Epigenetic Regulation of the Blimp-1 Gene (Prdm1) in B Cells Involves Bach2 and Histone Deacetylase 3*

B lymphocyte-induced maturation protein 1 (Blimp-1) encoded by Prdm1 is a master regulator of plasma cell differentiation. The transcription factor Bach2 represses Blimp-1 expression in B cells to stall terminal differentiation, by which it supports reactions such as class switch recombination of the antibody genes. We found that histones H3 and H4 around the Prdm1 intron 5 Maf recognition element were acetylated at higher levels in X63/0 plasma cells expressing Blimp-1 than in BAL17 mature B cells lacking its expression. Conversely, methylation of H3-K9 was lower in X63/0 cells than BAL17 cells. Purification of the Bach2 complex in BAL17 cells revealed its interaction with histone deacetylase 3 (HDAC3), nuclear co-repressors NCoR1 and NCoR2, transducin β-like 1X-linked (Tbl1x), and RAP1-interacting factor homolog (Rif1). Chromatin immunoprecipitation confirmed the binding of HDAC3 and Rif1 to the Prdm1 locus. Reduction of HDAC3 or NCoR1 expression by RNA interference in B cells resulted in an increased Prdm1 mRNA expression. Bach2 is suggested to cooperate with HDAC3-containing co-repressor complexes in B cells to regulate the stage-specific expression of Prdm1 by writing epigenetic modifications at the Prdm1 locus.

B lymphocyte-induced maturation protein 1 (Blimp-1) encoded by Prdm1 is a master regulator of plasma cell differentiation. The transcription factor Bach2 represses Blimp-1 expression in B cells to stall terminal differentiation, by which it supports reactions such as class switch recombination of the antibody genes. We found that histones H3 and H4 around the Prdm1 intron 5 Maf recognition element were acetylated at higher levels in X63/0 plasma cells expressing Blimp-1 than in BAL17 mature B cells lacking its expression. Conversely, methylation of H3-K9 was lower in X63/0 cells than BAL17 cells. Purification of the Bach2 complex in BAL17 cells revealed its interaction with histone deacetylase 3 (HDAC3), nuclear co-repressors NCoR1 and NCoR2, transducin ␤-like 1X-linked (Tbl1x), and RAP1-interacting factor homolog (Rif1). Chromatin immunoprecipitation confirmed the binding of HDAC3 and Rif1 to the Prdm1 locus. Reduction of HDAC3 or NCoR1 expression by RNA interference in B cells resulted in an increased Prdm1 mRNA expression. Bach2 is suggested to cooperate with HDAC3-containing co-repressor complexes in B cells to regulate the stage-specific expression of Prdm1 by writing epigenetic modifications at the Prdm1 locus.
B cells play important roles in humoral immunity. Upon stimulation by an antigen and additional stimulators, mature cells undergo clonal proliferation with concomitant somatic hypermutation and class switch recombination (CSR) 2 of immunoglobulin genes, followed by terminal differentiation into antibody-secreting plasma cells. These responses are coordinated by a network of transcription factor genes (1). Blimp-1 (B lymphocyte-induced maturation protein 1) is a master regulator driving the terminal differentiation of B cells into plasma cells (2)(3)(4). The expression level of Blimp-1 is low or absent in B cells but becomes high in plasma cells (5). The overexpression of Blimp-1 in B cells induces their differentiation into plasma cells (6,7). Conversely, the genetic ablation of Blimp-1 in B cells abolishes their terminal differentiation into plasma cells (3). Blimp-1 has been proposed to induce plasma cell differentiation by erasing the B cell phenotype (2,4,8). Blimp-1 also inhibits the events associated with activated B cells, including CSR (2). Thus, the timing and magnitude of Blimp-1 expression have been regarded to dictate the differentiation of B cells to plasma cells.
Transcription of the Blimp-1 gene (Prdm1) is regulated by both transcription activators and repressors (9). Two DNA binding factors are known to repress transcription of Prdm1. One is Bcl6 (B cell lymphoma 6) that is essential for the formation of a germinal center and is required for somatic hypermutation (10,11). Bcl6 represses Prdm1 by binding to its intron 5 (12). Another repressor is Bach2 (BTB and CNC homologue 2), which is a basic region-leucine zipper factor and forms heterodimers with small Maf proteins through their leucine zipper domain (1,13). The Bach2-Maf heterodimers then bind to a specific DNA element termed Maf recognition element (MARE) (14). Prdm1 carries two MAREs at its promoter upstream region and intron 5 region to which Bach2 binds with the small Maf proteins in vitro (15,16). Bach2 is expressed in B cells from pro-B to mature-B cell stages but not in plasma cells, showing a pattern totally opposite that of Blimp-1 (15,17). Genetic loss of Bach2 results in overexpression of Blimp-1 in activated B cells (18), suggesting that Bach2 is a bona fide repressor of Prdm1 in B cells. Bach2-mediated repression of Prdm1 is required for CSR (19). Although Bach2 is central to the regulation of plasma cell differentiation, the mechanism for the regulation of Prdm1 by Bach2 remains to be elucidated. More specifically, little is known about the co-regulators of Bach2 or the epigenetic regulation of Prdm1 in B and plasma cells. Here, we compare changes in acetylation and methylation of histones at the Prdm1 locus before and after plasma cell differentiation, and we purified the Bach2 protein complex to identify proteins involved in this epigenetic regulation.

Experimental Procedures
Bach2 Complex Purification and Mass Spectrometry Analysis-The Bach2 complex was purified from nuclear extracts prepared from BAL17 cells stably expressing FLAG-hemagglutinin (HA) epitope-tagged Bach2 (eBach2) as described previously (15,20). The eBach2-expressing cells were collected by centrifugation for 8 min at 1,865 ϫ g and were washed with phosphate-buffered saline. After centrifugation for 10 min at 1,190 ϫ g, the pellets were suspended in 6 volumes of hypotonic buffer (10 mM Tris-HCl (pH 7.3), 10 mM KCl, 1.6 mM MgCl 2 , 2 mM 2-mercaptoethanol (2-ME), 40 M phenylmethylsulfonyl fluoride (PMSF)) and were then collected by centrifugation for 5 min at 1,190 ϫ g. Next, the cells were resuspended in the same volume of hypotonic buffer, and the suspensions were homogenized and centrifuged for 15 min at 2,330 ϫ g to collect the nuclei. The obtained crude nuclei were then suspended in a half-volume of low salt buffer (0.02 M KCl, 20 mM Tris-HCl (pH 7.3), 25% glycerol, 1.5 mM MgCl 2 , 0.2 mM EDTA (pH 8.0)) for homogenization. The resulting suspension was dropped with a half-volume of high salt buffer (1.2 M KCl, 20 mM Tris-HCl (pH 7.3), 25% glycerol, 1.5 mM MgCl 2 , 0.2 mM EDTA) and then stirred gently for 60 min and centrifuged for 60 min at 48,384 ϫ g. The supernatants were dialyzed against 50 volumes of BC-0 buffer (20 mM Tris-HCl (pH 7.3), 20% glycerol, 0.2 mM EDTA (pH 8.0), 2 mM 2-ME, 40 M PMSF) until the conductivity reached 65-70 microsiemens/cm. The dialysate was centrifuged for 20 min at 23,700 ϫ g, and the supernatant was used as a nuclear extract. eBach2 was immunoprecipitated from nuclear extracts by incubating with M2 anti-FLAG-agarose (Sigma) for 4 h with rotation. After an extensive wash with 0.1 M KCl-containing buffer B, the bound proteins were eluted from FLAG M2-agarose by incubating for 60 min with the FLAG peptide in the same buffer (0.5 mg/ml). The eluates were then incubated with anti-HA antibody-conjugated beads for 4 h with rotation. After extensive washing with 0.1 M KCl-containing buffer B, the bound proteins were eluted from anti-HA antibody-conjugated beads by incubating with 100 mM glycine-HCl (pH 2.5) and neutralized with 1 M Tris-HCl (pH 8.0). Purified proteins were separated by SDS-PAGE on a 4 -20% gel, and the gel was subsequently stained with Coomassie Brilliant Blue. The stained bands were excised from the gel, and the proteins therein were subjected to in-gel reduction, S-carboxyamidomethylation, and digestion with trypsin (Promega). The molecular masses of the tryptic peptides were determined using LC-HCT plus (Bruker Daltonics), and protein identification was performed using the MASCOT search engine (Matrix Science).
Bach2 Complex Analysis by Reversible Cross-link Immunoprecipitation (ReCLIP)-The proteins interacting with Bach2 were isolated also by another method based on the one called "ReCLIP" (21). Approximately 1 ϫ 10 8 BAL17 cells were fixed in PBS containing 0.5 mM dithiobis(succinimidyl propionate) and 0.5 mM dithiobismaleimidoethane, and then they were treated with Tris and L-cysteine in PBS for quenching. The fixed cells were suspended in RIPA buffer, and after supersonic treat-ment and centrifugation, the cleared lysate was incubated with anti-FLAG magnetic beads (Sigma M8823) for 1.5 h at 4°C. The beads were then washed five times with RIPA buffer, and the proteins bound to the beads were eluted using 0.1 g/l FLAG peptide. The eluted proteins were treated with 50 mM DTT to dissolve the cross-link and were separated by SDS-PAGE using a 5-20% polyacrylamide gel (Oriental Instruments). After staining with 60 mg/liter Coomassie Brilliant Blue-G-250 and 35 mM HCl (22), each lane in the gel was divided into 12 sections, and the resulting gel blocks were treated with DTT for reduction and then with acrylamide for alkylation of the sulfhydryl groups. The proteins in every gel block were digested overnight by 10 -20 ng of trypsin (Promega). The resulting peptides were extracted from the gel with 75% acetonitrile and 0.5% formic acid and were concentrated in a SpeedVac. One-half of each sample was analyzed using an LTQ Orbitrap Velos with ETD mass spectrometer (Thermo Fisher Scientific) equipped with a PAL HTC-xt autosampler (AMR) and an ADVANCE HPLC system (AMR, Tokyo, Japan). The peptides were separated on a PepSwift Monolithic Column (100 m inner diameter ϫ 25 cm, Thermo Fisher Scientific) at a flow rate of 300 nl/min with a 60-min linear gradient generated by aqueous solvent A (0.1% formic acid) and organic solvent B (100% acetonitrile) as follows: 2.5-22.5% B in 52 min, to 35% B in 56 min, to 95% B in 58 min, and they were directly electrosprayed into the mass spectrometer. The data acquisition of every sample was carried out for 70 min after the LC gradient was started, in which MS 1 scans from m/z ϭ 321 to 1,800 were carried out in the orbitrap with the resolution set at 100,000 with a lock mass at m/z ϭ 445.120025, followed by sequential isolation of the 20 most intense precursor ions and MS 2 acquisition by collisioninduced dissociation in the ion trap in the normal resolution mode. The settings for MS 2 scans were as follows: minimal signal intensity required ϭ 500, isolation width ϭ 2 m/z, AGC target ϭ 10,000, maximum ion injection time ϭ 200 ms, normal collision energy ϭ 35, activation time ϭ 10 ms, microscan ϭ 1, and dynamic exclusion was enabled with a 60-s exclusion duration. The raw data files derived from samples in the same SDS-PAGE lane were converted together into a single MASCOT generic format file and used for the database search by MAS-COT (version 2.5.1, Matrix Science) against the human proteins in SwissProt (Feb., 2015), and a custom database, including contaminant proteins and the FLAG-tagged mouse Bach2 protein. A maximum of three trypsin miscleavages was allowed. The peptide mass tolerance and MS/MS tolerance were set at 5 ppm and 0.5 Da, respectively. Protein N-terminal acetylation (ϩ42.0106), oxidation of methionine (ϩ15.9949), phosphorylation at serine/threonine (ϩ79.9663), propionamidation at cysteine (ϩ71.0371), propionamidated dithiobis(succinimidyl propionate) at lysine (ϩ159.0354), and propionamidated dithiobismaleimidoethane at cysteine (ϩ246.0674) were considered to be variable modifications. Peptides identified with MASCOT expectation values of Ͻ0.05 were selected as significant hits. The false discovery rates calculated by decoy database search were 2.05% in mock IP and 1.79% in FLAG-Bach2 IP.
Immunoprecipitation-His-tagged mouse Bach2 protein (amino acid residues 119 -318) was overexpressed in Escherichia coli and purified using nickel resin. The recombinant Bach2 protein was used to immunize a rabbit, resulting in anti-Bach2 antisera (Bach2N-1 and N-2). These antisera were found useful for the immunoblot analysis and immunoprecipitation assays. Whole cell extracts of BAL17 cells were pre-cleared with protein G-Sepharose beads at 4°C for 2 h and immunoprecipitated with the anti-Bach2 antiserum (Bach2N-2) or anti-HDAC3 antibody (NB500-126; Novus) for 2 h to analyze the interaction of endogenous proteins. Immunoprecipitates were recovered with protein G-Sepharose beads and were washed seven times. Samples were analyzed by immunoblot analysis as described above (25). The primary antibodies were anti-Bach2 antiserum (F69-1 (14)) and anti-HDAC3 antibody (06-890; Upstate Biotechnology). The secondary antibody was HRP-conjugated anti-rabbit IgG (GE Healthcare). In some experiments, rabbit IgG TrueBlot (eBioscience) was used to eliminate the interference of signal detection by the immunoglobulin heavy and light chains. For primary B cell immunoprecipitation, ReCLIP was carried by using control IgG or anti-Bach2 antibodies (Bach2N-1) that were conjugated with beads, followed by immunoblot analysis. Anti-NCoR (ab24552; Abcam) was used as the primary antibody in addition to the above antibodies. Intensities of data images on the films were measured by ImageJ software.
Plasmids-Human HDAC3 expression plasmid was kindly provided by Dr. Yoshida and described previously (26). Mouse Rif1 expression plasmid was generated by N. Y. and H. M. and will be described elsewhere.
Statistical Analysis-The statistical analysis was performed using Student's t test or Wilcoxon test, as appropriate. p value Ͻ 0.05 was considered to be significant.

JOURNAL OF BIOLOGICAL CHEMISTRY 6319
revealed that the levels of acetylation of histone H3 and H4 at the intron 5 MARE of the Prdm1 gene were higher in X63/0 cells than in BAL17 cells. There was no significant difference between BAL17 and X63/0 in the amount of acetylation of histone H3 on the promoter MARE region, although the amount of acetylated histone H4 on the promoter region was low in X63/0 cells. We also compared individual acetylation of Lys-9, Lys-18, or Lys-27 of histone H3 using their respective antibodies. The levels of acetylation of H3-K9, -K18, and -K27 at the Prdm1 promoter MARE was slightly decreased in X63/0 cells compared with BAL17 (Fig. 1, D and F) or did not differ in these cells (Fig. 1E)

Bach2 Complex in Plasma Cell Differentiation
acetylation at the Prdm1 intron 5 MARE were higher in X63/0 plasma cells than in BAL17 cells (Fig. 1D). Taken together, these results suggested that hyperacetylation of Lys-9, Lys-18, and Lys-27 of histone H3 at the Prdm1 intron 5 MARE may be involved in the repression of Prdm1 in mature B cells.
Dynamic Changes of Histone Methylation Pattern at the Prdm1 Locus between Mature B Cells and PCs-Histone H3-K4 methylation is associated with regulatory regions of actively transcribed genes, whereas histone H3-K9 methylation is associated with inactive genes. We therefore assessed histone methylation at the promoter and intron 5 MARE regions of the Prdm1 locus. The levels of H3-K4 mono-, di-, and tri-methylation (me1, me2, and me3) at the promoter MARE were similar between BAL17 and X63/0 cells (Fig. 2, A-C). In contrast, these modifications at the intron 5 MARE were higher in X63/0 than in BAL17 cells (Fig. 2, A-C). In particular, H3K4me2 showed a profound difference. Next, we investigated the levels of histone H3-K9 methylation. The levels of H3K9me2 and H3K9me3 at the intron 5 MARE were significantly reduced in X63/0 cells (Fig. 2, D and E). These modifications at the promoter MARE tended to be lower in X63/0 cells than in BAL17 cells (Fig. 2, D  and E). Collectively, these results suggest that histone modifications involving acetylation and methylation may act together to regulate the plasma cell-specific Prdm1 expression. A competition between acetylation and methylation of H3-K9 may operate at the intron MARE region.
Bach2 Promotes Histone Deacetylation of Prdm1 Chromatin in B Cells-To examine whether Bach2 is involved in the epigenetic modifications of the Prdm1 locus in B cells, we carried out ChIP analysis of the locus using splenic B cells isolated from wild-type and Bach2 Ϫ/Ϫ mice and stimulated with LPS in vitro. The levels of H3K9ac at the promoter and intron 5 MAREs were significantly higher in Bach2 Ϫ/Ϫ splenic B cells than in wildtype B cells (Fig. 3A). To ascertain whether the elevated levels of H3K9ac directly resulted from the absence of Bach2, we reconstituted Bach2 expression in Bach2 Ϫ/Ϫ splenic B cells by a retroviral transduction, and we determined the levels of acetylation by ChIP analysis (Fig. 3B). The levels of H3K9ac at both the promoter and intron 5 MAREs were lower in the B cells reconstituted with Bach2 than those infected with the control virus. Taken together, these results suggested that Bach2 reduced H3-K9 acetylation at Prdm1 in B cells.
We examined whether Bach2 bound to the two regulatory regions of Prdm1 in B cells. We generated a new anti-Bach2

FIGURE 4. Purification of Bach2 complex in BAL17 mature B cells.
A, expression of FLAG-HA epitope-tagged Bach2 (eBach2). Immunoblotting analysis was carried out with anti-Bach2 antiserum using nuclear extracts from non-transduced (Mock) and eBach2-expressing BAL17 cells. IB, immunoblot. B, Bach2 complex was purified from nuclear extracts prepared from eBach2-expressing BAL17 cells. Mock purification from nuclear extracts prepared from nontransduced BAL17 cells was performed as a control. IP, immunoprecipitation. C-G, identification of Rif1, Tbl1x, HDAC3, MafG, and MafK by mass spectrometry. Amino acid sequences of peptides were determined by a mass spectrometry/mass spectrometry analysis.
antiserum (Bach2N-2), which recognized endogenous Bach2 in B cells in immunoblot analysis (see under "Materials and Methods"). Using this antiserum, we carried out ChIP assays of primary B cells stimulated with LPS. As shown in Fig. 3C, both Bach2 and MafK were found to bind to the promoter and intron 5 MAREs. Bach2 did not bind to the promoter region of the Mcm5 gene, confirming specific binding of Bach2 of the Prdm1 MAREs. These results together indicate that Bach2 promotes histone deacetylation, including H3-K9 of the Prdm1 locus by directly binding to the regulatory regions in B cells.
Characterization of Bach2 Complex in B Cells-To elucidate the factors involved in Prdm1 repression by Bach2, Bach2 was purified from BAL17 cells. FLAG-HA epitope-tagged Bach2 (eBach2) was stably expressed in BAL17 cells (Fig. 4A). The expression level of epitope-tagged Bach2 was seven times higher than that of endogenous Bach2 according to a densitometry analysis. The purification of eBach2 was performed by sequential steps of affinity chromatography using anti-FLAG antibody-conjugated agarose followed by anti-HA antibodyconjugated beads. The associated proteins were analyzed by SDS-PAGE along with samples that were purified with the same procedures from control cells (mock; Fig. 4B). Several proteins were found to associate with eBach2. These proteins were specific components of the Bach2 complex inasmuch as they were not enriched in the mock purification. Protein bands were excised from the gels, and their identities were determined by mass spectrometry analysis. Rif1, Tbl1x, HDAC3, MafG, and MafK were identified among these proteins (Fig. 4, C-G). In

TABLE 1 The list of proteins identified as Bach2-interacting proteins using ReCLIP
The proteins that were identified in the FLAG-Bach2 ReCLIP experiment with a protein score of Ͼ130 but were not detected in the mock experiment are listed in order of protein hit number, irrespective of their protein scores. Protein hit number indicates the rank of the MASCOT protein family (a group of proteins sharing at least one peptide) to which each protein belonged. Protein family member indicates the protein score rank of each protein in the protein family to which it belonged. Protein score indicates the reliability of the protein identification derived from sum of the high set ions score for each sequence match that exceeds the threshold where ions score is Ϫ10 logP, and p is the probability that the MS/MS spectral match is a random event. another set of purification, using the ReCLIP method (see "Experimental Procedures"), the lanes of the gels were excised as slices, and proteins in the slices were determined by mass spectrometry analysis. We focused on proteins that had been identified in the FLAG-Bach2 sample but absent from the mock purification. The list of these proteins included nuclear co-repressors NCoR1, NCoR2 (also know as SMRT) in addition to Rif1, HDAC3, Tbl1x, MafG, MafK and other proteins (Table 1). Bach2 was found to interact with several DNA binding transcription factors, chromatin-binding factors, RNA binding factors, and several proteins involved in ubiquitination such as Huwe1 (Fig. 5A). By integrating these results with the known protein interaction database, some of these proteins appear to interact with Bach2 by forming distinct complexes (Fig. 5B). The list of Bach2-interacting proteins will aid future analysis into the function and regulation of Bach2 in B cells. The presence of HDAC3, Tbl1x, and Rif1 in the Bach2 complex was verified by immunoblotting analyses using HDAC3-, Tbl1x-, or Rif1-specific antibodies (Fig. 6A). The possible interaction between endogenous Bach2 and HDAC3 in BAL17 cells was examined using an immunoprecipitation assay. Not only endogenous Bach2 but also endogenous HDAC3 were precipitated by the anti-Bach2 antibodies (Fig. 6B). In a reciprocal assay, endogenous HDAC3 and Bach2 were precipitated by the anti-HDAC3 antibodies (Fig. 6C). The enriched Bach2 migrated faster than the major form of Bach2. This band may Swi/Snf complex FIGURE 5. Proteins interacting with Bach2. A, schematic representation of candidate proteins interacting with Bach2. The mass spectrometry data from independent purifications were compiled. Proteins were selected based on two criteria, protein score of more than 130 and being absent in the mock samples. The protein score of HDAC3 was lower than 130 in this experiment but is indicated here to give its context. The thickness of each edge represents protein scores. Proteins are colored based on their representative functions as in the color key. B, interaction of Bach2 with known protein complexes. Known interactions are integrated with the results in A.

Bach2 Complex in Plasma Cell Differentiation
represent an alternative splicing form of Bach2 (14) or a posttranslationally modified Bach2. These results showed that endogenous Bach2 and HDAC3 interacted with each other.
Originally, Rif1 has been implicated in genome stability such as telomere DNA synthesis in yeast and regulation of DNA replication timing and DNA repair in mammalian cells (29). Recently, it has been reported that an interaction between Rif1 and 53BP1 is associated with DNA repair response during CSR in activated B cells (30 -33). However, its role in transcription remains less clear. To confirm the interaction between Rif1 and Bach2, we performed co-immunoprecipitation analysis of FLAG-Rif1 and Bach2 overexpressed in 293T human embryonic kidney cells. We detected their co-immunoprecipitation (Fig. 6D), confirming that Rif1 and Bach2 form a protein complex. We next examined whether HDAC3 and Rif1 interact with each other. When overexpressed in 293T cells, HA-HDAC3 was co-precipitated with FLAG-Rif1 (Fig. 6E).
Bach2-interacting Proteins Are Involved in the Repression of Prdm1-To examine whether the Bach2-mediated repression of Prdm1 involves HDACs, we first treated BAL17 cells with the HDAC inhibitor trichostatin A (TSA) for 12 h. The amount of Prdm1 mRNA increased following TSA treatment (Fig. 7A), suggesting that Prdm1 is repressed in mature B cells by TSAsensitive HDACs. To investigate whether HDAC3 is involved in the Prdm1 repression, we treated BAL17 cells with a HDAC3 selective inhibitor, RGFP966, at different concentrations for 12 h and assessed the Prdm1 expression (Fig. 7B). Quantitative RT-PCR analysis clearly demonstrated that inhibition of HDAC3 resulted in a significant up-regulation of Prdm1 expression, suggesting that HDAC3 is required for Prdm1 repression. To confirm this observation, an RNAi strategy was used to reduce the expression of HDAC3 mRNA in mouse splenic B220-expressing B cells to investigate whether HDAC3 was required for Prdm1 repression. Mouse splenic B cells were preactivated with LPS for 12 h to induce plasma cell differentiation and then electroporated with HDAC3 siRNA or control siRNA. The expression levels of HDAC3 and Blimp-1 mRNAs were determined after 36 and 48 h by RT-qPCR. Although the knockdown of HDAC3 was transient, it resulted in a mild but reproducible derepression of Prdm1 (Fig. 7C). Furthermore, ChIP experiments clearly demonstrated that HDAC3 binds to the MAREs at the Prdm1 locus (Fig. 7D). Twelve hours after treatment of BAL17 cells with RGFP966, the levels of H3K27ac increased dramatically at both the promoter and intron 5 MAREs (Fig. 7E). Collectively, these results suggested that HDAC3 was involved in the repression of Prdm1 in B cells through altering epigenetic modification.
HDAC3 is known to associate with co-repressors such as NCoR1 or its related factor NCoR2. Their complexes are recruited to target genes by DNA-binding transcription factors (34,35). To investigate whether NCoR1 is also involved in the repression of transcription of the Prdm1 gene, we knocked down the expression of NCoR1 by short hairpin RNA (shRNA). The NCoR1 mRNA expression levels were reduced to nearly 50% by two independent shRNAs (Fig. 7F, left panel). The expression levels of Blimp-1 mRNA were significantly increased (Fig. 7F, right panel). Together with the previously reported interaction between Bach2 and NCoR2 (17), these results suggest that HDAC3 forms co-repressor complexes with NCoR1 and/or NCoR2 at the Prdm1 locus.
To assess the putative function of Rif1 in the regulation of Prdm1, we performed ChIP assays of Rif1. We observed a recruitment of Rif1 to both of the Prdm1 regions examined (Fig.  8A). These experiments also revealed that Rif1 was recruited to the Mcm5 promoter region, suggesting that Rif1 may possess a widespread function. To explore the involvement of Rif1 in Prdm1 repression, we carried out knockdown of Rif1 by two independent short hairpin RNA (shRNA) clones in BAL17 cells. Although we confirmed a reduction of the Rif1 expression by RT-qPCR, the amount of Prdm1 transcript was virtually unaffected (Fig. 8B). The role of Rif1 for the Bach2-mediated transcription repression therefore remains to be clarified.
Bach2 Interacts with HDAC3 and Mediates the Repression of Prdm1 in Primary B Cells-To further confirm that Bach2 was involved in the repression of Prdm1 together with HDAC3, we utilized mature B cells isolated from B1-8 hi knock-in mice carrying immunoglobulin heavy chain gene that can produce an antibody to a pre-specified antigen (23). The mature B cells isolated from these mice show rather uniform reactions upon activation in vitro. The B1-8 hi mature B cells were stimulated with an antigen, and nuclear extracts were prepared. Endogenous Bach2 was immunoprecipitated. Immunoblot analyses of the precipitates revealed that endogenous MafK and HDAC3 45 were both co-precipitated along with Bach2 (Fig. 9A). The signal for NcoR1 was also detected, although it was not clear due to its large molecular weight. We confirmed their presence using mass spectrometry analysis of the immunoprecipitates. In contrast, we failed to detect an interaction of Bach2 with Rif1 in these experiments. Reducing Bach2 by an shRNA-mediated interference in B1-8 hi mature B cells resulted in increases in the expression of Prdm1 and plasma cell differentiation (Fig. 9, B and C). Therefore, these results established that Bach2 interacts with HDAC3 in primary B cells and suggest that this interaction is important for the repression of Prdm1 by Bach2.

Discussion
There has so far been no report on histone modification at the Prdm1 locus in B cells and plasma cells. By comparing histone modifications at the Prdm1 regulatory regions in BAL17 and X63/0 cells, we found that acetylation and methylation of H3-K9 at the intron MARE region correlated well with the Prdm1 expression in these cells. In addition, acetylation of H3-K18 and Lys-27 in this region was higher in X63/0 than in BAL17 cells. Therefore, we suggest that H3-K9 acetylation inhibits Lys-9 methylation, promoting Prdm1 expression in plasma cells. An important finding in this study is that Bach2 decreased the level of H3-K9 acetylation in primary B cells activated with LPS. Because the kinetics of transition between activated B cell and plasma cell is governed by Bach2 (19), our present observations suggest that Bach2 inhibits the process of plasma cell differentiation by promoting epigenetic modifications (low H3K9ac and high H3K9me) at the intron MARE region. Because H3-K9 acetylation at the promoter MARE region was also increased in Bach2-deficient primary B cells, Bach2 appears to regulate histone modification at the promoter region as well.
The analysis of Bach2-interacting proteins in BAL17 cells revealed several interesting candidate molecules for the epigenetic regulation by Bach2. Regarding the histone deacetylation, the presence of HDAC3, NCoR1, and NCoR2 is important. Because we have reported previously that Bach2 interacts directly with NCoR2 (17) and that they co-localize in nuclear foci (36), Bach2 may also directly interact with NCoR1. Both NCoR1 and NCoR2 are known to form co-repressor complexes with HDAC3 and other histone deacetylases (34). Therefore, Bach2 appears to recruit NCoR1 and NCoR2 complexes to its target genes to promote histone deacetylation. Consistent with this model, Prdm1 expression was increased in primary B cells upon knockdown of HDAC3 or NCoR1 and in response to the specific HDAC3 inhibitor RGFP966. It should be noted that Bcl6, with which Bach2 cooperates to repress the expression of Prdm1 in B cells, interacts with both NCoR1 and NCoR2 (37,38). Importantly, the binding sites of Bach2 and Bcl6 in the intron 5 are juxtaposed, and Bcl6 fails to repress Prdm1 reporter expression with a mutation at the MARE (16), suggesting that Bach2 is required for proper epigenetic regulation at the region by cooperating with Bcl6 in recruiting these co-repressor complexes. Recently, HDAC3 has been reported as a new therapeutic target of multiple myeloma (39). Considering that Prdm1 is essential for the terminal differentiation of plasma cells, an inhibition of HDAC3 in multiple myeloma may confer a therapeutic effect by promoting Prdm1 expression.
This study also identified other molecules that may be involved in the transcription repression by Bach2. Interestingly, Rif1 is known to interact with Sirt3, Sirt7 (40), and Setdb1 (41,42). Sirt3 and Sirt7 deacetylate histones, whereas Setdb1 methylates histone at H3-K9. Therefore, Rif1 may recruit additional histone-modifying enzymes such as those to the Prdm1 locus. Another interesting category of molecules includes Arid1a (BAF250) and Smarcd2 (BAF60B), which are the components of SWI/SNF chromatin remodelers (43), and Nasp, a histone chaperone (44). The presence of these molecules in the Bach2 complex suggests that chromatin remodeling may also be involved in the repression of Prdm1 and other target genes in B cells. Because Bach2 regulates the expression of diverse sets of genes important not only for B cells but also T cells (45)(46)(47)(48)(49)(50) and macrophages (51), further studies of the Bach2-interacting proteins will allow mechanistic insights into the regulation of the immune system in the light of a protein network for gene expression.