The MDM2 C-terminal Region Binds to TAFII250 and Is Required for MDM2 Regulation of the Cyclin A Promoter*

MDM2 proto-oncogene expression is aberrant in many human tumors. Its normal role is to modulate the functions of p53. The N terminus of MDM2 interacts with p53, whereas the properties of the rest of the molecule are poorly understood. We show that MDM2 binds to the general transcription factor TFIID in vivo. The C-terminal Ring finger interacts with TAFII250/CCG1, and the central acidic domain interacts with TBP. Expression of MDM2 activates the cyclin A gene promoter but not c-fos, showing that the effects of MDM2 are specific. Deletion of the C-terminal region of MDM2 abolishes activation, showing that the C-terminal domain of MDM2 is functionally important. We found that increasing MDM2 expression to higher levels inhibits the cyclin A promoter. Inhibition appears to result from titration of general transcription factors because MDM2 overexpression inhibits c-fos as well as other promoters in vivo and basal transcription in vitro. The mechanisms of repression of the cyclin A andfos promoters appear to be different. Cyclin A repression is lost by deleting the C terminus, whereas that of c-fosis lost by removal of the acidic domain. These results reinforce the conclusion that the C terminus of MDM2 mediates effects on the cyclin A promoter. MDM2 transformed cells contain elevated levels of cyclin A mRNA, showing that activation occurs under physiological conditions. There is a positive correlation between MDM2 binding to TAFII250 and MDM2 activation of the cyclin A promoter. The C-terminal region of MDM2, which contains the Ring finger, interacts with TAFII250 and is required for regulation of the cyclin A promoter by MDM2. Our results link the activity of MDM2, a transforming protein implicated in many human tumors, with cyclin A, a regulator of the cell cycle.

MDM2 has many features of transcription factors, including a nuclear localization signal, a central acidic region similar to a class of activation domains, an adjacent C4 zinc finger that could interact with nucleic acids, and a C-terminal variant C3HC4 Ring finger (34,35). Ring fingers fold into a characteristic motif with two molecules of zinc and mediate interactions with nucleic acids and proteins (36,37). The N-terminal region of MDM2 interacts with p53 (21), E2F1/DP1 (32), and SV40 large T (38). Recently, the acidic region has been shown to interact with the ribosomal protein L5 and the Ring finger has been shown to interact with RNA, suggesting that MDM2 may regulate protein synthesis (39). The effects of MDM2 on p53, pRb, p107, and E2F1/DP1 suggest that one of its major functions is to regulate the cell cycle.
Cell cycle progression depends upon cyclin-cdk complexes that are regulated in a complex manner by synthesis and degradation of the cyclins, association with inhibitory subunits, and phosphorylation of activatory and inhibitory sites on the cdks (40 -43). Cyclin A is required for the S phase, passage through G 2 , and mitosis (44 -48). Cyclin A synthesis at the G 1 /S border is tightly regulated at the transcriptional level (49), whereas degradation appears to be restricted to G 2 /M (50). The promoter is repressed during the G 1 phase by factors binding the cell cycle-dependent element (CDE) and cell cycle genes homology region (CHR), and dissociation of these repressors appears to release the activity of positive regulators, leading to activation during S phase (49,(51)(52)(53). There are similar CDE/CHR elements in the cdc2 and CDC25C gene promoters that mediate regulation during the S/G 2 phase (52), whereas related elements control G 1 and G 1 /S phase expression (54). The factors that bind to the CDE and the CHR have not been unambiguously established. E2F/DP1-like factors appear to bind weakly to the CDE (51)(52)(53), and there is evidence that the binding complex contains cyclin E, cdk2, p107, and an E2F-like factor (55). The association of E2F1 and p107 with both regulation of the cyclin A promoter and the activities of MDM2 suggests that cyclin A expression could be affected by MDM2.
We have found that MDM2 interacts physically with TAF II 250/CCG1, a specific transcription accessory factor for the cyclin A promoter (56) that is required for G 1 phase passage (57). MDM2 expression stimulates the cyclin A promoter. The C-terminal domain of MDM2, which interacts with TAF II 250, is required for activation of the promoter. Cyclin A expression is elevated in MDM2 transformed cells. These results provide a link between MDM2 and cyclin A, an important component of the cell cycle control machinery.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-GST fusion proteins were expressed in Escherichia coli BL21, grown at 28°C, and purified on gluthathione-Sepharose in the presence of 10 mM DTT as described previously (58). Insect cells (Sf9) were grown at 27°C in Grace's medium, infected with recombinant baculovirus producing MDM2, and harvested 72 h postinfection. The cells were pelleted, washed twice with ice-cold phosphate-buffered saline (1 mM Na 2 HPO 4 , 10.5 mM KH 2 PO 4 , 140 mM NaCl, 14 mM KCl, pH 6.2) extracted with 3 ml of lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 3 mM DTT, 0.35 mM phenylmethylsulfonyl fluoride, 10 g/ml ␣ 2 -macroglobulin, and 2.5 g/ml proteinase inhibitors (leupeptin, pepstatin A, antichymotrypsin, antipaïn, and aprotinin) and incubated for 30 min on ice. Lysates were centrifuged at 30,000 rpm for 30 min. Clarified supernatants were incubated for 4 h at 4°C with 1 mg of polyclonal antibody 365 crosslinked to protein A-Sepharose. The bound fractions were washed three times with 10 ml of ice-cold buffer A (20 mM HEPES, pH 8.0, 150 mM NaCl, 0.2% Tween 20, 5% glycerol, 2 mM DTT, 0.1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, and 2.5 g/ml proteinase inhibitors). Proteins were eluted for 12 h at 4°C in buffer A in the presence of 2 mg of the epitope peptide and concentrated on centricon-30. TFIID was immunopurified with the monoclonal antibody 2C1 according to Jacq et al. (59). The protein fractions were aliquoted and stored at Ϫ80°C.
Plasmids and in Vitro Transcription/Translation-In vitro transcription assay with the adenovirus major late promoter were as described by Fischer et al. (60). For in vitro binding studies, MDM2 constructs were made starting from a mouse cDNA clone obtained from a mouse testis library (Stratagene, generously provided by C. Theillet, CNRS Montpellier, France; see Ref. 31 for the sequence). Plasmids pGEX2TK-MDM2 (1-489), pGEX2TK-NMDM2 (1-177), pGEX2TK-ADMDM2 (221-272), pGEX2TK-RINGMDM2 (459 -489), and p␤-MDM2 were constructed using polymerase chain reaction (PCR). The 5Ј end of the cDNA in p␤MDM2 was modified to improve the translation efficiency in rabbit reticulocyte lysates by introducing a Kozak sequence that also results in a substitution of a glycine for a cysteine residue in the MDM2 sequence. The modified cDNA was cloned downstream from the 60-nucleotide 5Ј leader of the ␤-globin cDNA. Plasmids pGEX2TK-⌬NMDM2 and p␤-⌬NMDM2 were derived from pGEX2TK-MDM2 and p␤-MDM2, respectively, by removing an AccI 430-bp fragment, which leads to a deletion of MDM2 amino acids 10 -154. The proteins were produced in rabbit reticulocyte and TNT lysates, according the manufacturer's instructions (Promega). Plasmid pET11a-hTBP is derived from pET3b-TBP (61). pET11a-hTFIIB is described by Ha et al. (62). pGEX2T-TFIID was constructed by Nick Burton using the full-length human TBP cDNA. Plasmid constructs for expression in Sf9 cells were as follows: pVLHAX-hTAF II 250 (hemagglutinin-tagged human TAF II 250) derived from pHAX-hTAF II 250 (63) was constructed by Lazlo Tora; pVL␤-MDM2 (mouse MDM2) was constructed from p␤-MDM2. For the yeast two-hybrid interaction assay, the full-length mouse MDM2 cDNA was cloned in frame with the Lex DNA-binding domain in pBTM116, TAF II 250 constructs were generated by PCR and cloned in pVP16 (64). The cyclin A promoter construct p322 is described by Henglein et al. (49). Mammalian MDM2 expression vectors have been described by Dubs-Poterszman et al. (31). Additional constructs were made in the same vector by PCR-mediated mutagenesis. The sequences of oligonucleotides are available on request.
Protein-Protein Interaction Assays-Interaction assays were performed as described by Léveillard et al. (65). For batch assays, E. coli BL21 cells transformed with pET vectors coding for human TFIIB, TBP, and yeast TBP were rinsed with methionine-free minimal medium and induced for overexpression with isopropyl-1-thio-␤-D-galactopyranoside in the presence of [ 35 S]methionine. Bacterial extracts were incubated with bound GST-MDM2 full-length or GST alone, and, after washing three times in buffer B (20 mM Tris-HCl, pH 7.5, 50 mM KCl, 1 mM DTT, 5% glycerol, 0.1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, and 2.5 g/ml proteinase inhibitors), the bound proteins were analyzed by SDS-PAGE. For mini-columns assays, pipetteman tips were packed with 40 l of GST fusion proteins bound to glutathione-Sepharose. The mini-columns were equilibrated in buffer B, and the translation products were loaded and then washed with 10 volumes of buffer B. The different fractions were analyzed by SDS-PAGE and then either processed for fluorography or blotted onto nitrocellulose and revealed using a standard Western blotting protocol.
Immunodetection and Immunoprecipitation-For immunoprecipitation and immunodetection the following antibodies were used: 1) monoclonal antibody PAb3G3 raised against a N-terminal domain of human TBP; 2) PAb12CA5 anti-hemagglutinin epitope (Boehringer); 3) polyclonal antibodies against MDM2 (365 and 370) raised against a peptide (amino acids 165-183) and the N-terminal sequence (amino acids 1-177), respectively; pooled sera were first purified with caprylic acid and ammonium sulfate precipitations and then affinity purified on peptide immobilized on Sulfolink beads (Pierce); and 4) polyclonal antibodies against TAF II 250 as described by Sekiguchi et al. (66).
The protocol for co-immunoprecipitation of MDM2 and TAF II 250 after co-infection of Sf9 cells with bacoluviruses was adapted from Hannon et al. (67). Briefly, 10 7 cells were infected at a multiplicity per cell of 2 plaque-forming units for p53 and 5 plaque-forming units for MDM2 and TAF II 250. After 50 h at 27°C the medium was replaced with methionine-free Grace's medium containing [ 35 S]methionine and incubated for a further 5 h at 27°C. Lysates were prepared as described by Hannon et al. (67) and used for batch interaction assays as described above or for immunoprecipitation. The lysates were preincubated with protein A-Sepharose for 1 h at 4°C and then with antibodies bound to protein A-Sepharose for 4 h at 4°C. After washing in lysis buffer, the bound proteins were analyzed by SDS-PAGE and processed for fluorography. Dual hybrid interaction assays were performed according to Vojtek and Hollenberg (64). To quantify ␤-galactosidase expression, yeast clones were grown in the appropriate selective media and assayed using ONPG. ␤-Galactosidase units are normalized to the amount of protein.
Cells and Transfection-The BHK21 and tsBN462 cell lines are described by Sekiguchi et al. (66). Cells were transfected by the calcium phosphate technique described by Chen and Okayama (69, 70) with 1.5 g of the reporter plasmids, the expression plasmids described in the figure legends, and carrier DNA (empty expression vector) to make the total amount of DNA up to a 15 g. After 14 h, the cells were washed and refed with medium containing 0.07 or 10% fetal calf serum and incubated at the temperatures indicated in the figure legends. Luciferase activity was measured using a glow luminescence kit with either a Wallac luminometer or with flash luminescence reagents and an EG&G luminometer.

MDM2
Interacts with TFIID in Vivo and in Vitro-We investigated the possibility that MDM2 interacts with TFIID in vivo in 3T3DM cells that are transformed by MDM2 (71). Crude extracts were immunoprecipitated with antibodies against TBP, a component of TFIID, and MDM2 was revealed by Western blotting. The immunoprecipitate contained specifically associated MDM2 (Fig. 1A, lane 2), which migrated at the expected molecular weight (90,000, compare with input in lane 1). MDM2 was not detected in controls lacking either antibodies or extract (lanes 3 and 4). Another component of the TFIID complex, TAF II 250, also co-immunoprecipitated (lane 6). TAF II 250 was detected with specific antibodies (lane 6), was migrated at the expected molecular weight (compare with input, lane 5), and was absent in the controls (lanes 7 and 8). These results show that MDM2 complexes with TBP in vivo.
We studied MDM2-TFIID interactions in vitro. We initially investigated the interactions between GST-MDM2 and TBPassociated complexes in cell extracts. HeLa whole cell extracts (0.6 M heparin fraction) (72) were fractionated on GST-MDM2 or control GST columns, and TBP was detected by Western blotting. TBP was retained by GST-MDM2 (Fig. 1B, lanes 5-7) but not by GST (lanes 1-4), showing that MDM2 interacts with TBP or associated complexes in vitro. We investigated whether the TFIID component TAF II 250 is retained by GST-MDM2. TAF II 250 can be unambiguously identified by both its size and specific labeling in vitro when a phosphocellulose-D fraction is incubated with [␥-32 P]ATP (73,74). In vitro labeled TFIID was immunopurified with an antibody against TBP (59) and applied to GST or GST-MDM2 mini-columns (Fig. 1C, lanes 1 and 2). Bound proteins were analyzed by SDS-PAGE followed by autoradiography. A labeled protein of the expected size was specifically retained by MDM2 (lanes 1 and 2, upper band labeled with an arrow), showing that the TFIID complex binds to MDM2. Several other bands were also detected, only one of which was specific (see lower band labeled with an arrow). The identity of this protein has not been established.
TBP Interacts with the Acidic Domain of MDM2-To study whether TBP associates with MDM2 in the absence of other components of TFIID, we used [ 35 S]methionine-labeled proteins produced in E. coli. Human and yeast TBP were compared because they have highly related C-terminal domains. TFIIB was also studied because it is a basal transcription factor that mediates critical interactions with a number of activators (75,76). Crude extracts from E. coli containing the overexpressed and other labeled proteins ( Fig. 2A, lanes 1 and 4) were incubated with either GST or GST-MDM2. Human TBP was found to interact specifically with MDM2, as shown by both the amount retained on GST-MDM2 compared with GST (lanes 5 and 6) and the preferential retention over the labeled E. coli proteins. Yeast TBP also interacted specifically with MDM2, although with lower affinity than human TBP (data not shown), suggesting that the conserved domain mediates the interactions. TFIIB did not interact with MDM2 (lanes 1-3), providing further evidence for the specificity of the interaction with TBP and suggesting that TFIIB is not directly involved in regulation of transcription by MDM2.
Human TBP-MDM2 interactions were studied further using various combinations of in vitro translated proteins. MDM2 translated in vitro was retained specifically on GST-human TBP (Fig. 2B, lanes 1-6). Conversely, in vitro translated human TBP bound to GST-MDM2 (lanes 7-12), indicating that the nature of the fusion did not affect the interactions. A TBP-MDM2 complex was detected with purified proteins (results not shown). Taken together, these results show that the interactions are direct and are most probably not mediated through a protein bridge.
We used MDM2 deletion mutants to localize the sequences that interact with TBP. We initially established that the p53binding domain is not required.  (73), and TBP containing complexes were immunopurified with the 2C1 TBP monoclonal antibody followed by elution with the epitope peptide (59). The immunopurified proteins were applied to mini-columns packed with GST (lane 1) or GST-MDM2 (lane 2), and, after washing, the bound fractions were resolved on SDS-polyacrylamide gels. The arrows indicate two specifically retained polypeptides, of which the lower mobility protein is TAF II 250, identified by both size and auto-phosphorylation (73). The asterisks indicate nonspecific bands.
MDM2 Interacts with TAF II 250 -MDM2 and TAF II 250 are implicated in cell cycle control, suggesting that their activities could be linked. We investigated whether TAF II 250 and MDM2 interact physically using several in vivo and in vitro assays. Initially, interactions were studied by co-immunoprecipitation of MDM2 and hemagglutinin-tagged TAF II 250 expressed by baculoviruses in insect Sf9 cells. The cells were labeled with [ 35 S]methionine, and extracts were precipitated with MDM2 antibodies (Fig. 3A). The MDM2 antibodies were initially shown to be specific.  (Fig. 2B), did not retain TAF II 250 (Fig. 3A, lane 7). In contrast, the C-terminal Ring finger motif (amino acids 432-489) interacted specifically with TAF II 250 (Fig. 3A, lane 9) with an efficiency similar to both full-length MDM2 (lane 8) and TBP (lane 10), which is known to form a strong complex with TAF II 250. These results show that the Ring finger domain of MDM2 is sufficient to mediate efficient and specific interactions with TAF II 250.
The TAF II 250 sequence that interacts with MDM2 was localized by both in vivo and in vitro approaches. Initially we used the yeast dual hybrid system. TAF II 250 fragments containing functional domains predicted by homology (66) were fused to the VP16 transactivation domain (Fig. 3B). They were tested for their ability to interact with MDM2 fused to the LexA DNA-binding domain by the ␤-galactosidase assay. Among the ten fragments tested, only the HMG homology region (construct 5) gave a significant and reproducible increase in ␤galactosidase activity. The interactions were also studied in vitro using GST-MDM2, and TAF II 250 fragments were synthesized in reticulocyte lysates (Fig. 3C). MDM2 interacted specifically with the HMG region (lanes 5 and 6), with about onethird the efficiency of the acidic domain of p53 (compare lanes 4 and 6 with lanes 13 and 15). In contrast, it did not interact with mutants that were negative in the yeast assay (SWI4, DR1, and acidic domains; lanes 1-3, 7-9, and 10 -12). These results show that the TAF II 250 HMG-like region interacts with MDM2. It also interacts with TFIIF (78) and E1A (79), raising the possibility that MDM2 might regulate transcription by affecting TAF 11 250-TFIIF interactions.
MDM2 Activates the Cyclin A Promoter-TAF II 250 is specifically required for transcription from the cyclin A promoter (56). Because MDM2 interacts physically with TAF II 250, we tested the possibility that it may regulate the cyclin A promoter using transient transfection assays. BHK cells were transfected with various amounts of an MDM2 expression vector (Fig. 4A) (66,79), the N and C-terminal kinase domains (74), the histone acetyl transferase domain (HAT) (108), the G residue that is mutated to D in tsBN462 (57), and the binding domains for TBP (109,110) and RAP74 (74,78). C, the HMG-like domain of TAF II 250 interacts with MDM2 in vitro. TAF II 250 mutants analyzed in the yeast dual hybrid system (construct 3, lanes 1-3; construct 5, lanes 4 -6; construct 7, lanes 7-9; construct 10, lanes 10 -12) and p53 (amino acids 1-52; lanes [13][14][15] were translated in reticulocyte lysate and either analyzed directly (Load) or fractionated on GST or GST-MDM2 as indicated. The asterisk indicates a nonspecific protein labeled during translation. correction for transfection efficiency (data not shown and see below). Activation was lower with the highest amount of expression vector (1 g), suggesting that MDM2 overexpression titrates a limiting factor (see below). The activity of the c-fos promoter is not affected by a mutation in TAF II 250 (56), raising the possibility that MDM2 expression would not affect its activity. We co-transfected a c-fos promoter-CAT reporter together with the cyclin A-luciferase reporter. MDM2 expression had no effect on c-fos promoter activity with levels of MDM2 expression vector that maximally stimulated the co-transfected cyclin A promoter (Fig. 4B, 0.25 g of MDM2). However, the c-fos promoter was inhibited with the highest amount of MDM2 expression vector, again suggesting that MDM2 overexpression titrates a limiting factor. These results suggest that MDM2-TAF II 250 interactions may lead to specific activation of the cyclin A promoter.
The C Terminus and the Nuclear Localization Signal of MDM2 Are Required for Activation of the Cyclin A Promoter-We tested whether the MDM2 sequences that mediate interactions with TAF II 250 are required for activation of the cyclin A promoter (Fig. 5A). Mutant 1, lacking 30 amino acids from the C terminus, did not activate (Fig. 5A), even though it was expressed in similar amounts as the wild type protein, as determined by Western blotting of cell extracts with MDM2 antibodies (Fig. 5B, lanes 1-3). Mutant 1 was inactive, even when the quantity of expression vector was increased (up to 5 g; data not shown). Mutants 2 and 3, with further deletions from the C terminus, were also inactive (Fig. 5A) and were expressed at levels comparable with that of the wild type protein (Fig. 5B, lanes 4 -7). These results show that the C-terminal region (amino acids 390 -489) is required for cyclin A promoter activity. Because transcription occurs in the nucleus, we investigated the effect of deletion of the nuclear localization signal. Mutant 4 that lacks the nuclear localization signal was inactive (Fig. 5A), cytoplasmic (as determined by immunocytofluorescence; data not shown), and expressed at similar levels as the full-length protein (Fig. 5B, lanes 8 -10). These results show that MDM2 acts in the nucleus, in agreement with it having a direct effect on transcription. It is noteworthy that mutants 2 and 3 were inactive, even though they can inhibit p53-mediated transactivation (31). This indicates that activation does not result from titration of endogenous p53 that might inhibit the cyclin A promoter (80,81).
MDM2 Activation of the Cyclin A Promoter Is Mediated by the CDE-Serum stimulates cyclin A RNA levels, suggesting that it could affect activation by MDM2. We found that high serum stimulated the transfected cyclin A reported about 20fold (data not shown) and abrogated MDM2 activation with levels of expression vector that were optimal under low serum condition (Fig. 6A, see 0, 0.05, and 0.25 g of expression vector and compare with Fig. 4A). These results suggest that MDM2 and serum activate the cyclin A promoter through the same pathways. To test this hypothesis directly we investigated whether mutation of the CDE would affect activation by MDM2 in low serum. The transfections contained reporters with cyclin A promoter sequences from Ϫ214 to ϩ100, with or without a CDE mutation (52). The CDE mutation raised promoter activity about 6-fold (Fig. 6B), as expected from derepression. The CDE mutation also abrogated transactivation by MDM2 (compare wild type and CDE mutated reporters; Fig. 6B). These results suggest that MDM2 activates the cyclin A promoter through effects on the cell cycle-dependent repressor.
Overexpression of MDM2 Inhibits the Cyclin A and Various Other Promoters-In high serum, increasing the levels of MDM2 expression inhibited the cyclin A promoter (Fig. 6A, 1  and 5 g). Overexpressed MDM2 may inhibit by titrating limiting factors with which it interacts, such as TBP and TAF II 250. TBP is a general transcription factor that is required for basal transcription, raising the possibility that MDM2 overexpression could inhibit promoters other than cyclin A in vivo and basal transcription in vitro. To investigate these possibilities, we studied the effects of MDM2 overexpression in high serum on various promoters. MDM2 inhibited all the reporters that we tested, including those that contained the SV40 and TK promoters, the SV40 promoter with its enhancer, the Rous sarcoma virus long terminal repeat, and the c-fos promoter (Fig. 7A). We also studied the effects of MDM2 on basal transcription in vitro from the minimal adenovirus major late promoter. In two independent experiments, the addition of increasing quantities of purified MDM2 inhibited specific transcription (see Fig. 7B, lanes 1 and 4 -7), whereas a control extract that was purified in the same way from mock infected cells had no effect (lane 8). Transcription required TFIID and RNA polymerase II, as shown by rescue of heat-inactivated extracts with purified TFIID (lanes 3 and 9) and by inhibition with ␣-amanitin (lane 2). These results indicate that MDM2 inhibits through interactions with general transcription factors.

Different Sequences of MDM2 Are Required for Inhibition of the Cyclin A and c-fos Promoters-MDM2
inhibition is observed with both the cyclin A and c-fos promoters, whereas activation is specific for cyclin A, suggesting that different factors are involved in both processes. We tested this possibility using mutants of MDM2 that affect interactions with TBP and TAF II 250. BHK cells growing asynchronously in high serum were co-transfected with large amounts (5 g) of expression vectors and the cyclin A-luciferase and fos-CAT reporters, and extracts were assayed for both luciferase and CAT activity. Mutant 1 with a small deletion of C-terminal sequences (amino acids 461-489) repressed both promoters (Fig. 7C). Mutant 2, which lacks the whole of the C-terminal region, was substantially inactivated toward inhibition of the cyclin A promoter but still efficiently inhibited c-fos (Fig. 7C). Mutant 3, which lacks in addition the acidic domain, no longer repressed c-fos. These results show that the cyclin A and c-fos promoters respond differently to mutation of MDM2. They suggest that interactions of the C terminus with TAF II 250 and the acidic domain with TBP mediate repression. The results with mutant 1 are not necessarily incompatible with these conclusions. Mutant 1 has lost the ability to activate the cyclin A promoter (Fig. 5A) but not repression. It may lack only part of a larger functional domain, because a mutant with a deletion of the adjacent region (amino acids 365-426) did not activate the cyclin A promoter (data not shown). The small C-terminal deletion may only partially affect interactions, and in excess it may still titrate a limiting factor.
Elevated Levels of Cyclin A mRNA in MDM2 Transformed Cell Lines-The ability of MDM2 to activate the cyclin A promoter in transfection assays raised the possibility that cyclin A levels may be elevated in MDM2 transformed cells. We analyzed cyclin A gene expression in several MDM2 transformed cell lines. 3T3DM cells are spontaneously transformed fibroblasts that have multiple copies of the mdm2 gene and express high levels of MDM2 (71). N/mdm2 cells contain a mdm2 gene that was introduced on a cosmid and express lower levels of MDM2. N/pCV001 are the equivalent control cells that contain the empty vector (34). Cyclin A RNA was analyzed by both Northern blotting (Fig. 8A) and semi-quantitative PCR (Fig.  8B), using as internal controls either ribosomal RNA (Fig. 8A,  lanes 1-3) or glucose-6-phosphate dehydrogenase (Fig. 8B,  lanes 1-3). We found in serum deprived cells that there are elevated levels of cyclin A RNA in both MDM2 transformed cell lines and especially in 3T3DM cells, which express a greater amount of MDM2 (Fig. 8, A, lanes 4 -6, and B, lanes 1-3). These results suggest that one of the consequences of transformation by MDM2 is an increase in cyclin A gene expression.

MDM2
Interacts with TBP and TAF II 250 -We have found that MDM2 interacts in vivo and in vitro with the general transcription factor TFIID. In contrast, MDM2 does not bind to TFIIB, another general factor that interacts with a number of activators and co-activators (75,76,(82)(83)(84). Our in vivo and in vitro studies show that two different domains of MDM2 contact two subunits of the complex, the acidic domain TBP and the Ring finger TAF II 250. Importantly, these interaction domains lie in a poorly characterized part of MDM2, C-terminal to the p53-binding domain.
We detect MDM2-TBP interactions in a number of ways. MDM2 is present in complexes immunoprecipitated from 3T3DM cell extracts with antibodies against TBP. Immobilized MDM2 interacts with TBP in HeLa cell extracts. TBP forms complexes with different transcription accessory factors that have specific functions. These complexes are involved in transcription by the three RNA polymerases (85,86) and in specific functions of RNA polymerase II (59,(87)(88)(89). The TBP complexes isolated from both 3T3DM and HeLa extracts contain TAF II 250, a component of TFIID, suggesting that MDM2 is involved in transcription mediated by RNA polymerase II.
However, we cannot exclude the possibility that MDM2 also associates with other TBP complexes. MDM2 binds to isolated subunits of TFIID, suggesting that it interacts directly with TFIID, rather than indirectly through other factors. MDM2-TFIID complexes are present in transformed 3T3DM cells that overexpress MDM2 (34,71). It remains to be seen if TFIID-MDM2 interactions are important for transformation.
There are at least three distinct interaction domains in MDM2. The central region mediates interactions with TBP, because N-and C-terminal sequences (1-154 and 303-491) can be deleted without affecting binding (see "Results"). The major component of the interaction domain (219 -273) is within the acidic region (220 -300). It extends around a sequence (underlined in Fig. 2B) that resembles the TBP-binding motifs of c-fos, VP16, and EIA (77) (VEFEVESLD in MDM2, SSFVFTYPE in c-fos, EEFVLDYVE in EIA, and DDFDLDMLG in VP16, where acidic or identical amino acid in at least two of the sequences are underlined). The central region also binds the L5 ribosomal protein (39).
The C-terminal Ring finger region of MDM2 binds to TAF II 250 with similar efficiency as full-length MDM2, whereas the acidic region does not interact. The affinity is comparable with that of TBP, a well established interacting protein. The structures of several Ring fingers have been reported (90,91). MDM2 has a variant Ring finger motif, with a threonine in the place of the third cysteine of the C3HC4 motif. This could be a conservative substitution of a hydroxyl group for a sulfydryl that coordinates zinc (35). It remains to be seen if the Ring finger structure is important for the interaction of MDM2 with TAF II 250. Interestingly, the MDM2 Ring finger has recently been shown to interact specifically with RNA (39). Ring fingers were originally proposed to be nucleic acid-binding structures and have been shown to interact with nucleic acids. These  (34) were grown in low (0.07%) serum, and cyclin A mRNA levels were analyzed by Northern blotting (A) or semi-quantitative reverse transcription-PCR (B). RNA from 3T3DM cells was reverse transcribed with murine leukemia virus reverse transcriptase for 2 h at 37°C, heated to 94°C for 5 min, and then amplified with Vent polymerase and 30 cycles of 94°C for 1 min, 60°C for 1 min, and 72°C for 3 min with specific primers (VE 118 and VE 119 for cyclin A; PL114 and PL115 for glucose-6-phosphate dehydrogenase). The cyclin A-specific product (500 bp) was also used as a probe for the Northern blots after cloning in pBluescriptSKϩ and sequencing. Equal loading was verified by ethidium bromide staining of ribosomal RNA (A, 28S and 18S) or analysis of glucose-6-phosphate dehydrogenase (B, G6PDH).
containing proteins have been proposed to have an important role in forming large multi-protein complexes, such as the promyelocytic leukemia nuclear bodies. Ring fingers may participate in protein-protein interactions that act as a glue for assembly (37). However, MDM2 has not been described to form large assemblies, pointing to a different role for the MDM2 Ring finger. Elenbaas et al. (39) have proposed that the ability of MDM2 to contact both the L5 ribosomal protein and RNA through separate domains may indicate a role for MDM2 in the cytoplasm in regulating translation. Interestingly, the ability of the same two regions to contact two subunits of TFIID point to a nuclear function of MDM2 in regulating transcription.
The N-terminal domain of MDM2 has been most extensively studied. It is highly conserved across species and interacts with a number of proteins, most notably p53. The structure of the N-terminal region complexed to the activation domain of p53 has been determined (21). The binding domain englobes amino acids 17-125 and consists of a repetition of a ␤-␣-␤-␣-␤ motif that forms a deep hydrophobic cleft on which the amphipathic ␣ helix of p53 binds. The N-terminal domain (1-220) also binds to E2F1 and DP1 (32). There is a sequence resemblance between E2F1 and p53 that suggests that the interactions may be similar but not identical (21).
The interactions of MDM2 with other proteins are less well studied. MDM2 binds to SV40 T antigen through a domain that overlaps but is distinct from the one that binds to p53 (MDM2 amino acids 58 -220) (38). The binding site for pRb has not been mapped (30). Functional studies point to other interactions. MDM2 expression inhibits MyoD-dependent transcription (33). MDM2 enhances the expression of basic fibroblast growth factor and platelet-derived growth factor in rat astrocytes, although the mechanisms have not been studied (92). The identification of proteins that interact with MDM2 will help to establish its functions in both p53-dependent and -independent pathways.
MDM2 Activates the Cyclin A Promoter-We have found that MDM2 expression activates reporters containing the cyclin A promoter in transfection assays (see "Results" and data not shown) and that cyclin A mRNA levels are elevated in transformed cells that overexpress MDM2. Activation of the cyclin A promoter is specific because the c-fos promoter, which is upregulated earlier in G 1 , is not activated by MDM2. However, higher levels of MDM2 expression lead to general repression of many different reporters. Inhibition appears to result from the titration of general transcription factors because MDM2 inhibits basal transcription in vitro, and MDM2 sequences that mediate interactions with TBP and TAF II 250 are required for repression. Inhibition probably does not result from cell death, because MDM2 overexpression does not affect the morphology and the survival of the cells, as indicated by both immunocytochemical and FACS analysis. 2 The levels of MDM2 that activate the cyclin A promoter in our assays are similar to those in 3T3DM cells. MDM2 under these conditions can be expected to have effects that are physiological relevant, especially for situations in which it is expressed at elevated levels. MDM2 levels are high in various transformed cells, in response to DNA damage following p53 induction (7,9,27), and in some mouse tissues during development. 3 MDM2 may regulate the cyclin A promoter directly or indirectly. The cyclin A promoter has been shown to be repressed during the G 1 phase and activated during S 1 entry by derepression of the critical CDE and CHR elements (49,(51)(52)(53). We have found that MDM2 activates the cyclin A promoter when it is repressed in arrested cells but not when it is active in growing cells. Furthermore, promoter sequences that mediate repression in G 1 (the CDE) are required for activation, suggesting that MDM2 has an effect on the factors that mediate cell cycle control of the promoter. MDM2 may regulate the promoter indirectly through effects on cell cycle control mechanisms before S 1 entry or more directly through interactions with regulators of the promoter. MDM2 activates the promoter about 4-fold in our assays, which is comparable with the indirect effects of Myc in RAT1A cells (about 3-fold; see Ref. 93) and HPV E7 in NIH373 cells (about 7-fold, see Ref. 94). However, the relatively low but reproducible level of activation may reflect a high basal activity of the promoter in our assays, because promoter activity is substantially inhibited by a transdominant mutant of MDM2 that contains essentially the Cterminal region of MDM2 (⌬30 -388; data not shown). MDM2 probably activates transcription through interactions with proteins rather than DNA. There have been no reports of MDM2 binding specifically to DNA, and we did not see any evidence of MDM2-DNA binding in selection experiments using randomized DNA sequences and MDM2. 4 MDM2 may interact with the proteins that have been reported to bind to the CDE and CHR, namely cyclin E, cdk2, p107, and E2F (55). Interestingly, MDM2 can overcome a cell cycle block imposed by p107 (31), and MDM2 binds to p107 in our assays, 5 suggesting that MDM2 could target the p107 component of the complex. However, Xiao et al. (30) did not detect MDM2-p107 interactions in their experiments. Alternatively, MDM2 might interact with the E2F-like component, because it binds to the potentially related factors E2F1 and DP1 (32).
MDM2 interacts specifically with TAF II 250 both in vivo and in vitro. Deletion of the TAF II 250-binding site of MDM2 abolishes both activation of the cyclin A promoter by low levels of MDM2 and repression when it is overexpressed. The effects are specific for the cyclin A promoter, because MDM2 does not activate the c-fos promoter, and repression is still observed when the TAF II 250-binding sequences of MDM2 are deleted. Overexpression of the TAF II 250-binding site of MDM2 represses the cyclin A promoter but has no effect on c-fos (data not shown). TAF II 250 is specifically required for the activity of the cyclin A promoter but not c-fos, as shown by the effects of a temperature-sensitive mutant (56). These observations suggest that the effects of the mutation and MDM2 could be linked. However, we have found that MDM2 activation is not affected by the temperature-sensitive mutation, because activation is observed at both the permissive and nonpermissive temperatures (data not shown). MDM2 may directly overcome the effects of the mutation or may affect other functions of TAF II 250 that are not affected by the temperature-sensitive mutation. Interestingly, SV40 large T can also both rescue the TAF II 250 temperature-sensitive mutation and form complexes with TFIID in vivo (95). In contrast, EIA, which requires binding to p107 to activate the promoter (96), cannot overcome the block (95). Further studies will help to understand the mechanisms by which MDM2 activates the cyclin A promoter.
A Role for Activation of the Cyclin A Promoter in Cell Cycle Regulation by MDM2-Stimulation of cyclin A expression by MDM2 at the transcriptional level provides a potential link between MDM2 and positive regulation of the cell cycle. MDM2 targets other factors involved in cell cycle regulation. It inhibits pRb (30), stimulates E2F1/DP1 (38), enhances the expression of angiogenic mitogens (92), overcomes a cell cycle block by p107 (31), and inhibits differentiation induced by MyoD (33). Cyclin A is required for the G 1 /S transition and passage through S and G 2 (44,46,47). Transcriptional regulation of cyclin A expression is important for S phase entry (49) and is the target of many cellular signals and factors, such as cell adhesion, cAMP, TGF-␤1, p53, p107, cyclin E, p27 KIP1 , and Myc (55, 81, 93, 96 -101). It is also targeted by the viral oncogenes HPV 16 E7, adenovirus EIA, and SV40 T (94 -96). Cyclin A overexpression advances S phase entry (48,102) and confers anchorage-independent growth (97), whereas loss of cyclin A results in early embryonic lethality (103).
MDM2 levels are principally regulated by p53 in response to DNA damage (7,9,27). MDM2 induction is a late response that follows p21 Cip1/Waf1 , the major mediator of p53-induced G 1 cell cycle arrest (4,104,105). Induction of MDM2 correlates with recovery of normal DNA synthesis and may serve either to signal successful repair or to limit the severity or length of the arrest (7,98). MDM2 inhibits both the transcriptional and apoptotic functions of p53. Diminishing the levels of MDM2 in response to DNA damage by cisplatin induces apoptosis (27). MDM2 appears to have other effects in addition to down-regulation of p53, such as induction of cyclin A. p53 has been shown to suppresses cyclin A expression (80,81). It is unlikely that the effects of MDM2 are through p53, because expression of the p53-binding site of MDM2 does not activate the promoter. Interestingly, p53 regulates the expression of other factors that signal cell cycle re-entry, namely TGF␣ and cyclin G. The TGF␣ and cyclin G promoters are transcriptionally regulated by p53. TGF␣ production is proposed to stimulate proliferation of surrounding cells that replace the damaged cells that have been eliminated by apoptosis (106). Similarly, cyclin G induction promotes cell growth (107). MDM2 overexpression in tumors may mimic the normal mechanisms that re-establish cell growth following recovery from DNA damage.