Hetero-oligomerization with MdmX Rescues the Ubiquitin/Nedd8 Ligase Activity of RING Finger Mutants of Mdm2*

Mdm2 is a member of the RING finger family of ubiquitin ligases and is best known for its role in targeting the tumor suppressor p53 for ubiquitination and degradation. Mdm2 can bind to itself and to the structurally related protein MdmX, and these interactions involve the RING finger domain of Mdm2 and MdmX, respectively. In this study, we performed a mutational analysis of the RING finger domain of Mdm2, and we identified several amino acid residues that are important for Mdm2 to exert its ubiquitin ligase function. Mutation of some of these residues interfered with the Mdm2-Mdm2 interaction indicating that a homomeric complex represents the active form of Mdm2. Mutation of other residues did not detectably affect the ability of Mdm2 to interact with itself but reduced the ability of Mdm2 to interact with UbcH5. Remarkably, MdmX efficiently rescued the ubiquitin ligase activity of the latter Mdm2 mutants in vitro and within cells. Because the interaction of Mdm2 with MdmX is more stable than the Mdm2-Mdm2 interaction, this suggests that Mdm2-MdmX complexes play a prominent role in p53 ubiquitination in vivo. Furthermore, we show that, similar to Mdm2, the Mdm2-MdmX complex has Nedd8 ligase activity and that all mutations that affect the ubiquitin ligase activity of Mdm2 also affect its Nedd8 ligase activity. From a mechanistic perspective, this suggests that the actual function of Mdm2 and Mdm2-MdmX, respectively, in p53 ubiquitination and in p53 neddylation is similar for both processes.

Covalent modification of proteins by ubiquitin ("ubiquitination") is a selective process and plays an important role in the control of many fundamental cellular processes (1)(2)(3). Specific recognition of substrate proteins of the ubiquitin-conjugation system is mainly mediated by the action of ubiquitin-protein ligases (E3). 2 Based on the presence of distinct amino acid sequence motifs, proteins with E3 activity can be roughly grouped into three classes as follows: HECT domain E3s, RING finger E3s, and U-box E3s. Structural studies indicate that RING finger domains and U-box domains adopt similar structural folds providing an interface for the interaction with their cognate E2 ubiquitin-conjugating enzymes (4 -9). Because specific recognition of substrate proteins is mediated by regions other than the RING finger or the U-box, it is commonly assumed that RING finger/U-box E3s function as adaptor proteins bringing E2s and substrate proteins into proper orientation for E2-mediated ubiquitination. HECT domain E3s have a similar modular structure (i.e. the HECT domain is responsible for the interaction with cognate E2s; substrate recognition is mediated by regions other than the HECT domain), but in contrast to RING finger or U-box E3s, HECT domain E3s appear to play a direct catalytic role in the covalent attachment of ubiquitin to substrate proteins (10 -12).
At least some RING finger/U-box E3s have the ability to form homo-oligomeric complexes and/or hetero-oligomeric complexes with other RING finger/U-box E3s. Homo-oligomerization and/or hetero-oligomerization affect the activity of the respective E3s and involve physical contacts between the respective RING finger or U-box domains. Examples for hetero-oligomer formation between RING finger/U-box E3s include the BRCA1-BARD1 complex, the Mdm2-MdmX complex, and the Bmi1-Ring1b complex (13)(14)(15)(16). Thus, RING finger and U-box domains cannot only serve as interfaces for interaction with E2s but also for interaction with other E3s.
Mdm2 and MdmX are structurally related proteins insofar as both contain an N-terminal p53-binding domain, a central zinc-binding motif, and a C-terminal RING finger domain (15)(16)(17). Genetic experiments in mice have shown that both Mdm2 and MdmX play critical roles in the control of the growthsuppressive properties of the tumor suppressor p53 (18 -21). Mdm2-and MdmX-deficient mice, respectively, die early in embryogenesis, although at different stages, and the respective lethal phenotype is rescued by concomitant loss of p53 expression (18,19). In addition, data obtained with conditional knockout mice indicate that Mdm2 and MdmX act synergistically in p53 regulation but that MdmX has also Mdm2-independent functions in p53 regulation (22)(23)(24).
It is well established that Mdm2 acts as an E3 for p53, whereas MdmX does not appear to have appreciable E3 activity (25)(26)(27)(28)(29). As indicated above, Mdm2 and MdmX can form heteromeric complexes (15,16). However, the functional consequences of the interaction of Mdm2 with MdmX with respect to the E3 activity of Mdm2 are still controversially discussed. Results obtained in overexpression studies suggest that binding of MdmX interferes with the E3 activity of Mdm2 (15,30,31). In contrast, in vitro ubiquitination experiments indicate that binding of MdmX stimulates the E3 activity of Mdm2 (28,29). Furthermore, it was recently shown that, in addition to its function in p53 ubiquitination, Mdm2 facilitates the modification of p53 with the ubiquitin-like protein Nedd8 (32). At present it is not known whether MdmX affects the Nedd8 ligase activity of Mdm2.
To further characterize the ubiquitin and Nedd8 ligase activity of Mdm2 and the effect of MdmX on these Mdm2 activities, we performed a mutational analysis of the RING finger domain of Mdm2. The results obtained indicate that the ability of Mdm2 to form homomeric complexes is required for both its ubiquitin ligase and its Nedd8 ligase activity. Furthermore, we show that both the ubiquitin ligase and the Nedd8 ligase activity of some of the Mdm2 mutants are efficiently rescued by MdmX in vitro and within cells. This supports the notion that the Mdm2-MdmX complex has intrinsic E3 activity and is actively involved in ubiquitination and neddylation of p53 in vivo.

EXPERIMENTAL PROCEDURES
Cell Lines and Plasmids-H1299 cells and HEK293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum. Bacterial expression constructs for glutathione S-transferase (GST) fusion proteins of wild-type Hdm2 and HdmX were described previously (29). Bacterial expression constructs for GST fusion proteins of Mdm2, MdmX, and various RING finger mutants of Mdm2 and Hdm2, respectively (see Fig. 1), were generated by PCR-based approaches (further details will be provided upon request). The expression constructs used in transient transfection experiments ( Fig. 4) encoding wild-type Mdm2 (pCocMdm2), wildtype p53, His-tagged ubiquitin, and HA-tagged HdmX were described previously (29,33,34). For transient expression of Mdm2 mutants (see Fig. 4), the respective mutations were introduced by site-directed mutagenesis PCR into the wildtype Mdm2 cDNA within the frame of pCocMdm2. His-tagged Nedd8 was expressed from the eukaryotic expression vector pSG5.0 (Stratagene). For yeast two-hybrid assays, the respective cDNAs (Mdm2, Mdm2 mutants, HdmX, UbcH5b; see Table 1) were cloned into the expression vectors pGBKT7 and pGADT7 (Clontech).
Transfection and Antibodies-In transient expression experiments, cells were transfected with the respective constructs in the presence of a reporter construct encoding ␤-galactosidase by lipofection (Lipofectamine 2000; Invitrogen) according to the manufacturer's instructions. Protein extracts were prepared 24 h after transfection (see below), and transfection efficiency was determined by measuring ␤-galactosidase activity. Levels of p53 or ubiquitinated p53, HA-tagged HdmX, and Mdm2, respectively, were determined by Western blot analysis by using transfection efficiency adjusted protein amounts. The antibodies used for detection were the mouse monoclonal antibody HA.11 (Hiss Diagnostics, Freiburg, Germany) for HA-tagged HdmX, the mouse monoclonal antibody DO1 (Calbiochem) for p53, and the mouse monoclonal antibody SMP14 (Santa Cruz Biotechnology) that detects both Mdm2 and Hdm2.
Ubiquitination, Neddylation, and Degradation Assays-For in vitro ubiquitination experiments, wild-type Mdm2, Hdm2, and HdmX and the various Mdm2 and Hdm2 mutants were expressed as GST fusion proteins in Escherichia coli DH5␣. The ubiquitin-activating enzyme E1 and the ubiquitin-conjugating enzyme UbcH5b were expressed in the baculovirus system or in E. coli BL21 by using the pET expression system as described (35). For in vitro ubiquitination, 1 l of rabbit reticulocyte lysate extract-translated 35  For ubiquitination of p53 within cells, one 6-cm plate of H1299 cells or HEK293T cells was transfected with expression constructs encoding p53 (200 ng), His-tagged ubiquitin (1 g), Mdm2 or the respective Mdm2 mutants (200 ng), and HdmX (600 ng). 24 h after transfection, 30% of the cells were lysed under nondenaturing conditions as described (36) to determine the transfection efficiency (see above). The remaining cells were lysed under denaturing conditions and ubiquitinated proteins purified as described (34). For neddylation of p53 within cells, the same procedure was applied by using an expression construct for His-Nedd8.
To monitor degradation of ectopically expressed p53 within cells, one 6-cm plate of H1299 cells was transfected with expression constructs encoding p53 (10 ng), Mdm2 or the respective Mdm2 mutants (200 ng), and HdmX (600 ng). Protein extracts were prepared 24 h after transfection as described (36), and p53 levels were determined by Western blot analysis.
Yeast Two-hybrid Assays-Yeast two-hybrid experiments were performed by using the MatchMaker system and the Saccharomyces cerevisiae strain KF1 (37) according to the manufacturer's instructions (Clontech). Briefly, 1 g of the respective plasmids (see Table 1) was transformed, and transformed cells were selected by growth in media deficient of leucine and tryptophan. After 3 days, cells were streaked on three different reporter plates (deficient of histidine, adenine, and uracil, respectively), and growth of the cells followed for up to 5 days. The relative strength of the individual protein/protein interactions is indicated by the ability of the respective cells to grow on the different reporter plates. Growth on uracil-deficient plates requires stronger interaction of the respective proteins than growth on adenine-deficient plates, whereas growth on histidine-deficient plates requires the least efficient interaction (37).

Mutational Analysis of the Ubiquitin Ligase Activity of
Mdm2-The RING finger domain of Mdm2 is located within the C-terminal 60 amino acid residues and is highly conserved between mouse Mdm2 and its human ortholog Hdm2 (Fig. 1). In the following, there will be no differentiation between Mdm2 and Hdm2, because similar results were obtained with both Mdm2 mutants and the respective Hdm2 mutants. Based on the crystal structure of the U-box E3 Prp19, a dimeric structure has been proposed for the Mdm2 RING finger domain with four hydrophobic residues (Ile-448, Leu-456, Ile-483, and Leu-485; Fig. 1) being critically involved in the formation of the dimer interface (38). To test whether these residues may indeed be required for Mdm2-Mdm2 interaction and possibly the E3 activity of Mdm2, cDNAs encoding single point mutants of Mdm2, in which the respective amino acid residues are substituted by glutamate, were generated by PCR-based mutagenesis. In addition, a cDNA encoding a truncated form of Mdm2 devoid of the seven C-terminal residues was generated. Because the efficiency of Mdm2-Mdm2 interaction is rather low under standard in vitro coprecipitation conditions (data not shown), the binding abilities of the various mutants were determined in the yeast two-hybrid system (16). This revealed that, with the exception of Mdm2L485E, the ability of the mutants to interact with themselves or with wild-type Mdm2 was significantly reduced (Table 1).
To test the E3 activity of the various Mdm2 mutants, the respective proteins were expressed as GST fusion proteins in E. coli. Upon purification, the Mdm2 mutants were tested in vitro for their ability to ubiquitinate in vitro translated 35 S-labeled p53, MdmX, and wild-type Mdm2, respectively (Figs. 1 and 2). Similar to the results obtained in the yeast two-hybrid system, with the exception of Mdm2L485E, the E3 activity of the Mdm2 mutants was significantly reduced, regardless of the ubiquitination substrate. Furthermore, the Mdm2 mutants were inactive in in vitro auto-ubiquitination assays (data not shown). Although these data do not prove that Ile-448, Leu-456, Ile-483, and the very C-terminal 7 residues are directly involved in the formation of Mdm2 homomers (note that it is referred to homomers rather than homodimers because the exact oligomeric state of full-length Mdm2 is not known), they indicate that these residues are critical for both Mdm2-Mdm2 interaction and the E3 activity of Mdm2. The RING finger domain of Mdm2 contains a putative WalkerA motif (amino acid residues 446 -452; Fig. 1) and has been shown to bind were incubated with in vitro translated and radiolabeled p53 under standard ubiquitination conditions for 2 h as indicated. The reaction products were analyzed by SDS-PAGE followed by fluorography. Running positions of the nonmodified form and of the ubiquitinated forms of p53 are indicated by an arrowhead and an asterisk, respectively. Note that the activity of wild-type Mdm2 (or wild-type Hdm2) and of the Mdm2 mutants G451S, T453C, and T453A varies slightly from protein preparation to preparation (see also Fig. 3), but in general the mutants are significantly less active than the respective wild-type proteins.

TABLE 1 Binding and E3 ligase properties of Mdm2 mutants
For binding properties, the ability of the various Mdm2 mutants to interact with Mdm2, HdmX, UbcH5b, or with themselves (AD-self) was determined in the yeast two-hybrid system as indicated. The various interactions were monitored by growth selection on adenine-deficient media (Ade), uracil-deficient media (Ura), or histidinedeficient media (His). Mdm2 and the various Mdm2 mutants were expressed at similar levels under the conditions used as determined by Western blot analysis (supplemental Fig. 3). Note that growth on Ura requires more efficient interaction of the respective fusion proteins than growth on Ade, whereas growth on His requires the least efficient interaction. AD, Gal4 transactivation domain; BD, Gal4 DNA binding domain; Ϫ, no growth; ϩ, growth detectable after 3-4 days of selection; ϩϩ, growth detectable after 2 days of selection. For E3 ligase activity, this is a summary of the results obtained in the experiments shown in Figs. 1-4. Since similar results were obtained for the ubiquitin ligase activity (in vitro and within cells) and the Nedd8 ligase activity (within cells), it is not distinguished between both activities. ϪMdmX, E3 ligase activity in the absence of MdmX; ϩMdmX, E3 ligase activity in the presence of MdmX; ϩϩ, activity similar to wild-type Mdm2 (Mdm2); ϩ, less active than wild-type Mdm2 but considerable E3 activity; ϩ/Ϫ, significantly decreased E3 activity (Յ10% of wild-type Mdm2 activity); Ϫ, no detectable E3 activity.

AD-Mdm2
AD-HdmX AD-self AD-Mdm2 AD-HdmX AD-self BD-UbcH5 E3 ligase activity ؊MdmX ؉MdmX Ade Ura His ATP (39). Although the ability of Mdm2 to bind ATP is not required for its E3 activity, the glycine residues of the WalkerA motif (Gly-446 and Gly-451) and the threonine residue (Thr-453) adjacent to the WalkerA motif were subjected to mutational analysis, because Thr-453 has been assumed to be involved in coordinating Zn 2ϩ , and thus, the region surrounding Thr-453 may be critical for structural integrity of the RING finger domain (40). The respective point mutants were tested for their ability to interact with themselves and with wild-type Mdm2, respectively, and for E3 activity as described above. Surprisingly, Thr-453 was not essential for the ability of Mdm2 to form homomers (  (15,16,28,29) (in the following there will be no differentiation between MdmX and HdmX, because similar results were obtained with both proteins). Thus, we next tested the ability of MdmX to interact with the Mdm2 mutants described above and to rescue the E3 activity of these Mdm2 mutants. As reported previously, the interaction of MdmX with wildtype Mdm2 was more efficient than the Mdm2-Mdm2 interaction (16) ( Table 1; note that growth on uracildeficient media requires a stronger interaction of the binding partners than growth on adenine-deficient media). Similarly, MdmX efficiently interacted with the WalkerA motif mutants, the Thr-453 mutant, and the L456E mutant. In contrast, MdmX did not efficiently interact with Mdm2I448E, Mdm2I483E, and Mdm2⌬C7 supporting the notion that these residues/regions of Mdm2 are critically involved in the formation of homo-and heteromeric complexes. Importantly, MdmX rescued the E3 ligase activity of those Mdm2 mutants that efficiently bound to MdmX in the yeast two-hybrid system (Fig. 3).
MdmX Is a Positive Effector of Both the Ubiquitin Ligase and the Nedd8 Ligase Activity of Mdm2 within Cells-The data presented above support the previously reported findings that MdmX functions as a positive cofactor for the E3 activity of Mdm2 and that heteromeric Mdm2-MdmX complexes can mediate ubiquitination of p53 in vitro (28,29). However, no direct evidence had been provided that Mdm2-MdmX complexes can mediate ubiquitination of p53 within cells (for discussion of this issue, see Ref. 29). Therefore, we tested the ability of MdmX to rescue the ability of the E3-defective Mdm2 mutants G451S, T453C, and T453A to mediate ubiquitination and degradation of p53 in cotransfection assays (Fig. 4, A and  B). Similar to previously published data (29), coexpression of  MdmX did not significantly affect wild-type Mdm2-mediated ubiquitination and degradation of p53 under the conditions used. In contrast, MdmX efficiently rescued the ubiquitination defect of the Mdm2 mutants tested (Fig. 4A). In the presence of MdmX, the Mdm2 mutants tested were also able to target p53 for degradation, although the reaction was somewhat less efficient than wild-type Mdm2-mediated degradation of p53 (Fig. 4B).
Mdm2 was recently reported to function as a Nedd8 ligase for p53 (32). Because we have not yet been successful in setting up an in vitro system for Mdm2-mediated neddylation of p53 that would be efficient enough to reliably determine the effect of MdmX on the Nedd8 ligase activity of Mdm2, we studied the effect of MdmX on Mdm2-mediated neddylation of p53 in cotransfection experiments (Fig. 4C). As for p53 ubiquitination, coexpression of MdmX did not significantly affect Mdm2-mediated neddylation of p53 under the conditions used. Furthermore, the ability of the Mdm2 mutants tested to neddylate p53 was significantly reduced. Remarkably, the neddylation defect of the Mdm2 mutants was efficiently rescued by coexpression of MdmX. Taken together, the data indicate that, within cells, Mdm2-MdmX complexes function as E3 ligase in both p53 ubiquitination and p53 neddylation.
Binding of Mdm2 Mutants to UbcH5b-Mdm2 is known to functionally interact with members of the UBC4/UBC5 subfamily of E2s, and it was recently reported that, within cells, UbcH5b and UbcH5c are the most relevant E2s for Mdm2mediated ubiquitination of p53 (41). Because Mdm2 did not detectably interact with UbcH5b or UbcH5c in in vitro coprecipitation experiments (data not shown), we tested the ability of wild-type Mdm2 and the various Mdm2 mutants to bind to UbcH5b in the yeast two-hybrid system. This showed that most of the Mdm2 mutants with a defect in homomer formation interacted with UbcH5b with an efficiency similar to wild-type Mdm2 (Table 1). This suggests that the defect in E3 activity is not explained by the inability of the Mdm2 mutants to interact with their cognate E2 and that oligomerization of Mdm2 is not required for the interaction of Mdm2 with UbcH5b. Furthermore, those Mdm2 mutants whose E3 activity was efficiently rescued by MdmX (i.e. G451S, T453C, and T453A) interacted somewhat less efficiently with UbcH5b than wild-type Mdm2 providing a possible explanation for their defect in E3 activity. Finally, it should be noted that we did not observe an interaction of Mdm2 with the Nedd8-conjugating enzyme Ubc12 in the yeast two-hybrid system (data not shown). However, the relevance of this finding with respect to the mechanism involved in Mdm2-mediated neddylation of p53 is presently unclear.

DISCUSSION
Using a panel of E3-defective Mdm2 mutants, we provide evidence that the Mdm2-MdmX complex functions as an E3 for p53 within cells. Because it was recently shown that, upon DNA damage, Mdm2 targets MdmX for degradation and inactivation (42)(43)(44)(45), we propose that MdmX is both a positive cofactor and a substrate of the E3 activity of Mdm2 and that the interaction of Mdm2 with MdmX serves at least two functions as follows. Under normal growth conditions (i.e. in the absence of stress stimuli inducing p53 activation), the Mdm2-MdmX complex controls the growth-suppressive properties of p53 by targeting it for ubiquitination and degradation. Upon DNA damage, MdmX serves as substrate for Mdm2. As a result of  (His-ub), and wild-type (wt) Mdm2 or the indicated Mdm2 mutants in the absence or presence of expression constructs for HA-tagged HdmX as indicated. Protein extracts were prepared 24 h after transfection, and 70% of the extract was used to isolate ubiquitinated proteins by Ni 2ϩ -affinity chromatography. Upon affinity purification, levels of ubiquitinated p53 were determined by Western blot analysis with the p53-specific antibody DO1. The remaining 30% of the respective cell extracts were used to determine the expression levels of HA-HdmX. Note that under the conditions used, it was not possible to determine the expression levels of wild-type Mdm2 and the Mdm2 mutants. To ensure that HdmX expression does not affect the expression of Mdm2 and the Mdm2 mutants, similar experiments were performed with HEK293T cells. This showed that HdmX rescues the ubiquitin ligase activity of the Mdm2 mutants without affecting their expression levels (supplemental Fig. 5). Asterisk denotes ubiquitinated forms of p53. B, to score for p53 degradation, cotransfection experiments were performed as described in A but in the absence of the His-tagged ubiquitin expression construct and with lower amounts of the p53 expression construct (see "Experimental Procedures"). Protein extracts were prepared 24 h after transfection and adjusted according to transfection efficiency, and levels of p53 were determined by Western blot analysis using the p53-specific antibody DO1. Loading of the individual lanes is as indicated for A. C, H1299 cells were cotransfected with expression constructs for p53, His-tagged Nedd8, and wild-type (wt) Mdm2 or the indicated Mdm2 mutants in the absence or presence of expression constructs for HAtagged HdmX as indicated. Protein extracts were prepared 24 h after transfection and neddylated proteins purified by Ni 2ϩ -affinity chromatography. Upon affinity purification, levels of neddylated p53 were determined by Western blot analysis with the p53-specific antibody DO1. Asterisk denotes neddylated forms of p53. Note that the results presented in A-C are representative for results obtained in at least four independent experiments. this substrate/enzyme interaction, both the Mdm2-dependent and Mdm2-independent functions of MdmX are inactivated thereby contributing to the activation of the growth-suppressive properties of p53.
Mdm2 forms homomers, and we previously proposed that homomerization is required for Mdm2 to exert its E3 activity (29,46). Based on the structure of the U-box E3 Prp19, several hydrophobic amino acid residues within the RING finger domain of Mdm2 (i.e. Ile-448 and Leu-456) or adjacent to it (i.e. residues within the seven C-terminal residues of Mdm2) were suggested to be involved in dimerization of the Mdm2 RING finger (38). Indeed, replacement of the respective amino acid residues by glutamate or deletion of the C-terminal seven residues resulted in Mdm2 mutants that do not, or only inefficiently, form homomeric complexes and have significantly reduced E3 activity (Figs. 1 and 2). Furthermore, the ability of these Mdm2 mutants to interact with MdmX is negatively affected ( Table 1) and, accordingly, their E3 activity is not or only partially rescued by MdmX (Fig. 3). Because the ability of the Mdm2 mutants to interact with UbcH5b is not significantly affected, these data strongly support the notion that Mdm2 homomers rather than monomers display E3 activity.
While this work was prepared for submission, the NMR structure of the RING finger domain of Hdm2 (amino acid residues 429 -491) was published (47). The data obtained show that the RING finger domain forms a stable homodimer in solution and confirm the importance of the C-terminal seven amino acid residues and Leu-456 (numbering according to mouse Mdm2) for the formation of the dimer interface. Ile-448 does not appear to be directly involved in dimer formation, but because Val-449 is involved, it seems likely that introduction of a negative charge at position 448 has a significant impact on dimer formation. Surprisingly, the zinc ion is coordinated by His-450 rather than by Thr-453 that was previously proposed to be one of the zinc-binding residues of the Mdm2 RING finger (40). However, Thr-453 is located at the dimer interface and serves as hydrogen bond acceptor of the proton at the N-␦1 position of His-450 and thus appears to be important for the structural integrity of the Mdm2 RING finger. Indeed, mutation of Thr-453 profoundly affects the E3 activity of Mdm2 and, to a lesser degree, the ability of Mdm2 to form homomers (39; this study).
Similar to Thr-453, we found that mutation of Gly-451 interferes with the E3 activity of Mdm2. Gly-451 is part of the putative WalkerA motif of Mdm2, and thus it can be assumed that the ability of the respective mutants to bind ATP is reduced. However, because ATP binding is not required for the E3 activity of Mdm2 (39), this (reduced ATP binding) does not explain the defect in E3 activity of the Gly-451 mutant. Furthermore, in agreement with the NMR analysis (47), our data indicate that Gly-451 is not of critical importance for the integrity of the Mdm2 homodimer. However, it cannot be excluded that mutation of this residue results in subtle changes of the RING finger structure. Subtle structural changes may also explain the observation that, similar to Thr-453, mutation of Gly-451 decreases the ability of Mdm2 to interact with UbcH5b (note that based on available structures of RING finger domains with their cog-nate E2s, Gly-451 does not appear to be directly involved in the interaction with UbcH5b).
How does MdmX rescue the defective E3 function of the Gly-451 and Thr-453 mutants? An obvious explanation is provided by the notion that Mdm2-MdmX heteromers are more stable than Mdm2 homomers (16) (Table 1). Similarly, the NMR structure data indicate that the interface of the Hdm2 RING/HdmX RING heterodimer is similar to the Hdm2 homodimer interface but that the Hdm2 RING domain is more stable when bound to the HdmX RING domain (47). These data suggest that, compared with Mdm2 homomers, the Mdm2-MdmX interaction involves additional and/or slightly different contact sites between both proteins. This formation of additional and/or different contact sites between MdmX and Mdm2 is further supported by the observation that the HdmX mutants G452S and T454A can efficiently rescue the E3 function of the respective Mdm2 mutants (i.e. G451S and T453A) (supplemental Figs. 1 and 2). Thus, an attractive but purely speculative hypothesis is that local structural deviations are more likely to be tolerated by Mdm2-MdmX heteromers than by Mdm2 homomers, and consequently, mutant Mdm2-MdmX heteromers may more efficiently interact with UbcH5b than the Mdm2 mutant alone. Besides this structural role, MdmX may have additional yet unknown functions in Mdm2-mediated ubiquitination/neddylation processes. While this manuscript was under revision, it was reported that MdmX can cooperate with Mdm2 in Mdm2-mediated ubiquitination (48,49). Furthermore, it was shown that mutation of Phe-487 of MdmX does not interfere with its ability to bind to Mdm2 but abrogates its ability to support Mdm2-mediated p53 degradation (49). However, the reason for the inability of MdmXF487 to cooperate with Mdm2 in p53 degradation remains unclear. To eventually determine the mechanism(s) by which MdmX stimulates the E3 activity of Mdm2, structural analysis of the respective complexes (Mdm2-UbcH5 and Mdm2-MdmX-Ubch5) will be required.
Studies with transgenic mice and cells derived from these have shown that Mdm2 and MdmX have, at least in part, independent nonredundant functions in the control of p53 (21)(22)(23)(24). In addition, it appears that Mdm2-mediated degradation plays a major role in keeping p53 at low levels, whereas a role for MdmX in p53 degradation is not obvious, as loss of MdmX expression does not result in a significant increase in p53 levels (24). How is this reconciled with the notion that MdmX stimulates Mdm2-mediated ubiquitination/degradation of p53 within cells? As discussed above, Mdm2 has to form homomers or heteromers with MdmX to exert E3 activity, and Mdm2-MdmX heteromers are more stable than Mdm2 homomers. Thus, we propose that under normal growth conditions intracellular Mdm2 levels are too low to allow the efficient formation of Mdm2 homomers but are sufficient for Mdm2-MdmX complex formation. Furthermore, mdm2 gene expression is known to be directly regulated by p53 (27). Because MdmX may interfere with the transcriptional transactivation properties of p53 in an Mdm2-independent (and thus degradation-independent) manner (21,24,50), knockdown of MdmX expression should result in enhanced expression of the mdm2 gene and, consequently, enhanced Mdm2 levels. The increased Mdm2 levels are then sufficient to efficiently target p53 for degradation. Although it may be difficult, if not impossible, to prove this scenario, the available data are consistent with it.
Similar to Mdm2, the Mdm2-MdmX complex has both ubiquitin ligase and Nedd8 ligase activity. Thus, an alternative possibility is that within cells the Mdm2-MdmX heteromer functions as a Nedd8 ligase rather than a ubiquitin ligase. To prove this possibility, it will be important to identify Mdm2 and/or MdmX mutants that have ubiquitin ligase activity but not Nedd8 ligase activity and vice versa. However, characterization of the Mdm2 mutants used in this study (Fig. 4C and data not shown) indicates that, at least at the level of Mdm2, this will be a rather challenging task.