Nucleophosmin/B23 Negatively Regulates GCN5-dependent Histone Acetylation and Transactivation*

Nucleophosmin/B23 is a multifunctional phosphoprotein that is overexpressed in cancer cells and has been shown to be involved in both positive and negative regulation of transcription. In this study, we first identified GCN5 acetyltransferase as a B23-interacting protein by mass spectrometry, which was then confirmed by in vivo co-immunoprecipitation. An in vitro assay demonstrated that B23 bound the PCAF-N domain of GCN5 and inhibited GCN5-mediated acetylation of both free and mononucleosomal histones, probably through interfering with GCN5 and masking histones from being acetylated. Mitotic B23 exhibited higher inhibitory activity on GCN5-mediated histone acetylation than interphase B23. Immunodepletion experiments of mitotic extracts revealed that phosphorylation of B23 at Thr199 enhanced the inhibition of GCN5-mediated histone acetylation. Moreover, luciferase reporter and microarray analyses suggested that B23 attenuated GCN5-mediated transactivation in vivo. Taken together, our studies suggest a molecular mechanism of B23 in the mitotic inhibition of GCN5-mediated histone acetylation and transactivation.

Nucleophosmin/B23 is a multifunctional phosphoprotein that is overexpressed in cancer cells and has been shown to be involved in both positive and negative regulation of transcription. In this study, we first identified GCN5 acetyltransferase as a B23-interacting protein by mass spectrometry, which was then confirmed by in vivo co-immunoprecipitation. An in vitro assay demonstrated that B23 bound the PCAF-N domain of GCN5 and inhibited GCN5-mediated acetylation of both free and mononucleosomal histones, probably through interfering with GCN5 and masking histones from being acetylated. Mitotic B23 exhibited higher inhibitory activity on GCN5-mediated histone acetylation than interphase B23. Immunodepletion experiments of mitotic extracts revealed that phosphorylation of B23 at Thr 199 enhanced the inhibition of GCN5-mediated histone acetylation. Moreover, luciferase reporter and microarray analyses suggested that B23 attenuated GCN5-mediated transactivation in vivo. Taken together, our studies suggest a molecular mechanism of B23 in the mitotic inhibition of GCN5-mediated histone acetylation and transactivation.
The acetylation of nucleosomal histones by histone acetyltransferases (HATs) 5 has been known for several decades. Histone-modifying enzymes, such as GCN5 (general control of amino acid synthesis 5) in the Spt-Ada-Gcn5 acetyltransferase (SAGA) and Esa1p in the NuA4 complex, can acetylate specific lysine residues in histone N-terminal tails (1). According to the histone code hypothesis (2), site-specific acetylation of histone tails may trigger chromatin remodeling that leads to retention of effector proteins to active promoters and formation of transcriptionally active chromatin regions.
GCN5 was originally identified as a transcriptional coactivator in yeast and was proposed to contribute to transcription by establishing interactions between certain activators and transcriptional complexes (3). GCN5 enhances transcription through its intrinsic acetyltransferase activity (4), which facilitates acetylation of histones and nonhistone substrates (3). Recruitment of the SAGA complex greatly increases transcriptional activation in vitro (5), and the requirement of GCN5 for chromatin remodeling has been demonstrated in vivo (6). However, the mechanism(s) that regulates GCN5 activity particularly in the context of histone acetylation has yet to be examined in detail. Potentially, GCN5 HAT activity can be regulated through posttranslational modifications, interacting with other proteins or at the level of substrate modifications. For example, phosphorylation by the Ku-DNA-dependent protein kinase or sumoylation may regulate GCN5-HAT activity (7,8). Also, interaction with the SANT domain of the Ada2 affects GCN5-HAT activity in yeast (9). Likewise, modification of substrates, such as phosphorylation of H3 serine 10, increases GCN5 acetylation activity toward lysine 14 of H3 (10,11).
Nucleophosmin/B23, also known as NPM1, NO38, or numatrin, has been identified as a phosphoprotein that possesses multiple cellular functions (12). It plays important roles in ribosome assembly, protein folding, and centrosome duplication (13). During interphase, the majority of B23 protein localizes to the nucleolus, whereas in mitosis it distributes throughout the nucleoplasm and associates with condensed chromosomes (14). B23 is phosphorylated by CDC2-cyclin B during mitosis (15), and inhibition of mitotic phosphorylation leads to dissociation of B23 from mitotic chromosomes concomitant with their decondensation (16). B23 is frequently targeted for genetic mutations by chromosomal translocations and point mutations in lymphomas and leukemias and is one of the most frequently mutated genes in acute myeloid leukemia (13). It has also been established that B23 can function as an acidic histone chaperone that helps assemble nucleosomes in vitro (17). B23 can either repress (18,19) or stimulate transcription (20,21), depending on the promoter context and/or inter-acting transcription factors. There are controversial reports concerning the role of B23 in the regulation of tumor suppressor p53 stability and transcription. Depending on experimental settings, B23 can either increase or decrease p53 transcription (22)(23)(24). Recent knock-out studies suggest that deletion of B23 in mice leads to embryonic lethality, genomic instability, and activation of p53 and p21/Waf1 (25,26).
In this study, we identified GCN5 as a B23-interacting protein. B23 inhibited GCN5-dependent acetylation of free and nucleosomal histones and blocked histones from being acetylated by GCN5. The inhibitory effect became more evident in the presence of mitotic B23. Mitotic extract depleted of threonine 199-phosphorylated B23 barely inhibited GCN5-dependent histone acetylation. B23 inhibited GCN5-dependent transactivation of the p21/Waf1 and Sp1 promoter. Microarray expression analysis revealed that a considerable number of genes were counterregulated by B23 and GCN5. We propose that B23 may negatively regulate GCN5-dependent transactivation through inhibition of GCN5-mediated nucleosomal acetylation.
Cloning and Site-directed Mutagenesis-Human B23 open reading frame cDNA was amplified from the human testis Marathon-ready cDNA library (Clontech) using the following pair of primers: 5Ј-gca gtc gac gac acc aac ATG GAA GAT TCG ATG GAC-3Ј and 5Ј-cgc gtt aac AAG AGA CTT CCT CCA CTG-3Ј. Sequencing analysis of the resulting plasmid confirmed that the B23 cDNA was identical to human B23 in Gen-Bank TM (accession number BC012566). The PCR product was then cloned into PstI and SalI sites of pCMVtag2C (Stratagene) with a FLAG tag, which was subsequently transferred into pBItet (Clontech), and pET16b (Novagen), respectively. B23-T199A (T199A) mutant was generated by site-directed mutagenesis using the QuikChange site-directed mutagenesis kit (Stratagene) and verified by sequencing analysis. The following primers were used for the site-directed mutagenesis: 5Ј-GAA ATC TAT ACG AGA TGC TCC AGC CAA AAA  TGC-3Ј and 5Ј-GCA TTT TTG GCT GGA GCA TCT CGT  ATA GAT TTC-3Ј. Human Gcn5L2 open reading frame, the long form of human Gcn5 (27) in pCMVspot6 vector, was purchased from ATCC. Sequencing analysis confirmed that the cDNA was identical to human Gcn5L2 (GenBank TM accession number BC039907) and designated as Gcn5 in this study. Gcn5 open reading frame was PCR-amplified and subsequently cloned into pBIND (Promega), pGEX-3X (Amersham Biosciences), pBI-tet with an HA tag, and pCMVtag2C. Gcn5 deletion mutants were created by PCR amplification from pGEX-3X-Gcn5 using specific primers against the truncated regions of Gcn5 (primer sequences are available upon request), followed by in-frame ligation of PCR products into the EcoRI and XhoI sites of pGEX-4T1. To engineer a double inducible vector (pBI-tet-HA-Gcn5/B23-FLAG), B23-FLAG was PCR-amplified from pCMVtag2C-B23 and cloned into pBI-tet-HA-Gcn5 so that HA-Gcn5 and B23-FLAG are under the control of a doxycycline-inducible bidirectional promoter. Human histone H3 open reading frame in POTB7 vector was obtained from ATCC, which was then cloned into BamHI and EcoRI sites of pGEX-3X. Sequencing analysis verified that the H3 clone was identical to the human histone H3 in GenBank TM (accession number BC006497). The plasmid pG5luc was obtained from Promega.
Histone Acetyltransferase Assays-Histones H3 and H4 were purchased from Upstate Biotechnology. HeLa cell short oligonucleosomes were the kind gifts from Dr. Jerry Workman. To prepare mononucleosomes, 5 l of short oligonucleosomes (1 mg/ml) were mixed with 14 l of Buffer R (10 mM Hepes-K ϩ , pH 7.5, containing 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM EGTA, 10% (v/v) glycerol, 10 mM ␤-glycerophosphate, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride) and 0.6 l of CaCl (100 g/ml), followed by MNase (0.2 unit) digestion for 10 min at room temperature. To check the quality of the mononucleosomes, aliquots of the MNase-treated samples were treated with proteinase K at 37°C for 30 min. DNA was isolated and analyzed on 1.1% agarose gel. More than 90% of the samples were mononucleosomes (data not shown). HAT assays were carried out in a buffer containing 50 mM Tris-HCl (pH 8.0), 0.1 mM EDTA, 1 mM dithiothreitol, 0.25 g/l acetyl-CoA, 10% glycerol with protease inhibitor mixture (Roche Applied Science), and 0.05 g/l APHA Compound 8, a histone deacetylase (HDAC) inhibitor (Sigma). Each reaction contained 2 g of H3, H4, GST-H3, or 0.5 g of mononucleosomes as substrates and the indicated amount of GST-GCN5 and B23-His proteins. The reactions were carried out at 30°C for 15 min and stopped by 2ϫ SDS sample buffer followed by heating at 100°C for 5 min. Acetylated histone H3 or H4 was identified by Western blot using anti-acetylated histone antibodies as described above.
Kinase Reaction with Interphase, Mitotic Lysates, or Cyclin E-CDK2-In total, 100 or 2 g of purified His-tagged B23 (B23 WT -His or B23 T199A -His) bound to Ni 2ϩ -nitrilotriacetic acid beads was incubated with 250 g of U2OS interphase or mitotic lysates or 1 g of purified GST-cyclin E-CDK2 complex, respectively, for 90 min at 30°C in a kinase reaction buffer containing 10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , 50 M ATP, and protease inhibitor mixture. B23-His proteins were then eluted from Ni 2ϩ -nitrilotriacetic acid beads by washing the beads with 10 mM Tris-HCl, pH 7.5, buffer containing 250 mM imidazole. The proteins were desalted by filtration through Zeba Desalt Spin Columns (Pierce) before use.
Luciferase Assays-To prepare interphase cell lysates for luciferase assay, 12 h after transfection, cells were first switched to serum-free medium for 12 h followed by cultivation in 10% FBS for 8 h and then harvested for luciferase assay. To prepare mitotic cell lysates for luciferase assay, 24 h after transfection, cells were treated with 0.2 g/ml nocodazole for 12 h, released for 1 h, and then harvested by mitotic shake-off. The cells were lysed with 100 l of passive lysis buffer (Promega) and incubated at room temperature for 30 min. The luciferase activities were measured using a TD20/20 Luminometer (Turner Designs) and normalized by ␤-galactosidase activities of the co-transfected pCMV␤Gal.
RT-PCR-Total RNA was purified from cells using Trizol reagent (Invitrogen). The cDNA was synthesized from 2 g of total RNA with the use of Superscript II RNase H minus reverse transcriptase (Invitrogen) and oligo(dT) primer (Roche Applied Science) according to the manufacturer's instructions. Human p21/Waf1-specific primers (5Ј-CCA GTG GAC AGC GAG CAG-3Ј; 5Ј-CCC TGC AGC AGA GCA GGT-3Ј) were used to check p21 expression by RT-PCR. The ␤-actin primers (5Ј-TCC CTG GAG AAG AGC TAC GA-3Ј; 5Ј-AGC ACT GTG TTG GCG TAC AG-3Ј) were used in the same reactions as an internal control.
GST Pull-down, Immunopurification, and Depletion-For GST pull-down, 3 g of purified B23 WT -His or B23 T199A -His was incubated with 50 l of glutathione-agarose beads containing ϳ5 g of full-length or truncated GST-GCN5 proteins in 1 ml of pull-down buffer (1ϫ phosphate-buffered saline, 0.6 g of bovine serum albumin, 1% Triton X-100, and protease inhibitor mixture) at room temperature for 60 min. Beads were washed three times with the same buffer and boiled in 1ϫ SDS sample buffer for 5 min. Proteins were separated in 10% SDS-polyacrylamide gel followed by Western blot using anti-B23 antibody to identify the His-B23.
Interphase or mitotic shake-off U2OS cells (1 ϫ 10 6 ) were lysed in 150 l of phosphate-buffered saline buffer containing 2 M okadaic acid, phosphatase (Sigma), and protease inhibitor mixtures (Roche Applied Science) by brief sonication for 15 s. To immunopurify endogenous B23, the lysates were incubated with 2 g of anti-B23 antibody (Sigma) for 4 h at 4°C and the B23-antibody complex was then recovered by 30 l of protein G beads (Amersham Biosciences). To deplete phospho-B23-Thr 199 , 200 g of mitotic extracts were incubated with 2 g of anti-phospho-B23-Thr 199 antibody for 4 h at 4°C. The phospho-B23-Thr 199 protein-antibody complex was then depleted from the extracts by 50 l of protein A beads.
Microarray Analysis-U2OS-Tet-On inducible stable clones carrying pBI-tet vector, pBI-tet-HA-Gcn5, pBI-tet-B23-FLAG, and pBI-tet-HA-Gcn5/B23-FLAG were induced by the addition of 2 g/ml doxycycline for 48 h. Total RNA was extracted from both induced and uninduced cells and subject to microarray hybridization using human 30k oligonucleotide chips according to the protocols of the Vanderbilt Microarray Shared Resources (available on VMSR website). Microarray expression data were analyzed with GenePix4.0 and Acuity4.0 software (Axon Instruments).
Immunoprecipitation, Mass Spectrometric, and Data Analyses-Immunoprecipitation of endogenous B23, GCN5, and FLAG-tagged GCN5 was performed in essentially the same manner as described in our previous studies (29). For mass spectrometric analysis, asynchronized U2OS cells were treated with or without 0.2 g/ml nocodazole for 24 h and then lysed in a buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.25% Nonidet P-40 and protease inhibitor mixture (Roche Applied Science). Immunoprecipitation was conducted with 4 g of anti-B23 antibody (Sigma) cross-linked to protein G-agarose beads. Beads were then washed eight times with lysis buffer. Bound proteins were eluted with 0.1 M glycine, pH 3.5, and precipitated in 10% trichloroacetic acid followed by an acetone wash. Samples were subjected to digestion for mass spectrometry as described in our previous studies (30). Briefly, the dry pellet was brought up in 8.0 M urea, reduced with Tris(2carboxyethyl)phosphine (Bond-Breaker by Pierce), and treated with iodoacetamide to alkylate cysteines. The sample was then predigested with endoproteinase Lys-C, diluted, and digested with trypsin. The resultant peptides were protonated with formic acid and loaded onto the back column of a three-phase MudPIT (multidimensional protein identification technology) setup (back column, Aqua C18 Reverse Phase-Luna Strong Cation Exchange; front column, Jupiter C18 Reverse Phase, all from Phenomenex) using a pressure cell (31). The sample peptides were then separated and analyzed by nano-LC-MS/MS using an UltiMate TM LC pump (LC Packings) in line with a linear ion trap mass spectrometer (LTQ; Thermo Finnigan) operating in data-dependent MS/MS mode. Five separate LC-MS/MS cycles were performed per sample, each varying by increased salt pulse concentration and followed by a linear organic gradient to resolve peptides. Spectra obtained from the LTQ were analyzed by DBDigger (32) using the human IPI (International Protein Index) data base, version 3.05. DTASelect was used to filter and organize the search results, whereas Contrast (33) was used to differentiate proteins that appeared in either asynchronized or nocodazole-arrested cells.

Identification of GCN5 as a B23-interacting Protein in
Nocodazole-arrested Cells by Mass Spectrometry-In order to explore the potential role of B23 in mitosis, we set out to identify B23 interacting proteins through a mass spectrometric approach. U2OS cells were first arrested by nocodazole. A majority of the U2OS cells were synchronized at G 2 /M phase as shown by fluorescence-activated cell sorting analysis (Fig. S1). Protein complexes containing B23 were immunoprecipitated by anti-B23 antibody from both nocodazole-treated and -un-treated U2OS cells, digested, and analyzed by nano-LC-MS/MS (see "Experimental Procedures"). Among the known B23-interacting proteins, we identified Nop132 (34), NFB (12), ATR (24), cyclin B (35), and DNA-dependent protein kinase catalytic subunit (available on Human Protein Reference Database website), verifying the efficacy and specificity of our mass spectrometric approach (raw data provided upon request). Two peptides (K.AQVR.G and R.RQLLEKFRVEK.D; periods in sequences represent tryptic cleavage sites) that matched specifically to SwissPro ID number Q92830-1 (human GCN5 acetyltransferase) met our criteria of two peptides minimum per locus (30). Interestingly, peptides specific to GCN5 appeared only in the B23 immunoprecipitate of nocodazole-arrested but not asynchronized cells (Fig.  1A), suggesting that B23 may potentially interact with GCN5 in G 2 /M phase cells. However, it is quite possible that B23 may interact with GCN5 in other phases of the cell cycle, albeit below the detection limit of our mass spectrometry (see below).
B23 Interacts with GCN5 in Vitro and in Vivo-To confirm the mass spectrometric finding, we performed reciprocal immunoprecipitation experiments using antibodies against endogenous B23 and GCN5 in 293T cells, respectively. B23 and GCN5 were co-immunoprecipitated in either experiment (Fig.  1B, left), and the same results were obtained with their exogenous counterparts in U2OS-Tet-On cells stably expressing doxycycline-inducible HA-tagged GCN5 and FLAG-tagged B23 (Fig. 1B, right). Moreover, we showed that endogenous GCN5 associated with B23 in both nocodazole-treated and untreated 293T cells (Fig. 1C), indicating that the interaction may occur in both interphase and mitotic cells.
To verify the finding further, we performed a GST pull-down experiment using bacterially produced B23-His and GST-GCN5 fusion proteins. As shown in Fig. 2A, B23-His only presented in GST-GCN5 pull-down but not in GST control. This result indicates that B23 directly interacts with GCN5. Human GCN5 (the long form GCN5L2 used in this study) contains three major functional domains, the N-terminal PCAF-N, the HAT, and the C-terminal BROMO domains (Fig. 2B). In order to identify the functional domains of GCN5 that may interact with B23, we carried out a deletion mapping experiment. For this purpose, GST-GCN5 deletion mutants were generated as indicated in Fig. 2B. The purified wild-type or mutant GST-GCN5 proteins were incubated with purified B23-His. Our results showed that the N-terminal 93 amino acids of GCN5 were not required for B23 binding. However, amino acids 94 -337 encoding the PCAF-N domain were required for B23 binding (Fig. 2C). These data suggest that B23 may directly interact with GCN5 through the PCAF-N domain of GCN5.
B23 Regulates GCN5-mediated Histone Acetylation-Both B23 and GCN5 are involved in histone post-translational regulation (1,17). B23 can bind histones functioning as a histone chaperone and promote nucleosomal assembly in vitro (17), whereas recombinant full-length mammalian GCN5 is competent for the acetylation of nucleosomal histones (27). In order to scrutinize the biological and biochemical significance of the  (52). Manual analysis of this spectrum corroborates DBDigger's identification of putative B23-interacting partner GCN5. B, physical interaction between B23 and GCN5 in vivo. Left, co-immunoprecipitation of endogenous B23 and GCN5 from 293T cells. Cellular extract was immunoprecipitated (IP) with protein A/G beads conjugated with anti-B23, GCN5 antibodies, or IgG as a control. The immunoprecipitates were then separated by SDS-PAGE and probed with anti-B23 or GCN5 antibodies. Right, co-immunoprecipitation of exogenous B23 and GCN5 in pBI-HA-Gcn5/FLAG-B23 U2OS-tet-On stable clone. U2OS-tet-On cells carrying pBI-HA-Gcn5/FLAG-B23 were induced with 2 g/ml doxycycline for 24 h, and immunoprecipitation was performed using anti-FLAG antibody or IgG as a control. The immunoprecipitates were separated by SDS-PAGE and probed with either anti-FLAG (for B23) or anti-HA (for GCN5) antibodies. C, GCN5 interacts with B23 in both nocodazole-treated (ϩNZ) and -untreated (ϪNZ) cells. Cells (293T) were arrested with 0.3 g/ml nocodazole for 18 h, and an equal amount of protein lysates (200 g) from nocodazole-treated and -untreated cells were used for immunoprecipitation as described in B. FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9

B23 in GCN5-dependent Histone Acetylation and Transactivation
B23/GCN5 interaction, we performed an in vitro HAT assay using purified recombinant human GCN5. We first measured GCN5 HAT activity on lysine 14 of histone H3 (H3 K14 ), since previous studies have demonstrated that H3 K14 is the major histone acetylation site by GCN5 (36). As shown in Fig. 3A, in the presence of purified GST-GCN5, H3 K14 was efficiently acetylated (lane 2), whereas the addition of B23 (B23-His) sig-nificantly reduced H3 K14 acetylation (lane 4). Moreover, incubation with B23-His at increasing concentrations resulted in a progressive inhibition of both H3 K14 and H4 K8 acetylation (Fig. 3B), indicating a B23 dosage-dependent inhibition of GCN5 HAT activity. The PCAF-N domain has been shown to be required for PCAF autoacetylation in trans, which may in turn lead to enhancement of PCAF HAT activity (37). Whether GCN5 may use a similar mechanism to regulate its HAT activity or whether binding of B23 to the PCAF-N domain of GCN5 may impair this regulatory function remain to be determined.
All HAT assays throughout our experiments were performed in the presence of an HDAC inhibitor (APHA compound 8), which apparently did not perturb the B23-inhibitory function on GCN5-dependent HAT activity, excluding the possibility that B23 may serve as an HDAC. Since B23 is a core histonebinding protein (38), we wonder whether B23 may inhibit GCN5 activity through binding competition with GCN5 to histone substrates. As shown by the "order of addition experiment," the addition of B23 subsequent to the GCN5-dependent HAT reaction failed to inhibit GCN5-mediated H3 K14 acetylation (Fig. 3C, compare lanes 2 and 4, scheme A), which was in sharp contrast to the inhibitory effect when B23 was supplemented to the HAT reaction before the addition of GCN5 (Fig. 3C, compare  lanes 6 and 8, scheme B). This result indicates that B23 1) may not act at a postacetylation stage and 2) may as well block histones from being acetylated by GCN5 through direct binding competition with GCN5 to histone substrates, as shown previously for INHAT, a histone chaperone complex (39). The latter was supported by and in line with the fact that B23 is a histone chaperone (38) and can bind histone H3 independently of H3 acetylation (Fig. S2). This study, together with the interaction and deletion mapping data ( Figs. 1 and 2), suggests that B23 may negatively regulate GCN5-dependent acetylation at both enzyme (GCN5) and substrate (histone) levels. Purified B23-His (lane 3) was incubated with GST-GCN5, and B23-GCN5 complexes were affinity-purified with glutathione beads. GCN5 and B23 were detected by Western blot using anti-GCN5 and anti-B23 antibodies, respectively. B, schematics of human GCN5 and its respective deletion derivatives. C, PCAF-N domain of GCN5 is involved in GCN5-B23 interaction. Purified B23-His was incubated with the indicated GST-GCN5 fusion proteins, and GCN5-B23 complexes were affinity-purified by glutathione-agarose beads (GST pull-down). B23-His and GST-GCN5 fusion proteins were detected by immunoblot using anti-B23 and GST antibodies, respectively. To evaluate the effect of B23 on GCN5 activity to acetylate chromatin histones, we tested whether the inhibitory activity of B23 may also occur with nucleosomes. According to a previous study, full-length mammalian GCN5, in the absence of any other SAGA component, can efficiently acetylate nucleosomal histone H3 (27). In line with this study, we showed that the recombinant full-length GCN5 (the long form GCN5L2) acetylated HeLa cell mononucleosomal H3 K14 (Fig. 4, lane 2). In the presence of B23, the level of mononucleosomal H3 K14 acetylation decreased (Fig. 4, lane 4). This inhibition may have significant impact on GCN5-dependent chromatin structure remodeling, as shown by an S7 nuclease sensitivity assay in GCN5-and B23transfected cells (Fig. S3).
Phosphorylation of B23 at Threonine 199 Is Involved in B23mediated Mitotic Inhibition of GCN5-dependent Histone Acetylation-To further explore the role of B23 in the regulation of GCN5-mediated histone acetylation in cells, we used anti-B23 monoclonal antibody to immunoprecipitate B23 from U2OS interphase and mitotic extracts. The immunopurified antibody-B23 complex was then applied to the GCN5-mediated histone H3 acetylation reactions. As demonstrated in Fig.  5A (top), although both interphase and mitotic B23 inhibited GCN5-mediated acetylation, mitotic B23 was more effective to inhibit GCN5-mediated histone acetylation activity. Work from others shows that B23 is phosphorylated by CDC2 kinase during mitosis, and accumulation of the highly phosphorylated mitosisspecific B23 correlates with mitotic chromosome condensation (15). Indirect evidence suggests a correlation between B23 dephosphorylation and mitotic chromosome decondensation (16). To evaluate potential contributions of mitotic B23 phosphorylation on the regulation of GCN5-mediated histone acetylation, we tested the effect of in vitro phosphorylated B23-His protein in the GCN5-mediated HAT assay. After incubation with mitotic or interphase extracts in a kinase reaction buffer containing ATP, mitotic phosphorylated B23 again showed higher inhibitory activity than interphase phosphorylated B23 (Fig. 5A, bottom, lanes 3 and 4). Together, both in vitro and in vivo purified B23 exhibited strong mitotic inhibitory activity on GCN5 (Fig. 5A), implying that mitotic phosphorylation of B23 may enhance its inhibitory activity toward GCN5-mediated histone acetylation.
A recent study suggests that threonine 199 of B23 (B23 T199 ) is phosphorylated by CDK1 (CDC2) at the onset of mitosis, coinciding with chromosome condensation and nucleolar and nuclear envelope disassembly (40). The same phospho- FIGURE 5. Phosphorylation of B23 threonine 199 in mitosis enhances B23-inhibitory activity on GCN5-mediated histone acetylation. A, top, interphase-and mitosis-derived B23 inhibit GCN5 HAT activity. B23 was immunoprecipitated by anti-B23 antibody from interphase (I) and mitotic (M) extracts. The HAT reaction was then carried out in the presence or absence of the immunopurified B23 (IPed-B23). Bottom, purified B23-His was incubated with the interphase or mitotic extracts in a kinase reaction buffer containing ATP and then added to the HAT reaction. Acetylated histone H3 K14 (H3-AcK14), phosphorylated B23-Thr 199 (B23-pT199), B23-His, and immunopurified B23 were revealed by Western blot using acetyl-H3 K14 , phospho-B23-Thr 199 , and B23 antibodies, respectively. Coomassie stain shows equal loading of GST-H3 in each HAT reaction. The indicated -fold change of H3-AcK14 was calculated based on the densitometric values of each lane normalized against the GST-H3 loading control. B, top, threonine 199 of B23 was preferentially phosphorylated in mitosis. Endogenous B23 or B23-pT199 was detected in interphase (I) and mitotic (M) extracts as above. Bottom, specificity of phospho-B23-Thr 199 antibody. In vitro purified B23-His (WT or T199A) proteins were phosphorylated by baculovirus-produced cyclin E-CDK2 in the presence of ATP and visualized by Western blot using anti-phospho-B23-Thr 199 antibody. C, B23-Thr 199 phosphorylation is involved in mitotic inhibition of GCN5 HAT activity. Top, mitotic extracts were immunodepleted with anti-phospho-B23-Thr 199 antibody (depl) or control rabbit IgG (no depl) and then evaluated by an immunoblot using phospho-B23-Thr 199 , B23, and ␤-actin antibodies. Bottom, the resulting extracts were then added to the GCN5-mediated HAT reaction in the presence (lanes 3 and 4) or absence (lanes 1 and 2) of B23-His (WT) or B23-T199A-His (T199A) that had been prephosphorylated by cyclin E-CDK2 as in B. GCN5 activity was evaluated in an immunoblot using anti-acetyl-H3 K14 antibody. Coomassie stain shows equal loading of GST-H3 in each reaction. FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9 rylation site has also been shown to be a CDK2-cyclin E and CDK2-cyclin A substrate (41), playing an important role in the regulation of pre-mRNA processing and centrosome duplication (42,43). Since the anti-B23 antibody used for the immunopurification experiment (Fig. 5A, top) can immunoprecipitate both phosphorylated and unphosphorylated B23 T199 (42) and B23 phosphorylated by mitotic extract exhibited stronger inhibitory activity on GCN5-mediated histone acetylation than its interphase counterpart (Fig. 5A), we investigated whether the phosphorylation status of B23 T199 between interphase and mitosis could contribute to the differential effect of B23 on GCN5 HAT activity. Western blot analysis using anti-phospho-B23-Thr 199 -specific antibody revealed that endogenous B23 T199 in U2OS cells was more predominantly phosphorylated in mitotic than interphase cells (Fig. 5B, top). Similar results were obtained when in vitro purified B23-His was used as a substrate of interphase or mitotic extracts (Fig. 5A, bottom). To test the specificity of the phospho-B23-Thr 199 antibody, we created a B23 Thr 199 3 Ala (B23T199A-His) mutant. We showed that the antibody recognized wild-type B23-His but not B23T199A-His upon phosphorylation by cyclin E-CDK2 (Fig.  5B, bottom), verifying the specificity of the phospho-B23-Thr 199 antibody. To further investigate the effect of B23 Thr 199 phosphorylation on GCN5-mediated HAT activity, we performed a phospho-B23-Thr 199 depletion experiment in which mitotic extract was incubated with anti-phospho-B23-Thr 199 antibody, followed by depletion with protein-A beads (Fig. 5C,  top). The depleted mitotic extract was then supplemented to the GCN5-mediated HAT assay. As shown in Fig. 5C (bottom), mitotic extract depleted of phospho-B23-Thr 199 (depl) was much less effective in inhibition of GCN5 HAT activity than the extract depleted by IgG control (no depl). Since B23-Thr 199 is the major phosphorylation site of cyclin E-CDK2 at the onset of centrosome duplication (41), we tested if "add-back" of cyclin E-CDK2phosphorylated B23 may restore the inhibitory activity of phospho-B23-Thr 199 -depleted mitotic extract. Indeed, "add-back" of cyclin E-CDK2-phosphorylated wild-type B23-His but not B23T199A-His mutant significantly restored the inhibitory activity of phospho-B23-Thr 199 -depleted mitotic extract (Fig. 5C, bottom, lanes 3 and 4). Together, these studies suggest that phosphorylated B23-Thr 199 contributes to the mitotic inhibitory activity of B23. However, our study does not exclude the possibility that mitotic phosphorylation of other B23 phosphorylation sites may also play an important role in the regulation of GCN5-mediated histone acetylation.

B23 in GCN5-dependent Histone Acetylation and Transactivation
B23 Modulates GCN5-dependent Transactivation-It is generally believed that GCN5-dependent transcription relies on its HAT activity (2,3). The inhibitory effect of B23 on GCN5-dependent histone acetylation raised a question about whether B23 has any impact on GCN5-dependent transcriptional regulation. We thus evaluated the effect of B23 on GCN5-dependent transcription in the context of GCN5-regulated promoters. Our first experiment was designed to bring GCN5 to a reporter promoter followed by measuring the transactivation activity of GCN5 in the presence of B23. For this purpose, we fused GCN5 to GAL4 DNA-binding domain (pBIND-GCN5) and determined its activity on a pG5luc luciferase reporter that was under the control of five copies of GAL4 consensus binding sites. Previous studies demonstrated that fusion of GCN5 to the bacterial LexA DNA-binding domain activated transcription in yeast. It was suggested that this activity was attributable to the HAT and ADA2 binding domains of GCN5 (44). In line with the previous studies, transfection of pBIND-GCN5 into 293T or U2OS cells activated the pG5luc luciferase reporter in a GCN5-dependent manner (Fig. 6, pBIND-Gcn5). Overexpression of wild-type B23 significantly reduced the GCN5dependent GAL4 promoter activity in asynchronized 293T (t test, p Ͻ 0.005), U2OS interphase (p Ͻ 0.005), and mitotic cells (p Ͻ 0.02) (Fig. 6, pBIND-Gcn5 ϩ pCMVtag2C-B23). Moreover, wild-type B23 inhibits GCN5-mediated transactivation in a dosage-dependent manner in NIH3T3 cells (Fig.  S4). Interestingly, mutation of B23 Thr 199 to Ala partially compromised B23 inhibitory function in mitotic cells (Figs. 6 and S4, pBIND-Gcn5 ϩ pCMVtag2C-B23 T199A ; p Ͼ 0.05). This result is in agreement with our finding that phosphorylation of B23-Thr 199 is involved in mitotic inhibition of GCN5-mediated HAT activity (Fig. 5). In contrast, B23 T199A mutant was almost as competent as B23 WT in blocking GCN5-dependent transac- tivation in U2OS interphase cells (Fig. 6, p Ͻ 0.005). This finding is in line with our observation that B23 inhibits GCN5-dependent HAT activity in interphase (Fig. 5A), although B23-Thr 199 in interphase was not as heavily phosphorylated as in mitosis (Fig. 5B). Collectively, these results suggest that B23 may inhibit GCN5-dependent transactivation and are consistent with a potential role of mitotic phosphorylation of B23-Thr 199 in enhancing its inhibitory activity on GCN5-mediated histone acetylation.
Recent studies suggest that conversion to active chromatin by histone acetylation could be one of the best candidate mechanisms for p53-independent activation of p21/Waf1 (45,46). Using the pBI-tet-B23-and pBI-tet-Gcn5-inducible stable clones, we analyzed the effect of B23 on GCN5-dependent p21 activation. Quantitative RT-PCR using p21-specific primers demonstrated that p21/Waf1 mRNA level elevated 2-fold upon induction of GCN5 in pBI-tet-Gcn5-inducible cells (Fig. 7A, lane 4) but substantially decreased when GCN5 and B23 were simultaneously induced in pBI-tet-B23-Gcn5 double clones (Fig. 7A, lane 8). These results are consistent with the current model that p53-independent p21/Waf1 activation is regulated at the level of nucleosomal histone modification (47) and reinforce our finding that B23 may regulate GCN5dependent transcription on certain promoters.
The p21/Waf1 promoter contains 6ϫ Sp1-binding sites. Treatment of cells with sodium butyrate, an HDAC inhibitor, increases binding of GCN5 to the proximal region of p21/Waf1 promoter containing the Sp1 repeats concomitant with H3 hyperacetylation at the Sp1 sites (45). This suggests that GCN5 plays an important role in the regulation of Sp1 promoter activity. Histone H4 hyperacetylation has also been implicated in the regulation of TATA-less Sp1 promoter activity (46). We therefore utilized the Sp1 promoter activation as a model system to assess the roles of B23 and GCN5 in Sp1 promoter regulation. U2OS cells were transiently transfected with pSp1-luc reporter together with GCN5 and B23 alone or together (Fig. 7B). GCN5 activated the pSp1-luc reporter as exemplified by the 2-fold increase of luciferase activity (Fig. 7B, pCMVtag2C-Gcn5), whereas in the pCMVtag2C-B23 and pCMVtag2C-Gcn5 double-transfection experiment, the GCN5-dependent Sp1luc activation was significantly inhibited (t test, p Ͻ 0.01) with the luciferase activity reducing to a background level (Fig. 7B, pCMVtag2C-Gcn5 ϩ pCMVtag2C-B23). These results, in line with the inhibitory effect of B23 on GCN5-dependent histone acetylation, suggest that B23 may oppose GCN5-dependent activation of TATA-less Sp1 promoters.
To further evaluate the role of B23 in regulating GCN5-dependent transcriptional activation at a genome-wide scale, we carried out a microarray expression analysis in U2OS stable inducible clones carrying pBI-tet, pBI-tet-HA-GCN5, pBI-tet-B23-FLAG, or pBI-tet-HA-GCN5/B23-FLAG using Vanderbilt Microarray Shared Resources 30k human oligonucleotide chips. Gene expression profiles of GCN5-and B23-inducible clones were organized according to the dissimilarity of their effects on gene expression, and Pearson correlation coefficients were calculated for each pair of genes (Fig. 7C). Genes are considered to be significantly counterregulated only if differences in hybridization signals meet all the following criteria: gene expression is 1) over 2.4-fold induced in GCN5-inducible cells, 2) over 2.4-fold repressed in B23-inducible cells, and 3) changed less than 1.1-fold (1.0 equal to no change) in GCN5/B23 double-inducible cells after normalization against the vector con-   4). The gene expression cluster map represents 147 genes whose expression showed significant counterregulation between GCN5 and B23. The scale value did not include the -fold change of genes (see genes marked by an asterisk in Table S1) whose expression were induced or repressed over 9.9-fold. Values of normalized -fold changes are shown in Table S1. trols in at least two replicate experiments. Based on these criteria, we found about 147 genes that were counterregulated by GCN5 and B23 ( Fig. 7C and Table S1). These results are consistent with a role of B23 in negatively modulating at least some of the GCN5-dependent transcription.

DISCUSSION
In this study, we demonstrate that B23, a multifunctional phosphoprotein, can inhibit GCN5-mediated free histone and nucleosomal acetylation as well as transactivation in vitro and in vivo. Our study supports the possibility that B23 may inhibit GCN5 catalytic activity via a dual mechanism involving both a direct interaction with GCN5 and a blockage of histone substrates from being acetylated by GCN5 (Figs. 1, 2, and 3C).
Our finding that B23 may negatively regulate GCN5-dependent transcription corroborates its inhibitory role on GCN5-mediated histone acetylation. Although B23 inhibitory activity is much less dependent on Thr 199 phosphorylation in interphase (Figs. 5 and 6), phosphorylation of B23 Thr 199 may play a role in regulating its inhibitory activity toward GCN5-dependent histone acetylation and transcription in mitosis (Figs. 5 and 6). According to the current hypothesis, the presence of active HATs and HDACs in mitotic cells may help to prime the cells for transcriptional reactivation immediately after chromosomes become decondensed in late mitosis (48), and Gcn5p is particularly critical for reactivation of transcription after mitotic silencing in yeast cells (49). Since B23 Thr 199 becomes heavily phosphorylated at the onset of mitosis and dephosphorylated during anaphase (40), it is possible that this temporally regulated phosphorylation increases B23 inhibitory activity on GCN5 (Figs. 5 and 6), which may be a necessary step to prevent premature histone acetylation before the onset of mitotic transcriptional reactivation.
It has been demonstrated that B23 can either positively or negatively regulate transcription depending on the types of promoters and proteins it interacts with. For example, B23-RAR fusion proteins originated from acute promyelocytic leukemia cells can interact with co-repressors and co-activators, resulting in both transcriptional repression and activation (50). Similarly, B23 can both repress IRF1 (interferon regulatory factor 1)-mediated transcriptional activation (18,19) and relieve YY1induced transcriptional repression by direct interaction with YY1 transcription factor (20). Interestingly, a recent study shows that B23 works as an AP2␣-binding transcriptional corepressor by remodeling local chromatin structure (51). We documented that B23 attenuated GCN5-dependent transcription at the Sp1 promoter in this study. Possible functional relevance to this stems from our observation that B23 hampers GCN5dependent p21/Waf1 transcription, which is in agreement with the current model that nucleosomal histone deacetylation at p21/Waf1 promoter attributes to p53-independent inactivation of p21/Waf1 (47). A conceivable scenario is that inhibition of GCN5-mediated histone acetylation by B23 may prevail in a subset of genes depending on the contexts of their transcription activators and promoters. This idea is supported by our microarray data in which only a fraction of Gcn5-induced genes are suppressed by B23 (Fig. 7 and Table S1). Taken together, our observations demonstrate a novel function of B23 in the control of GCN5-dependent histone acetylation and transcription and may have significant impact on deciphering the mechanism(s) of GCN5-and B23-dependent transcriptional regulation during cell division.