Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Identification of novel androgen response genes in prostate cancer cells by coupling chromatin immunoprecipitation and genomic microarray analysis

Abstract

The androgen receptor (AR) plays a key role as a transcriptional factor in prostate development and carcinogenesis. Identification of androgen-regulated genes is essential to elucidate the AR pathophysiology in prostate cancer. Here, we identified androgen target genes that are directly regulated by AR in LNCaP cells, by combining chromatin immunoprecipitation (ChIP) with tiling microarrays (ChIP-chip). ChIP-enriched or control DNAs from the cells treated with R1881 were hybridized with the ENCODE array, in which a set of regions representing approximately 1% of the whole genome. We chose 10 bona fide AR-binding sites (ARBSs) (P<1e-5) and validated their significant AR recruitment ligand dependently. Eight upregulated genes by R1881 were identified in the vicinity of the ARBSs. Among the upregulated genes, we focused on UGT1A and CDH2 as AR target genes, because the ARBSs close to these genes (in UGT1A distal promoter and CDH2 intron 1) were most significantly associated with acetylated histone H3/H4, RNA polymerase II and p160 family co-activators. Luciferase reporter constructs including those two ARBSs exhibited ligand-dependent transcriptional regulator/enhancer activities. The present study would be powerful to extend our knowledge of the diversity of androgen genetic network and steroid action in prostate cancer cells.

Introduction

Androgen is a key regulator of male sexual differentiation as well as prostate development and carcinogenesis. Androgen-regulated gene expression is mediated by the action of androgen receptor (AR), which is a member of nuclear receptor superfamily that functions as a ligand-dependent transcription factor. Prostate cancer is originally an androgen-dependent tumor, whose growth and survival are under the control of AR signaling. Thus, androgen deprivation is the most common option of the cancer treatment. The therapy, however, eventually fails and most patients will relapse owing to adaptive progression of the surviving prostate cancer cells. The recurrent cancer is usually referred to as ‘androgen independent’ (Grossmann et al., 2001). Nevertheless, advanced prostate cancer often continues to express AR and androgen-regulated genes, suggesting a functional role of AR in the recurrent stage. Alterations of the AR gene including mutation and amplification are also shown in some recurrent tumors, but these mechanisms will not explain the hormone-refractory responses in the majority of patients after androgen deprivation. Indeed, a modest increase in AR mRNA has been shown to be associated with the resistance to anti-androgen therapy in isogenic prostate cancer xenograft models (Chen et al., 2004). Therefore, understanding the global aspects of AR signaling network and the distinct roles of AR target genes are essential for the development of new diagnostic procedures and therapeutic options for prostate cancer in various disease states.

AR regulates the expression of target genes by binding to androgen response elements (AREs) in the genome, or by interacting with other transcription factors bound to their specific recognition sites. AR-mediated gene transcription has been studied using prostate-specific antigen (PSA) as a prototypic model, and AREs in PSA promoter and enhancer have been shown to recruit various co-activators and general transcription factors including histone acetyltransferases, p160 family, mediator and RNA polymerase II (PolII) (Wang et al., 2005). Efforts have been paid to search various androgen target genes by using microarray techniques since last decade, identifying hundreds of genes with altered expression by hormone stimulation in cells. The gene expression profiling is powerful to depict the global function of androgen in a specified model; however, the technique will not be suitable to determine whether the alteration of gene expression is owing to direct or indirect action of AR transcription. Recent advance of human genome project enables to search putative AREs bioinformatically in the transcription regulatory regions of androgen target genes; yet, few AREs are identified as physiological elements in AR signaling (Horie-Inoue et al., 2004, 2006). Thus, the development of a new high-throughput method that identifies bona fide AR-binding sites (ARBSs) in the genome is a prerequisite for the elucidation of AR gene network.

Recently, a combined technique of chromatin immunoprecipitation (ChIP) analysis with DNA microarray has been established to identify chromatin-interacting domains of transcription factors in a genome-wide manner (Cawley et al., 2004; Bernstein et al., 2005). Regarding nuclear receptors, ligand-dependent estrogen receptor (ER)-binding sites have been recently shown by this ChIP-chip technique using Affymetrix-tiling oligonucleotide microarrays of chromosomes 21 and 22 (Carroll et al., 2005), or a custom-made promoter microarrays (Laganière et al., 2005). In this study, we have performed ChIP-chip using a sampler DNA microarray of the human genome, the so-called ENCODE chip. In this microarray, a set of regions representing approximately 1% (30 Mb) of the whole genome are included as the target for the pilot project that has been selected by the research consortium of the ENCyclopedia of DNA Elements (ENCODE Project Consortium, 2004). Fifty percent of the 30-Mb genomic regions, consisting of 14 regions (ENm001–ENm014), were manually selected, and the remaining 50% were composed of 30, 500 kb regions (ENr111–ENr334) selected according to a stratified random-sampling strategy based on gene density and level of non-exonic conservation.

Here, we find a discrete number of ARBSs in the selected regions of the ENCODE regions. Intriguingly, most of the AR-interacting regions have been shown to locate in non-promoter proximal regions; yet, they contained ARE sequences and were validated to recruit AR ligand dependently. In the vicinity of the functional ARBSs, we found several genes with upregulated transcript levels by hormone stimulation. Some of the AR target genes that have been identified in this study are previously known to be associated with AR expression, whereas some are novel targets. Our ChIP-chip analysis and transcriptional study indicate that non-promoter ARBSs play roles in the AR-dependent transcriptional regulation, potentially dissecting a series of AR-regulated mechanisms in a genome-wide manner.

Results

Screen of ARBSs on ENCODE DNA microarray

To perform a screen of ARBSs in AR-positive cells on tiling oligonucleotide microarrays, we first investigated the time course of ligand-dependent AR recruitment in human prostate cancer LNCaP cells. After 3-day hormone depletion, cells were stimulated with vehicle or a synthetic androgen R1881 (10 nM) for 2, 6 or 24 h. Cross-linked protein–DNA complexes extracted from the cells were immunoprecipitated with anti-AR antibody, and quantitative polymerase chain reaction (qPCR) for ARE regions in the proximal promoter and enhancer of PSA was performed using the purified precipitated DNAs as templates. AR binding in response to ligand stimulation exhibited maximal levels at 24 h (data not shown).

We next performed ChIP-chip analyses using the ENCODE tiling microarrays comprised of the total 30-Mb human genomic DNA, which corresponds to 1% of the genome. The chromatin DNAs immunoprecipitated by anti-AR or without ChIP (input control) were amplified unbiasedly by in vitro transcription (IVT), and the amplified DNAs were fragmented and biotin-labeled, then hybridized with the ENCODE chips for duplication. Using the Affymetrix Tiling Analysis Software, raw intensity data of duplicate arrays for each experimental group were transformed and signal and P-values for each genomic position interrogated were determined after quantile normalization. The results were mapped to genomic positions that could be visualized in the Affymetrix Integrated Genome Browser or the UCSC Genome Browser (NCBI Build 35). Applying a P-value cutoff of 1e-5 for a significant AR binding, we identified 10 ARBSs (Table 1) in the ENCODE genomic regions. Among them, five ARBSs were involved in the manually defined regions of the 30-Mb ENCODE regions (ARBSs no. 3–no. 7), whereas the remaining five binding sites were derived from the randomly selected regions.

Table 1 ARBSs identified in the encyclopedia of DNA element (ENCODE) regions by ChIP-chip experiments

Notably, most of the ARBSs were located within intronic regions or gene upstream regions at least 10 kb apart from the transcriptional start sites (TSSs) of their closest genes. One of the ARBSs included in the ENCODE chips was ARBS no. 1, which was located adjacent to UGT1A locus, in the 5′ upstream region >17 kb upstream of UGT1A1 gene TSS or in intron 1 of UGT1A3 on chromosome 2q37. As another example, ARBS no. 10 was situated in intron 1 of CDH2 on chromosome 18q11.2 (Figure 1).

Figure 1
figure1

Identification of in vivo ARBSs in LNCaP cells on the encyclopedia of DNA elements (ENCODE) array by ChIP-chip analysis. (a) An expanded view of the UGT1A locus on the ENCODE region ENr131 from chromosome 2q37 is shown in its genuine 5′−3′ orientation. ARBS no. 1 is located on the 5′ upstream region of UGT1A1, or on intron 1 of other UGT1A isoforms. (b) An expanded view of the CDH2 on the ENCODE region ENr213 from chromosome 18q11.2 is shown in its genuine 3′–5′ orientation. ARBS no. 10 is located on intron 1 of CDH2.

We next investigated whether the 10 ARBSs included sequences highly similar to the previously established consensus AREs. Using a weighted matrix-based finder TRANSFAC (Matys et al., 2003) with the matrix conservation >75% or a sequence analysis utility of JASPER with the relative profile score threshold >70% (Sandelin et al., 2004), we identified canonical ARE sequences in all of the ARBSs (Table 2).

Table 2 ARE sequences identified in the ARBSs detected by ENCODE chip

To verify whether the identified ARBSs in ChIP-chip were authentic ARBSs in the genome, we performed new independent ChIP experiments in LNCaP cells. We confirmed that >10-fold enrichment of R1881-dependent AR binding was shown in the all regions involved in the defined 10 ARBSs, targeting the identified ARE sequences (Figure 2). Thus, with the cutoff value 1e-5, we validated that our ChIP-chip results did not include false positives.

Figure 2
figure2

Validation of androgen-dependent AR enrichment by quantitative ChIP analysis on the identified ChIP-chip ARBSs in LNCaP cells. Hormone-deprivated cells were stimulated with R1881 (10 nM) or vehicle (0.1% ethanol) for 24 h. Cross-linked samples were immunoprecipitated with anti-AR antibody. The precipitated DNA fragments were subjected to qPCR. PCR primer sets were designed to include ARE sequences on individual ARBSs no. 1–no. 10. PCR products including ARBSs no. 8 and no. 9 were not distinguishable, as ARBSs no. 8 and no. 9 are located in genome duplication regions from the same origin. PSA promoter (PSA Pro) and enhancer (PSA Enh) regions including ARE sequences were used as positive controls. Data are fold enrichment compared with individual input non-enriched DNA (mean±s.d., n=2).

Identification of androgen target genes adjacent to AR-binding sites

To identify novel androgen target genes by using ChIP-chip data, we examined the alteration of gene expression closest to the ARBSs in LNCaP cells in response to R1881 (Figure 3). Eight of 10 genes adjacent to the ARBSs exhibited a ligand-dependent increase in expression levels by >2-fold compared with a vehicle-treated control by 48 h after treatment. Among them, UGT1A1 and Pepsinogen C (PGC) levels elevated by >30- and >1000-fold, respectively, with 48-h R1881 treatment (Figure 3a and b). Residual two of 10 genes adjacent to the ARBSs, SCAP2 and MET, exhibited a ligand-dependent decrease in expression levels (Figure 3c).

Figure 3
figure3

Androgen-dependent changes in expression of genes adjacent to the ChIP-chip ARBSs in LNCaP cells. Hormone-deprivated cells were stimulated with R1881 (10 nM) or vehicle for 12, 24 and 48 h. Quantitative RT–PCR analysis regarding the expression of 10 proximal genes close to ARBSs no. 1–no. 10 was performed using the reverse-transcribed cDNAs from the cells. Data are fold change compared with vehicle-treated cells at individual time point (mean±s.d., n=2). (a) UGT1A1 mRNA levels, (b) PGC mRNA levels and (c) mRNA levels of gene adjacent to ARBSs no. 3–no. 10.

UGT1A gene locus encodes nine distinct isoforms with unique exon 1 based on the difference of TSSs (Gong et al., 2001). Individual first exons are determinants for the structure of N-terminal UGT1A isoforms, which are important for substrate specificity. We examined whether androgen regulated the transcription of distinct UGT1A isoforms in LNCaP cells. Interestingly, only the mRNA levels of UGT1A1 (Figure 3a) and UGT1A3 isoforms, which have TSSs most adjacent to ARBS no. 1, were significantly increased up to 48 h after R1881 (10 nM) stimulation (UGT1A3 mRNA levels: 1.2±0.1 fold at 12 h, 5.8±0.2 fold at 24 h and 11.2±0.2 fold at 48 h after treatment). On the contrary, the mRNA levels of other UGT1A isoforms were not basically altered or rather decreased during the time course (data not shown).

Taken together, our data suggest that AR could regulate transcription of genes adjacent to ARBSs.

ARBSs are associated with histone acetylation and facilitate recruitment of RNA PolII

We next examined whether these ARBS regions recruited components indicative of transcriptional activation. ChIP analyses for acetylated histone H3/H4 (AcH3/H4) and RNA PolII were performed on the 10 ARBS regions in LNCaP cells (Figure 4). The AREs in PSA promoter and enhancer were used as positive controls. Histone acetylation was remarkable by R1881 treatment in ARBSs no. 1 and no. 10, in the vicinity of UGT1A1 and CDH2, respectively. Ligand-dependent RNA PolII recruitment was also significant in ARBS no. 1. In ARBS no. 10, PolII binding was enriched at basal levels and further enhancement of PolII binding was not observed by ligand stimulation (Figure 4c). Moderate histone acetylation and PolII recruitment in response to R1881 were also observed in ARBS no. 4, which was located in intron 3 of STEAP2. In the rest of seven ARBSs, all associated with ligand-dependent AcH3/H4 binding by >2-fold and three of seven recruited PolII ligand dependently by >2-fold. Although most of the identified ARBSs were located rather distal from known genes, our data suggest that a significant number of the ARBSs in the genome physically function as distal transcriptional regulatory domains during transcription of the adjacent genes.

Figure 4
figure4

Some of the ChIP-chip ARBSs associate with histone acetylation and RNA PolII recruitment. Hormone-deprivated LNCaP cells were stimulated with R1881 (10 nM) or vehicle for 24 h. Cross-linked samples were immunoprecipitated with anti-acetylated H3/H4 (AcH3/AcH4) or anti-RNA PolII antibodies. The precipitated DNA fragments were subjected to qPCR. Identical primer sets were used as described in Figure 2. Data are fold enrichment compared to individual input non-enriched DNA (mean±s.d., n=2).

Functional recruitment of p160 co-activators at ARBSs

The p160 SRC family co-activators play scaffold roles in forming co-activator complex involved in nuclear receptor-mediated transcription (Shang and Brown, 2002). In AR-mediated transcription, the p160 co-activators are shown to be recruited to AR complex and to facilitate AR transactivation by their histone acetylase activity (Shang et al., 2002). To delineate the functional roles of endogenous p160 co-activators in AR-mediated transcription from the identified ARBSs no. 1–no. 10, we performed ChIP analysis using antibodies against SRC1, GRIP1 and AIB1 (Figure 5). It is notable that all of the p160 co-activators were recruited by >10-fold in the ARE from PSA promoter. Consistent with the result of histone acetylation and RNA PolII recruitment, all of the p160 co-activators were recruited by >10-fold upon R1881 stimulation compared with vehicle in ARBS no. 1, adjacent to UGT1A1. The second potent binding site for AcH3/H4 and RNA PolII, ARBS no. 10 adjacent to CDH2, recruited SRC1 by >2-fold and GRIP1 and AIB1 by >20-fold upon ligand stimulation. Among other ARBSs, ARBS no. 5 close to PFTK1 recruited SRC1 by >2-fold in response to ligand stimulation. The data show that some of the authentic ARBSs may play roles as enhancers that recruit various transcriptional regulators and co-activators.

Figure 5
figure5

ChIP-chip ARBSs and p160 coactivator recruitment. Hormone-deprivated LNCaP cells were stimulated with R1881 (10 nM) or vehicle for 24 h. Cross-linked samples were immunoprecipitated with anti-SRC1, anti-GRIP1 or anti-AIB1 antibodies. The precipitated DNA fragments were subjected to qPCR. Identical primer sets were used as described in Figure 2. Data are fold enrichment compared to individual input non-enriched DNA (mean±s.d., n=2).

Distal and intronic ARBSs function as transcriptional regulators in androgen-dependent transcription

To further assess the possibility that the distal or intronic ARBSs function as bona fide transcriptional regulators in androgen-dependent transcription, we performed promoter activity assay using luciferase reporter constructs including ARE sequences derived from the ARBSs. Using the genomic DNA of LNCaP cells as a template, we amplified fragments included ARE sequences in ARBSs no. 1 and no. 10, corresponding to the 5′ upstream region of UGT1A1 (−17 kb) and intron 1 of CDH2 (Figure 6a and b). Note that ARBS no. 1 is also located in intron 1 of other UGT1A isoforms. The amplified fragments were ligated to a luciferase reporter plasmid pGL3-vector containing SV40 promoter. Regarding the 5′ upstream region of UGT1A1, we also generated a mutated construct including two substitutions at the positions −2C and +2G from the 3-bp spacer (UGT1A1 5′ Mut-Luc). Using LNCaP cells transfected with reporter constructs, the luciferase activities of UGT1A1 5′-Luc and CDH2 Int 1-Luc were increased 5- and 8-fold by R1881 treatment, respectively, whereas MMTV luciferase construct exhibited >100-fold activation in response to ligand stimulation (Figure 6c). UGT1A1 5′ Mut-Luc did not exhibit androgen-dependent transcriptional activation. These results suggest that a significant number of non-promoter ARBSs also play essential roles in AR-mediated gene transcription.

Figure 6
figure6

Transcriptional activity of UGT1A1 5′ upstream and CDH2 intronic ARBSs. (a) Construction of luciferase reporter plasmids containing UGT1A1 5′ upstream regions. A 519-bp fragment of the 5′ upstream of UGT1A1 (−17 753/−17 235 bp) amplified from LNCaP cells, and the 5′ fragment of UGT1A1 with mutation at the positions −2C and +2G from the 3-bp spacer in ARE sequences (ARBS no. 1) were cloned into pGL3 vector containing SV40 promoter (SV40 Pro), designated as UGT1A1 5′-Luc and UGT1A1 5′ Mut-Luc, respectively. (b) Construction of luciferase reporter plasmid containing CDH2 intron 1. A genomic fragment of CDH2 intron 1 (ARBS no. 10) derived from LNCaP cells (+20 197/+21 260 bp), containing 15 repeats of palindromic ARE sequences with the half-site GGTACA, was cloned into pGL3 vector (CDH2 Int 1-Luc). Note that the number of ARE sequences in LNCaP cells was larger than that in the genome database. (c) Androgen-stimulated luciferase activities of ARE sequences involved in UGT1A1 5′ upstream and CDH2 intron 1. LNCaP cells were stimulated with R1881 (10 nM) or vehicle 12 h after transfection with indicated plasmids together with a Renilla luciferase reporter gene, and incubated for another 24 h. Firefly luciferase activity was normalized to Renilla luciferase activity for each data set. MMTV luciferase reporter gene containing ARE sequences was used as a positive control. Data represent the mean±s.d., n=3. (d) Androgen-stimulated expression of UGT1A protein in LNCaP cells. Cells were stimulated with R1881 (10 nM) or vehicle for indicated times and whole cell lysates were separated by 8% SDS–PAGE. The PVDF membrane blotted with proteins were probed with anti-UGT1A or anti-β-actin antibodies.

Moreover, UGT1A protein expression could be regulated by androgen (Figure 6d). LNCaP cells after 72-h hormone deprivation were stimulated with R1881 or vehicle and cell lysates were prepared after 24 or 48 h. Although the isoform specificity of UGT1A was not shown by the antibody that we used, overall amounts of UGT1A protein were increased in response to androgen.

Discussion

This study aimed to identify novel androgen target genes in prostate cancer LNCaP cells by performing ChIP-chip analysis, identifying in vivo ARBSs in the selected ENCODE genomic regions. This scanning successfully identified 10 bona fide in vivo ARBSs with a P<1e-5. Notably, all of the 10 ARBSs included ARE sequences as determined by the sequence analysis utilities based on TRANSFAC or JASPER transcription factor-binding profiles (Matys et al., 2003; Sandelin et al., 2004), and ChIP-PCR validation confirmed that those ARBSs had abilities to recruit AR ligand dependently. Our ChIP-chip approach is a powerful high-throughput method that can be applied to the whole genome-wide screen of ARBSs.

Efforts have been paid to identify transcription factor-binding motifs for years by searching a consensus-like sequence through in silico or in vitro studies in the vicinity of transcription factor targets. In the days after the completion of the Human Genome Project, genomic DNA microarray has been developed and identification of in vivo binding targets of nuclear proteins is enabled by ChIP-chip analysis. For instance, in the ChIP-chip study for ERα on chromosomes 21 and 22, most of the binding sites were found at significant distances including several >100 kb removed from TSSs (Carroll et al., 2005). It has been suggested that these distal ERα-binding sites play an important role in estrogen-mediated regulation, as they could be physically associated with promoter-proximal regions. Similarly, in the present study, nine of 10 ARBSs that we identified by ChIP-chip were situated in the distal 5′ regions or intronic regions of known genes. Among those distal ARBSs, there were several sites that significantly recruited AcH3/H4 and RNA PolII (Figure 4). ARBSs no. 1 and no. 10, which are located at >17 kb upstream of UGT1A1 and in intron 1 of CDH2, respectively, could also associate with the p160 co-activators in a ligand-dependent manner (Figure 5). Based on our findings and previous evidence, a number of ARBSs may be located in non-promoter regions of the genome, and often associated with histone acethylation and co-activator recruitment.

In the vicinity of ChIP-chip identified ARBSs, we found several genes upregulated or downregulated by androgen stimulation. PGC, whose location is close to ARBS no. 2, is an androgen-upregulated gene that has been reported previously as a prognostic factor in prostate cancer (Diaz et al., 2002). It is an aspartyl protease and known as a protein involved in the digestion of proteins in the stomach. We identified a novel ARE sequence in ARBS no. 2, at −323 bp upstream of the TSS of PGC, indicating that ChIP-chip is particularly a powerful method to find out a novel transcription factor target regardless of the expression level of the target gene. As the ligand-dependent RNA PolII recruitment was not significant at ARBS no. 2 at the time we investigated (Figure 4c), other PGC regulatory region might be more important in the PolII activation; yet, ARBS no. 2 might play a role in the transcriptional regulation of PGC as the histone acetylation was at least promoted by ligand stimulation (Figure 4b). STEAP2, which includes ARBS no. 4 in intron 3, has been originally cloned as a STEAP homolog gene that encodes a transmembrane protein expressed in prostate cancer (Porkka et al., 2002). Although androgen responsiveness of STEAP2 was not reported previously, our data showed that it was a novel androgen target gene with a genuine ARBS in intron 3, which was also associated with histone acetylation.

ENCODE region ENm001 corresponds to chromosome 7q31, which is known as a fragile site with frequent loss of heterozygosity in advanced prostate cancer (Kawana et al., 2002). Among several genes at 7q31, TES (testis-derived transcript) has been shown as a candidate tumor suppressor gene in prostate cancer (Chene et al., 2004). In the present studies, we showed that TES was an androgen-upregulated gene with a genuine ARBS (ARBS no. 6) in its 3′ downstream region (−58 kb from the 3′-end). TES protein contains three conserved cysteine-rich zinc-binding motifs called LIM domains, suggesting that TES may play a role in protein-protein interaction and focal adhesion (Coutts et al., 2003). Methylation of the CpG island at the 5′ end of TES is frequently occurred in ovarian cancer cells, and overexpression of TES in culture cells was shown to be growth-inhibitory (Tobias et al., 2001). In contrast, we showed that MET was an androgen-downregulated gene with a novel ARBS (ARBS no. 7) in intron 17. MET encodes a receptor-like tyrosine kinase, c-met proto-oncogene product, which can be activated by hepatocyte growth factor as a receptor (Cooper et al., 1984). It has been reported that MET expression is upregulated by androgen deprivation and MET appears to be preferentially expressed in androgen-insensitive, high-grade prostate cancer cells (Pisters et al., 1995; Humphrey et al., 1995). It has been also shown that overexpression of AR in prostate cancer PC3 cells leads to MET downregulation (Maeda et al., 2006). Based on our bindings, we could propose that both TES and MET at 7q31 are regulated by AR in a way to exhibit negative feedback for prostate cancer progression, while the former is a tumor suppressor gene and the latter is a proto-oncogene.

CDH2 encodes one of the calcium-dependent cell adhesion molecules, N-cadherin. Whereas another calcium-dependent cell adhesion molecules, E-cadherin, is expressed in epithelial cells, N-cadherin is expressed in nerve system, skeletal muscle and mesenchymal cells (Jaggi et al., 2006). Recent evidence suggests that changes in cadherin expression or cadherin switching play a critical role during progression of various tumors including breast cancer (Hazan et al., 1997) and prostate cancer (Tomita et al., 2000). Loss of E-cadherin expression was seen in high-grade breast and prostate cancers, whereas high levels of N-cadherin expression was shown in invasive tumors. Androgen responsiveness of N-cadherin has been shown in neurons, in spinal motoneurons (Monks and Watson, 2001). Our finding demonstrates that CDH2 is an androgen target gene with a novel cluster of ARE repeats in intron 1.

Interestingly, we found a polymorphism in terms of the number of ARE sequences in intron 1 of CDH2. In regard to the 15-bp complete palindromes consisting of the half-site GGTACA motif, LNCaP cells had 15 ARE sequences as shown in Figure 6b, whereas the genomic data published in NCBI Genome Browser contained 13 ARE sequences. We also found that 15 and 14 ARE repeats were contained in the CDH2 intron 1 derived from prostate cancer DU145 cells and benign prostate hyperplasia BPH1 cells, respectively (data not shown). By generating luciferase constructs including the ARE repeats of the CDH2 intron 1 derived from DU145 and BPH1, we showed that the luciferase activities of those constructs were also induced by R1881 stimulation, although the response of BPH1 was smaller than that of LNCaP or DU145. Thus, the polymorphism of ARE repeats in intron 1 may be related to the intensity of androgen responsiveness.

In this study, we demonstrated that UGT1A was a novel androgen-regulated gene with a functional ARE sequence in the 5′ upstream region of UGT1A1 or intron 1 in other UGT1A isoforms. Among several UGT1A isoforms, only UGT1A1 and UGT1A3 have been shown as androgen-upregulated genes in LNCaP cells. Considering out results, it is possible that the isoform-specific androgen responsiveness is linked with the closeness of the functional ARE in ARBS no. 1 to each isoform TSS. Thus, ChIP-chip would be also useful to dissect the isoform specificity of transcription factor target genes that encode a number of isoforms.

The human UGT1A locus spans 200 kb on chromosome 2q37 and encodes nine UGT1A enzymes that play a crucial role in glucuronidation of xenobiotics and endobiotic substrates such as bilirubin (Chen et al., 2005). UGT1A proteins are expressed in liver, whereas also expressed in extrahaptic tissues like urinary bladder and large intestine (Giuliani et al., 2005). UGT1A gene products are generated by a strategy of exon sharing, resulting in divergent isoforms with a unique N-terminal domain and commonly shared C-terminal 245 amino acids. As UGT proteins are detoxifying enzymes, it is natural that this gene expression is regulated by xenobiotic receptor including pregnenolone X receptor and constitutive androstane receptor (Sugatani et al., 2001; Xie et al., 2003). Reduction of UGT1A expression is involved in the early phase of neoplastic transformation, such as in liver and biliary cancer, bladder cancer and colon cancer (Strassburg et al., 1997; Giuliani et al., 2005). In contrast, decrease in UGT1A1 expression seems to be associated with the reduced risk of endometrial cancer (Duguay et al., 2004). UGT1A1 promoter polymorphism with an A(TA)7TAA element instead of a normal A(TA)6TAA element is known to decrease the level of gene expression, and it has been shown that there was a significant inverse association with the seven dinucleotide repeat allele and endometrial cancer risk (Duguay et al., 2004). UGT proteins also glucuronidate steroid hormones, as the UGT1A enzymes showing specificity for estrogens, whereas androgens are substrates for another type of UGT family, UGT2B proteins (Lepine et al., 2004). It has been recently shown that UGT2B15 isoform is an estrogen-regulated gene that is involved in the glucuronidation of androgens as well as estrogens (Harrington et al., 2006). Similarly, there is a possibility that UGT1A1 and UGT1A3 play a role in glucuronidation of androgens as well as estrogens.

In summary, we performed ChIP-chip analysis for in vivo ARBSs in prostate cancer LNCaP cells, on the ENCODE regions in the human genome. A number of novel androgen target genes were identified adjacent to the ChIP-chip-based ARBSs. The present results show that ChIP-chip has an advantage over transcript-based microarray analysis, identifying a number of bona fide AR target genes regardless of their expression levels based on the data of functional ARBSs. The androgen target genes identified by the present study would play various important roles in the maintenance of prostate cancer, including detoxification, protein degradation, cell motility/migration and tumor suppression/progression. Our study could be extended to the whole genome search of ARBSs in different cell systems using various ligands for the receptor. Identification of novel androgen target genes by ChIP-chip will reveal the whole entity of androgen signaling network, and will be applied to develop new clinical methods of prevention, diagnosis and treatment for prostate cancer.

Materials and methods

Reagents

Methyltrienolone 17β-hydroxy-17α-methyl-estra-4,9,11-trien-3-one (R1881) was purchased from NEN Life Science Products (Boston, MA, USA). Anti-AR (H-280), anti-SRC1 (M341), anti-GRIP1 (M343), anti-UGT1A (H-300) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-AcH3 and anti-AcH4 were from Upstate Biotechnology (Lake Placid, NY, USA). Anti-RNA PolII (8WG16) was from Covance (Berkeley, CA, USA). Anti-β-actin monoclonal antibody was from Sigma (St Louis, MO, USA). Anti-AIB1 antibody was generated from rabbit serum using a glutathione S-transferase fusion protein with amino acids 1320–1420 of human AIB1 protein as an epitope.

Cell culture

Human prostate cancer LNCaP cells were purchased from American Type Culture Collection (Rockville, MD, USA). Cells were maintained in RPMI 1640 supplemented with 4.5 g/dl glucose, 1 mM sodium pyruvate, 10 mM HEPES and 10% fetal bovine serum (FBS). Before hormone addition, cells were cultured for 2 days in phenol red-free RPMI 1640 with 5% dextran-charcoal stripped FBS (dcc-FBS) and 1 day in phenol red-free medium supplemented with 2.5% dcc-FBS.

Chromatin immunoprecipitation

ChIP assay and qPCR were performed as previously described (Horie-Inoue et al., 2004, 2006). LNCaP cells after 72-h hormone depletion were treated with 10 nM R1881 or vehicle (0.1% ethanol) for the indicated times. Cells were fixed in 1% formaldehyde for 5 min at room temperature. Chromatin was sheared to an average size of 500 bp by sonication using a Bioruptor ultrasonicator (Cosmo-Bio, Tokyo, Japan). Lysates were rotated at 4°C for overnight with specific antibodies. Salmon sperm DNA/protein A–agarose (Upstate Biotechnology, Lake Placid, NY, USA) was added and incubated for 2 h. Precipitated DNA was used as templates for qPCR using Applied Biosystems 7000 sequence detector (Foster City, CA, USA) based on SYBR Green I fluorescence. Genomic fragments containing ARE in the promoter and enhancer regions of PSA (−250/−39 bp and −4170/−3978 bp from the TSS, respectively) were used as positive controls for AR binding (Horie-Inoue et al., 2004). Sequences of PCR primers are described in Supplementary Table 1.

DNA amplification and microarray preparation

ChIP-enriched DNA was amplified by two-step IVT as described previously (Katou et al., 2006). Briefly, alkali phosphatase-treated ChIP DNA was incubated with terminal transferase for poly-dT tailing, annealed with T7-poly A primer (5′-IndexTermGCATTAGCGGCCGCGAAATTAATACGACTCACTATAGGGAGAAAAAAAAAAAAAAAAAA[C/T/G]-3′), and used as a template for second-strand cDNA synthesis. Using this template DNA, first IVT amplification was performed by T7 RNA polymerase (Ambion Inc., Austin, TX, USA). The first-strand cDNA was synthesized using the amplified cRNA as a template. Second-strand cDNA synthesis and IVT amplification were carried out again. Second amplified RNA was converted into double-strand cDNA with random primers, fragmented with DNase I and end labeled with biotin. Hybridization was performed on the Affymetrix GeneChIP ENCODE01 1.0 Arrays (Santa Clara, CA, USA) using 2 μg of ChIP-enriched and non-enriched input control DNA.

Analysis of microarray data

Array intensity data were analysed by the Affymetrix Tiling Analysis Software based on the algorithm by Cawley et al. (2004), and the results were mapped to genomic positions in human genome assembly hg 17 (NCBI Build 35) or in Affymetrix Integrated Genome Browser. In ENCODE01 1.0 Arrays, sets of one probe pair, a perfect matched (PM) probe and a mismatch probe (MM) both 25 bases long are tiled at an average resolution of 22 bp as measured from the central position of adjacent 25-mer oligos, creating an overlap of approximately 3 bp. The (PM-MM) intensity value was recorded for each probe pair as a new probe value, and the distribution of probe value was adjusted to equal across all samples by conducting quantile normalization on each duplicate arrays for two groups, including non-enriched genomic input DNA or ChIP-enriched DNAs by AR antibody. To determine whether a probe x is ChIP-enriched, Wilcoxon rank sum test was applied to rank all the probe pairs within a 550-bp sliding window from x by their log2(max(PM-MM),1) values for checking whether the sum of ranks of all probe pairs in the ChIP samples were significantly higher than that in the controls (a P-value cutoff of 1e-5). For each window, a signal ratio was also estimated by the Hedges–Lehmann method computing the median of folds enrichment among the probe sets within the window.

Reverse transcription-qPCR

Total RNA was extracted from hormone-treated or 0.1% ethanol-treated cells for indicated times using ISOGEN reagent (Nippon Gene, Tokyo, Japan). First strand cDNA was generated from RNase-free DNase I-treated total RNA by using SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) and oligo-dT20 primer. Androgen responsiveness was analysed by quantitative reverse transcription-PCR (RT-qPCR) using Applied Biosystems 7000 sequence detector based on SYBR Green I fluorescence. Primer design and PCR protocol were as previously described (Horie-Inoue et al., 2004, 2006). Sequences of PCR primers are described in Supplementary Table 2.

Sequence analysis

The sequences of human RefSeq transcripts (hg 17, NCBI build 35) were retrieved from UCSC genome browser (http://www.genome.ucsc.edu/) (Kent et al., 2002). The presence of ARE sequences in the genomic DNA of every ChIP-enriched region were determined by a position weighted matrix method TRANSFAC (Matys et al., 2003) with the matrix conservation >75%. If no ARE sequence was predicted by this criteria, the search was performed by a sequence analysis utility of JASPER, an open-access database for eukaryotic transcription factor binding profiles, with the relative profile score threshold >70% (Sandelin et al., 2004).

Luciferase assay

Luciferase reporter genes containing ARE sequences in ARBSs no. 1 and no. 10 identified by ChIP-chip were constructed by ligating the fragments derived from UDP-glucuronosyltransferase (UGT) 1A1 5′ upstream region (−17 753/−17 235 bp from the TSS) and cadherin-2 (CDH2) intron 1 region (+20 197/+21 260 bp from the TSS) into pGL3 vector (Promega, Madison, WI, USA) at the sites between MluI and XhoI, designated as UGT1A1 5′-Luc and CDH2 Int 1-Luc, respectively. A mutated UGT1A1 5′ region construct (UGT1A1 5′ Mut-Luc) was also generated, including the identical region of UGT1A1 5′-Luc except two substitutions of conserved C and G for A and T, respectively, at the 2-bp apart positions from the 3-bp spacer of ARE sequence. Mouse mammary tumor virus luciferase construct (MMTV-Luc) was used as a positive control for AR transcription activity (Ogawa et al., 1995). LNCaP cells were plated at a density of 10 000 cells/well in a 24-well culture plate and cultured for 3 days in phenol red-free RPMI 1640 with 5% dcc-FBS. Cells were transfected with plasmids using the transfection reagent FuGENE6 (Roche Applied Science, Indianapolis, IN, USA), then 12 h later treated with R1881 (10 nM) or vehicle (0.1% ethanol) for 24 h. Luciferase activity of cell lysate was determined by the Dual Luciferase Assay Kit (Promega, Madison, WI, USA). A renilla luciferase reporter Tk-PRL was co-transfected as a control for evaluating transfection efficiency. Data represent means±s.d. from triplicate sets.

Western blotting

Whole cell lysates were prepared using lysis buffer (50 mM Tris–HCl, pH 8.0, 150 mM NaCl, 1% TritonX-100, 1.5 mM MgCl2, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 mM PMSF). Protein concentrations were analyzed using the BCA protein assay kit (Pierce Biotechnology, Rockford, IL, USA). Fifty microgram of proteins were resolved by 8% SDS–polyacrylamide gel electrophoresis and electroblotted onto Immobilon-P Transfer Membrane (Millipore, Billerica, MA, USA). Membranes were incubated with primary antibodies followed by incubation with secondary antibodies. Antibody–antigen complexes were detected using the Western Blotting Chemiluminescence Luminol Reagent (Santa Cruz Biotechnology, Santa Cruz, CA, USA).

References

  1. Bernstein BE, Kamal M, Lindblad-Toh K, Bekiranov S, Bailey DK, Huebert DJ et al. (2005). Genomic maps and comparative analysis of histone modifications in human and mouse. Cell 120: 169–181.

    CAS  Article  Google Scholar 

  2. Carroll JS, Liu XS, Brodsky AS, Li W, Meyer CA, Szary AJ et al. (2005). Chromosome-wide mapping of estrogen receptor binding reveals long-range regulation requiring the forkhead protein FoxA1. Cell 122: 33–43.

    CAS  Article  Google Scholar 

  3. Cawley S, Bekiranov S, Ng HH, Kapranov P, Sekinger EA, Kampa D et al. (2004). Unbiased mapping of transcription factor binding sites along human chromosomes 21 and 22 points to widespread regulation of noncoding RNAs. Cell 116: 499–509.

    CAS  Article  Google Scholar 

  4. Chen CD, Welsbie DS, Tran C, Baek SH, Chen R, Vessella R et al. (2004). Molecular determinants of resistance to antiandrogen therapy. Nat Med 10: 33–39.

    Article  Google Scholar 

  5. Chen S, Beaton D, Nguyen N, Senekeo-Effenberger K, Brace-Sinnokrak E, Argikar U et al. (2005). Tissue-specific, inducible, and hormonal control of the human UDP-glucuronosyltransferase-1 (UGT1) locus. J Biol Chem 280: 37547–37557.

    CAS  Article  Google Scholar 

  6. Chene L, Giroud C, Desgrandchamps F, Boccon-Gibod L, Cussenot O, Berthon P et al. (2004). Extensive analysis of the 7q31 region in human prostate tumors supports TES as the best candidate tumor suppressor gene. Int J Cancer 111: 798–804.

    CAS  Article  Google Scholar 

  7. Cooper CS, Park M, Blair DG, Tainsky MA, Huebner K, Croce CM et al. (1984). Molecular cloning of a new transforming gene from a chemically transformed human cell line. Nature 311: 29–33.

    CAS  Article  Google Scholar 

  8. Coutts AS, MacKenzie E, Griffith E, Black DM . (2003). TES is a novel focal adhesion protein with a role in cell spreading. J Cell Sci 116: 897–906.

    CAS  Article  Google Scholar 

  9. Diaz M, Rodriguez JC, Sanchez J, Sanchez MT, Martin A, Merino AM et al. (2002). Clinical significance of pepsinogen C tumor expression in patients with stage D2 prostate carcinoma. Int J Biol Markers 17: 125–129.

    CAS  Article  Google Scholar 

  10. Duguay Y, McGrath M, Lepine J, Gagne JF, Hankinson SE, Colditz GA et al. (2004). The functional UGT1A1 promoter polymorphism decreases endometrial cancer risk. Cancer Res 64: 1202–1207.

    CAS  Article  Google Scholar 

  11. ENCODE Project Consortium, EP (2004). The ENCODE (ENCyclopedia of DNA Elements) Project. Science 306: 636–640.

    Article  Google Scholar 

  12. Giuliani L, Ciotti M, Stoppacciaro A, Pasquini A, Silvestri I, De Matteis A et al. (2005). UDP-glucuronosyltransferases 1A expression in human urinary bladder and colon cancer by immunohistochemistry. Oncol Rep 13: 185–191.

    CAS  PubMed  Google Scholar 

  13. Gong QH, Cho JW, Huang T, Potter C, Gholami N, Basu NK et al. (2001). Thirteen UDP glucuronosyltransferase genes are encoded at the human UGT1 gene complex locus. Pharmacogenetics 11: 357–368.

    CAS  Article  Google Scholar 

  14. Grossmann ME, Huang H, Tindall DJ . (2001). Androgen receptor signaling in androgen-refractory prostate cancer. J Natl Cancer Inst 93: 1687–1697.

    CAS  Article  Google Scholar 

  15. Harrington WR, Sengupta S, Katzenellenbogen BS . (2006). Estrogen regulation of the glucuronidation enzyme UGT2B15 in estrogen receptor-positive breast cancer cells. Endocrinology 147: 3843–3850.

    CAS  Article  Google Scholar 

  16. Hazan RB, Kang L, Whooley BP, Borgen PI . (1997). N-cadherin promotes adhesion between invasive breast cancer cells and the stroma. Cell Adhes Commun 4: 399–411.

    CAS  Article  Google Scholar 

  17. Horie-Inoue K, Bono H, Okazaki Y, Inoue S . (2004). Identification and functional analysis of consensus androgen response elements in human prostate cancer cells. Biochem Biophys Res Commun 325: 1312–1317.

    CAS  Article  Google Scholar 

  18. Horie-Inoue K, Takayama K, Bono HU, Ouchi Y, Okazaki Y, Inoue S . (2006). Identification of novel steroid target genes through the combination of bioinformatics and functional analysis of hormone response elements. Biochem Biophys Res Commun 339: 99–106.

    CAS  Article  Google Scholar 

  19. Humphrey PA, Zhu X, Zarnegar R, Swanson PE, Ratliff TL, Vollmer RT et al. (1995). Hepatocyte growth factor and its receptor (c-MET) in prostatic carcinoma. Am J Pathol 147: 386–396.

    CAS  PubMed  PubMed Central  Google Scholar 

  20. Jaggi M, Nazemi T, Abrahams NA, Baker JJ, Galich A, Smith LM et al. (2006). N-cadherin switching occurs in high Gleason grade prostate cancer. Prostate 66: 193–199.

    CAS  Article  Google Scholar 

  21. Katou Y, Kaneshiro K, Aburatani H, Shirahige K . (2006). Genomic approach for the understanding of dynamic aspect of chromosome behavior. Methods Enzymol 409: 389–410.

    CAS  Article  Google Scholar 

  22. Kawana Y, Ichikawa T, Suzuki H, Ueda T, Komiya A, Ichikawa Y et al. (2002). Loss of heterozygosity at 7q31.1 and 12p13-12 in advanced prostate cancer. Prostate 53: 60–64.

    CAS  Article  Google Scholar 

  23. Kent WJ, Sugnet CW, Furey TS, Roskin KM, Pringle TH, Zahler AM et al. (2002). The human genome browser at UCSC. Genome Res 12: 996–1006.

    CAS  Article  Google Scholar 

  24. Laganière J, Deblois G, Lefebvre C, Bataille AR, Robert F, Giguère V . (2005). From the cover: location analysis of estrogen receptor alpha target promoters reveals that FOXA1 defines a domain of the estrogen response. Proc Natl Acad Sci USA 102: 11651–11656.

    Article  Google Scholar 

  25. Lepine J, Bernard O, Plante M, Tetu B, Pelletier G, Labrie F et al. (2004). Specificity and regioselectivity of the conjugation of estradiol, estrone, and their catecholestrogen and methoxyestrogen metabolites by human uridine diphospho-glucuronosyltransferases expressed in endometrium. J Clin Endocrinol Metab 89: 5222–5232.

    CAS  Article  Google Scholar 

  26. Maeda A, Nakashiro K, Hara S, Sasaki T, Miwa Y, Tanji N et al. (2006). Inactivation of AR activates HGF/c-Met system in human prostatic carcinoma cells. Biochem Biophys Res Commun 347: 1158–1165.

    CAS  Article  Google Scholar 

  27. Matys V, Fricke E, Geffers R, Gossling E, Haubrock M, Hehl R et al. (2003). TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res 31: 374–378.

    CAS  Article  Google Scholar 

  28. Monks DA, Watson NV . (2001). N-cadherin expression in motoneurons is directly regulated by androgens: a genetic mosaic analysis in rats. Brain Res 895: 73–79.

    CAS  Article  Google Scholar 

  29. Ogawa H, Inouye S, Tsuji FI, Yasuda K, Umesono K . (1995). Localization, trafficking, and temperature-dependence of the Aequorea green fluorescent protein in cultured vertebrate cells. Proc Natl Acad Sci USA 92: 11899–11903.

    CAS  Article  Google Scholar 

  30. Pisters LL, Troncoso P, Zhau HE, Li W, von Eschenbach AC, Chung LW . (1995). c-met Proto-oncogene expression in benign and malignant human prostate tissues. J Urol 154: 293–298.

    CAS  Article  Google Scholar 

  31. Porkka KP, Helenius MA, Visakorpi T . (2002). Cloning and characterization of a novel six-transmembrane protein STEAP2, expressed in normal and malignant prostate. Lab Invest 82: 1573–1582.

    CAS  Article  Google Scholar 

  32. Sandelin A, Alkema W, Engstrom P, Wasserman WW, Lenhard B . (2004). JASPAR: an open-access database for eukaryotic transcription factor binding profiles. Nucleic Acids Res 32: D91–94.

    CAS  Article  Google Scholar 

  33. Shang Y, Brown M . (2002). Molecular determinants for the tissue specificity of SERMs. Science 295: 2465–2468.

    CAS  Article  Google Scholar 

  34. Shang Y, Myers M, Brown M . (2002). Formation of the androgen receptor transcription complex. Mol Cell 9: 601–610.

    CAS  Article  Google Scholar 

  35. Strassburg CP, Manns MP, Tukey RH . (1997). Differential down-regulation of the UDP-glucuronosyltransferase 1A locus is an early event in human liver and biliary cancer. Cancer Res 57: 2979–2985.

    CAS  PubMed  Google Scholar 

  36. Sugatani J, Kojima H, Ueda A, Kakizaki S, Yoshinari K, Gong QH et al. (2001). The phenobarbital response enhancer module in the human bilirubin UDP-glucuronosyltransferase UGT1A1 gene and regulation by the nuclear receptor CAR. Hepatology 33: 1232–1238.

    CAS  Article  Google Scholar 

  37. Tobias ES, Hurlstone AF, MacKenzie E, McFarlane R, Black DM . (2001). The TES gene at 7q31.1 is methylated in tumours and encodes a novel growth-suppressing LIM domain protein. Oncogene 20: 2844–2853.

    CAS  Article  Google Scholar 

  38. Tomita K, van Bokhoven A, van Leenders GJ, Ruijter ET, Jansen CF, Bussemakers MJ et al. (2000). Cadherin switching in human prostate cancer progression. Cancer Res 60: 3650–3654.

    CAS  PubMed  Google Scholar 

  39. Wang Q, Carroll JS, Brown M . (2005). Spatial and temporal recruitment of androgen receptor and its coactivators involves chromosomal looping and polymerase tracking. Mol Cell 19: 631–642.

    CAS  Article  Google Scholar 

  40. Xie W, Yeuh MF, Radominska-Pandya A, Saini SP, Negishi Y, Bottroff BS et al. (2003). Control of steroid, heme, and carcinogen metabolism by nuclear pregnane X receptor and constitutive androstane receptor. Proc Natl Acad Sci USA 100: 4150–4155.

    CAS  Article  Google Scholar 

Download references

Acknowledgements

We thank T Suzuki and R Nozawa for their technical assistance. This work was supported in part by grants-in-aid from the Ministry of Health, Labor and Welfare; from the Japan Society for the Promotion of Science; from The Promotion and Mutual Aid Corporation for Private Schools of Japan. This work was supported in part by a grant of the Genome Network Project from the Ministry of Education, Culture, Sports, Science and Technology.

Author information

Affiliations

Authors

Corresponding author

Correspondence to S Inoue.

Additional information

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Takayama, K., Kaneshiro, K., Tsutsumi, S. et al. Identification of novel androgen response genes in prostate cancer cells by coupling chromatin immunoprecipitation and genomic microarray analysis. Oncogene 26, 4453–4463 (2007). https://doi.org/10.1038/sj.onc.1210229

Download citation

Keywords

  • androgen receptor
  • androgen response element
  • chromatin immunoprecipitation
  • prostate cancer
  • UGT1A

Further reading

Search

Quick links