MdmX Is a RING Finger Ubiquitin Ligase Capable of Synergistically Enhancing Mdm2 Ubiquitination*

It has been well documented that Mdm2 and its homologue MdmX not only are critical negative regulators of the tumor suppressor p53 but that both Mdm2 and MdmX interact to affect the function of the other. The mechanisms through which these effects are manifested, however, remain unclear. Although Mdm2 has been established as a RING finger ubiquitin ligase, MdmX has not been shown to possess this activity despite the extensive sequence homology between their respective RING finger domains. Here we demonstrate that MdmX acts as a ubiquitin ligase in vitro, being capable of autoubiquitination, as well as mediating the ubiquitination of p53. The addition of Mdm2 to in vitro ubiquitination assays containing MdmX results in a synergistic increase of ubiquitin conjugation. Analysis of the resulting ubiquitin conjugates reveals that this observed synergy reflects an increase in Mdm2 ubiquitination. This study also suggests that ubiquitination of Mdm2 and MdmX may not serve as a signal for degradation, as we show that each are capable of synthesizing non-lysine 48 polyubiquitin chains and, in fact, utilize multiple lysine linkages. Taken together, these findings suggest a more active role for MdmX in the Mdm2-MdmX-p53 regulatory network than has been proposed previously.

It has been well documented that Mdm2 and its homologue MdmX not only are critical negative regulators of the tumor suppressor p53 but that both Mdm2 and MdmX interact to affect the function of the other. The mechanisms through which these effects are manifested, however, remain unclear. Although Mdm2 has been established as a RING finger ubiquitin ligase, MdmX has not been shown to possess this activity despite the extensive sequence homology between their respective RING finger domains. Here we demonstrate that MdmX acts as a ubiquitin ligase in vitro, being capable of autoubiquitination, as well as mediating the ubiquitination of p53. The addition of Mdm2 to in vitro ubiquitination assays containing MdmX results in a synergistic increase of ubiquitin conjugation. Analysis of the resulting ubiquitin conjugates reveals that this observed synergy reflects an increase in Mdm2 ubiquitination. This study also suggests that ubiquitination of Mdm2 and MdmX may not serve as a signal for degradation, as we show that each are capable of synthesizing non-lysine 48 polyubiquitin chains and, in fact, utilize multiple lysine linkages. Taken together, these findings suggest a more active role for MdmX in the Mdm2-MdmX-p53 regulatory network than has been proposed previously.
The covalent attachment of ubiquitin to proteins plays a fundamental role in the regulation of cellular function and is achieved through three enzymatic steps (1)(2)(3)(4). In an ATP-dependent process, the ubiquitin activating enzyme (E1) 1 catalyzes the formation of a reactive thiolester bond with ubiquitin, followed by its subsequent transfer to the active site cysteine of a ubiquitin carrier protein (E2). The specificity of ubiquitin ligation arises from the association of the E2-ubiquitin thiolester with ubiquitin-protein isopeptide ligase (E3), which facilitates the formation of the isopeptide linkage between ubiquitin and its target protein. Eukaryotes possess a single ubiquitin activating enzyme and several families of E2 iso-forms that are easily identified by conservation of a bipartite E2 consensus motif (1). The recent discovery that RING finger proteins often exhibit ubiquitin ligase activity has led to the rapid identification of numerous novel ligases (5)(6)(7). One of the first of these RING finger ubiquitin ligases to be characterized was the oncoprotein Mdm2 (8). The mdm2 gene is amplified in a broad range of tumors including breast carcinomas, lung cancers, and osteosarcomas with the highest frequency (20%) in soft tissue tumors (9). It is well established that Mdm2 regulates the p53 tumor suppressor protein by inhibiting p53mediated transactivation and by promoting its degradation (10,11). For both of these mechanisms, elevated levels of Mdm2 can inhibit the protective functions of p53 and lead to oncogenesis. The ability of Mdm2 to negatively regulate p53 has sparked much interest in understanding how Mdm2 levels within the cell are controlled, demonstrated by numerous studies aimed at developing therapies designed to attenuate Mdm2 activity (12)(13)(14)(15). One obvious mechanism to control Mdm2 function is regulation of protein stability. Like several other RING finger ligases, Mdm2 is capable of catalyzing its own ubiquitination thus providing a possible mechanism for autoregulation (8).
The Mdm2-related protein MdmX shares several regions of homology with Mdm2 including the p53 binding domain, a zinc finger motif, and a C-terminal RING finger domain (16). Although the RING finger domains of MdmX and Mdm2 are very similar (46% identity over 52 residues), with the residues implicated in zinc chelation being absolutely conserved, there has been no reported ubiquitin ligase activity by MdmX. This apparent lack of ligase activity has contributed to the ambiguity surrounding the biological role of MdmX. Although structurally similar, genetic evidence demonstrates that MdmX cannot compensate for the loss of Mdm2 expression during embryonic development in mouse (17,18) nor can Mdm2 compensate for the loss of MdmX expression (19,20), suggesting that the two proteins have non-overlapping functions. The fact that the embryonic lethal phenotype displayed by MdmX null mice can be rescued in a p53 null genetic background has further demonstrated the importance of MdmX as a negative regulator of p53 function (19,20). The suggestion from these genetic studies that elevated MdmX levels may contribute to oncogenesis correlates with the recent reports of increased MdmX expression in tumor cell lines and malignant gliomas (21,22).
Previous studies have demonstrated that Mdm2 and MdmX form both homo-and hetero-oligomers in vitro and in vivo, with hetero-oligomers exhibiting a more stable association; in addition, deletions that include the RING finger domain of either protein abrogate this interaction (23). Recent studies have shown that the half-life of Mdm2 is increased when co-transfected with MdmX and that this stabilization is dependent on the RING finger of MdmX, suggesting a requirement for hetero-oligomer formation (16). One current model proposes that this observed stabilization of Mdm2 is the result of the Mdm2-MdmX interaction inhibiting the E3 ligase activity of Mdm2, thereby preventing its efficient ubiquitination and stabilizing Mdm2 protein levels.
Here we demonstrate that MdmX possesses ubiquitin ligase activity requiring the Ubc4/5 family of E2 isozymes and that the RING finger domain of MdmX is necessary for this activity. We further show that in addition to its capacity for autoubiquitination, MdmX catalyzes the ubiquitination of p53 in vitro. Our investigation into the proposed ability of MdmX to inhibit Mdm2 ubiquitination has revealed that the interaction between Mdm2 and MdmX results in a synergistic increase in Mdm2 ubiquitination. That MdmX autoubiquitination, as well as the MdmX-mediated increase in Mdm2 ubiquitination, may not target these proteins for degradation is suggested by our demonstration that both Mdm2 and MdmX are capable of forming non-lysine 48-linked polyubiquitin chains. These observations challenge the current hypothesis that the MdmX-mediated stability of Mdm2 reflects a decrease in Mdm2 ubiquitination and support the classification of MdmX as a new member of the RING finger family of ligases.

MATERIALS AND METHODS
Bovine ubiquitin, creatine phosphokinase, and yeast inorganic pyrophosphatase were purchased from Sigma. Thrombin, purified from bovine plasma, was purchased from Amersham Biosciences. The ubiquitin was purified to apparent homogeneity (24) and then radiolabeled by the chloramine T method (25). Alternatively, some of the purified ubiquitin was modified by reductive methylation (26). Carrier-free Na 125 I and [2, H]ATP were purchased from Amersham Biosciences. Human E1 from outdated human erythrocytes was purified to apparent homogeneity by modifying affinity chromatography and FPLC methods (27) and then quantitated by 125 I-ubiquitin thiolester assays and confirmed by the stoichiometric formation of ubiquitin [ 3 H]adenylate (28,29).
Plasmids-The entire coding sequence of human MdmX was amplified from a full-length cDNA (expressed sequence tag clone 177586) by polymerase chain reaction using Pfu polymerase (Stratagene). Appropriate primers were used during this amplification to allow for the in-frame insertion of the MdmX cDNA into the pGEX 4T-1 vector at the BamHI and XhoI sites. The construction of pGEX-MdmX⌬394 was achieved by digesting pGEX-MdmX with EcoRI and XhoI and blunting the 5Ј-overhangs with Klenow fragment, followed by re-ligation. This resulted in the deletion of the 96 carboxyl-terminal amino acids of MdmX. As was done with MdmX, the entire open reading frame of human p53 was amplified from a full-length p53 cDNA (expressed sequence tag clone 610186) utilizing appropriate primers to allow for the in-frame insertion of the p53 cDNA into the pGEX 4T-1 vector at the BamHI and XhoI sites.
A full-length human Mdm2 cDNA was obtained by polymerase chain reaction using a pGEX-Mdm2 cDNA construct kindly provided by Dr. Allan Weissman. Primers used in this amplification facilitated the insertion of this Mdm2 cDNA into the EcoRI and XhoI sites of a modified pGEX4T-1 vector containing a FLAG epitope fused to the C terminus of the GST coding sequence.
Recombinant Protein Production-The production and purification of human recombinant E2 proteins was performed as described previously (30). The concentration of active E2 protein in each preparation was determined by quantitation of the E1-catalyzed stoichiometric formation of the corresponding 125 I-ubiquitin thiolester. Ubiquitin lysine to arginine point mutants were produced and purified as described previously (24,31). DNA sequencing of the open reading frame of each plasmid, as well as mass spectrometry of each of the purified polypeptides, were performed to confirm that the expressed protein contained the appropriate lysine to arginine mutation. The human MdmX, MdmX⌬394, and FLAG-Mdm2 were expressed as N-terminal GST fusions in logarithmic phase Escherichia coli (Rosetta™; Novagen, Madison, WI) grown at 30°C. Expression at 30°C for 2 h was induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside. Following purification by glutathione affinity chromatography, the fusion proteins were further purified by FPLC. The fusion proteins were adsorbed to a Mono Q HR5/5 anion exchange FPLC column equilibrated with 50 mM Tris-HCl (pH 7.6) at a flow rate of 1 ml/min and then eluted with a linear 0 -0.5 M NaCl gradient (12.5 mM/min). Fractions containing the fusion protein, as determined by immunoblotting, were pooled and concentrated using an Ultrafree-4 centrifugal filtration device with a 5-kDa molecular mass exclusion (Millipore). The GST-FLAG-Mdm2 fusion protein was digested with thrombin (30 units/ml) for 3 h at room temperature to cleave the GST moiety, which was subsequently removed by glutathione affinity chromatography. Quantitation of GST-MdmX, GST-MdmX⌬394, and FLAG-Mdm2 was performed by silver staining using known concentrations of bovine serum albumin as standards.
Ubiquitin Conjugation Assays-Conjugation reactions were adapted from earlier kinetic methods (30). Reaction volumes of 25 l contained 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 1 mM dithiothreitol, 2 mM ATP, 10 mM creatine phosphate, 20 IU/ml creatine phosphokinase, 5 nM human ubiquitin activating enzyme, 400 nM of the indicated ubiquitin conjugating enzyme, and the indicated concentration of Mdm2 or MdmX. Reactions were initiated by addition of either 125 I-ubiquitin (typically ϳ10 4 cpm/pmol) or unlabeled ubiquitin and allowed to proceed for 15 min at 37°C. Reactions were quenched with 25 l of Laemmli sample buffer containing 2% (v/v) ␤-mercaptoethanol and boiled for 3 min. The conditions for conjugation reactions containing GST-p53 were similar to those stated above with the following modifications. Reaction volumes of 50 l containing 40 nM human ubiquitin activating enzyme, 150 nM HsUbc5A, 400 nM GST-p53, 4 nM FLAG-Mdm2, and 400 nM GST-MdmX were incubated for 2 h at 37°C either in the absence or presence of 10 M unlabeled ubiquitin. Reactions were quenched with 50 l of Laemmli sample buffer containing 2% (v/v) ␤-mercaptoethanol and boiled for 3 min.
Antibodies and Immunoprecipitation Assays-Commercially available mouse monoclonal antibodies were used for immunodetection of human Mdm2 (Ab-1; Oncogene Research Products) and human p53 (IgG 2a , clone BP53-12; Upstate Biotechnology). A ubiquitin-specific rabbit polyclonal antibody produced previously in the laboratory was used for detection of ubiquitin and polyubiquitin chains (32). An MdmXspecific rabbit polyclonal antiserum was produced against the peptide 142 ESSTSRKRTTEDDIP 156 synthesized on a four-branch multiple antigenic peptide resin (33). Following confirmation that this antiserum was immunospecific for MdmX, the antiserum was purified by peptide ligand affinity chromatography and used subsequently for immunoprecipitation and immunoblotting. Immunoprecipitation of MdmX-Ub conjugates was performed by incubating 20% of a ubiquitin conjugation reaction with a final concentration of 150 g/ml anti-MdmX antibody for 2 h at 4°C in phosphate-buffered saline containing 0.5% (v/v) Triton X-100 and 1 mg/ml carrier bovine serum albumin. The antibody-bound protein was precipitated with protein A-Sepharose CL-4B (Amersham Biosciences) and washed with phosphate-buffered saline containing 0.5% (v/v) Triton X-100 and 1 mg/ml carrier bovine serum albumin. Bound protein was eluted from the beads by boiling briefly with Laemmli sample buffer followed by resolution by SDS-PAGE.

RESULTS
MdmX Is a HsUbc5A-dependent Ubiquitin Ligase-Although there has been no E3 activity ascribed to MdmX to date, the high degree of homology between the RING finger domains of Mdm2 and MdmX suggested that MdmX might also function as a RING finger ubiquitin ligase. To investigate this possibility, we performed in vitro ubiquitin conjugation assays using a bacterially expressed GST-MdmX fusion protein as the only source of ubiquitin ligase activity. Previous studies have reported that E2 isoforms from the Ubc4/5 2 and HsUbc7 families interact with RING finger ligases, with Mdm2 displaying specificity for HsUbc5 family members (5,34). We assayed the ability of selected recombinant human E2 isoforms to support MdmX-catalyzed E3 activity. As expected, we observed no E3mediated ligase activity in reactions containing HsUbc2B, HsUbc3, or HsUbc7A as the source of E2 (Fig. 1A, lanes 1-4, 7, and 8); however the ability of HsUbc3 to catalyze its own monoubiquitination is demonstrated by the presence of a band migrating at ϳ42 kDa (Fig. 1A, lanes 3 and 4). In contrast, assays containing HsUbc5A, a member of the Ubc4/5 family of E2 ubiquitin conjugating enzymes (1), displayed robust ubiquitin ligation activity in the presence of GST-MdmX (Fig. 1A, lane 5 versus lane 6). As has been documented with other RING finger ligases, these ubiquitin conjugates were polyubiquitinated and migrated at the top of the gel (6). That the conjugates contained polyubiquitin chains was demonstrated by the complete abrogation of the high molecular weight adducts when 125 I reductively methylated ubiquitin was substituted for wildtype radiolabeled ubiquitin (data not shown).
We next addressed the nature of the ubiquitin conjugates by determining whether they represented MdmX autoubiquitination or the ubiquitination of trace contaminating proteins present in the enzyme preparations. After repeating the conjugation reaction of Fig. 1A (lane 6), we isolated GST-MdmX by immunoprecipitation with an anti-MdmX polyclonal antibody and found that GST-MdmX-Ub conjugates represented the majority (ϳ70%) of the total ubiquitin conjugates in the reaction (Fig. 1B). Because antibody concentrations were empirically optimized in all instances to ensure quantitative immunoprecipitation, the inability quantitatively to immunoprecipitate the ubiquitin adducts suggests MdmX either is capable of catalyzing the formation of free polyubiquitin chains, as found for E2 25K (35), or mediates conjugation to contaminating proteins present in the assay.
MdmX Ubiquitin Ligase Activity Is Dependent on Its RING Finger Domain-After we established that MdmX possessed E3 activity, we sought to confirm that this property was associated with its RING finger. Toward this end, we generated a truncated MdmX construct (MdmX⌬394) in which translation was terminated at codon 394, deleting the 96 C-terminal residues of the wild-type protein in which the RING finger is located ( Fig. 2A). MdmX⌬394 was expressed as an N-terminal GST fusion and compared with the full-length GST-MdmX to confirm the deletion of 11 kDa (Fig. 2B, lane 2 versus lane 3). MdmX has been reported to migrate at ϳ80 kDa contrary to its theoretical mass of 54 kDa (36), which is similar to the aberrant migration of Mdm2 (ϳ90-kDa apparent versus 54-kDa theoretical) (37). The apparent relative molecular masses of full-length GST-MdmX and of GST-MdmX⌬394 corresponding to 110 and 97 kDa, respectively, are consistent with the pre- dicted anomalous migration for deletion of the RING finger domain from GST-MdmX.
The ligase activity of MdmX⌬394 relative to that of fulllength MdmX was assayed using biochemically defined in vitro ubiquitin conjugation reactions. Although full-length MdmX displayed ligase activity, we observed no ubiquitin conjugates formed in the presence of an equal concentration of MdmX⌬394 or in the presence of MdmX⌬394 at a level three times greater than the full-length protein (Fig. 2C, lane 2 versus lanes 3 and  4). In an effort to determine whether less drastic alterations to MdmX would also abrogate its E3 activity, we attempted to selectively disrupt its RING finger. Because Zn 2ϩ stabilizes RING finger domains (38,39), we sequestered Zn 2ϩ from our MdmX protein preparations using the specific Zn 2ϩ chelator TPEN (40), which abrogates the E3 activity of Mdm2 (8). When MdmX was treated with 2 mM TPEN, we observed a biphasic decrease in ligase activity over time, presumably reflecting the time-dependent removal of Zn 2ϩ from the RING finger domain (Fig. 2D). Disruption of either Zn 2ϩ within the RING finger results in loss of ubiquitin ligase activity (8). Therefore, the biphasic kinetics for loss of MdmX activity most likely reflects different affinities for the two positions, the loss of either of which abrogates activity. Unlike Mdm2, whose activity is quantitatively restored after TPEN treatment by addition of ZnCl 2 (8), similar reconstitution of MdmX activity resulted in only a 4-fold increase in ubiquitin ligase activity (11% of the initial activity) after an overnight incubation in the presence of 3 mM ZnCl 2 , data not shown. These observations are consistent with an intact RING finger motif being required for MdmX-mediated ubiquitin ligase activity.
MdmX and Mdm2 Catalyze the Formation of Polyubiquitin Chains through Multiple Lysine Linkages-The ability of GST-MdmX to catalyze autoubiquitination suggests a mechanism for autoregulation. Because polyubiquitin chains linked via lysine 48 are the principle signal for targeting proteins for degradation (41) we sought to determine the linkage utilized by MdmX for chain formation. For this determination, we utilized our collection of ubiquitin lysine point mutants (24), each of which had one of the seven lysine residues contained within human ubiquitin mutated to arginine. These Lys 3 Arg point mutants have been used successfully in the past to demonstrate formation of unique chain linkages by ScUbc2/Rad6 and HsE2 epf (24). Using a ubiquitin-specific rabbit polyclonal antibody to detect the formation of polyubiquitin conjugates, we reconfirmed the ability of MdmX to form large molecular weight ubiquitin conjugates (Fig. 3A, lane 2). That these conjugates reflected the formation of polyubiquitin chains was confirmed by their elimination when reductively methylated ubiquitin, which is unable to form chains (24), was used (Fig.  3A, lane 3 versus lane 2). Substitution of ubiquitin Lys 3 Arg point mutants in these conjugation reactions revealed that GST-MdmX was capable of using each mutant to catalyze polyubiquitin chain formation (Fig. 3A, lanes 4 -10). This experiment demonstrates not only that MdmX is capable of synthesizing non-Lys 48 chains but that MdmX does not use one lysine exclusively for chain formation.
Given the homology between Mdm2 and MdmX, we asked whether Mdm2 could also utilize multiple lysine residues for chain formation. Using both wild-type and reductively methylated ubiquitin we were able to show that FLAG-Mdm2 catalyzed the formation of large molecular weight ubiquitin conjugates that were the result of polyubiquitin chain formation (Fig. 3B, lanes 2 and 3). Just as was seen with MdmX, when the ubiquitin Lys 3 Arg point mutants were substituted, Mdm2 was able to catalyze polyubiquitin chain formation (Fig. 3B,  lanes 4 -10). Recent reports have shown that HsUbc5A has been associated with the catalysis of Lys 29 -and Lys 48 -linked chains (42,43); however, the signaling function of Lys 29 ubiquitin chains remains unclear. It is worthwhile to note that other members of the RING finger family of ligases have been shown to generate non-Lys 48 ubiquitin chains, although their ability to utilize multiple lysine linkages have not been examined (44). The generation of ubiquitin Lys 3 Arg double mutants will be used in future experiments to determine whether both Lys 29 and Lys 48 are required for MdmX and Mdm2-mediated chain formation. (16,45). This has led to the suggestion that Mdm2-MdmX association results in the inhibition of Mdm2 autoubiquitination and a concomitant decrease in the rate of 26 S proteasome-mediated Mdm2 degradation (45,46). To test this directly, we examined the ability of MdmX to inhibit Mdm2-mediated autoubiquitination in vitro. That FLAG-Mdm2 and GST-MdmX associated with each other in our in vitro assays was confirmed by co-immunoprecipitation of GST-MdmX with FLAG-Mdm2 in the presence, but not in the absence, of an anti-Mdm2 monoclonal antibody (data not shown). To ensure that the FLAG-Mdm2 in the conjugation reaction was associated with GST-MdmX and not with other FLAG-Mdm2 molecules, GST-MdmX was added at a 200-fold molar excess. This large ratio of MdmX:Mdm2, in addition to the observation that Mdm2 has a stronger affinity for MdmX than for itself (23), should promote the formation of Mdm2-MdmX heterocomplexes over Mdm2 homo-oligomers. Instead of an MdmX-mediated reduction in ligase activity, we found that the addition of full-length MdmX to incubations containing Mdm2 resulted in a substantial non-additive increase in the initial rate of ubiquitin conjugate formation when compared with assays conducted with either protein alone (Fig. 4A, lanes 2 and  3 versus lane 4).

The Interaction Between Mdm2 and MdmX Enhances Ubiquitin Ligase Activity-Several studies have demonstrated through transfection assays that MdmX overexpression stabilizes Mdm2 in vivo
We examined whether this increased activity required the interaction of Mdm2 and MdmX by substituting MdmX⌬394 for full-length MdmX in these assays, because similar MdmX truncations fail to interact with Mdm2 both in vitro and in vivo (23,45). The failure of MdmX⌬394 to induce an increase in ubiquitin conjugation (Fig. 4A, lane 2 versus lane 6) suggests that the association of Mdm2 and MdmX is necessary for the increased conjugation and that this increase does not result from conjugation to contaminating bacterial proteins acting as nonspecific substrates. Quantitation of the initial rates of ligase activity revealed that the increase was not simply an additive effect of Mdm2 and MdmX activity but instead reflected an approximate doubling of total ligase activity (Fig. 4B). These results demonstrate that the Mdm2-MdmX association does not inhibit E3 activity as proposed in the literature; rather, it enhances the rate of ubiquitin ligation.
If the association between Mdm2 and MdmX induces this synergistic increase in ubiquitin ligation, one would expect to see this same effect at lower MdmX:Mdm2 ratios than were used in the previous assay. In the experiment of Fig. 4A, the concentration of GST-MdmX was twice that of HsUbc5A, leading to the potential that such a large excess could actually reduce the rate of conjugation in the reaction by competing  Fig. 1A contained the indicated combinations of 4 nM FLAG-Mdm2, 800 nM GST-MdmX, or 800 nM GST-MdmX⌬394. The resulting ligase activity was visualized by autoradiography of 125 I-ubiquitin conjugates following SDS-PAGE resolution. The mobility of molecular weight standards is shown to the left. B, the absolute amounts of ubiquitin conjugated in the reactions of A were quantitated by excising the poly-125 Iubiquitin conjugate bands from the gel and measuring the associated radioactivity by ␥-counting (25). C, in vitro ubiquitin conjugation assays contained 5 nM human E1, 400 nM HsUbc5A, 5 M 125 I-ubiquitin, and 2.5 nM FLAG-Mdm2. GST-MdmX was added to individual reactions to final concentrations ranging from 2.5 to 320 nM. Parallel reactions containing GST-MdmX at each concentration in the absence FLAG-Mdm2 were also performed to measure the initial rate of GST-MdmX ligase activity at the various concentrations (not shown). The resulting ligase activity was visualized by autoradiography of 125 I-ubiquitin conjugates following SDS-PAGE resolution. By excising the poly-125 I-ubiquitin conjugate bands from the gel and measuring the associated radioactivity by ␥-counting (25), the rate of ubiquitin conjugation was quantitated. The data are plotted as the net increase in ubiquitin conjugate formation above that observed for either Mdm2 or MdmX alone. The line through the data represents a hyperbolic regression fit of the data. transient Mdm2 2 ⅐MdmX 2 heterotetramer; therefore, it is not possible to interpret the physical meaning of the K1 ⁄2 value. However, the data demonstrate that a significant MdmX-mediated synergistic increase in ubiquitin conjugation occurs at a modest, presumably physiological, molar excess of MdmX (panel C) and that sequestering of HsUbc5A is a potential problem in interpreting the effects of MdmX when in excess over HsUbc5A (panel B).
MdmX Induces a Synergistic Increase in Mdm2 Ubiquitination-Although the experiments described in Fig. 4 showed that MdmX had a positive effect on ubiquitin ligation, they did not address the identity of the substrate for the enhanced ubiquitin conjugation. The tendency of Mdm2 and MdmX to co-precipitate, because of the formation of hetero-oligomers, precluded immunoprecipitation as a means to measure the relative ubiquitination of each protein. As an alternate approach, we performed Western blot analysis of parallel ligation reactions, using immunospecific anti-MdmX or anti-Mdm2 antibodies to visualize the ubiquitination state of MdmX and Mdm2, respectively. When we immunostained with anti-MdmX affinity-purified polyclonal antibody we were unable to detect high molecular weight MdmX-Ub conjugates, which were readily apparent by autoradiography (Fig. 5A, lane 6 versus Fig. 4A, lane 3). This is explained by a marked difference in the limit of detection for the two methods, because the Western blot reflects the percentage of MdmX that is ubiquitinated, whereas autoradiography reflects the number of 125 I-ubiquitin molecules ligated to the substrate, which favors detection of poly-ubiquitinated species because of the increased specific radioactivity of the ubiquitin chains with length. Therefore, the high molecular mass MdmX conjugates observed likely represent a small fraction of the total MdmX in the reaction and may not be easily detected by immunoblotting. We did, however, detect a small fraction of monoubiquitinated MdmX when incubated in the absence of Mdm2 that increased on addition of Mdm2 (Fig. 5A, lane 6 versus lane 4). In contrast, detection with anti-Mdm2 revealed a dramatic increase in the formation of polyubiquitinated Mdm2 on addition of MdmX that correlated with a decrease in non-ubiquitinated Mdm2 (Fig. 5B, lane 3 versus lane 4). From this observation, we conclude that the Mdm2-MdmX heterocomplex has an enhanced ability to ubiquitinate Mdm2 over that of the Mdm2 homocomplex. Taken together with the earlier observation that MdmX increases the overall rate of ubiquitin ligation (Fig. 4B), these results indicate that contrary to models based on in vivo overexpression of MdmX, the interaction between MdmX and Mdm2 results in a synergistic increase in the ubiquitination of Mdm2.
MdmX Catalyzes the Ubiquitination of p53 but Does Not Affect Mdm2-mediated Ubiquitination of p53-Because we had demonstrated that MdmX is a RING finger ubiquitin ligase with similar catalytic properties to Mdm2, we next addressed whether MdmX could ubiquitinate p53, a known Mdm2 substrate (5,8,47). Conjugation reactions containing GST-MdmX as the only source of ligase activity and GST-p53 revealed that MdmX was capable of ubiquitinating p53 in vitro, as detected by immunostaining of GST-p53 (Fig. 6, lane 4). The pattern and level of p53 ubiquitination by MdmX was similar to that of FLAG-Mdm2-mediated ubiquitination of p53 (Fig. 6, lane 4 versus lane 2); however, it must be noted that MdmX levels were 100-fold higher than Mdm2 in these assays, because MdmX consistently displayed lower ligase activity when compared with Mdm2.
In vivo, Mdm2, MdmX and p53 form a heterotrimeric complex (45). Our finding that the formation of the Mdm2-MdmX heterocomplex was associated with increased ligase activity raised the question of whether this complex, when associated with p53, would enhance the ubiquitination of the substrate. When both GST-MdmX and FLAG-Mdm2 were included in the p53 conjugation assay, we found no increase in the levels of p53-Ub conjugate formation when compared with levels catalyzed by either MdmX or Mdm2 alone (Fig. 6, lane 6 versus  lanes 2 and 4). These results not only provide the first evidence that MdmX is capable of ubiquitinating p53 in vitro but also suggest that the increased ligase activity associated with the MdmX-Mdm2 heterocomplex involves its autoubiquitination and not the ubiquitination of other substrates. DISCUSSION Genetic studies investigating the physiological consequences of the loss of either Mdm2 or MdmX function have demonstrated the requirement of both proteins for normal development and that each has non-overlapping roles (17)(18)(19)(20). Although one clear finding from these genetic studies is that each protein is involved in the negative regulation of the tumor suppressor p53, how these two related proteins affect the function of the other remains ambiguous. Recent studies have demonstrated that overexpression of MdmX stabilizes Mdm2 protein levels in vivo (16,45,46); however, the mechanism by which this occurs remains unclear. One proposed model asserts that the MdmX-Mdm2 interaction abrogates the E3 ligase activity of Mdm2, thereby reducing the ability of Mdm2 to target itself for 26 S proteasome-dependent degradation (45,46). Central to this model are the assumptions that MdmX does not possess intrinsic ubiquitin ligase activity and that the oligomerization of these proteins, presumably via their RING finger domains, inhibits Mdm2 ligase activity. Contrary to these assumptions, we have directly demonstrated in biochemically defined assays that MdmX is a RING finger ubiquitin ligase (see Figs. 1 and 2) and that the association of MdmX with Mdm2 results in the increased polyubiquitination of Mdm2 (see Figs. 4 and 5). At present, we cannot unambiguously distinguish whether MdmX directly conjugates Mdm2 or whether hetero-oligomer formation exerts a positive allosteric effect on the ligase activity of Mdm2. However, observation that heterodimer formation increases initial rates for MdmX, as well as Mdm2 conjugation (Fig. 5), tends to favor a mechanism of mutual allosteric activation.
The fact that the ligase activity of MdmX has been overlooked until now may be a reflection of its relatively low activity in autoubiquitination, ranging from 0.1 to 1% of that observed with Mdm2. The observed increase in ligase activity of the Mdm2-MdmX complex correlates with growing evidence that the dimerization of RING finger ligases does not interfere with their intrinsic E3 activity a priori. Recent studies with the breast and ovarian cancer tumor suppressor BRCA1 and one of its interacting proteins, BARD1, both of which are RING finger proteins possessing E3 activity, demonstrate that ubiquitin ligation is dramatically enhanced for the heterodimer (44,48). In fact, the structure of the BRCA1-BARD1 RING-RING complex reveals that the RING finger domains of the heterodimer make relatively few contacts (49).
Our finding that MdmX enhances the ubiquitination of Mdm2, as well as our demonstration of MdmX-mediated ubiquitination of p53, appears to be in conflict with studies demonstrating MdmX-mediated stabilization of both Mdm2 and p53, respectively. In the case of Mdm2, this discrepancy can be resolved by invoking a mechanism other than the inhibition of Mdm2 ubiquitination to explain the observed increase in the half-life of Mdm2. One possible mechanism is that MdmX acts to sequester Mdm2 to a subcellular compartment having a lower rate of Mdm2 degradation. Mdm2 contains a leucine-rich nuclear export sequence that has been shown to be required for its exit from the nucleus (50,51). In contrast, MdmX contains a nuclear export sequence similar to Mdm2 with the exception of a four-amino acid insertion (52) that could prevent MdmX from interacting with nuclear export machinery and may account for MdmX predominantly localizing within the nucleus (52). As a result, the association of MdmX with Mdm2 could impede nuclear export, thereby increasing the half-life of Mdm2. Stad et al. (46) have invoked a similar nuclear sequestration scenario to explain the observed stabilizing effect of MdmX on p53. In this way, the interaction between MdmX and p53 could stabilize p53 levels even if MdmX ubiquitinates p53.
Alternatively, resolution of these apparent conflicting observations may rest in the expression levels of MdmX during transfection assays. Evidence for the stabilization of Mdm2 by MdmX is predicated on overexpression of the latter (45,46); however, MdmX levels within cells are normally relatively constant (53). Quantitative biochemically defined rate assays in the present study demonstrate that, like Mdm2, MdmX is a RING finger ligase requiring ubiquitin conjugating enzymes of the Ubc4/5 family for activity (Fig. 1). Therefore, stabilization of Mdm2 by MdmX overexpression may simply represent the sequestering of Ubc4/5 isoforms by abnormally elevated levels of ectopic MdmX, thus interfering with the ability of Mdm2 to catalyze Ubc4/5-mediated autoubiquitination. The potential problems associated with overexpression of components of the ubiquitin ligation pathways for cellular dynamics have been discussed previously (1).
It is also important to note we have shown for the first time that both Mdm2 and MdmX can catalyze the formation of non-lysine 48 polyubiquitin chains. This strongly suggests that the increase in Mdm2 ubiquitination in the presence of MdmX also involves noncanonical chains, although this has not been experimentally established. If the Mdm2-MdmX complex does synthesize noncanonical chains, it raises the possibility that the increase in Mdm2 ubiquitination may not, in fact, serve as a signal for degradation as there is evidence that ubiquitin chains that utilize noncanonical lysine linkages are not appropriate signals for degradation (54). Again, a parallel can be drawn between the Mdm2-MdmX complex and the BRCA1-BARD1 complex, because autoubiquitination of the latter involves the formation of non-lysine 48 chains resulting in the mutual stabilization of BRCA1 and BARD1 (55).
Although it has clearly been shown in the literature that MdmX interacts with both p53 and Mdm2, the assignment of a specific biological role for MdmX has yet to be established. Although the function of MdmX remains elusive, its amplification in several tumor cell lines (22), and its requirement during development (20) leave little doubt of its biological significance. Indeed, the present observation that MdmX enhances Mdm2 ubiquitination redefines the functional relationships within this fundamentally important regulatory network. Our finding that MdmX is a new member of the RING finger family of ubiquitin ligases also offers new insights into possible function(s) of MdmX unrelated to Mdm2 and p53 interactions and suggests MdmX mediates the ubiquitination of novel substrates yet to be identified.