Methylation of Histone H3 Mediates the Association of the NuA3 Histone Acetyltransferase with Chromatin

The SAS3-dependent NuA3 histone acetyltransferase complex wasoriginally identified on the basis of its ability to acetylatehistone H3 in vitro. Whether NuA3 is capable of acetylatinghistones in vivo, or how the complex is targeted to the nucleosomesthat it modifies, was unknown. To address this question, weasked whether NuA3 is associated with chromatin in vivo andhow this association is regulated. With a chromatin pulldownassay, we found that NuA3 interacts with the histone H3 amino-terminaltail, and loss of the H3 tail recapitulates phenotypes associatedwith loss of SAS3. Moreover, mutation of histone H3 lysine 14,the preferred site of acetylation by NuA3 in vitro, phenocopiesa unique sas3 ${Delta}$ phenotype, suggesting that modification of thisresidue is important for NuA3 function. The interaction of NuA3with chromatin is dependent on the Set1p and Set2p histone methyltransferases,as well as their substrates, histone H3 lysines 4 and 36, respectively.These results confirm that NuA3 is functioning as a histoneacetyltransferase in vivo and that histone H3 methylation providesa mark for the recruitment of NuA3 to nucleosomes.

INTRODUCTION

Top

Introduction

In eukaryotes, DNA is packaged into chromatin, a nucleoproteinstructure consisting of DNA, histones, and nonhistone proteins.The assembly of DNA into chromatin modulates the access of cellularmachinery to DNA and thus regulates transcription, replication,repair, and recombination. Chromatin structure can be regulatedby the addition of posttranslational modifications to histones,including acetylation, methylation, phosphorylation, and ubiquitination.Consistent with this, several histone posttranslational modificationshave been linked to the regulation of gene expression.

Numerous multiprotein complexes are involved in the posttranslationalmodification of histones. These complexes vary in both theirprotein components and the modifications they effect. The mostwell-studied group of modifying complexes is the histone acetyltransferases(HATs), which use acetyl coenzyme A as a substrate for the acetylationof lysine residues within both the tail and globular domainsof histones (39, 43, 45, 73). In the budding yeast Saccharomycescerevisiae, there are at least eight proteins that have beenidentified as having HAT activity in vitro, including Gcn5p,Hat1p, Esa1p, Elp3p, Nut1p, Hpa2p, Sas2p, and Sas3p (45, 73),although for many of these proteins, whether histones representtheir true substrates in vivo is not known. Three of these proteinsare found in complexes that specifically acetylate the lysineswithin the tail of histone H3, including the GCN5-dependentSAGA, SLIK/SALSA, ADA, and HAT-A2 complexes (12, 16, 49, 58,60), the ELP3-dependent elongator complex (70), and the SAS3-dependentNuA3 complex (25). The acetylation of histone tails is associatedwith regions of transcriptional activity, and histone acetylationis thought to modulate chromatin structure through two differentmechanisms. First, the neutralization of charge associated withhistone acetylation is believed to relax the structure of chromatin,allowing access of other factors to nucleosomal DNA (11). Second,histone acetylation has been shown to regulate the binding ofchromatin-modifying factors to nucleosomes (9).

In a manner analogous to acetylation, histones are also modifiedby the addition of methyl groups to lysine and arginine residueswithin the amino-terminal tails of histones H3 and H4 (76) andwithin the globular domain of histone H3 (14, 32, 65). Althoughhistone methylation has been linked to transcriptional silencingin other organisms (18), in S. cerevisiae, methylation is foundassociated with regions of transcriptional activity (46). Consistentwith this, two yeast histone methyltransferases, Set1p and Set2p,bearing specificity toward histone H3 lysine residues 4 and36, respectively, are found associated with transcriptionallyactive regions of the genome (29, 41). Set1p and Set2p are foundassociated with RNA polymerase II (RNAPII), and these interactionsare dependent on the phosphorylation of the carboxyl-terminaldomain of Rpb1p, the largest subunit of RNAPII (20). The interactionof Set1p with the carboxyl-terminal domain requires the phosphorylationof serine 5 by Kin28p, a TFIIH-associated kinase that mediatesthe transition between transcription initiation and elongation(41). In contrast, Set2p interacts with RNAPII that has beenphosphorylated on serine 2, which is performed primarily byCtk1p at the later stages of elongation (29, 33, 34, 56, 72).Consequently, the level of H3 lysine 4 trimethylation peaksat the promoters and early coding regions of genes (40, 41,46, 54), whereas H3 lysine 36 di- and trimethylations are spreadthroughout the open reading frames (29, 40, 46, 56, 72). Recently,methylation of the histone H3 tail has been implicated in mediatingthe interaction of several HAT complexes (40, 48) and one histonedeacetylase complex (HDAC) (8, 26) with nucleosomes. Thus, itseems an attractive possibility that the methylation of specificlysine residues within histone H3 serves a universal role inmediating histone acetylation.

NuA3 is an approximately 400-kDa acetyltransferase complex containingthe catalytic subunit Sas3p, in addition to at least four othersubunits (22, 25). NuA3 was originally identified on the basisof its ability to acetylate histone H3 in vitro, but whetherNuA3 is capable of acetylating histones in vivo is unknown.Deletion of SAS3 is synthetically lethal with the loss of Gcn5p,the catalytic subunit of several histone H3-specific HAT complexes.However, this synthetic lethality is not due to the loss ofany of the known Gcn5p-dependent HATs, suggesting that Gcn5pmay have a function that is unrelated to histone acetylation(21). To understand the essential function of NuA3 in the cell,we addressed whether this complex is truly acting as a HAT invivo and asked how it is recruited to the histones that it modifies.With a chromatin pulldown assay, we were able to demonstratean interaction between NuA3 and chromatin and found that thisinteraction is dependent on the H3 tail. Consistent with theimportance of this interaction for NuA3 function, we found thatloss of the H3 tail recapitulates a unique sas3 ${Delta}$ -associated phenotype.In addition, we provide evidence that SET1- and SET2-dependentmethylation of histone H3 lysines 4 and 36, respectively, isrequired for NuA3 interaction with chromatin and for NuA3 functionin vivo. These data, together with observations that histoneH3 methylation mediates the acetylation of histones by otherHAT complexes (40, 48), suggest that SET1- and SET2-dependentmethylation may serve as universal marks for targeting histoneacetylation to active regions of the genome.

MATERIALS AND METHODS

Top

Materials and Methods

Yeast strains, plasmids, and genetic methods. All of the strains used in this study (Table 1) are isogenicto S288C. Yeast culture and genetic manipulations were carriedout by standard protocols (2, 37, 50). Genomic deletions orepitope tag integrations were verified by PCR analysis. Matingassays were performed as described by others (59). Briefly,strains were patched onto yeast extract-peptone-dextrose mediumand grown overnight at 30°C. Cells were replica plated ontoa lawn of YLH101 on yeast extract-peptone-dextrose medium andallowed to mate overnight at 30°C. Finally, plates werereplica plated onto medium lacking tryptophan and adenine toselect for diploids and incubated at 30°C overnight. Strainscarrying histone H3 mutations were derived by plasmid shufflefrom FY2162 (YLH224) (10). This strain carries deletions ofthe HHT1-HHF1 and HHT2-HHF2 loci and carries HHT2-HHF2 on aURA3 plasmid (pHHT2). Because this strain and all of the strainsderived from it carry a wild-type version of HHF2 on a plasmid,for simplicity, the genotypes of these strains will be referredto as hht1 ${Delta}$ hht2 ${Delta}$ in the figures and figure legends. Plasmid pHHT2was constructed by ligation of the SpeI fragment from pDM18(10) into the SpeI site of pRS414. phht2 ${Delta}$ 3-29 was generated byligating annealed phosphorylated oligonucleotides (5'-GATCCAAGCAAACACTCCACAATGGCCAGACCATCTA-3'and 5'-CCGGTAGATGGTCTGGCCATTGTGGAGTGTTTGCTTG-3') into the BamHIand AgeI sites of pHHT2. Plasmids containing mutations in HHT2were generated with the Stratagene QuikChange site-directedmutagenesis kit. Plasmid pSAS3 (pLP640) consists of the SAS3coding region, 1,300 bp of upstream and 100 bp of downstreamsequences inserted into pRS316 (21).

View this table:
[in this window]
[in a new window]
TABLE 1. Strains used in this study

Chromatin pulldown assays. All chromatin pulldown assays were performed with yeast whole-cellextracts prepared from mid-log-phase cells in IPP150 buffer(10 mM Tris-Cl [pH 8.0], 150 mM NaCl, 0.1% NP-40, 1 mM phenylmethylsulfonylfluoride, 2 µg/ml pepstatin A). Input protein amountswere normalized by Bradford protein assay, and 5 to 10 mg ofprotein was incubated with 10 µl of immunoglobulin G (IgG)Sepharose 6 Fast Flow resin (Amersham Biosciences, Piscataway,NJ). Samples were rotated at 4°C for 2 h, and the resinwas washed five times with 20 volumes of cold IPP150 buffer.Bound proteins were eluted by boiling in sodium dodecyl sulfatesample buffer and analyzed by immunoblotting with antihemagglutinin(anti-HA)-peroxidase (Roche Diagnostics Canada, Laval, Quebec,Canada), peroxidase-antiperoxidase (Sigma Chemical Co., St.Louis, MO), or anti-diacetylated histone H3 (Upstate Biotechnology,Lake Placid, NY) antibodies.

RESULTS

Top

Results

NuA3 interacts with chromatin in vivo. Previously it was shown that the NuA3 acetyltransferase canacetylate histones in vitro (25); however, the possibility existsthat NuA3 is acetylating a nonhistone substrate in vivo. Toprovide further support that NuA3 is acetylating histones, weasked whether NuA3 interacts with chromatin in vivo. To thisend, we introduced a tandem affinity purification (TAP) tagto the carboxyl terminus of H2B (Htb1p) to enable us to purifynative chromatin with IgG resin. Because of the presence ofnewly synthesized H2B in the cytoplasm, it was not expectedfor all of the TAP-tagged H2B that is bound by the IgG resinto be incorporated into nucleosomes. To ensure that at leastsome of the precipitated H2B is associated with nucleosomes,we immunoblotted precipitated material with anti-acetylatedH3 antibodies. As shown in Fig. 1A, the Htb1TAP-tagged strainwas capable of pulling down acetylated histone H3. Most of thepublished data support the hypothesis that nucleosome assemblyoccurs in a stepwise manner that involves the initial depositionof an H3-H4 tetramer on the newly synthesized DNA, followedby deposition of H2A-H2B dimers (19). If this is the case, thenone would have to assume that histones H2B and H3 would notexist in a complex other than in the form of the nucleosome.The fact that we are able to copurify histone H3 with H2B indicatesthat our TAP-tagged Htb1p strain is capable of pulling downintact nucleosomes.

View larger version (27K):
[in this window]
[in a new window]
FIG. 1. NuA3 interacts with chromatin in vivo. Chromatin pulldown assays were performed with strains YDM126 and YDM127 (A and B) and strains YDM126, YLH139 YDM127, and YDM210 (C), and the resulting samples were subjected to Western blotting with immunodetection with anti-diacetylated histone H3 antibodies (AcH3), anti-HA antibodies (Sas3HA), or peroxidase-antiperoxidase antibodies (Htb1TAP). Yeast whole-cell extract (1.4% of input) was also blotted for HA (INPUT).

To determine whether NuA3 is capable of binding chromatin, weHA tagged Sas3p, the catalytic subunit of NuA3, in our Htb1TAPstrain. Chromatin pulldown experiments were performed with IgGresin, followed by anti-HA immunoblot analysis for Sas3p. Supportingprevious in vitro evidence that NuA3 is capable of binding nucleosomes(22), we found that Sas3p coprecipitated with Htb1p in the TAP-taggedstrain but not the untagged strain (Fig. 1B). Additionally,we were also able to coprecipitate Nto1p, another componentof NuA3 (J. M. Dukowski et al., unpublished data), indicatingthat the entire NuA3 complex, and not just Sas3p, is capableof binding chromatin (Fig. 1C).

NuA3 function is dependent on the amino-terminal tail of histone H3. To provide further evidence that NuA3 is indeed interactingwith chromatin in vivo, we mapped the region of the nucleosomethat serves as an interaction site for NuA3 and asked whetherloss of this site creates a sas3 ${Delta}$ -specific phenotype. In lightof the fact that NuA3 is an H3-specific HAT, we surmised thata likely interacting target would be the H3 tail itself (25).To this end, we developed a mutant strain lacking amino acids3 to 29 of the H3 tail (hht2 ${Delta}$ 3-29) and performed chromatin pulldownassays. Samples were normalized for levels of Htb1TAP in thepulldown material, and Fig. 2A shows that coprecipitation ofSas3p with Htb1 TAP was abolished in the H3 tailless mutant.The loss of interaction seen in the H3 tailless mutant indicatesthat the H3 tail is required for NuA3 to bind chromatin. Wenext asked whether loss of the H3 tail recapitulates the phenotypeof a sas3 ${Delta}$ mutant. It has been shown that deletion of SAS3 aloneresults in only minor phenotypes (51). However, if SAS3 is deletedin conjunction with deletion of GCN5, the result is a syntheticlethality (21). Gcn5p is present in multiple HAT complexes inyeast, including SAGA, SLIK/SALSA, ADA, and HAT-A2, each ofwhich contains both unique and shared protein subunits (16,47, 49, 52). In addition to Gcn5p, the SAGA, SLIK/SALSA, ADA,and HAT-A2 complexes also share Ada2p, and experimental evidencesuggests that Ada2p is required to potentiate Gcn5p HAT activity(3, 63). Catalytically active ADA, SAGA, or HAT-A2 complexesfail to purify from strains in which ADA2 has been deleted (3,16, 58). Furthermore, phenotypes associated with deletions ofGCN5 are indistinguishable from those associated with deletionsof ADA2 (15). Thus, both in vitro and in vivo evidence supportsthe fact that Ada2p is required for the function of Gcn5p asa HAT. An interesting characteristic of the gcn5 ${Delta}$ sas3 ${Delta}$ syntheticlethality is that loss of SAS3 is compatible with deletion ofADA2, indicating that the gcn5 ${Delta}$ sas3 ${Delta}$ synthetic lethality is notdue to loss of SAGA, SLIK/SALSA, ADA, or HAT-A2 and thus maybe due to failure of Gcn5p to acetylate a nonhistone substrate.To date, no genes have been identified, other than those thatencode components of NuA3, which exhibit a genetic interactionwith GCN5 but not ADA2 (22). It has been previously demonstratedthat loss of the H3 tail is lethal in a gcn5 ${Delta}$ strain (75), sowe tested whether loss of the H3 tail resulted in the same phenotypein an ada2 ${Delta}$ background. GCN5 or ADA2 was deleted from strainswhich also carried deletions of both the HHT1 and HHT2 lociand expressed histone H3 from a URA3 plasmid. TRP1-based plasmidsexpressing HHT2 or hht2 ${Delta}$ 3-29 were introduced into these strains,and the resulting transformants were plated on 5-fluorooroticacid (5-FOA). When cells were subjected to negative selectionfor the URA3-based plasmid, no growth was observed for gcn5 ${Delta}$ hht2 ${Delta}$ 3-29 mutant cells, while ada2 ${Delta}$ hht2 ${Delta}$ 3-29 mutant cells wereviable (Fig. 2B). The fact that deletion of the H3 tail recapitulatedphenotypes seen in sas3 ${Delta}$ mutants supports the hypothesis thatthe interaction of NuA3 with chromatin is important for NuA3function. We also plated the cells on medium containing 3-aminotriazole(3-AT), which mimics histidine starvation. Strains carryingADA2 or GCN5 deletions are defective for growth on medium containing3-AT due to poor expression of the HIS3 gene (38, 44). Figure2B shows that the gcn5 ${Delta}$ and ada2 ${Delta}$ strains used in this study growpoorly on 3-AT compared to the wild type, demonstrating thatthese strains are exhibiting previously published phenotypes.

View larger version (22K):
[in this window]
[in a new window]
FIG. 2. NuA3 function is dependent on the H3 tail. (A) Chromatin pulldown assays were performed from the indicated strains (YDM141, YDM143, YDM142, and YDM144), and the resulting samples were subjected to Western blotting with immunodetection with anti-HA antibodies (Sas3HA) or peroxidase-antiperoxidase antibodies (Htb1TAP). Yeast whole-cell extract (1.4% of input) was also blotted for HA (INPUT). (B) Tenfold dilutions of yeast strains YLH170 (sas3 ${Delta}$ ), YLH356 (gcn5 ${Delta}$ sas3 ${Delta}$ ), YLH357 (ada2 ${Delta}$ sas3 ${Delta}$ ), YLH347 (hht1 ${Delta}$ hht2 ${Delta}$ ), YLH348 (gcn5 ${Delta}$ hht1 ${Delta}$ hht2 ${Delta}$ ), and YLH349 (ada2 ${Delta}$ hht1 ${Delta}$ hht2 ${Delta}$ ) containing the indicated plasmids were plated on synthetic complete medium (control), synthetic complete medium with 5-FOA, or synthetic histidine dropout medium with 50 mM 3-AT and incubated at 30°C for 3 days.

NuA3 function is dependent on lysine 14 of histone H3. Although chromatin binding is required for NuA3 function, itis still formally possible that NuA3 is recruited to nucleosomesbut is acetylating a nonhistone substrate. To deal with thispossibility, we asked whether loss of the site targeted by NuA3phenocopies a sas3 ${Delta}$ phenotype. NuA3 has been shown to preferentiallyacetylate lysine 14, and to a lesser extent lysine 23, on theH3 tail (21). Thus, we once again took advantage of the differentialgcn5 ${Delta}$ and ada2 ${Delta}$ genetic interactions with SAS3 mutants to testthe importance of histone H3 lysine 14 for NuA3 function. Asshown previously, mutation of lysine 14 to arginine createsa severe growth defect in gcn5 ${Delta}$ strains (75). However, we foundthat this mutation was tolerated in ada2 ${Delta}$ strains (Fig. 3A).To verify that the sas3 ${Delta}$ phenotype seen in lysine 14 mutantswas not due to loss of NuA3-nucleosome interaction, we performeda chromatin pulldown assay on the lysine 14 mutant. Sampleswere normalized for levels of Htb1TAP in the pulldown assays,and Fig. 3B shows that mutation of lysine 14 to arginine didnot affect Sas3p binding to chromatin, suggesting that lysine14 is important for NuA3 function at a step downstream of nucleosomebinding. Thus, the fact that mutation of lysine 14 recreateda sas3 ${Delta}$ -specific phenotype supports our hypothesis that NuA3modifies this residue in vivo. Interestingly, the data obtainedthus far indicate that in the absence of GCN5, Sas3p is essentialfor histone acetylation, while in the absence of SAS3, Gcn5pmay be required for acetylation of a nonhistone substrate.

View larger version (20K):
[in this window]
[in a new window]
FIG. 3. NuA3 function is dependent on lysine 14. (A) Tenfold dilutions of yeast strains YLH170 (sas3 ${Delta}$ ), YLH356 (gcn5 ${Delta}$ sas3 ${Delta}$ ), YLH357 (ada2 ${Delta}$ sas3 ${Delta}$ ), YLH347 (hht1 ${Delta}$ hht2 ${Delta}$ ), YLH348 (gcn5 ${Delta}$ hht1 ${Delta}$ hht2 ${Delta}$ ), and YLH349 (ada2 ${Delta}$ hht1 ${Delta}$ hht2 ${Delta}$ ) containing the indicated plasmids were plated on synthetic complete medium (control), synthetic complete medium with 5-FOA, or synthetic histidine dropout medium with 50 mM 3-AT and incubated at 30°C for 3 days. (B) Chromatin pulldown assays were performed with the indicated Sas3HA₃-expressing strains (YDM141, YDM142, YDM193), and the resulting samples were subjected to Western blotting with immunodetection with anti-HA antibodies (Sas3HA) or peroxidase-antiperoxidase antibodies (Htb1TAP). Yeast whole-cell extract (1.4% of input) was also blotted for HA (INPUT).

Histone H3 modifiers, Set1p and Set2p, are required for NuA3 interaction with chromatin. Our results obtained thus far support the hypothesis that theNuA3 HAT complex binds and acetylates nucleosomes in vivo. Ithas been previously shown that the binding of many chromatin-modifyingcomplexes to nucleosomes is regulated by histone posttranslationalmodifications. For example, the SAGA HAT complex has been shownto preferentially modify histones that are methylated at H3lysine 4 and phosphorylated at H3 serine 10 (36, 48). To determinewhether the binding of NuA3 to the H3 tail is regulated by otherhistone modifications, we once again employed the chromatinpulldown assay with mutants that lacked various posttranslationalmodifications of the H3 tail, including snf1 ${Delta}$ , gcn5 ${Delta}$ , set1 ${Delta}$ , andset2 ${Delta}$ strains. Samples were normalized for levels of Htb1TAPin the pulldown assays, and Fig. 4A shows that deletion of SNF1or GCN5, which, respectively, phosphorylates or acetylates specificresidues within the H3 tail (16, 17, 30, 35, 49), did not affectthe interaction of Sas3p with chromatin. Similarly, the levelof Sas3p precipitated with Htb1TAP did not appear to be dependenton SET1, which is required for methylation of lysine 4 on histoneH3. In contrast, deletion of SET2 had a significant impact onthis interaction. Set2p methylates lysine 36 within the H3 tail,suggesting that this modification may facilitate the bindingof NuA3 to histone H3. Interestingly, although deletion of SET1did not seem to have a significant impact on Sas3p binding,set1 ${Delta}$ strains consistently showed increased levels of Sas3HAin the yeast whole-cell extracts (Fig. 4A, INPUT). This maybe due to either increased expression of the SAS3 gene or increasedsolubility of the Sas3 protein upon extract preparation. Becausethe latter may be suggestive of a role for SET1 in the interactionof Sas3p with chromatin, we decided to investigate the involvementof SET1 in NuA3-chromatin interaction further. To this end,we tested the effect of a SET1 deletion on Sas3p binding tochromatin in a set2 ${Delta}$ background. We found that, despite increasedlevels of Sas3p in the input whole-cell extract, concomitantdeletion of SET1 and SET2 had an even greater effect than aSET2 deletion alone, suggesting that the binding of NuA3 tochromatin is dependent on both SET1 and SET2 (Fig. 4A). Thisresult suggests that methylation of lysines 4 and 36 mediatesthe interaction of NuA3 with chromatin.

View larger version (36K):
[in this window]
[in a new window]
FIG. 4. Histone H3 modifiers, Set1p and Set2p, are required for NuA3 function. (A) Chromatin pulldown assays were performed from the indicated Sas3HA₃-expressing strains (YDM126, YDM127, YDM151, YDM155, YDM153, YDM163, and YDM187), and the resulting samples were subjected to Western blotting with immunodetection with anti-HA antibodies (Sas3HA) or peroxidase-antiperoxidase antibodies (Htb1TAP). Yeast whole-cell extract (1.4% of input) was also blotted for HA (INPUT). (B) Deletion of SET2 restores silencing to strains with mutations in HMR-E. Wild-type (YLH164), sas3 ${Delta}$ (YLH165), set1 ${Delta}$ (YLH312), set2 ${Delta}$ (YLH324), and set1 ${Delta}$ set2 ${Delta}$ (YLH325) strains were assayed for mating efficiency with YLH101 as a tester strain.

To provide further support for the involvement of SET1 and SET2in NuA3 function, we asked whether deletion of these genes generateda sas3 ${Delta}$ phenotype. We initially looked for synthetic interactionsof mutations of SET1 and/or SET2 with mutations in GCN5 andADA2. However, we found that unlike ada2 ${Delta}$ sas3 ${Delta}$ mutants, ada2 ${Delta}$ set1 ${Delta}$ mutants did show a synthetic growth defect (data not shown),which suggested that Set1p has functions overlapping those ofone of the Gcn5p-dependent HAT complexes. This may be due toloss of histone H3 lysine 4 methylation or loss of another functionof Set1p, such as the recently demonstrated role for Set1p inmethylation of the kinetochore protein Dam1p (74). Therefore,to further address the role of SET1 and SET2 in NuA3 function,we looked for a different sas3 ${Delta}$ phenotype in set1 ${Delta}$ and set2 ${Delta}$ mutants.We and others have previously shown that deletion of genes encodingcomponents of NuA3 rescues the silencing defect resulting frommutations of the Rap1p and Abf1p binding sites at the HMR locus(22, 51). Although the molecular basis for this phenotype insas3 ${Delta}$ mutants has not been confirmed, it is most likely due todecreased levels of acetylation spreading past the impairedHMR-E silencer into the heterochromatic mating type locus (27,61). Figure 4B shows that deletion of SET2 also rescues thissilencing defect, suggesting that Set2p is required for thefunction of the NuA3 complex. In contrast, a set1 ${Delta}$ mutant didnot show this phenotype, and in fact SET1 has been previouslyshown to be required for the maintenance of silencing at thetelomeres, the rRNA genes, and the HML locus (6, 29, 42). Itwas later demonstrated that the role of Set1p in silencing wasindirect and was due to the fact that H3 lysine 4 methylationblocks the binding of the silence information regulator Sir3pto transcriptionally active regions of the genome (53). Lossof lysine 4 methylation therefore results in a titration ofSir3p away from heterochromatic regions, leading to a loss ofsilencing. A dual function for lysine 4 methylation in blockingthe binding of Sir3p, while enhancing the recruitment of NuA3,may explain why SET1 deletions did not rescue HMR silencingdefects. The loss of NuA3 recruitment in a set1 ${Delta}$ strain may becompensated for by a concomitant decrease in the levels of Sir3ppresent at HMR. Consistent with this hypothesis, deletion ofSET2 does not rescue silencing in a set1 ${Delta}$ strain (Fig. 4B). Consequently,these results are consistent with a role for SET2, and possiblySET1, in the function of NuA3.

Histone H3 lysine residues 4 and 36 are required for NuA3 function. Although it appears that methylation of H3 lysines 4 and 36mediates NuA3 interaction with chromatin, the possibility existsthat Set1p and Set2p are methylating nonhistone substrates,and it is this methylation that mediates NuA3 function. Set1phas recently been implicated in the methylation of the kinetochoreprotein Dam1p (74), and the possibility that Set2p also methylatesnonhistone substrates has not been ruled out. To clarify thisissue, three histone H3 substitution mutants were created bysite-directed mutagenesis, resulting in amino acid changes fromlysine to arginine at lysines 4, 36, and both 4 and 36. Chromatinpulldown assays were then performed on all mutants after epitopetagging the C termini of H2B with a TAP tag and Sas3p with atriple HA tag. The results of subsequent immunoblot analysisclosely mirrored the results obtained with the set1 ${Delta}$ and set2 ${Delta}$ mutants shown in Fig. 4A. Mutation of lysine 4 to arginine resultedin increased levels of Sas3HA in the whole-cell extract andan accompanying minor increase in the pulldown of Sas3p withHtb1p (Fig. 5A). Mutation of lysine 36 to arginine resultedin a modest decrease in the binding of Sas3p to chromatin, butwhen the higher levels of Sas3p present in these cells is takeninto account, this level represents a substantial decrease inthe fraction of Sas3p that is binding chromatin. Mutation ofboth lysines 4 and 36 resulted in almost total disruption ofSas3p-chromatin interactions. These results suggest that lysines4 and 36 are required for Sas3p, and consequently NuA3, to bindchromatin. To determine whether NuA3 directly interacts withmethylated histones, we examined the level of NuA3 acetylationof histone H3 tail peptides that were dimethylated on eitherlysine 4 or 36. The results showed no difference in the levelsof histone acetylation between these substrates and unmodifiedpeptides (data not shown). This may be indicative of a dependencyfor histone trimethylation for NuA3 acetylation, or the factthat the recruitment of NuA3 by histone methylation is mediatedby an intermediary factor.

View larger version (59K):
[in this window]
[in a new window]
FIG. 5. Histone H3 residues lysine 4 and 36 are required for NuA3 function. (A) Chromatin pulldown assays were performed with the indicated Sas3HA₃-expressing strains (YDM141, YDM142, YDM192, YDM207, and YDM208), and the resulting samples were subjected to Western blotting with immunodetection with anti-HA antibodies (Sas3HA) or peroxidase-antiperoxidase antibodies (Htb1TAP). Yeast whole-cell extract (1.4% of input) was also blotted for HA (INPUT). (B) Tenfold dilutions of yeast strains YLH347 (hht1 ${Delta}$ hht2 ${Delta}$ ), YLH348 (gcn5 ${Delta}$ hht1 ${Delta}$ hht2 ${Delta}$ ), and YLH349 (ada2 ${Delta}$ hht1 ${Delta}$ hht2 ${Delta}$ ) containing the indicated plasmids were plated on synthetic complete medium (control), synthetic complete medium with 5-FOA, or synthetic histidine dropout medium with 50 mM 3-AT and incubated at 30°C for 3 days. (C) Tenfold dilutions of yeast strains YLH224 (hht1 ${Delta}$ hht2 ${Delta}$ ) and YLH315 (sas3 ${Delta}$ hht1 ${Delta}$ hht2 ${Delta}$ ) containing the indicated plasmids were plated on yeast-peptone-dextrose medium and incubated at 20°C, 30°C, and 37°C for 3 days.

While histone H3 lysines 4 and 36 are required by NuA3 to interactwith chromatin, we were unsure as to whether these residuesare essential for NuA3 function. If these residues are indeedrequired by NuA3 for its function, it therefore follows thatmutations of these residues should result in sas3 ${Delta}$ phenotypes.With this question in mind, plasmids encoding wild-type andmutant versions of histone H3 were transformed into wild-type,gcn5 ${Delta}$ , and ada2 ${Delta}$ strains that carried a wild-type histone H3 geneon a URA3-based plasmid. Transformants were plated on 5-FOAto select for loss of the URA3 plasmid, and Fig. 5B shows thatmutation of both lysines 4 and 36 to arginine resulted in sensitivityto 5-FOA in a gcn5 ${Delta}$ mutant but not in an ada2 ${Delta}$ mutant. A similarresult was seen upon mutation of lysines 4 and 36 to alanines(data not shown). The strength of these phenotypes correlatedclosely with the effect of SET1 and SET2 deletion on the bindingof Sas3p to chromatin. Taken together, these results confirmthat the binding and acetylation of histone H3 by NuA3 are dependenton prior methylation of histones by Set1p and Set2p.

To verify that NuA3 functions in the same pathway as lysine4 and 36 methylation, we examined the phenotypes of wild-typeand sas3 ${Delta}$ strains carrying arginine substitutions of lysines4 and 36. If NuA3 functions in the same pathway as methylationof these residues, then we would expect that deletion of SAS3from K4R, K36R, and K4R K36R strains would not result in anyadditional phenotypes. Figure 5C shows that the growth of K4R,K36R, and K4R K36R mutants was indistinguishable from that ofwild-type histone H3 strains at 20°C and 30°C, and deletionof SAS3 did not affect this. However, mutation of lysine 36to arginine resulted in a mild temperature sensitivity, as indicatedby poor growth at 37°C. This temperature sensitivity wasfurther pronounced in the K4R K36R mutant. This synthetic phenotypeis not surprising when one considers the accumulating numberof factors that are recruited by methylation of histone H3 inyeast. These include the Isw1p ATPase (55) and the SAGA andSLIK/SALSA HATs (48), which bind methylated lysine 4, the Rpd3SHDAC (8, 26), which is targeted by methylation of lysine 36,and the NuA3 and Esa1p HATs (40, 57), which are targeted bymethylation of both lysines 4 and 36. Interestingly, deletionof SAS3 partially rescued the temperature sensitivity of theK36R mutant, as well as the K4R K36R mutant, albeit to a lesserextent. Because lysine 36 methylation is also required for targetingof Esa1p and the Rpd3S HDAC to transcribed genes (8, 26, 40),it is difficult to interpret these results. One possibilityis that NuA3 and Rpd3S have opposing functions in the cell.NuA3 may acetylate lysine 14 of histone H3 to facilitate thepassage of RNA polymerase, whereas the Rpd3S HDAC removes thismodification. It has been proposed that removal of transcriptionelongation-associated acetylation is required to suppress subsequentintragenic transcription initiation, and consistent with this,loss of the Rpd3S HDAC or mutation of lysine 36 results in crypticinitiation from intragenic sites within multiple genes (8).This may then result in a growth defect at higher temperatures.In the K36R mutant, however, NuA3 would still be recruited bylysine 4 methylation, and thus a lysine 36 mutant would maintainhigher histone acetylation levels than a wild-type strain. Thismay be rescued by deletion of SAS3, which removes the acetylationin the absence of Rpd3S recruitment.

The Set1p and Set2p methyltransferases mediate steady-state levels of histone acetylation on bulk histones. The fact that NuA3 binding to chromatin is dependent on SET1and SET2 suggests that these methyltransferases may mediatehistone acetylation. To address this, we asked whether lossof SET1 and/or SET2 results in a loss of steady-state histoneacetylation in vivo. This experiment was complicated by thefact that lysine 4 methylation has recently been shown to enhancehistone H3 acetylation by the GCN5-dependent SAGA and SLIK/SALSAHAT complexes (48). To rule out the possibility that any observedloss of histone acetylation in set1 ${Delta}$ mutants was due to a failureto recruit these HATs, we examined the histone acetylation ofbulk histones in wild-type, set1 ${Delta}$ , set2 ${Delta}$ , and set1 ${Delta}$ set2 ${Delta}$ strainsthat had been generated in a gcn5 ${Delta}$ background. We normalizedfor levels of histone H3 by immunoblotting with an antibodyspecific for the carboxyl terminus of histone H3 and probedan identical blot with an antibody that is specific for acetylatedlysine 14 of histone H3 (62). Figure 6A shows that there wasno detectable difference in the immunoblot signals seen in thewild type versus the set1 ${Delta}$ , set2 ${Delta}$ , or set1 ${Delta}$ set2 ${Delta}$ mutants when whole-cellextracts were probed with this ${alpha}$ AcK14 antibody. However, thisantibody also failed to detect a difference in the acetylationsignals seen in the wild type versus a K14R mutant (Fig. 6B),suggesting that the signals observed do not accurately representthe levels of lysine 14 acetylation seen in the cell. To circumventthis problem, we made use of a second antibody that was raisedagainst histone H3 that was diacetylated on lysines 9 and 14.This antibody also failed to detect a difference in histoneH3 acetylation in the wild type versus a K14R mutant (Fig. 6B);however, this was expected since this antibody also detectshistone H3 acetylated at lysine 9, which has been shown by othersto be the predominant epitope recognized by this antibody (75).However, since of all our strains were generated in a gcn5 ${Delta}$ background,we did not anticipate this being a problem since deletion ofGCN5 has been shown to result in the loss of the majority ofhistone H3 lysine 9 acetylation in vivo (75). When this antibodywas used to probe normalized levels of histone H3 from wild-type,set1 ${Delta}$ , set2 ${Delta}$ , and set1 ${Delta}$ set2 ${Delta}$ strains, the level of histone acetylationin set1 ${Delta}$ and set2 ${Delta}$ strains was seen to modestly decrease. However,concomitant deletion of SET1 and SET2 resulted in a more substantialdecrease in histone acetylation (Fig. 6A), suggesting that SET1and SET2 mediate bulk Gcn5p-independent histone acetylationin vivo. This, taken together with the decreased interactionof NuA3 with chromatin in set1 ${Delta}$ and set2 ${Delta}$ mutants, suggests thatNuA3 function is dependent on the methylation of lysine 4 or36 within the H3 tail.

View larger version (33K):
[in this window]
[in a new window]
FIG. 6. The Set1p and Set2p methyltransferases mediate the steady-state levels of histone acetylation of bulk histones. Whole-cell extracts from the indicated strains (YLH326, YLH327, YLH328, YLH329 [A], and Y224 transformed with pHHT2, phht2K14R, and phht2K9,14,18,23R [B]) were subjected to Western blotting and immunodetection with anti-acetylated histone H3 lysine 14 antibodies ( ${alpha}$ AcK14) and anti-diacetylated histone H3 lysine 9 and 14 antibodies ( ${alpha}$ AcK9,14). Samples were normalized for levels of histone H3 by immunoblotting with anti-carboxyl terminus of histone H3 antibodies ( ${alpha}$ H3).

DISCUSSION

Top

Discussion

NuA3 acetylates histones in vivo. The NuA3 HAT complex was originally identified on the basisof its ability to acetylate histone H3 in vitro (25). However,the possibility existed that histone H3 was not the main substratefor NuA3 in vivo. In this study, we have shown that Sas3p iscapable of interacting with chromatin in vivo and that thisinteraction is dependent on the H3 tail. Moreover, deletionof the H3 tail, or mutation of H3 lysine 14 to arginine, issynthetically lethal with gcn5 ${Delta}$ , indicating that the gcn5 ${Delta}$ sas3 ${Delta}$ synthetic lethality is due to loss of lysine 14 acetylationby NuA3. Finally, we have previously shown that a temperature-sensitivemutation of SAS3 in a gcn5 ${Delta}$ background results in a decreasein H3 acetylation levels seen in bulk histones (21). Althoughwe cannot rule out the possibility that Sas3p is also acetylatinganother substrate, this finding indicates that NuA3 is a physiologicalhistone H3 acetyltransferase.

The synthetic lethality between gcn5 ${Delta}$ and histone H3 ${Delta}$ 3-29 andK14R mutants has been reported previously (75). The same workshowed that mutation of H3 lysine 9 in a gcn5 ${Delta}$ background doesnot result in an enhanced phenotype over gcn5 ${Delta}$ alone, which isconsistent with the fact that Gcn5p is functioning in the samepathway as lysine 9 acetylation. Others have shown that Gcn5pacetylates H3 lysines 9, 18, 23, and 27 but not lysine 14 invivo (28, 62). This, taken together with the gcn5 ${Delta}$ K14R syntheticphenotype, suggests that H3 lysine 14 is acetylated by a HATother than Gcn5p. The fact that deletion of SAS3 recapitulatesphenotypes seen upon mutation of lysine 14 suggests that thisHAT is NuA3. The evidence that Gcn5p does not acetylate lysine14 in vivo is puzzling when one considers that, in vitro, nativeGcn5p HAT complexes target primarily lysine 14 when nucleosomalhistones are used as substrates (17). A similar result has beenseen with the elongator complex, which shows a preference foracetylating lysine 14 of peptides in vitro but a different substratespecificity in vivo (28, 68). One simple explanation for theseobservations is that Gcn5p and elongator acetylate lysine 14in vivo, but in the absence of these HATs, NuA3 is able to compensatesuch that no net loss of acetylation is observed. Alternatively,lysine 14 may be acetylated by NuA3 before other HATs have theopportunity to do so. NuA3 has been shown to interact with FACT,a complex with a role in DNA replication (25). It is possiblethat this interaction allows for acetylation of lysine 14 byNuA3 immediately following DNA replication, before HATs thatare targeted by other mechanisms can access this residue. However,this would argue that acetylation by NuA3 precedes transcription,which due to the link between transcription and histone H3 methylation,is not supported by the results of this study. A third hypothesisis that, in vivo, H3 lysine 14 is presented in such a way thatit cannot serve as a good substrate for Gcn5p complexes or elongator.

Histone methylation mediates global histone acetylation. Histone acetylation can be targeted to active regions of thegenome through a number of different mechanisms. The SAGA andNuA4 HAT complexes specifically interact with DNA-bound transcriptionalactivators, thus directing histone acetylation to the promoterregion of genes (7, 23, 24, 64, 66, 67). The elongator complexinteracts with the elongating form of RNAPII and thus wouldbe expected to be targeted to regions of transcription activity(70). The NuA3 HAT complex has been shown to interact with FACT,a protein complex which modulates chromatin structure (25).However, the abundance of FACT in the cell approaches that ofthe nucleosome (5, 69), and we estimate that NuA3 is presentin less than 5% of this amount (unpublished data). Thus, ifNuA3 is targeted to specific regions of the genome, anothermechanism must exist to control NuA3 recruitment. In supportof this, we find that the ability of NuA3 to bind chromatinis dependent on the Set1p and Set2p histone methyltransferasesand on histone H3 lysines 4 and 36, the substrates for thesemethyltransferases, respectively. Consistent with their rolein mediating histone acetylation, loss of SET1 and SET2 resultsin a moderate decrease in steady-state levels of histone acetylation,as shown by Western blot assays of bulk histones. These dataprovide support for a hypothesis whereby methylation of histonesprovides a signal for the recruitment of NuA3 and subsequencehistone acetylation.

The recruitment of Set1p and Set2p to genes is dependent onan association with RNAPII (20). Therefore, the ability of thesemethylation marks to mediate histone acetylation by NuA3 suggeststhat histone acetylation is dependent on transcription. A numberof studies have examined the correlation between transcriptionand histone H3 acetylation in yeast. However, the majority ofthese studies employed an antibody that was raised against H3peptide tails that were diacetylated at lysines 9 and 14, andwork with histone mutants suggests that these antibodies recognizeprimarily acetylated lysine 9 (46, 75). Two studies using antibodiesdirected specifically against acetylated H3 lysine 14 to examinegenome-wide acetylation profiles of open reading frames showeda positive correlation between H3 lysine 14 acetylation andtranscription (31, 46). Interestingly, a similar correlationwas seen between transcription and acetylation of H3 lysine23, which can also be acetylated by NuA3 in vitro (31, 46).

The ability of histone methylation to mediate histone acetylationhas also been observed by others. Methylation of H3 lysine 4has been shown to enhance histone acetylation by the SAGA andSLIK/SALSA HATs (48). Additionally, lysine 36 methylation mediatesthe interaction of the Rpd3S HDAC with transcribed genes (8,26), suggesting that histone acetylation can be positively andnegatively regulated by histone H3 methylation. SET1 and SET2are also required for the association of the NuA4 complex withthe MET16 and RPS11B promoters and the subsequent acetylationof lysine 8 on histone H4 (40). Interestingly, genome-wide histoneacetylation studies indicate that there is a strong positivecorrelation between histone H4 lysine 8 and histone H3 lysine14 acetylation, suggesting that NuA3 and NuA4 may be recruitedthrough a shared mechanism (31).

The recruitment of SAGA, SLIK/SALSA, Rpd3S, and NuA4 by methylatedhistones is not an unexpected result, as all three complexescontain protein subunits with chromodomains (8, 13, 26, 40,48). A subset of chromodomain proteins have been shown to bindspecifically to methylated lysines (4), but none of the knownsubunits of NuA3 contain this motif. Recent studies, however,have identified novel protein domains which are capable of bindingmethylated histones. These include the WD40 repeat protein WDR5,which is a common component of multiple H3 K4 methyltransferasecomplexes in higher eukaryotes (71). Thus, it is possible thatanother protein motif within NuA3 is responsible for recognizingmethylated histones. One candidate would be the PHD finger withinthe Yng1p subunit of NuA3. This zinc finger-like motif occursin various proteins thought to be involved in chromatin-mediatedgene regulation (1). Despite the intriguing possibility thatthis motif mediates the interaction of NuA3 with chromatin,we find that Yng1p lacking the PHD domain is still able to rescuethe gcn5 ${Delta}$ yng1 ${Delta}$ growth defect, suggesting that this region ofthe protein is not required for the function of the NuA3 complex(22). It is possible that NuA3 binding to methylated histonesis mediated by intermediary factors such as an unidentifiedprotein or another histone modification that is dependent onhistone methylation. A similar result was seen for the yeastIsw1p ATPase, which requires SET1 to bind chromatin but cannotbind methylated peptides in a purified system (55). Regardlessof the mechanism, our work clearly indicates that NuA3 bindingto methylated histones is an important means for directing theHAT activity of NuA3 to chromatin and explains how histone H3methylation can mediate transcription.

ACKNOWLEDGMENTS

Support for this work was provided by a grant to L.H. from theCancer Research Institute of the Canadian Institute for HealthResearch. L.H. is a scholar of the Michael Smith Foundationfor Health Research.

We gratefully acknowledge the valuable comments provided byJacques Côté, Jerry Workman, and members of theMolecular Epigenetics Group of the Life Sciences Institute atthe University of British Columbia. We are also grateful toFred Winston and Jasper Rine for providing yeast strains andplasmids.

FOOTNOTES

* Corresponding author. Mailing address: Department of Biochemistry and Molecular Biology, University of British Columbia, 2350 Health Sciences Mall, Vancouver, BC, Canada V6T 1Z3. Phone: (604) 822-6297. Fax: (604) 822-5227. E-mail: ljhowe{at}interchange.ubc.ca.

REFERENCES

Top