Post-translational Down-regulation of Melanoma Antigen-A11 (MAGE-A11) by Human p14-ARF Tumor Suppressor*

Background: Melanoma antigen-A11 (MAGE-A11) is a primate-specific steroid receptor coregulator and proto-oncogene expressed at increased levels in castration-resistant prostate cancer. Results: Human p14-ARF tumor suppressor promotes the proteasomal degradation of MAGE-A11 independent of ubiquitination. Conclusion: MAGE-A11 is post-translationally down-regulated by the p14-ARF tumor suppressor. Significance: Increased levels of MAGE-A11 associated with low p14-ARF promote the development of castration-resistant prostate cancer. X-linked primate-specific melanoma antigen-A11 (MAGE-A11) is a human androgen receptor (AR) coactivator and proto-oncogene expressed at low levels in normal human reproductive tract tissues and at higher levels in castration-resistant prostate cancer where it is required for androgen-dependent cell growth. In this report, we show that MAGE-A11 is targeted for degradation by human p14-ARF, a tumor suppressor expressed from an alternative reading frame of the p16 cyclin-dependent kinase inhibitor INK4a/ARF gene. MAGE-A11 degradation by the proteasome was mediated by an interaction with p14-ARF and was independent of lysine ubiquitination. A dose-dependent inverse relationship between MAGE-A11 and p14-ARF correlated with p14-ARF inhibition of the MAGE-A11-induced increase in androgen-dependent AR transcriptional activity and constitutive activity of a splice variant-like AR. Reciprocal stabilization between MAGE-A11 and AR did not protect against degradation promoted by p14-ARF. p14-ARF prevented MAGE-A11 interaction with the E2F1 oncoprotein and inhibited the MAGE-A11-induced increase in E2F1 transcriptional activity. Post-translational down-regulation of MAGE-A11 promoted by p14-ARF was independent of HDM2, the human homologue of mouse double minute 2, an E3 ubiquitin ligase inhibited by p14-ARF. However, MAGE-A11 had a stabilizing effect on HDM2 in the absence or presence of p14-ARF and cooperated with HDM2 to increase E2F1 transcriptional activity in the absence of p14-ARF. We conclude that degradation of MAGE-A11 promoted by the human p14-ARF tumor suppressor contributes to low levels of MAGE-A11 in nontransformed cells and that higher levels of MAGE-A11 associated with low p14-ARF increase AR and E2F1 transcriptional activity and promote the development of castration-resistant prostate cancer.


X-linked primate-specific melanoma antigen-A11 (MAGE-A11) is a human androgen receptor (AR) coactivator and protooncogene expressed at low levels in normal human reproductive tract tissues and at higher levels in castration-resistant prostate cancer where it is required for androgen-dependent cell growth.
In this report, we show that MAGE-A11 is targeted for degradation by human p14-ARF, a tumor suppressor expressed from an alternative reading frame of the p16 cyclin-dependent kinase inhibitor INK4a/ARF gene. MAGE-A11 degradation by the proteasome was mediated by an interaction with p14-ARF and was independent of lysine ubiquitination. A dose-dependent inverse relationship between MAGE-A11 and p14-ARF correlated with p14-ARF inhibition of the MAGE-A11-induced increase in androgen-dependent AR transcriptional activity and constitutive activity of a splice variant-like AR. Reciprocal stabilization between MAGE-A11 and AR did not protect against degradation promoted by p14-ARF. p14-ARF prevented MAGE-A11 interaction with the E2F1 oncoprotein and inhibited the MAGE-A11-induced increase in E2F1 transcriptional activity. Post-translational down-regulation of MAGE-A11 promoted by p14-ARF was independent of HDM2, the human homologue of mouse double minute 2, an E3 ubiquitin ligase inhibited by p14-ARF. However, MAGE-A11 had a stabilizing effect on HDM2 in the absence or presence of p14-ARF and cooperated with HDM2 to increase E2F1 transcriptional activity in the absence of p14-ARF. We conclude that degradation of MAGE-A11 promoted by the human p14-ARF tumor suppressor contributes to low levels of MAGE-A11 in nontransformed cells and that higher levels of MAGE-A11 associated with low p14-ARF increase AR and E2F1 transcriptional activity and promote the development of castration-resistant prostate cancer.
The MAGE-A11 gene at the Xq28 locus of the MAGE gene family on the human X chromosome evolved within the primate lineage by gene duplication and retrotransposition (12,13). The functional dependence on MAGE-A11 for increased human AR transcriptional activity is supported by the coevolution of X-linked human MAGE-A11 and X-linked human AR NH 2 -terminal sequence flanking the FXXLF motif that mediates AR interaction with MAGE-A11 and the androgen-dependent AR NH 2 -and COOH-terminal interaction (14,15). Rat and mouse AR Ala 33 evolved to Val 33 in human AR NH 2 -terminal sequence 23 FQNLFQSVREV 33 , a sequence change required for human AR to interact with MAGE-A11 (15). However, neither mouse nor rat AR Ala 33 or human AR Val 33 is required for the FXXLF motif-mediated androgen-dependent AR NH 2 -and COOH-terminal interaction. These findings suggest that the increase in human AR transcriptional activity resulting from the coactivator activity of MAGE-A11 evolved more recently among primates than evolution of the androgendependent AR NH 2 -and COOH-terminal interaction.
In addition to AR coregulator activity, MAGE-A11 has functions that directly impact the cell cycle and require complex regulatory control. MAGE-A11 interacts with and increases the transcriptional activity of E2F1, one of the most important transcription factors and oncoproteins that promotes unscheduled entry into S phase of the cell cycle (5). The increase in E2F1 activity by MAGE-A11 is similar to the effects of E1A, an adenovirus early protein that transforms cells (16). The ability of MAGE-A11 to increase AR and E2F1 transcriptional activity suggests that MAGE-A11 links AR to cell cycle progression and supports the concept that cancer cells continue to grow through the increased expression and activity of oncoproteins such as E2F1 and MAGE-A11.
The unrestrained growth of cancer cells also depends on deregulation of tumor suppressors, such as human p14-ARF (ARF, alternative reading frame) and p53, proteins that protect normal cells from tumorigenesis by arresting the cell cycle or inducing apoptosis. ARF derives from the INK4a/ARF locus in an alternative reading frame by alternate promoter usage and splicing that differs from the p16 cyclin-dependent kinase inhibitor that is more often mutated in cancer (17)(18)(19)(20)(21). Human p14-ARF shares only 50% homology with the p19-ARF mouse homologue (22), which indicates that the INK4a/ARF gene continued to evolve late within the mammalian lineage similar to the MAGE-A11 gene and AR NH 2 -terminal FXXLF motif flanking sequence required to interact with MAGE-A11. A primary function of ARF is to bind and inhibit HDM2 (human homologue of mouse double minute 2), an E3 ligase that ubiquitinates and promotes the degradation of p53 (23)(24)(25)(26)(27). Inhibition of HDM2 E3 ubiquitin ligase activity by ARF increases p53 levels, which induces cell cycle arrest or apoptosis. The inhibitory activity of ARF that promotes p53 accumulation requires a unique ARF exon 1B region (22). Loss of ARF or p53 activity in cancer cells allows cellular proto-oncogenes such as E2F1 and MAGE-A11 to promote uncontrolled cancer cell growth and the development of castration-resistant prostate cancer.
In this report we provide evidence that human ARF promotes the proteasomal degradation of MAGE-A11 independent of HDM2 E3 ubiquitin ligase or lysine ubiquitination. ARF inhibited the ability of MAGE-A11 to stabilize and stimulate AR or E2F1 transcriptional activity but did not overcome a stabilizing effect of MAGE-A11 on HDM2. The findings suggest that ARF maintains normal low levels of MAGE-A11 to limit cell proliferation and preserve the nontransformed phenotype, whereas low levels of ARF in prostate cancer cells contribute to higher levels of MAGE-A11 that promote the development of castration-resistant prostate cancer.
Transcription assays were performed in 12-well plates containing 1 ml medium/well plated with 5 ϫ 10 4 HeLa cells/well or 4 ϫ 10 4 CV1 cells/well and transfected using 0.6 l of X-tremeGENE DNA transfection reagent (Roche) and 50 l medium/well. AR transcriptional activity was measured in CV1 cells using 0.25 g of PSA-Enh-Luc, 25 ng of pCMV-AR, 10 ng of pCMV-AR-(1-660), and/or 25 ng of pCI-neo-HA-ARF. Endogenous E2F1 transcriptional activity was measured in HeLa cells using 0.1 g of E2F1-Luc, 0.1 g of pSG5-MAGE, and increasing amounts of pCI-neo-HA-ARF. Mammalian two-hybrid assays were performed in HeLa cells using 0.1 g of 5XGAL4Luc3, 50 ng of GAL-ARF, 50 ng of VP16 empty vector, or wild-type or mutant VP-MAGE. The day after transfection, cells were incubated for 24 h in serum-free medium with or without 10 nM DHT. Luciferase activity measurements represent the mean and S.E. of at least three independent experiments.
Control and human p14-ARF retrovirus derived from pBabepuro or pBabe-puro-hARF retroviral vectors were prepared using HEK293T cells. Retrovirus in 0.25 ml of HEK293T cell media was added to 2 ϫ 10 6 LAPC-4 cells/well or 6.2 ϫ 10 5 HeLa cells/well in 6-well plates containing 2 ml of medium/well and incubated for 48 h at 37°C. Confluent cells were selected over 9 days by initially passaging 2 wells into a 10-cm dish. After cell attachment, 3 g/ml puromycin was included in the medium that was exchanged every 2 or 3 days. LAPC-4 or HeLa cells not treated with retrovirus were killed within 5 days of treatment with 3 g/ml puromycin. Quantitative RT-PCR of RNA extracted using TRIzol from LAPC-4 and HeLa pBabecontrol and pBabe-ARF-expressing cells was performed using sequence specific primers for MAGE-A11 and peptidylprolyl isomerase A control as described (38).
Lentivirus shRNA-Endogenous ARF expression was inhibited using a pLK01-modified lentiviral vector that contains an RNA polymerase III-shRNA expression cassette and a puromycin resistance gene under the control of the phosphoglycerate kinase promoter. ARF shRNA targeted the 5Ј-CTCGTGCT-GATGCTACTGAgg-3Ј sense sequence in the unique ARF exon 1B (antisense sequence ccTCAGTAGCATCAGCAC-GAG-3Ј). Lentiviral vectors carrying nonspecific shRNA or empty vector with no shRNA served as negative controls. Vector particles (ϳ10 6 IU/ml) were generated by a three-plasmid transfection in HEK293 cells (39). HeLa cells (3 ϫ 10 5 cells/well in 6-well plates) were transduced using 0.1 or 0.25 ml of vectorcontaining conditioned media for 48 h at 37°C and selected for puromycin resistance over several days in the presence of 3 g/ml puromycin.
Cell Growth Assays-LAPC-4 cell growth assays were performed using cells that stably express retrovirus derived from pBabe-puro control or pBabe-puro-hARF retroviral vectors. LAPC4 control and LAPC-4-ARF cells were plated in triplicate (4 ϫ 10 5 cells/well) in 24-well plates with 0.5 ml of phenol redfree growth medium containing 10% charcoal-stripped serum (Atlanta Biologicals). The next day 0.1 ml of phenol red-free, serum-free medium without DHT was added or with DHT for a final concentration of 1 nM DHT. Cells were retreated with or without 1 nM DHT after 72 h on day 3. Plates were aspirated 24 h after addition of DHT and daily thereafter. Serum-free medium (0.2 ml) and 20 l of Cell Counting Kit-8 reagent (Dojindo Laboratories) were added/well and incubated at 37°C for 2.5 h. Optical absorbance was determined at 485 nm. Statistical significance was determined using two-way ANOVA.

Results
Interaction with Human ARF Promotes the Degradation of MAGE-A11-MAGE-A11 coregulator activity depends on interactions with the ligand-activated human AR and p300 transcriptional coactivator (1,6,10,11). MAGE-A11 also interacts with and increases the transcriptional activity of the E2F1 oncoprotein that promotes cell cycle progression through interactions with p107 of the retinoblastoma family (5). Based on the ability of MAGE-A11 to stabilize and increase E2F1 transcriptional activity, and the ARF tumor suppressor to promote E2F1 degradation and inhibit E2F1 transcriptional activity (41,42), we investigated whether MAGE-A11 is a direct target of human ARF or proteins regulated by ARF.
Interaction between MAGE-A11 and ARF was demonstrated by the coimmunoprecipitation of HA-ARF with FLAG-MAGE (Fig. 1A), and by MAGE-A11 coimmunoprecipitation with FLAG-ARF (Fig. 1B). There was no evidence for nonspecific association of ARF or MAGE-A11 with the FLAG antibody resin. However, expression of ARF decreased MAGE-A11 levels in cell extracts by 86% (Fig. 1A) or 44% (Fig. 1B) when normalized to the ␤-actin protein loading control.
An interaction between MAGE-A11 and ARF was also evident in a mammalian two-hybrid assay. Activity of a luciferase reporter gene linked to GAL4 DNA binding sites was increased by 6-fold with the coexpression of an ARF-GAL4 DNA-binding domain fusion protein and a full-length MAGE-A11-VP16 activation domain fusion protein (Fig. 1C). The similar activity induced by GAL-ARF in the presence of full-length 1-429amino acid VP-MAGE or VP-MAGE-(112-429) suggested that ARF interacts with a COOH-terminal region of MAGE-A11. This was consistent with a weak two-hybrid interaction between GAL-ARF and VP-MAGE-(2-110) that contains only the MAGE-A11 NH 2 -terminal region or between GAL-ARF and VP-MAGE-(112-362) that contained the central region of MAGE-A11. GAL-ARF, VP-MAGE, and VP-MAGE fragments expressed at similar levels, except for VP-MAGE-(2-110), which was expressed at a lower level (Fig. 1D). The results suggest an interaction between MAGE-A11 and ARF requires the COOH-terminal region of MAGE-A11.
The decline in MAGE-A11 associated with ARF expression (Fig. 1, A and B) suggested that ARF increases the degradation of MAGE-A11. This was supported by a dose-dependent inverse relationship between ARF and MAGE-A11 relative to ␤-actin ( Fig. 2A). In contrast, there was a direct relationship between ARF and AR ( Fig. 2B) that was lost with expression of MAGE-A11 (Fig. 2C). When expressed together, the levels of both AR and MAGE-A11 decreased with increasing ARF. The results suggest that the susceptibility of MAGE-A11 to degradation promoted by ARF was extended to AR through the interaction between AR and MAGE-A11.
Inhibition of protein synthesis using cycloheximide provided additional evidence that ARF increases the degradation rate of MAGE-A11 (Fig. 2D). The intracellular half-life of MAGE-A11 decreased from ϳ8 to 4 h with the expression of ARF.
The region in MAGE-A11 required for degradation promoted by ARF was investigated based on the similar susceptibility of MAGE-A11 ( Fig. 2A) and a full-length GAL-MAGE fusion protein (Fig. 2E, lanes 1 and 2) to degradation promoted by ARF. The stability of GAL-MAGE-(2-121) or GAL-MAGE-  in the presence of ARF (Fig. 2E, lanes 3-6) provided further evidence that the MAGE-A11 COOH-terminal region is targeted for degradation by ARF. In contrast, GAL-MAGE-(112-429) and GAL-MAGE-(112-276), which share the central region of MAGE-A11, were down-regulated by ARF similar to full-length GAL-MAGE (Fig. 2F).
The results show that ARF promotes the degradation of MAGE-A11 independent of the MAGE-A11 NH-terminal region or AR. Increased degradation of AR promoted by ARF was evident only in the presence of MAGE-A11.

MAGE-A11 Degradation by the Proteasome Promoted by
Human ARF Independent of Ubiquitination-Dependence on the proteasome and lysine ubiquitination for MAGE-A11 degradation promoted by ARF was investigated using chemical inhibitors and alanine substitution mutants. MAGE-A11 levels in the presence of ARF and normalized to ␤-actin increased 1.3-fold after incubation with 1 M MG132 and 2.7-fold with 10 M MG132 (Fig. 3A). A similar increase in MAGE-A11 was seen in the presence of ARF and 5 M bortezomib, another proteasome inhibitor (43) (Fig. 3B). The results provided evidence that ARF promotes MAGE-A11 degradation by the proteasome.
The requirement for lysine ubiquitination in the proteasomal degradation of MAGE-A11 promoted by ARF was examined by first establishing a minimal region of MAGE-A11 that could be subjected to lysine mutagenesis. In agreement with results in Fig. 2 (E and F), the MAGE-(2-252) NH 2 -terminal fragment did not interact with FLAG-ARF (Fig. 3C, lanes 4 and 11). This was in contrast to a strong interaction between ARF and MAGE-(112-429) and MAGE-(112-362) and a weaker interaction with MAGE-(112-276) (Fig. 3C, lanes 5-7 and 12-14).
The results suggest that the interaction between ARF and the central region of MAGE-A11 involves multiple domains that result in degradation by the proteasome independent of lysine ubiquitination. The ability of ARF to promote proteasomal degradation of MAGE-A11 is consistent with tumor suppressor of ARF that inhibits the cell growth promoting properties of MAGE-A11.
Inverse Relationship between Endogenous ARF and MAGE-A11-An increase in MAGE-A11 degradation promoted by ARF should decrease endogenous MAGE-A11 levels in normal cells that have higher levels of ARF. Similarly, higher levels of endogenous MAGE-A11 in prostate cancer cells may be associated with low levels of ARF to promote the growth of castration-resistant prostate cancer. We therefore investigated the relative levels of endogenous MAGE-A11 and ARF.
ARF was not easy to detect in LAPC-4 prostate cancer cells that have relatively high levels of MAGE-A1 using human ARF and MAGE-A11-specific antibodies (Fig. 4A, lane 4) (2, 5). However, a weak band corresponding to ARF could be detected with overexposure of the blot (not shown), which indicated a low level of ARF in LAPC-4 cells. An inverse relationship FIGURE 2. ARF promotes MAGE-A11 degradation. A, inverse relationship between MAGE-A11 and ARF was demonstrated by expressing 2 g of pSG5 (lane 1) or 2 g of pSG5-MAGE with 0.01-2 g of pCI-neo-HA-ARF/two 6-cm CV1 cell dishes (lanes 2-9). Cell extracts (80 g of protein/lane) pooled from two 6-cm dishes were analyzed on an immunoblot using MAGE1, HA, and ␤-actin antibodies. B, increase in AR and ARF was shown by expressing 2 g of pCMV5 (lane 1) or 2 g of pCMV-AR with 0.01-2 g of pCI-neo-HA-ARF balanced with up to 2 g of pCMV5 (lane 2-9)/two 6-cm CV1 cell dishes. Cell extracts (80 g of protein/lane) from two pooled 6-cm dishes were probed on an immunoblot using AR32, HA tag, and ␤-actin antibodies. C, that ARF promotes AR degradation in the presence of MAGE-A11 was shown by expressing 2 g of pSG5 (lane 1) or 2 g of pSG5-MAGE with 1 g of pCMV-AR and 0.01-2 g of pCI-neo-HA-ARF/ two 6-cm CV1 cell dishes (lanes 2-9). Cell extracts (80 g of protein/lane) pooled from two 6-cm dishes were analyzed on an immunoblot using AR32, MAGE1, HA, and ␤-actin antibodies. D, increase in MAGE-A11 degradation in the presence of ARF was shown by expressing 1 g of pSG5-MAGE with or without 0.5 g of pCI-neo-HA-ARF/6-cm COS1 cell dish. The day after transfection, cells were incubated for 24 h in serum-free medium and then in fresh serum-free medium containing 10 g/ml cycloheximide for 0, 2 between MAGE-A11 and ARF was also suggested by relatively low levels of MAGE-A11 and higher levels of ARF in human cervical carcinoma HeLa cells compared with LAPC-4 cells (Fig. 4A, lanes 3 and 4). ARF and MAGE-A11 were not detected in COS1 or CV1 cells (Fig. 4A, lanes 1 and 2), although endogenous MAGE-A11 can be detected in COS1 cells in a cell cycledependent manner (10).
PC-3 and DU145 prostate cancer cells had higher levels of ARF (Fig. 4B, lanes 5 and 6) (44, 45), and MAGE-A11 was not detected, consistent with ARF promoting the degradation of MAGE-A11. Low levels of MAGE-A11 in PC-3 and DU145 cells also result from CpG dinucleotide DNA methylation at the MAGE-A11 gene promoter transcription start site (2). LNCaP, CWR-R1, and 22Rv1 prostate cancer cells had intermediate levels of ARF relative to LAPC-4, PC-3, and DU145 cells, and MAGE-A11 was difficult to detect (Fig. 4B).
The cyclin-dependent kinase inhibitor and tumor suppressor p27 Kip1 is down-regulated by MAGE-A11 in LAPC-4 cells (5). Accordingly, p27 Kip1 was almost undetectable in LAPC-4 and was not detected in PC-3 or DU145 cells (Fig. 4B). A similar intermediate level of p27 Kip1 in CWR-R1 and LNCaP cells was less than 22Rv1 cells.
The results demonstrate that higher levels of MAGE-A11 are associated with low levels of ARF in support of ARF promoting the degradation of MAGE-A11. The results provide further evidence that MAGE-A11 contributes to low levels of p27 Kip1 and that p27 Kip1 may be down-regulated by ARF independent of MAGE-A11.
Down-regulation of MAGE-A11 by ARF-To address further a possible inverse relationship between MAGE-A11 and ARF,   OCTOBER 9, 2015 • VOLUME 290 • NUMBER 41 retrovirus derived from pBabe-control and pBabe-ARF vectors were used for stable expression of ARF in HeLa and LAPC-4 cells selected using puromycin. A ϳ3-fold or greater increase in ARF decreased endogenous MAGE-A11 levels by 64% in HeLa cells (Fig. 5A) and by 45% in LAPC-4 cells (Fig. 5B) relative to control cells and normalized to the ␤-actin protein loading control. Quantitative RT-PCR analysis of MAGE-A11 mRNA extracted from LAPC-4 and HeLa pBabe-control or pBabe-ARF expressing cells using MAGE-A11 and peptidylprolyl isomerase A control specific primers (38) showed no significant change in MAGE-A11 mRNA with the expression of ARF (Fig.  5C). The results demonstrate that lower levels of MAGE-A11 are associated with higher levels of ARF in support of an inverse relationship between MAGE-A11 and ARF consistent with MAGE-A11 degradation promoted by ARF.

Human p14-ARF Tumor Suppressor Promotes MAGE-A11 Degradation
We next determined whether stable retrovirus expression of ARF alters the growth of LAPC-4 cells in the absence or presence of androgen. DHT increased the growth of LAPC-4 pBabe-control cells analyzed using a colorimetric cell counting assay (Fig. 5D). However, stable expression of ARF inhibited LAPC-4 cell growth over 5 days in the absence or presence of 1 nM DHT.
An increase in MAGE-A11 degradation promoted by ARF would suggest that knockdown of ARF might increase endogenous levels of MAGE-A11. This was explored in HeLa cells that express low levels of MAGE-A11 and higher levels of ARF rel-ative to LAPC-4 cells. Lentivirus-expressed shRNA that targets the unique exon 1B region of the ARF gene decreased ARF levels and increased the levels of MAGE-A11 in HeLa cells treated without (Fig. 6A) or with the proteasome inhibitor, MG132 (Fig. 6B).
The results suggest that the low abundance MAGE-A11 regulatory protein is inversely regulated to ARF. Increased degradation of MAGE-A11 promoted by ARF was associated with inhibition of prostate cancer cell growth.
Inhibition of AR Transcriptional Activity by Human ARF-The inhibitory effect of ARF on coregulator activity of MAGE-A11 was investigated by measuring AR transcriptional activity that is increased by MAGE-A11 (1). AR transactivation of the prostate-specific antigen enhancer linked to a luciferase reporter gene was inhibited by ARF (Fig. 7A) in agreement with a previous report (46). ARF also blocked the stimulatory effect of MAGE-A11 on androgen-dependent AR transcriptional activity. The results suggest that an inhibitory effect of ARF on AR transcriptional activity is mediated by negative regulation of MAGE-A11.
Inhibition by ARF of the MAGE-A11-induced increase in AR transactivation suggested that the interaction between AR and MAGE-A11 does not protect against degradation promoted by ARF. This was suggested earlier (Fig. 2) where susceptibility of MAGE-A11 to degradation promoted by ARF was extended to AR. ARF also inhibited the stabilization of MAGE-A11 in the Retrovirus transduced LAPC-4 cells were selected using 3 g/ml puromycin as described under "Experimental Procedures." Cell extracts (60 g of protein/lane) were probed on the immunoblot using MAGE1 (10 g/ml), human p14-ARF 10437 (1:500 dilution), and ␤-actin antibodies. MAGE-A11 versus ␤-actin band intensity is shown in the right panel. C, quantitative RT-PCR was performed on RNA extracted from LAPC-4 or HeLa pBabe-control or pBabe-ARF expressing cells using specific amplification primers and shown for MAGE-A11 relative to the peptidylprolyl isomerase A control. D, increased ARF expression slows LAPC-4 cell growth was shown using LAPC-4 cells transduced with retrovirus as described in B. The day after plating, cells were incubated with or without 1 nM DHT and quantitated daily using a colorimetric assay as described under "Experimental Procedures." Statistical significance between LAPC-4 pBabe and pBabe-ARF cells was observed in the absence or presence of 1 nM DHT based on two-way ANOVA (p Ͻ 0.001).
presence of AR (Fig. 7B, lanes 3-6) and interfered with the increase in AR levels in the presence of MAGE-A11 (Fig. 7B,  lanes 2-4). Reciprocal stabilization between AR and MAGE-A11 in the absence or presence of androgen only partially rescued MAGE-A11 from degradation promoted by ARF (Fig. 7C).
The results suggest that ARF interferes with MAGE-A11induced AR stabilization and increased transcriptional activity of AR or a splice variant-like form of AR. MAGE-A11 remained susceptible to degradation promoted by ARF independent of its interaction with AR.
Human ARF Inhibits the MAGE-A11-induced Increase in E2F1 Transcriptional Activity-MAGE-A11 interacts with, stabilizes, and increases the transcriptional activity of E2F1, a transcription factor that promotes progression through the cell cycle (5). In contrast, ARF interacts with and inhibits E2F1 transcriptional activity through unknown mechanisms (42). We therefore determined whether ARF blocks the stabilization of E2F1 by MAGE-A11 or the MAGE-A11-dependent increase in endogenous E2F1 transcriptional activity.
The increase in E2F1 levels associated with MAGE-A11 was partially inhibited by ARF (Fig. 9A, lanes 4 -6). This did not appear to result simply from increased degradation of MAGE-A11 promoted by ARF. MAGE-A11 levels declined to a greater extent with ARF expression than with E2F1 stabilization (Fig.  9A, lanes 2-5). However, ARF did not decrease MAGE-A11 levels further in the presence of E2F1, and ARF did not promote the degradation of E2F1, in agreement and in contrast to previous reports (41,51).
Immunoprecipitation studies addressed whether the inhibitory effects of ARF on E2F1 stabilization by MAGE-A11 might result from ARF interference with MAGE-A11 interaction with E2F1. MAGE-A11 and ARF were each associated with FLAG-E2F1 (Fig. 9B, lanes 5 and 6). However, when expressed together, ARF inhibited the interaction between MAGE-A11 and E2F1 to a greater extent than could be accounted for by the lower levels of MAGE-A11 in the presence of ARF (Fig. 9B, lane  7). ARF also blocked the dose-dependent stimulatory effect of MAGE-A11 on endogenous E2F1 transcriptional activity when assayed in the absence or presence of p300 (Fig. 9C). There was a dose-dependent decrease in MAGE-A11-induced endogenous E2F1 transcriptional activity with increasing expression of ARF (Fig. 9D). Specificity for transcriptional repression by ARF was suggested by the absence of inhibition by ARF of CMV-Luc or pSG5-Luc constitutive transcriptional activity (not shown).
The results suggest that ARF interferes with the stabilizing and transcriptional enhancing effects of MAGE-A11 on E2F1 by blocking MAGE-A11 interaction with E2F1 and by promoting MAGE-A11 degradation. Inhibition of the stimulatory effects of MAGE-A11 on E2F1 transcriptional activity is consistent with the tumor suppressor activity of ARF.
Effects of MAGE-A11 on HDM2 Stabilization and E2F1 Transcriptional Activity-A well known function of ARF that contributes to cell cycle arrest is the inhibition of HDM2 E3 ubiquitin ligase activity, which results in up-regulation of p53 to promote cell cycle arrest or apoptosis (23)(24)(25)(26)(27)52) and downregulation of E2F1 activation of genes involved in cell cycle progression (53)(54)(55). This suggests that one mechanism by which cancer cells interfere with the tumor suppressor activity of ARF is through the stabilization of HDM2.
We found that MAGE-A11 increased the levels of ϳ90-kDa HDM2 by ϳ15-fold detected on an 8% acrylamide gel and an up-shifted self-ubiquitinated or sumoylated form of HDM2 (56,57) when normalized to the ␤-actin protein loading control (Fig. 10A). Stabilization of ϳ100-kDa HDM2 by MAGE-A11 on a 12% acrylamide gel was also increased in the presence of ARF (Fig. 10B, lanes 2-5), even though ARF promoted the degradation of MAGE-A11 in the absence or presence of HDM2 (Fig.  10B, lanes 5-7). The small increase in HDM2 by ARF (Fig. 10B,  lanes 2 and 4) was in agreement with a previous report that ARF increases the steady-state levels of HDM2 (58). Although stabilization of HDM2 by MAGE-A11 was inhibited by E2F1 (Fig.  10C, lanes 5 and 7), there was no evidence that HDM2 altered E2F1 levels, and HDM2 did not interfere with E2F1 stabilization by MAGE-A11 (Fig. 10C, lanes 2-5).
The increase in HDM2 caused by MAGE-A11 (Fig. 10A) and the ability of MAGE-A11 or HDM2 to increase E2F1 transcriptional activity (5,(53)(54)(55) suggested that MAGE-A11 and HDM2 may cooperate to up-regulate E2F1 transcriptional activity. This was supported by a similar increase in endogenous E2F1 transactivation in response to HDM2 or MAGE-A11 that was nearly additive when expressed together (Fig. 10D). MAGE-A11 and HDM2 functioned cooperatively with p300 to up-regulate endogenous E2F1 transcriptional activity, which reflects the ability of MAGE-A11 and HDM2 to interact with p300 during gene transcription (5,10,59). The increase in E2F1 transcriptional activity in response to MAGE-A11 and p300 was greater with HDM2 and inhibited by ARF with or without HDM2 (Fig. 10E). Cooperation between MAGE-A11, HDM2, and p300 in the up-regulation of E2F1 transcriptional activity suggested that p300 might influence HDM2 levels. However, unlike MAGE-A11, which stabilized HDM2 (Fig. 10F, lanes 2  and 3), p300 did not increase HDM2 or interfere with HDM2 stabilization in the presence of MAGE-A11 (Fig. 10F, lanes  3-5).
The results suggest that MAGE-A11 cooperates with HDM2 and p300 to up-regulate E2F1 transcriptional activity. MAGE-A11 in a complex with HDM2 and ARF may interfere with HDM2 inactivation by ARF.

Discussion
Post-translational Down-regulation of MAGE-A11 by Human ARF Tumor Suppressor-MAGE-A11 is a multifunctional protein involved in the regulation of cell cycle proteins. The evolution of the MAGE-A11 among primates, its increased expression during androgen deprivation therapy of prostate cancer, its function as an AR coregulator, and the requirement for MAGE-A11 in prostate cancer cell growth support the concept that MAGE-A11 is a proto-oncogene that hyperactivates human AR and promotes the development of castration-resistant prostate cancer (38). One mechanism for the increase in MAGE-A11 in prostate cancer clinical samples during androgen deprivation therapy and in the CWR22 human xenograft  model of prostate cancer that undergoes remission after castration but regrows after castration is progressive hypomethylation of CpG dinucleotides at the transcription start site of the MAGE-A11 gene promoter (2,3). MAGE-A11 expression is also up-regulated in prostate cancer during androgen deprivation therapy by increasing levels of cAMP associated with down-regulation of phosphodiesterases that degrade cAMP (2, 60 -63). In this report, we extend the family of MAGE-A11 interacting partners to include the human ARF tumor suppressor that targets MAGE-A11 for degradation by the proteasome independent of lysine ubiquitination.
Our studies suggest that down-regulation of MAGE-A11 by ARF represents a third mechanism that controls MAGE-A11, where low levels of ARF contribute to higher levels of MAGE-A11 during prostate cancer progression. Our findings are consistent with the tumor suppressor activity of ARF that protects normal cells from tumorigenesis and the proto-oncogene activity of MAGE-A11 that increases prostate cancer cell growth. The increase in MAGE-A11 in prostate cancer during androgen deprivation therapy provides an escape mechanism whereby prostate cancer cells survive and expand in an environment of low intratumoral active androgen biosynthesis.
Our studies suggest a model (Fig. 11) in which MAGE-A11 is central to a protein network involved in human cell growth regulation. We showed previously that MAGE-A11 increases AR transcriptional activity by binding the AR NH 2 -terminal FXXLF motif and recruiting p300 and TIF2 (NCOA2) transcriptional coregulators, which increases androgen-dependent prostate cancer cell growth (1,10,11,38). Here we provide evidence that MAGE-A11 interacts with and is targeted for degradation by the proteasome by the human ARF tumor suppressor. MAGE-A11 functions cooperatively with HDM2 E3 ubiquitin ligase to increase the transcriptional activity of E2F1, an oncoprotein that promotes progression through the cell cycle (5,64). ARF inhibits the increase in AR and E2F1 transcriptional activity induced by MAGE-A11. Multiple interaction regions between ARF and HDM2 (65) were seen between MAGE-A11 and ARF that resulted in MAGE-A11 degradation.
Inhibition of the stimulatory effects of MAGE-A11 on AR and E2F1 transcriptional activity by ARF provides new mechanisms of tumor suppression and maintenance of the nontransformed phenotype (5,42).
Low levels of ARF associated with cancer development result from destabilization of the ARF protein or from INK4a/ARF gene deletions, mutations, or methylation (66,67). Low levels of ARF in prostate cancer (19 -21) were also attributed to androgen-inducible microRNA-125b that negatively regulates ARF in prostate cancer (68). Our studies suggest that low levels of ARF in prostate cancer increase the levels of MAGE-A11 by decreasing MAGE-A11 degradation. Our findings are consistent with the proto-oncogene properties associated with increased levels of MAGE-A11 and with the tumor suppressor activity of ARF. Increased expression of ARF was proposed as a therapeutic approach to increase cancer sensitivity to chemotherapy (69). Our findings suggest that prostate cancers with elevated levels of MAGE-A11 are candidates for clinical intervention that increases ARF, which would decrease MAGE-A11 and induce cell cycle arrest or apoptosis. The association between higher levels of MAGE-A11 and low ARF in prostate cancer supports the concept that ARF maintains low levels of MAGE-A11 in normal cells by promoting degradation by the proteasome, and human ARF tumor suppressor activity depends on the post-translational down-regulation of MAGE-A11.
ARF Counteracts the Stimulatory Effects of MAGE-A11 on AR Transcriptional Activity-Increased AR transcriptional activity in response to low intratumoral androgen synthesis is a principal driver of castration-resistant prostate cancer (70,71). AR transcriptional activity also increases in response to higher levels of MAGE-A11, which has been observed in ϳ36% of castration-resistant prostate cancers (1,2). Here we show that ARF interferes with the stimulatory effects of MAGE-A11 on androgen-dependent AR transcriptional activity and the constitutive activity of a splice-variant-like AR reported in prostate cancer (47)(48)(49). The findings suggest that the inhibitory effects of ARF on AR transcriptional activity are mediated in part by down-regulation of MAGE-A11.
It was reported that ARF inhibits AR transcriptional activity by interfering with the androgen-dependent AR NH 2 -and COOH-terminal interaction (46) that stabilizes AR in the presence of AR agonists (14,72) and that AR is degraded by the proteasome in association with gene activation (73). We found that ARF interferes with a positive relationship between AR and MAGE-A11 by targeting MAGE-A11 for degradation. Although ARF did not directly decrease AR levels, ARF negatively regulated AR in the presence of MAGE-A11. Dependence on MAGE-A11 interaction with the AR NH 2 -terminal FXXLF motif for the effects of ARF provided evidence that ARF inter-feres with AR transcriptional activity by negatively regulating MAGE-A11. The inhibitory effects of ARF on AR transcriptional activity evident without the increased expression of MAGE-A11 suggest that negative regulation of AR by ARF occurs even when MAGE-A11 levels are low (10).
Up-regulation of HDM2 by MAGE-A11 Increases E2F1 Transcriptional Activity That Is Inhibited by ARF-HDM2 E3 ubiquitin ligase is a proto-oncogene overexpressed in human cancers and a target for anticancer therapy (74,75). HDM2 stabilizes E2F1 and mediates ubiquitin-dependent degradation of p53 (54). Here we show that MAGE-A11 cooperates with HDM2 to stabilize and increase the transcriptional activity of E2F1, an oncoprotein that up-regulates genes important for cell cycle progression (53)(54)(55). Our results suggest that MAGE-A11 is integral to a protein network involved in E2F1 regulation of cell cycle progression. The ability of MAGE-A11 to increase E2F1 transcriptional activity supports a critical role in prostate cancer cell growth.
A principal function of the ARF tumor suppressor is inhibition of HDM2, which results in the stabilization of p53 and loss of inhibition of E2F1. Both ARF and p53 tumor suppressor activity protect normal cells from oncogene-induced tumor formation (76). Previous studies suggested that ARF negatively regulates E2F1 through increased E2F1 ubiquitination and proteasome degradation (41,53,77). It was suggested that inhibition of E2F1-dependent transcriptional activity involves HDM2 and/or p53 (42,78). We found that one function of ARF is inhibition of MAGE-A11 interaction with E2F1, which interferes with the stabilizing effects and transcriptional enhancing effects of MAGE-A11. Negative feedback regulation is suggested by the up-regulation of ARF and p53 by E2F1 that contributes to the induction of apoptosis (79,80). ARF also interacts with c-myc and causes c-myc relocalization from the nucleoplasm to nucleoli, inhibits c-myc transcriptional activation of genes required for cell cycle progression (81), and induces sumoylation of its binding partners (57).
It remains to be determined what role MAGE-A11 has in the regulation of p53 and whether stabilization of HDM2 by MAGE-A11 interferes with ARF inhibition of HDM2 E3 ubiquitin ligase activity that promotes p53 degradation and cell cycle progression. It is not known how MAGE-A11 impacts feedback regulation of p53, such as the increase in ARF expression when p53 is inhibited or increased expression of ARF and p53 in response to cellular and viral oncogenes such as c-myc or E1A (82). Stabilization and sequestration of MAGE-A11 and HDM2 by ARF may be involved in feedback control (79). Furthermore, HDM2 interaction with p300 is required for the proteasomal degradation of p53 (75). MAGE-A11 forms a strong complex with p300 in association with gene transcription (10) that inhibited the stabilization of HDM2 by MAGE-A11. These findings suggest that MAGE-A11 interaction with HDM2 and p300 may be involved in HDM2 regulation of p53. Phosphorylation of HDM2 by Akt signals the nuclear localization of HDM2. ARF inhibits nuclear export of HDM2 by the sequestration of HDM2 in nucleoli, where HDM2 no longer can export p53 to the cytoplasm for degradation and thereby activates p53 (24,50,58,(83)(84)(85). The impact of MAGE-A11 on ARF and HDM2 regulation of p53 remains to be determined.
Author Contributions-J. T. M., G. G., and E. M. W. performed the experiments and analyzed the data. P. Z. and T. K. contributed retrovirus and lentivirus design and production. All authors reviewed the results and approved the final version of the manuscript.