Physical and Functional Interactions between PML and MDM2

, The tumor suppressor protein PML and oncoprotein MDM2 have opposing effects on p53. PML stimulates p53 activity by recruiting it to nuclear foci termed PML nuclear bodies. In contrast, MDM2 inhibits p53 by pro-moting its degradation. To date, neither a physical nor functional relationship between PML and MDM2 has been described. In this study, we report an in vivo and in vitro interaction between PML and MDM2 which is independent of p53. Two separate regions of PML are rec-ognized which can interact with MDM2. The C-terminal half of PML, encoded by residues 300–633, can interact with the central region of MDM2 which includes the MDM2 acidic domain. In addition, PML amino acids 1–200, which encode the RING-finger and most of the B box zinc binding motifs, can interact with the C-termi-nal, RING-finger containing region of MDM2. Interest-ingly, PML mutants in which sumoylation at lysine 160 was inhibited displayed an increased association with MDM2, suggesting that sumoylation at this site may be a determinant of PML-MDM2 binding. Coexpression with

Wild-type PML is a tumor suppressor and ubiquitously expressed nuclear phosphoprotein. The PML gene was originally identified as a result of a reciprocal translocation t (15:17) associated with acute promyelocytic leukemia (1)(2)(3)(4). The t (15:17) translocation disrupts the PML gene on chromosome 15 and the retinoic acid receptor (RAR) 1 ␣ (RAR␣) gene on chromosome 17 and is reciprocal in nature, resulting in the generation of novel fusion proteins PML-RAR␣ and RAR␣-PML (5). The most striking feature of wild-type PML is its localization to distinct nuclear foci that have been termed PML nuclear bodies (PML-NBs), Kremer bodies, ND10 or PODs (for PML oncogenic domains). These PML-NBs are multiprotein complexes that are 0.1-0.2 m in diameter, and cells typically contain 10 -30 PML-NBs/nucleus, although their number and size can vary during the cell cycle (6,7). In acute promyelocytic leukemia cells, expression of PML-RAR␣ causes disruption of PML-NBs and a redistribution of PML to a microspeckled nuclear localization pattern (8 -10). PML-RAR␣ can form heterodimers with wild-type PML (8,10) as well as retinoid X receptor, another retinoic acid receptor family member (11). Current models suggest that sequestration of normal PML by PML-RAR␣ inhibits the growth suppressive activity of PML, whereas sequestration of retinoid X receptor prevents the induction of differentiation (5). Inhibition of both of these pathways may be necessary for leukemogenesis.
The PML protein contains well characterized zinc binding domains in its N-terminal half, including a RING-finger adjacent to two cysteine/histidine-rich motifs known as B boxes. These domains, together with an ␣-helical coiled-coil domain, comprise a conserved motif known as RBCC (12). PML is covalently modified by SUMO-1, a small ubiquitin-like polypeptide also known as sentrin-1, UBL-1, or PIC-1 (13). Like ubiquitin, SUMO-1 is linked covalently to lysine residues on PML and other target proteins in an ATP-dependent manner (14). However, SUMO-1 modification seems to modulate the localization of its target proteins rather than induce their degradation. Three major sites for SUMO-1 modification were identified in PML: Lys-65 in the RING-finger, Lys-160 in the first B box, and Lys-490 in the nuclear localization signal (15,16). Sumoylation of PML is essential for formation of PML-NBs and for its ability to recruit NB-associated proteins (17,18). Sumoylation was also thought to be required for PML-dependent growth suppression. However, this latter hypothesis was challenged by a recent study in which a mutant PML deficient in sumoylation maintained its ability to inhibit growth when overexpressed (19).
The mechanisms by which PML functions as a tumor suppressor and inhibits growth have not been fully clarified. Mice with a targeted disruption of the PML locus (PMLϪ/Ϫ) display an increased incidence of carcinomas after treatment with tumor-promoting agents (20). Further, cells derived from PMLϪ/Ϫ mice are resistant to apoptosis in response to various apoptotic stimuli (21). These results suggest that PML plays a central role in apoptosis signaling. A relationship between PML and p53 was suggested by the finding that p53 could interact directly with PML and was recruited by PML into PML-NBs (22). Subsequent studies have shown that PML corecruits both p53 and CBP/p300 to PML-NBs. CBP/p300 then promotes the acetylation of p53, which increases p53 DNA binding affinity and thus leads to an activation of p53-responsive genes (23,24). Perhaps the most compelling evidence that links the PML and p53 growth-suppressive pathways comes from studies of oncogene-induced senescence. In these studies, p53 activity was measured in PMLϩ/ϩ and PMLϪ/Ϫ cells infected with retroviruses expressing an activated ras oncogene (Ras V12). Ras V12 expression in PMLϩ/ϩ cells resulted in the activation of p53, recruitment of p53 into PML-NBs, and the induction of premature senescence. In contrast, p53 was neither activated nor recruited into PML-NBs in PMLϪ/Ϫ cells, and the cells were resistant to ras-induced senescence (23,24). These studies provide strong evidence that PML is required for the efficient activation of p53 in response to aberrant oncogene signaling.
p53 levels and activity are controlled in large part by MDM2, the product of a p53-inducible gene. MDM2 binds to the N terminus of p53 and inhibits the activity of p53 as a transcription factor (25,26). Importantly, MDM2 binding also promotes the ubiquitination of p53 and its subsequent degradation by the proteasome, as well as the export of p53 from the nucleus to the cytoplasm (27)(28)(29)(30). Insofar as PML and MDM2 have opposing effects on p53 activity, it is of interest to investigate the functional relationship between PML and MDM2. However, to date neither a physical nor functional interaction between PML and MDM2 has been described. The current report demonstrates an in vivo and in vitro interaction between PML and MDM2 which can occur independent of p53. Mapping studies indicate that PML and MDM2 can interact through at least two separate protein regions and that these interactions can affect both the localization and activity of PML.
Tissue Culture, Immunoblots, Immunoprecipitation-Human osteosarcoma cell lines Saos-2 cells (p53-null) and U2OS (p53 wild-type), and 35-2 cells (murine p53 and MDM2 double knockout) were grown in minimum essential medium supplemented with 10% fetal bovine serum, 100 g/ml penicillin and streptomycin. Transfections in Saos-2 and U2OS cells were done using the calcium phosphate method in 35-mm dishes when the cells were ϳ80% confluent. 16 -20 h after addition of the DNA precipitate, cells were washed twice with minimum essential medium minus serum and refed with minimum essential medium plus 10% fetal bovine serum. Cell extracts were prepared 8 -10 h later. Transfections in 35-2 cells were done using FuGENE-6 (Roche Applied Science), according to the manufacturer's instructions. For immunoblot analysis and coimmunoprecipitations, cells were rinsed with phosphate-buffered saline and scraped into 500 l of lysis buffer (50 mM Tris, pH 7.5, 5 mM EDTA, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 5 g/ml leupeptin). The scraped cells were lysed on ice for 30 min with occasional light vortexing, followed by a 15-min centrifugation to remove cellular debris. Protein extracts were then either immunoprecipitated or resolved by SDS-PAGE, and transferred to a PolyScreen polyvinylidene difluoride transfer membrane (PerkinElmer Life Sciences). Antibodies used for immunoblotting were the anti-HA monoclonal antibody (HA.11 from Babco), an anti-MDM2 monoclonal antibody (SMP-14 from Santa Cruz), or an anti-FLAG monoclonal antibody (M5 from Sigma). For immunoprecipitations, 300 g of transfected cell extract was immunoprecipitated with 0.6 g of the anti-HA polyclonal antibody Y-11 (from Santa Cruz) or 0.2 g of anti-MDM2 (SMP-14). To detect sumoylated PML species, cells were rinsed with phosphate-buffered saline and scraped into 500 l of radioimmunoprecipitation assay buffer (20 mM Tris, pH 7.5, 2 mM EDTA, 150 mM NaCl, 0.25% SDS, 1% Nonidet P-40, 1% deoxycholic acid), and then sonicated for 10 pulses (setting 5, 50% output) using a Branson-450 sonifier. Lysates were centrifuged for 15 min prior to analysis to remove cellular debris.
Immunofluorescence Staining-For immunofluorescence staining, cells were plated on glass coverslips and transfected, washed, and refed as described above. 24 h after transfection, cells were rinsed with phosphate-buffered saline plus 0.1 mM CaCl 2 and 1 mM MgCl 2 and fixed with 4% paraformaldehyde for 30 min at 4°C. Paraformaldehyde was then replaced with 50 mM NH 4 Cl for 5 min, and cells were permeabilized with 0.1% Triton X-100 plus 0.2% bovine serum albumin. PML staining was carried out using the anti-HA monoclonal antibody HA.11 (Babco) as the primary antibody and rhodamine red-conjugated antimouse antibody (Jackson Labs) as the secondary antibody. MDM2 staining was carried out using the anti-MDM2 polyclonal antibody N-20 (Santa Cruz) as the primary antibody, and either 7-amino-4-methylcoumarin-3-acetic acid-conjugated anti-rabbit antibody, or fluorescein isothiocyanate-conjugated anti-rabbit antibody (Jackson Labs) as the secondary antibody. Specimens were then examined under a fluorescent microscope.
GST Fusion Protein Production-GST-tagged PML wild-type DNA was generated by PCR using HA-PML wild-type as a template. The 3Ј-primer for PCR was 5Ј-GGCGCGGCCGCCTCACCAGGAGAAC-CCACTTTCATTG-3Ј, and the 5Ј-primer was 5Ј-CCGGGATCCGAGC-CTGCACCCGCCCGATCTCCG-3Ј. The resulting PCR product was digested with BamHI and NotI restriction enzymes and cloned into the corresponding sites of pGEX-4T-3. GST-tagged PML K160R DNA was generated by PCR using the same primers but HA-PML K160R as a template. DNAs encoding GST alone or GST-PML were used to transform BL-21 bacterial cells, and transformed cells were grown at 37°C until reaching log phase. GST protein expression was induced by incubation in 0.2 mM isopropyl-1-thio-␤-galactopyranoside for 3 h. To purify the GST proteins, cells were lysed by sonication in lysis buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.1% Triton X-100, 150 mM NaCl), and the resulting lysate was incubated for 12 h at 4°C with glutathione-Sepharose beads. The beads were pelleted by centrifugation and washed with lysis buffer. For MDM2 binding, the GST-tagged proteins bound to beads were incubated with MDM2 protein in 300 l of lysis buffer for 5-6 h. Unbound MDM2 protein was removed by five washes (1 ml each) with lysis buffer. Bound proteins were eluted by boiling for 10 min in 1 ϫ loading buffer, resolved by SDS-PAGE, and examined by immunoblot analysis with the anti-MDM2 antibody SMP-14.
Expression and Isolation of MDM2 from Insect Cells-DNA encoding FLAG-tagged MDM2 that was either wild-type, lacked amino acids 217-246 (MDM2 ⌬p300BD), or lacked the C-terminal RING domain between residues 443 and 491 (MDM2 ⌬CT) was cloned into a baculovirus expression system and used for expression in insect cells. MDM2 protein was isolated by passing insect cell lysate over a column of Sepharose beads conjugated to an anti-FLAG antibody. Bound proteins were washed several times and eluted with a peptide encoding the FLAG epitope.

RESULTS
PML protein levels increase upon exposure of cells to interferons (IFNs) through direct transcriptional activation of the PML gene. To test for an in vivo interaction between PML and MDM2, H1299 cells (p53-null) were exposed to IFN-␥ for 48 h to induce endogenous PML expression, and the cells were then examined by immunofluorescence staining and confocal microscopy with antibodies against MDM2 and PML. These studies revealed colocalization between the endogenous PML and MDM2 proteins in IFN-␥Ϫtreated H1299 cells (Fig. 1A). Because H1299 cells lack p53 expression, these results suggest that MDM2 and PML can form an endogenous complex, either directly or indirectly, in the absence of p53. We also attempted to coimmunoprecipitate PML and MDM2 from IFN-␥-treated H1299 cells (not shown). However, we have been unable to immunoprecipitate large amounts of endogenous PML using our normal cell lysis conditions because the PML protein is insoluble under these conditions. We can solubilize PML under relatively harsh lysis conditions (high detergent), but these conditions are denaturing and therefore destroy any putative PML-MDM2 interactions. To examine the PML-MDM2 interaction further, Saos-2 cells (p53-null) were transfected with DNAs encoding MDM2 and epitope-tagged (HA-tagged) PML. Cell lysates were then immunoprecipitated with an anti-HA antibody and examined by immunoblotting with an antibody against MDM2. As shown in Fig. 1B, MDM2 coimmunoprecipitated with HA-PML in Saos-2 cells, again demonstrating that MDM2 and PML can interact in the absence of p53. To test the possibility that MDM2 and PML may interact directly, wildtype PML protein fused to GST was generated in bacteria, purified on glutathione-agarose beads, and mixed with partially purified, recombinant MDM2 protein produced in insect cells. Association between PML and MDM2 was then assessed in GST pull-down assays. As shown in Fig. 1C, wild-type (wt) MDM2 formed a complex with GST-PML wt, whereas no complex was observed between MDM2 and the GST protein alone. These results indicate that MDM2 and PML can interact with each other and suggest that this may be a direct interaction. An MDM2 protein lacking the C-terminal amino acids 443-491 (MDM2 ⌬CT) also formed a complex with GST-PML wt but not with the GST protein control, indicating that this in vitro binding between PML and MDM2 does not require the MDM2 C terminus. The arrow indicates MDM2 protein that was coimmunoprecipitated with HA-PML (upper panels). Steady-state levels of HA-PML and MDM2 were monitored by immunobloting of lysates without prior immunoprecipitation (lower panels). C, GST-PML wt or GST alone was generated in bacteria and tested for binding to partially purified MDM2 that was either wild-type or lacked either the C terminus (MDM2 ⌬CT). Binding complexes were resolved by SDS-PAGE and examined by immunoblotting with an antibody against MDM2. The Coomassie-stained gel indicates the relative amount of each GST protein used in the binding reaction.
We next wished to identify the regions of PML and MDM2 required for their interaction. PML contains several well characterized protein domains, including a RING-finger, two B boxes, and an ␣-helical coiled-coil ( Fig. 2A) (for review, see Ref. 12). To map the regions of PML required for MDM2 binding, N-terminal deletion mutants of HA-PML were expressed with MDM2 in transiently transfected cells. Cell lysates were then immunoprecipitated with an anti-HA antibody and examined by immunoblotting with an antibody against MDM2. As shown in Fig. 2B, MDM2 coimmunoprecipitated with HA-tagged wildtype PML, consistent with the results of Fig. 1. Interestingly, deletion mutants of PML lacking 100 or 200 amino acids from the N terminus displayed a much stronger interaction with MDM2 compared with wild-type PML, whereas a PML deletion mutant lacking 400 amino acids from the N terminus was unable to bind MDM2. Similar binding results were obtained in the reciprocal coimmunoprecipitation, which involved first immunoprecipitating with an anti-MDM2 antibody, followed by immunoblotting with an anti-HA antibody (Fig. 2C). In these studies, wild-type PML again displayed relatively weak binding to MDM2, whereas mutants lacking 100, 200, or 300 amino acids from the PML N terminus displayed stronger association with MDM2, and a mutant lacking 400 amino acids from the N terminus failed to bind MDM2. These results indicate that the C-terminal region of PML between residues 300 and 633 can form a complex with MDM2 and suggest that sequences within the N terminus of PML may be inhibitory to this PML-MDM2 complex formation.
Wild-type PML is post-translationally modified by the covalent attachment of SUMO, a small ubiquitin-like molecule, at three separate lysine residues in the PML sequence (Lys-65, Lys-160, and Lys-490; Fig. 2A) (13,15,16). Sumoylation is mediated by the sumo-conjugating enzyme Ubc9 and requires interaction between Ubc9 and the RING-finger domain within the first 100 amino acids of PML (16). The finding that PML mutants lacking the first 100 amino acids displayed an increased association with MDM2 suggested that sumoylation of PML may be inhibitory to PML-MDM2 binding. To investigate this possibility, PML mutants were generated in which each sumo site lysine was converted to arginine either singly or in combination, and association between these mutants and MDM2 was monitored by coimmunoprecipitation. As shown in Fig. 3A, MDM2 coimmunoprecipitated with wild-type PML to a relatively small extent in cells transiently expressing both proteins. MDM2 also coimmunoprecipitated with PML K65R and PML K490R to a comparable extent. In contrast, the PML K160R and PML 3KR mutants displayed a much stronger association with MDM2 compared with either wild-type PML or the other sumo site mutants. These findings suggested that PML sumoylation at lysine 160 may be an important regulator of MDM2 binding. In light of these results, we wished to confirm the sumoylation status of PML that coimmunoprecipitates with MDM2. Sumoylated PML is largely insoluble under gentle conditions of cell lysis. Therefore, to observe sumoylated PML, transfected cells were lysed in buffer containing a relatively high detergent concentration (radioimmunoprecipitation  (B1, B2), the predicted ␣-helical coiled-coil, and the three sites of sumoylation. B, U2OS cells were transfected with DNAs encoding wildtype MDM2 and either wild-type PML or the indicated deletion mutant of PML. PML protein was precipitated with an anti-HA antibody, and the immunoprecipitates (IP) were examined with an antibody against MDM2 (top panel). Steadystate levels of HA-PML and MDM2 were monitored by immunobloting (IB) of lysates without prior immunoprecipitation (middle and lower panels). The molecular mass (mw) in kDa of protein standard is indicated next to each blot. C, U2OS cells were transfected with DNAs encoding either wild-type MDM2 or an MDM2 mutant that localizes in the cytoplasm (MDM2 NLS-), and the indicated form of HA-PML. MDM2 protein was precipitated with an anti-MDM2 antibody, and the immunoprecipitates were examined with an antibody against HA (top panel). Steady-state levels of HA-PML and MDM2 were monitored by immunobloting of lysates without prior immunoprecipitation (middle and lower panels). buffer) and subsequently examined by immunoblotting with an anti-HA antibody. These studies revealed a series of high molecular mass PML species that migrated above wild-type HA-PML and each of the single sumo site mutants (Fig. 3B). Comparing the pattern of these species allowed us to identify each as PML that is sumoylated at Lys-65, Lys-160, or Lys-490. In contrast, these high molecular mass bands were not detected with expression of the PML 3KR mutant. To characterize the sumoylation state of PML bound to MDM2, wild-type PML and each of the sumo site mutants were expressed either alone or with MDM2. Transfected cells were then immunoprecipitated with either an anti-HA or anti-MDM2 antibody and examined by immunoblotting with an anti-HA antibody (Fig. 3C). In these studies the K160R and 3KR PML mutants again showed a much stronger association with MDM2 than the other forms of PML, and only the nonsumoylated PML species could be coimmunoprecipitated with MDM2. Taken together, these results indicate that nonsumoylated PML is the primary form of PML that coimmunoprecipitates with MDM2 and suggest that sumoylation at lysine 160 may be inhibitory to PML-MDM2 binding.
The strong interaction between HA-PML K160R and MDM2 allowed us to use this mutant in mapping other PML regions that may potentially interact with MDM2. To this end, Cterminal deletion mutants of HA-PML K160R were tested for MDM2 binding by coimmunoprecipitation. Because the nuclear localization signal (NLS) of PML is located in the C terminus between residues 487 and 493, C-terminal deletion mutants that lack the NLS may localize in the cytoplasm rather than the nucleus. Therefore, a cytoplasmic, NLS mutant of MDM2 was included in these studies to ensure colocalization of the transfected proteins. As shown in Fig. 4, full-length HA-PML (K160R) formed a strong complex with MDM2 in cotransfected cells. Similarly, C-terminal deletion mutants of HA-PML (K160R) also formed a strong complex with either wild-type MDM2 or the MDM2 NLS mutant, including a deletion mutant encoding only residues 1-200. In these studies, the C-terminal deletion mutants of HA-PML (K160R) displayed varying degrees of binding to either wild-type MDM2 or the MDM2 NLS mutant. Forms of PML (K160R) that are 1-633 and 1-500 contain the NLS and localize to the nucleus of transfected cells (not shown). These forms of HA-PML (K160R) bind most strongly to the nuclear localized wild-type MDM2 protein. In contrast, HA-PML (K160R) 1-400 lacks the NLS and localizes to the cytoplasm, unable to enter the nucleus without being actively imported. Not surprisingly, HA-PML (K160R) 1-400 associates strongly with the cytoplasmic MDM2 NLS mutant and only weakly with wild-type MDM2. The smaller HA-PML (K160R) mutants (1-300 and 1-200) also lack the NLS but are small enough to diffuse freely in and out of the nucleus, at least to some extent. These forms of HA-PML (K160R) display a diffuse localization pattern throughout the cell (not shown) and can bind both nuclear wild-type MDM2 and the cytoplasmic MDM2 NLS mutant. These results demonstrate that residues 1-200 in PML-K160R encode a separate MDM2 binding region. It should be noted that PML residues 1-200 (wild-type) displayed little to no binding to MDM2 when tested in a coimmunoprecipitation study similar to that shown in Fig. 4 (data not shown), consistent with the hypothesis that the K160R mutation uncovers an MDM2 binding site in this region of PML which would otherwise not be evident.
Our results indicate the presence of multiple MDM2 binding domains in PML, one located in the C terminus between residues 300 and 633, and a second that is revealed by the K160R mutation and is located between residues 1 and 200. We wished to define the regions of MDM2 required for interaction with either of these two regions of PML. To this end, MDM2 deletion mutants lacking various portions of the N terminus, C terminus, or central region were tested for their ability to coimmunoprecipitate with a form of PML which contained both binding domains (PML 1-633 (K160R)), or with PML mutants that expressed either MDM2 binding region alone. The results of these studies are summarized in Fig. 5. PML 1-633 (K160R), which contains both MDM2 binding regions, formed a strong complex with all of the MDM2 deletion mutants tested. This suggests the presence of multiple PML binding regions in MDM2, the deletion of any one of which may have little effect on PML binding. PML ⌬200N also formed a complex with all of the tested MDM2 mutants, although lesser binding was observed between this form of PML and an MDM2 mutant that lacked the acidic domain between residues 222 and 272. This suggests that MDM2 residues between 222 and 272 may be required for strong interaction between MDM2 and the Cterminal half of PML. PML 1-200 (K160R) formed a complex with each MDM2 mutant except one that lacked the C terminus (MDM2 6 -339), suggesting that MDM2 residues between 340 and 491 are required for interaction with HA-PML 1-200 (K160R). To map the PML interacting regions further, MDM2 mutants that expressed only the N terminus, C terminus, or central regions were tested in coimmunoprecipitation studies for binding to either PML 1-200 (K160R) or PML ⌬200N (Fig.  5). In these studies, PML 1-200 (K160R) formed a strong complex with the C terminus of MDM2 (residues 300 -488), but not with the MDM2 N terminus (residues 2-202) or central region (residues 100 -304), demonstrating that the MDM2 C terminus can form a specific complex with residues 1-200 in PML-K160R. In contrast, PML ⌬200N could be coimmunoprecipitated with MDM2 mutants encoding the central region between residues 100 and 304, but not with MDM2 mutants that encoded only the N (residues 2-202) or C (residues 300 -488) terminus. These results are consistent with the observation that PML ⌬200N displayed weak binding to an MDM2 mutant that lacked the central acidic domain (residues 222-272). In summary, these studies indicate that the C terminus of MDM2 can form a complex with PML residues 1-200 in the K160R mutant, whereas the MDM2 central region, which includes the acidic domain between residues 222 and 272, can interact with the C-terminal half of PML (Fig. 5).
Having identified multiple interactions between PML and MDM2, we next wished to determine the consequence of these interactions on the localization and/or activity of PML. The most striking feature of PML is its localization into multiprotein, nuclear foci structures, PML-NBs (33). Although the functions of PML are not fully understood, their proper localization and ability to recruit and interact with other NB components are considered essential for PML to act as a tumor suppressor (33,34). To assess the effect of MDM2 on PML localization, cells were transfected with DNAs encoding HA-tagged PML either alone or cotransfected with various forms of MDM2. Localization of the transfected PML and MDM2 proteins was then monitored by immunofluorescence staining. As illustrated in Fig. 6A, HA-PML (wild-type) and MDM2 localized almost exclusively to the nucleus when expressed alone. In contrast, coexpression with wild-type MDM2 caused a marked redistribution of PML to the cytoplasm, whereas the MDM2 protein remained nuclear (Fig. 6B). These results indicated that nuclear MDM2 could trigger a redistribution of PML from the nucleus to the cytoplasm. We scored the extent to which PML was localized to the cytoplasm when expressed alone or with MDM2, and the results are graphed in Fig. 7. We next wished to investigate the region of MDM2 which may be required to promote this nuclear exclusion of PML. As shown in Figs. 6 and 7, MDM2 deletion mutants lacking either the central acidic domain (⌬222-272) or the N-terminal p53 binding domain (⌬120N) could efficiently promote the nuclear exclusion of PML, indicating that neither the central domain nor p53 binding domain is required for this effect. Similar results were obtained in Saos-2 cells (p53-null; data not shown), further demonstrating that MDM2 can promote PML nuclear exclusion in the absence of p53. An MDM2 mutant lacking the C terminus (MDM2 6 -339) failed to promote PML nuclear exclusion (Figs. 6 and 7), suggesting that PML binding to the MDM2 C terminus may be required for this effect. Consistent with this hypothesis, an N-terminal deletion mutant of PML (HA-PML (⌬200N)) was completely resistant to MDM2-mediated nuclear exclusion. Given that the MDM2 C terminus can interact with the PML N terminus, these results suggest that binding between the N terminus of PML and the C terminus of MDM2 is required to promote the nuclear exclusion of PML.
One NB-associated protein with which PML interacts is the transcriptional coactivator CBP. PML binding has been reported to stimulate the ability of a GAL4-CBP fusion protein to activate transcription from a gene promoter containing GAL4 DNA binding sites (35). We wished to test whether MDM2 could block the ability of PML to activate CBP. As shown in Fig.  8, HA-PML activated GAL4-CBP-dependent transcription, consistent with previous studies (35). Coexpression of wild-type MDM2 decreased the ability of PML to activate GAL4-CBP, although having relatively little effect on the activity of GAL4-CBP alone. These results indicate that wild-type MDM2 can inhibit the ability of PML to activate the GAL4-CBP fusion protein. To address the region of MDM2 required for this effect, we tested various MDM2 deletion mutants for their ability to inhibit PML in this assay. As shown in Fig. 8, the C-terminal FIG. 5. Summary of PML-MDM2 binding reactions. The indicated forms of MDM2 and PML were coexpressed in transfected cells and examined for binding by coimmunoprecipitation. ϩ indicates that binding was observed, Ϫ indicates that no binding was observed, and ϩ/Ϫ indicates that weak binding was observed. nd indicates a combination of mutant MDM2 and PML that was not tested for binding (not done). The schematic on the right illustrates the proposed interactions between MDM2 and PML. The solid bars indicate the regions of each protein which are proposed to interact. PML amino acids 1-200 in the K160R mutant can interact with the C terminus of MDM2, probably between residues 340 and 488. PML amino acids 300 -633 contain a second MDM2 binding region (see Fig. 2). HA-PML ⌬200N could coimmunoprecipitate with the central region of MDM2 (amino acids 100 -304) but not with the MDM2 N terminus (amino acids 2-202), suggesting that MDM2 residues between 202 and 304 are required. The MDM2 acidic domain (residues 222-272) also appears to be required for strong interaction with the C-terminal region of PML because HA-PML ⌬200N displayed weaker interaction with an MDM2 mutant that lacks this region (MDM2 ⌬222-272). deletion mutant MDM2 (6 -339) efficiently inhibited the ability of PML to activate the GAL4-CBP fusion protein. These results indicate that binding between the MDM2 C terminus and PML is not required to inhibit PML in this assay. Given that MDM2 (6 -339) is unable to promote PML nuclear exclusion, these results also indicate that nuclear exclusion is not required for the MDM2-mediated inhibition of PML. An MDM2 mutant that lacks the N terminus (MDM2 ⌬120N) was also able to inhibit the ability of PML to activate the GAL4-CBP fusion protein. In contrast, an MDM2 deletion mutant lacking residues 222-272 had little to no effect on the ability of PML to activate the GAL4-CBP fusion protein. These results indicate that the ability of MDM2 to inhibit PML in this assay requires the central, acidic region of MDM2 and therefore likely involves binding between the central region of MDM2 and PML. DISCUSSION The tumor suppressor protein PML has received considerable attention recently due, at least in part, to its involvement in apoptosis signaling and its direct interaction with p53. PML binding recruits p53 to PML-NBs, multiprotein complexes that localize to discreet foci within the nucleus. The effect of p53-PML binding is to increase p53 activity as a transcription factor (22). Current models suggest that recruitment to PML-NBs brings p53 in close proximity with CBP/p300 (36). Acetylation of p53 by CBP/p300 then increases p53 DNA binding affinity, leading to an activation of p53-responsive genes. This activation of p53 most likely contributes to the tumor suppressor function of PML. In addition to p53, PML can interact with a variety of other proteins and colocalize with them in PML-NBs. Included among these are proteins involved in recombination (BLM, RAD51, RAD52), transcriptional activation and repression (CBP, Daxx, Sp100), protein degradation (HAUSP, PA28), and the regulation of cell growth (pRB) (for review, see Refs. 37 and 38). However, in most cases the physiologic relevance of these interactions and their effect on cell proliferation have not been clarified fully.
The current report identifies MDM2 as a novel PML-interacting protein. Confocal microscopy revealed colocalization between the endogenous PML and MDM2 proteins in IFN-␥Ϫtreated H1299 cells (p53-null), suggesting that they can complex one another in cells. Further, MDM2 immunoprecipitated with PML in cells transiently expressing both proteins, and recombinant MDM2 produced in insect cells formed a strong complex with a bacterially expressed GST-PML fusion protein. These results indicate that MDM2 and PML can interact with each other in a manner independent of p53 and suggest that this may be a direct interaction. Mapping studies revealed at least two separate regions of MDM2 and PML involved in their interaction. First, the central region of MDM2, which includes the MDM2 acidic domain, could form a complex with the C-terminal portion (residues 200 -633) of PML. The C terminus of PML is of particular interest when considering interactions between PML and other proteins. At least seven different PML isoforms have been described, all of which vary in their C termini and are generated by alternative splicing (12). It will be of interest in the future to determine whether interaction with the central region of MDM2 is an activity shared by all PML isoforms or occurs in an isoformspecific manner. The C terminus of MDM2 (residues 300 -488) could also form a complex with residues 1-200 in the PML K160R mutant. The N terminus of PML and the C terminus of MDM2 both contain a RING-finger domain within their sequences, and we considered that the interaction between these two regions might be mediated through their respective RINGfingers. However, preliminary results from our laboratory suggest that the PML RING-finger is not required for residues 1-200 to interact with MDM2 (not shown), suggesting that this interaction with MDM2 is not through their corresponding RINGs. Interestingly, the extent to which MDM2 could bind PML in coimmunoprecipitation experiments was increased markedly when sumoylation of PML at lysine 160 was inhibited, either through deletion of PML protein regions required for Lys-160 sumoylation or through mutation of the Lys-160 site. This finding raises the possibility that PML sumoylation at Lys-160 may in some way be inhibitory to PML-MDM2  binding. Lys-160 is located within the N-terminal region of PML which can interact with the MDM2 C terminus. One possibility, therefore, is that sumoylation at this site blocks access of this region of PML to MDM2. However, it should be noted that sumoylation at Lys-160 is required for the proper localization of PML into PML-NBs (39,40). A second possibility, therefore, is that sumoylation at Lys-160 may indirectly affect PML-MDM2 coimmunoprecipitation by directing PML into PML-NBs.
PML is the major component of multiprotein, nuclear foci structures, PML-NBs (33,34). PML appears to be absolutely required for NB formation, as other NB-associated proteins fail to localize in NBs when PML is not present (39,40). However, the biochemical and molecular functions of PML-NBs are still unknown. Most current models propose that PML-NBs are nuclear scaffolds that serve as either temporary storage sites for catalytically active proteins or as catalytic surfaces where specific biochemistries can occur (34). In either case, proper localization of PML, as well as its ability bind and recruit other NB-associated proteins, is considered essential for PMLs normal function. We observe that transient overexpression of MDM2 disrupts PML-NBs and promotes a redistribution of PML to the cytoplasm. Given the importance of proper localization in PML function, it is possible that MDM2 may inhibit one or more activities associated with PML through this disruption of its normal localization. The C terminus of MDM2 (residues 340 -488) is necessary to alter PML localization, and the N terminus of PML is also required for PML to be redistributed to the cytoplasm by MDM2 (Figs. 6 and 7). Given the interaction between the PML N terminus and MDM2 C terminus, these results suggest that binding between these two regions is responsible for the cytoplasmic redistribution of PML. It is interesting to note that the disruption of PML-NBs and cytoplasmic redistribution of PML observed here with MDM2 overexpression is strikingly similar to the effect of certain viral proteins on PML. During a variety of viral infections, including those of the herpes, adeno, and retro families, PML-NBs become disrupted, and, in some cases, PML is relocalized to the cytoplasm (41)(42)(43)(44)(45). This disruption of PML-NBs has been linked in some cases to host cell shut-off, inhibition of apoptosis, and establishment of a chronic viral infection (45). In contrast, PML and other proteins associated with PML-NBs are induced after exposure to antiviral IFNs, and the number of PML-NBs increases (46). Based on these findings it has been proposed that PML-NBs may represent preferential targets for viral infections and that PML could play a role in the mechanism of the antiviral action of IFNs. Viral proteins that disrupt PML-NBs and/or cause a relocalization of PML to the cytoplasm include the herpes simplex virus type-1 ICP0 protein, the adenovirus E4-ORF-3 protein, the human cytomegalovirus IE1 protein, the lymphocytic choriomeningitis virus Z protein, and the rabies virus P protein (45; for review, see Ref. 46). Like MDM2, the herpes simplex virus type-1 ICP0 and lymphocytic choriomeningitis virus Z proteins both contain a RING-finger domain. Further, lymphocytic choriomeningitis virus Z and rabies P protein cause a cytoplasmic relocalization of PML, and herpes simplex virus type-1 ICP0 promotes the proteasome-dependent degradation of PML. It remains to be seen whether the relocalization of PML observed here with MDM2 expression occurs through a similar or distinct mechanism as in cells expressing these viral proteins.
CBP is a transcriptional activator with histone acetyltransferase activity and is a component of PML-NBs (47). PML forms a direct complex with CBP, and this complex formation stimulates the ability of a GAL4-CBP fusion protein to activate gene transcription (35). In the current report, wild-type MDM2 inhibited the ability of PML to activate GAL4-CBP-mediated transcription. This inhibition required the central, acidic region of MDM2, suggesting that binding between the central region of MDM2 and PML is required to inhibit PML in this assay. An MDM2 mutant that lacked the C terminus, and was therefore unable to redistribute PML to the cytoplasm, also inhibited the ability of PML to activate CBP. These results indicate that the MDM2-mediated nuclear exclusion of PML is not required for MDM2 to inhibit PML, at least in this assay system. Functional interactions among PML, MDM2, and CBP are of particular interest with regard to the PML-mediated activation of p53. Current models suggest that PML corecruits p53 and CBP to PML-NBs, bringing them into close proximity (for review, see Ref. 36). CBP can then acetylate p53, resulting in an increased p53 DNA binding activity, thus leading to an activation of p53-responsive genes. This activation of p53 most likely contributes to the tumor suppressor function of PML. Conceivably, MDM2 could inhibit this activation of p53 by either disrupting PML-NBs or blocking the ability of PML to activate CBP. Such a mechanism could contribute to the MDM2-mediated inhibition of p53.