Oncogenic activation of c-Myb correlates with a loss of negative regulation by TIF1beta and Ski.

The c-myb proto-oncogene product (c-Myb) regulates proliferation of hematopoietic cells by inducing the transcription of a group of target genes. Removal or mutations of the negative regulatory domain (NRD) in the C-terminal half of c-Myb leads to increased transactivating capacity and oncogenic activation. Here we report that TIF1beta directly binds to the NRD and negatively regulates the c-Myb-dependent trans-activation. In addition, three corepressors (Ski, N-CoR, and mSin3A) bind to the DNA-binding domain of c-Myb together with TIF1beta and recruit the histone deacetylase complex to c-Myb. Furthermore, the Drosophila TIF1beta homolog, Bonus, negatively regulates Drosophila Myb activity. The Ski corepressor competes with the coactivator CBP for binding to c-Myb, indicating that the selection of coactivators and corepressors is a key event for c-Myb-dependent transcription. Mutations or deletion of the NRD of c-Myb and the mutations found in the DNA-binding domain of v-Myb decrease the interaction with these corepressors and weaken the corepressor-induced negative regulation of Myb activity. These observations have conceptual implications for understanding how the nuclear oncogene is activated.

The c-myb proto-oncogene is the cellular progenitor of the v-myb oncogenes carried by the chicken retroviruses avian myeloblastosis virus (AMV) 1 and E26, which cause acute myeloblastic leukemia or erythroblastosis (1,2). The level of c-myb expression is high in immature hematopoietic cells, and its expression is turned off during terminal differentiation (3). c-myb-deficient mice show a defect in definitive hematopoiesis in the fetal liver due to a severe reduction in the number of progenitor cells, indicating that c-myb is essential for the proliferation of immature hematopoietic cells (4). Analysis of homozygous null c-myb/Rag1 chimeric mice indicates that c-myb is also essential for early T-cell development (5). The myb gene is well conserved not only in vertebrates but also in other species. Drosophila melanogaster has one myb gene (dmyb), which is required in diverse cellular lineages throughout the course of development (6).
The c-myb gene product (c-Myb) is a transcriptional activator that recognizes the specific DNA sequence 5Ј-AACNG-3Ј (7)(8)(9)(10). Some of the c-Myb target genes, including c-myc, are required for the G 1 /S transition in the cell cycle (11,12). In contrast, dmyb is required for the G 2 /M transition (6), and cyclin B expression is directly regulated by dMyb (13). Several other target genes, including mim-1, GBX2, and bcl-2, are involved in lineage commitment in differentiation and blockage of apoptosis (10, 14 -16). c-Myb has three functional domains that are responsible for DNA binding, transcriptional activation, and negative regulation (8). The DNA-binding domain (DBD) in the N-terminal region of c-Myb consists of three imperfect tandem repeats of 51-52 amino acids, each containing a helix-turn-helix variation motif. Repeats 2 and 3 (R2 and R3) are sufficient for binding to the target DNA sequence (17). The transcriptional activation domain is adjacent to the DBD, to which the transcriptional coactivator CBP binds (18).
Analysis of various oncogenically activated myb genes suggests that truncation of either the N or C terminus of c-Myb can cause oncogenic activation. For example, the v-Myb protein encoded by AMV is N-and C-terminally truncated versions of c-Myb (1). Deletion of the negative regulatory domain (NRD) located in the C-terminal portion of the molecule increases both the trans-activation and transformation capacity of c-Myb, implying that the NRD normally represses c-Myb activity (8, 19 -21). The v-Myb encoded by AMV lacks the C-proximal region of the NRD. In addition, the mutations of only the leucine-rich region in the NRD result in oncogenic activation of c-myb (22). Thus, the NRD appears to contain multiple subdomains, and the deletion of any of these may result in the oncogenic activation of c-myb. However, the mechanism by which c-Myb is regulated by NRD still unclear.
Here, we demonstrate that four corepressors, including c-Ski, directly bind to c-Myb via multiple domains in the c-Myb molecule to negatively regulate c-Myb activity. Deletions or mutations of the NRD or the point mutations found in v-Myb reduces the affinity with these corepressors, leading to increased c-Myb activity. Thus, our results suggest that selection of coactivators or corepressors is a key event for oncogenic activation of c-Myb.

MATERIALS AND METHODS
Yeast Two-hybrid Screening and in Vitro Binding Assays-The yeast two-hybrid screening was performed using the mouse embryonic cDNA library as described previously (23). The protein containing the Cterminal 312 amino acids of mouse c-Myb was used as bait. GST pull-down assays were performed as described previously (23). To increase the solubility of GST fusion proteins expressed in bacteria, the thioredoxin coexpression system (34) was used. The binding buffer used for most of the experiments consists of 20 mM Hepes, pH 7.5, 1 mM dithiothreitol, 0.1% Nonidet P-40, and 100 mM KCl (for interactions between Myb and mSin3A or N-CoR) or 150 mM KCl (for interactions between Myb and c-Ski). The binding buffer used for the experiments with the R23 fragment consists of 50 mM phosphate buffer, pH 6.8, 20 mM dithiothreitol, and 100 mM KCl.
Coimmunoprecipitation and HDAC Assay-For coimmunoprecipitation of endogenous proteins, lysates were prepared from Molt-4 cells by mild sonication in NET buffer (20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.5% Nonidet P-40, protease inhibitor mixture) containing 150 mM NaCl. Anti-c-Myb monoclonal antibody 1-1, the rabbit anti-N-CoR antibody (23), the anti-mSin3A antibody (Santa Cruz Biotechnology, AK-11), or normal rabbit IgG were used for immunoprecipitation. The immunocomplexes were used in Western blotting with rabbit anti-TIF1␤ polyclonal antibodies raised against GST-TIF1␤C, anti-c-Ski monoclonal antibody, rabbit anti-mSin3A antibody (Santa Cruz Biotechnology, AK-11), or anti-c-Myb monoclonal antibody 1-1. To study the interaction between c-Ski and various forms of c-Myb, 293 cells were cotransfected via the CaPO 4 method with the c-Ski expression plasmid pact-Ski (10 g) and the c-Myb expression plasmid pact-FLAG-c-Myb (10 g). Forty hours after transfection, cells were lysed as described (23), and immunoprecipitation was performed with the anti-FLAG monoclonal antibody (M2, Sigma). Western blotting was performed using the anti-c-Ski monoclonal antibody. Assays for HDAC activity were performed essentially as described (23) using lysates prepared from 293 cells that were transfected with 10 g of the c-Myb expression plasmid.
Analysis of Repressor Domains in c-Myb-The cytomegalovirus promoter was used to express the Gal4-c-Myb fusion proteins consisting of the Gal4 DNA-binding domain fused to various portions of c-Myb. CV-1 cells were transfected with a mixture of 3 g of the luciferase reporter containing the TK promoter and six Gal4-binding sites, 0.33 g of the Gal4-c-Myb or Gal4 expression plasmids, and 1 g of the internal control plasmid pRL-TK (Promega). The luciferase assays were performed using the dual-luciferase assay system (Promega).
Subcellular Localization of c-Myb and Corepressors-CV-1 cells were transfected with a mixture of 1.5 g of the FLAG-c-Myb expression plasmid and 1.5 g of the plasmids that express c-Ski, mSin3A, or N-CoR. Forty hours after transfection, cells were fixed and stained as described (23) with anti-c-Myb, anti-c-Ski, and anti-FLAG antibodies. The signals for the different proteins were visualized by rhodamineand fluorescein isothiocyanate-conjugated secondary antibodies and analyzed by confocal microscopy.
Chromatin Immunoprecipitation Assays-The retroviral expression plasmids for wild-type c-Myb or CT3 were constructed using the MSCV (murine stem cell virus)-based retroviral vector, and viruses were prepared as described (35). To generate M1 cell clones that express c-Myb together with the neomycin resistance gene, M1 cells were infected with viruses and then grown in the presence of G418 (400 g/ml). ChIP assay was carried out essentially by using the method of Weinmann and Farnham (36). In brief, 1.5 ϫ 10 7 M1 cells were fixed with 1% formaldehyde for 10 min at room temperature. Nuclei were isolated and suspended in nuclei lysis buffer (50 mM Tris-HCl, pH 8.1, 10 mM EDTA, 1% SDS, protease inhibitors) and sonicated. After the centrifugation, the supernatant was diluted with IP dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8.1, 167 mM NaCl) and treated with antibody against each corepressor. The immunocom-plex was collected and was incubated at 65°C with IP elution buffer (50 mM NaHCO 3 , 1% SDS) to release the proteins and DNA complex. DNA was extracted and used for PCR. PCR reaction (94°C for 45 s, 55°C for 30 s, and 72°C for 3 min) was carried out with [ 32 P]dCTP for 30 cycles. PCR products were analyzed by electrophoresis on an 8% polyacrylamide gel. The primers used for the amplification of myc promoter were as follows: 782 GTGCCCAGTCAACATAACTGTACG 805 and 1101 GGCG-TATTGTGTGGAGCGAGGCAG 1124 .
Luciferase Reporter Assays-In the experiments using the reporter containing multiple Myb-binding sites, CV-1 cells (2 ϫ 10 5 cells per 60-mm dish) were cotransfected using LipofectAMINE Plus (Invitrogen) with the 6MBS-I-SV40-luc reporter (0.2 g), the c-Myb expression plasmid (0.03 g), the corepressor expression plasmid (0.5 or 1 g), and the internal control plasmid pRL-TK (0.05 g), followed by luciferase assays. The chicken ␤-actin promoter was used to express c-Myb and various corepressors. In the case of assays using the 6MBS-I-TK-luc reporter (1 g), the plasmid to express c-Myb (0.1 g), or the v-Myb (0.03 g) was used together with the same amounts of other plasmids as described above. The dominant negative form of the TIF1␤ expression plasmid was constructed by inserting the DNA fragment encoding the RBCC motifs and the artificial nuclear localization signal into the ␤-actin promoter-based vector. In the experiments using the reporter containing the c-myc promoter, CV-1 cells (4 ϫ 10 5 cells per 100-mm dish) were cotransfected by the CaPO 4 method with the myc-CAT reporter (4 g) (11), the c-Myb expression plasmid (4 g), the corepressor expression plasmid (0.5 or 2.5 g), and the internal control plasmid pact-␤-gal (0.3 g), followed by luciferase assays. The total amount of plasmid DNA was adjusted to 12 g by adding the control plasmid DNA lacking the cDNA.
Genetic Interaction between Bon and dMyb-Two alleles of dmyb mutants were described previously (13). The bon 241 and bon 487 mutants (37) were provided by H. J. Bellen. Eye imaginal discs were dissected from late third-stage larvae, fixed, and stained with the anti-CycB antibody (gift from C. Lehner) as described (13). For analysis of the lethal stages, dmyb 2507 (hypomorph) and dmyb 1 (hypomorph) were balanced by FM7c,y and bon 21B (amorph (37)) was balanced by TM6, Tb. dmyb 2507 /FM7c,y and dmyb 1 /FM7c,y females were mated with males of ϩ/Y or ϩ/Y; bon 21B /TM6, Tb. The offspring were grown at 25°C, and the males of third instar larvae hemizygous for dmyb and the FM7c balancer were selected by the size of the gonads embedded in the opaque fat body. Furthermore, the male larvae were separated by the presence or the absence of the y ϩ cuticular phenotypes. The bon 21B chromosome was isolated by the absence of the Tb phenotype. The viability was calculated as a standard value from the number of male larvae hemizygous for FM7c/Y or FM7c/Y; bon 21B /ϩ.

TIF1␤ Binds to the NRD of c-Myb-To identify putative
inhibitors that bind to the NRD of c-Myb, we performed yeast two-hybrid screening using the NRD of mouse c-Myb as bait. This resulted in the identification of clones encoding TIF1␤ (also called as KAP-1). TIF1␤, which contains a RING finger, B boxes, a coiled-coil region, and a plant homeodomain (PHD) finger, was originally identified by two groups as a protein that binds to the heterochromatin protein-1 (HP-1) (38) or to the KRAB repression domain (39). TIF1␤ was subsequently found to act as a corepressor that associates with the HDAC complex and HP-1 (40,41). The TIF1␤ clones isolated in our screening encoded the 92-amino acid region that includes the B1 box (Fig.  1A). In vitro translated c-Myb efficiently bound to the GST-TIF1␤N protein containing the N-terminal half of TIF1␤, but not to the GST-TIF1␤C protein that contains the C-terminal half of TIF1␤ (Fig. 1A, lower middle panel). The bacterially expressed recombinant NRD also bound to the GST-TIF1␤N protein (Fig. 1A, lower right panel).
We then performed GST pull-down assays using in vitro translated TIF1␤ and GST-NRD fusion proteins that contain the NRD of c-Myb. TIF1␤ efficiently bound to the GST-NRD, but mutation of the leucine-rich region (L34P) dramatically decreased but did not completely abrogate the affinity with TIF1␤ (Fig. 1B, lower left panel, see also Supplementary Fig.  1B for GST-NRD proteins). We also performed GST pull-down assays using GST-TIF1␤N and a series of in vitro translated C-terminally truncated forms of c-Myb bearing the mutation L34P in the leucine-rich region. The c-Myb protein truncated up to amino acid 500 still retained the capacity to interact with TIF1␤, but truncation up to amino acid 444 completely abrogated binding, indicating that TIF1␤ binds to the region between amino acids 444 and 500 (Fig. 1B, lower right panel). Thus, TIF1␤ interacts with c-Myb at both the leucine-rich region and a C-terminal region in the NRD. Supporting this is that TIF1␤ binds to AMV-v-Myb, which lacks the C-terminal region of the NRD, with lower affinity than to c-Myb (Fig. 1B).
Binding of c-Myb with Multiple Corepressors-A recent study on corepressors demonstrates that multiple corepressors bind to the same transcriptional factor. This raises the possibility that several other corepressors may bind to c-Myb together with TIF1␤. To investigate this, we analyzed the interaction between c-Myb and three other known corepressors, c-Ski, N-CoR, and mSin3A. All three in vitro translated pro-teins bound to GST-c-Myb ( Fig. 2A). In vitro translated mSin3A and c-Ski also bound to GST-TIF1␤N ( Fig. 2A). We previously demonstrated that c-Ski binds to both N-CoR and mSin3A (23). Thus, N-CoR, mSin3A, and TIF1␤ can all bind to c-Ski. Given that the four corepressors can interact with each other, these results suggest that these corepressors bind to c-Myb simultaneously or sequentially (see "Discussion").
To investigate the in vivo interaction between c-Myb and the corepressors, we performed the coimmunoprecipitation experiments using Molt-4 cell lysates (Fig. 2B). The anti-Myb antibody precipitated TIF1␤, c-Ski, and mSin3A. Furthermore, c-Myb and TIF1␤ were coprecipitated using the anti-N-CoR antibodies. The control IgG precipitated none of the corepressors or c-Myb. We then asked whether the c-Myb complex contains HDAC activity. We used anti-Myb to generate the immunocomplexes from 293 cells that had been transfected with the c-Myb expression plasmid and assessed their HDAC activity. A significant level of HDAC activity was observed compared with the immunocomplexes prepared with control IgG (Fig. 2C). The immunocomplexes of the c-Myb mutant containing the mutated leucine-rich region in the NRD (L34P) had slightly lower HDAC activity than those containing wildtype c-Myb (Fig. 2C). These observations are consistent with the fact that TIF1␤ only partly interacts with c-Myb through the leucine-rich region.
Using various forms of in vitro translated c-Myb, we determined which of the one or more regions of c-Myb is bound by c-Ski (Fig. 3A). The results obtained using GST-Ski fusion proteins indicate that c-Ski binds to the DBD of c-Myb (CT5), but not to the mutant lacking the DBD (⌬DB) ( Supplementary  Fig. 1, data are summarized in Fig. 3A). The region containing only R2 and R3 (R23) was sufficient for c-Ski interaction. Interestingly, removal of the NRD (CT3) or mutations in the leucine-rich region (L34A and L34P) almost completely abrogated the interaction with c-Ski, although c-Ski does not bind to NRD. It may be that the mutations or removal of NRD alters the conformation of some of the regions such as the transactivating domain, which then blocks access to c-Ski.
We then performed GST pull-down assays using various forms of in vitro translated c-Myb and GST-mSin3A-NCT. The latter protein contains the central region of mSin3A bearing the second and third putative paired amphipathic helical domains (PAH2 and PAH3). This is the region responsible for binding to c-Myb (Supplementary Fig. 2A). As with c-Ski, mSin3A binds to R2 and R3 of c-Myb ( Supplementary Fig. 2B, and data are summarized in Fig. 3A). The mutations of leucinerich region of the NRD (L34A and L34P) also partly decrease the affinity with mSin3A, but removal of the NRD (CT3) does not affect the affinity with mSin3A. To determine which region of c-Myb interacts with N-CoR, similar assays were performed. c-Myb binds to the N-terminal 427-amino acid region of N-CoR ( Supplementary Fig. 3A), and when GST-N-CoR, which contains this region, was used in GST pull-down assays, N-CoR was found to interact with R2 and R3 of c-Myb ( Supplementary  Fig. 3B, and data are summarized in Fig. 3A). However, the To confirm that the three corepressors directly bind to the DBD of c-Myb, GST pull-down assays were performed using the bacterially expressed c-Myb protein containing only the three DBD repeats (R123) and the c-Ski, mSin3A, or N-CoR GST fusion proteins. The R123 proteins bound directly to these GST-corepressor fusions (Fig. 3B). Thus, the three corepressors interact directly with the DBD.
c-Ski Abrogates the Interaction between c-Myb and CBP-The binding of c-Myb with the corepressor-HDAC complexes suggests that c-Myb bears a transcriptional repressor domain. We thus examined the repressor activities of fusion proteins consisting of the Gal4 DNA-binding domain fused to various portions of c-Myb (Fig. 4A). Both the Gal4 fusion proteins containing the NRD (NRD) or the c-Myb mutant lacking the transcriptional activation domain (⌬TA) significantly repressed the activity of the Gal4 site-containing promoter. The Gal4-NRD fusion containing mutations of the leucine-rich region had a slightly lower repressor activity than that containing the normal NRD (Fig. 4A). These results are consistent with our observations that the corepressors c-Ski, mSin3A, and N-CoR interact with c-Myb via the DBD and the corepressor TIF1␤ interacts via the NRD. Furthermore, because we know that the mutation of the leucine-rich region partly and completely disrupts the interaction with TIF1␤ and c-Ski, respectively, this may explain why L34P had lower repressor activity than the wild type NRD.
That the three corepressors c-Ski, mSin3A, and N-CoR all bind to the DBD of c-Myb suggests the corepressors could bind to c-Myb simultaneously or that they compete in their c-Myb binding. To discriminate these possibilities, we investigated the binding of in vitro translated [ 35 S]mSin3A to a small amount of GST-c-Myb resin in the presence of increasing amounts of in vitro translated c-Ski (Fig. 4B). The addition of c-Ski to the binding reaction did not decrease the amount of mSin3A bound to GST-c-Myb, suggesting that these corepressors do not compete in their binding to the c-Myb DBD. Given that the four corepressors can interact with each other, these results suggest that the corepressors bind to c-Myb simultaneously or sequentially (see "Discussion").
For the c-Myb-mediated transcriptional activation, CBP must bind to the transcriptional activation domain of c-Myb. The simultaneous binding of the four corepressors to c-Myb may mask the surface of the c-Myb protein and thereby block CBP interaction with c-Myb. To examine this hypothesis, we investigated whether the corepressors compete with CBP for binding to c-Myb (Fig. 4C). Thus, the binding of in vitro translated c-Myb to the GST-CBP fusion protein, which contains the Myb-binding domain (KIX) of the CBP molecule, was measured in the presence of an increasing amount of c-Ski translated in vitro. Wild-type c-Ski inhibited the interaction between c-Myb and GST-CBP in a dose-dependent manner. We know that the C-terminal coiled-coil region of c-Ski is responsible for interaction with c-Myb ( Supplementary Fig. 4). When increasing amounts of c-Ski mutant lacking this c-Myb-interacting region (⌬493-728) were added to c-Myb and GST-CBP, the affinity between c-Myb and CBP was only slightly decreased. We observed that in vitro translated Ski did not bind to the GST-CBP-KIX resin, indicating that the region of 493-728 amino acids of Ski does not mask the Myb-binding domain of CBP. Thus, c-Ski and CBP compete with each other for binding to c-Myb.
Loss of the NRD Lowers the Affinity between c-Ski and c-Myb-c-Ski and N-CoR have been reported to colocalize to dot-like domains in the nuclei (23). To confirm that c-Myb associates with the corepressors in vivo, we investigated whether c-Myb colocalizes with the corepressors. When c-Myb was coexpressed with c-Ski, N-CoR, or mSin3A in transfected CV-1 cells, most of the c-Myb signals were colocalized with the corepressors in dot-like structures (Fig. 5A). However, when c-Ski was coexpressed with the c-Myb mutant lacking the NRD (CT3) or containing a mutation in the leucine-rich region (L34P), the nuclear punctate structures of c-Ski was disrupted, resulting in c-Ski being mainly localized at the peripheral region in the nuclei (Fig. 5B). In these cells, c-Myb formed the nuclear dot-like structures, and the c-Ski and c-Myb signals did not overlap. Thus, removal of the NRD of c-Myb disrupts an interaction with c-Ski.
To confirm the effect of deleting the NRD on the interaction of c-Myb with c-Ski, coimmunoprecipitation experiments were performed (Fig. 5C). 293 cells were transfected with the plasmids that express c-Ski and the FLAG-linked wild-type or mutant forms of c-Myb. Lysates were prepared and used for coimmunoprecipitation. Anti-FLAG antibody coprecipitated c-Ski with wild-type c-Myb, but only a small amount of c-Ski was coprecipitated with the CT3 mutant that lacks the NRD. Although a significant amount of c-Ski was coprecipitated with L34P mutant, which has a mutant leucine-rich region, the number of c-Ski molecules per L34P molecule precipitated was still less than that with wild-type c-Myb. Thus, loss of the NRD or mutations of the leucine-rich region in the NRD decrease the affinity of c-Myb with c-Ski.
We then examined the recruitment of the corepressors by c-Myb to the c-myc gene promoter by chromatin immunoprecipitation (ChIP) assays. It was shown that the transcription of c-myc gene is directly regulated by c-Myb in a myeloblastic cell line, M1 (12). In addition, we previously demonstrated that the recombinant c-Myb proteins directly bind to the multiple sites in the c-myc promoter region (11). We generated M1 cell clones that constitutively express wild-type c-Myb or the CT3 mutant as well as control clones containing the empty vector by using the retrovirus vector (Fig. 5D). Anti-c-Myb antibody precipitated the c-myc promoter DNA fragment of the control M1 cells, which contains multiple Myb-bindings sites (Fig. 5E, left panels), and other regions that contain no Myb-binding sites were not precipitated (data not shown). These results indicate that the c-Myb proteins bind to the c-myc promoter region in M1 cells. Overexpression of wild-type or CT3 c-Myb increased the amounts of c-myc fragment precipitated by anti-Myb antibody (Fig. 5E, left panels), suggesting that the number of c-Myb molecules that bound to the c-myc promoter increased in Myboverexpressing M1 cells. The antibodies against Ski, mSin3A, or N-CoR precipitated a small or undetectable amount of c-myc DNA fragment of the control M1 cells (Fig. 5E, right panels).
Overexpression of wild-type c-Myb induced an increase in the occupancy by Ski and N-CoR of the c-myc gene promoter (Fig.  5E, right panels). Because c-Myb binds to either the corepressors or the coactivators, some of c-Myb molecules on the c-myc promoter probably interact with the corepressors. Therefore, an increase in the c-Myb molecule on the c-myc promoter leads to an increase in the corepressor molecule on the c-myc promoter. However, the amounts of mSin3A on the c-myc promoter were not dramatically increased by overexpression of wild-type c-Myb, suggesting that enough of the mSin3A occupies the c-myc promoter via the endogenous c-Myb. Furthermore, the amounts of c-Ski, N-CoR, and mSin3A recruited by the overexpressed CT3 mutant to the c-myc promoter were less than those by overexpressed wild-type c-Myb (Fig. 5E, right panels). These results suggest that CT3 has less affinity with these corepressors than wild-type c-Myb. Thus, loss of the NRD decreases the affinity of c-Myb with the corepressors on the c-myc promoter.
Negative Regulation of c-Myb-dependent Transcriptional Activation by Corepressors-We then investigated the effect of the corepressors on c-Myb-mediated activation of a promoter that contains the c-Myb-binding sites (Fig. 6). CV-1 cells were cotransfected with the luciferase reporter containing the SV40 promoter linked to six tandem c-Myb-binding sites (MBS-I). c-Myb stimulated luciferase expression 6.2-fold. This c-Mybmediated activation was inhibited by mSin3A and N-CoR in a c-Myb and corepressors were visualized by rhodamine-and fluorescein isothiocyanate-conjugated secondary antibodies, respectively, using laser confocal microscopy. In the right panel, the signals for both proteins are superimposed. B, c-Myb mutants localize separately from c-Ski. CV-1 cells were transfected with the plasmids that express c-Ski and the c-Myb mutants CT3 or L34P. The cells were then immuno-stained as described above. C, loss of NRD lowers the affinity of c-Myb with c-Ski. Lysates were prepared from 293 cells transfected with a mixture of c-Ski and FLAG-c-Myb expression plasmids. The wild-type c-Myb or the two mutants L34P and CT3 were used. Lysates were precipitated by anti-FLAG or control IgG and the immunocomplexes were analyzed by Western blotting using anti-c-Ski or anti-c-Myb antibodies. Samples from the lysates were also used directly for Western blotting (bottom). D, generation of M1 cell lines expressing wild-type c-Myb or CT3 mutant. Western blotting was performed with the anti-Myb antibody using the whole cell lysates prepared from the M1 cells containing the retrovirus vector to express wild-type c-Myb or CT3, or the control vector. E, ChIP assays. Soluble chromatin was prepared from the three types of M1 cells and immunoprecipitated with the antibodies indicated above each lane. The final DNA extractions were amplified using pairs of primers that cover the region of c-myc promoter. The relative densities of bands are indicated below.
dose-dependent manner to 23 and 40% of the control, respectively (Fig. 6, A and B). Both mSin3A and N-Cor did not affect the basal promoter activity in the absence of c-Myb (Supplementary Fig. 5, A and B). The two c-Myb mutants that lacked (CT3) or had mutations in the leucine-rich region (L34P) activated the same promoter better than the wild type c-Myb (15.4and 17.1-fold, respectively) as reported (8,22). mSin3A and N-CoR inhibited the trans-activation by CT3 and L34P less efficiently (to 69 -80% of the control) compared with their inhibition of wild-type c-Myb. We also examined the effect of mSin3A on the AMV v-Myb-mediated trans-activation using the luciferase reporter containing the thymidine kinase promoter linked to six tandem MBS-I sites (Fig. 6C). mSin3A did not inhibit the trans-activation by v-Myb.
The c-Myb-mediated activation of the MBS-I promoter was not inhibited or only weakly when TIF1␤ was overexpressed (data not shown), probably because sufficient levels of endogenous proteins were present (Supplementary Fig. 5C). We there-fore examined the effect on Myb-mediated activation of the dominant negative form of TIF1␤ (DN-TIF1␤), which contains only the N-terminal RING finger and B boxes (Fig. 6D). DN-TIF1␤ enhanced the wild-type c-Myb-mediated activation in a dose-dependent manner, but it did not affect the activation mediated by the CT3 mutant.
We also investigated the effect of the corepressors on c-Mybmediated activation of the human c-myc promoter (Fig. 6E). CV-1 cells were cotransfected with the c-myc-CAT reporter and the c-Myb expression plasmid. CAT expression was enhanced 5.0-fold. Coexpression of mSin3A decreased CAT expression to 20% of the control in a dose-dependent manner. When v-Myb encoded by AMV was used instead of c-Myb, mSin3A inhibited its activity more poorly (to 78% of the control) than when wild-type c-Myb was employed.
Although the less efficient inhibition of the trans-activating capacity of v-Myb by the corepressors may be, at least partly, due to the lack of the C-proximal TIF1␤-binding site in v-Myb, there might be an additional mechanism. Thus, we examined the effects of point mutations in the DBD of v-Myb on the interaction with the corepressors. We performed the GST pulldown assays using the in vitro translated c-Myb fragment containing only the R2 and R3 of the DBD (R23) under the less stringent condition compared with that in Fig. 3. The R23 fragment of c-Myb had a higher affinity with GST fusion proteins containing c-Ski, mSin3A, or N-CoR than the R23 containing the three point mutations found in v-Myb (Fig. 6F). In the wild-type c-Myb, these three residues are hydrophobic amino acids on the surface of DBD (42), whereas these residues are mutated to non-hydrophobic amino acids in v-Myb. These results suggest that c-Ski, mSin3A, and N-CoR have the lower affinity with v-Myb than c-Myb due to the point mutations in R2.
The Drosophila TIF1␤ Homolog Inhibits Drosophila Myb Activity-To confirm that TIF1␤ negatively regulates c-Myb, we used the Drosophila member of the TIF1␤ family and c-Myb, namely, Bonus (Bon) and dMyb. Like TIF1␤, the Bonus protein has RBCC motifs in its N-terminal half, whereas its bromo domain is preceded by its PHD finger. Bon exhibits 29% identity (50% similarity) with mouse TIF1␤ and is thought to be the Drosophila homolog of TIF1␤ (37). In the GST pull-down assays using the GST-dMyb fusion protein and in vitro translated Bon, Bon bound to GST-dMyb efficiently (Fig. 7A). Recently, we isolated two alleles of dmyb mutants, namely, el1 and el2507 (13). These two alleles encode 514-and 305-amino acid proteins. Bon bound to the dMyb 2507 protein but not to dMyb 1 (Fig. 7A), indicating that, like the vertebrate homolog, Bon interacts with dMyb via the C-terminal half of dMyb.
We then investigated the genetic interaction between Bon and dMyb by using their mutants (Fig. 7B). Because the dmyb gene is on the X chromosome, the dmyb mutation in males (hemizygotes) was lethal at the first or second larval instar stage. Upon reaching the third larval instar stage, the mutant larvae died in the ensuing 2-3 days without further growth and development. The mutant larvae displayed obvious morphologic abnormalities as judged by larval cuticle preparations (data not shown). Only 6.9% of the dmyb 2507 hemizygotes survived until the third instar larval stage. However, the loss of one copy of bon increased the viable population to 80.8%, suggesting that Bon negatively regulates dMyb activity. In contrast, the loss of one copy of bon did not affect the lethality of dmyb 1 hemizygotes. This is consistent with the observation that Bon binds to the dMyb 2507 protein but not to the dMyb 1 protein.
We found recently that dMyb directly regulates the expression of cycB in eye imaginal discs and the cycB expression in the posterior region of eye discs is lost in dmyb mutant clone cells (13). We observed here that cycB expression is also almost completely lost in the posterior region of the dmyb 2507 eye discs (Fig. 7C) but that loss of one copy of bon appeared to recover the cycB expression. Thus, Bon acts as a negative regulator of dMyb. DISCUSSION The present study demonstrate that four complexes containing each of four corepressors, TIF1␤, mSin3A, c-Ski, and N-CoR, directly bind to c-Myb via the NRD and DBD of c-Myb. These corepressor complexes block the trans-activating activity of c-Myb. Removal or mutation of the NRD and the mutations found in the DBD of v-Myb abrogate the interaction of c-Myb with these corepressor complexes, leading to increased c-Myb activity. These observations suggest that the oncogenic activation of c-myb due to the truncation of the c-Myb C terminus is caused by a loss of the corepressor-dependent negative regulation of c-Myb.
We have demonstrated that four corepressors directly bind to c-Myb. The non-liganded thyroid hormone receptor ␤ (TR␤) also binds to three corepressors (N-CoR/SMRT, mSin3A, and Ski), and a lack of either of these corepressors significantly decreases the repression capacity of TR␤ (23,24,30). These three corepressors are not the components of the same complex. The purified N-CoR complex contained neither mSin3A nor Ski (28), whereas the purified mSin3A complex involved neither N-CoR nor Ski (27). Therefore, the three different complexes containing either of these three corepressors bind to TR␤. However, it remains unknown whether the binding of these complexes to TR␤ occurs simultaneously or sequentially. Similarly, it is unknown whether the four complexes containing either four corepressors (TIF1␤, Ski, N-CoR, and mSin3A) bind to c-Myb simultaneously or sequentially. Because the mSin3A and N-CoR complexes contain the class I and II HDAC, respectively, the recruitment of these different types of HDACs may be needed for efficient transcriptional repression.
TIF1␤ binds to two sites in NRD, namely, the leucine-rich region and additional C-terminal site between amino acids 444 and 500. These results support our previous study indicating the presence of a putative leucine-zipper structure in the NRD, which interacts with some inhibitors (22). It is noteworthy that this region also contains two ⌽XX⌽⌽ (⌽: hydrophobic amino acids) motifs that were recently demonstrated to be critical for interaction between nuclear hormone receptors and their coactivators (43). Other proteins have also been found to bind to the leucine-rich region, namely, p67 and p160 (44). Recently, Ladendorff et al. (45) found that BS69 also binds to the C-terminal region of c-Myb. All of these proteins inhibit the trans-activating capacity of c-Myb. BS69 was originally identified as an adenovirus E1A-associated protein and was recently demonstrated to interact with N-CoR (46). Thus, other factors may also bind to c-Myb together with the four corepressors.
Although c-Ski directly binds to the DBD, not to the NRD, the interaction between c-Ski and c-Myb is almost completely abolished by either the deletion of the NRD or mutations of the leucine-rich region in the NRD. One possible mechanism is that the deletion or mutation of the NRD alters the conformation of the c-Myb protein, thereby causing the transcriptional activation domain to block the interaction between c-Ski and the DBD. mSin3A and N-CoR also directly bind to the DBD of c-Myb. In contrast to c-Ski, however, the deletion of the NRD does not disrupt their binding to c-Myb. However, both corepressors inhibit the trans-activating capacity of the CT3 mutant that lacks NRD less efficiently than they inhibit wild-type c-Myb. This is probably because all of the four complexes containing each corepressor bind simultaneously or sequentially to c-Myb and to each other. Thus, the loss of binding of any complex might result in the decreased interaction of the other corepressor complexes with the NRD and therefore less repression (Fig. 8).
c-Ski and CBP compete with each other for binding to c-Myb (Fig. 8). These observations raise the important questions, what causes c-Myb to be bound selectively by coactivators or corepressors and when does c-Myb act as transcriptional repressor. It is known that in the case of nuclear hormone receptors, the unliganded receptor binds to corepressors, but the binding of the ligand induces the release of the corepressors and the recruitment of the coactivators. In the case of c-Myb, certain specific signals may induce a similar exchange of the coactivators and corepressors. For example, the phosphorylation of c-Myb may result in such exchange. Another possibility is that the association with coactivators or corepressors depends on the target gene. Recently, it was found that a small difference in the target DNA sequence of the Pit-1 transcription factor leads to a preferential interaction with the corepressor N-CoR complex (47). Thus, the DNA sequence of the c-Mybbinding site could control the selection of coactivators and corepressors. That is, if a target gene had the c-Myb recognition sequence that allows c-Myb to associate only with the corepressor complex, c-Myb would act as a transcriptional repressor for that specific target gene. In addition, other transcription factors that can interact with c-Myb may affect the selection of coactivators and corepressors. For example, it has been recently suggested that the c-Myb complex with HES-1 mediates CD4 silencer activity (48).
The corepressors inhibit the trans-activating capacity of v-Myb encoded by AMV less efficiently than they inhibit wildtype c-Myb. This appears to be, at least partly, due to the lack of the C-proximal TIF1␤-binding site in v-Myb. In addition, three point mutations in the R2 of v-Myb decrease the affinity with the corepressors. In the wild-type c-Myb, these three residues are hydrophobic amino acids on the surface of DBD (42). In v-Myb, these residues are mutated to non-hydrophobic amino acids. It is known that these three wild type amino acids are involved not only in interacting with C/EBP␤ but also with Cyp-40 (49 -51). Thus, these three amino acids are involved in the interaction of c-Myb with various factors, including corepressors.
If the loss of corepressor-mediated negative regulation would result in the oncogenic activation of c-Myb, the corepressors would act as tumor suppressors. Recently, we demonstrated that Ski and its related gene product Sno indeed act as tumor suppressors in mice (52,35). Ski and Sno are already known to negatively regulate cellular proliferation by mediating the transcriptional repression by the two tumor suppressors Mad and Rb (23,53). Thus, the negative regulation of c-Myb could be another mechanism by which Ski and Sno act as tumor suppressors.