Deubiquitination of Histone H2B by a Yeast Acetyltransferase Complex Regulates Transcription*

Post-translational modifications of the histone protein components of eukaryotic chromatin play an important role in the regulation of chromatin structure and gene expression (1). Given the requirement of Rad6/Bre1-dependent ubiquitination of histone H2B for H3 dimethylation (at lysines 4 and 79) and gene silencing (2–7), removal of ubiquitin from H2B may have a significant regulatory effect on transcription. Here we show that a putative deubiquitinating enzyme, Ubp8, is a structurally nonessential component of both the Spt-Ada-Gcn5-acetyltransferase (SAGA) and SAGA-like (SLIK) histone acetyltransferase (HAT) complexes in yeast. Disruption of this gene dramatically increases the cellular level of ubiquitinated-H2B, and SAGA and SLIK are shown to have H2B deubiquitinase activity. These findings demonstrate, for the first time, how the ubiquitin moiety can be removed from histone H2B in a regulated fashion. Ubp8 is required for full expression of the SAGA- and SLIK-dependent gene GAL10 and is recruited to the upstream activation sequence (UAS) of this gene under activating conditions, while Rad6 dissociates. Furthermore, trimethylation of H3 at lysine 4 within the UAS increases significantly under activating conditions, and remarkably, Ubp8 is shown to have a role in regulating the methylation status of this residue. Collectively, these data suggest that the SAGA and SLIK HAT complexes can regulate an integrated set of multiple histone modifications, counteracting repressive effects that alter chromatin and regulate gene expression.

Enzymes responsible for the post-translational modification of histones via acetylation, methylation, and ubiquitination are rapidly emerging. Most of these, thus far, are components of stable multiprotein complexes that act on nucleosomal substrates (8), regulating specific biological processes through a possible histone code that reads distinct sequential and/or combinatorial patterns of modifications (1). The HAT enzyme, Gcn5, is the catalytic subunit of two related high molecular weight (ϳ1.8 MDa) protein coactivator complexes in Saccharomyces cerevisiae called SAGA (Spt-Ada-Gcn5-acetyltransferase) and SLIK (SAGA-like)/SALSA (SAGA altered, Spt8 absent) (9 -11). SAGA and SLIK have been shown to be required for the expression of a subset of genes as well as being functionally redundant with the general transcription factor, TFIID, in the transcription of most RNA polymerase II-transcribed genes (12). However, distinct and separate functions for the SAGA and SLIK HAT complexes have been established through compositional differences (10,11,13).
This study demonstrates that the yeast deubiquitinating enzyme, Ubp8, is a component of both SAGA and SLIK and mediates the removal of ubiquitin from histone H2B. The process of H2B monoubiquitination in yeast has recently been shown to regulate H3 methylation and gene silencing in a unidirectional, trans-histone fashion (2)(3)(4)(5)(6)(7)14). Here, we show that Ubp8 is required for full expression of the GAL10 gene and that Ubp8 is present at the promoter of this gene under activating conditions. Furthermore, we find that the removal of ubiquitin from H2B is involved in the regulation of histone H3 (Lys 4 ) trimethylation, a feature of transcriptional activation (15). Our data show a unique attribute of the SAGA coactivator complexes in the dual acetylation of histone H3 and deubiquitination of histone H2B, bridging two histone components of the nucleosome, while regulating the methylation of histone H3. These results have a profound implication for the understanding of how a cascade of post-translational histone modifications is generated in the process of gene expression.

EXPERIMENTAL PROCEDURES
Yeast Strains and Plasmids-SAGA and SLIK were isolated for mass spectrometric analysis from yeast strain CY396 (10). GST 1 -tagged Ubp8 protein was expressed in strain EJ758 (Research Genetics). Strain YZS236, expressing protein A-His 7 -tagged Rad6 protein, was derived from strain Y131 (hta1-htb1⌬::LEU2 hta2-htb2⌬ HTA1-HTB1 (2 URA3)). BY4741 and derivative ubp8⌬, spt20⌬, and rad6⌬ strains were obtained from Research Genetics. Strains MBY1217, YZS276 (FLAG-HTB1) and YZS277 (FLAG-HTB1-K123R) are described previously (2). Strain YZS282 (ubp8⌬::KAN) is derived from YZS276 and * This work was supported by NIDDK, National Institutes of Health (NIH) Grant DK58646 (awarded to P. A. G.), Grant RR11823-03 (to the National Center for Research Resources Yeast Center), National Science Foundation Science and Technology Center Grant BIR9214821AM (to J. R. Y.), and grants from the National Center for Research Resources (to C. D. A.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.   Purification of SAGA and SLIK Complexes and Mass Spectrometry-For mass spectrometric analysis, SAGA and SLIK were isolated and analyzed as described previously (10,16). The complexes were also partially purified from whole cell extracts of wild type, ubp8⌬ and GST-Ubp8 yeast strains grown to mid-log phase in 4 liters of yeast extract-peptone-dextrose medium, by sequential Mono Q HR 16/10, Mono Q HR 5/5 column, Superose 6 HR 10/30 columns, and a second Mono Q HR 5/5 column (Amersham Biosciences) (10).
Western Blot Analysis and HAT Assays-Proteins from column fractionations were processed for immunoblotting using Ada2, Taf9, or GST (Santa Cruz Biotechnology) antisera. HAT assays were performed with HeLa free histones or nucleosomes as described previously (17). For H2B studies, whole cell lysates were prepared and FLAG analysis was performed as described (2). Histones from wild type and ubp8⌬ mutant yeast were acid extracted and processed for immunoblotting using rabbit H3 antisera specific for the mono-, di-, or trimethylated states of lysine 4 (Abcam). Quantitation of relative fold changes (Western/Ponceau) was performed using ImageQuant software.
Deubiquitination Assay-Plasmid-borne HA-tagged ubiquitin and FLAG-tagged H2B were both expressed in strain YJD1 under galactose induction, and a trichloroacetic acid lysate preparation was immunoprecipitated using sequential FLAG M2-agarose (Sigma) and HA Affinity Matrix (Roche Applied Science) resins, largely as described previously (18). For the assay, 4 l of resuspended beads containing FLAG-HA-uH2B was mixed with 4 l of a Mono Q fraction containing SAGA, SLIK, or Mono Q buffer only, along with 100 g/ml bovine serum albumin, in a 20 l reaction containing 50 mM KCl, 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 5% glycerol, 1 mM dithiothreitol performed for 1 h at 30°C. Proteins were analyzed by Western blotting using FLAG M1 (Sigma) and HA-peroxidase (Roche Applied Science) antisera.
Chromatin Immunoprecipitation (ChIP) Assays and Northern Blot Analysis-Chromatin preparations and immunoprecipitations were performed as described previously (19). For ChIPs and Northerns, yeast were grown in synthetic complete (SC) medium containing 2% raffinose and then reseeded and grown in SC-raffinose medium or SC medium containing 2% galactose for an additional 4 h. Chromatin from whole cell extracts of strains expressing GST-Ubp8 or PrA-Rad6-His 7 , and respective no tag controls, was then precipitated using glutathione-Sepharose (Amersham Biosciences) or Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) and IgG-Sepharose 6 Fast Flow (Amersham Biosciences), respectively. ChIPs were also performed from wild type BY4741, ubp8⌬, and set1⌬ strains with H3 (Lys 4 ) mono-, di-, or trimethyl anitsera (Abcam). The input and precipitated DNA were amplified by PCR using oligonucleotide primers specific for the UAS within the GAL1-10 or ACT1 promoter regions. Quantitation was performed using an Al-phaImager 2000 documentation and analysis system from Alpha Innotech. The ratios of immunoprecipitated to input DNA were calculated and normalized to wild type or tag raffinose values, and these values are shown as relative amounts of IP/input signal. RNA purification and Northern blotting were performed as described previously (10).

Ubp8 Is a Stable Component of the Yeast SAGA and SLIK
HAT Complexes-To gain more insight into the transcriptional regulatory function of SLIK, we used mass spectrometric analysis of highly purified complex to identify a putative deubiquitinating enzyme, Ubp8, as a novel component (Fig. 1A). These data are in agreement with large-scale proteomics studies finding Ubp8 in a complex with Gcn5 and Spt7 (20,21) and also with a study demonstrating that HA-tagged Ubp8 can co-immunoprecipitate certain SAGA/SLIK components (22). To test whether Ubp8 was SLIK-specific, we utilized a strain expressing GST-tagged Ubp8 protein. Immunoblotting of partially purified SAGA and SLIK revealed that GST-Ubp8 pro- Ubp8 is a structurally nonessential component of the SAGA and SLIK HAT complexes. A, silver-stained gels of highly purified SAGA and SLIK. Known components are grouped by class of transcriptional regulator, namely Spts, Adas, and Tafs. The protein band identified by mass spectrometry as Ubp8 is highlighted. Numbers in the center of the panel are molecular masses in kilodaltons. B, Western blots, using GST and Ada2 antisera, of partially purified SAGA and SLIK Mono Q fractions prepared from GST-Ubp8 yeast extract. Through the purification scheme, the SAGA and SLIK fractions were pooled individually. C, Western blots, using Ada2 and Taf9 antisera, and HAT assays of Mono Q fractions prepared from wild type and ubp8⌬ strains. A type B H4 HAT complex elutes between SAGA and SLIK from these columns. tein co-fractionated with both complexes (Fig. 1B), although existing in higher stoichiometric proportion in SLIK, consistent with the identification of Ubp8 as part of the SLIK complex (Fig. 1A).
The polypeptides Spt20, Spt7, and Ada1 are required for the overall structural integrity of both SAGA and SLIK, because deletions of any of these genes cause severe phenotypes and do not allow normal chromatographic isolation of either complex (9,10,23). To test whether the Ubp8 protein is also required for structural integrity, we compared HAT complex fractionation profiles of wild type yeast to those bearing a UBP8 deletion by anion exchange chromatography. We found that SAGA and SLIK complexes lacking Ubp8 are intact and can acetylate free and nucleosomal histone H3 in vitro (Fig. 1C). Further anion exchange and size exclusion chromatography of the ubp8⌬ extract indicated that the two mutant complexes have similar charge and size to those of wild type. Therefore, Ubp8 is not necessary for the integrity or HAT activity of either complex, concurring with the lack of any marked phenotype seen with the ubp8⌬ yeast (Ref. 24 and data not shown).
Ubp8 Regulates H2B Deubiquitination at Lysine 123-A possible role of the Ubp8 protein within a HAT complex is to deubiquitinate chromatin-associated substrates. A good candidate substrate appeared to be histone H2B, which in yeast is monoubiquitinated at the highly conserved lysine 123 residue (14). Most transcription studies thus far have concentrated on the addition of ubiquitin to proteins (25), even though ubiquitination is an enzymatically reversible process, as evidenced in deubiquitinating activities associated with the proteosome (26). For regulation of rDNA and telomeric silencing, ubiquitination of lysine 123 in H2B by Rad6 has recently been demonstrated to be required for dimethylation of H3 at lysines 4 and 79 by their respective site-specific histone methyltransferase enzymes, Set1 and Dot1 (2)(3)(4)(5)(6)(7)14). Di-methylation of H3 at these residues and functional telomeric silencing are impaired in the absence of Rad6 or an intact H2B (Lys 123 ) ubiquitination site, indicating that conjugation of ubiquitin to H2B controls these methylation events on H3. Using FLAG-tagged H2B yeast strains, we found that disruption of UBP8 caused a dramatic increase in the monoubiquitinated form of H2B as compared with wild type (Fig. 2A). The disruption of UBP3, another putative ubiquitin-specific protease shown to be involved in transcriptional silencing (27) and to associate with the SAGA/SLIK component Ada3 (28), had no apparent effect on uH2B (see the Supplemental Figure). This result indicates that Ubp8 specifically regulates the deubiquitination of H2B at lysine 123 in vivo.
Since Ubp8 is a component of SAGA and SLIK, we wanted to test whether these two complexes were capable of directly hydrolyzing the ubiquitin moiety from histone H2B. To this end, we purified ubiquitinated-H2B from a yeast strain coexpressing FLAG-tagged H2B and HA-tagged ubiquitin proteins. Using this substrate, we show that SAGA and SLIK have in vitro H2B deubiquitinase activity (Fig. 2B). The detection of free HA-ubiquitin specifically after the addition of SAGA or SLIK confirms the ubiquitin-specific protease function of these two complexes. This result extends the in vivo finding that Ubp8 deubiquitinates H2B and establishes that this histone modification is also catalyzed, in vitro, in the context of SAGA and SLIK.
Ubp8 Regulates Bulk H3 Methylation at Lysine 4 -Given that ubp8⌬ yeast have elevated uH2B and that this modification is required for methylation of H3 (Lys 4 ), we hypothesized that the ubp8⌬ strain might have increased H3 methylation at this residue. To test this, we compared H3 (Lys 4 ) methylation of histones from wild type yeast to that of ubp8⌬ by Western analysis using antibodies that are specific for the mono-, di-, or trimethylated form of histone H3 at lysine 4. Although we see a slight change in dimethylation, monomethylation of H3 (Lys 4 ) increases significantly in the ubp8⌬ mutant, while trimethylation at this residue decreases (Fig. 2C). These results indicate that removal of ubiquitin from H2B regulates the methylation state of lysine 4 on H3, promoting a transition from a mono-to trimethylated state. A recent study using these antibodies in ChIP analyses suggests that trimethylated H3 (Lys 4 ) is present exclusively at active genes, while the dimethylated residue is in chromatin of both repressed and active genes (15). The biological consequence of monomethylation on H3 (Lys 4 ) has not yet been explored. Our results suggest that deubiquitination by Ubp8 affects bulk H3 lysine methylation status in vivo, promoting trimethylation of H3 (Lys 4 ) and as part of HAT/co-activator complexes may promote gene activation.
Ubp8 Is Recruited to a SAGA/SLIK-dependent Promoter during Activation, while Rad6 Dissociates-To utilize galactose as a carbon source, derepression of a number of GAL genes in  H3 (Lys 4 ). A, Western blot using Flag antiserum of whole cell lysates from the YZS276 wild type strain, and its derivatives YZS277 (H2B-K123R) and YZS282 (ubp8⌬). B, Western blots using FLAG and HA antisera of deubiquitination reactions performed with purified SAGA, SLIK, or control buffer and ubiquitinated-H2B substrate isolated from a strain expressing FLAG-tagged H2B and HAtagged ubiquitin. Eluates from the FLAG IP and subsequent HA IP are shown, along with whole cell lysate of strain YZS282. NS, nonspecific; asterisk, FLAG-HA-uH2B is not detected by the FLAG antiserum under these conditions. C, Western blots and ponceau staining of acid-extracted histones from wild type and ubp8⌬ strains using H3 (Lys 4 ) methyl antisera. Numbers in parentheses show quantitation of fold changes relative to wild type. yeast requires SAGA or SLIK. Binding of these complexes to the UAS-bound activator Gal4 influences recruitment of TATAbinding protein through interaction with the coactivator component Spt3 (10, 29 -32). Thus, we decided to use the SAGA/ SLIK-dependent gene GAL10 and its UAS GAL1-10 to study the effects of Ubp8 on transcription. To determine whether Ubp8 is present at the GAL1-10 UAS under activating conditions, we prepared chromatin from a strain expressing GSTtagged Ubp8 proteins. Cells grown in raffinose, or shifted to growth in galactose, were subjected to ChIP analysis using GST antisera and primers specific for the UAS within the GAL1-10 promoter region. Our results indicate that Ubp8 is present at the UAS specifically under activating conditions (Fig. 3A) and that this localization correlates well with recruitment of SAGA and SLIK and occurrence of H3 (Lys 9/14 ) acetylation (10), consistent with its function as part of these HAT complexes. In contrast, we observe that Rad6 is present at the GAL1-10 UAS under raffinose-noninducing conditions and dissociates from this promoter region during activation (Fig. 3A). As a control, significant changes in Rad6 or Ubp8 were not observed at the control ACT1 promoter. These data are suggestive of opposing functional roles of Rad6 and Ubp8 in histone ubiquitination during transcriptional activation of the GAL genes.
Trimethylation of H3 (Lys 4 ) at the GAL1-10 Promoter Is Dependent on Ubp8 and Set1 during Activation-We also investigated changes in the methylation state of H3 (Lys 4 ) under activating conditions at the GAL1-10 UAS, compared with the control ACT1 promoter, and whether Ubp8 functions in the regulation of this histone modification. Using H3 (Lys 4 ) antibodies, we performed ChIP analysis from wild type, ubp8⌬, and set1⌬ strains. Since Set1 is thought to be the only H3 (Lys 4 ) methylating enzyme in yeast (15), we used the set1⌬ mutant as a background control. In wild type yeast, although mono-and dimethylation are relatively unchanged upon activation, trimethylation of H3 (Lys 4 ) at the GAL1-10 UAS increases. However, in the ubp8⌬ or set1⌬ mutant, trimethylation of H3 (Lys 4 ) is impaired during activation (Fig. 3B). Thus, our data demonstrate, for the first time, that nucleosomal H3 (Lys 4 ) within the GAL1-10 UAS is trimethylated upon activation and that this modification is dependent upon Ubp8.

H2B Ubiquitination and H3 Methylation Affect
Transcription of the GAL10 Gene-Since the methylation changes within the GAL1-10 promoter region are regulated by Ubp8, we investigated whether Ubp8 and Set1 had a direct role in the transcription of a gene regulated by this element. To address this, we isolated total RNA from wild type, ubp8⌬, and set1⌬ strains grown in raffinose-noninducing or galactose-inducing media and performed Northern analysis with a probe specific for GAL10 expression. We find that the ubp8⌬, and more significantly, set1⌬ mutants have decreased GAL10 expression compared with wild type, suggesting the Ubp8 and Set1 proteins are required for full transcription of this gene (Fig. 3C,  compare lanes 17, 18, and 20). Northern analysis of RNA isolated from an spt20⌬ strain, a mutation that disrupts SAGA/SLIK complex integrity, showed a deficiency in GAL10 expression slightly more severe than the ubp8⌬ mutant, yet not as drastic as the set1⌬ mutant (Fig. 3C). Thus, we demonstrate co-activating functions for the histone H3 methyltransferase, Set1, and the deubiquitinase, Ubp8, in SAGA/SLIKdependent transcription of the galactose-regulated GAL10 gene.
In contrast, after a 4-hour induction time, transcription of GAL10 is increased in the rad6⌬ mutant compared with the isogenic wild type (Fig. 3C, compare lanes 17 and 21). This hyperactivation of GAL10 is visualized more dramatically when comparing two-hour galactose inductions (Fig. 3C, compare lanes 10 and 14). Interestingly, the rad6⌬ mutant does not cause induction of GAL10 in noninducing conditions, suggesting that RAD6 is not sufficient to maintain repression of the GAL genes. These studies offer more evidence for the contrasting functional roles of Rad6 and Ubp8 in the transcriptional activation of galactose-regulated genes. Surprisingly, H2B-K123R yeast, bearing a mutation in the H2B ubiquitination site, have decreased GAL10 gene expression (Fig. 3C, compare  lanes 15 and 16), demonstrating the need for a modifiable lysine at this position. This result may indicate that H2B ubiquitination, and subsequent deubiquitination, are both required for transcriptional activation. Alternatively, this lysine may be subject to other post-translational modifications such as acetylation or methylation. and activated transcription. A, ChIP assays performed from strains expressing protein A-and His 7 -tagged Rad6 (YZS236) or GST-tagged Ubp8 and two respective isogenic control strains grown in raffinose-noninducing (R) and galactose-inducing (G) conditions. Shown are PCR products using GAL1-10 and ACT1 primers with input and immunoprecipitated DNA templates. Numbers in parentheses show quantitation of relative amounts of (IP/input) signal. B, ChIP assays from wild type, ubp8⌬, and set1⌬ strains using H3 (Lys4) antisera. C, Northern analysis of total RNA prepared from noninduced and galactose-induced H2B-K123R (YZS277) and isogenic wild type (YZS276), ubp8⌬, spt20⌬, rad6⌬, and isogenic wild type (BY4741), and set1⌬ (MBY1217) strains hybridized with a GAL10 probe. RNA hybridization with an ACT1 probe is used as the loading control.

DISCUSSION
Our studies call attention to Ubp8, a component of the two yeast HAT complexes SAGA and SLIK. We present evidence that Ubp8 functions in the hydrolysis of ubiquitin from histone H2B and is recruited to the GAL1-10 UAS together with other SAGA and SLIK components during transcriptional activation (10,30,31). The association of Ubp8 at this locus coincides with (Lys 9/14 ) acetylation (10) and (Lys 4 ) hypermethylation of H3 allowing for normal expression of the GAL10 gene. In view of studies demonstrating that Rad6-mediated ubiquitination of H2B regulates H3 dimethylation and gene silencing, our data suggest that the SAGA and SLIK HAT complexes may counter Rad6 by influencing multiple modifications on more than one histone. During consideration of this manuscript, Henry et al. (33) also reported Ubp8 as an H2B deubiquitinase within the SAGA complexes and, in keeping with our results, provide evidence of a role for Ubp8 in transcriptional activation of the GAL genes; however, they show that deletion of UBP8 results in an increase in H3 (Lys 4 ) trimethylation at the GAL1-10 core promoter (33).
In summary, we propose that repression of some genes through ubiquitinated-H2B is overcome by an activator-targeted HAT complex containing deubiquitinase activity. We favor the view that deubiquitination of H2B occurs in concert or prior to both histone acetyltransferase activity and the intricate regulation of methyltransferase activity. In turn, changes in chromatin structure and/or the epigenetic pattern of modifications within a promoter region allow recruitment of the transcription machinery.