioned above, the H3 tail may be playing multiple roles in transcription at RNR3. Furthermore, cleavage of the H3 tail has been identified in mouse embryonic stem cells during differentiation (15). While the purpose of this cleavage is not well understood, the authors found that histone acetylation inhibited H3 proteolysis. Given that histones are found to be mostly deacetylated in differentiating cells, cleavage of the H3 tail may provide another mechanism to regulate lysine 36 methylation patterns during differentiation or under circumstances when changes in acetylation alone are insufficient. The novel intratail regulation described here provides insight into the depth of the regulatory mechanisms controlling histone modifications and may reveal a consequence of the recently described programmed proteolysis of H3.
We thank members of the Reese Lab and the Center for Eukaryotic Gene Regulation at Penn State for advice and comments on this work. We are grateful to Mitch Smith and LeAnn Howe for the yeast tail mutants and plasmids. Brian Strahl is acknowledged for providing the Set2 expression plasmids. We thank Bing Li for comments on the manuscript.
This research was supported by funds from National Institutes of Health (GM58672) to J.C.R.
Published ahead of print on 12 October 2009.