The androgen receptor (AR) is a critical transcriptional factor that contributes to the development and the progression of prostate cancer (PCa) by regulating the transcription of various target genes. Genome-wide screening of androgen target genes provides useful information to understand a global view of AR-mediated gene network in PCa. In this study, we performed 5′-cap analysis of gene expression (CAGE) to determine androgen-regulated transcription start sites (TSSs) and chromatin immunoprecipitation (ChIP) on array (ChIP-chip) analysis to identify AR binding sites (ARBSs) and histone H3 acetylated (AcH3) sites in the human genome. CAGE determined 13 110 distinct, androgen-regulated TSSs (P<0.01), and ChIP-chip analysis identified 2872 androgen-dependent ARBSs (P<1e-5) and 25 945 AcH3 sites (P<1e-4). Both androgen-regulated coding genes and noncoding RNAs, including microRNAs (miRNAs) were determined as androgen target genes. Besides prototypic androgen-regulated TSSs in annotated gene promoter regions, there are many androgen-dependent TSSs that are widely distributed throughout the genome, including those in antisense (AS) direction of RefSeq genes. Several pairs of sense/antisense promoters were newly identified within single RefSeq gene regions. The integration of CAGE and ChIP-chip analyses successfully identified a cluster of androgen-inducible miRNAs, as exemplified by the miR-125b-2 cluster on chromosome 21. Notably, the number of androgen-upregulated genes was larger in LNCaP cells treated with R1881 for 24 h than for 6 h, and the percentage of androgen-upregulated genes accompanied with adjacent ARBSs was also much higher in cells treated with R1881 for 24 h than 6 h. On the basis of the Oncomine database, the majority of androgen-upregulated genes containing adjacent ARBSs and CAGE tag clusters in our study were previously confirmed as androgen target genes in PCa. The integrated high-throughput genome analyses of CAGE and ChIP-chip provide useful information for elucidating the AR-mediated transcriptional network that contributes to the development and progression of PCa.
Androgen action is essential for the development, proliferation and subsequent progression of prostate cancer (PCa). Androgen binds to its cognate receptor, androgen receptor (AR), a member of the nuclear receptor superfamilies that functions as a ligand-dependent transcriptional factor (Shang et al., 2002; Wang et al., 2005; Dehm and Tindall, 2006). It has been shown that AR and its downstream signals are deeply involved in the pathophysiology of hormone-dependent and hormone-independent PCas (Suzuki et al., 2003; Chen et al., 2004; Debes and Tindall, 2004). Therefore, elucidation of the entire AR signaling pathways will reveal the precise mechanisms underlying the development and the progression of PCa and will identify novel molecular targets for cancer therapy. Recent advances in high-throughput gene analysis technology have enabled to identify a number of target genes that are associated with diseases at various statistical thresholds. Our group and others have successfully determined bonafide AR binding sites (ARBSs) in the human genome by chromatin immunoprecipitation (ChIP) analysis combined with genome tiling arrays (ChIP-chip) (Massie et al., 2007; Takayama et al., 2007, 2009; Wang et al., 2007, 2009). ChIP-chip analysis is also useful to determine epigenetic alterations in the genome. We have identified that amyloid β-precursor protein is an androgen-inducible gene that could be a novel prognostic and therapeutic target of PCa by screening genes in the vicinity of both ARBSs and acetylated histone H3 (AcH3) sites (Takayama et al., 2009).
In terms of transcriptome, cDNA microarray has been extensively used for years as a convenient and cost-effective technique for analyzing gene expression signatures. cDNA microarray, however, is not designed to identify novel genes without probes, thus it has limitations to analyze the expression of noncoding RNAs, including short RNAs. As some microRNAs (miRNAs) and novel RNA transcripts have been reported to be relevant to PCa (Louro et al., 2007; Shi et al., 2007), an alternative high-throughput transcriptome technology will be also required to analyze noncoding RNA expression as well as coding RNA profiles.
In this study, we investigated AR actions across the entire human genome to elucidate the androgen-mediated transcriptional network by performing ChIP-chip and cap analysis gene expression (CAGE). We could determine novel AR-mediated genomic actions, in particular, those dependent on antisense (AS) or intergenic transcription start sites (TSSs) that have not been previously identified by any other methods. Thus, our integrated approach identify novel androgen target genes and provides potential diagnostic and therapeutic molecular targets of PCa, which can be applied to the clinical management of the advanced disease.
Genome-wide analysis of androgen-activated promoters in prostate cancer cells
CAGE involves the preparation and sequencing of concatemers of 20-nucleotide (nt) DNA tags that were derived from the 5′-end of capped mRNA, and mapping the CAGE tags to the genome (Shiraki et al., 2003). This method is cost-effective and enables high-throughput genome-wide analyses of TSSs and promoter usage. For preparing CAGE samples, we stimulated LNCaP cells with R1881 (10 nM), which were cultured in a hormone-depleted medium for 72 h. We extracted total RNA from cells stimulated for 6 and 24 h, and synthesized first-strand cDNA by reverse transcription. After cutting and amplifying the 20-nt CAGE tags derived from the 5′-ends of mRNAs, we sequenced the tag-ligated concatemers. We then mapped the sequenced CAGE tags to the human genome (total tags at 0 h were 1 929 416; at 6 h, 682 111; and at 24 h, 586 083). Regions containing multiple tags were referred to as tag clusters (TCs); these clusters were diversely distributed throughout the genome. By using Fisher’s exact test, we identified TCs whose distributions were significantly regulated by androgen. The total number of androgen-dependent TCs obtained was 13 110 (P<0.01); 1572 TCs were obtained using a stringent threshold (1e-6.11 with Bonferroni correction). Among the androgen-regulated TCs, 78.9% were upregulated and the remaining was downregulated at 6 or 24 h after stimulation. In the promoter region of a representative androgen target gene FKBP5 (Velasco et al., 2004), we detected a time-dependent increase in tag numbers by androgen stimulation (21.2, 80.6 and 474.3 tags accumulated per million tags (t.p.m.) at 0, 6 and 24 h after R1881 treatment, respectively) (Figure 1a). Androgen-dependent increase in CAGE TCs was also shown in the promoter regions of PSA/KLK3 and TMPRSS2 (Supplementary Figure 1). The 5′-termini of RefSeq genes are often different from the TSSs obtained by CAGE analysis (Katayama et al., 2007). For example, our study shows that TRIM36 includes an intronic CAGE TC within the second AcH3 site, which is situated in the downstream of the intronic ARBS (Supplementary Figure 2). Only 35% of CAGE TCs were located at a distance of <1 kb from the TSSs of known RefSeq genes (Figure 1b). In total, 13% of the TCs were identified at intergenic regions that were located >100 kb upstream of TSSs and did not overlap with any RefSeq gene. In all, 7% of the TCs were genome-widely distributed at the antisense regions of RefSeq genes, and 33% were transcribed from the RefSeq regions, most of which were from novel TSSs. Global transcriptome analysis using the CAGE technique and large-scale cDNA sequencing in the FANTOM3 project has recently shown that a large proportion of the mammalian genome can transcribe genes from both sense (S) and antisense strands, and that many of the unknown transcripts are noncoding RNAs (Carninci et al., 2005). Taken together, these findings suggest that androgen-regulated transcripts are widely distributed across annotated genes and noncoding RNAs.
We also compared the androgen responsiveness of our CAGE TCs with RefSeq genes. Among protein-coding genes with an increasing number of CAGE TCs (>20 t.p.m. by 24 h) at their promoter regions (<0.1 kb from TSS), 51% of genes were also significantly upregulated (>1.5-fold) by 24-h treatment with R1881 (1 nM) in LNCaP cells based on the microarray data set GSE14028 (Yu et al., 2010, P=2.0e-24). In contrast, 40% of protein-coding genes with a decreasing number of TCs were significantly downregulated (<0.8-fold) by R1881 in LNCaP cells (Yu et al., 2010, P=2.7e-5). On the basis of another microarray data GSE7868 (Wang et al., 2007), 44% of the RefSeq genes with an increasing number of CAGE TCs were upregulated in LNCaP cells treated with dihydrotestosterone (100 nM) for 16 h (P=4.2e-28) and 45% of the RefSeq genes with a decreasing number of TCs were downregulated by dihydrotestosterone (P=1.7e-36).
Genome-wide identification of ARBSs in prostate cancer cells
ChIP-chip analysis revealed 6366 (P<1e-4), or 2872 (P<1e-5) specific ARBSs in the genome of LNCaP cells. To compare our ChIP-chip data with previous studies, we found that 77 and 74% of our AR binding (P<1e-5) were involved in the ChIP-chip data (Wang et al., 2009) and the ChIP-Seq data (Yu et al., 2010), respectively. We also identified 25 945 AcH3 sites (P<1e-4) as epigenetic markers that were associated with activated promoters (Bernstein et al., 2005). Among 2872 ARBSs (P<1e-5), 420 ARBSs (15%) were overlapped with AcH3 sites. In addition, approximately 37.6% of AcH3 sites were mapped to the proximal regions of the TSSs of RefSeq genes (<3 kb from the TSSs), including many sites located in the downstream proximal regions of TSSs (Figure 1c), as consistent with a previous report in regard to genome-wide analysis of active promoters (Kim et al., 2005). In terms of the overlap between the AcH3 sites and the CAGE TCs, 41% of the CAGE TCs were overlapped with AcH3 sites and 49% were overlapped with AcH3 sites within a distance <3 kb. On the other hand, most of the ARBSs were found to be widely distributed across the human genome, and only 4.8% of the ARBSs were mapped to promoter regions at <1 kb from the TSSs (Figure 1c).
We randomly selected 25 ARBSs in the vicinity of known androgen targets such as ACSL3, and intergenic regions on chromosomes 15, 20 and X to perform conventional ChIP analysis for AR binding in LNCaP cells (Figure 2a). When compared with four genomic regions as negative controls, most of the investigated regions were confirmed to be androgen-dependent ARBSs.
To further investigate the effects of the histone acetylation status on the AR-dependent transcriptional activity, we performed ChIP analysis for AR, SRC1, AcH3 and RNA Pol II recruitment for the group of ARBSs that overlapped with AcH3 sites (AcH3 overlap (+)) and the group of ARBSs distinct from the AcH3 sites (AcH3 overlap (−)) (Figure 2b). Significant androgen-dependent recruitment of both SRC1 and RNA Pol II was detected (>1.5-fold) in the AcH3 overlap (+) group, but not in the AcH3 overlap (−) group.
ChIP-chip and CAGE analyses reveal AR-dependent regulation of androgen target genes
Novel ARBSs and AcH3 sites adjacent to CAGE TCs were identified in the androgen-regulated genes. For instance, we found novel ARBSs in the intronic regions of α-methyl-CoA racemase, which is a prognostic marker for PCa (Luo et al., 2002) (Figure 3a). Interestingly, AR binding was also observed in an intronic region of insulin-like growth factor-1 receptor, whose expression has been previously reported to be dependent on the nongenomic action of androgen (Pandini et al., 2009); thus, we presume that the genomic activity of AR will also contribute to the transcriptional modification of insulin-like growth factor-1 receptor (Figure 3b). Identification of ARBSs in other androgen-regulated genes, such as NK3 homeobox 1 (Prescott et al., 1998) (Figure 3c) and jagged 1 (Santagata et al., 2004) (Figure 3d), also provide further evidence that direct actions of AR is required for their transcriptional regulation.
S/AS transcriptional regulation by androgen showed diverse alteration patterns
Genome-wide promoter exchange such as bidirectional or sense/antisense (S/AS) transcriptional regulation is known to occur in the human genome. Antisense transcripts associate with neighboring genes in complex loci to form chains of linked transcriptional units (Katayama et al., 2005). We investigated the number of antisense-directed TCs that were associated with ARBSs: of the 1572 TCs, 173 (11% with Bonferroni correction) were situated on the antisense strand of RefSeq genes and 34 of these 173 TCs (19%) were associated with ARBSs within a 100-kb region (P<1e-5).
Antisense TCs were mostly abundant in the 3′-regions of RefSeq genes (Figure 4a). However, sense TCs were abundant in the proximal regions of the TSSs of RefSeq genes (Figure 4b). We identified 39 RefSeq genes with bidirectional S/AS pairs of CAGE TCs in the human genome. Among these S/AS pairs, 15 of 39 S/AS pairs exhibited reciprocal alteration of tag numbers in response to androgen, whereas other pairs exhibited concordant alteration. With regard to the androgen-dependent reciprocal patterns of S/AS pairs, one pattern consisted of an upregulated sense TC present in the TSS region of a RefSeq gene and a downregulated intronic antisense TC; this pattern was observed for the androgen-inducible gene DNM1L (Figure 4c). The other pattern consisted of a downregulated sense TC present in the TSS region of a RefSeq gene and an upregulated intronic sense TC, as exemplified by the androgen-repressed gene SLC7A1 (Figure 4d). The contribution of an antisense promoter to the expression of a gene needs to be studied; nevertheless, CAGE analysis is a useful technology to clarify the complexity of androgen-regulated transcription in PCa cells.
ChIP-chip analysis reveals miRNA regulation by AR
Previous studies on expression profiling in PCa have revealed that ∼20-nt short RNAs or miRNAs are closely associated with the progression of the disease. On the basis of ChIP-chip data, we searched miRNAs in the vicinity of ARBSs; we identified four miRNAs in three miRNA clusters located at a distance of <100 kb from the ARBSs. We also identified four additional miRNAs that were relatively adjacent to the ARBSs or AcH3 sites, which were recently shown to be relevant in the progression of PCa (Shi et al., 2007; Leite et al., 2009; Ribas et al., 2009; Sun et al., 2009) (Table 1). Among them, miR-21 is a characterized AR-regulated miRNA that promotes hormone-dependent and -independent PCa growth (Ribas et al., 2009). In our analysis, miR-21 is surrounded by a 1-kb upstream AcH3 site and a 28-kb downstream ARBS (Supplementary Figure 3a). We validated that miR-21 expression was upregulated by R1881 in LNCaP cells (Supplementary Figure 3b). Furthermore, we focused on the miRNA-clustered region of miR-125b-2 on chromosome 21; miRNAs let-7c, miR-99a and miR-125b-2 were clustered in the intronic regions of C21orf34 gene (Figures 5a and b). These results will also provide evidence that androgen genomic actions regulate miRNAs, as miR-125b has been shown to be an androgen-inducible miRNA that can be associated with PCa cell proliferation (Shi et al., 2007), and let-7c is reported as an upregulated miRNA in metastatic PCa (Leite et al., 2009). miR-99a is located at 131 kb from the nearest ARBS, although the miRNA has not been well-characterized in PCa. Quantitative RT–PCR (qRT–PCR) revealed that miR-125b-2, let-7c and miR-99a exhibited androgen-dependent upregulation 48 h after R1881 treatment (Figure 5c), thereby indicating that all of the three miRNAs on the miR-125b-2 locus are putative AR-regulated targets in PCa cells.
Identification of AR regulated genes by ChIP-chip analysis
We explored androgen-regulated genes based on the data of ChIP-chip and CAGE analyses. Within a region of 100 kb from the ARBSs, 44, 66 and 134 androgen-upregulated CAGE TCs were identified by the comparison of time points at 6 versus 0 h, 24 versus 6 h and 24 versus 0 h, respectively (Figure 6a). mRNA microarray analysis revealed that 254 and 1477 genes were upregulated (>twofold) after 6 and 24 h of androgen treatment, respectively (Figure 6b). Among the genes upregulated by androgen, the rate of ARBS positivity (at <100 kb) was 21.6% at 24 h and 7.2% at 6 h (Figure 6b). AR binding bias was not significantly shown in upregulated genes at 6 h only versus the genomic background (P=0.23). In contrast, ARBS positivity was highly enriched in the upregulated genes at 24 h versus the genomic background (P=1.4e-19). These results show that the AR binding will contribute to the androgen-dependent transcription by 24 h. Moreover, representative qRT–PCR data showed that known and putative androgen target genes exhibited time-dependent upregulation and that the androgen-dependent upregulation was continuously observed up to at least for 48 h after treatment (Figure 6c). Taken together, we identified 258 novel upregulated genes (>twofold change at 24-h treatment), which may be putative direct targets of AR.
Androgen target genes are associated with tumorigenesis
To elucidate the role of putative androgen target genes in the tumorigenesis of PCa, we reviewed the expression profiles of 508 genes that were differentially regulated in our experiments (258 androgen-upregulated genes and 250 androgen-downregulated genes) in normal and cancer (PCa) regions of prostate tissues. cDNA microarray data for 62 primary tumors and 41 normal tissue samples (Lapointe et al., 2004) were downloaded from the Stanford University website (http://microarray-pubs.stanford.edu/prostateCA); genes upregulated and downregulated by androgen were hierarchically clustered using the pairwise average-linkage method depending on the gene expression features of normal against tumor predominancy (Figure 6d). The cluster of genes with higher expression in PCa tissues than in normal prostate tissues is depicted as normal <PCa (red square), and the other cluster of genes with higher expression in normal prostate than in PCa samples is depicted as normal >PCa (blue square). The normal <PCa cluster containing androgen-upregulated genes (group I) included genes such as ACACA and α-methyl-CoA racemase (related to amino acid metabolism) and the normal >PCa cluster containing androgen-upregulated genes (group II) included genes such as EGFR and EFNA5 (related to signal transduction). The normal <PCa cluster containing androgen-downregulated genes (group III) included genes such as NFIX and SOX4 (related to transcriptional regulation), and the normal >PCa cluster containing androgen-downregulated genes (group IV) included genes such as TGFBR3 and FOXO1A (related to signal transduction and transcriptional regulation). Among the androgen-upregulated genes, the proportion of group I genes was significantly higher than that of group II genes. The proportion of group IV was also substantially higher than that of group III genes among the androgen-downregulated genes. The rates of the genes with the expressions altered by androgen were significantly higher in the clinical microarray data overlapped with LNCaP data compared with the randomly selected genes overlapped with the clinical data (P=3.5e-9 for upregulated genes and P=0.001 for downregulated genes). This clustering analysis of microarray data will indicate that the androgen-dependent transcriptional regulation is closely associated with the development of PCa.
Finally, we extracted novel androgen-inducible genes that are regulated by direct AR actions. We selected 569 RefSeq genes that exhibited androgen-inducible CAGE TCs 24 h after treatment at <0.1 kb from their TSSs. Among them, 51% were induced by >1.5-fold at 24 h after R1881 treatment, indicating that the majority of CAGE TCs function as bonafide androgen-regulated promoters. We analyzed novel AR target genes with CAGE TCs in the promoter regions or within the genes. Of the AR target genes, 31% (82 genes) included androgen-regulated CAGE TCs (at least 20 t.p.m.) (Figure 6e). In the Oncomine database (Wilson and Giguère, 2007), we identified 112 genes that were upregulated by androgen (P<1e-4) in PCa tissues (Figure 6f). Among these androgen-upregulated genes, 48 of 82 genes (58%) are associated with adjacent CAGE TCs, whereas 64 of 176 (36%) did not exhibit CAGE TCs (P<0.001).
This study revealed a global approach for mapping of AR binding and AR-dependent promoters, which provides the systemic database regarding the AR-regulated gene network that essentially contributes to the development and progression of PCa. Notably, this study successfully confirmed that known androgen-regulated or PCa-specific genes are androgen-inducible, as exemplified by ACACA (Lapointe et al., 2004), ABCC4 (Ho et al., 2008), TRIM36 (Balint et al., 2004) and amyloid β-precursor protein (Takayama et al., 2009)(Table 2). On the basis of global mapping analysis, we identified novel androgen-upregulated genes, such as FGFRL1, CAMKK2 and C16orf60, although their functions remain to be studied in PCa (Table 2). Notably, the application of pathway analysis using Database for Annotation, Visualization and Integrated Discovery Bioinformatics Resources 6.7 (Dennis et al., 2003; Huang et al., 2009) further shows that metabolic pathways, such as diabetes pathways, metabolism of lipids and lipoproteins, or amino sugar and nucleotide sugar metabolism, were enriched in our CAGE data (Supplementary Table 2).
To understand the mechanism underlying AR regulation of the progression of PCa, we identified its transcriptional network in the genome of human PCa LNCaP cells. In the previous studies, we identified direct AR target genes by using the ChIP-cloning technique (Takayama et al., 2008) or tiling arrays for the Encyclopedia Of DNA Elements regions of the human genome (Takayama et al., 2007) or chromosomes 21 and 22 (Takayama et al., 2009). A number of genes adjacent to ARBSs were found to be upregulated or downregulated by androgen. In this study, we used the high-throughput CAGE method for mapping androgen-regulated promoters in the human genome. To validate the present CAGE and ChIP-chip results, we mapped the data on the genome browser adjacent to prototypic AR-regulated genes. Androgen treatment substantially increased the number of CAGE tags found in the promoter regions of AR-regulated genes such as PSA/KLK3, TMPRSS2 and FKBP5.
We performed both CAGE and microarray analyses for comparing the two transcriptome technologies. Microarray analysis is useful for understanding global expression profiles of protein-coding genes, whereas CAGE is useful for detecting alterations in expressions of individual promoters for both noncoding and protein-coding genes. The similarity between the two data sets is the expression alterations of protein-coding genes with CAGE tags in their promoters or within the genes. This tendency is well observed among the genes with CAGE tags in their promoters with a high TC. In contrast, the correlation between CAGE and microarray is lower for genes with multiple promoter regions as shown by the CAGE study in myeloid leukemia THP-1 cells (FANTOM Consortium, 2009).
Our combined analyses can clarify the direct role of AR in the transcriptional regulation of androgen target genes, including genes that were not previously considered to be regulated by androgen. Notably, we identified an intronic ARBS in the insulin-like growth factor-1 receptor gene, which indicates direct AR-mediated transcriptional regulation of this known androgen nongenomic target of androgen. The identification of a novel ARBS in the JAG1 gene, a canonical ligand for Notch1, also suggests that AR through this binding site has a critical role in the progression and metastasis of PCa. In this study, we also confirmed the involvement of transcription factors in the AR-mediated gene network, as SOX4, one of the developmental transcription factors overexpressed in PCa (Liu et al., 2006), was found to contain a functional ARBS within the gene. The calcium-dependent protein kinase CAMKK2 (Table 2) seems to be an interesting novel target of AR identified in this study. CAMKK2 can affect cell growth by upregulating AMPK phosphorylation (Yu et al., 2009), and it is also known to be upregulated in PCa.
In this study, we determined that transcriptional regulation was also mediated by antisense promoters in LNCaP cells. The antisense-directed TSSs were relatively abundant in the 3′-regions of RefSeq genes. This finding is consistent with those of previous studies that reported that antisense transcripts are often transcribed from the 3′-untranslated regions of genes. The relevance of antisense promoter-mediated transcriptional regulation in carcinogenesis has been reported in HCT116 colorectal cancer cells, as the binding of β-Catenin to its target E2F4 at the E2F4 3′-untranslated region induces an E2F4 antisense transcript, which downregulates E2F4 protein levels and suppresses the binding of E2F4 to its specific target genes (Yochum et al., 2007). We also found novel pairs of S/AS promoters in individual RefSeq genes. Both reciprocal and concordant alterations in promoter transcription are known to occur in PCa. Taken together, the abundant expression of androgen-regulated antisense promoters in LNCaP cells will indicate the physiological significance of antisense mechanisms in the AR-mediated gene network in PCa.
This integrated study also identified short RNAs or miRNAs that are regulated by AR. We observed that CAGE TCs and ARBSs accumulate in the vicinity of the miR-125b-2 cluster on chromosome 21. All miRNAs in the miR-125b-2 locus were upregulated by androgen. This result is consistent with a recent study in which let-7c was found to be overexpressed in PCa tissues with high Gleason score (Leite et al., 2009). The global mapping of AR-mediated gene regulation will further reveal novel short RNAs involved in the pathophysiology of PCa. Whether the adjacent ARBSs directly regulate the miRNA expression remains to be studied by performing such as chromosome conformation capture experiments.
Hierarchical cluster analysis of prostate tissue samples from a published microarray database and a comparative study of the Oncomine database revealed that the androgen-dependent overexpression of AR target genes and the increased activity of target gene promoters that is mediated by androgen are well-related to the signature of PCa. In contrast, a large cluster of androgen-downregulated genes is associated with a feature of normal prostate tissues.
It is also notable that partner genes involved in PCa-related chimeric transcripts were validated as direct AR targets in our study. We have previously shown that FOXP1 is a transcription factor that is also regulated by ARBSs (Takayama et al., 2008). Like the well-known TMPRSS2–ETV1 transcript in PCa, the upregulation of FOXP1–ETV1 transcript was reported in PCa cells (Hermans et al., 2008). Other androgen-regulated genes such as ACSL3 (Attard et al., 2008) and NDRG1 (Pflueger et al., 2009) were also identified as putative AR targets genes in this study (Table 2). In PCa, ACSL3 is fused to ETV1, whereas NDRG1 is fused to ERG. Further characterization will discover new partner genes for fusions from our database.
In summary, our integrated study is a powerful tool for elucidating the diversity of the AR-regulated transcriptional network, which alters the expression of various types of target genes throughout the genome, especially those present in the intergenic and antisense regions of RefSeq genes. The strand-specific transcriptome data combined with AR binding information will provide a basic pipeline for the screening of novel AR target genes and expand the global information available on the genome-wide data of direct AR-mediated transcriptional regulation; it will also offer new insights into unknown nongenomic regulations of genes and identify cancer-regulated biomarkers of unknown mechanisms. ‘Oncomining’ of the genome-wide landscape database will identify new classes of molecular targets that can be exploited for improving the diagnosis and treatment of PCa.
Materials and methods
Cell culture and reagents
Androgen-sensitive human PCa LNCaP cells (ATCC, Manassas, VA, USA) were maintained in RPMI medium supplemented with 10% fetal bovine serum, 50 U/ml penicillin and 50 μg/ml streptomycin. Before androgen treatment, cells were cultured in phenol red-free medium containing 5% charcoal-stripped fetal bovine serum for 48–72 h. Antibodies for AR, AcH3 and SRC1 and RNA Pol II were used for ChIP assay as described previously (Takayama et al., 2007).
CAGE libraries derived from total RNAs of LNCaP cells before (0 h) or after 6- or 24-h treatment with R1881 (10 nM) were generated as previously described (Maeda et al., 2008). Briefly, first-strand cDNAs, which were transcribed to the 5′-end of capped RNAs, were attached to CAGE ‘bar code’ tags (AATAG for 0 h, ATTAT for 6 h and ATTGG for 24 h), and digested into 20-nt tags by MmeI (New England Biolabs, Ipswich, MA, USA). The CAGE tags were concatenated, ligated to sequencer-specific DNA adaptors and analyzed by 454 FLX Sequencing (Roche Diagnostics, Penzberg, Germany). The positions of the CAGE tags on the human genome (NCBI version 35) were determined by using Vmatch alignment tool (by Professor Stefan Kurtz, University of Hamburg, Germany), and 1 929 416, 682 111 and 586 083 CAGE tags were determined from 0-, 6- and 24-h libraries, respectively. CAGE tags are grouped into TCs in which the member tags map to the same strand of a chromosome and overlap by at least 1 bp. Fisher's exact test was performed to compare tag distributions between the 0-, 6-, and 24-h samples; 2 × 2 contingency matrix constructed tag numbers in a TC against tag number in the other TCs by a sample pair. We determined whether there were significant differences (P<0.01) among tag distributions for each TC at 6 and 24 h after androgen treatment (total number of TCs=13 110). Also using the Bonferroni correction (P<0.01/n; where n is total TCs, 381 554), we identified 1572 androgen-regulated TCs. For calculating genomic locations of TCs relative to the closest Refseq genes or Refseq on the antisense strand, we defined the location of TC that has maximal frequency at the TSS regions. All data will be released from the website of the Genome Network Platform at National Institute of Genetics, Japan (http://genomenetwork.nig.ac.jp/download/dataset_e.html).
Total RNAs were extracted from LNCaP cells treated with R1881 (10 nM) for 6 and 24 h, using ISOGEN reagent (Nippon Gene, Tokyo, Japan). Affymetrix (Santa clara, CA, USA) U133Plus2.0 expression microarrays were used. Data analysis was performed using the Affymetrix Microarray Suite software. For comparing arrays, normalization was performed using data from all probe sets.
ChIP-chip analysis was performed in biological triplicates by using antibodies for AR and AcH3 as previously described (Takayama et al., 2007). Briefly, in vitro transcription was performed twice for amplification of ChIP materials. Amplified DNA was fragmented, labeled with biotin and hybridized to Human Tiling 1.0R microarrays (7 Chip Set) (Affymetrix). Data were analyzed as described previously. Using a stringent P-value cutoff 1e-5, 2800 AR ChIP-enriched regions were identified as significant ARBSs. Using a stringent P-value cutoff 1e-4, 25 954 AcH3 ChIP-enriched regions were identified as significant AcH3 sites. For comparison study of our data with previous AR binding data, we retrieved the ChIP-chip (Wang et al., 2009) and ChIP-Seq (Yu et al., 2010) data and mapped all the ARBS data to the human genome hg17 using liftOver tool in the UCSC Genome Bioinformatics (http://genome.ucsc.edu/).
ChIP and quantitative PCR (qPCR)
ChIP was performed as previously described (Takayama et al., 2007). Fold enrichments relative to immunoglobulin G or the input control were quantified by real-time PCR using SYBR green PCR master mix (Applied Biosystems, Foster City, CA, USA) and the ABI Prism 7000 system (Applied Biosystems) based on SYBR green É fluorescence. Relative differences in the amounts of PCR products among the treatment groups were evaluated by the comparative cycle threshold (Ct) method, using glyceraldehyde-3-phosphate dehydrogenase as an internal control. Primer sequences for ARBSs are listed in the Supplementary Table 1.
Quantitative reverse transcription–PCR (qRT–PCR)
First-strand cDNA was synthesized using the Primescript RT reagent kit (TAKARA, Kyoto, Japan). Primer sequences are listed in the Supplementary Table 1. qRT–PCR for miRNA was performed in triplicates by TaqMan miRNA RT–PCR system (Ambion, Austin, TX, USA), using RUN6B (Applied Biosystems) as an internal control.
Hierarchical cluster analysis for androgen-regulated genes in prostate samples
Unsupervised hierarchical clustering was performed for 508 genes selected by our experiments (258 androgen-upregulated and 250 androgen-downregulated genes). cDNA microarray data for 62 primary tumors and 41 normal samples (Lapointe et al., 2004) were retrieved from the Stanford University website (http://microarray-pubs.stanford.edu/prostateCA) and both androgen-upregulated and downregulated genes were clustered in a manner dependent on the gene expression features of normal or tumor predominancy, using a hierarchiral clustering by Cluster3.0 software (developed by Eisen et al., 1998, extended by de Hoon et al., 2004), based on the pairwise average-linkage method.
Pathway analysis was performed by Database for Annotation, Visualization and Integrated Discovery Bioinformatics Resources 6.7 (Dennis et al., 2003; Huang et al., 2009), using REACTOME (http://www.reactome.org/) and Kyoto Encyclopedia of Genes and Genomes (http://www.genome.jp/kegg/) databases as annotation sources.
Attard G, Clark J, Ambroisine L, Mills IG, Fisher G, Flohr P et al. (2008). Heterogeneity and clinical significance of ETV1 translocations in human prostate cancer. Br J Cancer 99: 314–320.
Balint I, Muller A, Nagy A, Kovacs G . (2004). Cloning and characterisation of the RBCC728/TRIM36 zinc-binding protein from the tumor suppressor gene region at chromosome 5q22.3. Gene 332: 45–50.
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.
Carninci P, Kasukawa T, Katayama S, Gough J, Frith MC, Maeda N et al. (2005). The transcriptional landscape of the mammalian genome. Science 309: 1559–1563.
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.
de Hoon MJ, Imoto S, Nolan J, Miyano S . (2004). Open source clustering software. Bioinformatics 20: 1453–1454.
Debes JD, Tindall DJ . (2004). Mechanisms of androgen-refractory prostate cancer. N Engl J Med 351: 1488–1490.
Dehm SM, Tindall DJ . (2006). Molecular regulation of androgen action in prostate cancer. J Cell Biochem 99: 333–344.
Dennis Jr G, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC et al. (2003). DAVID: database for annotation, visualization, and integrated discovery. Genome Biol 4: R3.
Eisen MB, Spellman PT, Brown PO, Botstein D . (1998). Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 95: 14863–14868.
FANTOM Consortium (2009). The transcriptional network that controls growth arrest and differentiation in a human myeloid leukemia cell line. Nat Genet 41: 553–562.
Hermans KG, van der Korput HA, van Marion R, van de Wijngaart DJ, Ziel-van der Made A, Dits NF et al. (2008). Truncated ETV1, fused to novel tissue-specific genes, and full-length ETV1 in prostate cancer. Cancer Res 68: 7541–7549.
Ho LL, Kench JG, Handelsman DJ, Scheffer GL, Stricker PD, Grygiel JG et al. (2008). Androgen regulation of multidrug resistance-associated protein 4 (MRP4/ABCC4) in prostate cancer. Prostate 68: 1421–1429.
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.
Huang DW, Sherman BT, Lempicki RA . (2009). Systematic and integrative analysis of large gene lists using DAVID Bioinformatics Resources. Nat Protoc 4: 44–57.
Katayama S, Tomaru Y, Kasukawa T, Waki K, Nakanishi M, Nakamura M et al. (2005). Antisense transcription in the mammalian transcriptome. Science 309: 1564–1566.
Katayama S, Kanamori-Katayama M, Yamaguchi K, Carninci P, Hayashizaki Y . (2007). CAGE-TSSchip: promoter-based expression profiling using the 5′-leading label of capped transcripts. Genome Biol 8: R42.
Kim TH, Barrera LO, Zheng M, Qu C, Singer MA, Richmond TA et al. (2005). A high-resolution map of active promoters in the human genome. Nature 436: 876–880.
Lapointe J, Li C, Higgins JP, van de Rijn M, Bair E, Montgomery K et al. (2004). Gene expression profiling identifies clinically relevant subtypes of prostate cancer. Proc Natl Acad Sci USA 101: 811–816.
Leite KR, Sousa-Canavez JM, Reis ST, Tomiyama AH, Camara-Lopes LH, Sanudo A et al. (2009). Change in expression of miR-let7c, miR-100, and miR-218 from high grade localized prostate cancer to metastasis. Urol Oncol (e-pub ahead of print).
Liu P, Ramachandran S, Ali Seyed M, Scharer CD, Laycock N, Dalton WB et al. (2006). Sex-determining region Y box 4 is a transforming oncogene in human prostate cancer cells. Cancer Res 66: 4011–4019.
Louro R, Nakaya HI, Amaral PP, Festa F, Sogayar MC, da Silva AM et al. (2007). Androgen responsive intronic non-coding RNAs. BMC Biol 5: 4.
Luo J, Zha S, Gage WR, Dunn TA, Hicks JL, Bennett CJ et al. (2002). Alpha-methylacyl-CoA racemase: a new molecular marker for prostate cancer. Cancer Res 62: 2220–2226.
Maeda N, Nishiyori H, Nakamura M, Kawazu C, Murata M, Sano H et al. (2008). Development of a DNA barcode tagging method for monitoring dynamic changes in gene expression by using an ultra high-throughput sequencer. Biotechniques 45: 95–97.
Massie CE, Adryan B, Barbosa-Morais NL, Lynch AG, Tran MG, Neal DE et al. (2007). New androgen receptor genomic targets show an interaction with the ETS1 transcription factor. EMBO Rep 8: 871–878.
Pandini G, Genua M, Frasca F, Vigneri R, Belfiore A . (2009). Sex steroids upregulate the IGF-1R in prostate cancer cells through a nongenotropic pathway. Ann NY Acad Sci 1155: 263–267.
Pflueger D, Rickman DS, Sboner A, Perner S, LaFargue CJ, Svensson MA et al. (2009). N-myc downstream regulated gene 1 (NDRG1) is fused to ERG in prostate cancer. Neoplasia 11: 804–811.
Prescott JL, Blok L, Tindall DJ . (1998). Isolation and androgen regulation of the human homeobox cDNA, NKX3.1. Prostate 35: 71–80.
Ribas J, Ni X, Haffner M, Wentzel EA, Salmasi AH, Chowdhury WH et al. (2009). miR-21: an androgen receptor-regulated microRNA that promotes hormone-dependent and hormone-independent prostate cancer growth. Cancer Res 69: 7165–7169.
Santagata S, Demichelis F, Riva A, Varambally S, Hofer MD, Kutok JL et al. (2004). JAGGED1 expression is associated with prostate cancer metastasis and recurrence. Cancer Res 64: 6854–6857.
Shang Y, Myers M, Brown M . (2002). Formation of the androgen receptor transcription complex. Mol Cell 9: 601–610.
Shi XB, Xue L, Yang J, Ma AH, Zhao J, Xu M et al. (2007). An androgen-regulated miRNA suppresses Bak1 expression and induces androgen-independent growth of prostate cancer cells. Proc Natl Acad Sci USA 104: 19983–19988.
Shiraki T, Kondo S, Katayama S, Waki K, Kasukawa T, Kawaji H et al. (2003). Cap analysis gene expression for high-throughput analysis of transcriptional starting point and identification of promoter usage. Proc Natl Acad Sci USA 100: 15776–15781.
Sun T, Wang Q, Balk S, Brown M, Lee GS, Kantoff P . (2009). The role of microRNA-221 and microRNA-222 in androgen-independent prostate cancer cell lines. Cancer Res 69: 3356–3363.
Suzuki H, Ueda T, Ichikawa T, Ito H . (2003). Androgen receptor involvement in the progression of prostate cancer. Endocr Relat Cancer 10: 209–216.
Takayama K, Horie-Inoue K, Ikeda K, Urano T, Murakami K, Hayashizaki Y et al. (2008). FOXP1 is an androgen-responsive transcription factor that negatively regulates androgen receptor signaling in prostate cancer cells. Biochem Biophys Res Commun 374: 388–393.
Takayama K, Kaneshiro K, Tsutsumi S, Horie-Inoue K, Ikeda K, Urano T et al. (2007). Identification of novel androgen response genes in prostate cancer cells by coupling chromatin immunoprecipitation and genomic microarray analysis. Oncogene 26: 4453–4463.
Takayama K, Tsutsumi S, Suzuki T, Horie-Inoue K, Ikeda K, Kaneshiro K et al. (2009). Amyloid precursor protein is a primary androgen target gene that promotes prostate cancer growth. Cancer Res 69: 137–142.
Velasco AM, Gillis KA, Li Y, Brown EL, Sadler TM, Achilleos M et al. (2004). Identification and validation of novel androgen-regulated genes in prostate cancer. Endocrinology 145: 3913–3924.
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.
Wang Q, Li W, Liu XS, Carroll JS, Janne OA, Keeton EK et al. (2007). A hierarchical network of transcription factors governs androgen receptor-dependent prostate cancer growth. Mol Cell 27: 380–392.
Wang Q, Li W, Zhang Y, Yuan X, Xu K, Yu J et al. (2009). Androgen receptor regulates a distinct transcription program in androgen-independent prostate cancer. Cell 138: 245–256.
Wilson BJ, Giguère V . (2007). Identification of novel pathway partners of p68 and p72 RNA helicases through Oncomine meta-analysis. BMC Genomics 8: 419.
Yochum GS, Cleland R, McWeeney S, Goodman RH . (2007). An antisense transcript induced by Wnt/beta-catenin signaling decreases E2F4. J Biol Chem 282: 871–878.
Yu J, Yu J, Mani RS, Cao Q, Brenner CJ, Cao X et al. (2010). An integrated network of androgen receptor, polycomb, and TMPRSS2-ERG gene fusions in prostate cancer progression. Cancer Cell 17: 443–454.
Yu SY, Chan DW, Liu VW, Ngan HY . (2009). Inhibition of cervical cancer cell growth through activation of upstream kinases of AMP-activated protein kinase. Tumour Biol 30: 80–85.
This study was supported by the Genome Network Project, Cell Innovation Program and Support Project of Strategic Research Center in Private Universities from the MEXT, grants from the Japan Society for the Promotion of Science, grants-in-aid from the MHLW and the Program for Promotion of Fundamental Studies in Health Sciences of the NIBIO. We are grateful to T Murata for helpful discussion and Hiromi Sano and Kazumi Yamaguchi for their assistance.
The authors declare no conflict of interest.
Supplementary Information accompanies the paper on the Oncogene website
About this article
Cite this article
Takayama, K., Tsutsumi, S., Katayama, S. et al. Integration of cap analysis of gene expression and chromatin immunoprecipitation analysis on array reveals genome-wide androgen receptor signaling in prostate cancer cells. Oncogene 30, 619–630 (2011). https://doi.org/10.1038/onc.2010.436
- androgen receptor
- prostate cancer
Breast Disease (2019)
The evolution of long noncoding RNA acceptance in prostate cancer initiation, progression, and its clinical utility in disease management
European Urology (2019)
Endocrine-Related Cancer (2019)
Cancer Letters (2019)
Expression profiles of mRNA and long noncoding RNA in the ovaries of letrozole-induced polycystic ovary syndrome rat model through deep sequencing