Tumor Suppressor ARF Promotes Non-classic Proteasome-independent Polyubiquitination of COMMD1*

Although the tumor suppressor ARF is generally accepted for its essential role in activating the p53 pathway, its p53-independent function has also been proposed. Here, we report that ARF associates with COMMD1 and promotes Lys63-mediated polyubiquitination of COMMD1 in a p53-independent manner. We found that ARF interacts with COMMD1 in vivo. Deletion analysis of ARF suggested that the N-terminal amino acids 15-45 are important for its interaction with COMMD1. In addition, we found that endogenous ARF redistributes from the nucleolus to the nucleoplasm and interacts with COMMD1 when DNA is damaged by actinomycin D. Interestingly, we found that ARF promotes the polyubiquitination of COMMD1 through Lys63 of ubiquitin but not the polyubiquitination of Lys48, which does not target COMMD1 for proteasome-dependent proteolysis. Moreover, ARF mutants lacking the domain interacting with COMMD1 did not promote COMMD1 polyubiquitination, indicating that physical association is a prerequisite condition for the polyubiquitination process. Together, these data suggest that the ability to promote Lys63-mediated polyubiquitination of COMMD1 is a novel property of ARF independent of p53.

The INK4a/ARF locus encodes two potent and distinct tumor suppressors, p16 INK4a and ARF (p14ARF in human and p19ARF in mouse) (1)(2)(3). p16 INK4a acts as an inhibitor of cyclindependent kinases to increase the growth-suppressive activity of pRb protein, whereas ARF activates the p53 pathway through its inhibitory effect on MDM2 (4). In response to oncogene activation, ARF binds to MDM2 and inhibits its E3 2 ligase activity for p53. Hence it protects p53 from degradation through a ubiquitin-proteasome pathway. In some conditions, sequestering MDM2 in the nucleolus by ARF may also contribute to stabilization and activation of p53 (5), but it has been demonstrated not to be essential for the inhibitory effect of ARF on MDM2 (6).
Apart from its major function in activating the p53 pathway, ARF does regulate other cellular activities such as ribosomal RNA processing and gene expression (7,8). Although it does not contain any lysine residue, ARF undergoes proteasome-dependent degradation (9). Moreover, ARF also regulates the turnover of its binding partners by affecting the ubiquitination process. ARF inhibits the function of B23, a nucleolar endoribonuclease involved in 28 S RNA maturation, through promoting B23 polyubiquitination and proteasomal degradation (10). In some circumstances, overexpression of ARF destabilizes MDM2, E2F1, and DP1 by facilitating the polyubiquitination of the proteins (11)(12)(13)(14). Strikingly, in addition to promoting ubiquitin conjugation, ARF also targets a number of its binding partners including MDM2, NPM (B23), Werner helicase, HIF-1␣, and E2F1 for small ubiquitin-like modifier modification (11,15,16). The extensive effect of protein sumoylation induced by ARF may include control of protein stability, formation of subnuclear structure, and regulation of transcriptional activities in a p53-and MDM2-independent manner (17,18). COMMD1 (copper metabolism gene MURR1 domain-containing protein 1, previously known as MURR1) is a multifunctional factor that was identified to be involved in copper metabolism (18). Some of its functions are demonstrated related to the factors involved in apoptosis. XIAP, a potent inhibitor of apoptosis, can regulate cellular copper levels by promoting polyubiquitination and degradation of COMMD1 (19). In addition, COMMD1 blocks NF-B activation by either inhibiting the proteasome-dependent degradation of IB or interfering with the interaction between the NF-B component RelA and chromatin (20,21).
In this study, we have characterized COMMD1 as a novel binding partner of ARF. ARF colocalizes with COMMD1 in the nucleoplasm and promotes Lys 63 -mediated, proteasome-independent polyubiquitination of COMMD1.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-The following antibodies were used in this study: mouse anti-FLAG M2 monoclonal Ab (Sigma), rabbit anti-FLAG polyclonal Ab (Sigma), rabbit anti-HA polyclonal Ab (Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti-Myc monoclonal Ab (Cell Signaling Technology), goat anti-actin polyclonal Ab (Santa Cruz Biotechnology), rabbit anti-ARF polyclonal Ab (Neomarkers), mouse anti-␣- tubulin monoclonal Ab (Molecular Probes), rabbit anti-GCN5 Ab (Cell Signaling Technology), and mouse anti-COMMD1 Ab (Abnova). Rhodamine-and fluorescein isothiocyanate-conjugated secondary antibodies were purchased from Molecular Probes. Horseradish peroxidase-conjugated protein G was purchased from Bio-Rad. MG132 and cycloheximide were purchased from Sigma. Yeast Two-hybrid Screening-A Gal4-based system (Matchmaker, Clontech) was used according to the manufacturer's instructions. Full-length human p14ARF was cloned into pGBKT7, which carries a DNA-binding domain, to create a bait protein. This plasmid was transformed into yeast strain AH109, and no spontaneous transactivation was detected. The transformed AH109 cells were then mated with yeast strain Y187 cells pretransformed with a Clontech Matchmaker human skeletal muscle cDNA library fused with a DNA activation domain in the pACT2 plasmid. After screening, plasmids were extracted from positive colonies, propagated in Escherichia coli, and analyzed by automated sequencing.
Plasmid Construction-To create the mammalian expression plasmids, full-length ARF was amplified using pGBKT7-ARF as the template and subcloned into the pcMV5-HA vector. COMMD1 cDNA was amplified using reverse transcription-PCR from HeLa total RNA extracts and subcloned into the pcDNA3-FLAG vector. pCMV5-Myc-ub and pCMV5-Myc-ubK48R were kindly provided by Dr. Peter Cheung. The K63R mutant of ubiquitin was generated by introducing the mutation site into the reverse primer and then subcloning it into the pCMV5 backbone.
Cell Culture and Transfection-Human NCI-H1299 cells were grown in RPMI 1640 medium with 2 mM L-glutamine, 2 g/liter sodium bicarbonate, 4.5 g/liter glucose, 10% fetal bovine serum, 100 g/ml penicillin, and 100 g/ml streptomycin (Invitrogen). Transfection of cells with mammalian expression constructs by Lipofectamine 2000 (Invitrogen) was according to the methods recommended by the manufacturer.
Protein Stability Assay-H1299 cells grown on 6-well plates were transfected with the plasmid as indicated. 24 h after transfection, cells were treated with 25 g/ml cycloheximide or 2 M MG132 for the indicated period of time. Protein levels were analyzed by Western blotting.
Western Blotting and Immunoprecipitation-Cells were washed twice with PBS and lysed in Pierce M-PER lysis buffer supplemented with protease inhibitor mixture (Roche Applied Science). The protein samples were boiled in SDS sample buffer, resolved on 12% SDS-polyacrylamide gel, and transferred to nitrocellulose membrane (Bio-Rad). The membranes were blocked with 10% nonfat milk in 20 mM Tris-HCl, pH 7.6, 150 mM NaCl, and 0.1% Tween 20 for 1 h at room temperature. After blocking, membranes were incubated with antibodies as indicated for 3 h at room temperature or overnight at 4°C, followed by incubation with horseradish peroxidase-conjugated secondary antibodies (Amersham Biosciences) or horseradish peroxidase-conjugated protein G (Bio-Rad) for 1 h at room temperature. Antibody detections were performed with a Pierce ECL detection kit or Amersham Biosciences ECL advanced Western blotting system according to the manufacturer's instructions.
For FLAG-tagged protein immunoprecipitation, FLAG M2 affinity gels were added to the cell lysates and incubated overnight at 4°C. For HA-tagged ARF or endogenous COMMD1, cell lysates were first incubated with anti-HA or anti-COMMD1 polyclonal antibody for 6 h, followed by protein G-Sepharose bead incubation for another 3 h at 4°C. The beads were then washed five times with wash buffer (20 mM Tris-HCl, pH 7.6, and 500 mM NaCl), boiled in SDS sample buffer, and subjected to Western blot analysis.
Indirect Immunofluorescence Staining and Microscopy-H1299 cells grown on a glass coverslip were transfected with plasmids as indicated. 1 day after transfection, cells were washed twice with PBS, fixed with 4% paraformaldehyde in PBS for 10 min, permeabilized with 0.1% Nonidet P-40 in PBS for 10 min, and blocked with 4% bovine serum albumin for 30 min. Fixed cells were first incubated with anti-HA antibody for 2 h, followed by incubation with fluorescein isothiocyanate-conjugated secondary antibody for another 1 h, incubation with anti-FLAG M2 antibody for 2 h, and incubation with rhodamineconjugated secondary antibody for another 1 h. To detect localization of endogenous proteins, H1299 cells were treated with specific drugs or left untreated for the indicated period of time. After stimulation, cells were fixed and stained first with anti-ARF Ab for detection of ARF and then with anti-COMMD1 Ab for detection of COMMD1. Fluorescence signals were visualized using a Carl Zeiss LSM 510 confocal microscope.

RESULTS
ARF Interacts with COMMD1 in Vivo-We performed a yeast two-hybrid screening to identify novel ARF-interacting proteins. Full-length p14ARF fused with the Gal4 DNA-binding domain was used to screen a human skeletal muscle library fused with the DNA activation domain. This screen led to identification of a protein named COMMD1 as an ARF-binding partner candidate. The interaction between ARF and COMMD1 in the mammalian cells was verified by co-immunoprecipitation (co-IP) assay. As shown in Fig. 1A (panel a), HA-ARF could be detected in the FLAG-COMMD1 immunoprecipitate but not in FLAG mock. In a reciprocal co-IP experiment when HA-ARF was precipitated from cell lysates (Fig. 1A, panel b), FLAG-COMMD1 could be pulled down together only with HA-ARF but not with HA, demonstrating that ARF interacts with COMMD1 specifically in mammalian cells.
To delineate the region of ARF involved in the interaction with COMMD1, full-length ARF and various deletion mutants were constructed (Fig. 1B). The N-terminal domain (amino acids 1-64) of ARF encoded by exon 1␤ was found to be sufficient for the interaction with COMMD1 (Fig. 1C, lane 3), whereas the exon 2-encoded C-terminal domain (amino acids 65-132) did not show any binding to COMMD1 (lane 4). To better define the interacting domain, ARF mutants with more discrete deletions in the N terminus were generated. Deletion of the first 14 residues of ARF (ARF⌬N14) had no discernible effect on the interaction with COMMD1, whereas deletion of residues 1-29 (ARF⌬N29) still resulted in binding to COMMD1. Deletion of residues 1-45 of ARF (ARF⌬45) totally abolished the interaction with COMMD1, suggesting that the binding site may lie between amino acids 30 and 45. However, the ARF⌬N30 -45 mutant still retained some capacity to bind COMMD1, whereas deletion of residues 15-45 completely abrogated the associ-ation. These results suggest that amino acids 15-45 contribute to the interaction of ARF with COMMD1 ( Fig. 1C).
COMMD1 Colocalizes with ARF in the Nucleoplasm-ARF normally localizes in the nucleolus in cancer cell lines and colocalizes with its associated proteins such as MDM2, E2F1, and HIF-1␣ in the nucleolus (5,13,22,23). Therefore, we examined the localization of COMMD1 upon simultaneous expression with ARF. pCMV5-HA-ARF and pcDNA3-FLAG-COMMD1 or the corresponding empty vector were transfected into H1299 cells, followed by immunostaining using anti-FLAG antibody for FLAG-COMMD1 and anti-HA antibody for HA-ARF. As shown in Fig. 2A, HA-ARF localized in the nucleolus when it was coexpressed together with the empty vector. In contrast, FLAG-COMMD1 exhibited a mainly cytoplasmic localization pattern with accumulation of the protein in some discrete perinuclear regions. FLAG-COMMD1 could also be detected in the nucleus, but it was excluded from the nucleolus. However, when FLAG-COMMD1 was coexpressed along with HA-ARF, FLAG-COMMD1 was found in the nucleus and colocalized with HA-ARF in the nucleoplasm. Previous studies have shown that although ARF is capable of sequestering its binding partners into the nucleolus, the nucleolar localization signal of the binding partners may also be required for the nucleolus colocalization (23,24). We found that FLAG-COMMD1 did not localize to the nucleolus when it was coexpressed with HA-ARF; instead, they both localized in the nucleoplasm. It could be due to the lack of the nucleolar localization signal in COMMD1. Nevertheless, this result indicates that formation of the ARF-COMMD1 complex alone is not sufficient for ARF to sequester protein to the nucleolus.
Previously it was reported that the DNA damage reagent actinomycin D can induce nucleolus disruption and rapid redistribution of endogenous ARF into the nucleoplasm (25). It is possible that ARF would colocalize and interact with FIGURE 1. ARF interacts with COMMD1 in vivo. A, reciprocal co-IP assay. H1299 cells were cotransfected with FLAG-COMMD1 and HA-ARF. 24 h after transfection, cells were lysed in Pierce M-PER lysis buffer. Immunoprecipitation was performed using mouse anti-FLAG M2 or rabbit anti-HA antibody as indicated, and then cells were immunoblotted with rabbit anti-FLAG or anti-HA antibody. To test the input of different proteins, 2% cell lysates were loaded and probed with antibodies for specific proteins. WB, Western blotting. B, schematic representation of ARF and ARF truncated mutants used in the deletion mapping experiment. C, deletion mapping. H1299 cells were cotransfected with FLAG-COMMD1 and HA-ARF (full-length or various truncated mutants as indicated). Cell lysates were immunoprecipitated with anti-FLAG M2 affinity gel. The pulled down materials were separated by SDS-PAGE and immunoblotted with anti-HA or anti-FLAG antibody. 2% cell lysates were loaded and probed with specific antibodies for detection of the inputs of different proteins. *, IgG light chain.
COMMD1 in the nucleoplasm after DNA damage stimulation. To test this idea, immunostaining and co-IP assay were performed after treatment of H1299 cells with actinomycin D for 8 h. As shown in Fig. 2B (lower panelS), actinomycin D induced a complete redistribution of ARF in the nucleus and subsequently led to more ARF colocalizing with COMMD1 in the nucleoplasm. H1299 cells are deficient of p53, so the redistribution of ARF in response to actinomycin D treatment is independent of p53. Consistent with this observation, the complex of endogenous ARF-COMMD1 could be detected only in H1299 cells with actinomycin D stimulation but not in untreated cells (Fig. 2C). Thus, the interaction of ARF and COMMD1 under physiological conditions appears to be dependent, at least partially, on their colocalization in the nucleoplasm.
ARF Stabilizes COMMD1 by Regulating Its Turnover-We noticed that the level of FLAG-COMMD1 appeared to be higher in cells transfected with HA-ARF than in cells transfected with the empty vector (Fig. 3A). The accumulation of FLAG-COMMD1 was observed in both A549 (p53 ϩ/ϩ ) and H1299 (p53 Ϫ/Ϫ ) cells, indicating that it is p53-and MDM2independent. Conversely, in Fig. 3B, knockdown of endogenous ARF by RNAi led to a decrease in COMMD1 protein levels. We Scale bar ϭ 10 m. C, H1299 cells were treated with actinomycin D or left untreated for 8 h. Cells were then harvested, and COMMD1 was immunoprecipitated with mouse anti-COMMD1-conjugated protein G-agarose for 6 h. Endogenous ARF coprecipitated with COMMD1 was detected by rabbit anti-ARF antibody. Immunoprecipitated COMMD1 was analyzed by Western blotting (WB) with mouse anti-COMMD1 antibody, followed by horseradish peroxidase-conjugated protein G to avoid detection of IgG light and heavy chains. next examined whether elevation of FLAG-COMMD1 by ARF occurs at the transcriptional or post-translational level. Cycloheximide was used to block protein synthesis in cells. H1299 cells transfected with HA-ARF or the empty vector were incubated with cycloheximide and harvested at the time indicated (Fig. 3C). The cells were lysed, and the cells lysates were then analyzed by Western blotting. As shown in Fig. 3B, in the presence of cycloheximide, FLAG-COMMD1 appeared to be quite unstable, with a half-life of ϳ1 h as described previously (19). However, in cells ectopically expressing HA-ARF, FLAG-COMMD1 decayed at much slower rate, with a half-life of 4 h. This result suggests that COMMD1 is stabilized in the presence of ARF, which occurs at the post-translational state.
COMMD1 is a proteasome substrate, and ubiquitination is essential for its specific degradation (19). Therefore, we reasoned that ARF might affect COMMD1 expression by regulating the ubiquitination process. We tested whether up-regulation of COMMD1 by ARF is sensitive to the treatment with MG132, a potent proteasome inhibitor. FLAG-COMMD1 levels increased dramatically after a 2-h proteasome inhibition (Fig. 3D, second lane), indicating that degradation of COMMD1 is mediated by the 26 S proteasome complex. However, no obvious accumulation of FLAG-COMMD1 was detected after MG132 treatment in the presence of ectopically expressed HA-ARF, suggesting that ARF may stabilize COMMD1 by either down-regulating the level of polyubiquitinated COMMD1 that is targeted for proteasomal degradation or acting similarly as a proteasome inhibitor to enhance and sustain COMMD1.
ARF Promotes Proteasome-independent Polyubiquitination of COMMD1-Next we examined the effect of ARF expression on the level of ubiquitinated COMMD1 by performing in vivo ubiquitination assay. H1299 cells were transfected with Myc-ub along with FLAG-COMMD1 and HA-ARF or the corresponding empty vector. Transfected cells were treated with 2 M MG132 or buffer as a control for another 8 h. FLAG-COMMD1 was then immunoprecipitated from cell extracts, followed by Western blot analysis. Consistent with previous findings (19), COMMD1 was ubiquitinated in the presence of MG132 (Fig.  4A, upper right panel). Surprisingly, we found that in the sample coexpressing FLAG-COMMD1, Myc-ub, and HA-ARF (Fig.  4A, upper left panel) without MG132 treatment, the level of ub-conjugated COMMD1 was increased significantly, demonstrating that ARF promotes polyubiquitination of COMMD1. When MG132 was added to allow accumulation of polyubiquitinated COMMD1 that ultimately was degraded by the proteasome, a high molecular mass smear that is characteristic of polyubiquitination was detected in cells transfected with both the HA vector and HA-ARF, but transfection with HA-ARF led to greater recovery of ubiquitinated COMMD1 (Fig. 4A, upper  right panel). The ubiquitination of COMMD1 in the absence or presence of MG132 was also confirmed using anti-Myc antibody to probe the precipitated COMMD1-ub conjugates (Fig. 4A, middle panels). We noted that the amount of COMMD1-ub conjugates induced by ARF was not elevated after proteasome blockade. In many cases, the ubiquitinated form of proteasome substrate can be detected only in the presence of proteasome inhibitors. Therefore, the possible explanation for this result is that ARF-facilitated ubiquitination of COMMD1 is proteasome-independent, and it does not target the protein for degradation.
To further confirm that ARF-induced COMMD1 ubiquitination is proteasome-independent, the K48R point mutant of ubiquitin, which cannot form Lys 48 -conjugated polyubiquitin chains, was used for the in vivo ubiquitination assay. As shown in Fig. 4B (left panels), the ubK48R mutant did not inhibit ARFinduced COMMD1 ubiquitination. Interestingly, we found that overexpression of HA-ARF still enhanced the ubiquitination of COMMD1 in the presence of the Myc-ubK48R mutant, demonstrating that ARF induces proteasome-independent non-Lys 48 ubiquitin chain conjugation to COMMD1. In addition to Lys 48 -mediated ubiquitination, ubiquitination could also occur at Lys 63 of ubiquitin. It has been reported that Lys 63linked polyubiquitination does not involve proteolysis. Therefore we sought to investigate whether the apparent stability of ubiquitinated COMMD1 induced by ARF is due to the assembly of the polyubiquitin chain involving Lys 63 but not Lys 48 of ubiquitin. As shown in Fig. 4B (right panels), overexpression of HA-ARF did not enhance polyubiquitination of FLAG-COMMD1 as compared with control transfection in the presence of Myc-ubK63R. This result suggests that Lys 63 of ubiquitin is important for ARF-induced COMMD1 polyubiquitination. We found that ARF promotes non-Lys 48 ubiquitination of COMMD1, which is mediated through the Lys 63 -Gly 76 linkage.
Previously it was reported that the RNAi of XIAP, an E3 ligase that ubiquitinates and degrades COMMD1 through Lys 48 linkage, could enhance non-Lys 48 ubiquitination of COMMD1 (19). This result suggests that XIAP-mediated Lys 48 ubiquitin chains may be assembled at the same lysine residues as those when ubiquitination occurs via non-Lys 48 linkage. Hence, reduction in Lys 48 ubiquitination could result in greater ubiquitination through non-Lys 48 linkage. In our study, we have shown that ARF-facilitated Lys 63 ubiquitination stabilizes COMMD1; COMMD1 is destabilized in cells deficient in ARF. It is possible that the Lys 63 ubiquitination induced by ARF may compete with Lys 48 ubiquitination and thus protect COMMD1 from proteasomal degradation. Therefore, we sought to determine COMMD1 ubiquitination pattern in cells depleted of ARF. As shown in Fig. 4C (left panels), knockdown of endogenous ARF by RNAi led to an obvious reduction in the total amount of COMMD1 ubiquitination, demonstrating that ARFmediated ubiquitination constitutes a large part of COMMD1 ubiquitination. A decrease of ubiquitinated COMMD1 was also observed in cells cotransfected with ubK48R but not in cells cotransfected with the K63R mutant (Fig. 4C, right panels), indicating that Lys 63 -linked ubiquitination is mediated at least partially by endogenous ARF. We noticed that in contrast to the effect of XIAP RNAi, knockdown of ARF did not result in greater Lys 48 ubiquitination of COMMD1. Consistent with this, the destabilization of COMMD1 caused by ARF RNAi was also not sensitive to the proteasome blockade (Fig. 4D). Together, these results suggest that ARF-induced Lys 63 ubiquitin chains are assembled at some specific lysine residues that are not responsible for Lys 48 ubiquitination; hence, a decrease in Lys 63 ubiquitination would not enhance Lys 48 -mediated ubiquitination and degradation. The possible explanation for the destabilization of COMMD1 in ARF-deficient cells is that ARF-induced Lys 63 ubiquitination could protect COMMD1 from degradation in some proteasome-independent manners; thus, COMMD1 would be degraded as a result of the loss of protection from Lys 63 ubiquitin chains.
Meanwhile, the above data also imply that the interaction of ARF and COMMD1 may already exist in the normal state, and thus, endogenous ARF could induce polyubiquitination of COMMD1 without any stimulation. Recently, it was found that the non-nucleolar form of the ARF mutant is inherently unstable compared with nucleolar ARF (26). Like other proteins, ARF also undergoes proteasomal proteolysis (9). We reasoned that proteasome inhibition would result in recovery of the ARF-COMMD1 complex in the nucleoplasm in the normal state. As shown in supplemental Fig. 1A, treatment with MG132 led to great recovery of ARF in the nucleoplasm, which was not observed in non-stimulated cells. This indicates that nucleoplasmic ARF is less stable than nucleolar ARF. Similarly, in the co-IP experiment (supplemental Fig. 1B), no endogenous ARF could be coprecipitated with FLAG-COMMD1 in the absence of proteasome inhibition. However, after MG132 treatment, ARF could be readily detected in the FLAG-COMMD1 precipitates. Taken together, these data demonstrate that ARF specifically interacts with COMMD1 in vivo but is subjected to rapid degradation; endogenous ARF could induce Lys 63 polyubiquitination of COMMD1 in the normal state.
Finally, as ARF associates with COMMD1 in the nucleus, we determined whether ARF specifically induces ubiquitination of COMMD1 in the nucleus. As shown in Fig. 4E, compared with the cytoplasmic fraction, the nuclear COMMD1 protein level was dramatically down-regulated in ARF-depleted cells. Consistent with this, the decreased COMMD1 polyubiquiti- anti-FLAG affinity gel and then immunoblotted with anti-FLAG or anti-HA antibody. C, H1299 cells were transfected with ARF, control small interfering RNA (siRNA), or FLAG-COMMD1 along with wild-type ub (wt-ub), ubK48R, or ubK63R for 48 h. Cells cotransfected with wild-type ub were treated with MG132 for another 8 h. Cells were then lysed and immunoprecipitated with mouse anti-FLAG M2 affinity gel. The immunoprecipitates were immunoblotted with rabbit anti-FLAG or anti-Myc antibody. *, IgG light chain. D, H1299 cells were transfected with ARF or control small interfering RNA together with FLAG-COMMD1 for 2 days. Cells were then split into two parts. Each part was treated with MG132 or left untreated for 8 h. COMMD1 protein levels were detected by Western blotting with anti-FLAG antibody. Endogenous actin levels were used as loading controls. The relative densities of COMMD1/actin were analyzed by Photoshop software. E, H1299 cells were transfected with ARF or control small interfering RNA and FLAG-COMMD1 for 2 days and treated with MG132 for an additional 8 h. The cytoplasmic and nuclear fractions were separated using a Pierce cytoplasmic and nuclear extraction kit, followed by immunoprecipitation with anti-FLAG M2 antibody. Immunoprecipitated COMMD1 in each fraction was analyzed by anti-FLAG antibody. ␣-Tubulin and GCN5 protein levels were used as loading controls for the cytoplasmic and nuclear fractions, respectively. nation in the nuclear fraction was also observed in cells depleted of ARF. Hence, these results suggest that ARF specifically induces COMMD1 polyubiquitination in the nucleus.
ARF Mutants Are Defective in Promoting COMMD1 Polyubiquitination-To determine whether the physical interaction of ARF with COMMD1 is a prerequisite condition for COMMD1 ubiquitination, we tested the effect of two ARF deletion mutants: ARF⌬N15-45, which is defective in binding COMMD1, and ARF-(1-64), which binds to COMMD1 efficiently but lacks the C-terminal domain encoded by exon 2. As shown in Fig. 5, full-length ARF promoted COMMD1 ubiquitination as observed previously, whereas expression of ARF⌬N15-45 did not result in any ubiquitination of COMMD1, indicating that ARF-facilitated ubiquitin conjugation of COMMD1 depends on their physical association. Previous study has established that the functional domain of ARF is encoded by exon 1␤ (amino acids 1-64), and ARF interacts with most of its binding partners through the N-terminal region (6,27,28). In contrast, the C terminus contributes little to the function of ARF except nucleolus localization (24). However, our observation showed that although the ARF-(1-64) mutant strongly associated with COMMD1, it could not promote ubiquitination of COMMD1. We believe that deletion of the C terminus impairs the association of ARF with some other unknown factor that also participates in the ubiquitination process.

DISCUSSION
It was reported that overexpression of ARF arrests proliferation of p53-null or p53/Mdm2-null cells (29). Mice deficient in Arf, p53, and Mdm2 develop tumors at a greater frequency compared with mice lacking both p53 and Mdm2 or p53 alone (8). These observations suggest that ARF also possesses tumor suppressor functions that are independent of p53. In this study, we have characterized a novel p53-independent interaction between the tumor suppressor ARF and COMMD1. The physical association of the two proteins has been confirmed by co-IP assay and found to depend on their colocalization in the nucleoplasm. ARF normally localizes in the nucleolus, and it was proposed that physical sequestration of MDM2 by ARF can stabilize p53 by separating it from MDM2-mediated degradation (5). However, recent data demonstrated that nucleolar relocalization of MDM2 is not required for p53 activation and that redistribution of ARF into the nucleoplasm induced by DNA damage enhances its association with MDM2-and p53dependent growth-suppressive activity (5,6). Here, our results showed that ectopically expressed ARF relocalizes COMMD1 into the nucleoplasm without nucleolar sequestration. In nonstressed cells, endogenous ARF and COMMD1 have differential subcellular localization. However, the association of these two proteins could be observed when ARF was redistributed into the nucleoplasm in response to actinomycin D treatment. These observations suggest that under certain circumstances such as DNA damage stimulation, ARF would be released from the nucleolus into the nucleoplasm, and therefore, it can perform its function by interacting with its binding partners such as MDM2 and COMMD1.
Overexpression of ARF was reported to induce degradation of certain E2F family members through a ubiquitin-proteasome pathway. It was suggested that ARF may act conversely to trim down the level of E2F and thus protect cells from tumorigenesis upon oncogene activation (13). Previous studies have shown that ARF promotes proteasomal degradation of nucleolar protein B23 and therefore inhibits B23-mediated ribosome biogenesis (10). However, our results showed that instead of facilitating ubiquitination, which leads to degradation of its binding partners, ARF promotes Lys 63 -linked polyubiquitination of COMMD1. Lys 63 ubiquitination is known as non-classic ubiquitination, as it does not target proteins for degradation. Instead, it has distinct roles in regulating cellular functions such as protein kinase activation, DNA repair, and vesicle trafficking (17,30,31). We found that ARF stabilizes COMMD1 through its ability to promote Lys 63 -mediated ubiquitination of COMMD1. Because non-proteasomal proteolysis is also utilized for COMMD1 degradation, it is possible that Lys 63 ubiquitin chains may be capable of regulating and stabilizing the structure of the protein or protecting it from being recognized by some unknown protease. Lys 63 -linked ubiquitination could also regulate protein-protein interactions (31,32), which may have additional functions, such as affecting the binding affinity of COMMD1 for other factors. COMMD1 has been demonstrated as a ubiquitously expressed inhibitor for NF-B (20,33). It was reported that COMMD1 interacts with and accelerates degradation of RelA in the nucleus, which terminates the association of RelA with chromatins (33). We found that ARF specifically promotes nuclear COMMD1 ubiquitination through Lys 63 linkage. Therefore, it is possible that ARF-induced Lys 63 polyubiquitination may regulate the inhibitory effect of COMMD1 on NF-B-mediated transcription in the nucleus.
Does ARF act as an E3 ligase for the ubiquitination process? Previous studies demonstrated that overexpression of ARF induces sumoylation of its binding partners such as MDM2, E2F1, B23, and Werner helicase (11,15,16). ARF also binds to UBC9, the only E2 found to be responsible for sumoylation (16), which suggests that ARF may serve as an E3 ligase to bridge the E2 complex to its binding partners. In addition, overexpression of ARF promotes ubiquitination of its binding partners such as E2F and B23. Although ARF does not possess the typical E3 ligase structural feature or known catalytic activity for ubiquitination, it is possible that ARF mediates ubiquitination of these proteins indirectly by bringing different proteins such as E3 ligase together. We found that physical association of ARF and COMMD1 is necessary for the ubiquitination; therefore, we do not exclude the possibilities that the C terminus of ARF interacts with other factors, such as E3 ligase for COMMD1 ubiquitination, or that ARF itself serves as an E3 ligase for COMMD1 ubiquitination.