CCN3/NOV gene expression in human prostate cancer is directly suppressed by the androgen receptor

Abstract

Androgen receptor (AR) has essential roles during prostate cancer progression. With genome-wide AR-binding sites mapped to high resolution, studies have recently reported AR as a transcriptional repressor. How AR inhibits gene expression and how this contributes to prostate cancer, however, are incompletely understood. Through meta-analysis of microarray data, here we nominate nephroblastoma overexpressed (NOV) as a top androgen-repressed gene. We show that NOV is directly suppressed by androgen through the AR. AR occupies the NOV enhancer and communicates with the NOV promoter through DNA looping. AR activation recruits the polycomb group protein EZH2, which subsequently catalyzes histone H3 lysine 27 tri-methylation around the NOV promoter, thus leading to repressive chromatin remodeling and epigenetic silencing. Concordantly, AR and EZH2 inhibition synergistically restored NOV expression. NOV is downregulated in human prostate cancer wherein AR and EZH2 are upregulated. Functionally, NOV inhibits prostate cancer cell growth in vitro and in vivo. NOV reconstitution reverses androgen-induced cell growth and NOV knockdown drives androgen-independent cell growth. In addition, NOV expression is restored by hormone-deprivation therapies in mice and prostate cancer patients. Therefore, using NOV as a model gene we gained further understanding of the mechanisms underlying AR-mediated transcriptional repression. Our findings establish a tumor-suppressive role of NOV in prostate cancer and suggest that one important, but previously underestimated, manner by which AR contributes to prostate cancer progression is through inhibition of key tumor-suppressor genes.

Introduction

Prostate cancer is a highly prevalent disease in American men. Although organ-confined prostate cancer can be effectively treated with surgical or radiation therapy, late-stage castration-resistant prostate cancer is essentially lethal. The growth and progression of prostate cancer is highly dependent on the transcriptional activities mediated by the androgen receptor (AR), a hormonal transcriptional factor.¹ Once bound by androgen, AR turns on prostate-specific gene expression such as prostate-specific antigen (PSA). With recent advances in high-throughput sequencing approaches, the genomic landscape of AR has been carefully mapped to high resolution.^2–5 By integrating with genome-wide expression profiling data, much progress has been made recently in the understanding of AR transcriptional regulation. In addition to its well-understood roles in gene activation, AR was recently shown to act as a transcriptional repressor.^6,7 However, in contrast to the plethora of AR-induced genes reported in the literature, only few have been shown directly inhibited by the AR.^8–10 A majority of the AR-repressed genes are yet to be identified. Further, how AR directly suppresses gene expression is not fully understood and how AR-mediated repression contributes to prostate cancer remains to be characterized.

Nephroblastoma overexpressed (NOV), also known as CCN3, was initially discovered in nephroblastoma.^11–13 Owing to structural similarity, CCN3/NOV belongs to the CCN family of matrix proteins that also include CCN1/CYR61, CCN2/CTGF, CCN4/WISP-1, CCN5/WISP-2 and CCN6/WISP-3 (refs 14,15). NOV is a critical regulator of cell differentiation.¹⁶ It regulates mesenchymal stem cell differentiation and maturation of committed blood cells from hematopoietic stem or progenitor cells.¹⁶ NOV has been implicated in a range of cancers including gliomas, melanoma, Ewing's sarcoma, leukemia and breast cancer.^14,17–19 It has been reported to exhibit both tumor-suppressive or oncogenic roles depending on the cellular context.^14,17,20 A majority of studies showed that NOV inhibits cell proliferation in cancer cells such as Ewing's sarcoma, melanoma and breast cancer, while conflicting reports are also abundant.^18,21,22 Full-length NOV is secreted to the extracellular space and is able to regulate integrins to increase cell adhesion to extracellular matrix. In melanoma, this increase in cell adhesion was found to enhance metastasis.^21,23 In breast cancer, however, NOV has been shown to interact with gap junction protein connexin43 (Cx43)^22,24 and increases intercellular adhesion and thus decreases cell migration.²² Functional assays showed that NOV overexpression also reduces breast cancer cell proliferation.²² Concordantly, NOV was found downregulated in aggressive subtypes of breast cancer²⁵ and high CCN2/NOV ratio, often because of NOV downregulation, was associated with breast cancer metastasis to the bone.¹⁴ Therefore, NOV can have either tumorigenic or tumor-suppressive roles depending on the cellular context. The role of NOV, however, has not been carefully examined in prostate cancer.

Furthermore, only few studies have examined how NOV expression is regulated. Using a colon carcinoma cell line Bohlig et al.²⁶ showed that the p53 tumor-suppressor activates the NOV gene transcription. NOV was thought to contribute to the tumor-suppressor role of p53 in inhibiting cell proliferation and increasing cell adhesion. NOV has also been shown to be positively regulated by the gap junction protein Cx43 (refs 22,24) and by cytokines through STAT5A/B transcription factors.²⁷

In this study, we nominated NOV as one of the top target genes that are directly repressed by the AR. Using NOV as a model gene, we demonstrated that AR directly binds the enhancers of target genes and controls the promoter through DNA looping with enhancer. In addition, we showed that, while the enhancer elements are important in recruiting AR, the promoter dictates gene repression through repressive chromatin remodeling that is mediated in part by the histone H3 lysine 27 (H3K27) methyl-transferase EZH2. Furthermore, this study characterized the role of NOV in prostate cancer cells through overexpression and knockdown assays and found that NOV downregulation contributes to prostate cancer progression.

Results

NOV expression in prostate cancer cells is suppressed by androgen

Recent studies have suggested that AR directly inhibits the expression of AR itself and many other genes.^6,7 How this AR-mediated repression contributes to prostate cancer, however, is not clear. It has thus become important to characterize critical AR-repressed genes. To nominate such genes, we performed meta-analysis of three microarray data sets including two data sets that profiled LNCaP and VCaP cells treated with androgen over a time course using Agilent arrays² and another data set that analyzed androgen response in LNCaP cells using Illumina bead arrays.⁴ With a meta-z-score cutoff of 6.0, we identified 22 genes that are consistently repressed by androgen in all three studies (Figure 1a and Supplementary Table S1). To further narrow down the target genes and also to circumvent cross-platform variability, we restricted our analysis to the first two data sets that used the same microarray platform² and selected the top 25 probes by fold change from each study. Out of these 25 probes, we found in common 10 probes corresponding to 4 unique genes, which were also identified in the meta-analysis. These include UGT2B17, MET, CCDC83 and NOV. Out of these, UGT2B17 and MET have been previously shown inhibited by androgen,^9,28 while very little information is available for CCDC83 in the literature. We chose to study NOV as its role in tumorigenesis has been well characterized in a number of cancer types but has not been carefully examined in prostate cancer.

Figure 1.

NOV gene expression is inhibited by androgen in prostate cancer cells. (a) NOV is among the top genes that are consistently inhibited by androgen across multiple data sets. Meta-analyses were performed using three expression microarray data sets^2,4 profiling androgen response in LNCaP and VCaP cells. For the genes passed the threshold cutoff of meta-z-score, heatmap view was generated to reflect their expression across all data points. (b) Androgen inhibits NOV expression in LNCaP cells in a time-dependent manner. LNCaP cells were hormone starved for 2 days and treated with 1 nM synthetic androgen R1881 for up to 48 h. (c) Androgen represses NOV expression in a dose-dependent manner. LNCaP cells were hormone starved for 2 days and treated with varying concentrations of R1881 for 48 h. (d) VCaP cells were hormone starved for 2 days and treated with R1881 for up to 12 h. In (b–d), glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used a control. Error bars: n = 3, mean ± s.e.m. (e) Western blot of AR, PSA and NOV in LNCaP cells treated with varying concentration of R1881 for 48 h as described in c. Tubulin was used as loading control. (f) Western blot of AR, PSA and NOV in LNCaP cells treated with 1 nM R1881 for up to 48 h as described in b. Tubulin was used as loading control.

To confirm androgen inhibition of NOV, we performed quantitative reverse transcriptase–PCR analysis of LNCaP cells treated with 1 nM synthetic androgen R1881 for 0, 6, 12, 24 and 48 h. Our data showed that NOV was markedly inhibited by androgen for almost 20-fold in a time-dependent manner, whereas PSA, an androgen-induced gene, was dramatically induced (Figure 1b). Similar level of repression was also observed in a set of other five top androgen-repressed genes derived from aforementioned microarray analysis including AMIGO2, PKIB, LRRN1, CXCR7 and SERPINI1 (Supplementary Figure S1A). To demonstrate that this regulation is of physiological relevance, we stimulated LNCaP cells with 0, 0.01, 0.1, 1 and 10 nm (saturating amount) of androgen for 48 h and indeed observed a dose-dependent NOV inhibition. NOV transcript level reached the lowest with 1 nM of androgen where PSA level reached the highest (Figure 1c). The robustness of this regulation is confirmed through the analysis of several additional AR-expressing cell lines including VCaP (Figure 1d), C4-2B and 22RV1 cells (Supplementary Figures S1B and C). Furthermore, immunoblot analysis revealed a concordant decrease in NOV protein following androgen stimulation and PSA upregulation (Figures 1e and f). Having verified that NOV is robustly inhibited by androgen, we next examined whether this regulation is directly mediated by the AR.

AR directly suppresses NOV gene expression in prostate cancer cells

To test whether androgen inhibits NOV expression through the AR, we performed RNA interference of AR in LNCaP cells. AR knockdown was confirmed by quantitative reverse transcriptase– PCR and immunoblot analysis of AR itself and known AR-induced genes PSA and TMPRSS2 (Figures 2a and b). Importantly, although AR knockdown has little effect on NOV expression in the absence of androgen, it markedly restored NOV expression in androgen-treated LNCaP cells (Figure 2a). The remaining repression is likely due to incomplete AR knockdown leaving approximately 20% of AR. Concordantly, about 80–90% AR-induced gene expression was lost during AR knockdown, suggesting that AR-mediated repression of NOV is as robust as AR activation of well-known targets. Furthermore, immunoblot analysis confirmed recovered NOV protein expression following AR knockdown (Figure 2b). We also treated LNCaP cells with AR antagonist MDV3100 and observed that MDV3100 significantly upregulated NOV transcription for approximately 14-fold, whereas inhibited PSA for about 6-fold (Supplementary Figure S2A). On the other hand, overexpression of ectopic AR in LNCaP cells led to further suppression of NOV expression (Supplementary Figures S2B and C). Therefore, androgen inhibits NOV expression through the AR. As AR is able to directly bind to the chromatin to induce gene expression, we next investigated whether AR similarly binds to the regulatory elements of NOV for transcriptional repression. AR chromatin immunoprecipitation (ChIP)-Seq data indeed revealed one strong AR-bound enhancer at 63-kb upstream and two weaker enhancers at 71- and 13-kb upstream of the transcriptional start site (TSS) of the NOV gene (Figure 2c). Although all three enhancers may contribute to gene regulation, we chose to focus on the primary enhancer at 63-kb upstream as a representative in later analysis of the mechanisms. To validate that AR binds to this enhancer, we performed AR ChIP in LNCaP cells before and after androgen stimulation. As ChIP–PCR has detected AR binding at the promoters of AR-induced genes in addition to the primary binding events at the enhancers,² we analyzed both the NOV and PSA promoters and enhancers. As expected, following androgen stimulation AR binds the PSA enhancer strongly and, to a lesser degree, the PSA promoter. Similarly, we observed 16-fold enrichment of AR on the NOV enhancer on androgen stimulation, thus confirming direct AR occupancy (Figure 2d). In addition, our data also revealed significant (1.8-fold, P<0.05) AR enrichment on the NOV promoter. This direct AR protein occupancy on the NOV enhancer and promoter was further confirmed in the VCaP cells (Figure 2e).

Figure 2.

NOV is directly bound and suppressed by the AR. (a) AR knockdown restored androgen-suppressed NOV expression. LNCaP cells were hormone starved for 2 days, followed by transfection of control small interfering RNA (siRNA) or siRNA targeting AR in the presence or absence of androgen for 48 h. *Denotes P<0.05 between indicated samples. (b) Western blot of AR, PSA and NOV in LNCaP cells with AR knockdown by siRNA. Tubulin was used as loading control. (c) ChIP-Seq plot showing AR binding to the NOV promoter and enhancer. Previously described ChIP-Seq data were used for the analysis.² (d) AR binds the NOV promoter and enhancer in LNCaP cells. LNCaP cells were hormone starved for 3 days before treatment with ethanol or 10 nM R1881. Cells were harvested for AR ChIP after 16 h of treatment. ChIP-enriched DNA was analyzed by qPCR. Inset shows significant increase in AR recruitment to NOV promoter on R1881 treatment. (e) AR binds the NOV promoter and enhancer in VCaP cells. Error bars: n = 3, mean ± s.e.m.

To evaluate their transcriptional roles, we sub-cloned the NOV enhancer or promoter into a luciferase reporter and transfected them into LNCaP cells. Interestingly, we found that androgen significantly inhibited the activities of both the NOV promoter and enhancer luciferase reporters (Figures 3a and b). This is consistent with previous study showing that androgen increased both the promoter and enhancer luciferase activities of AR-induced TMPRSS2 gene.²⁹ Both studies showed that the changes in luciferase activities are in a much lesser extend than AR binding, suggesting the importance of in vivo environment for most efficient regulation. Moreover, we show ectopic AR overexpression in LNCaP cells further suppressed NOV promoter activity while inducing the activities of the PSA and TMPRSS2 promoters (Supplementary Figures S3A and B). We next examined whether the androgen response elements (AREs) were involved in recruiting AR to the repressed genes. Analysis of AR-binding sites at the NOV enhancer and promoter revealed several ARE motifs (Supplementary Figures S3C and D). To determine whether ARE is crucial for the transcriptional regulation, we generated ARE-mutant NOV promoter construct. Luciferase assay showed that the mutant construct no longer responded to androgen stimulation (Figure 3c). These data suggest that the promoters themselves are sufficient to dictate an either stimulatory or inhibitory response to androgen. However, the responses were only of two- to threefold, which is a magnitude less than the response in gene expression, suggesting the importance of the in vivo environment in amplifying the signal.

Figure 3.

Androgen inhibits NOV promoter activity and induces DNA looping between the NOV enhancer and promoter. (a) Androgen inhibits NOV promoter activity. The NOV promoter activity was measured in LNCaP cells in the presence/absence of androgen using a luciferase reporter. Error bars: n = 3, mean ± s.e.m. *Denotes P<0.05. (b) Androgen inhibits NOV enhancer activity. The NOV enhancer (− 63 kb) was cloned into a luciferase reporter. The enhancer activity was measured in LNCaP cells in the presence/absence of androgen. Error bars: n = 3, mean ± s.e.m. (c) Mutation of the core ARE motif within the NOV promoter abolishes androgen inhibition. Luciferase activities of wild type (WT) and ARE-mutant NOV promoters were determined in LNCaP cells in the presence or absence of androgen. (d) Schematic diagram showing the locations of chromatin conformation capture (3C) primers around the NOV regulatory elements. The black curve represents the looped DNA linked together by AR and cofactor complex. The spikes on the curve indicate the BglII sites (further detailed in Supplementary Table S2). The arrows represent the location (immediately before a BglII site) and the direction of the primers used for PCR analysis (further detailed in Supplementary Table S3). AP stands for the anchor primer that was designed at the promoter and used in all PCR reactions along with one of the enhancer primers (F1 to F5). The two black bell-shaped curves on the looped DNA represent the ChIP-Seq AR binding on the enhancers at −63 and − 17-kb upstream of NOV TSS, respectively. (e) Androgen induces interaction between the NOV enhancer and promoter in LNCaP cells. The 3C experiments were performed in LNCaP cells in the presence or absence of androgen (R1881). Equal amount of control- and R1881-treated chromatin (as confirmed by glyceraldehyde 3-phosphate dehydrogenase (GAPDH)) was digested with BglII followed by re-ligation that favors proximity rather than random ligation. The re-ligated DNA was then subjected to PCR using AP primer and one of the F1-5 primers. The PCR products were visualized using 2% agarose gel. To assure that the differential PCR amplification in 3C experiments was not due to primer failure, a BAC clone (RP11-840I14) that covers the entire NOV regulatory region shown in d was used as a positive control. The BAC DNA was digested with BglII followed by random re-ligation. PCR analysis of this randomly relegated BAC DNA confirmed that all primer pairs generated one specific PCR product.

Previous studies have reported promoter–enhancer DNA looping as a major mechanism for AR transcriptional activation in vivo.²⁹ To test if this holds true for AR-mediated repression, we examined DNA looping between the NOV enhancer and promoter. Chromatin conformation capture experiments were carried out in hormone-depleted LNCaP cells stimulated with vehicle or androgen. The chromatin was isolated and digested at BglII restriction enzyme sites (Supplementary Table S2). The BglII-digested DNA was subjected to proximity re-ligation reaction that used a very low concentration of DNA favoring intramolecular ligation rather than the ligation of random fragments. The re-ligated DNA was then purified and subjected to PCR analysis using an anchor primer at the promoter and another primer (F1 to F5) on the enhancer or control regions (Figure 3d). As shown in Figure 3e, a PCR product is detectable only between the promoter (anchor primer) and enhancer (F3) primers and is significantly augmented by androgen stimulation. To assure that the lack of PCR products between anchor primer and other primers was not due to the amplification efficiency, we tested the primers using a bacterial artificial chromosome (BAC) clone that covers the entire region (137-kb upstream to 47-kb downstream of the NOV TSS). The BAC DNA was digested with BglII followed by random re-ligation. PCR analysis of equal amount of DNA confirmed that all primer pairs generated single-PCR product (Figure 3e). Therefore, similar as in the case of AR-induced genes, chromosomal looping brings AR-bound enhancers to close proximity of the promoters of AR-repressed genes to modulate transcription. We next attempted to understand why these AR-binding events lead to NOV repression instead of activation as in the case of PSA.

Androgen induces repressive chromatin remodeling around the NOV promoter through the recruitment of EZH2

To depict the potentially distinct chromatin environment around the NOV and PSA genes, we performed ChIP–PCR analysis of representative active or repressive marks. Concordant with its inhibition, NOV promoter was occupied by greatly decreased amount of RNA polymerase II on AR activation, whereas polymerase II occupancy on the PSA gene was markedly increased (Figure 4a). Similarly, ChIP analysis showed that active histone marks such as H3K4me3 and H3K9 acetylation were dramatically reduced on the NOV promoter but increased on the PSA promoter (Figures 4b and c). In addition, H3K4me1 and me2 were markedly induced on PSA enhancer but not on NOV enhancer (Supplementary Figures S4A and B). By contrast, repressive H3K27me3 was remarkably increased on the NOV promoter (Figure 4d). Therefore, androgen stimulation and AR occupancy specifically induce repressive chromatin remodeling around the AR-repressed genes.

Figure 4.

Androgen induces repressive chromatin remodeling through EZH2. (a–e) ChIP–PCR analysis of polymerase II (PolII) (a), H3K4me3 (b), H3K9 acetylation (c), H3K27me3 (d) and EZH2 (e) on NOV and PSA promoters and enhancers. LNCaP cells were hormone starved for 3 days and then treated with ethanol or R1881 for 16 h. Cells were then harvested for ChIP using corresponding antibodies. ChIP enrichment on target regions were analyzed by qPCR. (f) AR knockdown and EZH2 inhibition synergistically restore NOV expression. LNCaP cells were transfected with control or AR-targeting small interfering RNA (siRNA) in combination with 5 μm SAHA, 5 μm 5-azacytidine (AZA) or both. Cells were harvested 48 h after treatment for quantitative reverse transcriptase (qRT)–PCR analysis. (g) EZH2 knockdown restores NOV expression. LNCaP cells were transfected with control siRNA or siRNA targeting EZH2 (siEZH2) and analyzed by qRT–PCR for gene expression. (h) AR and EZH2 synergistically regulate NOV protein. LNCaP cells were transfected with control siRNA, siAR, siEZH2 or both. Protein expression was analyzed by immunoblotting.

To further understand the molecular basis mediating the repressive chromatin remodeling on NOV promoter, we performed ChIP analysis of the polycomb group protein EZH2, a primary H3K27 methyltransferase. Androgen stimulation dramatically increased EZH2 binding on the NOV enhancer and promoter but not that of PSA (Figure 4e). This suggests that EZH2 is recruited by AR activation specifically to the repressed genes. To confirm the importance of EZH2 in this regulation, we treated LNCaP cells with histone deacetylase inhibitor suberoylanilide hydroxamic acid (SAHA) and methylation inhibitor 5-azacytidine that have been previously shown to block EZH2-mediated H3K27me3 and epigenetic silencing.³⁰ Our data revealed remarkable (P<0.01) NOV upregulation for 2.3-fold by 5-azacytidine alone (Supplementary Figure S4C), 8-fold by SAHA alone and for 25-fold by SAHA and 5-azacytidine together (Figure 4f). The degree of NOV upregulation by SAHA was much more than that of SLIT2, a previously characterized EZH2 target gene.³¹ In addition, AR knockdown in these cells further upregulated NOV expression, demonstrating a synergistic effect with EZH2 inhibition and supporting the involvement of both EZH2 and AR in suppressing NOV expression.

To directly show that EZH2 is required for NOV repression, we performed RNA interference of EZH2 in LNCaP cells. Consistently, following EZH2 knockdown NOV transcript was markedly (8.5-fold) increased, a level comparable to the induction of well-known EZH2 target gene CNR1 (Figure 4g). We next performed RNA interference of AR alone, EZH2 alone or both in LNCaP cells and assayed NOV expression. Immunoblot and quantitative reverse transcriptase–PCR analysis demonstrated that the loss of AR and EZH2 synergistically increased NOV expression at both the protein and transcript levels (Figure 4h and Supplementary Figure S4D). Therefore, AR and EZH2 are both required for inhibiting NOV gene expression. As they are both dys-regulated in prostate cancer, we next examined the pathological relevance of NOV expression and suppression during prostate tumorigenesis.

NOV is downregulated in poorly differentiated prostate cancer

To gain insights into the potential role of NOV in prostate cancer, we first examined NOV expression in a panel of benign and prostate cancer cell lines (Supplementary Figure S5). We found that NOV is highly expressed in benign cell lines such as PrEC and BPH1 and is dramatically downregulated in prostate cancer cell lines. Furthermore, NOV expression is the lowest in AR-positive prostate cancer cells supporting its negative regulation by AR. To confirm this in human specimens, we analyzed NOV expression in a microarray data set that profiled gene expression in a panel of benign prostate, clinically localized and metastatic prostate cancer tissues.³¹ We found that NOV mRNA was remarkably (P< 0.001) downregulated in metastatic tumors (Figure 5a). Furthermore, the level of NOV was negatively correlated with the level of androgen signaling represented by PSA expression (r = − 0.41; P<1e−7; Figure 5b). As EZH2 cooperates with AR in suppressing NOV in vitro, we examined its pathological relevance in vivo. Concordantly, NOV expression was also negatively (r = − 0.44; P<1e−7) correlated with that of EZH2 in human prostate cancer (Figure 5c).

Figure 5.

NOV is downregulated in poorly differentiated prostate cancer. (a) The expression of NOV decreases in metastatic prostate cancer. NOV transcript level was determined in an expression microarray data set³⁸ profiling a cohort of human prostate tissues including normal/benign, localized and metastatic prostate tumors. (b) Negative correlation between the expression levels of NOV and AR signaling, exemplified by KLK3 expression, in human prostate cancer. (c) Negative correlation between the expression levels of NOV and EZH2 in human prostate cancer. (d) NOV is downregulated in poorly differentiated prostate cancer. Quantitative reverse transcriptase (qRT)–PCR analysis was carried out to determine NOV expression in a panel of low Gleason (Gleason 6 and 7) and high Gleason (Gleason 8–10) human prostate tumors. (e) Immunoblot analysis of NOV, AR and EZH2 proteins in a set of benign adjacent prostate tissues, localized and metastatic human prostate cancers. Tubulin was used as loading control.

To confirm that NOV is downregulated in aggressive prostate cancer, we performed quantitative reverse transcriptase–PCR analysis of NOV transcript in a set of prostate tumors including 17 moderately differentiated Gleason 6 and 14, poorly differentiated Gleason 8–10 prostate tumors (Figure 5d). Our data showed that NOV expression was significantly (P =0.0168) suppressed in poorly differentiated prostate cancer. Furthermore, immunoblot analysis demonstrated that NOV protein is remarkably downregulated in advanced prostate cancer relative to localized prostate cancer and benign prostate tissues (Figure 5e). In addition, the loss of NOV is associated with increased expression of AR and/or EZH2. Therefore, in prostate cancer NOV may function as a tumor-suppressor gene, the loss of which may contribute to tumorigenesis. To investigate this, we next carried out functional assays to examine NOV's role in prostate cancer cells.

NOV inhibits prostate cancer cell growth

We next attempted to examine the function of NOV in the context of AR signaling in prostate cancer. We first ectopically reexpressed NOV in LNCaP cells that have very low endogenous NOV (Figure 6a). Cell growth assay showed that NOV reconstitution drastically inhibited LNCaP cell growth both in the presence and absence of androgen, supporting a growth inhibitory role of NOV (Figures 6a and b). This effect of NOV in inhibiting cell growth was further confirmed in two additional cell lines: the androgen-independent AR-positive 22RV1 cells and AR-negative DU145 cells (Supplementary Figures S6A and B). To test NOV function in vivo, control cells and LNCaP cells with stable NOV overexpression were inoculated subcutaneously into nude mice. Tumors developed in four out of the five control mice at 5 weeks after injection of LNCaP cells (Figure 6c). The tumors continued to grow until we had to terminate the experiment because of the large tumor volume. Remarkably, none of the mice injected with NOV-expressing LNCaP cells formed any tumors during the whole experimental period. Therefore, our results provide compelling evidence that in prostate cancer NOV inhibits tumorigenic properties both in vitro and in vivo.

Figure 6.

NOV inhibits prostate cancer growth in vitro and in vivo. (a) NOV overexpression inhibits prostate cancer cell proliferation. LNCaP cells that stably overexpress NOV was generated and confirmed by immunoblot (inset figure). Equal number of control and NOV-overexpressing LNCaP cells were plated for WST-1 cell growth assay. (b) NOV reconstitution inhibits androgen-independent prostate cancer cell growth. Equal number of control and NOV-overexpressing LNCaP cells were plated in phenol-free RPMI supplemented with charcoal stripped fetal bovine serum, and then subjected for WST-1 cell growth assay. (c) NOV overexpression inhibits xenograft LNCaP prostate cancer formation. LNCaP cells stably overexpressing NOV or control were injected subcutaneously into the dorsal flank of nude mice. Tumors size was evaluated weekly. (d) NOV knockdown promotes androgen-independent prostate cancer cell growth. Equal amount of LNCaP cells with NOV short hairpin RNA (shRNA) knockdown and corresponding control cells were plated in hormone-starved medium for WST-1 cell proliferation assay. (e, f) PSA is decreased but NOV expression is restored by castration in mice. Microarray data profiling prostate gene expression in mice castrated for 0, 3, 14 days with or without 3 days of testosterone re-implantation. NOV and PSA expression in the data set is re-analyzed. (g, h) PSA is decreased but NOV restored by hormone-deprivation therapy. Microarray data profiling localized prostate cancer in patients with or without hormone-deprivation therapy are re-analyzed.

In order to further determine whether NOV downregulation mediates androgen-stimulated cell growth, we performed NOV knockdown using short hairpin RNA. Importantly, NOV knockdown in LNCaP cells significantly induced androgen-independent cell growth (Figure 6d). This is consistent with the data that NOV reconstitution blocks LNCaP cell proliferation (Figure 6a). Therefore, NOV inhibition has important roles in mediating androgen-induced cell growth and its blockage is sufficient to drive androgen-independent prostate cancer cell growth.

We thus asked whether androgen-deprivation therapies restore NOV expression, which may subsequently contribute to tumor regression. To address this we first examined NOV expression in castrated mice.³² We found that NOV expression is indeed significantly restored following surgical castration whereas PSA expression was eliminated (Figures 6e and f). Moreover, testosterone add-back once again inhibited NOV but induced PSA. Interestingly, although EZH2 level was not significantly regulated by initial castration, it became remarkably upregulated following testosterone add-back (Supplementary Figure S6C). This is consistent with previous reports of EZH2 in metastatic prostate cancer, which has often undergone castration and has relapsed. This high level of EZH2 may also explain the observed NOV repression during testosterone add-back to a level even lower than the control mice. To assure that this holds true in prostate cancer patients we analyzed NOV and PSA expression in a microarray data set that compared gene expression in a cohort of tissues from localized prostate cancer patients with or without hormone-deprivation therapy.³³ Indeed, we found that hormone-deprivation therapy significantly decreased PSA level but increased NOV expression (Figures 6g and h). By contrast, EZH2 is only marginally significantly increased (Supplementary Figure S6D). Taken together, NOV has important roles in androgenmediated tumorigenesis and NOV restoration may be useful in treating prostate cancer.

Discussion

Several studies have recently begun to show AR as a transcriptional repressor, in addition to its well-documented role in gene activation.^6,7 How AR directly suppresses gene expression, however, are not fully understood. In addition, AR-repressed genes that are essential in conveying the role of AR in prostate cancer progression and castration resistance are yet to be identified. In this study, through integrative genomic analysis we nominate NOV as a top target gene that is directly suppressed by the AR. Using NOV as a model gene, we further analyzed the mechanisms underlying AR-mediated transcriptional repression. We found that AR inhibits gene expression by binding to the enhancer and forming DNA looping with the target promoter, a mechanism that was also utilized in AR-mediated gene activation. The promoters and enhancers alone are able to dictate either a stimulatory or inhibitory response to androgen. However, these responses are of only two- to threefold. The in vivo chromatin environment such as DNA looping is essential to amplify the signal, eventually leading to magnitudes higher response in gene expression, such as NOV inhibition or PSA activation.

AR mediates transcriptional repression by specifically recruiting EZH2 to the repressed genes, but not to the induced genes. EZH2 occupancy on the repressed genes further leads to strong H3K27 trimethylation around the promoter, inducing repressive chromatin remodeling and epigenetic silencing. Concordantly, active marks such as polymerase II, H3K4 methylation and H3K9 acetylation are dramatically decreased around NOV promoter following AR activation. By contrast, they are substantially induced around the PSA promoter. It will be of great interest to determine in future studies the molecular mechanisms that recruit EZH2 specifically to a subset of AR target genes.

AR-mediated transcriptional activities are essential for prostate cancer growth and progression. In castration-resistant prostate cancer, AR remains required and sufficient for the tumor growth.¹ In fact, AR is often hyperactive in advanced prostate cancer in the milieu of very low androgen because of mechanisms such as AR amplification, overexpression and hypersensitivity. A comprehensive understanding of the different pathways that AR could contribute to cell growth is exceedingly important in guiding novel therapeutic strategies to target castration-resistant prostate cancer and in measuring treatment responses. Our study suggests that one important but previously unknown mechanism, by which AR mediates prostate cancer progression, is via inhibition of tumor-suppressor genes, the loss of which could contribute to prostate cancer. In addition, they may be further suppressed by secondly hits such as EZH2 upregulation, which is highly prevalent in metastatic prostate cancer.³⁴ AR-repressed genes are downregulated in aggressive prostate cancer wherein AR and EZH2 are upregulated. The restored expression of these genes is at least in part accountable for the efficacy of hormone-deprivation therapies. AR-repressed gene expression thus may be useful biomarkers to assay treatment response. Reconstitution of these genes and their pathways may be novel avenues for treating prostate cancer. In addition to AR and EZH2 cooperation on repressed genes, histone deacetylases have also been shown required for AR-induced gene expression.³⁵ This suggests that it may be beneficial to combinatorial target AR and EZH2 pathways to treat late-stage prostate cancer.

Since its discovery two decades ago,^12,36 NOV has been investigated in a number of cancer types.¹⁷ NOV has been shown to regulate oncogenic properties in a cell type- and context-dependent manner.¹⁷ In breast cancer that shares many homologous oncogenic mechanisms with prostate cancer, NOV has been shown to inhibit cell growth and tumor metastasis.¹⁴ Our study is the first to characterize NOV functions through NOV overexpression and knockdown assays in prostate cancer. We demonstrate that NOV inhibits prostate cancer growth in vitro and in vivo. This growth inhibitory role of NOV is consistent with its inhibition by AR and EZH2 oncogenes that are often upregulated in advanced prostate cancer. NOV downregulation mediates androgen-induced prostate cancer cell growth and NOV reconstitution reverses it. NOV knockdown by RNA interference drives androgen-independent cell growth. Therefore, NOV has a critical role in AR-mediated prostate cancer cell growth.

In conclusion, our study used meta-analysis to nominate NOV as a top target of AR-repressed genes. Using NOV as a model gene, we illustrate the mechanisms by which AR functions as a transcriptional repressor. We further show that NOV inhibits prostate cancer progression and has important roles in mediating AR-induced tumorigenesis. It may be clinically important in future studies to explore novel therapeutic strategies to combinatory target AR and EZH2 pathways. AR-repressed genes such as NOV may be useful biomarkers for treatment response. Reconstitution of AR-repressed genes and their pathways offers innovative avenues to treat advanced prostate cancer.

Materials and Methods

Human tissue specimens

Prostate cancer tissues were collected by the Northwestern University Prostate Cancer Specialized Program of Research Excellence Tissue Core (SPORE). Tissue samples were collected with informed consent of the patients and prior Institutional Review Board approval.

Cell lines and treatments

The prostate cancer cell lines LNCaP, VCaP and DU145 were obtained from ATCC (Manassas, VA, USA). For RNA interference assays, non-targeting small interfering RNA (D-00110-01, Dharmacon, Lafayette, CO, USA), small interfering RNA specific to EZH2 (P-002079-01, Dharmacon) or AR (L-003400, Dharmacon), and short hairpin RNA targeting NOV (V2LHS_152302, Open Biosystems, Huntsville, AL, USA) were used. All transfections were performed using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). SAHA (5 μm) and 5-azacytidine (5 μm) treatments were carried out using dimethylsulphoxide as control.

Western blots and ChIP

Protein extraction and western blot analysis were performed as previously described.² The following antibodies were used: anti-glyceraldehyde 3-phosphate dehydrogenase (ab9385, Abcam, Cambridge, MA, USA), anti-AR (#06–680, Millipore, Billerica, MA, USA), anti-PSA (K2889, Sigma-Aldrich, St Louis, MO, USA), anti-EZH2 (612667, BD Biosciences, San Jose, CA, USA) and anti-NOV (K19M). ChIP were carried out as previously described.^30,31 ChIP antibodies used include AR (#06–680, Millipore), H3K4me1 (#07–436, Millipore), H3K4me2 (#07–030, Millipore), H3K4me3 (ab8580, Abcam), H3K27me3 (#07–449, Millipore), EZH2 (#49–1043, Invitrogen), acetyl-H3K9 (ab10812, Abcam) and RNA polymerase II (ab817, Abcam). All primers (Supplementary Table S3) were designed using primer 3 and synthesized by Integrated DNA Technologies (Coralville, IA, USA).

Functional assays

Cell growth and luciferase assays were performed as previously described.^2,30 The NOV promoter with ARE mutation was generated using QuikChange II Site-Directed Mutagenesis Kits (Agilent Technologies, Santa Clara, CA, USA). The canonical ARE (5′-GTCTGCTACAGCAGG TGCTTCA-3′) that corresponding to the nucleotide from − 729 to − 708 of NOV promoter was mutated to 5′-GTCTGCTTCTGCAGCTCCTTCA-3′. Approximately, a 1-kb fragment around the NOV enhancer at − 63-kb upstream was also cloned into a pGL3 luciferase reporter vector.

Chromosome conformation capture assay

We analyzed the digestion map of commonly used restriction enzymes from − 100 to +22-kb enhancer/promoter region of NOV locus. We used BglII digestion, as BglII sites show a distribution that will enable appropriate primers to be designed to generate ∼250-bp PCR products on re-ligation. All primers are designed based on the forward strand immediately upstream of a BglII restriction site. The BglII digestion sites, primers and their distance to the NOV TSS are shown in Supplementary Tables S2 and 3. The chromatin conformation capture experiments were conducted in LNCaP cells according to a standard chromatin conformation capture protocol.³⁷ Briefly, fixed chromatin of LNCaP cells (5×10⁶) was digested with BglII overnight. To favor the intramolecular ligation, the reaction was done with 800 units of T4 DNA ligase (New England Biolabs, Ipswich, MA, USA) for 4 h in a volume of 7.2 ml to keep the DNA concentration at 2∼3 ng/μl. For positive control experiment to test the primers, 10 μg of BAC clone (RP11-840I14) was digested and re-ligated at 100 ng/μl for 4 h to encourage random re-ligation.

Bioinformatics-analysis of expression microarray and ChIP-Seq data

ChIP-Seq and expression microarray data were downloaded from GEO database with accession numbers GSE14092 (ref. 2), GSE14097 (ref. 2), GSE18684 (ref. 4), GSE5901 (ref. 32), GSE2443 (ref. 33) and GSE35988 (ref. 38). Expression data were log2-transformed. For time course data sets, a moving average with n = 2 was calculated over a series of time points for each gene. We defined response of a gene as the maximum of moving averages across time points. As responses of genes are observed approximately normal distributed, a z-score transformation was then performed to generate final response score for each gene in each data set. Similarly, a z-score transformation was applied to gene responses in VCaP treated with R1881 relative to ethanol. In order to identify commonly repressed AR targets across studies, a meta-analysis was performed. Briefly, a meta z-score was calculated based on un-weighted Stouffer's method ( $Z=\frac{\sum _{i=1}^k{Z}_i}{\sqrt{k}}$, k is the number of studies), which follows a standard normal distribution under the null hypothesis. Top androgen-repressed genes were selected using a meta-z-score threshold of 6 (a Bonferroni corrected P<4e−05).

Murine prostate tumor xenograft model

All procedures involving mice were approved by the Institutional Animal Care and Use Committee at Northwestern University and conform to their relevant regulatory standards. Five-week-old male nude athymic BALB/c nu/nu mice (Charles River Laboratory, Wilmington, MA, USA) were used for evaluating the role of NOV in tumor formation, the LNCaP cells stably overexpressing NOV or the control cells were inoculated by subcutaneous injection into the dorsal flank of ten mice (n = 5 per group). Tumor size was measured every week, and tumor volumes were estimated using the formula (π/6) (L × W²), where L = the length of tumor and W=the width.