The androgen-specific probasin response element 2 interacts differentially with androgen and glucocorticoid receptors.

The nuclear receptors constitute a large family of transcription factors characterized by a well conserved DNA-binding domain. The receptors for glucocorticoids, progestins, mineralocorticoids, and androgens constitute a subgroup because they bind in vitro with high affinity to DNA elements containing a partial palindrome of the core sequence 5′-TGTTCT-3′. In vivo, however, the corresponding steroids differentially regulate the expression of their target genes, even when more than one receptor type is present in a particular cell. The DNA-binding domains of the androgen and of the glucocorticoid receptors bind most androgen response elements with similar relative affinities. In contrast, one element (5′-TGG-3′) which was recently described in the promoter region of the probasin gene selectively interacts with the DNA-binding domain of the androgen receptor and not with that of the glucocorticoid receptor. From studies with chimeric elements, it can be deduced that it is the left subsequence 5′-GGTTCT-3′ which excludes the glucocorticoid receptor domain from binding. In co-transfection experiments where the ARE of the C3(1) gene is responsive to both androgens and glucocorticoids, the probasin element is induced only by androgens and not by glucocorticoids. The existence of response elements which are recognized preferentially by the androgen receptor provides yet another possible mechanism to explain the differences of the in vivo effects between androgens and other steroids of the subgroup.

* This work was supported in part by Grant Geconcerteerde Onderzoeksactie van de Vlaamse Gemeenschap and by Grant 3.0048.94 from the Belgian Fonds voor Geneeskundig Wetenschappelijk Onderzoek. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Senior assistant of the Nationaal Fonds voor Wetenschappelijk Onderzoek.
§ Supported by a scholarship of the Vlaams Instituut voor de Bevordering van het Wetenschappelijk-Technologisch Onderzoek in de Industrie.
ʈ To whom correspondence should be addressed. Tel.: 32-16-345-700; Fax: 32-16-345-995; E-mail: frank.claessens@med.kuleuven.ac.be. assays, in which the PB-ARE-2 element acts as ARE and not GRE in conditions were the C3(1)ARE construct is responsive to both androgens and glucocorticoids. This is the first report of a simple ARE which is specifically recognized by the AR and therefore provides evidence for one of the older hypotheses put forward to explain steroid specificity of in vivo responses, i.e. the existence of receptor-specific response elements (17). EXPERIMENTAL PROCEDURES Enzymes were purchased from Pharmacia Biotech Inc., Promega, and Boehringer Mannheim. [␣-32 P]dATP was purchased from Amersham. X-Omat S x-ray films were from Kodak. Oligonucleotides were synthesized on a Biosearch Cyclone DNA synthesizer (Milligen Corp., Bedford, MA).
The fragment of the AR cDNA coding for amino acids 533 to 637 (numbering according to Ref. 4) and the corresponding fragment from the GR were cloned 3Ј to an expression cassette for protein A in pRIT2T (Pharmacia). Procedures for expression in Escherichia coli and subsequent purification of the proteins have been described elsewhere (6,16). The only modification applied here was the addition of 1 M ZnCl 2 to the lysis buffer, which resulted in an increase of the affinity of the prepared proteins for DNA. To exclude variations in quality of proteins, experiments were repeated with three independent protein preparations. No clear differences have been observed.
DNA binding was studied by gel retardation analysis. Constant amounts of the labeled oligonucleotides were incubated with increasing amounts of the fusion protein in 20 l of binding buffer (10 mM Hepes, pH 7.9, 2.5 mM MgCl 2 , 0.05 mM EDTA, 10% glycerol, 50 mM NaCl, 1 g of poly(dI/dC), 0.1% Triton, 1 mM dithiothreitol). Subsequently, free probe was separated from bound probe by a 90-min electrophoresis at 120 V in a nondenaturing 4% polyacrylamide gel.
The African green monkey kidney cell line CV-1 was obtained from the American Type Culture Collection and maintained at 37°C under 5% CO 2 in Dulbecco's modified Eagle's medium supplemented with 4.9 g/liter glucose, 10 mM Hepes, 10 units/ml penicillin, and 10 g/ml streptomycin (Life Technologies, Inc.), and 10% fetal calf serum (Seralab). Transient transfections were done as described (14). Oligonucleotides (Table II) were cloned as dimers in the SalI site of a pBLCAT2derived vector. To lower background activity, the RsrII-BamHI fragment, containing part of the thymidine kinase promoter, as well as the NdeI-HindIII fragment, containing a forskolin-responsive element, were deleted from the original pBLCAT2 (18).

RESULTS AND DISCUSSION
DNA Binding by AR-DBD and GR-DBD Is Very Similar-A receptor fragment containing the DNA-binding domain and the first 30 amino acids from the hinge region of the AR was expressed in E. coli as part of a fusion product with protein A. Sequence-specific DNA binding by this protein has been well documented for the AREs of the MMTV LTR promoter and the C3(1) intron (6,16). Since then, we and others have described several new sequences that bind the AR or act as AREs (Table  I and references therein). These AR-binding sequences were tested for their relative affinity for AR-DBD in comparative gel retardation experiments (Fig. 1). Based on our results, we can classify these sequences as either high or low affinity binding elements (Table I).
The low affinity binding group contains elements which do not display any retardation. Nevertheless, we either found them to bind AR-DBD in footprinting experiments (19 -21) or we considered them as candidate AREs in genomic fragments with in vitro affinity for the AR (22,23). Additional binding sites in the vicinity of these AREs in the genomic context seem to be necessary for the functionality of these elements in transfection experiments (20,21).
Elements displaying obvious AR binding in our gel retardation experiments ( Fig. 1) were classified as high affinity binding. The consensus derived from these motifs (Table I) was identical to the GRE consensus (10), except for a A/G difference at position Ϫ7 relative to the central nucleotide. This difference, when introduced in the distal GRE of the MMTV LTR, has no dramatic effect on the recognition by GR or PR (7,9). The high resemblance to the GRE consensus is in agreement with the results from a DNA-binding site selection assay performed with the AR-DBD fusion protein used in this study (24). Furthermore, when the elements described in Table I were examined for GR-DBD binding, it became clear that all sequences with high affinity for AR-DBD also have a high affinity for GR-DBD (results not shown). Similar observations have been reported for the Slp, TAT, and C3(1) elements (25).
AR-DBD, but Not GR-DBD, Binds to the ARE-2 from the Probasin Gene Promoter-In the second part of this study, we compared the relative affinities of the AR-DBD as well as the GR-DBD for the C3(1)ARE and the PB-ARE-2 (Table II and Fig. 2). The C3(1)ARE has a high affinity for both DBDs, in agreement with the results of transfection experiments in which this ARE can act as a GRE as well (13). The PB-ARE-2 is recognized by the AR-DBD with high affinity, and, when aligned with the C3(1)ARE as in Table II, its partial palin-  (19); C3I, C3(1)ARE, C3(2)ARE, C3pr-, C3prϩ, and C1ABS originate from AR-binding fragments of the C3(1), C3(2), and C1 genes (19,22,23); GRE 177 contains the distal GRE from the MMTV LTR HRE (8); flX, PSA, and hGK1 sequences were taken from Refs. 21 and 32. These sequences were synthesized as the central part of double strand oligomers. The codes used in this study are given at the left. The homologies with the GRE consensus 5Ј-GGTACAnnnTGTTCT-3Ј are underlined. The oligonucleotides are divided into two groups according to their low or high affinity for the AR-DBD as determined in Fig. 1. The consensus of the high affinity binding oligomers is given at the bottom.  Table I were labeled to the same specific activity. Equal amounts of DNA (4 ϫ 10 4 cpm) were incubated with increasing amounts of AR-DBD protein A fusion protein: lanes 1, 0 pmol; lanes 2, 2.5 pmol; lanes 3, 5 pmol; lanes 4, 10 pmol; lanes 5, 20 pmol. After electrophoresis, the gels were dried and exposed for 2 h to a X-Omat S film. dromic structure and resemblance with the consensus for the high affinity binding AREs (Table I) becomes apparent. However, in gel retardations with the GR-DBD, no binding was observed (Fig. 2). This is the first description of a sequence which is recognized by the AR-DBD and not by the GR-DBD.
When compared to the MMTV LTR-GRE 177, the PB-ARE-2 has a T instead of an A at position Ϫ4 and an A instead of a T at position ϩ2 (5Ј-GGTTCTnnnAGTACT-3Ј). These substitutions, when introduced in the MMTV LTR-GRE 177, have a dramatic effect on the glucocorticoid-inducibility (7,9). Our GR-DBD binding data corroborate these observations.
The PB-ARE-2 Is a Specific Androgen Response Element-In transient transfection experiments, pb promoter constructs containing the ARE-2, co-transfected with receptor-coding plasmids, are more effectively induced by androgens than by glucocorticoids (15). Transgenic mice, containing a construct in which the promoter region of pb regulates CAT expression, synthesize this enzyme in an androgen-dependent way (26). Although this is in agreement with our in vitro binding data, the mechanism of androgenic induction of the pb gene is more complex since another ARE, as well as additional transcription factors, are involved (27).
To verify whether the PB-ARE-2 is a functional ARE, but not GRE, we cloned the oligonucleotides containing the PB-ARE-2 and the C3(1)ARE (Table II) in duplicate immediately upstream of the tk-TATA-box followed by a CAT-coding region. Transient co-transfection of these constructs with expression plasmids for the rat AR or GR demonstrated that, at 10 Ϫ9 M concentrations, both androgens and glucocorticoids induce transcription of the C3(1)ARE construct, whereas only androgens induce the transcription of the PB-ARE-2 element (Fig. 3). These experiments clearly indicate that the PB-ARE-2 is a functionally selective ARE, and that this selectivity is due to a difference at the level of receptor-DNA binding by AR versus GR.
Which Characteristics of PB-ARE-2 Are Important for the AR-DBD-specific Binding?-To verify whether the selectivity of the AR-DBD for the PB-ARE-2 was due to the partial palindromic element, and not to the surrounding nucleotides, the underscored elements in the sequence 5Ј-AATAGGTTCTTG-GAGTACTTTAC-3Ј were inserted in the C3(1)ARE environment (PB/C3; Table II). This new sequence, when compared to the PB-ARE-2, has slightly decreased affinity for the AR-DBD (Fig. 2), which indicates that the stabilization of the DNAprotein interaction (28) by the surrounding nucleotides is sequence-dependent. Similar observations have also been made for other AREs (24,29). However, the PB/C3 element has no affinity for the GR-DBD as measured in our gel retardation experiments (Fig. 2), indicating that it is the imperfect palindrome (as underlined) of the PB-ARE-2 which is involved in the specific receptor interactions.
Gel retardation experiments with other chimeric oligonu-TABLE II Sequences of the PB-ARE-2 and C3(1)ARE derived oligonucleotides Oligonucleotides containing parts of the PB-ARE-2 element inserted in the C3(1)ARE environment were tested in gel retardation experiments. Underscored are the 5Ј-AGTACT-3Ј (normal) and the 5Ј-GGT-TCT-3Ј subsequences from the PB-ARE-2. PB/C3 contains the partial palindrome of the PB-ARE-2 element inserted in the C3(1)ARE environment; IR1 contains an inverted repeat (in this case the same as a direct repeat) of the element 5Ј-AGTACT-3Ј; IR2 and DR1 contain an inverted and direct repeat, respectively, of the element 5Ј-GGTTCT-3Ј, 4N contains the PB-ARE-2 with an additional A inserted between the two palindromic parts. DR3 contains a triplet of the 5Ј-GGTTCT-3Ј spaced by 3 nucleotides.  Table II were incubated without proteins (lanes 1) or with 5 pmol of AR-DBD (lanes 2), 10 pmol of AR-DBD (lanes 3), 5 pmol of GR-DBD (lanes 4), 10 pmol of GR-DBD (lanes 5). Gels were dried and exposed to X-Omat S films.

FIG. 3. Induction of PB-ARE-2 and C3
(1)ARE containing constructs by androgens and glucocorticoids. 10 6 CV-1 cells were transfected with the empty reporter vector (tkcat) or with constructs containing two C3(1)AREs (C3(1)ARE) or two PB-ARE-2 (PB-ARE-2), and CAT assays were performed as described (19). The amount of acetylated chloramphenicol was determined on a Molecular Dynamics PhosphorImager device and expressed as the sum of all pixel values in the area of the acetylated chloramphenicol, minus background. Cells were grown in the absence of hormone (white bars), or induced with 10 Ϫ9 M R1881 (as androgen; black bars) or dexamethasone (as glucocorticoid; striped bars). All numbers are the mean of three independent experimental values. cleotides (Table II) revealed that an inverted repeat (IR1; in this case identical to a direct repeat) of the 5Ј-AGTACT-3Ј subsequence is recognized by both the GR-DBD and AR-DBD (Fig. 2), while the inverted repeat (IR2) of the 5Ј-GGTTCT-3Ј motif retains the AR-DBD specificity. Much to our surprise, a direct repeat (DR2) of this element, which has a G at position ϩ2, is also bound by the AR-DBD, but not by the GR-DBD. In conclusion, it seems to be the 5Ј-GGTTCT-3Ј motif which excludes the GR-DBD from binding.
The importance of the length of the spacer was illustrated by the absence of retardation of oligomers containing a four-instead of three-nucleotide spacer. Adding a third copy of the 5Ј-GGTTCT-3Ј core to the PB-ARE-2 sequence (DR3) does not increase the amount dramatically, nor does it change the position of retarded oligonucleotide (Fig. 2).
All these data indicate that the AR-DBD is binding the PB-ARE-2 in a classical way (1, 10), as two molecules oriented head to head interacting with the two parts of an imperfect palindrome separated by a three-nucleotide spacer.
Which Characteristics of the Receptor Fragments Could Explain This Difference in Sequence Specificity?-In the DNAbinding domain, defined as the 75-amino acid stretch starting at the first Cys of the first zinc-finger, the GR, PR, and MR differ only at 7 positions, but the AR differs in 17 residues from the GR (25). Possibly, these differences between AR-DBD and GR-DBD have an effect on the DNA recognition. Furthermore, in the receptor fragments as used in this study, 30 amino acids from the less conserved hinge region were included. Mader et al. (30) reported that the affinity of the ER-DBD increases significantly when parts of the hinge region were included in the proteins, but changes in sequence specificity have not been reported.
In summary, the AR-DBD has a similar but not identical sequence specificity when compared to the GR-DBD. With the PB-ARE-2 element, we have a first example of a selective ARE. Further investigation of the many recently described androgenregulated genes (summarized in Ref. 31) may reveal whether other examples of AR-specific elements exist.