Androgen Receptor Serine 81 Phosphorylation Mediates Chromatin Binding and Transcriptional Activation*

Background: AR Ser-81 phosphorylation correlates with transcriptional activity and can be mediated by CDK9 and CDK1, but its function is unknown. Results: Chromatin-associated AR is enriched for Ser-81 phosphorylation, and an S81A mutation abrogates AR transcription and chromatin binding. Conclusion: Ser-81 phosphorylation is required for AR chromatin binding. Significance: These findings identify a critical function for Ser-81 phosphorylation and a mechanism through which CDK1 may enhance AR activity. Our previous findings indicated that androgen receptor (AR) phosphorylation at serine 81 is stimulated by the mitotic cyclin-dependent kinase 1 (CDK1). In this report, we extended our previous study and confirmed that Ser-81 phosphorylation increases during mitosis, coincident with CDK1 activation. We further showed blocking cell cycle at G1 or S phase did not disrupt androgen-induced Ser-81 phosphorylation and AR-dependent transcription, consistent with a recent report that AR was phosphorylated at Ser-81 and activated by the transcriptional CDK9. To assess the function of Ser-81 phosphorylation in prostate cancer (PCa) cells expressing endogenous AR, we developed a ligand switch strategy using a ligand-binding domain mutation (W741C) that renders AR responsive to the antagonist bicalutamide. An S81A/W741C double mutant AR stably expressed in PCa cells failed to transactivate the endogenous AR-regulated PSA or TMPRSS2 genes. ChIP showed that the S81A mutation prevented ligand-induced AR recruitment to these genes, and cellular fractionation revealed that the S81A mutation globally abrogated chromatin binding. Conversely, the AR fraction rapidly recruited to chromatin after androgen stimulation was highly enriched for Ser-81 phosphorylation. Finally, inhibition of CDK1 and CDK9 decreased AR Ser-81 phosphorylation, chromatin binding, and transcriptional activity. These findings indicate that Ser-81 phosphorylation by CDK9 stabilizes AR chromatin binding for transcription and suggest that CDK1-mediated Ser-81 phosphorylation during mitosis provides a pool of Ser-81 phosphorylation AR that can be readily recruited to chromatin for gene reactivation and may enhance AR activity in PCa.

Androgens and the androgen receptor (AR) 2 play essential roles in the development and progression of prostate cancer (PCa). AR can be structurally divided into an N-terminal domain (NTD) that harbors a major transcriptional activation function (AF-1), a central DNA-binding domain (DBD), a hinge region that contains a nuclear localization signal, and a C-terminal ligand-binding domain (LBD) that has a minor transcriptional activation function (AF-2) (1). Binding of androgen (testosterone or dihydrotestosterone (DHT)) causes a conformational change in the LBD and generates a hydrophobic cleft that initially binds a hydrophobic helix in the AR NTD (AR N-C interaction) and subsequently mediates binding of LXXLL motifs in transcriptional coactivator proteins. The androgen-liganded AR forms homodimers and accumulates in the nucleus, where it binds to androgen-responsive elements (AREs) in the enhancers of androgen-regulated genes and recruits coactivator proteins through interactions with both the NTD and LBD (2,3). Androgen deprivation therapy (surgical or medical castration) is used for the initial systemic treatment of PCa, but the disease invariably recurs, and these castrationresistant prostate cancers (CRPCs) are generally more aggressive. Many patients with CRPC will respond to secondary hormonal therapies with AR antagonists or agents that further suppress androgen synthesis, in particular the recently FDA-approved CYP17A1 inhibitor abiraterone, but most of these patients still relapse within a year (4,5). Significantly, the AR remains highly expressed and transcriptionally active in these advanced tumors, but the molecular mechanisms contributing to this activity at low androgen levels are poorly understood (6). AR is phosphorylated constitutively at Ser-94 in the NTD and Ser-650 in the hinge region prior to androgen exposure (7)(8)(9). Androgen stimulates AR phosphorylation at multiple additional sites in the NTD, including serines 81, 256, 308, and 424, and increases Ser-650 phosphorylation (9). Recent reports indicate that Ser-650 phosphorylation is stimulated by a stressinduced kinase pathway and enhances AR nuclear export and that this site is dephosphorylated by protein phosphatase 1 (10,11). However, the functional significance of other androgenstimulated phosphorylation sites and the kinase pathways regulating phosphorylation of these sites remain to be fully defined (8,(12)(13)(14)(15). Previous mass spectrometry studies have shown that Ser-81 is the most highly phosphorylated site in response to androgen stimulation and shown that Ser-81-phosphorylated AR (Ser(P)-81-AR) accumulates gradually over ϳ8 h in PCa cells after androgen stimulation (9,16). We have reported that CDK1 activity was increased in CRPC (17) and that CDK1 could phosphorylate AR at Ser-81 and sensitize AR to low levels of androgens (16). A recent study found that, in addition to CDK1, AR was associated with CDK5 and CDK9 and that CDK9 could also phosphorylate AR at Ser-81 (18). In contrast to CDK1, which is activated in mitosis, CDK9 associates with cyclin T to form the P-TEFb complex that stimulates transcriptional elongation through phosphorylation of substrates, including Ser-2 in the RNA polymerase II C-terminal domain (19). Significantly, this latter study showed that an S81A mutant AR was less effective at stimulating growth and, when expressed in AR-negative PCa cells, was decreased in its ability to transactivate a subset of AR-regulated genes (18).
In this report, we first extend our previous studies by establishing that AR Ser-81 phosphorylation is increased during the M phase of the cell cycle and mediated by CDK1 activation. Moreover, by preventing androgen-stimulated PCa cells from progressing to G 2 /M, we confirmed that Ser-81 also can be phosphorylated independently of CDK1 activation, consistent with the findings showing CDK9-mediated Ser-81 phosphorylation (18). To address the functional significance of Ser-81 phosphorylation in AR-expressing PCa cells, we exploited a W741C mutation in the AR LBD that allows the LBD to fold into the agonistic conformation and AR to be activated by the antagonist bicalutamide (20). Using this ligand switch strategy to stably express and selectively activate a S81A mutant in LNCaP PCa cells expressing endogenous AR, we found that the S81A/W741C double mutant versus the W741C single mutant AR was unable to stimulate endogenous AR-regulated genes in response to bicalutamide. Chromatin immunoprecipitation (ChIP), in conjunction with immunofluorescence and cellular fractionation studies, showed that the molecular basis for this defect was an inability of the S81A mutant to bind stably to chromatin. By ChIP and biochemical analyses, we further found that AR tightly associated with chromatin was enriched for Ser-81 phosphorylation. In addition, both CDK1 and CDK9 antagonists decreased AR Ser-81 phosphorylation, chromatinbinding, and transactivation. Based on these findings, we propose that Ser-81 phosphorylation by CDK9 in non-mitotic cells stabilizes AR binding to chromatin and is required for the subsequent chromatin remodeling and transcriptional activation. Moreover, we suggest that Ser-81 phosphorylation by CDK1 during mitosis provides a pool of Ser(P)-81-AR that can be rapidly recruited to AR-regulated genes when the cells enter G 0 /G 1 . Through this mechanism, increased CDK1 activity may contribute to AR activation in CRPC.
Reporter and Real-time RT-PCR Assays-The luciferase reporter and real-time RT-PCR analyses were described previously (10,16).
DNA Mutagenesis, Generation of Stable Lines, Immunoblotting, and Immunofluorescence Analyses-These experiments were performed as described previously (10).
Cellular Fractionation Assay-Cellular fractionation analysis using the NE-PER kit was carried out as described by the manufacturer (catalog no. 78835, Pierce), and using the Triton lysis buffer (TLB) (10). For extraction, LNCaP cells grown in a 10-cm dish were washed once in cold PBS, harvested for resuspension in 800 l of TLB containing protease inhibitor and phosphatase inhibitor but no salt (NaCl), vortexed, and kept on ice for 15 min, followed by centrifugation at 11,000 ϫ g for 5 min. The supernatant was removed and saved, and the pellet was resuspended in 100 l of TLB containing 50 mM of NaCl and gently shaken for 10 min in at 4°C, followed by centrifugation at 11,000 ϫ g for 5 min. The pellet was similarly extracted for several more rounds, in sequential steps with TLB containing increased NaCl up to 600 mM, as indicated. The micrococcal nuclease digestion was performed as follows. The insoluble fraction (pellet left from cytoplasmic and nuclear protein extraction) was suspended in 40 l of micrococcal nuclease digestion buffer (M0247, New England Biolabs), kept at room temperature, and treated with 600 units of micrococcal nuclease for 2 min. The reaction was stopped by adding EGTA (5 mM, final concentration), followed by centrifugation at 11,000 ϫ g for 5 min.
Statistical Analysis-Results are represented as mean Ϯ S.D. for replicate samples. Data are representative of at least three experiments. Significant differences (p Ͻ 0.05 in Student's t test) are indicated (*) in the experiments.

RESULTS
AR Ser-81 Phosphorylation Is Mediated Physiologically by CDK1 during Mitosis-Our previous report showed that AR Ser-81 phosphorylation could be increased by co-expression with CDK1 or a constitutively active CDK1 mutant (16). CDK1 is activated physiologically during mitosis. Therefore, to determine whether AR is phosphorylated at Ser-81 by endogenous CDK1 during mitosis, we used nocodazole to arrest LNCaP PCa cells in the G 2 /M phase of the cell cycle. As shown in Fig. 1A, nocodazole led to a time-dependent accumulation of cells in G 2 /M, with an increase in the G 2 /M fraction occurring within 2 h and the majority of the cells arrested in G 2 /M by 24 h. CDK1 activation, as indicated by phosphorylation at Thr-161, was also induced by 2 h and was markedly increased by 24 h (Fig. 1B). Coincident with the time courses of G 2 /M arrest and CDK1 activation, AR phosphorylation at Ser-81 also increased progressively over the 24-h period of nocodazole treatment. In contrast, there was no increase in AR phosphorylation at Ser-213, a site that is phosphorylated independently of cell cycle by AKT (24).
To address further whether Ser-81 phosphorylation during mitosis was mediated by CDK1, we challenged nocodazole-treated LNCaP cells with two different CDK1 inhibitors, CGP74514 and RO-3306 (25,26). Treatment of nocodazolearrested cells with CDK1 inhibitors led to time-dependent reduction in nocodazole-induced Ser-81 phosphorylation (Fig.  1C). Consistent with our previous findings, the down-regulation of Ser(P)-81-AR by CDK1 inhibition was also associated with reduced levels of AR protein (16). However, quantification of the ratio of Ser(P)-81-AR over total AR indicated that the reduction in Ser-81 phosphorylation was predominant and not solely due to decreased AR protein levels. Together, these findings supported the conclusion that AR Ser-81 is phosphorylated by CDK1 during mitosis.
S or G 1 Phase Arrest Does Not Prevent DHT-stimulated AR Ser-81 Phosphorylation and Transcriptional Activation-Previous studies have shown that androgen-induced AR Ser-81 phosphorylation increases progressively and does not reach maximal levels until ϳ8 h (9,16). Because androgen treatment also stimulates androgen-deprived PCa cells to cross the G 1 /S checkpoint and enter the cell cycle (27)(28)(29), we considered that the DHT-stimulated increase in Ser-81 phosphorylation could reflect cells progressing to G 2 /M phase. Therefore, to determine whether the DHT-stimulated increase in Ser(P)-81-AR was dependent on cell cycle progression, we used hydroxyurea to arrest DHT-stimulated LNCaP cells in S phase. Fig. 2A shows that hydroxyurea treatment led to a time-dependent accumulation of cells in S phase. Immunoblotting analysis showed that AR protein, which is increased by DHT due to increased stability of the androgen-liganded AR, was decreased by hydroxyurea treatment in the absence and presence of DHT (Fig. 2B). However, DHT still increased Ser(P)-81-AR in the hydroxyureatreated cells, indicating that Ser-81 phosphorylation is not dependent upon cell cycle progression to M phase. Overall AR transcriptional activity also appeared to be repressed by hydroxyurea, based on immunoblotting for PSA protein (Fig.  2B) and RT-PCR for the AR-regulated PSA and TMPRSS2 genes (Fig. 2C). These decreases in AR expression and activity may reflect S phase arrest or other effects of hydroxyurea, but the results still indicated that the androgen-stimulated increase in Ser(P)-81 was not dependent on cells reaching M phase.
The observation that AR Ser-81 phosphorylation is not dependent on cell cycle progression to M phase was further verified by treating LNCaP cells with a reversible cell cycle inhibitor, mimosine. This amino acid analog reversibly represses the initiation of DNA replication and synchronizes cells in G 1 phase of the cell cycle (30). As shown in Fig. 2D, mimosine treatment had no pronounced effect on DHT-stimulated AR protein expression and transcriptional activity and, most importantly, Ser-81 phosphorylation. Together these findings demonstrated that Ser-81 can be phosphorylated in response to androgen stimulation independently of cell cycle progression and CDK1 activation and were consistent with the recent study identifying CDK9 as another mediator of Ser-81 phosphorylation (18).

AR Transcriptional Activity in Transient Transfection Assays Is Not Enhanced by CDKs and Is Not Dependent on Ser-81
Phosphorylation-As noted above, a recent study confirmed an association between AR and CDK1 and further showed that CDK9 was associated with AR and could mediate Ser-81 phosphorylation (18). Another transcriptional CDK, CDK7, also has been reported to phosphorylate AR (at Ser-515) and increase AR transcriptional activity (31). To assess and compare the effects of these CDKs, we carried out AR-dependent reporter assays in LNCaP cells with co-transfected transcriptional CDKs (CDK7, -8, and -9 (32)) or activated CDK1 (CDK1-AF (16)). As shown in Fig. 3A, all tested CDKs failed to enhance the DHTstimulated activity of AR-regulated ARE4-Luc or PSA-Luc reporter genes in LNCaP cells. We further determined that substitution of Ser-81 by alanine (S81A) did not detectably impair DHT induction of ARE4-Luc or PSA-Luc reporter genes in LNCaP cells, consistent with previous reporter gene studies of this mutant in AR-deficient cells (Fig. 3B) (8,16). We further examined whether Ser-81 phosphorylation has an impact on the androgen-mediated interaction between the AR NTD and LBD (AR N-C interaction), which is required for AR-mediated transcription and chromatin interaction (33,34). Using mammalian two-hybrid protein interaction assays with Gal4-AR-LBD and VP16-AR-NTD versus VP16-AR-NTD(S81A), we found that the S81A mutant has comparable N-to-C activity as its wild-type counterpart (Fig. 3C).
Ser-81 Phosphorylation Is Required for Induction of Endogenous Androgen-regulated Genes-Although previous studies and the studies described above have shown that the S81A mutation does not impair the ability of AR to stimulate transiently transfected ARE-regulated reporter genes, it is well

Ser-81 Phosphorylation Is Required for AR Nuclear Functions
established that these assays may not effectively reflect the activities of transcription factors on endogenous genes. Indeed, a recent study examining a transfected S81A mutant AR in AR negative PCa cells found a decreased ability to transactivate a subset of endogenous AR regulated genes (18). An approach that has been used to assess the function of a mutant AR in PCa cells expressing endogenous AR has been to express shRNA targeting endogenous AR in conjunction with a stably expressed mutant AR that is not targeted by the shRNA, an approach referred to as the codon switch strategy (35). An alternative approach we employed is based on a well characterized mutation in the AR LBD (W741C) that results in activation by bicalutamide, a clinically used antagonist for the wild-type AR and for the T877A AR mutant expressed by LNCaP cells (20,36). Importantly, crystal structure analyses of the AR LBD have shown that the bicalutamide-liganded AR W741C mutant folds in the agonist conformation that generates the coactivator binding site, similarly to the androgen-liganded wild-type AR LBD (37,38). The incorporation of this mutation into an AR that is either wild type or mutated at other sites, which we refer to as a ligand switch strategy, allows us to assess selectively AR mutants by stimulating with bicalutamide without interference from the endogenous AR that is not activated by bicalutamide.
For the ligand switch approach, we generated FLAG epitopetagged AR expression vectors that were wild type, W741C single mutant, or W741C/S81A or W741C/S81D double mutant (Fig. 4A). Transient transfection with an AR-driven reporter gene (ARE4-Luc) in AR-deficient PC3 cells or AR-expressing LNCaP cells confirmed that bicalutamide could potently activate the W741C mutant AR but not the wild-type AR, whereas  MARCH 9, 2012 • VOLUME 287 • NUMBER 11 both were stimulated by DHT (Fig. 4B). We next tested the S81A/W741C double mutant by transient transfection in LNCaP cells and found that the AR-S81A/W741C and W741C control yielded similar reporter gene activities (ARE4-Luc or PSA-Luc) in response to bicalutamide (Fig. 4C). An AR with an S81D mutation to mimic Ser-81 phosphorylation (AR-S81D/ W741C) also showed comparable activity (Fig. 4C). Using a mammalian two-hybrid protein interaction assay based on the W741C backbone, we also verified that the wild-type AR NTD linked to the VP16 transactivation domain and the corresponding S81A and S81D mutants interacted equally with the bicalutamide-liganded AR-W741C LBD (Fig. 4D). These findings further confirmed that Ser-81 phosphorylation is not essential for AR activation or N-to-C interaction in exogenous reporter assays.

Ser-81 Phosphorylation Is Required for AR Nuclear Functions
We next used this ligand switch strategy to assess the requirement for Ser-81 phosphorylation in regulation of endogenous AR-stimulated genes. LNCaP stable lines expressing the W741C mutant AR versus the S81A/W741C double mutant AR were generated. These cells were cultured for 3 days in steroid-depleted medium (RPMI 1640 with 10% charcoal/ dextran-stripped FBS (CDS)) to suppress activity of the endogenous and transfected ARs and were then treated with bicalutamide to selectively activate the transfected AR or with DHT to activate both the endogenous and transfected ARs. Quantitative real-time RT-PCR showed that endogenous AR-regulated genes (PSA and TMPRSS2) could be stimulated by bicalutamide in the W741C single mutation control but not in the W741C/ S81A double mutant cells (Fig. 4E). Importantly, these genes could be activated in both lines by DHT, which stimulates both the endogenous and transfected ARs (Fig. 4E). Fig. 4F shows expression of the transfected AR (detected by both anti-FLAG antibody and anti-AR antibody; upper arrowhead) and endogenous AR (detected by an anti-AR antibody only; lower band) in these stable lines and further shows that bicalutamide stimulates Ser-81 phosphorylation of the W741C single mutant control but not the double mutant or the endogenous AR. Overall, although the S81A mutant AR is expressed at lower levels, its inability to stimulate PSA or TMPRSS2 gene expression suggested a critical function of Ser-81 phosphorylation in AR-mediated transcription of endogenous but not exogenous genes.
Ser-81 Phosphorylation Is Required for AR Chromatin Binding-To identify the molecular mechanisms by which Ser-81 phosphorylation may regulate AR transcription, we next carried out ChIP to assess effects of the Ser-81 mutation on binding to AREs in the enhancer locus of endogenous AR-regulated genes. We initially examined an LNCaP stable cell line expressing FLAG-tagged wild-type AR. As shown in Fig. 5A, using anti-FLAG antibody, we confirmed DHT-stimulated binding of the transfected AR to the major ARE in the PSA gene enhancer. Next, we used ChIP to study bicalutamide-stimulated recruitment of the W741C control versus S81A/W741C double mutant ARs in LNCaP stable cells. Bicalutamide induced binding of the W741C control AR to the PSA enhancer ARE but did not increase AR binding to an irrelevant nonspecific (NS) region of chromatin (Fig. 5B). In contrast, there was no detectable recruitment of the S81A/W741C double mutant AR.
We also carried out similar experiments in transiently transfected LNCaP cells and examined the AREs in both the PSA and TMPRSS2 enhancers (TMPRSS2-ARE5) (39). Bicalutamide stimulated binding of the transiently transfected AR-W741C AR to both AREs, although the -fold increase was modest due to transfection of only a subset of cells (Fig. 5C). In contrast, bicalutamide did not stimulate binding of the transfected

Ser-81 Phosphorylation Is Required for AR Nuclear Functions
AR-S81A/W741C double mutant to either site. Interestingly, bicalutamide caused a slight but consistent decrease in binding of the S81A/W741C mutant AR. This may reflect basal (in the absence of bicalutamide) weak chromatin engagement by the highly overexpressed transfected ARs, which then was decreased by bicalutamide due to competition from bicalutamide-liganded endogenous AR that can associate weakly with AREs (40 -42). Finally, we also assessed in LNCaP stable lines the binding of a FLAG-tagged S81D/W741C double mutant AR. ChIP results showed that the S81D mutation, which may simulate Ser-81 phosphorylation, did not prevent AR binding to chromatin and may instead enhance binding (Fig. 5D). Together, these findings indicated that Ser-81 phosphorylation was required for AR occupancy to AREs in endogenous AR-regulated genes.
Ser-81 Phosphorylation Is Involved in Cellular Distribution of AR-We next performed indirect immunofluorescence to determine whether the defect of the S81A mutation on AR binding to particular AREs was reflected in an effect of Ser-81 phosphorylation on cellular distribution in response to ligand stimulation. It has been well established that in androgen-depleted PCa cells, AR is distributed diffusely in the cytoplasm and nucleus and becomes strongly concentrated in the nucleus in response to androgen (43,44). Similarly, the FLAG-AR-W741C single mutant in androgen-starved LNCaP stable transfectants was distributed in both the cytoplasm and nucleus, and nuclear expression was enhanced by bicalutamide (Fig. 6A, top). In contrast, the S81A/W741C double mutant AR was almost entirely cytoplasmic in the absence of ligand, and substantial AR remained in the cytoplasm after bicalutamide treatment, with a minor population migrating into the nucleus (Fig. 6A, middle). Finally, the S81D/W741C mutant showed predominant nuclear expression in the absence and presence of ligand (Fig. 6A, bottom).
Although immunofluorescence studies show that wild-type AR in androgen-depleted cells is distributed in both the nucleus FIGURE 6. Ser-81 phosphorylation regulates the distribution of AR to the nuclear and chromosomal pools. A, LNCaP stable cell lines expressing FLAG-AR-W741C (C), FLAG-AR-W741C-S81A (S81A), or FLAG-AR-W741C-S81D (S81D) were grown in CDS-containing medium and treated for 2 h without or with 10 M bicalutamide, followed by immunofluorescence. Red, anti-FLAG (M2) antibody; blue, DAPI staining. B, similarly treated LNCaP stable cells were harvested and fractionated using the NE-PER kit for proteins distributed in the cytoplasm (Ct) and the nucleoplasm (Nu) (10). The pellet containing the insoluble fraction (INS) was extracted by boiling in 2% SDS. C, LNCaP stable cells grown in FBS-containing medium were sequentially extracted using Triton lysis buffer containing increased NaCl, and the insoluble pellet fraction was isolated by boiling in 2% SDS. and cytoplasm, it is only weakly associated with the nucleus. Therefore, in cellular fractionation studies, the unliganded wild-type AR is found predominantly in the cytoplasmic fraction, whereas a substantial fraction is recovered in the nuclear fraction in response to androgen stimulation (10). Similarly to the wild-type AR, cellular fractionation studies showed that the FLAG-AR-W741C single mutant in androgen-starved LNCaP cells was recovered primarily in the cytoplasmic fraction (Fig.  6B, top). As expected, bicalutamide decreased the cytoplasmic expression and increased the nuclear recovery of the W741C single mutant AR. The S81D/W741C double mutant behaved similarly to the single mutant, although the nuclear recovery in the absence of ligand was somewhat increased relative to the single mutant (Fig. 6B, bottom). In contrast to these results, only a small fraction of the S81A/W741C double mutant was in the nuclear fraction in the absence of ligand, and this fraction was not increased by bicalutamide (Fig. 6B, middle).
In addition to analyzing the nuclear proteins that are extracted at moderate stringency, we also addressed whether AR was associated with the insoluble fraction that contains histones and other proteins that are tightly bound to chromatin. By boiling the insoluble pellet fractions in SDS to release histones, we found that low levels of the W741C single mutant and W741C/S81D double mutant ARs were associated with this insoluble fraction after bicalutamide stimulation (Fig. 6B). In addition, the W741C/S81D double mutant AR gained enhanced recovery from the insoluble fraction in the absence of ligand, compared with the W741C single mutant counterpart. In contrast, we did not detect any of the W741C/S81A double mutant AR in this fraction, in the absence or presence of ligand.
As a complementary approach to assess nuclear binding, stable cells cultured in complete medium containing steroid hormones (RPMI 1640 with 10% FBS) were lysed in a low salt Triton X-100 lysis buffer, and increasing salt concentrations were used to extract sequentially nuclear proteins. In LNCaP cells expressing the W741C single mutant AR, one pool of AR was released at low salt (0 -50 mM NaCl). Additional AR then could be extracted progressively at higher salt concentrations along with the nuclear protein FOXA1, an AR-interacting transcriptional pioneer factor (Fig. 6C, top) (45,46). The insoluble material that remained after extraction with 600 mM NaCl contained histones and additional AR and FOXA1. The S81D/W741C double mutant AR showed a similar distribution of AR with slightly enhanced nuclear distribution (Fig. 6C, bottom). In contrast, nearly all of S81A/W741C double mutant AR was extracted at low salt (Fig. 6B, middle). These findings, in conjunction with the above immunofluorescence and cellular fractionation analyses, indicate that Ser-81 phosphorylation is not required for nuclear translocation but does enhance AR nuclear retention and is required for stable AR association with chromatin.
Androgen Stimulates Increased Binding of Ser(P)-81-AR to Chromatin-The data described above using the S81A mutant AR indicated that Ser-81 phosphorylation may be required for AR recruitment to AREs in endogenous androgen-regulated genes. ChIP studies of AR binding have shown that androgen stimulates very rapid recruitment of AR, with substantial binding occurring within 15 min (41). In contrast, the increase in Ser-81 phosphorylation in response to androgen is relatively slow, with cellular levels increasing over about 8 h (Fig. 7A), consistent with previous studies (9,16). Androgen-stimulated accumulation of Ser(P)-81-AR in the nucleus is similarly slow (Fig. 7B), so it was unclear whether these low levels of Ser(P)-81-AR could be contributing substantially to the androgenstimulated recruitment of AR to AREs. Therefore, to directly address the contribution of Ser-81 phosphorylation in androgen-stimulated AR recruitment to chromatin, we next carried out AR ChIP studies using the Ser(P)-81-AR antibody.
As an initial control to assess the specificity of the Ser(P)-81specific AR antibody in the ChIP assay, we transfected PC3 cells (an AR-negative PCa cell line) with the S81D mutant AR. This AR in androgen-starved or DHT-stimulated cells was then examined by ChIP for binding to the AREs in the FKBP5 or OPRK1 gene loci (the PSA gene is not activated by transfected AR in this cell line) (22). Using an anti-AR antibody directed at the extreme N terminus, we detected DHT-stimulated AR recruitment to both AREs but not to a control nonspecific region (Fig. 7C). In contrast, no binding was detected using the anti-Ser(P)-81 AR antibody in the absence and presence of ligands, indicating that this antibody did not react with nonphosphorylated AR associated with chromatin.
We next stimulated LNCaP cells with DHT for 8 h followed by ChIP with the anti-AR or anti-Ser(P)-81 AR antibody. Quantitative real-time PCR for the ARE in the PSA gene enhancer showed that the Ser(P)-81 antibody precipitated levels of this ARE that were comparable with the levels precipitated by the AR antibody (Fig. 7D, Round 1). Although higher affinity of the Ser(P)-81-AR could make it chromatin-immunoprecipitate more efficiently, this result suggested that a substantial portion of the AR binding to this ARE may be Ser-81-phosphorylated. To further test this hypothesis, we repeated the ChIP on the same lysate using the same antibodies to further deplete the chromatin (Round 2), followed by an additional round of anti-AR ChIP on the depleted lysate (Round 3). Recovery of the ARE by the Ser(P)-81 antibody was decreased in Round 2 (compare lane 6 in Round 1 versus Round 2), indicating that the antibody was at least partially depleting Ser(P)-81-AR-bound chromatin. ARE recovery by the anti-AR antibody in Round 2 was similarly decreased in the lysate that was immunoprecipitated with the AR antibody in Round 1 (compare lane 4 in Round 1 versus Round 2). Immunoprecipitation of this latter anti-AR-depleted lysate for a third time (Round 3) with the same AR antibody showed that ARE recovery was substantially reduced relative to AR ChIP from the control anti-IgG-treated lysate (Round 3, lane 4 versus lane 2, respectively). AR ChIP from the Ser(P)-81-AR antibody-depleted lysate also showed some depletion (Round 3, lane 6 versus lane 2), but it was clear that a large portion of the AR-associated ARE was not depleted by the Ser(P)-81-AR antibody. Although differences in antibody efficiency, possible in vitro AR dephosphorylation during the sequential immunoprecipitations, and other technical factors preclude precise quantitative conclusions, these data indicate that this ARE may be binding substantial levels of both Ser-81-phosphorylated and non-phosphorylated AR.
Previous reports indicated that a high stoichiometric phosphorylation occurred at AR Ser-81 upon 6 -8 h of DHT stimu-lation (9, 16), so the above ChIP results could be consistent with random chromatin binding of Ser-81-phosphorylated or nonphosphorylated AR. Therefore, we next performed ChIP after 15 min of DHT stimulation, at which time whole cell or nuclear extracts do not show detectable increases in Ser(P)-81-AR. Significantly, despite extremely low cellular levels of Ser(P)-81-AR at this time, ChIP indicated that a substantial proportion of the recruited AR was Ser-81-phosphorylated (Fig. 7E). This observation suggested that the very low cellular levels of Ser(P)-81-AR after short term DHT stimulation may preferentially be recruited to chromatin.
To further test this hypothesis, we examined the levels of total and Ser-81-phosphorylated AR that were in cytoplasmic and nuclear extracts and that were tightly associated with chromatin at 15 min versus 8 h of DHT stimulation. To more definitively identify the latter tightly chromatin-bound fraction, we used micrococcal nuclease digestion of the insoluble material after nuclear extraction to selectively solubilize proteins that tightly occupy the chromatin. By overexposing the gels, Ser(P)-81-AR could be found prior to DHT treatment in the cytoplasmic fraction, with a very weak band in the nuclear fraction and nothing detectable in the chromatin fraction (Fig. 7F, left). In contrast, upon DHT stimulation for 15 min, Ser(P)-81-AR was increased in the nuclear fraction and was detected in the chromatin fraction (Fig. 7F, middle). The levels of Ser(P)-81-AR in the cytoplasmic and nuclear extracts were much higher after 8 h FIGURE 7. Dynamics of chromatin-binding and cellular distribution for AR and Ser-81-phosphorylated AR. A, LNCaP cells grown in CDS-containing medium were treated without or with 10 nM DHT for the indicated times. Total proteins were harvested in 2% SDS for Western blotting. B, LNCaP cells grown in CDS-containing medium were treated with 10 nM androgens for the indicated times, and nuclear proteins were isolated using the NE-PER kit for Western blotting. C, PC3 cells transfected of FLAG-AR-S81D were split in CDS-containing medium for 1 day, treated without or with 10 nM DHT for 15 min, followed by ChIP using anti-Ser(P)-81 or anti-AR antibody, with rabbit IgG as control. D, LNCaP cells grown in CDS-containing medium were treated for 8 h without or with 10 nM DHT, followed by sequential ChIP analysis of PSA enhancer ChIP using anti-Ser(P)-81 or anti-AR antibody, with rabbit IgG as control. The depleted supernatant from ChIP Round 1 was applied for ChIP Round 2, with the same antibodies as indicated. The depleted supernatant from ChIP Round 2 was applied for ChIP Round 3, using anti-AR antibody. E, LNCaP cells grown in CDS-containing medium were treated for 15 min without or with 10 nM DHT, followed by ChIP analysis for binding of PSA enhancer by AR versus Ser-81 phosphorylated AR. F, LNCaP cells grown in CDS-containing medium were treated without DHT or with 10 nM DHT for 15 min or 8 h, and cytoplasmic (Ct) and nuclear (Nu) proteins were harvested using the NE-PER kit. The insoluble pellet fraction (INS) was digested without or with micrococcal nuclease (M). Error bars, S.D. P-AR-S81, Ser(P)-81-AR. MARCH 9, 2012 • VOLUME 287 • NUMBER 11 but were not substantially further increased in the chromatin fraction (Fig. 7F, right). Finally, comparisons of Ser(P)-81-AR band intensities in the nuclear extracts and chromatin-bound fractions versus total AR in these fractions indicated that the stoichiometry of AR Ser-81 phosphorylation was markedly higher in the chromatin-bound fractions.

Ser-81 Phosphorylation Is Required for AR Nuclear Functions
These fractionation studies indicated that Ser(P)-81-AR binds tightly to a relatively small number of sites on chromatin and that these sites are rapidly saturated after androgen stimulation. Based on the similar rapid binding of Ser(P)-81-AR to the PSA enhancer ARE, we suggest that these high affinity sites include AREs. However, it was not clear from these data whether AR is phosphorylated prior to chromatin binding or is phosphorylated by CDK9 subsequent to binding.
Inhibition of CDK1 or CDK9 Decreases AR Ser-81 Phosphorylation and Chromatin Binding-We showed previously that CDK1 antagonists could decrease AR protein expression, Ser-81 phosphorylation, and transcriptional activity (16). A recent report showed that CDK9 inhibition with flavopiridol similarly could reduce Ser-81 phosphorylation and AR-dependent transcription of endogenous genes (18). Consistent with these previous data, a CDK1-specific inhibitor (R-3306) decreased levels of basal and DHT-stimulated AR protein and Ser-81 phosphorylation (Fig. 8A). In contrast to the effects of R-3306, a CDK9-specific inhibitor, CDK9 inhibitor II (47) decreased DHT-stimulated Ser-81 phosphorylation without decreasing overall AR protein levels.
To determine whether the decrease in CDK1-mediated basal Ser-81 phosphorylation impaired the initial rapid DHT-stimulated binding of AR to chromatin, we examined LNCaP cells that were pretreated for 8 or 24 h with R-3306. Significantly, the R-3306 pretreatment prevented the rapid AR recruitment to the PSA and TMPRSS2 enhancers after 15 min of DHT stimulation (Fig. 8B). We next determined whether inhibiting the DHT-stimulated Ser-81 phosphorylation mediated by CDK9 would impair AR binding to chromatin. For these studies, androgen-starved LNCaP cells were pretreated for 30 min with CDK9 inhibitor II and then stimulated with DHT for 2 or 8 h. As shown in Fig. 8C, the CDK9 inhibitor impaired AR recruitment to both the PSA and TMPRSS2 enhancers.
Finally, we studied the effects of CDK1 and CDK9 inhibition on AR-mediated transactivation. CDK1 inhibition for 8 or 24 h suppressed the DHT-stimulated increase in PSA and TMPRSS2 mRNA message (Fig. 8D). Pretreatment with the CDK9 inhibitor similarly impaired DHT-stimulated expression of PSA and TMPRSS2, although TMPRSS2 expression appeared to recover FIGURE 8. CDK1 and CDK9 inhibitors repress AR Ser-81 phosphorylation and functional activities. A, LNCaP cells grown in CDS-containing medium were treated without or with 10 nM DHT for the indicated times. The CDK1 inhibitor (R-3306; 10 M) was added as indicated, and the CDK9 inhibitor (CDK9iII; 50 M) was added 30 min before DHT treatment. Total proteins were harvested in 2% SDS for Western blotting. B, LNCaP cells grown in CDS-containing medium were treated with DHT and CDK1 inhibitor for the indicated times, followed by ChIP analysis of PSA and TMPRSS2 enhancers. C, similarly to B, LNCaP cells were treated without or with DHT, together with CDK9 inhibitor for a 30-min pretreatment as indicated, followed by ChIP analysis. D, LNCaP cells grown in CDS-containing medium were similarly treated with DHT and CDK1/9 inhibitors as indicated, followed by RNA isolation for real-time RT-PCR analyses of PSA and TMPRSS2 mRNA, which were normalized to GAPDH mRNA. Error bars, S.D.; *, p Ͻ 0.05 in Student's t test. P-AR-S81, Ser(P)-81-AR. after 8 h (Fig. 8D). Because CDK9 is generally required for transcriptional elongation (19), this apparent recovery after 8 h may instead reflect a decrease in the GAPDH internal control used for normalization. In any case, these results support the conclusion that both CDK1 and CDK9 participate in Ser-81 phosphorylation that is required for AR activity. Based on these findings, we propose that Ser-81 phosphorylation is necessary to achieve stable AR binding and initially activate transcription. This model is further detailed under "Discussion."

DISCUSSION
The AR undergoes serine phosphorylation at multiple sites in response to ligand binding, with Ser-81 being the most heavily stimulated site and generally correlating with transcriptional activity (9,16). We reported previously that CDK1 could mediate Ser-81 phosphorylation and enhance the sensitivity of AR to androgens (16). Because CDK1 is activated during mitosis, in this study we first established that AR Ser-81 phosphorylation was increased during mitosis, coincident with CDK1 activation. A more recent report identified an interaction between AR and both CDK1 and CDK9 (a transcriptional CDK) and showed that CDK9 could mediate Ser-81 phosphorylation (18). Consistent with additional kinases, including CDK9, mediating Ser-81 phosphorylation, we found that androgen-stimulated Ser-81 phosphorylation could occur independent of cell cycle progression and CDK1 activation. We also confirmed that inhibition of either CDK1 or CDK9 resulted in decreased Ser-81 phosphorylation. In agreement with previous studies using transiently transfected overexpressed AR mutants and reporter genes (8,16), we found that the S81A mutation did not interfere with AR stimulation of reporter genes or the AR N-C interaction. To assess the function of Ser(P)-81 in the more physiological setting of PCa cells that are expressing endogenous AR and androgen-regulated genes, we developed a ligand switch strategy by taking advantage of a mutation in the AR-LBD(W741C) that renders the AR responsive to bicalutamide. Significantly, although the S81A/W741C double mutant AR could stimulate the expression of ARE-regulated reporter genes in response to bicalutamide, this mutant AR stably expressed in LNCaP cells was unable to stimulate expression of the endogenous AR-regulated PSA or TMPRSS2 genes. Moreover, ChIP studies identified a defect in chromatin binding, because the S81A mutant AR was not recruited to the AREs in these genes. A recent study found that the S81A mutant AR transiently transfected into AR-negative PCa cells was less active at transactivating only a subset of genes (18). However, our cellular fractionation and immunofluorescence studies indicated that this defect in chromatin binding in the AR-expressing LNCaP PCa cells was not isolated to a subset of genes because the S81A mutation globally abrogated ligand-stimulated chromatin binding.
Taken together, our findings indicated that Ser-81 phosphorylation was required for androgen-stimulated AR binding to chromatin. Previous studies have shown that mutations in the AR DBD that prevent DNA binding also impair Ser-81 phosphorylation (15,48), which suggests that Ser-81 phosphorylation may occur subsequent to DNA binding. It also should be noted that the levels of cellular Ser(P)-81 do not increase rapidly in response to androgen but instead rise progressively over about 8 h, consistent with CDK9 phosphorylation of this site as AR cycles on and off of chromatin (9,16,18,41). In contrast to this slow increase in Ser(P)-81, ChIP studies have shown that androgen-stimulated AR recruitment to chromatin occurs rapidly and is nearly maximal within 15 min. Significantly, despite the low cellular levels of Ser(P)-81 after 15 min of androgen stimulation, we found that a substantial proportion of the AR bound to chromatin was phosphorylated at Ser-81. Nonetheless, immunodepletion experiments carried out after 8 h, when Ser(P)-81 levels were maximal, suggested that a large fraction of AR bound to chromatin was not phosphorylated at Ser-81. To reconcile these results, we suggest that Ser-81 phosphorylation is not required for an initial weak transient interaction with chromatin that is not readily detected by ChIP and that Ser-81 phosphorylation by promoter-associated CDK9 stabilizes this AR binding. We further suggest that this Ser-81 phosphorylation permits the recruitment of additional coactivators and subsequent chromatin remodeling required to initiate transcription. Finally, once chromatin has been remodeled and transcription has been initiated, we suggest that Ser-81 phosphorylation may no longer be critical to maintain transcription.
Because cells are not transcriptionally active during mitosis, there may be a distinct function for CDK1-mediated phosphorylation of Ser-81. Interestingly, recent studies indicate that CDK1 plays an important role in retaining gene expression patterns by phosphorylating EZH2 (49 -51) and lineage-specific transcription factors, such as Runx2, that remain associated with chromatin during mitosis (52). Significantly, a previous report suggested that AR also remains associated with mitotic chromatin (53). Based on these findings, we suggest that AR Ser-81 phosphorylation by CDK1 may normally function to mark AR-regulated genes during mitosis, allowing daughter cells to rapidly resume their differentiated functions. Additionally, although AR protein levels have been reported to decline after mitosis (54), CDK1 may generate a pool of Ser-81-phosphorylated AR that can efficiently initiate transcription of androgen-regulated genes. Both of these models are consistent with our previous data indicating that CDK1 can sensitize PCa cells to lower levels of androgen (16) and are supported by our data in this study showing that CDK1 inhibition prevents the rapid DHT-stimulated binding of AR to chromatin. Because CDK1 activity is increased in CRPC (17,45), this may be one mechanism contributing to the maintenance of AR transcriptional activity. We are currently exploring the therapeutic potential of combined CDK1 inhibition and androgen deprivation therapy.