Functional Analysis of the Rat N -Methyl- D -aspartate Receptor 2A Promoter MULTIPLE TRANSCRIPTION START POINTS, POSITIVE REGULATION BY Sp FACTORS, AND TRANSLATIONAL REGULATION*

N -Methyl - D -aspartate (NMDA) receptor subunit 2A (NR2A) is an important modulatory component of the NMDA subtype of glutamate receptors. To investigate the transcription mechanism of the NR2A gene, we cloned the 5 (cid:1) -flanking sequence from a rat genomic library. RNA mapping with rat brain RNA revealed two sets of major and several minor transcription start points in a single exon of 1140 bp. Reporter gene and mutation studies indicated that core promoter activity resided in exon 1, whereas the 5 (cid:1) -flanking sequence up to 1.5 kb showed no significant impact on promoter activity. Fragments containing minor transcription start points were able to drive a reporter gene in transfected cells and produce nascent RNAs in an in vitro transcription system. All fragments tested showed more promoter activity in dissociated neurons of the rat embryonic cerebrocortex and cell lines expressing NR2A mRNA than that in glial cultures and non-neuronal cells. Within exon 1 there are three GC-box elements that displayed distinct binding affinity to both Sp1- and Sp4-like factors. Overexpression of Sp1 or Sp4, but not Sp3, significantly increased the activity of the promoter containing these elements. Inclusion of exon 2 and 3 sequences, which contain five short open-reading frames, attenuated promoter-driven reporter

The N-methyl-D-aspartate (NMDA) 1 subtype of glutamate receptors plays a critical role in neuronal plasticity and neuro-nal differentiation, as well as in the excitotoxicity of glutamate on neurons (1,2). Functional NMDA receptor complexes are composed of subunits from three gene families: NR1, NR2A-D, and NR3A-B (3). Although NR1 serves as an essential component, members of the NR2 family bind ligand and differentially modulate receptor functions. The NR3 subunits exhibit an inhibitory role and form a glycine receptor when co-expressed with NR1 (4,5). In comparison to other family members, the NR2A subunit participates in the formation of an NMDA receptor complex that exhibits a rapid conduction (6 -8). NR2A is also the component responsible for high affinity Zn 2ϩ (9) or nitric oxide inhibition (10), glycine-independent desensitization of the receptor (11) as well as for mediating a trophic effect on neurons (12).
The NR2A gene, like other NR genes, is expressed preferentially in neurons and its distribution pattern in the developing and adult brain is similar to that of the NR1 mRNA (7). Several physiological and pathological factors have been shown to regulate its expression at the mRNA level. In neurons of the neonatal brain, NR2A mRNA is progressively increased during development (7). This developmental increase is dependent upon synaptic activity (13) and may be promoted by brainderived neurotrophic factor (14) or high KCl or NMDA in cultured neurons (15,16). NR2A mRNA in the developing visual cortex is up-regulated by light and down-regulated by darkness (17). Chronic alcohol treatment delays developmental expression in cultured cerebellar granule neurons (18). Exposure to lead results in specific down-regulation of NR2A mRNA in the developing brain (19), whereas prolonged exposure to nicotine induces an increase of NR2A mRNA in auditory cortex of postnatal rats (20). In the adult brain, NR2A mRNA is up-regulated by treatments of growth hormone (21), substance P (22), and dioxin (23); withdrawal of butorphanol infusion (24); estrogen reduction in acyclic animals (25); ginsenoside Rg1 infusion (26); spinal cord injury (27); and inflammation-induced hyperalgesia (28). NR2A mRNA is down-regulated in the CA1 and CA3 regions of the hippocampus by butorphanol-induced tolerance (24), in the hypothalamus between cycling and acyclic female rats, possibly by estrogen (29), in the pyramidal neurons of the visual cortex by focal thermolesion (30), and in the dentate gyrus granule cells by kindling-induced epileptogenesis (31). Furthermore, abnormal expression of the NR2A gene has been reported in the dorsolateral prefrontal cortex and the occipital cortex of patients with schizophrenia (32) and in entorhinal cortex of Alzheimer's patients (33).
The cellular mRNA level of most genes is predominately controlled by the transcription rate that is in turn determined by promoter activity, although other mechanisms may be involved. The mRNA level of the NR2A gene in neurons is regulated by various factors, as mentioned above, but the mechanisms underlying the regulation remain for the most part unknown. As a first step toward elucidating these mechanisms, we characterized the promoter of the rat NR2A gene. We defined the transcription start points (TSPs) using RNase protection assay (RPA), rapid amplification of 5Ј cDNA ends (5Ј-RACE), and in vitro transcription experiments. Then, we tested abilities of the isolated promoter to drive a reporter gene in cultured cells and to interact with nuclear Sp factors, as well as examined the impact of Sp factors on the promoter capability to drive the reporter gene. We further studied the effects of the 5Ј-untranslated region (5Ј-UTR) on NR2A gene expression at transcriptional and posttranscriptional levels.

EXPERIMENTAL PROCEDURES
Materials-A rat genomic library harboring EcoRI partially digested DNA (Sprague-Dawley) in the lambda phage was purchased from Clontech (Palo Alto, CA). Antibodies against Sp1, Sp3, and Sp4 were purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and from Upstate Biotechnology (Waltham, MA). DNA primers were synthesized either by Genemed Synthesis (South San Francisco, CA) or by the core facility of University of Maryland at Baltimore.
Isolation of Genomic Sequence-To screen the rat genomic library for the NR2A promoter a probe was generated by PCR with an upstream primer GB54 (5Ј-GCTGCGAGCCCTGCTGCAGAG) and a downstream primer GB98 (5Ј-CAATAACGCCACCACGTTCACATC). The reaction employed rat liver genomic DNA as a template, and Ex Taq DNA polymerase (TaKaRa/PanVera, Madison, WI) to amplify DNA for 30 cycles of 30 s at 95°C and 2 min at 68°C. The resulting PCR product was treated with Klenow enzyme, cloned into pGEM3Zf(ϩ) at the HincII site (Promega, Madison, WI), and sequenced as described previously (34). The PCR product was confirmed to be NR2A genomic sequence, containing exons and introns, and named GB5498. Excised DNA probe (GB5498) was labeled in the presence of [␣-32 P]dATP (ICN Biomedicals, Irvine, CA) with a random primer extension kit following the manufacturer's instruction (Amersham Biosciences, Piscataway, NJ). Library screening was carried out as described previously (34). One positive clone named NR2A32E was digested completely by EcoRI to produce three fragments that were subcloned into pGEM3Zf(ϩ) for sequencing. These subclones were named as NR2A32E32, NR2A32E11, and NR2A32E22. DNA sequencing was performed as described previously (34) or by the core facility of University of Maryland at Baltimore. Sequencing data were analyzed with Vector NTI software (InforMax, Bethesda, MD). DNA sequence of clone NR2A32E plus GB5498 was deposited in GenBank TM with an accession number of AY167029.
In Vitro Transcription-In vitro transcription was performed on the basis of a protocol described by Schwartz et al. with modifications (36). Briefly, nuclei were isolated from adult rat brain (Sprague-Dawley) by centrifugation on a sucrose phase and resuspended in nuclear buffer (10 mM Hepes, pH 7.9, 3 mM MgCl 2 , 0.1 M EDTA, 1 mM DTT, 0.1 M phenylmethylsulfonyl fluoride, 15% glycerol, 10 g/ml each of bestatin, leupeptin, and aprotinin). Crude nuclear proteins were obtained by addition of an equal volume of nuclear buffer supplemented with 0.7 M KCl, rotation for 15 min at 4°C, and centrifugation at 125,000 ϫ g for 20 min. Crude nuclear proteins underwent ammonium sulfate precipitation and dialysis against a buffer consisting of 25 mM Hepes, pH 7.9, 100 mM KCl, 15% glycerol, 0.25 mM EDTA, 1 mM DTT, and 0.5 mM phenylmethylsulfonyl fluoride. Protein concentration was determined by the Brandt method (Bio-Rad). For in vitro transcription, 0.1 pmol of NR2A promoter-luciferase constructs were mixed with 50 g of nuclear proteins in a transcription buffer (40 mM Hepes, pH 7.9, 1 mM MgCl 2 , 1 mM DTT). Reactions were pre-run at 30°C for 30 min. Then, 10 mM rNTP and 40 units of RNasin (Promega) per reaction were added for a further incubation at 30°C for 30 min. Then, 1 l of RNasin (40 units, Promega) and 2 l of RNase-free DNase (2 units, Promega) were added for incubation at 37°C for 20 min to remove DNA templates. The reaction was quenched by addition of a stopping buffer (40 mM Hepes, pH 7.9, 20 mM EDTA, 25 g/ml yeast tRNA, 0.2 M sodium acetate, and 1% SDS). RNA in the reactions was then extracted with a conventional phenol/chloroform method and subjected to RNase protection assay with riboprobe 8, 9, or 11 as described above. RNase A and T1 were used to digest single-stranded RNA. Protected RNAs were denatured and fractionated on 6% PAGE-7 M urea-TBE gel. 32 P-Labeled X174 DNA digested by HinfIII was used as a DNA ladder. Autoradiography was recorded with Kodak BioMax x-ray film.
Cell Culture, Transfection, and Reporter Gene Assay-PC12, HeLa, HEK293, and P19 cells were maintained and transfected as described previously (37,38) except that LipofectAMIN2000 (Invitrogen) was used according to the manufacturer's instruction. Co-transfection of luciferase and ␤-galactosidase constructs and assay of reporters were done as described previously (37,38). In some experiments, cells were used for RNA extraction 2 days after co-transfection. Co-transfection of cDNA expression constructs was conducted with a 4:1 ratio to luciferase constructs in addition to a CMV-LacZ construct (pCMV␤gal) at onetenth of total DNA, and pDNA3 was used as vector DNA for co-transfection. Luciferase activity was normalized to ␤-galactosidase activity. Sp factor constructs were described in previous studies (39).
Neuron-rich primary culture was prepared as described by Plesnila et al. (40). Briefly, cells were dissociated from the cerebral cortex of rat embryos at gastrulation day 15 as described previously (38) and plated at 10 5 /cm 2 on poly-L-lysine-coated surface with Neurobasal media plus 2% B27 serum-free supplement (Invitrogen). At day 8 of culture in vitro, more than 95% of cells were neurons proved by both morphology and microtubule-associated protein 2 immunostaining. At the same time, cells were transfected with DNA using LipofectAMIN2000. Lipo-fectAMIN2000 was mixed with DNA at a ratio of 2 l:1 g, and 2.5 g of DNA was used for cells cultured in a 10-cm 2 surface. DNA constructs of the NR2A promoter luciferase gene and pCMV␤gal (9:1) were transfected to cells. Two days after transfection, cells were lysed for reporter gene assays as described previously (39). Transfection efficiency was about 5% determined by green fluorescent cells after transfection of pEGFP (Invitrogen). Glial cells were isolated from the cerebrocortex of rats at postnatal day 1 and cultured for 8 days before transfection as described by Plesnila et al. (40). Transfection and reporter gene assays were performed as described for neurons.
Reverse Transcription-PCR-Total RNA was extracted from rat and mouse brains as described above and from cultured cells with TRIzol (Invitrogen) following the manufacturer's instructions. Human brain RNA was obtained from Dr. John W. Kusiak at the National Institutes of Health. RNA was treated with RNase-free DNase (Promega) and recovered with an RNA clean kit (Zymo Research, Orange, CA). 2 g of total RNA was annealed to a poly(dT) primer (Promega) at 65°C and reversibly transcribed into cDNA by SuperScript II reverse transcriptase (Invitrogen) in 20 l at 42°C for 30 min. Three l of the reaction of cell lines, 2 l of cultured neurons, or 1 l of brain tissues was used in a PCR to detect NR2A mRNA with human primers, GB509 (5Ј-TGATGAGGGAACCTCTTTGG)/GB510 (5Ј-GGTTGCTCTTCTCCA-TCAGC), rat primers, GB508 (5Ј-TGTGAAGAAATGCTGCAAGG)/G-B511 (5Ј-GAACGCTCCTCATTGATGGT), or mouse primers, GB512 (5Ј-CCATCAGCAGGGGCATCTACAG)/GB513 (5Ј-ACTGTGTTGGGGTT-GGACTCAT). Primers of GB450 (5Ј-ACCACAGTCCATGCCATCAC) and GB449 (5Ј-TCCACCACCCTGTTGCTGTA) were used to detect the glyceraldehydes-3-phosphate dehydrogenase (GAPDH) mRNA from human, rat, and mouse. To detect reporter gene mRNAs, primers GB149 (5Ј-GCGGTTGCAAAACGCTTCCATCTTCCAGGGATAC)/G-B148 (5Ј-AAATAGGATCTCTGGCATGCGAGAATCTGACGCAG) were used for luciferase transcripts and GB122 (5Ј-ATGTGCGGCGAGTTG-CGTGACTACCTA)/GB123 (5Ј-ATTCATTGGCACCATGCCGTGGGTT-TC) for Escherichia coli ␤-galactosidase. PCR was performed with Ex Taq DNA polymerase and cycled with annealing temperatures between 50 and 60°C and extension at 72°C for 45 s. 40 cycles were used for NR2A mRNA, 30 cycles for luciferase mRNA, and 26 cycles for ␤-galactosidase. DNA products were visualized with UV illumination on agarose gel containing 10 g/ml ethidium bromide. Images were taken on a Kodak Image Station 440CF and analyzed with 1D Image Analysis software as stated previously (39).
DNA-Protein Interaction Analysis-Crude nuclear extractions were prepared from cells as previously described (37). DNA sequences used as probes in electrophoretic mobility shift assays (EMSA) are as follows: GC-box-a, 5Ј-GCCTCCGGGGTGGGCGACTGC; GC-box-b, 5Ј-CCGGG-GAGCGGGCGGAGAGCGTGGT; GC-Box-c, 5Ј-GCATCCTGGGCGGG-TGTGTGC. Probes were labeled by T4 DNA kinase in the presence of [␥-32 P]ATP (ICN Biomedicals), and EMSA was conducted in a method reported previously (37). Competition and supershift of EMSA were performed as described previously with antibodies against Sp1, Sp3, or Sp4 (37). A consensus of GC-Box (Sp1 site) from Promega was used as a competitor. Results of EMSA were analyzed and quantified with a PhosphorImager as described in our previous studies (41).

RESULTS
Isolation of the NR2A Promoter-To study the mechanism of transcriptional regulation of the NR2A gene, we cloned its promoter from a rat genomic library from which we previously isolated the NR1 promoter (34). To prepare a probe for library screening, we chose a sequence near the 5Ј-end of the NR2A cDNA (GenBank TM M91561). To avoid skipping intronic sequences in a probe derived from cDNA and hence having a mismatch between the probe and the genomic clones, we conducted PCR on genomic DNA with primers surrounding the chosen sequence. This guaranteed a probe that exactly matched the genomic sequences. A 1.8-kb product, termed GB5498, was isolated. Sequencing of this product revealed an internal intron (Fig. 1). Using this probe, we screened 3 ϫ 10 6 plaques of a rat genomic library and obtained three positive clones of ϳ13 kb each. Restriction analysis suggested that all three clones were identical (data not shown). We then sequenced one clone and named it NR2A32. Sequence analysis showed that the 3Ј-end of NR2A32 or its subclone NR2A32E22 contained at least one intron and two exons overlapping the 5Ј-UTR of the NR2A cDNA. The sequence of this region, derived from NR2A32E22 and GB5498, is shown in Fig. 1. All exon-intron junctions follow the major GT/AG rule (42). In addition, our analysis suggested that the most 5Ј-end of the GenBank TM cDNA belongs to an upstream exon that may be wholly contained in our clone. These results suggest that the regulatory sequences and the promoter of the NR2A gene reside in the clone NR2A32.
Identification of TSP(s) or 5Ј End(s) of NR2A mRNA-Next, we searched for the TSP(s) by mapping the 5Ј-end(s) of NR2A mRNA. First we performed 5Ј-RACE experiments on rat brain poly(A) ϩ RNA using an NR2A gene-specific primer, either GB97 or GB98. After two rounds of PCR, we obtained fragments that contained at most an additional 236-bp sequence beyond the 222-bp 5Ј-UTR of the NR2A cDNA. Sequence analysis indicated that all of the newly obtained sequences are extensions of NR2A cDNA in the clone NR2A32E22 and thus belong to the same exon that encodes the 5Ј sequence of the cDNA. These results suggest that the NR2A gene has multiple TSPs.
Considering that the reverse transcriptase is sensitive to the secondary structure of mRNA and very often generates truncated products (43), the cDNAs in our 5Ј-RACE experiments may have resulted from incomplete reverse transcription. To confirm results obtained from the 5Ј-RACE above and search for possible upstream TSPs, we performed an RPA to map the 5Ј-end(s) of the NR2A mRNA. In addition, RPA is also able to determine TSP strength by quantifying the relevant transcripts (44). We designed 11 complementary riboprobes for use in the RPA based on the NR2A32E22 sequence (Fig. 1). RPAs with riboprobes 4 -8 and rat brain or cerebral cortex RNA each resulted in a major protected band or clusters representing full protection of the NR2A sequence in the probes. In addition, each probe also yielded a number of small, weak protected bands representing minor TSPs (summarized in Table I). Riboprobes 1-3, however, repeatedly produced multiple packed bands indicating an interference of RNA secondary structure of the probes (data not shown). As a NR2A mRNA-free control, RNA extracted from the rat liver did not show any protection. Fig. 2 (A and B) shows representative results obtained from riboprobes 5 and 6 after digestion by RNase I. The same results were obtained from RNase A/T 1 digestion (not shown). As shown in the Fig. 2 (A and B), RNA extracted from the adult rat brain generated strong single clusters with bands that varied by 2-4 nt, and these clusters represent the full sequence of the NR2A gene in the probes. A number of weak bands followed these clusters and delineated minor TSPs. These results suggest that the major TSP(s) must be upstream of the 5Ј-end of riboprobe 8 (Ϫ1073, Fig. 1). To uncover the major TSP(s), we employed riboprobes 9 and 10, both of which end at Ϫ1230 (Fig.   1). As shown in Fig. 2C, riboprobe 9 revealed a strong band that is 7 bp less than the NR2A sequences in the riboprobe and corresponds to a protected sequence ending at Ϫ1222/Ϫ1223. Riboprobe 10 produced the same results (data not shown). RPAs with riboprobe 11, which covers further upstream sequences, resulted in multiple protected bands (Fig. 2D). Among them, there are two groups of strong bands delineating major TSPs at positions Ϫ1233/Ϫ1234 and Ϫ1222/Ϫ1223. Both upstream and downstream of these strong bands are several weak bands indicating minor TSPs ( Figs. 1 and 2). Density analysis of these protected bands indicated that the utilization difference between the major and minor points was more than 3.55 times in pair comparison. Detection of major TSPs at Ϫ1222/ Ϫ1223 by riboprobes 9 -11 strongly supports their utilization by the NR2A gene in the rat brain. The largest protected band of riboprobe 11 represents a minor TSP at Ϫ1343 (Figs. 1 and 2D and Table I). On the basis of RPA results, we designed new primers of GB390, GB391, and GB462 (see "Experimental Procedures" for sequences) complementary to NR2A mRNA and performed a highly specific RLM-5Ј-RACE, which reacts only with 5Ј-caped mRNA. This experiment revealed multiple 5Јends of the NR2A mRNA from rat brain poly(A) ϩ RNA. All newly obtained sequences are extensions of the previous 5Ј-RACE products in NR2A32E22. Most of these sites exist in the Ϫ1243/Ϫ726 region (Table I), with the most distal position at Ϫ1349 relative to the first codon of open-reading frame (ORF). This position is 7 bp upstream of the site revealed by RPA (Figs. 1 and 2D) and may represent a very weak TSP that could be uncovered only by 5Ј-RACE. As a result of combination of these two methods, many of the minor TSPs were revealed by both 5Ј-RACE and RPA or more than two different riboprobes in RPAs, and are indicated by underlining in Table I. Taken together, our data suggest that the NR2A gene starts transcription at multiple sites with differential strength in exon 1 and that the NR2A transcript has a 5Ј-UTR with a maximum of 1349 nt encoded by a combination of exon 1 (1140 bp), exon 2 (191 bp), and part of exon 3 (18 bp).
Sequence analysis of the proximal 5Ј-flanking region revealed a long polypyrimidine track, a direct repeat sequence, and an exonic TATA box upstream of the major TSPs (Fig. 1). Although the flanking sequence is TA-rich, exon 1 is over 50% GC in the region close to the 5Ј-end. There is no DNA sequence with more than 65% homology to the initiator (Inr) consensus, YYANWYY, because the initiator may direct transcription initiation for most genes (45). We also did not find an RE1/NRSE site using the same searching criteria to the consensus, TTCAGMACCDYGSASAGHRCC (46), considering that the RE1/NRSE DNA cis-element plays an important role in neuronal expression (41).
The Promoter of the NR2A Gene-To test whether the isolated 5Ј-flanking sequence has promoter activity, we utilized reporter gene constructs, because they are highly sensitive and capable of testing cell-type specificity (47). First, using an RT-PCR technique, we examined NR2A mRNA in cultured rat cerebrocortical neurons to determine whether an endogenous mechanism for NR2A transcription was active. Our results showed that NR2A mRNA increased about 2.5-fold from 2 days in vitro (DIV) to 8 DIV (Fig. 3A). These cultures retained more than 95% neurons by cell morphology and MAP-2 immunostaining (data not shown). We also employed RT-PCR to examine several cell lines for expression of NR2A mRNA. We designed our PCR reactions to include an exon-exon junction in the product and thus were able to distinguish cDNA from products generated by contaminating genomic DNA. The fidelity of the PCR products was confirmed by DNA sequencing. As reported by others (48,49), the mouse embryonic cell line, P19, in the non-differentiated state, and the rat pheochromacytoma cell line, PC12, express NR2A mRNA, albeit at a level lower than brain tissues (Fig. 3B). In contrast, RNA from human HEK293 or HeLa cells showed no NR2A transcript even after 40 PCR cycles but exhibited robust and stable GAPDH mRNA after 30 cycles (data not shown).
Next, considering that the promoter of a given gene transcribed by RNA polymerase II usually includes a core promoter, approximately between nucleotides Ϫ35/ϩ35 relative to the TSP and ϳ50 -200 bp of flanking sequences (50 -52), we fused a 1548-bp 5Ј-flanking sequence from clone NR2AE22 with various exonic sequences from the 5Ј-UTR (1349 bp in total) of the NR2A gene as illustrated schematically in Fig. 4A. We then tested the ability of these constructs to drive transcription of a luciferase reporter gene in neural cells (Fig. 4B) and cell lines (Fig. 4C). The largest fragment (pNR2A2897⌬1) exhibited significantly higher promoter activity in neuronal culture or PC12 cells than that in glial or HEK293 cells. The same results were obtained from differentiated P19 (neuronal) and HeLa (nonneuronal) cells (data not shown). Removal of 338 bp of proximal sequence, which comprises part of exon 1, the entire exon 2, and part of exon 3, significantly increased promoter activity in TABLE I Transcription start points of the rat NR2A gene Three experimental strategies were used to reveal TSPs of the rat NR2A gene: 5Ј-RACE, RPA, and in vitro transcription (IVT) plus RPA. TSPs are numbered following the rule described in Fig. 1 and underlined for those to be proved by more than two strategies or two riboprobes. The major TSPs are bold and italic. R followed by a number or numbers represents riboprobes used in RPA to reveal the TSP. Constructs used in IVT are indicated. Selective minor TSPs revealed by various riboprobes were presented. At least three independent experiments were performed for each probe in RPA. all cells tested. Further removal of the 3Ј 1072-bp exonic sequence encoding 5Ј-UTR (pNR2A2897⌬1072), while preserving all major TSPs, decreased promoter activity to a level comparable to that of pNR2A2897⌬1 in all types of cells (Fig. 4, B and C). Promoter activity was totally lost when all exonic sequences, bases Ϫ1350/Ϫ1 (pNR2A2897⌬1350), were deleted or DNA fragments tested were in the reverse orientation (e.g. pNR2A2897⌬822, 3Ј35Ј). In comparison to the SV40 early gene promoter, the NR2A promoter showed relatively higher activity in neurons but lower activity in cell lines and comparable activity in glial cells. These results indicated that the sequences studied here demonstrate promoter activity preferentially in neurons or neuronal cell lines and that this activity requires both major TSPs and their downstream sequences of exon 1.
From results above (Fig. 4, B and C), it can bee seen that the maximal promoter activity was given in neuronal cells by NR2A fragments consisting of the 1548-bp 5Ј-flanking sequence plus exons lacking 338 -822 bp proximal to the 5Ј-UTR. On the basis of these constructs (pNRL2897⌬338 and pNRL2897⌬822), we deleted the 5Ј-flanking sequence to different sizes and examined the contribution of the 5Ј-flanking sequence to the NR2A promoter activity in cultured neurons and PC12 cells. Results obtained from deletions of the parent construct pNRL2897⌬338 are shown in Fig. 4D. Surprisingly, deletion of the entire 5Ј-flanking sequence and 5Ј 121-bp sequence of exon 1 (pNRL1228⌬338) did not cause any significant loss of promoter activity. However, deletion of the most 5Ј region of exon 1 (pNRL580⌬338) resulted in a reduction of, but did not abolish, promoter activity in cultured neurons. In nonneuronal cells (glial, HEK293, and HeLa cells), similar results were obtained (data not shown), although the activity remained lower than that in neuronal cells, consistent with results obtained from 3Ј deletion studies (Fig. 4, B and C). In PC12 cells, an exonic fragment from bases Ϫ1228 to Ϫ339 (pNRL1228⌬338) generated higher activity than any other fragments tested. This fragment contains the proximal major TSPs and most of the internal exon 1 sequence but not the TATA-box. Taken together, these results suggest that exon 1 of the NR2A gene contains basal promoter activity that is neuron- RPA experiments and probe preparation are described in detail under "Experimental Procedures." A, RPA with riboprobe 5. Samples were loaded as indicated. R5: undigested riboprobe 5. B, RPA with riboprobe 6. Samples were loaded as indicated. R6: undigested riboprobe 6. C, RPA with riboprobe 9. Samples were loaded as indicated. R9: undigested riboprobe 9. D, RPA with riboprobe 11. Samples are loaded as indicated. R11: undigested riboprobe 11. Cx: cortex. NR2A gene sequences in the probes are illustrated in Fig. 1, and vector sequences are described under "Experimental Procedures." DNA ladders were prepared with M13 Ϫ20 or Ϫ40 primers and fractionated as indicated for reactions to be stopped by ddGTP (G), ddATP (A), ddTTP (T), or ddCTP (C). Solid arrowheads represent the major 5Ј-ends, open ones denote the minor 5Ј-ends, arrow lines imply fully protected NR2A sequences in riboprobes, and lines indicate the undigested riboprobes. preferential and independent of the TATA-box. Furthermore, the 5Ј-flanking sequence up to 1548 bp may slightly enhance the promoter (Fig. 4B), and the exonic sequence encoding the 3Ј portion of 5Ј-UTR may play a negative role.
In view of exon 1 containing multiple TSPs (Table I) and exhibiting promoter activity of the NR2A gene (Fig. 4D), we next searched the core promoter within the exon 1 sequence. We generated small, sequential fragments of the exonic sequence that encodes the 5Ј-UTR of the NR2A gene (left panel, Fig. 5A) and tested their ability to drive the reporter gene transcription in cultured cells. These fragments vary from 157 to 342 bp in length, and most of them were derived from  Fig.  1. Three representative constructs are presented following their positions corresponding to the genomic sequence. The sites of major products protected by each riboprobe from RNA transcribed in vitro from NR2A promoter constructs are presented by an arrowed line followed by a solid circle for riboprobe 8, by an open circle for riboprobe 11, and by an open end for riboprobe 9. C, example of in vitro transcription products detected by RPA. Nascent RNAs transcribed from indicated NR2A promoter constructs are used for RPA with riboprobes as displayed. X 174 DNA digested with HinfIII and labeled with 32 P at the 5Ј end is used as molecular weight markers. Riboprobes without enzyme digestion were loaded to indicate size and integrity of the probe as indicated. Three independent experiments were performed for each in vitro transcription and relevant RPAs. exon 1. Interestingly, all fragments independently showed more promoter activity in neurons or neuronal cells line (PC12) than that in dissociated glial cells or non-neuronal cell line (HEK293) (Fig. 5A). The high activity remains within three fragments spanning from positions Ϫ1077 to Ϫ339, but is lower than that in a fragment consisting of the most sequence of exon 1 seen in Fig. 4D (pNRL1228⌬338) that contains major TSPs. These data suggest that the core promoter resides in exon 1 and requires major TSPs as well as downstream sequences for full activity. As negative controls, fragments in reverse orientation, e.g. 1077⌬822 or 827⌬528 in front of the luciferase gene, failed to show significant promoter activity (data not shown).
To further confirm whether fragments tested above are able to directly transcribe RNA, we performed in vitro transcription experiments. We used nuclear proteins prepared from adult rat brains to transcribe RNA from varied constructs as listed in Table I, and then detected nascent RNA products using RPA. Based on the size of protected RNA probes, we estimated the TSP locations in these fragments. Interestingly, all tested constructs that had promoter activity transcribed nascent RNAs. Fig. 5B schematically illustrates positions of the fragments tested and TSPs revealed by RPA. Fig. 5C is representative of in intro transcripts analyzed by RPA. As can be seen, nascent RNAs were transcribed not only from large fragments showing maximal promoter activity above but also from exonic fragments exhibiting partial promoter activity. As summarized in Table I, all TSPs revealed from these internal fragments repeated or were close to the results revealed by 5Ј-RACE or RPA above (Table I and Fig. 5C). Major TSPs showed again the strongest activity (Fig. 5C, lanes 6 and 7, and Table I). These data further confirmed that the core promoter along with multiple TSPs resides in exon 1 of the NR2A gene.
Regulation of NR2A Promoter Activity by Sp Factors-Sequence analysis of bases Ϫ1227/Ϫ577 revealed three putative GC-boxes (Fig. 1). GC-boxes in a promoter region have been shown to interact with Sp1, Sp3, and Sp4 transcription factors (39) as well as other transcription factors, such as Egr1 (53). To test their role in forming the NR2A promoter, we co-transfected Sp cDNAs with NR2A promoter/luciferase gene constructs into PC12 cells. As shown in Fig. 6A, promoter activity was increased by overexpression of Sp1 or Sp4 but not by Sp3.
We then examined the binding capacity of these cis-elements to endogenous nuclear proteins in EMSA experiments. Results in Fig. 6B show that all three GC-boxes formed three bands with nuclear proteins extracted from P19, PC12, or HeLa cells. Similar binding patterns were formed by all experimental probes as well as by a consensus GC-box oligonucleotide, although there are varying binding intensities among different cell lines. To identify proteins involved in producing these bands, we preincubated nuclear extracts with unlabeled oligonucleotide probes and found that the binding was diminished only by specific probes (lane 3, Fig. 6C) but not by nonspecific probes, such as Oct 1 (lane 4, Fig. 6C). Pretreatment of nuclear proteins with antibodies specific to each of the Sp factors showed that bands 1 and 3 were disrupted by an Sp1 antibody (lane 5, Fig. 6C), whereas band 2 was disrupted by an Sp4   FIG. 6. Regulation of the NF2A promoter by GC-boxes and Sp factors. A, up-regulation of the NR2A promoter by overexpressed Sp factors in PC12 cells. Expression constructs of Sp factors were cotransfected into PC12 cells with pNR2A2897⌬1223 lacking GC-boxes or pNR2A2897⌬822 retaining all three GC-boxes as described under "Experimental Procedures." The -fold increases in relative promoter activity were calculated on the basis of promoter activity without co-expression of Sp factors and are shown for constructs as indicated. B, differential binding of GC-boxes-a, -b, and -c with nuclear proteins of cell lines. Individual labeled probes of GC-boxes-a (lanes 1-3), -b (lanes  4 -6), or -c (lanes 7-9) reacted with nuclear proteins extracted (N.E.) from cells indicated. The solid arrows denote specific binding bands 1-3. antibody (lane 6, Fig. 6C), suggesting that these Sp factors are involved in specific band formation.
As shown in Fig. 6B, there are differential binding intensities among the three GC-boxes, even within the same cell line. To more accurately evaluate their binding capacity, we used equal amounts of each probe to react with the equal amounts of P19 cell nuclear proteins in EMSA and quantified the signal using a PhosphorImager. We combined signals from free probe, band 1 and band 2 as the total signal of the probe, and then calculated the signal shifted by each band as a percentage of the total signal in cognate lanes. As indicated in Fig. 6D, GC-box-c generated the largest percentage shift, whereas the GC-box-a generated the least, suggesting that the GC-box-c has the best binding capacity to Sp proteins.
Role of 5Ј-UTR in Regulation of NR2A Gene Expression-As shown in Fig. 4, the construct containing the full-length 5Ј-UTR had significantly lower reporter activity than those lacking the 338-bp proximal sequence of the 5Ј-UTR. Sequence analysis revealed that the 338-bp 5Ј-UTR has five small ORFs (Fig. 1). A previous study with a limited 282-bp sequence indicated that the mouse NR2A 5Ј-UTR interfered with the translation rate of that mRNA, even though the complete upstream 5Ј-UTR was not available, and thus its impact on this interference could not be fully determined (54). Because reporter activity really represents activity of protein products that are regulated by both transcription and translation, we explored the mechanism by which inclusion of the 338-bp proximal 5Ј-UTR might inhibit reporter activity by reducing translation of the message RNA. We transfected cells with pNR2A2897⌬1 or pNR2A2897⌬338 and then extracted RNA for RT-PCR of luciferase and ␤-galactosidase messages. To avoid interference by the transfected plasmids, we pretreated the RNA with RNasefree DNase. Our results, as summarized in Fig. 7 indicated that inclusion of the 337-bp sequence slightly reduced luciferase mRNA level (Ͻ50%). These reductions were much smaller than those seen in reporter gene activity: a 4.3-fold reduction in PC12 cells and a 2-fold in P19 cells (Fig. 4C). These results suggest that sequences of exons 2 and 3 may not only negatively regulate transcription but also attenuate NR2A gene expression by a different mechanism, very likely at the translational level, after transcribed into mRNA. DISCUSSION In the present study, we have isolated part of genomic and 5Ј-flanking sequence of the rat NR2A gene and mapped the 5Ј-ends of its mRNA using 5ЈRACE, RPA, and in vitro transcription experiments. Aligning the mapped 5Ј-ends and the isolated genomic sequences, we determined the location of TSPs, exon 1, and the 5Ј-flanking sequence. Our data indicate that this gene has two sets of major TSPs spanning a 13-bp region within exon 1 and multiple minor TSPs spreading most of exon 1. Interestingly, exon 1 alone retains promoter activity preferential in neurons or neuronal cells, whereas the 5Ј-flanking sequences of the NR2A gene up to 1.5 kb have no significant impact on the promoter activity. Further studies of deletion fragments with reporter gene and in vitro transcription experiments demonstrated that the core promoter resides in the exon 1 of the rat NR1 gene and the full promoter activity depends upon the major TSPs for cell-type preference.
The core promoter in eukaryotic cells includes Ϫ35/ϩ35 bp sequences relative to the TSP and contains three common elements: 1) an upstream TATA motif plus a TFIIB recognition element in some cases, 2) one or more short initiator (Inr) sequences, and 3) a downstream promoter element (DPE) or other elements (45,50,52,55). Among these elements, the DPE is well conserved from Drosophila to human and may replace the TATA-box by interacting with several basal transcription factors in the transcription initiation of TATA-less genes (56). Sequence analysis of the NR2A exon 1 revealed only one TATAbox consensus located among multiple TSPs, and this site seems unimportant for promoter activity (Figs. 4 and 5). However, there is no consensus for either an Inr or a DPE in the exon 1. Currently, the Inr consensus remains loose or partially unidentified yet (45). Additionally, Kutach and Kadonaga (55) searched 205 promoters in a Drosophila core promoter data base and found that ϳ31% of the promoters did not have either a TATA motif or a DPE. They proposed that these genes might utilize other adjacent cis-elements, such as GC-boxes, to direct basal transcription. In the sequence of rat NR2A exon 1, we uncovered three putative GC-boxes, a putative AP-2 site, and a putative NFB site. We further demonstrated that these GCboxes are able to actively regulate promote transcription ( Figs.  1 and 7). In addition, a sequence of CA repeats was found in exon 1 (Fig. 1). Although the significance of CA repeats has yet to be fully elucidated, it was reported that the GAPDH gene in P. anserine is positively regulated by a CA repeat sequence (57), and therefore this sequence may also function as a regulatory element for the NR2A gene. All of these elements may act as part of the core or proximal promoter of the NR2A gene by supporting transcription initiation, however, this hypothesis needs to be explored. Although several lines of evidence obtained from rat brain RPA, reporter gene assay, and in vitro transcription demonstrate that transcription of the rat NR2A gene can be started from multiple points internal to exon 1, only two clusters of major TSPs can be protected by brain RNA and are required for full activity of the promoter. Therefore, we believe that the NR2A core promoter resides in a signal exon.
We also noted that the functional fragments of the exon 1 sequence of the NR2A gene exhibit high activity in cultured neurons, suggesting a cell-type preference. NR2A is an important neuronal protein, and its mRNA is found only in neuronal cells (3) or at a very low level in limited non-neuronal tissues (58,59). Our studies further confirmed that the NR2A promoter is active only in neuronal tissue or cell lines (Fig. 3). The core promoter plus adjacent regulatory cis-element(s) located within Ϫ250/ϩ250 bp regions to the TSP constitutes the proximal promoter (50). Some of these regulatory elements may be gene-specific and participate in tissue-specific or combinatorial regulation of the promoter (51). Previous studies indicate that, for some neuronal genes, the proximal promoter exhibits neuron-preferential activity (60). Very often these short promoters contain crucial cis-element(s) near the TSPs. For example, a 356-bp NR1 promoter, including an RE1/NRSE site downstream of multiple TSPs, showed cell-type specificity in PC12 cells (37). The RE1/NRSE site is believed to specifically suppress neuronal promoters in non-neuronal cells and thus restrict promoter activity in neurons (61). Additionally, a 386-bp genomic sequence of the growth-associated protein 43 gene encompassing multiple TSPs confers high promoter activity in cultured neurons (62). A silencer sequence found in this 386-bp promoter restricts promoter activity in neurons. It is possible that the short fragments of the NR2A gene studied here may use the sequence in front of the TSPs as proximal 5Ј-flanking sequence to harbor gene-specific regulatory element(s) and form individual proximal promoters (50). Then, these proximal promoters, instead of the core promoters, confer upon the NR2A gene a neuron-preferential expression, although the regulatory elements involved remain elusive. It also should be noted that the involvement of the core promoter in tissuespecific expression has been recently proposed (51). Whether neuron-preferential expression of the NR2A gene is controlled by the core promoter or the proximal promoter and which regulatory sequences are involved are interesting questions, currently under further investigation in our laboratory.
In addition, our results showed that deletion of the entire 5Ј-flanking sequence did not significantly alter rat NR2A promoter activity (Fig. 4). This may be caused by a number of mechanisms. First, the most upstream TSP is a very weak site that can be revealed only by a highly sensitive RLM-5Ј-RACE method (Fig. 1). Therefore, it is possible that the contribution from this site and its 5Ј-flanking sequence to promoter activity is too weak to be detected by the reporter gene technique. Second, the sequences proximal to the TSPs within exon 1 are sufficient to drive NR2A promoter activity in neurons. Third, other regulatory cis-elements for the rat NR2A gene reside either further upstream or downstream regions and are not in the sequences studied here. Using sequences already cloned in our laboratory, we are currently investigating this possibility.
Recent studies have shown that exonic regions of many genes contain various types of cis-acting regulatory elements. Among them, GC-boxes seem universal (37,63). In this study of the NR2A promoter, three functional GC-boxes were found to interact with Sp1 and Sp4 factors and to be important for the activity of the core promoter. Sp factors are known to interact with GC-boxes located in proximal promoter regions of most neuronal genes and participate in the regulation of these genes. Sp proteins, including Sp1, Sp3, and Sp4, generally share the same binding element, although whether the mechanism is in coordination or competition remains unknown. Sp1 may function by interacting with other nuclear factors to form enhancersomes or may recruit the basal transcriptional machinery independently (64). For example, the interaction of Sp1 with TAFII130 participates in the regulation of the dopamine D2 receptor. In Huntington's patients, huntingtin interferes with this interaction, resulting in a deficit of neuronal gene expression, including the dopamine D2 gene (65). Sp1 factor is also involved in the regulation of promoters of other glutamate receptor genes. For example, Sp1 interacts with MEF2c and up-regulates the NR1 promoter in neurons (38). However, whether Sp1 and Sp4 further regulate the NR2A promoter via protein-protein interactions with other transcription factors is unknown. Putative GCG and NFB elements are also found in the exonic region of the NR2A gene. Their functional impact on the promoter needs to be further investigated.
Structurally, the exonic organization of the NR2A gene shows a certain similarity to other members of the NR2 family by starting the ORF in an internal exon. For example, mNR2B and 2C begin their ORF in exon 4 (66). In addition, the promoters of NR2B and NR2C genes also lack TATA-box and contain multiple TSPs and GC-rich sequences. These structures are common features of neuronal genes. However, the NR2A gene has a large initiative exon that contains several unusual structures, including five small ORFs, a CA-rich region, and an internal GC-rich region that have been discussed above. Wood et al. (54) previously reported that small ORFs in a limited fragment of the NR2A 5Ј-UTR interfered with the rate of translation. Our results, obtained from reporter gene transcripts and protein activity assays, further support this concept. In the present study, the full-length 5Ј-UTR has been tested, and therefore, a possible counteractive effect from the upstream 5Ј-UTR can be ruled out. More interestingly, the mouse NR2B gene also has multiple ATGs in the 5Ј-UTR (66). This similarity suggests a common evolutionary origin of NR2 subunit genes and supports their classification as a gene family. It also suggests the possibility of a common regulatory mechanism operating at the post-transcriptional level. Sequence comparisons among human, rat, and mouse NR2A genes deposited in GenBank TM indicate that they do share high homology in exonic region but display low homology in the 5Ј-flanking regions.
During preparation and review of this manuscript, two studies regarding regulatory sequence of the NR2A gene were reported. Richard et al. (67) mapped the 5Ј-most ends of the rat NR2A gene at Ϫ591 nt relative to the first codon in ORF, whereas Desai et al. (68) estimated that TSPs are located in two regions, Ϫ1133/Ϫ1078 and Ϫ486/Ϫ447 relative to the first codon of the mouse NR2A ORF. Consistent with our observation, studies of Desai et al. indicate that the NR2A neuronpreferential promoter activity resides in the sequence between bases Ϫ1253/Ϫ210. Our studies indicate that exon 1 of the rat NR2A gene covers all of these sequences ( Figs. 1 and 2). Using three overlapping deletion fragments, they claimed that a 40-bp sequence between bases Ϫ486 and Ϫ447 determines neuronal specificity. However, this 40-bp sequence was shown to have promoter activity only after it was linked with a 133-bp upstream or with 237-bp downstream sequences. Therefore, whether this 40-bp sequence initiates transcription for both fragments above or regulates TSPs that are located in either upstream or downstream sequences remains to be demonstrated.
In our studies, we demonstrate that the rat NR2A gene has multiple TSPs in a single, large initiative exon. Two clusters of these TSPs play major roles in initiation of mRNA in rat brain and in control of promoter activity preferential in neuronal cells. Our studies also reveal that three GC-boxes located in these small regions positively regulate promoter activity and interact specifically with Sp1 and Sp4 transcriptional factors. Finally, we confirm that the 5Ј-UTR sequence may participate in post-transcriptional regulation of the NR2A gene independently upon any upstream sequence either at transcriptional level or mRNA level.