Antizyme Targets Cyclin D1 for Degradation: A Novel Mechanism for Cell Growth Repression

cell In We Our results document a new pathway for cyclin D1 degradation, which differs from previously described cyclin degradation pathways in being independent of ubiquitination as demonstrated by cyclin D1 degradation in vitro in the presence of only antizyme and purified proteasomes. In addition, overexpression of antizyme mediates degradation in vivo of a cyclin D1 mutant that is not susceptible to ubiquitin-mediated degradation. In the alternative cyclin degradation pathway we have described, antizyme replaces ubiquitin as the activity responsible for delivering the protein substrate to the proteasome for degradation. We suggest that antizyme has the ability to form complexes with cyclin D1 and that binding of antizyme to this target protein does not require additional factors. This is consistent with the demonstrated mechanism for antizyme-mediated degradation of ornithine decarboxylase (1,2,26). Evidence in support of this model has been obtained in hepatoma cells, malignant oral keratinocytes, in reticulocyte extracts, and in purified systems. In vivo , the ability of antizyme to induce degradation of cyclin D1 is not contingent on its artificial overexpression, but is observed when endogenous antizyme is induced through the physiological mode of increased polyamine levels. There is some variability of response when a large population of cells is tested and we believe that this is due, in part, to differing levels of the endogenous antizyme inhibitor (AZI) in these cells. Further studies will be required to determine whether other cellular properties also influence the ability of antizyme to facilitate cyclin D1 degradation. Additionally we show that even the low endogenous levels of antizyme present in the cells may also contribute to fluctuations in cyclin D1 levels.


Introduction
The regulatory protein antizyme (AZ) has been studied primarily in the context of its role in facilitating degradation of the enzyme ornithine decarboxylase (ODC), which catalyzes the rate-limiting step in polyamine synthesis (reviewed in [1][2][3]. AZ is therefore thought to be dedicated principally to the feedback regulation of polyamine levels. AZ synthesis is controlled via an unusual mechanism of translational frameshifting. The ribosomal frameshift required for the translation of full-length AZ is directly induced by polyamines when their levels rise (4,5). Polyamines can thus inhibit their own synthesis via AZmediated down-regulation of ODC. This mechanism serves to prevent extreme fluctuations in polyamine levels, which are thought to be toxic (6)(7)(8). Despite this homeostatic mechanism, the levels of polyamines and ODC vary markedly during the cell cycle, indicating that additional factors control polyamine levels and suggesting a role for this pathway in the regulation of cell proliferation (9).
Overproduction of AZ in a variety of cell types, including malignant oral keratinocytes, hepatoma cell lines and prostate cancer cells, coincides with growth inhibition (10)(11)(12) and cell cycle arrest in the G1 phase (13,14). Furthermore, overexpression of antizyme in both mouse skin cancer and gastric epithelia models has been shown to result in tumor suppression (15,16). These and other observations of antiproliferative effects of antizyme prompted us to test whether antizyme may have a specific role in cell cycle regulation, thereby accounting for its potential role as a tumor suppressor. 4 The cell cycle arrest previously seen in prostate carcinoma and malignant oral keratinocyte models and the growth inhibition seen in a variety of cell types upon antizyme overexpression could be explained if antizyme modulated intracellular levels of cell cycle regulatory proteins such as cyclins, as it does for ODC. We previously showed that treatment of cells with the polyamine spermine resulted in both antizyme upregulation and cell cycle arrest (14). In the current study, we have studied the ability of antizyme to complex with and degrade cyclin D1. We report that antizyme binds to cyclin D1 and facilitates its proteasomal degradation. Using purified components, we find that antizyme-mediated degradation of cyclin D1 proceeds in the absence of ubiquitin. As the degradation of this cyclin has been previously characterized and found to be dependent on the SCF family of ubiquitin-protein ligases (17,18), our data indicate that the same protein may be delivered to the proteasome via two distinct targeting mechanisms. More generally, our results suggest that antizyme may indeed act as a tumor suppressor by controlling progression through the G1 phase of the cell cycle by a novel mechanism.
5 AZ PCR product was also cloned into pCMV-Flag using the same restriction enzymes.

35S labeling and pulse-chase
For short metabolic labeling, AT2.1 cells were treated with spermine as described above. The lysates were then used for immunoprecipitation as described above using antibody against the FLAG epitope of AZ, cyclin D1, or β-actin. Proteins were separated by SDS-PAGE, visualized and quantitated using a phosphoimaging system. The signal intensity of each band was normalized against that of β-actin.

In vitro protein degradation in reticulocyte lysates
Human cyclin D1 and rat antizyme were synthesized in vitro using reticulocyte lysates as

In vitro degradation assay using purified proteasomes
Human 26S proteasomes were purified from red blood cells as described previously (20).

Upregulation of endogenous antizyme in prostate cancer cells results in degradation of cell cycle regulatory proteins
Addition of exogenous polyamines has been shown to upregulate endogenous antizyme levels in a variety of different cell lines. We previously showed that induction of antizyme within two hours after addition of the polyamine spermine is followed by G1 growth arrest in prostate carcinoma cell lines (14). As cyclins are key determinants of cell cycle transitions, we investigated whether growth arrest was due to a change in the levels of specific cyclins responsible for the G1-S transition. Cyclin levels were measured in cell lines grown in defined media in the presence or absence of 10 µM spermine. After 24 hours we observed a marked decrease in the levels of cyclin D1, in spermine-treated cells relative to untreated controls (Fig. 1a). Under the conditions of this experiment, antizyme is efficiently induced, as previously shown (14). As expected, ODC levels are reduced following antizyme induction (Fig. 1a). Levels of other cell cycle regulatory proteins, such as cdk2, and cyclin A, remained unchanged, as did the level of 13 actin (Fig. 1a). Transcriptional regulation could not account for the decrease in cyclin D1 levels in these cell lines, since treated and untreated cells contained equivalent levels of mRNA encoding this protein (Fig. 1b). Furthermore, loss of cyclin D1 could not be attributed to a block in translation as synthesis of cyclin D1 protein remained constant during spermine treatment in these cell lines (Fig. 1c).
A critical function of the cyclin D1/CDK complex is phosphorylation of the retinoblastoma protein (pRb), which is required for entry into the S phase of the cell cycle (21)(22)(23). After spermine-induced loss of cyclin D1, pRb was found in a hypophosphorylated form (Fig. 1d). By this functional criterion, it is likely that cyclin D1/CDK kinase complexes are effectively inhibited by spermine treatment. The loss of G1-specific cyclin D1/CDK complexes and the resulting decrease in Rb phosphorylation may account for the predominant G1 arrest seen in these cells after spermine treatment (14).

Antizyme is required for the destruction of cyclin D1
The loss of cyclin D1 seen in spermine-treated prostate carcinoma cells could be a consequence of spermine-induced cell cycle arrest rather than a consequence of antizyme upregulation. To determine if antizyme alone is sufficient to promote degradation of cyclin D1, we looked at the effect of blocking antizyme synthesis on the levels of cyclin keratinocytes transfected with AZ-specific siRNA (Fig. 2c).

Antizyme overexpression leads to loss of cyclin D1
Similar to the effects seen in prostate carcinoma cells, stimulation of antizyme expression in rat hepatoma cells has been shown to result in growth inhibition (12). Additionally, overexpression of antizyme in malignant oral keratinocytes and prostate carcinoma has previously been shown to lead to G1 growth arrest and a reversal of the malignant phenotype (13,14). We overexpressed antizyme in rat hepatoma cells (Fig. 3a) and malignant oral keratinocytes (Fig. 3b) to investigate whether the growth inhibition previously seen in these cell types could be correlated with a change in cyclin D1 levels.
Full-length antizyme is efficiently expressed in these transfectants. We observed a marked decrease in the levels of cyclin D1 in cell lines overexpressing AZ as compared to control cells. Levels of another cell cycle regulatory protein, cdk2, remained unchanged, as did the level of actin. These results show that overexpression of antizyme alone is sufficient to promote specific degradation of cyclin D1 in these cell types.
To further investigate the ability of AZ to degrade cyclin D1 and to dissociate this process from the well-studied-ubiquitin dependent degradation pathway, we employed a mutant form of cyclin D1 (T286A) that cannot be degraded via the ubiquitin pathway (24,25). As shown in figure 4, levels of this mutant are stable in control AT2.1 cells, but fell significantly following spermine-induced upregulation of antizyme. These results were subject to immunoprecipitation using antibodies against cyclin D1. Degradation of the T286A mutant was seen only in cells expressing AZ and the overexpressed protein was stable in control cells (Fig. 5a, b). Because control cells showed no alteration in the rate of degradation of the T286A cyclin D1 mutant, enhanced degradation of the protein seen in AZ overexpressing cells could be attributed directly to AZ. These results confirm the ability of AZ to mediate cyclin D1 degradation in the absence of ubiquitination. To investigate this relationship further, we employed a variety of in vitro assays to study whether AZ can bind directly to cyclin D1 and influence cyclin D1 stability.

Antizyme complexes with cyclin D1 in vitro
The principal known function of antizyme is to bind to ornithine decarboxylase (ODC) and increase the affinity of ODC for the proteasome, where it is degraded in an ubiquitinindependent manner. The finding that cyclin D1 degradation correlates with increased antizyme levels led us to test whether antizyme can directly target additional regulatory proteins for degradation. A prediction of this model is that antizyme should be capable of complex formation with proteins whose degradation it stimulates. Here we use three independent methods to show that antizyme is capable of direct binding to cyclin D1 both in vitro and in vivo.
First, the cyclin D1 and antizyme proteins were synthesized in 35  show evidence of an antizyme-cyclin D1 complex (Fig. 6a). Control experiments indicated that all antibodies used are non-crossreactive (Fig. 6b).
Secondly, we assessed complex formation between antizyme and cyclin D1 by coimmunoprecipitation of overexpressed proteins from mammalian cell lysates. Direct binding of antizyme to cyclin D1 was confirmed by the ability of antibodies against antizyme to isolate a complex of antizyme and cyclin D1 (Fig. 6c).
Finally, complex formation between antizyme and cyclin D1 was assayed using purified recombinant proteins. Direct binding of antizyme to purified cyclin D1 was confirmed using GSH beads to isolate proteins bound to a GST-antizyme fusion protein (Fig. 6d).
Specificity of this interaction is suggested by the fact that antizyme fails to bind negative control proteins such as citrate synthase (CS) (Fig. 6d). Furthermore, ODC effectively inhibits the interaction between cyclin D1 and antizyme (Fig. 6e), indicating that AZ may have a higher affinity for ODC as compared to cyclin D1.

Antizyme promotes proteasomal degradation of cyclin D1 in vitro
Our initial experiments indicated that cells containing high levels of antizyme have reduced levels of cyclin D1. We therefore examined the effect of antizyme on cyclin D1 protein stability in an in vitro degradation assay using rabbit reticulocyte lysate as a source of proteasomes. Antizyme and cyclin D1 were individually synthesized and 35 Slabeled in reticulocyte lysates, then incubated in the presence of an ATP-regenerating system for 1.5 hours at 30°C. Both proteins were stable under these conditions (Fig. 7a).

19
In contrast, when antizyme-containing lysate was mixed with cyclin D1-containing lysate, efficient degradation of cyclin D1 was observed. These data thus indicate that the ability of antizyme to facilitate protein degradation is not limited to ODC and SMAD1 but may be a more general mechanism for ubiquitin-independent degradation. The application of this mechanism to cell cycle regulatory proteins suggests a novel means of modulating the cell cycle through antizyme-mediated protein degradation.
It is unclear from the above studies whether antizyme binding alone is sufficient to target proteasomal degradation of cyclin D1. In particular, cyclin D1 can be modified by ubiquitination, thus raising a question as to whether the antizyme-dependent pathway of cyclin D1 turnover is independent of the ubiquitin-dependent pathway. To confirm that antizyme alone can promote degradation of cyclin D1 in an ubiquitin independent manner, we reconstituted cyclin D1 degradation in vitro using purified components.
Purified human proteasomes alone do not degrade cyclin D1 (Fig. 7c). However, cyclin D1 was unstable in the presence of both antizyme and proteasomes (Fig. 7b). When proteasomes were omitted from the reaction, cyclin D1 was stabilized (Fig. 7d). In support of the conclusion that antizyme targets cyclin D1 for degradation by the proteasome, the reaction was sensitive to both proteasome inhibitors and ATP-depletion ( Fig. 7e, f). It should be noted that the degradation experiments were carried out at 30˚C rather than 37˚C, which may account for the slower progression of cyclin D1 degradation than has been previously reported. These characteristics of antizyme-dependent cyclin D1 degradation closely mimic those of ODC degradation (26). Given that purified by guest on March 24, 2020 http://www.jbc.org/ Downloaded from 20 proteins were used in these experiments, no factors other than antizyme, ATP and the proteasome appear to be strictly required for cyclin D1 degradation by this pathway.

Discussion
Studies from several laboratories have shown that perturbation of the levels of antizyme, ODC, or polyamines can affect cell cycle progression (9,14,27,28). However, the mechanisms by which antizyme may influence the cell cycle have remained unclear. It has often been assumed that the effects of antizyme are mediated via enhanced ODC degradation, and consequently, suppression of polyamine levels. Our data suggest that the effects of antizyme can also be mediated through a pathway that involves a direct interaction between antizyme and cell cycle regulators. This is consistent with the report that ectopic overexpression of antizyme in malignant oral keratinocytes results in an increased doubling time coupled with G1 arrest (13). Murakami and colleagues have also shown that AZ overexpression inhibits growth of HTC cells (29). Addition of putrescine does not completely reverse that growth inhibition as it does for cells treated with the ODC inhibitor DFMO (10). Similar results were obtained with S. pombe antizyme overexpression (30). The growth phenotype is apparently not due simply to reduced polyamine levels, because it is only partially relieved by the addition of polyamines to the growth medium (30). These data imply that direct regulation of the cell cycle by antizyme is an evolutionarily conserved biological mechanism.
Our results suggest the possibility that altered antizyme regulation could lead to improperly regulated cell growth in human malignancy or other diseases characterized by by guest on March 24, 2020 http://www.jbc.org/ Downloaded from 21 abnormal cell proliferation. This view is consistent with our previous finding that spermine negatively regulates prostate cancer cell growth (31) and that sperminemediated growth arrest of prostate cancer cells is apparently dependent on the ability of those cells to upregulate antizyme after spermine exposure (14). Our results may provide an explanation for the previous findings that overexpression of antizyme in rastransformed NIH3T3 cells as well as in malignant oral keratinocytes inhibits cell growth and tumor formation both in vitro and in vivo (11,13). Additionally, overexpression of antizyme in a mouse skin cancer and gastric epithelia models was shown to inhibit tumor formation (16). These findings suggest that antizyme may have a role in retarding tumor development.
Whether antizyme can target the degradation of G1 cell cycle regulators in other cell types and cancer models besides hepatoma cells, malignant oral keratinocytes and in prostate carcinoma cells remains to be determined. However, the required components of this pathway are ubiquitously expressed. Additionally, loss of antizyme expression has been reported in the progression of prostatic carcinoma (14), as well as for progression of keratinocyte-derived oral tumors (32). Considerable work has been done to develop polyamine analogs that suppress cell growth and induce apoptosis (33). Consistent with the work of Mitchell et al (12), our work suggests that antizyme may be a principal target through which such drugs exert their antiproliferative effects and provides a potential mechanism for their action.

22
Our results document a new pathway for cyclin D1 degradation, which differs from previously described cyclin degradation pathways in being independent of ubiquitination as demonstrated by cyclin D1 degradation in vitro in the presence of only antizyme and purified proteasomes. In addition, overexpression of antizyme mediates degradation in vivo of a cyclin D1 mutant that is not susceptible to ubiquitin-mediated degradation. In the alternative cyclin degradation pathway we have described, antizyme replaces ubiquitin as the activity responsible for delivering the protein substrate to the proteasome for degradation. We suggest that antizyme has the ability to form complexes with cyclin D1 and that binding of antizyme to this target protein does not require additional factors.
This is consistent with the demonstrated mechanism for antizyme-mediated degradation of ornithine decarboxylase (1,2,26). Evidence in support of this model has been obtained in hepatoma cells, malignant oral keratinocytes, in reticulocyte extracts, and in purified systems. In vivo, the ability of antizyme to induce degradation of cyclin D1 is not contingent on its artificial overexpression, but is observed when endogenous antizyme is induced through the physiological mode of increased polyamine levels. There is some variability of response when a large population of cells is tested and we believe that this is due, in part, to differing levels of the endogenous antizyme inhibitor (AZI) in these cells. Further studies will be required to determine whether other cellular properties also influence the ability of antizyme to facilitate cyclin D1 degradation. Additionally we show that even the low endogenous levels of antizyme present in the cells may also contribute to fluctuations in cyclin D1 levels.
During the cell cycle, cyclin D1 is phosphorylated at Thr-286 by the glycogen synthase kinase 3ß (GSK-3 ß) and subsequently polyubiquitinated and degraded by the 26S proteasome. Our data shows that this phosphorylation-ubiquitination pathway is dispensable for AZ-mediated proteasomal degradation. However, in addition to promoting the ubiquitination and degradation, phosphorylation of cyclin D1 at Thr-286 also mediates nuclear export of cyclin D1 to the cytoplasm. This redistribution of cyclin D1 is dependent on the nuclear exportin CRM1 (34). The non-phosphorylatable D1-T286A mutant has been shown to remain in the nucleus throughout the cell cycle (34).
Therefore, in order for antizyme to promote degradation of D1-T286A it may be required to enter the nucleus. Intriguingly, localization of both antizyme and ODC to the nucleus has been demonstrated during mouse development (35), and nuclear localization signals have been identified in the antizyme protein (36). Like cyclin D1, AZ has been shown shuttle between the nucleus and cytoplasm in a CRM-1 dependent manner (36).
Furthermore, antizyme is able to enter the nucleus in response to bone morphogenic protein type-1 receptor activation by forming a complex with the Smad1 transcription factor and the proteasome subunit HsN3. It has been suggested that AZ could bind to ODC and other proteins in the nucleus and escort them to the cytoplasm for degradation (36). Taken together these finding strengthen a role for AZ in regulation of cyclin D1 levels in the absence of ubiquitin.
The implication of our results is that alterations in antizyme levels may provide one form of cell cycle regulation that operates by influencing intracellular levels of cyclin D1.
Recent evidence that antizyme may also interact with SMAD1 indicates that ubiquitin-by guest on March 24, 2020 http://www.jbc.org/ Downloaded from 24 independent regulation of protein turnover may be a more prevalent mechanism of protein degradation than previously thought (37). It is also worth noting that an endogenous inhibitor of antizyme (antizyme inhibitor, AZI) has been identified (38)(39)(40)(41) and that intracellular levels of AZI will likely influence the ability of antizyme to regulate protein degradation. The ability of antizyme to suppress cyclin D1 levels provides one explanation for the previous observations that antizyme upregulation results in cell cycle arrest in a variety of cell types. Because AZ levels are low in certain late stage cancers, it is conceivable that antizyme may act as a tumor suppressor by helping to finely regulate cyclin levels in vivo.