DAX-1 functions as an LXXLL-containing corepressor for activated estrogen receptors.

We have discovered that the orphan receptor DAX-1 (NROB1) interacts with the estrogen receptors ERalpha and ERbeta. Interaction occurs with ligand-activated ERs in solution and on DNA and is mediated by the unique DAX-1 N-terminal repeat domain. Each of the three repeats contains a leucine-rich receptor-binding motif, known as the LXXLL motif, which is usually found in nuclear receptor coactivators. We have demonstrated that DAX-1 functions as an inhibitor of ER activation in mammalian cells and suggest a mechanism involving two sequential events, occupation of the ligand-induced coactivator-binding surface and subsequent recruitment of corepressors. Accordingly, we propose that DAX-1 itself acts as a corepressor for ERs. Because DAX-1 is coexpressed with ERs in reproductive tissues, these interactions could play significant roles by influencing estrogen signaling pathways. Our results point at functional similarities between DAX-1 and the orphan receptor SHP (NROB2) in that they have acquired features of transcriptional coregulators that are unique for members of the nuclear receptor family.

We have discovered that the orphan receptor DAX-1 (NROB1) interacts with the estrogen receptors ER␣ and ER␤. Interaction occurs with ligand-activated ERs in solution and on DNA and is mediated by the unique DAX-1 N-terminal repeat domain. Each of the three repeats contains a leucine-rich receptor-binding motif, known as the LXXLL motif, which is usually found in nuclear receptor coactivators. We have demonstrated that DAX-1 functions as an inhibitor of ER activation in mammalian cells and suggest a mechanism involving two sequential events, occupation of the ligand-induced coactivator-binding surface and subsequent recruitment of corepressors. Accordingly, we propose that DAX-1 itself acts as a corepressor for ERs. Because DAX-1 is coexpressed with ERs in reproductive tissues, these interactions could play significant roles by influencing estrogen signaling pathways. Our results point at functional similarities between DAX-1 and the orphan receptor SHP (NROB2) in that they have acquired features of transcriptional coregulators that are unique for members of the nuclear receptor family.
To transduce hormone and metabolic signaling to target genes, nuclear receptors require transcriptional cofactors that are collectively referred to as coregulators (1). These proteins exist in multiple complexes, possess multiple enzymatic activities, and bridge receptors to chromatin components or to the basal transcription machinery or to both. Multiple candidate proteins exist that are believed to be critical for the proper function of nuclear receptor signaling (1)(2)(3). The majority of coregulators bind to the receptor ligand-binding domain (LBD), 1 which is able to adopt different conformations depending on the ligand status and thereby discriminate between coactivators and corepressors. Functional and structural studies in particular elucidated the precise mechanisms of coactivator interaction with the ligand-inducible activation domain (AF-2) via short leucine motifs known as LXXLL or NR-Box (4 -8).
We became interested in coregulators in particular that influence the transcriptional activity of estrogen receptors (ER). Two related subtypes, ER␣ and ER␤, play crucial roles in sex development and reproduction in multiple physiological processes as well as in cancer (9 -11). Previous research has provided detailed insights into structural and functional aspects of their interplay with coregulators (8,(12)(13)(14). Although agonist binding usually is associated with ER activation caused by coactivator recruitment, regulatory mechanisms have been proposed that could play a role in modulation and feedback control of estrogen signaling (15,16). Recent work has revealed an unexpected role of the orphan receptor SHP (NROB2) in inhibiting transactivation of ERs (17,18). Particularly, we have provided evidence that SHP, which consists only of an LBD and thus cannot bind target genes directly, has instead acquired a novel coregulator function by antagonizing the interactions of ERs with associated coactivators (18,19).
The closest relative to SHP within the nuclear receptor family is the orphan receptor DAX-1 (NROB1) (20), which has a homologous LBD but contains a unique three repeat domain in the N terminus representing a novel type of single strand DNA/RNA-binding domain (21)(22)(23). Mutations in the human gene encoding DAX-1 cause adrenal hypoplasia, a rare inherited male disorder that frequently is associated with hypogonadotropic hypogonadism (22). Intriguingly, many mutations abolish the potent silencing function within the LBD and have lost the ability to recruit corepressors such as N-CoR and Alien (24 -27). Multiple evidence suggests key roles for DAX-1 in mammalian sex development, reproduction, and steroidogenesis (28 -33). DAX-1 is predominantly expressed in adrenal, ovary, testes, hypothalamus, and pituitary and functionally antagonizes SF-1, an essential orphan receptor for the development of the hypothalamic-pituitary-gonadal axis and regulator of male-specific gene expression (27,28,30,(32)(33)(34)(35). Together with demonstration of direct interaction, these findings suggest DAX-1 specifically acts as a cofactor for SF-1 (25,32).
In this study we have provided evidence that DAX-1 may play roles in regulation of ER transactivation. We have demonstrated that DAX-1 directly binds to ER␣ and ER␤ via the N-terminal repeat domain that contains LXXLL motifs. Functionally, DAX-1 inhibits the transcriptional activity of liganded ERs by a sequential mechanism, possibly involving the recruitment of corepressors. Accordingly, we propose that DAX-1 itself acts as an ER corepressor.
Protein-Protein Interaction Assays-Interaction assays were performed essentially as described previously (18,19,36). For GST pulldown assays, GST⅐DAX-1 fusion proteins bound to Sepharose beads (Amersham Pharmacia Biotech) were incubated with 35 S -labeled receptors in the absence (Me 2 SO) or presence of 1 M E 2 (17␤-estradiol) for 2-3 h at 4°C. After extensive washing, bound proteins were analyzed by SDS-polyacrylamide gel electrophoresis and visualized by autoradiography. For yeast two-hybrid assays, GAL4 plasmids transformed HF7c (MAT␣) were mated with GAL4AD plasmids transformed Y187 (MAT␣). All cotransformants were grown in selective media in the absence (Me 2 SO) or presence of 1 M E 2 . Interactions were monitored as relative ␤-galactosidase activity. For electrophoretic mobility shift assay supershift studies, ϳ200 ng of purified GST⅐DAX-1 R3 or GST⅐TIF2, respectively, were incubated with 10 ng of purified ER␤ (Panvera) in the absence or presence of 1 M E 2 . Radiolabeled ERE oligonucleotide was added last, and the binding reaction was allowed to proceed for 20 min on ice. Reactions were loaded on a 4% nondenaturing polyacrylamide gel and electrophoresed for 2 h in 0.5ϫ Tris-borate-EDTA at 4°C. Gels were dried and exposed to x-ray film. For coimmunoprecipitation, COS-7 whole cell extracts expressing DAX-1 and FLAG-tagged ERs were incubated with FLAG-M2 affinity matrix (Sigma) for 2 h at 4°C in the presence of 1 M E 2 in IP buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.2% Nonidet P-40, 1 mM EDTA, and 10% glycerol). Beads were subsequently washed using IP buffer lacking Nonidet P-40. Western analysis was performed using a rabbit polyclonal anti-DAX-1 serum (Santa Cruz Biotechnology) at 1:1000 dilution.
Mammalian Cell Transfections-Transient transfections were performed as described previously (18,19). Briefly, COS-7 cells were transfected with appropriate expression and luciferase reporter plasmids as indicated in the figure legends by using Lipofectin (Life Technologies) as instructed by the manufacturer. Cell extracts were prepared after 24-h expression and analyzed for relative luciferase activity. Individual transfections were performed in triplicate, and all experiments were repeated at least three times.

Interaction of DAX-1 with Liganded ERs via the N-terminal
Repeat Domain-Inspection of the human DAX-1 sequence revealed that each of the three N-terminal repeats contains a leucine-rich motif resembling the NR-box (Fig. 1A). Whereas the third motif (Box 3) matches the consensus LXXLL core with leucines in critical ϩ1, ϩ4, and ϩ5 positions, the other two motifs contain methionine at the ϩ4 position (Fig. 1C). Intriguingly, only Box 3 is conserved between species (e.g. mouse and human DAX-1) and has extensive homology (i.e. outside the leucine core) to the SHP-Box1 (Fig. 1, A and C), which we have previously demonstrated to be functional as an ER-binding motif (19). Therefore, we reasoned that the DAX-1 repeat region, which functions as a DNA/RNA-binding domain (21,23), may additionally mediate binding to ERs.
We made several constructs expressing NR-Box-containing DAX-1 fragments fused to GST (Fig. 1B) and assessed their binding to radiolabeled ERs in pull-down assays (Fig. 2). First, GST⅐DAX-1 DBD was found to interact with estradiol-bound ER␣ and ER␤ ( Fig. 2A) as well as with various other receptors (data not shown), indicating that the N-terminal repeat region of DAX-1 indeed may serve as receptor-binding domain. Second, because only repeat 3 contains a perfect LXXLL motif, we assessed binding of ERs to the repeat 3 region alone (Fig. 2B). Both ERs bound equally well, and estradiol enhanced the interactions. Third, to see whether these interactions were mediated by the LXXLL motif, we assessed binding to 14-mer NR-Box peptides fused to GST (Fig. 2C). For comparison, we analyzed peptide constructs expressing the second LXXLL motif of the p160 coactivator TIF2, which represents a highaffinity ER-binding motif (8,19), and we included a DAX-1-Box1 peptide with the core sequence LXXML (see Fig. 1C). Interestingly, although liganded ER␣ apparently displayed a higher affinity for DAX-1-Box3, the two ERs interacted with all peptides including DAX-1-Box1. This indicates that possibly all three NR Boxes in the repeat region may be involved in ER binding. Consistently, mutation of the third LXXLL motif (Fig.  1B, R3mut) abolished interaction with ERs when analyzed in the context of the isolated repeat 3 as expected but not when analyzed in the context of the three repeat domain or the entire DAX-1 protein (data not shown).
To investigate whether DAX-1 can interact with DNA bound and liganded ER dimers in vitro, we performed gel shift experiments using purified proteins (Fig. 3A). Whereas estradiol in the presence of GST control protein induced a characteristic downshift of the ER␤⅐DNA complex caused by conformational changes upon ligand binding (lanes 1 and 2), addition of GST⅐DAX-1 R3 (see Fig. 1B) protein led to a significant upshift (compare lane 4 with lanes 2 and 3) indicating ternary complex formation. For comparison, the LXXLL domain of TIF2 fused to GST (6) promoted a ligand-dependent supershift as expected (lane 6). Ternary complex formation was similarly observed with ER␣ homodimers, with RXR heterodimers and with monomeric SF-1 (data not shown). These data indicate that DAX-1 interaction with other nuclear receptors is not interfering with dimerization and DNA binding but instead resembles coactivator-type interactions with the AF-2 domain via LXXLL motifs. Further evidence for the DAX-1 interaction with ERs was observed in two additional experimental settings. First, in a coimmunoprecipitation assay, whole-cell extracts expressing FLAG-tagged ERs and wild-type DAX-1 were incubated with an ␣FLAG-affinity matrix and after washing, analyzed for the presence of DAX-1 protein (Fig. 3B). Overexpressed DAX-1 (lanes 2 and 4) as well as a protein possibly representing endogenous DAX-1 (lanes 1 and 3) were specifically coprecipitated only when FLAG⅐ERs were present but not in their absence (lanes 5 and 6). Second, in a yeast two-hybrid assay (Fig.  3C), GAL4⅐DAX-1 strongly interacted with both activation domain (GAD)-tagged ERs but not with GAD alone. Apparent differences between ER␣ and ER␤ with regard to their ligandindependent interaction in two-hybrid assays are not a peculiarity of the DAX-1 interaction but have been observed with SHP as well (19).
Functional Consequences of the DAX-1 Interaction with ERs-To investigate functional consequences of the interactions of DAX-1 with ERs on the activity of estrogen-responsive promoters, we performed transient transfections (Fig. 4). We compared three different DAX-1 expression constructs: (i) wildtype DAX-1, which is known to be a nuclear protein and a potent transcriptional repressor (21,23,24,27,32), (ii) the naturally occurring DAX-1 R267P mutation, which is nonrepressing possibly caused by its inability to bind the corepressors N-CoR and Alien (24,26), and (iii) a GAL4⅐DAX-1 fusion protein lacking the N-terminal receptor interaction domain but containing the repressor function within the LBD. We observed that DAX-1 WT and surprisingly also DAX-1 R267P inhibited ER␣ (Fig. 4A) or ER␤ activity (Fig. 4B), respectively, in a dose-dependent manner. However, the repression-defective DAX-1 variant appeared to be less effective than the wild-type DAX-1, particularly when using higher amounts of ER expression plasmid (data not shown), indicating that active repression may contribute to the inhibitory effects. Inhibition did not occur with GAL4⅐DAX-1 LBD lacking the N-terminal repeat domain, indicating that the inhibitory effect requires a direct interaction of DAX-1 with ERs and furthermore that the DAX-1 LBD cannot serve as an ER interaction domain.
To obtain further evidence for a possible recruitment of corepressors to ERs via DAX-1 as bridging protein, we performed a mammalian two-hybrid experiment in analogy to the experiments described for SF-1 (25). As seen in Fig. 4C, GAL4⅐ER␣ and VP16 activation domain-tagged N-CoR encompassing the DAX-1 interaction domain were cotransfected in the absence or presence of DAX-1. Consistent with the inability of N-CoR to bind to ERs in the presence of agonists (37,38), we found that VP16⅐N-CoR in the absence of cotransfected DAX-1 had no effect on estradiol-induced ER␣ activity. However, VP16⅐N-CoR could partially restore reporter gene activity in the presence of inhibitory amounts of DAX-1. Because control Western blot analysis shows equal DAX-1 protein levels in both the absence or presence of VP16⅐N-CoR (Fig. 4C, lower), this result suggests that DAX-1 may serve as a bridging protein between liganded ER␣ and the corepressor N-CoR (see Fig. 5).

DISCUSSION
The results presented in this study provide insights into previously uncovered aspects of DAX-1 structure and function and substantially expand the regulatory potential of DAX1. The interactions with ERs may be physiologically relevant because DAX-1 is expressed in multiple estrogen target tissues (9,39). Summarizing the results from independent immuno-

FIG. 3. Analysis of DAX-1 interactions with ER on DNA in cell extracts and in vivo.
A, electrophoretic mobility shift assays using 10 ng of purified ER␤ and ϳ200 ng of partially purified GST⅐DAX-1 R3-(115-199), GST⅐TIF2-(594 -766) or GST alone. Proteins were incubated in the absence or presence of 1 M E 2 with 32 P-labeled ERE oligonucleotide and analyzed for complex formation as described under "Experimental Procedures." B, coimmunoprecipitation assay analyzing the association of full-length DAX-1 with FLAG-tagged ERs in COS-7 cell extracts. Approximately 0.5 mg of whole cell extracts was incubated with 20 l of FLAG-affinity matrix in the presence of 1 M E 2 , and the DAX-1 protein (55 kDa) in the immunoprecipitates was analyzed by Western blot as described under "Experimental Procedures." C, yeast two-hybrid analysis of interactions between DAX-1 and ERs in vivo. Yeast whole cell extracts coexpressing GAL4⅐DAX-1 DBD-(1-253) and activation domain-tagged ERs in the absence or presence of 1 M E 2 were analyzed in a liquid ␤-galactosidase assay as described under "Experimental Procedures." The relative ␤-galactosidase activity observed with DAX-1 DBD and ER␣ in the presence of E 2 was set to 100%. As seen with GAD control, GAL4⅐DAX-1 DBD displayed no background activation nor interacted with the activation domain alone. Values shown are the mean Ϯ S.D. from three independent experiments. histochemical analyses (40,41), 2 it is quite striking that ER␣ and ER␤ proteins are apparently differentially expressed in distinct cell types of male and female reproductive tissues. For example, in testis ER␣ seems predominantly expressed in Leydig cells, whereas high ER␤ expression seems restricted to Sertoli and germ cells. In ovary, ER␤ is mainly expressed in granulosa cells, whereas ER␣ expression is much lower and restricted to interstitial theca cells. Although a comparative analysis of DAX protein expression needs to be accomplished, the known mRNA expression pattern suggests DAX-1 coexpression in all of these cell types (22,29,31,32,35). Notably, until now little is known about female-specific roles of DAX-1. Gene inactivation in mice surprisingly did not affect ovarian development and fertility but instead caused male infertility (28), a phenotype which intriguingly has been observed in mice lacking ER␣ (42). Moreover, developmental studies suggest coexpression of ERs and DAX-1 in testis and ovary during certain stages of embryogenesis (35,43). Possibly, DAX-1 serves as a tissue-or stage-specific ER coregulator involved in modulation of estrogen signaling. Comparatively, much less is known about expression of ERs in adrenal gland, a major site of DAX-1 function in steroidogenesis. However, estrogens are known to affect adrenal development, and ERs have been detected in the cortex of fetal primate glands (44) as well as in all cell types of the adult rat gland (45). Because DAX-1 expression could be hormonally regulated (29), feedback mechanisms in steroidogenesis involving DAX-1 may principally resemble the recently discovered feedback loop in bile acid biosynthesis involving SHP (46,47). Furthermore, our results indicate nuclear receptor binding as a novel feature of the DAX-1 repeat domain and thereby it reveals additional functional similarities between DAX-1 and SHP, the only two members of the nuclear receptor family that have acquired characteristics of transcriptional coregulators (18,19,48). Interestingly, the chicken DAX-1 homolog apparently lacks the entire mammalian repeat domain but contains in its short N terminus a single ILYSIL motif (49). This possibly suggests that binding to nuclear receptors is evolutionarily conserved between species whereas binding to DNA is not. Intriguingly, the mechanism we propose here for ERs may provide an alternative explanation for the inhibitory effects of DAX-1 on retinoic acid receptor transcription first reported in the original study (22), supporting the idea that DAX-1 may serve broader functions in nuclear receptor signaling (25,39).
In Fig. 5 we provide a discussion model by integrating our findings with current theories of DAX-1 mechanisms of action. In this model, DAX-1 is envisaged to mediate functional interactions between upstream coregulator complexes consisting of corepressors or, if ligands exist, coactivators and downstream target genes. These genes may be regulated directly via DAX-1 (21) or indirectly via nuclear receptors such as ERs and SF-1. It will be necessary to determine the relative importance of these two alternatives in vivo and to investigate whether these two events occur simultaneously or competitively. Because DAX-1 interacts through LXXLL motifs and possesses an intrinsic silencing function, it is likely that DAX-1 may inhibit receptor activation by a sequential mechanism involving coactivator displacement and subsequent corepressor recruitment. Therefore, DAX-1 itself may be defined as an LXXLL-containing corepressor and shares this feature with SHP. Although DAX-1 is a true orphan receptor, endogenous ligands could possibly convert DAX-1 from a repressor to an activator and thereby possibly activate, for example, estrogen target genes. Indeed, indirect evidence for a role of ligand-activated DAX-1 comes from the recent discovery of a patient with X-linked congenital adrenal hypoplasia carrying a missense mutation in the activation domain helix 12 (50), which is dispensable for corepressor binding but indispensable for coactivator binding to liganded receptors (1). Also, the DAX-1 helix 12 contains a glutamate residue that is conserved in all ligand-activatable receptors and was suggested to be critical for coactivator LXXLL binding (7). Future structural information may be required to reveal the presence of a ligand-binding pocket. Until then, it is an exciting possibility that the coregulatory potential of DAX-1 could be hormonally, metabolically, or pharmaceutically regulated, a novel aspect of both receptor and coregulator function.