Loss of Androgen Receptor Transcriptional Activity at the G1/S Transition*

Androgens are essential for the differentiation, growth, and maintenance of male-specific organs. The effects of androgens in cells are mediated by the androgen receptor (AR), a member of the nuclear receptor superfamily of transcription factors. Recently, transient transfection studies have shown that overexpression of cell cycle regulatory proteins affects the transcriptional activity of the AR. In this report, we characterize the transcriptional activity of endogenous AR through the cell cycle. We demonstrate that in G0, AR enhances transcription from an integrated steroid-responsive mouse mammary tumor virus promoter and also from an integrated androgen-specific probasin promoter. This activity is strongly reduced or abolished at the G1/S boundary. In S phase, the receptor regains activity, indicating that there is a transient regulatory event that inactivates the AR at the G1/S transition. This regulation is specific for the AR, since the related glucocorticoid receptor is transcriptionally active at the G1/S boundary. Not all of the effects of androgens are blocked, however, since androgens retain the ability to increase AR protein levels. The transcriptional inactivity of the AR at the G1/S junction coincides with a decrease in AR protein level, although activity can be partly rescued without an increase in receptor. Inhibition of histone deacetylases brings about this partial restoration of AR activity at the G1/S boundary, demonstrating the involvement of acetylation pathways in the cell cycle regulation of AR transcriptional activity. Finally, a model is proposed that explains the inactivity of the AR at the G1/S transition by integrating receptor levels, the action of cell cycle regulators, and the contribution of histone acetyltransferase-containing coactivators.

cases of prostate cancer, androgen ablation is the primary therapy used (4,5), yet with time, androgen-independent tumors arise in individuals who undergo this therapy (6,7). This has led to an intense investigation of the molecular mechanisms involved in androgen signaling.
The actions of androgens are mediated by the androgen receptor (AR), 1 a transcription factor that belongs to the nuclear hormone receptor superfamily. In the absence of androgens, the AR protein is primarily cytosolic and is found complexed with heat shock proteins that keep it inactive (8). Upon binding to androgens, the receptor undergoes a conformational change that releases it from this inhibitory complex (9). AR then localizes to the nucleus, where it binds as a dimer to androgen response elements found on the promoters of target genes (10). The ability of the AR to modulate gene transcription is enhanced by the recruitment of coactivators and possibly by the release of corepressors (11,12). Coactivators can provide enhanced interactions with the basal transcriptional machinery through activation domains of their own. They also contribute intrinsic or associated histone acetyltransferase activities, thus allowing for chromatin remodeling (13). We have previously shown that activation of AR brings about such nucleosome rearrangements on the mouse mammary tumor virus (MMTV) promoter and that this remodeling correlates with transcriptional activity (14). We have also reported that the hyperacetylation of histones enhances the ability of the AR to remodel chromatin and modulate transcription (15) and that anti-androgens inhibit chromatin remodeling, consequently blocking AR transcriptional activity (16). Thus, the functions of AR require the activity of histone acetylases. The AR itself seems to also be the target of acetylation, and its transcriptional activity may be enhanced in vivo by this modification (17).
The rate of mammalian cell growth is largely determined by the length of the G 1 phase of the cell cycle. Progression from G 1 phase through the G 1 /S transition 2 and into S phase is governed by the action of cyclins and cyclin-dependent kinases (CDKs) on the retinoblastoma protein (Rb) (18,19). Cyclin D1-CDK4 complexes in middle to late G 1 and then cyclin E-CDK2 complexes in G 1 /S and early S phase phosphorylate Rb, diminishing its ability to bind and repress the S-phase-promoting factor E2F (20 -25). It is known that androgens influence growth, shortening the length of G 1 /G 0 and accelerating entry into S phase, by affecting the expression and/or activity of cyclins and CDKs (3,26). Recently, it has been demonstrated that some cell cycle regulatory proteins can, in turn, influence AR transcriptional activity by acting as AR coregulators. These include the retinoblastoma protein, and cyclins D1 and E, molecules that show altered expression in many human cancers.
Our laboratory and others have reported that expression of the retinoblastoma protein restores AR function in Rb-deficient cells (27,28). Additionally, Knudsen et al. (29) and Reutens et al. (30) have shown that overexpression of cyclin D1 (and to a lesser extent cyclin D3) inhibits AR function in a CDK-independent manner. Furthermore, Yamamoto et al. (31) determined that cyclin E overexpression, independently of its association with CDK2, results in the positive regulation of AR activity. Generally, these experiments used transient transfection techniques to introduce into cells expression vectors of both the AR and the cell cycle regulator under study and measured transcriptional effects on transient templates (27)(28)(29)(30)(31). This approach results in overexpression of the cell cycle regulators throughout the cell cycle rather than the phasespecific expression found in normal cells. We have taken a more physiological approach by investigation of the regulation of the transcriptional activity of endogenous AR on integrated promoters during the cell cycle.
In this report, we show that the transcriptional activity of endogenous AR varies through the cell cycle. We demonstrate that the AR is transcriptionally active in G 0 , loses over 90% of its activity during the G 1 /S transition, and then regains the ability to enhance transcription in S phase. We show that this transient negative regulation at the G 1 /S transition is specific for the AR, since the related glucocorticoid receptor (GR) maintains transcriptional activity at this boundary. The down-regulation of AR protein that we observe at G 1 /S may partially explain the lack of transcriptional activity. However, chemical inhibition of histone deacetylases rescues AR activity during G 1 /S without increasing the level of AR protein, suggesting that regulation of AR activity during the cell cycle also involves acetylation/deacetylation pathways.

EXPERIMENTAL PROCEDURES
Cell Culture and Development of Stable Cell Lines-L929 cells (American Type Culture Collection) were cultured in Dulbecco's modified Eagle's medium supplemented with 3% calf serum, 100 IU/ml penicillin, and 100 g/ml streptomycin. For the development of the L929-MMTVCAT stable cell line, L929 cells were transfected using Dosper liposomal reagent (Roche Molecular Biochemicals) with pM-MTVCAT and pSV2neo (20:1 ratio), according to the manufacturer's protocol. To obtain the L929-ProbasinLuc cell line, L929 cells were transfected using LipofectAMINE 2000 reagent (Invitrogen) with pϪ286/ϩ28PB-luciferase (32) and pSV2neo (20:1 ratio), according to the manufacturer's protocol. In both cases, the cells were split 48 h after transfection and selected in growth media supplemented with 400 mg/ liter G418 sulfate (Cellgro). Single clones were picked with sterile pipette tips and expanded. Clones were screened for chloramphenicol acetyltransferase (CAT) or luciferase activity after a 24-h hormone induction. Single clones showing low basal reporter activity and at least 5-fold activation with DHT were used for further studies (clones L929-MMTVCAT #31 and L929-ProbasinLuc 2.9 were used in this study).
Cell Cycle Arrests and Fluorescence-activated Cell Sorting Analysis-L929-MMTVCAT and L929-ProbasinLuc cells were cultured in Dulbecco's modified Eagle's medium supplemented with 3% calf serum. All cell cycle arrests were carried out on ϳ80% confluent cells (300,000 cells/ well of a 24-well plate incubated overnight or the equivalent density on larger surface areas) according to the methods shown in Fig. 2A. After plating and overnight growth, cells were serum-starved (grown in 0.1% calf serum) for 48 h to induce entry into G 0 . (For control experiments, G 0 cells were exposed to 1 or 2 mM hydroxyurea for the second 24 h of starvation and during hormone induction as illustrated in Fig. 3B). During an additional 24 h of serum deprivation, cells were treated with steroids or left untreated. For G 1 /S arrests, cells were starved for 48 h as above. After this starvation, cells were exposed to 1 or 2 mM hydroxyurea in 10% serum for 24 h and for an additional 24 h in the presence or absence of hormone. Cells growing in serum were arrested along S phase by treatment with 1 or 2 mM hydroxyurea for 48 h. Cells were then induced with steroids for 24 h in the presence of hydroxyurea. In all cases, cells were washed with PBS after hormone treatment and harvested in 0.25 M Tris-HCl, pH 7.8 (when only CAT or luciferase assays were performed) or trypsinized and collected (when additional fluorescence-activated cell sorting (FACS) or Western analysis was to be performed). Collected cells were aliquoted, spun down, and resuspended. For CAT/luciferase assays, cells were resuspended in 0.25 M Tris-HCl, pH 7.8; aliquots for FACS were resuspended in citrate buffer (250 mM sucrose, 40 mM trisodium citrate-2H 2 O, 5% Me 2 SO, pH 7.6), and cells for Western analysis were lysed in modified radioimmune precipitation buffer (see below). All samples were stored frozen until analyzed. Citrate buffer samples were analyzed for DNA content at the Lombardi Cancer Center Flow Cytometry/Cell Sorting Shared Resource by propidium iodide staining in a FACSort (Becton Dickinson) (33). Computer modeling of cell cycle phase distribution was performed at this facility using the software package ModFit (Verity).
CAT Assays/Luciferase Assays-Cell extracts in 0.25 M Tris-HCl buffer, pH 7.8, were frozen/thawed three times to lyse the cells. Protein concentrations were measured by the Bradford method (34). For CAT assays, equal amounts of protein from each extract (typically 1 or 2 g) were combined with 1 l of 3 H-labeled acetyl coenzyme A (1.33 Ci/mmol specific activity; DuPont), 19.0 l of 2 mg/ml chloramphenicol, and 80 l of 0.25 M Tris-HCl, pH 7.8. One ml of organic scintillation mixture Econoflour-2 (Packard Instrument Co.) was overlaid on the reaction mix, and vials were placed in a scintillation counter. As the reaction proceeds, the acetylated product is incorporated into the organic phase and is counted (35). Samples were counted for three consecutive cycles in a ␤ counter, and the results were expressed as the increase of the counts produced/min (cpm/min). For luciferase assays, equal amounts of protein from each extract were combined with 100 l of luciferase assay substrate (Promega) and immediately counted in a luminometer.
Western Analysis-Cells were collected, spun, and washed in cold PBS. Cell pellets were dissolved in modified radioimmune precipitation buffer (20 mM Tris-HCl pH 7.8, 140 mM NaCl, 1 mM EDTA, 0.5% sodium deoxycholate, 0.5% Nonidet P-40) supplemented with 0.66 mg/ml Pefabloc (Invitrogen), 3.3 g/ml leupeptin, and 1 mM dithiothreitol. Typically, 30 g of total protein was loaded in each lane of a 4 -20% gradient SDS-polyacrylamide gel and separated by electrophoresis. Protein was transferred to nitrocellulose membranes and confirmed by Ponceau Red staining. After blocking for at least 2 h in 5% milk, 0.2% polyvinyl pyrrolidone, membranes were blotted with the corresponding first antibody. For AR detection, 4 g/ml PA1-111A, a rabbit polyclonal antibody that recognizes the N terminus of the AR (Affinity Bioreagents) was used. A 1:200 dilution of sc-1616, a goat polyclonal antibody, was used to probe for actin (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Immunoreactive bands were visualized using anti-rabbit or antigoat horseradish peroxidase-conjugated secondary antibodies and ECL reagents (Amersham Biosciences). Immunoreactive bands were quantified using the software package ImageQuant.

Construction of a Cell Line with an Integrated Reporter That
Responds to Both Androgens and Glucocorticoids-Several reports over the past few years show the involvement of G 0 , G 1 /S, and S phase cell cycle regulators in the control of androgen receptor activity. Generally, these experiments used transient transfection techniques to introduce into cells expression vectors of both the AR and the cell cycle regulator under study and measured transcriptional effects on transient templates (27)(28)(29)(30)(31). Although such studies provide important information on the interaction of cell cycle regulators and the AR, they do not distinguish between effects seen due purely to overexpression and those that reflect interactions that occur during normal cell growth. Our approach was to investigate the regulation of the transcriptional activity of endogenous AR on integrated promoters during the cell cycle. To do this, we developed a cell line with an integrated AR-responsive CAT reporter gene. L929 cells that express endogenous AR were stably transfected with the androgen-and glucocorticoid-responsive reporter pMMTV-CAT. The resulting clones were expanded and characterized. A cell line was established from a representative clone and is referred to here as L929-MMTVCAT. The presence of functional AR in these cells is shown in Fig. 1A (left panel), where treatment with androgens (1 nM dihydrotestosterone (DHT)) resulted in over 30-fold induction of CAT activity. This transcriptional activity was fully blocked by the anti-androgen cyproterone acetate, demonstrating the involvement of the AR in this response. Since L929 cells are known to also express endogenous GR, we measured transcriptional activity in response to glucocorticoids. CAT activity was induced over 20fold in the presence of dexamethasone (DEX). This induction of MMTVCAT was due to the action of the GR, since the antiglucocorticoid ZK 98.299 fully blocked the response (Fig. 1A, right panel).
The Androgen Receptor Loses Transcriptional Activity at the G 1 /S Boundary-To optimize the androgen response for cell cycle studies, a time course of AR activation was performed in serum-starved L929-MMTVCAT cells. Androgens clearly induced measurable CAT activity after 24 h (Fig. 1B). This time point was used in cell cycle experiments, since we observed that cells lose synchrony during prolonged arrest (data not shown).
To measure the transcriptional activity of the AR in G 0 , G 1 /S, or S phase, cells were arrested prior to receptor activation, and cell cycle blocks were maintained during hormone treatment as described under "Experimental Procedures" and outlined in Fig. 2A. To ensure effective cell cycle arrests throughout the length of the experiments, we performed FACS analysis on arrested cells both prior to (data not shown) and after hormone induction as well as on uninduced controls (Fig. 2, B-D, right panels). Importantly, we observed that 24-h androgen treatment had no discernible effect on cell cycle distribution (compare NH histograms with DHT histograms, in Fig. 2, for example). This was expected, since the growth of L929 cells is affected negatively by glucocorticoids and positively by androgens only under chronic long term exposure (36).
As seen in Fig. 1A, we found that unsynchronized cells growing in the presence of 3% serum routinely showed 20 -30-fold induction of CAT activity in response to 1 nM DHT. AR consistently had the highest activity in serum-starved G 0 cells, inducing CAT activity up to 100-fold in the presence of DHT (Fig.  2B, left panel). In contrast, the AR showed no detectable activity after treatment with 1 nM DHT, in cells arrested at the G 1 /S boundary (Fig. 2C). AR regained transcriptional activity when the cells were released from G 1 /S arrest (not shown) or were blocked along S phase by direct treatment with hydroxyurea without prior serum starvation (Fig. 2D). These data indicate that there is a transient regulatory event that prevents AR transcriptional activity at the G 1 /S boundary. The anti-androgen cyproterone acetate inhibited DHT-induced activity in G 0 cells and did not show any agonistic activity in cells synchronized at the G 1 /S boundary (data not shown). As seen in Table  I, in three independent experiments, the transcriptional activity of the AR at the G 1 /S junction was decreased 92-100% compared with its activity in G 0 . This shows that at the G 1 /S transition AR function is strongly and consistently inhibited.
To ensure that the inactivity of the AR in cells arrested at the G 1 /S transition was not the result of nonspecific actions of the arresting drug, we tested the effects of hydroxyurea on AR activity during G 0 . L929-MMTVCAT cells were prearrested in G 0 by serum starvation for 24 h. During the next 24 h, the cells were exposed to 2 mM hydroxyurea with continued serum starvation. In the final 24 h of treatment, cells were induced with androgens during continued exposure to hydroxyurea and serum starvation (Fig. 3B). AR transcriptional activity in G 0 cells was unaffected by the presence of hydroxyurea, giving androgen inductions within the range usually obtained with cells in this phase of the cell cycle (Fig. 3A, left panel). These data demonstrate that the loss of AR function observed at G 1 /S is not a nonspecific or toxic effect of the drug per se. Indeed, a similar lack of AR inhibition by hydroxyurea is seen in cells arrested along S phase with this drug (Fig. 2D). To test whether the prolonged treatment of G 1 /S cells (96 h compared with 72 h for G 0 and S phase cells; see Fig. 2A) could account for the inactivity of the AR, cells were serum-starved for 72 h and then exposed to hormone during an additional 24 h of starvation. As can be seen in Fig. 3D, the transcriptional activity of the AR was unaffected by the 96-h starvation treatment. Indeed, we have prolonged starvation for an additional 24 h as well as performed 72-h hydroxyurea treatments in serum with no effects on AR activity (not shown).
The Glucocorticoid Receptor Is Transcriptionally Active in G 1 /S Cells-To test whether there was a general shut down of transcription or translation at the G 1 /S boundary or whether this regulation was specific to the androgen pathway, we arrested cells at the G 1 /S boundary and then treated them with either glucocorticoids or androgens. Treatment of cells synchronized at the G 1 /S boundary with 100 nM DEX or 1 nM DHT for 24 h did not alter their distribution along the cell cycle (Fig.  3C). GR was transcriptionally active in cells arrested at the G 1 /S transition, inducing CAT activity over 20-fold, yet no AR activity was detected in androgen-treated cells in the same experiment (Fig. 3A, right panel). These data show that there is a preferential negative regulation of the AR over the GR at the G 1 /S transition. They also demonstrate that there is not an inherent deficiency in the transcription or the translation of the  CAT message or protein, respectively, in G 1 /S boundary-arrested cells, since glucocorticoid treatment results in CAT activity.
G 1 /S Regulation of Transcription from an Androgen-specific Promoter-The MMTV long terminal repeat is a promiscuous promoter that not only responds to androgens and glucocorticoids but also to mineralocorticoids and progestins (37,38). The results presented above demonstrate that the strong inhibition of transcriptional activity seen on the MMTV promoter at the G 1 /S boundary is specific for the AR. To evaluate whether a similar temporal regulation of AR is observed on promoters that respond only to the AR, we obtained a luciferase reporter construct, driven by the androgen-responsive region of the natural probasin promoter (32). An L929 cell line with integrated copies of this construct was developed as outlined under "Experimental Procedures." We then tested the specificity of this promoter in our cells (referred to here as L929-ProbasinLuc cells). Treatment with 1 nM DHT for 24 h resulted in a 3-5-fold induction of luciferase activity in asynchronous cultures. Cyproterone acetate inhibited this transcriptional activity, confirming the involvement of the AR (Fig. 4A). Both synthetic (dexamethasone) and natural glucocorticoids (cortisol) completely lacked the ability to induce transcription at this promoter (Fig. 4A).
We then evaluated the regulation of transcription from this androgen-specific promoter during the cell cycle. We found that luciferase activity was induced 3-5-fold in response to 1 nM DHT in G 0 cultures (Fig. 4, B and C, leftmost panels). However, there was almost no induction in cells arrested at the G 1 /S transition (Fig. 4, B and C, middle panels). As seen in the case of the MMTV long terminal repeat, AR retained transcriptional activity on the probasin promoter in cells arrested along S phase (Fig. 4, B and C, rightmost panels). Thus, in cells blocked at the G 1 /S boundary, the AR is transcriptionally inactive not only on promiscuous promoters but also on promoters which are specifically androgen-responsive.
AR Protein Levels Decrease at G 1 /S but Retain Their Ability to Be Up-regulated by Androgens-To determine whether there was a correlation between AR transcriptional activity and the levels of AR protein during the cell cycle, cells were arrested in G 0 , at the G 1 /S boundary, and along S phase as in Fig. 3 and treated with androgens or left untreated. Cells were harvested, one aliquot was used to determine CAT activity, and another aliquot was used for Western analysis. As can be seen in Fig.  5A, AR levels are regulated across the cell cycle, with the lowest levels occurring at G 1 /S (Fig. 5A, top and middle panels) when AR transcriptional activity is at its lowest (Fig. 5A, bottom panel). Hormone treatment results in increased levels of AR in all stages of the cell cycle examined, including G 1 /S. The reproducibility of the hormone induction of AR levels in G 1 /S cells is shown in Fig. 5B. Stabilization of the AR protein in the presence of androgens has been shown to occur in L929 and other cells previously (14,39), but this is the first demonstration that it occurs in G 0 , in G 1 /S, and in S phase and that the stabilization itself does not correlate with the transcriptional activity of the AR. Despite the increase in AR protein seen with androgens, DHT-treated G 1 /S cells still only contain 20 -25% of the receptor levels present in DHT-treated G 0 cells (Fig. 5, A  and B, middle panels). It has been shown previously that the transcriptional activity of steroid receptors closely correlates with the number of bound receptors based upon the Michaelis-Menten equation adapted for ligand-receptor interaction (40,41). Therefore, to determine whether the decrease in total receptor levels fully accounts for the loss of transcriptional activity of the AR seen in G 1 /S-arrested cells, cells in G 0 were treated with decreasing concentrations of DHT, and transcriptional activity was determined (Fig. 5C). Substantial transcriptional activity was found in response to 1 nM DHT, where 45% of receptors are occupied (K d ϭ 1.4 nM (27), where 50% of receptors are theoretically occupied by DHT), and to DHT levels 10-fold below this, where only 10% of receptors are occupied. Indeed, measurable transcriptional activity was detected at DHT levels 100-fold below the K d , where only 1% of receptors are predicted to be occupied by hormone. These data indicate that although decreased receptor levels at G 1 /S may play a significant role, they do not fully account for the almost complete loss of transcriptional activity of the AR at this stage of the cell cycle.
Histone Hyperacetylation Rescues AR Activity in Cells Arrested at the G 1 /S Boundary without Increasing AR Protein Levels-It is known that steroid receptor action is mediated by the recruitment of histone acetyltransferase-containing coactivators that bring about chromatin remodeling by acetylating histones (42)(43)(44)(45)(46). We have previously shown that chromatin remodeling is a necessary step in AR transcriptional activity (14) and that the hyperacetylation of histones facilitates this process, whereas anti-androgens prevent it (15,16). For these reasons, we decided to test the hypothesis that AR complexes are unable to induce chromatin remodeling during the G 1 /S transition and that this inability partly accounts for the lack of AR transcriptional activity. If this is true, inhibition of histone deacetylases should restore partial AR activity. To evaluate this, we chemically blocked L929-MMTVCAT cells at the G 1 /S boundary and then treated them simultaneously with androgens and with trichostatin A (TSA), an inhibitor of histone deacetylases (47), or with either one alone. Cells induced with androgens showed no more than 2-3-fold induction of CAT activity over background levels (Fig. 6A, right panel). This represents a greater than 90% inhibition of AR activity compared with the corresponding DHT-treated G 0 samples shown in Fig. 6A (middle panel). In contrast, AR activity was induced more than 20-fold in cells co-treated with androgens and TSA, whereas TSA alone had no effect (Fig. 6A, right panel). This enhanced transcriptional activity of the AR was not the result of cells progressing through G 1 /S and entering S phase, since cells remained arrested at the G 1 /S boundary during treatments, as shown by FACS analysis (Fig. 6C). Furthermore, when the effects of TSA on DHT induction were measured in other stages of the cell cycle and compared, it was clear that TSA preferentially enhanced DHT action at the G 1 /S transition (Fig. 6B). In the presence of TSA, androgen induction levels increased almost 10-fold in G 1 /S cells, compared with 3-5-fold increases in asynchronous cells and in G 0 cells (Fig. 6A, left and middle panels). In all cases, this enhanced activity was fully blocked by CA, demonstrating that it was mediated through the AR (Fig. 6A, all panels). In contrast, the effects of TSA on DEX induction of the GR remained constant throughout the cell cycle, showing no preferential enhancement in G 1 /S (Fig. 6D).
Since G 1 /S cells have decreased levels of AR protein, one possible explanation for the rescue of transcriptional activity in these cells by TSA would be an induction of AR levels by TSA. This was not the case, however, since treatment of G 1 /S cells with TSA partly restored AR transcriptional activity in response to DHT without altering receptor levels, as shown in Fig. 6E. These data show that the reduced levels of AR found at G 1 /S are capable of activating transcription in TSA-treated cells. Indeed, when only 1% of the receptors present in G 0 are occupied with hormone, AR activity is maintained (Fig. 5C), further demonstrating that a low number of activated receptors can be transcriptionally functional in G 0 . This suggests that a transient regulatory event involving acetylation/deacetylation pathways prevents AR from activating transcription during the G 1 /S transition. In addition, these data indicate that the reduced levels of AR protein seen at G 1 /S are the result of a regulatory event at the level of AR expression and/or stability  Fig. 3B) or in G 1 /S or along S phase (as in Fig. 2A). AR transcriptional activity was measured by luciferase assays as described under "Experimental Procedures. and not due to the decreased transcriptional activity of the AR itself, since TSA increases AR transcriptional activity without increasing receptor levels. DISCUSSION This is the first report to measure the transcriptional activity of endogenous AR during the cell cycle. We have demonstrated that the AR is fully active in G 0 -arrested mouse L929 cells and inactive in cells blocked at the G 1 /S boundary and that it regains transcriptional activity in cells arrested along S phase. We have shown that this transient negative regulation at the G 1 /S transition preferentially affects the AR, since the related GR is active in these cells. Androgens were able to up-regulate receptor protein during G 1 /S boundary arrest, demonstrating that at least one androgenic function remains intact. AR protein levels were found to be regulated through the cell cycle, with the lowest levels present at G 1 /S. This down-regulation of AR protein may partly explain the lack of AR activity in these cells. However, the partial recovery of AR activity in cells at the G 1 /S transition treated with TSA, without a concomitant increase in AR levels, indicates that this low level of AR can be active in the context of hyperacetylated histones and that de-creased AR levels are not the only reason for AR inactivity in G 1 /S. Thus, the inactivity of the AR at G 1 /S seems to be the result of two regulatory events: down-regulation of receptor levels and transient inactivation of the receptor's transcriptional activity. The second but not the first effect can be rescued by inhibiting deacetylases with TSA, providing evidence for the involvement of acetylation/deacetylation pathways in the cell cycle regulation of AR transcriptional activity. An earlier study reported that exogenously expressed AR was transcriptionally active on a transient template in cells treated with hydroxyurea and simultaneously induced with androgens (31). Although these authors termed this a G 1 /S arrest, it most closely resembles what we term an S phase arrest, since they did not perform a prior G 0 synchronization. For this reason, the data from the two studies do not disagree.
The AR is not the only transcription factor outside the family of cell cycle control proteins whose regulation is cell cycle-dependent. The closely related GR, for example, has been shown to be transcriptionally inactive in G 2 /M in many cell types (48 -50). During this part of the cycle, it has been reported that the pattern of GR phosphorylation is altered and that these FIG. 5. AR levels are low at G 1 /S but retain their ability to be up-regulated by androgens. L929-MMTVCAT cells were synchronized as described for Fig. 3B. Cells were harvested in trypsin and aliquoted for FACS analysis, Western blotting, and CAT assays. FACS analysis confirmed effective cell cycle arrests (histograms not shown). A, cells for Western analysis (top panel) were suspended in modified radioimmune precipitation buffer and separated by electrophoresis as described under "Experimental Procedures." The same amount of protein was loaded in each lane. Membranes were blotted with PA1-111A, a polyclonal antibody that recognizes the N terminus of the AR. Actin was detected with sc-1616, a goat polyclonal antibody. Immunoreactive bands were visualized by chemiluminescence as described under "Experimental Procedures." Bands were quantified (middle panel) using the software package ImageQuant. AR levels are expressed relative to actin control bands. CAT activity (bottom panel) of an aliquot of the cells used in A was determined as described under "Experimental Procedures." The same amount of protein was used for each CAT assay. The relative CAT activity of S phase cells varies between experiments but is within 20 -80% of the activity of cells in G 0 . B, Western blot (top panel) of an independent experiment showing the up-regulation of AR levels in the presence of androgens in G 1 /S-arrested cells at two different hormone concentrations. Quantifications of bands (middle panel) and CAT activity of the samples shown in A (bottom panel) were performed as described above. C, AR activity in G 0 cells in response to decreasing DHT concentrations. AR activity is present in response to 1 nM DHT (K d ϭ 1.4 nM, where 50% of receptors are occupied). Activity is still measurable even at more than 100-fold below the K d , where less than 1% of receptors are predicted to be occupied by hormone. changes may prevent GR from being properly retained in the nucleus (50). Phosphorylation also regulates the activity of other transcription factors through the cell cycle. MEF, a member of the ETS family, for example, is controlled by cyclin A-dependent phosphorylation that restricts its activity to G 1 (51). The DNA binding ability of the Cut homeodomain transcription factor is mainly seen during S phase. In this case, cell cycle regulation is the result of increased transcription of the cut gene and of dephosphorylation of the Cut protein by the Cdc25A phosphatase during S phase (52). We cannot rule out an involvement of phosphorylation in the cell cycle control of AR; however, the CDK-independent effects of cyclin D1 and cyclin E on receptor activity suggest that mechanisms other than phosphorylation are at play (29,31). Indeed, the partial reversal of G 1 /S inhibition of AR by TSA suggests an involvement of histone acetylation in the response. Regardless of the mechanism of regulation, it is of particular interest that there is a class of transcription factors that affects the dynamics of the cell cycle by controlling the expression of proliferative/ differentiation genes and that these transcription factors are regulated by the molecules whose activities define the phases of the cell cycle. Down-regulation of AR protein levels during the G 1 /S transition may be one mechanism by which cells modulate the transcriptional activity of the receptor, since androgen sensitivity in various tissues and cell lines has been correlated with receptor protein levels. The factors that control androgen receptor expression are poorly understood, however, and seem to be highly tissue-and cell type-specific. It has been shown that NF-B and NF1 negatively regulate AR gene expression (53,54), whereas c-Myc and Sp1 increase AR expression (55,56). Androgens themselves regulate AR expression at several levels, and this regulation has been only partly characterized. Androgens have been observed to have a variety of effects in vivo and in tissue culture according to cell type. These effects include down-regulating steady state levels of AR mRNA (57,58), stabilizing the AR message (59,60), and increasing or decreasing the rate of transcription (60,61). In general, AR protein levels are increased by androgens regardless of the effect on mRNA levels. This increase in AR protein is brought about either by stabilization of the protein as measured by longer receptor half-life (62)(63)(64) or indirectly by increased translation as a result of altered mRNA levels or potentially by a combination of both effects (65)(66)(67). The present study shows that this basic function of androgens is maintained throughout the cell cycle but is not in itself sufficient to elicit a measurable AR transcriptional response.
The finding that the AR is inactive at G 1 /S on both the MMTV and the probasin promoters implies that this regulation is a general feature of AR action. The MMTV long terminal repeat is a promiscuous promoter responsive to androgens, glucocorticoids, mineralocorticoids, and progestins (37,38). In contrast, the probasin promoter is AR-specific. Specificity for AR on the probasin promoter has been associated with the arrangement of one of its two androgen response elements as a direct repeat rather than as the inverted repeats found on MMTV (32,38,68). It has been suggested that the exact manner of AR dimer formation, AR N-and C-terminal interactions, and recruitment of coactivator complexes may be different on AR-specific direct repeats compared with general steroid-responsive inverted repeats (69). Regardless of the differences that may indeed exist, the mechanism(s) responsible for AR inactivity at the G 1 /S transition are at work in both cases. The preferential inactivation of the AR over the GR on MMTV at the G 1 /S boundary further suggests that this temporal regulation may contribute to transcriptional specificity, reducing or abolishing the androgen responsiveness of some genes while maintaining their glucocorticoid responsiveness.
The observation that histone hyperacetylation restores AR activity at the G 1 /S junction (Fig. 6) suggests that histone modifications may repress transcription in a manner that can be overcome by GR-recruited protein complexes but not by AR complexes during this transition. This possibility is particularly appealing, since the dynamics of chromatin remodeling at the MMTV promoter have been shown to differ in response to glucocorticoids and androgens (14). The GR rapidly and tran- FIG. 7. Model of AR regulation through the cell cycle. This model suggests possible mechanisms for the specific negative regulation of the AR at the G 1 /S transition. In general, AR-activated transcription involves the recruitment of coactivators to the promoter, including proteins that remodel chromatin. In G 1 /S, we propose that the AR-recruited complexes inefficiently remodel chromatin, leading to loss of receptor activity. Additionally, receptor levels are decreased, further contributing to transcriptional inactivity. The model shows how cell cycle specific coactivators and corepressors may bring about this regulation. Left and bottom right panels, in G 0 and S phase, when cyclin D1 levels are low, the AR coactivator complexes that are formed include p/CAF, providing acetylation activities. Cell cycle-specific coactivators such as hypophosphorylated Rb (in G 0 ) and cyclin E (in S phase) further promote AR activity. Top right and bottom left panels, in G 1 /S, binding of cyclin D1 to AR prevents the binding and/or action of p/CAF, causing a loss in acetylation of histones and/or of AR. Additionally, phosphorylation of Rb at the G 1 /S junction inhibits its coactivator function. Thus, the low amounts of AR that are expressed are transcriptionally inactive. Histone hyperacetylation by TSA obviates the requirement for chromatin-remodeling complexes, partly restoring AR activity in G 1 /S. Note that the sizes of the diagrams do not represent relative protein dimensions. Sites of interaction between proteins (where known) are not necessarily accurately drawn. The AR is shown as a monomer for simplicity of presentation. siently remodels MMTV chromatin during transcriptional activation (70), whereas the AR gradually induces chromatin remodeling over time (14), suggesting that distinct complexes mediate these two remodeling events. Any coregulator required by AR but not by GR may be modified at the G 1 /S transition, altering its activity. This could affect the AR directly by posttranslational modifications and/or indirectly through chromatin remodeling defects or other inhibitory events. These inhibitory activities may be prevented in G 0 and reversed or compensated for in S phase by cell cycle-specific components of AR coactivator complexes.
A number of cell cycle-specific proteins have the potential for regulating AR activity during the cell cycle. These include Rb, cyclin D, and cyclin E. Hypophosphorylated Rb has been shown to be an essential AR coactivator in some cell lines but is not required by GR (27,71). Low levels of hypophosphorylated Rb are consistent with the inactivity of the AR and the activity of the GR at the G 1 /S transition but do not explain the presence of AR activity in S phase, since hypophosphorylated Rb levels remain low throughout this stage (20,23). The loss of hypophosphorylated Rb in late G 1 and G 1 /S is due to the increased activity of cyclin D1-CDK4 complexes at these points of the cell cycle. Interestingly, cyclin D1 strongly inhibits the AR (29,30). Thus, at the G 1 /S transition, these two separate but interrelated events may conspire to decrease AR activity (Fig. 7). In S phase, there is decreasing cyclin D-CDK4 activity and increased cyclin E-CDK2 activity (24). Since cyclin E activates the AR (31), it is possible that increased cyclin E levels in S phase compensate for the low levels of hypophosphorylated Rb and, together with decreasing cyclin D1 levels, explain the S phase activity of the AR.
AR activity at the G 1 /S transition can be partly restored by treating cells with the histone deacetylase inhibitor TSA. This is particularly interesting, since cyclin D1 inhibition of AR activity is also overcome by TSA treatment (72). Taken together, these data suggest that cyclin D1 inhibits AR activity during G 1 /S by inhibiting an AR-specific acetylation event(s) that can be overcome with the use of TSA. Two acetylation events have been proposed to increase AR activity. One is the recruitment of coactivator-associated histone acetyltransferase activity, leading to chromatin rearrangement (73,74). The other is the acetylation of specific lysines in the AR by p/CAF (17). Mutation of these lysines severely reduces AR activity. Cyclin D1 has recently been shown to strongly disrupt p/CAF-AR interactions (30). Thus, we propose that the inactivity of the AR at G 1 /S is due to competition between cyclin D1 and p/CAF, leading to the failure of p/CAF to be recruited to AR complexes, resulting in decreased histone and AR acetylation. A model incorporating this idea is shown in Fig. 7, where the balance between the permissive effects of RB, cyclin E, and acetylation and the inhibitory effects of cyclin D1 leads to AR activity in G 0 and S phase and inactivity at G 1 /S. The biological significance of AR down-regulation and inactivity at the G 1 /S transition is unclear to us, yet it could be advantageous for cells to have a mechanism for controlling the action of growth-promoting or differentiation factors such as the AR at this check point. The functional meaning of this regulation at G 1 /S may become clear only in situations where it is lacking due to abnormal coregulators or mutations in AR. Aberrant expression or function of AR coregulators, including proteins of the cell cycle machinery and acetylases/deacetylases is thought to occur in a range of tumors (75)(76)(77)(78)(79)(80). This abnormal environment could potentially alter the ability of the AR to modulate its target genes in a proper temporal manner, leading to defects in growth control or differentiation, even in the presence of wild type AR. It will be interesting to evaluate if AR mutants found, for example, in benign hyperplasias or tumors of the prostate bypass cell cycle regulation, being active during the G 1 /S transition and/or inactive in G 0 or S phase. (For a summary of AR mutants, see Refs. 81 and 82). If such mutants are identified, it would be important to also evaluate their ability to interact with cell cycle-specific AR coregulators. Additionally, it is possible that nonsteroidal activators of the AR (83-86) may bypass G 1 /S control or further restrict the action of AR during the cell cycle. Whether G 1 /S regulation of AR activity is necessary for proper control of growth or differentiation in androgen-sensitive tissues awaits further investigation. The development of methods to simultaneously measure AR transcriptional activity and cell cycle position in single cells would greatly facilitate such studies.