Research Resource: Interplay between the Genomic and Transcriptional Networks of Androgen Receptor and Estrogen Receptor α in Luminal Breast Cancer Cells
Materials and Methods
Cell lines and culture, quantitative PCR (QPCR), microarray, and immunoblot analysis
ZR-75–1 cells (American Type Culture Collection, Manassas, VA) were maintained in RPMI 1640 medium (Life Technologies, New South Wales, Australia) containing 10% fetal bovine serum (Sigma-Aldrich, New South Wales, Australia). Cells, plated for 48 h in six-well dishes in phenol red-free RPMI 1640 containing 10% hormone-stripped fetal calf serum (1.5 × 10 cells per well; Sigma-Aldrich) were treated for 16 h with 10 nm E2 or 10 nm DHT alone or in combination, or equivalent vehicle (ethanol). RNA was extracted using an RNeasy kit with DNA digestion (Qiagen, Victoria, Australia). Microarray analysis was performed on quadruplicate samples randomly distributed to Illumina HumanWG-6v3 chips by the Australian Genome Research Facility (St. Lucia, Australia). Raw transcript expression data were exported from IlluminaBeadStudio software and analyzed using the BioconductorLimma package implemented in R (28). Briefly, array data were normalized using variance stabilization normalization (29), corrected with ComBat (30), filtered to likely expressed transcripts (∼24,000), and subjected to linear model fitting. Regulation compared with vehicle was accepted for an empirical Bayes moderated t statistic incorporating Benjamini-Hochberg correction of equal to or less than 0.05. The effect of E2 + DHT cotreatment on transcripts only regulated by E2 or DHT was determined using a two-tailed Student's t test with significance accepted at P ≤ 0.05. Validation by QPCR was performed on replicate sets of RNA independently derived in each of two laboratories. Immunoblot analysis was performed using AR-N20, PGR-H190, ERα-HC20, cathepsin D (CTSD)-H75, and FKBP5-H100 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), calnexin (Thermo Scientific, Victoria, Australia), and antitubulin α (Millipore, Victoria, Australia) antisera as previously described (31). All primer sequences used are provided in Supplemental Table 1 published on The Endocrine Society's Journals website at http://mend.endojournals.org.
Chromatin immunoprecipitation (ChIP) and ChIP-sequencing
ChIP was performed as previously described (31) with the following modifications. ZR-75–1 cells (1.2 × 10) were plated for 72 h in phenol red-free RPMI1640 containing 5% hormone-stripped fetal calf serum and treated for 4 h with 10 nm DHT or 45min with 10 nm E2. Immunoprecipitation was performed with AR-N20, ERα-HC20, or normal rabbit IgG antisera (all from Santa Cruz Biotechnology), and collected on protein G sepharose beads (GE Healthcare, New South Wales, Australia) blocked with yeast tRNA (Sigma-Aldrich). RNase digestion at 37 C for 1.5 h with 50 U RNase per sample was performed before proteinase K digestion. Enrichment analysis was performed by QPCR as previously described (31). For sequencing, DNA from four independent ChIP experiments after treatment with E2 (ERαChIP) or DHT (AR ChIP), each validated by QPCR for enrichment at the known ERα binding site in progesterone receptor (PGR) (19, 32) and AR binding site in FKBP51 (1), respectively, were pooled and concentrated by ethanol precipitation. Each sample and input pooled for each experiment was subjected to 36 bp read DNA sequencing on a Solexa Genome Analyzer II by the Australian Genome Research Facility.
Peak calling and analyses
Sequence output in repeat-masked eland-extended format (UCSC hg18 human genome) was generated using Illumina Genome Analyzer GAPipeline 1.4 and trimmed to uniquely mapped sequence reads. Genomic regions with a peak height of 3 (minimum of three independent reads per site) were recorded using FindPeaks4 (Vancouver Short Read Analysis Package; http://vancouvershortr.sourceforge.net/) (33). Subsequent analysis was performed in R using custom algorithms. To study SR cross talk in a single breast cancer cell line, ad hoc threshold adjustment recommended for different antibodies/conditions (34) was used to select equivalent numbers of ERα and AR peaks for further processing. Briefly, an integer peak height (PH) threshold was used to reduce the binding sites for ERα and AR to a minimum of 10,000. The resulting 15,188 AR (PH ≥ 7) and 13,101 ERα (PH ≥ 5) peaks were selected subject to having 1) no overlap with any of the 274,378 input peaks (PH ≥ 3); 2) <10% sequence overlap with input and a peak ratio > 2; and 3) any overlap with input and a peak ratio more than 4. This yielded 10,130 and 7835 putative AR and ERα binding sites, respectively. Manipulation of intervals for analyzing overlaps between different ChIP-seq datasets was performed in R or using Galaxy (35). Conservation of binding sites among vertebrates was performed using the Cistrome Analysis Pipeline (http://cistrome.dfci.harvard.edu/ap).
Gene ontology (GO) analysis and motif searches
GO enrichment in gene sets was determined in R/Bioconductor by testing the hypergeometric distribution for a greater than expected number of regulated genes per category compared with chance alone. A P value cut-off for category enrichment of P < 0.005 was applied. The presence of known transcription factor motifs from the MatBase database (release 8.3) was predicted for AR and ERα peaks using RegionMiner (Genomatix; http://www.genomatix.de/en/index.html), scored as enrichment in each set over genomic background. Only matrices with continuity-corrected Z scores equal to or greater than 10 were presented. Known motifs in the JASPAR CORE vertebrata database were scanned with CisGenome (36, 37). Fold enrichment and significance (Fisher's exact test) were estimated compared with an equal number of 1-kb control regions with matched physical distribution (36). Only matrices with P ≤ 0.05 and enrichment values equal to or greater than 1.15 are reported. Identification of de novo sequence motifs in AR- and ERα-binding sites was performed using the Gibbs Motif Sampling approach (37), with enrichment and significance calculated as above.
DATA
Included in the supplemental material are normalized/filtered microarray expression data (.csv file, Supplemental Data 1), processed and filtered sets of AR and ERα binding sites (.csv files, Supplemental Data 2–3; and AR.bed, ERα .bed), and R/Bioconductor code used for analysis/graphing bundled with specific input files called by that analysis (Supplemental Data 4). Raw and normalized primary expression data and QSEQ sequencing files are available online at the Gene Expression Omnibus ({"type":"entrez-geo","attrs":{"text":"GSE38132","term_id":"38132"}}GSE38132).
Cell lines and culture, quantitative PCR (QPCR), microarray, and immunoblot analysis
ZR-75–1 cells (American Type Culture Collection, Manassas, VA) were maintained in RPMI 1640 medium (Life Technologies, New South Wales, Australia) containing 10% fetal bovine serum (Sigma-Aldrich, New South Wales, Australia). Cells, plated for 48 h in six-well dishes in phenol red-free RPMI 1640 containing 10% hormone-stripped fetal calf serum (1.5 × 10 cells per well; Sigma-Aldrich) were treated for 16 h with 10 nm E2 or 10 nm DHT alone or in combination, or equivalent vehicle (ethanol). RNA was extracted using an RNeasy kit with DNA digestion (Qiagen, Victoria, Australia). Microarray analysis was performed on quadruplicate samples randomly distributed to Illumina HumanWG-6v3 chips by the Australian Genome Research Facility (St. Lucia, Australia). Raw transcript expression data were exported from IlluminaBeadStudio software and analyzed using the BioconductorLimma package implemented in R (28). Briefly, array data were normalized using variance stabilization normalization (29), corrected with ComBat (30), filtered to likely expressed transcripts (∼24,000), and subjected to linear model fitting. Regulation compared with vehicle was accepted for an empirical Bayes moderated t statistic incorporating Benjamini-Hochberg correction of equal to or less than 0.05. The effect of E2 + DHT cotreatment on transcripts only regulated by E2 or DHT was determined using a two-tailed Student's t test with significance accepted at P ≤ 0.05. Validation by QPCR was performed on replicate sets of RNA independently derived in each of two laboratories. Immunoblot analysis was performed using AR-N20, PGR-H190, ERα-HC20, cathepsin D (CTSD)-H75, and FKBP5-H100 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), calnexin (Thermo Scientific, Victoria, Australia), and antitubulin α (Millipore, Victoria, Australia) antisera as previously described (31). All primer sequences used are provided in Supplemental Table 1 published on The Endocrine Society's Journals website at http://mend.endojournals.org.
Chromatin immunoprecipitation (ChIP) and ChIP-sequencing
ChIP was performed as previously described (31) with the following modifications. ZR-75–1 cells (1.2 × 10) were plated for 72 h in phenol red-free RPMI1640 containing 5% hormone-stripped fetal calf serum and treated for 4 h with 10 nm DHT or 45min with 10 nm E2. Immunoprecipitation was performed with AR-N20, ERα-HC20, or normal rabbit IgG antisera (all from Santa Cruz Biotechnology), and collected on protein G sepharose beads (GE Healthcare, New South Wales, Australia) blocked with yeast tRNA (Sigma-Aldrich). RNase digestion at 37 C for 1.5 h with 50 U RNase per sample was performed before proteinase K digestion. Enrichment analysis was performed by QPCR as previously described (31). For sequencing, DNA from four independent ChIP experiments after treatment with E2 (ERαChIP) or DHT (AR ChIP), each validated by QPCR for enrichment at the known ERα binding site in progesterone receptor (PGR) (19, 32) and AR binding site in FKBP51 (1), respectively, were pooled and concentrated by ethanol precipitation. Each sample and input pooled for each experiment was subjected to 36 bp read DNA sequencing on a Solexa Genome Analyzer II by the Australian Genome Research Facility.
Peak calling and analyses
Sequence output in repeat-masked eland-extended format (UCSC hg18 human genome) was generated using Illumina Genome Analyzer GAPipeline 1.4 and trimmed to uniquely mapped sequence reads. Genomic regions with a peak height of 3 (minimum of three independent reads per site) were recorded using FindPeaks4 (Vancouver Short Read Analysis Package; http://vancouvershortr.sourceforge.net/) (33). Subsequent analysis was performed in R using custom algorithms. To study SR cross talk in a single breast cancer cell line, ad hoc threshold adjustment recommended for different antibodies/conditions (34) was used to select equivalent numbers of ERα and AR peaks for further processing. Briefly, an integer peak height (PH) threshold was used to reduce the binding sites for ERα and AR to a minimum of 10,000. The resulting 15,188 AR (PH ≥ 7) and 13,101 ERα (PH ≥ 5) peaks were selected subject to having 1) no overlap with any of the 274,378 input peaks (PH ≥ 3); 2) <10% sequence overlap with input and a peak ratio > 2; and 3) any overlap with input and a peak ratio more than 4. This yielded 10,130 and 7835 putative AR and ERα binding sites, respectively. Manipulation of intervals for analyzing overlaps between different ChIP-seq datasets was performed in R or using Galaxy (35). Conservation of binding sites among vertebrates was performed using the Cistrome Analysis Pipeline (http://cistrome.dfci.harvard.edu/ap).
Gene ontology (GO) analysis and motif searches
GO enrichment in gene sets was determined in R/Bioconductor by testing the hypergeometric distribution for a greater than expected number of regulated genes per category compared with chance alone. A P value cut-off for category enrichment of P < 0.005 was applied. The presence of known transcription factor motifs from the MatBase database (release 8.3) was predicted for AR and ERα peaks using RegionMiner (Genomatix; http://www.genomatix.de/en/index.html), scored as enrichment in each set over genomic background. Only matrices with continuity-corrected Z scores equal to or greater than 10 were presented. Known motifs in the JASPAR CORE vertebrata database were scanned with CisGenome (36, 37). Fold enrichment and significance (Fisher's exact test) were estimated compared with an equal number of 1-kb control regions with matched physical distribution (36). Only matrices with P ≤ 0.05 and enrichment values equal to or greater than 1.15 are reported. Identification of de novo sequence motifs in AR- and ERα-binding sites was performed using the Gibbs Motif Sampling approach (37), with enrichment and significance calculated as above.
DATA
Included in the supplemental material are normalized/filtered microarray expression data (.csv file, Supplemental Data 1), processed and filtered sets of AR and ERα binding sites (.csv files, Supplemental Data 2–3; and AR.bed, ERα .bed), and R/Bioconductor code used for analysis/graphing bundled with specific input files called by that analysis (Supplemental Data 4). Raw and normalized primary expression data and QSEQ sequencing files are available online at the Gene Expression Omnibus ({"type":"entrez-geo","attrs":{"text":"GSE38132","term_id":"38132"}}GSE38132).
Results
Bidirectional shaping of estrogen- and androgen-responsive transcription networks
To investigate how androgens might affect E2-regulated transcriptional programs in breast cancer, we used the human metastatic ductal epithelial breast carcinoma model, ZR-75–1, which exhibits a relatively more equitable expression of AR and ERα than other breast cancer cell lines (38, 39). Immunoblot analysis of known E2/ERα targets, PGR, and cathepsin D (CTSD), and the AR/DHT target FK506 binding protein 5 (FKBP5), confirmed regulation by these steroids in ZR-75–1 cells, as well as known changes in AR and ERα steady-state protein levels by their cognate ligands (16) (Fig. 1A). Of note, E2 cotreatment had no effect on DHT up-regulation of FKBP5, whereas DHT cotreatment affected E2-induction of PGR but not CTSD (Fig. 1A). QPCR analysis (Supplemental Fig. 1) suggested that the divergent response to DHT cotreatment was the result of gene-specific effects on transcriptional regulation rather than modification of protein stability.
To further characterize the interplay between estrogen and androgen signaling, we performed global Illumina microarray expression profiling of ZR-75–1 cells after treatment with vehicle (ethanol), E2, DHT, or both ligands in combination. Of the 240 E2 and 185 DHT-responsive genes (Benjamini-Hochberg corrected Bayesian moderated t-statistic q < 0.05; Fig. 1B), steroid cotreatment significantly (P < 0.05) affected regulation of 64 E2 (26%) and 29 DHT (15%) targets [Supplemental Tables 2 and 3; Fig. 1, C(i) and D(i)]. For example, up-regulation of E2 target genes RTN1 and PGR was antagonized by DHT, whereas others including GREB1 and CTSD were not significantly affected [Fig. 1C(i)]. Similarly, E2 affected DHT-regulated transcription at certain genes (e.g. CD9 and SCIN) but not others (e.g. AR) [Fig. 1D(i)]. Irrespective of whether genes were up- or down-regulated by E2 or DHT, the effect of cotreatment was generally inhibitory to that regulation [50/64 = 78% of affected E2 targets and 24/29 = 83% of affected DHT targets; Fig. 1, C(i) and D(i)]. QPCR analysis of independently generated RNA samples validated the nature and specificity of this hormonal cross talk (Fig. 1, C(ii) and D(ii)]. Direct comparison of QPCR and microarray results, and additional genes, are presented in Supplemental Fig. 2.
GO analysis identified enrichment of cellular pathways of mammary gland development, cell growth, antiapoptosis, and proliferation in the E2-regulated gene set, and androgen binding and activity, prostate gland development, and a number of metabolic processes in the DHT-regulated gene set. When the E2-regulated set was depleted of the 50/64 DHT-antagonized genes, 15 pathway categories were lost including those involved in the growth, proliferation, and survival of breast cells (e.g. antiapoptosis, proliferation, epithelial cell maturation, phosphoinositide 3-kinase activity; Supplemental Table 4A). Conversely, depletion of the 24/29 E2-antagonized genes from the DHT-regulated set had a similarly profound effect (Supplemental Table 4B). These results demonstrate pathway-level antagonism between androgen and estrogen signaling in this cell line.
A more detailed analysis of selected DHT-antagonized E2 target genes revealed this effect to be dose dependent on both up- and down-regulated E2 target genes, and to occur with 100 fold less steroid (e.g. 0.1 nm DHT in the presence of 10 nm E2) in both ZR-75–1 (data not shown) and another model of luminal breast cancer, T47-D (Fig. 2). DHT activity was verified by analysis of FKBP5 expression, whereas quantitation of CTSD confirmed that the response to cotreatment is restricted to a subset of E2-target genes. We previously demonstrated the central role of AR in this process by showing that DHT-mediated inhibition of PGR regulation by E2 could be reversed with both an AR-specific antagonist (i.e. bicalutamide), and by AR antisense oligonucleotides (20).
Characterization of AR and ERα cistromes in ZR-75–1 cells
To examine the interplay between AR and ERα signaling at the cistromic level, we generated genome-wide binding profiles of each receptor by ChIP-seq in ZR-75–1 cells. DNA pooled from four independently validated ChIP experiments after DHT (AR ChIP) or E2 (ERα ChIP) treatment (Supplemental Fig. 3) was subjected to Illumina high-throughput sequencing. After adjusting for input (see Materials and Methods), 7835 ERα and 10,130 AR binding sites were identified. Well-characterized ERα-binding sites from breast cancer cell lines and AR-binding sites from prostate cancer cell lines were identified, suggesting robustness in ChIP-seq and peak calling. A complete list of sites are available as supplemental material, with examples presented in Supplemental Fig. 4. Using independently generated ChIP samples, we confirmed AR and ERα binding as significantly (P < 0.05) enriched over vehicle at 11 of 12 ERα and eight of 10 AR sites chosen (Fig. 3, A and B). Those analyses included several genomic regions common to both AR- and ERα-binding sites.
ERα- and AR-binding sites were more highly conserved among vertebrates compared with immediate flanking regions (Fig. 3C), suggesting that the identified cistromes contain functional elements. For both SRs, binding sites were predominantly distributed in introns and distal intergenic regions, with only a small proportion found in traditional promoters [<10 kb upstream of a transcriptional start site (TSS)] (Fig. 3D and Supplemental Fig. 4). Nonetheless, both receptors were more highly enriched around TSSs compared with an equivalent number of random genomic regions (Fig. 3E). Although this physical distribution of binding sites in relation to genomic elements is similar to that reported previously for AR or ERα in other cell lines (10, 19, 40, 41), the cistrome composition of both receptors in ZR-75–1 cells is unique. Specifically, only 6.5% of AR sites in ZR-75–1 cells overlap the combined AR cistromes from prostate cancer LNCaP and breast cancer MDA-MB-453 cell lines (Fig. 4A). In contrast, 13.2% of ERα-binding sites are in common with those reported in breast cancer MCF-7 cells (Fig. 4A) (42), which closely equates to the overlap in E2-regulated genes (34 of 245; 13.9%) between these cell lines. Importantly, the overlap with previously reported cistromes increases for both AR and ERα as the peak stringency is raised (Fig. 4B). For AR, the ZR-75–1 cistrome is more similar to that in LNCaP than in LNCaP-abl cells, particularly at higher stringencies, but in both cases the overlap with the top tenth of ZR-75–1 sites is less than 17%. By contrast, 86% of the top tenth of the ERα ZR-75–1 cistrome is common with the ERα MCF-7 cistrome.
Recent studies suggest that pioneer factors such as FOXA1 are critical in binding and activity of AR and ERα and coassociate in close proximity to the receptors on DNA (8). We tested the intersection between AR- and ERα-binding sites and the recently published ZR-75–1 cistrome for FOXA1 (43). Compared with AR, ERα was more likely to colocate with FOXA1, with approximately 20% of all peaks and more than 60% of high-stringency sites showing a direct overlap, and a greater proportion located within 5 kb (Fig. 4C). AR overlap was only 8% for all peaks and about 35% for high-stringency sites, which may reflect a lesser reliance on FOXA1 for AR signaling in these cells compared with apocrine lines, the growth of which is driven by AR (41).
Motif enrichment in AR and ERα cistromes in ZR-75–1 cells
Binding sites were analyzed for enrichment of transcription factor motifs using both de novo scanning and candidate approaches. Using a Gibbs motif sampling approach, the most highly enriched de novo motifs in ERα- and AR-binding sites resembled canonical estrogen and androgen response elements (EREs and AREs), respectively (Fig. 5A). These de novo ERE and ARE motifs were overrepresented, respectively, 2.38-fold (P =1.38) and 1.15-fold (P =3.34) compared with the background genome average, centered within their respective binding sites (Fig. 5B), and associated with peak stringency (Fig. 5C).
Using all AR- and ERα-binding sites, we next scanned for known transcription factor binding motifs in the MatBase and JASPAR CORE vertebrata databases (see Materials and Methods). Supplemental Tables 5 and 6 provide a complete list of enriched motifs. Of the ERα-binding sites, 4934 (63.0%) contained a canonical ERE, and a further 650 (8.3%) contained an ERE half-site (AGGTCA) described as sufficient for ERα binding (19), a distribution that is similar to previous genome-wide binding studies of ERα (5, 32). Motifs recognized by known ERα collaborators, including GATA2, AP-1, AP-2, and ETS and Forkhead (including FOXA1) factors were also enriched in the ERα cistrome (15, 19, 44, 45). Of the AR-binding sites, 6283 (62.0%) contained at least one sequence resembling a canonical ARE, and a further 959 (9.5%) contained ARE half-sites (TGTTCT) that could be targeted by AR (40, 46). Motifs recognized by GATA2, AP-1, FOXA1, NF1, ETS1, and YY1, all of which are known genomic cofactors of AR (1, 10, 41, 46, 47), were enriched in the AR cistrome (Supplemental Table 6, A and B). Notably, AREs were enriched in ERα-binding sites, and EREs were enriched in AR-binding sites (Supplemental Tables 5 and 6), highlighting the potential for cistromic cross talk between these two factors.
We next extracted the subset of AR- and ERα-binding sites lacking AREs/ARE half-sites or EREs/ERE half-sites, respectively, and scanned those for enriched binding motifs that may represent tethering or pioneer factors of the receptors (Supplemental Table 7). Of particular interest, AR binding sites lacking AREs were enriched for the retinoic acid receptor (RAR)-related orphan receptor α motif, which contains the ERE half-site core sequence (AGGTCA), and these sequences were centered within the peaks (data not shown). These data suggest that, in the absence of its cognate response element, AR might target ERE-like motifs in these cells or interact with receptors capable of recognizing these sequences. By contrast, ERα-binding sites lacking EREs did not show differential enrichment of AREs or ARE-like sequences but were enriched for motifs recognized by various FOX factors, including FOXA1 and FOXA2.
Genomic interplay between ERα and AR
We next analyzed the relationship between AR- and ERα-binding events and genes regulated by DHT and E2. Approximately 35% (85 of 245) of the E2 regulated genes have an ERα-binding site within 50 kb of their TSS (Fig. 5D), which is enriched compared with non-E2-regulated genes and comparable to the relationship described in MCF-7 cells (3). Conversely, there was no significant enrichment of AR-binding sites around DHT-regulated compared with non-DHT-regulated genes. This latter result is in contrast to findings in prostate cancer cells (10, 45), possibly reflecting a greater role for longer-range enhancers in AR-driven transcription in breast cells, or less reliance on AR signaling in maintenance of phenotype and/or function. However, AR-binding sites were enriched near E2-regulated genes affected by DHT cotreatment compared with those that were not (Fig. 5E), suggesting that AR may antagonize or cooperate with ER through proximal cis-regulatory elements. By contrast, ERα-binding sites were no more frequent at DHT genes affected by E2 than those unaffected by cotreatment (Fig. 5E).
A comparison of the ZR-75–1 AR and ERα cistromes revealed 984 (9.7–12.6%) 10-kb genomic regions capable of binding both receptors (Fig. 6A, Euler diagram). Within those, there is a strong correlation between the center of the AR- and ERα-binding sites, and the respective AREs and EREs (Fig. 6A, plot), providing the opportunity for antagonism or cooperativity between the two receptors at proximal or shared cis-regulatory elements. To test for such interplay, we analyzed a subset of overlapping binding sites found near genes affected by hormone cotreatment using ChIP and qRT-PCR. At the CD9 locus, ERα binding antagonized the AR-DNA interaction, concomitant with the effect of E2 on DHT induction of CD9 expression (Fig. 6B). For the HAPLN2 locus, ligand cotreatment enhanced ERα, but not AR binding, and resulted in E2-mediated reversal of DHT regulation (Fig. 6C). For THOC5, binding of each receptor was unaffected by cotreatment, but there was a synergistic response in the up-regulation of THOC5 expression (Fig. 6D). These data verify that chromatin interactions between AR and ERα could contribute to the transcriptional interplay between androgen and estrogen signaling in breast cancer cells, but suggest site-specific mechanisms in coregulation of gene expression.
Bidirectional shaping of estrogen- and androgen-responsive transcription networks
To investigate how androgens might affect E2-regulated transcriptional programs in breast cancer, we used the human metastatic ductal epithelial breast carcinoma model, ZR-75–1, which exhibits a relatively more equitable expression of AR and ERα than other breast cancer cell lines (38, 39). Immunoblot analysis of known E2/ERα targets, PGR, and cathepsin D (CTSD), and the AR/DHT target FK506 binding protein 5 (FKBP5), confirmed regulation by these steroids in ZR-75–1 cells, as well as known changes in AR and ERα steady-state protein levels by their cognate ligands (16) (Fig. 1A). Of note, E2 cotreatment had no effect on DHT up-regulation of FKBP5, whereas DHT cotreatment affected E2-induction of PGR but not CTSD (Fig. 1A). QPCR analysis (Supplemental Fig. 1) suggested that the divergent response to DHT cotreatment was the result of gene-specific effects on transcriptional regulation rather than modification of protein stability.
To further characterize the interplay between estrogen and androgen signaling, we performed global Illumina microarray expression profiling of ZR-75–1 cells after treatment with vehicle (ethanol), E2, DHT, or both ligands in combination. Of the 240 E2 and 185 DHT-responsive genes (Benjamini-Hochberg corrected Bayesian moderated t-statistic q < 0.05; Fig. 1B), steroid cotreatment significantly (P < 0.05) affected regulation of 64 E2 (26%) and 29 DHT (15%) targets [Supplemental Tables 2 and 3; Fig. 1, C(i) and D(i)]. For example, up-regulation of E2 target genes RTN1 and PGR was antagonized by DHT, whereas others including GREB1 and CTSD were not significantly affected [Fig. 1C(i)]. Similarly, E2 affected DHT-regulated transcription at certain genes (e.g. CD9 and SCIN) but not others (e.g. AR) [Fig. 1D(i)]. Irrespective of whether genes were up- or down-regulated by E2 or DHT, the effect of cotreatment was generally inhibitory to that regulation [50/64 = 78% of affected E2 targets and 24/29 = 83% of affected DHT targets; Fig. 1, C(i) and D(i)]. QPCR analysis of independently generated RNA samples validated the nature and specificity of this hormonal cross talk (Fig. 1, C(ii) and D(ii)]. Direct comparison of QPCR and microarray results, and additional genes, are presented in Supplemental Fig. 2.
GO analysis identified enrichment of cellular pathways of mammary gland development, cell growth, antiapoptosis, and proliferation in the E2-regulated gene set, and androgen binding and activity, prostate gland development, and a number of metabolic processes in the DHT-regulated gene set. When the E2-regulated set was depleted of the 50/64 DHT-antagonized genes, 15 pathway categories were lost including those involved in the growth, proliferation, and survival of breast cells (e.g. antiapoptosis, proliferation, epithelial cell maturation, phosphoinositide 3-kinase activity; Supplemental Table 4A). Conversely, depletion of the 24/29 E2-antagonized genes from the DHT-regulated set had a similarly profound effect (Supplemental Table 4B). These results demonstrate pathway-level antagonism between androgen and estrogen signaling in this cell line.
A more detailed analysis of selected DHT-antagonized E2 target genes revealed this effect to be dose dependent on both up- and down-regulated E2 target genes, and to occur with 100 fold less steroid (e.g. 0.1 nm DHT in the presence of 10 nm E2) in both ZR-75–1 (data not shown) and another model of luminal breast cancer, T47-D (Fig. 2). DHT activity was verified by analysis of FKBP5 expression, whereas quantitation of CTSD confirmed that the response to cotreatment is restricted to a subset of E2-target genes. We previously demonstrated the central role of AR in this process by showing that DHT-mediated inhibition of PGR regulation by E2 could be reversed with both an AR-specific antagonist (i.e. bicalutamide), and by AR antisense oligonucleotides (20).
Characterization of AR and ERα cistromes in ZR-75–1 cells
To examine the interplay between AR and ERα signaling at the cistromic level, we generated genome-wide binding profiles of each receptor by ChIP-seq in ZR-75–1 cells. DNA pooled from four independently validated ChIP experiments after DHT (AR ChIP) or E2 (ERα ChIP) treatment (Supplemental Fig. 3) was subjected to Illumina high-throughput sequencing. After adjusting for input (see Materials and Methods), 7835 ERα and 10,130 AR binding sites were identified. Well-characterized ERα-binding sites from breast cancer cell lines and AR-binding sites from prostate cancer cell lines were identified, suggesting robustness in ChIP-seq and peak calling. A complete list of sites are available as supplemental material, with examples presented in Supplemental Fig. 4. Using independently generated ChIP samples, we confirmed AR and ERα binding as significantly (P < 0.05) enriched over vehicle at 11 of 12 ERα and eight of 10 AR sites chosen (Fig. 3, A and B). Those analyses included several genomic regions common to both AR- and ERα-binding sites.
ERα- and AR-binding sites were more highly conserved among vertebrates compared with immediate flanking regions (Fig. 3C), suggesting that the identified cistromes contain functional elements. For both SRs, binding sites were predominantly distributed in introns and distal intergenic regions, with only a small proportion found in traditional promoters [<10 kb upstream of a transcriptional start site (TSS)] (Fig. 3D and Supplemental Fig. 4). Nonetheless, both receptors were more highly enriched around TSSs compared with an equivalent number of random genomic regions (Fig. 3E). Although this physical distribution of binding sites in relation to genomic elements is similar to that reported previously for AR or ERα in other cell lines (10, 19, 40, 41), the cistrome composition of both receptors in ZR-75–1 cells is unique. Specifically, only 6.5% of AR sites in ZR-75–1 cells overlap the combined AR cistromes from prostate cancer LNCaP and breast cancer MDA-MB-453 cell lines (Fig. 4A). In contrast, 13.2% of ERα-binding sites are in common with those reported in breast cancer MCF-7 cells (Fig. 4A) (42), which closely equates to the overlap in E2-regulated genes (34 of 245; 13.9%) between these cell lines. Importantly, the overlap with previously reported cistromes increases for both AR and ERα as the peak stringency is raised (Fig. 4B). For AR, the ZR-75–1 cistrome is more similar to that in LNCaP than in LNCaP-abl cells, particularly at higher stringencies, but in both cases the overlap with the top tenth of ZR-75–1 sites is less than 17%. By contrast, 86% of the top tenth of the ERα ZR-75–1 cistrome is common with the ERα MCF-7 cistrome.
Recent studies suggest that pioneer factors such as FOXA1 are critical in binding and activity of AR and ERα and coassociate in close proximity to the receptors on DNA (8). We tested the intersection between AR- and ERα-binding sites and the recently published ZR-75–1 cistrome for FOXA1 (43). Compared with AR, ERα was more likely to colocate with FOXA1, with approximately 20% of all peaks and more than 60% of high-stringency sites showing a direct overlap, and a greater proportion located within 5 kb (Fig. 4C). AR overlap was only 8% for all peaks and about 35% for high-stringency sites, which may reflect a lesser reliance on FOXA1 for AR signaling in these cells compared with apocrine lines, the growth of which is driven by AR (41).
Motif enrichment in AR and ERα cistromes in ZR-75–1 cells
Binding sites were analyzed for enrichment of transcription factor motifs using both de novo scanning and candidate approaches. Using a Gibbs motif sampling approach, the most highly enriched de novo motifs in ERα- and AR-binding sites resembled canonical estrogen and androgen response elements (EREs and AREs), respectively (Fig. 5A). These de novo ERE and ARE motifs were overrepresented, respectively, 2.38-fold (P =1.38) and 1.15-fold (P =3.34) compared with the background genome average, centered within their respective binding sites (Fig. 5B), and associated with peak stringency (Fig. 5C).
Using all AR- and ERα-binding sites, we next scanned for known transcription factor binding motifs in the MatBase and JASPAR CORE vertebrata databases (see Materials and Methods). Supplemental Tables 5 and 6 provide a complete list of enriched motifs. Of the ERα-binding sites, 4934 (63.0%) contained a canonical ERE, and a further 650 (8.3%) contained an ERE half-site (AGGTCA) described as sufficient for ERα binding (19), a distribution that is similar to previous genome-wide binding studies of ERα (5, 32). Motifs recognized by known ERα collaborators, including GATA2, AP-1, AP-2, and ETS and Forkhead (including FOXA1) factors were also enriched in the ERα cistrome (15, 19, 44, 45). Of the AR-binding sites, 6283 (62.0%) contained at least one sequence resembling a canonical ARE, and a further 959 (9.5%) contained ARE half-sites (TGTTCT) that could be targeted by AR (40, 46). Motifs recognized by GATA2, AP-1, FOXA1, NF1, ETS1, and YY1, all of which are known genomic cofactors of AR (1, 10, 41, 46, 47), were enriched in the AR cistrome (Supplemental Table 6, A and B). Notably, AREs were enriched in ERα-binding sites, and EREs were enriched in AR-binding sites (Supplemental Tables 5 and 6), highlighting the potential for cistromic cross talk between these two factors.
We next extracted the subset of AR- and ERα-binding sites lacking AREs/ARE half-sites or EREs/ERE half-sites, respectively, and scanned those for enriched binding motifs that may represent tethering or pioneer factors of the receptors (Supplemental Table 7). Of particular interest, AR binding sites lacking AREs were enriched for the retinoic acid receptor (RAR)-related orphan receptor α motif, which contains the ERE half-site core sequence (AGGTCA), and these sequences were centered within the peaks (data not shown). These data suggest that, in the absence of its cognate response element, AR might target ERE-like motifs in these cells or interact with receptors capable of recognizing these sequences. By contrast, ERα-binding sites lacking EREs did not show differential enrichment of AREs or ARE-like sequences but were enriched for motifs recognized by various FOX factors, including FOXA1 and FOXA2.
Genomic interplay between ERα and AR
We next analyzed the relationship between AR- and ERα-binding events and genes regulated by DHT and E2. Approximately 35% (85 of 245) of the E2 regulated genes have an ERα-binding site within 50 kb of their TSS (Fig. 5D), which is enriched compared with non-E2-regulated genes and comparable to the relationship described in MCF-7 cells (3). Conversely, there was no significant enrichment of AR-binding sites around DHT-regulated compared with non-DHT-regulated genes. This latter result is in contrast to findings in prostate cancer cells (10, 45), possibly reflecting a greater role for longer-range enhancers in AR-driven transcription in breast cells, or less reliance on AR signaling in maintenance of phenotype and/or function. However, AR-binding sites were enriched near E2-regulated genes affected by DHT cotreatment compared with those that were not (Fig. 5E), suggesting that AR may antagonize or cooperate with ER through proximal cis-regulatory elements. By contrast, ERα-binding sites were no more frequent at DHT genes affected by E2 than those unaffected by cotreatment (Fig. 5E).
A comparison of the ZR-75–1 AR and ERα cistromes revealed 984 (9.7–12.6%) 10-kb genomic regions capable of binding both receptors (Fig. 6A, Euler diagram). Within those, there is a strong correlation between the center of the AR- and ERα-binding sites, and the respective AREs and EREs (Fig. 6A, plot), providing the opportunity for antagonism or cooperativity between the two receptors at proximal or shared cis-regulatory elements. To test for such interplay, we analyzed a subset of overlapping binding sites found near genes affected by hormone cotreatment using ChIP and qRT-PCR. At the CD9 locus, ERα binding antagonized the AR-DNA interaction, concomitant with the effect of E2 on DHT induction of CD9 expression (Fig. 6B). For the HAPLN2 locus, ligand cotreatment enhanced ERα, but not AR binding, and resulted in E2-mediated reversal of DHT regulation (Fig. 6C). For THOC5, binding of each receptor was unaffected by cotreatment, but there was a synergistic response in the up-regulation of THOC5 expression (Fig. 6D). These data verify that chromatin interactions between AR and ERα could contribute to the transcriptional interplay between androgen and estrogen signaling in breast cancer cells, but suggest site-specific mechanisms in coregulation of gene expression.
Discussion
Estrogen action through ERα is a well-established mediator of proliferation and differentiation in most breast cancers. In contrast, androgen action through the AR appears to be dichotomous depending on the breast cancer subtype and ERα status. In the MDA-MB-453 model of apocrine ERα-negative disease, AR can stimulate cell growth (16, 48, 49), whereas in luminal ERα-positive breast cancers it has an antiproliferative role (20, 48, 50, 51). Indeed, studies from our laboratory and others have shown that AR expression is a determinant of survival in ERα-positive, but not ERα-negative, breast cancer samples (20, 27). Moreover, we recently found that AR could target classical EREs, suggesting that it may antagonize ERα at the genomic level to effect estrogen-driven growth of breast cancer cells (20). Here, we extend on those findings by utilizing ChIP-seq and expression profiling to analyze the transcriptional cross talk between AR and ERα signaling in a cell line model of luminal disease, ZR-75–1 (52). These analyses are the first to characterize the genomic functions of AR and ERα in a single cell line, revealing extensive interaction between these two SRs and identifying genes and pathways by which AR is likely to exert its protective role.
The transcriptional responses to estrogen and androgen alone in ZR-75–1 cells are highly disparate. Pathway analysis demonstrated that estrogens drive proliferation and inhibition of the cellular apoptotic program, whereas androgens regulate androgen metabolism and differentiation. In cells exposed to both androgens and estrogens, however, many of those key cellular pathways may no longer be effectively regulated. Notably, E2 cotreatment disrupted androgen regulation of pathways involved in testosterone binding and activity, whereas DHT cotreatment antagonized estrogen-driven pathways in mammary gland development, growth stimulation, antiapoptosis, proliferation, and phosphoinositide 3-kinase activity. These findings provide insight into the transcriptional programs underlying androgen and estrogen signaling in the breast and identify genes and pathways by which androgens may inhibit proliferation of luminal breast cancer cells. Furthermore, they demonstrate that steroid receptor interplay is not necessarily additive but can generate cellular responses distinct from the sum of component activities.
Genome-wide binding profiles generated by ChIP coupled to tiling microarrays or high-throughput sequencing is proving crucial in elucidating the normal and pathological functions of nuclear receptors. For ERα and AR in particular, cistrome studies have been used to help unravel regulatory functions that contribute to breast and prostate cancers, respectively (5, 7, 9, 10, 41). To date, all genome-scale studies of AR or ERα binding sites have been performed in isolation in cells of different origins: mostly prostate and breast cancer cell lines expressing high levels of AR or ERα, respectively (14, 19, 40, 42, 47, 53, 54). Of particular relevance, the complement of ERα-binding sites has been shown to be highly dependent on cell lineage (42), which makes extrapolation between cells and receptors challenging. In this study, we have generated and integrated the cistromes of AR and ERα from the same cell line model of breast cancer, ZR-75–1. We anticipate that these data will be used as a reference set for future work aimed at mapping AR- and ERα-binding sites in other cancer cell lines and/or tissues.
Although unique to ZR-75–1 cells, the ERα cistrome was more similar to that of another luminal breast cancer cell line, MCF-7, than the AR cistrome was to those characterized in either apocrine MDA-MB-453 breast cancer or LNCaP prostate cancer cells. In each case, however, the overlap increased significantly when only high-stringency binding sites were considered. Our data are consistent with a model in which a common set of high-affinity SR-binding sites are active in most cells, but lower-affinity lineage-specific binding sites, which rely on pioneer factors, dictate unique cell and tissue responses to those steroids (55). Indeed, both the AR and ERα cistromes overlapped substantially with a published ZR-75–1 genome-wide binding profile of FOXA1, a pioneer transcription factor known to play a critical role in defining lineage-specific SR-binding sites (10, 19, 41). Our data are also consistent with analogous modes of ERα action and response in breast cancer cell lines with the same classification but divergent actions of AR in luminal vs. apocrine breast cancer cells (41).
An emerging paradigm in the field of SR transcription is the notion that different receptors can target the same cis-regulatory elements, which represents another level of complexity in steroid signaling that could mediate integrated, fine-tuned responses as the hormone environment changes. For example, Hua et al. (14) found that RARs bind to a subset of EREs and antagonize ERα signaling in the MCF-7 breast cancer cell line. Moreover, ERα and ERβ compete for many of the same elements in the MCF-7 genome and, in doing so, can be shifted to new sites that are not targeted when either receptor is expressed alone (15). By contrast, glucocorticoid receptor (GR) and an ERα variant engineered to resemble GR in the DNA binding domain can dynamically and noncompetitively access the same RE (56), whereas RAR can enhance ERα transcription by binding to ERα-targeted EREs (57). By comparing genome-wide binding profiles of ERα and AR in ZR-75–1 cells, our study has revealed the potential for cross talk between these receptors at the level of individual cis-regulatory elements. Specifically, we found that 1) ER-binding sites were enriched for AREs and motifs recognized by collaborative coregulators of both SRs; 2) the AR and ERα cistromes share many binding sites at the physical and target gene level; 3) proximal (< 10 kb) AR- and ERα-binding sites and hormone response elements tend to directly overlap; and 4) AR-binding sites are enriched around estrogen-regulated genes that are affected by DHT cotreatment. This led us to hypothesize that recruitment of AR and ERα to common REs might explain the interference between androgen- and estrogen-signaling pathways in ZR-75–1 cells. Indeed, two recent studies found that the AR cistrome in the ERα-negative MDA-MB-453 breast cancer cell line overlaps, to a significant extent, with the ERα cistrome from MCF-7 cells, highlighting the potential for cistromic cross talk (9, 41). To test this hypothesis, we activated both receptors and measured their association with common REs near to genes affected by hormone cotreatment. Although those experiments showed that E2 cotreatment decreased AR binding near the CD9 gene, concomitant with E2 inhibition of DHT regulation, we did not observe any robust instances of direct binding antagonism at any of the other candidate sites tested. AR and ERα might therefore be able to co-occupy chromatin at specific loci in a dynamic, noncompetitive process, but positively or negatively influence transcriptional outcome based on their site-specific relationship with accessory factors and/or the transcription machinery. Supporting this idea of site-specific transcriptional capacity, an ERα engineered with a GR DNA-binding domain can bind to glucocorticoid response elements, but is incapable of remodeling those sites for activation (56).
There are several other plausible mechanisms, unrelated to interference via overlapping cis-elements, by which transcriptional antagonism between AR and ERα might occur. For example, androgen signaling through the AR could indirectly affect the ERα-driven transcriptional program. AR-mediated induction or repression of genes, transcription factors, or micro RNAs that collectively influence ERα targets could have a profound effect on estrogenic output. Alternatively, AR and ERα may bind to independent cis-regulatory elements that regulate the same gene, with distinct transcriptional hubs and gene looping acting to localize disparate loci (6). This mechanism would explain the relatively small co-occurrence of AR- and ERα-binding sites within 50 kB of E2-regulated genes. Finally, AR interference of ERα signaling could be mediated by competition for shared coregulators (17, 18), although that explanation is not consistent with our previous observation that the AR DNA-binding domain is sufficient to inhibit ERα signaling in T-47D cells (20). It is likely that none of these potential mechanisms are mutually exclusive and that all may occur within breast cancer cells within different temporal or spatial contexts.
A precise understanding of steroid receptor cross talk in breast cancer is of significant importance for clinical practice. For example, we show here that estrogen regulation of PGR is decreased in a dose-dependent manner by DHT, which reinforces the association between AR and PGR status reported in ERα-positive disease (58, 59). Indeed, reanalysis of our previously published radioligand binding data (59) revealed a significant inverse correlation between AR and PGR status in clinical breast cancer samples (Supplemental Fig. 5). Critically, PGR expression is used diagnostically as a marker of ERα activity, thereby identifying women as candidates for antiestrogen therapies such as tamoxifen. Our data suggest that if AR and androgen levels were high in a particular cancer then PGR would be correspondingly low. In extreme cases, women who would likely be sensitive to antiestrogen treatment may not be offered tamoxifen in the course of clinical management.
In summary, this study has revealed a dynamic interplay between AR and ERα signaling in breast cancer cells that results in the shaping of a unique transcriptional network. The discovery of specific genes and pathways by which AR is likely to oppose ERα action, and identification of potential mechanisms underlying this antagonism, provides a new perspective on hormone receptor profiles in breast cancer. We envision that this study will serve as a platform for future studies aimed at better understanding the overall response of breast cancer cells to the hormonal milieu.
Supplementary Material
NURSA Molecule Pages†:
Ligands: Dihydrotestosterone17β-estradiol. Bicalutamide
Abstract
The cellular response to circulating sex steroids is more than the sum of individual hormone actions, instead representing an interplay between activities of the evolutionarily related steroid hormone receptors. An example of this interaction is in breast cancer, where the risk of dying from estrogen receptor-α (ERα)-positive disease decreases approximately 4-fold when androgen receptor (AR) expression is high. In this study, we used chromatin immunoprecipitation sequencing (ChIP-seq) and microarray expression profiling to investigate the genomic and transcriptional cross talk between AR and ERα signaling in a luminal breast cancer cell line model, ZR-75–1. Expression profiling demonstrated reciprocal interference between 5α-dihydrotestosterone (DHT)- and 17β-estradiol (E2)-induced transcriptional programs. Specifically, regulation of 26% of E2 and 15% of DHT target genes was significantly affected by cotreatment with the other hormone, in the majority of cases (78–83%) antagonistically. Pathway analysis suggested that DHT cotreatment, for example, depleted E2-regulated pathways in cell survival and proliferation. ChIP-seq identified substantial overlap between the steroid receptor cistromes in ZR-75–1 cells, with 10–13% of AR- and ERα-binding sites located within 10 kb of the other receptor. Enrichment of androgen response elements in ERα-binding sites and vice versa was revealed by motif analysis, and AR-binding sites were enriched about E2-responsive genes affected by DHT cotreatment. Targeted ChIP and expression analysis revealed locus-specific outcomes when AR and ERα bind to the same DNA region. This work provides the first cistrome data for two steroid receptors in the same cell, insight into the antagonistic interplay between estrogens and androgens in luminal breast cancer, and an important resource for future work aimed at evaluating interrelated steroid receptors in different cellular systems.
The development, growth, and homeostatic maintenance of many tissues depends on the action of sex steroids, including estrogens such as 17β-estradiol (E2), and androgens such as testosterone (T) or 5α-dihydrotestosterone (DHT). At a genomic level, the action of these steroids is mediated via binding to, and activation of, two related steroid nuclear receptors (SRs), estrogen receptor-α (ERα) and androgen receptor (AR). In classical models, agonist activation results in binding of these receptors to specific DNA response elements (REs) within enhancers and promoters of target genes and the subsequent regulation of specific transcriptional programs. Recently, however, a more contemporary understanding of SR action has been revealed by genome-wide binding profiles (cistromes) coupled with expression profiling after ligand treatment. Specifically, the extent, diversity, and cell-specific transcriptional activity of each SR is derived from the availability and nature of the bound ligand, interaction with distinct subsets of accessory cofactors, a dynamic relationship with chromatin that depends on lineage-specific chromatin organization/modification, and the presence of other DNA-binding proteins that regulate receptor-DNA interactions (1–8). In addition to yielding information on the fundamental mechanisms by which SRs regulate transcription, those genomic studies have identified many downstream effector genes and pathways and thereby provided insights into the physiological processes of SR action in various cell types (9–12).
Functional interactions between different nuclear receptors provide another level of complexity to the gene programs mediated by these potent transcription factors. For the most part, those interactions appear to be competitive or antagonistic (13–15), which is perhaps not surprising given the common evolutionary origin and similar modes of transcriptional regulation of the nuclear receptor superfamily. There are several ways in which the action of steroid receptors might be competitive, including structural similarity of activating ligands, formation of homo- and heterodimers, sequestration of limiting transcriptional coregulators that modulate chromatin structure and gene transcription, or even targeting of shared REs (16–20). Indeed, many of these mechanisms have been demonstrated in cellular systems. For example, the orphan nuclear receptor DAX-1 competes for a coactivator-binding site on the orphan nuclear receptor, Nur77, with the resultant heterodimer repressing transcription (21). The AR has been shown to inhibit both liver X and vitamin D receptors by competing for coactivator proteins (22, 23), whereas ERβ and the retinoic acid receptors antagonize ERα-driven transcription on a proportion of genomic sites by competition for shared REs (14, 15).
A physiological example of steroid receptor cross talk is suggested by studies of breast cancer, in which androgens oppose the proliferative effects of estrogens on normal and malignant breast cell growth in vitro and in vivo (reviewed in Ref. 24). Indeed, androgens such as fluoxymesterone were successfully used for the hormonal management of metastatic breast cancer until supplanted by the selective ERα modulator, tamoxifen. When used in combination with tamoxifen, however, androgens do not significantly improve outcome over tamoxifen alone (25, 26), which may indicate a common mode of action. Clinically, women with invasive ERα-positive ductal carcinoma have an approximately 4-fold decreased risk of cancer-related death if AR levels are high (20, 27). Finally, we showed in a recent study that AR can bind to a classical ERα RE (ERE), suggesting a potential for direct genomic interplay between these factors (20).
In this study, we used ChIP-seq and microarray expression profiling to investigate the interplay between ERα/estrogen and AR/androgen signaling in the AR-positive/ERα-positive ZR-75–1 cell line, a model of luminal breast cancer. This work revealed extensive cross talk between the androgen- and estrogen-signaling networks, which may be partly mediated at a cistromic level, and that androgens antagonize estrogen progrowth responses. These results provide insight into how AR confers antiestrogenic effects in breast cancer and emphasize how a combination of hormones can elicit a cellular response that is distinct from the sum of their component parts.
Click here to view.Acknowledgments
We thank Dr. Lisa Butler (The University of Adelaide, Australia) and Professors Geoffrey Greene (The University of Chicago) and Jason Carroll (Cambridge Cancer Centre) for their critical evaluation of this manuscript. We also thank Andrew Trotta, Melissa O'Loughlin, Vanessa Thompson, and Phing Lee (all The University of Adelaide) for technical assistance; Illumina Australia and Deniz Kollhofer for bioinformatics support and Annette Benning, Paul Gooding, Artem Men, Kirby Siemering, and Tim Bruxner for facilitating the ChIP-sequencing (all The Australian Genome Research Facility).
This work was supported by the Australian Research Council (Grant DP110101101 to G.B. and W.D.T.), the National Health and Medical Research Council (Grant 1008349 to W.D.T.), and the Prostate Cancer Foundation of Australia (Grant YI0810; to L.S.; Grants YI02; and PG2210 to G.B.). G.B. and L.S. hold Freemasons Foundation Postdoctoral Research Fellowships.
Disclosure Summary: The authors have nothing to disclose.
Notes
NURSA Molecule Pages†:
Ligands: Dihydrotestosterone17β-estradiol. Bicalutamide
Annotations provided by Nuclear Receptor Signaling Atlas (NURSA) Bioinformatics Resource. Molecule Pages can be accessed on the NURSA website at www.nursa.org.
Abbreviations:
- AR
- Androgen receptor
- ARE
- androgen response element
- ChIP
- chromatin immunoprecipitation
- CTSD
- cathepsin D
- DHT
- dihydrotestosterone
- E2
- 17β-estradiol
- ER
- estrogen receptor
- ERE
- estrogen response element
- GR
- glucocorticoid receptor
- PGR
- progesterone receptor
- PH
- peak height
- QPCR
- quantitative PCR
- RAR
- retinoic acid receptor
- RE
- response element
- SR
- steroid nuclear receptor
- TSS
- transcriptional start site.
References
- 1. Jia L, Berman BP, Jariwala U, Yan X, Cogan JP, Walters A, Chen T, Buchanan G, Frenkel B, Coetzee GA. 2008 Genomic androgen receptor-occupied regions with different functions, defined by histone acetylation, coregulators and transcriptional capacity. PLoS ONE3:e3645. [Google Scholar]
- 2. Kim TH, Barrera LO, Zheng M, Qu C, Singer MA, Richmond TA, Wu Y, Green RD, Ren B. 2005 A high-resolution map of active promoters in the human genome. Nature436:876–880 [Google Scholar]
- 3. Carroll JS, Brown M. 2006 Estrogen receptor target gene: an evolving concept. Mol Endocrinol20:1707–1714 [[PubMed][Google Scholar]
- 4. Wang Q, Carroll JS, Brown M. 2005 Spatial and temporal recruitment of androgen receptor and its coactivators involves chromosomal looping and polymerase tracking. Mol Cell19:631–642 [[PubMed][Google Scholar]
- 5. Cicatiello L, Mutarelli M, Grober OM, Paris O, Ferraro L, Ravo M, Tarallo R, Luo S, Schroth GP, Seifert M, Zinser C, Chiusano ML, Traini A, De Bortoli M, Weisz A. 2010 Estrogen receptor α controls a gene network in luminal-like breast cancer cells comprising multiple transcription factors and microRNAs. Am J Pathol176:2113–2130 [Google Scholar]
- 6. Fullwood MJ, Han Y, Wei CL, Ruan X, Ruan Y. 2010 Chromatin interaction analysis using paired-end tag sequencing. Curr Protoc Mol Biol21:Unit 21.15.1–25 [[PubMed][Google Scholar]
- 7. Hurtado A, Holmes KA, Geistlinger TR, Hutcheson IR, Nicholson RI, Brown M, Jiang J, Howat WJ, Ali S, Carroll JS. 2008 Regulation of ERBB2 by oestrogen receptor-PAX2 determines response to tamoxifen. Nature456:663–666 [Google Scholar]
- 8. Jozwik KM, Carroll JS. 2012 Pioneer factors in hormone-dependent cancers. Nat Rev Cancer12:381–385 [[PubMed][Google Scholar]
- 9. Ni M, Chen Y, Lim E, Wimberly H, Bailey ST, Imai Y, Rimm DL, Liu XS, Brown M. 2011 Targeting androgen receptor in estrogen receptor-negative breast cancer. Cancer Cell20:119–131 [Google Scholar]
- 10. Wang Q, Li W, Zhang Y, Yuan X, Xu K, Yu J, Chen Z, Beroukhim R, Wang H, Lupien M, Wu T, Regan MM, Meyer CA, Carroll JS, Manrai AK, Jänne OA, Balk SP, Mehra R, Han B, Chinnaiyan AM, Rubin MA, True L, Fiorentino M, Fiore C, Loda M, Kantoff PW, Liu XS, Brown M. 2009 Androgen receptor regulates a distinct transcription program in androgen-independent prostate cancer. Cell138:245–256 [Google Scholar]
- 11. Welboren WJ, van Driel MA, Janssen-Megens EM, van Heeringen SJ, Sweep FC, Span PN, Stunnenberg HG. 2009 ChIP-Seq of ERα and RNA polymerase II defines genes differentially responding to ligands. EMBO J28:1418–1428 [Google Scholar]
- 12. Wyce A, Bai Y, Nagpal S, Thompson CC. 2010 Research resource: the androgen receptor modulates expression of genes with critical roles in muscle development and function. Mol Endocrinol24:1665–1674 [Google Scholar]
- 13. Bonofiglio D, Gabriele S, Aquila S, Catalano S, Gentile M, Middea E, Giordano F, Andò S. 2005 Estrogen receptor α binds to peroxisome proliferator-activated receptor response element and negatively interferes with peroxisome proliferator-activated receptor γ signaling in breast cancer cells. Clin Cancer Res11:6139–6147 [[PubMed][Google Scholar]
- 14. Hua S, Kittler R, White KP. 2009 Genomic antagonism between retinoic acid and estrogen signaling in breast cancer. Cell137:1259–1271 [Google Scholar]
- 15. Charn TH, Liu ET, Chang EC, Lee YK, Katzenellenbogen JA, Katzenellenbogen BS. 2010 Genome-wide dynamics of chromatin binding of estrogen receptors α and β: mutual restriction and competitive site selection. Mol Endocrinol24:47–59 [Google Scholar]
- 16. Poulin R, Simard J, Labrie C, Petitclerc L, Dumont M, Lagace L, Labrie F. 1989 Down-regulation of estrogen receptors by androgens in the ZR-75-1 human breast cancer cell line. Endocrinology125:392–399 [[PubMed][Google Scholar]
- 17. Lanzino M, De Amicis F, McPhaul MJ, Marsico S, Panno ML, Andò S. 2005 Endogenous coactivator ARA70 interacts with estrogen receptor α (ERα) and modulates the functional ERα/androgen receptor interplay in MCF-7 cells. J Biol Chem280:20421–20430 [[PubMed][Google Scholar]
- 18. Migliaccio A, Di Domenico M, Castoria G, Nanayakkara M, Lombardi M, de Falco A, Bilancio A, Varricchio L, Ciociola A, Auricchio F. 2005 Steroid receptor regulation of epidermal growth factor signaling through Src in breast and prostate cancer cells: steroid antagonist action. Cancer Res65:10585–10593 [[PubMed][Google Scholar]
- 19. Carroll JS, Liu XS, Brodsky AS, Li W, Meyer CA, Szary AJ, Eeckhoute J, Shao W, Hestermann EV, Geistlinger TR, Fox EA, Silver PA, Brown M. 2005 Chromosome-wide mapping of estrogen receptor binding reveals long-range regulation requiring the forkhead protein FoxA1. Cell122:33–43 [[PubMed][Google Scholar]
- 20. Peters AA, Buchanan G, Ricciardelli C, Bianco-Miotto T, Centenera MM, Harris JM, Jindal S, Segara D, Jia L, Moore NL, Henshall SM, Birrell SN, Coetzee GA, Sutherland RL, Butler LM, Tilley WD. 2009 Androgen receptor inhibits estrogen receptor-α activity and is prognostic in breast cancer. Cancer Res69:6131–6140 [[PubMed][Google Scholar]
- 21. Song KH, Park YY, Park KC, Hong CY, Park JH, Shong M, Lee K, Choi HS. 2004 The atypical orphan nuclear receptor DAX-1 interacts with orphan nuclear receptor Nur77 and represses its transactivation. Mol Endocrinol18:1929–1940 [[PubMed][Google Scholar]
- 22. Ting HJ, Bao BY, Hsu CL, Lee YF. 2005 Androgen-receptor coregulators mediate the suppressive effect of androgen signals on vitamin D receptor activity. Endocrine26:1–9 [[PubMed][Google Scholar]
- 23. Krycer JR, Brown AJ. 2011 Cross-talk between the androgen receptor and the liver X receptor: implications for cholesterol homeostasis. J Biol Chem286:20637–20647 [Google Scholar]
- 24. Dimitrakakis C, Bondy C. 2009 Androgens and the breast. Breast Cancer Res11:212. [Google Scholar]
- 25. Kennedy BJ. 1958 Fluoxymesterone therapy in advanced breast cancer. N Engl J Med259:673–675 [[PubMed][Google Scholar]
- 26. Ingle JN, Suman VJ, Mailliard JA, Kugler JW, Krook JE, Michalak JC, Pisansky TM, Wold LE, Donohue JH, Goetz MP, Perez EA. 2006 Randomized trial of tamoxifen alone or combined with fluoxymesterone as adjuvant therapy in postmenopausal women with resected estrogen receptor positive breast cancer. North Central Cancer Treatment Group Trial 89-30-52. Breast Cancer Res Treat98:217–222 [[PubMed][Google Scholar]
- 27. Hu R, Dawood S, Holmes MD, Collins LC, Schnitt SJ, Cole K, Marotti JD, Hankinson SE, Colditz GA, Tamimi RM. 2011 Androgen receptor expression and breast cancer survival in postmenopausal women. Clin Cancer Res17:1867–1874 [Google Scholar]
- 28. Smyth G, Gentleman R, Carey V, Dudoit S, Irizarry R, Huber W. 2005 Limma: linear models for microarray data. In: Bioinformatics and Computational Biology Solutions using R and Bioconductor. New York: Springer; 397–420 [PubMed][Google Scholar]
- 29. Huber W, von Heydebreck A, Sultmann H, Poustka A, Vingron M. 2002 Variance stabilization applied to microarray data calibration and to the quantification of differential expression. Bioinformatics 18 Suppl1:S96–S104 [[PubMed][Google Scholar]
- 30. Johnson WE, Li C, Rabinovic A. 2007 Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics8:118–127 [[PubMed][Google Scholar]
- 31. Need EF, Scher HI, Peters AA, Moore NL, Cheong A, Ryan CJ, Wittert GA, Marshall VR, Tilley WD, Buchanan G. 2009 A novel androgen receptor amino terminal region reveals two classes of amino/carboxyl interaction-deficient variants with divergent capacity to activate responsive sites in chromatin. Endocrinology150:2674–2682 [Google Scholar]
- 32. Carroll JS, Meyer CA, Song J, Li W, Geistlinger TR, Eeckhoute J, Brodsky AS, Keeton EK, Fertuck KC, Hall GF, Wang Q, Bekiranov S, Sementchenko V, Fox EA, Silver PA, Gingeras TR, Liu XS, Brown M. 2006 Genome-wide analysis of estrogen receptor binding sites. Nat Genet38:1289–1297 [[PubMed][Google Scholar]
- 33. Fejes AP, Robertson G, Bilenky M, Varhol R, Bainbridge M, Jones SJ. 2008 FindPeaks 3.1: a tool for identifying areas of enrichment from massively parallel short-read sequencing technology. Bioinformatics24:1729–1730 [Google Scholar]
- 34. Hawkins RD, Hon GC, Ren B. 2010 Next-generation genomics: an integrative approach. Nat Rev Genet11:476–486 [Google Scholar]
- 35. Goecks J, Nekrutenko A, Taylor J. 2010 Galaxy: a comprehensive approach for supporting accessible, reproducible, and transparent computational research in the life sciences. Genome Biol11:R86. [Google Scholar]
- 36. Ji H, Jiang H, Ma W, Johnson DS, Myers RM, Wong WH. 2008 An integrated software system for analyzing ChIP-chip and ChIP-seq data. Nat Biotechnol26:1293–1300 [Google Scholar]
- 37. Bryne JC, Valen E, Tang MH, Marstrand T, Winther O, da Piedade I, Krogh A, Lenhard B, Sandelin A. 2008 JASPAR, the open access database of transcription factor-binding profiles: new content and tools in the 2008 update. Nucleic Acids Res36:D102–D106 [Google Scholar]
- 38. Engel LW, Young NA, Tralka TS, Lippman ME, O'Brien SJ, Joyce MJ. 1978 Establishment and characterization of three new continuous cell lines derived from human breast carcinomas. Cancer Res38:3352–3364 [[PubMed][Google Scholar]
- 39. Zhu Y, Wang A, Liu MC, Zwart A, Lee RY, Gallagher A, Wang Y, Miller WR, Dixon JM, Clarke R. 2006 Estrogen receptor α positive breast tumors and breast cancer cell lines share similarities in their transcriptome data structures. Int J Oncol29:1581–1589 [[PubMed][Google Scholar]
- 40. Wang Q, Li W, Liu XS, Carroll JS, Jänne OA, Keeton EK, Chinnaiyan AM, Pienta KJ, Brown M. 2007 A hierarchical network of transcription factors governs androgen receptor-dependent prostate cancer growth. Mol Cell27:380–392 [Google Scholar]
- 41. Robinson JL, Macarthur S, Ross-Innes CS, Tilley WD, Neal DE, Mills IG, Carroll JS. 2011 Androgen receptor driven transcription in molecular apocrine breast cancer is mediated by FoxA1. EMBO J30:3019–3027 [Google Scholar]
- 42. Krum SA, Miranda-Carboni GA, Lupien M, Eeckhoute J, Carroll JS, Brown M. 2008 Unique ERα cistromes control cell type-specific gene regulation. Mol Endocrinol22:2393–2406 [Google Scholar]
- 43. Hurtado A, Holmes KA, Ross-Innes CS, Schmidt D, Carroll JS. 2011 FOXA1 is a key determinant of estrogen receptor function and endocrine response. Nat Genet43:27–33 [Google Scholar]
- 44. Davis DL, Burch JB. 1996 The chicken vitellogenin II gene is flanked by a GATA factor-dependent estrogen response unit. Mol Endocrinol10:937–944 [[PubMed][Google Scholar]
- 45. Lin CY, Vega VB, Thomsen JS, Zhang T, Kong SL, Xie M, Chiu KP, Lipovich L, Barnett DH, Stossi F, Yeo A, George J, Kuznetsov VA, Lee YK, Charn TH, Palanisamy N, Miller LD, Cheung E, Katzenellenbogen BS, Ruan Y, Bourque G, Wei CL, Liu ET. 2007 Whole-genome cartography of estrogen receptor alpha binding sites. PLoS Genet3:e87. [Google Scholar]
- 46. Massie CE, Adryan B, Barbosa-Morais NL, Lynch AG, Tran MG, Neal DE, Mills IG. 2007 New androgen receptor genomic targets show an interaction with the ETS1 transcription factor. EMBO Rep17:17 [Google Scholar]
- 47. Bolton EC, So AY, Chaivorapol C, Haqq CM, Li H, Yamamoto KR. 2007 Cell- and gene-specific regulation of primary target genes by the androgen receptor. Genes Dev21:2005–2017 [Google Scholar]
- 48. Birrell SN, Bentel JM, Hickey TE, Ricciardelli C, Weger MA, Horsfall DJ, Tilley WD. 1995 Androgens induce divergent proliferative responses in human breast cancer cell lines. J Steroid Biochem Mol Biol52:459–467 [[PubMed][Google Scholar]
- 49. Doane AS, Danso M, Lal P, Donaton M, Zhang L, Hudis C, Gerald WL. 2006 An estrogen receptor-negative breast cancer subset characterized by a hormonally regulated transcriptional program and response to androgen. Oncogene25:3994–4008 [[PubMed][Google Scholar]
- 50. Macedo LF, Guo Z, Tilghman SL, Sabnis GJ, Qiu Y, Brodie A. 2006 Role of androgens on MCF-7 breast cancer cell growth and on the inhibitory effect of letrozole. Cancer Res66:7775–7782 [[PubMed][Google Scholar]
- 51. Søreide JA, Lea OA, Varhaug JE, Skarstein A, Kvinnsland S. 1992 Androgen receptors in operable breast cancer: relation to other steroid hormone receptors, correlations to prognostic factors and predictive value for effect of adjuvant tamoxifen treatment. Eur J Surg Oncol18:112–118 [[PubMed][Google Scholar]
- 52. Mandal S, Davie JR. 2007 An integrated analysis of genes and pathways exhibiting metabolic differences between estrogen receptor positive breast cancer cells. BMC Cancer7:181. [Google Scholar]
- 53. Takayama K, Kaneshiro K, Tsutsumi S, Horie-Inoue K, Ikeda K, Urano T, Ijichi N, Ouchi Y, Shirahige K, Aburatani H, Inoue S. 2007 Identification of novel androgen response genes in prostate cancer cells by coupling chromatin immunoprecipitation and genomic microarray analysis. Oncogene26:4453–4463 [[PubMed][Google Scholar]
- 54. Valouev A, Johnson DS, Sundquist A, Medina C, Anton E, Batzoglou S, Myers RM, Sidow A. 2008 Genome-wide analysis of transcription factor binding sites based on ChIP-Seq data. Nat Methods5:829–834 [Google Scholar]
- 55. Magnani L, Eeckhoute J, Lupien M. 2011 Pioneer factors: directing transcriptional regulators within the chromatin environment. Trends Genet27:465–474 [[PubMed][Google Scholar]
- 56. Voss TC, Schiltz RL, Sung MH, Yen PM, Stamatoyannopoulos JA, Biddie SC, Johnson TA, Miranda TB, John S, Hager GL. 2011 Dynamic Exchange at regulatory elements during chromatin remodeling underlies assisted loading mechanism. Cell146:544–554 [Google Scholar]
- 57. Ross-Innes CS, Stark R, Holmes KA, Schmidt D, Spyrou C, Russell R, Massie CE, Vowler SL, Eldridge M, Carroll JS. 2010 Cooperative interaction between retinoic acid receptor-α and estrogen receptor in breast cancer. Genes Dev24:171–182 [Google Scholar]
- 58. Agoff SN, Swanson PE, Linden H, Hawes SE, Lawton TJ. 2003 Androgen receptor expression in estrogen receptor-negative breast cancer. Immunohistochemical, clinical, and prognostic associations. Am J Clin Pathol120:725–731 [[PubMed][Google Scholar]
- 59. Buchanan G, Birrell SN, Peters AA, Bianco-Miotto T, Ramsay K, Cops EJ, Yang M, Harris JM, Simila HA, Moore NL, Bentel JM, Ricciardelli C, Horsfall DJ, Butler LM, Tilley WD. 2005 Decreased androgen receptor levels and receptor function in breast cancer contribute to the failure of response to medroxyprogesterone acetate. Cancer Res65:8487–8496 [[PubMed][Google Scholar]
- 60. Crooks GE, Hon G, Chandonia JM, Brenner SE. 2004 WebLogo: a sequence logo generator. Genome Res14:1188–1190 [Google Scholar]