Differential Modulation of Androgen Receptor Action by ...

0888-8809/03/$15.00/0 Printed in U.S.A.

Molecular Endocrinology 17(9):1738–1750 Copyright © 2003 by The Endocrine Society doi: 10.1210/me.2002-0379

Differential Modulation of Androgen Receptor Action by Deoxyribonucleic Acid Response Elements CHRISTOPH GESERICK, HELLMUTH-ALEXANDER MEYER, KARINA BARBULESCU, BERNARD HAENDLER

AND

Research Laboratories of Schering AG, D-13342 Berlin, Germany In addition to the steroid response elements (SREs), which are recognized by several steroid receptors, a second class of DNA elements exhibiting selectivity for the androgen receptor (AR) and named androgen response elements (AREs) has been identified. Here we provide evidence for the differential role of these element classes in modulating AR function. AR complexes attached to response elements representative of each class were purified. Limited protease digests of ARE- or SRE-bound AR complexes led to the generation of different patterns, in line with differential accessibilities. In transactivation assays, mutations in the AR dimerization interface of the DNA-binding domain had various effects, depending on the response elements tested. The R598D mutant displayed much enhanced activity on SREs, whereas far less effect was seen on the selective AREs. The A596T mutant had reduced activity on AREs but not on SREs. Ectopic expression of the coactivators transcriptional intermediary factor 2 (TIF2) and

ARA55 stimulated AR activity to different extents, depending on the response element. When using cysteine-rich secretory protein 1 (CRISP-1) SRE as reference, the most significant difference was observed with Pem ARE-2. A differential response of each element class was furthermore observed in the presence of two enzymes involved in the sumoylation pathway. Ubiquitin-conjugating enzyme 9 (Ubc9) overexpression enhanced AR action conveyed by SREs, whereas little effect was seen on Pem ARE-1 and repression on Pem ARE-2. Protein inhibitor of activated STAT (PIAS)x␣ overexpression had little influence on SRE-mediated AR activity but was repressive when using AREs. Altogether, these results demonstrate that DNA response elements play an important modulatory role in transmitting AR action and may be determinative for specificity of gene expression in cell or tissue types. (Molecular Endocrinology 17: 1738–1750, 2003)

S

elucidated. Examination of natural, functional SREs shows that most do not exactly fit the consensus, suggesting that individual sequence variations may be important for specificity of gene control. Recently, DNA response elements mediating a preferential response to androgen stimulation have been described in some androgen-regulated genes including probasin and Pem (4–6). Unlike the SREs, which form imperfect inverted repeats of the 5⬘-TGTTCT-3⬘ half-site, these selective androgen response elements (AREs) have direct repeat features (4–6). Mutational analyses indicate that AREs, but not SREs, are recognized by the AR in an assymetrical mode (4, 5). Furthermore, domain-swapping experiments between the AR and GR demonstrate that the second zinc finger and the C-terminal extension of the hinge region are essential for recognition of selective elements (7). Altogether, this suggests a different mode of interaction between the AR and the two classes of response elements, which might be instrumental for the preferential recruitment of cofactors and hence the selective stimulation of target genes. To analyze this, we studied the implications of the interaction between the AR and the two response element classes by different approaches. After purifi-

TEROID RECEPTORS ARE ligand-dependent transcription factors organized in distinct functional domains that can communicate in an elaborate fashion (1–3). They exist in different states, either complexed to chaperone proteins or activated by ligand and bound to DNA response elements. Two different consensus response elements for steroid receptors have been identified (1–3). The steroid response element (SRE) with the consensus sequence 5⬘-GGTACAnnnTGTTCT-3⬘ is bound by the androgen receptor (AR), glucocorticoid receptor (GR), progesterone receptor, and mineralocorticoid receptor, whereas the estrogen response element with the sequence 5⬘AGGTCAnnnTGACCT-3⬘ is only recognized by the estrogen receptor (ER). How the promiscuous SRE differentially transmits cues originating from steroid receptors with partially overlapping expression patterns but very different functions has so far not been Abbreviations: AR, Androgen receptor; ARE, androgen response element; CRISP-1, cysteine-rich secretory protein 1; DBD, DNA-binding domain; DTT, dithiothreitol; ER, estrogen receptor; GR, glucocorticoid receptor; PIAS, protein inhibitor of activated STAT; SRE, steroid response element; TIF2, transcriptional intermediary factor 2; Ubc9, ubiquitin-conjugating enzyme 9.

1738

Geserick et al. • Modulation of AR Action by Response Elements

cation of DNA-bound complexes, we probed the accessibility of the AR to proteases, as an indicator of the protein conformation induced by response elements. Furthermore, we determined the importance of the dimerization interface of the AR DNA-binding domain (DBD) for transactivation potential and compared the regulatory effects of different cofactors in the presence of representative response elements. The results strongly suggest that AR conformation varies depending on the bound response element, thereby altering the role of the dimerization interface and adjusting the response to cofactors. These results underscore the importance of DNA elements as modulators of AR function.

RESULTS

Mol Endocrinol, September 2003, 17(9):1738–1750

1739

to the context, we exchanged both response elements for the 2.43 motif, a sequence identified by virtue of its high binding affinity for the AR (8) and exhibiting preferential glucocorticoid response (see Fig. 2 below). We found that the mutated Pem promoter was far less responsive to androgen treatment (down to 4-fold) and more sensitive to glucocorticoid treatment (up to 4-fold), in comparison to the wild-type form (Fig. 1). Indeed, the responses of the mutated Pem promoter to androgen and glucocorticoid were comparable, indicating a complete loss of androgen selectivity. For a more detailed analysis of the mechanisms underlying preferential androgen stimulation and to avoid interference of other promoter components, we analyzed minimal reporter constructs containing the Pem elements ARE-1 or ARE-2, the ⫺1253 SRE found in the promoter of the strongly androgen-regulated cysteine-rich secretory protein 1 (CRISP-1) gene (9),

DNA Elements Specify Preferential Transcriptional Response to the AR or GR Transactivation assays were performed with the murine Pem proximal promoter linked to a luciferase reporter gene. We used the Green Monkey kidney CV-1 cells because they do not express steroid receptors. In the presence of the adequate receptor, the Pem promoter exhibited a preferential response to androgen as opposed to glucocorticoid treatment (11-fold vs. 2-fold), in stark contrast to the mouse mammary tumor virus promoter (8-fold vs. 29-fold) (Fig. 1). Two response elements, named ARE-1 and ARE-2, which mediate AR response, have been identified in the Pem promoter (5). To verify that the selectivity was not due

Fig. 1. Selective Androgen Response of the Murine Pem Promoter Reporter plasmids harboring the wild-type or mutated Pem proximal promoter were cotransfected with expression vectors for the AR or GR into CV-1 cells. The mouse mammary tumor virus (MMTV) promoter was used as control. The luciferase activity [relative light units (R.L.U.)] measured before (white bars) and after treatment with R1881 (black bars) or dexamethasone (gray bars) is given. The results are a representative of three separate experiments. Values are the means ⫾ SD of sextuplicate values.

Fig. 2. Selectivity of DNA Response Elements for Steroid Receptors A, Reporter plasmids carrying two copies of the indicated DNA elements were cotransfected with expression vectors for the AR or GR into CV-1. B, Reporter plasmids carrying two copies of the indicated DNA elements were cotransfected with expression vectors for the AR or GR into PC-3 cells. The luciferase activity [relative light units (R.L.U.)] measured before (white bars) and after treatment with R1881 (black bars) or dexamethasone (gray bars) is given. The results are a representative of three separate experiments. Values are the means ⫾ SD of sextuplicate values.

1740 Mol Endocrinol, September 2003, 17(9):1738–1750

the 2.43 SRE described above, or the consensus SRE identified by in vitro selection based on maximal AR binding (10). The constructs contained two copies of each element separated by the identical spacing sequence and at the same distance from the minimal promoter region, to exclude positional effects. When transfecting CV-1 cells with an AR expression vector and stimulating them with R1881, we found Pem ARE-1 and Pem ARE-2 to transmit a much higher response than the CRISP-1 SRE, 2.43 SRE, or the consensus SRE (Fig. 2A). In the presence of GR and dexamethasone, the CRISP-1 SRE, 2.43 SRE, and consensus SRE were the best responders (Fig. 2A). When considering each DNA element separately, it appeared that Pem ARE-1 and Pem ARE-2 displayed a highly selective response to the AR (7- and 4-fold selectivity compared with the response to GR, respectively), whereas the CRISP-1 SRE, 2.43 SRE, and consensus SRE were preferentially stimulated by the GR. To verify that this selectivity was not unique to CV-1 cells, we performed similar experiments in the human prostate adenocarcinoma cell line PC-3, which contains no endogenous AR or GR. The different reporter constructs were introduced into the cells together with an AR-expressing plasmid. After androgen treatment, a stronger response of Pem ARE-1- and ARE-2-driven plasmids than of CRISP-1 SRE-, 2.43 SRE-, or consensus SRE-containing plasmids was observed (Fig. 2B). The opposite situation was seen after transfection of a GR-expressing plasmid and glucocorticoid treatment (Fig. 2B). The results show that small variations in the DNA sequence of response elements lead to important differences in AR effects. They also allow the definition of two classes of elements: those exhibiting a selective response to androgen stimulation when compared with glucocorticoid treatment (Pem ARE-1 and Pem ARE-2), and those with the opposite pattern (CRISP-1 SRE, 2.43 SRE, and consensus SRE). Distinct AR Protease Digestion Patterns Result from Binding to Different Response Elements To determine whether the two response element classes influenced the conformation of bound AR, we purified complexes formed of DNA elements and associated proteins (Fig. 3A). Biotinylated doublestranded oligonucleotides containing the various response elements as tandem repeats were bound to streptavidin sepharose. Aliquots were incubated with nuclear extracts prepared from androgen-treated PC3/AR cells that stably express the AR. After thorough washing to remove nonspecifically attached proteins, the captured AR-protein complexes were eluted from the DNA using high-salt buffer and precipitated with trichloroacetic acid. The samples were resuspended, separated on sodium dodecyl sulfate gel, and analyzed by Coomassie staining (not shown). The pattern of the major protein bands was highly similar for all the response elements, strongly suggesting that no sys-


Fig. 3. Isolation of Response Element-Protein Complexes A, Schematic representation of the strategy used to purify response element-AR protein complexes. B, Determination of AR content. The amount of AR bound to different response elements was determined using a specific antibody directed against the N-terminal domain. The results are a representative of three separate experiments.

tematic bias due to nonspecific binding to a given DNA sequence had taken place. Western blot analysis using an antibody directed against the AR N-terminal domain was performed (Fig. 3B). Starting from identical amounts of nuclear extract, we found that significantly less AR was associated with the two selective than with the three nonselective elements, in line with our previous EMSAs using in vitro translated AR (5). To test whether DNA elements had an influence on the conformation of bound AR, we probed the accessibility by protease digestions using chymotrypsin or trypsin. The purified complexes were separated on a denaturing gel, and the bands corresponding to digested AR were detected by an antibody recognizing the C-terminal moiety. As observed above, more AR was associated with the SREs than with the AREs in the undigested complexes (Fig. 4, A and B, 0 ng/␮l protease, lanes 1 and 2 compared with lanes 3 and 4). Some digestion products closely reproduced this pattern so that a differential sensitivity to proteases could not necessarily be inferred. There were instances, however, where more product was formed from the ARE complex than from the SRE complex. Using chymotrypsin at a concentration of 0.2 ng/␮l, a rapid degradation of AR associated with the selective elements was apparent for band C1 (Fig. 4A, lanes 3 and 4), whereas only a very weak band corresponding to intact AR remained (band AR). Considering that more AR tethered to the nonselective elements was originally present, the product C1 was proportionally less represented than in the samples where ARE-bound complexes were digested (lanes 3 and 4). Also, the fact that a band of lower molecular weight (C2) exhib-



1741

Mutations in the AR Dimerization Box Differentially Affect Transactivation Transmitted by Different Response Elements

Fig. 4. Protease Digestion Patterns of AR Attached to DNA Elements Response element-protein complexes were purified from nuclear extracts prepared from PC-3/AR cells and digested using the indicated amounts of protease. A, Chymotrypsin digest. B, Trypsin digest. Total protein concentration was identical for all elements. Lane 1, Consensus SRE; lane 2, 2.43 SRE; lane 3, Pem ARE-2; lane 4, Pem ARE-1. The AR digestion patterns were visualized using a specific antibody directed against the C-terminal end of the ligand-binding domain. The digestion products discussed in the text are indicated (C1, C2, T1, T2). The results are a representative of three separate experiments.

ited a pattern opposite to that of C1 and comparable to that of undigested AR indicated that the ARE-bound complexes were not altogether more accessible to chymotrypsin. When trypsin was used at a concentration of 2 ng/␮l, a prominent digestion product labeled T1 was generated from selective element-bound complexes (Fig. 4B, lanes 3 and 4) but was less apparent for the nonselective elements (lanes 1 and 2). The opposite pattern was seen for a digestion product with a lower molecular weight (band T2). This band was stronger for the nonselective than for the selective elements (lanes 1 and 2 compared with lanes 3 and 4). The inverted ratios seen for the bands AR and C2 compared with C1, and T1 compared with T2, indicate that differences in protease sensitivity exist for the AR, depending on the bound DNA motif. This suggests element-induced alterations in AR conformation, in the set of associated cofactors, or in both.

AR-selective DNA elements have an opposite half-site orientation compared with nonselective DNA elements and are recognized in an asymmetrical way by the AR (4, 5). These observations and the different protease digestion patterns generated in the present study suggest that the AR homodimer adopts a different conformation, depending on the bound element class. This, in turn, may impinge on the role of domains directly implicated in receptor dimerization. The second zinc finger and especially its D box have been shown to partake in the dimerization interface of steroid receptors (11). From the crystal structure of the GR DBD and from mutagenesis experiments, it is known that two charged amino acids of the D box are engaged in intermolecular salt bridges (11). Mutation for a residue of the opposite charge disrupts this interface, leading to a dramatic loss of activity on a single DNA element but, surprisingly, to an enhanced activity on multiple elements (12). The equivalent mutations R598D and D600R were introduced into the AR (Fig. 5). The activity of each mutant was compared with that of wild-type AR after transfection into CV-1 cells and using reporter plasmids harboring the different response elements as tandem repeats (Fig. 6A). AR D600R behaved similarly to the wild-type form for transactivation through Pem ARE-1 and Pem ARE-2, but was twice as active in the presence of CRISP-1 SRE and 2.43 SRE (Fig. 6B). With AR R598D, the effects were much more pronounced. Here, a doubling of activity was seen in the presence of the Pem elements and an 8- to 9-fold increase for the two other elements, when compared with wild-type AR (Fig. 6B). Western blot analysis showed that the expression levels of wild-type AR and AR mutants were comparable

Fig. 5. Schematic Drawing of AR Homodimer Salt Bridges and Localization of Point Mutations Only the AR DBD of each monomer is drawn. Potential salt bridges are indicated with dotted lines. Positions of the A596T, R598D, and D600R mutations are shown.



Fig. 6. Role of Intermolecular Salt Bridges in AR Transactivation Mediated by Selective or Nonselective DNA Elements A, Reporter plasmids carrying two copies of the indicated elements were cotransfected into CV-1 cells together with an expression plasmid for wild-type AR (wt AR), AR R598D, or AR D600R. The luciferase activity measured in the absence (white bars) or presence of androgen (black or striped bars for wild-type or mutant AR, respectively) is given. The results are representative of three separate experiments, and the bars are the mean ⫾ SD of sextuplicate values. B, Fold inductions calculated for wild-type and mutant AR in the presence of androgen. C, Western blot analysis of transfected CV-1 cells using an AR-specific antibody directed against the N-terminal domain. R.L.U., Relative light units.

after transfection of the corresponding expression plasmids into CV-1 cells (Fig. 6C). Another D box residue implicated in AR dimerization is A596 (Fig. 5). Its mutation to threonine has been discovered in patients suffering from Reifenstein syndrome, a disease characterized by partial androgen insensitivity (13). The activity of AR A596T mediated by the different elements present as tandem repeats was determined (Fig. 7A). For the Pem elements, we measured a decreased activity of AR A596T in comparison to wild-type AR. In contrast, we saw little change when testing the CRISP-1 SRE and more than a 5-fold increase for the 2.43 SRE (Fig. 7A). Western blot analysis indicated the level of AR A596T to be lower than that of wild-type AR (Fig. 7B), so that the enhancement observed might be an underestimation.

Altogether, these results demonstrate that disruption of the DBD dimerization box is still compatible with AR function and that in the presence of some SREs it even leads to enhanced activity. They also show the role of the AR dimerization interface to vary considerably, depending on the bound response element. The Coactivators Transcriptional Intermediary Factor 2 (TIF2) and ARA55 Differentially Modulate AR Activity Depending on the Response Element In the next step we examined whether the different conformations elicited by binding to DNA elements affected AR response to cofactors. TIF2 and ARA55 stimulate the ligand-dependent activity of the AR and



1743

Fig. 7. Impact of the Reifenstein Mutation on AR Transactivation Mediated by Selective or Nonselective DNA Elements A, Reporter plasmids carrying two copies of the indicated elements were cotransfected into CV-1 cells together with an expression plasmid for wild-type AR (wt AR) or AR A596T. The luciferase activity measured in the absence (white bars) or presence of androgen (black or striped bars for wild-type or mutant AR, respectively) is given. The results are a representative of three separate experiments, and the bars are the mean ⫾ SD of sextuplicate values. B, Western blot analysis of transfected CV-1 cells using an AR-specific antibody directed against the N-terminal domain. R.L.U., Relative light units.

of other steroid receptors (6, 14). CV-1 cells were cotransfected with reporter plasmids containing two copies of the different DNA elements and expression plasmids for the AR and for the coactivators TIF2 or ARA55. In the control experiments, the same amount of an expression vector expressing ␤-galactosidase, rather than the cofactor, was added. For both cofactors and for all elements tested, enhancement of androgen-driven transcriptional activity was noted after overexpression (Fig. 8). A peak at 0.1 nM followed by a small decline of stimulation was observed for TIF2 (Fig. 8, left column). This was not the case for ARA55, where an increase of stimulation followed by a plateau effect was noted (Fig. 8, right column). In absolute terms, cofactor overexpression led to stronger AR action through AREs than through SREs, parallelling the activity seen without added cofactor. When analyzing the effects elicited by TIF2 or ARA55 via each DNA element, we found that, in comparison to the natural CRISP-1 SRE, the increase in response mediated by Pem ARE-2 was less at all concentrations tested. For Pem ARE-1, highly significant differences were only noted at low androgen levels. Conversely, the cofactor effects mediated by the 2.43 SRE did not differ significantly from those mediated by the CRISP-1 SRE. Thus, both TIF2 and ARA55 enhance ligand-dependent AR activity through all DNA elements, but not to the same extent. The significant differences observed

between Pem ARE-2- and CRISP-1 SRE-mediated effects support the notion that cofactor action is affected by the modulatory effects of DNA elements on AR conformation. Sumoylation Enzymes Modulate AR Function in a DNA Element-Dependent Fashion Sumoylation of AR has been reported to either stimulate or repress AR activity, depending on the experimental setting (15, 16). Both ubiquitin-conjugating enzyme 9 (Ubc9) and the PIAS (protein inhibitor of activated STAT) family are involved in this process (15–18). Using cotransfection experiments as above, we analyzed the effect of ectopic Ubc9 or PIASx␣ expression on AR function mediated by various response elements. When overexpressing Ubc9, AR action transmitted by SREs, but not by AREs, was stimulated at all hormone concentrations from 0.01 nM onward. Using CRISP-1 SRE as reference, we found that both Pem AREs, but not the 2.43 SRE, interpreted Ubc9 overexpression in a significantly different way (Fig. 9, left column). Thus, AREs and SREs fall into two classes with regard to transmission of Ubc9 effects. When overexpressing PIASx␣, virtually no effect on AR stimulation was visible at any concentration for the nonselective elements. Conversely, a marked inhibitory effect on AR activity mediated by the selective



Fig. 8. Effects of TIF2 and ARA 55 Overexpression on AR Function Mediated by Selective or Nonselective DNA Elements CV-1 cells were cotransfected with reporter plasmids carrying two copies of Pem ARE-1, Pem ARE-2, 2.43 SRE, or CRISP-1 SRE and an AR expression vector. The concentrations of R1881 used for treatment are indicated. In the control curve the fold inductions obtained in presence of R1881 and of a neutral plasmid encoding ␤-galactosidase (␤-Gal) are indicated with white diamonds. In the cofactor curve the inductions seen in the presence of R1881 and of an expression plasmid for TIF2 (left column) or ARA55 (right column) are shown with black dots. Each point is the mean ⫾ SD of two independent experiments in which sextuplicate values were determined. Statistical analysis was performed using ANOVA and t tests to compare cofactor effects for a given hormone dosis between 0.1 and 100 nM, using the CRISP-1 SRE as reference. A P value keeping the global significance level of 0.05 is marked by *, and a P value keeping the global significance level of 0.01 is marked by **. A detailed interpretation based on the graph is given in Results.



1745

Fig. 9. Effects of Ubc9 and PIASx␣ Overexpression on AR Function Mediated by Selective or Nonselective DNA Elements CV-1 cells were cotransfected with reporter plasmids carrying two copies of Pem ARE-1, Pem ARE-2, 2.43 SRE, or CRISP-1 SRE and an AR expression vector. The concentrations of R1881 used for treatment are indicated. In the control curve, the fold inductions obtained in presence of R1881 and of a neutral plasmid encoding ␤-galactosidase (␤-Gal) are indicated with white diamonds. In the cofactor curve the inductions seen in the presence of R1881 and of an expression plasmid for Ubc9 (left column) or PIASx␣ (right column) are shown with black dots. Each point is the mean ⫾ SD of two independent experiments in which sextuplicate values were determined. Statistical analysis was performed using ANOVA and t tests to compare cofactor effects for a given hormone dosis between 0.1 and 100 nM, using the CRISP-1 SRE as reference. A P value keeping the global significance level of 0.05 is marked by *, and a P value keeping the global significance level of 0.01 is marked by **. A detailed interpretation based on the graph is given in Results.


AREs was observed at 0.1 nM and higher androgen concentrations (Fig. 9, right column). Statistical analysis confirmed the significance of these results, allowing again the distinction between the ARE and the SRE classes. Altogether, these data show the effects of Ubc9 and PIASx␣ on AR function to greatly vary, depending on the response elements tested. Ubc9 enhances AR activity transmitted by nonselective elements but not by selective elements. Conversely, PIASx␣ hardly affects AR activity conveyed by nonselective elements but significantly represses it in the presence of selective ones. It can therefore be concluded that the modulatory effects of two enzymes involved in the sumoylation pathway are differently interpreted by each class of DNA response elements.

DISCUSSION In the present study we have demonstrated that DNA elements do not merely position the AR receptor on the promoter but represent, in addition to the ligand and the receptor-associated proteins, other interacting partners with important regulatory functions. The selective AREs and the nonselective SREs represent two classes of response elements recognized by the AR for stimulation of target genes. Using mutational analysis, we have previously shown that, with regard to AR binding, the importance of the DNA half-sites differs for both element classes (5). Here we present evidence for the important part played by the two element classes in modulating AR action. The Pem AREs mediated a higher androgen response than the SREs. Even though the AR affinities for the different element classes were not quantified, our data suggest that AR binding and transactivation potential do not necessarily correlate. Altered intraand interdomain interactions as well as differential cofactor recruitment may be instrumental in modulating AR function, and a model in which DNA elements impart conformational modifications onto bound AR leading to such novel interactions can be postulated. Very recent results documenting the differential importance of the amino/carboxyl-terminal interaction in AR-mediated transactivation via specific and nonspecific response elements are in line with this (19). It should be noted, however, that a competitive effect of AR-binding factors or of other proteins present in the complexes cannot be excluded at this point. Limited protease digests were performed on DNA-AR complexes purified from PC-3/AR nuclear extracts. This had the advantage of keeping the AR in its native conformation, bound to DNA and associated with cofactors. We found differences in the sensitivity to chymotrypsin and trypsin depending on whether the AR was attached to AREs or SREs. The chymotrypsin digestion experiments showed a differential accessibility of ARE-associated AR than of SRE-bound AR.


Similar results were obtained after trypsin treatment. The cleavage sites were difficult to localize precisely, but all were in the AR N-terminal domain. The C1 band was generated by means of cutting near the N-terminal end, and the T1 and T2 bands were generated by cutting close to the C-terminal end of the AR N-terminal domain. Even though an effect linked to differential association and dissociation at beads carrying the various oligonucleotides cannot be ruled out, our results suggest that the DNA elements induced changes outside of the DBD and that interdomain communication existed in the AR. Another, non-mutually-exclusive possibility is that different coregulator sets were recruited by the AR tethered to different response elements, thereby altering protease accessibility. Recent data indicate that estrogen response element sequences may also affect the conformation of bound ER (20, 21). The AR D box is part of the dimerization interface. Disruption of salt bridges by mutating specific residues in this region of the GR or mineralocorticoid receptor leads to enhanced activity on multiple SREs, indicating a role in restraining transcriptional synergy rather than in stabilizing the receptor homodimer (12, 22). We found that the equivalent mutations R598D and D600R increased AR activity mainly on the nonselective SREs. Far fewer (R598D) or no effects (D600R) were observed when testing the selective Pem AREs. This documents that self-synergy varies depending on response elements and opens up the possibility that control of AR activity may be achieved at this level. Another residue of the AR D box likely to play a role in receptor dimerization is alanine at position 596. The A596T mutation has been identified in the AR of patients suffering from Reifenstein syndrome, a form of partial androgen insensitivity (13). Due to its inability to bind to promoters containing a single response element, the mutant receptor cannot transactivate. It is fully active, however, on promoters with multiple response elements in vitro (13). We found that AR A596T possessed a reduced transactivation potential toward the selective AREs. The situation was not the same for the SREs. Little change was noted for the CRISP-1 SRE and a much enhanced activity for the 2.43 SRE. This indicates that, here again, disruption of the dimerization interface had very different effects on AR function, depending on the response elements analyzed. Even though the results were biased by the lower expression of the mutant AR as compared with the wild-type form, it can be concluded that the selective AREs behaved similarly. It is also noteworthy that all D box mutations we tested had the strongest stimulating effects on transactivation potential mediated by the 2.43 SRE, which was originally identified as the optimal AR binding sequence using an in vitro selection procedure (8). Conversely, the weak AR binders generally responded less, or even negatively, to disruption of the dimerization interface. Altogether, the results show the effects of the mutations to be context dependent and are in line with an influence of DNA elements on the contacts formed


within the dimer interface. Previous studies on the ER and GR have shown this interface to form only after DNA binding, implying that its conformation is not fixed beforehand (23, 24). By affecting the dimerization interface, DNA elements might furthermore allosterically influence the position and activity of the synergy control motifs described in the AR N-terminal domain (25). These motifs are operative only when the dimerization interface is intact and have recently been shown to harbor sites for sumoylation (18). The changes in AR conformation brought about by DNA elements might, in turn, affect the recruitment and activity of coregulatory proteins. TIF2 and ARA55 have been identified as potent AR coactivators (6, 14). We found their ectopic expression to enhance AR activity in transactivation assays. The peak noted at 0.1 nM for TIF2 overexpression showed the role of this coactivator to be more important at low than at high ligand concentrations. In absolute terms, the effects of both cofactors were much more pronounced on the AREs than on the SREs, with Pem ARE-1 exhibiting the strongest response to overexpression of both TIF2 and ARA55. Interestingly, the cofactor-enhanced response of the SREs attained about the levels reached by AREs without added cofactors. This demonstrates that the respective strengths of response elements may be dramatically influenced by cofactors. It should also be noted that despite the high overall induction, the relative enhancement of androgen effects by TIF2 and ARA55 was significantly less in the presence of Pem ARE-2 than for the other response elements. The fact that the effects of cofactors known to associate with the AR N-terminal region or ligand-binding domain are affected by the response element recognized by the DBD indicates that, indeed, information originating from DNA elements can be transmitted to distal domains. Since the contact interface between nuclear receptors and cofactors is generally small (26), it is conceivable that even limited conformational changes affect recruitment of regulatory proteins. An influence of DNA elements on TIF2 association with the ER (27, 28) as well as on SRC-1 interaction with the thyroid hormone receptor (29) have already been reported. A function of DNA elements in inducing changes in receptor conformation leading to selective cofactor recruitment may thus represent a general mechanism for control of gene expression. Posttranslational modifications are emerging as important regulatory steps in AR function. Among these, the process of sumoylation has been shown to dramatically repress AR transactivation (18). Recent data indicate that Ubc9 and the PIAS proteins are implicated in AR sumoylation as E2 conjugase and E3 ligase, respectively (15, 17). We found that the effects of Ubc9 differed, depending on the class of DNA elements tested. A parallel enhancement of AR response was seen for the nonselective SREs. For the selective AREs, little effect or even repression was observed. Concerning PIASx␣ cotransfection, we noted no effect for the nonselective SREs but a dramatic repression of


1747

AR activity when the selective AREs were tested. Altogether, our data show that the response to cofactors implicated in protein sumoylation is different for the nonselective SREs and the selective AREs. The parallel shift toward repression of AR action observed for a given element when comparing the effects of Ubc9 and PIASx␣ is striking. A possible explanation is that in the sumoylation pathway, Ubc9 acts upstream of different PIAS proteins, some with repressing and some with stimulating effects on AR activity, indicating the existence of a regulatory network (15, 16, 30, 31). In addition, sumoylation-independent effects of Ubc9 on AR activity have also been described (17) so that the effects of both enzymes need not necessarily be identical. Finally, sumoylation affects the function of many proteins, including TIF2, leading to complex effects within the cell (15). The AR sumoylation sites have recently been identified within the synergy control motifs present in the N-terminal region. Mutation of these motifs has effects similar to mutations of the DBD dimer interface on AR activation, suggesting a functional link between both domains (18). The recent discovery that ablation of the AR sumoylation sites leads to loss of recruitment of the corepressor silencing mediator of retinoid and thyroid hormone receptor in the presence of an antiandrogen might explain, at least in part, how control of self-synergy is achieved (32). Finally, the observation that sumoylation also has repressive effects on progesterone receptor and GR function (33, 34) suggests that this posttranslational modification represents a general mechanism for regulation of steroid receptor function. Our finding that DNA elements differentially interpret the effects of two sumoylating enzymes adds another level of complexity in transcriptional control. In summary, we have shown that DNA elements convey information to bound AR and act as regulators affecting the effects of cofactors and protein-modifying coregulators. As proposed for other transcription factors (35), allosteric modification induced by DNA elements may represent an essential mechanism used to control the activity of individual genes, and future studies will show whether modulation of transcription can be achieved at this level.

MATERIALS AND METHODS Chemicals and Cell Culture Media R1881 (methyltrienolone) was from Dupont NEN (Boston, MA) and dexamethasone was from Sigma (St. Louis, MO). RPMI 1640, MEM, OPTI-MEM, streptomycin, penicillin, Geneticin, and L-glutamine were obtained from Life Technologies (Gaithersburg, MD). Fetal calf serum was from PAA Laboratories (Linz, Austria). The oligonucleotides were purchased from Roth (Karlsruhe, Germany). Trypsin, chymotrypsin, Pefabloc SC, Complete Protease Inhibitor Cocktail tablets, and FuGene 6 were from Roche Molecular Biochemicals (Indianapolis, IN).


Plasmids


Preparation of DNA-Protein Complexes and Protease Digestions

Expression vectors containing cDNAs for full-length AR, GR, TIF2, ARA55, Ubc9, and PIASx␣ inserted in the pSG5 plasmid (Stratagene, La Jolla, CA) were generated and used for transfections. Site-directed mutagenesis of pSG5-AR was carried out using the QuikChange kit (Stratagene) and the heat-stable polymerase mix from the Advantage kit (CLONTECH, Palo Alto, CA). The luciferase reporter plasmids were based on the pGL3-Basic plasmid (Promega, Madison, WI). The murine Pem proximal promoter construct has been described (5). Replacement of both AREs by the 2.43 SREs was performed by site-directed mutagenesis using PCR fragments as megaprimers (36). Briefly the primers 5⬘-CATGAACTGTGTCCATCTTGCAGGAACAAAATGTCCCTTACATCCCCAAACTGCTATCAC-3⬘ and 5⬘-GGGCCCCTGGTGTCCCCGGGGACATTTTGTTCCGTGATGCTCAACCTGG- CAGGG-3⬘ were used to amplify a 221-bp long fragment of the Pem promoter flanked by the novel 2.43 SRE sequences (underlined). After purification on gel, this amplification product was mixed with the Pem promoter luciferase plasmid for insertion mutagenesis using the QuikChange kit, as described (36). In the minimal reporter constructs two copies of the DNA response elements Pem ARE-1 (5), Pem ARE-2 (5), CRISP-1–1253 ARE, hereafter named CRISP-1 SRE (9), 2.43 SRE (8), or of the consensus high-binder SRE element (10) were placed upstream of the thymidine kinase minimal promoter and of the luciferase reporter gene (Table 1). Spacing and sequence of the regions between the response elements and upstream of the minimal promoter were identical in all cases. DNA sequencing was performed using standard procedures. Preparation of Nuclear Extracts PC-3/AR cells were grown as described (5) and induced with 100 nM R1881 for 24 h before harvesting. The nuclear extracts were prepared essentially as described (12). Approximately 108 cells were treated with lysis buffer [20 mM HEPES (pH 7.6), 10% glycerol, 1.5 mM MgCl2, 10 mM NaCl, 0.1% Triton X-100, 2.5 mM dithiothreitol (DTT), 0.2 mM R1881, and protease inhibitor tablets as required]. They were then centrifuged for 4 min at 13,000 ⫻ g, and the pelleted nuclei were resuspended in lysis buffer containing 600 mM NaCl and 10 mM MgCl2, and gently vortexed for 30 min at 4 C. Undissolved material was removed by centrifugation. The nuclear extract was loaded on a NAP25 column (Pharmacia Biotech, Piscataway, NJ) equilibrated with binding buffer [20 mM HEPES (pH 7.5), 0.5 mM EDTA, 2.5 mM Mg acetate, 130 mM NaCl, 0.1% Nonidet P-40, 10% glycerin, 2 mM DTT] and protease inhibitor tablets, and eluted in the same buffer.

Table 1. DNA Sequence of AR-Selective and Nonselective Response Elements Name

Sequence

Reference

Pem ARE-1 Pem ARE-2 CRISP-1 SRE 2.43 SRE Consensus SRE

AGATCTcattcTGTTCC AGCACAtcgTGCTCA GGTACAttgTGTTCA GGAACAaaaTGTCCC GGTACAttgTGTTCT

5 5 9 8 10

The sequence is given in the 5⬘-to-3⬘-orientation. The halfsites are in uppercase letters; the spacing is in lowercase letters.

DNA-protein complexes were purified and analyzed for AR content using an adapted avidin-biotin complex DNA-binding assay (37, 38). The following biotinylated oligonucleotides containing two copies of each element (underlined) and their complementary sequence were mixed in annealing buffer [40 mM Tris-HCl (pH 7.5), 20 mM MgCl2, 50 mM NaCl], heated to 100 C, and cooled slowly to room temperature. They were then incubated with Streptavidin Sepharose beads (Sigma) overnight in binding buffer. Pem ARE-1: 5⬘-GATAATTCTTAAAGATCTCATTCTGTTCCATTGATTTTTAGAGATCTCATTCTGTTCCAATTCATTGATT-3⬘; Pem ARE-2: 5⬘-GATAATTCTTAAAGCACATCGTGCTCAATTGATTTTTAGAGCACATCGTGCTCAAATTCATTGATT -3⬘; CRISP-1 SRE: 5⬘-GATAATTCTTAAGGTACATTGTGTTCAATTGATTTTT AGGGTACATTGTGTTCAAATTCATTGATT-3⬘; 2.43 SRE: 5⬘-GATAATTCTTAAGGAACAAAATGTCCCATTGATTTTTAGGGAACAAAATGTCCCAATTCATTGATT-3⬘; Consensus SRE: 5⬘-GATAATTCTTAAGGTACATTGTGTTCTATTGATTTTTAGGGTACATTGTGTTCTAATTCATTGATT-3⬘. The flowthrough containing excess oligonucleotides was removed, and the oligonucleotide-loaded sepharose was washed with binding buffer. Binding of the oligonucleotides was verified by comparing the A260 values of the input and the flowthrough to ensure saturation of the streptavidin beads. After that, 70 ␮l of the 1:1 oligonucleotide-loaded sepharose solution were mixed with 100 ␮l of 1 mg/ml nuclear extract supplemented with 2.6 ␮g of poly(dI-dC) and 2.6 ␮g of poly(dA-dT) in a Micro Bio-Spin chromatography column (Bio-Rad Laboratories, Inc., Hercules, CA). Incubation was for 20 min at room temperature followed by 2 h at 4 C under constant shaking. After removal of the flowthrough, the beads were washed twice with 1 ml of binding buffer without protease inhibitor. Elution of bound proteins was performed with high-salt buffer (binding buffer supplemented with 1.2 M NaCl) which was followed by trichloroacetic acid precipitation. The protein content of the complexes was assessed by gel separation (4–12% polyacrylamide gel, Invitrogen, San Diego, CA) and Coomassie blue staining. For determination of AR content, the proteins were transferred from the gel onto a polyvinylidene fluoride membrane (Invitrogen) and probed with an anti-AR antibody (sc-7305; Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Processing was carried out with the Enhanced Chemiluminescence detection kit (Perkin Elmer Corp., Norwalk, CT). For the protease digestions 75 ␮l of chymotrypsin or trypsin dilutions in binding buffer were added to the beadsassociated protein-DNA complexes. After 10 min at room temperature the reaction was stopped by adding 35 ␮l of sample buffer (4% sodium dodecyl sulfate, 20% glycerin, 120 mM Tris base, 100 mM DTT, 0.02% bromophenol blue, 150 mM Pefablock SC) and heating to 95 C for 5 min. After centrifugation, aliquots of the supernatants were loaded onto polyacrylamide gels for Coomassie blue or Western blot analyses. Coomassie blue staining was performed to check that the total protein amounts were identical. Western blot analysis of the AR digestion products was carried out after transfer to a polyvinylidene fluoride membrane by incubating with the anti-AR C-19 antibody (Santa Cruz Biotechnology, Inc.) and processing with an enhanced chemiluminescence detection kit, as above.


Transfections CV-1 and PC-3 cells were cultured and transfected as described (5) with 100 ng of pSG5-AR or pSG5-GR plasmids. Hormone treatment was with 1 nM R1881 or dexamethasone, unless indicated otherwise. Luciferase activity was measured 24 h later in the presence of LucLite Plus reagent (Packard Instruments, Meriden, CT) in a Lumicount luminometer (Packard Instruments). The average value of six wells treated in parallel was determined, and the experiments were repeated at least three times independently. The amount of reporter plasmid harboring the different response elements was identical (100 ng) in all experiments. For the cofactor assays, 25 ng of pSG5-AR and 70 ng of cofactor-expressing plasmid or of ␤-galactosidase control plasmid were used. Statistical Analysis The significance of the results obtained after cofactor overexpression was determined by performing an ANOVA for each cofactor. The fixed factors used were: DNA element, hormone level, experiment, and statistical interaction between DNA element and hormone level. If the interaction factor was significant using ANOVA, a comparison of cofactor effects on response elements was carried out. The local ␣ levels of the t tests used were adapted for multiple comparisons by means of a Bonferroni adjustment.

Acknowledgments The support of Professor W.-D. Schleuning (Paion GmbH, Berlin, Germany) in the initial phase of this project was much appreciated. We thank Professor A. Cato (Forschungszentrum Karlsruhe, Germany) for AR and AR A596T cDNAs and for the PC-3/AR cells. We are grateful to Dr. M. Obendorf and Dr. S. Wolf (Jenapharm, Jena, Germany) for the PIASx␣ and ARA55 cDNAs. We are indebted to M. Preuss, I. Schu¨ ttke, and F. Knoth for excellent technical assistance. The help of S. Schwenke with statistical analysis is gratefully acknowledged.

Received November 15, 2002. Accepted May 29, 2003. Address all correspondence and requests for reprints to: Dr. B. Haendler, Corporate Research Business Area Oncology, Schering AG, Mu¨ llerstrasse 178, D-13342 Berlin, Germany. E-mail: bernard.haendler@ schering.de. This work was supported in part by Research Grants 0310681B from the German Federal Ministry of Education and Research (BMBF) and QLRT-2000-00602 from the 5th Framework Program of the European Union. Responsibility for the content of this publication resides solely with the authors.

REFERENCES 1. Lucas PC, Granner DK 1992 Hormone response domains in gene transcription. Annu Rev Biochem 61:1131–1173 2. Tsai MJ, O’Malley BW 1994 Molecular mechanisms of action of steroid/thyroid receptor superfamily members. Annu Rev Biochem 63:451–486 3. Freedman LP 1992 Anatomy of the steroid receptor zinc finger region. Endocr Rev 13:129–145 4. Verrijdt G, Haelens A, Claessens F 2003 Selective DNA recognition by the androgen receptor as a mechanism for hormone-specific regulation of gene expression. Mol Genet Metab 78:175–185


1749

5. Barbulescu K, Geserick C, Schu¨ ttke I, Schleuning WD, Haendler B 2001 New androgen response elements in the murine Pem promoter mediate selective transactivation. Mol Endocrinol 15:1803–1816 6. Haendler B 2002 Androgen-selective gene regulation in the prostate. Biomed Pharmacother 56:78–83 7. Schoenmakers E, Alen P, Verrijdt G, Peeters B, Verhoeven G, Rombauts W, Claessens F 1999 Differential DNA binding by the androgen and glucocorticoid receptors involves the second Zn-finger and a C-terminal extension of the DNA-binding domains. Biochem J 341:515–521 8. Roche PJ, Hoare SA, Parker MG 1992 A consensus DNA-binding site for the androgen receptor. Mol Endocrinol 6:2229–2235 9. Schwidetzky U, Schleuning WD, Haendler B 1997 Isolation and characterization of the androgen-dependent mouse cysteine-rich secretory protein-1 (CRISP-1) gene. Biochem J 321:325–332 10. Nelson CC, Hendy SC, Shukin RJ, Cheng H, Bruchovsky N, Koop BF, Rennie PS 1999 Determinants of DNA sequence specificity of the androgen, progesterone, and glucocorticoid receptors: evidence for differential steroid receptor response elements. Mol Endocrinol 13: 2090–2107 11. Luisi BF, Xu WX, Otwinowski Z, Freedman LP, Yamamoto KR, Sigler PB 1991 Crystallographic analysis of the interaction of the glucocorticoid receptor with DNA. Nature 352:497–505 12. Liu W, Wang J, Yu G, Pearce D 1996 Steroid receptor transcriptional synergy is potentiated by disruption of the DNA-binding domain dimer interface. Mol Endocrinol 10: 1399–1406 13. Kaspar F, Klocker H, Denninger A, Cato ACB 1993 A mutant androgen receptor from patients with Reifenstein syndrome: identification of the function of a conserved alanine residue in the D box of steroid receptors. Mol Cell Biol 13:7850–7858 14. Heinlein CA, Chang C 2002 Androgen receptor coregulators: an overview. Endocr Rev 23:175–200 15. Kotaja N, Karvonen U, Ja¨ nne OA, Palvimo JJ 2002 PIAS proteins modulate transcription factors by functioning as SUMO-1 ligases. Mol Cell Biol 22:5222–5234 16. Nishida T, Yasuda H 2002 PIAS1 and PIASx␣ function as SUMO-E3 ligases towards androgen receptor and repress androgen receptor-dependent transcription. J Biol Chem 277:41311–41317 17. Poukka H, Aarnisalo P, Karvonen U, Palvimo JJ, Ja¨ nne OA 1999 Ubc9 interacts with the androgen receptor and activates receptor-dependent transcription. J Biol Chem 274:19441–19446 18. Poukka H, Karvonen U, Ja¨ nne OA, Palvimo JJ 2000 Covalent modification of the androgen receptor by small ubiqutin-like modifier 1 (SUMO-1). Proc Natl Acad Sci USA 97:14145–14150 19. Callewaert L, Verrijdt G, Christiaens V, Haelens A, Claessens F 2003 Dual function of an amino-terminal amphipatic helix in androgen receptor-mediated transactivation through specific and nonspecific response elements. J Biol Chem 278:8212–8218 20. Klinge CM, Jernigan SC, Smith SL, Tyulmenkov VV, Kulakosky PC 2001 Estrogen response element sequence impacts the conformation and transcriptional activity of estrogen receptor ␣. Mol Cell Endocrinol 174:151–166 21. Hall JM, McDonnell DP, Korach KS 2002 Allosteric regulation of estrogen receptor structure, function and cofactor recruitment by different estrogen response elements. Mol Endocrinol 16:469–486 22. Chen S, Wang J, Yu G, Liu W, Pearce D 1997 Androgen and glucocorticoid receptor dimer formation. A possible mechanism for mutual inhibition of transcriptional activity. J Biol Chem 272:14087–14092 23. Schwabe JW, Chapman L, Rhodes D 1995 The oestrogen receptor recognizes an imperfectly palindromic re-


24.

25. 26.

27.

28.

29.

30.

sponse element through an alternative side-chain conformation. Structure 3:201–213 Hard T, Dahlman K, Carlstedt-Duke J, Gustafsson JA, Rigler R 1990 Cooperativity and specificity in the interactions between DNA and the glucocorticoid receptor DNA-binding domain. Biochemistry 29:5358–5364 In˜ iguez-LluhI´ JA, Pearce D 2000 A common motif within the negative regulatory regions of multiple factors inhibits their transcriptional synergy. Mol Cell Biol 20:6040–6050 Feng W, Ribeiro RCJ, Wagner RL, Nguyen H, Apriletti JW, Fletterick RJ, Baxter JD, Kushner PJ, West BL 1998 Hormone-dependent coactivator-binding to a hydrophobic cleft on nuclear receptors. Science 280:1747–1749 Loven MA, Likhite VS, Choi I, Nardulli AM 2001 Estrogen response elements alter coactivator recruitment through allosteric modulation of estrogen receptor ␤ configuration. J Biol Chem 276:45282–45288 Wood JR, Likhite VS, Loven MA, Nardulli AM 2001 Allosteric modulation of estrogen receptor configuration by different estrogen receptor elements. Mol Endocrinol 15: 1114–1126 Takeshita A, Yen PM, Ikeda M, Cardona GR, Liu Y, Koibuchi N, Norwitz ER, Chin WW 1998 Thyroid hormone response elements differentially modulate the interactions of thyroid hormone receptor with two receptor binding domains in the steroid receptor coactivator-1. J Biol Chem 273:21554–21562 Kotaja N, Aittoma¨ ki S, Silvennoinen O, Palvimo JJ, Ja¨ nne OA 2000 ARIP-3 and other PIAS proteins differ in their ability to modulate steroid receptor-dependent transcriptional activation. Mol Endocrinol 14:1986–2000


31. Gross M, Liu B, Tan J, French FS, Carey M, Shuai K 2001 Distinct effects of PIAS proteins on androgen-mediated gene action in prostate cancer cells. Oncogene 20: 3880–3887 32. Dotzlaw H, Moehren U, Mink S, Cato ACB, In˜ iguez Lluhı´ JA, Baniahmad A 2002 The amino terminus of the human androgen receptor is target for corepressor action and antihormone agonism. Mol Endocrinol 16:661–673 33. Abdel-Hafiz H, Takimoto GS, Tung L, Horwitz KB 2002 The inhibitory function in human progesterone receptor N termini binds SUMO-1 protein to regulate autoinhibition and transrepression. J Biol Chem 277: 33950–33956 34. Tian S, Poukka H, Palvimo JJ, Ja¨nne OA 2002 Small ubiquitin-related modifier-1 modification of the glucocorticoid receptor. Biochem J 367:907–911 35. Lefstin JA, Yamamoto KR 1998 Allosteric effects of DNA on transcriptional regulators. Nature 392:885–888 36. Geiser M, Ce`be R, Drewello D, Schmitz R 2001 Integration of PCR fragments at any specific site within cloning vectors without the use of restriction enzymes and DNA ligase. Biotechniques 31:88–92 37. Glass CK, Franco R, Weinberger C, Albert VR, Evans RM, Rosenfeld MG 1987 A c-erb-A binding site in rat growth hormone gene mediates trans-activation by thyroid hormone. Nature 329:738–741 38. Liu Y, Asch H, Kulesz-Martin MF 2001 Functional quantification of DNA-binding proteins p53 and estrogen receptor in cells and tumor tissues by DNA affinity immunoblotting. Cancer Res 61:5402–5406