|Home | About | Journals | Submit | Contact Us | Français|
Asf1 is a conserved histone H3/H4 chaperone that can assemble and disassemble nucleosomes and promote histone acetylation. Set2 is an H3 K36 methyltransferase. The functions of these proteins intersect in the context of transcription elongation by RNA polymerase II: both contribute to the establishment of repressive chromatin structures that inhibit spurious intragenic transcription. Here we characterize further interactions between budding yeast (Saccharomyces cerevisiae) Asf1 and Set2 using assays of intragenic transcription, H3/H4 posttranslational modification, coding region cross-linking of Asf1 and Set2, and cooccurrence of Asf1 and Set2 in protein complexes. We find that at some genes Asf1 and Set2 control chromatin metabolism as components of separate pathways. However, the existence of a low-abundance complex containing both proteins suggests that Asf1 and Set2 can more directly collaborate in chromatin regulation. Consistent with this possibility, we show that Asf1 stimulates Set2 occupancy of the coding region of a highly transcribed gene by a mechanism that depends on Asf1 binding to H3/H4. This function of Asf1 promotes the switch from di- to trimethylation of H3 K36 at that gene. These results support the view that Set2 function in chromatin metabolism can intimately involve histone chaperone Asf1.
Transcriptional regulation in eukaryotes critically involves the regulation of histone-DNA complexes. Two conserved proteins that modulate the properties of histone-DNA complexes in budding yeast (Saccharomyces cerevisiae) are Asf1 and Set2. Asf1 and Set2 have quite distinct biochemical activities. Asf1 is a histone chaperone that binds tightly to the dimer formed by histones H3 and H4 (13, 35). It can deposit H3/H4 onto DNA and remove H3/H4 from nucleosome core particles (35, 44). Asf1 also modulates the enzymatic activity of some histone-directed lysine acetylases (KATs). It stimulates Rtt109 acetylation of H3 K56 (10, 20) by a mechanism that requires Asf1 binding to H3/H4 (21). Asf1 can also inhibit acetylation: this activity has been reported in the context of H3/H4 acetylation by the SAS complex (49). In contrast to Asf1, Set2 is an enzyme. It catalyzes mono-, di-, and trimethylation of H3 K36 in nucleosome core particles (11, 16, 47).
Since Asf1 and Set2 regulate the properties of histone-DNA complexes and these properties affect all steps in the transcription cycle of nuclear genes, it is not surprising that Asf1 and Set2 have both been implicated in the control of transcriptional activation, initiation, and elongation by RNA polymerase II (RNAP II). These regulatory functions of Asf1 and Set2 are best understood in budding yeast. Asf1 potentiates early steps in the transcription cycle by stimulating H3 K56 acetylation, which destabilizes nucleosomes and makes them easier to evict (55). Set2 somehow modulates promoter binding by TATA binding protein (TBP) and transcription factor IIA (TFIIA), although the precise mechanism remains unknown (4). As far as we are aware, there is no evidence that Asf1 and Set2 function in the same pathway of nucleosome metabolism in promoters and cis-acting control regions.
The critical function of Asf1 during transcription elongation is to disassemble nucleosomes ahead of elongating RNAP II and reassemble them behind it (42). By virtue of these two activities, Asf1 promotes elongation and reestablishes a chromatin architecture that disfavors spurious initiation within coding regions.
In contrast, Set2 has a single known function in coding regions: it inhibits elongation by catalyzing methylation of H3 K36. Although Set2 catalyzes mono-, di-, and trimethylation of H3 K36, dimethylation is fully sufficient for inhibition of elongation by Set2 (32, 59). This inhibition is a stepwise process. First, Set2 dimethylates K36 of nucleosomal H3. Second, the Rpd3S histone deacetylase (HDAC) complex binds to H3 K36-methylated nucleosomes (6, 31). Third, deacetylation of nucleosomes by Rpd3S makes chromatin less permissive for elongation by RNAP II. Both Asf1 and Set2 can modulate the properties of chromatin in a way that inhibits transcription during elongation. By virtue of this activity, both proteins are able to dampen spurious transcription from cryptic promoters in coding regions (7, 42). Whether Asf1 and Set2 can collaborate in this regulation, either as part of the same pathway or converging pathways, remains unknown.
The regulation of Set2-dependent H3 K36 trimethylation is just beginning to be understood. The available evidence suggests that H3 K36 trimethylation of chromatin is directly coupled to transcriptional elongation, since it requires the following: (i) the C-terminal domain (CTD) of RNAP II, (ii) an enzyme that phosphorylates the CTD during elongation (Ctk1), and (iii) a domain in Set2 that interacts with the CTD (59). Set2 binding to H4 (11) and rotation of H3 around proline 38 (36, 59) are also important for H3 K36 trimethylation.
In this report, we further examine the functions of Asf1 and Set2 in the coding regions of active genes. Our results suggest that Asf1 can function separately from Set2 in chromatin regulation during elongation or directly modulate the activity of Set2.
Except as noted under “Synthetic genetic analysis” below, all strains used are derived from S. cerevisiae BY4741 (MATa his3Δ1 leuΔ0 met15Δ0 ura3Δ0) (5). Chromosomal mutations were generated by one-step integration using PCR products obtained from previously described plasmids (18, 33, 45). The addition of sequences encoding the myc and hemagglutinin (HA) epitope tags was verified by PCR using three primer sets (flanking the target gene, internal to the target gene, and flanking primer plus primer specific for the marker; sequences available upon request). The asf1Δ::kanMX, set2Δ::kanMX, eaf3Δ::kanMX, and rtt109Δ::kanMX deletion mutants from the S. cerevisiae haploid nonessential gene deletion library (56) were verified to be correct by PCR. The asf1Δ::kanMX set2Δ::natr, asf1Δ::kanMX rpd3Δ::HIS3MX6, asf1Δ::natr eaf3Δ::kanMX, asf1Δ:: kanMX sds3Δ:: natr, and rtt109Δ::kanMX set2Δ::natr double mutants were used. The rpd3Δ::HIS3MX6 and sds3Δ:: natr deletion mutants were constructed by R. Friis in the laboratory of M. C. Schultz. The strains expressing Set2-TAP and Asf1-TAP are described by Ghaemmaghami et al. in reference 17. pRS314-based and pRS316-based plasmids harboring ASF1 and asf1V94R (34) were transformed into an asf1Δ set2Δ strain for Northern blotting analysis and into a SET2-TAP asf1Δ strain for chromatin immunoprecipitation (ChIP) analysis. A strain expressing HA-tagged Asf1V94R (Asf1V94R-HA) from its normal chromosomal location was constructed in multiple steps by H. Mewhort. Briefly, the asf1V94R mutation was created using site-directed mutagenesis and propagated as an EcoRI/XhoI insert in pGEX6P-1. The correct GT → CG mutation at 280 bp of ASF1 was confirmed by sequencing. A cassette containing (i) asf1V94R with sequence encoding the 3HA epitope tag at its 3′ end and (ii) the kanMX G418 resistance marker was then assembled in Escherichia coli plasmid pHM3. PCR was used to amplify this cassette and appropriate targeting sequences up- and downstream of the ASF1 open reading frame (ORF). The PCR product was transformed into S. cerevisiae BY4741. Successful incorporation of both the V94R mutation and the 3HA epitope was confirmed by genomic sequencing and immunoblotting, respectively. Further details are available on request. All media were prepared as described previously, and standard genetic methods for mating, sporulation, transformation, and random spore analysis were used throughout this study (53). For testing sensitivity to 6-azauracil (6-AU), strains were transformed to Ura+ (40) using plasmid pRS316 (45). They were spotted onto medium lacking uracil and containing 25 to 100 μg of 6-AU per ml. Sensitivity to hydroxyurea (HU) and methane methylsulfonate (MMS) was tested by spotting strains onto medium containing 50 and 100 mM HU and 0.005 and 0.01% MMS.
Automated synthetic genetic analysis (SGA) was performed by the method of Tong et al. (52) using an asf1Δ query strain (Y2454 background) and the yeast haploid gene deletion collection. To generate the query strain, ASF1 was replaced with a nourseothricin resistance cassette (18) in strain Y2454 (MATα mfa1Δ::MFA1pr-HIS3 can1Δ ura3Δ0 leu2Δ0 his3Δ1 lys2Δ0) using homologous recombination. Oligonucleotides bearing sequences flanking the ASF1 coding region and plasmid pAG25 as a template (52) were used to generate the deletion cassette by PCR amplification. Correct replacement of ASF1 was verified by PCR analysis.
Cellular DNA content was determined by flow cytometry as described previously (39). Briefly, cells were stained with propidium iodide, sonicated, and analyzed using a FACScan flow cytometer (Becton-Dickinson).
Total proteins were prepared by trichloroacetic acid precipitation (39). Identical cell equivalents of protein in samples were compared. Antibodies that recognize total histone H3 (ab1791), monomethylated histone H3 K36 (ab9048), and trimethylated histone H3 K36 (ab9050) were obtained from Abcam, and antibodies that recognize calmodulin binding protein (CBP) (07-482), dimethylated histone H3 K36 (07-369), acetylated H3 K9 (07-352), acetylated H3 K14 (07-353), and tetra-acetylated H4 (06-866) were obtained from Upstate. The antibody against actin (MAB 1501) was obtained from Millipore. The 12CA5 antibody for detecting the HA epitope was from Roche. Quantitative analysis of total H3 trimethylated K36 (K36me3) was performed using ImageJ 1.38x after optical scanning of appropriately exposed films. H3 K36me3 signals were normalized to bulk H3 recovery, which was assessed by analysis of blots probed with the total histone H3 antibody (ab1791).
Yeast strains were grown at 30°C in yeast extract-peptone-dextrose (YPD) to a concentration of 1 × 107 cells per ml. Chromatin immunoprecipitation was performed essentially as described previously (3), using antibodies that recognize the myc tag (9E10) or total H3 (ab1791) or are specific for dimethylated H3 K36, trimethylated K9/K14 acetylated H3 (06-599; Upstate), tetra-acetylated H4 (K5, K8, K12, and K16), or K16 acetylated H4 (07-329; Upstate). Set2-TAP was recovered from sonicates using IgG-Sepharose 6 Fast Flow (17-0969-01; GE) based on the IgG-agarose protocol in reference 28. Immunoprecipitation was performed overnight at 4°C. Input and immunoprecipitated DNA was purified using Qiagen PCR purification columns and analyzed by standard PCRs to which 1 μCi of 32P-labeled dCTP had been added. Primers used for ChIP assays are described in reference 8. A region proximal to the telomere of chromosome 5 (TELV) was used as a control (8). PCR signals were captured using a Storm 840 phosphorimager and quantified using ImageQuant software (Molecular Dynamics). For localization of Set2-TAP, Asf1-myc, and H3, ChIP was quantified by normalizing band intensities for each sample using the following calculation as described in reference 6: (specific gene IP/TELV IP)/(specific gene input/TELV input). The normalized ChIP value calculated for H3 dimethylated K36 (K36me2), H3 K36me3, H3 acetylated K9 or acetylated K14 (K9ac/K14ac), or acetylated H4, was divided by the normalized ChIP value calculated for total H3. Statistical significance was assessed by applying Student's unpaired (independent) two-tailed t test for three independent samples of each strain under consideration.
Tandem affinity purification of protein complexes was performed using extracts from the indicated strains essentially as described previously (27). Proteins stained with Sypro Ruby (Invitrogen) were in-gel digested with trypsin. Spectra were obtained by matrix-assisted laser desorption ionization-time of flight (MALDI-ToF) mass spectrometry and analyzed using Mascot software in the Institute for Biomolecular Design at the University of Alberta.
Total RNA was isolated by hot phenol extraction (14) from cells grown in YPD to a concentration of 1 × 107 cells per ml. DNA probes were prepared by random primed labeling of PCR products for FLO8 (+1672/+2399), SPB4 (+1100/+1820), ACT1, and SCR1.
In a high-throughput genetic screen, we uncovered a synthetic sick interaction between ASF1 and SET2, which was confirmed by random spore analysis (Fig. (Fig.1A)1A) (this interaction has been reported by others ). We also found that an asf1Δ set2Δ strain generated by successive one-step gene replacements grows at a much lower rate than either single mutant at 30°C and that this phenotype is slightly exacerbated at 37°C (Fig. (Fig.1B).1B). These results prompted us to examine in more detail the nature of functional interactions between Asf1 and Set2.
We used flow cytometry to assess the impact of individual and combined deletion of ASF1 and SET2 on cell cycle progression (Fig. (Fig.1C).1C). Consistent with previous work (54), we find that asf1Δ cells accumulate with a G2/M content of DNA and take longer than wild-type cells to complete S phase. In contrast, the flow cytometry profile of set2Δ cells is very similar to the profile of wild-type cells. Furthermore, the cell cycle profile of the asf1Δ mutant is unaffected by deletion of SET2. Therefore, Asf1 and Set2 do not perform the same function in parallel pathways of chromatin regulation that affect cell cycle progression under normal culture conditions.
Because the regulation of chromatin structure is important for protection from DNA damage, we tested whether Asf1 and Set2 both function in pathways that protect cells from DNA structure abnormalities. As previously reported (54), loss of ASF1 enhances the sensitivity of yeast to two genotoxins, hydroxyurea (HU) and methane methylsulfonate (MMS) (Fig. 2A and B). SET2 null mutants, on the other hand, grow similarly to wild-type cells in the presence of these agents. Furthermore, the sensitivity of the asf1Δ set2Δ double mutant to HU and MMS is only marginally greater than that of the asf1Δ single mutant. Therefore, Set2 is probably not in a major Asf1-dependent pathway of chromatin metabolism that protects cells from genotoxin-induced DNA structure abnormalities. This conclusion is supported by the observations that set2Δ cells are more sensitive to UV irradiation than asf1Δ cells are and that the asf1Δ set2Δ double mutant is no more sensitive to UV than the set2Δ single mutant is (Fig. (Fig.2C2C).
Like wild-type cells, SET2 null mutants arrest in early S phase when treated with 0.2 M HU or 0.1% MMS (Fig. 2D and E). Despite the fact that they progress slowly through the division cycle, asf1Δ cells also arrest predominantly in early S phase in response to HU and MMS. The response of asf1Δ set2Δ cells to HU and MMS is identical to the response of asf1Δ cells. Therefore, Asf1 and Set2 do not have similar functions in the control of cell cycle progression under environmental conditions that cause DNA structure abnormalities. Collectively, these results suggest that Asf1 and Set2 do not have similar core functions in steps of chromatin regulation that affect cell cycle progression or DNA structure checkpoint responses.
Since Asf1 and Set2 are known to function in steps of chromatin regulation that are coupled to elongation by RNAP II (see the introduction), we predicted that they might share phenotypes connected to misregulation of this phase of the transcription cycle. One such phenotype of set2Δ cells is resistance to the elongation inhibitor 6-azauracil (6-AU) (24, 26). We therefore tested the effect of 6-AU on set2Δ and asf1Δ cells in a side-by-side spotting experiment (Fig. (Fig.2F).2F). As expected (24, 26), set2Δ cells exhibited increased resistance to 6-AU compared to wild-type cells. Surprisingly, however, asf1Δ cells are hypersensitive to 6-AU. Compared to the single mutants, asf1Δ set2Δ cells have intermediate sensitivity to 6-AU. On the basis of these results, we consider it unlikely that Asf1 and Set2 have similar global roles in the regulation of RNAP II transcription.
Both Asf1 and Set2 contribute to the regulation of acetylation of the core histones. Asf1 promotes acetylation of H3 at K56 by stimulating the KAT activity of Rtt109 (10, 20). Set2 promotes deacetylation of H3 and H4 by a mechanism that affects histone deacetylase (HDAC) recruitment to chromatin (reviewed in reference 46). In other words, Asf1 and Set2 have opposite effects on H3 acetylation. Accordingly, we predicted that the combined effect of deleting ASF1 and SET2 on overall H3 K9 and K14 acetylation would be intermediate between the effects of the individual deletions.
To test this prediction, we analyzed total protein extracts from wild-type, asf1Δ, set2Δ, and asf1Δ set2Δ strains by immunoblotting, using antibodies that specifically detect K9- and K14-acetylated H3 (Fig. (Fig.3A).3A). As previously reported (1, 2), overall acetylation of H3 at these sites is lower in asf1Δ cells than in wild-type cells. In set2Δ cells, H3 K9/K14 acetylation is elevated. The overall H3 acetylation phenotype of the asf1Δ set2Δ double mutant falls between that of the single mutants (higher than the asf1Δ mutant but lower than the set2Δ mutant). Therefore, on a global scale, the effects of ASF1 and SET2 deletion on H3 tail acetylation seem to offset one another.
We next turned our attention to the control of histone methylation as a possible node at which the functions of Asf1 and Set2 intersect. By immunoblotting, Cheung et al. observed no effect of deletion of ASF1 on overall H3 K36 methylation (7). Our results are not as clear-cut. That is, while H3 monomethylated K36 (K36me1) and H3 K36me2 were unaffected by ASF1 deletion, in 7 of 10 independent comparisons between log-phase wild-type and asf1Δ cells, H3 K36me3 was reduced in the mutant (Fig. (Fig.3B3B shows a representative example). Because three comparisons suggested equivalent H3 K36me3 expression in wild-type and asf1Δ cells, the experiment was repeated by analyzing total protein obtained in parallel from four independent wild-type cultures and four independent null mutant cultures. H3 K36me3 and total H3 expression were assayed by immunoblotting and quantitated. The results, in Fig. Fig.3C,3C, support the conclusion that deletion of ASF1 is associated with lower overall H3 K36 trimethylation. Since Set2 protein levels are the same in wild-type and asf1Δ cells (Fig. (Fig.3D),3D), this phenotype does not reflect Asf1 regulation of steady-state Set2 expression. These findings led us to further explore the possibility that Asf1 does modulate methylation of H3 K36.
H3 K36 methylation by Set2 plays an important role in the suppression of spurious transcription within the coding regions of genes. Specifically, chromatin deacetylation which inhibits spurious transcription depends on binding of the Rpd3S HDAC complex to nucleosomes that have been dimethylated at H3 K36 by Set2 (46).
Asf1 also inhibits spurious transcription within coding regions (7, 42), raising the possibility that it functions in the Set2-dependent pathway of elongation-coupled chromatin remodeling. We tested this possibility by several approaches. First, we used Northern blotting to compare the effects of individual and combined deletions of ASF1 and SET2 on spurious intragenic transcription at FLO8.
In some chromatin metabolism mutants which cannot fully suppress spurious transcription within coding regions, two novel FLO8 transcripts have been detected: the major abnormal transcript produced by conditional SPT6 mutants under restrictive conditions (23) likely corresponds to the “short” transcript in Fig. Fig.4A,4A, while the slower-migrating spurious transcript previously observed in set2Δ cells (6) likely corresponds to the “long” transcript. As expected (6), a long intragenic FLO8 transcript is induced in cells lacking Set2 (Fig. (Fig.4A).4A). Under the conditions used here, we observe little or no induction of these species when ASF1 is deleted (Fig. 4A and C to E). Therefore, if Asf1 functions at FLO8 in a Set2-dependent pathway for suppression of spurious intragenic transcription, it is not an essential component of this pathway.
The notion that Asf1 is not necessary for Set2-dependent inhibition of spurious intragenic transcription at FLO8 is supported by the finding that combined deletion of both ASF1 and SET2 is associated with strong induction of the short intragenic transcript and modest induction of the long transcript. It is apparent from the result for the double mutant (Fig. (Fig.4A)4A) that Asf1 can contribute to inhibition of spurious intragenic transcription in FLO8 by a mechanism that does not require H3 K36 methylation. In other words, Asf1 plays an important role in suppression of intragenic transcription when H3 cannot be methylated at K36.
The same is likely true at another gene, SPB4, where inactivation of SPT6 is known to be associated with induction of a single spurious transcript (23). A tightly clustered population of abnormal SPB4 transcripts also accumulates in cells lacking SET2 (Fig. (Fig.4B).4B). These transcripts are not induced in an asf1Δ single mutant but are very abundant in cells lacking both ASF1 and SET2. Collectively, these results suggest that Asf1 and Set2 modulate spurious intragenic transcription in FLO8 and SPB4 as components of independent pathways.
This idea was supported by experiments in which we assessed the genetic interactions between ASF1 and genes encoding components of the Rpd3S complex, whose role in suppression of intragenic transcription depends on Set2 (6, 24). Since Rpd3S is in the same pathway as Set2, its inactivation in asf1Δ cells is expected to have an additive effect on intragenic transcription similar to the effect of deleting SET2. Rpd3S was inactivated by deleting the gene encoding its catalytic subunit (RPD3). As predicted, spurious intragenic transcription at FLO8 is more pronounced in asf1Δ rpd3Δ cells than in either single mutant (Fig. (Fig.4C).4C). The Eaf3 subunit of Rpd3S is required for interaction of Rpd3S with Set2-methylated nucleosomes (31). Therefore, in asf1Δ cells, deletion of EAF3 is also expected to have the same effect on spurious intragenic transcription as deletion of SET2. Figure Figure4D4D shows that spurious intragenic transcription at FLO8 is indeed more pronounced in asf1Δ eaf3Δ cells than in either single mutant. Rpd3 exists in an additional complex (Rpd3L), which does not function in Set2-dependent chromatin deacetylation (6). Deletion of SDS3, which encodes a protein found in Rpd3L, but not Rpd3S, does not exacerbate spurious intragenic transcription in asf1Δ cells (Fig. (Fig.4E).4E). We conclude that Asf1 functions separately from Set2 and Rpd3S in a pathway that affects suppression of spurious intragenic transcription at FLO8.
The latter conclusion was reinforced by a chromatin immunoprecipitation (ChIP) experiment in which we assessed H3 K36 di- and trimethylation at FLO8 after correction for total H3 occupancy (Fig. (Fig.5A).5A). As shown in Fig. 5B and C, deletion of ASF1 has no effect on H3 K36 methylation at the 5′ or 3′ end of the coding region of FLO8.
On the basis of the known biochemical activities of Asf1, we hypothesized that its contribution to transcriptional regulation at FLO8 depends on the conserved mechanism by which its core domain binds to the H3/H4 heterodimer (13, 35). Accordingly, we tested, using a genetic complementation approach, whether the capacity of Asf1 to bind to H3/H4 is important for suppression of spurious intragenic transcription in the coding region of FLO8. Mutation of valine 94 of Asf1 to arginine almost completely eliminates its histone binding activity (34). As expected, wild-type Asf1 expressed from a low-copy vector fully suppresses expression of the short intragenic transcript in asf1Δ set2Δ cells (Fig. (Fig.6A).6A). Asf1V94R, although expressed at the same level as the wild-type protein (Fig. (Fig.6B),6B), fails to suppress this phenotype (Fig. (Fig.6A).6A). Therefore, the ability of Asf1 to bind to H3/H4 is important for suppression of spurious intragenic transcription.
Since induction of cryptic intragenic transcription is associated with histone hyperacetylation (see the introduction) and Asf1 can inhibit H3/H4 acetylation (49), we reasoned that in the absence of ASF1, cryptic intragenic transcription may be increased in FLO8 due to histone hyperacetylation. Analysis by ChIP, however, revealed that loss of Asf1 expression has little or no effect on H3 K9ac/K14ac in FLO8 (Fig. 6C and D). H4 acetylation at the 5′ end of FLO8 is also unaffected by deletion of ASF1 (Fig. (Fig.6E).6E). At the 3′ end of the coding region of FLO8, however, deletion of ASF1 causes a substantial increase in H4 acetylation (Fig. (Fig.6F).6F). This effect of ASF1 deletion on H4 acetylation is additive with the induction of acetylation caused by SET2 deletion. Therefore, Asf1 modulates H4 acetylation by a Set2-independent mechanism. Since bulk H3 cross-linking was not strongly affected by deletion of ASF1 (Fig. (Fig.5A),5A), overall our results indicate that Asf1 modulates spurious intragenic transcription at FLO8 by a mechanism that affects H4 acetylation but is independent of the regulation of H3 K36 methylation by Set2.
The data we obtained for FLO8 and SPB4 (Fig. (Fig.4)4) support the notion that Asf1 is not an essential component of a linear pathway of chromatin regulation that depends on Set2. They do not, however, rule out the possibility that Asf1 and Set2 collaborate in chromatin regulation at these or other genes.
PMA1 is a frequently transcribed gene that is known to be occupied by Set2 (28, 57). In view of the evidence that Asf1 might modulate overall H3 K36 trimethylation by Set2 (Fig. (Fig.3C),3C), we tested whether deletion of ASF1 affects H3 K36me3 in the coding region of PMA1 by ChIP (Fig. (Fig.7).7). Surprisingly, in contrast to FLO8 (Fig. (Fig.5C),5C), trimethylation of H3 K36 throughout the coding region of PMA1 partly depends on Asf1 (Fig. (Fig.7A)7A) (Fig. (Fig.7B7B shows the H3 cross-linking control). This dependency is not observed for H3 K36 dimethylation (Fig. (Fig.7C).7C). These results suggest that Asf1 modulates the switch from di- to trimethylation of H3 K36.
Using ChIP, we also determined the effect of the Asf1 V94R mutation on H3 K36 trimethylation. Figure Figure7D7D shows that H3 K36 trimethylation in the coding region of PMA1 is significantly impaired in the asf1V94R mutant, as it is in the ASF1 null mutant (Fig. (Fig.7A).7A). H3 cross-linking remains unaffected (Fig. (Fig.7E).7E). We conclude that the ability of Asf1 to stimulate the switch from di- to trimethylation of H3 K36 in PMA1 requires robust H3/H4 binding by Asf1.
The activity of Rtt109, the lysine acetylase (KAT) that acetylates K56 in the histone fold domain of H3, is known to depend on Asf1 (10, 20). Furthermore, Rtt109 activity is impaired by the Asf1 V94R mutation (21). These facts and the recent observation that H3 acetylated K56 (K56ac) has a strong destabilizing effect on nucleosomes (37) led us to hypothesize that mutation of ASF1 may affect H3 K36me3 through an effect on H3 K56ac. To test this hypothesis, we asked whether H3 K36me3 in PMA1 is misregulated in an rtt109Δ strain by using ChIP. It is not (Fig. (Fig.7F).7F). Furthermore, we do not detect a synthetic sick interaction between SET2 and RTT109 in plating assays (Fig. (Fig.7G).7G). Therefore, Asf1 does not modulate H3 K36me3 by virtue of an effect on acetylation at H3 K56.
Asf1 could stimulate H3 K36 trimethylation in coding regions by multiple mechanisms. Two mechanisms that seem likely are (i) indirect stimulation of Set2 activity and (ii) direct enhancement of Set2 recruitment to target genes. We have not exhaustively explored possible indirect mechanisms for induction of Set2 activity in asf1Δ cells, because our data (below) suggest that Asf1 directly controls Set2 occupancy of PMA1.
We used ChIP to test whether Asf1 contributes to the association of Set2 with PMA1. As shown in Fig. Fig.8A,8A, Set2 cross-linking in the coding region of this gene is significantly dampened in asf1Δ cells. The converse is not true: using the ChIP assay validated in Fig. Fig.8B,8B, we find that cross-linking of Asf1 is unaffected by SET2 deletion (Fig. (Fig.8C).8C). Since deletion of ASF1 does not inhibit the expression of Set2 (Fig. (Fig.3D3D and and8D),8D), these findings are consistent with a direct role for Asf1 in promoting Set2 occupancy in the coding region of PMA1.
We tested whether the capacity of Asf1 to bind to the H3/H4 dimer is important for Set2 localization to PMA1 by ChIP. Figure Figure8E8E shows that Set2 cross-linking to PMA1 is impaired in cells that express Asf1V94R in place of wild-type Asf1 (expression of relevant proteins in this experiment is shown in Fig. Fig.8D).8D). Therefore, H3/H4 binding by Asf1 is essential for normal Set2 association with PMA1.
Because a soluble complex that includes both Asf1 and Set2 has not been reported in the literature, it seems unlikely that these molecules are recruited to PMA1 as components of a stable assemblage of proteins. Our ChIP results do not, however, rule out the possibility that interactions (possibly indirect) between Asf1 and Set2 contribute to Set2 cross-linking to genes such as PMA1. These interactions would likely be strong but short-lived or persistent but weak.
The notion that some functional interactions between Asf1 and Set2 could involve their direct or indirect physical association is supported by our analysis of Asf1-associated proteins obtained by tandem affinity purification (TAP). We isolated Asf1-TAP complexes and identified some of the species in this preparation by mass spectrometric analysis of trypsin-digested gel slices. Many proteins already known to associate with Asf1 were identified (Hir1, Hir2, Hir3, Rad53, Hpc2) (Fig. (Fig.9A)9A) (12, 19, 22, 38). One gel slice also yielded a peptide that was assigned to Set2. This result suggested that Asf1 and Set2 coexist in a “rare” complex.
We further tested the hypothesis that Asf1 and Set2 can occur in the same complex by purifying Asf1- and Set2-TAP from strains that also expressed HA epitope-labeled Set2 (Set2-HA) or myc epitope-labeled Asf1 (Asf1-myc). We detected Set2 in association with Asf1-TAP and Asf1 in association with Set2-TAP (Fig. 9B and C). In each case, the amount of HA- or myc-tagged protein in the final eluate represented only a very small fraction of the respective protein present in the input. This finding is consistent with cooccurrence of a subset of Asf1 and Set2 molecules in low-abundance protein complexes.
Trimethylation of H3 K36 by Set2 requires phosphorylation of the CTD of elongating RNAP II by protein kinase Ctk1 and phosphorylation of the domain of Set2 that is known to bind to RNAP II (59). It follows that if complexes containing Asf1 and Set2 form only on DNA, then their existence is likely to depend on expression of Ctk1. To test this possibility, we determined whether the recovery of Set2-HA in Asf1-TAP complexes is affected by deletion of CTK1. Despite the low expression of Set2 in ctk1Δ strains (59), Set2-HA was recovered in Asf1-TAP complexes obtained from a CTK1 null mutant (Fig. (Fig.9D).9D). We conclude that complexes that include Asf1 and Set2 can exist independently of Ctk1-phosphorylated RNAP II.
In the context of our goal to gain a better overall appreciation of possible functional interplay between pathways of chromatin metabolism that depend on Asf1 and Set2, we further analyzed the effect of deleting ASF1, SET2, or both, on H3/H4 acetylation in the coding region of PMA1. These results are presented in Fig. 10A to D. Figure 10F summarizes the results of this experiment and of analysis of FLO8 acetylation, which is shown in Fig. 6C to E.
Consistent with published evidence that Set2 promotes histone deacetylation (25, 46), deletion of SET2, when it has an effect, causes acetylation in coding regions to increase. In addition to the induction of H4 acetylation at the 3′ end of FLO8 (Fig. (Fig.6F),6F), deletion of ASF1 is associated with increased H4 acetylation at the 5′ and 3′ ends of PMA1 (Fig. 10A and B) and increased H3 acetylation at the 5′ end of PMA1 (Fig. 10C). We conclude that Asf1 can inhibit tail acetylation of nucleosomal H3/H4 throughout coding regions. At the 3′ end of PMA1 where Asf1 inhibits H4 acetylation, loss of Asf1 is associated with decreased H3 acetylation (Fig. 10D). Therefore, it appears that within a single region of a gene, Asf1 can stimulate H3 acetylation and inhibit H4 acetylation.
As noted in the introduction, Asf1 is a potential inhibitor of an H4 K16-specific KAT, the SAS complex (49). Increased H4 acetylation detected by ChIP using the pan-specific tetra-acetylated H4 antibody mixture (anti-K5ac/8ac/12ac/16ac) could be due to loss of this inhibition. To test this possibility, the effect of ASF1 deletion specifically on H4 K16ac in PMA1 was determined by ChIP. H4 K16 acetylation does not differ between wild-type and asf1Δ cells at either the 5′ or 3′ end of PMA1 (Fig. 10E). Therefore, Asf1 regulates H4 acetylation in PMA1 at sites other than those preferentially acted on by the Sas2 catalytic subunit of the SAS complex.
The general nature of interplay between gene products in the generation of a specific phenotype can often be inferred by comparing the effects of individual deletions and combined deletions of the genes of interest. The genetic interactions between ASF1 and SET2 suggest that Asf1 and Set2 can separately contribute to the regulation of H3/H4 acetylation. For H3, there is an additive stimulatory effect of ASF1 and SET2 deletion on acetylation at the 5′ ends of FLO8 and PMA1 (Fig. (Fig.6C6C and 10C). Deletion of ASF1 and SET2 also has an additive positive effect on H4 acetylation, but in this case, it occurs at the 3′ ends of FLO8 and PMA1 (Fig. (Fig.6F6F and 10B). We conclude that Asf1 and Set2 can separately inhibit H3/H4 acetylation. Considering these findings, it is intriguing that deletion of ASF1 is associated with decreased H3 acetylation at the 3′ end of PMA1 and that this phenotype is suppressed by deletion of SET2 (Fig. 10D).
This study shows that chromatin regulation in the coding regions of genes can involve separate contributions of Asf1 and Set2 or collaboration of Asf1 and Set2. The collaboration between Asf1 and Set2 enhances trimethylation of H3 K36.
Recent work by Youdell et al. (59) and Li et al. (32) revealed that Set2 dimethylation of H3 K36 is sufficient for suppression of intragenic transcription by the Rpd3S complex. Asf1 does not affect H3 K36 dimethylation in the coding region of FLO8 (Fig. (Fig.5B)5B) and therefore is unlikely to suppress spurious intragenic transcription as a component of the Set2 pathway. Consistent with this interpretation, we also observe the following. (i) Asf1 does not affect Set2 cross-linking in the coding region of FLO8 (Fig. (Fig.8A).8A). (ii) The function of Asf1 in transcription at FLO8 is genetically separable from the functions of Set2 and components of the Rpd3S complex (Fig. (Fig.44 and and6A6A).
Deletion of ASF1 is associated with decreased histone acetylation in some locations and increased acetylation in others (this study). If we take the evidence that Asf1 can directly stimulate H3 K56 acetylation by Rtt109 (10, 20) to indicate a more general ability for Asf1 to stimulate KAT reactions (for example, acetylation of other H3 residues), then decreased histone acetylation in asf1Δ cells is not surprising. However, the locus specificity of this effect remains unexplained (for example, Asf1 appears to stimulate H3 acetylation at the 3′ end of PMA1 [Fig. 10D] but not at the 3′ end of FLO8 [Fig. [Fig.6D6D]).
Our observation that deletion of ASF1 is often associated with increased histone acetylation (for example, H4 at the 3′ end of FLO8 and at the 5′ end of PMA1; Fig. Fig.6F6F and 10A, respectively) suggests that in some contexts, Asf1 can inhibit acetylation. Asf1 can inhibit H4 K16 acetylation by the SAS complex in vitro (49), but this regulation is not important at the gene we studied because its H4 K16 acetylation state is unaffected by deletion of ASF1 (Fig. 10E). An indirect mechanism could also account for increased histone acetylation in the absence of ASF1. For example, considering the finding that Asf1 inhibits spurious intragenic transcription in FLO8 (7), a simple explanation for increased H4 acetylation at the 3′ end of this gene in asf1Δ cells (Fig. (Fig.6F)6F) would be that hyperacetylation is caused by intragenic transcription. This model demands an explanation for increased intragenic transcription in asf1Δ cells. At this time, however, induction of cryptic transcription in FLO8 as a result of loss of Asf1 expression is difficult to explain. Most importantly, although decreased nucleosome assembly activity could induce spurious transcription in the absence of Asf1, we observe little effect of ASF1 deletion on H3 cross-linking in the coding region of FLO8 (Fig. (Fig.5A5A).
Despite the wealth of published information about the regulation of histone acetylation (30, 43), at this time we cannot envisage a straightforward mechanism by which deletion of ASF1 differentially stimulates H3 K9/K14 acetylation and inhibits H4 tail acetylation at the 3′ end of PMA1 (Fig. 10B and D). Similarly, the locus-specific genetic interactions between ASF1 and SET2 summarized in Fig. 10F remain unexplained at a molecular level. We cannot rule out complex indirect effects by which the balance of KAT/HDAC activity toward H3 and H4 is differentially affected by deletion of ASF1.
The results of previous studies have not yet provided a robust understanding of the mechanism for specific association of Set2 with the coding regions of genes. A large body of early work supported the idea that the critical step in recruitment of Set2 to chromatin is its binding to the serine 2 phosphorylated form of the CTD of RNAP II (reviewed in reference 15). More recent results, however, strongly suggest that Set2 association with one well-studied gene (STE11) does not involve this mechanism (59). An alternative mechanism for Set2 association with STE11 was not elucidated.
Our analysis of PMA1 and data in the literature suggest that Asf1 can play an important role in Set2 association with the coding region of an active gene. First, deletion of ASF1 is associated with decreased cross-linking of Set2 to PMA1 (Fig. (Fig.8A).8A). Second, an expected functional consequence of reduced Set2 cross-linking to PMA1 is readily apparent in asf1Δ cells, namely, decreased H3 K36 trimethylation (Fig. (Fig.7A7A).
How might Asf1 stimulate Set2 association with PMA1 and H3 K36 trimethylation in its coding region? Since H3 K36 trimethylation by Set2 requires the elongating form of RNAP II (59), it would be reasonable to ascribe Asf1 stimulation of H3 K36 trimethylation in PMA1 to stimulation of RNAP II occupancy. This possibility seems unlikely, however, since loss of Asf1 does not cause lower RNAP II cross-linking in the coding region of PMA1 (42) (our unpublished data). Nonetheless, it does seem reasonable to suppose that the effect of Asf1 on Set2 association with PMA1 is mechanistically related to its stimulation of H3 K36 trimethylation. Under this assumption, recent work on the role of histone binding by Set2 suggests a plausible model for Set2 regulation by Asf1.
Du et al. demonstrated that di- and trimethylation of H3 K36 by Set2 depends on the ability of Set2 to bind to histone H4 (11). This binding occurs in a loop region between the α1 and α2 helices of the H4 histone fold domain, and the binding critically requires K44 within the loop. Since the C terminus of H2A surrounds the side chain of H4 K44 in nucleosomes, it was proposed that efficient interaction of Set2 with H4 could require removal of H2A-H2B dimers. Asf1 binds to H3-H4 dimers in a way that prevents H2A-H2B association and leaves the loop containing H4 K44 exposed (13, 35). We propose that the complex of Asf1 with a H3/H4 dimer could be a preferred substrate for H4 binding by Set2 (Fig. 10G, diagram 1). Two of our findings support this idea. First, the existence of a complex containing Asf1 and Set2 (albeit of very low abundance) is consistent with a reaction pathway involving the formation of an intermediate comprised of Asf1, H3/H4, and Set2. Second, Asf1 must be able to tightly bind to H3/H4 in order to stimulate Set2 association with PMA1 and H3 K36 trimethylation (Fig. (Fig.7D7D and and8E).8E). We envisage an overall model in which Asf1 stimulates H3 K36 trimethylation in the course of an elongation-coupled nucleosome reconfiguration cycle by transiently promoting H3/H4 association with Set2 (Fig. 10G, diagram 1). Reorganization of H2A/H2B dimers in nucleosomes, required for execution of this regulation, could be performed by the H2A/H2B chaperone FACT, which is known to function in the coding regions of transcribed genes (58).
In addition to functioning in a single pathway, Asf1 and Set2 have independent roles which impinge on chromatin in a way that suppresses spurious intragenic transcription. Importantly, Asf1 does not inhibit unwanted intragenic transcription as a component of the pathway which involves HDAC recruitment by Set2-dependent H3 K36 dimethylation (Fig. 10G, diagram 2). We know that suppression of intragenic transcription by Asf1 depends on its ability to bind to H3/H4, and therefore involves regulation of the nucleosome (Fig. (Fig.6A6A and Fig. 10G, diagram 2). The possibility that disruption of the latter regulation has indirect effects that induce spurious intragenic transcription in ASF1 mutants remains to be investigated.
Since H3 K36me3 is not necessary for suppression of spurious intragenic transcription (32, 59), it likely has other roles in chromatin regulation. One might be to control gene silencing by the Sir proteins (41). Set2 and H3 K36 methylation are known to antagonize silencing by a mechanism that does not involve suppression of spurious intragenic transcription (51). It follows that the ability of Asf1 to promote the switch from di- to trimethylation of H3 K36 might be important for control of silencing. In favor of this possibility is the evidence that silencing is modulated by Asf1 (29, 48). While H3 K36me3 might be important for silencing, it is unlikely that silencing-related functions of H3 K36me3 are relevant to its function at PMA1. We base this proposal on the finding that the nonconserved C-terminal domain of Asf1 contributes to the regulation of silencing (50) but is not needed for the H3 K36 dimethylation to trimethylation switch at PMA1 (L. Lin, L. V. Minard, and M. C. Schultz, data not shown). Accordingly, we suspect that Asf1-dependent switching from di- to trimethylation of H3 K36 in PMA1 has a role in the regulation of chromatin function not yet revealed by published studies of Set2.
Darren Hockman is thanked for excellent technical assistance, and members of the Schultz lab are acknowledged for helpful discussions. We thank Carl Mann for plasmids.
This work was supported by grants to M.C.S. from the Canadian Institutes of Health Research and the Alberta Heritage Foundation for Medical Research and by grants to G.C.J. and R.A.S. from the Canadian Institutes of Health Research. L.L. was supported in part by a QEII Scholarship from the Government of Alberta, and L.V.M. was supported in part by a Master's Award from the Canadian Institutes for Health Research.
Published ahead of print on 4 January 2010.