Differential regulation of the orphan nuclear receptor small heterodimer partner (SHP) gene promoter by orphan nuclear receptor ERR isoforms.

The orphan nuclear receptor small heterodimer partner (SHP; NR0B2) interacts with a wide array of nuclear receptors and represses their transcriptional activity. SHP expression is regulated by several other members of the nuclear receptor superfamily, including the orphan receptors SF-1 and LRH-1, and the bile acid receptor FXR. We have found that the SHP promoter is also activated by the estrogen receptor-related receptor gamma (ERRgamma) but not the related ERRalpha and ERRbeta isoforms. SHP and ERRgamma mRNAs are coexpressed in several tissues, including pancreas, kidney, and heart, confirming the potential relevance of this transactivation. ERRgamma transactivation is dependent on only one of five previously characterized DNA-binding sites for SF-1, and this element differs from previously reported ERR response elements. However, treatment with the histone deacetylase inhibitor trichostatin A significantly increased ERRalpha and ERRbeta activity on this element indicating that the lack of activity of ERRalpha and -beta may depend on their association with co-repressor in vivo. Furthermore, using protease sensitivity assays on DNA bound receptors it was demonstrated that DNA sequence of different response elements may cause allosteric modulation of ERR proteins, which in turn may be responsible for the differential activities of these receptors on different response elements. SHP inhibits ERRgamma transactivation and physically interacts with all three members of ERR subfamily, as demonstrated by both yeast two-hybrid and biochemical assays. As with other SHP targets, this interaction is dependent on the AF-2 coactivator-binding site of ERRgamma and the previously described N-terminal receptor interaction domain of SHP. Several recently described SHP mutations associated with moderate obesity in humans block the inhibition of ERRgamma activity. Overall, these results identify a new autoregulatory loop controlling SHP gene expression and significantly extend the potential functional roles of the three ERRs.

The nuclear receptor superfamily is a diverse group of tran-scription factors that includes both conventional receptors with known ligands and orphan nuclear receptors that lack them (for reviews, see Refs. 1 and 2). Both conventional and orphan nuclear receptors share a very similar structure. At the N terminus is a domain that is not conserved in sequence in different receptor families, but in many cases contains a ligand independent transcriptional activation domain termed activation function 1. The central DNA-binding domain (DBD) 1 is strongly conserved and includes two zinc-binding units based on invariant cysteine residues. A conserved helical region within the DBD termed the P box (3) makes base specific contacts with hormone response elements. The sequence of this motif serves as one of the main criteria for dividing the nuclear receptor superfamily into subgroups. The C-terminal ligandbinding domain (LBD) functions not only in ligand binding, but also dimerization and ligand mediated transcriptional activation. This transactivation is based on an additional activation function referred to as AF-2. In most receptors, allosteric effects of ligand binding result in the formation of an appropriate binding site for a series of transcriptional coactivators. In a number of orphans, however, coactivator binding is not dependent on the presence of a ligand. Members of this subset of orphans, which includes HNF-4, and ERR␥, function as constitutive transactivators (4,5).
SHP is an atypical member of nuclear receptor superfamily that lacks a DBD. Various studies have reported SHP to be a repressor of transcriptional activities of a number of nuclear receptors, including both ligand responsive receptors like, ER, TR, RAR, and RXR, and orphan receptors like CAR, HNF-4, and FTF (6 -10). The very broad range of receptors sensitive to inhibition by SHP suggests a central role for SHP in modulation of nuclear receptor signaling pathways. Although the mechanisms underlying this repressive function remain unclear, recent results (7,11) demonstrate that SHP can compete with coactivators for binding to the AF-2 surface. In addition, a direct transcriptional repressor domain contributes significantly to the inhibitory function of SHP.
The SHP gene is structurally conserved in all the species from which it has been characterized. It consists of 2 exons interrupted by an intron and is located on human chromosome 1 at position 1p36.1 (12). Initial studies indicated that SHP is expressed predominantly in the liver, spleen, small intestine, and pancreas (8,12). However, later studies with more sensitive approaches have demonstrated the presence of SHP mRNA in a wide variety of tissues, with highest expression in heart, brain, liver, lung, and adrenal gland (11). Although very little is known about SHP gene regulation, some recent reports have suggested important roles for the bile acid receptor FXR, and AP-1 transcription factors in regulating SHP gene expression and subsequent regulation of cholesterol homeostasis (13)(14)(15). A recent report also demonstrated the activation of the SHP promoter by SF-1 and its close relative LRH-1, both of which bind five distinct sites present in the region spanning from 453 to 68 base pairs upstream of the transcriptional start (16). SHP is coexpressed with SF-1 or LRH-1 only in gonads, adrenal gland, and liver, however, indicating that other transcription factors are involved in SHP gene expression in other tissues, such as pancreas, heart, and brain.
The ERR subfamily includes the first orphan receptors described (17). These proteins form a distinct subgroup within the superfamily with striking sequence similarity in the DBD with estrogen receptors (ER). The similarity in the LBD is more limited, and the ERRs do not bind natural estrogen. However, a recent report has demonstrated the binding of the synthetic estrogen analogue diethylstilbestrol to the ERR subfamily members (18). In this case, diethylstilbestrol acts as an inverse agonist by disrupting ERR-coactivator interaction (18). Three different ERR genes have been characterized (5,17,19,20). The ERR␣, ERR␤, and ERR␥ isoforms show considerable homology among themselves, with the highest similarity between ERR␤ and ERR␥. The hypervariable N-terminal A/B domain of ERR␤ shares 58.2% amino acid identity with ERR␥, and in the DBD and the LBD they share 98.9 and 73.1% identity, respectively (5). All three isoforms have been reported to bind and transactivate both estrogen response elements (ERE) and SF-1 response elements (SF-1RE) (5, 20 -27). ERR␣ has been implicated in various roles in bone and muscle development, modulation of aromatase expression in breast tissues, regulation of thyroid hormone receptor ␣ (TR␣) expression, overall development of mouse embryos, and fatty acid metabolism (21, 23, 28 -30). Inactivation of the murine ERR␤ gene caused severe placental abnormalities and embryonic mortality at 10.5 days post-coitum, and ERR␤ has also been implicated in repression of glucocorticoid receptor-mediated transcriptional activities (31,32). The biological roles of ERR␥, the newest member of the subfamily, remain to be elucidated. The expression profiles of the three ERR isoforms show an overlapping pattern. ERR␣ mRNA is primarily expressed in tissues associated with fatty acid metabolism such as kidney, heart, and brown adipocytes. ERR␥ is expressed at a high level in heart, brain, kidney, pancreas, and at a lower level in liver (5,20,26), while the expression of its closer relative ERR␤ is barely detectable in few tissues in adult rat including kidney, testis, heart, brain, and prostate. In the mouse ERR␤ expression has been demonstrated in a subset of extraembryonic tissues in a small window between 5.5 days post-coitum and 8.5 days post-coitum (17).
Since the SHP gene promoter contains several SF-1 response elements that are predicted to be ERR response elements, we investigated whether ERR subfamily members can regulate the SHP gene promoter. Unexpectedly, transient transfection studies demonstrated a preferential effect of ERR␥ on this promoter. Mapping studies revealed that only one of the five previously reported SF-1REs on SHP gene promoter is responsible for the ERR␥ mediated activation of the SHP promoter. Electrophoretic mobility shift assays, and mutational studies confirmed the binding of ERR␥, to this element. As expected, SHP coexpression inhibits ERR␥ transactivation of its own promoter, and SHP competes with the coactivator GRIP-1/ SRC-2 for binding to ERR␥. We conclude that ERR␥ is a component of a new potential autoregulatory loop controlling SHP gene expression.

EXPERIMENTAL PROCEDURES
Plasmids-pCMX ERR␣ and pCMX ERR␤ were kind gifts from Dr. Vincent Giguere. pSG5HA ERR3, pSG5HA GRIPI, and ERE II Luc were kind gifts from Dr. Michael R. Stallcup (to avoid confusions regarding the nomenclature, ERR3 will be indicated as ERR␥ in this report henceforth). Five copy SF-1RE Luc was a kind gift from Dr. Jean Mark Vanacker. B42AD fused yeast expression constructs of ERR␣ and ERR␤ were made by inserting PCR products encoding the open reading frame of ERR␣ or ERR␤ containing a 5Ј EcoRI and a 3ЈXhoI sites into pJG4-5 (CLONTECH). B42AD fused construct of ERR␥ was constructed by inserting a PCR product encoding the complete open reading frame of ERR␥ with flanking XhoI sites into pJG4-5. LexA DNAbinding domain fused ERR␥ LBD was created by inserting a PCR product encoding the ligand-binding domain in the appropriate restriction sites in a pEG202 vector (33). For bacterial expression, GST-fused full-length ERR␣ and ERR␤ were constructed by inserting EcoRI-XhoI fragments of ERR␣ or ERR␤ from B42 ERR␣, or B42 ERR␤, into pGEX4T-1 vector (Amersham Bioscience, Inc.), GST-fused full-length ERR␥ was constructed by inserting a XhoI fragment encoding complete open reading frame from B42ERR␥ in-frame into pGEX.4T-1. GST fused ERR␣, -␤, and -␥ ligand-binding domains were created by inserting PCR products into appropriate restriction sites in pGEX.4T-1 vector. To put all the ERR isoforms into same mammalian expression vector pcDNA3ERR␣, pcDNA3ERR␤, and pcDNA3ERR␥ were constructed by inserting fragments from B42 constructs of ERR␣, ERR␤, and ERR␥ as described, for the GST-fused constructs. ERR␥ S-N mutant was created by PCR-based site-directed mutagenesis from pcDNA3ERR␥ using a proofreading Vent polymerase (New England Biolabs), the resulting mutant PCR product was confirmed by restriction digest as well as by sequencing and cloned into pcDNA3 at HindIII-EcoRV sites. ERR␥ dAF-2 construct was created by PCR from pcDNA3 ERR␥ and cloned into pcDNA3, the pcDNA3 ERR␥ dAF-2 was then digested with HindIII and NotI and the resulting fragment was cloned in-frame into pYesTRP2 (Invitrogen), a B42 yeast vector. SHP mutants are described elsewhere (35). Yeast expression constructs of naturally occurring SHP mutants were created by inserting respective EcoRI-XhoI fragments into pEG202 (33). pcDNA3 HA ERR␣ and ERR␤ was constructed by subcloning EcoRI-XhoI fragments from respective B42 constructs in-frame into a pcDNA3 vector containing an HA coding sequence. Three-copy sft4 reporter element was created by compatible end ligation at BamHI and BglII sites followed by cloning into BglII site of pGL2 promoter vector (Promega). All the clones were verified by sequencing.
In Vitro Translation-SHP or ERR␥ S-N cDNAs in pcDNA3 were transcribed and translated in vitro by using a coupled rabbit reticulocyte system (TNT, Promega) in the presence or absence of [ 35 S]methionine (Amersham Bioscience, Inc.) according to the manufacturer's instructions.
GST Pull-down Assay-A GST pull-down assay was performed according to the method described previously (10). Briefly the GST fusion proteins or GST protein only were expressed in Escherichia coli BL21(DE3) pLys bacterial culture and recovered on glutathione-Sepharose-4B beads (Amersham Bioscience, Inc.), the GST fusion proteins bound to glutathione-Sepharose 4B beads were incubated in a 100-l reaction for 2 h at 4°C with 35 S-labeled receptors expressed by coupled in vitro transcription and translation (TNT, Promega). Specifically bound proteins were eluted from beads with 15 mM reduced glutathione in 50 mM Tris (pH 8.0) and analyzed by SDS-polyacrylamide gel electrophoresis and visualized by a PhosphorImager analyzer (BAS-1500, Fuji).
Gel Mobility Shift Assays-Bacterially expressed GST-fused ERR␣, ERR␤, and ERR␥, were purified using GST-Sepharose beads (Amer-sham Bioscience, Inc.) as described elsewhere (10), ERR␥ S-N was transcribed and translated in vitro, using a coupled rabbit reticulocyte system. Gel mobility shift assays (20 l) contained 10 mM Tris (pH 8.0), 40 mM KCl, 0.05% Nonidet P-40, 6% glycerol, 1 mM dithiothreitol, and 1 g of poly(dI-dC). Either 1 g of GST purified proteins, or 2.5 l of programmed/mock-programmed reticulocyte lysates were used in each reaction. Competitor oligonucleotides were included at a 10 -250-fold molar excess as indicated in the figure legends. After 10 min incubation on ice, 10,000 cpm of end labeled oligonucleotide probes were added and the incubation continued for another 10 min. DNA-protein complexes were analyzed on 5% polyacrylamide gel in 1 ϫ TBE (1 ϫ TBE ϭ 90 mM Tris, 90 mM boric acid, 2 mM EDTA). Gels were dried and analyzed by autoradiography. The sequences of the oligonucleotides used are described elsewhere (16).
Limited Proteolysis Analysis by Gel Shift Assay-Protease sensitivity assays were performed as described earlier (37) with minor modifications. Briefly, 5,000 cpm of end labeled oligonucleotides corresponding to sft4, (GATCCACATGACTTCTGGAGTCAAGGTTGTTTGGA) or consensus SF-1RE (GATCCACATGACTTCTGGAGTCAAGGTCATTT-GGA) were incubated on ice with GST purified ERR␥ in DNA-binding buffer in a reaction volume of 10 l. After 20 min incubation on ice, 0, 1, 2.5, 5, 10, or 50 ng of proteinase K (Promega) was added to the binding reaction and the reaction was incubated for a further 15 min at room temperature. The reactions were stopped by mixing 10 l of 1 ϫ DNA binding buffer and putting the samples on ice. The samples were then immediately subjected to electrophoresis through a 8% nondenaturing polyacrylamide gel, and analyzed as described in the case of gel mobility shift assays.
Limited Proteolysis Analysis by SDS-PAGE-Six l of in vitro transcribed and translated ERR␥ was incubated with 40 pmol of sft4 or SF-1RE oligonucleotides in the presence of DNA binding buffer in a reaction volume of 10 l. After incubation on ice for 20 min, indicated amounts of proteinase K was added, and the incubation was continued for a further 15 min at room temperature. The reactions were stopped by mixing with equal volume of 2 ϫ SDS loading buffer, heated at 95°C for 5 min, and subjected to electrophoresis on a 12% denaturing polyacrylamide gel. The gels were then dried and analyzed by autoradiography.
Transient Transfection Assays-HeLa (Human cervical carcinoma), CV-1 (green monkey kidney), and HEK 293 (human embryonic kidney) cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) in the presence of 10% fetal bovine serum (Invitrogen). For Luciferase assays, 24 h before transfection cells were plated in 24-well plates and transfected, transfections were carried out with Superfect reagent (Qiagen) according to the manufacturer's instructions. Total DNA used in each transfection was adjusted to 1 g by adding appropriate amount of pcDNA3 vector. Approximately 48 h post-transfection, cells were harvested, and the luciferase activity was measured as described (33) and normalized against ␤-galactosidase activity. Western Blot Analysis-HA-tagged ERR␣, -␤, or -␥ (5 g) were transfected into HeLa cells, in 10-cm dishes using Superfect reagent. The cell lysates were prepared 48 h after transfection. 20 g of protein from each cell lysates were loaded and separated on a 12% denaturing polyacrylamide gel. The proteins were blotted on to Hybond-C extra nylon membranes (Amersham Bioscience, Inc.), and visualized with monoclonal HA antibody (Roche Molecular Biochemicals) and ECL detection kit (Amersham Bioscience, Inc.).

Activation of the SHP Gene Promoter by ERR␥, but Not
ERR␣ or ERR␤-Since a previous report demonstrated that SF-1 activates the SHP gene promoter via five characterized SF-1REs (16) and SF-1 and ERR family members share similar DNA binding characteristics (23,34), we investigated whether ERR family members can also regulate the SHP gene promoter. Transient transfection studies using ϳ2 kb pairs of either mouse or human SHP promoter sequence fused to a luciferase reporter demonstrated a strong activation of SHP gene promoter by ERR␥, whereas ERR␣ and ERR␤ did not show a significant effect. ERR␥ showed a strong transactivation of SHP promoter in both HeLa (Fig. 1A) and CV-1 (Fig. 1B) cells. However, ERR␥ showed a moderate activity in HEK-293 cells, inducing the promoter by 5-fold at the highest dose of receptor DNA (Fig. 1C). Unexpectedly SF-1, the positive control for the transfection assays, failed to activate the SHP gene promoter in CV-1 cells, although it did so in HeLa and HEK-293 cells (Fig. 1, A and C). It is possible that CV-1 cells lack a specific signaling pathway or coactivator required for SF-1 activity. Consistent with the ERR␥-dependent response of the mouse promoter, ERR␥ also substantially activated the human SHP gene promoter (Fig. 1D), demonstrating the conservation of ERR␥ responsiveness.
To ensure that all the ERR constructs were effectively expressed, Western blot analysis was performed with whole cell extracts from HeLa cells transfected with the HA epitopetagged ERR isoforms. As demonstrated in Fig. 1E, comparable levels of the three ERR isoforms were expressed, indicating that their relative expression levels were not responsible for their differential activity on SHP promoter. To further ensure that all the ERR constructs are transcriptionally active, HeLa cells were transfected with a two-copy estrogen response element-driven reporter element (ERE II Luc) (Fig. 1F) along with the ERR expression vectors. In agreement with other observations (18) ERR␣ showed a modest activity on an estrogen response element-driven reporter (ERE-II-Luc), whereas ERR␤ and -␥ showed a stronger activity. These results indicate that the differences in the activity of the ERR isoforms exhibited on SHP promoter are not dependent on their relative expression, but may depend on the sequences of respective DNA response elements.
The DBDs of ERR␤ and ERR␥ differ at only a single residue. To determine whether this variation is involved in the differential response observed, an ERR␥ construct (ERR␥ S-N) in which this serine residue was converted to the asparagine residue of ERR␤ was generated. Transient transfection assays demonstrated ERR␥ S-N still retained the capacity to activate SHP promoter (data not shown). Thus, the lack of activity of ERR␤ must be dependent on sequences outside the DBD. This could be due to a lack of transactivation function or to indirect effects of other domains on DNA binding or interaction with other factors. Taken together, these results demonstrate that ERR␥ specifically induces SHP promoter.
Identification of the Sequences Responsible for ERR␥ Response on SHP Gene Promoter-To determine the sequences required for ERR␥ mediated activation, a previously described series of 5Ј deletions of the mouse SHP promoter were used. ERR␥ responsiveness was retained with a deletion retaining 139 bp upstream of the putative transcriptional start site, but was lost with the construct containing only 68 bp upstream of the putative transcriptional start point ( Fig. 2A). Thus, sequences required for ERR␥ response lie in the region between Ϫ139 and Ϫ68, which contains a single SF-1 responsive site previously designated sft4 (16).
To specifically identify whether the ERR␥ response requires sft4, a series of mutant constructs containing individual GG to TT double substitution mutations in the core of the five SF-1binding sites in the Ϫ0.453 promoter (17) were tested for ERR␥ activity (Fig. 2B). As shown in Fig. 2B, the sft4 mutation completely abolished ERR␥ response, whereas mutations of other SF-1-binding sites did not significantly alter response, indicating that sft4 is the only one of these sites required for ERR␥ response.
To further clarify whether ERR␥ specifically binds sft4, a series of gel mobility shift assays were conducted using oligonucleotides for all the five SF-1REs as probes. As demonstrated in Fig. 3A, both ERR␥ and ERR␥ S-N formed specific complexes with sft4 (lanes 2, 5, and 12) and a 250-fold molar excess of unlabeled sft4 oligonucleotide successfully competed with the DNA-protein complex (lanes 3, 9, and 13). As expected a 250fold molar excess of msft4 did not affect the DNA-protein interaction (lanes 4 and 14). To identify whether ERR␥ binds to other sft sequences, a series of competition assays were performed with 250-fold molar excess of unlabeled sft sequences. Excess unlabeled oligonucleotides corresponding to sft1, -2, -3, and -5 had no effect on ERR␥-sft4 interaction (lanes 6 -8 and 10, respectively). Direct binding assays using labeled sft1, -2, -3, and -5 also confirmed that ERR␥ forms specific DNA-protein complex with sft4 only (data not shown). Since all the ERR isoforms exhibit a highly homologous DBD sequence we examined whether ERR␣ and -␤ also bind to sft4 sequence, and whether their apparent lack of activity on SHP promoter is a result of differential binding affinity of these two proteins on sft4 sequence. Surprisingly, both ERR␣ and ERR␤ formed specific complex with sft4 sequence and their affinity for sft4 sequence was similar to ERR␥ (Fig. 3B). In all the cases at least  (17) and in the figure, were transfected in CV-1 cells with 500 ng of ERR␥ or vector alone. The luciferase activity was measured and normalized against ␤-galactosidase activity. One representative experiment is shown, and similar results were obtained from at least three independent experiments. 100-fold molar excess of unlabeled sft4 was required to successfully compete with the DNA-protein complex (compare lanes 6, 12, and 18) whereas purified GST protein alone did not form any complex with the probes (lanes 1 and 14). Intriguingly, ERR␣ and ERR␤ did not bind to other sft sequences as evidenced by both direct binding and competition analysis (data not shown). In agreement with a previous report (20), all three isoforms produced two specific complexes. We assume the upper complex may be a result of dimerization between two ERR molecules, and both yeast two-hybrid and biochemical interac-tion assays indicated that all the ERR family members can form homodimeric complexes. 2 To specifically determine whether the sft4 element by itself is sufficient to confer complete ERR␥ activity, a reporter construct containing three copies of the sft4 element cloned upstream of a luciferase gene was transiently transfected with ERR isoforms in CV-1 and HeLa cells. As expected, ERR␥ 2 S. Sanyal and H-S. Choi, unpublished observation.

FIG. 3. All the ERR isoforms demonstrate specific DNA-protein complex with sft4 only.
A, purified GST-ERR␥ or in vitro transcribed and translated ERR␥ S-N, was incubated with end labeled sft4 as indicated. In competition assays 250-fold molar excess of unlabeled wild type or mutant sft sequences were used as indicated. B, GST-ERR␣, GST-ERR␤, or GST-ERR␥ were incubated with end labeled sft4 as indicated in the figure. GST, purified GST protein only protein, 5, 10, 50, 100, and 250 indicate the -fold of molar excess of unlabeled specific competitors. Msft4 indicates 250-fold molar excess of mutant sft4 (msft4). Lys, mock programmed reticulocyte lysate. C, HeLa and CV-1 cells were co-transfected with a three-copy sft4 luciferase and 500 ng of ERR␣, ERR␤, or ERR␥. The luciferase activity was measured and normalized against ␤-galactosidase activity. One representative experiment is shown, and similar results were obtained from at least three independent experiments. D, a comparison of human, mouse, and rat SHP promoter sequences, the sft4 motif is in bold letters. Numbers indicate corresponding nucleic acid positions. substantially induced this reporter whereas, both ERR␣ and ERR␤ failed to do so (Fig. 3C).
Taken together, we conclude ERR␥ confers its response solely through sft4. The high degree of sequence identity of the sft4 site and surrounding sequences in mouse, rat, and human SHP promoters (Fig. 3D) indicates that this ERR␥ response is conserved.
Potential Mechanism for the Differential Activation of SHP Promoter by ERR Isoforms-Differential corepressor recruitment is one mechanism that could explain why all the three ERR isoforms bind the sft4 DNA element with comparable affinity but only ERR␥ activates it significantly. The effects of corepressors can be blocked by TSA, a selective inhibitor of histone deacetylases as well as a disruptor of histone deacetylase co-repressor complexes. Thus, the observation that TSA treatment significantly increased ERR␣ and ERR␤ activity on sft4, but failed to potentiate ERR␥ activity (Fig. 4A), is consistent with the possibility that ERR␣ and ERR␤ but not ERR␥, associate with corepressors in vivo.
Since ERR␥ can bind to a consensus SF-1RE (20, 29) we tried to determine whether ERR␥ can also transactivate other gene promoters containing such sites. Unexpectedly, ERR␥ demonstrated no significant activity on TR␣ gene promoter (data not shown), which contains a consensus SF-1RE (23) that is reported to be a bona fide target of ERR␣ (23) and binds ERR␥ efficiently (Ref. 20, and see below). To determine whether these discrepancies in ERR␥ mediated activities on different promoters depend on the sequence of the available binding sites, a comparison of ERR␥ activity on the tandem repeats of such binding sites was executed. Interestingly, ERR␥ demonstrated distinctly different activities on these elements. Thus, although ERR␥ activated a three-copy sft4-Luc in HeLa and CV-1 cells by 12-15-fold and 35-40-fold, respectively, it activated the five-copy SF-1 RE Luc by 2-3-fold in HeLa and 4 -5-fold in CV-1 cells (Fig. 4B).
These results indicate that binding of ERR␥ to different DNA elements may induce specific allosteric modulations of the tertiary structure of this protein as reported previously in cases of GR (36), ER (37,38), and TR (39). To investigate this possibility, a protease sensitivity experiment was performed. GST purified ERR␥ was incubated with 32 P-labeled DNA fragments containing either sft4 or consensus SF-1RE followed by digestion with increasing concentrations of proteinase K for a fixed time period as described under "Experimental Procedures" and resolved on a nondenaturing polyacrylamide gel. As demonstrated in Fig. 4C, ERR␥ formed specific DNA-protein complexes with both sft4 and SF-1RE (lanes 1 and 8). Surprisingly the proteinase K-digested sft4 and SF-1RE bound ERR␥ demonstrated different patterns, with additional partial protease products observed with the latter (arrows, Fig. 4C). This result was also confirmed by a different approach in which 35 S-labeled in vitro transcribed and translated ERR␥ were incubated with FIG. 4. DNA sequence mediated structural and functional alterations of ERR isoforms. A, HeLa cells were co-transfected with a three-copy sft4 luc and 500 ng of ERR␥ expression plasmids. 24 h after transfection cells were treated with 100 nM TSA. After incubating the cells in presence of TSA for a further 24 h, cells were lysed, luciferase activity was measured, normalized against ␤-galactosidase activity, and plotted as relative luciferase activity (RLU). B, HeLa and CV-1 cells were transfected with indicated reporter plasmids and 500 ng of ERR␥, as indicated in the figure. After 48 h incubation luciferase activity was measured, normalized against ␤-galactosidase activity, and plotted as -fold basal activity with 1-fold basal activity defined as the luciferase activity with pcDNA3. One representative experiment is shown, and similar results were obtained from at least three independent experiments. C, 32 P-labeled DNA fragments containing sft4 or SF-1RE were combined with GST-purified ERR␥, digested with 0, 1, 2.5, 5, 10, 20, or 50 ng of proteinase K, and fractioned on an 8% nondenaturing polyacrylamide gel. Note, to get a better resolution, the right gel was exposed for a 6 times lower time period than the left gel. The proteinase K-digested fragments are indicated by arrows. D, 35 S-labeled in vitro transcribed and translated ERR␥ was incubated with excess of sft4 or SF-1RE oligonucleotides, and followed by digestion with the indicated doses of proteinase K the complexes were resolved on a 12% denaturing polyacrylamide gel. The arrows indicate differentially digested fragments. M indicates apparent molecular mass in kDa. The results were reproducible and one representative result from three independent experiments is shown. unlabeled sft4, or SF-1RE, followed by proteinase K digestion. In agreement with the gel shift results the labeled ERR␥ demonstrated distinctly different digestion patterns when bound to sft4 or SF-1RE (Fig. 4D).
These results demonstrate that the sequences of response elements may act as a determining factor for the tertiary conformation of the DNA-bound ERR␥. It remains to be determined whether such DNA induced conformational changes can alter co-activator or co-repressor binding.
Coexpression of SHP and ERR␥-Based on previous reports, SHP and ERR␥ are apparently coexpressed in several tissues, including pancreas and heart (5,6,20). To further explore their expression patterns, Northern blot analyses were performed using SHP or ERR␥ probes. As demonstrated in Fig. 5A, among the human tissues examined both SHP and ERR␥ are expressed relatively highly in pancreas, as expected, and also in the stomach. SHP was coexpressed with relatively lower level of ERR␥ in adrenal medulla, adrenal cortex, and small intestine. To examine expression of ERR␥ and SHP in mouse tissues, a quantitative dot blot (CLONTECH Inc.) was hybridized with either ERR␥ or SHP cDNAs, and the respective signals were measured by densitometry and plotted as relative density. As demonstrated in Fig. 5, B and C, ERR␥ and SHP coexpression was observed in pancreas and heart, as well as submaxillary gland, kidney, epididymus, and prostate.
SHP Represses ERR␥ Transactivation of Its Own Promoter-Since SHP is known to interact with ER␣ and ER␤ and repress their activity (6, 10), we investigated whether SHP can also repress ERR␥ mediated transactivation of its own promoter. Co-transfection studies in CV-1 cells demonstrated that SHP inhibits the ERR␥ response of the SHP promoter in a dose-dependent manner, as expected (Fig. 6). This inhibition was also observed in HEK293 cells, and ERR␣ and ERR␤ mediated transactivation of an ERE was also found to be repressed by SHP (data not shown).
The inhibitory effect of SHP on its various nuclear receptor targets is based on a direct interaction with their AF-2 surface. To determine whether SHP can interact directly with ERR family members, a yeast two-hybrid interaction study was performed. As demonstrated in Fig. 7A LexA fusion proteins including either human or mouse SHP strongly interacted with B42 activation domain fused to ERR␥. ERR␣ also showed a significant interaction with mouse SHP, but its interaction with human SHP was comparatively weak. Both human and mouse SHP showed a much weaker interaction with ERR␤ in this assay. A similar result was obtained by a biochemical approach, which demonstrated interaction of in vitro translated mouse SHP with GST fusions to each of the three ERR subfamily members (Fig. 7B). In agreement with the yeast two-hybrid results, SHP showed a somewhat weaker interaction with ERR␤ in comparison to other ERR family members.
To investigate whether the interaction between SHP and ERR␥ is also dependent on the AF-2 pocket of ERR␥, an AF-2 domain deletion construct of ERR␥ was prepared. As expected from results with other receptors, the interaction of LexASHP with B42 fused wild type ERR␥ is lost with B42 ERR␥ dAF-2 (ERR␥ minus the conserved C-terminal helix 12 motif) (Fig. 8A).
To determine whether the SHP sequences required for interaction with ERR␥ are the same as those required for interaction with other receptors, a series of previously described LexAfused deletion constructs of SHP were used as outlined in Fig.  8B. Yeast two-hybrid interactions revealed that dN-148, which contains the entire interaction domain of SHP, was sufficient for the interaction, whereas, d210 which contains the repressor FIG. 6. SHP represses ERR␥ activity. CV-1 cells were co-transfected with 100 ng of Ϫ2.2 SHP-luc, 500 ng of ERR␥, and increasing doses of SHP, as indicated in the figure. The luciferase activities were measured and normalized against ␤-galactosidase activity. One representative experiment is shown, and similar results were obtained from at least three independent experiments.

FIG. 5. SHP and ERR co-localize both in human and mouse.
A, a human endocrine tissue blot (CLONTECH Inc.) was hybridized with a human SHP fulllength cDNA probe, followed by stripping and hybridization with a full-length mouse ERR␥ probe and ␤-actin was used as a control. Adr.Med, adrenal medulla; Adr.Ctx, adrenal cortex; Small.Int, small intestine. B and C, a mouse master blot (CLONTECH) was hybridized with a fulllength mouse SHP cDNA probe (B), followed by stripping and hybridization with a mouse ERR␥ full-length cDNA probe (C). Following the autoradiogram the blots were analyzed by a densitometric analysis using a Tina 2.0 software and plotted after substracting the background value. Submax.Gland, submaxillary gland; Sk.Muscle, skeletal muscle; Sm.muscle, smooth muscle. domain only, failed to interact with ERR␥ and thus did not increase the reporter activity (Fig. 8C). Overall, these data are consistent with a number of previous reports. As with its other targets, we conclude that the SHP receptor interaction domain interacts directly with the AF-2 surface of ERR␥.
To explore the role of coactivator competition in the repression of ERR␥ activity, a co-transfection assay was performed in HEK293 cells, which were found to exhibit significant coactivator mediated increase in ERR␥ activity. As demonstrated in Fig. 8D, GRIP Ϫ1/SRC-2, a previously described coactivator for all the ERR family members (5,26,27), significantly increased ERR␥ activity on the SHP promoter, and coexpression with SHP repressed this induction in a dose-dependent manner. This result suggests that GRIP-1 and SHP compete for binding the AF-2 pocket of ERR␥, as demonstrated for other receptors (6,10,11).
Recently, a series of mutations affecting the human SHP protein were identified in Japanese subjects with mild obesity (35). These include deletion mutants removing large portions of the protein. An apparent polymorphism present in both normal and obese subjects was also identified. All of the mutant proteins, but not the polymorphic variant, showed significantly decreased ability to inhibit transactivation by the orphan nu-clear receptor HNF-4 (35). As expected the truncated SHP mutants H53 and L98, did not inhibit ERR␥ activity, whereas both wild type SHP, and the polymorphic form R216H substantially reduced ERR␥ activity on SHP promoter (Fig. 9). To identify whether this inability of the truncated SHP mutants to inhibit ERR␥ activity was due to any defect in interaction between the mutants and ERR␥, yeast two-hybrid interactions were performed. As demonstrated in Table I, although wild type SHP and the polymorphic form R216H strongly interacted with ERR␥, both L98 and H53 showed no interaction with ERR␥ confirming that the receptor interacting domain of SHP is essential for its interaction with ERR␥. DISCUSSION Several recent results have identified three nuclear receptors, SF-1, LRH-1, and FXR, as potential regulators of SHP expression. This is consistent with the relatively high levels of expression of SHP in tissues that express these three receptors including liver, adrenal, and pancreas. However, SHP is expressed in additional tissues, including heart, small intestine, stomach, epididymus, and prostate, several of which also express the orphan receptor ERR␥. Thus, it has been previously reported that murine ERR␥ is expressed relatively highly in heart, brain, and kidney, with a lower expression in liver (5). Human ERR␥ transcripts were found in heart, brain, placenta, kidney, pancreas, with a lower level of expression detected in spleen, thymus, prostate, testis, and small intestine (20). In our study of human endocrine tissues, we found that both ERR␥ and SHP are strongly expressed in pancreas, with a lower level of expression for both observed in adrenal medulla and cortex, stomach, and intestine. In mouse tissues, ERR␥ expression was highest in the submaxillary gland, which also showed relatively high levels of SHP. Coexpression of ERR␥ and SHP was also observed in heart, kidney, epididymus, prostate, and pancreas. Based on both this coexpression and its ability to transactivate the SHP promoter, we conclude that ERR␥ is an additional potential regulator of SHP expression. As a consequence of the broad inhibitory functions of SHP, such regulation by ERR␥ could have significant implications for nuclear receptor signaling pathways in a number of tissues.
In addition to identifying a novel target gene for ERR␥, to our knowledge the first for this as yet poorly characterized receptor, these results also define unexpected DNA binding and functional properties for ERR␥. SF-1 requires at least three high affinity binding sites to efficiently transactivate the SHP promoter (16). Although such SF-1 sites have been proposed to be targets for ERR family members, only the sft4 element showed such binding. Unexpectedly, although all three ERR family members bound this element with comparable affinity, only ERR␥ significantly transactivated the SHP promoter in several different cell lines examined. Somewhat similar differential effects have been described in other contexts. Thus, a previous report demonstrated that ERR␥ could efficiently bind a consensus SF-1RE/ERRE, TCAAGGTCA, but could only transactivate it weakly (20). In agreement with this, we found that ERR␥ exhibited insignificant activity on the TR␣ promoter, which contains a perfect ERRE (23), whereas, on a five-copy SF-1RE reporter ERR␥ demonstrated modest activity relative to that on a three-copy sft4 Luc (Fig. 4B). An additional study demonstrated ERR␥ could not activate a palindromic thyroid hormone response element (TRE-Pal) (5), whereas both ERR␣ and ERR␤ are known to bind this element and efficiently activate a reporter containing it (26). In our hands, ERR␥ shows a modest activity on TRE inducing it by 1.5-2-fold in HeLa cells and DNA binding studies indicate that ERR␥ binds TRE-Pal with a similar affinity as it binds sft4 (data not shown). The basis for the differential activities of the three FIG. 7. SHP physically interacts with the ERR isoforms. A, a yeast strain EGY48 which contains an integrated ␤-galactosidase reporter gene controlled by the LexA-binding site was transformed with the indicated LexA and B42 plasmids. The transformants were selected on plates containing appropriate selection markers, and assayed for ␤-galactosidase activity. The result shown is the mean of ␤-galactosidase value from six independent transformant colonies. The error bars indicate standard deviation. B, GST fused ERR␣, ERR␤, and ERR␥ were isolated from bacterial culture and immobilized on glutathione-Sepharose beads. In vitro transcribed and translated [ 35 S]methioninelabeled SHP was incubated with purified GST fused receptors or GST alone as indicated in the figure. The interaction complexes were resolved by a 12% denaturing polyacrylamide gel electrophoresis, and analyzed by autoradiography.
ERR isoforms observed on SHP promoter is yet to be completely elucidated, but our results demonstrate differential effects of different DNA sequences on ERR␥ tertiary structure. Protease sensitivity experiments performed on consensus SF-1RE, sft4, and TRE-Pal bound ERR␥ or ERR␣ (Fig. 4, B and C, and data not shown) confirm this conclusion and suggest that DNA sequences may also act as allosteric modulator of ERR␣ structure. These effects are similar to those previously described for the GR and other nuclear receptors (36 -39).
Although on sft4 TSA induced the activities of ERR␣ and -␤ but not ERR␥, we found that the activities of all three isoforms could be augmented by TSA treatment on ERE and TRE-Pal (Fig. 4A, and data not shown). The allosteric effects of DNAbinding sites on ERR structure suggest a specific mechanism for such differential transactivation. In this hypothesis, the structure of ERR␣ and -␤ bound to sft4 would be relatively permissive for corepressor binding, while the ERR␥-sft4 would be nonpermissive. It is also possible, of course, that allosteric effects of the DNA-binding site could alter interactions with coactivators, or with other DNA bound transcription factors. Detailed further studies will be required to test the potential importance of the differential effects of DNA sequence on ERR functions suggested by the initial results described here.
The current study also demonstrates that SHP inhibits ERR␥ transactivation. As with other receptors (5, 6) this inhibition is a consequence of a direct physical interaction of SHP with the AF-2 surface of ERR␥. Thus, in tissues where ERR␥ is active, SHP should autoregulate its own gene expression. This mechanism should limit the response of not only SHP, but other ERR␥ target genes to signals that increase ERR␥ trans- The luciferase activities were measured and normalized against ␤-galactosidase activity. One representative experiment is shown, and similar results were obtained from at least three independent experiments.
FIG. 9. Naturally occurring SHP mutants fail to inhibit ERR␥ activity. CV-1 cells were transfected with 100 ng of Ϫ2.2-kb SHP promoter, 500 ng of ERR␥ expression plasmid, and 100 ng of indicated naturally occurring SHP mutants as described in text and elsewhere (35). The luciferase activities were measured and normalized against ␤-galactosidase activity. One representative experiment is shown, and similar results were obtained from at least three independent experiments.

TABLE I Interaction of ERR␥ with naturally occurring SHP mutants in yeast
The indicated B42-and LexA-plasmids were transformed into a yeast strain containing an appropriate LacZ reporter gene, as described in "Experimental Procedures." At least six separate transformants from each transformation were transferred to indicator plates containing 5-bromo-4-chloro-3-indoyl ␤-D-galactosidase, and reproducible results were obtained using colonies from two separate transformations. SB, strongly blue colonies after 6 -8 h of incubation; W, white colonies. N indicates empty vectors. Wt, wild type human SHP; R216H, replacement of arginine codon 216 by histidine; H53, deletion of 10 bases starting at codon 53 for histidine; L98; deletion of 9 bases and insertion of a dinucleotide AC at codon 98 for leucine. activation. The identification of such signals and such gene targets will be necessary to test this hypothesis. While this report was under review two reports demonstrated an antiestorgenic therapeutic drug, 4-hydroxytamoxifen, as a ligand for ERR␤ and ERR␥ (40,41). These reports, together with our report indicate that the SHP gene promoter may be a bona fide target of 4-hydroxytamoxifen. Further studies are needed to characterize such a possibility.