The ETO Protein Disrupted in t(8;21)-Associated Acute Myeloid Leukemia Is a Corepressor for the Promyelocytic Leukemia Zinc Finger Protein

Abstract

The ETO protein was originally identified by its fusion to the AML-1 transcription factor in translocation (8;21) associated with the M2 form of acute myeloid leukemia (AML). The resulting AML-1–ETO fusion is an aberrant transcriptional regulator due to the ability of ETO, which does not bind DNA itself, to recruit the transcriptional corepressors N-CoR, SMRT, and Sin3A and histone deacetylases. The promyelocytic leukemia zinc finger (PLZF) protein is a sequence-specific DNA-binding transcriptional factor fused to retinoic acid receptor α in acute promyelocytic leukemia associated with the (11;17)(q23;q21) translocation. PLZF also mediates transcriptional repression through the actions of corepressors and histone deacetylases. We found that ETO is one of the corepressors recruited by PLZF. The PLZF and ETO proteins associate in vivo and in vitro, and ETO can potentiate transcriptional repression by PLZF. The N-terminal portion of ETO forms complexes with PLZF, while the C-terminal region, which was shown to bind to N-CoR and SMRT, is required for the ability of ETO to augment transcriptional repression by PLZF. The second repression domain (RD2) of PLZF, not the POZ/BTB domain, is necessary to bind to ETO. Corepression by ETO was completely abrogated by histone deacetylase inhibitors. This identifies ETO as a cofactor for a sequence-specific transcription factor and indicates that, like other corepressors, it functions through the action of histone deactylase.

Myeloid and hematopoietic cell development is a complex process regulated by an extensive network of transcription factors (reviewed in references 58 and 61). These proteins coordinate the sequential expression of gene products which results in progressive stages of progenitor cell commitment and differentiation (14, 57, 59). In hematological malignancies, transcription factors are often disrupted by chromosomal translocations and fused to genes encoding other transcriptional regulators (42, 51, 52). The resulting aberrant factors are oncoproteins that yield altered transcriptional patterns leading to the development of leukemia (54, 61).

One such event disrupts ETO (for eight-Twenty One), a protein identified as part of a fusion product resulting from the translocation (8;21) found in 50% of patients with the M2 variant of acute myelogenous leukemia (AML) (see reference 48 and references within). Translocation (8;21) fuses ETO to AML-1, a critical regulator of hematopoiesis (36) that activates a number of myeloid genes, including those coding for granulocyte/macrophage–colony-stimulating factor (CSF), macrophage-CSF, and myeloperoxidase (61) through recruitment of the CREB binding protein (CBP) or p300 and other histone acetyl transferases to the promoters of these genes (31). In contrast, the AML-1–ETO oncoprotein is a dominant-negative form of AML-1 which represses the promoters of genes normally activated by AML-1 (16, 17, 44, 46). This model is highly supported by the similar phenotypes of AML-1 knockout mice and heterozygous AML-1/ETO knockin mice (49, 66), which include a severe block in hematopoiesis at the fetal liver stage and fatal hemorrhages within the central nervous system. At the molecular level, the dominant-negative effect of AML-1–ETO is due to the ability of the ETO moiety of the fusion protein to associate with the corepressors N-CoR, SMRT, and Sin3A, as well as histone deacetylases 1 and 2 (HDAC1 and -2) (17, 44, 62). Despite its ability to interact with other corepressors and HDAC, ETO itself was not previously identified as a corepressor for any sequence-specific transcription factor.

The promyelocytic leukemia zinc finger (PLZF) protein is fused to the retinoic acid receptor α (RARα) in the retinoic acid-resistant t(11;17)(q23;q21) variant of acute promyelocytic leukemia (APL) (6, 19, 38). As in the case of t(8;21), this translocation yields an aberrant transcription factor. While RARα activates key genes required for normal myelopoiesis, PLZF-RARα represses expression of such genes in a dominant-negative manner (7, 9, 40, 45). We showed that PLZF was a sequence-specific DNA binding transcriptional repressor (2, 37, 67). This is due to the ability of the PLZF moiety to attract corepressor molecules, such as N-CoR, Sin3A, and SMRT, as well as HDAC1 (8, 20, 22, 25, 41). This interaction is, at least in part, mediated through the N-terminal POZ/BTB (poxvirus and zinc finger/Broad Complex, tramtrack, Bric a Brac) domain of PLZF (25) and indicates that PLZF may repress transcription by altering chromatin conformation. In its basal state, RARα also recruits N-CoR, SMRT, and HDACs to its target promoters, thus keeping them repressed in the absence of ligand. In the presence of the ligand all-trans retinoic acid (ATRA), corepressors are released and coactivators are recruited, resulting in transactivation of RARα target genes (5, 23, 26). However, in APL, the association of the PLZF portion of PLZF/RARα with corepressors and HDACs prohibits activation of RARα targets, even in the presence of high doses of ATRA (18, 20).

PLZF is expressed in CD34⁺ myeloid progenitor cells and is down-regulated during differentiation of myeloid cell lines (53). In addition, PLZF causes growth suppression, differentiation blocking and cell cycle delay and/or arrest in myeloid cell lines (56, 67). These findings suggest that the transcriptional repression mediated by PLZF needs to be switched off for cells to differentiate and proliferate. The identity of potential PLZF targets, including cyclin A and the interleukin 3 receptor alpha (IL-3R α) chain, supports this hypothesis (45). Like PLZF, ETO is expressed in CD34⁺ cells and several leukemic cell lines (10, 12) and is down-regulated as hematopoietic progenitors mature (12). However, in contrast to PLZF, ETO was not found to bind to a specific DNA sequence (11, 35, 44). The facts that ETO can function as a powerful transcriptional repressor when fused to AML-1 and that it associates with corepressors suggest that it may normally act as a cofactor for certain sequence-specific DNA binding repressors. In light of this notion, we demonstrated that ETO interacts with PLZF in vivo and in vitro through specific domains of the two proteins. Coexpression of ETO with PLZF augments the ability of PLZF to repress transcription through its cognate binding site. This effect was abrogated by sodium butyrate and trichostatin A (TSA), inhibitors of HDACs. Our results suggest that, like the N-CoR and SMRT corepressors, ETO amplifies the transcriptional effects of PLZF by enhancing recruitment of HDACs to target promoters.

MATERIALS AND METHODS

DNA constructs.

PLZF and PLZF deletion mutants were expressed in mammalian cells by utilizing either the pSG5 (Stratagene, La Jolla, Calif.) or pCDNA3.1+myc/his (Invitrogen, Carlsbad, Calif.) expression vectors. PLZF Δ^POZ/BTB lacks the first 120 N-terminal amino acids of PLZF and was described previously (9). The PLZFΔ^RD2 deletion mutant was created by cutting pSG5 and pCDNA-PLZF with XcmI endonuclease (New England Biolabs, Beverly, Mass.), thus deleting sequences between positions 596 and 939 of the PLZF cDNA. The resulting large fragment of PLZF was religated in the presence of an excess of the linker oligonucleotides 5′ GAGAGTGCCGAGCAGGTGCCACCCCCAGCT 3′ and 3′ CTCTCACGGCTCGTCCACGGTGGGGGTCGA 5′. This resulted in a PLZF sequence lacking amino acids 199 to 313. A plasmid containing the full PLZF POZ/BTB domain (amino acids 1 to 137) was generated by PCR (28) with an N-terminal primer containing a BamHI site (5′ CGCGGATCCGTATGGATCTGACAAAAATG 3′) and a C-terminal primer containing an SfiI site and an XbaI site (5′ TCACTCTAGAGCGGCCATGGTGGCCTCCGTGTCATT 3′). The resulting PCR products were ligated into the PCR II vector and then subcloned into the pBXG1 GAL4(1–147) mammalian expression vector (39) and the pAS2-1 and pACTII yeast two-hybrid vectors (Clontech, Palo Alto, Calif.). All constructs used in this study were confirmed by automated sequencing (Utah State University Biotechnology Center, Logan). The GAL4-PLZF^1–400 and GAL4-RD2 fusion construct and expression vectors for ETO deletion mutants were described previously (37, 43). Expression of all of these proteins was confirmed by immunoblotting as described below.

Immunoprecipitation and immunoblotting.

To detect protein-protein interactions between full-length overexpressed PLZF and ETO (see Fig. 1 and 5), COS-7 cells were transfected, lysed, and subjected to immunoprecipitation with anti-ETO or anti-PLZF antibodies as described previously (44). To detect interaction between ETO and PLZF Δ^POZ/BTB or PLZFΔ^RD2, 293T cells were plated at approximately 6 × 10⁵ cells per well in a six-well dish. After 24 h, each well was transfected with 0.75 μg of expression constructs and 0.75 μg of carrier plasmid DNA by using Superfect (Qiagen, Valencia, Calif.). After 48 h, the cells were harvested for immunoprecipitation and lysed in 0.5% Triton buffer for immunoprecipitation (21). ETO Ab-1 antibody (Oncogene Research, Cambridge, Mass.) was covalently linked to protein A-agarose beads at a ratio of 1 mg/ml. After 1 h of incubation, the beads were washed twice with 0.2 M sodium borate (pH 9.0) and once with 0.2 M triethanolamine (pH 8.5) (Sigma, St. Louis, Mo.). The beads were incubated for 1 h with 40 mM dimethylpimelimidate (pH 8.5) (Sigma), followed by one wash with triethanolamine and two with 0.2 M sodium borate (pH 8.2). The beads were next blocked with mouse immunoglobulin G (Dako, Carpinteria, Calif.) for 1 h, washed, and then added to lysates of transfected 293T cells for 3 h at 4°C. The precipitated proteins were released by boiling in sodium dodecyl sulfate (SDS) loading buffer, separated by electrophoresis through an SDS–12% polyacrylamide gel, and transferred to an Immobilon P membrane (Millipore, Bedford, Mass.). Immunoblotting was performed with a 1-μg/ml concentration of monoclonal PLZF antibody (40) or a 2.5-μg/ml concentration ETO AB-1 antibody and a 1:7,500 concentration of horseradish peroxidase-conjugated antimouse or antirabbit secondary antibody (Roche Molecular Biochemicals, Indianapolis, Ind.) and developed with the ECL enhanced chemiluminescence system (Amersham Pharmacia, Buckinghamshire, United Kingdom). Lysates from the same cells were immunoblotted with PLZF or ETO Ab-1 antibodies as described above to confirm expression of ETO and PLZF deletion mutants. To detect interaction between ETO and or GAL4-PLZF fusion proteins, cells transfected as described above were lysed and subjected to immunoprecipitation with polyclonal anti-GAL4-DNA binding domain (DBD) antibodies (Zymed, San Francisco, Calif.), with the resulting proteins subjected to electrophoresis and immunoblotting with anti-ETO antibodies as described above.

FIG. 1.

The PLZF protein associates with the ETO gene product in mammalian and yeast cells. (A) Lysates from transiently transfected COS-7 cells were immunoprecipitated (IP) with PLZF antibody, and the bound proteins were immunoblotted with PLZF or ETO antibodies as indicated. Lanes: 1, mock-transfected cells; 2, ETO-transfected cells; 3, PLZF-transfected cells; 4, lysates from cells transfected with both PLZF and ETO. (B) Yeast two-hybrid experiments, in which β-galactosidase assays were performed to detect an interaction between ETO and PLZF. The results are normalized to the interaction between PLZF bait and prey (lane 2). Lanes: 1, full-length GAL4 as a positive control; 3, interaction of ETO fused to the GAL4(DBD) with PLZF fused to the GAL4-AD; 4, lack of interaction between ETO and the SV40 T antigen. (Because there was no growth on Ade⁻ Leu⁻ Try⁻ media, there were no colonies to assay for β-galactosidase activity.) (C) Endogenous PLZF and ETO proteins physically interact. Lanes: 1, immunoblot of ETO from lysates of transfected 293T cells; 2, immunoblot of endogenous ETO in HEL cells; 3, immunoblot of PLZF from lysates of transfected 293T cells; 4, immunoblot of PLZF after immunoprecipitation from HEL cells with PLZF antibodies; 5, immunoprecipitation of HEL cell lysates with a polyclonal ETO antibody Ab-1 and immunoblotting with PLZF monoclonal antibody; 6, immunoprecipitation of HEL lysates with an irrelevant antibody directed against the GAL4(DBD) and immunoblotting with PLZF monoclonal antibody.

FIG. 5.

Identification of sequences with PLZF and ETO required for functional interaction. (A) The presence of both PLZF repression domains is required for functional interaction with ETO. In lanes 1 to 4, 293T cells were cotransfected with the IL3R-tk-Luc reporter (100 ng) and 400 ng of the pCDNA expression vector or pCDNA vector containing sequences encoding wild-type PLZF, PLZF-Δ^POZ/BTB, or PLZF-Δ^RD2. In lanes 5 to 8, the same reporter vector was coexpressed along with wild-type or mutant PLZF expression vector and 400 ng of ETO expression vector. (B) ETO potentiates transcriptional repression by the PLZF POZ/BTB domain. 293T cells were cotransfected in duplicate with the GAL4-tk-Luc reporter (100 ng), 400 ng of the GAL4(1–147) or GAL4-POZ/BTB expression vector, and increasing amounts of ETO (100, 400, and 800 ng, respectively). Repression of the GAL4-tk-Luc reporter was measured relative to expression of luciferase directed by the reporter in the presence of GAL4(1–147). (C) C-terminal domains of ETO are required to enhance repression by PLZF. In lanes 1 to 5, wild-type and or mutant ETO expression vectors (400 ng) were cotransfected into 293T cells along with the PLZF expression vector and IL3R-tk-Luc reporter (100 ng). Lanes: 1, PLZF; 2, PLZF plus ETO; 3, PLZF plus ETO^{(ΔHR,Ner,MYND)}; 4, PLZF plus ETO^(1–322); 5, PLZF plus ETO^(1–322) plus wild-type ETO. In lanes 6 to 10, the GAL4-tk-Luc reporter (100 ng) was cotransfected with a vector encoding GAL4-POZ/BTB and wild-type or mutant ETO. Lanes: 6, GAL-POZ/BTB alone; 7, GAL4-POZ/BTB plus ETO; 8, GAL4-POZ/BTB plus ETO∖^{(ΔHR,Ner,MYND)}; 9, GAL4-POZ/BTB plus ETO^(1–322); 10, GAL4-POZ/BTB plus ETO^(1–322) plus wild-type ETO. (D) ETO enhances repression by the second repression domain of PLZF. The GAL4-tk-Luc reporter (100 ng) was cotransfected in 293T cells along with a GAL4-RD2 (PLZF200–300) vector and an expression vector for wild-type ETO or ETO^(1–322). All data represent the average of two to six individual transfection experiments performed in duplicate.

Coimmunoprecipitation of endogenous PLZF and ETO was performed with human erythroleukemia (HEL) cells. Approximately 6 × 10⁷ cells were lysed in 0.5% Triton buffer for each experiment and incubated with either 2.5 μg of rabbit polyclonal ETO antibody Ab-1, PLZF monoclonal antibody, or irrelevant anti-GAL4 polyclonal antibody (Clontech, Palo Alto, Calif.) for 2 h at 4°C. This was followed by the addition of 30 μl of protein A-conjugated agarose bead slurry (Roche Molecular Biochemicals, Indianapolis, Ind.), which was incubated overnight at 4°C. The precipitated proteins were resolved through an SDS–12% polyacrylamide gel, transferred to an Immobilon P membrane, and blotted with PLZF monoclonal antibody or polyclonal ETO Ab-1 antibody as described above. Direct immunoblotting was performed with 100 μg of protein lysate from HEL and EML (gift of S. Tsai, Mount Sinai School of Medicine) cells or 1/5 (20 μl) of the lysate from transfected 293 T cells. All lysates were submitted to SDS-polyacrylamide (12%) gel electrophoresis (PAGE), transferred to Immobilon P, and blotted with either 1 μg of PLZF monoclonal antibody per ml, 4 μg of our ETO rabbit polyclonal antibody per ml (44), or 2.5 μg of ETO Ab-1 antibody per ml.

Yeast two-hybrid assay.

The PJ69-4A strain of S. cerevisiae (29) was transformed with plasmids encoding the GAL4 DNA binding domain [GAL4(DBD)] linked to ETO or ETO mutants, or the GAL4 acidic activation domain (GAL4-AD) linked to PLZF or the PLZF-POZ/BTB domain. The yeast cells were then grown on media lacking leucine (Leu), tryptophan (Trp), and adenine (Ade). To control for transformation efficiency, an aliquot of the same yeast was also grown on Leu⁻ Trp⁻ media as well. Yeast colonies were counted and then selected in duplicate for liquid β-galactosidase assays according to the manufacturer's protocol (Clontech). Results were normalized relative to the level of β-galactosidase generated by the self-association of PLZF. Positive controls included transformation of yeast with a full-length GAL4 plasmid or transformation with a p53-GAL4(DBD) plasmid and a vector encoding simian virus 40 (SV40) large T antigen fused to the GAL4-AD (Clontech). The same p53 and T antigen plasmids were used as negative controls for binding to PLZF and ETO. Immunoblots were performed to confirm expression of noninteracting proteins.

Protein affinity chromatography.

Sequences encoding amino acids 1 to 114, 217 to 385, 339 to 499, or 494 to 559 of ETO were cloned into the pGEX4T-3 glutathione S-transferase (GST) vector (Pharmacia Biotech AB, Uppsala, Sweden). These constructs were then transformed into Escherichia coli strain DH5α. Single colonies were expanded, and GST fusion proteins were induced, extracted, and collected on glutathione-agarose beads as described previously (1). Full-length PLZF protein was produced by coupled in vitro transcription-translation in the presence of [³⁵S]methionine (1,175 Ci/mmol) (New England Nuclear, Boston, Mass.) from pSG5-PLZF by using T7 RNA polymerase (TNT; Promega, Madison, Wis.). Protein affinity chromatography was performed as previously described (15). The bound proteins were separated by SDS-PAGE (12% polyacrylamide), and the gel was fixed, soaked in sodium salicylate, dried, and subjected to autofluorography at −80°C. Subsequently, the same gel was rehydrated and stained with Coomassie blue to detect expression of the GST fusion proteins.

Transfection assays.

To determine the transcriptional effects of the GAL4 fusion constructs, we used a reporter containing five GAL4 binding sites linked 5′ to the herpes virus thymidine kinase (tk) promoter and the firefly luciferase gene (GAL4-tk-Luc) (gift of P. Traber, University of Pennsylvania). To assay the transcriptional effects of native PLZF, a reporter gene containing four high-affinity PLZF sites found in the IL-3Rα chain promoter were linked 5′ to the tk-luciferase reporter (IL3R-tk-Luc) (2). Several experiments were performed with a reporter containing a large fragment of the cyclin A promoter, which contains two binding sites for PLZF (67). A tk-luciferase construct lacking specific binding sites was used as a negative control, and all experiments contained the renilla tk-luciferase plasmid as an internal control. Cells were plated in 12-well tissue culture dishes at a density of 2 × 10⁵ per well or in 6-well dishes at a density of 4 × 10⁵ per well, and after 24 h, transient transfections were performed with Lipofectamine (Gibco BRL, Rockville, Md.) or Superfect with the combinations and quantities of plasmids noted in the figure legends. Transcriptional activity was determined by dual luciferase assays (Promega, Madison, Wis.), with activity measured with an MLX microtiter plate luminometer (Dynex Technologies, Chantilly, Va.). Experiments were performed in duplicate from 3 to 10 times, and the duplicate luciferase activities were normalized to those of the internal control and averaged. These results are presented as the fold repression of normalized luciferase activity in the presence of PLZF and/or ETO expression vectors compared to normalized luciferase activity in the presence of empty expression vectors. For inhibition of HDAC, 293T cells were transfected as described above in the presence of 200 nM TSA (Sigma) or 1 mM sodium butyrate (gift of Sam Waxman, Mount Sinai School of Medicine). Exposure to each drug was maintained throughout the 48-h culture period.

RESULTS

PLZF interacts in vivo with ETO.

Because ETO and PLZF are expressed in early myeloid cells and can be associated with complexes that contain HDACs, we determined whether ETO and PLZF could physically interact. COS-7 cells were transiently transfected with expression vectors for both proteins. The cell lysates were then immunoprecipitated with a monoclonal PLZF antibody and immunoblotted with a polyclonal ETO antibody (Fig. 1A). Overexpressed ETO is visualized as a dual band of 77 and 80 kDa in COS and 293T cells, which may represent posttranslationally modified species or degradation products. ETO was observed only in immunoprecipitates from lysates containing both ETO and PLZF, but not in lysates lacking either PLZF, ETO, or both proteins. No coimmunoprecipitation of ETO was seen when an irrelevant antibody was used as a negative control. The ETO-PLZF interaction was also studied in the context of the yeast two-hybrid system. GAL4-AD-PLZF was used as a “prey” construct to avoid any false-positive results conferred by the ability of GAL4(DBD)-PLZF to mildly activate transcription (data not shown). Yeast transformed with both the ETO bait and the PLZF prey grew in adenine-deficient media, indicating a protein-protein interaction in vivo. To confirm this interaction, the yeast cells were expanded, lysed, and subjected to a solution β-galactosidase assay. These studies showed a strong interaction between ETO and PLZF, similar to the self-association of PLZF (Fig. 1B). When ETO or PLZF was transformed alone, yeast cells did not grow on adenine-deficient media and yeast cells from leucine- and tryptophan-deficient media, harboring only ETO or PLZF, did not express β-galactosidase.

To establish the potential physiological relevance of this interaction, we determined whether the endogenous PLZF and ETO proteins, expressed in hematopoietic cells, could associate. ETO and PLZF are coexpressed in several hematopoietic cell lines, including HEL and EML (Fig. 1C, lanes 2 and 4, and data not shown). While the presence of ETO was well documented in the HEL cell line (10, 35, 47), PLZF was not previously known to be expressed by these cells, nor was either protein previously shown in the multipotent EML cell line. PLZF was expressed at low levels in HEL cells, requiring prior immunoprecipitation of lysate derived from 6 × 10⁷ cells with monoclonal antibody to concentrate the available gene product prior to immunoblotting (Fig. 1C, lane 4). Immunoprecipitation of the same amount of lysate with an ETO polyclonal antibody coprecipitated a modest, but readily detectable portion of the available PLZF as compared to the amount of PLZF detected by immunoprecipitation blot assay for PLZF (Fig. 1C, lanes 4 and 5). An irrelevant rabbit antibody to the GAL4 (DBD) did not precipitate PLZF (Fig. 1C, lane 6).

ETO potentiates transcriptional repression mediated by PLZF.

Having established that ETO and PLZF can form a complex in vivo, we determined whether the proteins could cooperate to mediate transcriptional repression. PLZF, driven by the SV40 promoter and enhancer was coexpressed in 293T cells or NIH 3T3 cells along with a reporter containing four copies of a high-affinity PLZF binding site found in the IL-3R promoter (IL3R-tk-Luc) (Fig. 2A) (2). PLZF specifically repressed this reporter two- to threefold, as we reported previously (Fig. 2B) (2). When ETO was cotransfected with PLZF, there was an up to 30-fold enhancement of transcriptional repression (Fig. 2B, lanes 5 to 8). This effect was not seen when PLZF and ETO were transfected together with a tk-Luc reporter plasmid lacking PLZF binding sites (Fig. 2B, lanes 1 to 4). Enhancement of repression by ETO was also observed when PLZF was expressed from the cytomegalovirus (CMV) promoter (Fig. 2C). In this experiment, expression of PLZF led to 2.5-fold repression of the promoter, and expression of ETO had a less than 2-fold effect. Together the proteins repressed the IL3-tk-Luc reporter nearly sevenfold. These results were not cell-type specific, since they could be reproduced in 3T3 cells (data not shown). When increasing amounts of ETO were introduced into the cells, there was progressive transcriptional repression, suggesting a cooperative effect between the two proteins. In contrast, increasing doses of ETO alone had relatively little effect on the reporter (Fig. 2D).

FIG. 2.

ETO enhances transcriptional repression mediated by PLZF. (A) Schematic representation of a reporter plasmid containing four binding sites for PLZF from the IL-3Rα chain promoter and a reporter plasmid containing five multimerized GAL4 operators. (B and C) Repression by PLZF through its cognate binding site. Transient transfection assays performed in duplicate in 293 T cells. Results are expressed as fold repression of luciferase expression from the reporter gene in the presence of PLZF and ETO compared to luciferase activity in the presence of parental expression vectors. (B) Lanes 1 to 4, 293T cells in six-well dishes were transfected with the tk-Luc reporter (100 ng) without PLZF binding sites and with pSG5 (800 ng [lane 1]), pSG5 (800 ng) plus CMV-ETO (400 ng) (lane 2), pSG5-PLZF (800 ng [lane 3]), and pSG5-PLZF (800 ng) plus CMV-ETO (400 ng) (lane 4). Lanes 5 to 8 contained IL3R-tk-Luc reporter with similar effectors as lanes 1 to 4. (C) 293T cells in 12-well dishes were transfected in duplicate with the IL3R-tk-Luc reporter (50 ng) and pCDNA (800 ng) plus pCMV (300 ng) (lane 1), pCDNA-PLZF (800 ng) plus pCMV (300 ng) (lane 2), pCDNA (800 ng) plus pCMV-ETO (300 ng) (lane 3), and pCDNA-PLZF (800 ng) plus pCMV-ETO (300 ng) (lane 4). (D) Dose-dependent corepression by ETO. 293T cells in six-well dishes were transfected in duplicate with the IL3R-tk-Luc reporter (100 ng) and pSG5 alone (400 ng [lane 1]) or pSG5 (400 ng) plus 200 ng (lane 2), 400 ng (lane 3) and 1,600 ng (lane 4) of pCMV ETO. Lanes: 5, pSG5-PLZF (400 ng); 6 to 8, pSG5-PLZF (400 ng) plus 200 ng (lane 6), 400 ng (lane 7), and 1,600 ng (lane 8) of ETO. (E) ETO enhances transcriptional repression by a Gal4-PLZF^1–400 fusion protein. 293T cells were transfected with the Gal4-tk-Luc reporter (100 ng) and Gal(1–147) (400 ng [lane 1]) or Gal(1–147) (400 ng) plus CMV-ETO (400 ng) (lane 2). Lanes 3 and 4: Gal-PLZF^1–400 (400 ng [lane 3]) and Gal-PLZF^1–400 (400 ng) plus CMV-ETO (400 ng) (lane 4).

Based on the current model of PLZF function, the nine C-terminal zinc fingers bind specific target promoters, while more N-terminal repression domains can recruit corepressors and HDACs (45). Therefore a GAL4-PLZF fusion containing amino acids 1 to 400 (lacking the C-terminal zinc fingers) was tested for its ability to cooperate with ETO. In concordance with the results observed with wild-type PLZF, GAL4- PLZF^1–400 repressed transcription of a reporter containing five GAL4 operators four- to fivefold (Fig. 2A). In the presence of ETO, nearly 15-fold repression was observed (Fig. 2E). ETO only had a modest, approximately twofold repressive effect on its own in the absence of GAL4-PLZF^1–400. Furthermore, there was no significant effect of ETO and GAL4-PLZF^1–400 on a reporter lacking GAL4 sites (data not shown). These results indicate that the N terminus of PLZF, bound to a specific promoter sequence, is sufficient to functionally interact with the ETO corepressor.

Mapping of ETO sequences required for binding to PLZF.

Previous work on the transcriptional properties of ETO indicated that its ability to repress transcription when fused to AML-1 was largely due to its ability to bind the N-CoR and SMRT corepressors through a C-terminal zinc finger-like motif known as the MYND (Myeloid, nervy, DEAF) domain (17, 43, 44, 63). ETO can also bind the Sin3A corepressor, mainly through a region between amino acids 217 and 387 (B. Lutterbach and S. W. Hiebert, unpublished data). ETO contains three other motifs conserved within members of the ETO family of proteins, including a TAF¹¹⁰ homology domain (13), a hydrophobic heptad repeat (HR) domain which mediates homo- and heterodimerization (30, 43), and the Nervy homology domain 3 (Ner), a conserved domain of unknown function (43). ETO mutants lacking each of the conserved domains (ETO-Δ^MYND, ETO-Δ^Ner, ETO-Δ^HR, and ETO-Δ^TAF110) and a mutant containing amino acids 1 to 345 (and therefore lacking the MYND, Ner, and HR domains) were fused to the GAL4(DBD) and used as baits in the yeast two-hybrid system. These were cotransformed into yeast with a PLZF-AD prey plasmid. Each one of the ETO mutants interacted with PLZF as determined by the growth of yeast colonies on deficient media and by activation of β-galactosidase expression (Fig. 3A). ETO did not interact with a control prey of the SV40 T antigen, and PLZF did not interact with a control bait of p53 (data not shown). This indicted that the N terminus of ETO including amino acids 1 to 345 was sufficient for interaction with PLZF and that the TAF110 homology domain between ETO residues 125 and 196 was not required for the interaction. To further define the region of interaction between ETO and PLZF, we performed protein affinity chromatography using GST-ETO fusion proteins. Glutathione-agarose beads were coated with GST-ETO^1–114, GST-ETO^217–387, GST-ETO^339–499, GST-ETO^494–559, or GST itself and incubated with [³⁵S]methionine-labeled, in vitro-translated PLZF. PLZF was specifically retained only on beads containing a central fragment of ETO, including amino acids 217 to 387 (Fig. 3B). All of the GST-ETO fusion proteins were produced to a comparable extent, as indicated by staining with Coomassie blue (Fig. 3C). Together these data suggest that the minimal region of ETO required to interact with PLZF is localized to a central region of the protein between amino acids 217 and 345. This region is distinct from the C-terminal sequences required for interaction of ETO with N-CoR and SMRT, but similar to the sequences required for interaction with Sin3A.

FIG. 3.

Mapping the region of ETO required for interaction with PLZF. (A) Yeast two-hybrid assays were performed with wild-type and mutant forms of ETO fused to the GAL4(DBD) as bait and full-length PLZF fused to the GAL4-AD as prey. The diagrams to the left depict the ETO mutants used. The results are expressed as β-galactosidase activity normalized to the positive control of PLZF self-association averaged from three to six experiments. Lanes: 1, GAL4-positive control. 2, PLZF self-association; 3, ETO plus PLZF; 4, ETO^ΔMYND plus PLZF; 5, ETO^ΔNer plus PLZF; 6, ETO^ΔHHR plus PLZF; 7, ETO^ΔTAF110 plus PLZF; 8, ETO^1–345 plus PLZF. (B) GST-ETO fusion proteins were produced in E. coli and allowed to bind to glutathione-conjugated beads. The protein-coated beads were allowed to bind to [³⁵S]methionine-labeled in vitro-translated PLZF. The top panel shows a fluorogram of radiolabeled PLZF retained by the glutathione beads. Lanes: 1, naked beads; 2, GST-coated beads; 3 to 6, beads coated with the indicated GST-ETO fusion proteins; 7, reticulocyte lysate (10% of input). (C) Coomassie blue staining of the same gel as in panel B to demonstrate expression of the GST fusion proteins.

A second repression domain, but not the POZ/BTB domain, is essential for interaction of PLZF with ETO.

Previous structure-function analysis of the PLZF protein revealed the presence of two repression domains (37). The first corresponds to the N-terminal evolutionarily conserved POZ/BTB domain between residues 1 and 120, and the second (RD2), located between amino acids 200 and 300, does not resemble any other known repression domains. When fused to the GAL4(DBD), both domains mediate transcriptional repression (37). The N-terminal POZ/BTB domain of PLZF is an interaction site for the N-CoR and SMRT corepressors, but the protein partners of RD2 have not yet been characterized (20, 27). Deletion of either the POZ domain or RD2 blocks the ability of PLZF to repress transcription from its cognate binding site (see Fig. 5A). Therefore we determined which one of these domains interacted with ETO (Fig. 4A). COS-7 cells were transiently transfected with the PLZFΔ^POZ/BTB or the PLZFΔ^RD2 mutant and wild-type ETO. Cell lysates were immunoprecipitated with PLZF monoclonal antibody and then blotted with an ETO polyclonal antibody (Fig. 4B). Full-length PLZF was readily coprecipitated with ETO. PLZFΔ^POZ/BTB could also be coprecipitated with ETO, but to a lesser extent than wild-type PLZF, while PLZFΔ^RD2 did not complex with ETO. This indicated that the second repression domain of PLZF was required for interaction with ETO in vivo and that the POZ/BTB domain contributes to, but is not absolutely required for, the interaction. In a complementary, reciprocal experiment, 293T cells were transfected with PLZFΔ^POZ/BTB or PLZFΔ^RD2 and ETO, followed by immunoprecipitation with ETO and immunoblotting with PLZF antibody. Again, wild-type PLZF exhibited the strongest interaction with ETO, while relatively less PLZFΔ^POZ/BTB could be coprecipitated with ETO. No complex could be demonstrated between PLZFΔ^RD2 and ETO (Fig. 4C). Expression of PLZF deletion mutants and ETO was confirmed in all of these experiments by immunoblotting the cell lysates with the appropriate antibodies (Fig. 4B and C). We next determined which domain of PLZF was sufficient for interaction with ETO. The 293T cell line was transfected with the GAL4-PLZF^1–400, GAL4-POZ, or GAL4-RD2 expression vector, and the resulting lysates were subjected to immunoprecipitation with an antibody directed to the GAL4(DBD) and blotted with ETO antisera. GAL4-PLZF^1–400, containing both repression domains of PLZF, coprecipitated a significant amount of ETO (Fig. 4D). GAL4-RD2 was sufficient to form a complex with ETO, but precipitated less ETO than the longer PLZF fusion protein. In contrast, GAL4-POZ was unable to form a complex with ETO in this assay. The inability of the POZ domain to form a complex with ETO was confirmed in a two-hybrid experiment in which yeast cells were transformed with a GAL(AD)-POZ/BTB prey vector along with GAL4(DBD) bait vectors encoding ETO or full-length PLZF (Fig. 4E). As previously established, PLZF interacted with both itself and ETO. However, yeast cells expressing the PLZF POZ/BTB prey and ETO bait were unable to grow in Leu⁻ Trp⁻ Ade⁻ media, and β-galactosidase levels from yeast grown in Leu⁻ Trp⁻ media, harboring ETO and the POZ/BTB domain, were at background levels (data not shown). Expression of the GAL-POZ fusion product was confirmed by immunoblotting (data not shown). Taken together, these experiments indicate that RD2 of PLZF is required for interaction with ETO and that, on its own, the POZ domain of PLZF does not interact with ETO. However, the POZ domain, when present along with RD2, strengthens the interaction between PLZF and ETO.

FIG. 4.

The second repression domain of PLZF is required for interaction with ETO. (A) Diagram of PLZF deletion mutants used in coimmunoprecipitation (IP) and reporter assays (Fig. 6). PLZF-Δ^POZ/BTB lacks the 100 N-terminal amino acids, and PLZF-Δ^RD2 lacks amino acids 199 to 313. (B) COS-7 cells were cotransfected with plasmids encoding PLZF, PLZF-Δ^RD2, or PLZF-Δ^POZ/BTB. The lysates were subjected to immunoprecipitation with anti-PLZF monoclonal antibody, and the bound proteins were immunoblotted with anti-ETO or anti-PLZF antibodies as indicated. Also shown are immunoblots of the same lysates for PLZF and ETO demonstrating expression of both proteins. (C) 293T cells were transfected with ETO and PLZF, PLZF-Δ^RD2, or PLZF-Δ^POZ/BTB. Lysates were immunoprecipitated with ETO polyclonal antibody and blotted with PLZF monoclonal antibody. Direct immunoblots of the same lysates for PLZF and ETO are shown. (D) 293T cells were transfected with GAL4-PLZF^1–400, GAL4-POZ, or GAL4-RD2 and ETO expression vectors. Lysates were immunoprecipitated with anti-GAL4(DBD) polyclonal antibody and blotted with ETO antibody. (E) Two-hybrid interaction between PLZF and ETO. Yeast cells were transformed with vectors encoding GAL4(DBD) full-length PLZF bait, GAL4-POZ/BTB bait, or ETO bait as indicated and a PLZF-AD or POZ/BTB-AD prey. Yeast colonies growing in Leu⁻ Trp⁻ Ade⁻ media were expanded and assayed for β-galactosidase activity. GAL4 was included as a positive control for activation, and the activity generated by self-association of PLZF was set to 1.

The repression domains of both ETO and PLZF are required for their ability to functionally cooperate.

Having defined the protein sequences required for interaction between ETO and PLZF, we determined whether the same sequences were required for functional interaction and augmentation of transcriptional repression by PLZF. First, we found that PLZFΔ^POZ/BTB and PLZFΔ^RD2 were severely deficient for transcriptional repression of the IL3R-tk-Luc reporter compared to wild-type PLZF (Fig. 5A). Furthermore, both mutants were defective in their ability to cooperate with ETO (Fig. 5A). This correlates with the fact that strong interaction between PLZF and ETO requires the presence of the critical RD2 domain plus the POZ domain. Given the contrast between the lack of a direct interaction of the PLZF POZ domain and ETO, but the requirement of the POZ/BTB domain of PLZF for the ability of ETO to augment repression by PLZF, we further analyzed the functional relationship between the POZ/BTB domain and ETO. Transcriptional repression of the GAL4-tk-Luc reporter by the POZ domain of PLZF fused to GAL4 (GAL4-POZ) was augmented in a dose-dependent manner by coexpression of ETO (Fig. 5B), in a manner similar to that observed for full-length PLZF. Therefore, even though ETO could not be shown to bind to the POZ/BTB domain, it was still able to interact functionally with this repression domain. This could be mediated via bridging factors, such as SMRT and N-CoR, present in mammalian cells.

We next determined the functional consequences of deleting the conserved regions of ETO on transcriptional cooperation between PLZF and ETO. ETO^1–322 and ETOΔ^HHR,Ner,MYND were both defective in their ability to enhance repression of the IL3R-tk-Luc reporter mediated by PLZF (Fig. 5C, lanes 1 to 4). This supports an important role for the N-CoR and SMRT corepressors in ETO-PLZF repression, since these deletion mutants can still interact with PLZF, but have had the C-terminal binding sites required for interaction with N-CoR and SMRT deleted. When wild-type ETO was cotransfected with ETOΔ^HHR,Ner,MYND, enhancement of repression was restored (Fig. 5C, lane 5), indicating that these mutants did not act in a dominant-negative manner to inhibit repression by PLZF or cooperation with ETO. Similar results were obtained in an analysis performed with the GAL4-POZ/BTB fusion protein on the GAL4-tk-Luc reporter (Fig. 5C, lanes 6 to 10). ETO mutants lacking C-terminal sequences associated with corepressor binding were unable to cooperate with the POZ/BTB repression domain. This supports the notion that the POZ domain of PLZF can functionally interact with ETO through a bridging interaction with N-CoR and/or SMRT. Finally, the ability of the RD2 domain to functionally interact with ETO was tested. A GAL4-RD2 fusion containing PLZF amino acids 200 to 300 was an extremely efficient transcriptional repressor of GAL4-tk-Luc, fivefold more powerful than the GAL4-POZ fusion protein (Fig. 5D). The RD2 domain was also potentiated by ETO, and again, this was dependent on the C-terminal sequences of ETO required for interaction with N-CoR and SMRT. Hence, although only RD2 of PLZF appears to be required to form a complex with ETO, both repression domains of PLZF can be affected by ETO.

Inhibitors of HDACs antagonize the repression effect mediated by ETO and PLZF.

To determine whether the cooperative effect by ETO and PLZF was mediated through an HDAC-dependent mechanism, the IL3R-tk-Luc reporter was cotransfected with PLZF and ETO in the presence or absence of TSA and sodium butyrate (4, 68), both inhibitors of HDAC. The inhibitors completely abrogated transcriptional repression by both PLZF and the PLZF-ETO combination (Fig. 6). The effect was less evident when lower doses of inhibitor were used (data not shown). This strongly suggests that the cooperative transcriptional effect mediated by the combination of ETO and PLZF is mediated through the recruitment of HDAC activity to target promoters.

FIG. 6.

ETO modulation of repression by PLZF depends on the action of HDACs. 293T cells were cotransfected with the IL3R-tk-Luc reporter (100 ng), PLZF expression vector (100 ng), and ETO expression vector (200 ng) as indicated in the presence or absence of sodium butyrate at a concentration of 1 mM (A) or TSA at a concentration of 200 nM (B). In lanes 1 and 3 of both panels, fold repression by PLZF and ETO was measured relative to the expression of luciferase in the presence of the parental pCDNA vector in the absence of HDAC inhibitors. In lanes 2 and 4 of both panels, fold repression was calculated relative to luciferase expression in the presence of pCDNA and HDAC inhibitors.

DISCUSSION

Recent reports indicate that deacetylation of nucleosomal histones represents a major mechanism of transcriptional repression (34, 50, 65). Nucleosomes deprived of acetyl groups may have altered interactions with DNA and other proteins that may hinder access of the RNA polymerase II transcriptional machinery to target gene sequences (50). This fundamental mechanism has come under increased scrutiny with the discovery that several leukemogenic chimeric transcription factors are aberrant transcriptional repressors (17, 44, 45, 62). Prominent among these fusion products are the AML-1/ETO fusion generated in t(8;21) M2 leukemia and the RARα fusion proteins of M3, APL. Among these partners of RARα, PLZF stands out as a transcriptional repressor, usually expressed in myeloid progenitor cells. In t(11;17)-associated APL, PLZF-RARα attracts an HDAC-corepressor complex to RARα target promoters, resulting in a leukemia refractory to ATRA (18, 20, 41). ETO, a protein whose exact role in cellular physiology remains uncertain, in a similar manner inappropriately forces a repressor to the promoters of complex AML-1 targets in t(8;21)-associated AML (43, 46).

We showed, for the first time, that ETO can associate with a sequence-specific DNA binding protein, namely PLZF. This indicates that, like N-CoR, SMRT, and Sin3A, ETO can be considered a transcriptional corepressor. This finding is particularly significant, since this interaction may be a marker for a repressor complex active in early myeloid cells, which disappears as cells differentiate. Expression of PLZF was shown to mediate growth suppression, inhibition of myeloid differentiation, and cell cycle arrest (56, 67). Such effects may be mediated by the ability of PLZF to repress target genes, such as those coding for cyclin A2 and the IL-3Rα chain, through the actions of complexes which contain ETO, other corepressors, and HDACs. Thus, further characterization of the transcriptional complex containing ETO and PLZF may contribute to understanding the role of transcriptional repression in myeloid progenitor cells. These statements are supported by several lines of evidence. (i) Both ETO and PLZF are expressed in CD34⁺ progenitor cells, but decrease as myeloid cells differentiate (12, 53; A. M. Melnick, G. Carlile, and J. D. Licht, unpublished data). (ii) Endogenous coexpressed ETO and PLZF interact in vivo, as shown by coimmunoprecipitation in a hematopoietic cell line. (iii) In contrast to PLZF, ETO is unable to bind specifically to DNA (11, 35, 44, 45). (iv) PLZF and ETO bind to a similar group of corepressor and HDAC proteins (8, 17, 18, 20, 41, 44, 62). (v) ETO strongly and consistently potentiates transcriptional repression mediated by PLZF. (vi) The transcriptional repression mediated by PLZF plus ETO is abrogated by treatment with HDAC inhibitors.

It should be emphasized that, in this report, we indicated that the endogenous ETO and PLZF proteins can interact. In contrast, although N-CoR, SMRT, and Sin3A were shown to interact with PLZF in vitro (8, 18, 20, 22, 25, 41, 64) and overexpressed PLZF-RARα and PLZF were found to interact with these corepressor molecules in immunoprecipitation experiments (8, 18, 22, 41), no experiments have yet been published indicating endogenous PLZF interacts with any of these corepressors. Although coimmunoprecipitations of overexpressed protein and yeast two-hybrid assays indicated a strong and consistent interaction between ETO and PLZF, similar studies of the native proteins in HEL cells indicate that only about 10% of endogenous PLZF coprecipitates with ETO (Fig. 1C). This finding may have several explanations. First, not all repressive complexes containing PLZF may require or contain ETO. Second, ETO may target other proteins to mediate repression. Furthermore, ETO may serve in other roles besides transcription. ETO binds to poly(G) RNA and can be localized in the nucleolus, consistent with a role for ETO in RNA metabolism (11). PLZF may have other functions as well. PLZF can colocalize with the PML protein of t(15;17)-associated APL (32, 55) in the nuclear body structure. Therefore, PLZF may be involved in PML-mediated processes such as apoptosis and control of mRNA transport and translation (3, 24, 45). Finally, the interaction between ETO and PLZF may depend on posttranscriptional modifications, which may change depending on the cell cycle, stage of differentiation, and cell growth conditions. Indeed, both ETO and PLZF proteins interact with regulatory kinases and are subject to phosphorylation (2, 12, 33, 60).

ETO consistently potentiated transcriptional repression mediated by native PLZF as well as the PLZF repression domains fused to a heterologous DBD. At least part of this transcriptional function is dependent on the ability of PLZF to associate with the same corepressors and HDACs as ETO (8, 18, 20, 41). The potentiation of transcriptional repression by ETO can be further enhanced by coexpression of the N-CoR or SMRT corepressors and HDACs, indicating a combinatorial effect in which all of these proteins functionally interact (data not shown). Finally, the transcriptional repression mediated by ETO and PLZF is blocked by TSA and sodium butyrate, indicating that the enhancement of PLZF repression mediated by ETO is dependent on HDAC activity. This suggests that ETO enhances the ability of PLZF to recruit a transcriptional repression complex which functions, at least in part, through histone deacetylation (Fig. 7A).

FIG. 7.

Model of interaction between ETO and PLZF. (A) PLZF binds to specific sequences in its target promoters through its zinc finger (ZF) domains. PLZF then recruits ETO through sequences in the second repression domain of PLZF (residues 200 to 300), binding to sequences of ETO localized between amino acids 217 and 345. Both PLZF and ETO are able to bind to N-CoR, SMRT, Sin3A, and HDACs, resulting in a multiprotein complex able to mediate nucleosomal modifications and transcriptional repression. ETO binds to N-CoR and SMRT through its C-terminal MYND domain. The PLZF-POZ/BTB domain stabilizes the interaction, though does not bind directly to ETO, possibly through interaction with N-CoR and/or SMRT. The effect of ETO may be to either stabilize the complex or to help recruit HDACs and/or as-yet-unidentified additional proteins, augmenting transcriptional repression. (B) GAL4-POZ/BTB recruits N-CoR and SMRT, which then bind to ETO. This results in an N-CoR–SMRT bridge (not present in yeast) between POZ/BTB and ETO. ETO also interacts with HDACs and Sin3A. Augmentation of repression would then occur as described for panel A.

The POZ/BTB domain of PLZF mediates homodimerization, transcriptional repression, and heterologous protein-protein interactions, including those with corepressors (8, 20, 45). ETO does not associate with the PLZF-POZ/BTB domain in yeast and cannot form a stable complex with this domain in mammalian cells. Furthermore, ETO can still bind to PLZF with the POZ/BTB domain deleted. Despite this fact, ETO still could enhance repression mediated by the POZ/BTB domain of PLZF. This paradox can be solved by hypothesizing that ETO is able to mediate this effect indirectly, through interaction with the N-CoR and SMRT or Sin3A corepressors, all of which can interact with the POZ/BTB domain (Fig. 7B). ETO fails to enhance transcription of PLZF mutants with either the POZ/BTB domain or the second repression domain deleted, and PLZF with the POZ/BTB domain deleted exhibited diminished binding of ETO. It therefore seems likely that the POZ/BTB domain is important for stabilizing a multiprotein complex containing ETO. The second repression domain of PLZF was absolutely required for interaction with ETO. This was intriguing, given that the mechanism of action of this domain of PLZF had not previously been characterized. A PLZF deletion mutant lacking the RD2 was severely deficient for transcriptional repression and was not significantly potentiated by ETO (Fig. 6A), while ETO was able to augment repression mediated by the second repression domain of PLZF fused to GAL4 (Fig. 6C). Together, this information indicates that there are multiple sites for corepressor interaction in the PLZF protein. The binding of N-CoR, SMRT, and Sin3A to the N-terminal POZ/BTB domain of PLZF and ETO to the more C-terminal repression domain may lead to the formation of a multiprotein repression complex, stabilized by multiple cross-protein contacts (Fig. 7A).

Surprisingly, none of the four conserved domains of ETO were required to associate with PLZF, while a previously poorly characterized, central region of ETO was sufficient to complex with PLZF. Interestingly, the Sin3A corepressor binds to ETO through similar sequences (B. Lutterbach and S. W. Hiebert, unpublished data). The functional relationship between Sin3A, ETO, and PLZF is underscored by the fact that ETO and Sin3A (44), ETO and PLZF (Fig. 1C), and PLZF and Sin3A (A. M. Melnick, G. Carlile, and J. D. Licht, unpublished data) can be coimmunoprecipitated in HEL cells. Although our yeast two-hybrid (Fig. 1B) and protein affinity chromatography studies (Fig. 4C) could indicate a direct interaction between PLZF and ETO, it remains a strong possibility that Sin3A could act as a bridging factor between these proteins, since Sin3A is present in yeast and Sin3A can be detected in reticulocyte lysates (B. Lutterbach and S. W. Hiebert, unpublished data). It must be emphasized that the N-terminal half of ETO alone, despite its ability to bind to both PLZF and Sin3A, was unable to mediate corepression, suggesting that the participation of Sin3A is not sufficient for repression by ETO. The ability of ETO to augment repression by PLZF required C-terminal sequences of ETO. This suggests a model in which the central portion of ETO forms complexes with PLZF, while the C-terminal region binds to the N-CoR and SMRT corepressors, which in turn can bind to the POZ/BTB domain of PLZF. The resulting collection of proteins recruit HDACs and mediate transcriptional repression (Fig. 7A).

Given the fact that ETO and PLZF physically and functionally interact, how is this relationship disrupted in leukemia? If for example, the AML-1-ETO oncoprotein interacts with wild-type PLZF in myeloid progenitor cells, would this affect PLZF-mediated repression? Alternatively, does ETO participate in the repression of RARα target genes? In fact, our preliminary results indicate that AML-1–ETO is a dominant-negative inhibitor of PLZF repression. Hence in t(8;21)-associated AML, not only might AML-1 target genes be aberrantly repressed, but PLZF targets associated with control of cell growth might be derepressed as well. In the case of APL, our initial data indicate that ETO can potentiate repression of reporter genes containing RAR binding sites in the absence of retinoic acid. Whether PLZF-RARα might affect some other essential function of ETO or sequester ETO, making it unavailable for use by other transcriptional repressors or other cellular processes, is uncertain. ETO and PLZF may represent components of a functional repression complex present in myeloid cells. Disruption of this complex could represent a common pathway of malignant transformation. A fuller understanding of other components of this complex will lead to the identification of mechanisms of transcriptional repression and hence offer new opportunities for therapeutic intervention in leukemia.