Distinct expression and activity of GSK-3α and GSK-3β in prostate cancer



Glycogen synthase kinase (GSK-3) is upregulated in many types of tumor, including prostate cancer. GSK-3 inhibitors reduce prostate tumor cell growth; however, it is not clear if both isoforms, GSK-3α and GSK-3β, are involved. Here, we compared their expression in prostate tumors and used gene silencing to study their functions in 22Rv1 prostate cancer cells. Compared to normal prostate, GSK-3α and GSK-3β were upregulated in 25/79 and 24/79 cases of prostate cancer, respectively, with GSK-3α elevated in low Gleason sum score tumors and GSK-3β expressed in high Gleason tumors, and both isoforms correlating with high expression of the androgen receptor (AR). Gene silencing of GSK-3α and, to a lesser extent, GSK-3β reduced AR transcriptional activity. In addition, silencing of GSK-3β, but not GSK-3α, reduced Akt phosphorylation. Acute and chronic silencing of either isoform reduced 22Rv1 growth in colony formation assays; however, this did not correlate with effects on AR activity. The GSK-3 inhibitor CHIR99021 reduced 22Rv1 colony formation by 50% in normal growth medium and by 15% in hormone-depleted medium, suggesting that GSK-3 is required both for hormone-dependent and hormone-independent proliferation. In addition, CHIR99021 enhanced growth inhibition by the AR antagonists bicalutamide and MDV3100. Finally, expression of GSK3A and GSK3B mRNAs correlated with a gene expression signature for androgen-regulated genes. Our observations highlight the importance of the GSK-3/AR signaling axis in prostate cancer and support the case for development of isoform-specific GSK-3 inhibitors and their use, in combination with AR antagonists, to treat patients with prostate cancer.

Glycogen synthase kinase (GSK-3) is a widely expressed, multifunctional serine (Ser)/threonine (Thr) protein kinase that transduces signals activated by Wnts, Hedgehog, Notch and growth factors such as insulin and epidermal growth factor, and thus plays roles in cell proliferation, differentiation, survival and apoptosis.1 Most cells express two isoforms of GSK-3, GSK-3α and GSK-3β, which have almost identical catalytic domains. Both isoforms are activated by tyrosine phosphorylation (Tyr279 in GSK-3α and Tyr216 in GSK-3β) and inhibited by Ser/Thr phosphorylation (Ser21 in GSK-3α and Ser9 and Thr390 in GSK-3β).1

Several studies have highlighted the importance of GSK-3 for cancer cell survival and/or proliferation. GSK-3β expression is upregulated in many types of cancer, including myeloma,2 bladder cancer,3 glioblastoma,4, 5 colon cancer,6 pancreatic cancer,5–7 ovarian cancer4, 8 and renal cancer.9 The evidence supporting a role for GSK-3β in prostate cancer is particularly strong.10–14 A recent analysis of tumors from 499 patients with prostate cancer found a correlation between increased levels of cytoplasmic GSK-3β and clinical stage, Gleason score and high expression of the androgen receptor (AR).15 There are few reports on the role of GSK-3α in cancer. However, in those studies where GSK-3 isoforms have been compared, GSK-3α is more important than GSK-3β for survival of pancreatic cancer cells16 and for resistance of multiple myeloma cells to bortezomib-induced cytotoxicity,17 whereas GSK-3β is more important than GSK-3α in mixed-lineage leukemia (MLL) acute leukemia.18 Importantly, inhibition of GSK-3 inhibits the survival and/or proliferation of many types of cancer, both in vitro9, 10 and in vivo in tumor xenografts.7, 14, 18

GSK-3α and GSK-3β also have overlapping and specific roles in normal cells. For example, mice lacking GSK-3β die late in development, whereas mice lacking GSK-3α have abnormalities in glucose metabolism and in the brain.19 These isoform-specific functions could reflect differences in tissue expression, subcellular localization20 and/or substrate specificity.21

There are a number of GSK-3 substrates that may be relevant to prostate cancer. GSK-3β phosphorylates C/EBPα, for example, promoting its degradation, and as a result activates E2F target genes and promotes cell cycle progression in the androgen-independent PC3 prostate cancer cell line.14 AR may be another important substrate: it can be phosphorylated in cells by ectopically expressed GSK-3β and in vitro by purified GSK-3β.22, 23 However, both gene silencing of GSK-3β or treatment of prostate cancer cells with GSK-3 inhibitors10, 11 and ectopic expression of GSK-3β22, 23 reduce AR transcriptional activity. These apparently contradictory observations suggest that GSK-3 regulates AR at multiple levels. Indeed, GSK-3 has demonstrated effects both on AR stability and nuclear localization.11, 13, 24 Some effects of GSK-3 are likely to be indirect, mediated by GSK-3 substrates that regulate AR, for example, Akt/PKB, which was recently reported to be regulated by GSK-3.25, 26 Alternatively, or in addition, there may be isoform-specific effects, since most studies have focused on GSK-3β rather than GSK-3α, and GSK-3 inhibitors target both isoforms.

In our study, we compared the expression levels of GSK-3α and GSK-3β proteins in normal prostate and prostate tumors and used inhibitors and isoform-specific gene silencing to study the role of GSK-3 in the regulation of AR and prostate cancer cell proliferation. We find increases in cytoplasmic GSK-3β and GSK-3α in prostate cancer, both correlating with AR levels, but in low and high Gleason sum score tumors, respectively. The link between GSK-3 and AR is underlined by gene expression data that find both GSK-3 isoforms associated with activation of AR-regulated genes. In addition, we observe differential effects of silencing of each isoform on AR activity and Akt phosphorylation, suggesting that it may be useful to consider GSK-3α and GSK-3β as separate therapeutic targets for the treatment of prostate cancer.

Material and Methods

Cell culture and analysis

The human prostate cancer cell line 22Rv1 was purchased from the American Type Culture Collection (ATCC, Rockville, MD) and cultured at 37°C, 5% CO2 in a 1:1 mixture of high glucose (4.5 g/l) DMEM and RPMI-1640 media (Life Technologies, Paisley, UK) supplemented with 20% fetal bovine serum (FBS; Life Technologies, Paisley, UK) and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin; Life Technologies, Paisley, UK). For studies of the effects of hormone depletion, cells were cultured in phenol-red free medium containing 5% charcoal-stripped serum (CSS; First Link (UK) Ltd., Wolverhampton, UK). The human prostate cancer cell line LNCaP (ATCC) was cultured in RPMI-1640 medium with 10% FBS and antibiotics. Cells were lysed in radio-immunoprecipitation assay (RIPA) lysis buffer, as previously described,27, 28 and probed by Western blotting using antibodies to GSK-3α (H-12; Santa Cruz Biotechnology, Heidelberg, Germany), GSK-3β (27C10; New England Biolabs, Hitchin, UK), phospho-GSK-3 Tyr216/279 (Abcam plc, Cambridge, UK), GSK-3α+β (4G-1E; Millipore UK, Watford, UK), pan-Akt (New England Biolabs, Hitchin, UK), phospho-Akt Ser473 (New England Biolabs, Hitchin, UK), GAPDH (clone 71.1; Sigma-Aldrich, Dorset, UK) and HRP-conjugated secondary antibodies (Jackson Labs, Stratech Scientific Ltd, Newmarket, UK). Each experiment was repeated three times and representative blots are shown.

Cell growth assays were done by seeding 75,000 cells per well in triplicate in six-well plates overnight and then adding GSK-3 inhibitors or an equal volume of carrier [dimethyl sulfoxide (DMSO)] for 72 hr. The GSK-3 inhibitors were purchased from Merck (Nottingham, UK), except for SB216763 (Sigma, Gillingham, UK), CHIR99021 (Stemgent, Cambridge, MA) and CT73911 and CT118637,29 which were kindly provided by Dirksen Bussiere (Novartis, Emeryville, CA). For GSK-3 shRNA clones (see below), cells were counted 16, 72 and 96 hr after plating. Cells were harvested by trypsinization and counted using a hemocytometer. Colony formation assays of stable cell lines were carried out by seeding 1,000 cells in duplicate in 100-mm plates and culturing for 14 days. Colony formation assays for drug treatments were done by plating 7,500 cells per well in 12-well plates for 7 days. In addition to the GSK-3 inhibitors above, we used bicalutamide and MDV3100 (Selleck Chemicals LLC, Houston, TX). Cells were then rinsed with phosphate buffered saline, stained using crystal violet, rinsed again and air dried, and the number of colonies was counted using ImageJ software (http://rsb.info.nih.gov/ij/). In some experiments, colonies were subsequently solubilized in 10% acetic acid and diluted in water, and then absorbance was measured at 590 nm.

Tissue analysis

Formalin-fixed, paraffin-embedded tissue microarray (TMA) slides containing either 24 cores (nine prostate cancer specimens, four paired, one unpaired; 15 normal prostate specimens, seven paired, one unpaired) or 80 cores (74 prostate cancer, six normal prostate specimens), with patient pathology data provided when available, were purchased from US BioMax (Rockville, MA). All sections were independently scored for Gleason sum score by a histopathologist (M.M.W.). Slides were deparaffinized in xylene and serially rehydrated in graded alcohols. Antigen retrieval was done by incubation in 10 mM sodium citrate pH 6.0 and heating in a microwave oven at 560 W for 8 min. Endogenous peroxidase was quenched using 3% H2O2 for 30 min. Antibodies were used at 1:50 [anti-GSK-3α H-12 (Santa Cruz) and anti-Tyr-phosphorylated GSK-3, 612313 (BD Biosciences, Oxford, UK)] or at 1:20 (GSK-3β, 27C10; New England Biolabs, Hitchin, UK) dilution. To determine the specificity of these antibodies for use in immunostaining, they were applied to 22Rv1 cells silenced for each isoform using specific shRNAs. The GSK-3α and GSK-3β antibodies were found to be specific, but the anti-Tyr-phosphorylated GSK-3 antibody recognized nuclear antigens in some cells, and therefore, they could only be used to determine expression levels in the cytoplasm.27 Blocking, antibody incubations and washes were done using the Vectastain Elite ABC Standard kit, according to the manufacturer's instructions (Vector Labs, Peterborough, UK). The DAB Chromagen system (DAKO UK Ltd., Ely, UK) was used to visualize the signal. Slides were briefly counter stained using hematoxylin. TMAs that had been stained for AR as described30 were kindly provided by Professor Manohar Ratnam (Medical University of Ohio, Toledo, OH). TMA analysis was performed using an image analyzer (ChromaVision ACIS II Zeiss, Welwyn Garden City, UK). Tissue samples were scored by two observers for signal strength and intensity.

Analysis of microarray data

In our study, we analyzed the human prostate cancer gene expression dataset published by Taylor et al.31 (available at http://www.cbioportal.org/public-portal/index.do/). Tumors were sorted based on Z-score values of GSK-3A and/or GSK3B. Tumors with a Z-score value of GSK3A or GSK3B < −1 were classified as low GSK3A or GSK3B, and tumors with a Z-score value of GSK3A or GSK3B > 1 were classified as tumors with high GSK3A or GSK3B. In addition, tumors with a Z-value of GSK3A and GSK3B < −0.5 were classified as low total GSK3, and tumors with a Z-value > 0.5 for both isoforms were classified as high total GSK3. The intensity of the androgen-responsive gene signature32 was scored by summing the expression Z-scores per tumor within the human prostate cancer cohort, as previously described.33 To compare the mRNA expression levels in primary prostate tumors, we took advantage of the CGDS-R package provided by the Computational Biology Center of the MSKCC. Gene set enrichment analysis34 was performed with the normalized gene expression levels from the dataset tumors classified as above, and the level of enrichment of the androgen-responsive and GSK-3 gene signatures were calculated using Student's t-test on the collapsed probe sets and by running 1,000 permutations. The control GSK-3 gene signature was obtained by taking 40 genes with the highest score when comparing the gene expression dataset from GSK-3 null double knockout (DKO) and wild-type mouse embryonic stem (mES) cells (data accessible at NCBI GEO database, accession number GSE27395).

Plasmids and transfections

GSK-3α (αsh1 and αsh2), GSK-3β (βsh1 and βsh2) and control shRNA plasmids in pSM2c (Open Biosystems, Madrid, Spain) have been previously described and characterized.27, 28 22Rv1 and LNCaP cells were transfected using Lipofectamine Plus, as previously described.27, 35 For colony formation assays, shRNA-transfected cells were cultured in the presence of 2 μg/ml puromycin for up to 3 weeks and then stained using crystal violet. For generation of stably silenced cell lines, shRNA-transfected 22Rv1 cells were replated and selected for growth in media containing 2 μg/ml puromycin. Six independent 22Rv1 cell clones for each shRNA were amplified and analyzed for expression of GSK-3 by Western blotting and densitometry. Three GSK-3α-silenced clones and five GSK-3β-silenced clones were used for further studies. A pooled cell line expressing a noncoding shRNA and the parental 22Rv1 line were used as controls. Luciferase assays were carried out as previously described27, 36 using the Dual-Glo Luciferase Assay System (Promega Biotech, Madrid, Spain) and SuperTOP/FOPFlash,37 kindly provided by Randall Moon (University of Washington, Seattle, WA), mouse mammary tumor virus (MMTV)-luciferase and pRL-TK (Promega).

Statistical analysis

Cell proliferation, transcription assays and gel quantitation results are presented as means ± standard deviation (SD). Statistical significance was determined by Student's t-test, with a p-value below 0.05 considered to be significant. Immunohistochemical staining results were analyzed by one-way χ2 test or Fisher's exact test (available in the VassarStats web site; http://faculty.vassar.edu/lowry/VassarStats.html).


Increased levels of GSK-3α and cytoplasmic GSK-3β in prostate tumors correlate with AR expression and with low and high Gleason score, respectively

High levels of cytoplasmic GSK-3β have been reported to correlate with clinical stage, Gleason score and AR expression in prostate cancer.15 However, the expression of GSK-3α in prostate cancer has not been reported. We therefore used immunohistochemistry to compare the expression and cellular localization of GSK-3α and GSK-3β in sections of normal prostate and prostate cancer. GSK-3α was present at low levels in normal prostate, where it was found in the cytoplasm of both epithelial (Figs. 1a and 1b, solid arrowhead) and stromal cells. In general, there was more GSK-3α in stromal cells than in epithelial cells (Fig. 1b, dashed arrowhead). In prostate cancer, GSK-3α was detected in 77/79 cases (97%) and was predominantly cytoplasmic (Figs. 1c1f). Elevated GSK-3α expression in tumors (Figs. 1d and 1e), compared to in normal prostate epithelial cells, was observed in 25/79 cases (31.6%), and this correlated with low Gleason sum score (≤7; Table 1). GSK-3α was also detected in stromal cells in 24% of tumors. There was a trend for tumors with stromal GSK-3α to have elevated GSK-3α in tumor cells; however, this did not reach significance or correlate with Gleason score. GSK-3α was present in prostate tumor cell nuclei in 10/79 tumors (12.7%), the majority of which (7/10) were also Gleason score ≤ 7 (Figs. 1g and 1h and Table 1). These results indicate that increased expression of GSK-3α is a feature of low Gleason score prostate cancer, in contrast to what has been reported for GSK-3β.15

Figure 1.

Increased levels of GSK-3α in prostate tumors. (a and b) Low- and high-magnification images showing weak expression in prostate epithelial cells (red arrow) and expression in smooth muscle cells (black arrow). (c and d) Low- and high-magnification images showing strong staining for GSK-3α in a tumor of Gleason score 7. (e) Example of moderate GSK-3α expression in a tumor of Gleason score 7. (f) Example of weak GSK-3α staining in a tumor of Gleason score 10. (g and h) Examples of nuclear GSK-3α (red arrows) in tumors of Gleason score 7; scale bars = 500 μm (a and c) and 50 μm (b, d, e, f and h). (i) Graph showing the proportion of cases with differing levels of GSK-3α in normal prostate and tumors of Gleason score ≤ 7 and Gleason score > 7.

Table 1. Comparisons of GSK-3 expression with Gleason score and AR signal strength
inline image

In our sample set, GSK-3β was not detected in normal prostate epithelial cells, although it was found expressed by stromal cells in some cases (Figs. 2a and 2b). In prostate cancer, GSK-3β was found in the cytoplasm of tumor cells in 24/79 cases (30.4%; Figs. 2c2f), where it correlated with high Gleason sum score (>7; Fig. 2i and Table 1). GSK-3β was rarely detected in prostate tumor stroma; however, the number of normal prostate samples with stromal GSK-3β was too low to determine if stromal GSK-3β is reduced in prostate cancer. GSK-3β was detected in tumor cell nuclei in six cases (Figs. 2g and 2h), all of which were Gleason score ≤ 7. In total, 48% of tumors contained elevated levels of GSK-3α and/or expressed GSK-3β, with 14% showing strong staining for GSK-3α and expression of GSK-3β.

Figure 2.

Increased levels of cytoplasmic GSK-3β in prostate tumors. (a and b) Low- and high-magnification images showing low/negative staining in prostate epithelial cells; arrow shows strongly positive stromal cells. (c and d) Low- and high-magnification images showing cytoplasmic GSK-3β in a tumor of Gleason score 9. (e) Example of cytoplasmic GSK-3β in a tumor of Gleason score 7. (f) Example of cytoplasmic GSK-3β in a tumor of Gleason score 10. (g) Example of nuclear GSK-3β in a tumor of Gleason score 6. (h) Example of nuclear GSK-3β in a tumor of Gleason score 7; scale bars 500 μm (a and c) and 50 μm (b, d, e, f and h). (i) Graph showing the proportion of cases expressing GSK-3β in the nucleus or cytoplasm in normal prostate and in tumors of Gleason score ≤ 7 and Gleason score > 7.

The active form of GSK-3 is Tyr phosphorylated.38 Although there are no antibodies that distinguish between Tyr-phosphorylated GSK-3α and GSK-3β, immunostaining for Tyr-phosphorylated GSK-3 might be useful to identify tumors with high levels of “active” GSK-3. However, as we have previously reported, antibodies raised to Tyr-phosphorylated GSK-3 recognize nonspecific nuclear antigens in prostate cancer cell lines.27 This was also the case in prostate tumors, since tumor cell nuclei stained positively in 39 cases, more than the 16 cases that had nuclear GSK-3α or GSK-3β (an example of a tumor with strong nuclear ‘Tyr-phosphorylated GSK-3’ and low levels of GSK-3α and GSK-3β is shown in Supporting Information Figure S1c). For this reason, we focused on those tumors with cytoplasmic Tyr-phosphorylated GSK-3, which we have shown by gene silencing to be specific.27 Tyr-phosphorylated GSK-3 was detected in the cytoplasm in prostate cancer cells (Supporting Information Fig. S1a), and in some cases both in tumor and stromal cells (Supporting Information Fig. S1b).

As GSK-3 inhibitors reduce AR levels in prostate cancer cell lines,11 we reasoned that the expression of GSK-3 might correlate with that of AR. We therefore examined AR expression in the tumor array composed of 74 tumors. AR was detected in tumor cell nuclei in 55 cases (74%) and was strongly positive in 25 cases (Figs. 3c and 3d). The strength of the GSK-3α signal correlated with that of AR: among the 25 tumors strongly positive for GSK-3α (Figs. 3a and 3b), AR was strongly positive in 12 cases (48%), whereas in the remaining 49 tumors, AR was strong in 13 cases (26%; Fig. 3i and Table 1). The weaker staining of GSK-3β prevented comparison of GSK-3β and AR signal strengths (Figs. 3e3h). However, among high Gleason score tumors, AR expression was strong or moderate in 10/13 GSK-3β-positive cases (77%) compared to 15/37 GSK-3β-negative cases (41%; Fig. 3j). In contrast, there was no such difference among low Gleason score tumors.

Figure 3.

Correlation of GSK-3 and AR expression in prostate tumors. (ad) Low- and high-magnification images of GSK-3α (a and b) and AR (c and d) in a prostate tumor of Gleason score 6. (eh) Low- and high-magnification images of GSK-3β (e and f) and AR (g and h) in a prostate tumor of Gleason score 7; scale bars 500 μm (a, c, e and g) and 50 μm (b, d, f and h). (i and j) Graphs showing the proportion of tumors with (i) weak and strong staining for GSK-3α compared to AR and (j) the proportion of tumors positive or negative for GSK-3β, comparing Gleason scores, compared to AR.

In summary, GSK-3α and GSK-3β are upregulated in a significant number of tumors, with GSK-3α being predominant in those with low Gleason score and cytoplasmic GSK-3β in those with high Gleason score, and both correlating with high expression of AR.

Inhibition of GSK-3 enhances the effects of AR antagonists

We previously reported that two distinct GSK-3 inhibitors, SB216763 and SB415286, reduce the growth of AR-positive prostate cancer cells.11 The effect of SB216763 was confirmed in another study, which also showed that the unrelated GSK-3 inhibitor, AR-A014418, weakly stimulates growth by activating AR.12 More recently, two non-ATP-competitive GSK-3 inhibitors, TDZD-8 and lithium chloride, were reported to inhibit growth of both AR-positive (C4-2) and AR-negative (PC3) prostate cancer cells in tumor xenograft assays.14 We compared the effects of seven ATP-competitive GSK-3 inhibitors [SB216763, AR-A014418, 1-azakenpaullone, 6-bromoindirubin-3′-acetoxime (BIO-acetoxime), CT73911, CT118637 and CHIR99021 (CT73911 and CT118637 are related to CHIR9902129, 39)] and two non-ATP-competitive inhibitors (TWS119 and TDZD-8) on proliferation of the 22Rv1 prostate cancer cell line, which is frequently used as a model for AR-dependent prostate cancer.10 22Rv1 cell proliferation was significantly reduced by SB216763 (48% average reduction in cell number at 72 hr compared to cells treated with the carrier DMSO), CT73911 (27% reduction), CT118637 (19%), AR-A014418 (44%), 1-azakenpaullone (32%) and CHIR99021 (37%) (Supporting Information Fig. S2). These inhibitors did not significantly affect proliferation of HEK293 or PC3 cells at the doses used (data not shown). In contrast, BIO-acetoxime (2.5 μM), TWS119 (3 μM) and TDZD-8 (20 μ) inhibited growth to a similar extent in HEK293 cells (79, 37 and 66% reduction) and 22Rv1 cells (75, 40 and 60%, reduction) (data not shown). Based on these results and given its reported high specificity for GSK-3,40 CHIR99021 was selected for further studies.

The AR antagonists bicalutamide and MDV3100 inhibit proliferation of AR-expressing prostate cancer cells.41 Bicalutamide has weak effects in 22Rv1 cells compared to in androgen-dependent cells,42 whereas MDV3100 inhibits proliferation of castrate-resistant AR-expressing cells43; however, to our knowledge, it has not been tested in 22Rv1 cells. Given the correlation we observed between GSK-3 and AR expression levels, we examined the effects of combining CHIR99021 with either bicalutamide or MDV3100 on 22Rv1 cell growth in colony formation assays (Fig. 4a). Treatment with bicalutamide and MDV3100 alone reduced cell growth by 30–40%, whereas CHIR99021 inhibited growth by 50%, and combined treatment with each AR antagonist and CHIR99021 inhibited growth by 80%. These results suggest that GSK-3 inhibitors could be useful when combined with AR antagonists to treat prostate cancer. To determine if the effects of GSK-3 inhibition on cell proliferation were hormone dependent, experiments were also carried out in cell culture medium containing CSS. Colony formation of 22Rv1 cells was unaffected by MDV3100, slightly increased by bicalutamide and inhibited to 15% by CHIR99021 (Supporting Information Fig. S3a), indicating that GSK-3 may also play a role in hormone-independent proliferation of 22Rv1 cells.

Figure 4.

Effects of silencing of GSK-3 isoforms on AR transcriptional activity and prostate cancer cell growth. (a) Colony formation assays of 22Rv1 cells treated for 7 days with either CHIR99021 (2 μM) or DMSO (carrier) and the AR antagonists MDV3100 (10 μM) or bicalutamide (10 μM). Image shows cells stained using crystal violet; graph shows quantification of solubilized crystal violet with DMSO/DMSO set to 100%; *p < 0.01 vs. DMSO/DMSO; **p < 0.01 vs. DMSO/CHIR99021. Graph shows means ± SD normalized to DMSO/DMSO from three independent experiments. (b) 22Rv1 cells transfected with shRNA plasmids for GSK-3α (αsh), GSK-3β (βsh) or control (ctrlsh) were selected using puromycin for 7 days and the numbers of colonies counted; *p < 0.01; **p < 0.1 vs. ctrlsh. Graph shows means ± SD from three independent experiments. (c) AR gene reporter assays from extracts of 22Rv1 cells transfected with the indicated shRNA plasmids, MMTV-luciferase and renilla and treated ± ligand (DHT) for 48 or 72 hr. Graph shows means ± SD normalized to ctrlsh + DHT. GSK-3α shRNA reduced AR activity at 48 and 72 hr and GSK-3β reduced AR activity at 72 hr; *p < 0.01 vs. ctrlsh+ DHT; n = 3. (d) AR gene reporter assays from extracts of LNCaP cells transfected with the shRNA plasmids described in (b) plus MMTV-luciferase reporter and renilla and treated ± ligand (DHT) for 48 hr. Graph shows means ± SD (normalized to ctrlsh + DHT); n = 2. GSK-3α shRNA reduced AR activity (*p < 0.01 vs. ctrlsh + DHT). The effect of GSK-3β shRNA was not significant at 48 hr. (e) β-Catenin/Tcf gene reporter assays from extracts of 22Rv1 cells transfected with the same shRNA plasmids described in (a) together with Super8xTOPFlash or Super8xFOPflash luciferase reporters and renilla. Silencing GSK-3α and GSK-3β together increased β-catenin/Tcf activity, *p < 0.05 vs. ctrlsh. (f) 22Rv1 cells transfected with shRNA plasmids for GSK-3α (αsh), GSK-3β (βsh), both isoforms together (α/βsh) or control (ctrlsh) were selected for 48 hr with puromycin and extracts probed by Western blotting for total GSK-3, Akt phosphorylated at Ser473 (P-Akt 473), total Akt and GAPDH as a loading control. Graph shows quantitation of blots, normalized to GAPDH, relative to ctrlsh cells; *p < 0.05 vs. ctrlsh; n = 2. (g) Extracts from 22Rv1 cell clones stably expressing shRNAs were probed for total GSK-3, stripped and reprobed for γ-tubulin. Blot quantitation is shown in Supporting Information Figure S4a. (h) 22Rv1 cell clones stably expressing shRNAs were plated at low density for 2 weeks, and colonies were stained using crystal violet and counted; the graph shows colony number for the indicated cell clones normalized to ctrlsh cells; *p < 0.05 vs. ctrlsh; n = 3.

Regulation of AR transcriptional activity and prostate cancer cell growth by GSK-3α and GSK-3β

Our observations are consistent with previous studies showing that inhibition of GSK-3 reduces AR transcriptional activity and prostate cancer cell growth.11, 12 However, the relative importance of each isoform for the activation of AR is not known. We therefore used gene silencing to address this question in 22Rv1 cells. Cells were transiently transfected with plasmids that express shRNAs to silence GSK-3α and/or GSK-3β. Western blotting indicated the extent of silencing of GSK-3α and GSK-3β (Fig. 4f), consistent with our previous studies using the same shRNA plasmids.43, 44 To determine the effects of silencing on cell growth, we conducted colony formation assays. Silencing of either GSK-3α or GSK-3β reduced 22Rv1 colony formation, although the effect was less consistent for GSK-3β (Fig. 4b). We also conducted colony formation assays in hormone-depleted medium. 22Rv1 cells did not tolerate transfection followed by culture in hormone-depleted medium containing puromycin. Nevertheless, we observed a trend for a reduction in the number of colonies formed on silencing of either GSK-3 isoform (Supporting Information Fig. S3b), suggesting that both isoforms of GSK-3 may also play roles in hormone-independent prostate cancer cell growth.

To determine the effects of GSK-3 silencing on AR transcriptional activity, 22Rv1 cells were cotransfected with shRNAs and MMTV-luciferase and treated with the AR ligand dihydrotestosterone (DHT). Silencing GSK-3α reduced AR activity at 48 hr (p < 0.01), and a further reduction was noted at 72 hr (p < 0.001; Fig. 4c). In contrast, silencing GSK-3β had no effect on AR activity at 48 hr, but did reduce it at 72 hr (p < 0.001), albeit to a lower extent, compared to silencing of GSK-3α. Silencing both GSK-3 isoforms together did not further reduce AR activity. Similar results were obtained in LNCaP cells (Fig. 4d). These results suggest that GSK-3α is more important than GSK-3β for maintaining AR transcriptional activity and that GSK-3β may play a less direct role as its effect was observed only at the later time point.

Silencing of both GSK-3α and GSK-3β is normally required to activate β-catenin/Tcf-dependent transcription. To test if this was also the case in 22Rv1 cells, we examined the effects of silencing each isoform on β-catenin/Tcf transcriptional activity using the luciferase reporters Super8xTOPFlash and its control, Super8xFOPFlash, which has mutations at the Tcf binding sites.37 Consistent with previous studies, β-catenin/Tcf activity was not affected by silencing either GSK-3α or GSK-3β alone, but was increased on silencing of both GSK-3 isoforms together (p < 0.05; Fig. 4e). These results suggest that GSK-3 regulation of AR transcriptional activity is independent of Wnt/β-catenin signaling, as the former is observed upon silencing of a single isoform, whereas the latter requires silencing of both isoforms.

We were interested in why silencing of GSK-3α reduced AR activation to a greater extent than silencing of GSK-3β. Although there are many possible reasons for this, recent studies have reported isoform-specific effects of GSK-3 on Akt phosphorylation.25, 42 As Akt/PKB signaling inhibits AR,33, 44 we examined Akt phosphorylation in 22Rv1 cells silenced for each GSK-3 isoform. Interestingly, silencing GSK-3β, but not GSK-3α, reduced Akt Ser473 phosphorylation (Fig. 4f). This may, in part, account for the different effects of silencing each GSK-3 isoform on AR activity, as downregulation of GSK-3β, by reducing Akt phosphorylation, could reduce Akt inhibition of AR.

To determine the effects of chronic silencing of each GSK-3 isoform in prostate cancer cells, we generated stable shRNA-expressing 22Rv1 cell lines, using the control shRNA and two independent shRNAs for each GSK-3 isoform. Selection and expansion of control and GSK-3β shRNA clones was straightforward; however, several GSK-3α shRNA clones could not be maintained (R.M.K., unpublished observations), in keeping with the more consistent effects of acute silencing on colony formation (Fig. 4b). Among the selected cell lines, densitometry analysis of Western blots for GSK-3 indicated a reduction of between 25 and 80% compared to cells expressing control shRNA (Fig. 4g and Supporting Information Fig. S4a). There were no significant differences in growth of the cell lines in short-term proliferation assays (data not shown). However, cell clones with reduced levels of either GSK-3α or GSK-3β formed fewer colonies than control cells in colony formation assays (Fig. 4h), except for clone αsh12, in which silencing was less efficient (Supporting Information Fig. S4a). When colony formation assays were conducted in hormone-depleted medium, cell clones with reduced levels of GSK-3α formed similar numbers of colonies as control cells, whereas cell clones with reduced levels of GSK-3β formed fewer colonies (Supporting Information Fig. S4b). Thus, in colony formation assays, chronic silencing of GSK-3α affects hormone-dependent growth, whereas chronic silencing of GSK-3β affects both hormone-dependent and hormone-independent growth. To determine if the effects of chronic silencing of GSK-3 isoforms are related to AR transcriptional activity, we carried out gene reporter assays in the cell clones. In contrast to the acute silencing experiments, AR transcriptional activity did not correlate with loss of GSK-3 expression (Supporting Information Fig. S4c). In addition, β-catenin/Tcf transcriptional activity was compared, and no consistent differences were observed (Supporting Information Fig. S4d). Together, these results suggest that chronic downregulation of GSK-3 isoforms reduces growth in colony formation assays; however, this may not be directly related to AR or β-catenin/Tcf signaling.

GSK3A and GSK3B mRNA expression levels correlate with an androgen response gene expression signature

To determine if the increases in GSK-3 isoform protein expression also occured at the level of mRNA, we examined the expression of GSK3A and GSK3B in prostate TMA datasets available in Oncomine (www.oncomine.org). We found no consistent differences in GSK3A expression between normal prostate and prostate cancer (2/14 datasets showed an increase and two showed a decrease), whereas GSK3B expression was elevated in prostate cancer in 7/14 datasets, with two showing a decrease (data not shown). Given the correlation we observed between GSK-3 and AR at the protein level in prostate tumors, we analyzed microarray data for expression of GSK3A and GSK3B and genes that define an established androgen-responsive gene signature.32 The androgen-responsive gene expression signature was elevated in tumors with high expression of either GSK3A or GSK3B, as well as with high total GSK3 expression, compared to tumors with low GSK3A or GSK3B (Fig. 5). To extend these results, we examined a gene expression signature obtained from a study of mES cells lacking both isoforms of GSK-3 (DKO cells). This showed that genes that are upregulated in mES cells lacking GSK-3, compared to in control mES cells, are expressed at low levels in those prostate tumors that have high expression of GSK-3 and at high levels in those prostate tumors with low levels of GSK-3. Altogether, these data support a model in which GSK-3 plays an important role in AR-dependent gene expression in prostate cancer.

Figure 5.

Correlation of GSK3A and GSK3B mRNA expression and the AR androgen-responsive gene expression signature. (a) Mean summed Z-score ± SD for the androgen-responsive gene signature, as defined by Hieronymus et al.,32 in tumors with the indicated Z-scores for GSK3A and/or GSK3B. Tumors with low GSK3A and/or GSK3B mRNA expression have a reduced androgen-response gene signature. Significance was determined by the Mann-Whitney U-test (p < 0.001). (b) Gene set enrichment analysis plots of the androgen-responsive gene signature32 are shown for tumors sorted by low GSK3A, low GSK3B or low total GSK3 mRNA expression. (c) mRNA expression levels of the indicated genes in 131 primary tumors from the MSKCC prostate cancer dataset. Gene expression was normalized to Z-score and compared using the CGDS-R package. (d) Heat maps showing enrichment of an androgen-responsive gene signature in tumors with high levels of GSK3A and/or GSK3B mRNAs and enrichment of a GSK3A/B null (DKO) mES gene signature in tumors with low levels of GSK3A and/or GSK3B mRNAs.


We and others have previously shown that GSK-3, and GSK-3β in particular, can play important roles in prostate cancer.10–12, 14, 15, 22, 23, 30, 45 Our study indicates that the two isoforms of GSK-3 are both upregulated in prostate cancer and suggests that they play distinct roles. GSK-3α is predominant in low Gleason score tumors, where it correlates with AR expression, suggesting that it is more important in hormone-dependent prostate cancer. This would be consistent with the stronger inhibitory effect of GSK-3α shRNA on AR transcriptional activity (Figs. 4c and 4d). In contrast, GSK-3β expression correlates with high Gleason score, suggesting that it is more important in advanced prostate cancer, consistent with its survival function in androgen-independent disease.45 Our analysis of microarray gene expression data, comparing expression in normal prostate and prostate cancer, indicated that GSK3B mRNA is upregulated in prostate cancer, but that GSK3A mRNA is not. The reason for the latter is not known, it may reflect differences in postranscriptional regulation or different levels of expression in the stromal cells. Nevertheless, the androgen-responsive gene expression signature correlated with high expression of GSK3A or GSK3B (Fig. 5b), in agreement with the immunohistochemistry data correlating GSK-3 and AR protein levels in tumors.

The cellular localization of GSK-3 can provide clues about its function. Nuclear GSK-3α was detected in 10/79 cases, seven of which were low Gleason score, and nuclear GSK-3β was found only in low Gleason score tumors, consistent with the study of Li et al.15 However, in contrast to that report, we did not detect nuclear GSK-3β in benign or normal prostate tissue, except in one case, where there may be weak nuclear staining (Fig. 2b). GSK-3β can shuttle between the nucleus and the cytoplasm, accumulating in the nucleus in response to proapoptotic stimuli.46 In contrast, GSK-3α does not shuttle between the nucleus and the cytoplasm, but may enter the nucleus on activation of the calcium/calpain pathway, which is implicated in pathological conditions associated with cell death.20 Interestingly, expression of a GSK-3β mutant that is unable to enter the nucleus protects cells from TNF-induced apoptosis.46 Thus, the presence of nuclear GSK-3 in low Gleason score tumors could indicate prostate cancer cells that are under stress and/or preapoptotic.

The predominant cytoplasmic localization of GSK-3α and GSK-3β in prostate tumors suggests that substrates relevant to prostate cancer are cytoplasmic. AR is predominantly nuclear, but shuttles between the nucleus and the cytoplasm. Ligand-induced accumulation of AR in the nucleus is accompanied by phosphorylation at several sites,47, 48 and GSK-3 inhibitors can reduce DHT-induced AR phosphorylation.13 Mutation of Ser650 in AR results in its rapid nuclear export,47 and this has led to the suggestion that GSK-3β regulates AR localization by phosphorylating this site.12 This would be consistent with the effects of GSK-3 inhibitors on nuclear export of AR.13, 24 Further studies will be required to confirm this and to identify isoform-specific GSK-3 phosphorylation sites in AR and other relevant substrates.

The contrasting effects of silencing of GSK-3α and GSK-3β further highlight the possibility that each isoform plays a different role in prostate cancer cells. The acute silencing experiments indicated that AR transcriptional activity is more dependent on GSK-3α than on GSK-3β. One possible reason for this may be the differential effects of each isoform on Akt phosphorylation. Consistent with recent reports,25, 26 silencing of GSK-3β reduced Akt phosphorylation (Fig. 4f), and this should lead to increased AR signaling.33, 44 However, the signaling crosstalk that occurs among GSK-3, Akt and AR is complex, and more detailed studies are required to determine which of them plays a dominant role, and when, as their interactions are likely to change during prostate cancer progression.

The chronic silencing experiments revealed a trend for increased AR activity in cells silenced for GSK-3α and increased β-catenin/Tcf activity in cells silenced for GSK-3β (Supporting Information Fig. S4; data not shown). The reasons for this are unknown, but the latter effect does not involve GSK-3α, as silencing of GSK-3β did not affect expression of GSK-3α, and neither acute nor chronic silencing of GSK-3α affected β-catenin/Tcf activity. Further studies will be required to identify the signals that allow prostate cancer cells to bypass the effects of GSK-3 inhibition. Chronic silencing of GSK-3 isoforms may lead to the emergence of subsets of cells, effectively “resistant” clones that activate AR or promote hormone-independent proliferation. Thus, the development of GSK-3 inhibitor-resistant prostate cancer cell lines may provide useful models for studying prostate cancer progression.

Several GSK-3 inhibitors specifically inhibited the growth of AR-expressing prostate cancer cells (Supporting Information Fig. S2), largely supporting previous observations.10–12 In addition, we showed that combined treatment of cells with AR antagonists and the most specific of the GSK-3 inhibitors, CHIR99021, inhibited prostate cancer cell growth to a greater extent than each alone. It would therefore be worth considering therapies for prostate cancer that combine AR antagonists and GSK-3 inhibitors, particularly as there is a GSK-3 inhibitor in clinical trials for the treatment of neurodegenerative disorders.49 Our observations further suggest that if/when isoform-specific inhibitors become available, GSK-3α inhibitors may be best used to treat early stage prostate cancer, and GSK-3β inhibitors to treat the subset of more advanced tumors that express this isoform.

In conclusion, we have shown that both GSK-3 isoforms are upregulated in prostate cancer; GSK-3α is more prevalent in low Gleason score tumors, where it might promote androgen-dependent proliferation by activating AR, whereas GSK-3β is more important in high Gleason score tumors, where it could be driving androgen-independent AR activation and/or activating other signals, such as those involving Akt.


The authors thank their colleagues for generously providing reagents, in particular Professor Manohar Ratnam (Medical University of Ohio, Toledo, OH) for the tissue microarray stained for androgen receptor.