The molecular contacts between the ETO zinc finger motif and the nuclear hormone corepressors N-CoR and SMRT have been defined by yeast two-hybrid assays, mammalian two-hybrid assays, and purification studies (10
). However, our approach of defining domains of ETO that are sufficient for corepressor binding by using chimeric ETO/ETO-2 and GAL4-ETO proteins has defined multiple contacts between ETO, N-CoR, mSin3A, HDAC-1, HDAC-2, and HDAC-3 (the known molecular contacts for ETO are summarized in Fig. A). By using overlapping fragments of ETO as GAL4-ETO fusion proteins, we were able to separate closely adjoining binding sites for corepressors. For example, HDAC-3 bound both GAL4-ETO(259-343) and GAL4-ETO(300-385), indicating that the binding site likely resides in the region of overlap between these fragments (residues 300 to 343). By contrast, mSin3A bound to only GAL4-ETO(300-385), placing this binding site C terminal to HDAC-3 (Fig. A). These results are consistent with the recent identification of an mSin3A binding site that overlapped the HHR domain (16
). However, our results using point mutations indicated that the HHR-dimerization domain of ETO contacts mSin3A (Fig. ). Although we were able to define a second mSin3A binding site between the HHR and nervy domain by using the ETO/ETO-2 fusion proteins, we were not able to use this strategy to map the second HDAC-2 binding site because HDAC-2 also bound ETO-2 (Fig. B).
ETO is the prototype of a family of highly homologous evolutionarily conserved proteins. Like ETO, ETO-2 is a potent repressor when tethered to a promoter by fusing it to the GAL4 DNA binding domain (Fig. ). Therefore, it was unexpected that ETO, but not ETO-2, interacted with mSin3A. In fact, while the human family members ETO and MTG16a are 67% identical, the regions that specify mSin3A binding include one of the least conserved regions of ETO and a subtle alteration in the HHR-oligomerization motif. Because HDAC-1 and HDAC-2 heterodimerize and copurify with mSin3, whereas HDAC-3 copurified with N-CoR (24
), we propose a model in which ETO can recruit at least two different corepression complexes (Fig. B). This model is consistent with the sucrose gradient sedimentation pattern of ETO because ETO cosedimentated with both N-CoR and mSin3A but was found in fractions that contained mSin3A and not N-CoR (29
). However, as ETO can bind HDACs independent of either mSin3A or N-CoR, it remains possible that ETO is a component of multiple corepression complexes.
The lack of mSin3A binding by ETO-2 and its ability to associate with five HDACs indicates that it is a component of a corepressor complex(es) that is distinct from ETO-containing complexes (Fig. C). If ETO and ETO-2 have distinct mechanisms of action or distinct targets for repression, this would explain why multiple family members are expressed in the same cell type. Alternatively, the fact that ETO family members can homo- and heterodimerize through the HHR domain adds a further level of complexity to transcriptional regulation by this family and may be a way of fine tuning transcriptional control in a particular tissue (21
It has been postulated that oligomerization of ETO is critical for interactions with N-CoR or SMRT and for AML-1–ETO biological actions (33
). However, in light of our definition of corepressor binding sites within and adjacent to the HHR-dimerization domain of ETO, it is more likely that it was the deletion of these corepressor binding sites that affected AML-1–ETO function (46). In fact, the AML-1–ETO deletion that was used to suggest a biological role for dimerization of AML-1–ETO (deletion of ETO residues 340 to 440 as numbered here) impinged upon one HDAC-1 binding motif, both mSin3A binding domains, and possibly binding sites for HDAC-2 and HDAC-3 (46) (Fig. A). By contrast, a more specific deletion of the HHR-dimerization motif that affected only one mSin3A-binding site had modest effects on GAL4-ETO (45
) and AML-1–ETO-mediated transcriptional repression (28
). This mutation reduced, but did not eliminate, binding to mSin3A, N-CoR, or HDACs (references 28
and this work). Thus, it is possible that dimerization contributes to repression, but dimerization is not required for corepressor binding. If indeed ETO family proteins heterodimerize, our results indicate that association between ETO and ETO-2 could lead to the recruitment of a distinct set of corepressors and HDACs given that ETO binds mSin3A, but ETO-2 does not, and ETO-2 binds HDAC-6 and HDAC-8. This might broaden the biological action of ETO and, by extension, AML-1–ETO.
Though counterintuitive, inhibition of cell cycle progression by AML-1–ETO has a precedent in that the inv
) fusion protein, which also represses AML-1 target genes, caused a slowing of cell cycle progression in the G1
). In fact, t(8;21)-containing leukemia cell lines (e.g., Kasumi-1) grow very slowly. Although it is difficult to ascertain whether this phenotype is related to leukemogenesis, it does afford us a highly reproducible biological assay with which to probe the mechanism of action of AML-1–ETO. TSA induces cell differentiation, cell death, and, in some cases, cell cycle inhibition, perhaps due to induction of p21waf1/cip1
). Indeed, upon longer exposure of these MEL cells to TSA, we observed partial differentiation and cell cycle arrest in both control and AML-1–ETO-expressing cells (data not shown). However, in the short term, we observed that TSA biologically inactivated the t(8;21) fusion protein.
The t(15;17) and t(11;17) fusion proteins also used HDACs to repress transcription and preliminary work using inhibitors of these enzymes in therapy was successful (42
). In addition, t(8;21)-containing blasts are sensitive to a combination of retinoic acid and TSA (8
). Our results begin to provide a mechanistic basis for the sensitivity to TSA and further emphasize that HDAC inhibitors can biologically inactivate AML-1–ETO. This work also demonstrates the necessity of the development of HDAC inhibitors that target specific classes of HDACs or specific HDACs. Although N-CoR and mSin3A bind class I and class II HDACs, the independent association of HDAC-1, HDAC-2, and HDAC-3 with ETO suggests that treatment of patients harboring t(8;21) may require only inhibition of the class I deacetylases, thereby decreasing any potential toxicity associated with a broad-spectrum HDAC inhibitor such as TSA. Moreover, our work demonstrates the interactions that can
take place, but the constellation of corepressors and HDACs that are expressed in t(8;21)-containing leukemic blasts will ultimately define the critical therapeutic targets.