Phosphorylation and Activation of Androgen Receptor by Aurora-A*

Aurora-A kinase is frequently overexpressed/activated in various types of human malignancy, including prostate cancer. In this study, we demonstrate elevated levels of Aurora-A in androgen-refractory LNCaP-RF but not androgen-sensitive LNCaP cells, which prompted us to examine whether Aurora-A regulates the androgen receptor (AR) and whether elevated Aurora-A is involved in androgen-independent cell growth. We show that ectopic expression of Aurora-A induces AR transactivation activity in the presence and absence of androgen. Aurora-A interacts with AR and phosphorylates AR at Thr282 and Ser293 in vitro and in vivo. Aurora-A induces AR transactivation activity in a phosphorylation-dependent manner. Ectopic expression of Aurora-A in LNCaP cells induces prostate-specific antigen expression and cell survival, whereas knockdown of Aurora-A sensitizes LNCaP-RF cells to apoptosis and cell growth arrest. These data indicate that AR is a substrate of Aurora-A and that elevated Aurora-A could contribute to androgen-independent cell growth by phosphorylation and activation of AR.

Androgens, functioning through the androgen receptor (AR), 2 are essential for the initiation and progression of prostate cancer (1), and androgen ablation therapy usually achieves significant clinical responses in the beginning. Under the selective pressure of androgen withdrawal, however, prostate cancers progress to an androgen-independent stage. Because AR is expressed in the majority of androgen-dependent prostate cancer and androgen-independent prostate cancer (1,2), and decreasinglevelsofARproteinexpressionreducesbothandrogendependent prostate cancer and androgen-independent prostate cancer growth in model systems (2,3), it appears that the AR signaling pathway plays a critical role in both androgen-dependent prostate cancer and androgen-independent prostate cancer.
The mechanism for this progression to androgen independence is not completely understood. Although androgen-independent progression has been correlated with mutation of the AR gene (4), most androgen-independent prostate cancers express AR and the androgen-dependent gene prostate-specific antigen (PSA), implying that these cells maintain a functional AR signaling pathway. Furthermore, it has been shown that activation of MAPK, AKT, and tyrosine receptor kinases is involved in the development of androgen escape (5)(6)(7)(8). These pathways act to directly phosphorylate AR, altering the sensitivity of AR to both androgens and anti-androgens.
Aurora-A, a serine/threonine protein kinase, has been shown to be crucial in chromosome segregation and centrosome functions (9). Aurora-A has attracted intense interest after the discovery of its chromosomal localization (20q13.2), a common amplicon in epithelial cancers (10,11). It has been shown that 20q13.2 amplifications involving the Aurora-A gene occur in as many as 12-50% of ovarian, breast, colorectal, and gastric cancers (12)(13)(14). Moreover, up to 57 and 62% of ovarian and breast cancers, respectively, show overexpression and/or activation of Aurora-A, even where gene amplification is not detected (15)(16)(17)(18)(19). In addition, ectopic expression of Aurora-A in murine fibroblasts as well as mammary epithelia induces centrosome amplification, aneuploidy, and oncogenic phenotype (20,21).
Previous studies show that Aurora-A is overexpressed in high grade prostatic intraepithelial neoplasia and prostate cancer, and such expression patterns correlate with tumorigenicity, clinical staging, surgical margin status, and seminal vesicle invasion (22)(23)(24). Aurora-A expression in the radical prostatectomy after neoadjuvant hormonal therapy specimens correlated significantly with the preoperative value of the serum PSA, cell-proliferative activity, and pathological stage (23). The biochemical recurrence-free survival in patients with a persistent Aurora-A expression in radical prostatectomy specimens was significantly lower than that in those with a weak Aurora-A expression (23,24). After treatment with Aurora kinase inhibitor VX680, human prostate cancer cell lines (PC3, LNCaP, and mouse C1A) exhibited attenuation of the phosphorylation of histone H3 and reduced survival (25). However, the mechanism of Aurora-A in prostate carcinogenesis and cancer progression has not been thoroughly addressed.
In this study, we demonstrated that Aurora-A interacts with AR and phosphorylates AR-Thr 282 /Ser 293 in the transactivation domain. Aurora-A induces AR transactivation activity in the presence and absence of androgen in phosphorylation-dependent manner. Ectopic expression Aurora-A in LNCaP cells increases the PSA expression and induces cell survival, whereas knockdown of Aurora-A in androgen-independent LNCaP-RF cells sensitizes cells to death. Therefore, we provide direct evidence that Aurora-A could contribute to prostate carcinogen-esis and androgen-independent growth through phosphorylation and activation of AR.

EXPERIMENTAL PROCEDURES
Reagents and Plasmids-Anti-Aurora-A antibody was generated by immunization of rabbit with GST-Aurora/box-2 fusion protein (16). AR antibody was from Upstate. PSA antibody was purchased from Abcam. Anti-HA and -FLAG antibodies were from Sigma. Recombinant AR and Aurora-A proteins were from Stressgen and Cell Signaling, respectively. HA-tagged (pHM6) Aurora-A expression plasmids were described previously (18,19). FLAG-tagged full-length and truncated AR mutants and GST-AR constructs were created by PCR and subcloned into FLAG-pCMV2, pcDNA3, and pGEM-4T-1 vectors, respectively. Mutations of AR Thr 282 /Ser 293 to alanine (AR-2A) and aspartic acid (AR-2D) were prepared with the QuikChange site-directed mutagenesis kit (Stratagene) and were confirmed by sequence analysis. Aurora-A/shRNA and scramble shRNA were obtained from Sigma. Androgen response region (ARR3)tk-Luc was kindly provided by Robert Matusik (Vanderbilt University).
In Vitro Kinase Assay, in Vivo [ 32 P]P i Cell Labeling-An in vitro Aurora-A kinase assay was performed as described previously (16,17) using recombinant AR as substrate, which is inactive. For in vivo labeling, HEK293 cells were transfected with FLAG-AR/HA-Aurora-A. After serum starvation overnight, cells were labeled with [ 32 P]P i (0.5 mCi/ml) in phenol red-free minimum Eagle's medium without phosphate for 4 h. FLAG-AR was then immunoprecipitated and separated in SDS-PAGE. After transferring to membrane, phospho-AR was detected and quantified.
Luciferase Reporter, Cell Proliferation, Apoptosis, and Chromatin Immunoprecipitation (ChIP) Assays-The luciferase reporter assay was carried out as described previously (17). Briefly, cells were transiently cotransfected with ARR3-Luc, Aurora-A, wild-type or mutant AR, and ␤-galactosidase. The amount of DNA in each transfection was kept constant by the addition of empty pcDNA3 vector. After 36 h of transfection, luciferase activity was measured using a luciferase assay reagent (Promega). Transfection efficiency was normalized by cotransfection with ␤-galactosidase expression vector. The experiment was repeated three times in triplicate.
For cell proliferation assay, cells were seeded in the 6-well (1 ϫ 10 5 cells/well) plate cultured in phenol red-free RPMI 1640 medium containing 10% charcoal-stripped FBS in the absence or presence of androgen. At days 0, 1, 2, 3, and 4, cells were trypsinized and counted. Apoptosis was determined using a cell death detection ELISAPLUS kit (Roche Applied Science) according to the manufacturer's protocol. The results are expressed as the enrichment factor relative to the untreated controls. The data represent the mean value of at least three independent experiments.
A ChIP assay was performed as described previously (17,28). Briefly, cells were cross-linked with 1% formaldehyde and sonicated. The soluble chromatins were immunoprecipitated with anti-AR or -FLAG antibody. The PCR primers from PSA promoter were as follows: 5Ј-AGGGATCAGGGAGTCTCACA-3Ј and 5Ј-GCTAGCACTTGCTGTTCTGC-3Ј. Notably, a higher level of Aurora-A was detected in androgen-independent LNCaP-RF than in parent LNCaP cells. B, AR is phosphorylated by Aurora-A in vitro. In vitro kinase was performed by incubation of recombinant AR with and without recombinant Aurora-A (top). Panels 2 and 3 are immunoblots showing the proteins used for in vitro kinase assay. C, Aurora-A phosphorylates AR in vivo. HEK293 cells were transfected with FLAG-AR together with and without HA-Aurora-A. After 36 h of transfection, cells were labeled with [ 32 P]P i (0.5 mCi/ml) in phenol red-free minimum Eagle's medium without phosphate and serum for 4 h. FLAG-AR was immunoprecipitated, separated on SDS-PAGE, and exposed (top). Expression of the transfected plasmids is shown in panels 2 and 3. Actin was used as a loading control (bottom). D and E, Aurora-A interacts with AR. LNCaP-RF cells were immunoprecipitated (IP) with anti-AR and detected with anti-Aurora-A antibody (D) and vice versa (E). NS, nonspecific band. Each experiment was repeated three times.
Statistical Analysis-Statistical significance was analyzed by unpaired Student's t test, and p Յ 0.05 was considered to be statistically significant.

Aurora-A Interacts with and Phosphorylates AR in Vitro and
in Vivo-Previous studies have shown frequent overexpression of Aurora-A in prostate cancer, which is associated with tumor progression and poor prognosis (22). Moreover, Aurora-A expression levels were shown to be correlated significantly with the value of the serum PSA and cell proliferative activity (23). Mounting evidence indicates that AR signaling plays a critical role in the initiation and progression of prostate cancer (22,23). We examined Aurora-A expression in a panel of prostate cancer cell lines. Immunoblotting analysis revealed that Aurora-A was elevated in LNCaP-RF (androgenindependent growth) when compared with parental LNCaP (androgen-dependent), whereas DU145 and PC3 (AR-negative/androgen-independent) also express high levels of Aurora-A (Fig. 1A). These findings prompted us to examine if Aurora-A regulates AR. We first performed an in vitro kinase assay by incubation of recombinant AR and Aurora-A. Fig.  1B shows that AR is phosphorylated by Aurora-A. To determine if Aurora-A phosphorylates AR in vivo, HEK293 cells, which are AR-negative and express low levels of Aurora-A (29,30), were co-transfected with FLAG-AR and HA-Aurora-A. After 36 h of transfection, cells were labeled with [ 32 P]orthophosphate for 4 h. Fig.  1C shows that Aurora-A phosphorylates AR in vivo. In addition to the centrosome, Aurora-A is also detected in the cytoplasm and the nucleus (22). Therefore, we examined whether Aurora-A interacts with AR. Co-immunoprecipitation was carried in LNCaP-RF cells. Endogenous Aurora-A was readily detected in AR immunoprecipitates ( Fig. 1D) and vice versa (Fig. 1E). These findings suggest that AR is a substrate of Aurora-A.
Thr 282 and Ser 293 of AR Are Phosphorylated by Aurora-A-Sequence analysis revealed that AR contains three Aurora-A phosphorylation consensus motifs ((R/K)X(S/T)(I/ V/L/P)) at residues Thr 282 , Ser 293 , and Thr 662 ( Fig. 2A). To determine if Aurora-A phosphorylates these sites, we created GST fusion proteins that contain wild type and Thr/Ser 3 Ala mutant for each putative phosphorylation site (e.g. GST plus 20 -30 amino acids of AR with a putative phosphorylation Thr or Ser site in the middle). An in vitro kinase assay revealed that wild-type Thr 282 and Ser 293 but not Thr 662 , T282A, and S293A were phosphorylated by Aurora-A ( Fig. 2A). To further confirm phosphorylation of Thr 282 and Ser 293 by Aurora-A, we created FLAG-tagged deletion mutants of AR by truncation of Thr 282 or/and Ser 293 sites (Fig. 2B, top) and then transfected them into HEK293 cells. Following immunoprecipitation with anti-FLAG antibody, an in vitro Aurora-A kinase assay was performed In vitro Aurora-A kinase was carried out by incubation of recombinant Aurora-A with each GST-fused wild-type and Thr/Ser 3 Ala-mutated AR motif as substrate (middle). Panels 3 and 4 are Western blots showing the proteins used for the kinase assay. B and C, further definition of the phosphorylation sites with truncated and point-mutated AR. HEK293 cells were transfected with the indicated truncation (B) and point mutation (C) of full-length AR and then immunoprecipitated with anti-FLAG antibody. The immunoprecipitates were used as substrates for in vitro Aurora-A kinase assay (top panels). Bottom panels are immunoblotting with anti-FLAG antibody showing expression of transfected plasmids. Note that the size of the nonspecific band (NS) is comparable with FLAG-AR-N3 (B). D, Aurora-A phosphorylates AR-Thr 282 /Ser 293 in vivo. HEK293 cells were transfected with the indicated plasmids and labeled with [ 32 P]P i (0.5 mCi/ml). FLAG-AR was immunoprecipitated, separated in SDS-PAGE, and exposed (top). Panels 2 and 3 show expression of the transfected AR and Aurora-A. E, knockdown of Aurora-A reduces phosphorylation of AR. PC3 cells were transfected with FLAG-AR and shRNA of Aurora-A, labeled with [ 32 P]P i (0.5 mCi/ml), and analyzed as described in D (top). Panels 2-4 are Western blots hybridized with the indicated antibodies.
using FLAG-AR immunoprecipitates as substrates. Fig. 2B shows that Aurora-A phosphorylates fragments N2 (Thr 282 ) and N3 (Thr 282 and Ser 293 ) but not fragment N1 in which both Thr 282 and Ser 293 are deleted. Moreover, we mutated Thr 282 or/and Ser 293 to alanine in full-length FLAG-AR. An in vitro Aurora-A kinase assay using FLAG-AR immunoprecipitates as substrates revealed that mutation of either Thr 282 or Ser 293 decreased the phosphorylation by Aurora-A compared with wild-type AR. However, Aurora-A could not phosphorylate AR-T282A/S293A (Fig. 2C). To further demonstrate Aurora-A phosphorylation of AR at these two sites, we performed in vivo labeling by transfection of HEK293 cells with FLAG-AR and AR-T282A/S293A together with and without Aurora-A. Fig. 2D shows that Aurora-A phosphorylates wild-type AR but not AR-T282A/S293A. In addition, in vivo labeling shows that knockdown of Aurora-A considerably reduces the AR phosphorylation in PC3 cells (Fig. 2E). Thus, we concluded that Thr 282 and Ser 293 are major AR sites that are phosphorylated by Aurora-A.
Aurora-A Induces AR Transcriptional Activity-Because Aurora-A phosphorylates AR in vitro and in vivo, we next investigated whether AR transactivation activity is regulated by Aurora-A. We introduced AR-responsive reporter (ARR3-Luc) into AR-positive LNCaP and LNCaP-RF and AR-negative PC3 and DU145 lines, which express different levels of Aurora-A (Fig. 1A). A luciferase assay revealed that the promoter activity in LNCaP-RF is 6-fold higher than that in LNCaP. Because DU145 and PC-3 are ARnegative, no ARR3-Luc activity was observed, although they expressed high levels of Aurora-A (Fig. 3A), suggesting that AR transactivation activity is regulated by Aurora-A.
To further demonstrate the importance of Aurora-A in AR transactivation activity, LNCaP cells were co-transfected with ARR3-Luc and wild-type or dominant-negative Aurora-A. After culture for 48 h in phenol red-free medium supplemented with 10% charcoal-treated FBS, we performed a luciferase assay and observed that the ARR3-Luc activity was induced by wildtype Aurora-A about 7-fold but was suppressed by dominant-negative Aurora-A (Fig. 3B). Moreover, we examined whether Aurora-A potentiates androgen-stimulated AR activity. Following transfection with ARR3-Luc and Aurora-A or pcDNA3 vector, LNCaP cells were treated with and without low concentrations of R1881 (0.05 and 0.1 nM) and then assayed for luciferase activity. As expected, AR activity was induced by androgen about 10-fold at 0.1 nM concentration. Further, ectopic expression of Aurora-A significantly enhanced androgen-induced AR transactivation activity (p Ͻ 0.05), which was comparable with that of constitutively active SrcY527F (7), whereas dominant-negative Aurora-A reduced the androgen action (Fig. 3C).

. Aurora-A induces AR activity and potentiates androgen action in AR.
A, basal level of AR activity is higher in LNCaP-RF than LNCaP cells. The indicated cells were transfected with ARR3-Luc and ␤-galactosidase. Following 48 h of incubation, luciferase activity was measured and normalized to ␤-galactosidase. Results are the mean Ϯ S.E. (error bars) of three independent experiments performed in triplicate. B, wild-type Aurora-A stimulates and DN-Aurora-A inhibits AR activity. LNCaP cells were transfected with indicated plasmids and assayed for luciferase activity as described above. C, androgen-induced AR activity is enhanced by wild-type Aurora-A and inhibited by DN-Aurora-A. LNCaP cells were transfected with the indicated plasmids. After 36 h of incubation, cells were treated with and without R1881 and then subjected to a luciferase assay. D, expression of Aurora-A in AR-negative HEK293 cells had no effect on ARR3-Luc activity; however, co-expression of AR and Aurora-A stimulates the reporter activity. HEK293 cells were transfected with the indicated plasmids and assayed for luciferase activity. The bottom panels of A-D show the expression of transfected plasmids. E, knockdown of Aurora-A reduces AR transactivation activity. LNCaP cells were transfected with Aurora-A shRNA and control shRNA. After treatment with and without R1881, cells were subjected to a luciferase assay. Each experiment was repeated three times in triplicate. *, p Ͻ 0.05.
Subsequently, we investigated if AR is directly regulated by Aurora-A by transfection of HEK293 cells with AR, ARR3-Luc, and Aurora-A. A luciferase assay revealed that expression of AR but not Aurora-A alone induces ARR3-Luc activity. However, the reporter activity is stimulated by co-expression of AR and Aurora-A in a dose-dependent manner (e.g. ϳ3.0-fold). Dominant negative Aurora-A reduces AR-induced ARR3-Luc activ-ity (Fig. 3D). In addition, knockdown of Aurora-A reduces basal and androgen-induced AR activity more than 50% in LNCaP cells (Fig.  3E). Collectively, these findings indicate that Aurora-A activates AR and potentiates androgen-induced AR activity.
Aurora-A Transactivation of AR Depends on Thr 282 /Ser 293 Phosphorylation-Both Thr 282 and Ser 293 are located in the transactivation domain of AR ( Fig. 2A). Thus, we next examined if Aurora-A transactivation of AR relies on the phosphorylation of T282A/S293A. Because AR-negative PC3 cells express a high level of Aurora-A (Fig. 1A), we co-transfected AR, AR-2A, or AR-2D with ARR3-Luc into PC3 cells and compared the luciferase activities. After 48 h of incubation in phenol redfree/10% charcoal-treated FBS medium, a luciferase assay revealed that expression of wild-type AR and phosphomimic AR-2D induces ARR3-Luc activity, whereas AR-2A has much less effect on the reporter activity. Moreover, transactivation activity of wild-type AR and AR-2D but not AR-2A was enhanced by R1881 treatment (Fig. 4A).
To further demonstrate Aurora-A activation of AR through phosphorylation of Thr 282 and Ser 293 , HEK293 cells were transfected with ARR3-Luc, different forms of AR, and increasing amounts of Aurora-A. As shown in Fig. 4B, ectopic expression of Aurora-A induced AR but not AR-2A and AR-2D transactivation activity. The activity of AR-2D is comparable with that of AR plus Aurora-A. These data suggest that phosphorylation of Thr 282 and Ser 293 not only is required for Aurora-A activation of AR but also modulates androgen response.

Aurora-A Regulates PSA Expression, AR DNA Binding Activity, and Androgen-independent
Growth-PSA is a major androgen-stimulated gene that is used to monitor treatment response and progression in patients with prostate cancer. A previous study has shown that Aurora-A expression correlates with PSA levels and cell-proliferative activity (23). In addition, phosphorylation of AR by protein kinases, such as Src, Akt, and MAPK, has been demonstrated to associate with prostate cancer androgen-independent growth

FIGURE 5. Ectopic expression of Aurora-A in LNCaP cells induces a PSA-and androgen-independent phenotype.
A and B, effect of Aurora-A and/or androgen on PSA expression. LNCaP cells were stably transfected with HA-Aurora-A or pHM6 vector and immunoblotted with the indicated antibody (A). Aurora-A-and pHM6-transfected cells were treated with and without R1881 and subjected to RT-PCR (top panels) and immunoblotting (bottom panels) analysis (B). C and D, overexpression of Aurora-A renders LNCaP cells androgenindependent. Aurora-A-and pHM6-transfected LNCaP cells were cultured in 24-well plates with phenol redfree RPMI 1640 medium containing 10% charcoal-stripped serum in the absence or the presence of androgen. The cell number was counted in the indicated time. *, p Ͻ 0.05 (C). The same cells were cultured in either regular FBS or charcoal-stripped serum (CS) for 3 days and examined for apoptosis using a Cell Death Detection ELISAPLUS kit (D). Error bars, S.E. (5-7, 31, 32). Thus, we reasoned that Aurora-A phosphorylation of AR could regulate PSA expression and androgen-independent growth. To test this, we ectopically expressed Aurora-A in androgen-dependent LNCaP cells. After G418 selection, stable Aurora-A and pHM6 vector-transfected LNCaP cell lines were established (Fig. 5A). Following treatment with and without R1881, cells were subjected to semiquantitative RT-PCR and immunoblotting analyses. Fig. 5B shows that expression of Aurora-A or treatment with R1881 induces PSA expression. Combination of Aurora-A expression with R1881 treatment increases PSA expression more significantly than either one alone. Further, we examined if expression of Aurora-A renders LNCaP cells capable of androgenindependent growth. The cells were cultured in phenol red-free RPMI 1640 medium containing 10% charcoal-stripped serum in the absence or presence of androgen. A cell proliferation assay revealed that depletion of androgen inhibits pHM6 vector-transfected cell growth. Expression of Aurora-A overrode the androgen depletion effect on cell growth. Notably, treatment of Aurora-A-transfected cells with R1881 induces cell proliferation even more robust than cells treated/transfected with R1881 or Aurora-A alone (Fig. 5C). In addition, we com-pared androgen depletion-induced apoptosis in Aurora-A-and pHM6transfected LNCaP cells. As shown in Fig. 5D, the cells expressing Aurora-A largely abrogated the androgen depletion-induced programmed cell death. To further demonstrate the role of Aurora-A in AR-mediated PSA expression and androgen-independent growth, Aurora-A was knocked down in LNCaP-RF cells. Cells treated with scramble shRNA were used as controls (Fig. 6A). The knockdown of Aurora-A considerably reduces PSA expression, cell growth, and cell survival (Fig. 6, A-C). Collectively, these data suggest that Aurora-A regulates androgen/AR-mediated PSA expression and cell growth.
Because Aurora-A phosphorylation of AR-Thr 282 /Ser 293 induces AR activity, we reasoned that Aurora-A-phosphorylated AR-2D promotes but nonphosphorylatable AR-2A reduces cell proliferation induced by androgen. To test this, AR-negative PC3 cells were transfected with AR, AR-2A, or AR-2D. Following culture for 4 days in phenol red-free medium containing 10% charcoal-stripped serum in the presence and absence of R1881, the number of cells was counted. Fig.  6D shows that expression of AR-2D but not AR-2A induces cell growth both in the presence and absence of androgen, whereas, as expected, wild-type AR only promotes cell proliferation in the presence of R1881. Moreover, AR-2A largely abrogates androgen-induced cell growth. To further ascertain the significance of Aurora-A-phosphorylated AR in prostate cancer cell growth, we knocked down AR in LNCaP cells and then reconstituted with different forms of AR (Fig.  6E). Depletion of AR largely reduced R1881-stimulated cell growth. Reintroduction of wild-type AR rescued the effect of knockdown of endogenous AR. Further, expression of AR-2D but AR-2A significantly stimulated cell growth both in the absence and presence of androgen (Fig. 6E). These results further support the notion that Aurora-A phosphorylation of AR enhances AR-mediated cell proliferation and could play an important role in androgen-independent growth.
We also investigated if Aurora-A regulates the DNA binding activity of AR. Chromatin immunoprecipitation revealed that ectopic expression of Aurora-A in LNCaP cells induces, whereas knockdown of Aurora-A in LNCaP-RF decreases, endogenous AR binding to the PSA promoter (Fig. 7A), which indicates Aurora-A regulation of AR DNA-binding activity and further supports our findings that Aurora-A regulates PSA

FIGURE 6. Knockdown of Aurora-A in LNCaP-RF cells reduces PSA expression and androgen-independent phenotype; Aurora-A-phosphorylated AR induces cell growth in the presence and absence of androgen.
A, knockdown of Aurora-A decreases PSA expression. LNCaP-RF cells were transfected with shRNA-Aurora-A and scramble shRNA and detected for expression of Aurora-A (top two panels) and PSA (panels 3 and 5) as described in the legend to Fig. 5. Actin was used as a control. B and C, knockdown of Aurora-A reduces LNCaP-RF androgen-independent phenotype. Aurora-A/shRNA-and scramble shRNA-treated LNCaP-RF cells were cultured in phenol red-free RPMI 1640 medium containing 10% charcoal-stripped serum and then assayed for cell growth (B) and apoptosis (C). D and E, AR-2D but not AR-2A induces cell proliferation in the absence and presence of R1881. PC3 (D) and LNCaP (E) cells were seeded in a 6-well plate (10 5 cells/well) and then transfected with the indicated plasmids (bottom panels). Following incubation for 4 days in phenol redfree RPMI 1640 medium containing 10% charcoal-stripped serum in the absence or the presence of R1881, cell number was counted. The experiments were repeated three times in triplicate. *, p Ͻ 0.05. Error bars, S.E. expression (Figs. 5 and 6). We next examined if the AR DNA binding activity regulated by Aurora-A depends on phosphorylation of Thr 282 /Ser 293 . HEK293 cells were transfected with different forms of AR together with and without Aurora-A. Fig.  7B shows that the DNA binding activity of AR-2A is lower than AR, whereas phosphomimic AR-2D exhibits higher activity than AR and AR-2A. Ectopic expression of Aurora-A induces AR but not AR-2A DNA binding. Thus, we conclude that Aurora-A regulates AR DNA binding activity through phosphorylation of Thr 282 and Ser 293 .
In addition, we have selected 62 AR-positive prostate cancers, 7 of which were resistant to anti-androgen therapies, and examined Aurora-A expression. Immunoblotting and immunohistochemical staining analysis revealed high levels of Aurora-A in five resistant tumors (e.g. 5 of 7, 71%) as well as in 22 of the remaining specimens (e.g. 22 of 55, 40%; Fig. 7, C and D). These data further support the notion that elevated Aurora-A contributes to androgen-independent growth through regulation of AR.

DISCUSSION
Previous studies have demonstrated frequent overexpression of Aurora-A in prostate cancer (22)(23)(24)(25). In particular, elevated levels of Aurora-A are associated with PSA value, late stage and high grade tumors, and poor prognosis (22,23). In this report, we demonstrate that Aurora-A phosphorylates and activates AR in vitro and in vivo. As a result, Aurora-A potentiates andro-gen-stimulated AR and PSA expression. Further, we show that Aurora-A is elevated in androgenindependent LNCaP-RF cells. Ectopic expression of Aurora-A in androgen-dependent LNCaP cells induces cell proliferation and cell survival in the absence of androgen. In contrast, knockdown of Aurora-A in LNCaP-RF cells inhibits cell growth and promotes apoptosis. These findings are important for several reasons. First, they provide a direct link between Aurora-A and the androgen/AR pathway. Second, the mechanism by which elevated Aurora-A is associated with high levels of PSA-and androgen-refractory prostate cancer has now been uncovered. Finally, they further support the recent findings of Aurora-A function outside centrosome (33,34).
The critical role of androgen and AR in prostate cancer development has been well documented. It has been shown that ligand binding induces AR phosphorylation at multiple sites, primarily serine/ threonine residues at the N-terminal transactivation domain, which include Ser 16 , Ser 81 , Ser 213 , Ser 256 , and Ser 308 . Additionally, phosphorylation of Ser 424 , Ser 650 , and Ser 791 was also described in response to androgen (1,5,6,31,32,35). Our study shows that Aurora-A phosphorylates AR on Thr 282 and Ser 293 in the transactivation domain. While they are not androgen-responsive sites, the phosphorylation of Thr 282 and Ser 293 potentiates androgen-induced AR activation. Nonphosphorylatable Thr 282 /S293A mutation reduces androgenstimulated AR activity (Fig. 3). These findings suggest that phosphorylation of Thr 282 and Ser 293 by elevated Aurora-A enhances androgen/AR action in prostate cancer development and progression.
Androgen deprivation is an effective therapy in AR-positive prostate cancer by inducing apoptosis and cell growth arrest; however, in the majority of patients, the cancer eventually loses its dependence on androgen and progresses to an androgenindependent state. Various mechanisms have been postulated to account for the conversion of prostate cancer into a castration-resistant state. One of them is aberrant phosphorylation and activation of AR by protein kinases, which include serine/ threonine kinases Erk1/2 (Ser 514 ) (36), Akt (Ser 213 and Ser 791 ) (8,38,39), Cdk1 (Ser 81 ) (40), and Src family kinases (Tyr 534 ) (7). These phosphorylation events modulate AR activity and result in androgen-independent growth (1,31,32,40). The AR has an N-terminal transactivation domain with a hormone-independent transcriptional activation function (NTD/AF-1), a central DNA-binding domain, and a C-terminal ligand-binding do-main with a hormone-dependent transcriptional activation function (AF-2) (41,42). Because Thr 282 and Ser 293 locate in the NTD region ( Fig. 2A), activation of AR by Aurora-A phosphorylation of these two sites was shown to be ligand-independent (Figs. 3D and 6, D and E). However, the phosphorylation potentiates androgen-stimulated AR activity (Figs. 3C, 4A, and 5, B and C). In addition, we demonstrate that androgen-independent LNCaP-RF expresses an elevated level of Aurora-A compared with parental LNCaP. Knockdown of Aurora-A in LNCaP-RF cells decreases PSA expression and AR DNA binding activity (Figs. 6 and 7). Thus, Aurora-A phosphorylation and activation of AR contributes to androgen-independent growth, and elevated levels of Aurora-A could play a pivotal role in development of hormone refractory prostate cancer. This notion is further supported by our finding that the frequency of Aurora-A overexpression in androgen ablation-resistant tumors is much higher than that in the responsive cases (71% versus 40%).
Accumulated studies have shown that Aurora-A plays an important role in promoting timely mitotic entry by controlling initial centrosomal activation of cyclin B/Cdk1 (43,44). Recently, it was shown that Aurora-A promotes the G 2 activation of Polo-like kinase-1 (Plk1) through direct phosphorylation within the T-loop of Plk1 (45,46). Active Plk1 controls cyclin B/Cdk1 activity through the phosphorylation of Cdc25B phosphatase and the degradation of the Cdk1 inhibitory kinase Wee1 (47,48). We previously showed that Aurora-A phosphorylates p53 in the centrosome and abrogates p53 DNA binding activity (17). Although Aurora-A is a centrosome kinase, it has been shown to phosphorylate the molecules outside the centrosome, including RalA, TRF1, and HDAC6 (33,34,49). Although AR is an important substrate of Aurora-A, Aurora-A function in prostate cancer development and progression must be the sum of its substrates inside and outside the centrosome. This perception is evidenced by the observation of elevated levels of Aurora-A in AR-negative and androgen-independent DU145 and PC3 cells (Fig. 1). Finally, Aurora-A inhibitors are currently used in clinical trials (37). Our findings indicate that inhibition of Aurora-A or the combination with other therapeutic modalities could have great potential to overcome hormone-refractory prostate cancer.