Mdm2 is a RING finger-dependent ubiquitin protein ligase for itself and p53.

Mdm2 has been shown to regulate p53 stability by targeting the p53 protein for proteasomal degradation. We now report that Mdm2 is a ubiquitin protein ligase (E3) for p53 and that its activity is dependent on its RING finger. Furthermore, we show that Mdm2 mediates its own ubiquitination in a RING finger-dependent manner, which requires no eukaryotic proteins other than ubiquitin-activating enzyme (E1) and an ubiquitin-conjugating enzyme (E2). It is apparent, therefore, that Mdm2 manifests an intrinsic capacity to mediate ubiquitination. Mutation of putative zinc coordination residues abrogated this activity, as did chelation of divalent cations. After cation chelation, the full activity could be restored by addition of zinc. We further demonstrate that the degradation of p53 and Mdm2 in cells requires additional potential zinc-coordinating residues beyond those required for the intrinsic activity of Mdm2 in vitro. Replacement of the Mdm2 RING with that of another protein (Praja1) reconstituted ubiquitination and proteasomal degradation of Mdm2. However, this RING was ineffective in ubiquitination and proteasomal targeting of p53, suggesting that there may be specificity at the level of the RING in the recognition of heterologous substrates.

Mdm2 has been shown to regulate p53 stability by targeting the p53 protein for proteasomal degradation. We now report that Mdm2 is a ubiquitin protein ligase (E3) for p53 and that its activity is dependent on its RING finger. Furthermore, we show that Mdm2 mediates its own ubiquitination in a RING finger-dependent manner, which requires no eukaryotic proteins other than ubiquitin-activating enzyme (E1) and an ubiquitinconjugating enzyme (E2). It is apparent, therefore, that Mdm2 manifests an intrinsic capacity to mediate ubiquitination. Mutation of putative zinc coordination residues abrogated this activity, as did chelation of divalent cations. After cation chelation, the full activity could be restored by addition of zinc. We further demonstrate that the degradation of p53 and Mdm2 in cells requires additional potential zinc-coordinating residues beyond those required for the intrinsic activity of Mdm2 in vitro. Replacement of the Mdm2 RING with that of another protein (Praja1) reconstituted ubiquitination and proteasomal degradation of Mdm2. However, this RING was ineffective in ubiquitination and proteasomal targeting of p53, suggesting that there may be specificity at the level of the RING in the recognition of heterologous substrates.
Increases in cellular levels of p53 in response to stress induces cell growth arrest and apoptosis (1)(2)(3). Accumulating evidence suggests that this crucial protein is regulated primarily through post-translational mechanisms (4,5). Central to the regulation of p53 is Mdm2, which is itself a transcriptional target of p53 (4,6). Two functional domains of Mdm2 are involved in regulating p53 levels. Its N terminus binds to p53 and targets p53 for degradation (7,8). Binding of p53 through this region also conceals the trans-activation domain of p53 and, therefore, inhibits its transcriptional activity (6). Diverse signals regulate p53 levels by regulating the interaction between Mdm2 and p53 (4). Stress-induced phosphorylation of serines in the trans-activation domain of p53 attenuates the binding and stabilizes p53 (4,9,10). Several cellular kinases, including Raf, DNA-PK, ATM, CAK, and JNK, can catalyze this phosphorylation (11)(12)(13)(14). Recent studies also indicate that nucleocytoplasmic shuttling of Mdm2 plays a significant role in p53 degradation (15)(16)(17)(18). Mutation of the nuclear export signal of Mdm2 abrogates its ability to target p53 for degradation, raising the possibility that the shuttling of p53 by Mdm2 from nucleus to cytoplasm is required for p53 to be subject to proteasome-mediated degradation (15,17). This is supported by a recent finding that ARF sequesters Mdm2, but not p53, into nucleoli and blocks export of Mdm2 from the nucleus to cytoplasm (17,19,20). Several proteins including E2F1, Myc, Ras, and adenoviral protein E1A stabilize p53 through the ARFmediated pathway (21)(22)(23)(24)(25). Other Mdm2-associated proteins also affect Mdm2-mediated p53 degradation; the interaction of p300 with Mdm2 promotes p53 degradation (26), whereas the interaction of Mdm2 with pRB and c-Abl stabilizes p53 in cells (27,28).
It is becoming clear that Mdm2-mediated degradation of p53 is proteasome-dependent (7,8). Recent studies have suggested that Mdm2 has the capacity to function as a ubiquitin (Ub) 1 protein ligase (E3) both for itself and for p53 in vitro (29,30). E3s provide specificity to Ub conjugation and are the final components in the multienzyme process that eventually leads to the covalent modification of proteins with Ub (31). The first step in Ub conjugation involves the ATP-dependent activation of Ub by a Ub-activating enzyme (E1), where a transient thiol ester linkage is formed with the C terminus of Ub. The activated Ub is then transferred to an active site cysteine of a Ub-conjugating enzyme (Ubc or E2). The final step requires an E3 that catalyzes the formation of an isopeptide linkage between the C terminus of Ub and ⑀-amino group of a lysine on a substrate (31). In some cases, such as the HECT (homologous to E6-AP C terminus) family E3s (32), a thiol ester intermediate between Ub and the E3 is formed (31). In other instances it is believed that the E3 binds the E2-Ub complex and mediates the direct transfer of Ub from E2 to substrates (31).
In this study, we demonstrate that Mdm2 is an E3 and that it functions both in ubiquitination of p53 and itself. The E3 activity of Mdm2 is dependent on its RING finger domain. Unlike the binding of the RING finger domain to RNA (33), coordination of zinc by the RING finger is required for E3 activity of Mdm2. Furthermore, by substitution of the RING finger from a heterologous protein, we demonstrate that another RING finger can substitute for that of Mdm2 for mediating its own ubiquitination in vitro and proteasomal targeting in cells.
All mutations were created using the Quick Change site-directed mutagenesis kit (Stratagene). Restriction sites were created in the mutation sites to facilitate screening by restriction digestion. All mutations were confirmed by sequencing. Oligonucleotide sequences used for mutagenesis are available upon request.
Cell Culture and Transfection-COS cells and U2OS osteosarcoma cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Cells were transfected using calcium phosphate precipitation. As previously reported (34), for examination of the degradation of p53 by Mdm2 and its mutants, U2OS cells were co-transfected with 3 g of Mdm2 or its mutant-encoding plasmids (pCHDM, pCHDM mutants, pCIneoMdm2, pCIneo-Mdm2 mutants, and pCIneo-Mdm2 PR ), 9 g of pcDNA3FLAG-p53, and 1 g of pEGFP. Cells were harvested 48 h after transfection. The total cell lysates were used for immunoblotting using an ECL detection kit as described (34).
Recombinant Protein Preparation-Glutathione S-transferase (GST) fusions of Mdm2 and its mutant were expressed in log phase Escherichia coli BL-21 (DE3) (Novagen, Madison, WI) that had been grown overnight at room temperature (RT) and induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 1 h, also at RT. Bacterial pellets were resuspended in 50 mM Tris, pH 7.4, 1 mM EDTA, 1% Triton X-100, 5 mM DTT, 2 mM phenylmethylsulfonyl fluoride (sonication buffer) and lysed by probe sonication, using 4 ml of sonication buffer per 100 ml of bacterial culture. The sonicate was clarified by centrifugation at 4°C for 15 min at 18,000 ϫ g, aliquoted, and stored at Ϫ70°C. Levels of expressed GST fusion proteins were estimated by incubating sonicates with glutathione-Sepharose beads (GS), washing, and quantification by SDS-PAGE followed by staining with Coomassie Brilliant Blue R-250. Known amounts of bovine serum albumin were used as standards. Preparation of wheat E1 and UbcH5B have been previously reported (36,37). 32 P-Labeled Ub was prepared using a plasmid encoding GST-Ub. GST-Ub was expressed in BL21 cells. After immobilized on GS, the GST-Ub was labeled with 32 P as described (38), followed by cleavage of the GST portion using thrombin (Amersham Pharmacia Biotech). Thrombin was depleted using benzamidine-Sepharose (Amersham Pharmacia Biotech) according to the manufacturer's protocol.
Recombinant mouse E1 in a baculovirus vector containing a His tag was a gift of Kazuhiro Iwai. Recombinant virus was added at a multiplicity of infection of 5 and allowed to infect the cells for 2 h at room temperature with occasional rocking. The viral stock was removed and replaced with supplemented Grace's medium (Invitrogen) for 2 days at 27°C. Cells were harvested and centrifuged at low speed for 10 min. The cell pellet was lysed with 1 ml of lysis buffer (100 mM Tris, pH 8, 100 mM NaCl, 1% Nonidet P-40, 1 mM DTT, 10% glycerol) per 8 ϫ 10 6 cells (ϳ1.5 ml buffer per flask). Lysate was clarified by centrifugation and stored at Ϫ70°C. Mouse E1 was purified using nickel-nitrilotriacetic acid-agarose (Qiagen), using the protocol supplied by the manufacturer. Briefly, 1 ml of clarified extract containing 8 mM imidazole was bound to nickel-nitrilotriacetic acid beads for 1 h at 4°C. Beads were washed three times with buffer containing 50 mM NaH 2 PO 4 , pH 8, 300 mM NaCl, 10 mM imidazole. The His-tagged mouse E1 was eluted in elution buffer (50 mM NaH 2 PO 4 , pH 8, 300 mM NaCl, 250 mM imidazole). The eluate was dialyzed against 50 mM NaH 2 PO 4 , 2 mM DTT overnight at 4°C using Pierce Slidealyzer dialysis cassettes with a molecular weight cut-off of 10 kDa that had been soaked in 0.5% bovine serum albumin before use. Final dialyzed eluate was evaluated by Coomassie Blue staining and by its ability to catalyze ubiquitination.
In Vitro Ubiquitination Assay-GST-Mdm2 (500 ng) was immobilized on GS. Ubiquitination assays were carried out by adding 20 ng each of bacterially expressed wheat E1, UbcH5B, and 2 ϫ 10 4 cpm of 32 P-labeled Ub in ubiquitination buffer containing 50 mM Tris, pH 7.4, 2 mM ATP, 5 mM MgCl 2 , 2 mM DTT, and 2 l of bacterial cell lysate from BL-21 cells transformed with "empty" pET15B. Reactions were in 30 l and were carried out for 1.5 h at 30°C with agitation in an Eppendorf Thermomixer. Following the ubiquitination reaction, the samples were heated to 95°C in SDS-PAGE sample buffer containing 2-mercaptoethanol prior to resolution on 7.5% SDS-PAGE. Ubiquitinated products were revealed by exposure to Phosphor screens and analyzed using a Storm PhosphorImager and ImageQuant software (Applied Biosystems).
For p53 ubiquitination assays, confluent COS cells were treated with MG132, 30 M, overnight, to accumulate endogenous p53. The cells were then lysed using Triton X-100 lysis buffer. The lysate was centrifuged, and the supernatant was collected and stored at Ϫ70°C. 200 l of 50 mM Tris buffer, pH 7.5, and 30 -50 l of COS cell lysate were added to GS-immobilized GST-Mdm2. The mixture was tumbled for 2 h at 4°C to form GST-Mdm2⅐p53 complex. Following extensive washing with 50 mM Tris buffer, the GST-Mdm2⅐p53 complex was used for ubiquitination assay that was carried out in ubiquitination buffer containing mouse E1 (200 ng), UbcH5B (150 ng), Ub (10 g) in a total volume of 30 l. The reaction proceeded at 30°C for 1.5 h in an Eppendorf Thermomixer at 800 rpm to keep the beads in suspension. Following the reaction, the samples were processed for SDS-PAGE and Western blot for p53.
A similar ubiquitination assay was carried out using [ 35 S]methionine-labeled p53. The latter was generated by in vitro transcription and translation in reticulocyte lysate using a T7-driven coupled TnT Kit (Promega). 2 ϫ 10 4 cpm of in vitro translated p53 were added to GST-Mdm2 in the presence of ubiquitination components as described above. The entire reaction mix was resolved on 7.5% SDS-PAGE and analyzed using a Storm PhosphorImager.
For alkylation experiments, GS-bound GST-Mdm2 and GST-Nedd4 were treated with the indicated concentration of N-ethylmaleimide (NEM) (Fluka, Ronkonkoma, NY) in PBS for 30 min at room temperature on a rotator. Negative controls were treated with PBS only. NEM was removed by two washes in 10 mM DTT/PBS, followed by 3 washes in PBS alone.

Mdm2 Mediates Ubiquitination of p53 and Mdm2
Itself-A GST fusion of Mdm2 was expressed in E. coli and assessed for its capacity to catalyze ubiquitination in vitro. GST-Mdm2 was immobilized on glutathione-Sepharose beads (GS) followed by a ubiquitination reaction using recombinant 32 P-labeled Ub in the presence of E1 and E2 (UbcH5B), both of which were also expressed as recombinant proteins in E. coli. As is evident (Fig.  1A), GST-Mdm2 mediated ubiquitination that was dependent on E1 and E2. Although all of the proteins present in the bacterial cell lysate are potential ubiquitination substrates, at least 90% of the high molecular weight species detected in this reaction represent ubiquitination of GS-immobilized GST-Mdm2 itself (data not shown). These findings are consistent with a previous report that Mdm2 expressed in eukaryotic cells is able to mediate its own ubiquitination in vitro (30) and demonstrate that this activity is not dependent on other proteins that might co-purify with Mdm2 when expressed in eukaryotic cells. Rather these findings establish that the capacity to mediate E2-dependent ubiquitination is intrinsic to Mdm2.
The activity of the recombinant GST-Mdm2 fusion protein toward p53 was evaluated in an in vitro p53 ubiquitination assay using p53 derived from cells. In this assay p53 was bound to GST-Mdm2 and assessed for ubiquitination followed by resolution on SDS-PAGE and immunoblotting with anti-p53. Higher molecular weight species indicative of addition of multiple Ub moieties were seen only in the presence of added E1 and E2 (Fig. 1B) and Ub. A similar result was obtained using 35 S-labeled p53 generated by in vitro translation, with a smear of ubiquitinated p53 extending to the top of the gel (Fig. 1C).
Effects of Alkylation and Zinc Chelation on E3 Activity of Mdm2-The initial description of Mdm2 as an E3 suggested that the activity of this protein is dependent on a HECT-like domain within the C-terminal region with evidence of a thiol ester linkage of Ub with Mdm2 (29). Cysteine 464 (Cys-464) was suggested to be the active cysteine analogous to that in HECT E3s, such as E6-AP and Nedd4 (29,32). However, the cysteine-rich C-terminal region, including Cys-464, of Mdm2 has also been shown to be a RING finger and to coordinate zinc (33, 40) ( Fig. 2A). Since RING finger proteins may also participate in mediating E2-dependent ubiquitination (37,(41)(42)(43)(44)(45)(46)(47), it may be that the zinc-coordinating RING finger of Mdm2 and not a HECT-like domain is required for its capacity to mediate ubiquitination. To begin to evaluate these alternative possibilities, we tested the sensitivity of Mdm2 to the alkylating agent N-ethylmaleimide (NEM). As is evident, when exposed to concentrations of this reagent as low as 0.1 mM, the activity of a HECT E3 (Nedd4) was largely abrogated (85%), whereas the activity of Mdm2 was not significantly affected (Fig. 2B). However, at higher concentrations of NEM, some diminution in activity was also evident for Mdm2, possibly due to alkylation of cysteines within the RING finger that participate in zinc coordination. Nevertheless, this result is consistent with observations that ubiquitination mediated by RING finger proteins is relatively insensitive to alkylating agents (45). 2 Further differentiation between HECT and RING-mediated ubiquitination took advantage of the fact that HECT domain E3s do not require divalent cations for their activity, whereas RING finger proteins require zinc (37). As shown in Fig. 2C, treatment of GST-Mdm2 with the potent metal chelator, TPEN (39), led to a total loss of Mdm2 ubiquitination. The ubiquitination of Mdm2 was readily restored by addition of ZnCl 2 . Similarly, the capacity to catalyze p53 ubiquitination in vitro was also abrogated by TPEN and restored by ZnCl 2 (Fig. 2D). To rule out the possibility that the requirement for zinc reflected a need for the zinc finger of Mdm2, a zinc-coordinating cysteine (Cys-305) within the zinc finger was mutated to serine. The mutant retained the capacity to catalyze E2-dependent ubiquitination of p53 (Fig. 2E). Thus, an intact zinc finger is not required for p53 ubiquitination by Mdm2 in vitro. These results suggest that a zinc-coordinated RING finger likely provides the basis for the E3 activity of Mdm2.
RING Finger Mutations Abolish E3 Activity of Mdm2-RING finger domains are conserved cysteine-rich amino acid sequences that form two interleaved zinc-binding sites with a total of eight cysteines and histidines constituting the sites of metal coordination (48,49) (Fig. 3A). RING fingers are classified as RING-HC or RING-H2 depending on whether they have a cysteine or a histidine in the fifth coordination site. Based on the crystal structure of the RING-HC finger of promyelocytic leukemia protein (PML) (50), cysteines or histidines in the first, second, fifth, and sixth positions coordinate one zinc ion (Site 1) and those in the third, fourth, seventh, and eighth positions coordinate the second (Site 2) (Fig. 3A). The data presented in Fig. 2 indicate that chelation of zinc from the Mdm2 RING finger results in loss of E3 activity. Thus, we would predict that mutation of zinc coordination sites in the RING finger would similarly abrogate E3 activity. Because the Mdm2 RING does not fit within RING finger consensus se- Mdm2 RING Finger-dependent E3 Activity quences, particularly with regard to spacing between the third and fourth coordination residues, it has been speculated that a threonine (Thr-455) rather than a cysteine together with His-457 might constitute the third and fourth zinc-coordinating residues, respectively (40) (Fig. 3B). However, based on metal binding studies in which Cys-449 and His-452 were simultaneously mutated, it has been suggested that Cys-449 and His-452 instead represent the third and fourth residues (33) (Fig.  3B). We therefore mutated each of the 10 potential zinc coordination residues and evaluated the effects on the E3 activity of Mdm2.
Mutation of eight different potential zinc coordination residues resulted in complete loss of Mdm2 ubiquitination. These amino acids are Cys-438, His-452, His-457, Cys-461, Cys-464, Cys-475, Cys-478 (Fig. 3C, lanes 4, 6, and 8 -12), and Cys-441 (data not shown). Mutation of Thr-455 (Fig. 2C, lane 7) resulted in a significant decrease in overall ubiquitination with a pattern that was different from that seen with wild type Mdm2. The significance of this altered pattern of Mdm2 ubiquitination is unknown. In Fig. 3C the ubiquitination seen with Thr-455 was 49% of wild type and is representative of results from multiple experiments. Mutation of Cys-449, which was predicted to be a coordinating residue based on metal binding studies (33), did not result in a substantial decrease in ubiquitination. As expected, mutation of Cys-373 (first cysteine upstream of the RING finger) affected neither the level nor pattern of ubiquitination (Fig. 3C, lane 3).
The effects of RING finger mutations on p53 ubiquitination were next examined. As is evident, RING finger mutations that abolished the intrinsic E3 activity of Mdm2 (Fig. 3C) similarly prevented ubiquitination of p53 derived from cells (Fig. 3D,  lanes 3 and 5 and 7-11) or p53 translated in vitro (Fig. 3E, lane  3 and 6 -8). Additionally, the Thr-455 mutation exhibited a marked diminution in p53 ubiquitination (Fig. 3C, lane 7, D,  lane 6, and E, lane 5). As expected (51), none of these mutations affected the physical association of Mdm2 with p53 (Fig. 3D). These in vitro findings implicate three (His-452, Thr-455, and His-457) of the four residues postulated to be the third and fourth zinc-binding sites as being important for the E3 activity of Mdm2 in vitro.

RING Finger Mutations Stabilize p53 and Mdm2 in Cells-
Previous studies have shown that overexpression of Mdm2 reduces p53 levels in cells due to its ability to target p53 for proteasome-dependent degradation (7,8,34). We have shown that the intact RING finger is required for Mdm2 to ubiquitinate p53 in vitro. Therefore, we next examined whether an intact RING finger is required for Mdm2 to target p53 for degradation in cells. U2OS cells were co-transfected with plasmids encoding wild type or RING finger mutants of Mdm2 together with FLAG-tagged p53. A plasmid encoding green fluorescent protein (GFP) was used to control for transfection efficiency. Forty eight hours after transfection, p53 and Mdm2 levels were examined by Western blot. As expected, Mdm2 down-regulated p53 (Fig. 4, middle panel, lane 2), as did the mutant in the zinc finger cysteine (Cys-305) (Fig. 4, middle  panel, lane 3). Consistent with the in vitro results, Mdm2 RING finger mutations that failed to ubiquitinate p53 in vitro did not target p53 for degradation when expressed in cells (Fig. 4,  middle panel, lanes 4 -9). However, the C449S mutant, which catalyzed ubiquitination of p53 in vitro, also did not target p53 for degradation in cells (Fig. 4, middle panel, lane 5). In contrast, the zinc finger mutant, C3055, was not impaired in targeting p53 for degradation (Fig. 4, middle panel, lane 3). Additionally, when the levels of the transfected wild type and mutant Mdm2 were evaluated, all of the RING finger mutants accumulated including C449S and T455A (Fig. 4, upper panel,  lane 4 -9). These data support the idea that the E3 activity of Mdm2 in vivo requires an intact RING finger, confirms the importance of Thr-455 in the function of this molecule, and also implicates Cys-449 as being important for the in vivo function of Mdm2.
Effect of Heterologous RING Finger on E3 Activity of Mdm2-The finding that the E3 activity of Mdm2 is dependent on its RING finger brings about the interesting question of whether a heterologous RING finger could substitute for that of Mdm2 in mediating ubiquitination. We have recently deter- FIG. 3. GST-Mdm2 E3 activity is dependent on its RING finger. A, RING finger consensus sequence and schematic presentation of RING finger. B, sequence of Mdm2 RING finger. The potential zinc coordination residues are in bold and numbered. C, GST-Mdm2 and mutants were evaluated for mediating their own ubiquitination as in Fig. 1A. D, GST-Mdm2 and mutants were evaluated for ubiquitination of p53 derived from COS cells as in Fig. 1B. E, GST-Mdm2 and mutants were evaluated for ubiquitination of 35 S-p53 translated in vitro as in Fig. 1C. mined that a number of otherwise unrelated RING finger proteins of unknown function all have the capacity to mediate E2-dependent ubiquitination (37). Where evaluated, the RING finger by itself was not sufficient for ubiquitination, and converting from a RING-H2 to a RING-HC consensus sequence does not ensure retention of activity. Among the RING-H2 finger proteins that have the capacity to mediate ubiquitination is the 398-amino acid protein Praja1, which like Mdm2 has its RING finger at its C terminus (35,37).
When the C terminus of Praja1, including the RING finger and 17 amino acids as linker, was substituted for that of Mdm2 (Fig. 5A), this chimeric protein (Mdm2 PR ) manifested substantial E2-dependent ubiquitination (Fig. 5B, lanes 5 and 6). The ubiquitinated species formed a smear beginning above the point where the chimera migrates, suggesting a high level of E2-dependent ubiquitination of the chimeric protein. The ubiquitination of the chimera was confirmed by immunoprecipitation for Mdm2 PR following a ubiquitination assay (data not shown). Like GST-Mdm2, ubiquitination of GST-Mdm2 PR occurred regardless of GS immobilization. These data demonstrate that a heterologous RING finger domain can substitute for that of Mdm2 in mediating ubiquitination.
In accord with the high level of ubiquitination observed in vitro, Mdm2 PR was barely detectable when expressed in cells (Fig. 5C, upper panel, lane 4, the faint band is endogenous Mdm2), whereas a mutant in which the fifth coordination residue histidine was mutated to serine (Mdm2 PR(H3 S) ) accumulated to high levels (Fig. 5C, upper panel, lane 5, upper band). Equal transfection efficiency was confirmed by blotting for co-transfected GFP (Fig. 5C, lower panel). Consistent with proteasomal degradation in vivo, Mdm2 PR levels increased markedly when cells were exposed to the proteasome inhibitor MG132 for 4 h, whereas levels of the RING-mutated form were not altered (Fig. 5D, upper panel, compare the upper bands in  lane 2 with 5 and lane 3 with 6). The co-transfected GFP showed no change after MG132 treatment (Fig. 5D, lower panel). Thus, it is apparent that this chimeric protein efficiently targets itself for degradation in a proteasome-dependent manner.
The possibility that the heterologous Praja1 RING can target p53 for ubiquitination was next evaluated. As expected, both GST-Mdm2 and GST-Mdm2 PR bound p53, whereas GST-Praja1 did not (Fig. 6A). Despite the ability of GST-Mdm2 PR to bind p53 and its high intrinsic capacity to mediate its own ubiquitination (Fig. 5B), little evidence of higher molecular weight forms of p53 was observed (Fig. 6A, compare lanes 1, 4,  and 5). Consistent with this, unlike wild type Mdm2, Mdm2 PR failed to down-regulate co-transfected p53 (Fig. 6B, panel p53). Thus, although the Praja1 RING is sufficient to target Mdm2 for ubiquitination and proteasomal degradation, the presence of an active RING at the C terminus of Mdm2 is not adequate for substantial ubiquitination of p53 nor apparently to target p53 for degradation in cells.

DISCUSSION
In this study we demonstrate that the intrinsic E3 activity of Mdm2 is dependent on its zinc-coordinated RING finger domain. The capacity to mediate Mdm2's own ubiquitination requires no eukaryotic protein other than E1 and E2. This finding is consistent with recent observations that multiple otherwise unrelated RING finger proteins, including AO7, BRCA1, NF-X1, Siah-1, TRC8, kf-1, and Praja1, have the intrinsic capacity to mediate ubiquitination (37) and the finding of RING fingers within other E3s. Among known E3s, the N-end rule E3s (Ubr1, E3␣) have RING finger domains, as does the Apc11 subunit of the anaphase-promoting complex (52,53). A protein containing a RING finger-like domain variously referred to as Rbx1, ROC1, or Hrt1 functions as a component of SCF (Skp1, Cdc53/cullin, F box) and VHL (von Hippel-Lindau tumor suppressor protein) E3 complexes (41)(42)(43)(44)(45)(46). More recently, the RING finger domain of Cbl has been implicated in ubiquitination of the epidermal growth factor receptor and the platelet-derived growth factor receptor (47,54).
The importance of the RING finger domain in ubiquitination is also supported by the fact that several RING finger proteins have been found to associate with Ubcs and/or to target specific proteins for proteasome-dependent degradation, although their E3 activities toward these proteins have not been shown. This is the case for the Drosophila RING finger protein SINA, which associates with UbcD1 and targets Tramtrack for degradation (55,56). SINA and its mammalian homologues, Siah-1 and Siah-2, interact with Ubc9 through their N termini, which includes the RING finger domain. Siah-2 was found to target nuclear receptor corepressor for degradation (57). Furthermore, the RING finger is required for targeting Siah-1 itself and the deleted in colorectal cancer gene product for degradation in cells (58,59), which is consistent with the RING finger and E2-dependent auto-ubiquitination of Siah-1 (37). Additionally, the Vmw110 protein of herpes simplex virus type 1 has the RING finger-dependent capacity to target several cellular proteins, such as DNA-PK, SUMO-conjugated PML, and CENP-C, for degradation after infection (60 -62). In yeast, Hrd1p (Der3) is found to target misfolded ER proteins for degradation in RING finger-dependent manner (63,64). RAD6, an E2, associates with a RING finger protein RAD18 (65). However, the significance of this association remains to be determined.
Whereas several RING finger proteins have been directly shown to coordinate zinc (33,66), for the most part assignment of a RING finger domain to proteins has been based on the presence of consensus sequences, with neither direct evidence of metal binding nor functional correlations (48,49). The determination that RING finger proteins mediate E2-dependent ubiquitination (37) now allows an assessment of structurefunction relationships for the RING finger domain. Based on primary amino acid sequence, Mdm2 does not fall within the limits of RING finger consensus sequences, as there is no appropriately spaced cysteine-histidine pair in the region of the third and fourth coordination sites. This led Freemont and colleagues (40) to postulate, based on sequence alignment, that Thr-455 and His-457 represent the third and fourth coordina- tion sites. However, metal binding studies in which Cys-449 and His-452 were simultaneously mutated have implicated at least one of these residues as a zinc ligand at the third and/or fourth coordination sites (33). Our in vitro analysis of Mdm2 auto-ubiquitination implicates His-452 and His-457 as being required for activity. However mutation of Thr-455 alters both the pattern of auto-ubiquitination and the efficiency of this process, markedly decreases in vitro ubiquitination of p53, and results in the stabilization of both Mdm2 and p53 in cells. Mutation of Cys-449 also stabilizes both Mdm2 and p53 in vivo. Thus, mutation of any of the four amino acids postulated to be the third and fourth zinc ligands adversely affects the function of Mdm2. Whether only two of these four residues actually participate in metal coordination, with the others affecting the structure or stability of this atypical RING, or whether there is an additional degree of complexity to metal coordination in this region requires further analysis. Regardless, these results demonstrate that cysteine-and histidine-rich RING-like regions that mediate ubiquitination can vary significantly from canonical RING consensus sequences.
Although in vitro ubiquitination of p53 by Mdm2 requires addition of only E1, E2, and Ub, we cannot exclude the participation of additional proteins in the binding to, or ubiquitination of, p53. Indeed, it has been suggested that p300 may function as a platform, bringing together the necessary catalytic and regulatory factors needed for p53 ubiquitination (26). However, it is clear from the present study that RING fingerdependent ubiquitination of Mdm2 requires only E1 and E2. Whereas otherwise unrelated RING finger proteins have the capacity to mediate ubiquitination, it is also the case that the RING fingers of these proteins vary substantially in sequence (37). This raises the possibility that, in addition to serving as regions that interact with and activate E2s, some degree of substrate specificity also resides within the RING finger. Consistent with this, Mdm2 binds to the p53-related molecule p73 but does not target it for degradation (67)(68)(69). Therefore, physical association alone is apparently not sufficient to target Mdm2-associated proteins for degradation. One explanation for this is that the tertiary structure of the Mdm2 RING specifically facilitates the juxtaposition of a RING-associated complex of E2 and Ub with target lysines in p53. The finding that Mdm2 PR mediates its own ubiquitination in vitro and proteasomedependent degradation in cells, but is ineffective in ubiquitinating p53 in vitro or in targeting p53 for degradation in cells, is in agreement with the concept of substrate specificity at the level of the RING. An alternative explanation for the failure of Mdm2 PR to ubiquitinate p53 is that unanticipated steric constraints generated by fusing the Praja1 RING to Mdm2 through the Praja1-derived linker precludes access to target lysines within p53. However, this latter possibility seems less likely as a second chimera, in which the Mdm2 RING was replaced by the Praja1 RING without the Praja1derived linker, also targets itself, but not p53, for degradation in cells. 3 In addition to playing roles in ubiquitination, RING fingers have other cellular functions. The RING finger of PML has been implicated either directly or indirectly in its association with Ubc9 and in modification of PML with the Ub-like molecule Sumo-1 (70). Zinc-coordinated RING fingers have been shown to mediate dimerization of proteins such as BARD1 and FIG. 6. Mdm2 PR does not ubiquitinate p53 in vitro or target p53 for degradation in cells. A, p53 derived from COS cells was incubated with the indicated GST fusions followed by a ubiquitination assay and Western blotting for p53. B, cells were transfected with plasmids encoding either wild type Mdm2, Mdm2 PR , or Mdm2 H452A together with plasmid encoding FLAG-p53. Expression of Mdm2 and p53 was examined by Western blot. Equal transfections were confirmed by blotting for co-transfected GFP. BRCA1, Mdm2 and MdmX, and constitutive photomorphogenic protein 1 and CIP8 (71)(72)(73). Recently a RING finger-like domain, the FYVE domain, has been determined to bind specifically phosphatidylinositol 3-phosphate (74). The FYVE domain contains a highly conserved motif that provides specificity for binding to phosphatidylinositol 3-phosphate (74). This ability to bind phosphatidylinositol 3-phosphate is also dependent on zinc coordination, and the removal of zinc results in the unfolding of this domain (74). In the case of Mdm2 the RING finger not only mediates ubiquitination (this study) but also binds to RNA (33). Notably, this RNA binding has been found to occur whether or not zinc is present (33). An interesting question that now arises is whether the physical association of Mdm2 with RNA affects the ability of this RING finger protein to mediate ubiquitination.
The mdm2 proto-oncogene is overexpressed and amplified in a variety of human tumors (75). The tumorigenic activity associated with a high level of Mdm2 may be due to its ability to target p53 for degradation, leading to inhibition of p53-induced cell growth arrest and apoptosis (6). Mdm2 has also been found to induce spontaneous tumorigenesis independent of p53 (76). Identification of the requirement for a zinc-coordinated RING finger in the E3 activity of Mdm2 toward p53 may provide a potential new target for assessing the tumorigenic activity of Mdm2 and for re-activation of p53 in tumor therapy.