Androgen receptor-Ets protein interaction is a novel mechanism for steroid hormone-mediated down-modulation of matrix metalloproteinase expression.

Matrix metalloproteinases belong to a family of structurally related enzymes that plays important role in tissue morphogenesis, differentiation, and wound healing. Their expression is negatively regulated by several members of the steroid hormone receptor family. This is thought to occur through interaction of the steroid receptors with the transcription factor AP-1 that is otherwise required for positive regulation. Here, we demonstrate that AP-1 is not always a target for down-regulation of expression of matrix metalloproteinases by steroid receptors. Androgen receptor negatively regulates matrix metalloproteinase-1 expression not through AP-1 but through a family of Ets-related transcription factors that are also required for positive regulation. This negative regulation is specific for the androgen receptor. It does not require the DNA binding activity but needs amino-terminal sequences of the receptor. These results identify a novel regulatory pathway for negative regulation utilized by a member of the steroid hormone receptor family for down-regulating the expression of matrix metalloproteinases.

negative regulation of expression of MMP genes by a variety of growth factors, cytokines, oncogenes, tumor promoters, and steroid hormones are required (2,3). The steroid hormones in particular, negatively regulate the expression of the MMPs through the action of their corresponding receptors (4,5).
Glucocorticoid receptor (GR) exerts its negative regulation by interacting with the transcription factor AP-1, an important regulator of expression of several MMP genes (4 -8). Progesterone acting through its receptor in stromal cells of the endometrium down-regulates the levels of MMP mRNAs of the stromelysin family through a mechanism that is not yet known (9,10). In the epithelial cells of the endometrium, progesterone down-regulates the expression of matrilysin through the release of transforming growth factor ␤1 from the stromal cells (10). Although androgen receptor (AR) represses the expression of several genes (11)(12)(13)(14)(15)(16)(17), it is not clear how this is effected nor is it clear whether MMPs belong to the genes that are negatively regulated by this receptor. The only link between androgens, their receptor, and MMPs is the finding that collagen content in ventral prostate of the rat is increased in the presence of androgens and decreased after orchiectomy. These results are partly interpreted as reflecting a negative effect of the AR on MMP expression (18,19). It however remains to be demonstrated which MMPs are involved and the mechanism by which this down-regulation of expression is brought about.
In this communication we show that androgens negatively regulate the expression of interstitial collagenase (MMP1), stromelysin I (MMP3), and matrilysin (MMP7). Sequences on MMP1 gene that mediate this effect do not encompass the regulatory element for AP-1 but rather a motif that binds Ets family of proteins. Several Ets-related proteins are targets for this repression. With the use of the Ets family member ERM, we have shown that binding of the Ets protein to DNA is necessary for the repression. We have further shown that the AR requires its amino-terminal sequences for the repression. The ERM-mediated repression of collagenase I expression is specific for the AR because it does not occur with other receptors such as the GR nor the mineralocorticoid receptor. The AR has therefore adopted a unique pathway for down-modulating the expression of a MMP gene through interference of action of Ets transcription factors.

MATERIALS AND METHODS
Cell Culture and Transfection-Human prostate carcinoma DU145 cells and simian CV-1 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Human prostate carcinoma cells LNCaP were cultured in RPMI 1640 medium also supplemented with 10% fetal calf serum. All culture media contained 100 units/ml penicillin and 100 g/ml streptomycin. The cells were all kept at 37°C in a 5% CO 2 atmosphere. Unless otherwise stated, the DU145 and CV-1 cells were transfected by the calcium phosphate coprecipitation procedure in 3.5-cm diameter plates with 0.6 g of reporter plasmid, 1.8 g of expression vector for the Ets protein, and 0.76 g of expression vector either alone or containing receptor cDNA sequences. After 5 h of incubation of the cells with DNA and 2 min of glycerol shock with 10% glycerol in phosphate-buffered saline (PBS), the cells were treated with TPA (80 ng/ml) and 10 Ϫ7 M steroid hormones.
They were harvested 40 h after transfection and disrupted five times by freezing and thawing, and the cellular extracts were used for CAT assay as described (20). For luciferase assay, the transfected cells were washed twice with ice-cold PBS lacking calcium and magnesium and scraped into 200 l of lysis buffer (0.1 M Tris acetate, pH 7.5, 2 mM EDTA, 1% Triton X-100). The extract was obtained by centrifugation for 5 min at 14,000 rpm in a microfuge. Luciferase assay was performed in a Berthold Lumat LB9501 luminometer with 100 l of sample, 300 l of assay buffer (25 mM glycylglycine, 15 mM MgSO 4 , 4 mM EGTA, 2 mM ATP, 1 mM dithiothreitol, pH 7.8) and 100 l luciferine assay solution (0.2 mM luciferine in 25 mM glycylglycine, 15 mM MgSO 4 , 4 mM EGTA, pH 7.8).
Plasmid Constructs-Human ERM, chicken c-ets-1, and mouse PEA3 cloned into the expression vector pSG5 have been described (26 -28). ERM⌬C and ERM⌬N lack the carboxyl-terminal amino acids 448 -510 and the amino-terminal amino acids 1-86 (29). The construct mERM419 was generated by subcloning the EcoRI fragment of pcDNA3-mERM419 into pSG5. The plasmid pcDNA3-mERM419 contains pcDNA3 vector into which are cloned ERM sequences with a mutation that alters a tyrosine at position 419 into a proline.
The indicator constructs Ϫ517/ϩ63 Coll CAT and Ϫ73/ϩ63 Coll CAT are described by Angel et al. (21). The luciferase version was constructed by exchanging the CAT cassette with a luciferase cassette. The constructs Ϫ160/ϩ63 Coll CAT, Ϫ127/ϩ63 Coll CAT, and Ϫ107/ϩ63 Coll CAT contain the indicated region of the human collagenase I promoter sequence linked directly to a CAT gene. The 5Ј deletion mutants were generated by PCR amplification of segments of the Ϫ517/ ϩ63 Coll CAT construct and were made available to us by H.-P. Auer. The construct mEBS(Ϫ517/ϩ63) contains the Ϫ517/ϩ63 collagenase I promoter region with a mutation at the Ets binding site (EBS) that changes the sequence 5Ј-CAAGAGGATGT-3Ј into 5Ј-CAAGAACATGT-3Ј. This mutation generated an Af13 site at the EBS. The whole promoter region bearing the mutation was linked to the CAT sequence (61).
The GR and mineralocorticoid receptor expression vectors have been described (30,31). The expression vector containing wild-type androgen receptor and the mutant mAR596 have been previously described (32). The 5Ј-truncated receptor expression vectors ⌬473AR and ⌬536AR have also been been described (33). The mutant AR expression vectors mAR615 and mAR617 were constructed by PCR mutagenesis. Starting with the wild-type AR cDNA, two fragments were amplified with the use of the primer pairs 1 and 2 (5Ј-AGCTACTCCGGACCTTACG-3Ј and 5Ј-CATAACATTTCGGAAGACGAC-3Ј) and 3 and 4 (5Ј-TGTCCATC-TTGTCATCTTCGG-3Ј and 5Ј-CTTTCATGCACAGGAATTCCTGG-3Ј). The resulting two fragments were ligated by PCR overlap extension and reamplified with the primers 1 and 4. This produced a fragment containing either the mutation CGT to CAT at triplet 615 or CGG to CCG at triplet 617 within a HindIII/AspI restriction fragment. This restriction fragment was isolated and cloned into the corresponding sites of the AR expression vector. The resulting clones were identified by DNA sequencing.
The 5Ј-truncated AR expression vectors ⌬440AR, ⌬488AR, and ⌬510AR were constructed by PCR amplification. The 5Ј end primer contained a BamHI adaptor with an artificial Kozak sequence (34) (5Ј-CCGGATCCACCATGACTGAG-3Ј) linked to the following sequences: for ⌬440, 5Ј-GAAGGCCAGTTGTATGGACCG-3Ј; for ⌬488, 5Ј-CTG-GCGGGCCAGGAAAGCGAC-3Ј; and for ⌬510, 5Ј-GTGCCCTATCCC-AGTCCCACT-3Ј. The 3Ј primer was derived from the sequence 5Ј-TG-CAGCTTCCACATGTGAGAGCTGGATAGTG-3Ј immediately behind the HindIII site at amino acid position 565. After the PCR amplification reaction, the products were cleaved with BamHI/HindIII. The resulting fragment was used to replace a BamHI/HindIII sequence of a wild-type AR cDNA cloned into the BamHI/BglII site of the vector pSG5. The cloned sequences were verified by DNA sequencing.
The constructs ⌬510 -536, ⌬488 -536, and ⌬473-536 were carried out with the method of overlap extension with the polymerase chain reac-tion of Ho et al. (35) in two steps. In the first step, the 5Ј BstEII primer 5Ј-GCCTGCATGGCGCGGGTGCAGCGGGAC-3Ј and the following primer combinations: 5Ј-GTCCCTGGCAGTCTCCAAACGCATTCTGC-TCACCATGCC-3Ј for ⌬510, 5Ј-GTCCCTGGCAGTCTCCAAACGCATC-CCCTGAGGGGGCCGAGTGTAGCC-3Ј for ⌬488, and 5Ј-TCCCTGGC-AGTCTCCAAACGCATCTCGCCGCCGCCGCCGCC-3Јfor ⌬473 were used for the amplification reaction. In the second step, the HindIII primer 5Ј-TGCAGCTTCCACATGTGAGAGCTCCATAGTG-3Ј was combined with the complementary sequences for the oligonucleotides indicated above for ⌬510, ⌬488, and ⌬473. The pair of fragments generated for each deletion mutation was combined and reamplified with the BstEII and HindIII primers. The resulting fragments were cleaved with BstEII and HindIII enzymes and used to replace the corresponding region in the wild-type AR cDNA. The deletions were verified by DNA sequencing.
GST-ERM Fusion Protein and Pull-down Experiments-ERM sequences were amplified by PCR from pSG5-ERM with the use of the oligonucleotides 5Ј-CTAGGGATCCATGGACGGGTTTTATGA-3Ј and 5Ј-GACTGAATTCTTAGTAAGCAAAGCCTTCG-3Ј. These oligonucleotides contain the initiation and termination codons of the open reading frame of ERM as well as BamHI and EcoRI ends. The resulting fragment from the amplification reaction was digested with BamHI and EcoRI and subcloned between the respective sites of the GST fusion vector pGEX-4T-1 (Pharmacia Biotech Inc.). To avoid PCR errors, the BglII-EcoRI fragment from the resulting construct was replaced by the corresponding fragment from pSG5-ERM construct. The BamHI-BglII fragment and the GST/ERM boundary were confirmed by DNA sequencing.
Expression of the GST-ERM fusion protein was carried out in the bacterial strain BL21. Synthesis of the fusion protein was induced in the log phase of the bacterial cultures with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 2 h. The cells were harvested and lysed by sonification in PBS containing 2 mM EDTA. After addition of 1% Triton X-100, the lysate was cleared by centrifugation, and the expressed GST-ERM fusion protein was adsorbed onto glutathione-Sepharose beads with gentle shaking at 4°C for 1 h. This was followed by extensive washing with PBS, after which the fusion protein was eluted off the beads with 50 mM Tris, pH 8.0, 5 mM glutathione and dialyzed against PBS. Aliquots of the GST-ERM fusion protein were supplemented with 10% glycerol and 1 mM dithiothreitol and stored at Ϫ70°C.
The protocol for the GST pull-down assay was essentially as described (36). 20 g of either purified GST or GST-ERM were first bound to 20 l of glutathione-Sepharose beads for 1 h at 4°C. After extensive washing with LBST buffer (20 mM Hepes-KOH, pH 7.9, 100 mM NaCl, 2.5 mM MgCl 2 , 0.1 mM EDTA, 0.05% Nonidet P-40, 1% Triton X-100), the volume was set to 180 l with LBST buffer. 20 l of in vitro transcribed/translated radioactively labeled proteins (either the wildtype androgen receptor, the ⌬536 mutant receptor, or the LexA-VP16 fusion protein) were added with gentle mixing at room temperature for 30 min, followed by a further 30-min incubation with gentle shaking at 4°C. The beads were washed four times with 1 ml of ice-cold LBST buffer, and the bound proteins were analyzed by SDS-polyacrylamide gel electrophoresis followed by autoradiography.

RESULTS
Androgens Down-regulate MMP Gene Expression-Increase in collagen expression by androgens has been suggested to occur from a negative action of the hormone on MMP gene expression (18,19). To verify this, we carried out Northern blot experiments with mRNA isolated from androgen receptor positive LNCaP cells previously treated with androgens and the phorbol ester TPA to enhance MMP expression. The androgens used were mibolerone, methyltrienolone, and dihydrotestosterone (DHT). The Northern blot filters were hybridized with probes for interstitial collagenase (MMP-1), stromelysin (MMP-3), matrilysin (MMP-7), and as a control, glyceraldehyde 3-phosphate dehydrogenase.
With the exception of matrilysin, the collagenase I and stromelysin I genes were not expressed in the absence of TPA ( Fig. 1, lanes 1-4). TPA, however, enhanced the expression of all three MMP genes (Fig. 1, lane 5). Androgens exhibited a ligand-dependent repression of the activity of all three MMP genes in the order of DHT Ͼ methyltrienolone Ͼ mibolerone. The repressing activity of methyltrienolone and DHT was particularly noticeable on the uninduced expression of matrilysin (Fig. 1, compare lanes 3 and 4 with 1) as well as on the TPA-enhanced levels of collagenase I and stromelysin I (Fig. 1, compare lanes 7 and 8 with 5). After correcting for the amount of mRNA loaded on the gels with the glyceraldehyde-3-phosphate dehydrogenase signal, the level of repression of the uninduced expression of matrilysin by DHT was 70% (Fig. 1, lane  4) but only 50 and 20%, respectively, for the TPA-induced activity of collagenase I and stromelysin I (Fig. 1, lane 8). None of the androgens examined repressed the TPA-induced matrilysin expression ( Fig. 1 compare lanes 6 -8 with 5). The reason for the lack of androgen effect on the TPA-induced level of matrilysin expression is not known, but it may reflect the uniqueness of the promoter region of this gene compared with the other MMPs.
AP-1 Is Not a Target for Negative Regulation by the AR-To determine the target of negative regulation of MMP gene expression by the AR, we first transiently transfected receptor negative CV-1 cells with a wild-type AR and a human collagenase-I-CAT construct (Ϫ517/ϩ63 Coll-CAT). This indicator gene was employed as a prototype of the MMP genes because of its extensive use in studies on the negative regulation of MMP expression by steroid hormones such as glucocorticoids (4,37). Collagenase I activity was stimulated with the phorbol ester TPA. This compound activates the transcription factor AP-1, which in turn cooperates with Ets proteins in binding to the promoter region of the collagenase I gene (38). The AR was activated by the synthetic androgen mibolerone.
In the absence of TPA, the transfected collagenase I gene construct was expressed at a low but detectable level in the CV-1 cells (results not shown). TPA enhanced this level of expression, and the stimulated activity was down-regulated by the activated AR (Fig. 2). The level of repression in the transient transfection assay even exceeded that in the Northern blot experiment. This made the transient transfection assay a relatively useful system for the determination of regulatory elements on the collagenase promoter that contribute to the repression.
Deletions of the 5Ј-terminal region of the collagenase I promoter to position Ϫ107 from the start of transcription produced mutant constructs that were still repressed by the AR (Fig. 2). A deletion mutant that ends at position Ϫ73 and contains only the AP-1 binding site of the collagenase I gene no longer showed a repressed activity in the presence of the AR but rather exhibited a slightly activated response (Fig. 2). This slight activation was confirmed in another experiment in which we measured the effect of the AR on the activity of a multimerized collagenase I AP-1 site cloned in front of a minimal promoter construct (results not shown). Other investigators have also previously reported an activating influence of AR on the activity of the dimeric partners of AP-1, c-Fos, and c-Jun (39).
Because the promoter activity of the Ϫ107 collagenase construct but not the Ϫ73 construct was repressed by the AR, the sequence necessary for the repression must be located between position Ϫ107 and Ϫ73. This region is known to be important for the transcriptional activity of collagenase I and to contain a binding site for the transcription factors from the Ets family (38). Thus the repressive action of the AR may be mediated by Ets proteins.
Negative Regulation by AR Is Mediated through Ets Proteins-To confirm the involvement of Ets proteins in the repression by the AR, we transiently co-transfected expression vectors containing ets and AR cDNA sequences along with the Ϫ517/ϩ63 collagenase promoter construct into receptor negative DU145 human prostatic cells. Collagenase I activity was enhanced in this case by overexpression of ets sequences. This mode of induction of collagenase I promoter activity functioned more effectively in DU145 as opposed to CV-1 cells. This is possibly due to the relatively lower levels of endogenous Ets proteins in DU145 cells compared with the CV-1 cells (results not shown).
Several Ets proteins, among them Ets-1, PEA3, and ERM, enhanced collagenase expression in the DU145 cells in the absence of TPA, and this response was repressed by the activated receptor (Fig. 3A). The AR-mediated repression of collagenase I activity through Ets proteins appears to be specific for this type of steroid receptor. Other members in the steroid receptor family such as the GR or the mineralocorticoid receptor do not exhibit such an effect as demonstrated in transfection experiments with ERM as the inducing Ets protein (Fig.  3B). These other receptors, however, functioned efficiently in the same cells in transactivation studies (results not shown), indicating that their inability to transrepress is not due to lack of expression in the transfected cells. It therefore appears that different members of the steroid receptor family use different means to negatively regulate expression at the collagenase I promoter. The GR does so through AP-1 sequences (4), but the AR accomplishes this negative function through the Ets family of transcription factors.
DNA Binding Activity of ERM Is Necessary for AR-mediated Repression-The Ets transcription factor family comprises of more than 20 members that are involved in normal development and in oncogenesis (for reviews see Refs. 40 -42). They possess a conserved DNA binding domain, the ETS domain, of some 85 amino acids (43). The degree of homology in this region has allowed them to be grouped into subfamilies (42). One of these subfamilies is further characterized by a conserved 32acidic amino acid transactivation sequence at its amino-terminal region and consists of PEA3, ER81, and ERM (26,29).
Generally, the Ets proteins possess transactivation function at their amino-and carboxyl-terminal regions, although they may also have other functions such as a negative regulatory region at their carboxyl-terminal region as in the case of Ets-1 and Erg (44,45). In principle all these regulatory regions are potential targets for the negative action of the AR. To determine the necessary regions of ERM for the repression, we analyzed different deletion mutants of this protein in our transfection experiments.
Deletion of the acidic domain at the amino terminus or the carboxyl-terminal region downstream of the Ets domain lowered the transactivation potential of ERM in agreement with published data (29). The reduced transactivation function of the mutant ERM protein can, however, still be repressed by the AR (Fig. 4), indicating that the deleted sequences do not contribute to the repression. A mutation at position 419 of ERM that converts a conserved tyrosine in the ETS domain into a proline destroyed its transactivation potential and the transrepressing action of the AR (Fig. 4). Because this mutation destroys the structure of the DNA binding domain of ERM (29), we concluded that binding of ERM to the collagenase promoter is important for the repression. This finding is further corroborated by studies in which mutation of the Ets-binding site in the collagenase promoter by nucleotide exchanges [mEBS(Ϫ517/ϩ63)] destroyed the ability of ERM to transactivate and of the AR to transrepress (Fig. 4).
Regions of the AR Important for Repression-To determine what regions of the AR contribute to the repression, we co- transfected various deletion constructs of the receptor with the wild-type ERM expression vector and the Ϫ517/ϩ63 collagenase luciferase construct into DU145 cells. Deletions of the amino-terminal sequence of the AR up to position 510 (⌬510) produced mutant receptors that still repressed collagenase expression (Fig. 5). A mutant construct lacking sequences that encode the first 536 amino acids (⌬536) was, however, unable to repress (Fig. 5), indicating that sequences between 510 and 536 are important for repression. Surprisingly, deletion of this region in the context of the wild-type receptor (⌬510 -536) did not destroy the ability of the mutant receptor to repress (Fig. 5). This suggests that further amino-terminal sequences are involved in the repression.
To determine these sequences, we deleted the amino acids that precede the DNA binding domain ⌬488 -536 and ⌬448 -536. Both mutant receptors still transrepressed (Fig. 5), pointing to a contribution of more distal amino-terminal sequences to the repression. Removal of the first 440, 473, or 488 amino acids in the context of the ⌬510 -536 mutant did not provide a hint as to which region in the amino terminus contributes to the transrepression activity of the AR (Fig. 5). These results together led us to conclude that no single entity in the amino terminus is necessary for repression. The amino terminus must contain highly redundant motifs that together provide the repressing interface for the down-modulation of the collagenase I activity. These motifs could either function directly through interaction with the Ets-related proteins or through prior recruitment of other regions of the receptor.
This latter possibility was investigated by the use of natu-rally occurring mutations outside the amino terminus that alter the function of the AR. We examined three mutant receptors that carry single amino acid exchanges in their DNA binding domains. One of them carries an amino acid exchange at position 596 in the D loop of the receptor that prevents dimerization and DNA binding activity of the receptor (32,33). The other two carry amino acid exchanges at positions 615 and 617 in the nuclear localization signal (46,47) and are therefore defective in cytoplasmic nuclear translocation (results not shown).
In transfection experiments in which ERM-activated collagenase I promoter activity, the mAR596 mutant receptor was as effective as the wild-type receptor in transrepression, whereas the mAR615 and mAR617 did not repress but rather produced a slight activation (Fig. 5). The repression by the 596 mutant receptor indicates that dimerization and DNA binding activity of the AR are not necessary for the repression. Nevertheless nuclear transport is important, as indicated by the results of the two nuclear translocation defective mutants. Our results with these mutant receptors do not provide any evidence of participation of the DNA binding domain in the repression. Thus the amino terminus either interacts directly with ERM or with auxilliary proteins to inhibit transactivation by ERM.
AR Interacts Physically with ERM-In a GST pull-down assay, we investigated whether the AR interacts physically with ERM. Bacterially produced GST-ERM fusion protein was linked to glutathione-Sepharose beads and incubated with radioactively labeled in vitro translated AR. As a nonreceptor control, a radioactively labeled lexA-VP16 fusion protein was used. After extensive washing of the glutathione beads to eliminate unspecific interactions, specifically bound proteins were eluted and visualized by gel electrophoresis.
In this assay, the AR specifically interacted with the GST-ERM fusion protein but not with the GST protein alone (Fig. 6, compare lane 5 with lane 4), indicating an interaction with the ERM portion of the fusion protein. The interaction appears to be specific because the nonreceptor lexA-VP16 protein did not interact either with the GST-ERM fusion nor with the GST protein alone (Fig. 6, lanes 8 and 9). The mutant AR (⌬536AR) that lacks the whole amino-terminal domain and does not repress ERM-activated collagenase expression still interacted specifically in the assay with ERM (Fig. 6, lanes 6 and 7). This indicates that the interaction of the AR with ERM does not occur through the amino-terminal region of the receptor but through the DNA or hormone binding domains. The mere interaction of the AR with ERM may not result in repression but possibly allows the apposition of the two proteins to enable the amino-terminal region of the AR to interfere with ERM function. This can take place through steric interferance of the transactivation function of ERM or through titration of auxilliary proteins that are needed for transactivation by this protein.

DISCUSSION
Negative regulation of MMP expression by androgens has been suggested to occur in the prostate where dynamic tissue remodelling occurs as a result of alterations in extracellular matrix, cellular surface integrins, and protease activity (48). Several gelatinolytic proteinase activities have been detected by zymographic analysis of extracts of ventral prostate of the rat. These activities increase after castration and decrease upon treatment of the castrated animals with testosterone (49,50). However, it has not been determined which MMPs are targets for the negative regulation by androgens nor how their activity is repressed by this hormone.
The main message of this paper is that androgens acting through their receptor negatively regulate the expression of MMP-1, -3, and -7 in a prostate tumor cell line. Of these three genes, MMP-7 is known to be expressed in human prostatic adenocarcinoma (48,51,52), and we have preliminary evidence that MMP1 is expressed in this tissue as well. 2 On the whole very little information is available on mechamisms of negative regulation of gene expression by the AR. Two studies are so far reported in the literature. In the first case, AR has been reported to negatively regulate the expression of glycoprotein ␣-subunit gene through binding of the receptor to DNA and displacement of an important regulatory factor (11). Previous reports on GR down-regulation of expression of this gene by a mechanism of receptor-DNA interaction (53) were later retracted and replaced by another model involving protein-protein interaction (54). Whether this also applies to the proposed mechanism of DNA binding by the AR in the negative regulation of the glycoprotein ␣ subunit gene is yet to be determined. In the second case, experiments designed to examine AR-mediated negative regulation of a low affinity neurotrophin receptor chimeric gene identified AP-1 sequences on the plasmid vector but not in the promoter region as the target for the repression (55).
We have in the present study identified a binding site for the Ets-related family of proteins as a target for the negative regulation of expression of MMP-1. Because this regulatory element is present in the promoter regions of the other MMP genes, we take it that it participates in the AR-mediated repression of activity of these genes. Ets binding sites are bound by several members in the Ets family and these proteins cooperate with AP-1 in the activation of expression of several genes. Collagenase I, stromelysin I (38,56), human urokinase type plasminogen activator (57), or the polyoma virus enhancer element (58) have all been reported to be positively regulated by a cooperative action of Ets proteins and AP-1. In many MMP promoters, the Ets binding sites are so close to the AP-1 site that it is tempting to postulate a negative action of AR by an Ets protein-AP-1 complex. This idea is further strengthened by recent studies that show that ERM, the Ets protein we have used in our studies, functionally interacts with c-Jun, one of the heterodimeric partners that constitute AP-1 (59). Whether the AR functions only through Ets proteins or through an Ets-AP-1 complex has not been analyzed in this present work.
We have, however, clearly shown in our experiments that with regards to ERM-activated expression of collagenase I, a fundamental difference exists between the transrepression activities of the AR and the GR. The AR negatively regulates ERM's activity, but the GR is unable to do so. Whether other Ets-related proteins behave similarly to ERM in mediating ARbut not GR-directed repression is yet to be determined. There are, however, several lines of evidence that suggest that certain FIG. 6. The androgen receptor physically interacts with ERM. 20 l radioactively labeled rabbit reticulocyte lysate-translated wildtype AR, mutant AR (⌬536AR), and LexA VP16 were incubated with bacterially expressed GST and GST-ERM immobilized on glutathione-Sepharose 4B beads. After extensive washing, the bound proteins were eluted and analyzed by SDS-polyacrylamide gel. Equal amounts (20 g) of GST and GST-ERM were used. Lanes 1-3 show one-tenth of the input amount of radioactively labeled proteins. M indicates the lane containing radioactively labeled protein markers. members of the Ets protein family may even mediate negative regulation by the GR. First, the Ets member Sp1-1/PU.1-mediated transcriptional activation through PU.1 site on an artificial reporter gene is repressed by the GR (60). Second, in in vivo footprinting experiments in human skin fibroblast CRL 1497 where the negative effect of the GR was examined at the collagenase I promoter, no change in factor occupancy at the AP-1 was observed, although there was a clear displacement of a protein binding to the Ets site (37). The Ets protein binding to this site is not known, but its displacement may have resulted from an Ets protein-GR interaction. Future experiments on the functional interaction of different Ets proteins with members of the steroid hormone receptor family will reveal similarities or differences in the negative action of the steroid hormone receptors.
Physical interaction of AR with ERM has been demonstrated in this work and proposed to form the basis of the negative action of AR on the transactivation function of ERM. Surprisingly an amino-terminal deletion mutant of the receptor that does not repress also interacted with this protein. We also found an interaction of GR with ERM, although this receptor does not repress the activity of ERM (results not shown). It therefore appears that mere association of the receptors with the Ets proteins is not enough for the repression. AR association with ERM through its DNA or hormone binding domains may allow its amino-terminal sequences to gain assess to and interfere with the action of other factors necessary for the transactivation function of ERM. Whether this occurs through a secondary protein-protein interaction or through auxilliary proteins is yet to be determined.
The ERM-mediated negative regulation of collagenase I expression by AR described in this work opens up another chapter in the diverse mechanisms for transrepression by members of the steroid hormone receptor family. Because the functions of Ets proteins range from normal development to oncogenesis, their interaction with members of the steroid receptors may reveal a whole repertoire of regulatory possibilities that could have major impact on various physiological and pathophysiological processes.