Phosphoinositide 3-Kinase-independent Non-genomic Signals Transit from the Androgen Receptor to Akt1 in Membrane Raft Microdomains*

The serine-threonine kinase, Akt1/protein kinase Bα is an important mediator of growth, survival, and metabolic signaling. Recent studies have implicated cholesterol-rich, lipid raft microdomains in survival signals mediated by Akt1. Here we address the role of lipid raft membranes as a potential site of intersection of androgenic and Akt1 signaling. A subpopulation of androgen receptor (AR) was found to localize to a lipid raft subcellular compartment in LNCaP prostate cancer cells. Endogenous AR interacted with endogenous Akt1 preferentially in lipid raft fractions and androgen substantially enhanced the interaction between the two proteins. The association of AR with Akt1 was inhibited by the anti-androgen, bicalutamide, but was not affected by inhibition of phosphoinositide 3-kinase (PI3K). Androgen promoted endogenous Akt1 activity in lipid raft fractions, in a PI3K-independent manner, within 10 min of treatment. Fusion of a lipid raft targeting sequence to AR enhanced localization of the receptor to rafts, and stimulated Akt1 activity in response to androgen, while reducing the cells' dependence on constitutive signaling through PI3K for cell survival. These findings suggest that signals channeled through AR and Akt1 intersect by a mechanism involving formation within lipid raft membranes of an androgen-responsive, extranuclear AR/Akt1 complex. Our results indicate that cholesterol-rich membrane microdomains play a role in transmitting non-genomic signals involving androgen and the Akt pathway in prostate cancer cells.

The serine-threonine kinase, Akt1/protein kinase B␣ is an important mediator of growth, survival, and metabolic signaling. Recent studies have implicated cholesterol-rich, lipid raft microdomains in survival signals mediated by Akt1. Here we address the role of lipid raft membranes as a potential site of intersection of androgenic and Akt1 signaling. A subpopulation of androgen receptor (AR) was found to localize to a lipid raft subcellular compartment in LNCaP prostate cancer cells. Endogenous AR interacted with endogenous Akt1 preferentially in lipid raft fractions and androgen substantially enhanced the interaction between the two proteins. The association of AR with Akt1 was inhibited by the anti-androgen, bicalutamide, but was not affected by inhibition of phosphoinositide 3-kinase (PI3K). Androgen promoted endogenous Akt1 activity in lipid raft fractions, in a PI3K-independent manner, within 10 min of treatment. Fusion of a lipid raft targeting sequence to AR enhanced localization of the receptor to rafts, and stimulated Akt1 activity in response to androgen, while reducing the cells' dependence on constitutive signaling through PI3K for cell survival. These findings suggest that signals channeled through AR and Akt1 intersect by a mechanism involving formation within lipid raft membranes of an androgen-responsive, extranuclear AR/Akt1 complex. Our results indicate that cholesterol-rich membrane microdomains play a role in transmitting nongenomic signals involving androgen and the Akt pathway in prostate cancer cells.
Androgen receptor (AR), 3 a member of the nuclear receptor superfamily of transcription factors, is an important regulator of reproductive physiology as a mediator of the biological activities of androgen (1). AR also plays a role in the pathogenesis of prostate cancer (PCa), including progression to the androgen-independent state and metastasis (2,3). Upon androgen binding, the AR undergoes conformational changes and translocates from the cytosol to the nucleus (4), where the receptor-hormone complex regulates the expression of target genes (5).
In addition to this well recognized genomic function, AR has been demonstrated to activate intracellular signaling pathways by a rapid (occurring in minutes) non-genomic process activated in response to hormonal stimulation (6,7). The mechanisms underlying the non-canonical pathways involving the AR and other nuclear receptors are still poorly understood, however, possible nongenomic roles for the AR have been described (8 -12). For example, androgen promotes oocyte maturation, in both Xenopus (10) and mouse (12), by a transcription-independent mechanism that involves the classical AR (12,14). Another report showed that cells treated with 17␤-estradiol (E2) or androgen resulted in cell proliferation mediated by the rapid formation of a cytosolic signaling complex containing estrogen receptor-␣ or -␤ (ER␣ or ER␤ (depending on the cell type)), AR, and the non-receptor tyrosine kinase, Src (8). Similarly, Kousteni et al. (9) showed that ER␣, ER␤, or AR are capable of activating the Src-Shc-ERK signaling pathway resulting in the protection of bone cells (osteoblasts and osteocytes) from apoptotic stimuli. Membrane-associated AR was also shown to regulate the activity of both ERK and Akt in C6 glial cells (13). The recent demonstrations of membrane-or cytoplasm-initiated rapid androgenic signals from several groups indicate that one or more AR signaling pathways exist in parallel with the genomic pathway in some physiologic settings (14).
Studies from our laboratory have demonstrated that epidermal growth factor receptor activation of Akt mediates cell survival signaling in LNCaP PCa cells, in part, through cholesterolenriched membrane domains (21,22). Cholesterol-rich membranes, generally referred to as "lipid rafts," exhibit a liquid-ordered structure and are insoluble in non-ionic detergents (23). Lipid rafts sequester and exclude certain types of signaling proteins and are thought to act as membrane platforms for signal transduction (24). Caveolae, membrane structures that harbor caveolin proteins, are a form of lipid raft (24,25). Although considerable heterogeneity has been found in lipid rafts isolated from different cell backgrounds (26), signaling through lipid rafts does not require the presence of caveolins, which regulate many signal transduction proteins, particularly kinases, by direct binding (22). A previous study demonstrated that enforced expression of caveolin-1 in (caveolin-null) LNCaP cells resulted in enhanced ligand-dependent AR transcriptional activity (27). Enforced expression of AR in PC3 cells was also shown to result in complex formation and co-localization with epidermal growth factor receptor (28). These findings suggest the possibility that either AR itself, or signaling proteins that communicate with the AR, reside in or transit through cholesterol-rich microdomains.
Published studies have demonstrated that the treatment of AR-expressing cells with androgen rapidly activates Akt1 in a transcription-independent manner (11,29). This non-genomic signal from AR to Akt was demonstrated to occur as a result of PI3K activation. In the present study, we identify the lipid raft subcellular compartment as a site where AR transmits a PI3Kindependent signal to Akt1.

EXPERIMENTAL PROCEDURES
Plasmids and Constructs-Hemagglutinin-Myr-Akt was from Dr. Thomas Franke (Columbia University), and T7-Akt was from Dr. William R. Sellers (Dana-Farber Cancer Institute). The DNA-binding domain-deleted AR (AR-⌬DBD) was obtained from Dr. John Isaacs (Johns Hopkins Medical Institute). The p/m-AR fusion construct was generated by inserting the PCR product of a full-length AR cDNA into the NotI site of the pcDNA3.1-p/m-myc/his vector. We engineered the pcDNA3.1-p/m-myc/his by inserting double-stranded 5Ј-phosphorylated DNA containing palmitoylation and myristoylation sequences from the Lck gene into the EcoRV site of the pcDNA3-myc/his vector. pcDNA3-hAR was described previously (30). Bacterial transformations and plasmid preparations were conducted according to standard methods. All point mutants were generated using the QuikChange site-directed mutagenesis kit according to the manufacturer's instructions (Stratagene, La Jolla, CA). Fidelity of all constructs was confirmed by DNA sequencing.
Western Blotting and Immunoprecipitations-Cell fractionations were performed according to published methods (22). Cell lysates were prepared in lysis buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, and complete protease inhibitors or phosphatase inhibitor, 1 mM sodium orthovanadate). Protein concentrations were determined by standard methods (Bio-Rad). For immunoprecipitation, cleared protein lysates were incubated with an appropriate antibody overnight at 4°C. Antibody-antigen complexes were collected on Protein A-or G-Sepharose, and immune complexes were washed three times with lysis buffer. Immunoprecipitates were resolved by SDS-PAGE, transferred to nitrocellulose membranes, and blocked either with PBST (0.1% Tween 20) containing 5% (w/v) skim milk powder (pan-specific antibodies) or PBST containing 5% BSA (Sigma) (phospho-specific antibodies). Following incubation with primary antibodies, membranes were incubated with species-specific secondary antibodies, and proteins were visualized using the SuperSignal chemiluminescence system and exposure of membranes to film.
Cell Cultures and Transfections-RPMI medium was used for LNCaP, and Dulbecco's modified Eagle's medium was used for HEK 293 and COS cells. All media were supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin, and cells were incubated at 37°C and 5% CO 2 . Where appropriate, complete serum-free RPMI medium was used in all experimental conditions. Cell transfections were performed using Lipofectamine 2000 according to the manufacturer's instructions with some modifications. Stable transfections were performed as described (3).
Density Gradient Analysis and Cholesterol Measurement-Monolayer cells treated with androgen (R1881) or vehicle (EtOH) for 10 min were washed with PBS, followed by lysis in TNET buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1 mM EGTA, 1 mM sodium orthovanadate, sodium pyrophosphate, 1% Triton X-100, and complete protease inhibitors). Cell lysis was performed on ice for 30 min. Lysates were passed through a 21-gauge syringe 12 times to break up the DNA. Lysates were cleared by (3000 ϫ g) centrifugation at 4°C for 10 min. Ice-cold OptiPrep reagent was added to cleared lysates to obtain 40% OptiPrep/sample mixture in a 1-ml volume. The samples were placed at the bottom of ultracentrifuge tubes, and then layered with 30, 20, 10, and 5% OptiPrep solution, 1 ml each. OptiPrep solutions were prepared in TNET buffer. Samples were centrifuged at 48,000 ϫ g for 18 h using a swinging bucket rotor (MLS-50). Fractions were collected from the top. To measure cholesterol content, cholesterol was extracted adding equal volumes of phenol into sample (1:1 ratio), mixed, and centrifuged. The bottom phase was collected; phenol extractions were repeated three times, and the final extracts were washed two times with water. Cholesterol extracts were obtained by removing phenol in a SpeedVac (SPD1010, Thermo Savant). The cholesterol pellet was eluted with cholesterol reagent (Thermo Scientific, Inc.). Cholesterol content (pink color) was determined at A490 nm).
Immunocytochemistry and Microscopy-LNCaP, COS, or HEK 293 cells were plated on 8-well tissue culture slides (BD Biosciences) at 70% confluence. Individual wells were transfected with either 100 ng of pcDNA3-hAR or p/m-hAR. In a separate experiment, LNCaP cells were also plated on 8-well tissue culture chamber slides at 70% confluence. Following transfection, cells were starved 24 h and incubated with 1 nM R1881 for 10 min where appropriate. Cells were labeled in serum-free growth media at 4°C for 30 min with 20 ng/ml of FITC-or 10 ng/ml Alexa Fluor 594-CTB to visualize lipid rafts. Cells were fixed in 3% paraformaldehyde for 30 min at room temperature, washed in PBS, and incubated with anti-phospho-Akt1 (Ser 473 ) antibody (1:100 dilutions) followed by goat antirabbit Cy3 antibody (1:1000 dilutions). Sequential labeling with anti-AR monoclonal antibody diluted at 1:100 and Cy5 goat anti-mouse diluted at 1:2000 was used to detect AR. Nuclei were visualized with 4Ј,6-diamidino-2-phenylindole containing mounting media (Vector Laboratories, Inc., Burlingame, CA). Cells were imaged at ϫ63 with a Plan-Apochromat oil immersion lens on either an Axioplan 2 Apotome epifluorescence microscope (Zeiss, Germany) equipped with Axiovision 4.4 and 4.5 software or an LSM 510 confocal microscope (Zeiss) with LSM 510 Browser software.
Kinase and Apoptosis Assays-For in vitro Akt kinase reactions, Akt immune complexes were mixed with a reaction mixture containing an Akt substrate (H2B; 1 g/reaction), 10 Ci of [␥-32 P]ATP, and 100 M unlabeled ATP were incubated at 30°C for 30 min. The reaction mixture was mixed with sample buffer and resolved by SDS-PAGE, followed by autoradiography. Apoptosis or cell death assays were performed using the Cell Death Detection ELISA Plus kit according to the manufacturer's instructions (Roche Applied Science).
Statistical Analysis-For biochemical and biological experiments, values are expressed as mean Ϯ S.D. Where appropriate, a t test was conducted to assess the significance of differences between treatments. Statistical significance was determined at p Յ 0.05.

AR Is Present Within Lipid Raft
Membranes-We initially sought to determine whether Akt1 (hereafter Akt) could be rapidly activated by androgen in androgen-sensitive LNCaP cells. Serum-depleted LNCaP cells were incubated with the AR agonist R1881 (1 nM) for varying times (0, 5, 10, 15, and 30 min). Akt activation, as assessed by measurement of phosphorylation at the Ser 473 site, was stimulated by androgen in a time-dependent manner. Akt activation was evident as early as 5-10 min of treatment (Fig. 1A). To study the potential role of lipid raft membranes in this signaling event, we determined whether AR was present in Triton X-100-insoluble, octylglucoside-soluble biochemical fractions, which are enriched in lipid raft compo-nents (22,31). Triton X-100-soluble (TS) and Triton X-100insoluble/octylglucoside-soluble (TI) fractions were isolated from LNCaP cells in log-phase growth, and the distribution of AR within these fractions was examined by Western blot. The results (Fig. 1B, upper panel) showed that substantial levels of AR partition into the TI fraction under these conditions. Correct partitioning of the lipid raft protein marker, G␣ i2 , into the raft fractions confirmed the integrity of the biochemical fractionation ( Fig. 1B, lower panel).
We next examined the effect of androgen on AR localization to the TI fraction. Serum-depleted LNCaP cells were treated with androgen for varying times (0, 15, 30, and 60 min), and the relative levels of AR partitioning into TI and TS fractions were assessed (Fig. 1C). Although the majority of AR was found in the TS fraction, substantial levels of AR were detected in the TI fraction at time 0, and levels of AR partitioning into lipid raft fractions were enhanced further by androgen. AR in the TI fraction reached a maximum 15 min following androgen treatment, with no significant difference in AR levels seen between 15 and 60 min. Western blots for ␤-tubulin and G␣ i2 , markers for cytosol/non-raft membrane and lipid raft membrane, respectively, showed no cross-contamination between the fractions (Fig. 1C, lower panels). A similar result was obtained using HEK 293 cells transiently transfected with wild-type (WT) human AR (supplemental Fig. S1).
To verify the above findings, LNCaP cells were treated with androgen or vehicle for 10 min, and lipid raft fractions were isolated by gradient ultracentrifugation (32). Raft fractions in the gradient were identified by immunoblots of gradient frac-FIGURE 1. Androgen promotes rapid Akt activation and AR localization to lipid rafts. A, Akt phosphorylation level. LNCaP cells were treated with androgen (1 nM R1881, a synthetic androgen analog) at the indicated times. Equal amounts of total protein were resolved by SDS-PAGE, and Western blots were performed with antibodies against phospho-S473 or total Akt proteins. B, Western blot analysis of AR or raft marker G␣ i2 in Triton X-100-soluble (TS) or Triton X-100-insoluble (TI) fractions isolated from LNCaP cells grown under normal serum condition. The G␣ i2 blot represents a lipid raft marker. C, androgen increases the levels of AR in lipid rafts. LNCaP cells were serum depleted and treated with either 1 nM R1881 or vehicle (EtOH) control. TS and TI fractions were isolated at the indicated times. Western blot shows AR, soluble marker ␤-tubulin, and raft marker G␣ i2 . Data are representative of two individual experiments.
tions with antibodies against the raft markers G␣ i2 ( Fig. 2A, lower panel) and flotillin-2 (not shown). In response to androgen, AR accumulated in G␣ i2 -containing fractions ( Fig. 2A). These results are largely in agreement with the data obtained using the successive detergent extraction method, although endogenous AR in the raft material obtained from gradient preps was more difficult to detect in the unstimulated condition by immunoblotting. G␣ i2 -reactive fractions contained high levels of cholesterol, verifying the cholesterol-rich nature of the raft material (Fig. 2B). Immunofluorescence cell staining of AR along with the Alexa Fluor 594-CTB labeling of GM1, a lipid raft-resident ganglioside, revealed that the AR co-localizes with the raft marker in LNCaP cells (Fig. 2, C and D). Taken together, these observations indicate that AR resides at lipid raft membranes and this association is enhanced within minutes in the presence of androgen.
AR and Akt Interact in Lipid Rafts-A subpopulation of Akt is present in lipid rafts of LNCaP cells (22,33). This observation, along with previous studies indicating that AR and Akt can assemble into the same multiprotein complex (27,34), suggest the hypothesis that AR and Akt may interact within lipid raft membranes. Co-immunoprecipitation experiments indicated that endogenous AR and Akt1 could be co-precipitated with anti-Akt1 specific antibody (IG1) from LNCaP cells (Fig. 3A). The AR/Akt interaction within lipid raft fractions was markedly stimulated by androgen treatment (Fig. 3A). Complex for-mation between AR and Akt was also seen by co-immunoprecipitation (IP) in total cell lysates from HEK 293 cells that were transiently co-transfected with AR and T7-Akt1 expression FIGURE 2. Analysis of AR in density gradient fractions. A, Western blots of AR in gradient fractions. LNCaP cells grown in the absence of serum were treated with vehicle (EtOH) or R1881 (1 nM) for 10 min. Cell lysates were collected in buffer containing 1% Triton X-100 and subjected to OptiPrep (Sigma) density gradient centrifugation. Fractions were collected from the top of the gradient. Western blots of AR and G␣ i2 were performed to assess the amount of AR in raft membranes (predominantly fractions 5 and 6). B, the levels of cholesterol in the indicated fractions from "A." Cholesterol measurement was performed as described under "Experimental Procedures." C and D, immunofluorescence of AR along with the Alexa Fluor 594-CTB labeling of GM1 in LNCaP cells treated with vehicle (EtOH) or R1881 (1 nM) for 10 min in a serum depleted condition. Alexa Fluor 594-CTB conjugate was used to stain the lipid raft marker GM1 (red). AR was detected using FITC-conjugated antirabbit (green) secondary antibody. Co-localization of AR and GM1 are evident in the absence of androgen (yellow). Nuclei (blue) are stained with 4Ј,6-diamidino-2-phenylindole (DAPI). These micrograph images are the representative of multiple images from at least three different experiments. constructs and treated with androgen or vehicle control (Fig.  3B). To investigate whether Akt activity is required for the kinase to interact with the AR, two kinase-inactivation mutants (20) were analyzed in AR/Akt co-transfected HEK 293 cells. Both mutants interacted in an indistinguishable manner from the wild-type form, indicating that Akt activation is not required for complex formation (Fig. 3C). To determine whether AR and Akt interact directly, we carried out an in vitro binding experiment. Increasing amounts (0, 200, 400, and 800 ng) of pre-activated Akt1 recombinant protein were reconstituted with AR immune complex from LNCaP cells, and the presence of Akt1 in the precipitated complex was assessed by anti-Akt1 antibody. We observed that pre-activated Akt physically interacted with AR in a dose-dependent manner (Fig. 3D).
To examine the extent to which the androgen-regulated AR/Akt complex forms in other subcellular locations, LNCaP cells were stimulated with androgen or vehicle. Akt was subsequently immunoprecipitated from cytoplasmic, nuclear, and TI fractions using a different Akt1-specific antibody (2H10). The results shown in Fig. 3E indicate that AR was primarily detected in Akt immune complexes generated from TI fractions. Although Akt and AR predominantly localized to cytoplasmic and nuclear fractions, levels of AR/Akt complexes in these fractions were low in comparison to those seen in the TI fractions. G␣ i2 , which in LNCaP cells is found exclusively in cholesterolrich fractions in density gradients ( Fig. 2A), was used as a raft marker in these experiments. These observations suggest that complex formation between AR and Akt, as measured by co-IP with two different anti-Akt antibodies, occurs preferentially in lipid raft membrane fractions.
Published studies have demonstrated that androgen stimulates the AR to signal to Akt by a PI3K-dependent mechanism in LNCaP cells (29) and in vas deferens epithelial cells (11). To determine the significance of the interaction of AR and Akt in lipid rafts, and to assess the extent to which PI3K might be involved in the regulation of the complex, LNCaP cells were treated with androgen alone or together with the androgen antagonist Casodex (bicalutamide) or the PI3K inhibitor LY294002. Drugs were added to the cultures 60 min prior to androgen stimulation (for 10 min), and the effect of these agents on AR/Akt complex formation was analyzed in TI fractions. The results shown in Fig. 3F demonstrated that treatment with Casodex prevented AR/Akt complex formation in response to androgen (compare lane 2 to lane 1), however, the PI3K inhibitor had only a marginal effect (compare lane 3 to lane 1), suggesting that binding of AR to Akt induced by androgen may bypass PI3K.
AR Promotes Akt Signaling in Lipid Rafts-To establish whether androgen can transmit a rapid signal to Akt in lipid rafts, Akt phosphorylation and activity were measured in raft fractions isolated from LNCaP cells following treatment with androgen or vehicle for 10 min. The results indicated that androgen enhanced the levels of Akt phosphorylation 2-fold (Fig. 4A, upper panel), and kinase activity 3-fold (Fig. 4B, blot  and graph), over control conditions. Accumulation of AR in the TI fraction induced by androgen was also observed (Fig. 4A,  lower panel). Consistent with the binding data (Fig. 3D), Caso-dex also inhibited the androgen-induced Akt kinase activity in lipid raft fractions (Fig. 4, C and D).
In a separate experiment, Akt phosphorylation levels were also measured in density gradient fractions by Western blot. Results from this experiment showed that a 10-min pulse of androgen induced Akt phosphorylation only in G␣ i2 -reactive and cholesterol-rich fractions, corresponding to rafts, compared with the control treatment (Fig. 4E, fractions 5 and 6). However, no significant changes in Akt phosphorylation levels were observed in fractions corresponding to non-raft material (Fig. 4E, fractions 2 and 3 or 7 and 8), which were devoid of G␣ i2 and low in cholesterol. Taken together, the data suggest that

. Androgen enhances Akt activity in lipid rafts in LNCaP cells.
A and B, Akt phosphorylation levels and kinase activity in raft fractions. LNCaP cells were treated with R1881 (1 nM) or vehicle (EtOH) following serum depletion. Cell fractionation was performed 10 min post-treatment. A, equal amounts of total protein were resolved by SDS-PAGE. Blots represent corresponding proteins detected by antibodies against the total or phospho-Akt or AR proteins. B, Akt kinase assays were performed with H2B (1 g), [␥-32 P]ATP, and Akt immune complex. The graph represents the intensities of Akt activity from the autoradiograph. Error bars represent the mean Ϯ S.D. (n ϭ 2) per data point. C and D, Akt kinase activity in the TI fraction. LNCaP cells treated with androgen alone or together with bicalutamide (Cas.). Drug treatments were initiated 60 min prior to androgen stimulation. Kinase assay was performed as in B. The H2B blot shows Coomassie Blue-stained H2B protein levels used in the kinase assay. The graph represents the intensities of Akt activity from the autoradiograph of C. Error bars represent the mean Ϯ S.D. (n ϭ 3) per data points. E, Akt phosphorylation levels in density gradient fractions. LNCaP cells grown in the absence of serum were treated with vehicle (EtOH) or R1881 (1 nM) for 10 min. Cell lysates were fractionated by Opti-Prep (Sigma) density gradient centrifugation. The fractions were separated by SDS-PAGE. Western blots with phospho-Ser 473 antibody were performed to assess the amount of phospho-Akt in fractions. Blots are total or phospho-Akt. G␣ i2 is a lipid raft marker.

complex formation between AR and Akt promoted by androgen resulted in Akt activation in lipid rafts.
Our results showed that inhibiting PI3K was unable to prevent AR/Akt complex formation in TI fractions (Fig. 3F) or in whole cell extracts (not shown). To test whether chemical inhibition of PI3K is capable of altering androgen-induced Akt activation, LNCaP cells were treated with vehicle, LY294002 alone, or LY294002 plus R1881, and the effect on Akt activity determined. Consistent with the binding data, LY294002 was only partially effective at inhibiting Akt activation by androgen (Fig.  5, A and B). A similar result was observed when Akt activity was assayed in lipid raft-enriched fractions (Fig. 5C). These observations suggest that there are both PI3K-dependent and -independent pathways from AR to Akt1.
Lipid Raft-targeted AR Enhances Akt Activity-To further clarify the origin of the androgenic signal to Akt in lipid rafts, we engineered a lipid raft-targeted variant of AR by fusing the palmitoylation (p) and myristoylation (m) sequences from the Lck kinase to the N-terminal domain of AR. Myristoylation is the permanent modification of a protein by the addition of the 14-carbon fatty acid myristate to the N-terminal glycine via an amide bond (35). Palmitoylation is a reversible post-translational modification involving the addition of the 16-carbon fatty acid palmitate to variably located cysteine residues via a thioester bond (36). This strategy has resulted in targeting of proteins to lipid raft membranes in published studies (20,37). The AR fusion construct, designated p/m-AR (Fig. 6A), was stably expressed in HEK 293 cells and its subcellular location compared with wild-type AR. Biochemical fractionation and Western blot analysis revealed that p/m-AR primarily localized to the TI (raft-enriched) fraction, in marked contrast to WT-AR, which was found primarily in the TS fraction (Fig. 6A). Caveolin-1 and -2 (Cav1/2) was used as a lipid raft marker in this experiment. These data were confirmed by immunofluorescence staining in a second cell type (COS) (Fig. 6B). A similar staining pattern was observed in HEK 293 cells expressing stable p/m-or WT-AR (supplemental materials Fig. S2). Furthermore, sucrose density gradient analysis also demonstrated that p/m-AR was extensively enriched in light, lipid raft fractions, whereas the majority of unmodified AR was in the heavy fractions (Fig. 6C).
We then examined the role of p/m-AR on endogenous Akt activity and its ability to form a complex with Akt in response to androgen. HEK 293 cells were transiently transfected with p/m-AR, followed by stimulation with or without androgen for 10 min. Similar to the data obtained with endogenous AR in LNCaP cells, androgen induced the formation of a complex between p/m-AR and endogenous Akt in TI fractions (Fig. 7A). In addition, assessment of Akt activity by phosphorylation (Fig.  7B) and kinase activity measurements (Fig. 7, C and D) revealed that androgen induced Akt activity 3-4-fold in raft fractions. Despite the Akt activating activity seen in these experiments, p/m-AR failed to promote transcription from androgen-responsive promoter-reporter constructs, consistent with a nongenomic mode of signaling to Akt (Fig. 7E and supplemental materials Fig. S3). Furthermore, to eliminate the possibility that androgen action on Akt was mediated by the genomic function of AR, we assayed the Akt activity in TI fractions from HEK 293 cells that were transiently transfected with a mutant AR where the DNA binding domain was deleted (AR-⌬DBD). Western blot analysis of Akt phosphorylation levels showed that deletion of the DNA binding domain did not affect AR-mediated Akt activation when the results were compared with WT-AR (Fig.  7F). Furthermore, the above data demonstrated that the majority of p/m-AR mainly localized to lipid rafts. We envisioned that alteration in subcellular localization of raft-targeted AR from the WT-AR pattern would diminish the inhibitory effect of LY294002 on Akt activity. Western blot analysis of Akt phosphorylation levels showed that the androgen treatment of cells expressing p/m-AR dramatically reversed the LY294002-induced Akt inhibition (Fig. 7G) compared with conditions seen with the WT-AR (Fig. 7H). Taken together, these observations indicate that AR action on Akt in raft fractions does not involve the AR genomic function.
Signals from AR to Akt Override PI3K-dependent Cell Survival-The above findings indicate that signals from AR to Akt may bypass PI3K. Constitutive PI3K activity mediates an obligatory survival signal in LNCaP cells (38). To test the hypothesis that p/m-AR may promote cell survival independently of PI3K, LNCaP cells were transiently transfected with p/m-AR or vector control, followed by the treatment of cells with 10 M LY294002 for 5 h in serum-free medium. Under these conditions, the level of p/m-AR expression was assayed by Western blot (Fig. 8A), and the number of viable cells was determined by cell counting (Fig. 8B). The results demonstrated that relative cell number over control was significantly higher in the presence of p/m-AR expression than the vector control. In an alternative assay, we also determined the apoptotic efficacy of LY294002 in LNCaP cells. The apoptotic effect of LY294002 was significantly less in p/m-AR-expressing cells than in vector was included as a control to confirm the validity of raft isolation. Below, a schematic representation of the p/m-AR. B, immunofluorescence of p/m-AR or WT-AR together with the FITC-CTB labeling of GM1 in COS cells expressing transient p/m-AR or WT-AR. FITC-CTB conjugate was used to stain the lipid raft marker GM1 (green). AR was detected using Texas Redconjugated anti-mouse (red) secondary antibody. Co-localization of AR and GM1 are evident in the absence of androgen (yellow). Nuclei (blue) are stained with 4Ј,6-diamidino-2-phenylindole. C, density gradient analysis of p/m-hAR and WT-hAR. The fractionation was performed as described in the legend to Fig. 2A. Blots are the corresponding proteins detected with a protein-specific antibody. Western blots were probed with anti-Akt or anti-Myc antibody. B, Akt phosphorylation levels. Cells were transiently transfected with p/m-AR and treated with androgen (ϩ) or vehicle control (Ϫ) following serum depletion. TS and TI fractions were isolated at 10 min post-treatment, and equal amounts of protein were resolved. Western blots show proteins detected by antibodies against total and phospho-Akt proteins. C and D, Akt kinase activity in TS and TI fractions. Akt kinase activity was determined using H2B and [␥-32 P]ATP in TS and TI fractions from cells transfected with p/m-AR, followed by R1881 or vehicle (EtOH) treatment for 10 min. The graph in D represents normalized Akt activity from the autoradiograph (C). Error bars represent the mean Ϯ S.D. (n ϭ 3) per data points. E, AR-responsive pGRE4-TATA-Luc promoter-reporter activity in COS cells. Cells were co-transfected with p/m-AR or WT-AR and pGRE4-TATA-Luc. Cells were treated with vehicle (EtOH) or R1881 (1 nM). Luciferase signal normalized to total protein was expressed as -fold induction with respect to control vector. Data points represent mean Ϯ S.D. of triplicate determinations. F, Akt phosphorylation level in TI fractions of HEK 293 cells. Cells were transiently transfected with WT-AR and DBD-deleted AR (AR-⌬DBD), followed by treatment with R1881 or vehicle (EtOH) for 10 min. Equal amounts of TI fractions were resolved by SDS-PAGE. Western blots were probed with antibody against total or phospho-Akt. G and H, Akt phosphorylation levels in HEK 293 cells. Cells were transiently transfected with p/m-AR or WT-AR and treated with 10 M LY294002 alone or together with R1881 (1 nM) or vehicle control for 10 min. Equal amounts of total protein were resolved. Western blots were probed with antibodies against total and phospho-Akt. Drug treatments were initiated 60 min prior to androgen stimulation. Data are representative of two individual experiments.
control cells (Fig. 8C). Furthermore, androgen also significantly abolished the apoptotic efficacy of LY294002 in LNCaP cells over control (not shown) (38). These findings demonstrate that recruitment of AR to lipid raft membranes activates survival signals that are independent of the PI3K pathway.

DISCUSSION
Our study demonstrates for the first time that AR forms a hormone-sensitive complex with the serine-threonine kinase Akt1 in lipid raft microdomains. The evidence that this complex mediates transmission of a non-genomic signal from androgen to Akt is the following: (i) a subpopulation of endogenously expressed AR in LNCaP cells was found in a Triton X-100-insoluble, cholesterol-enriched (lipid raft) subcellular compartment; (ii) rapid interaction of AR with Akt occurred preferentially in detergent-resistant (TI) lipid raft fractions in comparison to cytosol-or nuclei-enriched fractions; (iii) AR and Akt complex formation was enhanced by androgen within 10 min of treatment of cell cultures with the hormone; (iv) AR/Akt complex formation and rapid Akt activation by androgen were inhibited by the anti-androgen, bicalutamide (Casodex), but were insensitive to PI3K inhibition; (v) the major effect of androgen on rapid Akt phosphorylation and kinase activity occurred in raft fractions; (vi) a genomically inactive, raft-targeted AR fusion protein (p/m-AR) mediated androgeninduced Akt activity in AR-negative cells; and (vii) p/m-AR was capable of reducing LNCaP cells' dependence on constitutive PI3K signaling for cell survival. Although significant levels of p/m-AR were found in TS fractions, the Akt activating signal attributable to AR was greater in TI fractions. Because androgen effects on AR binding to Akt, and on Akt activation status, occurred within a short time period (10 min), and because these effects were elicited by an AR mutant incapable of binding DNA, we conclude that this is a non-genomic signaling mech-anism. Furthermore, our findings indicate that the androgen-AR complex activates Akt directly, not via a pathway that is dependent on the upstream Akt activator, PI3K.
Steroid hormones and their receptors have been demonstrated to rapidly activate intracellular signal transduction pathways in different cell backgrounds (6 -12, 14, 29, 39, 40). For example, E2 was shown to rapidly stimulate endothelial nitric-oxide synthase by activating plasma membrane-associated ER␣. This action of E2 on endothelial nitric-oxide synthase required a coordinated event involving ER␣, Src, and PI3K-Akt (41,42). This mechanism has been invoked to explain the rapid cardioprotective effects of estrogen in the vascular system. Membrane ER␣ was shown to form signaling complexes with the proto-oncoprotein ErbB2, with subsequent activation of the PI3K-Akt pathway (43). In addition, ER␣, ER␤, or AR was demonstrated to attenuate apoptosis of osteoblasts and osteocytes, embryonic fibroblasts, and HeLa cells by activating a c-Src-Shc-ERK signaling pathway (9). Similarly, membrane-associated AR has also been shown to be involved in rapid signaling processes (10,12,13).
Observations from the present study and from the literature taken together suggest that functionally distinct forms of AR exist, which signal from discrete subcellular locations. At least two independent studies, using LNCaP and vas deferens epithelial cells, have described a non-genomic mechanism of AR activation of PI3K-Akt pathway signaling (11,29). In both cases, AR was shown to form a complex with the p85␣ regulatory subunit of PI3K. Sun et al. (29) demonstrated that the interaction between AR and p85␣ was dependent on Src activity. Akt activation by androgen in the Sun et al. study (29) occurred within 24 h following exposure to androgen. Similarly, Pandini et al. (44) showed the up-regulation of insulin-like growth factor-IR expression in HEK 293 cells within 24 h following androgen  OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40 treatment. Consistent with the present results, this effect was induced in AR-negative cells by a non-genomic signal from mutant forms of AR that are transcriptionally inactive and/or retained in the cytoplasm (AR-C619Y or AR-C574R). This androgen-induced insulin-like growth factor-IR up-regulation was shown to be sensitive to the Src inhibitor, PP2, although a demonstration that the non-genomic effects of androgen occurred rapidly was not a component of the Pandini et al. (44) study. In our study, which focused on rapid effects of androgen, we find that AR and Akt form a complex in a PI3K-independent manner. Formation of this complex was also largely insensitive to Src inhibition, suggesting that non-genomic signals from AR to Akt may bypass both PI3K and Src. Taken together, these results indicate that non-genomic signals involving AR likely diverge at the level of cytoplasmic and plasma membranes. Because lipid raft membranes have been shown to harbor preexisting signaling complexes, our results imply that raft membranes play a prominent role in this divergence of androgenic signals to downstream mediators.

Androgen Receptor Signaling in Lipid Rafts
The molecular events controlling recruitment of AR to the plasma membrane or lipid raft microdomains specifically are unclear. One potential mechanism is association of AR with an integral membrane component through direct interaction. In support of this possibility, Lu et al. (27) demonstrated that enforced expression of Cav1 in Cav1-negative LNCaP cells resulted in the localization of AR to Cav1-enriched membrane fractions and resulted in enhanced androgen-dependent AR activation. In the same study, androgen was also demonstrated to promote Cav1/AR interaction. Although Cav1 may promote AR localization to membranes (Fig. 5) (27), we demonstrate here that AR partitioning into lipid rafts and signaling from membrane AR can occur in a Cav1 negative cell background. This suggests that AR might be recruited to membranes by other membrane proteins, thereby facilitating the flow of signals from AR to Akt and possibly to other signaling proteins. Consistent with this idea, ER␣ was found to be capable of activating signaling from the insulin-like growth factor-1 receptor by a mechanism involving direct binding of ER␣ to the insulinlike growth factor-1R (45). Post-translational modifications, such as phosphorylation (46) or palmitoylation (47), would be another mechanism by which AR might be recruited to membrane sites. Phosphorylation of ER␣ on Ser 522 was previously demonstrated to promote ER␣ recruitment to the plasma membrane (46). The role of phosphorylation in AR localization to membranes remains to be investigated. Palmitoylation of the ER␣ ligand binding domain at Cys 477 mediates ER␣ plasma membrane localization (47). In that study, ectopic expression of a C477A ER␣ mutant in HeLa cells abolished ER␣ localization to the plasma membrane while reducing E2-induced rapid activation of ERK signaling. A mechanism of membrane localization that is similarly dependent on palmitoylation may apply to the AR. 4 Thus, post-translational modifications may provide the opportunity for facilitated interactions between AR and integral membrane proteins, such as caveolins or ErbB receptors, with implications for regulation of downstream signaling.
The p/m-AR described in the current study is structurally different from the membrane-targeted ER described by Rai et al. (48). Membrane-targeted-ER was engineered by N-terminal myristoylation and C-terminal palmitoylation, while lacking a nuclear localization sequence (48). In contrast, we engineered p/m-AR to express a short peptide with a myristoylation and palmitoylation site linked to the N-terminal domain of AR. Although structurally distinct, both p/m-AR and membranetargeted ER displayed similar characteristics: localization to the plasma membrane and cytoplasm, transcriptional inactivity, and rapid response to their cognate ligands. This suggests the possibility that minimal post-translational modification of the AR may be sufficient to elicit non-genomic effects of androgen (48). We present a model ( Fig. 9) in which a pool of classical AR associates with lipid raft-like membrane domains and forms a complex with Akt. Interactions between the two proteins may lead to Akt activation and other downstream effects by recruiting other factor(s) into the kinase complex. Our results also support the notion that signaling from AR to Akt within lipid rafts is PI3K-independent. This, however, does not preclude the existence of one or more parallel PI3K-dependent pathways from AR to Akt (11,29). Because many hormone-resistant prostate cancers are AR-positive, our results suggest the possibility that membrane-associated AR may contribute to androgen-dependent growth of PCa cells in a manner that does not obligatorily involve the AR transcription function (39,49). Consistent with this idea, a cytoplasmic form of AR has recently been linked to human PCa metastasis (50).
In summary, this investigation reveals that signals channeled separately through AR and Akt intersect by a mechanism involving a direct interaction between the two proteins within lipid raft microdomains. The finding of an alternative pathway of Akt activation by androgen may be relevant to situations, such as aggressive PCa, where both proteins, or their activated forms, may be overexpressed. The existence of unconventional signals from AR to Akt should be taken into account in the context of approaches toward novel drug development targeted to either the androgen or Akt signaling pathways in patients. 4 E. Levin, personal communication.