Two nonconsensus sites in the Epstein-Barr virus oncoprotein EBNA3A cooperate to bind the co-repressor carboxyl-terminal-binding protein (CtBP).

CtBP (carboxyl-terminal binding protein) has been shown to be a highly conserved co-repressor of transcription that is important in development, cell cycle regulation, and transformation. Viral proteins E1A and EBNA3C and all the various Drosophila and vertebrate transcription factors to which CtBP has been reported to bind contain a conserved "PXDLS" CtBP-interaction domain. Here we show that EBNA3A binds CtBP both in vitro and in vivo but that this interaction does not require a near consensus (98)PLDLR(102) motif in the NH(2) terminus of EBNA3A. However, further deletion and mutation analysis revealed that CtBP interacts with this viral protein through a cryptic, bipartite motif located in the COOH terminus of EBNA3A. The two components of this binding domain are similar to the canonical PXDLS motif but do not include the highly conserved, and normally critical, first proline residue. These nonconsensus sites, (857)ALDLS(861) and (886)VLDLS(890), synergize to produce very efficient binding to CtBP. Interaction with CtBP was shown to be important in the repression of transcription by EBNA3A and in the ability of EBNA3A to cooperate with activated Ras to immortalize and transform primary rat embryo fibroblasts. Similar bipartite sequences can be found in other viral and cellular proteins that can interact with CtBP, including the retinoblastoma-interacting protein-methyltransferase RIZ, the oncoprotein EVI1, and Marek's disease virus transforming protein Meq.

E1A COOH-terminal binding protein (CtBP1) 1 was initially identified as a cellular factor interacting with the COOH terminus (amino acids 225-238) of adenovirus E1A oncoproteins. Although the precise significance of this interaction remains unknown, it is essential for the immortalization of primary rodent cells by E1A and it has also been reported to negatively modulate E1A-mediated transformation, tumorigenicity, and metastasis (1,2). More recently it has been shown that this E1A-binding protein is one of a highly conserved family of co-repressors of transcription. In Drosophila melanogaster, Drosophila CtBP mediates transcriptional repression by numerous factors involved in developmental programs of gene expression including Knirps, Kruppel, Snail, Polycomb, Hairy, and Hairless (3)(4)(5)(6). CtBP also interacts with Xenopus laevis (xPC) and human (HPC2) polycomb proteins and thus may be involved in vertebrate gene silencing (7). Human CtBP also acts as a co-repressor for the ZEB factors that are involved in the regulation of lymphocyte and muscle differentiation (8), and it mediates the repressor and transforming activities of the oncoprotein EVI1 (9,10). In mice, CtBP is a co-repressor for the NET member of the Ets family of transcription factors and oncogenes (11). Furthermore, murine CtBP participates in the Ikaros repression complex (12). Angustifolia, the first plant member of the CtBP family was recently identified in Arabidopsis thaliana (13,14).
It has been reported that in some situations CtBP can recruit chromatin-modifying histone deacetylase enzymes HDAC1, -4, -5, and -7 and that it can also bind Sin3A. However, on some promoters CtBP-mediated repression is independent of HDAC activity (15), so presently the precise molecular mechanism by which CtBP inhibits transcription is unknown and may turn out to be quite different in different situations (11,12,(15)(16)(17). All these various factors that regulate transcription and bind to CtBP contain a conserved Pro-X-Asp-Leu-Ser ("PXDLS") CtBPinteraction domain that is necessary and probably sufficient for the interaction. The conserved proline and leucine residues are generally considered essential for efficient binding (15,18,19). A second mammalian CtBP was recently described and the two family members, which are referred to as CtBP1 and CtBP2, are largely homologous although they may have distinct tissue distributions (20). The protein described in this report is human CtBP1 and hereafter will be referred to as CtBP.
The PXDLS amino acid motif is also found in a human cellular protein called CtIP (CtBP-interacting protein). CtIP was shown to bridge an interaction between the retinoblastoma tumor suppressor protein and CtBP, forming a complex that can repress E2F-regulated genes and thus participate in the regulation of the cell proliferation cycle; CtIP probably bridges p130 and CtBP in a similar manner (16,21). Recently it was shown that CtBP and CtIP form a link between human polycomb group proteins and retinoblastoma tumor suppressor protein and this repressor complex appears to inhibit cyclin A and cdc2 expression and block entry into mitosis (22). CtIP has also been shown to bind to the carboxyl-terminal region of the breast cancer-associated tumor suppressor and transcription factor BRCA1 and may be involved in the regulation of p21 WAF1 and GADD45 genes by BRCA1 (17,(23)(24)(25). The recent demonstration that the binding of CtBP to viral and cellular transcriptional repressors is regulated by the nicotinamide adenine dinucleotides NAD ϩ and NADH has led to the suggestion that CtBP may serve as a redox sensor for transcription (26).
The EBNA3A gene has been shown by genetic analysis using recombinant viruses, to be one of the five viral genes that are absolutely essential for the efficient activation and immortalization of human B cells by Epstein-Barr virus (EBV) (27)(28)(29). The large (944-amino acid) nuclear protein it encodes is one of a family of three nuclear antigens that probably arose by gene duplication events. EBNAs-3A, -3B, and -3C have limited but significant amino acid sequence homology, have the same gene structure (a short 5Ј exon and a long 3Ј exon), and are tandemly arranged in the genome (28). Genetic studies using recombinant viruses have shown that EBNA3A and EBNA3C are essential for in vitro immortalization of B cells, whereas EBNA3B is dispensable (29). All three proteins bind to a cellular DNAbinding protein known as CBF1 or RBP-JK. This factor also binds to, and targets to DNA, the EBV transactivator protein EBNA2 and also the IC-Notch effector of the Notch signaling pathway (30). EBNA3A, -3B, and -3C can all repress EBNA2mediated transactivation of the EBV LMP-2 promoter (31) and EBNA3A and -3C can repress reporter plasmids containing the RBP-JK/CBF1 sites from the EBV Cp promoter independently of EBNA2-mediated activation (32). Because Cp is the promoter for all EBNA mRNA initiation in lymphoblastoid cells immortalized by EBV (LCLs), the EBNA3 proteins probably contribute to a negative autoregulatory control loop. In addition, both EBNA3A and EBNA3C exhibit robust repressor activity when targeted to DNA by fusion with the DNA-binding domain of Gal4 (33)(34)(35). The repression by EBNA3C is partly mediated by interactions with cellular histone deacetylase and possibly CtBP, but the mechanism of EBNA3A-mediated repression is unknown.
It was recently shown that EBNA3C can bind CtBP in vitro and in vivo through a PLDLS motif located in its COOH terminus (amino acids 728 -732). This binding is required for EBNA3C to act as a co-operating oncogene (36). We were therefore prompted to consider whether EBNA3A may also bind this co-repressor and whether it has similar oncogenic activity. CtBP generally binds to proteins including a PXDLS motif but there are several reports that binding can occur through a variant amino acid sequence, PLDLR (18). Because EBNA3A has PLDLR located at amino acids 98 -102, GST pull-down experiments were performed to determine whether, like EBNA3C, this related protein binds CtBP in a similar manner.
We report that EBNA3A can physically and functionally interact with CtBP but that this does not require the 98 PLDLR 102 motif. The interaction was shown to depend on two cryptic sites located near the COOH terminus of the protein that bind CtBP synergistically. These sites appear to be necessary for EBNA3A-mediated repression of transcription and, as with EBNA3C, binding to CtBP correlated with the ability of EBNA3A to co-operate with oncogenic Ras in primary rodent fibroblasts. Consistent with this, EBNA3A was co-immunoprecipitated with rat CtBP from the EBNA3A-positive immortalized and transformed rat embryo fibroblasts.
pcDNA3-EBNA3B and pcDNA3-EBNA3C contain full-length EBNA3B and -3C cDNAs, respectively, and have been described previously (36). The plasmid pSG5-HA-CtBP contains cDNA encoding the full-length human CtBP1 fused in-frame with an HA epitope tag. The plasmid was constructed by PCR from pGex-CtBP using the primers: CGGAATTCCGATGGGCAGCTCGCACTTG and CGGAATTCCGC-TACAACTGGTCACTGGC and cloned as an EcoRI fragment into pSG5-HA.
The plasmid SV␤-gal for the expression of ␤-galactosidase in mammalian cells has been described previously (37). The plasmid pGL-UAS (a gift from Roger Watson) contains luciferase cDNA under the control of a Gal4 responsive promoter. EJ6.6 includes an activated human (Ha)-ras under the control of the SV40 early promoter (38). pGEX-CtBP for bacterial expression of a GST-CtBP fusion protein was described previously (36). Similarly, pGEX-RBP-JK was used for the expression of GST-RBP-JK in bacteria as described previously (32,34,37). pcDNA3-HPC2 contains a full-length HPC2 cDNA, pcDNA3-HDAC4 contains a full-length HDAC4 cDNA, and pcDNA3-myc-CtIP contains a Myc-tagged CtIP cDNA (16).
GST Pull-down Assays-Assays were performed essentially as described previously (37). In vitro translated test proteins were adsorbed to GST-coated Sepharose beads for 1 h by rotation at 4°C in 200 l of EBC buffer (140 mM NaCl, 0.5% Nonidet P-40, 50 mM Tris-Cl, pH 8.0, 100 mM NaF, 200 M Na 3 VO 4 , 1 mg/ml bovine serum albumin). The beads were sedimented by centrifugation, and the cleared supernatant was incubated with GST or GST-CtBP protein bound-Sepharose beads at 4°C for 90 min. The beads were washed 4 times with 1 ml of NETN (300 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 20 mM Tris-Cl, pH 8.0). The beads were heated at 95°C for 5 min in SDS sample buffer, and bound proteins were resolved by SDS-PAGE and visualized by autoradiography. Results were quantified by PhosphorImager analysis. Coupled in vitro transcription-translation reactions and the purification of GST proteins were carried out as previously described (33,36).
Antibodies-For Western blot and immunofluorescent detection of EBNA3A a sheep anti-EBNA3A polyclonal antibody (Exalpha Biologicals) was used. For immunoprecipitation of proliferating cell nuclear antigen from cell extracts the mouse monoclonal antibody PC10 was used, which was a kind gift from Dr. X. Lu. The antibodies horseradish peroxidase-conjugated goat anti-rabbit IgG, horseradish peroxidaseconjugated rabbit anti-sheep, and FITC-conjugated rabbit anti-sheep were used in the detection of CtBP and EBNA3A by Western blot analysis or immunofluorescence and were supplied by Dako. The rabbit anti-CtBP serum was raised against GST-CtBP.
Immunoprecipitations-Immunoprecipitations were performed essentially as described previously (37). Briefly 10 7 cells were harvested and washed once in ice-cold PBS. The cells were collected by centrifugation and the pellet re-suspended in 600 l of IP lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10% glycerol, 0.5% Triton X-100, 2 mM phenylmethylsulfonyl fluoride, 2 mM mixture of proteinase inhibitors (Roche Molecular Biochemicals)). The cell suspensions were incubated at 4°C for 20 min and the debris was pelleted by centrifugation. The supernatant was split into 200-l aliquots per immunoprecipitation reaction and 20 l of protein G-Sepharose beads was added before the mixture was mixed at 4°C for 1 h. The beads were next pelleted by centrifugation and the supernatant was transferred to a fresh tube. Complexes were then precipitated with an antibody specific for the protein of interest and the mixture was incubated at 4°C on an orbital rotor for 2 h. After this, 30 l of protein G-Sepharose beads were added to the mixture and it was left to rotate at 4°C for 1 h. The beads were next washed 4 times in immunoprecipitation lysis buffer. 30 l of SDS sample buffer was then added to the pelleted beads, boiled, centrifuged, and the supernatant was loaded on a SDS-PAGE gel for Western blot analysis.
Immunofluorescence and Western Blot Analysis-EBNA3A and Rastransformed rat embryo fibroblasts (REF) were grown on coverslips before being washed 3 times with PBS. The cells were fixed with 2% formaldehyde for 15 min and then washed twice with PBS. Cells were then incubated with 0.5% Triton X-100 in PBS for 5 min before being washed twice with PBS and then blocked with 100 mM glycine. After washing twice in PBG (0.2% gelatin, 0.5% bovine serum albumin in PBS) the cells were incubated with sheep anti-EBNA3A antibody for 1 h. After 4 washes in PBG the cells were incubated with a FITCconjugated donkey anti-sheep antibody for 1 h before a further 4 washes with PBG and a final wash in PBS. The coverslips were mounted on slides and viewed on a confocal microscope. Western blots were performed as previously described (39) and probed with a sheep anti-EBNA3A antibody for the detection of EBNA3A or with rabbit anti-CtBP serum for the detection of CtBP.
Cell Culture, Transient Transfections, ␤ϪGalactosidase, and Reporter Assays-DG75 cells (a Burkitt's lymphoma-derived cell line) and LCL lines were grown in suspension and maintained in RPMI 1640 (Invitrogen) supplemented with 10% heat-inactivated fetal calf serum, 2 mM L-glutamine (Invitrogen), and 100 units of penicillin and streptomycin (Invitrogen)/ml. REFs were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% heat-inactivated fetal calf serum, 2 mM L-glutamine (Invitrogen), and 100 units of penicillin and streptomycin (Invitrogen)/ml with transformed REFs being selected by the addition of G418 (400 g/ml) to the media. DG75 cells were transfected with 10 g of each of the EBNA3A expressing vectors and a CtBP expressing vector for the immunoprecipitation experiments. For reporter assays, 2 g of SV␤-gal, 7 g of pGL-UAS, and varying amounts of pcDNA3-HA-GAL4DBD-EBNA3A wt or 2ϫ mutant. The total amounts of DNA were balanced throughout with control vector DNA. Electroporation was used to transfect DG75 cells. Cells were split at a density of 2 ϫ 10 5 /ml 24 h prior to transfection. DNA was added to 0.4-cm 3 cuvettes (Bio-Rad) and the volume was made up to 50 l with unsupplemented RPMI. 10 7 cells were harvested per transfection. These cells were pelleted at 1400 rpm for 4 min at 4°C. The supernatant was retained as conditioned medium. The pellet was resuspended in 150 l of unsupplemented RPMI, added to the cuvette, and incubated on ice for 10 min. Electroporation was performed at 250 V, 960 microfarads using a Gene Pulser (Bio-Rad). The cuvettes were then incubated at 37°C for 15 min after which the transfected cells were resuspended by pipette and added to 10 ml of medium (7.5 ml of conditioned medium and 2.5 ml of fresh RPMI with 10% fetal calf serum). Transfected cells were incubated at 37°C for 24 -48 h before harvesting.
Transfected cells were collected by centrifugation and washed in PBS. Cells were then pelleted by centrifugation and the pellet was lysed in 100 l of reporter lysis buffer (Promega) for 15 min at room temperature. Debris was removed by centrifugation and the supernatant was transferred to a fresh Eppendorf tube. Twenty l of supernatant per transfection was used in the luciferase assay or in the ␤-galactosidase assay. Luciferase activity was measured on an Autolumat luminometer using a luciferase detection reagent (Promega). To measure ␤-galactosidase activity 20 l of cell lysate per transfection was added to one well of a 96-well tissue culture plate. 150 l of reaction mixture (1.25 mM CaCl 2 , 25 M chlorophenol red-␤-galactopyranoside) was added to each well and plates were incubated at 37°C until the yellow color of the mixture began changing to red. Absorbance at 570 nm was taken as a measure of ␤-galactosidase activity. All luciferase activity values were normalized to ␤-galactosidase activity in the lysate.
Rat Embryo Fibroblast Immortalization⁄Transformation Assays-Primary Fisher rat embryo fibroblasts (Biowhittaker UK Ltd.) were transfected by the calcium phosphate method with 5 g of EJ6.6 plasmid expressing Ha-ras and 10 g of pcDNA3-HA-EBNA3A wt or mutants essentially as described previously (39). Cells were supplied with fresh, supplemented media containing G418 (400 g/ml) every 3 days and after 3-4 weeks morphologically transformed colonies of cells were scored by microscope observation.

RESULTS
EBNA3A Binds to GST-CtBP through a Sequence(s) Located in the COOH Terminus, Not Through 98 PLDLR 102 -An analysis of the predicted amino acid sequence of EBNA3A had revealed a sequence in the NH 2 terminus (amino acids 98 -102) similar to the PLDLR CtBP binding motif in HDAC4, HPC2, and xPC (18). Because CtBP is a co-repressor of transcription and EBNA3A can repress transcription and is a close relative of EBNA3C (that has been shown to bind CtBP), the ability of in vitro translated EBNA3A to bind to a GST fusion with CtBP was tested. Similar experiments were also performed in parallel with EBNA3C and also EBNA3B.
This revealed that although EBNA3B showed no binding, EBNA3A appeared to bind to CtBP with even greater affinity than EBNA3C (Fig. 1A). However, mutagenesis of the PLDLR to ALDAR had no effect on binding to CtBP and parallel experiments showed that the binding activity was localized within the COOH, not the NH 2 terminus of EBNA3A (Fig. 1,  B-D). EBNA3A binding to CtBP is clearly not mediated through the 98 PLDLR 102 motif.
EBNA3A Binds to GST-CtBP through Two Cryptic "PLDLSlike" Sequences in the COOH Terminus That Co-operate to Enhance Binding-Further inspection of the amino acid sequence of the COOH terminus of EBNA3A revealed two potential, but variant, CtBP binding motifs at amino acids 857-861 (ALDLS) and 886 -890 (VLDLS). A recent study of the CtBPbinding site in E1A indicated that the initial proline residue in PXDLS was extremely important for the efficient binding and mutation to alanine dramatically reduced the affinity between CtBP and a peptide (19). Consistent with this observation, various cross-species comparisons of well characterized CtBPbinding sites showed that the initial proline is almost invariably conserved (15,18). Nevertheless, in the absence of alternative candidates these sites in EBNA3A were investigated.
Initially, by introducing stop codons, deletions of the COOH terminus of EBNA3A were produced. The truncated mutant EBNA3A 1-875 includes only 857 ALDLS 861 , whereas the truncated mutant EBNA3A 1-847 does not include either potential binding site. Pull-down experiments showed that EBNA3A 1-875 binds to GST-CtBP with reduced affinity and EBNA3A  shows no binding in similar assays (data not shown). To analyze the relative importance of these sites further, 857 ALDLS 861 was mutated to ALDAA or 886 VLDLS 890 was changed to VLDAA. In addition, another mutant was constructed in which both sites were mutated (ALDAA, VLDAA). The resulting pulldown experiments (Fig. 2A) demonstrated that when either site was mutated there was a decrease in binding and when both sites were mutated, no binding was detected. Quantitation of the binding using a phosphorimaging system revealed the relative binding to each site and showed that binding was synergistic rather than additive (Fig. 2D). The data are consistent with 857 ALDLS 861 being a very weak binding site, 886 VLDLS 890 having moderate binding activity, and together these variant sites cooperating to bind CtBP with very high efficiency. Furthermore, GST pull-down experiments using GST-RBP-JK/ CBF1 showed that the mutations specifically altered binding to CtBP and had no detectable effect on the interaction between EBNA3A and RBP-JK/CBF1 (Fig. 2B). This was not surprising because the interaction with RBP-JK/CBF1 has been localized to the NH 2 -terminal 556 amino acids of EBNA3A (34).
EBNA3A Binds to GST-CtBP with Greater Affinity Than Several Known Cellular CtBP-binding Proteins-Multiple GST pull-down experiments were then performed to compare the relative binding activity of EBNAs3A, -3B, -3C, and three cel-lular proteins that are known to bind CtBP, HPC2, HDAC4, and CtIP (see Introduction). Fig. 3 shows that in these assays that EBNA3A exhibits the most robust interaction. This presumably reflects the co-operative binding of two CtBP molecules by the two relatively low affinity sites on each EBNA3A.
CtBP and EBNA3A Interact in Vivo-Further experiments were performed to determine whether EBNA3A and CtBP interact in cells. Initially a series of co-transfections were performed using plasmids encoding CtBP and EBNA3A or EBNA3A mutants. DG75, Burkitt lymphoma-derived cells, were transfected with plasmid DNA encoding either wild type EBNA3A or the double mutant EBNA3A (ALDAA,VLDAA) with or without, plasmid DNA-encoding CtBP. Protein extracts from the transfected cells were subjected to immunoprecipitation using a rabbit polyclonal serum directed against CtBP or an irrelevant serum. Representative results shown in Fig. 4A demonstrate that more than 10% of EBNA3A wild type (WT) is co-precipitated with the specific serum only in the presence of CtBP. In contrast, none of the EBNA3A (ALDAA,VLDAA) double mutant was precipitated by the CtBP-specific serum. In these experiments it was noted that routinely a small but significant amount of WT EBNA3A was co-immunoprecipitated from extracts of cells transfected with only the EBNA3A encoding plasmid. This presumably represents the interaction of EBNA3A with the relatively low level of endogenous CtBP present in the DG75 cells.
Co-immunoprecipitation experiments were also performed on extracts from LCL cells using the anti-CtBP rabbit serum. The LCL cells express the complete range of EBV latent proteins expressed at their biologically normal level from viral episomes. The products of immunoprecipitation were then Western blotted and probed with sheep anti-EBNA3A serum. These experiments confirmed that in vivo, some fraction of Bacterially expressed GST or GST-CtBP fusion protein (CtBP) were incubated with equal amounts of [ 35 S]methionine-labeled EBNA3A, -3B, or -3C. Ten percent of the input protein in each binding reaction is shown. Bound proteins were resolved by 7.5% SDS-PAGE. B, the PLDLR motif in the NH 2 terminus of EBNA3A does not contribute to CtBP binding. GST or GST-CtBP were incubated with equal amounts of [ 35 S]methionine-labeled wild type EBNA3A (3A) or a mutant of EBNA3A in which the PLDLR motif has been mutated to form an ALDAR motif (Pro 98 to Ala 98 and Leu 101 to Ala 101 )(3A ALDAR). C, schematic representation of EBNA3A illustrating the truncations and mutants used. D, the COOH-terminal half of EBNA3A interacts with CtBP. GST or GST-CtBP were incubated with equal amounts of [ 35 S]methionine-labeled wild type EBNA3A (3A), a fragment including amino acids 1-524 (3A 1-524), or a fragment including amino acids 500 -915 of EBNA3A (3A 500 -915).

FIG. 2. Nonconsensus CtBP binding motifs in the COOH terminus of EBNA3A act cooperatively to bind CtBP.
A, GST or GST-CtBP was incubated with equal amounts of [ 35 S]methionine-labeled wild type EBNA3A (3A), a mutant in which the ALDLS motif was point mutated to ALDAA (Leu 860 to Ala 860 and Arg 861 to Ala 861 ) (AL-DAA), a mutant in which the VLDLS motif was mutated to VLDAA (Leu 889 to Ala 889 and Arg 890 to Ala 890 ) (VLDAA), or a mutant in which the ALDLS was mutated to ALDAA and the VLDLS was mutated to VLDAA (2ϫ mutant). B, a similar experiment to that illustrated in A was performed using GST-RBP-JK and showed that these mutations had no significant effect on the interaction between EBNA3A and RBP-JK/CBF1. C, schematic representation of EBNA3A illustrating the mutants used. D, the results of the binding experiments shown in panel A were quantified using a Storm 860 (Molecular Dynamics). The amount of wild type EBNA3A bound to GST-CtBP was given an arbitrary value of 100. endogenous CtBP and the physiologically normal level of EBNA3A interact and can be co-immunoprecipitated (Fig. 4B). A Western blot of the LCL extract with the anti-CtBP serum confirmed the specificity of this reagent and showed that there was no cross-reaction with the endogenous EBNA3A (data not shown).
An EBNA3A Homologue from a Nonhuman Primate ␥-Herpesvirus Also Binds GST-CtBP and the Introduction of a Second Binding Site Increases the Efficiency of Binding-An inspection of the predicted amino acid sequence of an EBVrelated ␥-herpesvirus from Rhesus monkeys revealed a homologue of EBNA3A that has a single predicted CtBP-binding site (amino acids 896 -900) (40). This Rhesus EBNA3A binds human CtBP with relatively low efficiency (Fig. 5) and mutating the 896 PIDLS 900 to 896 AIDAS 900 destroyed binding. However, introduction by site-directed mutagenesis of an additional (ALDLS) site at a similar relative position to the one in EBNA3A produced co-operative binding to CtBP. Again the ALDLS, which is itself a poor binding site, synergized with a second CtBP-interaction motif ( 896 PIDLS 900 ) located nearby.
CtBP Binding Is Involved in the Repression of Transcription by EBNA3A-It has been shown that EBNA3A fused to the DNA-binding domain of Gal4 has repressor activity on the Gal4-responsive reporter, pGL-UAS. To determine the effect of deleting or mutating the CtBP interaction domain of EBNA3A on this repression, DG75 Burkitt lymphoma-derived B cells were transfected with pGL-UAS together with pcDNA3GAL4-EBNA3A WT or pcDNA3GAL4-EBNA3A(ALDAA,VLDAA). The results of multiple experiments summarized in Fig. 6 show that whereas wild type EBNA3A consistently repressed expression from the pUAS-reporter by up to 6 -7-fold, no such repression was seen with the double mutant of EBNA3A (ALDAA,V-LDAA) that no longer binds CtBP. A COOH-terminal deletion mutant EBNA3A1-915 that retains the CtBP-binding domain retained the ability to repress the pGL-UAS reporter (data not shown). This result suggests that merely altering the structure of the COOH terminus is insufficient to ablate transcriptional repression. In these DG75 B cells, efficient repression of transcription by the GAL4-EBNA3A fusion proteins requires an interaction with CtBP.
EBNA3A Cooperates with Ha-ras in the Immortalization and Transformation of REFs but CtBP-binding Mutants of EBNA3A Are Impaired in this Activity-We previously showed that the other family member, EBNA3C, cooperates with the oncogenic ras in the immortalization and transformation of primary REFs and that this activity correlated with the ability of EBNA3C to bind CtBP (36,39). Therefore REFs were cotransfected with plasmid DNA encoding activated Ha-ras together with EBNA3A expression vectors to determine whether After resolution on 7.5% SDS-PAGE, the proteins were transferred to nitrocellulose and probed with an anti-EBNA3A antibody. B, EBVencoded EBNA3A interacts with cellular CtBP in LCL cells. Extracts from Chuck-LCL were subjected to immunoprecipitation with the rabbit polyclonal antibodies (␣-CtBP or unrelated rabbit anti-mouse, Dako) as indicated. After resolution on 7.5% SDS-PAGE, the proteins were transferred to nitrocellulose and probed with an anti-EBNA-3A antibody. EBNA3A also overcomes the premature senescence induced by the oncogenically activated ras in primary rodent cells (41). We also determined whether the interaction with CtBP is involved. The results of multiple experiments (n ϭ 6), summarized in Table I, show that EBNA3A can also rescue primary rodent cells from premature senescence induced by oncogenic ras to allow the outgrowth of immortalized and transformed colonies. EBNA3A expression was demonstrated in randomly selected clones by immunofluorescence staining and Western blotting with an anti-EBNA3A serum (Fig. 7, A and B). Consistent with the hypothesis that the interaction with CtBP may contribute to this phenotype, CtBP-binding mutants of EBNA3A were severely compromised in their ability to co-operate with Ha-ras and furthermore, co-immunoprecipitation assays using anti-CtBP serum showed that the endogenous rat CtBP could associate with WT EBNA3A and be co-precipitated from the transformed clones (Fig. 7C). Mutants EBNA3A(ALDAA), EBNA3A-

TABLE I EBNA3A cooperates with activated Ras
In each assay 10 g of each plasmid was co-transfected with 5 g of EJ6.6 (Ha-Ras). Cells were grown for 21-28 days with G418 (400 g/ml). The highly refractile morphologically transformed colonies that survive drug selection were then counted under a phase-contrast microscope. (VLDAA), and the 2ϫ mutant EBNA3A(ALDAA,VLDAA) were all expressed at similar levels to WT EBNA3A in transient assays (Fig. 4). 2 Immunofluorescence staining confirmed that all the mutant proteins had a nuclear localization (data not shown). DISCUSSION It is emerging that CtBP family members are important components of many protein complexes involved in the repression of transcription in cells from diverse species (15). To our knowledge nearly all the proteins that have been shown to bind directly to CtBP include the consensus PXDLS amino acid sequence and this is necessary (and probably sufficient) for binding. Most mutation studies suggest that the initial proline residue is essential for an efficient interaction (see Introduction). Exceptions to this rule are the very recently identified -GLDLS-motif in the tumor suppressor HIC1 (42) and perhaps the VLDLS sequence located in the minimal repression region of the Drosophila repressor Giant (43,44). In this report we have extended the variety of CtBP interaction motifs and shown that two relatively weak interaction sites that do not include the initial proline residue (ALDLS and VLDLS) can co-operate to efficiently recruit CtBP to the viral oncoprotein EBNA3A. Although in the assays performed here the VLDLS motif seems to have moderate CtBP binding activity, it should be noted that in Drosophila Giant, VLDLS may recruit an as yet unidentified co-repressor protein (43) and we cannot rule out this sequence also recruiting a protein other than CtBP to EBNA3A. It was possible to show that the second weak binding site ALDLS acts synergistically with VLDLS. This is consistent with the suggestion that CtBP can dimerize (7) and stabilize the binding of two molecules to one EBNA3A. It was also possible to confirm that a second weak (ALDLS) site can act synergistically with the resident PIDLS in another viral protein. Engineering an additional sequence into the EBNA3Alike protein from the Rhesus monkey ␥-herpes virus produced binding that appeared to be more efficient than the additive effect of two individual sites.
A surprising discovery was that the 98 PLDLR 102 sequence located in the NH 2 -terminal end of EBNA3A was not involved in the interaction with CtBP. Although this motif creates an effective CtBP-binding site in other proteins (e.g. HPC2 and HDAC4) we can only assume that in EBNA3A it is occluded by protein folding or perhaps other protein-protein interactions.
The study also showed for the first time that EBNA3A cooperates with activated Ha-ras in the immortalization and transformation of rodent fibroblasts. This activity appears to be dependent on the interaction between EBNA3A and CtBP. Thus EBNA3A resembles the related EBNA3C protein in two further respects: in its ability to rescue primary rodent cells from the premature senescence induced by oncogenic ras and in its ability to form complexes with CtBP. In both of these EBV proteins their oncogenic activity correlated remarkably well with their ability to bind CtBP.
Data base searches for proteins that include two PLDLS-like sequences separated by 30 or less amino acids identified the tumor suppressor RIZ and two other oncogenes, Evi1 and Marek's disease virus Meq (Fig. 8). EVI1 is a particularly interesting protein in that it has PFDLT and PLDLS with almost the same spacing as the bipartite domain in EBNA3A. EVI1 has been extensively studied and both sites have been shown to bind CtBP independently. Also both the transcriptional repressor and transforming/cell cycle disrupting activities of EVI1 are mediated by the CtBP interaction (9). The AML1/MDS/EVI1 fusion gene expressed in various hematolog-ical malignancies encodes a protein that binds CtBP. The interaction between this oncoprotein and CtBP is also biologically important and appears to be involved in growth deregulation and abnormal differentiation of murine hematopoietic cells (45). The similarities between EVI1 and EBNA3A and EBNA3C are quite striking and will be studied further. RIZ and Meq on the other hand have been characterized less well (46,47), nevertheless, both bind to GST-CtBP 2 and are currently the subject of further studies. Because Meq is encoded by another oncogenic herpesvirus (that infects and causes lymphomas in chickens) it will be very interesting to discover the role of CtBP in the transforming function of Meq.
Both EBNA3A and EBNA3C also bind to the cellular DNAbinding protein RBP-JK/CBF1, which is an effector component of the Notch signaling pathway and equivalent to the D. melanogaster suppressor of Hairless. In EBV-immortalized cells, EBNA2 mimics activated Notch by binding to RBP-JK/CBF1 and activating transcription of viral and cellular genes (30). In Drosophila, Notch signaling is antagonized by the Hairless protein and it has been proposed that EBNAs3A and -3C function in a similar manner to Hairless to repress transcriptional activation mediated by EBNA2 and RBP-JK/CBF1 (30). It has recently been revealed that Hairless acts as an adapter, targeted to DNA by suppressor of Hairless, and recruits the long range co-repressor Groucho and the short-range co-repressor Drosophila CtBP (6,48). The interaction between Hairless and Drosophila CtBP is through a PXDLS-like motif (PLNLS), so it would seem that EBV has faithfully mimicked several components of the Notch effector pathway to regulate gene expression tightly and precisely.
In summary, a novel strategy has evolved in EBNA3A for efficient binding to the cellular co-repressor CtBP. Whereas the precise significance of this interaction in the immortalization of B cells and in the life cycle of EBV remains to be determined, at least two mechanisms for how EBNA3A might function to disrupt the regulation of cell proliferation and contribute to cell transformation are suggested. EBNA3A may bind CtBP and target the associated repression complex to a cellular gene(s) and so down-regulate a critical function required for the inhibition of cell proliferation. In this case, EBNA3A could be directed to DNA by its association with RBP-JK/CBF1. Alternatively, because EBNA3A has a relatively high affinity for CtBP, it might disrupt or modify cellular repression complexes that normally regulate progression through the cell cycle and so release the cell from the effect of anti-proliferative signals. Currently both of these possibilities are being explored.