The AR (androgen receptor) is known to influence the expression of its target genes by binding to different sets of AREs (androgen-response elements) in the DNA. One set consists of the classical steroid-response elements which are partial palindromic repeats of the 5′-TGTTCT-3′ steroid-receptor monomer-binding element. The second set contains motifs that are AR-specific and that are proposed to be partial direct repeats of the same motif. On the basis of this assumption, we used an in silico approach to identify new androgen-selective AREs in the regulatory regions of known androgen-responsive genes. We have used an extension of the NUBIScan algorithm to screen a collection of 85 known human androgen-responsive genes compiled from literature and database searches. We report the evaluation of the most promising hits resulting from this computational search by in vitro DNA-binding assays using full-size ARs and GRs (glucocorticoid receptors) as well as their isolated DBDs (DNA-binding domains). We also describe the ability of some of these motifs to confer androgen-, but not glucocorticoid-, responsiveness to reporter-gene expression. The elements found in the aquaporin-5 and the Rad9 (radiation-sensitive 9) genes showed selective AR versus GR binding in band-shift assays and a strong activity and selectivity in functional assays, both as isolated elements and in their original contexts. Our data indicate the validity of the hypothesis that selective AREs are recognizable as direct 5′-TGTTCT-3′ repeats, and extend the list of currently known selective elements.
- androgen receptor (AR)
- androgen-response element (ARE)
- aquaporin-5 (AQP5)
- radiation-sensitive 9 (Rad9)
Being a ligand-induced transcription factor, the AR (androgen receptor) influences gene expression by binding specific DNA motifs called AREs (androgen-response elements) in the regulatory regions of its target genes. The AR is known to interact functionally with two different sets of AREs . One set is that of the cAREs (classical steroid-hormone-response elements), which are three-nucleotide-spaced partial-palindromic repeats of the AR monomer consensus binding site 5′-TGTTCT-3′. The cAREs are also recognized by the other steroid-hormone receptors that belong to the same subgroup of the nuclear receptor superfamily as does AR .
The AR is able to interact with another set of motifs that are essentially three-nucleotide-spaced partial direct repeats of the same monomer-binding element. In that respect they resemble the binding elements for the VDR (vitamin D receptor) and the RAR/RXR (retinoic acid receptor/retinoid X receptor)-type of nuclear receptors . These motifs are called sAREs (selective AREs). They have so far been characterized in the rat probasin promoter [4,5], the human secretory component gene upstream enhancer , the mouse sex-limited protein enhancer [7–9], the mouse Rhox5 (reproductive homeobox 5) promoter  and the human selective androgen-responsive gene . In each of these genes, the sARE was identified as necessary to confer selective androgen versus glucocorticoid stimulation of gene expression to a heterologous promoter, since its deletion or mutation results in a strong decrease in androgen-responsiveness. These elements are also sufficient to mediate selectivity of androgen versus glucocorticoid stimulation to reporter-gene expression when taken out of the contexts of their original enhancers and tested individually in transient transfection assays [4–11].
Sequence logos illustrating the conservation of bases in cognate sARE and cARE motifs, created using the WebLogo technique [12,13] are shown in Figure 1. In a previous study we demonstrated the in vivo relevance of at least one sARE (the Rhox5 ARE 1) in the so-called SPARKI (specificity-affecting AR knockin) mouse model in which the wild-type AR was exchanged for a mutant receptor that had lost its ability to interact with sAREs . We demonstrated that Rhox5 expression in mutant mice was more than 10-fold lower compared with that in wild-type littermates, whereas the expression levels of other known androgen-responsive genes, driven by cARE motifs, were unchanged.
In the present paper we describe the application of a combined computational and experimental approach to discover previously unknown sAREs in the promoter regions of androgen-responsive genes. Using an extension of the NUBIScan algorithm , we examined the promoter region of a collection of genes that had previously been described to respond to activation of the AR.
After computational analysis, we selected seven motifs from seven genes, which, according to both the algorithm and visual inspection, are the best candidate androgen-selective AREs. A first assessment of whether or not these motifs are functional sAREs was done by evaluating their roles in androgen regulation of reporter-gene expression in transient transfection experiments using Luc (luciferase) promoter constructs containing four copies of each element. Next we performed in vitro DNA-binding assays using isolated DBD fragments of either AR or GR (glucocorticoid receptor) and comparing their affinities for each element. Three elements that in both tests displayed significant androgen-selectivity were tested in band-shift assays using the full-size receptors. Finally, the functionality of the Rad9 (radiation-sensitive 9) and ABCC1 (ATP-binding cassette subfamily C member 1) motif was evaluated in the original genomic contexts by performing transfection assays using reporter constructs containing the enhancer region encompassing the ARE. We were able to confirm the androgen-specificity of the AQP5 (aquaporin-5) ARE by assessing AQP5 expression in the SPARKI mouse model, which expresses an AR mutant that is not able to bind sAREs.
We have thus succeeded in identifying and characterizing two new functional androgen-selective response elements and enhancers in two human genes that were previously known to respond to androgens but so far had lacked more data on the DNA regions or motifs that are responsible for their androgen regulation.
Luc reporter constructs are all derived from the pGL3 basic reporter vector (Promega, Madison, WI, U.S.A.). The empty promoter vector pTK-TATA-Luc construct was described previously  (TK refers to thymidine kinase and TATA to the the consensus sequence TATAA/TAA/T). The empty pE1b-TATA-Luc vector was made by excising, by PstI and BglII digestion, the two copies of the TAT-GRE (tyrosine aminotransferase glucocorticoid-response element) from the (ARE)2-E1b-Luc vector, which was described in . Four copies of each ARE were inserted in pTK-TATA-Luc by cloning the corresponding oligonucleotides, having SacI- and HindIII-compatible overhangs, in the SacI-digested vector. Oligonucleotides having NheI- and XhoI-compatible overhangs were inserted in the NheI-digested pTK-TATA-Luc vector. A genomic fragment containing the Rad9 putative ARE was amplified by PCR from genomic DNA from HeLa cells using the following primers (5′-ccccggtaccgctggtgacgaggggagcag-3′ and 5′-ccccctcgagtggcctctcaaagtggagtgag-3′). For the enhancer fragment encompassing the ABCC1 motif, in a similar PCR reaction, the primers (5′-ccccggtacccagtgcctggctcttgaataatg-3′) and (5′-ccccctcgagggaggaagaggctttcaaactga-3′) were used. The resulting PCR products were inserted in the pE1b-TATA-Luc vector. All constructs were checked by sequencing.
The AR and GR mammalian expression vectors were kindly provided by Professor Alessandro (‘Sandro’) Rusconi (Institute of Biochemistry, University of Fribourg, Fribourg, Switzerland).
HeLa cells were plated on day 1 in 96-well microtitre plates in Dulbecco's modified Eagle's medium containing 1000 mg/ml glucose, penicillin (100 i.u./ml), streptomycin (100 μg/ml) and 5% dextran-coated charcoal-stripped fetal-calf serum. On day 2 the cells were transfected with a DNA mixture of 100 ng of reporter vector, 10 ng of an expression plasmid for β-galactosidase and 1 ng of the appropriate AR or GR expression vector per well. Transfections were performed using the Genejuice® transfection reagent (Novagen, Merck, Darmstadt, Germany) according to the manufacturer's instructions. On day 3, the medium was replaced with fresh medium, with or without the addition of R1881 (a synthetic androgen methyltrienolone; 1 nM) or dexamethasone (10 nM), as appropriate. At 24 h after the addition of the hormone, cells were harvested in 25 μl of Passive Lysis Buffer (Promega) per well. Luc light emission of 2 μl samples of each well were determined using the Luc assay reagent (Promega) in a Luminoskan Ascent Luminometer (Thermo Labsystems, Franklin, MA, U.S.A.). To correct for transfection efficiency, Luc values were always related to the β-galactosidase value as measured by the Galacto-Light Chemiluminscent Reporter Gene Assay System (Tropix, Bedford, MA, U.S.A.). Results are expressed as corrected Luc values relative to the Luc value of the non-stimulated samples (which was arbitrarily set at 1). These values therefore also reflect fold induction as a ratio of the value for the stimulated samples versus the non-stimulated samples transfected with the same reporter construct. Experiments were performed in triplicate and repeated at least three times independently.
Purification of receptor DBDs and full-size receptors
AR-DBD and GR-DBD fragments were grown and purified as described previously . Full-size receptors for use in the band-shift assays were obtained by transfecting 15-cm-diameter plates of African-green-monkey kidney COS7 cells with 7 μg of the appropriate AR or GR expression plasmid (see above). After 36 h and an additional 12 h stimulation of the cells with the appropriate hormone (either 1 nM R1881 or 10 nM dexamethasone), the cells were harvested as described previously .
In vitro DNA-binding assays
For the calculation of the Ks (apparent dissociation constant) value for each motif–receptor–DBD combination, an equal amount of radiolabelled oligonucleotide was incubated with increasing amounts (from 8 nM to 1.6 μM) of AR-DBD or GR-DBD in 20 μl of binding buffer as described in . After non-denaturing PAGE, gels were dried and scanned in a STORM 840 PhosphorImager (GE Healthcare Bio-sciences, Uppsala, Sweden). The percentage of binding was calculated for each lane by calculating the ratio of the intensity of the bound probe versus the total amount of radioactivity in that lane. Binding curves were constructed by plotting percentages of binding as a function of the concentration of AR-DBD or GR-DBD that was used. For each curve, best fits to four-parameter curves with allosteric Hill kinetics were calculated using the SigmaPlot software package (SPSS Inc., Chicago, IL, U.S.A.) and Ks values were determined accordingly.
Band-shift assays detecting the binding of the full-size receptors were performed as described above, except that the radiolabelled probe was incubated with approx. 5 μg of total extract of COS7 cells transiently transfected with the appropriate AR or GR mammalian expression plasmid. The double-stranded oligonucleotides used as specific or non-specific competitors for AR or GR interaction had the following sequences: C3(1) ARE, 5′-agcttacatagtacgtgatgttctcaagctcga-3′; and NF1 (nuclear factor 1), 5′-attttggctacaagccaatatgat-3′. Anti-AR is an in-house antibody against the first 21 amino acids of the AR, and anti-GR antibody was kindly given by Dr Ann-Charlotte Wikström, Karolinska Institutet, Stockholm, Sweden.
Gene array analysis
Total RNA was extracted from ventral prostate glands of 11-week-old wild-type or SPARKI mice using the RNeasy Mini Kit (Qiagen, Hilden, Germany) with an additional on-column DNase treatment. RNA quality was checked in a NanoDrop Technologies (Centerville, DE, U.S.A.) spectrophotometer and in an Agilent (Palo Alto, CA, U.S.A.) 2100 BioAnalyzer. Total RNA from ventral prostate glands of three wild-type or three SPARKI littermates were pooled, creating paired control and SPARKI samples. These pools were used for hybridization with the Agilent Whole Mouse Genome 4×44K Dual-Mode Microarray. Two biological repeats were done in the dual-label mode, creating four datasets. Normalized transcript signals in each sample were compared with background signals and scored either present or absent. A Student's t test was performed to filter all transcripts that were significantly (P<0.01) different in SPARKI versus wild-type. Transcripts that were more than two-fold up- or down-regulated in SPARKI were regarded as differentially expressed. Overall, 364 signals were significantly down-regulated, and 344 signals were up-regulated in SPARKI prostate glands.
RESULTS AND DISCUSSION
Selection of androgen-responsive genes
Androgen-regulated genes were compiled from a VEAT (Virtual Expression Analysis Tool) comparison of two datasets of expression profiles of androgen-starved and androgen-stimulated LNCaP (human prostate adenocarcinoma) cells in the PEDB (Prostate Expressed DataBase) at http://www.pedb.org/AR/microarray/ , as well as from two other comparative microarray analyses of androgen-starved and androgen-stimulated LNCaP cells [20,21]. This resulted in the compilation of a set of 85 known androgen-regulated genes in the promoter regions of which we searched for putative androgen-selective AREs. Using the SOURCE website , LocusLink (now Entrez Gene) identifiers were retrieved for all but nine of the genes in the set. Using these identifiers, the first exon of each gene and 5 kb of the 5′ flanking region were retrieved from TRASER (Transcript Sequence Retreiver; http://genome-www6.stanford.edu/cgi-bin/Traser/traser) and saved for further analysis.
Analysis of promoter regions
We have previously published NUBIScan, an algorithm for the sensitive detection of nuclear-receptor binding sites . In the present study we have used a slightly modified version of the NUBIScan algorithm that allows the definition of different positional weight matrices for each of the two hexamer half-sites.
For each of the half-sites, we created positional weight matrices. Initially, a nucleotide distribution matrix was calculated on the basis of the aligned known positive binding-site sequences. Next, an arbitrarily scaled positional weight factor was calculated on the basis of the degree of conservation at each position of the alignment. Because of this term, highly conserved positions contribute more to a site's scores than variable positions. Next, promoter sequences were scanned with these matrices, obtaining match quality scores at each sequence position. Finally, the scores from the relevant matrices, representative of the individual half-sites, were combined by multiplication for a final score.
A major challenge in this approach was that at the time of design of the study, only three confirmed naturally occurring and two synthetic selective AREs were described [23,24], and, therefore, our matrices had to be composed using this small training set. We have used leave-one-out cross-validation to investigate the sensitivity of our algorithm. Here, one of the ARE elements is chosen as the validation data point and the algorithm is trained with all other ARE elements. Next, the validation data point is classified using the algorithm. This procedure is then repeated with all data points and allows a rough estimate of the algorithm's predictive quality, in spite of the small sample size. A preliminary study, furthermore, indicated that the algorithm was able to discriminate known positive sequences from randomized sequence and thus to identify binding sites within large sequence contexts. For each predicted binding site in the analysed candidate promoter regions, expectation values were estimated by performing the same analysis on 100 randomly shuffled sequences of the same length and base composition. We are aware, however, that such a small and consistent training set may miss some of the less-well-conserved binding sites. Notably, elements organized as direct versus inverted repeats are not easy to distinguish, as individual binding sites are degenerate, and share sequence overlap. For this reason, some of the sites detected by the algorithm may serve as binding sites in either of the two configurations.
For 76 of the candidate AR-responsive genes collected from the literature, 5 kb of the 5′ flanking region sequence and the first exon were retrieved. All of these sequences were subjected to computational analysis, a search being made for the direct repeat-type arrangement of sAREs. After this search, motifs from seven genes from the top-ranked predictions by NUBIScan, namely ABCC1, AQP5, INPP5A (inositol polyphosphate 5-phosphatase), SREBF2 (sterol-regulatory-element-binding transcription factor 2), Rad9A (Rad homologue 9A), THRA (thyroid hormone receptor α) and UCK2 (uridine–cytidine kinase 2) (Table 1), were chosen for an initial analysis. The binding sites in this set all had a raw score >0.75 and an expectation value of <1×10−5. Next, we set out to verify experimentally the ability of these elements to be activated by AR and, more specifically, to ascertain their specificity towards AR.
Functional assays using the isolated ARE motifs
Four copies of each putative ARE were cloned in a Luc reporter vector upstream of the TK minimal promoter. Theoretically they would thereby confer steroid-responsiveness to Luc expression in transient transfection experiments in cells co-transfected with the appropriate receptor.
The SREBF2, INPP5A, THRA and UCK2 motifs did not confer androgen or glucocorticoid stimulation to the minimal promoter (Figure 2). Because of their intrinsic inability to confer androgen-responsiveness, these motifs were excluded from subsequent experiments.
The ABCC1 motif displayed strong responsiveness to androgens (a 20-fold induction), but no selectivity towards glucocorticoids (20-fold induction). The slpHRE2 [an ARE in the upstream region of the slp (sex-limited protein) gene] control, Rad9 and AQP5 motifs display a clear androgen-inducibility (9.7-, 14.1- and 14-fold induction respectively), whereas glucocorticoid responsiveness was significantly lower (4.2-, 1.2- and 4.4-fold induction respectively). These motifs are therefore able to selectively confer androgen- as opposed to glucocorticoid-responsiveness to a heterologous promoter.
Band-shift assays using isolated DBDs
Oligonucleotides carrying the Rad9, AQP5 and ABCC1 motifs were radiolabelled and incubated with increasing amounts of recombinant AR- or GR-DBD and subjected to non-denaturing PAGE (Figure 3). From the binding patterns for each element, binding curves were obtained and apparent dissociation constant (Ks) values were calculated as has been described, for example, for slpHRE2 .
The Rad9 and AQP5 motifs displayed the most striking AR- versus GR-DBD selectivity: their dissociation constants (Ks) for AR-DBD interaction were 124 and 112 nM respectively, while the Ks values for their interaction with the GR-DBD could not be calculated (>1000 nM). This correlates well with their functional androgen- versus glucocorticoid-selectivity in transfection assays.
The ABCC1 motif displayed a relatively high affinity for the AR-DBD (Ks value is 65 nM) and approximately 6-fold lower affinity for the GR-DBD (Ks 405 nM).
Confirmation of selective receptor interaction with the Rad9, ABCC1 and AQP5 motifs
To confirm the results obtained in the two previous experiments, we performed band-shift experiments using the Rad9, ABCC1 and AQP5 motifs and full-size AR and GR (Figure 4). As a positive control, band-shift experiments were performed using the non-selective C3(1)ARE  and the selective slpHRE2 . All four motifs were bound by the AR, as demonstrated by the supershifted band on the addition of an anti-AR antibody and the disappearance of this supershifted band on the addition of an excess of an unlabelled specific competitor oligonucleotide [C3(1)ARE], but not of a non-specific competitor. When using extracts of cells not transfected with either receptor, no supershifts appear on the addition of anti-AR or anti-GR antibody. Only the C3(1)ARE, but none of the putative selective AREs, were significantly recognized by the GR, thereby confirming the binding assays using the isolated DBDs as well as the transfection assays. For ABCC1, a very faint supershift was visible when using cell extracts containing GR. Apparently the GR is a stronger transcriptional activator than is the AR, as demonstrated, for example, for the non-specific TAT-GRE and C3(1)ARE. Despite having similar affinities for both DBDs, the fold induction that these elements and promoter regions confer upon glucocorticoid stimulation is always much higher compared with their androgen effects [23,24]. It is therefore not surprising that the ABCC1 motif displays low androgen- versus glucocorticoid-specificity in transient transfections. On the other hand, the affinity of the GR-DBD for the Rad9 and AQP5 motifs is probably below the threshold for the GR to function as a transcriptional activator, which makes them genuine androgen-specific AREs.
Functional confirmation of androgen-selectivity of the AREs in their original contexts
We subsequently cloned the enhancer regions, or ARUs (androgen-regulatory units), that surround the RAD9 and ABCC1 motifs in the E1b minimal promoter-driven Luc reporter vector and performed transient transfection experiments in HeLa cells (Figure 5). The Rad9 ARU shows a clear responsiveness to androgens (a 5.7-fold induction), whereas it does not respond to glucocorticoids (1.1-fold induction).
We also determined the expression levels of the AQP5 gene in the previously developed SPARKI mouse model containing a mutated AR that is able to interact with cAREs but not with sAREs . In a comparative microarray analysis on ventral prostate, the mouse AQP5 gene was found to be approx. 3.6-fold down-regulated in the SPARKI mice compared with the wild-type, confirming that its expression is strongly dependent on an AR able to interact with (a) selective ARE(s). This difference in AQP5 expression level is not due to a generally lower abundance of transcripts in mutant versus wild-type mice or a general lower activity of the androgen signalling pathway, since other known androgen-responsive genes (for example NKx3.1) display equal expression levels in the SPARKI versus wild-type mice.
The AQP5 ARE located approx. 570 bp downstream of the AQP5 transcripition initiation site is well conserved between human (5′-tgttcgcagagttct-3′) and mouse (5′-tgttcgcagagttcc-3′). The T-to-C change at position +7 (underlined) most probably does not significantly affect AR interaction.
The RAD9 motif, however, is not conserved between human and mouse. Although it is located in a region with 60% sequence identity between human (at +/−1800 bp upstream of the transcription initiation site) and mouse (at +/−3800 bp upstream of the promoter), the element is changed (5′-ggctctggggctggg-3′; underlined nucleotides are different from the human motif), so that it no longer corresponds to a consensus cARE or sARE.
Androgen regulation of Rad9 and AQP5
Rad9 was initially found in fission yeast as a gene involved in resistance to ionizing radiation . Apart from its function as a checkpoint-control protein, it also plays a more direct role in DNA repair. As part of a complex with Rad1 and HUS1 (hydroxyurea-sensitive 1; also termed ‘9-1-1 complex’), it is thought to act as a DNA damage sensor . Next to its functions in cell-cycle control and DNA damage repair, it was also found to play a role in apoptosis and have dual functions in transcription. On the one hand, it acts as a transcriptional activator to stimulate the expression of genes that could be essential in cell cycle regulation or DNA damage repair [27–29]. On the other hand, it has been proposed to be a co-repressor of ligand-activated AR . Rad9 could therefore theoretically play an important role in the onset and evolution of prostate cancer, protecting prostate epithelial cells from the influence of androgens on their proliferation.
AQP5, one of 13 aquaporin transmembrane fluid channels, is expressed, among others, in salivary and lacrimal glands, sweat glands and lung and uterus epithelium. It functions as a channel through which fluids can flow from the inside of the cell to the external environment . AQP5-knockout mice show lower rates of saliva production, with a higher osmolality and viscosity of saliva . In the progesterone-primed uterus, its expression can be significantly up-regulated by oestrogens and a functional oestrogen-response element was characterized in its proximal promoter . In mouse lung epithelial cells its expression can be regulated at the transcriptional level by all-trans-retinoic acid . A conserved transcriptional enhancer region was identified in the first intron of the AQP5 gene . The motif that we have discovered here resides in the first exon of the gene, which is well conserved between species. AQP5 also seems to be involved in the pathology of Sjögren's syndrome . Since Sjögren's syndrome affects mainly women (9:1 female/male ratio) and often displays a late onset of disease, the presumed androgen regulation of AQP5 expression could be an important factor in the aetiology of Sjögren's syndrome. In the acinar cells of normal individuals, AQP5 is expressed primarily at the apical membrane, whereas in cells of Sjögren's syndrome patients, AQP5 molecules are redistributed to the basal membrane, diverting a large portion of the fluid flow to the interstitium instead of to the acinar lumen .
Genome-wide searches for AREs
Recent results from several genome-wide searches for AR interaction sites indicate that the AR is more promiscuous than previously thought in the selection of motifs with which it interacts to stimulate gene expression. Wang et al.  reported the results of a ChIP (chromatin immunoprecipitation)-on-chip analysis on human chromosomes 21 and 22. Only 10% of genomic regions interacting with the AR were found to contain a cARE, 22% did not contain an ARE-like motif, whereas the majority (68%) were found to contain a non-canonical ARE, including isolated half-sites and everted and direct repeat elements. One of the genomic regions in which a non-canonical ARE was found was the TMPRSS2 (transmembrane protease, serine 2) gene . Using a similar technique, Bolton et al.  found a number of AREs involved in the androgen regulation of a set of androgen-responsive genes. In that study, 69% of all AR-interacting DNA fragments contained sequence motifs resembling cAREs. However, the sequence logo presentation of the enriched sites in the AR-binding regions does not allow discrimination between the type of ARE. The relatively high abundance of non-canonical or half-site ARE motifs in genomic regions that were found to bind the AR was confirmed by Massie et al. . In ChIP-on-chip analysis of androgen-deprived and -stimulated LNCaP cells, this group found only a minority of 27% of the AR binding regions to be significantly enriched in 15 bp palindromic cAREs. That study also claimed that half-sites are sufficient for the AR to bind and transactivate. It is important to consider, however, that not all half-sites will be sufficient for AR binding and transactivation, since the present study indicated that even motifs with a partial repeat structure do not always bind AR or confer androgen-responsiveness to a promoter (Figure 2). We have also shown previously that, for example, hormone-response element 1 in the mouse sex-limited protein ARU, which contains a perfect 5′-TGTTCT-3′ motif without a proper upstream half-site, does not bind the AR-DBD, nor does it transactivate in reporter assays . AR binding to DNA fragments containing 5′-TGTTCT-3′ motifs was demonstrated by our group as long ago as 1990 , but most of the elements identified in that study did not function as AREs in transient transfection asays. The possibility of a (near-) consensus ARE half-site acting as an ARE therefore needs to be investigated more thoroughly, and the putative modulatory role of flanking and spacer sequences should not be overlooked [42–44].
From seven motifs found in the vicinity of androgen-regulated promoters using NUBIScan (Table 1), we characterized three putative AREs by transactivation assays (Figure 2) and by band-shift (Figures 3 and 4). We have identified two new androgen-selective AREs, one in the human Rad9 and one in the AQP5 genes. For the RAD9 gene, we confirmed that its androgen regulation was impaired in the SPARKI mouse model. Our work also indicates that selective, non-canonical AREs are probably more general than was previously thought. What the precise criteria are for an ARE to be able to recruit AR and function as a transcriptional regulator in vivo warrants further investigation.
We gratefully acknowledge the excellent technical assistance of Ms R. De Bruyn, Ms R. Bollen and Ms K. Bosmans. F. C. is the beneficiary of a grant from the Association for International Cancer Research (AICR). G. V. is a postdoctoral fellow of the Fonds voor Wetenschappelijk Onderzoek-Vlaanderen. This work was supported by grants from the Geconcerteerde Onderzoeks Actie and the FWO-Vlaanderen.
Abbreviations: ABCC1, ATP-binding cassette subfamily C member 1; AQP5, aquaporin-5; AR, androgen receptor; ARE, androgen-response element; ARU, androgen-regulatory unit; C3(1), prostatic steroid binding protein; cARE, classical steroid-hormone-response element; ChIP, chromatin immunoprecipitation; DBD, DNA-binding domain; GR, glucocorticoid receptor; INPP5A, inositol polyphosphate 5-phosphatase; Luc, luciferase; Rad9, radiation-sensitive 9; Rad9A, Rad9 homologue A; Rhox5, reproductive homeobox 5; sARE, selective ARE; SC, secretory component; Slp, sex-limited protein; slpHRE2, and slpHRE3, two AREs in the upstream region of the slp gene; SPARKI, specificity-affecting AR knockin; SREBF2, sterol-regulatory-element-binding transcription factor 2; TATA, the consensus sequence TATAA/TAA/T; TAT-GRE, tyrosine aminotransferase glucocorticoid-response element; THRA, thyroid hormone receptor α; TK, thymidine kinase; UCK2, uridine–cytidine kinase 2
- © The Authors Journal compilation © 2008 Biochemical Society