Mutual transcriptional interference between RelA and androgen receptor.

Cross-modulation between androgen receptor (AR) and NF-kappaB/Rel proteins was studied using various androgen- and NF-kappaB-regulated reporter genes under transient transfection conditions. In COS-1 cells, elevated expression of RelA (p65) repressed AR-mediated transactivation in a dose-dependent manner, whereas NFkappaB1 (p50), another major member of the NF-kappaB family, did not influence transactivation. The repression of AR appeared to involve the N-terminal region of the protein between residue 297 and the DNA-binding domain. RelA-mediated transrepression could not be overcome by increasing the amount of AR. Transcriptional interference between RelA and AR was mutual in that cotransfected AR was able to attenuate transactivation by RelA in a dose- and steroid-dependent fashion. An excess of RelA was able to rescue the repression to some extent. Immunological analyses of RelA and AR protein levels indicated that transrepression was not due to reciprocal decrease in their amounts. Neither did AR increase the concentration of IkappaBalpha, which can sequester and inactivate RelA. Electrophoretic mobility shift assays using extracts from cotransfected cells and purified recombinant proteins showed that AR and RelA did not significantly influence each other's DNA binding activity. Nevertheless, protein-protein interaction experiments demonstrated a weak association between AR and RelA. Collectively, these data suggest that the mutual repression in intact cells is due to formation of AR-RelA complexes that are held together by another partner or to competition for a coactivator required for transcription.

processes, ranging from the development of neural tissues to the modulation of immune function. Once bound to androgen, AR acquires a new conformational state, which renders the receptor capable of interacting not only with DNA sequences specified by androgen response elements (AREs) but also with other transcription-regulating proteins. This interaction will eventually result in activation and/or repression of specific gene transcription, depending on the physiological context (3).
The importance of a phenomenon called cross-modulation, unexpected interactions of distinct transcription factors, is illustrated by well documented interactions of AP-1 family members with the nuclear receptors (3)(4)(5)(6). Molecular mechanisms that underlie these interactions have mostly remained elusive. Cross-talk between signaling pathways may link processes occurring in different cellular compartments, increase regulatory diversity, and provide opportunities for cell-and tissue-specific responses. AP-1 and NF-B families of transcription factors regulate an array of gene networks that are induced in response to growth factors, mitogens, tumor promoters, DNAdamaging agents, or oxygen radicals (7,8). Although NF-B is a ubiquitous transcription factor, its function plays a central role in cells of the immune system (7). It has been recently reported that one member of the nuclear receptor superfamily, the glucocorticoid receptor (GR), is able to repress NF-B function probably through a direct physical association that generates transcriptionally inactive complexes (9 -11). The transactivation properties of AR and GR are in many aspects similar; however, some of their cross-modulatory properties with AP-1 family members are clearly distinguishable (3,12).
The interleukin-6 (IL-6) gene is one of the many cytokine genes, the regulation of which involves activation of NF-B (13), and androgens have been reported to inhibit IL-6 production by murine bone marrow-derived stromal cells (14). Additionally, several other genes down-regulated by androgens appear to contain potential NF-B binding sites in their promoter regions. 2 On the other hand, NF-B has been implicated in the repression of the AR gene promoter, and an age-dependent increase in NF-B content in the liver may relate to androgen desensitization of this tissue that occurs during senescence (15). Tumor necrosis factor ␣, an activator of NF-B, has been shown to inhibit proliferation of androgen-dependent prostate cancer cells, whereas cancer cells devoid of AR were not affected (16), suggesting that NF-B interferes with the function and/or synthesis of the AR. In the present work, we have investigated cross-modulation between NF-B and AR by using transactivation assays in intact cells and by examining the possibility that these proteins are physically associated in vitro.
Cell Culture and Transfections-CV-1 and COS-1 cells (from ATCC) were maintained and transfected using the calcium phosphate precipitation method as described previously (20,22). In short, 1.5 ϫ 10 6 cells were plated on a 10-cm dish 24 h before adding the precipitate containing indicated amounts of expression and reporter vectors. A ␤-galactosidase expression plasmid, pSV-␤gal (Promega), was used as an internal control for transfection efficiency (4 g/10-cm plate). For preparation of whole cell or nuclear extracts, 5 g of both pCMV-hAR and pCMV-p65 per 10-cm dish was used to transfect COS-1 (2.5 ϫ 10 6 cells) using DOTAP transfection reagent according to manufacturer's instructions (Boehringer Mannheim). The cells received 18 h later fresh medium containing 2% (v/v) charcoal-stripped fetal bovine serum in the presence or the absence of 10 nM testosterone. ␤-Galactosidase activity and protein concentration were assayed as described previously (20). Luciferase activity was determined with reagents from Promega using Luminoskan RT reader (Labsystems, Helsinki, Finland) (23). Statistical analyses of the data were carried out using two-tailed Student's t test.
Immunoblotting, Coimmunoprecipitation, and Protein-Protein Interaction Experiments-Whole cell extracts from COS-1 cells were resolved by electrophoresis on polyacrylamide gels under denaturing conditions, proteins transferred onto Immobilon-P membrane (Millipore, Bedford, MA) and processed as described previously (20). Coimmunoprecipitation of in vitro translated proteins (RelA in wheat germ lysate and AR in rabbit reticulocyte lysate) was performed using AR333 antibody 4 (raised against purified full-length rAR) or a RelA-specific anti-body. After a 30-min incubation at 22°C, the cross-linking agent dithiobis(succinimidylpropionate) was added to a final concentration of 5 mM, and the incubation continued for additional 30 min, essentially under the conditions described by Stein et al. (25). Protein-protein affinity chromatography using GST-NT, GST-DBD, and GST-LBD with [ 35 S]methionine-labeled RelA and NFB1 proteins was carried out as described (25), except that incubations were performed at 22°C and washes at 4°C. Triton X-100 was omitted from buffers, and the washing buffer contained 200 mM NaCl.

Repression of Androgen Receptor by RelA-Transcriptional
cross-modulation between AR and NF-B was studied in COS-1 cells, which are devoid of endogenous androgen receptor. The cells were cotransfected with an AR expression vector and increasing amounts of RelA expression plasmid, together with an androgen-regulated LUC reporter gene driven by two AREs in front of the thymidine kinase promoter (pARE 2 tk-LUC). Under these conditions, 10 nM testosterone in culture medium increased LUC activity 10-fold over that without androgen (Fig. 1A). Cotransfection of increasing amounts of RelA expression plasmid (pCMV-p65) brought about a consistent, dose-dependent decrease in androgen-induced transactivation. Already low amounts of pCMV-p65 (0.1 g/10-cm plate) decreased the transactivation by AR over 50%, and higher amounts of RelA (1-2 g/plate) almost completely blocked the effect of testosterone. RelA (1 g/plate) influenced only marginally basal LUC activity in the absence of testosterone (data not shown). Interestingly, NFB1, which is in many aspects homologous to RelA but lacks the transactivation domain (19), was not able to repress AR-mediated reporter gene activation. Comparable results were obtained with other reporter constructs transactivated by the AR, such as pHH-LUC containing the region Ϫ203/ϩ103 of the MMTV promoter (Fig. 1B), the rat probasin gene proximal promoter (Fig. 1C) or pMMTV-CAT ( Fig. 1D), indicating that the effect of RelA on AR function was not a feature peculiar to artificial tk-based promoters. Moreover, the function of an AR expression vector driven by the SV40 promoter (pSG5-rAR) instead of the CMV promoter was repressed by RelA in a similar fashion in CV-1 cells, verifying that transrepression by RelA was not due to inhibition of CMVdriven expression vectors or specific for transformed COS-1 cells (Fig. 1D).
When both RelA and NFB1 (in equal amounts) were transfected together, the effect on AR-mediated transactivation was similar to that of RelA alone (Fig. 1A). The repression by RelA could not be overcome by the addition of excess AR, suggesting that the two proteins competed for coactivator(s) present in limiting amounts in the cell, rather than putative AR-RelA complexes being incapable of transactivation and that the affinity of RelA for this plausible coactivator is higher than that of AR. An alternative explanation is that RelA induces expression of an unknown repressor of AR, even though there is currently no experimental evidence for the presence of such a repressor.
Domains of AR Mediating the Repression by RelA-The effect of RelA on transactivation mediated by a few mutated receptor forms was examined in CV-1 cells to delineate domains of AR involved in the repression. A receptor mutant devoid of the LBD (rAR⌬641-902 mutant) and a form that lacked both LBD and N-terminal amino acids 38 -296 (rAR⌬38 -296/⌬641-902 mutant) were inhibited by cotransfected RelA to a degree similar to that of the native rat AR (Fig. 2), which, in turn, behaved the same way as the human receptor (c.f. Fig. 1). Additional deletion of amino acids from 296 to 408 generated a receptor form (rAR⌬46 -408/⌬641-902 mutant) that retained approximately 50% of the activity of the wild-type rAR in the presence of androgen but that was no longer significantly affected by cotransfected RelA. These results suggest that transrepression of AR by RelA does not involve LBD or most of N-terminal transactivation domain (20), whereas the N-terminal region between residue 297 and the DNA-binding domain is mandatory.
RelA-induced Transactivation Is Inhibited by AR-A B 6 tk-LUC construct containing six B-binding sites in front of the minimal tk promoter was cotransfected along with RelA and increasing amounts of AR expression vector. In the absence of AR, RelA activated reporter gene expression 100-fold over that with an empty expression plasmid (Fig. 3). Cotransfection of increasing amounts of AR attenuated RelA-mediated transactivation, and the ligand-bound AR was always more potent than the aporeceptor (Fig. 3A). A similar inhibition and androgen dependence were observed when 10-fold lower amounts of RelA and AR expression vectors were used (Fig. 3B). Unlike in the case with AR, excess RelA was capable of rescuing the repression to some extent (Fig. 3A). The repression was not limited to a tk-based promoter, because AR was also capable of blunting the RelA-induced activation of the IL-6 promoter (Fig.  3C). Repression of RelA function by AR in the absence of androgen could be due to the fact that the regions of AR needed for the transcriptional interference reside outside the ligandbinding domain, as illustrated by studies with AR mutants (Fig. 2). In addition, a significant portion of immunoreactive AR protein resides in nuclei in transfected COS-1 cells even in the absence of androgen, 4 which may permit interaction of the aporeceptor with other nuclear proteins under these conditions.
Analysis of RelA, AR, and IB␣ Protein Levels-One possible explanation for the above findings is that expressed AR and RelA proteins decrease each other's cellular concentration in COS-1 cells. Overexpression of RelA increased the level of immunoreactive AR, both in whole cell extracts and nuclear extracts (Fig. 4, upper panel, lanes 2 and 3). This is likely due to activation of the CMV promoter by RelA, because CMV promoter-driven ␤-galactosidase activity (from pCMV␤ reporter vector) was also increased 2-fold by overexpression of RelA. Immunoblotting analysis of whole cell and nuclear extracts revealed that immunoreactive RelA protein was not decreased by cotransfected AR (Fig. 4, middle panel, lanes 2 and  3). If anything, the amount of RelA was increased by coexpression of AR, especially in nuclear extracts. Because AR and RelA expression vectors were driven by the CMV promoter and were affected in a similar fashion, RelA responsiveness of these promoters should not complicate interpretation of our data, which is that the mutual transrepression is not due to decreased levels of RelA and AR proteins.
It has been reported that in addition to direct interaction between GR and NF-B, glucocorticoids are also capable of inducing the IB␣ gene (26,27), the product of which is an inhibitor of NF-B. Immunoblotting experiments revealed that ligand-free or ligand-occupied receptors do not influence markedly the concentration of IB␣ protein in COS-1 cells (Fig. 4,  bottom panel, lanes 2 and 3). By contrast, cells transfected with RelA had a substantially elevated IB␣ content, especially in whole cell extracts (Fig. 4, bottom panel, lanes 1-3).
Binding of AR and RelA to Their Cognate DNA Elements-Transrepression of RelA function could derive from its altered DNA binding activity for the NF-B element. Transfection of COS-1 cells with RelA resulted in the formation of a new NF-B element-binding complex that migrated somewhat faster than other main complexes present in mock-or ARtransfected cells (Fig. 5A and B, lanes 1-3). Combined expression of RelA and AR did not result in decreased amount of this complex or other complexes either in whole cell extracts or nuclear extracts (Fig. 5, A and B, lanes 2 and 3; cf. lane 1); rather, cotransfection of AR increased NF-B binding activity in whole cells extracts (Fig. 5A, lanes 2 and 3). The reason for this latter phenomenon is currently unknown. Fig. 5C shows identification of DNA-protein complexes specific for cells transfected with RelA. Several protein complexes were bound to NF-B element in a specific manner in that they were abolished by excess of cold NF-B oligomer but not by an AP-1 oligomer (Fig. 5C). A RelA-specific antibody supershifted only the complex that appeared upon transfection with the RelA expression plasmid (Fig. 5C, lane 10), whereas a nonspecific antibody did not have the same effect (Fig. 5C, lane 9). The nature of the other proteins interacting with NF-B element was not investigated.
We also examined the ability of purified full-length AR and its DBD to interfere with DNA binding of the Rel homology domain of RelA. Incubations were performed under the buffer conditions that we have previously used to demonstrate inhi-bition of c-Jun/AP-1 element interaction by AR (3). Preincubation of increasing amounts of AR or AR-DBD with RelA decreased slightly DNA binding of the latter protein (Fig. 6, lanes [3][4][5][6][7][8]; this decrease was maximally 25% with the highest amount of rAR protein used. The Rel homology domain did not modify binding of full-length AR or AR-DBD to AREs (Fig. 6, lanes 10 -12 and 14 -16), further testifying for a lack of significant interaction between RelA and AR proteins. The same result was obtained when in vitro translated full-length AR and RelA proteins were used (data not shown). It is also worth emphasizing that RelA was not able to bind to a consensus ARE or vice versa; AR did not recognize the NF-B element (Fig. 6,  lanes 2, 9, and 13).
Analysis of Protein-Protein Interaction-A direct interaction between AR and RelA is still a plausible mechanism to explain our findings. To examine whether RelA and AR indeed associate in the absence of specific DNA elements, as suggested for the GR (9 -11), protein-protein affinity chromatography and coimmunoprecipitation experiments were performed. 35 S-Labeled RelA or NFB1 proteins were tested for their ability to bind to purified GST fusion proteins containing various domains of AR. GST-NT, GST-DBD, and GST-LBD were immobilized onto glutathione-Sepharose matrix, incubated separately with labeled RelA and NFB1, and bound proteins were analyzed by polyacrylamide electrophoresis under denaturing conditions. The DNA-binding domain of AR showed weak association with both RelA and NFB1 (Fig. 7), but only low amounts (0.1%) of the input proteins were recovered as complexes. NFB1 that had no effect in cotransfection assays bound somewhat better to GST-DBD than RelA, whereas RelA but not NFB1 showed weak association with GST-NT. No binding of a control protein, 35 S-labeled luciferase, was observed to any of the GST matrices used in these experiments. 5 We could not detect specific association between AR and RelA by coimmunoprecipitation, even in the presence of the protein cross-linking agent dithiobis(succinimidylpropionate). 5 Taken together, these results question the formation of stable complexes between RelA and AR in vitro but do not necessarily exclude the possibility that some AR-RelA interactions occur in intact cells. DISCUSSION We have shown in the present work for the first time that AR-mediated transactivation can be markedly repressed by another transcription factor, RelA. Our data on transrepression between RelA and AR are similar to those recently re-5 P. Reinikainen, and J. J. Palvimo, unpublished observations.  1 and 3-5) or vehicle (lane 2), and whole cell extracts or nuclear extracts were prepared. The total amount of DNA per dish was kept constant by adding empty vector DNA as needed. Immunoblot analyses of whole cell (45 g of protein/lane, corresponding to 7 ϫ 10 4 cells) and nuclear (45 g/lane, corresponding to 3 ϫ 10 5 cells) extracts were performed using ARp3 antibody (␣-ARp3; Refs. 20, 22, and 24), and affinity-purified antibodies against human RelA (␣-RelA, Santa-Cruz A) or IB␣ (␣-IB, Santa Cruz C-15). The anti-RelA and anti-IB blots on nuclear extracts were developed 3-5 times longer than those on whole cell extracts. The specificity of the immunocomplexes obtained with IB␣ antibody were verified by preincubation of the antibody with an excess of the peptide corresponding to the IB␣ antigen (Santa Cruz). The stars depict complexes that were not abolished by excess of IB␣ peptide. FIG. 3. RelA-induced transactivation is repressed by AR. A and B, COS-1 cells were transfected with pB 6 tk-LUC (5 g) and indicated amounts (in g) of RelA and AR expression vectors. C, the IL-6 promoter vector pIL-6(Ϫ226/ϩ11)-LUC was used as the reporter in COS-1 cells. The cells received 18 h later fresh medium with vehicle or 10 nM testosterone (T) as depicted by minus or plus signs. After 30 h of culture, the cells were harvested and reporter gene (LUC) and internal control gene (␤-galactosidase) activities were determined. LUC activities are expressed relative to that of RelA expression plasmid (1 g in A and C, and 0.1 g in panel B) in the absence of AR expression plasmid (100%), and the mean Ϯ S.E. values of three independent experiments are given in percentages. Transactivation by RelA was 100-, 50-, and 30-fold in A, B, and C, respectively. Total amount of DNA per dish was kept constant by adding pCMV DNA as required.
ported for GR and estrogen receptor (9 -11, 21). Although transrepression between AR and RelA was mutual, our results failed to demonstrate the presence of stable stochiometric complexes between the two transcription factors. In comparison with the findings with GR (9 -11), we could detect only a relatively weak association between AR and RelA under our in vitro experimental conditions. Another disparity between AR and GR was that AR did not decrease significantly the binding of RelA to its cognate DNA element (9,10).
In the case of steroid receptor-AP-1 interaction, the exact mechanism of cross-modulation is still poorly understood. Several groups have either failed to detect interaction by coimmunoprecipitation or shown just marginal interactions under in vitro conditions (reviewed in Ref. 5). It is of note that the ligand concentration needed to repress AP-1 function by GR appears to be significantly lower than that required for transactivation, and hormone-independent repression has been reported to occur as well (28,29). The partially ligand-independent repression of RelA by AR observed in this work could be explained, at least in part, by the findings that a significant portion of AR resides in nuclei of transfected cells even in the absence of androgen and that regions other than LBD of AR appear to participate in transrepression.
Direct interaction is not the only mechanism for the GRmediated repression of NF-B, because glucocorticoids also induce expression of the IB␣ gene (26,27), the product of which is an inhibitor of NF-B. Increased levels of IB␣, in turn, reduce the amount of active NF-B that is able to translocate to the nucleus. In contrast to the situation with GR, our results indicate that AR does not affect IB␣ protein level, and therefore, the mechanism of AR-mediated repression should not involve inhibition of RelA translocation to nucleus, a fact that was also verified by immunoblotting experiments. The observation that RelA itself increased IB␣ levels is in accordance with the recent reports showing the existence of an autoregulatory loop between RelA and IB␣ (30,31).
Although RelA and NFB1 share a conserved Rel homology domain needed for DNA binding and dimerization, only RelA was capable of repressing AR-mediated gene activation. The C-terminal parts of RelA and NFB1 differ significantly from each other (7,8,19). RelA contains a potent activation domain in its C terminus, which enables it to transactivate as both homo-and heterodimers, whereas NFB1 is devoid of transcription activation domain and thus inactive as a homodimer.  Fig. 4. A, aliquots of whole cell extracts (15 g) were incubated with 32 P-labeled NF-B element oligomer in the presence or the absence of 100-fold molar excess of competing cold oligomer as depicted by plus and minus signs (verifying the specificity of DNA-protein complexes) and electrophoresed on 4% polyacrylamide gels under the conditions described previously (24). B, aliquots of nuclear extracts (10 g). Lane 1, RelA alone in the presence of testosterone; lane 2, RelA and hAR in the absence of testosterone; lane 3, RelA and hAR in the presence of testosterone; lane 4, hAR alone in the presence of testosterone; and lane 5, empty pCMV vector. C, identification of DNA-protein complexes specific for RelA-transfected cells. Lanes 1-5, nuclear proteins from cells transfected with pCMV-hAR; lanes 6 -10, nuclear proteins from cells transfected with the RelA expression vector pCMV-p65. The extracts (10 g of protein) were incubated in the presence or the absence of 100-fold molar excess of AP-1 or NF-B oligomers or RelA-specific antibody (0.5 g) or control IgG as depicted by plus and minus signs. The transactivation domain of RelA has been recently shown to contact some general transcription factors, such as TFIIB and TATA-binding protein, and functionally interact with them (32,33). However, overexpression of TATA-binding protein in COS-1 cells does not affect androgen-induced transactivation or rescue the repression of AR by RelA. 5 In addition to the general transcription factors, coactivators such as PC1 have been reported to activate RelA-dependent transcription (33). In view of this, it tempting to speculate that mutual repression of transactivation by AR and RelA is due to their competition for PC1-like coactivators that are utilized by both transcription activators. Alternatively, a coactivator may mediate or stabilize complex formation between RelA and AR in intact cells. This latter alternative is supported by our findings (Fig. 3B) that mutual transcriptional interference also occurred when 10-fold lower amounts of RelA and AR expression plasmids were used in transfections.
Our findings offer a mechanistic explanation for the ARmediated down-regulation of IL-6 gene expression (14), which is inhibition of transcriptional activation of this gene by NF-B. In the case of estrogen receptor, transcription factor C/EBP␤ is also involved in the repression of the IL-6 gene (21). It remains to be elucidated whether C/EBP␤ plays a role in the transrepression of this gene by AR. NF-B has been implicated in the negative regulation of the rat AR gene promoter (15). As a consequence, the age-dependent increase in NF-B activity could, at least in part, explain androgen desensitization of the liver of aged rats owing to a decreased receptor gene transcription (15). Our results offer an alternative mechanism for attenuated androgen responsiveness with increasing age, in that age-related elevation in RelA (NF-B) content could interfere with AR-mediated transactivation events at large, rather than being specific for the AR gene promoter. Finally, the results from this work add another piece of evidence to the notion that down-regulation of gene expression by the androgen receptor occurs via mechanisms other than those involving "negative" DNA elements.