Identification of Nuclear Orphan Receptors as Regulators of Expression of a Neurotransmitter Receptor Gene*

Nuclear orphan receptors are known to be important mediators of neurogenesis, but the target genes of these transcription factors in the vertebrate nervous system remain largely undefined. We have previously shown that a 500-base pair fragment in the first intron of the GRIK5 gene, which encodes the kainate-preferring glutamate receptor subunit KA2, down-regulates gene expression. In our present studies, mutation of an 11-base pair element within this fragment resulted in a loss of nuclear protein binding and reverses negative regulation by the intron. Using yeast one-hybrid screening, we have identified intron-binding proteins from rat brain as COUP-TFI, EAR2, and NURR1. Gel shift studies with postnatal day 2 rat brain extract indicate the presence of COUP-TFs, EAR2, and NURR1 in the DNA-protein complex. Competition assays with GRIK5-binding site mutations show that the recombinant clones exhibit differential binding characteristics and suggest that the DNA-protein complex from postnatal day 2 rat brain may consist primarily of EAR2. The DNA binding activity was also observed to be enriched in rat neural tissue and developmentally regulated. Co-transfection assays showed that recombinant nuclear orphan receptors function as transcriptional repressors in both CV1 cells and rat CG4 oligodendrocyte cells. Direct interaction of the orphan receptors with and relief of repression by TFIIB indicate likely role(s) in active and/or transrepression. Our findings are thus consistent with the notion that multiple nuclear orphan receptors can regulate the transcription of a widely expressed neurotransmitter receptor gene by binding a common element in an intron and directly modulating the activity of the transcription machinery.

Nuclear orphan receptors are known to be important mediators of neurogenesis, but the target genes of these transcription factors in the vertebrate nervous system remain largely undefined. We have previously shown that a 500-base pair fragment in the first intron of the GRIK5 gene, which encodes the kainate-preferring glutamate receptor subunit KA2, down-regulates gene expression. In our present studies, mutation of an 11-base pair element within this fragment resulted in a loss of nuclear protein binding and reverses negative regulation by the intron. Using yeast one-hybrid screening, we have identified intron-binding proteins from rat brain as COUP-TFI, EAR2, and NURR1. Gel shift studies with postnatal day 2 rat brain extract indicate the presence of COUP-TFs, EAR2, and NURR1 in the DNA-protein complex. Competition assays with GRIK5-binding site mutations show that the recombinant clones exhibit differential binding characteristics and suggest that the DNA-protein complex from postnatal day 2 rat brain may consist primarily of EAR2. The DNA binding activity was also observed to be enriched in rat neural tissue and developmentally regulated. Co-transfection assays showed that recombinant nuclear orphan receptors function as transcriptional repressors in both CV1 cells and rat CG4 oligodendrocyte cells. Direct interaction of the orphan receptors with and relief of repression by TFIIB indicate likely role(s) in active and/or transrepression. Our findings are thus consistent with the notion that multiple nuclear orphan receptors can regulate the transcription of a widely expressed neurotransmitter receptor gene by binding a common element in an intron and directly modulating the activity of the transcription machinery.
Glutamate is the principal excitatory neurotransmitter in the vertebrate central nervous system ,and its physiological effects are mediated through the activation of N-methyl-Daspartate, ␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA), 1 and kainate receptors (1). Although assigning a specific role for kainate receptors has been difficult, the presence of functional kainate receptors in neural cells (2)(3)(4) and their involvement in synaptic transmission (5)(6)(7) have been demonstrated. It has also been shown that kainate/AMPA receptor agonists promote the survival of rat spinal cord cells (8) and kainate-generated currents were detected in human fetal astrocytes (9). Despite overlapping binding properties between kainate and AMPA receptor classes, kainate preferentially promotes neurite arborization in rat cerebellar neurons in culture, an effect not reproduced by AMPA, thus demonstrating differential effects on neural cell growth and survival (10). In addition, kainate receptors are found both in the rat embryonic neural tube as well as in cell cultures derived from this tissue (11), and expression of the subunits GluR6 and KA2 is upregulated by culture conditions that stimulate cell differentiation (11). Taken together, these observations support the notion that kainate receptors may play important role(s) in neural cell differentiation.
Nuclear orphan receptors constitute the majority of members in the nuclear receptor superfamily (12). These proteins are of great interest, because the lack of identified ligands presents a challenge to the elucidation of their function. COUP-TFI was shown to antagonize retinoic acid-induced differentiation of teratocarcinoma cell lines (13). More recently, targeted gene ablation has provided evidence that COUP-TFI is an important regulator of neuronal development, because COUP-TFI-null mice display deficiencies in the formation of cranial nerves IX and X and do not survive after birth (14). Deletion of NURR1 resulted in a failure to generate midbrain dopaminergic neurons, a condition that was also lethal in neonatal mice (15). To understand the function of members of the nuclear orphan receptor family in neural development, it is important to identify neural-specific genes that are targets of orphan receptor action and whose products are themselves involved in neural cell function and possibly maturation.
Studies have previously demonstrated the expression of the neural-specific KA2 subunit to be widespread in the rat brain (16,17); however, the factors that regulate KA2 subunit expression have not been characterized. Elucidation of these factors is necessary to understand the physiological basis of changes in kainate receptor composition in neural cells. In our effort to address regulatory factors of KA2 gene expression, we have cloned the rat GRIK5 gene (glutamate receptor ionotropic kainate 5), which encodes the KA2 receptor, and shown that neural cell-specific activity of its TATA-less promoter is largely dictated by sequences contained within 2 kb of its upstream region (18). Although most studies of transcriptional regulation have localized cis-acting regulatory regions in 5Ј-flanking sequences, there are many examples of specific regulatory functions, mediated by both positively and negatively acting genetic elements, that have been attributed to introns (Refs. 19 -21 and references therein). Such elements appear to be located most commonly in the first intron, perhaps because of its proximity to the promoter and to the cap site of the gene. Transfection studies with rat GRIK5 constructs also revealed the presence of negative regulatory element(s) within the first intron (18). Inclusion of this intron in reporter constructs represses, in a position-independent manner, GRIK5 promoter activity not only in neural cells but also in HeLa cells (18). In contrast to RE1-silencing transcription factor, REST/neuron-restrictive silencer factor, NRSF, which was found to control neural-specific expression of a sodium channel (22,23) and the AMPA receptor subunit GluR2 (24), the repressor(s) acting on the GRIK5 intron is not a determinant of neural-specific KA2 expression but rather a modulator of transcript levels in KA2-expressing cells.
In this study, we show that binding of nuclear proteins to an 11-bp element in this intron is sufficient for transcriptional repression of GRIK5 promoter activity. We identify several nuclear orphan receptors from rat brain that are capable of binding this GRIK5 element. In addition, we also examine their binding characteristics in various neural cell types and, using binding site mutations, establish a profile of binding activity in the rat brain and neural cells to determine the major component of the repressor-DNA complex. Protein interaction assays with TFIIB suggest that each of these receptors can participate in active repression of transcription. Previous studies of other neural-specific genes, mouse arrestin (25), and preproenkephalin (26) have implicated COUP-TF as a positive modulator of gene transcription. However, there is now much evidence supporting the notion that transcriptional repression is as common and important as transcriptional activation (27). Our observations thus represent the first description of negative regulation of a widely expressed neurotransmitter-gated ion channel gene by members of the nuclear orphan receptor family.

Preparation of Nuclear Protein Extracts from Cultured Cells and
Tissues-CG-4 cells were cultured as described previously by Louis et al. (28) and Patneau et al. (3) and were harvested 2 days after replating (80% confluency). Cortical type-1 astrocyte cultures were prepared as described previously in accordance with the National Institutes of Health Animal Welfare guidelines (29) and were harvested after 12 days in culture. Cerebellar granule cell cultures were prepared as described previously (30) and were harvested after 2 and 7 days in culture. Nuclear protein extracts from CG-4 cells, astrocytes, and granule cells were prepared with the protocol described by Dignam et al. (31). The nuclear proteins from tissues were extracted by using a rapid method developed by Deryckere and Gannon (32) with modifications. Rat tissues were dissected out, frozen in liquid nitrogen, and shattered between layers of aluminum foil. Approximately 500 mg of tissue powder was homogenized in 5 ml of solution A (0.6% Nonidet P-40, 150 mM NaCl, 10 mM HEPES, pH 7.9, 1 mM EDTA, 0.5 mM AEBSF and homogenized with five strokes with a Teflon pestle. The homogenate was centrifuged for 30 s at 2,000 rpm. The supernatant was centrifuged for 5 min at 5,000 rpm. The pelleted nuclei were resuspended in 500 l of solution B (25% glycerol, 20 mM HEPES, pH 7.9, 420 mM NaCl, 1.2 mM MgCl 2, 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM AEBSF, 5 g/ml pepstatin, leupeptin, and aprotinin) and incubated on ice for 20 min. The lysed nuclei were transferred to a microcentrifuge tube and centrifuged for 15 s. The supernatant was aliquoted in small fractions, frozen rapidly in liquid nitrogen, and stored at Ϫ80°C. The protein concentration of nuclear extracts was determined by the Bradford assay.
DNA Gel Mobility Shift Assays-The sequences of probe 62 and other competitors are given in Table II. Probe 62 is a 27-bp probe containing the putative 11-bp GRIK5 silencer element and corresponds to the sequence between positions ϩ3206 and ϩ3232 in the rat GRIK 5 gene relative to the start of exon 1 (18). The probes were end-labeled with [␥-32 P]ATP (NEN Life Science Products), purified on G50 columns, and used in gel shift assays. The reactions were carried out in a total volume of 20 l of binding buffer containing 25 mM HEPES, pH 7.5, 60 mM KCl, 10% glycerol, 5 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, 50 g/ml poly(dI-dC) (Amersham Pharmacia Biotech). For in vitro translated recombinant clones, the reaction buffer contained 20 mM HEPES, pH 7.9, 50 mM KCl, 10% glycerol, 2 mM MgCl 2 , 1 mM dithiothreitol, 0.05% Nonidet P-40, 0.1 mg/ml bovine serum albumin, and 0.1 mM EDTA. The binding reactions were incubated on ice for 30 min with 6 g of nuclear proteins from cell and tissue extracts or 3-8 l of crude translation mixture. Then 10 fmol of labeled probe was added to each reaction and incubated for another 30 min on ice. For supershifts, 1-2 l of antiserum was added to the reactions 15-30 min prior to loading. After adding 2 l of 0.1% bromphenol blue loading dye, each reaction was directly loaded onto a 4% nondenaturing polyacrylamide gel and electrophoresced at 100 V. The gels were then dried and autoradiographed on x-ray film (Kodak) for 16 -24 h. In competition experiments, the wild type probe was 32 P-labeled, and unlabeled competitor oligonucleotides were used in 50-and 100-fold molar excess. The intensity of the shifted bands was quantitated on a PhosphorImager (Molecular Dynamics). In a separate set of experiments, complementary Sp1 consensus oligonucleotides (5Ј-ATTCGATCGGGGCGGGGCGAGC-3Ј and 5Ј-GCTCGCCCCGCCCCGATCG-3Ј) were used as a control in gel shift assays. RXR antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).
Construction of Mutated Intron 1-CAT Plasmids-A common 5Ј primer (5Ј-ATGGAGCTCAATCTGGCTTTGAA-3Ј) and a series of 3Ј primers containing different mutations (m11-21) were used for polymerase chain reaction (PCR). All primers contained a SacI site at their 5Ј ends for cloning of PCR products into SacI sites. PCR was performed with these primers using 5.4-CAT plasmid DNA as template (18). The annealing temperatures were different, depending on the primer pairs (m11-13,15-21 at 45°C; m11-13 and m15-17 at 50°C; all others at 55°C). Eight PCR products with different mutations at distinct positions were generated, and each one was used as a cassette to replace the corresponding wild type sequence in 5.4-CAT between two SacI sites. The 5.4-CAT DNA lacking the 2-kb SacI fragment (5.4-⌬SacI-CAT) was generated by digestion of 5.4-CAT with SacI, and individual SacI-digested PCR products containing mutations were ligated with 5.4-⌬SacI-CAT.
Transformants were screened by hybridizing with individual mutated 32 P-labeled primers as probes to select for desired clones. The orientation of positive clones was determined by restriction digestion, and all the mutations verified by dideoxynucleotide DNA sequencing. The resulting plasmid DNAs were used in transient transfection assays (see below).
Transient Transfections-All expression plasmids used in transfections were purified with Qiagen DNA purification kits. A plasmid containing a lacZ gene driven by the RNA polymerase II gene promoter (pPolIIplacF.␤-gal) was simultaneously transfected with all the CAT constructs to correct for variations in transfection efficiency. CAT constructs (10 g) and pPolIIplacF⅐␤-gal (2 g) were used to transiently transfect CG-4 cells using LipofectAMINE reagent (Life Technologies, Inc.). Transfections were carried out for 8 h, and the culture medium was replaced. Cells were harvested 38 h after transfection and assayed for CAT and ␤-galactosidase activities as described previously (18). Levels of CAT activity were determined on a PhosphorImager. For nuclear orphan receptor co-transfections, reporter plasmid was transfected with varying amounts of orphan receptor cDNA cloned into the expression vector pcDNA3 (Invitrogen). DNA concentrations were balanced with the empty expression vector pcDNA3. pBLCAT51 was generated by ligating, between the HindIII and XbaI sites of the thymidine kinase promoter-driven reporter pBLCAT5, synthetic oligonucleotides containing the sequence of wild-type probe 62 (see Table I). The construction of ␤-actin-I-CAT has been previously described (18). For TFIIB co-transfections, reporter plasmids were transfected with either 0.2 or 2 g of RSV-TFIIB or negative control RSV-Luciferase. CAT activity was determined by scintillation counting, and values were normalized against total protein content instead of ␤-galactosidase activity, because COUP-TF expression vectors have been shown to repress the expression of ␤-galactosidase expression vectors (33).
One-hybrid Screen for Binding Protein(s)-The in vivo screening strategy was carried out using the Yeast Matchmaker One-hybrid system (CLONTECH). This required generating a yeast reporter strain containing three tandem copies of wild type or mutated probe 62 sequences (to eliminate false positives) cloned upstream of yeast promoters driving two different reporter genes, HIS3 and lacZ. The dual reporter yeast strain containing the wild type 62 sequence was used as a host for introducing a postnatal day 2 rat brain GAL4 activation domain cDNA library in pGAD10 (custom generated by CLONTECH). After introduction of the cDNA library (corresponding to approximately 150 ϫ 10 6 independent clones) into yeast cells by the LiAc method, cells were selected on SD/ϪHis/ϪLeu plates in the presence of 45 mM 3amino-1,2,4-triazole. The transformation efficiency was 12 ϫ 10 4 cfu/g DNA, and putative positives were screened for lacZ expression by a filter assay (CLONTECH). The cDNA clones from these colonies were isolated, amplified, and reintroduced into the yeast reporter strain carrying the mutated probe 62 reporter. The confirmed positives were analyzed by restriction mapping and sequencing. The inserts were subsequently cloned into pcDNA3 (Invitrogen) for in vitro translation and co-transfection assays. Expression plasmid for rat EAR2 cDNA carried a modified Kozak sequence upstream of the coding region for improved translation efficiency.
In Vitro Translation and Immunoprecipitations-Coupled transcription-translation of nuclear receptor constructs was carried out in rabbit reticulocyte lysates according to the manufacturer's directions (Promega). For recombinant proteins used in gel shift assays, a 1 mM amino acid stock (Promega) was used in the translation reactions. For immunoprecipitation, translation products were labeled with [ 35 S]methionine (Amersham Pharmacia Biotech). Equal volumes of the labeled products were analyzed by SDS-polyacrylamide gel electrophoresis and autoradiographed. Relative intensities of the bands were estimated by scanning densitometry, so that equivalent amounts could be used in immunoprecipitation reactions. Approximately 3-6 l of crude translation mixture was mixed with 4 l of TFIIB antibody (Santa Cruz) and 20 l of protein A-agarose (Santa Cruz) in a total volume of 150 l, containing binding buffer of the following composition: 10% glycerol, 50 mM HEPES, pH 7.9, 100 mM KCl, 6 mM magnesium acetate, 5 mM EDTA, 0.5 mM dithiothreitol, 0.5% nonfat dry milk. The mixtures were rocked at 4°C for 1-2 h, after which the tubes were centrifuged for 5 min at 500 ϫ g and washed three times with cold binding buffer. The agarose pellets were then resuspended in 20 l of SDS loading buffer and boiled prior to SDS-polyacrylamide gel electrophoresis analysis.

An 11-bp Segment Is Responsible for Repressor Activity of the
First Intron-GRIK5 promoter constructs have previously been shown to be expressed in cells of the oligodendrocyte line CG4 (18). Fig. 1A shows two GRIK5 constructs used in our studies. 2-CAT consists of the GRIK5 promoter and 2 kb of upstream sequences cloned into a promoterless CAT vector, pCATbasic, whereas 5.4-CAT contains the 3.4 kb first intron cloned downstream of the 2-kb promoter fragment. As shown in Fig. 1B and as described previously (18), inclusion of the 3.4-kb first intron resulted in a 3-fold repression of reporter gene activity. Preliminary analysis of mice bearing 7.7 kb of GRIK5 sequences that contain the 4-kb promoter fragment and the 3.4-kb first intron cloned upstream of a ␤-galactosidase reporter gene indicates that the inclusion of the intron results in a generalized decrease in reporter gene expression in the brain, when compared with transgenic mice bearing only a 4-kb promoter fragment. 2 This suggests that the intronic element also acts to repress GRIK5 transcription in vivo. Extensive deletion analysis and transfection assays as well as DNase 1 footprint assays have previously delineated a single 24-bp region (3207-3231) in this intron that binds nuclear protein from CG4 cells. The role of this region in transcriptional repression is thus the main focus of our present studies.
A series of mutations was generated in 5.4-CAT in which 1-3 nucleotides were systematically replaced from 5Ј to 3Ј with their respective complements (Fig. 1B). Transfection of these constructs in CG-4 cells showed that changes in 10 nucleotides significantly affected repressive capacity of the intron (Fig. 1B). When clustered together in m11-13,15-21, reporter activity was restored to levels similar to 2-CAT, indicating a complete loss of repressor function of the intron. Although the CAT activities of mutant constructs m18,19, m15-17, m11-13, and m20,21 were clearly lower than that of 2-CAT, they were still significantly higher than that of 5.4-CAT, indicating that these mutations had partially affected the activity of the repressor. In contrast, mutations in m22-25, m9,10, and m14 were ineffective, suggesting that these nucleotides were not important for mediating repressor activity (Fig. 1B). Thus the positions 11-13,15-21 form a putative GRIK5 silencer element.
To correlate the transfection results with DNA binding, a se-ries of oligonucleotide probes containing 5Ј to 3Ј nucleotide changes was also tested in gel shift assays, using nuclear extracts from CG4 cells. The wild type probe 62 corresponds to the region between ϩ3206 and ϩ3232 relative to the start of transcription, a region previously shown to bind nuclear protein from CG4 cells (18). As summarized in Table I, the binding potential of probes that carried mutations outside of the 11-bp silencer region identified by transfections (Fig. 1B) was not significantly different from that of wild type probe 62. Changing 10 bp in m11-13,15-21 rendered the probe incapable of binding nuclear protein (Table I). This is consistent with the complete loss of negative regulation by the same cluster of mutations in transient transfections (Fig. 1B). Mutations in probe 62, which affected 3 bp or fewer within this 11-bp silencer region, such as m11,12, m18,19, and m20 -23, severely impaired but did not abolish protein binding. These mutations also resulted in elevated reporter activity in transfections, indicating that repressor activity of the nuclear protein(s) can still be partially mediated by weakened binding to the element. It is also noteworthy that m14, despite being situated within the 11-bp region, was no different from wild type, both in transfections and in gel shift assays ( Fig. 1B and Table I).
Characterization of Binding Activity in Rat Tissues and Neural Cells-Because the GRIK5 promoter is known to be active only in neural cells, we asked whether there was a differential expression of the intron-binding proteins in vivo. Nuclear extracts were prepared from various tissues of P6 rats and equivalent amounts of nuclear protein used in gel shift assays. As illustrated in Fig. 2A, a clear abundance of binding activity in the brain over peripheral tissue is observed, suggesting a neural-selective role for the binding protein(s). To assess possible developmental changes in binding activity, gel shift assays were performed using equivalent amounts of brain and cerebellar nuclear protein isolated from rats at various ages. Fig.  2B shows that in both these regions, binding activity dramatically declines toward adulthood, indicating that binding activity of these protein(s) is closely associated with maturation of neural tissue. To determine whether the binding activity exhibited cell-type selectivity among neural cells, nuclear extracts were prepared from purified glial cells, immature dorsal root ganglion neurons, cerebellar granule neurons, and PC12 cells. As shown in Fig. 2 (C and D), binding activity was abundant in all of these preparations, indicating that multiple cell types present in the developing central and peripheral nervous systems possess nuclear factor(s) potentially capable of regulating GRIK5 expression. In addition, the decrease in binding activity from brain tissue during development is also observed in cultured cells that are allowed to undergo neuronal differentiation (Fig. 2D), providing evidence for an early role in neural cell development.
Isolation of Nuclear Orphan Receptor cDNA from Rat Brain-Using the yeast matchmaker one-hybrid screen, with tandem repeats of probe 62 as bait and a P2 rat brain cDNA library (without cerebellum), we isolated a number of overlapping clones encoding members of the family of nuclear orphan receptors. The most highly represented clones, upon BLAST analysis, showed 95% homology to the recently cloned rat COUP-TFI, EAR2, and NURR1. The sequences for rat COUP-TFI, EAR2 (a distant relative of the COUP family), and NURR1 have been previously reported (35,36). 3 The full-length cDNA clones were subcloned into expression vectors and translated in vitro, generating products between 40 and 66 kDa (see Fig. 7), which agrees with the known molecular masses of these receptors. Gel shift assays performed with the recombinant orphan receptors indicated that each of the recombinant orphan receptors forms specific DNA-protein complexes in a dose-dependent fashion on wild type probe 62 (data not shown).
Characterization of Nuclear Proteins in Brain and Cell Extracts-To confirm the presence of COUP-TF family members and NURR1 in the brain DNA complexes, supershift experiments with recombinant orphan receptors were first designed to demonstrate antibody specificity among members of the COUP family. Fig. 3A shows that the COUP-TFI antibody does not discriminate between COUP-TFI and COUP-TFII, whereas those for COUP-TFII and EAR2 show no cross-reactivity with other members of the family. The activity of the NURR1 antibody has been previously demonstrated (37).
Gel shift experiments were then performed with these antibodies on probe 62, using nuclear extracts prepared from P2 rat brain as well as CG4 cells. As shown in Fig. 3B, supershifts were produced by each of the antibodies, except for retinoid X receptor (RXR). In contrast to recombinant proteins, the intensity of the I Effect of probe mutation on nuclear protein binding Summary of effect of changes in probe sequence on binding by CG4 nuclear protein, as analyzed by gel shift assays. Equivalent amounts of probe and nuclear protein were used, and the intensities of the complexes formed are expressed relative to the intensity of the wild type complex. ϩϩϩ, intensity of complex formed on wild type probe 62 (wt 62); ϩϩ, between 50 and 80% of wild type complex; ϩ, Ͻ50% wild type complex; ϩ/Ϫ, barely detectable specific complex; Ϫ, specific complex undetectable. Numbers refer to positions of replaced nucleotides relative to the 5Ј end of the DNA sequence. Underlining indicates the 10 nucleotides determined by transient transfection (Fig. 1B)   DNA-protein complex from cell extracts was surprisingly enhanced by addition of specific antisera, so that inferences on the major component of the heterogeneous complex could not be made from these results. The supershifts nonetheless indicate that in cells that express the KA2 gene, COUP-TFII, EAR2, NURR1, and, very likely, COUP-TFI are able to bind the KA2 element without forming a heterodimer with RXR, although heterodimerization with RXR has been reported previously (38). As a positive control for the RXR antibodies, the binding of liver RXR to ␤RA probe was demonstrated in a separate experiment (Fig.  3C). Finally, Fig. 3D illustrates that different neural cells in primary culture possess both COUP and NURR1 binding activity. Further supershift assays of equivalent protein concentrations using specific antibodies against COUP-TFI and COUP-TFII, showed that rat brain and cerebellum at P2 contained a greater relative abundance of both COUP-TFs in the brain than in cerebellum. Rat dorsal root ganglia at E15 contained markedly higher levels of COUP-TFII activity than COUP-TFI, whereas EAR2 activity showed little relative difference among the tissues examined (data not shown).
DNA Binding Characteristics of CG4 and P2 Brain Extracts-Based on supershift analysis, it is clear that more than one nuclear orphan receptor forms a complex on probe 62. However, it is still possible that one of the receptors may constitute a predominant component of the complex, because orphan receptor expression patterns are distinct in the brain (39). Characterization of individual binding profiles may pro-vide clues to the composition of the heterogeneous complex from mixed extracts. It is known that COUP-TFs exist in solution as dimers and also bind to consensus response elements as dimers in gel mobility shift assays (40). COUP-TFs are able to assume a variety of conformations to accommodate spatial changes, such as the spacing between direct repeats, although the relative affinities for direct repeats have been ranked as follows: DR1 Ͼ DR4 Ͼ DR0 (40). This is confirmed in gel shift experiments with recombinant receptors using probe 62 and 100-fold molar excess of consensus DR0 -5 as competitors (Fig.  4A). As expected, COUP-TFI and COUP-TFII binding is competed by all DRs tested, but relative differences in binding affinity among these DRs are also apparent. The strongest competitors are DR1 and DR2. A similar profile is exhibited by EAR2, which is considered a more distant member of the COUP family (41). NURR1, on the other hand, shows weaker binding to probe 62 and is not observed to be affected by competitors with perfect DRs. The lack of competition may be explained by the absence of the extended half-site AAAGGTCA in the competitors and would be consistent with reports that members of the NGFI-B/Nor-1/NURR1 family bind DNA as monomers and recognize an extended consensus sequence called the NGFI-Bresponse element (AAAGGTCA; Ref. 42).
Because direct repeats effectively compete for binding to probe 62, this GRIK5 element should be expected to contain AGGTCA repeats as well. The GRIK5 element contains an imperfect DR1, as represented by probe 62 in Table II. To study  1-7) and CG4 cell nuclear extracts (lanes 8 -14) R p , pan-specific anti-RXR antibody; R␣, RXR␣-specific antibody. C, presence of RXR binding to ␤RA probe (Table II)  the binding preferences of each of the recombinant clones to probe 62, the sequence of the half-sites was mutated, and the resulting oligonucleotides were used in gel shift experiments as unlabeled competitors. Changing the relative orientation of these half-sites to an inverted repeat (IR) or everted repeat (ER) produces differential effects by each of the recombinant orphan receptors as illustrated in Fig. 4B. Neither IR nor ER is able to compete for binding to probe 62 by COUP-TFI and COUP-TFII, indicating that both receptors bind as obligate dimers with a strict requirement for direct repeats. On the other hand, EAR2 binds IR partially but not ER, indicating asymmetrical binding or a preference for inverted repeats over everted ones, although direct repeats still bind best. Similarly, NURR1 does not bind ER, but IR competes as effectively as probe 62 itself. This provides evidence that NURR1 binds as a monomer to the right half-site only, because it bears the NGFI-B-response element. In the remaining competitors, the left halfsite is scrambled, leaving only the right half-site intact (N1 and N2). Once again, COUP-TFI and -TFII were indistinguishable in binding to these mutants. EAR2 is able to bind partially as a monomer requiring a 5Ј 2A extension, because it binds N1 but not N2. Interestingly, the requirement for the extension can be overridden when a DR1 is present (binding to N3; Fig. 4B). NURR1 binds only N1 and not N2 or N3, confirming its binding as a monomer with an absolute requirement for the extended halfsite. Taken together, the results show that the binding of EAR2 to probe 62 appears unique in its versatility, i.e. partial binding to one half-site with a conditional dependence on a 2A extension.
Using the above battery of competitors as a tool to distin-guish the various recombinant receptors, gel shift experiments were performed with nuclear extracts from P2 rat brain and CG4 cells to determine whether one particular orphan receptor formed a predominant component of the DNA-binding complex. Fig. 5 shows that the DNA-protein complexes from cells and tissue extracts are effectively competed by DRs and N3. This rules out NURR1 as a major component. Partial binding to IR and N1 by the extracts closely resembles the binding profile of EAR2 (compare Fig. 5 with Fig. 4B). Thus, although COUP-TFI, TFII, and NURR1 were all detected by supershift analysis, these gel shift competitions provide evidence that one receptor type, which is likely to be homomeric EAR2, may predominate in the DNA-bound complex.

Differential Effects of Recombinant Nuclear Orphan Receptors in Transient
Transfections-To determine whether each of the recombinant clones functions as a repressor on the GRIK5 binding site, transient co-transfections were performed in Green monkey kidney cells (CV1) using a CAT construct carrying a single copy of the GRIK5 silencer cloned immediately upstream of the thymidine kinase promoter (pBLCAT51). In contrast to previous reports showing NURR1 in transactivation (43,44), NURR1 represses pBLCAT51 activity in CV1 cells (Fig. 6A). It was observed that the basal activity of pBLCAT51 was not significantly different from that of the parent vector pBLCAT5 (not shown), indicating that endogenous binding and transcriptional modulation at the GRIK5 silencer in these cells, if any, was undetectable. Fig. 6A shows that co-transfection of recombinant COUP-TFI, EAR2, and NURR1 represses the activity of pBLCAT51 in a dose-dependent manner. However,  (Table II) for binding to wild type probe 62 by recombinant orphan receptors. B, comparison of competitive potential of half-site changes (Table II)  when the recombinant orphan receptors are instead co-transfected with reporter constructs containing the full-length intron 1 downstream of the ␤-actin promoter (␤-actin-I-CAT), only COUP-TFI and EAR2 showed repressor activity (Fig. 6B). The reason for this discrepancy for NURR1 is unclear, but the results appear consistent with gel shift competitions which implicate EAR2, a distant member of the COUP-TF family, as a major component of the complex bound to the GRIK5 element.
We also wanted to examine the effects of recombinant orphan receptors on GRIK5 promoter-driven reporter expression. Transcriptional repression of GRIK5 promoter activity in CG4 cells is shown to be dependent on the silencer element. In the experiments depicted in Fig. 6C, CG4 cells were transfected with 5.4-CAT or 5.4-CAT with a mutated orphan receptorbinding site (m11-13,15-20 -CAT; see Fig. 1B) in the presence of expression vectors for COUP-TFI, EAR2, or NURR1. The results show that the repression of CAT activity by recombinant COUP-TFI and EAR2 is significantly more pronounced with 5.4-CAT than with the mutant construct m11-13,15-21-CAT. This indicates that a functional intronic COUP-binding site facilitates transcriptional repression in the context of the native GRIK5 promoter. NURR1 overexpression also repressed reporter activity of both the wild type and mutated reporter constructs; however, binding site dependence could not be demonstrated. It may be possible that the activity of recombinant NURR1 is promoter-dependent, and our result may suggest a FIG. 5. Gel shift competition analysis of CG4 cell and P2 rat brain nuclear extract using the DR oligonucleotides and halfsite sequence mutations as in Fig. 4. 6 l of nuclear extract was used in binding reactions for all lanes. Comp, 100-fold molar excess unlabeled competitor whose sequences are shown in Table II. The autoradiographs shown are have been deliberately underexposed (note DR0 -DR5) to accommodate the intense control and N2 signals. Longer exposures show preferential competition by DR1 and DR2 more clearly.

TABLE II
Oligonucleotides used for gel mobility shift assays and competition studies The core sequence is shown for all oligonucleotides used in this study for gel mobility shift competition assays. AGGTCA-like motifs are underlined, and orientations are indicated by arrows. Mutated nucleotides are in lowercase bold lettering. DR0 -5, direct repeats with 0 -5 spaces; 62, probe sequence corresponding to region between ϩ3206 and ϩ3232 in the GRIK5 gene; IR, inverted repeat; ER, everted repeat; N1, N2, and N3, novel variants of probe 62; UR, unrelated sequence in the first intron, corresponding to region between ϩ2660 and ϩ2683 in the GRIK5 gene. ␤RA, retinoic acid-responsive element (25).

FIG. 6. Effect of co-transfection of recombinant orphan receptors on CAT activity in CV1 and CG4 cells.
A, co-transfection of recombinant receptors with pBLCAT51. Data are expressed as percentages of controls, in which pBLCAT51 is co-transfected with 4 g (for COUP-TFI and EAR2) or 6 g (for NURR1) of the empty expression vector pcDNA3. *, p Ͻ 0.005 versus control (Student's t test). Values are presented as the means Ϯ S.E. of at least three independent experiments. B, co-transfection of receptors with ␤-actin-I-CAT. Controls contained 4 g (for COUP-TFI) or 6 g (for EAR2 and NURR1) of pcDNA3. *, p Ͻ 0.005 versus controls (Student's t test). Values are presented as the means Ϯ S.E. of at least three independent experiments. C, effect of co-transfection of recombinant orphan receptors in CG4 cells using CAT constructs driven by the GRIK5 promoter. 20 g of 5.4-CAT (empty bars) or mutated construct m11-13,15-21-CAT (filled bars) were cotransfected with expression plasmids for COUP-TFI (2 g), EAR2 (1 g), or NURR1 (1g). Data are expressed as percentages of respective controls in which the empty expression vector pcDNA3 was co-transfected in place of recombinant orphan receptors. *, p Ͻ 0.05 versus mutated construct (Student's t test). Values are presented as the means Ϯ S.E. of at least three independent experiments. preference for transrepression in CG4 cells. It should also be noted that COUP-TFs have been shown to repress basal and activator-dependent transcriptional activities of various promoters with binding sites placed upstream or downstream of these promoters (63), whereas DNA-dependent NURR1 action appears to favor an upstream site (Fig. 6, B and C).
TFIIB Is a Target for COUP-TFs, EAR2, and NURR1-In the course of the co-transfection experiments, it was also observed that although basal activity of pBLCAT51 was not significantly different from that of the vector pBLCAT5 in CV1 cells in the absence of recombinant orphan receptors, co-transfection of the recombinant receptors repressed the activity of the parent vector pBLCAT5 by approximately 50% of control values for COUP-TFI and EAR2 and by 54% for NURR1 (not shown). Repression of CAT activity in both wild type 5.4-CAT and m11-13,15-21-CAT was also observed in CG4 cells. There is evidence in the literature that COUP-TF acts by active repression on DR3-DR5 response elements (33,45). The mechanism of active repression involves transcriptional silencing by interfering with the formation of the transcription complex or competing with other activators for limiting coregulators. One member of the transcription initiation complex, TFIIB, has been demonstrated to be a target of COUP, estrogen receptor, and progesterone receptor (46,47). However, the detailed mechanism by which COUP-mediated transcriptional repression occurs remains unclear. We wanted to determine whether there were detectable differences in the interaction of the various orphan receptors with the basal transcription machinery which might explain the transfection results. Co-immunoprecipitation experiments were performed using 35 S-labeled recombinant receptors and human TFIIB. As shown in Fig. 7, each of the receptors is co-precipitated with TFIIB by the anti-TFIIB antibody, indicating that each orphan receptor is capable of directly interacting with TFIIB in vitro. This indicates that each orphan receptor is capable of either transrepression and/or active repression. 35 S-Labeled luciferase did not interact with TFIIB and serves as a negative control against nonspecific binding (Fig. 7).
Based on the interaction of recombinant nuclear orphan receptors with TFIIB, we hypothesized that the overexpression of TFIIB in intact cells should be expected to enhance GRIK5 promoter activity by preventing the repressor function of endogenous nuclear orphan receptors. Fig. 8 shows that co-transfection of a human RSV-TFIIB expression vector into CG4 cells relieves repression by the wild type intron (5.4-CAT), while having an insignificant effect on the mutated intron (m11-12, 13-15-CAT). This indicates that the effect of recombinant TFIIB is specifically dependent on the presence of endogenous DNA-bound nuclear orphan receptors and that this interaction is likely to contribute to the mechanism of negative regulation of GRIK5 promoter activity.
These results suggest that the interaction of nuclear orphan receptors with TFIIB may form a general mechanism by which these receptors mediate transcriptional repression in vivo, and apparently additional factors may be involved in determining the specificity of action for members of the COUP-TF family in the context of the intact first intron (Fig. 6B) and the GRIK5 promoter (Fig. 6C).

DISCUSSION
Nuclear orphan receptors have been proposed to play a key role in regulating organogenesis, neurogenesis, and cellular differentiation during embryonic development. Genetic ablation of nuclear orphan receptors including COUP-TF and NURR1 is lethal, and perinatal mortality arises from severe deficiencies in nervous system maturation (14,15). Revealing the target genes of orphan receptor action will therefore help us to understand the mechanisms of nervous system development. Studies in rat embryonic neural tubes have shown the expression of the kainate receptor subunits GluR6 and KA2 to be up-regulated following the onset of differentiation (11). Kainate-activated receptor function was also shown to be present in the human fetus as early as 5.5 weeks of gestation, indicating functional role(s) during neural cell maturation (48). In our present study, we describe the identification of several members of the orphan nuclear receptor family as potential participants in the process of neural development, through their demonstrated action on the promoter activity of a widely expressed, developmentally regulated neurotransmitter receptor gene encoding the KA2 subunit.
The documented tissue distribution of COUP-TFI, COUP-TFII, EAR2, and NURR1 (25, 36, 49 -51) is consistent with the brain-enriched nuclear protein binding to the GRIK5 silencer, as shown in Fig. 2A. In supershift experiments (data not shown), we have observed differences in the relative abundance of COUP-TFI and COUP-TFII between brain, cerebellum, and peripheral ganglia, supporting the notion of distinct functional roles for each nuclear orphan receptor. Thus, consistent with the expression patterns of the nuclear orphan receptors, it is reasonable to speculate that negative modulation of GRIK5 expression could vary among distinct neural cell populations according to the cellular orphan receptor repertoire.
Nonetheless, the overall developmental changes in binding activity of brain nuclear protein(s), as illustrated in Fig. 2B, overlaps with the observed increase in endogenous GRIK5 gene expression in the rat CNS between embryonic day 14 and postnatal day 1 (11). From previous in situ hybridization studies (16) and preliminary RNA analysis in our lab, 4 it was observed that expression of the rat GRIK5 gene is induced early in development but remains constant between neonate and adult when binding activity of nuclear orphan receptors is clearly diminished. The early expression of COUP-TFs predominantly in the developing rodent central nervous system (50,72,73), albeit in partially overlapping domains, is largely in agreement with a regulatory role for these proteins in endogenous GRIK5 expression. However, because an inverse relationship between orphan receptor binding activity and GRIK5 expression is not immediately apparent in the developmental profile of steady state GRIK5 RNA, it is more likely that nuclear orphan receptors contribute to maintaining postnatal levels of GRIK5 RNA at a plateau following an embryonic period of induction. It is plausible that COUP-TFs play very early roles in the developmental control of target neural genes. The low levels of COUP-TF expression and activity later in mature neuronal tissue (50, 72, 73, Fig. 2, B and D) is in keeping with the demonstrated antagonistic effect of COUP-TFI overexpression on neuronal differentiation of mouse teratocarcinoma PCC7 cells (13). Furthermore, additional positive regulatory factors acting elsewhere on the native GRIK5 gene could help maintain balanced KA2 RNA levels in vivo early in development, whereas other uncharacterized mechanisms may later take over as neural cells mature and acquire properties of the adult phenotype.
Yeast one-hybrid screening and supershift analysis (Fig. 3) indicate that multiple nuclear orphan receptors can regulate GRIK5 gene expression by binding the GRIK5 silencer. However, gel shift assays suggest that homomeric EAR2 may predominate in the DNA-protein complexes from rat brain and CG4 cells (Figs. 4B and 5). In addition, transfection assays have indicated that in the context of GRIK5 promoter activity, the physiological relevance of members of the COUP-TF family may be determined by flanking sequences of the intron (Fig. 6). In contrast to COUP-TF, relatively little is known about EAR2 and NURR1 as repressors of neural gene expression, apart from reports of EAR2 in the modulation of oxytocin gene expression in mouse Neuro-2A (52) and rat uterine cells (53). However, its demonstrated inhibitory effects on myeloid cell differentiation (54) strongly suggest that EAR2 may be involved in an important switch mechanism that determines the choice between growth or differentiation for myeloid cells. This may have implications for EAR2 in cells of the central nervous system and will be an important question to address in future studies.
Mutational analyses of the GRIK5 repressor binding site by transfections and binding assays were not only essential for delineation of the silencer ( Fig. 1 and Table I), but they also provided clues to the possible mode of repression mediated by orphan receptors at the GRIK5 promoter. Based on studies of COUP-TF, transcriptional repression by nuclear orphan recep-tors can occur by several distinct mechanisms, namely, competition for occupancy of activator sites, competition for RXR, active repression, and transrepression (for review see Ref. 40). In a recent report of another neuronal gene, the thyroid hormone (T3)-responsive cerebellar specific gene, PCP-2, it was suggested that COUP-TF repressed T3-dependent activation of the PCP-2 promoter by competing with thyroid hormone receptor for binding the thyroid hormone-responsive element (55). Several lines of evidence suggest that the first two scenarios of competitive binding and RXR sequestration are unlikely in the repression of GRIK5 expression. If repression occurs through competitive displacement of an activator, removal of the activator binding site would be expected to reduce the basal transcriptional activity. This was not observed with m11-13,15-21, whose activity was higher, not lower, than that of the construct containing the wild type intron, 5.4-CAT (Fig. 1). Even if activation at an uncharacterized site on the intron was being repressed, mutation of the repressor-binding site would be expected to raise reporter activity significantly above that of the intronless 2-CAT. This was not the case (Fig. 1B), and the lack of other protein-binding sites elsewhere in the first intron (18) strongly suggests that repressing such modes of transactivation are unlikely for the GRIK5 gene.
It is well documented that RXR is a universal heterodimeric partner for retinoic acid receptor, thyroid hormone receptor, vitamin D receptor, peroxisome proliferator-activated receptor, and orphan receptors (38, 56 -59). In our studies, RXR was not detected in DNA-bound complexes from CG4 and P2 rat brain (Fig. 3B) and its cDNAs not isolated in the yeast one-hybrid screen. This indicates that contrary to previous studies describing COUP-RXR heterodimers (33,38,60), COUP-TFs, as well as EAR2 and NURR1, are likely to bind the silencer as either as homodimers or monomers. This is also supported by studies with COUP-TFII, which form homodimers in preference to heterodimers with RXR␣ or thyroid hormone receptor ␤ in intact cells (61). Thus, it appears unlikely that COUP-TF, EAR2, and NURR1 in the brain repress GRIK5 promoter activity by sequestering RXR.
The remaining possible mechanisms, active repression and transrepression, both involve the direct interaction of repressor protein with the general transcription machinery, except that the latter model requires the tethering of repressor protein indirectly to DNA by interacting with another DNA-bound protein. Evidence for transrepression by COUP-TF was also provided by the demonstration that COUP-TF interacts directly with the estrogen receptor (62). The results of the CV-1 transfections shown in Fig. 6A may be explained by both DNA-dependent and DNA-independent mechanisms, because the activity of the parent pBLCAT5 vector was found to be reduced when repressors are overexpressed (not shown). The results of the CG4 transfections in Fig. 6C may be similarly explained. Indeed, COUP-TF has been shown to be capable of both active repression and transrepression (63), and the demonstrated physical interaction of orphan receptors with TFIIB ( Fig. 7 and Ref. 46) would be consistent with both hypotheses, but the strong competition in gel shift assays with unlabeled consensus DR1 and DR2, which are the preferred binding sites for COUP-TF (Fig. 5), would argue against transrepression in GRIK5 regulation.
Eukaryotic activators have been found to contact several general transcription factors, including subunits of TFIID, TFIIB, TFIIF, and TFIIH (64). Our studies have shown the physical interaction of nuclear orphan receptors with TFIIB ( Fig. 7) to have functional consequences in the intact cell (Fig.  8). This relief of nuclear orphan receptor-mediated repression is apparently DNA-dependent, because overexpression of TFIIB had no significant "squelching" effect on basal activity of the mutant reporter construct. The pivotal role of TFIIB in the apparently symmetrical processes of transcriptional activation and repression have received increasing support; TFIIB not only interacts with transcription coactivators, such as CBP/ p300 and p/CAF (65,66), but also interacts with co-repressors N-CoR (nuclear receptor co-repressor) (67) and SMRT (silencing mediator for retinoic acid receptor and thyroid hormone receptor), which in turn are involved in diverse regulatory mechanisms (68). In fact, repressor activity of COUP-TFI has been shown to be potentiated by overexpression of both N-CoR and SMRT (69). It remains to be seen, however, whether there are differences in the interaction of the individual orphan receptors with corepressors in the control of GRIK5 expression.
Our studies have thus far shown that neurotransmitter receptor proteins are subject to a form of transcriptional repression that is not REST-mediated. REST binding sites are found in many neuronal genes encoding proteins of fundamental importance, i.e. ion channels and synaptic vesicle proteins (for review see Ref. 74). The mechanistic distinction between nuclear receptors and REST as repressor proteins has recently been demonstrated by the lack of interaction of REST with N-CoR (34), whose action is often associated with repressor function of members of the nuclear receptor family. Further work on the action of nuclear orphan receptors at the GRIK5 promoter may also help elucidate the nature of negative regulatory mechanisms operating at TATA-less promoters, which are prevalent among the family of glutamate receptor genes (24,70,71).
In summary, we have shown that the transcription of a widely expressed neurotransmitter receptor gene is regulated by nuclear orphan receptors. The observations described support a model involving a form of active repression, which may provide a simple mechanism by which transcription can be down-regulated once cell-specific kainate receptor gene expression is induced during development (27). Further studies of the molecular events occurring at the promoter will be necessary to understand the role of active repression in the context of developmentally regulated neural-specific gene expression.