The human L(3)MBT polycomb group protein is a transcriptional repressor and interacts physically and functionally with TEL (ETV6).

H-L(3)MBT, the human homolog of the Drosophila lethal(3)malignant brain tumor protein, is a member of the polycomb group (PcG) of proteins, which function as transcriptional regulators in large protein complexes. Homozygous mutations in the l(3)mbt gene cause brain tumors in Drosophila, identifying l(3)mbt as a tumor suppressor gene. The h-l(3)mbt gene maps to chromosome 20q12, within a common deleted region associated with myeloid hematopoietic malignancies. H-L(3)MBT contains three repeats of 100 residues called MBT repeats, whose function is unknown, and a C-terminal alpha-helical structure, the SPM (SCM, PH, MBT domain, which is structurally similar to the SAM (sterile alpha motif) protein-protein interaction domain, found in several ETS transcription factors, including TEL (translocation Ets leukemia). We report that H-L(3)MBT is a transcriptional repressor and that its activity is largely dependent on the presence of a region containing the three MBT repeats. H-L(3)MBT acts as a histone deacetylase-independent transcriptional repressor, based on its lack of sensitivity to trichostatin A. We found that H-L(3)MBT binds in vivo to TEL, and we have mapped the region of interaction to their respective SPM/SAM domains. We show that the ability of TEL to repress TEL-responsive promoters is enhanced by the presence of H-L(3)MBT, an effect dependent on the H-L(3)MBT and the TEL interacting domains. These experiments suggest that histone deacetylase-independent transcriptional repression by TEL depends on the recruitment of PcG proteins. We speculate that the interaction of TEL with H-L(3)MBT can direct a PcG complex to genes repressed by TEL, stabilizing their repressed state.

The human lethal(3)malignant brain tumor (h-l(3)mbt) 1 gene, a recently cloned human homolog of the Drosophila l(3)mbt gene, encodes a member of the polycomb group (PcG) of proteins (1). Drosophila l(3)mbt functions as a tumor suppressor gene that, in the homozygous mutated state, leads to malignant transformation of adult optic neuroblasts and ganglion mother cells in the larval brain (2,3). Because these mutant alleles have never been characterized, the biochemical role of the Drosophila L(3)MBT protein in suppressing tumorigenesis is not known.
The h-l(3)mbt gene is located on chromosome 20q12 (1), within the common deleted region identified in patients with deletions of the long arm of chromosome 20. 20q abnormalities are found in hematological disorders including the myeloproliferative disorders, especially polycythemia vera where it is identified in ϳ10% of the patients (4), the myelodysplastic syndromes, and in a fraction of acute myeloid leukemias cases (5)(6)(7)(8).
H-L(3)MBT is a member of the PcG of chromatin-associated proteins originally defined by their ability to maintain long term repression of the homeotic genes that govern axial patterning during Drosophila embryogenesis (9 -12). More than 15 PcG genes have been characterized to date, and subfamilies can be identified based on the presence of common protein domains. The human L(3)MBT protein contains three MBT repeats, making it a member of a subfamily that includes at least three other genes: h-l(3)mbt-like (13), SCMH1 (14), and SCML2 (15). PcG proteins function in large multiprotein complexes to maintain long term repression of gene expression thereby permitting stable and heritable transmission of gene activity. This function is crucial to maintaining differentiated identity of cells over subsequent generations. The mechanism by which the PcG proteins maintain repression is unknown; however, recent evidence indicates that an important component of this repression occurs via chromatin remodeling. Another set of proteins, the trithorax group, is thought to antagonize PcG-mediated repression and maintain active gene expression. The trithorax group complex include proteins with diverse activities including histone acetyltransferase (16) and ATP-dependent nucleosome remodeling (SWI/SNF) activities (17).
The h-l(3)mbt gene gives rise to three major mRNA species that encode protein isoforms that differ at their C-terminal region and contain respectively, 772, 752, and 738 amino acids. H-L(3)MBT protein contains three MBT repeats, which are 95-105-amino acid motifs of unknown function. Although analysis of the MBT motif has not shown homology with any catalytic domains or sequence-specific DNA binding domains (13), these repeats are highly conserved between the Drosophila and human homologs, suggesting that they have an important function.
The SPM domain present at the H-L(3)MBT C terminus is an ␣-helical structure of ϳ60 amino acids. The SPM domains in the Drosophila Sex Comb on Midleg (SCM) and polyhomeotic (PH) proteins have been found to mediate protein-protein interactions (18). The predicted ␣-helical secondary structure and conservation of hydrophobic residues prompt comparison of the SPM domain with the helix-loop-helix type motifs used for homotypic and heterotypic protein interactions in other transcriptional regulators. The SPM domain belongs to the extended family of SAM (sterile alpha motif) domains (also known as HLH or pointed domains) which are found in several regulatory proteins including kinases, adaptor proteins, and transcription factors (19,20). These more distantly related proteins include several members of the ETS family of transcription factors (for review, see Ref. 21), such as TEL. The tel (translocation Ets leukemia) gene was identified and located at the chromosomal breakpoints of several leukemia-associated translocations (22)(23)(24)(25). The fusion proteins derived from these translocations contain the N-terminal portion of TEL, including the SAM domain, fused to a variety of tyrosine kinase domains, such as the platelet-derived growth factor receptor (PDGFR) ␤, ABL, Janus kinase (JAK) 2, and NTRK3, or to transcription factors such as AML1 (22,26,27). Dimerization of the TEL-SAM domain results in the constitutive activation of the TEL-ABL (25), TEL-PDGFR (28), and TEL-JAK2 (29) tyrosine kinase activity and is essential for transcriptional repression by TEL-AML1 (30 -32).
Even though the PcG-SPM and the TEL-SAM domains share very little sequence identity, the polyhomeotic SPM domain has been recently reported to form a helical polymer structure similar to the one formed by TEL-SAM, with both domains displaying closely related structural architecture (33). This homology prompted us to screen several SAM-containing ETS proteins for their ability to interact with H-L(3)MBT.
We report that H-L(3)MBT is an HDAC-independent transcriptional repressor, and we have characterized its in vivo and in vitro physical association with TEL; this interaction requires their respective SPM and SAM domains. Using both the stromelysin-1 (matrix metalloproteinase-3) promoter, which is physiologically regulated by TEL, and an artificial TEL-regulatable promoter, we show that H-L(3)MBT is targeted to these promoters through its interaction with TEL, enhancing the repressive effect of TEL on transcription. We believe that this is the first report of a functional interaction between a PcG protein and a mammalian DNA sequence-specific binding transcription factor. Given the pivotal role that ETS factors (and TEL in particular) play in human malignancies, the functional interaction between H-L(3)MBT and TEL provides a clue as to the potential role of PcG proteins such as L(3)MBT in cancer.

MATERIALS AND METHODS
Plasmid Construction-To generate a full-length h-l(3)mbt cDNA clone, total RNA from placenta (Research Genetics) was reverse transcribed using a GeneAmp RNA PCR Core kit (Roche Diagnostics). cDNA products were PCR-amplified with the Expand High Fidelity PCR System (Roche Diagnostics) in 5% dimethyl sulfoxide. using forward (5Ј-GAGTCTTGAGGCCTGCTGAG-3Ј) and reverse (5Ј-GGAT-CCTTGCAGTAAATCACAA-3Ј) primers for the entire open reading frame of the type I h-l(3)mbt transcript (1). Full-length h-l(3)mbt cDNA was FLAG-or HA-tagged at the 5Ј-end using PCR methodology.
To generate an h-l(3)mbt cDNA lacking the SPM domain, the pCR2.1-FLAG/HA-h-l(3)mbt construct was digested with StuI and SspI to liberate a 525-bp fragment containing the SPM domain sequences. Blunt end ligation of the larger fragment generated an h-l(3)mbt clone with an intact reading frame, lacking the SPM domain. To generate an h-l(3)mbt cDNA lacking the MBT repeats, the splicing by overlap extension (SOEing) method was used (34). The SOEing process involves amplification of two products (P1/P2 and P3/P4) in which a tail is added to one product (P1/P2) that is complementary to the end of the second sequence. The two fragments are then able to anneal by virtue of the complementary terminal tails and can self-prime with subsequent amplification of the full-length product by flanking primers (P1 and P4). Hence a template pCR2.1-FLAG-h-l(3)mbt construct DNA was PCR amplified using primer P1 (5Ј-TCGTTCGACCCCGCCTCGAT-3Ј) and P2 (5Ј-TGAGAGGGGGACAGCCCCCAGCAGGCTGGCTGCTGCTCCA-3Ј) to generate a P1/P2 fragment, and P3 (5Ј-TGGGGGCTGTCCCCC-TCTCA-3Ј, which is complementary to the end of P2) and P4 (5Ј-TCC-AAGCGGCTTCGGCCAGT-3Ј) to generate a P3/P4 fragment. Equimolar amounts of P1/P2 and P3/P4 were mixed and PCR amplified using P1 and P4. The resulting P1/P4 fragment encodes an in-frame H-L(3)MBT polypeptide lacking 330 amino acids corresponding to the MBT repeats. The GAL4-DBD-H-L(3)MBT construct was generated by cloning the h-l(3)mbt cDNA in-frame into pFA vector (Stratagene). For creation of the MBT mutant version of this GAL4 fusion construct, SOEing was applied using the P2 and P3 primers mentioned above, with P1 (5Ј-TGCGACATCATCATCGGAAGAGAG-3Ј) and P4 (5Ј-GGG-CCTGAAATGAGCCTTGG-3Ј) amplifying within the pFA vector. For the GAL4-H-L(3)MBT-C2HC deletion mutant 5Ј-CTTCTTGCGGGCT-GAGGCCCCAGGGGAGGCAGAGC-3Ј was used for P2, and 5Ј-GCCT-CAGCCCGCAAGAAGA-3Ј was used for P3. For the SPM mutant the pFA-h-l(3)mbt DNA was digested with BamH1 and StuI, and the resulting fragment lacking SPM domain sequences was subcloned into the BamHI and SmaI sites of pFA.
Cell Culture, Transient Transfection, and Luciferase Assay-NIH3T3, U2OS, and 293 cells were cultured in Iscove's modified Dulbecco's medium with 10% fetal bovine serum, 2 mM glutamine, and antibiotics and were trypsinized and replated to the needed density the day before transfection. NIH3T3 cells were transfected with Lipo-fectAMINE Plus reagent (Invitrogen); 293 cells were transfected using a calcium phosphate DNA precipitation technique (ProFection, Promega, Madison, WI). Unless otherwise specified, 1 g of expression construct (or empty vector), 2 g of the reporter construct, and 0.2 g of the cytomegalovirus-driven GFP expression plasmid (pEGFP-C3) control plasmid (Stratagene) were used for each transfection; the empty pcDNA3 expression vector was added, as needed, to keep the amount of transfected DNA the same for each sample. For the luciferase assays, cells were collected 48 h after transfection and lysed; luciferase activity was assessed using Luciferase Assay Substrate (Promega) on a LUMAT LB9501 (Berthold) luminometer. Luciferase activity was normalized according to the percent GFP-positive cells (as assessed by flow cytometry) and the total protein concentration. To rule out general effects of the transfected proteins on the cytomegalovirus promoter the mean fluorescence of the GFP peaks was determined for the samples and found to be comparable.
Immunoprecipitation Assays-After collection, cells were washed with PBS once, lysed by brief sonication in NET-N lysis buffer (1% Nonidet P-40, 150 mM NaCl, 50 mM Tris pH 8, 1 mM EDTA) in the presence of proteinase inhibitors (10 g/ml leupeptin, 10 g/ml aprotinin, 0.5 mM phenylmethylsulfonyl fluoride) and phosphatase inhibitors (10 mM NaF, 20 mM ␤-glycerophosphate), and then clarified by centrifugation. The protein concentration was adjusted to 1 mg/ml, and the lysates were incubated with 2 g of specific antibodies for 1 h at 4°C. Protein AϩG-agarose beads (Santa Cruz) were added to the extracts and mixed for at least 1 h at 4°C. The immune complexes were washed four times with NET-N lysis buffer, once with high salt buffer (1 M NaCl, 20 mM Tris-HCl, pH 8), and once with low salt buffer (10 mM MgCl 2 , 50 mM HEPES-KOH, pH 7.5); they were released by adding SDS sample buffer and boiled for 4 min. Samples were subjected to SDS-PAGE with 7.5% gel, transferred to a polyvinylidene difluoride membrane, and Western blotted with the following antibodies: FLAG-horseradish peroxidase-conjugated (M2, Sigma), HA mouse monoclonal antibody (12CA5, Roche Molecular Biochemicals), GAL4-DBD (RK5C1, Santa Cruz), anti-TEL (gift of P. Marynen), and anti-H-L(3)MBT (MBT1) generated in our laboratory (see below). In each experiment, the membranes were stripped and reprobed with the appropriate antibody to assess the correct expression of the cotransfected expression plasmids.
Nuclear extracts were prepared in 0.4 M NaCl, according to the protocol of Dignam and co-workers (35), and the immunoprecipitates from nuclear extracts were washed with a buffer containing 0.4 M NaCl.
Expression of GST Fusion Proteins in Bacteria and GST Pulldown Assays-The GST fusion protein plasmids were transformed into the Escherichia coli BL21 strain (Novagen, Inc., Madison, WI). After overnight culture, protein expression was induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside for 2-4 h. The bacterial pellets were then lysed in 1 ml of PBS-T buffer (140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.3, 1% Triton X-100), with the aid of sonication. 5 g of GST-TEL fusion protein (or GST) was incubated with 20 l of 50% slurry glutathione-Sepharose 4B beads (Amersham Biosciences) in a total volume of 1 ml of PBS-T for 45 min at 4°C. The beads were washed three times with 1 ml PBS-T and then incubated with 2 l of in vitro transcribed and translated 35 S-labeled protein in 1 ml of NET-N buffer for 2-4 h at 4°C. The beads were then washed with 1 ml of NET-N six times. The bound proteins were released from the beads by boiling in SDS-gel loading buffer for 4 min. Proteins were analyzed by SDS-PAGE with 7.5% gel and autoradiography. The TNTcoupled reticulocyte lysate system (Promega) was used to generate in vitro translated proteins, following the procedures specified by the manufacturer.

Preparation of Anti-H-L(3)MBT Antisera and Coupling to Protein A Gel-Anti-H-L(3)
MBT antisera were generated by immunizing rabbits with either a 174-amino acid fragment encoding the second half of the MBT repeats (MBT1) or a protein fragment containing the three MBT repeats (MBT2). Each fragment was bacterially expressed as a GST fusion protein using pGEX-4T expression plasmid (Amersham Biosciences) and purified using glutathione-Sepharose 4B beads. The antisera were affinity purified using N-hydroxysuccinimide-activated Sepharose columns (Amersham Biosciences) coupled to GST alone (first passage for negative selection) and then GST-MBT-coupled beads (second passage for positive selection). The affinity-purified MBT2 antibody (suitable for immunoprecipitation) was then coupled to recombinant protein A gel, using the Immunopure rProtein A IgG Plus Orientation kit from Pierce.  (Fig. 1b, 10th lane), whereas removal of the three MBT repeats had no effect on homodimerization (Fig. 1c, 4th lane). The protein membrane was stripped and probed for HA antibody to confirm the presence of equal amount of the HA-tagged proteins (not shown). We conclude that H-L(3)MBT self-associates in vivo and that self-association requires the SPM domain.

H-L(3)MBT Has General Transcriptional
Repressor Activity-PcG proteins maintain a chromatin-repressed state, but unlike classic transcription factors they do not recognize specific DNA target sequences. Thus far, the only PcG protein shown to have sequence-specific DNA binding activity is pleiohomeotic, the Drosophila homolog of mammalian YY1 (36). Although PcG proteins act on chromatin, their repressive function can also be detected in transcriptional assays using transiently transfected reporter gene plasmids (which are believed to form less organized chromatin-like structures) (37).
To evaluate the transcriptional regulatory activity associated with H-L(3)MBT, we fused it to the DNA-binding domain (DBD, amino acids 1-147) of GAL4. The GAL4-H-L(3)MBT construct was tested using a luciferase reporter plasmid that contains four copies of the GAL4 DNA recognition sequence positioned immediately upstream of a thymidine kinase (tk80) minimal promoter (diagrammed in Fig. 2a, upper panel). In-  Fig. 3a) were tested for expression by Western blot analysis (Fig. 3c) and were shown to localize in the cell nucleus by immunofluorescence staining (data not shown). Mutants lacking either the zinc finger or the SPM domain retained the ability to repress the GAL4-reporter gene (Fig. 3b). In contrast, the mutant missing the three MBT repeats had a minimal capacity to repress the GAL4-TK-LUC reporter (30% repression compared with ϳ90% repression for the wild type protein). Thus, the transrepressive ability of H-L(3)MBT requires the presence of the MBT repeats for full activity, whereas the zinc finger and SPM domains appear to be dispensable for this function. Repression by H-L(3)MBT-⌬-SPM suggests that self-dimerization is not necessary for trans-repression; however, a role for protein dimerization cannot be completely excluded because the GAL4-DBD itself can direct the oligomerization of GAL4 and GAL4derived proteins.

H-L(3)MBT and TEL Physically Interact through Their SPM/SAM Domains-The SPM domain contained in several
PcG proteins shares structural homology with the SAM domain found in a subset of members of the ETS family of transcriptional factors (e.g. TEL) (33,38,39). To determine whether H-L(3)MBT and TEL interact in vivo we performed immunoprecipitation assays, using a rabbit polyclonal antibody (MBT2) to precipitate H-L(3)MBT followed by Western blotting for TEL. To avoid interference from the heavy chain of the antibody used to pull down H-L(3)MBT, it was necessary to crosslink covalently both the H-L(3)MBT antibody and the normal rabbit polyclonal IgG to protein A-agarose gel (see "Materials and Methods"). Using the osteosarcoma-derived U2OS cells, which express readily detectable levels of both proteins, we showed that the MBT2 anti-H-L(3)MBT antibody, but not the preimmune antiserum, could coprecipitate H-L(3)MBT and TEL (Fig. 4). The intensity of the bands in the immunoblot suggests that more TEL protein than H-L(3)MBT protein is present in the immunoprecipitate, however, potential differences in the affinities of the two antisera may account for some of the differences seen.   3, 7, and 9). Thus, these domains are involved in the physical association of TEL and H-L(3)MBT.

Functional Interaction between H-L(3)MBT and TEL
Recently, TEL2 (or TELB), a new member of the ETS family of transcription factors which is very similar to TEL, was cloned (39,40). TEL2 is predicted to be structurally similar to TEL and has been shown to form homodimers as well as heterodimers with TEL (41). We tested whether H-L(3)MBT could also interact with TEL2 and found that HA-TEL2 coimmunoprecipitates with FLAG-H-L(3)MBT when the two proteins are coexpressed (Fig. 5b).
To further study the interaction between TEL and H-L(3)MBT, we used GST-based in vitro protein binding assays. 35 S-Radiolabeled H-L(3)MBT (and deletion-mutant proteins) were produced by in vitro transcription/translation (IVT) and tested for their ability to bind to a bacterially expressed GST-TEL fusion protein. The self-association of TEL with GST-TEL was used as a positive control. These studies determined that GST-TEL specifically binds H-L(3)MBT but not H-L(3)MBT-⌬-SPM (Fig. 5c) Table I, where the interactions between GST-TEL and the IVT proteins are assigned a ϩ, ϩ/Ϫ, or a Ϫ based on the relative amounts of input IVT protein detected after the pulldown.

. The interaction of TEL with H-L(3)MBT-⌬-MBT and with H-L(3)MBT-⌬-C2HC appears to be much less efficient than its interaction with the intact H-L(3)MBT. This is shown in
TEL and FLI-1 are known to bind to each other via their SAM domains, but they do not interact with ETS-1 (42). We investigated whether in vitro translated H-L(3)MBT can bind to either of these SAM domain-containing ETS proteins. As shown in Fig. 5d, neither GST-FLI-1 nor GST-ETS-1 binds H-L(3)MBT in this assay. The association of in vitro translated TEL with GST-FLI-1 served as the positive control.

Cooperation of H-L(3)MBT and TEL in the Repression of Transcriptional Targets-To assess the functional impact of the physical interaction between H-L(3)MBT and TEL on tran-
scriptional regulation, we generated a reporter gene construct whose luciferase expression is regulated by a minimal tk80 promoter element, and three concatemerized ETS binding sites (43). This artificial promoter/enhancer sequence has been shown to drive high levels of luciferase activity in all of the cell lines tested (compared with the control tk80Luc reporter plasmid), probably reflecting the binding of endogenous transactivating ETS proteins (44). Transfection of the pcDNA3-H-L(3)MBT construct had no effect on the activity of the 3xEBStk80Luc reporter plasmid, whereas pcDNA3-TEL somewhat repressed its activity. However, the combination of H-L(3)MBT and TEL significantly repressed luciferase expression (Fig. 6a), indicating that TEL binding its consensus binding site can recruit H-L(3)MBT to a promoter or enhancer regulatory region. Although previous studies have shown that the Drosophila PcG proteins can interact with the basic transcriptional apparatus (e.g. transcription factor IID) (45), we did not find any appreciable effect of H-L(3)MBT on the activity of either an SV40 promoter or a herpes simplex virus TK promot-  lanes 1 and 5) and GST-FLI-1 (lanes 2 and 6) pulldown assays. In vitro translated TEL and H-L(3)MBT were used for lanes 1-4 and 5-8, respectively. [ 35 S]Methionine-radiolabeled in vitro translated proteins were incubated with 5 g of bacterially produced GST or GST fusion proteins immobilized on GSH-Sepharose. After incubation and washing, the bound proteins were separated by SDS-PAGE (7.5% gel), fixed, and stained with Coomassie to show the presence and purity of the purified GST fusion proteins. Input represents 5% of the total protein used in the pulldown assay.

Functional Interaction between H-L(3)MBT and TEL
er-driven reporter construct (Fig. 6a). It was shown recently that TEL can repress transcription of stromelysin-1 (matrix metalloproteinase-3) gene by direct binding to its promoter (46). To study the functional relevance of the H-L(3)MBT-TEL interaction on a natural target of TEL, we cotransfected pcDNA3-h-l(3)mbt, pcDNA3-TEL, and pcDNA3-TEL-⌬-SAM alone, or in combination, with a rat stromelysin-1 promoter-regulated luciferase reporter gene plasmid (shown schematically in Fig. 6b). This promoter contains 754 bp of 5Ј-flanking sequence and at least two TEL binding sites (46); overexpression of TEL represses the stromelysin-1 promoter 4 -5-fold, whereas the TEL-⌬-SAM mutant has no effect on promoter activity (Fig. 6b). H-L(3)MBT, itself, was able to repress transcription from the stromelysin-1 promoter 3-fold (Fig. 6b), and coexpression of TEL with H-L(3)MBT caused further repression of the stromelysin-1 reporter (8.5-fold), suggesting an additive rather than synergistic activity on this promoter (Fig. 6b). No additive repression was seen when TEL-⌬-SAM was coexpressed with H-L(3)MBT (Fig. 6b). Similarly, when H-L(3)MBT mutants lacking either the SPM domain or the MBT repeat region were coexpressed with TEL, they had no effect on repression (Fig. 6c). Thus, the combined effects of TEL and H-L(3)MBT on the stromelysin-1 promoter require the presence of their respective SAM and SPM domains, and the MBT repeats (the region required for repression by the GAL4-H-L(3)MBT fusion protein (see Fig. 3b)). DISCUSSION Koga et al. (1) and our group (6) have previously isolated the h-l(3)mbt gene, which we showed is located within the commonly deleted region of 20q-seen in patients with myeloid hematologic malignancies. We now report that H-L(3)MBT is a transcriptional repressor and has the ability to direct repression of specific promoters by binding to the DNA sequencespecific transcription factor TEL. We show that two protein motifs in H-L(3)MBT, the MBT repeats and the SPM domain, which are conserved in other PcG family members, are required for its repression and protein binding activities.
The Trans-repressing Ability of H-L(3)MBT Requires Mainly the Presence of the MBT Repeats-Removal of the MBT repeats from H-L(3)MBT eliminated much of its trans-repressing ability, demonstrating the role for these repeats in establishing repression, likely through protein-protein contacts. Future studies will determine whether all three MBT repeats are required for H-L(3)MBT function; however, the known MBTcontaining PcG proteins possess two, three, or four MBT repeats, suggesting that the functional unit may require a minimum of two repeats. Little is known about the biological FIG. 6. Coexpression of H-L(3)MBT potentiates the repression by TEL of the stromelysin-1 and 3xETS BS-TK promoter activities. a, NIH3T3 cells were cotransfected with the pcDNA3 constructs indicated at the bottom of the graph and either the SV40 luciferase or the three ETS binding sites-tk80Luc construct (diagrammed above the graphs). Cells were collected at 48 h after transfection; luciferase activity was normalized to the percentage of GFP-positive cells as assessed by flow cytometry and the total protein concentration. The -fold repression was calculated as the -fold decrease in luciferase activity compared with the empty vector. Values shown are the average Ϯ S.E. of two experiments performed in triplicate. b and c, NIH3T3 cells were cotransfected with the pcDNA3 constructs indicated on the bottom of the graphs and the diagrammed rat stromelysin-1 promoter sequence (1-754 bp) luciferase construct, which contains at least two TEL binding sites (TBS), (shown at the top). Cells were collected 48 h after transfection; luciferase activity was normalized to a percentage of GFP-positive cells as assessed by flow cytometry and total protein concentration. The -fold repression was calculated as the -fold decrease in luciferase activity compared with the empty vector. Values shown are the average Ϯ S.E. of two experiments performed in triplicate. relevance of the MBT repeats; however, these repeats appear to be indispensable to the function of the SCM protein in Drosophila (47). Three hypomorphic SCM alleles, which are mutated in the MBT repeats, interact genetically with PcG mutations more strongly than SCM null alleles, and the strongest interactions produce partial syntenic lethality (47). These SCM mutant proteins can still associate with polytene chromosomes, suggesting that although the MBT repeats have a critical function in the PcG complex, they may not be essential for protein localization.
The In addition to binding H-L(3)MBT, the TEL-SAM domain is involved in several other interactions, such as the binding of TEL to TEL2 (39) or to FLI-1 (42). TEL is conjugated to SUMO-1 after interacting with UBC9, and this modification requires the SAM domain and largely involves a specific lysine residue (Lys-99) in the TEL-SAM domain (49,50). Finally, the TEL-SAM domain is required to target TEL to specific subnuclear structures called TEL bodies (50).  (44), suggesting that dimerization but not corepressor binding to the SAM domain is essential. Yet TEL point mutants that fail to dimerize can retain their ability to bind mSin3A and maintain partial trans-repressor activity (51). An explanation for these apparently conflicting results could be that the SAM domain constitutes an interface for numerous protein contacts, including the H-L(3)MBT polycomb protein, which provide TEL with several mechanisms of repression. Transcriptional repression by TEL has been shown to involve at least two distinct regions and mechanisms of action. Repression by the central region of TEL involves the recruitment of a complex that includes N-CoR, SMRT, and mSin3A (which is also contacted by the SAM domain (51)); these corepressors recruit HDAC activity to DNA-binding proteins to actively repress transcription (52). The SAM domain of TEL has been shown to repress gene transcription through a different mechanism, which appears to be HDAC-independent (53).
We have shown that H-L(3)MBT functions as an HDAC-independent repressor and interacts with the TEL-SAM domain, which supports a model in which TEL recruits components of two separate repressor complexes through different regions of the protein. In an analogous manner, the retinoblastoma protein (pRB) represses cyclin E expression via an HDACdependent mechanism but utilizes PcG complexes for the HDAC-independent, long term silencing of cyclin A and cdc2 expression, which leads to arrest in G 2 (54). In fact, this G 2 block is maintained even if p16 is depleted and pRB becomes hyperphosphorylated, which suggests that PcG complexes help mediate irreversible repression by pRB. Short term inhibition of transcription might therefore be achieved through the recruitment of corepressors and HDACs to the promoter. Later, long term silencing might be established by an H-L(3)MBTcontaining PcG protein complex that can maintain repression and would make HDACs available to other target genes in the cell. Additional work aimed at identifying other components of this complex is under way, which will help address the respective contributions of HDAC-dependent and -independent transcriptional repression to the regulation of TEL target genes.
PcG Protein Recruitment to DNA through Canonical DNAbinding Factors-With the exception of YY1 (36), none of the PcG proteins has been shown to bind DNA in a sequencespecific fashion (55). Notwithstanding, PcG complexes are found in stable association with chromatin on polytene chromosome in Drosophila (18) and with core promoter regions in chromatin immunoprecipitation assays (45). Although formation of a multiprotein complex could generate a DNA binding activity, more likely, PcG complex formation at DNA target sites is dependent on a mosaic of interactions of different PcG proteins with multiple DNA-binding proteins that act as recruiters (55). The physical and functional interaction of H-L(3)MBT with TEL presents some analogy to the interaction of PcG proteins in the Drosophila polycomb repressive complex 1 (PRC1) with the Zeste protein (56). Zeste is a sequencespecific DNA-binding factor, which has consensus binding sites in the promoter and regulatory regions of several homeotic genes (57); it has been proposed that the interaction of Zeste with PRC1 aids in the targeting of PcG proteins to repressed gene loci. TEL is a sequence-specific DNA-binding protein that is directly involved in establishing the repression of specific target genes. By binding both H-L(3)MBT and specific promoter sequences, TEL may bridge the PcG complex with specific regulatory elements.
Significance of H-L(3)MBT/TEL Interaction-Silencing by PcG proteins seems to depend on the state of transcriptional activity of the target gene because PcG complexes generally require a silenced gene as template (58). When the target is transcriptionally active, silencing is not established (59); whereas once established, the repressed state persists throughout embryonic development. Thus, the PcG complex has been proposed to function as a memory system that can stabilize previous regulatory events by "locking" them in. However, this model may be too simplistic, given that the vertebrate EED⅐EZH2 complex possesses HDAC activity (60), which could be involved in initiating gene repression.
We found that substituting a thymidine kinase minimal promoter with the much stronger SV40 promoter in the 3xEB-StkLuc construct led to a loss of repression by TEL (and, as a consequence, by TEL plus H-L(3)MBT) (data not shown). This is consistent with the principles of gene silencing observed in Drosophila, where high doses of trans-activators can antagonize PcG-mediated silencing (61).
The molecular mechanisms through which PcG proteins achieve and maintain repression are largely unknown. Purification of two different PcG protein complexes has identified proteins that could interact functionally with PcG proteins to repress gene expression, such as HDAC2, which is found in the EED⅐EZH2 complex (60) and the corepressor molecules SMRT and Sin3A, which are found in the Drosophila PRC1 (56). Although repression mediated through the EED⅐EZH2 complex is relieved by trichostatin A (a potent HDAC inhibitor), the PRC1 complex blocks chromatin remodeling by the trithorax group-related SWI⅐SNF complex in vitro through a mechanism that appear to be HDAC-independent (62). Trichostatin A is unable to relieve the repression brought about by H-L(3)MBT; this suggests that the mechanism of repression of H-L(3)MBT may be more similar to that of the PRC1 complex.
PcG Genes in Hematopoiesis-Several lines of evidence support a critical role for PcG proteins in regulating the early and late stages of hematopoiesis. Mice lacking bmi-1 display a progressive marrow hypoplasia similar to aplastic anemia and have a hypoplastic spleen and thymus (63). On the contrary, eed ϩ/Ϫ mice have a propensity to develop both myeloproliferative and lymphoproliferative diseases later in life (64). Mice lacking mel-18 have a severe combined immunodeficiency, and mel-18 Ϫ/Ϫ lymphocyte precursors respond poorly to interleukin-7 stimulation (65). PcG genes regulate Hox gene expression, and there are numerous examples where dysregulated Hox gene expression profoundly perturbs hematopoiesis or leads to acute myeloid leukemia, underscoring the importance of maintaining the timely and tight down-regulation of homeobox gene expression.
The stage-specific expression of PcG genes in the human bone marrow compartment suggests that the regulation of homeobox, cell cycle, and other critical regulatory genes by PcG proteins is important in hematopoietic processes (66). H-L(3)MBT is expressed in CD34 ϩ blood progenitor cells, 2 and TEL is absolutely required for the establishment of definitive hematopoiesis in the murine bone marrow (67). The contemporaneous expression of TEL and H-L(3)MBT in the hematopoietic stem cell compartment may signify that their physical and functional interactions play a critical role in the transcriptional regulation of the commitment and differentiation processes.