INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Nearly 40% of cases of acute myeloid leukemia (AML) M2 are associated with the t(8;21)(q22;q22) chromosome translocation (39). This translocation creates a fusion between the AML1 gene on chromosome 21 and the ETO gene (also known as MTG8/CDR) on chromosome 8. The resulting chimeric protein AML1-ETO contains the DNA-binding domain (DBD) of AML1 and nearly all of ETO (5, 24, 38, 42). The underlying mechanism of AML1-ETO leukemogenic activity is not fully understood.

AML1 is a hematopoietic cell-specific transcription factor and is essential for definitive hematopoietic development (43, 44, 51). At least two potentially related mechanisms have been proposed for the AML1 function. The first is that AML1 synergistically interacts with other adjacent transcription factors, including C/EBP (55) and myb (2, 53). The second is that AML1 is able to recruit the p300/CBP coactivator complex to activate its target gene expression (23). In t(8;21), the activation domain of AML1 is replaced by the ETO protein. Transient-transfection assays indicate that AML1-ETO interferes with AML1 transactivation from certain potential AML1 target genes (9, 36).

ETO is expressed at high levels in brain (38) and has also been detected in hematopoietic cells (6). Recently, we and others discovered that ETO as well as AML1-ETO physically associates with nuclear receptor corepressors called N-CoR (nuclear receptor corepressor) and SMRT (silencing mediator for retinoid and thyroid receptors) (11, 33, 49). This discovery provides a mechanism for the repression activity of AML1-ETO. N-CoR and SMRT recruit class I and class II histone deacetylases (HDACs) (19-21), and this enzyme activity leads to a repressive chromatin state. Consistent with this mechanism, ETO and AML1-ETO are both associated with cellular HDAC activity (11, 33, 49).

ETO is homologous to the Drosophila melanogaster protein Nervy in four regions (7). Both N-CoR-HDAC and SMRT-HDAC associations require the most C-terminal Nervy-homologous region (NHR), NHR4, also called the MYND domain (31). This domain contains two putative zinc (Zn) fingers, and point mutations in either of these block interaction with N-CoR and SMRT (11). NHR4 is also required for the AML1-ETO fusion protein to block hematopoietic differentiation of U937 cells. Interestingly, an NHR in the middle of the molecule, NHR2, also plays a role in the dominant-negative activity of AML1-ETO (17, 25) as well as its ability to repress basal transcription (31). Biochemically, NHR2 forms an amphipathic helix and is proposed to mediate homodimerization (31) as well as heterodimerization with an ETO-related protein, MTGR1 (22).

Here, we show that both NHR2 and NHR4 are required for ETO interaction with SMRT as well as functional repression. Removal of NHR2 diminishes the ability of AML1-ETO to form oligomers and block hematopoietic differentiation of U937 cells. Fusion to a heterologous dimerization domain allows NHR4 to interact with SMRT. These results are consistent with a model in which productive corepressor association with ETO or AML1-ETO requires at least two NHR4 polypeptides brought together by the NHR2 oligomerization function. Recruitment of the nuclear receptor corepressor complex is thus dependent upon oligomerization of AML1-ETO via NHR2.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Plasmids. Full-length mouse ETO cDNA as well as DNA lacking amino acids 488-525 and ETO-C488S and -C508S mutants have been described previously (11). The HA-ETO403-577 and HA-ETO295-577 constructs were truncated at amino acids 404 and 296, respectively (made through deletion at the EcoRI or Bsu36I site). The NHR2 deletion shown in Fig. 2, 3, 5, and 6 removes amino acids 325 to 345 of ETO. The plasmid pcDNA3-AML1-ETO has been previously described (11, 37). The AML1-ETONHR2 deletion carries an internal deletion that eliminates the NHR2 region (amino acids 313 to 413 of ETO). Retroviral expression of AML1-ETO and AML1-ETONHR2 was obtained by cloning the appropriate cDNA at the EcoRI site of the Pinco virus (11). N-CoR lacking amino acids 1 to 161 and repression domain 3 (RD3) (amino acids 1007 to 1445) were generated from full-length N-CoR by PCR. SMRT plasmids have been previously described (11). All mutants were made using PCR and confirmed by sequencing.

In vitro interaction assays. Glutathione S-transferase (GST) pulldown assays were performed as previously described (56).

Cell culture and transfection. 293T and C33A cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. 293T cells were transfected by the calcium phosphate precipitation method as described previously (56). C33A cells were transfected with Lipofectamine reagents (GIBCO) according to the manufacturer's instructions. The Gal4 upstream activation sequence × 5-simian virus 40-luciferase reporter contains five copies of the Gal4 17-mer binding site. Light units were normalized to the expression of a cotransfected -galactosidase expression plasmid. Mammalian two-hybrid assays were performed and interpreted as previously described (56). Experimental results are expressed as the mean and range of duplicate samples, and each experiment was performed multiple times.

Cell differentiation experiments. Differentiation experiments of U937 cells were performed as described previously (11). Cells were infected with a retrovirus that expressed AML1-ETO or AML1-ETONHR2 (lacking amino acids 313 to 413 of the ETO moiety) as described previously (11) except that green fluorescent protein (GFP) was expressed under the control of an internal cytomegalovirus promoter. In essentially all cases for U937 cells, we have found that the intensity of the GFP fluorescence reflects the levels of the fusion protein. Comparable expression of AML1-ETO and AML1-ETONHR2 was ascertained by semiquantitative reverse transcription-PCR of the infected cells, Western blot analysis of tagged pcDNA3 AML1-ETO and AML1-ETONHR2 in transfected U937 cells, and comparable GFP levels in the two populations of infected cells. The percentage of GFP-positive cells, the differentiation of antigen-positive cells (either GFP-positive or GFP-negative cells), and the fluorescence intensity were evaluated by FACScan.

Size-exclusion chromatography. In vitro-translated AML1-ETO and deletion derivatives were analyzed by size-exclusion chromatography on a Superose 6 column (SMART system; Pharmacia, Uppsala, Sweden) equilibrated in column buffer (20 mM HEPES [pH 7.4], 1 mM EDTA, 1 mM dithiothreitol, aprotinin [10 µg/ml], leupeptin [10 µg/ml], pepstatin [2 µg/ml], 1 mM phenylmethylsulfonyl fluoride, 1% glycerol, 5 mM NaF, 0.4 M KCl). Fractions were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by autoradiography.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

ETO C terminus is necessary but not sufficient for interaction with SMRT. We previously reported that point mutations in either of the two C-terminal Zn fingers of ETO abolished interaction with corepressors N-CoR and SMRT. This is in agreement with reports from other investigators (33, 49) and indicates that the zinc finger-containing C terminus of ETO is necessary for corepressor interaction. Here we tested whether the C terminus of ETO was sufficient for interaction. In vitro-translated ETO interacted with SMRT fused to GST, and as expected, deletion of the C terminus of ETO (403-577) abolished this interaction (Fig. 1a). However, although required for interaction, the C-terminal 174 amino acids (404 to 577) containing the zinc fingers and NHR4 did not interact with SMRT, indicating that this polypeptide was insufficient for interaction (Fig. 1a). Similar observations were made when full-length SMRT was translated in vitro and used in GST pulldown assays with GST-ETO, GST-ETO403-577, and GST-ETO(404-577). In this context as well, the corepressor interacted with full-length ETO but not with the N- or C-terminally deleted mutants (Fig. 1b).

View larger version (34K): [in this window] [in a new window]   FIG. 1.   The C-terminal Zn finger domain of ETO is necessary but not sufficient for interaction with SMRT in vitro. (a) Interaction of ETO proteins with GST-SMRT RD3 (amino acids 1041 to 1476) (11). (b) GST pulldown assay of full-length SMRT (45) by GST fusions to various ETO proteins.

The NHR2 domain is required for interaction with SMRT in vitro. We next explored the importance of NHR2 for interaction between ETO and nuclear hormone receptor corepressors. This region is encompassed by 28 amino acids (Fig. 2a). Interestingly, deletion of a highly conserved 20-amino-acid polypeptide (amino acids 325 to 345) constituting the bulk of NHR2 greatly reduced the ability of in vitro-translated ETO to interact with SMRT (Fig. 2b). These results indicate that NHR2 is required for SMRT interaction with ETO in vitro.

View larger version (58K): [in this window] [in a new window]   FIG. 2.   NHR2 of ETO is required for interaction with SMRT in vitro. (a) Schematic of ETO showing the location of the NHRs. (b) Interaction of ETO proteins with GST-SMRT. ETONHR2 lacks amino acids 325 to 345 of ETO. Migration of molecular mass markers is indicated.

Both NHR2 and NHR4 contribute to SMRT interaction in vivo. We next explored the relative contributions of the NHR2 domain and the NHR4 Zn finger domain in vivo. These experiments utilized a mammalian two-hybrid assay in which the ETO interaction domain of SMRT was fused to Gal4, and ETO and various mutants were fused to VP16, which contains a strong transcription activation domain. The strong interaction between SMRT and ETO was confirmed in 293T cells by using this assay (Fig. 3a). Mutation of either zinc finger (C488S, C508S) essentially abolished any detectable interaction. In addition, deletion of the 20-amino-acid NHR2 polypeptide dramatically reduced the interaction between ETO and SMRT, confirming the importance of the ETO-ETO interaction for high-affinity, stable corepressor interaction (Fig. 3a). Similar results were observed in C33A cells (Fig. 3b).

View larger version (15K): [in this window] [in a new window]   FIG. 3.   NHR2 and NHR4 are required for ETO interaction with SMRT in cells. A mammalian two-hybrid assay of interaction between Gal4-SMRT and various VP16-ETO fusion proteins in 293T cells (a) and C33A cells (b) is shown. Results shown are normalized luciferase activities (see Materials and Methods). ETONHR2 lacks amino acids 325 to 345 of ETO.

Deletions encompassing NHR2 abolish oligomerization of ETO and AML1-ETO. NHR2 has previously been demonstrated to function as a self-association domain (22, 31). C-terminal truncation of ETO at amino acid 403 did not abolish interaction with GST-ETO, whereas a mutant ending at amino acid 295 was unable to interact with ETO (Fig. 4a). This shows that amino acids 295 to 403, encompassing NHR2, are required for in vitro-translated ETO to interact with GST-ETO. We next turned our attention to the AML1-ETO fusion protein. Superose 6 chromatography analysis of AML1-ETO showed that, consistent with previous reports (33, 37, 49), the AML1-ETO fusion protein was found in high-molecular-weight complexes (Fig. 4b, top). These complexes were also observed with bacterially expressed AML1-ETO (data not shown), suggesting that these represent AML1-ETO oligomers. An internal deletion that included NHR2 of the ETO moiety shifted the chromatographic profile of the fusion protein to lower-molecular-weight forms (Fig. 4b, bottom). These results are consistent with a role of NHR2 in the formation of high-molecular-weight complexes by the AML1-ETO fusion protein.

View larger version (31K): [in this window] [in a new window]   FIG. 4.   Deletions encompassing NHR2 abolish oligomerization of ETO and AML1-ETO. (a) Interaction of ETO proteins by using GST pulldown assay. (b) Size-exclusion chromatography products of AML1-ETO (top) and AML1-ETONHR2 (bottom). AML1-ETONHR2 lacks amino acids 313 to 413 of the ETO moiety. Elution of globular protein molecular weight standards (arrows) is shown for comparison.

NHR2 is required for AML1-ETO to block hematopoietic differentiation of U937 cells. Ectopic expression of AML1-ETO interfered with vitamin D₃-dependent monocytic differentiation of U937 cells (Fig. 5), consistent with our earlier results (11). However, AML1-ETO-NHR2 was much less capable of blocking hematopoietic differentiation under identical conditions (Fig. 5). Thus, this ETO mutant lacking the NHR2 oligomerization domain was functionally similar to mutants lacking the C-terminal Zn finger corepressor-interaction domain (11). By contrast, an AML1-ETO mutant lacking the NHR3 region retains the ability to block differentiation (data not shown).

View larger version (39K): [in this window] [in a new window]   FIG. 5.   NHR2 is required for AML1-ETO to block differentiation of U937 cells. U937 cells were infected with the retroviral vector alone (control) or with retroviral vector expressing AML1-ETO or AML1-ETONHR2, and then cells were induced to differentiate with 1, 25-(OH2)-D3 (VD3) and transforming growth factor  as previously described (11). AML1-ETONHR2 lacks amino acids 313 to 413 of the ETO moiety. GFP positivity and the presence of the surface marker CD14 were monitored by fluorescence-activated cell sorting. GFP was expressed from the GFP expression cassette of the vector, and it marked infected cells. CD14 was used as a marker of differentiation (11).

Oligomerization is required for ETO to bind corepressor. Earlier we showed that the isolated C terminus of ETO is not sufficient for corepressor interaction (Fig. 1). Remarkably, fusion of the identical polypeptide (amino acids 404 to 577) to Gal4 allowed robust interaction with GST-SMRT (Fig. 6a). This result is consistent with the observation that Gal4-ETO(484-577) interacted with N-CoR in yeast (49). The Gal4 DBD is a potent protein-protein interaction domain, such that Gal4 fusion proteins are obligate oligomers both in vitro and in vivo (34). Indeed, Gal4-ETO(404-577) as well as Gal4 DBD behaved as oligomeric species when chromatographed using Superose 6 (Fig. 6b). These data are consistent with the conclusion that the heterologous dimerization domain of Gal4 allowed the C-terminal polypeptide of ETO to interact with SMRT.

View larger version (38K): [in this window] [in a new window]   FIG. 6.   Dimerization is required for SMRT interaction and repression function of ETO NHR4. (a) Fusion of NHR4 zinc fingers to the Gal4 dimerization domain is sufficient for SMRT interaction in vitro. Shown are results of a GST pulldown assay of Gal4-ETO(404-577) using GST and GST-SMRT. (b) Size-exclusion chromatography of Gal4 DBD and Gal4-ETO(404-577). Elution of globular protein molecular mass standards (asterisks) is shown for comparison. (c) Deletion of NHR4 but not NHR2 does not block repression function of Gal4-ETO in 293T cells. (d) Deletion of both NHR2 and NHR4 abolishes repression by Gal4-ETO in 293T cells. (e) ETO, but not ETONHR2, potentiates repression by Gal4-ETO. Results shown are normalized luciferase activities (see Materials and Methods). ETONHR2 lacks amino acids 325 to 345 of ETO.

Oligomerization is required for ETO to repress transcription. We next examined the ability of various ETO polypeptides to repress transcription as Gal4 fusion proteins. As predicted, wild-type ETO was a strong repressor (Fig. 6c). Unlike wild-type ETO, ETO403-577 did not interact with corepressors or the HDAC complex in vivo when expressed as hemagglutinin (HA) epitope-tagged proteins in 293T cells (11). Remarkably, this ETO polypeptide did repress transcription in 293T cells when expressed as a Gal4 fusion protein (Fig. 6c). This was surprising, since the Gal4-ETO403-577 mutant lacks the corepressor-interaction surface. To explain this result, we hypothesized that the Gal4 dimerization domain would keep the two ETO403-577 molecules together on the reporter gene while the NHR2 interaction domain of each interacted with endogenous ETO and/or ETO dimerization partners, such as MTGR1 (22). Consistent with this, further truncation of amino acids 295 to 403 (ETO295-577) abrogated repression activity (Fig. 6c). Deletion of only the 20-amino-acid NHR2 region also did not block repression (data not shown).

A subdomain of RD3 of nuclear receptor corepressors is necessary and sufficient for ETO interaction. N-CoR and SMRT have multiple functional domains, including three RDs. We and others have previously shown that ETO interacts with RD3 of N-CoR and SMRT (11, 49). The results thus far suggest that at least two ETO NHR4 polypeptides are required for interaction with nuclear receptor corepressors. This raised the question of whether multiple regions within the larger N-CoR and SMRT proteins interact with ETO. A nearly full-length N-CoR protein lacking all but the first 160 amino acids did interact with ETO (Fig. 7a). This was expected since we have previously shown that RD1 of N-CoR (amino acids 1 to 312) does not interact with ETO (11). However, deletion of RD3 (amino acids 1007 to 1445) within this nearly full-length N-CoR abolished interaction with ETO (Fig. 7a). This showed that RD3 was necessary as well as sufficient for interaction with ETO. We have recently found that the C-terminal region of SMRT RD3 (amino acids 1242 to 1476) is responsible for repression and interaction with HDACs 4 and 5 (20). Interestingly, this subdomain of RD3 does not interact with ETO (Fig. 7b). Rather, the N-terminal portion of RD3 (amino acids 1041 to 1258) are sufficient to mediate the interaction between SMRT and ETO (Fig. 7b). These data indicate that this polypeptide contains the entire surface or surfaces required for interaction with ETO oligomers.

View larger version (35K): [in this window] [in a new window]   FIG. 7.   Corepressor RD3 is necessary and sufficient for ETO interaction. (a) Interaction of in vitro-translated N-CoR and N-CoR lacking RD3 with GST or GST-ETO. (b) Interaction of ETO with GST, GST-SMRT RD3 (amino acids 1041 to 1476), and GST fusions to indicate polypeptides derived from RD3.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this report, we demonstrated that the pathogenic recruitment of corepressors by AML1-ETO requires the ETO oligomerization motif, NHR2. Moreover, NHR2 is essential for AML1-ETO to block hematopoietic differentiation. Most strikingly, we found that a heterologous dimerization motif can rescue an NHR2 deletion to restore the interaction of ETO with corepressors and restore transcriptional repression activity.

These results support a model in which recruitment of N-CoR or SMRT requires AML1-ETO to present at least two NHR4 polypeptides (Fig. 8). The N-terminal subdomain of RD3 constitutes the interaction surface(s) of the corepressor. Oligomerization of AML1-ETO is normally mediated by NHR2, but heterologous dimerization domains can substitute. This model explains why dimeric Gal4-NHR4 but not NHR4 itself is capable of high-affinity interaction with SMRT (Fig. 8c and d; compare Fig. 1a and 6a). It also explains why provision of the Gal4 dimerization surface to NHR4-mutant ETO allows the NHR2 function to repress transcription in certain cell types (Fig. 6c, modeled in Fig. 8e). This is consistent with the ability of ectopically expressed MTGR1 to enhance the activity of AML1-ETO in disrupting differentiation of L-G myeloid cells, in which a dimerization partner, and hence corepressor recruitment, might be limiting (22).

View larger version (24K): [in this window] [in a new window]   FIG. 8.   Role of dimerization, NHR2, and NHR4 in repression and recruitment of the SMRT-N-CoR-HDAC complex by ETO. (a) Wild-type ETO is an oligomer presenting two NHR4 surfaces to SMRT (or N-CoR). (b) The dimeric ETONHR4 does not bind SMRT or N-CoR in vitro because the Zn finger-structured NHR4 interaction surface is not present. (c) The monomeric ETONHR2 does not bind SMRT or N-CoR in vitro because at least two Zn finger-structured NHR4 interaction surfaces need to be presented. (d) Gal4-NHR4 binds SMRT and N-CoR in vitro and represses in vivo because the Gal4 dimerization domain replaces NHR2. (e) Gal4-ETONHR4 binds SMRT and N-CoR and represses in cells containing endogenous ETO or ETO-like molecules which can interact with NHR2, allowing the putative oligomeric complex to present two NHR4 interaction surfaces to the corepressors. Similar mechanisms pertain to the AML1-ETO fusion protein.

Interestingly, dimerization of nuclear hormone receptors is also required for productive interaction with N-CoR and SMRT (54). In that case, the interaction surface on the transcription factor is a hydrophobic pocket formed by the folding of -helices, and a single corepressor contains two interaction domains, each containing an amphipathic helix called a CoRNR box that binds in this pocket (19, 41, 46). The structure of the ETO NHR4 polypeptide has not been solved, but it is likely to contain zinc fingers that are necessary for corepressor interaction (11). We have localized the region of N-CoR and SMRT that interacts with ETO to a subdomain within RD3. Intriguingly, this subdomain contains two copies of a GSI motif that were first noted in the Drosophila corepressor SMRTER (48). The significance of this is unclear, however, because GSI motifs are also found in the C-terminal RD3 polypeptide that interacts with class II HDACs but not ETO. Nevertheless, the oligomerization dependence of corepressor recruitment appears to be a general phenomenon. In this regard, it is noteworthy that TBL1, a component of the endogenous high-molecular-weight SMRT complex (13, 27), also forms oligomers in solution (M. Guenther and M. A. Lazar, unpublished data).

Since we, and others, first identified a role for nuclear receptor corepressor pathways in acute promyelocytic leukemia (APL), it has become evident that aberrant recruitment of these pathways is a recurring event in leukemogenesis (12, 15, 29). In addition to the 8;21 translocation described in this paper, the 16;21 translocation creates a fusion between AML1 and the ETO family member MTG16 (10). While the resulting fusion protein has not formally been shown to interact with corepressors, the protein retains the domains necessary for corepressor recruitment and presumably produces disease through similar mechanisms. In APL associated with t(15;17), t(11;17), and t(5;17), leukemogenic retinoic acid receptor  (RAR) fusion proteins associate with Sin3, SMRT, N-CoR, and HDAC1 (4, 12, 14, 15, 18, 29, 47). The responsiveness of these APL variants to differentiation therapy is correlated with the degree to which the fusion proteins surrender corepressor following treatment (12, 14, 18, 29, 47). Oligomerization also plays a role in APL, although in that case the oligomerization domain is present in one fusion partner (PML) while the coregulator binding domain is provided by the other (RAR) (28, 37). PLZF, the RAR fusion protein in t(11;17) APL, has also been shown to interact directly with ETO (35). The leukemogenic fusion proteins, including TEL-AML1 (8) and MYH11-CBF (30), have transcriptional repression functions associated with Sin3 recruitment. In each of these cases, the domains responsible for corepressor interaction are necessary for in vitro activity. Collectively, these reports underscore the importance of transcriptional repression pathways in oncogenesis.

Nevertheless, it is overly simplistic to suggest that AML1 derives its leukemogenic activity solely from interactions with ubiquitous repression pathways. The AML1 regions deleted by the t(8;21) include an activation domain, which binds the histone acetyltransferase p300 (23), and at least two independent repression domains (26, 32). A C-terminal repression domain interacts with members of the TLE/Groucho family (26). Groucho, in turn, can bind the class I HDAC Rpd3 (3). Interestingly, we have recently found that TBL1, a histone-binding protein containing WD40 repeats similar to those of Groucho, is a component of the core SMRT corepressor complex (13). The second AML repression domain utilizes Sin3 and functions through a Groucho-independent mechanism (11, 33, 49). Sin3 is thought to deliver HDAC1 to N-CoR and SMRT (1, 16, 40), which also interact directly with HDAC3 (13) and with class II HDACs 4, 5, and 7 (20, 21). Therefore, pathology resulting from t(8;21) may be the result of the loss of the AML1 activation domain or the replacement of the repression domains by a mistargeted ETO repression domain, which recruits a different repression complex via SMRT and N-CoR. It remains to be determined which repression activities or combinations thereof are responsible for producing leukemia.

The importance of understanding transcriptional repression for the management of disease is already becoming evident. An AML-ETO-transformed cell line is responsive to HDAC inhibitors (50), and recently, therapies targeting HDACs have been initiated in patients with relapsed APL (52). Our present work reveals the essential activity of multimerization in corepressor recruitment and marks the dimerization domain as a potential therapeutic target in the treatment of leukemia associated with AML1-ETO. Undoubtedly, additional targets will present themselves as the mechanisms underlying recruitment of corepressors by fusion protein transcription factors are elaborated.