Characterization of the rat GRIK5 kainate receptor subunit gene promoter and its intragenic regions involved in neural cell specificity.

The GRIK5 (glutamate receptor ionotropic kainate-5) gene encodes the kainate-preferring glutamate receptor subunit KA2. The GRIK5 promoter is TATA-less and GC-rich, with multiple consensus initiator sequences. Transgenic mouse lines carrying 4 kilobases of the GRIK5 5'-flanking sequence showed lacZ reporter expression predominantly in the nervous system. Reporter assays in central glial (CG-4) and non-neural cells indicated that a 1200-base pair (bp) 5'-flanking region could sustain neural cell-specific promoter activity. Transcriptional activity was associated with the formation of a transcription factor IID-containing complex on an initiator sequence located 1100 bp upstream of the first intron. In transfection studies, deletion of exonic sequences downstream of the promoter resulted in reporter gene activity that was no longer neural cell-specific. When placed downstream of the GRIK5 promoter, a 77-bp sequence from the deleted fragment completely silenced reporter expression in NIH3T3 fibroblasts while attenuating activity in CG-4 cells. Analysis of the 77-bp sequence revealed a functional SP1-binding site and a sequence resembling a neuron-restrictive silencer element. The latter sequence, however, did not display cell-specific binding of REST-like proteins. Our studies thus provide evidence for intragenic control of GRIK5 promoter activity and suggest that elements contributing to tissue-specific expression are contained within the first exon.

In situ hybridization and immunohistochemical studies have demonstrated that, in different brain regions, specific combinations of members of the KA1/2 and GluR5-7 gene subfamilies are coexpressed (11)(12)(13). This indicates that, similar to ␣-amino-3-hydroxymethylisoxazole-4-propionic acid and Nmethyl-D-aspartate receptors, kainate receptor subunits assemble to form diverse subtypes of heteromeric channels. For example, co-immunoprecipitation experiments with anti-GluR6 and anti-KA2 antibodies have shown that these subunit proteins associate to form receptor channel complexes in brain cell membranes (14).
These molecular studies are consistent with the finding that functional homomeric and hetero-oligomeric kainate-preferring receptors are expressed in cultured neurons and glia (15)(16)(17). Despite the widespread expression of kainate receptor subunits in the mammalian nervous system, their function has remained elusive. Kainate receptor-mediated synaptic currents have been found in various areas of the central nervous system (18 -23), and presynaptic kainate-preferring receptors appear to be involved in the modulation of neurotransmitter release (24 -28). Recently, however, some light has finally been shed on a role for kainate receptors at thalamocortical synapses by the work of Kidd and Isaac (29), showing a decrease in kainate modulation of synaptic plasticity during development.
The kainate receptor subunit KA2 is encoded by the GRIK5 gene (30,31). KA2 transcripts can be detected in the rat as early as embryonic day 10 (E10) in the neural tube (32) and at E12 in the cortical plate (9). These findings suggest that KA2 may play a developmental role prior to synapse formation (33). At birth, the GRIK5 gene is expressed throughout the central nervous system, although its level of expression varies considerably between different brain areas and distinct cell types (11,12). Since KA2 associates with other kainate receptor subunits, it follows that different functional receptor subtypes can be * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
formed as a result of a stringent qualitative and quantitative control of GRIK5 expression (34).
To gain an understanding of how kainate receptor subunit expression is regulated, we have previously cloned the GRIK5 gene and characterized its structure (31). We have also identified an intronic element of this gene, which displays functional features of a silencer (31). More recently, we described the binding of nuclear orphan receptor proteins to this sequence to down-regulate GRIK5 transcription (35).
In this study, we analyzed the GRIK5 promoter and regulatory regions sufficient for tissue-specific transcription of the gene in both cultured cells and transgenic mice. We also found that GRIK5, like other glutamate receptor genes of the ␣-amino-3-hydroxymethylisoxazole-4-propionic acid and N-methyl-Daspartate classes (36), possesses a TATA-less and initiator (Inr)-containing promoter region that is GC-rich. This promoter is regulated by elements located within its first exon, and its selective expression in neural cells may involve a mechanism of preferential repression in non-neural cells.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes and DNA-modifying enzymes were from New England Biolabs Inc. (Beverly, MA). Polyacrylamide gel electrophoresis-purified oligonucleotides were purchased from Life Technologies, Inc. Radionuclides were from PerkinElmer Life Sciences. Large-scale plasmid DNA preparations were carried out using a QIA-GEN Plasmid Maxi kit. All animal procedures were in accordance with the National Institutes of Health Animal Welfare Guidelines.
GRIK5 Reporter Gene Constructs and Transient Transfection Assays in Cultured Cells-The isolation and characterization of GRIK5 genomic clones from a rat Sprague-Dawley genomic library (DASH II; Stratagene, La Jolla, CA) have been described previously (31). The nucleotide sequence reported in this paper has been previously submitted to the GenBank™/EBI Data Bank with accession number U81010. The BamHI site in the 2-kb EcoRI-BamHI fragment defines the exon 1/intron 1 boundary (see Fig. 3A).
The GRIK5 2Kb-CAT construct was generated as previously described (31). GRIK5-CAT deletion constructs were generated by restriction enzyme digestion and religation of GRIK5 2Kb-CAT (see Fig. 2). Apa5-CAT was generated by subcloning an 800-bp ApaI fragment from 2Kb-CAT into pLITMUS39 (New England Biolabs Inc.) to form pLI-TApa5 and subsequently inserting into pCATBasic (Promega, Madison WI) which will be subsequently referred to as Basic-CAT. Apa5A-, Apa5B-, Eag-, Sph-Apa-, and 2Kb⌬Eag-CAT constructs were derived by digestion and religation of Apa5-CAT. Deletion constructs of Apa5A-CAT such as Apa5A⌬SP1(a)Inr(a), Apa5A⌬Inr(a)(b), and Apa5A⌬Inr(b) were made by site-directed mutagenesis to generate appropriate restriction enzyme sites at the positions of Inr(a) (ϩ1) and/or Inr(b) (ϩ130) (see Fig. 3A and Table I). Apa5A-KCSE-CAT was formed by ligating a double-stranded oligonucleotide between the XbaI and XbaI sites of Apa5-CAT from which the XbaI-XbaI fragment had previously been removed.
Transient transfections were performed in all cells in serum-free medium by lipofection (2 l of LipofectAMINE/g of DNA; Life Technologies, Inc.) as previously described (31). Cells were plated in 100-mm dishes, transfected with 10 g of plasmid DNA for each GRIK5 construct, and cotransfected with 2 g of pPolIIplacF.gal (Grant MacGregor, Emory University) to normalize for transfection efficiency. Cells were harvested 40 h after transfection, and cell extracts were prepared in 0.25 M Tris buffer (pH 8.0) and assayed for CAT activity after heat inactivation of endogenous deacetylase for 10 min at 60°C. CAT assays were carried out using n-butyryl-CoA as specified by the manufacturer (Promega). CAT activity results were obtained by liquid scintillation counting of xylene extracts. ␤-Galactosidase assays were performed as described by Nielsen et al. (40). For all cell types, at least three independent transfection experiments were performed for each CAT construct. Statistical analysis was performed using Statview Version 5.0.
Northern Blot Analysis-Total RNA was prepared from tissues, CG-4 cells and astrocytes. Poly(A) ϩ RNAs were purified by an Oligotex kit (QIAGEN Inc.). RNA samples (2 g/lane) were resolved by electrophoresis through a formaldehyde-containing 1% denaturing agarose gel, electrotransferred to Nytran membranes (Schleicher & Schü ll), crosslinked to the membranes by UV irradiation (Stratalinker, Stratagene), and hybridized with a random-primed [␣-32 P]dCTP-labeled EcoRI-StuI KA2 cDNA fragment (595 bp). The specific activity of the probe was 10 8 cpm/g of DNA. The blot was hybridized in 50% formamide at 42°C and washed at high stringency with 0.1ϫ SSC at 60°C.
Ribonuclease Protection Assay-To generate the pSP72Apa5A template plasmid, a HindIII-XbaI fragment from the subclone pLITApa5 was inserted into the corresponding sites of pSP72 (Promega). The insert was subsequently shortened at the 3Ј-end by digestion with EagI and EcoRI, filled in with Klenow enzyme, and religated. The template pSP72Apa5B was constructed by inserting an XbaI-StuI fragment from pLITApa5 into the SmaI and EcoRV sites of pSP72. Both templates were linearized with HindIII, and antisense [␣-32 P]UTP-labeled RNA probes were generated by transcription with T7 RNA polymerase (Maxiscript kit, Ambion Inc., Austin, TX). Total RNA was isolated by a single-step method using RNAzol (41). Following gel purification of probes, RNase protection assays were performed with 30 g of total RNA using the RPAIII kit from Ambion Inc. Hybridization was performed overnight at 55°C. Reaction products were digested with a 1:100 dilution of RNase A/T1 mixture at 37°C for 30 min, precipitated, and resolved on a 6% denaturing polyacrylamide gel alongside an M13 sequencing ladder. The ladder was generated with 35 S-ATP using the Ϫ40 sequencing primer supplied with the U. S. Biochemical Corp. sequencing kit.
Dnase I Footprinting and DNA Electrophoretic Mobility Shift Assays-Nuclear extracts from CG-4 cells were made and DNase I footprinting was performed as described by Huang and Gallo (31) and Chew et al. (35). Footprinting probes were generated by polymerase chain reaction and are shown in Fig. 8. The sequences of gel-shift probes, competitors, and mutagenic primers are given in Table I. For gel-shift assays, the probes were end-labeled by T4 polynucleotide kinase with [␥-32 P]ATP (PerkinElmer Life Sciences), and purified on Sephadex G-50 columns. 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% Transgenic Analysis-The 4.3-kb BamHI GRIK5 fragment was cloned upstream of the bacterial lacZ coding region (31). Following restriction digestion and purification from agarose gel, the DNA fragment containing 4.3-␤-gal was purified by three consecutive ethanol precipitations, a chloroform extraction, and a final ethanol precipitation. The DNA was finally filtered through a Millex GV4 0.22-m filter (Millipore Corp., Marlborough, MA) before microinjection into fertilized eggs and implantation into pseudo-pregnant foster mothers, after injected eggs developed into blastocysts (42). Putative founders were screened for transgene integration by both polymerase chain reaction and Southern blotting of genomic tail DNA. Two stable transgenic lines were obtained for 4.3-␤-gal. Expression of the bacterial lacZ gene in transgenic mice was detected using the luminescence assay (CLON-TECH, Palo Alto, CA). Tissue homogenate was prepared by sonication in lysis solution containing 100 mM KHPO 4 , 0.2% Triton X-100, and 1 mM dithiothreitol. After centrifugation for 5 min, 25 l of the supernatant was assayed by incubation with 200 l of reaction buffer at room temperature for 60 min, and luminescence was measured in a luminometer according to the manufacturer's instructions (CLONTECH). Luminescence was normalized to protein concentration in each sample.
Histochemical staining was performed as follows. Mice were anesthetized and perfused transcardially with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Tissues were removed, immersed in the same fixative for 2 h at 4°C, and stored in phosphate-buffered saline ϩ 0.05% sodium azide. Tissue sections (100 m) were cut using a Vibratome (Technical Products International Inc., St. Louis, MO). ␤-Galactosidase expression was analyzed by overnight incubation at 37°C in X-gal staining solution (5 mM K 3 Fe(CN) 6 , 5 mM K 4 Fe(CN) 6 , 2 mM MgCl 2 , 0.02% Nonidet P-40, 0.01% sodium deoxycholate and 1 mg/ml X-gal). For whole embryo tissue sections, pregnant mice were perfused with phosphate-buffered saline, and the embryos (E15) were removed, immersion-fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4), and stained as described above for brain tissue sections. After staining, frozen sections (100 m) were then obtained on a freezing microtome (Microm International GmbH, Walldorf, Germany).
In Situ Hybridization Histochemistry-In situ hybridization protocols using 20-and 50-m fresh frozen cryostat tissue sections were performed as described by Hayes and Loh (43) and Gallo et al. (44). The probe used was derived from the BamHI-SacI fragment (bp ϩ452 to ϩ1463) of KA2 cDNA (9) inserted between the BamHI and SacI sites of pSP72. Antisense riboprobes were generated by in vitro transcription with T7 RNA polymerase following linearization of the template with HindIII.

Analysis of the GRIK5 Proximal 5Ј-Flanking Region in
Transgenic Mice-To delineate the genomic sequences responsible for tissue-specific GRIK5 expression in vivo, transgenic mouse lines were generated using 4.3 kb of 5Ј-flanking region. According to the numbering system used in this study (see Figs. 2 and 3), this flanking region spans Ϫ3231 to ϩ1100. Fusion genes were constructed that contained the promoter fragment and the bacterial lacZ coding region. Of 14 4.3-kb construct founders analyzed, two were found to carry the transgene by both polymerase chain reaction and Southern blotting of tail DNA (data not shown). These animals were bred to produce two independent lines, which displayed germ-line transmission of the transgene. In both lines, the pattern of ␤-galactosidase activity was clearly restricted to the central nervous system (Fig. 1C) (data not shown), demonstrating that the 4.3-kb GRIK5 fragment containing the promoter is sufficient to direct expression to the rodent central nervous system. In addition, transgenic mice were also generated with the 2Kb-CAT and 4Kb-CAT constructs (31) with very similar results (data not shown).
Expression of the 4.3-␤-gal transgene was analyzed in different areas of the mouse brain ( Fig. 1, D-I) and compared with expression of endogenous KA2 transcripts (Fig. 1, A and B). The distribution of ␤-galactosidase activity detected in the embryonic brain at E16 was similar to that of the endogenous GRIK5 mRNAs (Fig. 1, A and D). Both transgenic lines showed similar transgene expression patterns. In the adult brain, high levels of ␤-galactosidase activity were found in areas with abundant GRIK5 mRNA (Fig. 1, compare B and E). The hippocampus (Fig. 1H) and pyriform cortex (Fig. 1I) were among the areas that displayed the highest levels of GRIK5 mRNA and ␤-galactosidase activity.
Characterization of the GRIK5 Promoter-A series of GRIK5 promoter constructs (31) was assayed in cells of neural origin (central glia and CG-4 cells) and non-neural origin (NIH3T3 and HeLa cells). CG-4 cells are central glia with properties of oligodendrocyte progenitors and were shown to support high levels of GRIK5 promoter activity (31,35). As reported previously (31), CAT activity driven by 2 kb of the GRIK5 promoter was significant in CG-4 cells and undetectable in NIH3T3 and HeLa cells (Fig. 2) (data not shown). Inclusion of an additional 2.3 kb of 5Ј-flanking sequence to generate a 4.3-kb construct did not significantly modify GRIK5 promoter activity (31), indicating that 2 kb is sufficient for robust neural cell-specific GRIK5 expression.
Deletion analysis of 2Kb-CAT indicated that there is a negative regulatory element(s) between the EcoRI (Ϫ931) and ApaI (Ϫ112) sites and that the core promoter(s) lies within the 1200-bp fragment (Fig. 2). (Reference numbers assigned to restriction enzyme sites are based on physical mapping of the transcription initiation cluster shown in Fig. 3, with ϩ1 indicating the start of the first exon.) A large ApaI deletion in the 2Kb-CAT construct (2Kb⌬Apa5-CAT) completely abolished transcriptional activity (Fig. 2), thus providing evidence that minimal promoter sequences lie between Ϫ112 and ϩ723. Indeed, this ApaI fragment (Apa5-CAT) was found to maintain cell type-specific transcriptional activity in CG-4 cells (Fig. 2). Interestingly, removal of sequences 3Ј of XbaI from Apa5-CAT (Apa5A-CAT) resulted in a significant 2-3-fold increase in CAT activity compared with 1200-CAT, whereas the 3Ј-segment (Apa5B-CAT) was found to be inactive, demonstrating a core promoter element within Apa5A-CAT (Ϫ112 to ϩ331) and repressor activity of Apa5B-CAT. Further analysis of these upstream sequences showed that sequences between Ϫ112 and ϩ223 (Eag-CAT) could function as a minimal promoter (Fig. 2). In support of this observation, further 5Ј-deletion to the EagI site (2Kb⌬Eag-CAT) dramatically attenuated activity. Importantly, the emergence of significant CAT activity in non-neural cells, NIH3T3 (Fig. 2) and HeLa (data not shown), with Apa5A-CAT, Apa-Sph-CAT, and Eag-CAT supports the notion that transcription could initiate within the EagI fragment, and neural cell specificity was determined by sequences located downstream of SphI (ϩ503) (Fig. 2).
Since the GRIK5 promoter region contained neither TATAA nor CCAAT consensus sequences, it was possible that multiple transcripts could initiate from the GRIK5 promoter region. To map the transcription initiation site(s), ribonuclease protection assays were performed using non-overlapping riboprobes corresponding to ApaI-XbaI (Ϫ112 to ϩ331) and XbaI-ApaI (ϩ331 to ϩ723). The former 443-bp probe produced a protected band too large for precise reading against the sequencing ladder (data not shown). A 3Ј-deletion in this riboprobe template to generate the 335-bp ApaI-EagI probe yielded the results shown in Fig. 3B, where the predominant protected band with RNA from adult rat cerebral cortex and CG-4 cells mapped to a consensus initiator (Inr) at ϩ1 in Fig. 3A. This Inr is given the arbitrary location ϩ1 and will be referred to as Inr(a). No protection was observed with yeast RNA or rat liver RNA with this probe. Fig. 3C shows that, excluding vector sequences, the XbaI-ApaI probe (ϩ331 to ϩ723) was protected along its entire length, which is in agreement with the results in Fig. 3B, that the predominant initiation window lies upstream of the XbaI site at ϩ331. Additional downstream probes spanning SphI-XhoI (ϩ503 to ϩ1070) and EcoNI-BamHI (ϩ551 to ϩ1100) also yielded similar results as in Fig. 3C, with complete protection of the riboprobes with CG-4 and rat brain RNAs (data not shown).
These ribonuclease protection assay results interestingly did not show the presence of multiple transcription start sites, which are characteristic of TATA-less promoters. This is in agreement with the results from Northern blot analysis of GRIK5 mRNA, demonstrating a major transcript of 4.2 kb in CG-4 cells, astrocytes, and rat cerebral cortex (Fig. 3D).
Characterization of the Minimal Promoter-To study the promoter region in greater detail, 100-bp deletions were made in the Apa5A fragment, and the resulting CAT constructs were analyzed in CG-4 cells (Fig. 4). Only the Apa5A⌬SP1(a)Inr(a)-CAT construct showed transcriptional activity that was significantly lower compared with full-length Apa5A-CAT, indicating that the participation of Inr(a) along with its upstream regions is involved in the positive control of GRIK5 transcriptional activity. Other internal deletions, such as Apa5A⌬Inr(a)(b) and Apa5A⌬Inr(b) (Fig. 4), did not show any significant reduction in reporter activity. The activities of the latter two constructs were also found not to differ significantly from each other.
Initiator sequences in general are thought to bind at least three proteins, which can form an active transcription complex (45). We tested for the presence of two of these proteins, TFIID and YY1. Fig. 5A shows that Inr(a) formed sequence-specific complexes (lanes 1-5) that were not affected by addition of sequence 62, which is an intronic repressor previously shown to bind members of the nuclear orphan receptor family (35). Fur-thermore, recombinant TFIID demonstrated dose-dependent, sequence-specific binding to Inr(a) (Fig. 5B, lanes 1-4), which was partially competed by a consensus TFIID sequence (lane FIG. 2. Expression studies of the rat GRIK5 5-flanking region reveals a core promoter that is active in neural and non-neural cells. Shown are the results from transfection studies of a 2-kb genomic fragment in CG-4 and NIH3T3 cells. Data for HeLa cells are not shown. The activities of the positive control, SV40-CAT (not shown), and of 1200-CAT were 54-and 10-fold greater than that of promoterless Basic-CAT, respectively. Note that the CAT activity of constructs in CG-4 cells is expressed as a percentage of 1200-CAT, whereas that in NIH3T3 cells is expressed as a percentage of promoterless Basic-CAT. For CG-4 cells: a, p Ͻ 0.05 versus 1200-CAT; b, p Ͻ 0.05 versus 2Kb⌬Apa5-CAT and 2Kb⌬Eag-CAT (ANOVA followed by Fisher's PLSD). Note that these latter two constructs lack the putative promoter region. For

5), but not by a consensus SP1-binding site (lane 6).
The supershift experiments in Fig. 5C show that TFIID was indeed a component of the CG-4 complex binding to Inr(a). In control experiments using consensus SP1 and YY1 probes, the anti-YY1 antibody produced distinct supershifts with nuclear extract from CG-4 cells (data not shown), although YY1 could not be detected in the complex binding GRIK5 Inr(a) (Fig. 5C). As negative controls, two consensus initiators present downstream of Inr(a), viz. Inr(b) at ϩ127 and Inr(c) at ϩ397 (Fig.  3A), did not exhibit sequence-specific binding (data not shown).
Since it is well documented that many initiators function in conjunction with SP1 sites, we sought to examine the binding of nuclear proteins to these sites in the fragment between bp Ϫ112 and ϩ1. The SP1 site adjacent to Inr(a), referred to as SP1(a) (Ϫ29 in Fig. 3A), showed sequence-specific binding not only to an unlabeled SP1(a) competitor, but also to a consensus SP1-binding site (Fig. 6A, lanes 1-3). Furthermore, the protein binding to SP1(a) was supershifted by specific antisera against SP1 and SP3 (Fig. 6B, lanes 1-5), suggesting a possible role for SP1 family proteins in the activation of GRIK5 transcription.
An Intragenic Region Is Involved in Neural Cell-specific Expression-Since, as shown in Fig. 2, the absence of sequences downstream of SphI at ϩ503 (Apa-Sph-CAT) resulted in the loss of neural cell-specific transcription, we sought to delineate neural cell-specific elements in the 600-bp fragment located downstream of this SphI site. Functional analysis was performed in CG-4, NIH3T3, and HeLa cells. The latter two cell lines gave very similar results; therefore, only data obtained with NIH3T3 cells are shown. Fig. 7A shows that, surprisingly, ϳ70% of the reporter gene activity was retained within the 600-bp fragment (600-CAT) as compared with 1200-CAT. A further 150-bp 5Ј-deletion abolished CAT activity in CG-4 cells (Fig. 7A). Since, as mentioned above, endogenous transcription start sites were not detected within the 600 bp by ribonuclease protection assays with SphI-XhoI and EcoNI-BamHI probes (data not shown), we believe that the transcriptional activity of the 600-CAT construct is due to a secondary promoter, which is active only in the absence of the GRIK5 minimal promoter. Nonetheless, this fragment also retains neural cell-specific reporter expression (see below).
Two different internal deletions of 600-CAT were found not to significantly alter reporter gene activity in CG-4 cells (600⌬Apa3-CAT and 600⌬Nae-CAT) (Fig. 7B); but in NIH3T3 cells, the activity of 600⌬Nae-CAT was consistently elevated to almost 3-fold that of Basic-CAT (Fig. 7B). Compared with the Apa3 deletion, the NaeI deletion spans an additional 77 bp from ϩ646 to ϩ723. Because it appears to be involved in cell specificity, we have referred to it as a kainate cell-specific enhancer (KCSE) region.
The putative KCSE was further analyzed in the context of the upstream core promoter region. Placing the KCSE downstream of the Apa5A fragment in the same orientation as in the native gene was effective in silencing the activity of the GRIK5 promoter in NIH3T3 cells to below that of Basic-CAT (Fig. 7C). The same construct was still significantly active in CG-4 cells, although its activity was also attenuated. We then reasoned that there might be differences in the binding of nuclear factors that give rise to this preferential repression in NIH3T3 cells.
DNase I Footprinting of the 600-bp GRIK5 Promoter Region-We studied DNA-protein interactions within the 600-bp region by DNase I footprint analysis using CG-4 nuclear extracts. Several protected sites were identified, including Inr, AP2, MED1, and sequences known to bind members of the SP1 family (Fig. 8). The protected region over the Inr between ϩ610 and ϩ638 (Figs. 3A and 8, B and C) indicates that this Inr is indeed capable of binding nuclear proteins and might be involved in driving the transcriptional activity of 600-CAT. The region between ϩ646 and ϩ723, which lies in the 3Ј-end of the putative KCSE, shows a DNase I footprint corresponding to a consensus SP1 (SP1(b) in Fig. 3A)-binding site (Fig. 8C).
To further analyze the KCSE, two probes designated K(u) and K(d), spanning bases ϩ653 to ϩ690 and ϩ692 to ϩ724, respectively, were designed for studies of DNA-protein interactions by electrophoretic mobility assays. The sequence contained in K(u) bears some resemblance to a neuron-restrictive silencer element (NRSE) (Table II). However, as shown in Fig.  9, K(u) bound to nuclear proteins in extracts from both CG-4 and NIH3T3 cells. Sequence-specific DNA complexes were formed in both cell types, which were not competed by the NRSE-binding protein REST or by consensus SP1 or AP2 sequences. The identity of the protein(s) binding to K(u) remains unclear, but these results suggest that they are present in both CG-4 and NIH3T3 cells. On the other hand, K(d) was shown to bind members of the SP1 family (Fig. 9, B and C). A consensus SP1 sequence specifically competed for binding (Fig. 9B, lanes  2 and 8). mK(d), an unlabeled competitor in which the SP1 sequence of K(d) has been mutated, failed to compete, clearly defining the nucleotides involved in SP1 binding.
Since there is now evidence that differing SP1/SP3 ratios affect gene expression (46,47), we investigated the presence of individual SP1 proteins and their potential role in GRIK5 promoter binding. As shown in Fig. 9B (lanes 4 -6 and 10 -12), anti-SP1 and anti-SP3 antibodies showed differential effects. The anti-SP1 antibody produced a clear supershift with the CG-4 extract, whereas little effect was observed with the NIH3T3 extract (lanes 4 and 10). In contrast, an SP3-specific antibody produced supershifts in both extracts ( lanes 5 and 11), whereas AP2 was not detected in either cell type (lanes 6 and  12). We also tested for the presence of SP4 in these complexes. Unlike whole brain extract, in which a clear supershift with the anti-SP4 antibody was observed (Fig. 9C, lane 9), a similar SP4 supershift was not obtained with either CG-4 or NIH3T3 extract (lanes 3 and 6). Since brain tissue consists of diverse neural cell types, of which some displayed detectable SP4binding activity, it appears from these experiments that SP1 and SP3 (but not SP4) are involved in regulating GRIK5 expression in CG-4 and NIH3T3 cells. This suggests that subtle differences in the cellular repertory and behavior of SP1 family proteins may contribute to the determination of neural cell specificity in GRIK5 promoter activity. DISCUSSION We have previously shown that a 2-kb genomic fragment of the GRIK5 gene could support reporter gene activity specifically in neural cell cultures (31). These studies have also included the identification of a negative regulatory sequence in the first intron that binds nuclear orphan receptors such as chicken ovalbumin upstream promoter transcription factor I (31,35) in both neural and non-neural cells. However, the functional promoter elements and the genomic sequences that determine neural cell specificity of GRIK5 gene expression still remained to be characterized.
In this study, our analysis revealed that the in vivo distribution of endogenous mouse GRIK5 transcripts in the central nervous system could be closely mimicked by reporter expression driven by a 4.3-kb fragment of the rat GRIK5 gene (31). This indicates that, in addition to directing expression to the central nervous system, elements specifying regional expression within the rat brain are also present within this fragment. It should be noted that the transcriptional activity of this fragment and of the smaller 2-kb fragment (31) in neuron-like cells (nerve growth factor-differentiated PC12) was significantly lower than in CG-4 cells, although the promoter activity was at least 10-fold higher in both of these neural cell types than in non-neural cells (NIH3T3 and HeLa) (31). This difference between CG-4 and PC12 cells suggests that sequences within the 4-kb flanking region perhaps favor a glial component of GRIK5 expression. Nonetheless, our analysis thus far has shown the neural cell specificity of the rat GRIK5 promoter fragment; however, further analysis will be needed to investigate potential differences in GRIK5 gene regulation between neurons and glia.
In contrast with other TATA-less glutamate receptor genes, which clearly exhibit multiple initiation sites, the small cluster of transcription initiation sites around ϩ1 and the lack of multiple transcripts upon Northern blot analysis suggest that the use of multiple promoters and/or alternative processing is unlikely for the GRIK5 gene. In addition, the utilization of the GRIK5 initiation site is identical in both CG-4 cells and brain tissue, indicating that a wide variety of neural cell types are likely to share a common mechanism of transcriptional control. On the other hand, like other TATA-less promoters, an initiator element appears likely to be important for transcription (48,49). Fig. 4 shows that removal of Inr(a) and an upstream SP1(a) site significantly reduces promoter activity, whereas removal of downstream segments containing other Inr sequences does not. The demonstration of SP1 binding to the ) and Inr (600Inr) sequences, respectively. Region ϩ646 to ϩ723 is designated as the KCSE. C, core promoter activity in NIH3T3 is silenced by the KCSE. A 77-bp putative KCSE preferentially represses promoter activity in NIH3T3 cells. **, p Ͻ 0.005 compared with each of the other constructs (ANOVA and Fisher's PLSD). Note that in NIH3T3 cells, the activity of Apa5A-KCSE-CAT was lower than that of Basic-CAT and is hence considered inactive. All data were normalized to cotransfected pPolIIplacF.gal. Values shown are means Ϯ S.E. of at least three independent experiments.
FIG. 8. DNase I footprinting assay of the 600-bp GRIK5 fragment of the first exon. A, four overlapping polymerase chain reaction-amplified DNA fragments spanning the entire 600-bp region used for DNase I footprinting assays. The sizes of the probes ranged from 148 to 248 bp. FP, footprint. B, footprinting assay with the FP1 probe. FP1 was labeled at the 5Ј-ends of each strand (BamHI and EcoRI, respectively) and analyzed by DNase I footprinting. Reactions were performed with increasing amounts of CG-4 nuclear extracts, and products were resolved on a sequencing gel. Two protected regions could be discerned in both strands of the FP1 fragment, corresponding to nucleotides 546 -609 and 610 -638. Binding to these sites was confirmed by gel-shift assays (data not shown). Consensus sequences corresponding to an Inr and an SP1 site lie within this protected region (see below). C, diagram showing a summary of the DNase I footprint analysis of the 600-bp region of GRIK5. The putative transcription factor-binding sites were identified based on consensus sequence associated with the footprint. Their positions relative to the upstream transcription start site are indicated. SP1(a) site suggests a functional role for SP1 in Inr-mediated activation of GRIK5 transcription. This is consistent with studies of other TATA-less promoters (48,49). Interestingly, removal of an EagI-XbaI fragment containing two putative AP2 sequences at ϩ248 and ϩ266, respectively (compare Apa5A-CAT and Eag-CAT in Fig. 2), reduces promoter activity in both neural and non-neural cells, suggesting that these sites may be functionally important for basal transcription. This is in contrast to the description of AP2 as a tissue-specific transcription factor primarily expressed in neural crest cell and epidermal cell lineages (50).
Although the repertoire of proteins involved in the activation of transcription at TATA-containing promoters is now well established, the factor requirements for TATA-less, Inr-containing promoters are less defined. An issue of continued interest regarding Inr-containing promoters is the identity of Inr-binding proteins. Previous studies have indicated that TFIID is involved in the recognition of Inr elements, but TFIID binding to a consensus Inr has not been shown in a gel-shift assay unless the DNA probe also contained an upstream TATA box (48). Our demonstration of sequence-specific binding of recombinant TFIID to Inr(a) and the presence of TFIID in the Inr(a) DNA-protein complex (Fig. 5, B and C) supports the involvement of TFIID-Inr interaction in the expression of a TATA-less neural cell-specific receptor gene. The presence of YY1 binding was also tested, although none was detected on this GRIK5 Inr(a) site, confirming previous observations that YY1 recognizes a small subset of Inr elements (51,52). Taken together, our observations suggest that SP1 function at the SP1(a) site may help drive the TFIID-mediated basal activity of the rat GRIK5 promoter.
Unlike the minimal promoter of GluR2, the basal promoter region of GRIK5 is unable to support neural cell-specific expression (see Apa5A-CAT in Fig. 2). Since the 2Kb-CAT and 1200-CAT constructs are clearly active only in CG-4 cells, such regulatory sequences for the GRIK5 gene must be located downstream of the promoter. Although the presence of regulatory sequences within the 5Ј-untranslated region of genes is not unprecedented (53), our finding of a strong non-neural determinant within the 5Ј-untranslated region places GRIK5 among a few neural genes whose cell-specific expression is controlled by downstream elements.
The GRIK5 regulatory region, KCSE, which silenced the activity of the upstream promoter preferentially in NIH3T3 cells, contains not only a consensus SP1-binding site, but also a sequence in K(u) that resembles an RE1/NRSE, an element recognized by the RE1-silencing transcription factor (REST/ neuron-restrictive silencer factor). The NRSE is found in the promoter regions of many neural cell-specific genes and mediates the repression of neural gene expression in non-neural cells (54 -59). Interestingly, it has been shown that members of the N-methyl-D-aspartate family of glutamate receptor genes all harbor consensus NRSEs within their 5Ј-untranslated regions (60 -62). However, analysis of the GRIK5 KCSE by binding assay (Fig. 9A) failed to establish the identity of the binding factor as a typical REST-like protein. Furthermore, unlike with 600⌬Nae-CAT (Fig. 7B), deletion of the same NaeI region (bp ϩ646 to ϩ971) from 1200-CAT did not result in the emergence of reporter expression in non-neural cells (data not shown). Our observations suggest that the KCSE is not a REST-binding site and that it is by itself insufficient to confer neural cell-specific expression in the context of the 1200-bp genomic fragment. In light of studies of the dopamine ␤-hydroxylase gene showing that its NRSE does not bind recombinant REST (56) and showing both general and cell-specific repressor activities (58), our findings with the KCSE therefore support the possibility for a more complex multifactorial mechanism of neural gene specification.
The presence of several SP1-binding sites along the first exon prompted the investigation of a potential role for SP1 in nonneural repression of GRIK5 gene expression. Although the SP1 transcription factors are often involved in the activation of expression of TATA-less genes, the SP1 protein has been shown to have negative regulatory roles when bound to non-consensus FIG. 9. Differential binding of nuclear proteins to the kainate cell-specific enhancer region. A, sequence-specific binding of CG-4 and NIH3T3 nuclear proteins to the upstream KCSE probe, K(u). comp, unlabeled competitor; cons 200 and cons 500, consensus NRSE sequence present at 200-and 500-fold molar excesses, respectively. All other competitors were present at a 100-fold molar excess. B, SP1 binding to the downstream probe, K(d). Lanes 2, 3, 8, and 9 contain a 100-fold molar excess of unlabeled competitor. cons, consensus SP1binding site. Single arrowheads indicate the specific DNA-protein complex. Double arrowheads shows the positions of the supershifted complexes produced by addition of antibodies (Ab) against SP1 and SP3. No supershift was observed with the anti-AP2 antibody. C, absence of SP4 in CG-4 and NIH3T3 nuclear extracts. Gel-shift assays were performed with a labeled K(d) probe. The single arrowhead denotes the sequencespecific DNA-protein complex. The double arrowhead indicates SP1and SP4-supershifted complexes in lanes 2 and 9. P2, postnatal day 2. GGCACTGGGGGTGAGGACTGGGT sites (63). It has also been shown that SP1 and SP3 specifically modulate the expression of the neuronal nicotinic acetylcholine receptor ␤4 subunit gene (64). The negative activity of SP3 is well documented, and a change in the SP1/SP3 ratio results in different levels of gene expression (46). The observed differential binding of SP1 family members to the GRIK5 KCSE, specifically K(d), supports this possibility (Fig. 9B). An alternative mechanism of SP1 action includes the repression of promoter activity through competition by other proteins with the SP1 protein for its cognate binding site (65).
In conclusion, we have characterized the TATA-less, Inrcontaining promoter of the rat GRIK5 gene, delineating a 4.3-kb fragment capable of directing reporter expression to the mouse central nervous system. We have demonstrated a role for the 5Ј-untranslated region in cell-specific promoter activity and described an element that contributes to the determination of neural cell-specific GRIK5 gene expression. Since our analysis has focused on a relatively small region of DNA, it is possible that elements that regulate other aspects of GRIK5 gene expression, such as in embryonic development or in response to other physiological cues, may lie outside of the sequences examined in our experiments. Nonetheless, in our efforts toward understanding the role of kainate receptors, this study constitutes the first report that analyzes the DNA regulatory sequences directing the transcription of a widely expressed kainate-preferring receptor subunit gene in the mammalian nervous system.