RUNX1 Associates with Histone Deacetylases and SUV39H1 to Repress Transcription

Oncogene (2006) 25, 5777–5786 & 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc ORIGINAL ARTICLE RUNX1 associates with histone deacetylases and SUV39H1 to repress transcription

E Reed-Inderbitzin1,8,9, I Moreno-Miralles1,9, SK Vanden-Eynden1,9, J Xie2, B Lutterbach1,10, KL Durst-Goodwin1,11, KS Luce1, BJ Irvin1,12, ML Cleary3, SJ Brandt2,4,5,6,7 and SW Hiebert1,6

1Department of Biochemistry, Vanderbilt University School of Medicine, Nashville, TN, USA; 2Department of Medicine, Vanderbilt University School of Medicine, Nashville, TN, USA; 3Stanford University Medical School, Palo Alto, CA, USA; 4Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, TN, USA; 5Cancer Biology, Vanderbilt University School of Medicine, Nashville, TN, USA; 6Vanderbilt-Ingram Cancer Center, Vanderbilt University School of Medicine, Nashville, TN, USA and 7Tennessee Valley VA Healthcare System, Nashville, TN, USA

RUNX1 (AML1) is a gene that is frequently disrupted by Introduction chromosomal translocations in acute leukemia. Like its Drosophila homolog Runt, RUNX1 both activates and Translocations that disrupt the RUNX1 gene are among represses transcription. Both Runt and RUNX1 are the most common aberrations found in human leuke- required for gene silencing during development and a mia. The t(8;21) is found in 10–15% of myeloid central domain of RUNX1, termed repression domain 2 leukemia and gives rise to a fusion protein that contains (RD2), was defined as being required for transcriptional the N-terminal portion of RUNX1 fused to nearly all of repression and for the silencing of CD4 during T-cell myeloid translocation gene on chromosome 8 (MTG8, maturation in thymic organ cultures. Although transcrip- also known as eight-twenty-one (ETO)) (Miyoshi et al., tional co-repressors are known to contact other repression 1991, 1993; Erickson et al., 1994). The fusion protein domains in RUNX1, the factors that bind to RD2 had not appears to function as a transcriptional repressor of been defined. Therefore, we tested whether RD2 contacts RUNX1-regulated genes (Peterson and Zhang, 2004). histone-modifying enzymes that may mediate both repres- The t(12;21) is found in up to 25% of pediatricB-cell sion and gene silencing. We found that RD2 contacts acute leukemia and creates a chimeric gene encoding the SUV39H1, a histone methyltransferase, via two motifs TEL–RUNX1 fusion protein (Golub et al., 1995; and that endogenous Suv39h1 associates with a Runx1- Nucifora et al., 1995; Raynaud et al., 1995; Romana regulated repression element in murine erythroleukemia et al., 1995). TEL–RUNX1 also appears to act as a cells. In addition, one of these SUV39H1-binding motifs is transcriptional repressor that dominantly interferes with also sufficient forbinding to histone deacetylases 1 and 3, RUNX1-specific transactivation (Hiebert et al., 1996; and both of these domains are required for full RUNX1- Chakrabarti and Nucifora, 1999; Fenrick et al., 2000; mediated transcriptional repression. The association Guidez et al., 2000). RUNX1 function is also impaired between RUNX1, histone deacetylases and SUV39H1 by the inv(16), which fuses the RUNX1 associating provides a molecular mechanism for repression and factor, core binding factor b (CBFb or polyoma possibly gene silencing mediated by RUNX1. enhancer binding protein 2 beta) to the smooth muscle Oncogene (2006) 25, 5777–5786. doi:10.1038/sj.onc.1209591; myosin heavy-chain gene MYH11, in approximately 8% published online 1 May 2006 of acute myeloid leukemia (Liu et al., 1993). Like the t(8;21) and the t(12;21), the inv(16) also encodes a Keywords: Suv39h1; RUNX1; AML1; histone; deacetylase transcriptional repressor, but this factor must associate with RUNX1 to regulate transcription (Lutterbach et al., 1999; Durst et al., 2003). RUNX1 is a member of the mammalian Runt domain Correspondence: Dr SW Hiebert, Department of Biochemistry, Vanderbilt-Ingram Cancer Center, Vanderbilt University School of family of transcription factors, which were recently Medicine, Preston Research Building, Rm 512, 23rd and Pierce, renamed the RUNX factors after the Runt protein that Nashville, TN 37232, USA. controls segmentation in the Drosophila embryo (Ger- E-mail: [email protected] gen and Wieschaus, 1985; van Wijnen et al., 2004). Runt 8Current address: University of Oregon Health Sciences Center, either activates or represses transcription in a promoter- Beaverton, OR, USA. 9These authors contributed equally to this work. or enhancer-specific context and cell type-specific 10Current address: Merck Inc., West Point, PA, USA. manner, but more frequently it is required for repressing 11Current address: Indiana University School of Medicine, Indiana- transcription (Wheeler et al., 2002; Swantek and polis, IN, USA. Gergen, 2004). For instance, during embryogenesis 12Current address: The Blood Center of Wisconsin, Milwaukee, WI, USA. the establishment of Runt-mediated repression of Received 19 April 2005; revised 21 February 2006; accepted 28 February engrailed is dependent on the dosage of the zinc-finger- 2006; published online 1 May 2006 containing transcription factor Tramtrack. However, RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5778 the maintenance of repression of engrailed is dependent RUNX1 and RUNX3 may associate with factors that on the presence of the Groucho and dCtBP co-repressors regulate the ‘histone code’ (Durst et al., 2003). The and the Rpd3 histone deacetylase (Wheeler et al., 2002). histone code hypothesis suggests that deacetylation of Thus, Runt acts in a context-dependent manner, key lysine residues in the conserved tails of histones is activating or repressing gene expression depending on followed by methylation that allows the association of the constitution of a particular promoter/enhancer or silencing factors such as heterochromatin protein-1 the availability of a given co-factor in a specific cell type. (HP1), which binds to methylated histone H3 (Firestein Likewise, RUNX1 acts as an ‘organizing’ factor for et al., 2000; Jacobs et al., 2001; Jenuwein and Allis, many promoters and enhancers. RUNX1 is a poor 2001; Lachner et al., 2001). Therefore, we asked whether activator of transcription when expressed alone or when SUV39H1, a histone H3 lysine 9-specific methyltrans- tested using artificial promoters containing multiple ferase (Rea et al., 2000), and histone deacetylases RUNX-binding sites. However, in concert with other contact RUNX1 within RD2 or other domains that DNA-binding factors, RUNX family members coop- are known to be required for transcriptional repression erate to activate transcription (Zhang et al., 1996; and gene silencing. We found that both SUV39H1 and Britos-Bray and Friedman, 1997; Goetz et al., 2000; histone deacetylase-1 and -3 contacted RUNX1 within Gu et al., 2000; Cameron and Neil, 2004). By contrast, RD2. Deletion of either of the two SUV39H1 contact RUNX1 is a potent transcriptional repressor, even when sites or the HDAC-binding motif within repression expressed alone, albeit in a cell type-specific manner. In domain 2 of RUNX1 significantly impaired RUNX1- fact, three distinct domains, termed repression domains mediated transcriptional repression. In addition, 1, 2 and 3, are required for RUNX1-mediated repression RUNX1 and RUNX3, but not Runx2, associated with (Aronson et al., 1997; Lutterbach et al., 2000). RUNX1 SUV39H1 and Runx1 recruited Suv39h1 to an endo- recruits the mSin3A co-repressor through a domain genous target gene, which raises the possibility that adjacent to the DNA-binding domain (repression recruitment of histone methyltransferases contributes to domain 1) (Lutterbach et al., 2000). In addition, RUNX1-mediated silencing of gene expression. RUNX1 contacts Groucho family members to repress transcription in a promoter-dependent manner through a C-terminal motif termed repression domain 3 Results (Aronson et al., 1997; Levanon et al., 1998; Javed et al., 2000; Lutterbach et al., 2000; Nishimura et al., 2004). RUNX1 and RUNX3 associate with SUV39H1 Repression domain 2 was simply defined as an As RUNX1 and RUNX3 both contribute to gene approximately 100 amino-acid sequence that was silencing of CD4 at distinct stages of T-cell development required for repression (Lutterbach et al., 2000). (Taniuchi et al., 2002), we tested whether RUNX family In addition to transcriptional repression, genetic members could associate with SUV39H1. Each RUNX studies have implicated Runx1 and Runx3 in the family protein was co-expressed with epitope-tagged silencing of gene transcription. During T-cell develop- SUV39H1 (FLAG-SUV39H1), SUV39H1 was immu- ment, CD4 and CD8 are silent in immature T-cells noprecipitated, and the co-purifying RUNX proteins (CD4À/CD8À), and after these cell surface markers detected by immunoblot analysis (Figure 1a). Both expressed either are CD4 or CD8 are silenced as T-cells RUNX1 and RUNX3, which are required for CD4 move from the ‘double-positive’ to the ‘single-positive’ silencing, associated with SUV39H1 (Figure 1a). By stage (e.g. CD4À/CD8 þ ). Using site-directed mutagen- contrast, Runx2, which does not contribute to CD4 gene esis of the CD4 silencer elements in transgenic mouse silencing, did not co-purify with SUV39H1 under these studies, binding sites for Runx family members were conditions (Figure 1a). In addition, using these same found to be necessary for silencing CD4 expression. conditions we were unable to detect an association Moreover, mice lacking expression of Runx1 in T-cells between RUNX1 and HP1, which associates with failed to establish CD4 silencing in immature T-cells, SUV39H1 (Lachner et al., 2001) (data not shown), whereas mice lacking Runx3 failed to maintain silencing suggesting that SUV39H1 contacts RUNX1 and of CD4 in cytotoxic T-cells (Taniuchi et al., 2002). RUNX3 specifically. Conversely, Runx2 was not required for the silencing of Independently, this same interaction was identified, CD4, even though it may be expressed in T-cells. Thus, but it was found that when SUV39H1 was over- there may be differences between the factors that the expressed, it contacted the Runt domain and inhibited Runx family members recruit to repress and/or silence the ability of RUNX1 to bind to DNA in vitro, thus transcription. creating an apparently inactive complex (Chakraborty The development of an in vitro-based system for et al., 2003). Given that the Runt domains of the RUNX studying the mechanism of RUNX1-mediated silencing family members are nearly identical, the association of CD4 allowed the demonstration that repression between SUV39H1 and RUNX1 and RUNX3, but not domains 1 and 3 were dispensable for silencing of RUNX2, suggested that a domain(s) outside the Runt CD4, but that repression domain 2 (RD2) was required domain must contribute to or mediate the association for silencing of CD4 (Telfer et al., 2004). Given the roles with SUV39H1. In addition, the levels of SUV39H1 are of RUNX family members in both establishment and very low in many cell types that express RUNX1 or maintenance of repression, as well as gene silencing RUNX3 at high levels. Thus, it seemed possible that during T-cell development, we hypothesized that there would not be enough SUV39H1 in most cell types

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5779

a V a 1 480 RUNX1 RHD RD1 RD2 1 RD3 469 ∆469-480 RUNX1+SURUNX1 RUNX1+SUVRUNX1 RUNX3+SUVRUNX3RUNX3+SUVRUNX3 RUNX2+SUVRUNX2 RUNX2+SUVRUNX2 1 290 351 381 1-290/351-381 1 290 432 480 Anti- ∆290-432 RUNX1,2,3 b n WCL IP WCL IP WCL IP o

-480 NX1 C U UNX1 469 290-432 290/351-381 b R R ∆ ∆ 1-

Jurkat RUNX1NE DBSControlDBS IP Anti-RUNX1 *

α-SUV39H1

α-RUNX1 WCL * Anti-RUNX1

Figure 1 RUNX1 and RUNX3, but not RUNX2, associate with SUV39H1. (a) RUNX1 and RUNX3 bind to SUV39H1. RUNX1, RUNX2, or RUNX3 were co-expressed with or without FLAG- Figure 2 Repression domain 2 of RUNX1 contributes to the SUV39H1. 40 h after transfection, cell lysates were prepared in PBS recruitment of SUV39H1. (a) Schematic diagram of the RUNX1 containing 0.5% Triton X-100, 0.1% sodium deoxycholate, 0.1% deletion mutants used in (b). Repression domains (RD)1, 2, and 3 sodium dodecyl sulfate (SDS), and SUV39H1 was immunopreci- represent the three known repression domains within RUNX1. pitated with anti-FLAG agarose beads. Co-purifying RUNX RHD, the runt homology domain that mediates DNA binding. proteins were detected with antibodies specific to each family (b) RUNX1 or the indicated mutant was co-expressed with member. (b) RUNX1 associates with SUV39H1 in Jurkat T cells. SUV39H1 in Cos7 cells and immuno-purified using anti-FLAG. RUNX1 was purified using DNA affinity chromatography and the RUNX1 was detected by immunoblot analysis using an antibody eluate analysed for RUNX1 or SUV39H1 by immunoblot analysis. directed against the N-terminus, which detects all of the mutants. Control, a DNA binding site for T cell factors was used as a non- specific control. might contact RUNX1 within the C-terminal repression domains by using previously created C-terminal deletion to inhibit RUNX1 DNA-binding functions. Therefore, mutants (Figure 2a) (Lutterbach et al., 2000). These we asked whether additional complexes between deletion mutants all retain the Runt domain, which is RUNX1 and SUV39H1 might be present in cells that required for DNA binding and nuclear localization. could contact DNA and mediate transcriptional repres- Each RUNX1 mutant was co-expressed with FLAG- sion. We used DNA affinity chromatography to purify SUV39H1 and cell lysates were immunoprecipitated endogenous RUNX1 from Jurkat T cells and asked with anti-FLAG crosslinked to sepharose beads to whether SUV39H1 co-purified. In numerous experi- avoid interfering immunoglobulin heavy-chain reactiv- ments we consistently observed more SUV39H1 in the ity with the secondary antibody. As a control, RUNX1 eluate from the RUNX DNA affinity column than was expressed in the absence of SUV39H1 (Figure 2b, from a non-specific DNA control column (Figure 1b). lane labeled RUNX1 Con). RUNX1 was immunopre- However, there was often background contamination cipitated using anti-FLAG, only when FLAG- from our control DNA-binding site column, possibly SUV39H1 was co-expressed (Figure 2b). In addition, reflecting the ability of SUV39H1 to associate with deletion of residues 469–480 (RD3) did not affect the many DNA-binding factors that contact DNA non- association with SUV39H1. However, a deletion of specifically (e.g. histones). RD2, residues 290–432, appeared to significantly reduce the association between RUNX1 and SUV39H1 (Figure 2b), although it retained some ability to co- Identification of the RUNX1: SUV39H1 interaction purify, likely due to the association between SUV39H1 domain and the RUNX1 DNA-binding domain (Chakraborty The observation that SUV39H1 might contact RUNX1 et al., 2003). Thus, RUNX1 may contact SUV39H1 in a manner that allowed RUNX1 to still contact DNA through multiple domains, one of which lies within and the finding that RUNX2 failed to associate with a previously identified repression domain (residues SUV39H1 encouraged us to ask whether SUV39H1 290–432) (Lutterbach et al., 2000).

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5780 The possibility of multiple contact sites in RUNX1 Next, we sought to define the motifs in the C-terminus for recruitment of SUV39H1 diminishes the effective- of RUNX1 that are sufficient to contact SUV39H1. ness of deletion-mapping studies. Therefore, we created GAL4 fusion proteins were constructed that span the a series of GAL4–RUNX1 fusion proteins to define the region between residues 210 and 480 of RUNX1 domains in RUNX1 that are sufficient for associating (Figure 3a) and tested for SUV39H1 recruitment by with SUV39H1 (Figure 3a). The GAL4 DNA-binding co-purification as described above. RUNX1 residues domain is especially useful for these studies as it 210–290, 290–351 and 432–480 failed to associate with functions as an epitope tag and contains a nuclear SUV39H1 under these conditions (Figure 3b, middle localization signal. The initial deletion studies suggested and right-hand panels). By contrast, sequences that were the presence of SUV39H1-binding sites in both the previously defined as required for RUNX1-dependent N-terminal and C-terminal portions of RUNX1. There- transcriptional repression (residues 351–381 and 380– fore, GAL4–RUNX1 fusion proteins spanning the first 430) were sufficient to co-purify with SUV39H1 250 amino acids (aa) of RUNX1 were tested by co- (Figure 3b, middle and right panels, note that GAL4- expression with FLAG-SUV39H1, immunoprecipita- 290–351 migrates above the light-chain band, lower tion, and immunoblot analysis to detect the purified panel, to allow a clear determination of binding). The proteins (Figure 3b, left-hand panels). As expected, the observation that residues 351–381 could contact GAL4 DNA-binding domain did not associate with SUV39H1 was consistent with the initial deletion SUV39H1 and served as a negative control. Also, analysis in which the 1–290/351–381 mutant associated residues 210–250 of RUNX1 were not sufficient to with SUV39H1 (Figure 2b). Thus, at least four motifs of mediate an association with SUV39H1. By contrast, RUNX1 can contact SUV39H1. fusion proteins containing either residues 1–170 or residues 172–212 of RUNX1 were sufficient to co-purify with SUV39H1. SUV39H1 associates with RUNX1 through multiple domains The observation that multiple domains of RUNX1 contact SUV39H1 raised the possibility that more than a 1 RHD 480 one domain of SUV39H1 might contact RUNX1. To RUNX1 GAL4 RD1 RD2 RD3 test this possibility, we fused fragments of SUV39H1 1-170 GAL4 * 172-250 GAL4 * to the GAL4 DNA-binding domain (Figure 4a) and 172-212 GAL4 * co-expressed these epitope-tagged proteins with 210-250 GAL4 432-480 GAL4 RUNX1. After immunoprecipitation using anti-GAL4, 351-381 GAL4 * co-purifying RUNX1 was detected by immunoblot 210-290 GAL4 analysis. Both N-terminal (DSet) and C-terminal do- 172-349 GAL4 * 290-351 GAL4 mains (242–412) of SUV39H1 were sufficient to contact 380-432 GAL4 * RUNX1 (Figure 4b, upper right-hand panel), which is consistent with multiple contact sites for SUV39H1 on b RUNX1. However, we consistently observed that

50 49 80 -480 -381 -351 -432 deletion of the first 104 aa of SUV39H1 impaired the 2 1 0-290 0 0 GAL4GAL4 Con RUNX1RUNX1 Con 1-170 172-250172-212210-250 43 35 21 172-2 172-3 29 38 432-4 association with RUNX1 (Figure 4b, third lane, upper right-hand panel). Thus, it is possible that the N- terminus contains sequences that contribute to RUNX1 IP Anti-GAL4 binding or that this mutant assumes a conformation that IgG does not associate as well with RUNX1. 380-432 Next, we attempted to extend this mapping to an in vitro system. We expressed GST–SUV39H1 fusion proteins in bacteria and used glutathione agarose beads containing these fusion proteins to purify 35S-labeled RUNX1 or RUNX1a (an isoform containing essentially WCL Anti-GAL4 the first 250 aa of RUNX1) that were synthesized by in vitro transcription and translation. Both RUNX1 and RUNX1a associated with GST–SUV39H1 when compared to GST alone. Although we consistently observed Figure 3 Defining the SUV39H1 contact sites in RUNX1. some background with RUNX1 on the GST-containing (a) Schematic diagram of the RUNX1 GAL4-fusion proteins used beads (Figure 4c), more of the 480 aa isoform as in (b). The * in the diagram indicate the fragments that were sufficient to contact SUV39H1. RD1, RD2, and RD3 represent the compared to the 250 aa isoform (RUNX1a) was three known repression domains within RUNX1. (b) Identification consistently precipitated from the lysates by GST– of SUV39H1 binding sites in the C-terminus of RUNX1. The SUV39H1. However, the majority of this association GAL4 DNA binding domain (residues 1–147), GAL4–RUNX1 or was lost when GST-DSet or GST-243–412 were used in the indicated mutants were co-expressed with SUV39H1 in Cos7 cells and immuno-purified using anti-FLAG beads. The GAL4– the binding assay (Figure 4c), although we did observe RUNX1 fusion proteins were detected by immunoblot analysis somewhat more binding of RUNX1 to GST-DSet over with anti-GAL4. GST or GST-242–412. Upon longer exposure, we also

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5781

a 1 SET 412 SUV39H1 GAL4 1 250 ∆Set GAL4 GAL4 HDAC1290-351351-381380-432 432-480 GAL4 HDAC3 290-351 351-381380-432432-480Gal-Runx1 1 147 242 412 ∆Pre GAL4 104 412 104-412 GAL4 242 412 242-412 GAL4 Anti-GAL4 (FLAG IP) b 2 2 IgG pre set pre set Con SUV104-41∆ ∆ 242-412 Con SUV104-41∆ ∆ 242-412

αRUNX1

IgG α GAL4 Anti-GAL4 (WCL)

WCL IP c

ControlRUNX1RUNX1aControlRUNX1RUNX1aControlRUNX1RUNX1aControlRUNX1RUNX1aControlRUNX1RUNX1a Anti-Flag (FLAG IP) HDAC1 HDAC3 Figure 5 Definition of the sites in RUNX1 that contact HDAC1 and HDAC3. The GAL4 DNA binding domain (residues 1–147), GAL4–RUNX1 or the indicated mutants (see Figure 3a) were co- expressed with FLAG-HDAC1 (left panel) or FLAG-HDAC3 (right panel) in Cos7L cells and immuno-purified using anti-FLAG beads. The GAL4-tagged fusion proteins were detected by TnT GST SUV ∆ Set 243-412 immunoblot analysis with anti-GAL4. Figure 4 SUV39H1 contains at least 2 domains that contact RUNX1. (a) Schematic diagram of the GAL4-SUV39H1 fusion screening these factors against the collection of GAL4– proteins used in (b). The C-terminal SET domain that mediates methyltransferase activity is indicated by a shaded box. (b) The RUNX1 fusion proteins. Both HDAC1 and HDAC3 GAL4 DNA binding domain (residues 1–147), GAL4-SUV39H1 associated with residues 380–432 when linked to GAL4, or the indicated mutants were co-expressed with RUNX1 in Cos7L but not with residues 351–381, 290–350 or 432–480 cells and immuno-purified using anti-GAL4 beads. Co-purifying (Figure 5). Given that this motif also was sufficient to RUNX1 was detected by immunoblot analysis with anti-RUNX1. (c) RUNX1 binds to SUV39H1 in vitro. RUNX1 or the 250 aa bind to SUV39H1, this sequence may contain two isoform RUNX1a were transcribed and translated in reticulocyte binding sites for histone-modifying enzymes, or lysates and affinity purified using glutathione-agarose beads SUV39H1 and HDACs may form a complex that can containing the GST–SUV39H1 fusion proteins indicated below contact this domain. each panel. TnT, unfractionated transcribed and translated Given that the sequences between RUNX1 residues RUNX1 proteins. 350–432 within RD2 contain two motifs that are sufficient for contacting SUV39H1, one of which binds observed a small amount of binding of RUNX1a to both HDAC1 and HDAC3, we created smaller deletions both of the SUV39H1 deletion mutants, which appeared within this region to further explore the function of to be specific, as there was no background in the these subdomains. As SUV39H1 contacts residues corresponding GST lane (data not shown). 351–381 and this sequence lies within a region of RUNX1 that is required for silencing of CD4 in thymic Histone deacetylases also contact RD2 organ cultures (Telfer et al., 2004), we deleted this region In addition to the association between SUV39H1 and up to residue 387 to match the mutation assessed in the RUNX1, RUNX1 can bind to HDAC1, HDAC3 and thymicorgan culture studies ( D347–387, Figure 6a) HDAC9 and to a lesser degree HDAC2, HDAC5 and (Telfer et al., 2004). In addition, we deleted the domain HDAC6 (Durst et al., 2003). Therefore, we tested that was sufficient to contact all three of these factors whether HDAC1 or HDAC3 might also contact the (D379–430). When these mutants were co-expressed with motifs that are required for repression within RD2 by HDAC1 or HDAC3, co-immunoprecipitation analysis

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5782 indicated that D347–387 retained the ability to bind to contains multiple RUNX1-binding sites (Lutterbach HDAC3 as expected, but that HDAC1 was severely et al., 1999; Durst et al., 2003). RUNX1 was able impaired in its ability to associate with this mutant to repress this promoter up to five-fold in the absence (Figure 6b, upper panels). By contrast, neither HDAC of its co-factor CBFb, and the co-expression of was able to bind to the mutant lacking residues 379–430. CBFb stimulated RUNX1-mediated repression to nearly For SUV39H1, our initial analysis suggested that the 10-fold (Figure 6d). By contrast, deletion of residues 351–381 motif of RUNX1 required higher levels of 347–387 or 379–430 impaired RUNX1-mediated repres- expression to co-purify with SUV39H1 than did residues sion (Figure 6d). The inclusion of CBFb in these assays 380–432 (Figure 2). Consistent with this observation, the did augment the action of these deletion mutants, but D379–430 mutant was severely impaired for binding to they were still impaired relative to the wild-type protein. SUV39H1 (Figure 6c), whereas 347–387 was still able to Thus, each subdomain that contacts SUV39H1 and/or associate with SUV39H1 to some degree. In this assay, HDAC1 or HDAC3 is required for RUNX1-mediated we also included an additional mutation, which deletes repression. the region between residues 290 and 387 (D290–387). This mutant, similar to D347–387, was impaired in association with SUV39H1, but not to the degree of Suv39h1 associates with a RUNX1-regulated repression 379–430. element To begin to dissect the function of the subdomains The observation that RUNX1 is required for gene that contact HDACs and SUV39H1, we performed silencing (Taniuchi et al., 2002), together with the reporter assays using the p21 cyclin-dependent kinase established role of SUV39H1 in transcriptional silen- inhibitor promoter linked to firefly luciferase, which cing, and our observation that SUV39H1 can be purified

a c

3147-387379-430290-387 347-387379-430290-387 RUNX∆ ∆ ∆ RUNX1∆ ∆ ∆ Suv39H1 + + + + + + + + 1RHD 480 RUNX1

∆347-387 RUNX1 * * ∆379-430

347-387379-430 347-387379-430 RUNX1 ∆ ∆ RUNX1 ∆ ∆ HDAC1 + + + + + + HDAC3 Suv

Anti-RUNX1

WCL FLAG IP

d 10.0 9.0 Anti-FLAG 8.0 7.0 6.0 5.0 FLAG IP FLAG IP 4.0 3.0 2.0 Fold repression 1.0 Anti-RUNX1 0.0

Con RUNX1290-387347-387379-430RUNX1290-387347-387379-430 WCL WCL ∆ ∆ ∆ ∆ ∆ ∆ +CBFß Figure 6 Both SUV39H1 and HDAC contact sites in repression domain (RD)2 are required for RUNX1-mediated repression. (a) Schematic diagram of the RUNX1 deletion mutants used. (b) Deﬁnition of sequences required for RUNX1 association with HDAC1 and HDAC3. RUNX1, D347–387,or D379–430 were expressed with the indicated FLAG-tagged HDAC for co- immunoprecipitation analysis and immunoblot assays as indicated beside each panel. (c) Deﬁnition of RUNX1 C-terminal sequences required for RUNX1 association with SUV39H1. RUNX1, D347–387, D379–430, or D290–387 were expressed with FLAG-SUV39H1 for co-immunoprecipitation analysis and immunoblot assays as indicated beside each panel. (d) RUNX1 residues 347–387 and 379–430 are required for RUNX1-dependent transcriptional repression. Transcription repression assays were performed using NIH 3T3 cells transfected with the WWP-Luc p21-promoter-luciferase reporter (el-Deiry et al., 1994) and 300 ng of pCMV5-RUNX1 or the indicated mutant in the absence or presence of 100 ng of pCMV5-core binding factor (CBF)b. pCMV-secreted alkaline phosphatase (SEAP) was used as an internal control. The normalized light units are shown from a representative experiment performed in triplicate. Error bars indicate standard deviation.

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5783 using RUNX1 DNA-binding sites (Figure 1b), sug- Discussion gested that Suv39h1 may be recruited to RUNX1- regulated genes to mediate repression and/or gene RUNX1 and RUNX3 both activate and repress silencing. Although CD4 would be an ideal target to transcription depending on the cellular or gene context. assess the recruitment of Suv39h1, an appropriate cell During T-cell development, these factors are required model is not available for this analysis. However, the for the silencing of CD4 (Taniuchi et al., 2002). These Band 3 upstream regulatory element (B3URE) contains factors can associate with the mSin3 and Groucho co- five consensus RUNX1-binding sites and mediates repressors, but it appears that repression domain 2 is RUNX1-dependent repression of Band 3 in murine critical for RUNX1-mediated silencing of CD4 (Telfer erythroleukemia (MEL) cells (J Xie and SJ Brandt, in et al., 2004). Thus, our identification of two SUV39H1 preparation). Therefore, we asked whether Suv39h1 contact points within repression domain 2 of RUNX1, could associate with this control region using biotiny- with one of these RUNX1 domains also contacting lated double-stranded oligonucleotides derived from the HDAC1 and HDAC3, begins to provide a mechanistic B3URE, which contain two of the RUNX1-binding basis for gene silencing mediated by RUNX1. The sites. We found that both RUNX1 and CBFb were recent identification of Band 3 as a target for repression enriched when wild-type sequences were used, with only by RUNX1 (J Xie and SJ Brandt, in preparation) has modest or no binding when the RUNX1 sites were allowed us to confirm that Suv39h1 is associated with a mutated (Figure 7a). Although some non-specific bind- silencing element that is controlled by RUNX1. ing of Suv39h1 was observed, endogenous Suv39h1 was Co-repressors often associate with target proteins greatly enriched using the B3URE-containing wild-type through multiple interaction domains and it is often the RUNX1-binding sites (Figure 7a). Finally, we con- case that each contact point with the target factor is firmed that Suv39h1 associates with the B3URE in MEL critical for full activity. This may be due to the need to cells in which Band 3 expression is suppressed using create the correct context, conformation or spacing for chromatin immunoprecipitation. Whereas control IgG the associated histone-modifying enzymes to function. did not immunoprecipitate the B3URE, anti-Suv39h1 For example, ETO/MTG8 contains multiple contact was able to dramatically enhance the amount of B3URE sites for mSin3, nuclear hormone co-repressors and immunoprecipitated (Figure 7b and c), indicating that histone deacetylases (Lutterbach et al., 1998b; Amann Suv39h1 associates with this regulatory element. et al., 2001; Hildebrand et al., 2001), but mutation of only one site can significantly impair RUNX1-ETO- mediated repression (Lutterbach et al., 1998b; Amann et al., 2001). In fact, a completely inactive form of RUNX1-ETO retains the ability to bind to the nuclear hormone co-repressor N-CoR and histone deacetylase-1 (Lenny et al., 1995; Amann et al., 2001). Our mutagen- esis study indicates that elimination of residues 379–430 or 347–387 impaired RUNX1-mediated repression. While the deletion of residues 379–430 impaired the association of SUV39H1, HDAC1 and HDAC3, the deletion of residues 347–387 did not greatly impact binding to HDAC3, but affected HDAC1 association (Figure 6). It is notable that the D347–387 and D379–430 deletions overlap by 7 aa, which might suggest that HDAC1 binds to the N-terminal portion of the 380–432 fragment (Figure 5). Nevertheless, we conclude that simply associating with co-repressors and histone- Figure 7 SUV39H1 associates with the upstream regulatory modifying enzymes is not sufficient for RUNX1- region of the Band 3 upstream regulatory element (B3URE) by mediated repression and perhaps gene silencing, because interacting with Runx1 both in vitro and in vivo.(a) DNA affinity all of these mutants retain the ability to interact with purification of Runx1 and Suv39h1 using the B3URE. Biotinylated mSin3 and Groucho (Aronson et al., 1997; Lutterbach double stranded oligonucleotides containing wild type or mutated RUNX1 DNA binding sites from the B3URE were incubated with et al., 2000) and their associated histone-modifying nuclear extract from murine erythroleukemia (MEL) cells, and enzymes. precipitated with streptavidin agarose beads. Bound RUNX1, core Thymicorgan cultures were used to determine that binding factor (CBF)b, and Suv39h1 were detected by Western blot silencing of CD4 required RUNX1 sequences between analysis. (b) ChIP analysis of Suv39h1 binding to the B3URE in residues 290 and 387, but also indicated that the MEL cells. Ethidium bromide staining of the agarose gel-separated polymerase chain reaction (PCR) products amplified after 31 cycles C-terminal 90 aa as well as residues 239–289 contributed by using templates from input DNA, control IgG and SUV39H1- to silencing (Telfer et al., 2004). This latter domain is antibody precipitated DNA. The number of cycles was selected adjacent to sequences that are required for RUNX1 to from within the linear range as judged by real-time PCR using SyBr contact mSin3A, but deletion of residues 180–210 in Green. (c) Quantitation of Suv39h1 occupancy at B3URE in MEL cells by real-time PCR. Data was normalized as a percentage of murine Runx1, which is analogous to the deletion of input, and presented as the average of triplicates with standard residues 208–237 in human RUNX1 that impaired deviations shown as error bars. mSin3A association (Lutterbach et al., 2000), had little,

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5784 if any, effect on silencing of CD4 (Telfer et al., 2004). Thus, it is possible that SUV39H1 competes for these However, the sequences that mediate the association sites on RUNX1 with activating factors and provides a with mSin3A have not been pinpointed and it remains molecular switch for turning RUNX1 into a transcrip- possible that disruption of residues 239–289 may also tional repressor. Alternatively, RUNX1 may have affect the association with this co-repressor as phos- promoter/enhancer-specific functions dependent on the phorylation of residues in this domain affected mSin3A context of the RUNX1-binding sites. In addition, our association (Lutterbach et al., 2000; Imai et al., 2004). data raise the possibility that RUNX1 forms two distinct Also, the initial studies were performed in high complexes with SUV39H1, one that contacts DNA to concentrations of detergent (Lutterbach et al., 2000), silence gene expression in addition to the complex that leaving open the possibility that mSin3A may bind a cannot bind to DNA (Chakraborty et al., 2003). second domain within RUNX1. Alternatively, differ- By defining multiple binding sites within RUNX1 for ences between the murine and human RUNX1 may SUV39H1, our work also implicates SUV39H1 as an account for differences in the requirement of these important effector of transcriptional repression or gene sequences for RUNX1 action. Further analysis of silencing mediated by the chromosomal translocation mSin3A binding will be required to determine whether fusion proteins generated by the t(12;21), t(8;21), and the it contributes to silencing of CD4 or whether another co- inv(16). The t(12;21) fusion protein retains nearly all of factor also is required for RUNX1 to act. RUNX1 and, therefore, was predicted to contact In our analysis, residues 380–432 (within the C- SUV39H1. We have confirmed that it does indeed terminal 100 aa) were sufficient for binding to associate with SUV39H1. However, TEL contributes SUV39H1, HDAC1 and HDAC3, and deletion of these repression domains to this fusion protein, which limits sequences impaired the association with these factors our ability to determine whether SUV39H1 recruitment and repression. Given the tight links between histone contributes to repression (Reed-Inderbitzin and Hiebert, deacetylation, histone methylation of H3 lys 9 and gene unpublished data). RUNX1 residues 1–204 are also silencing (Jenuwein and Allis, 2001), it is somewhat retained in the t(8;21) fusion protein, which again unexpected that deletion of this domain only modestly predicts that this fusion protein can recruit SUV39H1, affected silencing of CD4 (Telfer et al., 2004). However, although this association may negatively affect DNA given that removal of the C-terminal 90 aa only removes binding (Chakraborty et al., 2003). Although the inv(16) one SUV39H1-binding site and that this mutant retains contains no homology to RUNX1, it acts as a co- the ability to recruit mSin3A and its associated histone repressor for RUNX1 (Lutterbach et al., 1999; Durst deacetylases, it is possible that, in overexpression assays, et al., 2003). Our initial analysis failed to detect an this mutant retains sufficient activity to silence CD4. association between SUV39H1 and the inv(16) fusion In addition to residues 380–432, a second motif within protein. However, it remains possible that this fusion repression domain 2 that was sufficient to contact protein stabilizes the interaction between RUNX1 and SUV39H1 and that was required to bind to HDAC1 is SUV39H1. Given the tight links between histone residues 351–381, which were identified as critical for deacetylation, methylation and gene silencing, it is RUNX1-mediated gene silencing (Telfer et al., 2004). possible that inhibitors of this histone methyltransferase While the simple interpretation of this result is that and/or histone deacetylases may have therapeutic poten- SUV39H1 and perhaps HDAC1 recruitment are critical tial for the leukemia containing disruptions of RUNX1. for RUNX1-mediated repression, and possibly silencing of CD4, this motif is also sufficient to target a Materials and methods heterologous protein to the nuclear matrix (Kanno et al., 1998; Zeng et al., 1998). The dual function for this Cell culture and plasmids motif clouds the interpretation of this result, but this Cos-7 cells were maintained in Dulbecco’s modified Eagle domain does not appear to be absolutely required to medium (DMEM, BioWhittaker Inc., Walkersville, MD, direct RUNX1 to the nuclear matrix, as RUNX1 may USA) containing 10% fetal bovine serum, 50 U/ml penicillin, contain a second nuclear matrix targeting domain 50 mg/ml streptomycin and 2 mML-glutamine (BioWhittaker). NIH 3T3 cells were maintained in DMEM containing 10% (Kanno et al., 1998). Thus, the loss of transcriptional bovine serum. Jurkat cells were cultured in RPMI 1640 repression upon deletion of residues 347–387 may be (BioWhittaker) containing 10% fetal bovine serum, antibiotics due to disrupting the conformation or stability of the and L-glutamine. pCMV5-RUNX1 and deletion plasmids have histone-modifying complex, rather than inappropriate been previously described (Lutterbach et al., 2000), with the subnuclear localization. Conversely, our work raises the exception of the D347–387 and D379–430 mutants. These latter possibility that the nuclear matrix targeting that is plasmids were created by polymerase chain reaction (PCR) by mediated by residues 351–381 may be due to the amplifying sequences including residues 1–347 and subcloning association with SUV39H1. this fragment into the BglII site in RUNX1 D290–387. RUNX1 activates transcription by associating with The D379–430 deletion was created by PCR amplification of multiple classes of DNA-binding transcription factors. the 1–379 and 430–480 fragments with the inclusion of a unique NheI restriction site for cloning. Other deletions and Most of these interactions are mediated by sequences in small fragments that were fused to the GAL4 DNA-binding the N-terminus of RUNX1 or the Runt domain. We domain were created by PCR. The GAL4 fusion cDNAs were have mapped two SUV39H1-binding sites to these same prepared by subcloning the indicated fragments, in frame, with domains (residues 1–170 and 172–212), which is con- the GAL4 DNA-binding domain (1–147) using the pCMV-M sistent with previous results (Chakraborty et al., 2003). series vectors (Amann et al., 2001).

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5785 Co-immunoprecipitations and immunoblotting TTA ACCACA TTA ACCACA TT-30, 500 mM of each Cos-7 cells (3 Â 106 cells in 100 mm dishes) were transfected oligonucleotide), or oligonucleotides derived from the using lipofectamine or lipofectamine 2000 (Invitrogen, Carls- B3URE (termed S2, JX and SB, in preparation), and bound bad, CA) with up to 5 mg of expression plasmids. Cells were to streptavidin-agarose (Sigma, St Louis, MO, USA) in extracted with phosphate-buffered saline supplemented with suspension overnight in binding buffer (BB) (12 mM N-2- 1mM ethylenediamine tetraacetate (EDTA), 1.5 mg/ml iodoa- hydroxyethylpiperazine-N0-2-ethane sulfonicacid(HEPES), cetamide, 0.2 mM polymethyl sulfonyl fluoride (PMSF) and pH 7.9, 4 mM Tris, pH 7.9, 100 mM NaCl, 0.5 mM EDTA, 0.1 T.I.U/ml aprotinin, and containing 0.5% Triton X-100 1mM MgCl2). For Jurkat cells, non-specific DNA binding was unless otherwise noted as described (Amann et al., 2001). blocked by the addition of 100 mg of Salmon sperm DNA to Immunoblot analysis was performed as described (Lutterbach 7 mg of nuclear extract. Proteins were allowed to mix with the et al., 1998a, 2000; Amann et al., 2001). Antibodies to CBFb, resin overnight at 41C, the resin was poured into a column and RUNX1, RUNX2 and RUNX3 are available from EMD the column washed two times with 10 column volumes of BB. Biosciences, San Diego, CA. Anti-SUV39H1 has been Bound proteins were eluted with the nuclear lysis buffer described previously (Firestein et al., 2000). Anti-GAL4and supplemented with 1 M NaCl. The samples were concentrated Anti-FLAG were purchased from Sigma (St Louis, MO, to 100 ml and 20–30 ml analysed by immunoblot for RUNX1 USA). and SUV39H1. For MEL cells, 500 mg of nuclear extract was In vitro purification studies were performed using the incubated in 1 ml BB (1 Â BB, 20 mM HEPES, pH 7.9, 100 mM bacterially expressed GST-fusion proteins, which were created KCl, 1 mM MgCl2, 0.5 mM EDTA, 5% glycerol, 1 mM DTT), by subcloning the full-length SUV39H1 or the DSet and supplemented with 20 mg/ml poly(dI-dC), 5 mg/ml sonicated 243–412 fragments from the GAL4 plasmids into pGEX4T. sperm DNA and 100 pmol biotinylated oligo. After 1-h The purified proteins were quantified after sodium dodecyl incubation at 41C, a 100 ml slurry of streptavidin-agarose sulfate (SDS)–PAGE using Coomassie Brilliant Blue, and beads that were pre-blocked with 20 mg/ml salmon sperm DNA approximately equal amounts of protein were used for each and 40 mg/ml acetated BSA in 1 Â BB were added and further reaction. RUNX1 or RUNX1a was transcribed and translated incubated overnight. Beads were washed four times with 1 ml using the quick coupled system (TnT; Promega, Inc., Madison, cold washing buffer (BB supplemented with 0.025% nucleo- WI, USA) in the presence of 35S-methionine. An equal amount protein (NP)-40) and directly analysed by Western blot of the reticulocyte lysate was diluted into PBS containing 0.5% analysis with the indicated antibodies. Triton X-100 and added to the glutathione beads and incubated on ice for 4–12 h. Samples were washed three times with PBS containing 0.5% Triton X-100 and analysed by Chromatin immunoprecipitation SDS–PAGE. ChIP analysis was carried out using a kit from Upstate Biotechnology following the manufacturer’s procedure with Transcription assays slight modifications. After 1% formaldehyde crosslink of NIH 3T3 cells were transfected using the Superfect or Polyfect 1.0 Â 107 MEL cells, the cell pellet was suspended in lysis reagent (Qiagen, Valencia, CA) with 300 ng WWP-Luciferase buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.05% SDS, (p21Waf1/Cip1), 300 ng pCMV5-RUNX1 (unless otherwise noted) or 0.5% Triton X-100, 0.5% NP-40, 0.1% sodium deoxycholate, deletion constructs, and 300 ng pCMV5-secreted alkaline phos- 1mM EDTA), and sonicated to shear the DNA to an average phatase (SEAP) plasmids (Lutterbach et al., 2000). Empty pCMV5 size of 500 base pairs. The chromatin was then pre-cleared was used as a control and pBluescript KS was added to bring the with protein A-agarose and incubated with normal immuno- total amount of DNA in each sample to 1.5 mg. Luciferase activity globulin G (IgG) or anti-Suv39h1 (Upstate Biotech) at 41C was measured using the Luciferase Assay System (Promega, overnight. Purified DNA was resuspended in 20 ml of 0.1 Â TE Madison, WI, USA) and normalized to SEAP activity. and analysed by real-time quantitative PCR using SyBr Green.

DNA afﬁnity chromatography Jurkat or MEL cells were lysed with IsoHI buffer containing Acknowledgements 0.2% Triton X-100 for 5 min on ice and the nuclei collected by centrifugation for 5 min at 1400 r.p.m. The cell pellet was We thank members of the Hiebert lab for critical evaluation of washed one time before the nuclei were extracted in a buffer the manuscript. We thank the Vanderbilt-Ingram Cancer containing 0.5% Triton X-100, 0.1% SDS, 67 mM Tris-Cl, pH Center and the VICC DNA sequencing core for support. This 6.8, 10 mM EDTA and 0.2 mM PMSF. The DNA was work was supported by NIH/NCI grants RO1-CA64140, fragmented by sonication and the extract clariﬁed by high- RO1-CA77274, and RO1-87549 (SWH), RO1-HL49118 (SJB), speed centrifugation. Biotinylated sense strand oligonucleo- a Merit Review Award from the Department of Veterans tides containing perfect RUNX-binding sites (50-biotin-AA Affairs (SJB), and a Center grant from the National Cancer TGTGGT TAA TGTGGT TAA TGTGGT TAA-30) were Institute (CA68485). BJI was a fellow of the Leukemia and hybridized to the antisense oligonucleotide (50-TTA ACCACA Lymphoma Society of America.

References

Amann JM, Nip J, Strom DK, Lutterbach B, Harada H, Chakrabarti SR, Nucifora G. (1999). Biochem Biophys Res Lenny N et al. (2001). Mol Cell Biol 21: 6470–6483. Commun 264: 871–877. Aronson BD, Fisher AL, Blechman K, Caudy M, Gergen JP. Chakraborty S, Sinha KK, Senyuk V, Nucifora G. (2003). (1997). Mol Cell Biol 17: 5581–5587. Oncogene 22: 5229–5237. Britos-Bray M, Friedman AD. (1997). Mol Cell Biol 17: 5127–5135. Durst KL, Lutterbach B, Kummalue T, Friedman AD, Cameron ER, Neil JC. (2004). Oncogene 23: 4308–4314. Hiebert SW. (2003). Mol Cell Biol 23: 607–619.

Oncogene RUNX1 recruits histone-modifying enzymes E Reed-Inderbitzin et al 5786 el-Deiry WS, Harper JW, O’Connor PM, Velculescu VE, Liu P, Tarle SA, Hajra A, Claxton DF, Marlton P, Freedman Canman CE, Jackman J et al. (1994). Cancer Res 54: M et al. (1993). Science 261: 1041–1044. 1169–1174. Lutterbach B, Hou Y, Durst KL, Hiebert SW. (1999). Proc Erickson PF, Robinson M, Owens G, Drabkin HA. (1994). Natl Acad Sci USA 96: 12822–12827. Cancer Res 54: 1782–1786. Lutterbach B, Sun D, Schuetz J, Hiebert SW. (1998a). Mol Fenrick R, Wang L, Nip J, Amann JM, Rooney RJ, Walker- Cell Biol 18: 3601–3611. Daniels J et al. (2000). Mol Cell Biol 20: 5828–5839. Lutterbach B, Westendorf JJ, Linggi B, Isaac S, Seto E, Firestein R, Cui X, Huie P, Cleary ML. (2000). Mol Cell Biol Hiebert SW. (2000). J Biol Chem 275: 651–656. 20: 4900–4909. Lutterbach B, Westendorf JJ, Linggi B, Patten A, Moniwa M, Gergen JP, Wieschaus EF. (1985). Dev Biol 109: 321–335. Davie JR et al. (1998b). Mol Cell Biol 18: 7176–7184. Goetz TL, Gu TL, Speck NA, Graves BJ. (2000). Mol Cell Biol Miyoshi H, Kozu T, Shimizu K, Enomoto K, Maseki N, 20: 81–90. Kaneko Y et al. (1993). EMBO J 12: 2715–2721. Golub TR, Barker GF, Bohlander SK, Hiebert SW, Ward DC, Miyoshi H, Shimizu K, Kozu T, Maseki N, Kaneko Y, Ohki Bray-Ward P et al. (1995). Proc Natl Acad Sci USA 92: M. (1991). Proc Natl Acad Sci USA 88: 10431–10434. 4917–4921. Nishimura M, Fukushima-Nakase Y, Fujita Y, Nakao M, Gu TL, Goetz TL, Graves BJ, Speck NA. (2000). Mol Cell Biol Toda S, Kitamura N et al. (2004). Blood 103: 562–570. 20: 91–103. Nucifora G, Begy CR, Erickson P, Drabkin HA, Rowley JD. Guidez F, Petrie K, Ford AM, Lu H, Bennett CA, MacGregor (1995). Proc Natl Acad Sci USA 92: 4917–4921. A et al. (2000). Blood 96: 2557–2561. Peterson LF, Zhang DE. (2004). Oncogene 23: 4255–4262. Hiebert SW, Sun W, Davis JN, Golub T, Shurtleff S, Buijs A Raynaud S, Cave H, Baens M, Bastard C, Cacheux V, et al. (1996). Mol Cell Biol 16: 1349–1355. Grosgeorge J et al. (1995). Gene 159: 245–248. Hildebrand D, Tiefenbach J, Heinzel T, Grez M, Maurer AB. Rea S, Eisenhaber F, O’Carroll D, Strahl BD, Sun ZW, (2001). J Biol Chem 276: 9889–9895. Schmid M et al. (2000). Nature 406: 593–599. Imai Y, Kurokawa M, Yamaguchi Y, Izutsu K, Nitta E, Romana SP, Mauchauffe M, Le Coniat M, Chumakov I, Le Mitani K et al. (2004). Mol Cell Biol 24: 1033–1043. Paslier D, Berger R et al. (1995). Blood 85: 3662–3670. Jacobs SA, Taverna SD, Zhang Y, Briggs SD, Li J, Eissenberg Swantek D, Gergen JP. (2004). Development 131: 2281–2290. JC et al. (2001). EMBO J 20: 5232–5241. Taniuchi I, Osato M, Egawa T, Sunshine MJ, Bae SC, Komori Javed A, Guo B, Hiebert S, Choi JY, Green J, Zhao SC et al. T et al. (2002). Cell 111: 621–633. (2000). J Cell Sci 113: 2221–2231. Telfer JC, Hedblom EE, Anderson MK, Laurent MN, Jenuwein T, Allis CD. (2001). Science 293: 1074–1080. Rothenberg EV. (2004). J Immunol 172: 4359–4370. Kanno T, Kanno Y, Chen L, Ogawa E, Kim W, Ito Y. (1998). van Wijnen AJ, Stein GS, Gergen JP, Groner Y, Hiebert SW, Mol Cell Biol 18: 2444–2454. Ito Y et al. (2004). Oncogene 23: 4209–4210. Lachner M, O’Carroll D, Rea S, Mechtler K, Jenuwein T. Wheeler JC, VanderZwan C, Xu X, Swantek D, Tracey WD, (2001). Nature 410: 116–120. Gergen JP. (2002). Nat Genet 32: 206–210. Lenny N, Meyers S, Hiebert SW. (1995). Oncogene 11: Zeng C, McNeil S, Pockwinse S, Nickerson J, Shopland L, 1761–1769. Lawrence JB et al. (1998). Proc Natl Acad Sci USA 95: Levanon D, Goldstein RE, Bernstein Y, Tang H, Goldenberg 1585–1589. D, Stifani S, Paroush Z, Groner Y. (1998). Proc Natl Acad Zhang DE, Hetherington CJ, Meyers S, Rhoades KL, Larson Sci USA 95: 11590–11595. CJ, Chen HM et al. (1996). Mol Cell Biol 16: 1231–1240.

Oncogene