Polycomb-mediated disruption of an androgen receptor feedback loop drives castration-resistant prostate cancer

Abstract

The lethal phenotype of castration-resistant prostate cancer (CRPC) is generally caused by augmented signaling from the androgen receptor (AR). Here we report that the AR-repressed gene CCN3/NOV inhibits AR signaling and acts in a negative feedback loop to block AR function. Mechanistically, a cytoplasmic form of CCN3 interacted with the AR N-terminal domain to sequester AR in the cytoplasm of prostate cancer cells, thereby reducing AR transcriptional activity and inhibiting cell growth. However, constitutive repression of CCN3 by the Polycomb Group Protein EZH2 disrupted this negative feedback loop in both CRPC and enzalutamide-resistant prostate cancer cells. Notably, restoring CCN3 was sufficient to effectively abolish CPRC cell growth in vitro and in vivo. Taken together, our findings establish CCN3 as a pivotal regulator of AR signaling and prostate cancer progression and they establish a functional intersection between Polycomb and AR signaling in CRPC.

Introduction

Androgen receptor (AR) is an androgen-stimulated transcription factor that critically promotes prostate cancer (PCa) development and progression. Inhibition of AR signaling through hormone-deprivation therapies (ADT) is effective originally on androgen-dependent prostate cancer (ADPC) and yet resistance develops within several months to years of treatment, leading to castration-resistant prostate cancer (CRPC). Metastatic CRPC with resistance to new-generation anti-androgen enzalutamide (also called MDV3100) is a lethal disease. Taking advantage of chromatin immunoprecipitation (ChIP)-based assays, genomic analysis has determined the AR cistrome in various prostate cancer cell lines as well as in primary PCa tissues (1–4). We and other have reported that, in addition to its conventional role in transcriptional activation, AR also acts as a transcriptional repressor that directly inhibits target gene expression, an activity that is collaborated by the epigenetic silencer EZH2 (5,6). Our previous study has nominated CCN3/NOV as one of the most strongly repressed genes by AR and EZH2 (7).

CCN3/NOV is structurally similar to other CCN family of matrix proteins that also include CCN1/CYR61, CCN2/CTGF, CCN4/WISP-1, CCN5/WISP-2 and CCN6/WISP-3 (8,9). CCN3 contains an N-terminal IGFBP domain, a vWF type C (VWC) domain, a trombospondin type I (TSP1) domain, and a C-terminal (CT) cysteine knot domain. It was first discovered in myeloblastosis-associated virus-induced nephroblastoma in chickens (10) and later implicated in a wide-range of cancer types and in a variety of cellular events such as cell proliferation, adhesion, migration and differentiation in a context-dependent manner (11,12). While CCN3 is known as a matricellular, secreted protein, previous studies have reported cytoplasmic and even nuclear CCN3 proteins, the function of which is unknown (13,14). Further, immuocytochemistry experiments have revealed cytoplasmic immunostaining of CCN3 in the epithelial compartment of human prostate tissues (15). However, to date, a majority of studies of CCN3 function in prostate cancer has focused almost exclusively on its secreted form in AR-negative cells (16–18). The roles of cytoplasmic CCN3 and potential regulation of clinically relevant AR-positive prostate cancer cells have not been investigated.

In the present study, we demonstrated substantial amount of intracellular CCN3 protein, which physically interacts with the AR protein to sequester it in the cytoplasm, thereby inhibiting AR nuclear translocation and suppressing AR chromatin targeting and transcriptional activation. CCN3 continues to be repressed in CRPC as well as enzalutamide-resistant prostate cancer, potentially due to increased EZH2 expression in these cells. Importantly, re-expression of CCN3 strongly suppresses enzalutamide-resistant prostate cancer cell proliferation, colony formation, and xenograft tumor growth. Taken together, our data reveal a novel function of the matricellular protein CCN3 in inhibiting AR signaling and thus CRPC progression through its cytoplasmic form and suggest a novel mechanism by which the epigenetic regulator EZH2 indirectly activates AR signaling in CRPC.

Materials and Methods

Cell culture, Constructs, and Stable Cell Lines

LNCaP and 22RV1 cell lines were recently obtained from the American Type Culture Collection (ATCC). C4-2B and abl cells were kind gifts from Drs. Arul Chinnaiyan and Qianben Wang, respectively. We authenticated all the cell lines by morphology check and growth curve analysis, tested free of mycoplasma, and grown in RPMI supplemented with 10% fetal bovine serum, 1% penicillin and streptomycin. MDV3100 was from Selleck Chemicals. Wildtype and mutant CCN3 constructs and EZH2-full length construct were first cloned in pCR8 Gateway Entry vector. The CCN3 constructs were then transferred into the doxycycline-inducible lentiviral vector pLenti-SFB (S-protein, Flag, and a streptavidin-binding peptide), while the EZH2 construct were transferred into and the pLVX Gateway-compatible lentiviral destination vector by LR clonase (Invitrogen). AR-full length and AR fragments were cloned into the pcDNA3.1 vector. pGIPZ Non-silencing Lentiviral shRNA Control (RHS4346) and CCN3-targeting shRNA (V2LHS_152302) were purchased from Open Biosystems as used in previous study (7). Lentiviral supernatants were collected from 293T cells that were transfected with lentiviral constructs along with packaging plasmids pSPAX2 and pMD2G for 48 hours. Supernatants were passed through a 0.45-µm filter, combined with 4 µg/mL polybrene, and then used to infect prostate cancer cells. To generate doxycycline-inducible pLenti-SFB CCN3 in LNCaP/C4-2B cells, 400µg/mL G-418 was used to treat the infected cells to select for stable clones. To induce CCN3 expression, the stable cells were treated with 1mg/ml doxycycline (+DOX), while vehicle-treated cells (−DOX) served as a negative control. To generate LNCaP cells with stable expression of pGIPZ control or CCN3-targeting shRNAs, LNCaP cells were treated with 1µg/mL puromycin 2 days after infection to select for stable clones.

Cell fractionation, co-IP, Western Blotting, Immunostaining, and qPCR

Cell pellet was washed twice by ice cold PBS, resuspended in 10× pellet volume of buffer A (10mM HEPES, 10mM KCl. 1.5mM MgCl₂, 10% glycerol and 1mM EDTA) and incubated in ice for 10min. Then, final concentration of 0.5% Triton X-100 was added to cell suspension, vortexed for 15 sec, and cell homogenate was spun down at 2min at 2,000 × g at 4°C. The supernatant was saved as cytoplasmic fraction and the resulting pellet was the nuclear fraction. S-beads pull down were performed using 30ul S-protein agarose beads (Millipore) for 3hr at 4°C. Co-IP, western blotting and immunostaining were carried out using standard procedure. QPCR was performed using SYBR Green by StepOne Plus. Details are provided in the Supplementary materials and methods. PCR primers and antibodies used in this study are listed in Supplementary Table S1 and S2.

ChIP, ChIP-seq and data analysis

ChIP and ChIP-seq library preparation and sequencing were performed as described previously (2). ChIP-seq reads were analyzed using the HOMER suite. Weighted venn diagrams were created by R package Vennerable. The motif percentage of occurrence was created by R packages: ggplot2, scales and gridExtra. Microarray expression profiling was performed using HumanHT-12 v 4.0 Expression BeadChip (Illumina). Bead-level data were preprocessed and normalized by GenomeStudio. Differentially expressed genes were identified by Bioconductor limma package (cutoff p<0.005). Heatmap view of differentially expressed genes was created by Cluster and Java Treeview. GO term enrichment was analyzed using DAVID and plot was drawn by R package ggplot2. GSEA was performed as described previously (2). The GEO accession number for the data reported in this study is GSE79358.

In vitro Functional assays

Cell proliferation, cell invasion, and luciferase reporter assays were carried out using the WST-1 kit according to the manufacturer’s instruction (Clontech), the Boyden Chamber assay (Millipore), and the Dual-Luciferase Reporter Assay System (Promega), respectively. For soft agar assay, 5000 cells were resuspended in 0.4% agar in culture medium and layered on bottom agar containing 0.7% agar in the same medium. After 3 weeks of incubation, the cells on soft agar were fixed and stained by 0.05% crystal violet. The colonies were then imaged and counted for quantification.

Results

Genomic analysis reveals CCN3/NOV regulation of AR signaling

We have recently reported CCN3 as a prototype target gene of AR-mediated transcriptional repression(7). To gain insights into the potential functions of CCN3, we generated doxycycline (Dox)-inducible CCN3 overexpression in LNCaP cells, which express a relative lower level of endogenous CCN3 compared to benign cell lines, and performed microarray analysis (Fig. 1A and Supplementary Fig. S1A). Microarray analysis of replicated experiments identified 594 genes that were upregulated upon CCN3 overexpression and 675 genes that were downregulated. Gene Ontology (GO) analysis revealed that CCN3-repressed genes are strongly enriched in the regulation of cell cycle, mitosis, DNA replication, and condensed chromosome, while CCN3 appears to induce chemotaxis, cell death, and immune response, indicating potentially important roles of CCN3 in prostate oncogenesis (Fig. 1B and C). Notably, CCN3-regulated genes are involved in steroid metabolism and hormone biosynthesis, suggesting that CCN3, a target gene of AR, might in turn alter androgen signaling. In agreement with this, hierarchical clustering and heatmap analysis of CCN3-induced and -repressed genes revealed that CCN3 overexpression substantially abolished androgen-stimulated gene expression patterns (Fig. 1D). Further, Gene Set Enrichment Analysis (GSEA) demonstrated that androgen-induced genes were markedly down-regulated following CCN3 overexpression (Fig. 1E), whereas androgen-repressed genes were significantly upregulated (Fig. 1F), supporting an inhibitory role of CCN3 on AR-mediated transcriptional activities. Concordantly, qRT-PCR analysis confirmed that androgen-induced genes, such as PSA and KLK2, were significantly repressed by CCN3 (Fig. 1G). By contrast, androgen-repressed genes, including OPRK1 and AMIGO2, were increased following CCN3 overexpression (Fig. 1H). Moreover, CCN3 overexpression is able to further inhibit residual AR activity even under androgen-deprived conditions. These findings were confirmed in another prostate cancer cell line C4-2B, wherein GSEA analysis showed that androgen-induced genes were enriched for downregulation by CCN3, whereas androgen-repressed genes were restored upon CCN3 overexpression (Supplementary Fig. S1B and S1C). QRT-PCR analysis further corroborated that CCN3 overexpression inhibits AR-mediated transcriptional regulation of target genes in C4-2B cells (Supplementary Fig. S1D). Taken together, our data suggest that CCN3 plays important roles in negative regulation of AR signaling.

Figure 1. CCN3/NOV inhibits AR transcriptional activities.

A–C. GO analysis suggests CCN3 as a regulator of cell cycle and steroid hormone metabolism. LNCaP cells were transduced with lentivirus containing inducible CCN3-SFB construct and CCN3 expression was induced by 1ug/ml doxycycline and confirmed by western blotting (A). CCN3-repressed (B) and –induced (C) genes were determined by replicate microarray profiling of LNCaP cells with or without Dox induction and subsequently subjected to GO analysis. The GO-enrichment p values and the number of genes in each GO category are indicated at the x-axis and next to the bar, respectively.

D. Heatmap view showing CCN3-regulation of androgen-regulated genes. Ethanol and R1881-stimulated LNCaP cells with control or Dox-induced CCN3 expression were profiled by duplicate microarray experiments and clustered based on CCN3-induced/-repressed gene expression. Genes also regulated by R1881 are indicated by side bars. Each row represents one gene, and each column corresponds to one microarray experiment.

E–F. CCN3 inhibits AR-mediated gene expression program. GSEA was performed to determine the enrichment of AR-induced (E) and AR-repressed (F) gene sets in gene expression dataset profiling control and CCN3-overexpressing LNCaP cells.

G–H. CCN3 inhibits AR transcriptional activity. Control and CCN3-overexpressing LNCaP cells stimulated with ethanol or R1881 was subjected to qRT-PCR analysis of representative AR-induced (G) and -repressed (H) genes. Data were normalized to GAPDH. Data shown is mean (±SEM) of technical replicates from one representative experiment out of 3. *p<0.05; ** p<0.01.

CCN3 inhibits AR nuclear translocation

To elucidate the mechanism by which CCN3 inhibits AR-mediated transcriptional activities, we first examined whether CCN3 affects AR mRNA and protein levels. QRT-PCR and western blot analysis showed that AR mRNA and protein levels remained unchanged upon CCN3 depletion (Supplementary Fig. S2A and S2B). Although CCN3 is originally known as a secreted protein, previous studies have suggested intracellular CCN3 in prostate epithelial cells (15). We hypothesized that CCN3 might inhibit AR signaling through impairing its nuclear translocation. Western blot analysis indeed revealed intracellular CCN3 localized exclusively in the cytoplasm, whereas AR is present mostly in the nuclei but also abundant in the cytoplasm (Supplementary Fig. S2C). Immunofluorescent (IF) staining further showed cytoplasmic localization of CCN3 in both hormone-deprived and androgen-simulated PCa cells (Supplementary Fig. S2D).

To determine the effects of CCN3 on AR nuclear translocation, we overexpressed CCN3 in LNCaP cells. Importantly, western blot analysis showed a significant decrease of AR protein level in the nuclear fraction following CCN3 overexpression (Fig. 2A). Concordantly, IF staining demonstrated remarkable sequestration of AR in the cytoplasm of the cells with ectopic CCN3 overexpression, which, similar to endogenous CCN3, also localized in the cytoplasm (Fig. 2B and Supplementary Fig. S2E). This regulation is likely independent of ligand-AR interaction, since CCN3 retains AR in the cytoplasm in both ethanol and R1881-stimulated LNCaP cells. Similar inhibition of AR nuclear translocation by CCN3 was confirmed using another PCa cell line C4-2B (Fig. 2C–D; Supplementary Fig. S2F). A major mechanism to prostate cancer resistance to androgen deprivation therapy is alternative splicing of the AR gene leading to constitutively active AR variants (AR-Vs) lacking the ligand-binding domain (19,20). It is thus of interest to test whether CCN3 regulates nuclear localization of AR-Vs. We examined 22RV1 and VCaP cells that have been shown to express AR-Vs and found that CCN3 overexpression greatly decreased not only full-length AR but AR-V protein levels in the nuclei (Fig. 2E and F).

Figure 2. CCN3 reduces AR nuclear translocation.

A, C. CCN3 overexpression decreases nuclear AR protein level. Ethanol or R1881-stimulated control and CCN3-overexpressing LNCaP (A) and C4-2B (C) cells were subjected to cytoplasmic and nuclear protein fractionation followed by western blog analysis using anti-Flag (CCN3) and anti-AR antibodies, respectively. GAPDH and H3 were used as cytoplasmic and nuclear protein loading controls, respectively.

B, D. CCN3 inhibits AR translocation into the nucleus. LNCaP (B) and C4-2B (D) cells transiently transfected with CCN3 were hormone-starved for 3 days and treated with ethanol or 1nM R1881 for 24 hours before subjected to immunofluorescent staining using anti-Flag (stain for CCN3) and anti-AR antibodies. Scale bar, 10um.

E–F. CCN3 decreases AR and AR-V7 nuclear translocation in 22RV1VCaP cells. Cytoplasmic and nuclear protein fraction of 22RV1 (E) and VCaP (F) cells were analyzed by western blotting using anti-Flag (CCN3) and anti-AR antibodies, respectively. GAPDH and H3 were used as cytoplasmic and nuclear protein loading controls, respectively.

G. Schematic representation of CCN3 signal peptide deletion mutant (CCN3_ΔSP) used in this study.

H. Western blot shows the distribution of protein in cell pellet or culture medium fraction. Anti-Flag (upper panel) and anti-CCN3 (lower panel) antibodies were used to detect CCN3_WT-SFB and CCN3_ΔSP-SFB.

I. Intracellular CCN3 inhibits AR translocation into the nucleus. LNCaP cells ectopically expressing CCN3_ΔSP-SFB were hormone-starved for 3 days and treated with ethanol or 1nM R1881 for 24 hours before subjected to immunostaining using anti-Flag (stained for CCN3_ΔSP) and anti-AR antibodies. Scale bar, 10um.

J. QRTPCR analysis of AR downstream genes in ethanol or R1881-stimualted LNCaP_Ctrl, LNCaP_CCN3_WT or LNCaP_ CCN3_ΔSP cells. Data were normalized to GAPDH. Data shown is mean (±SEM) of technical replicates from one representative experiment out of 3. *p<0.05; ** P<0.01”.

To further assure that this regulatory role is mediated by cytoplasmic, but not the secreted form of CCN3, we created a non-secretive mutant of CCN3 (CCN3_ΔSP) by deleting the N-terminal signal peptide (Fig. 2G). Western blot confirmed that all CCN3_ΔSP proteins remained in the cell pellet, while ~60% of the wildtype (CCN3_WT) was secreted into cell supernatant (Fig. 2H). Further, CCN3_ΔSP recapitulated the ability of full-length CCN3 to detain AR in the cytoplasm and inhibit AR downstream gene expression (Fig. 2I–J; Supplementary Fig. S2G). On the contrary, addition of synthetic CCN3 into culture medium had minimal effect on AR nuclear translocation and target gene expression (Supplementary Fig. S3A and S3B). Taken together, our data suggest abundant cytoplasmic CCN3 that inhibits nuclear translocation of AR and decreases AR and AR-V protein levels in the nuclei.

CCN3 and AR proteins physically interact

Next, we attempted to determine how CCN3 prevents AR nuclear translocation. We hypothesized that cytoplasmic CCN3 binds to the AR protein to sequester AR in the cytoplasm. To test this, we performed co-immunoprecipitation (co-IP) assay to pull down endogenous CCN3 from LNCaP cell lysates. Western blot analysis readily detected AR in CCN3-containing complex (Fig. 3A). Further, we co-expressed AR protein with SFB-tagged FOXA1 (as positive control), SFB-CCN3, or SFB-empty vector in 293T cells, and the cell lysates were subjected to pull down by S-beads. Western blot analysis confirmed previously reported interaction between AR and FOXA1 and, most importantly, clearly detected AR protein being pulled down by SFB-CCN3 but not the SFB-vector (Fig. 3B). To further map out the subdomains of CCN3 responsible for its interaction with AR, we expressed a series of SFB-tagged deletion mutants of CCN3 along with the full-length AR protein in 293T cells. IP pull down by S-beads followed by western blot detection revealed that only the TSP1 and CT domain-containing fragment (D2), but not other CCN3 domains, is able to pull down AR (Fig. 3C–D). On the other hand, HA-tagged AR or its deletion mutants were co-expressed along with SFB-tagged CCN3 in 293T cells and the cell lysates were subjected to S-beads pull down. Western blot analysis using anti-HA demonstrated that N-terminal domain (NTD) of AR interacts with the CCN3 protein (Fig. 3E and F). Lastly, to address whether CCN3-AR interaction is necessary for CCN3 to inhibit AR nuclear translocation, we overexpressed full-length CCN3 (FL) or its deletion mutants (D1 and D4) that are incapable of interacting with AR protein in LNCaP cells. Immunofluorescent staining showed that, unlike CCN3-FL, CCN3-D1 and -D4 mutants failed to sequester AR in the cytoplasm (Fig. 3G and Supplementary Fig. S2H). Therefore, our data strongly argue that CCN3 physically interacts with the AR protein to retain it in the cytoplasm, thereby reducing AR nuclear translocation.

Figure 3. CCN3 interacts with the N-terminal domain of the AR protein.

A. Endogenous CCN3 and AR proteins interact. LNCaP cells were subjected to co-immunoprecipitation (co-IP) using an anti-CCN3 antibody or normal rabbit IgG control followed by western blot detection using anti-AR and anti-CCN3 antibodies.

B. Ectopically expressed CCN3 and AR proteins interact with each other. Empty SFB vector, FOXA1-SFB (positive control), or CCN3-SFB were co-expressed with pcDNA3.1 AR in HEK293T cells. Whole cell lysates (input) and lysates immunoprecipitated for SFB-tagged proteins by S-beads were subjected to western blot analysis using anti-Flag (for SFB-tagged proteins) and anti-AR antibodies.

C–D. C-terminal end of CCN3 protein interact with AR. Various SFB-tagged CCN3 deletion mutants (C) were co-transfected with pcDNA3.1 AR into HEK293T cells. Cell lysates were then subjected to pull-down experiments using S-beads and eluted protein complexes analyzed by western blotting using anti-Flag (for SFB-tagged proteins) and anti-AR antibodies (D).

E–F. AR N-terminal domain (NTD) interacts with the CCN3 protein. Various HA-tagged AR deletion mutants (E) were co-transfected with CCN3-SFB constructs in to HEC293T cells, which were then subjected to S-beads pull down (of CCN3) and western blot detection using anti-Flag (for SFB-tagged CCN3) and anti-HA (for AR mutants) antibodies (F).

G. CCN3-AR protein interaction is necessary to inhibit AR nuclear translocation. Full-length CCN3 or CCN3 mutants (D1 and D4) lacking C-terminal AR binding domain were transiently expressed in LNCaP cells and then subjected to immunostaining using anti-Flag (green, for CCN3) and anti-AR (red) antibodies. Scale bar, 10um.

CCN3 reduces AR chromatin binding and transactivation

Since CCN3 decreases nuclear AR level, next we attempted to examine how CCN3 regulates AR cistrome. We performed AR ChIP-seq in LNCaP cells with either vector or CCN3 overexpression. ChIP-seq analysis revealed significantly reduced total number of AR binding events in LNCaP cells following CCN3 overexpression (Fig. 4A). Heatmap view and average intensity plots demonstrated a global decrease of ChIP enrichment or AR binding intensity at nearly all target sites in CCN3-overexpressing cells (Fig. 4B and C). UCSC Genome Browser view confirmed remarkable decrease of AR binding at several known AR-occupied regions including PSA, KLK2 and TMPRSS2 enhancers (Fig. 4D and Supplementary Fig. S4A). To further validate this, we performed AR ChIP-qPCR on several known AR target genes using site-specific primers. Indeed, AR recruitment to enhancers of PSA, KLK2, TMPRSS2 and FKBP5 was significantly decreased upon CCN3 overexpression (Fig. 4E). Similar results were also observed in androgen-independent cell lines C4-2B (Supplementary Fig. S4B). Further, ChIP-qPCR demonstrated that CCN3 knockdown, on the other hand, greatly potentiated AR binding at target enhancers (Fig. 4F). This occurred in both androgen-deprived and androgen-stimulated cells, being concordant with the ability of CCN3 loss to induce nuclear AR level under both conditions. Moreover, luciferase reporter assay showed that, while PSA promoter/enhancer construct can be efficiently activated by androgen stimulation, such response was significantly hampered by CCN3 overexpression (Fig. 4G). Taken together, our results support important roles of CCN3 in suppressing AR chromatin occupancy and transcriptional activation.

Figure 4. CCN3 expression inhibits AR chromatin occupancy.

A–C. CCN3 overexpression inhibits AR binding events. Control and CCN3-overexpressing LNCaP cells were analyzed by AR ChIP-seq to determine the number of control-only, shared, or CCN3-only AR binding sites (A). AR ChIP-seq read intensity around the three categories of AR binding sites in both cell lines were comparatively visualized using Heatmap View (±5kb) (B) and Average Read Intensity Plots (C).

D. Genome browser view of ChIP-seq AR binding events around known AR target gene PSA (upper panel) and KLK2 (bottom panel) in control and CCN3-overexpressing LNCaP cells.

E. ChIP- qPCR analysis of known AR target genes. AR ChIP was performed in LNCaP_Ctrl and LNCaP_CCN3 cells followed by qPCR analysis using primers flanking AR binding sites of several AR downstream genes. KIAA0066 is used as a negative control. Data shown is mean (±SEM) of technical replicates from one representative experiment out of 3. *p<0.05; ** p<0.01.

F. CCN3 knockdown enhances AR genome occupancy at target genes. AR ChIP was performed in androgen-depleted or –stimulated LNCaP_shCtrl or LNCaP_shCCN3 cells followed by qPCR using primers flanking AR-bound enhancers of target genes. KIAA0066 is used as a negative control. Data shown is mean (±SEM) of technical replicates from one representative experiment out of 3. *p<0.05; ** p<0.01.

G. CCN3 inhibits AR transcriptional activity. A luciferase reporter construct consisted of PSA enhancer and promoter elements was transfected into control, CCN3 overexpression (G). Data shown is mean (±SEM) of technical replicates from one representative experiment out of 3. *p<0.05; ** p<0.01.

CCN3 inhibits prostate cancer cell proliferation and xenograft tumor growth

As AR is a key driver of prostate cancer growth, we next investigated the roles of CCN3 in prostate cancer cells. Being consistent with its role in inhibiting AR signaling, CCN3 overexpression markedly suppressed LNCaP cell growth, which was fully recapitulated by the overexpression of its cytoplasmic form CCN3_ΔSP (Fig. 5A). Indeed, addition of synthetic CCN3 into the culture medium had no effect on cell proliferation (Supplementary Fig. S5A). Next, we examined the various forms of CCN3 in regulating cell invasion. Our data showed that overexpression of cytoplasmic CCN3_ΔSP, but not full-length CCN3, markedly inhibited prostate cancer cell invasion (Fig. 5B), suggesting that extracellular CCN3 may have stimulatory effects on cell invasion, as previously reported (16). Addition of synthetic CCN3 into the culture medium, indeed, stimulated cell invasion (Supplementary Fig. S5B). Therefore, wildtype CCN3, comprised of both intracellular and secreted forms, inhibits PCa cell growth, without overall effects on cell invasion.

Figure 5. CCN3 inhibits prostate cancer cell proliferation in vitro and tumor growth in vivo.

A. Intracellular CCN3 inhibits prostate cancer cell growth. Equal numbers of LNCaP_Ctrl, LNCaP_CCN3_WT or LNCaP_CCN3_ΔSP cells were subjected to WST-1 cell growth assay.

B. Intracellular CCN3 reduces cell motility. LNCaP_Ctrl, LNCaP_CCN3_WT, LNCaP_CCN3_ΔSP cells were evaluated for cell invasion using Boyden transwell assay. Representative images of invaded cells are shown in the left panel and the quantitative result shows the mean (±SEM) number of cells in each view in the right panel.

C. CCN3 expression is downregulated in CRPC relative to localized, hormone-naïve prostate cancer. Boxplot analyses showed the expression of CCN3 in three publically available datasets.

D. CCN3 expression suppresses anchorage-independent C4-2B cell growth in a soft agar assay. C4-2B_Ctrl or C4-2B_CCN3 cells were plated in soft agar for colony formation assays. The number of colonies in each well were quantified and shown as mean (±SEM). ** p<0.01.

E. CCN3 overexpression attenuates CRPC cell growth. Equal numbers of control or CCN3-overexpression C4-2B cells were plated for WST-1 cell growth assay.

F. CCN3 knockdown enhances LNCaP anchorage independent cell growth. LNCaP_shCtrl or LNCaP_shCCN3 cells were plated in soft agar for colony formation assays. The number of colonies in each well were quantified and shown as mean (±SEM). ** p<0.01.

G. CCN3 knockdown drives androgen-independent LNCaP cell growth. Control and CCN3-knockdown LNCaP cells were hormone-starved and analyzed by WST-1 cell growth assay.

H. CCN3 knockdown promotes xenograft LNCaP prostate cancer growth. LNCaP cells stably expressing shCtrl or shCCN3 were injected subcutaneously into the dorsal flank of nude mice. Tumor size was evaluated weekly. Tumors were excised from the euthanized mice and representative images were shown. Tumor weight was presented as mean (±SEM). *p<0.05.

To evaluate the clinical relevance of this regulation, we examined CCN3 expression in primary patient samples and observed highly significant down-regulation of CCN3 transcript in CRPC relative to localized prostate tumors in three independent microarray profiling datasets (Fig. 5C). Consistently, immunohistochemistry staining showed that CCN3 protein expression is significantly downregulated in CRPC relative to PCa (Supplementary Fig. S5C and S5D). To examine whether CCN3 loss is causatively related to CRPC, we first tested whether its re-expression abolishes CRPC cell growth. In an anchorage-independent cell growth assay, colony formation of the androgen-independent C4-2B cells was diminished following CCN3 overexpression (Fig. 5D). Further, WST-1 assay showed that CCN3 overexpression greatly suppressed the growth of CRPC cell lines including C4-2B, 22Rv1, and LNCaP-abl (Fig. 5E; Supplementary Fig. S5E and S5F). CCN3 knockdown, on the other hand, significantly increased androgen-independent cell growth (Fig. 5F) and colony formation (Fig. 5G) of LNCaP cells under hormone-depleted conditions. Moreover, we found that CCN3-knockdown cells formed significantly bigger xenograft tumors than the control cells (Fig. 5H), further supporting the role of CCN3 in suppressing CRPC progression.

Further, to evaluate the potential role of the CCN3-AR axis in the clinically challenging CRPC with resistance to the new-generation antiandrogen enzalutamide (MDV3100), we generated enzalutamide-resistant prostate cancer cells by continuously growing C4-2B cells in 10µM MDV3100 over the course of several months. PSA was downregulated by MDV3100 as expected and, although recurred after long-term treatment, remained at a lower level in MDV3100-resistant (MDV^R) C4-2B cells (Fig. 6A). CCN3 mRNA level was initially increased upon short-term MDV3100 treatment, potential due to AR inhibition, but was further down-regulated in MDV^R cells. This is potentially due to EZH2, another upstream repressor of CCN3, which was markedly up-regulated in enzalutamide-resistant prostate cancer cells. Western blot analysis confirmed the increase of EZH2 protein and a remarkable decrease of CCN3 protein in MDV^R C4-2B and VCaP cells (Fig. 6B). In addition, CCN3 expression levels were reduced in patients that were treated with enzalutamide in a recently-published dataset (21) (Fig. 6C).

Figure 6. CCN3 re-expression suppresses enzalutamide-resistant prostate cancer growth.

A. QRT-PCR showing CCN3, EZH2 and PSA expression in C4-2B cells following time-course MDV3100 treatment. Gene expression was normalized to GAPDH. Error bars are mean (±SEM).

B. CCN3 downregulation in MDV3100-resistant PCa. C4-2B and VCaP cells were treated with DMSO, 10uM (C4-2B) or 30uM of MDV3100 (VCaP) for several months to establish resistant cell lines. Cell lysates were subjected to western blot analysis using anti-CCN3, anti-EZH2, and anti-GAPDH antibodies.

C. CCN3 expression is downregulated in PCa patients post enzalutamide therapy. CCN3 expression in publicly available datasets was comparatively plotted in metastatic prostate cancer with or without receiving abiraterone and enzalutamide therapy.

D. CCN3 re-expression inhibits enzalutamide-resistant CPRC cell growth. Colony formation assay was conducted to monitor the growth of enzalutamide-resistant MDV^R C4-2B cells with or without CCN3 overexpression.

E. Schematic diagram illustrating C4-2B xenograft tumor establishment and treatment.

F–G. CCN3 re-expression inhibits xenograft CRPC tumor growth in vivo. Dox-inducible MDV^R C4-2B cells were inoculated subcutaneously to the right flank of nude mice. Upon initial tumor establishment, the mice were then castrated. Once CRPC tumors re-grew to its pre-castration size, the mice were then treated with 10mg kg−1 MDV3100 alone or in combination with Dox feed to induce CCN3 overexpression for 5 days per week. Tumor volumes were measured every 4 days and shown as mean ±SEM (F). *p<0.005 by linear mixed effects model. At the end point, xenograft tumors were harvested and subjected to IHC staining for Ki-67 and AR levels (G). Scale bar, 10um.

Next, we investigated whether CCN3 re-expression is able to suppress enzalutamide-resistant CRPC. Using a doxycycline (Dox)-inducible system, we showed that CCN3 overexpression upon doxycycline treatment dramatically reduced MDV^R C4-2B colony formation, as compared to the control cells (Fig. 6D). To further test whether CCN3 re-expression is able to rescue EZH2-induced effects, we first established an EZH2-overexpressing LNCaP cells and confirmed that EZH2 overexpression markedly promoted cell growth and colony formation, along with reduced CCN3 expression (Supplementary Fig. S6A–S6D). Importantly, CCN3 re-expression in these cells strongly blocked the oncogenic effects of EZH2 overexpression. To examine the therapeutic effects of CCN3 restoration in vivo, we inoculated Dox-inducible MDV^R C4-2B subcutaneously into nude mice. Once initial tumors were established, the mice were castrated and CRPC tumors were randomized to receive enzalutamide alone or in combination with Dox feed that induces CCN3 overexpression (Fig. 6E). Our data showed that Dox-induced CCN3 overexpression significantly retarded enzalutamide-resistant CRPC tumor growth (Fig. 6F). Western blot analysis of the xenograft tumors confirmed CCN3 overexpression in the dox-fed mice (Supplementary Fig. S6E). Immunostaining demonstrated markedly decreased nuclear AR level and Ki-67-positive cells in tumors with CCN3 overexpression, suggesting that CCN3 re-expression inhibited AR nuclear translocation and thus prostate cancer cell proliferation in vivo (Fig. 6G). Taken together, our results support that CCN3 loss plays an important role in driving enzalutamide-resistant CRPC growth.

Discussion

CCN3 is previously known as a secreted protein. Although intracellular CCN3 has been reported (15), their expression has not been well demonstrated and their roles poorly understood. In this study, we demonstrated cytoplasmic CCN3 protein in prostate cancer cells, which interacts with the N-terminal domain of the AR protein to inhibit nuclear translocation of both AR and its variants, thereby abrogating AR-mediated transcriptional program. Using a CCN3 mutant lacking the N-terminal signal peptide for secretion, we demonstrated that the secreted form of CCN3 does not regulate AR signaling. Concordantly, we found that the intracellular CCN3 is able to inhibit prostate cancer cell growth as well as cell invasion, due to its role in suppressing AR signaling. On the contrary, secreted CCN3 does not regulate cell growth, but promotes cell invasion, which is consistent with previously reported roles of CCN3 as a matricellular protein supporting cell adhesion and invasion (17). Altogether, CCN3, including its intracellular and extracellular forms, does not strongly affect cell invasion. We found the role of CCN3 on cell proliferation highly important, as CCN3 overexpression strongly suppresses prostate cancer cell growth, whereas CCN3 depletion is a main driver of AR signaling and androgen-independent CRPC progression and drug resistance.

Of clinical relevance, we observed a significant decrease of CCN3 expression in CRPC compared to localized prostate cancer in various PCa patient cohorts. In addition, we found that CCN3 is further down-regulated when CRPC progresses to enzalutamide resistance. This continued down-regulation is in large extent due to the upregulation of EZH2, another important repressor of CCN3. Recently, the expression of AR variants, especially AR-V7, that are constitutively active in the absence of androgen, have been shown to be a critical driver of the AR oncogenic transcription program necessary for cell survival, leading to enzalutamide resistance (22,23). Molecular targets and therapeutic strategies that are capable of deactivating both AR and AR variants hold great promise in effective treatment of enzalutamide-resistant prostate cancer. Here, we showed that CCN3 re-expression is able to sequester both AR and AR variants in the cytoplasm, and thus may be effective in the treatment of late-stage enzalutamide-resistant prostate cancer. Indeed, our in vitro and in vivo data demonstrated that CCN3 restoration strongly inhibited Enzalutamide-resistant prostate cancer cell proliferation, colony formation, and xenograft tumor growth, with concomitant reduction of nuclear AR levels.

In summary, our study suggests a working model wherein AR-repressed tumor suppressor gene CCN3 sequesters AR in the cytoplasm, forming a negative feedback loop, which however is disrupted in CRPC due to EZH2 up-regulation and continued CCN3 repression (Supplementary Fig. S7). CCN3 loss contributes significantly to androgen-independent AR activation and prostate cancer progression and CCN3 restoration is highly effective in suppressing late-stage enzalutamide-resistant CRPC growth. Therefore, our results establish CCN3 as an important regulator of prostate cancer and indicate a novel mechanism by which the epigenetic regulator EZH2 indirectly activates AR signaling and favoring CRPC progression.