The Conserved Mynd Domain of BS69 Binds Cellular and Oncoviral Proteins through a Common PXLXP Motif*

BS69 is a transcriptional co-repressor protein and a potential tumor suppressor that binds to the adenoviral oncoprotein E1A. We show that the C-terminal Mynd domain of BS69 (amino acids 516–561) or the closely related Mynd domains of theCaenorhabditis elegans proteins Bra-1 and Bra-2 bind not only to E1A but also to the Epstein-Barr virus EBNA2 oncoprotein and the Myc-related cellular protein MGA. Interaction depends on intact PXLXP motifs present in all three proteins. Moreover, viral proteins compete for binding of BS69 to MGA in a PXLXP-dependent fashion. Because deletions in E1A or EBNA2 that cover the PXLXP motifs are non-transforming, our observations suggest a role for BS69 in cell growth control that is reminiscent of abrogation of the Rb function by various oncoproteins.

Oncoviral proteins of transforming DNA tumor viruses have evolved to induce viral replication in resting host cells. They do so by selective interaction with a restricted number of cellular proteins that govern the cell cycle, differentiation, or cell survival. The same cellular proteins are often mutated in various cancers demonstrating their importance in cellular growth control.
An example of such a viral oncoprotein that interferes with the function of critical growth regulators is the Epstein-Barr nuclear antigen EBNA2. 1 EBNA2 interferes with transcriptional regulation and is required for B-cell immortalization by the Epstein-Barr virus. Association of EBNA2 with cellular transcription factors such as CBF1/RBP-J or PU-1 activates cellular and viral target genes by EBNA2-mediated recruitment of the pre-initiation complex, histone acetyltransferases, as well as the SWI⅐SNF remodeling complex (1,2).
Another eminent example and an invaluable tool for unraveling proliferation and differentiation control is the adenoviral E1A oncoprotein. E1A is expressed early in the viral replication cycle and interferes with the function of multiple cellular proteins involved in gene regulation. E1A can immortalize rodent fibroblasts and transforms them in collaboration with E1B or with activated versions of Ras (3). A comparison of E1A proteins from various viral serotypes highlights three conserved regions, CR1, CR2, and CR3, that bind to and modulate the activities of cellular proteins implicated in cell cycle progression (the retinoblastoma protein (Rb)), in chromatin remodel-ing (SWI⅐SNF complex, histone acetyltransferases) and in gene transcription (TFIID, mediator complex, transcription factors) (3,4).
BS69 was originally identified as a nuclear protein that binds to E1A and inhibits its trans-activation potential (5). An alternatively spliced BS69 isoform, termed BRAM1, encompasses the C terminus of BS69. BRAM1, as well as its Caenorhabditis elegans orthologs, Bra-1 and Bra-2, were proposed to participate in BMP and TGF␤ signaling pathways (6,7). The BS69 C terminus is related to a cysteine-rich structure that has been termed the Mynd domain. The Mynd domain is present in the AML1 chromosomal translocation partner ETO or in several developmentally important proteins, such as Nervy and DEAF1 (8). The ETO-Mynd domain recruits a histone deacetylase complex and mediates gene repression (9,10). Similarly, the BS69-Mynd domain has been suggested to be involved in repression through N-CoR recruitment (11).
Here we show that the BS69-specific Mynd domain shares structural and binding features with the two C. elegans proteins Bra-1 and Bra-2. BS69-type Mynd domains bind not only to E1A but also to the EBNA2 oncoprotein or to the Myc-related cellular transcription factor, MGA. Binding depends on a common PXLXP peptide motif found in all three proteins. Moreover, E1A competes for the interaction between BS69 and MGA. Our data suggest that the respective peptide motifs in oncoviral proteins have evolved to abrogate cellular functions of BS69-type Mynd domains. The competitive binding of cellular and viral proteins is reminiscent to the abrogation of functions of Rb-type proteins.
GST Pull-down Assay and Coimmunoprecipitation Experiments-GST pull-down assays were performed as described previously (14). Immunoprecipitation was performed in 100 mM NaCl, 20 mM Tris, pH 8, and 0.5% Nonidet P-40 supplemented with protease inhibitors in experiments employing E1A and a BS69 C-terminal peptide. In exper-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence may be addressed. Tel.: ϩ49-30-9406-3735; Fax: ϩ49-30-9406-3298; E-mail: ansieau@mdc-berlin.de; aleutz@mdcberlin.de. 1 The abbreviations used are: EBNA, Epstein-Barr nuclear antigen; iments employing the full-length BS69 protein, cells were lysed in 350 mM NaCl, 20 mM Tris, pH 8, 30 mM MgCl 2 , and 1% Nonidet P-40, sonicated, cleared by centrifugation, and diluted to a 20 mM Tris, pH 8, 100 mM NaCl, 8.5 mM MgCl 2 , and Nonidet P-40, 0.5% final concentration. Lysates were incubated with 6 l of anti-BS69 ascites at 4°C for 3 h. Protein A-Sepharose (Amersham Biosciences Inc.) was added and incubated at 4°C for 1 h. Beads were washed six times in 100 mM NaCl, 20 mM Tris, pH 8, 0.5% Nonidet P-40, and proteins were separated by SDS-PAGE. BS69 and EBNA2 proteins were visualized with an M2 anti-FLAG antibody (Integra Bioscience) and an anti-EBNA2 monoclonal antibody, respectively (15). Yeast Two-hybrid Screen-Yeast two-hybrid screens were performed with the Matchmaker system (CLONTECH). A PCR fragment encoding amino acids 411-to 561 of human BS69 was generated and subcloned in-frame with the Gal4 DNA binding domain in the pGBT9 vector (CLONTECH). The HF7c yeast strain (CLONTECH) was sequentially transformed with the bait and an Epstein-Barr viral human B-cell library or a human lung library (CLONTECH) and plated on selective plates. Colonies were tested for ␤-galactosidase activity using a standard filter assay. To control the interaction specificity, plasmids were isolated and used to transform the SFY526 yeast strain (CLONTECH) in combination with the bait or unrelated Gal4 fusion proteins.
Antibodies-A BS69-specific antibody was generated by expressing a peptide encompassing residues 358 -441 of the human BS69 protein as a GST fusion protein. The fusion protein was purified by affinity chromatography on glutathione beads. An antiserum against the peptide was raised in guinea pig.

Adenoviral E1A Proteins Bind to BS69-type Mynd Domains-
The human BS69 protein displays signatures of proteins that are involved in chromatin and gene regulation such as the PHD zinc-finger (16), bromo-(17), PWWP (18), and Mynd domains (8) (Fig. 1A). The C-terminal third of BS69 corresponds to the splice variant BRAM1 and has been described as a target of E1A (5). To determine the interaction between BS69 and E1A more precisely, we expressed BS69 deletion mutants as GST fusion proteins in bacteria (Fig. 1A) and examined their interaction with in vitro translated 13S E1A. The BS69 Mynd domain was found to be required and sufficient for E1A binding (Fig. 1B). Mutation of a single conserved cysteine residue in the BS69 Mynd domain (Cys-532 to serine, Fig. 1A) abrogated E1A binding (Fig. 1, B and C). BRAM1 orthologs have also been identified in C. elegans as Bra-1 and Bra-2, implicated in TGF␤/BMP signaling (6,7). Whereas the Mynd domains of Bra-1 and Bra-2 also bind to E1A, the Mynd domains of RACK7 (KIAA1125) and of ETO (9, 10), respectively, failed to do so (Fig. 1E). These data suggest that BS69-type Mynd domains are specific and conserved targets of E1A.
The Epstein-Barr Viral Protein EBNA2 Binds to BS69-type Mynd Domains-To identify additional proteins that interact with BS69-type Mynd domains, we performed yeast two-hybrid screens of an Epstein-Barr virus-immortalized B-cell library using the BS69 Mynd domain as bait. Eight His ϩ , ␤-Gal ϩ clones were isolated from 1 ϫ 10 6 independent transformants. All eight clones encoded different C-terminal portions of the Epstein-Barr viral protein, EBNA2. Fig. 1B shows that similar to E1A, binding of EBNA2 is restricted to and depends on the integrity of the BS69-Mynd domain. Fibroblasts were co-transfected with BS69 and EBNA2 expression vectors to examine whether EBNA2 also binds to BS69 in cells. As shown in Fig.  1D, BS69 immunoprecipitates EBNA2. Similar to E1A, EBNA2 also binds to Bra-1 and Bra-2 but not to ETO or RACK7 (Fig.  1E).
BS69 Binds to a PXLXP Peptide Motif Present in Both Viral Proteins-Binding of two different oncoviral proteins to the BS69 Mynd domain prompted a search for a common target site. A series of oncoprotein mutants was generated and examined for interaction with the BS69 Mynd domain (Fig. 2). Amino acid (aa) residues 86 -128 in E1A ( Fig. 2A) and two independent protein parts in EBNA2 (aa 379 -408 and 409 -439, Fig. 2B) were identified. A comparison of the corresponding amino acid sequences revealed a common pentamer peptide motif, PXLXP (Fig. 2C). To determine whether this motif is involved in binding to BS69, the central invariable leucine residue was mutated into an alanine in E1A. As shown in Fig.   FIG. 1. E1A and EBNA2  and EBNA2 proteins to various BS69 deletion mutants (depicted in panel A) expressed as GST fusion proteins or to the GST moiety (G-) as control. Input is 10%. C, a single point mutation in the Mynd domain abrogates BS69 binding to E1A. QT6 fibroblasts were transfected with E1A and/or FLAG-tagged BS69 cDNA expression vectors as indicated. Cellular protein complexes were immunoprecipitated with M73 anti-E1A antibody. Co-immunoprecipitated (Co-IP) and ectopic expression of BS69 were revealed by immunoblotting using an anti-FLAG M2 antibody. D, BS69 binds to EBNA2 in cells. HeLa fibroblasts were transfected with EBNA2 and FLAG-tagged BS69 cDNA expression vectors as indicated. Cellular protein complexes were immunoprecipitated with a polyclonal anti-BS69 antibody. Co-immunoprecipitated and ectopic expression of EBNA2 and BS69 were revealed by immunoblotting using monoclonal anti EBNA2 and anti-FLAG M2 antibodies, respectively. E, E1A and EBNA2 bind specifically to the BS69-type Mynd domain. Binding of E1A and EBNA2 to Bra-1, Bra-2, RACK7, and ETO Mynd domains expressed as GST fusion proteins or to the GST moiety (G-) as control.
3A, the resulting 12S and 13S E1A (E1A,L115A) mutant proteins failed to interact with BS69. Similarly, EBNA2, mutated within both of the PXLXP motifs (EBNA2 L385A,L439A), failed to interact with BS69 (Fig. 3B). Mutation of a single motif in EBNA2 (EBNA2 L385A and EBNA2 L439A), however, did not abrogate BS69 binding, suggesting the redundancy of EBNA2 PXLXP motifs (data not shown). These data suggest that both oncoviral proteins interact specifically with the BS69 specific Mynd domain via a common structural motif.
E1A Competes the Interaction between BS69 and MGA-We searched for cellular proteins that interact with the BS69 Mynd domain in a lung library using the yeast two-hybrid assay. From one million transformants two His ϩ , ␤-Gal ϩ clones displayed very fast staining kinetics and were therefore examined further. Sequence analysis revealed that both clones were identical and encompassed amino acid residues 2566 -2975 of the Myc-related protein MGA (19). Sequence analysis revealed that PXLXP motifs are present in duplicate in the MGA C terminus (Fig. 2C). Deletion analysis, as shown in Fig. 3C, confirmed that the PXLXP motifs in MGA are essential for binding to the BS69 Mynd domain.
Next we determined whether E1A proteins compete for the interaction between BS69 and MGA. A GST-BS69 protein construct was preincubated with total cellular extracts of mocktransfected or E1A-transfected QT6 fibroblasts and tested for its ability to pull down radiolabeled MGA protein. As shown in Fig. 3D, binding of either 12S E1A or 13S E1A proteins to BS69 abrogated the binding of MGA, whereas E1A mutants that fail to bind to BS69 (12 S E1A L115A and 13 S E1A L115A) do not compete for BS69-MGA interaction. E1A and EBNA2 Neutralize BS69 Functions-E1A has previously been shown to neutralize the repressing function of BS69 (11). Repressor activity of BS69 can be determined in an assay in which reporter expression is driven by the oncoviral protein (5). To examine whether E1A and EBNA2 similarly affect BS69 function, we took advantage of the fact that both oncoviral proteins activate reporter gene expression through the CBF1/RPB-J transcription factor (14, 15, 20 -22), a downstream target of Notch signaling (23,24). As shown in Fig. 4A, expression of BS69 strongly reduces reporter activation through E1A or EBNA2 but did not affect reporter activation through the active form of Notch1 (23,24). Furthermore, E1A or EBNA2 mutants that failed to bind BS69 were also not suppressed by BS69. To validate the inhibitory effect of BS69 on EBNA2 trans-activation, EBNA2 was fused to the DNA binding domain of the Gal4 transcription factor (Gal-), and the A, disruption of the PXLXP motif in E1A abrogates binding to BS69. QT6 fibroblasts were transfected with various E1A and/or FLAG-tagged BS69 (residues 411-561) cDNA expression vectors, as indicated. Cellular protein complexes were immunoprecipitated with an anti-E1A monoclonal antibody. Co-immunoprecipitated (Co-IP) and ectopic expression of BS69 were revealed by immunoblotting using an anti-FLAG M2 antibody. B, disruption of both PXLXP motifs in EBNA2 abrogates binding to BS69. In vitro radiolabeled EBNA2 proteins were incubated with total cellular extracts from mock-transfected (Ϫ) or from HeLa fibroblasts transfected with a FLAG-tagged BS69 C-terminal peptide (residues 411-561). Cellular protein complexes were immunoprecipitated with a monoclonal anti-FLAG antibody. Co-immunoprecipitated EBNA2 and BS69 were revealed by autoradiography and by immunoblotting using a monoclonal anti-HA antibody, respectively. C, deletion of both PXLXP motifs in MGA abrogates binding to BS69. Binding of MGA peptides (as depicted on the top) to BS69 expressed as GST fusion protein (G-⌬410) or to the GST moiety (G-) as control. Input, 10%. D, E1A and MGA compete for binding to BS69. GST (G-) or GST-BS69 (G-⌬410)-loaded glutathione beads were preincubated with total cellular extracts of mock-transfected or E1A-transfected QT6 fibroblasts and tested for their ability to pull down an in vitro radiolabeled MGA protein. Top, binding of MGA. Bottom, E1A expression revealed by immunoblotting using a monoclonal M73 anti-E1A antibody. Input, 10%. activity of the resulting fusion protein (Gal-EBNA2) was examined on an appropriate reporter. As shown in Fig. 4B, BS69 expression strongly reduces reporter activation of the Gal-EBNA2 fusion protein yet not of a Gal-VP16 control. These data show that specific interaction between BS69 and its target proteins is a prerequisite for BS69-mediated gene repression. The results also suggest that the inhibitory function of BS69 is specifically neutralized by E1A and EBNA2. DISCUSSION BS69 mediates transcriptional inhibition and represents a potential tumor suppressor protein (11). For gene repression, BS69 essentially requires the function of its C-terminal Mynd domain (5,11), which also represents a conserved target of different viral oncoproteins. We found that BS69-type Mynd domains interact through conserved PXLXP peptide motifs with the oncoviral proteins E1A and EBNA2 and with the cellular MGA protein. MGA is a member of the Myc-Max network of bHLH-ZIP transcription factors that contains a T-box domain and presumably links cell proliferation control to mesoderm specification during embryogenesis (19). Our data also show that E1A competes the interaction between BS69 and MGA.
Whereas the C. elegans Bra-1 and Bra-2 Mynd domains shared the same binding specificities as BS69, the related Mynd domains of RACK7 and ETO proteins failed to do so, suggesting that various types of Mynd domains mediate highly specific protein interactions. Comparison of the BS69, RACK7, and ETO Mynd domains revealed differences in amino acids and cysteine spacing that might account for distinct protein interactions (Fig. 1A). In agreement with this notion is the observation that the interaction between E1A or EBNA2 with BS69 is abrogated by a point mutation of a single cysteine residue (C523S) in the Mynd domain.
PXLXP motifs were identified as essential components of the BS69 Mynd targets. The first PXLXP target was found in E1A between conserved regions CR1 and CR2 ( Figs. 2A and 3A). Others have suggested that a BS69 interaction site is located between CR1 and CR2 of E1A in addition to binding to CR3 (5). In our hands, however, CR3 contributed poorly to BS69 bind-ing. This conclusion is based on the observations that comparable BS69 immunoprecipitation was obtained with 12S and 13S E1A (Fig. 3A) and that weak residual binding of the PX-LXP mutant 13S E1A L115A was only observed under low stringency binding conditions (data not shown). Moreover, two PXLXP motifs in EBNA2, located in regions CR7 and CR8, were also identified as critical for BS69 binding (Figs. 2B and 3B). Mutations of the PxLxP motifs in either E1A or EBNA2, however, did not inhibit binding to and trans-activation through CBF1 (Fig. 4 and data not shown). Furthermore, E1A PXLXP mutants specifically disrupted the interaction with BS69 without affecting the ability of E1A to associate with Rb (data not shown). Thus, we conclude that PXLXP motifs are indispensable for BS69-type Mynd domain interactions.
Conservation of PXLXP motifs in various viral serotypes suggests that the motifs mediate important oncoprotein functions. This notion is supported by the fact that others have shown that deletions covering CR7 or CR8 in EBNA2 and deletion of sequences comprising the BS69 binding sites in E1A abrogate the transformation potential of the respective oncoproteins (25,26). A question that emerges is whether the PX-LXP sequence would be sufficient for BS69 interaction. Data base searches identified many proteins that contain PXLXP motifs. One of them is the human TATA-box-binding protein, TBP, that contains PXLXP at its C terminus. TBP, however, failed to bind to BS69 (data not shown). This suggests that additional structural constraints are required for specific BS69 Mynd domain/PXLXP interaction that await further analysis. Nevertheless, we believe that the remarkable conservation of protein sequence and interaction specificity across species is intriguing and will help to elucidate Mynd functions in development and tumorigenesis.
Data presented here also include identification of MGA as a cellular binding partner for BS69. E1A competed with MGA for BS69 binding, and the interaction of MGA with BS69 depended on PXLXP motifs in the C terminus of MGA. The competitive interaction between BS69 and a critical cellular protein with two entirely different viral oncoproteins via identical protein motifs is reminiscent of the interaction between cellular pocket-binding proteins of the Rb family with several DNA tumorviral oncoproteins through LxCxE motifs (27)(28)(29). In keeping with that idea, it has been speculated that BS69 is a tumor suppressor that is frequently deleted in cancers and leukemia and that inhibits transcription through proto-oncogene products (11). Of particular interest is also the genetic connection between the TGF␤/BMP pathways and Bra-1/Bra-2 (6, 7). Our data show that the C. elegans Bra-1/Bra-2 Mynd domains display the same binding specificity as BS69. Because Bra-1 has been implicated in TGF␤ signaling, it will be interesting to investigate whether E1A and EBNA2 might disrupt TFG␤ signaling and alter cell proliferation and differentiation through abrogating functions of BS69-type Mynd domains.