A Mechanism of Repression by Acute Myeloid Leukemia-1, the Target of Multiple Chromosomal Translocations in Acute Leukemia*

AML1 is one of the most frequently translocated genes in human leukemia. Here we demonstrate that acute myeloid leukemia-1 (AML-1) (Runx-1) represses transcription from a native promoter, p21 Waf1/Cip1 . Unexpectedly, this repression did not require interactions with the Groucho co-repressor. To define the mechanism of repression, we asked whether other co-repressors could interact with AML-1. We demonstrate that AML-1 interacts with the mSin3 co-repressors. Moreover, endogenous AML-1 associated with endogenous mSin3A in mammalian cells. A deletion mutant of AML-1 that did not interact with mSin3A failed to repress transcription. The AML-1/mSin3 association suggests a mechanism of repression for the chromosomal translocation fusion proteins that disrupt AML-1.

AML-1 is a member of the Runt-like transcription factors (Runx-1, -2, and -3), named after the Runt protein that controls segmentation in the Drosophila embryo (11,12). The highly conserved Runt domain is the DNA binding motif in this family of proteins. AML-1 appears to act as an "organizing" factor for many promoters and enhancers by interacting with co-activators and other DNA binding transcription factors (18 -27).
Both Runt and a second Drosophila runt-like protein Lozenge can repress as well as activate transcription (17,28). In addition, when fused to the DNA binding domain of GAL-4, Runx family proteins repressed transcription through GAL4 DNA-binding sites (28). AML-1 interacted with the Groucho co-repressor in yeast two-hybrid assays, and the Groucho binding domain contributed to repression, but it was not sufficient for full activity (28). As well, AML-1 cooperated with LEF-1 to repress transcription of the T-cell receptor ␣ enhancer, and this repression was partially dependent on the Groucho binding domain (29). Thus, AML-1 may act as a master regulator that facilitates the assembly of protein complexes for transcriptional activation or transcriptional repression in a cell typespecific manner.
We have established an assay for AML-1␤-dependent repression using the cyclin-dependent kinase inhibitor p21 Waf1/Cip1 promoter (AML-1B is the largest isoform of AML-1(9)). Repression was not mediated by the Groucho binding motif. This led us to identify an association between AML-1B and the mSin3 co-repressors. An AML-1B protein containing a deletion of the mSin3A binding domain lacked the ability to repress transcription. These results define a mechanism for AML-1-dependent repression.

EXPERIMENTAL PROCEDURES
Cell Culture and Plasmids-C33A cells and Cos-7 cells were maintained in Dulbecco's modified Eagle's medium (BioWhittaker Inc, Walkersville, MD) containing 10% fetal bovine serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM L-glutamine (BioWhittaker). NIH 3T3 cells were maintained in Dulbecco's modified Eagle's medium containing 10% bovine serum. HEL cells were cultured in RPMI 1640 (Bio-Whittaker) containing 10% fetal bovine serum, antibiotics, and L-glutamine. pCMV5-AML-1B and deletion plasmids have been previously described (30). AML-1B-(del 208 -237) and AML-1B-(del 275-314) were created by polymerase chain reaction with the insertion of a BglII restriction site to join the fragments. Each fragment was fully sequenced to confirm that no other mutations were introduced by polymerase chain reaction. The GAL4 fusion cDNAs were prepared by subcloning the indicated fragments, in frame, with the GAL4 DNA binding domain (residues 1-147) using the pCMVM series vectors (31).
Yeast Two-hybrid Assays-AML-1B residues 1-237 were subcloned in frame with GAL4 residues 1-147 in the pAS2 vector. Human Sin3A residues 76 -693 linked to the transcriptional activation domain of GAL4 was constructed by subcloning the blunt-ended ApaI-EcoRV fragment of Sin3A into the SmaI site of pACTII. Sin3A 76 -480 containing PAH domains 1 and 2 was constructed by deleting the 3Ј StuI to BamHI fragment from the previous construct. Protein interactions in PJ69-4A strains were measured by growth on adenine-deficient media and by growth on histidine-deficient media containing 2 mM 3-aminotriazole as described (26).

AML-1B Represses the p21 Waf1/Cip1
Promoter-When the C terminus of Runt or AML-1 family proteins was fused to the GAL4 DNA binding domain, these fusion proteins repressed transcription 3-4-fold (28). To establish an AML-1-dependent repression assay, we tested native promoters containing AML-1-binding sites in various cell types. The p21 Waf1/Cip1 cyclindependent kinase inhibitor promoter contains four perfect AML-1-binding sites and several sites with only one substitution (Fig. 1A). To determine whether AML-1B could repress the p21 Waf1/Cip1 promoter, a 2.4-kilobase pair fragment of the promoter (containing three of the AML-1-binding sites) or 5Ј promoter deletions were linked to the firefly luciferase reporter gene (WWP-luciferase (33)) and transiently transfected with a plasmid expressing AML-1B. We found that AML-1B repressed transcription from the p21 Waf1/Cip1 promoter by 5-10-fold (Figs. 1 and 5). Deletion analysis indicated that sequences 5Ј to the second AML-1-binding site were required for promoter-specific repression (Fig. 1A).
AML-1 Associates with the mSin3 Co-repressors-To determine the mechanism of AML-1B-dependent repression, we tested whether AML-1B interacts with known co-repressors in mammalian cells. Cell lysates from Cos-7 cells transiently expressing AML-1B were immunoprecipitated with two different AML-1 antibodies (␣-RHD and ␣-AML-1-N), and the presence of co-repressors in the immune complexes was determined by immunoblot analysis. Both AML-1 antibodies co-immunoprecipitated the mSin3A co-repressor only when AML-1B was expressed ( Fig. 2A, left panel, last 2 lanes).
To confirm this result, we performed the reciprocal experiment. Because AML-1B co-migrates with the heavy chain, we could not use immunoblotting. Therefore, cells transiently expressing AML-1B were metabolically labeled, and cell lysates were immunoprecipitated with anti-mSin3A in the presence of 0.5% Triton X-100, 0.5% deoxycholic acid, and 0.5% SDS. AML-1B efficiently co-immunoprecipitated with anti-mSin3A (Fig. 2B, middle lane), and the co-immunoprecipitation of AML-1B and mSin3A was blocked by the presence of the immunizing mSin3A peptide (Fig. 2B, last lane).
The amount of radiolabeled mSin3A immunoprecipitated from cells labeled for 3 h was low relative to AML-1B (Fig. 2B, middle lane). This could be due to a slow synthesis rate of the endogenous mSin3A or due to more than one AML-1B molecule associating with mSin3A. To test the former possibility, we performed the same experiment, but we labeled the cells for 30 h (Fig. 2C). Under these conditions, similar amounts of radiolabeled AML-1B and mSin3A were precipitated with anti-mSin3A IgG (Fig. 2C, lane ␣-Sin3A). The ␣-mSin3A antibody did not cross-react with AML-1B in immunoblot analysis and did not immunoprecipitate bacterially produced AML-1. 2 In addition, boiling the cell lysate abrogated the AML-1B/mSin3A co-immunoprecipitation. 2 The interaction of AML-1B with mSin3A was stable under highly stringent conditions (0.5% Triton X-100, 0.5% deoxycholic acid, and 0.5% SDS), but these previous experiments relied on overexpressed AML-1B and endogenous mSin3A. To determine whether the endogenous proteins associate, we immunoprecipitated mSin3A from metabolically labeled HEL cells. The mSin3A antibody recovered a number of associated proteins (Fig. 2D, first lane). Two of these proteins co-migrated with AML-1 proteins immunoprecipitated with anti-AML-1 (Fig. 2D, 1st lane, note that AML-1 migrated as a doublet). AML-1 could be re-immunoprecipitated from the mSin3A immune complexes (Fig. 2D, last lane) but not when the immunizing peptide was added to the lysate prior to antibody addition (Fig. 2D, next to last lane).
The mSin3 protein family includes mSin3A and the homologous mSin3B protein. By having established an interaction with mSin3A, we asked whether AML-1B could also interact with mSin3B. After transient expression of AML-1B and metabolic labeling, AML-1B was co-immunoprecipitated with endogenous mSin3B using a C-terminal (Fig. 2E, last lane) but not an N-terminal antibody (Fig. 2E, middle lane). It is possible that by interacting with mSin3B, AML-1B obscures the anti-N-terminal antibody epitope.
We next localized the AML-1B interaction domain on mSin3A. The N-terminal 237 aa of AML-1B were fused to the GAL4 DNA binding domain and tested for interaction with mSin3A and C-terminal truncations of mSin3A (Fig. 4A) in the yeast two-hybrid system. These AML-1B sequences interacted with mSin3A aa 76 -693 when fused to the GAL4 transcriptional activation domain as measured by growth on adeninedeficient media (Fig. 4, B and C). Deletion of mSin3A residues 480 -693 eliminated the AML-1/mSin3A interaction (Fig. 4, B and C), indicating that these sequences, including the third amphipathic helix (PAH 3), are needed for AML-1 interactions (Fig. 4, B and C). In addition, the interaction of AML-1 aa 1-237 with mSin3A further localizes the AML-1B SID to residues 172-237.
sion-To determine whether the AML-1B SID is required for repression, we tested the panel of AML-1B deletion mutants (Fig. 3A) for repression of the p21 Waf1/Cip1 promoter (Fig. 5A). Deletion of residues 208 -237, which eliminates the interaction between mSin3 and AML-1B in vivo, ablated repression of the p21 Waf1/Cip1 promoter. By contrast, deletion of the C-terminal WRPY motif or deletion of the C-terminal 99 aa had little or no effect on AML-1B-mediated repression (Fig. 5A). However, deletion of aa 290 -387 or 290 -432 also eliminated AML-1Bmediated repression (Fig. 5A). We conclude that the mSin3A interaction is necessary for repression, and a second C-terminal domain contributes to repression of the p21 Waf1/Cip1 promoter by AML-1B. mSin3 proteins repress transcription by directly interacting with the basal transcriptional machinery (38,39) and by linking site-specific DNA-binding proteins to histone deacetylases (40). AML-1B is tightly associated with the nuclear matrix and requires ionic detergents and/or sonication to extract it from cells efficiently. Under these conditions, we were unable to demonstrate consistently an in vivo interaction between AML-1 and histone deacetylases. 2 Therefore, we tested the ability of a histone deacetylase inhibitor, trichostatin A, to affect AML-1B-mediated repression. Importantly, the levels of trichostatin A used in this analysis did not markedly alter expression from the p21 Waf1/Cip1 promoter in NIH 3T3 cells (Fig. 5B). However, trichostatin A impaired AML-1B-dependent repression (Fig. 5B), suggesting that histone deacetylases contribute to AML-1B-mediated repression. DISCUSSION AML-1B is a transcriptional activator of numerous tissuespecific promoters and enhancers. In this report, we show that AML-1B also acts as a transcriptional repressor of the p21 Waf1/Cip1 promoter. AML-1B and mSin3A interacted in yeast two-hybrid assays and by co-immunoprecipitation from mammalian cells. Although this association is observed in multiple assays, and it is maintained in the presence of ionic detergents, it is possible that this interaction is not direct but that it is mediated by an as yet unidentified protein. As well as the AML-1B SID, a second C-terminal region of AML-1B is required for repression (residues 290 -387). This region contains putative protein interaction sites and a nuclear matrix targeting signal (41). Thus, it is likely that other co-repressors contribute to AML-1-mediated repression or that correct subnuclear localization is required.
The mSin3 co-repressors often act in cooperation with accessory proteins such as the nuclear hormone co-repressors N-CoR and SMRT and histone deacetylases (40). We determined that AML-1B (residues 1-237) and N-CoR (residues 1019 -2061) interact in the yeast two-hybrid system, but we have not been able to consistently co-immunoprecipitate AML-1B and N-CoR from mammalian cells. 2 Because residues 1-237 of AML-1B also interact with mSin3 proteins, this association between AML-1 and N-CoR may be indirect. Similarly, we have not been able to co-immunoprecipitate consistently histone deacetylase-1 or -2 with AML-1B. 2 Again, these associations may be indirect. It is also possible that the conditions required to solubilize AML-1B (AML-1B is a component of the nuclear matrix (41)) make it difficult to preserve these weak protein interactions. Although mSin3A can also repress transcription by interacting with TFIIB (38, 39), our results with trichostatin A indicate that histone deacetylases contribute to AML-1Bmediated repression.
AML-1B can function as both a repressor and an activator of transcription. AML-1B is a weak transactivator in C33A, CV-1, and Jurkat cells but cooperates with transcription factors such as C/EBP␣ or Ets-1 to more strongly activate transcription (25,42,43). For the p21 Waf1/Cip1 promoter, we observed a modest transactivation by AML-1B in the K562 myeloid cell line (approximately 7-fold). AML-1-dependent repression is cell typespecific as it represses p21 Waf1/Cip1 in NIH 3T3 cells (Figs. 1 and 5). We have also observed AML-1-dependent repression of the Multidrug Resistance-1 and Rous sarcoma virus promoters in NIH 3T3 cells. 3 Because mSin3 levels are similar between these cell lines, 2 other factors must influence AML-1 transcrip-tional activity.
Whereas AML-1 can both activate and repress transcription in a cell type-specific manner, the AML-1 fusion proteins found in acute leukemia are constitutive repressors (44). The work presented here has implications for the mechanism of action of these chromosomal translocation fusion proteins. The entire SID is retained in the t(12;21), which retains nearly all of AML-1 (6,45). We have shown that the t(12;21) can bind to mSin3 proteins, and this interaction is mediated by the AML-1 SID as well as by the pointed domain in TEL (46). Interestingly, transcriptional repression by TEL/AML-1 was abrogated by deletion of the AML-1 SID and the TEL pointed domain (8,46,47). Although the t(3;21) deletes most of the C terminus of AML-1 in much the same manner as the t(8;21), the t(3;21) breakpoint can retain the entire AML-1 SID (5). We are currently investigating a role for mSin3 in AML/EVI1 transcriptional repression. Coupled with the utilization of the mSin3 co-repressors by the t(8;21), the t(15;17), and t (11;17) in acute promyelocytic leukemia (44), our work suggests that this corepressor may be used by a significant proportion of chromosomal translocation fusion proteins. If so, mSin3A, mSin3B, and their downstream effectors (e.g. histone deacetylases) are 3 S. W. Hiebert, unpublished observations.  Fig. 3. B, analysis was performed in the absence or presence of trichostatin A (TSA). potential therapeutic targets in a large proportion of acute leukemias.