Interaction of Androgen Response Elements with the DNA-binding Domain of the Rat Androgen Receptor Expressed in Escherichia coZi*

A fragment of the rat androgen receptor (amino acids 633-637) containing the DNA-binding domain was produced in Escherichia coli as a fusion product with protein A of Staphylococcus uureus. The fusion protein was purified on IgG-Sepharose, a method that does not involve the use of denaturing agents. Approximately 4 mg of fusion protein was obtained from 500 ml of bacterial culture. gel shift recombinant displays an for a fragment of the of mouse mammary tumor and for an intronic of the for the C3 component of rat a C3 a functional androgen interact DNA- binding

The androgen receptor (AR)' belongs to a superfamily of transcription regulating proteins which comprises not only the steroid hormone receptors but also receptors for thyroid hormones, retinoic acids, 1,25-dihydroxyvitamin DB, and receptors for a number of other identified and unidentified ligands (1-4). These transcription factors contain distinct domains responsible for ligand binding, DNA binding, and transcriptional activation. They recognize response elements (RES) localized in or adjacent to the controlled genes. Such RES may act as hormone-dependent enhancers and confer hormone responsiveness to homologous and heterologous promoters (1,(5)(6)(7).
Although a large number of genes have been demonstrated *This work was supported by the Belgian National Incentive Program on Fundamental Research in Life Sciences initiated by the Belgian State-Prime Minister's Office Science Policy Programming, Grant 3.0015.88 from the Nationaal Fonds voor Wetenschappelijk Onderzoek van Belgie, and a grant from the Nationale Loterij. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
to respond to androgens (6,(8)(9)(10)(11), relatively little is known on the interaction of androgen receptors with the corresponding androgen response elements (AREs). Purification of the AR has proven more difficult than purification of the other steroid receptors. This is at least in part related to the high protease activity in most androgen target tissues. Moreover, although binding of crude or relatively crude AR preparations to specific regions of androgen responsive genes has been described (12-14), transfection studies have often failed to demonstrate the ability of these regions to confer androgen responsiveness to reporter genes. At present, only three gene fragments have been shown to act as androgen-dependent enhancers: the long terminal repeat (LTR) of mouse mammary tumor virus (MMTV) ( E ) , a fragment of the first intron of C3(1), a gene which codes for one of the constituent peptides of prostatic binding protein (16), and a 0.75-kilobase fragment in the upstream region of the mouse sex-limited protein gene (17). This might mean that a number of the studied androgen effects are not directly mediated by androgens, that adequate function of the putative ARE relies on the presence of transcription factors which are absent in the transfected cells, or that expression of the relevant genes is specifically suppressed in the transfected cells.
The major aim of the present study was to improve our understanding of the interaction of the AR with putative AREs. To this end, the DNA-binding domain of the rat AR, fused to a fragment of protein A from Staphylococcus aureus (18), was expressed in Escherichia coli. It is demonstrated that this fusion protein interacts specifically with putative RES in the MMTV-LTR and with a similar region in the first intron of C3(1).

MATERIALS AND METHODS
Bacterial Strains and Biochemicak-The E. coli strain N4830-1 (19) carries the temperature-sensitive cI857 repressor integrated in the genome and is intended for use with vectors containing the X PL or X PR promoter. A temperature shift from 30 to 42 "C initiates protein synthesis. E. coli N99cI' (20) is a X-lysogen that carries the wild type X cI repressor and is recommended for initial production of recombinant plasmids. Both strains were purchased from Pharmacia (Uppsala, Sweden) and were transformed according to the protocol provided by the supplier. Restriction enzymes were purchased from Gibco-BRL Life Technologies (Ghent, Belgium) or Boehringer (Mannheim, Federal Republic of Germany). Klenow DNA polymerase and T4 ligase were from Boehringer Mannheim. DNA manipulations were performed according to standard procedures (21).
Construction of an Expression Plasmid Coding for the DNA-binding Domain of the Rat Androgen Receptor-A rat prostate cDNA library constructed in X gtl0 (Clontech, Palo Alto, CA) was screened with two 31-nucleotide long 5"''P-labeled synthetic probes, homologous to regions of the DNA-binding and the hormone-binding domains of the human AR. Some 200,000 plaques were screened and one of them hybridized with both probes. The 500-base pair EcoRI insert, derived from the positive phage, was subcloned into pGEM7Zf(+) (Promega,

to Androgen Response Elements
Madison, WI) and sequenced using the Sequenase DNA sequencing kit (U. S. Biochemicals, Cleveland, OH). The sequence of the insert corresponded to the entire DNA-binding domain and part of the steroid-binding domain of the rat AR. It was identical to the sequence published by Chang et al. (22) except for 3 single base changes that did not affect the amino acid sequence of the DNA-binding domain.
pGEM7Zf(+) containing the rat AR insert, was digested with TaqI and AoaII to isolate the DNA-binding domain and filled in with Klenow DNA polymerase. The 300-base pair fragment was isolated on a 6% polyacrylamide gel. The protein A expression vector pRIT2T (Pharmacia) was BamHI digested and filled in. pRIT2T and the 300base pair AR fragment were ligated and E. coli N99cI' was transformed with the ligation mixture. The correct construct, named pRITPTAR, was selected by restriction fragment analysis and partial dideoxy sequencing, and introduced into E. coli N4830-1.
Expression and Purification of the Fusion Protein-E. coli N4830-1 containing pRIT2TAR was grown overnight a t 30 "C in 250 ml of Luria Broth containing 50 pg/ml ampicillin. The cells were heatinduced by addition of another 250 ml of Luria Broth, preheated to 60 "C. The ampicillin concentration was adjusted to 50 pg/ml and the culture was incubated for another 2 h a t 42 "C. The cells were pelleted by centrifugation at 4 "C and resuspended in 20 ml of 100 mM Tris-HCI, pH 7.6,200 mM NaCI, 10 mM MgC12,O.l mM EDTA, and 10% glycerol containing 1 mM phenylmethylsulfonyl fluoride. Lysozyme was added to a final concentration of 2.5 mg/ml and after incubation for 30 min on ice, the cell suspension was sonicated twice for 30 s (Vibra Cell, Sonics & Materials, Danbury, CT). The disrupted cells were centrifuged for 30 min a t 48,000 X g. The fusion protein was purified from the supernatant on an IgG-Sepharose Fast Flow column (Pharmacia) following the manufacturer's specifications. Briefly, the cleared extract was applied to the column and the column was washed extensively, first with TST (50 mM Tris-HCI, pH 7.6, 150 mM NaCI, 0.05% Tween 20), then with 5 mM ammonium acetate, pH 5.0. The fusion protein was eluted with 0.5 M acetic acid, pH 3.4. The eluate was neutralized with 1 volume of 1 M Tris-HCI, pH 8.0, and dialyzed overnight a t 4 "C against 25 mM Hepes-KOH, pH 7.6, 40 mM KCI, 0.1 mM EDTA, 10% glycerol, and 1 mM dithiothreitol. The protein concentration was measured with the Coomassie protein assay (Pierce Chemical Co.). The extract was aliquoted, frozen in liquid nitrogen, and stored at -70 "C. Samples from different steps of the extraction and purification procedure were electrophoresed on a 10% sodium dodecyl sulfate (SDS)-polyacrylamide gel as described (23). The proteins were either stained with Coomassie Blue dye or blotted onto a nitrocellulose membrane (24) and visualized with peroxidase-conjugated rabbit anti-human immunoglobulin G (IgG) (Dakopatts, Copenhagen, Denmark).
Gel Retardation Assay-Complementary oligonucleotide probes were synthesized on a Biosearch Cyclone DNA synthesizer (Milligen, Bedford, MA) and annealed by heating to 95 "C in 10 mM Tris-HCI, pH 8.0, 100 mM NaCI, 1 mM EDTA and cooling slowly to room temperature. Restriction fragment probes and oligonucleotides were labeled with Klenow polymerase and [a-"PIdATP or [a-'*P]dCTP to a specific activity of a t least 10' cpmlpg. For gel retardation, 5,000-10,000 cpm of each probe was incubated in a total volume of 20 pl with 100-200 ng of purified fusion protein, 2 pg of poly[d(I-C)] (Boehringer), and 10 pl of 2 X TM buffer (TM: 100 mM Tris-HCI, pH 7.9, 12.5 mM MgC12, 1 mM EDTA, 1 mM dithiothreitol, 20% glycerol) at room temperature for 20 min. For competition experiments, 8-32 ng of unlabeled double-stranded oligonucleotide was added before the incubation. The samples were loaded on a 4% polyacrylamide gel (acrylfiisacryl = 19/1) and electrophoresed in 0.25 X TBE (TBE: 0.1 M Tris, 0.083 M boric acid, 1 mM EDTA) a t 150 V for 90 min. After electrophoresis, the gel was slab-dried and autoradiographed overnight (Kodak X-OMAT AR).
DNase I Footprinting Assay-Probes for footprinting were prepared as described above. Each probe (20,000 cpm) was incubated in a final volume of 50 pl with 400-3,200 ng of fusion protein, 2 pg of poly[d(I-C)], 2% polyvinylalcohol, and 25 pl of 2 X TM buffer. After incubation a t room temperature for 20 min, 50 pl of 10 mM MgCI?, 5 mM CaCI2 was added. Digestion was initiated with 0.05 units of DNase I (Boehringer). After 60 s a t room temperature the reaction was stopped with 100 pl of 10% SDS, 200 mM NaCI, 20 mM EDTA. The samples were extracted with phenol and ether, precipitated with ethanol/isopropyl alcohol after addition of 5 pg of sonicated herring sperm DNA and dissolved in 4 pl of formamide/dye mixture (98% formamide, 10 mM EDTA, 0.2% bromphenol blue, 0.2% xylene cyanol). After heating to 95 "C for 2 min and rapid cooling on ice, the samples were loaded on a 7% polyacrylamide (acrylfiisacryl = 19/1), 6 M urea sequencing gel and run in 1 X TBE. After electrophoresis, the gel was dried and autoradiographed for 2 days.

RESULTS
Construction of the PlasmidpRIT2TAR-A cDNA fragment corresponding to amino acids 533-637 of the rat AR was inserted into the polylinker site of the prokaryotic expression vector pRIT2T (Fig. 1). This cDNA fragment codes for the DNA-binding domain and some flanking amino acids. The AR gene fragment was fused in-frame with the first 265 codons of the protein A gene. A TAA sequence located immediately behind the polylinker acts as a stop codon. The resulting fusion protein is synthesized under control of the phage X P R promoter, which is temperature inducible in E. coli cI857 strains.
Expression and Purification of the Fusion Protein-E. coli cells carrying pRIT2TAR expressed large amounts of fusion protein at 42 "C. The expected molecular mass of the fusion product is approximately 40 kDa and a protein of this size was detected on SDS-polyacrylamide gels following induction (Fig. 2, lane 2). A small amount of fusion protein was also found in uninduced cells (Fig. 2, lane I), indicating some promoter activity at temperatures below 42 "C.
After cell lysis, more than 90% of the hybrid protein was found in the soluble fraction (Fig. 2, lane 3). Because of the A, 10% SDS-polyacrylamide gel, Coomassie-stained; B, 10% SDSpolyacrylamide gel, blotted onto nitrocellulose and stained with peroxidase-conjugated rabbit anti-human IgG. Lane 1, crude extract from bacterial culture grown overnight at 30 "C. Lane 2, crude bacterial extract after growth a t 42 "C for 2 h. Lane 3, cleared extract, obtained after centrifugation a t 48,000 X g for 30 min. The prominent low molecular weight band represents lysozyme that was added to disrupt the cell walls. Lanes 4-6,3 consecutive fractions eluted from the IgG-Sepharose column with 0.5 M acetic acid, pH 3.4. The oertical numbers indicate the molecular mass of the reference proteins in kilodaltons.
affinity of protein A for immunoglobulins, the fusion product could be purified on an IgG-Sepharose column. The cleared extract was applied to the column and after extensive washing, the fusion protein was eluted with 0.5 M acetic acid, pH 3.4, and neutralized immediately. Coomassie-stained gels and blots reacted with peroxidase-conjugated IgG (Fig. 2, lanes 4-6), showed that the eluate contained a highly purified fusion product. In addition to the 40-kDa band, some slower migrating bands appeared on the stained gel. These bands were not detected on the blot; they may represent contaminating E. coli proteins. Both the gel and the blot also showed some faster migrating bands, probably resulting from degradation of the fusion protein during cell growth.
Using this purification strategy, approximately 4 mg of a protein A-DNA-binding domain fusion protein was obtained from 500 ml of bacterial culture. DNA-binding Properties of the Fusion Protein-Gel retardation and DNase I footprinting techniques were used to study the DNA-binding activity of the recombinant protein.
A fragment of the MMTV-LTR, extending from position -207 to -71 and containing 4 imperfect glucocorticoid/progesterone responsive element (GRE/PRE) consensus sequences (15), was used as a first probe. In a gel retardation assay, protein-DNA complexes were detected after incubation of this MMTV fragment with the fusion protein (Fig. 3A, lane  3). Protein A, purified from an E. coli culture harboring the parental pRIT2T vector, did not bind to the probe (Fig. 3A, lane 2). This indicates that the binding activity observed with the fusion protein was not due to a contaminating E. coli protein in the purified extract.
For the following experiments, a 162-nucleotide long ScaI-PuuII fragment from the first intron of the C3(1) gene of the androgen-regulated rat prostatic binding protein was used. This fragment contains two sequences, named core I and core 11, that resemble the GRE/PRE consensus element. It has previously been shown to display in vitro affinity for partially purified AR and to confer androgen inducibility to a thymidine kinase-chloramphenicol acetyltransferase construct (16).  In lanes 2 and 3, 500 and 2,000 fmol of unlabeled oligonucleotide containing the native core I1 of C3(1) were added. Reactions in lanes 5 and 6 contained 500 and 2000 fmol of oligonucleotide with the mutated core I1 of C3(1).
In gel retardation assays, the C3(1) probe was recognized by the fusion protein (Fig. 3A, lane 6), whereas the control with purified protein A proved negative (Fig. 3A, lane 5). A probe in which the TGTTCT sequence of core I1 (i.e. the GRE/ PRE located most downstream) was mutated to TTTTCT showed no retardation (Fig. 3A, lane 9), suggesting that the recognition site is located at the level of core 11. Interestingly, this single base substitution was shown by Claessens et al. (16) to abolish androgen responsiveness, whereas a similar mutation in core I had no effect.
These observations were confirmed by a gel shift assay with a double-stranded oligonucleotide (5"GATCATAGTACGT-GATGTTCTCAAGATC-3') containing core I1 (underlined) and rsimilar oligonucleotide in which the G (double underlined) of the TGTTCT sequence was replaced by a T, as in the mutated intron fragment.
As expected, the wild type oligonucleotide, but not its mutated counterpart, was recognized by the fusion protein (Fig. 3B).
Furthermore, addition of a 200-fold molar excess of the unlabeled oligonucleotide containing the wild type core I1 sequence led to competition with the C3(1) probe for binding to the fusion protein (Fig. 3C, lanes 1-3). In similar experiments with the oligonucleotide containing the mutated core 11, no competition was seen (Fig. 3C, lanes 4-6).
Finally, a DNase I footprinting assay on the wild type C3 (1) fragment showed that the fusion protein protected a stretch of 22 nucleotides covering core I1 (Fig. 4, lanes 2-5). This suggests that the recognition site is restricted almost exactly to the actual 15-nucleotide long consensus sequence. As expected, a footprint was not detected when the mutated C3( 1) probe was used (Fig. 4, lanes 6-9).

DISCUSSION
Several authors have reported the expression of diverse steroid and thyroid hormone receptor domains in E. coli (25-34). To the best of our knowledge, this is the first report describing the production of milligram quantities of the DNAbinding domain of the AR. In the present experiments, the DNA-binding domain is produced as a fusion protein, containing part of protein A. The presence of a prokaryotic protein or peptide fused to the amino terminus of the eukaryotic protein to be expressed can protect the latter from degradation and usually guarantees a high level of expression (35,36).
Many eukaryotic proteins do not fold correctly when expressed in E. coli and form insoluble aggregates (37,38). Solubilization of such aggregates requires the use of strong denaturing agents and correct refolding upon renaturation is not evident. Conveniently, the protein A-DNA-binding domain was found in the soluble fraction after cell lysis, which obviated the need for denaturation-renaturation procedures.
For the initial verification of the DNA-binding activity of the fusion protein, a DNA fragment derived from the MMTV-LTR was used. This LTR fragment can mediate glucocorticoid, progesterone, androgen, and mineralocorticoid responses (15) and contains 4 imperfect GRE/PRE consensus sequences which are recognized i n vitro by glucocorticoid and progesterone receptors (39,40). The LTR fragment was also recognized by the AR-DNA-binding domain, as demonstrated in a gel shift assay. Further studies will be required to determine which of the 4 GRE/PREs are involved in AR binding.
In the present series of experiments, one of the two nonallelic genes that code for the C3 component of prostatic binding protein, a major androgen-regulated protein of the rat prostate, was used as a model system. An intronic fragment of this C3(1) gene has been demonstrated to display i n uitro affinity for androgen, progesterone, and glucocorticoid receptors (13,14,41). When introduced before a thymidine kinasechloramphenicol acetyltransferase construct, the same fragment confers responsiveness to 5a-dihydrotestosterone, progesterone, and dexamethasone after transfection in appropriate cells (16). The relevant fragment contains 2 elements (core I and core 11) that resemble the GRE/PRE consensus sequence. Only the most downstream located element (core 11) acts as a functional androgen, progesterone, and glucocorticoid-responsive element (16). In agreement with these findings, the experiments reported in this paper show that the protein A-DNA-binding domain fusion product recognizes only core 11. The validity of these observations is further strengthened by the demonstration that a single base pair mutation in core I1 abolishes not only the ability of this fragment to act as an ARE (16), but also its ability to interact with the protein A-DNA-binding domain fusion product. The data further indicate that glucocorticoid, progesterone, and androgen receptors, which have highly homologous DNAbinding domains, can recognize a common response element.
Although the fusion protein recognizes putative ARES, it is obvious that its affinity for these RES is rather low. Large amounts of fusion protein or of competitor DNA are needed in all binding and competition experiments. In gel shift experiments, for instance, a 100-fold protein to probe molar excess is required. In footprinting experiments, a 1,000-to 2,000-fold excess of protein has to be used. These observations are not entirely unexpected. Several explanations can be considered. Since the fusion protein is expressed in a prokaryotic system and contains only the DNA-binding domain of the AR and since this domain is fused to a prokaryotic protein that is almost 3 times as large, uncorrect folding of the DNAbinding domain is an obvious possibility. Alternatively, protein-protein interactions which enhance the binding of intact receptor to its RES in uiuo may be missing in the present in uitro system. There is considerable evidence that in normal target cells intact receptor proteins may interact with a number of other factors involved in transcriptional regulation and that this interaction may increase the specificity as well as the affinity of receptors for the cognate RES ( 5 , 33,42).
Moreover, some members of the steroid receptor superfamily have the ability to form dimers and bind cooperatively to the two half-sites of the palindromic RES (33,(43)(44)(45)(46)(47). The segments which act as determinants for protein-protein interaction in such dimeric complexes may differ from receptor to receptor. In the estrogen (46) as well as in the glucocorticoid receptor (33,(43)(44)(45) the DNA-binding region itself seems to contain a determinant which favors cooperative binding. It is conceivable that the DNA-binding domain of the AR contains no dimerization site, or that cooperative interaction is impeded by the fact that this domain is fused to a prokaryotic protein. Moreover, other determinants have been localized in the amino-terminal domain for the glucocorticoid receptor (47) and in the hormone-binding domain for the estrogen receptor (46). It is obvious that the fusion protein, used in the present experiments, may lack essential dimerization sites in these regions.