Transcriptional regulation of the mouse gene encoding the alpha-4 subunit of the GABAA receptor.

The gamma-aminobutyric acid type A receptors (GABAA-Rs) mediate fast inhibitory synaptic transmission in the brain. The alpha4 subunit of the GABAA-R confers distinct pharmacological properties on the receptor and its expression pattern exhibits plasticity in response to physiological and pharmacological stimuli, including withdrawal from progesterone and alcohol. We have analyzed the promoter region of the mouse GABRA4 gene that encodes the alpha4 subunit and found that the promoter has multiple transcriptional initiation sites and lacks a TATA box. The minimal promoter for GABRA4 spans the region between -444 to -19 bp relative to the coding ATG and shows high activity in cultured mouse cortical neurons. Both Sp3 and Sp4 transcription factors can interact with the two Sp1 binding sites within the minimal promoter and are critical for maximal activity of the promoter in neurons.

In the adult brain, the expression of the GABA A -R subunits varies among brain regions and neuron types, and this regional specificity determines the pharmacological and functional properties of local GABAergic transmission (12)(13)(14). The expression pattern of the GABA A -Rs is not static, and alterations of subunit expression are observed in a number of neurological, psychiatric, and physiological conditions (reviewed in Ref. 12) including temporal lobe epilepsy (15), Huntington's disease (16), focal ischemia (17), progesterone withdrawal (11), and alcohol withdrawal/dependence (18 -22).
GABRA4, the gene that encodes the ␣4 subunit, is normally expressed at low levels in the brain, except in the ventrobasal thalamus, striatum, nucleus accumbens, olfactory tubercle, dentate gyrus, CA1 region of the hippocampus, and the outer layers of the cortex (14,23). GABA A -Rs containing the ␣4 subunit have unusual pharmacological properties that are distinct from those of receptors containing the more common ␣1 or ␣2 subunits (24 -26). These include insensitivity to benzodiazepines (26), high affinity for GABA (25), slower desensitization kinetics (24), and high sensitivity to facilitation by alcohol (27,28). An increase in levels of GABRA4 mRNA and membrane protein is observed in an animal model of premenstrual syndrome and postpartum depression, in which rats undergo withdrawal from high levels of progesterone treatment (11,29,30). Withdrawal of progesterone is accompanied by a series of behavioral changes, including reduced seizure threshold, increased anxiety, and locomotor activity, all of which occur in parallel with the increase of GABRA4 expression. An elevation of GABRA4 expression is also observed in the rat following electroshock (31), in an animal model of temporal lobe epilepsy (15), and during withdrawal from alcohol (21,22,32). These findings have led to the speculation that the ␣4 subunit of the GABA A -R might be involved in neuronal hyperexcitability.
Despite accumulating reports on the plasticity of the ␣4 subunit, the molecular mechanisms controlling this process are little understood at present. In this report we have analyzed the promoter region of GABRA4 in cultured cortical neurons by a luciferase reporter assay and identified two Sp1 binding sites within the proximal promoter region that are critical for promoter activity. The Sp1 binding site, also known as the GC box, contains an asymmetric hexanucleotide core GGGCGG sequence that is recognized by the members of Sp/KLF family of transcription factors with zinc finger motifs (33)(34)(35). The closely related Sp1, Sp3, and Sp4 have a similar binding affinity to the classic Sp1 site, as do other subfamily members, such as BTEB1, TIEG1, and TIEG2 (33)(34)(35). Our results indicate that both Sp3 and Sp4 can interact with the Sp1 sites in the GABRA4 promoter and are critical for promoter activity in neurons.
Wisconsin) at default settings. Restriction sites for MluI and NheI were included at the 5Ј-end of the PCR primers to facilitate subsequent cloning steps. PCR products were gel purified using the QIAquick gel extraction kit (Qiagen, Valencia, CA) and ligated into pGEM-T Easy vector (Promega, Madison, WI). Mini-prep DNA was isolated from white colonies and screened by MluI and NheI digestion. Fragments from the MluI and NheI digestion were gel purified and ligated into the MluI and NheI sites of pGL3-basic vector (pGL-BV) (Promega). Mutant vectors pLuc-P7⌬1 through pLuc-P7⌬8 were generated from the wildtype plasmid pLuc-P7 using the ExSite PCR-based system (Stratagene, La Jolla, CA) with primers listed in Table I. Constructs pLuc-P7A, pLuc-P7B, and pLuc-P7AB were generated from the pLuc-P7 vector using the QuikChange site-directed mutagenesis kit (Stratagene) with primers listed in Table I. All promoter constructs were confirmed by DNA sequencing.
Mouse and human genomic DNA sequences were obtained by searching the genome data base using Blat program. Both mouse and human genomic sequences were based on 2002 assembly. DNA sequence analysis was performed using the AliBaba2 program and TRANSFAC data base for predicting transcription factor binding sites. Repetitive DNA sequence was analyzed using a tandem repeats finder program. The 5Ј-untranslated region putative stem-loop structure was analyzed by a RNA secondary structure prediction program.
5Ј-Rapid Amplification of cDNA Ends (RACE) and Primer Extension Assays-Total RNA was isolated from the brain tissue of adult C57/BL6 mice (Jackson Laboratory and Charles River Laboratories, Wilmington, MA) using the TRIzol reagent (Invitrogen). 5Ј-RACE analysis of transcription initiation sites was carried out by synthesizing RACE-ready cDNA from total brain RNA using the SMART RACE cDNA amplification kit (Clontech, Palo Alto, CA) and by using the FirstChoice RLM-RACE kit (Ambion, Austin, TX). The RACE-ready cDNA was then subjected to PCR amplification using the universal anchor primer and the GABRA4 specific primer anti-4 (5Ј-CGCAATCGCGGGTACCTTCT-GGACAG-3Ј), which anneals to the region ϩ8 to ϩ33 relative to the coding ATG. PCR amplification was performed by denaturing the cDNA for 2 min at 94°C followed by 35 cycles of 94°C for 10 s, 55°C for 30 s, and 72°C for 2 min. PCR products were subcloned into pGEM-T Easy vector (Promega) and clones containing the PCR inserts were sequenced to locate the 5Ј-end of the transcript for GABRA4.
The primer extension assay was carried out by using the Primer Extension System (Promega). 3 ϫ 10 6 cpm of the [␥-32 P]ATP (Perkin-Elmer Life Sciences, Boston, MA) labeled anti-4 primer was annealed to 30 -40 g of total brain RNA at 55°C for 30 min following denaturation at 95°C for 5 min. Annealed samples were slowly cooled to room temperature prior to addition of reverse transcriptase. After incubating for 1 h at 42°C, the reverse transcription (RT) reaction mixture was purified by phenol-chloroform extraction followed by ethanol precipitation. The primer extension products were analyzed on a 6% denaturing TBE gel containing 8 M urea using a DNA sequencing size marker prepared with the AmpliCycle sequencing kit (Applied Biosystems, Foster City, CA).
RT-PCR and Real-time PCR Analyses-For RT-PCR analysis, cDNA was reverse transcribed from 0.5 to 1 g of total RNA in a 30-l reaction containing 0.25 g of random hexamer (Qiagen), 1 mM dNTP, 2.5 units of RNaseOut (Invitrogen), and 5 units of reverse transcriptase (New England Biolabs, Beverly, MA) with 1ϫ RT reaction buffer. The RT reaction mixture was purified by one step of phenol-chloroform extraction. The primers for GAPDH, GB450 (5Ј-ACCACAGTCCATGCCATC-AC-3Ј) and GB449 (5Ј-TCCACCACCCTGTTGCTGTA-3Ј), which amplify both human and mouse genes, were described previously (37). The primers for GABRA4 mRNA were individually designed and the resulting PCR products were confirmed by DNA sequencing. GABRA4 RT-PCR primers for human (HEK293 cells) were: sense (5Ј-CTGGAC-AAAAGGTCCTGAG-3Ј) and antisense (5Ј-CTCTCTGCACTGGAGCAG-C-3Ј), and for mouse (neurons and astrocytes) were: sense (5Ј-TACAC-CTGGACCAAAGGCCC-3Ј) and antisense (5Ј-TCTGTGTGTTTCTCCT-TCAGCACA-3Ј). The 50-l PCR reaction mixture contained 2 l of cDNA, 0.2 M of each primer, 0.2 mM of each dNTP, 2.5 units of Taq polymerase, and 1ϫ PCR buffer with Mg 2ϩ (Roche Applied Science). PCR amplification was carried out by denaturing the cDNA for 2 min at 94°C followed by 28 cycles (GAPDH) or 35 cycles (GABRA4) of 94°C for 10 s, 58°C for 30 s, 72°C for 2 min, and a final step of elongation for 7 min at 72°C.
For real-time PCR analysis, 0.5-1 g of total RNA was isolated by TRIzol reagent (Invitrogen) and was treated with RNase-free DNase I (Promega) for 15 min followed by heat inactivation. cDNA was prepared using the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen) with random hexamer followed by RNase H digestion. Real-time PCR primers were designed by OligoPerfect Designer (Invitrogen) using default settings, which generates T m matched primer pairs that amplify 80 -250 bp of target sequence. The GAPDH primers were forward (5Ј-AACTTTGGCATTGTGGAAGG-3Ј) and reverse (5Ј-ACACATTGGGGGTAGGAACA-3Ј) and amplified a product of 223 bp. The GABRA4 primers were forward (5Ј-TAAACG-AATCCCCAGGACAG-3Ј) and reverse (5Ј-GACGCAGCCTGTTGT-CATAA-3Ј) and amplified a product of 105 bp. The amplified PCR products were confirmed by DNA sequencing. Real-time PCR were carried out in a DNA Engine Thermocycler (MJ Research, Reno, NV) using SYBR green I dye. The 50-l reaction mixtures contained 0.5 M of each primer, 1 to 2 l of cDNA, and 25 l of Platinum Quantitative PCR SuperMix-UDG (Invitrogen). The conditions for realtime PCR were 2 min at 95°C, and 45 (GAPDH) or 50 (GABRA4) cycles for 15 s at 95°C, 30 s at 60°C, 30 s at 72°C followed by a gradient step from 55 to 95°C to obtain a melting curve for each sample. Plasmids containing GAPDH and GABRA4 cDNA were linearized by EcoRI digestion and were used as PCR standards in 5-fold serial dilutions. The cycle threshold (C t ) line was positioned manually at the point where the fluorescent signals surpassed background noise and started to increase linearly. Standard curves were determined for each reaction by plotting C t as a function of log [DNA]. C t for each sample was obtained from the fluorescence trace where it intersected with the C t line. DNA concentration was obtained by converting the C t to an initial quantity by interpolation from the standard curve run on the same plate. The homogeneity of PCR products was analyzed by the observation of a single peak of the fluorescence versus temperature trace at the melting temperature and by agarose gel electrophoresis. In each experiment, the average values of triplicate samples were used for each data point, and a minimum of three independent neuronal cultures were included for each analysis to obtain the final results. A control sample with no reverse transcriptase in the RT reaction was included in each experiment to monitor for genomic DNA contamination. Cell Culture, Transfection, and Dual Luciferase Assay-Mouse cortical neurons were prepared as previously described with modifications (38). Briefly, cerebral cortices were isolated from embryonic day 17 or 18 C57/BL6 mice (Jackson Laboratory and Charles River Laboratories). The dissected cortices were minced in 1ϫ Dulbecco's phosphate-buffered saline (Mediatech, Herndon, VA) and treated for 30 min at 37°C with 20 units/ml papain (Calbiochem, San Diego, CA) in 1ϫ Earle's balanced salts solution (Sigma) containing 1 mM L-cysteine (Calbiochem) and 0.5 mM EDTA (Sigma). Digested cortical cells were further dissociated in Earle's balanced salts solution containing 2 mg/ml albumin (Sigma) and 100 units/ml DNase I (Worthington, Lakewood, NJ) by gentle trituration with a fire-polished Pasteur pipette. The resulting single cell suspension was layered on top of a 20 mg/ml albumin solution in Earle's balanced salt solution for a single step of discontinuous gradient centrifugation at 800 rpm for 10 min at room temperature. The cell pellet was then resuspended in phenol red-free minimum essential medium (PRF-MEM) (Invitrogen) supplemented with 5% horse serum (Sigma) and 0.5 mM L-glutamine (Sigma). Cells were seeded at a concentration of 1 ϫ 10 6 cells/ml in 12-well plates (Corning, Corning, NY) pre-coated overnight with 0.05 mg/ml poly-D-lysine (Sigma). On day 3 in vitro, half of the media was replaced with phenol red-free neurobasal medium (Invitrogen) containing 0.5 mM glutamine, 1ϫ B27 supplement (Invitrogen), and the anti-mitotic agents 10 Ϫ4 M uridine and 10 Ϫ5 M 5-fluoro-2-deoxyuridine (Sigma). Fresh PRF neurobasal medium was added to replace half of the media every other day thereafter. Astrocytes were prepared by culturing the cortical cells at a density of 2 ϫ 10 5 cells/ml in 5% equine serum (HyClone, Logan, UT) PRF-MEM. Astrocytes continued to propagate and constituted more than 90% of the cells in the mixed culture population after 2 weeks of expansion. Transfection of neurons was generally performed on days 7-9 in vitro for luciferase analysis, whereas transfection of astrocytes was performed after 2 weeks in culture. HEK293 cells were cultured as described previously (39) and transfected when the cells reached 70 -80% confluence. All cells were maintained at 37°C in a humidified environment of 5% CO 2 .
Plasmid DNA used for transfection was obtained using the HiSpeed maxi-prep kit (Qiagen) and re-purified by sodium acetate and ethanol precipitation. For each construct, at least three independent plasmid DNA preparations were used for transfection experiments. For each transfection, 1 g of plasmid DNA was added to 5 ng of CMV promoter driven Renilla luciferase construct (pRL-CMV) (Promega) for co-transfection. The pRL-CMV construct served as the internal control for the dual luciferase assay (Promega). Transfection was performed using a combination of the Nupherin (Biomol Research Laboratories, Plymouth Meeting, PA) and LipofectAMINE 2000 (Invitrogen) reagents. 1 g of plasmid DNA was first mixed with 6 g of Nupherin in 150 l of PRF-MEM for 15 min and then combined with 150 l of PRF-MEM containing 3 l of LipofectAMINE 2000 for another 15 min at room temperature. The culture media was replaced with 300 l of transfection media containing the LipofectAMINE-Nupherin-DNA complex. After incubating for 1 h, transfection media was replaced with 1 ml of culture media. Luciferase activity was analyzed 20 -24 h post-transfection with a Turner Design 20/20 luminometer (Turner Designs, Sunnyvale, CA) using the dual-luciferase reporter assay system (Promega). Triplicate samples were measured for each construct, and the average values of luciferase light units were used for data analysis. The negative control pGL-BV plasmid was the pGL3-basic vector (Promega), which contained no eukaryotic promoter or enhancer sequence. The luciferase activity was defined as the ratio of the light units of GABRA4 promoter driven firefly luciferase versus CMV promoter driven Renilla luciferase. The relative luciferase activity for each sample was normalized to that of the SV40 minimal promoter (P SV40 ) driven luciferase construct for comparison among different experiments. The expression vectors pEVR2/Sp1, pRC/CMV-Sp3, and pRC/CMV-Sp4 were obtained from Professor G. Suske (40,41).
Gel Shift Assay-Mouse brain nuclear extract was prepared as described previously with modifications (42). Briefly, the brain tissue was frozen in liquid nitrogen and shattered into tissue powder. About 500 mg of tissue was lysed in buffer A containing 10 mM HEPES, pH 7.9, 150 mM NaCl, 1 mM EDTA, 0.5 mM dithiothreitol and protease inhibitor mixture (Sigma) and homogenized for 10 strokes using a glass homogenizer with loose type pestle. Tissue debris was discarded by centrifuging the homogenate at 2,000 rpm for 30 s at 4°C, and nuclei were collected by centrifuging the supernatant at 5,000 rpm for 5 min at 4°C. The nuclei were re-suspended in 200 -500 l of buffer B containing 20 mM HEPES, pH 7.9, 25% glycerol, 420 mM NaCl, 1.2 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, and protease inhibitor mixture and incubated for 30 min at 4°C by gentle shaking. Nuclear extract was collected by centrifuging the lysate at 14,000 rpm at 4°C for 15 s.
Oligos Sp1-A (5Ј-GAGCGGCGAGGGAGGGGGCGGGCGCGCAGG-TC-3Ј) and Sp1-B (5Ј-GCGCGCAGGTCCCGCCTCCCCTGGCCGCG-T-3Ј) were synthesized (Qiagen) and double-stranded oligos were prepared by annealing the complementary strands in 10 mM Tris-HCl, 100 mM NaCl, 1 mM EDTA buffer, pH 8.0. The consensus Sp1 binding sites were underlined in the oligo sequences. The sequences of mutant oligos MutA and MutB were listed in Table I. The Sp1 core oligo was obtained from Promega. The in vitro binding reactions were carried out using a gel shift kit (Roche Applied Science). The oligos were either end-labeled with [␥-32 P]ATP or digoxigenin. About 10 ϫ 10 3 cpm or 0.4 ng of labeled oligo was used for each binding reaction. Human rSp1 and HeLa nuclear extract were obtained from Promega. 0.8 g of antibodies against c-Myc (C-19), Sp1 (PEP 2), Sp3 (D-20), and Sp4 (V-20) (Santa Cruz Biotechnology) were used in supershift analysis. For the digoxigenin-labeled probes, the DNA-protein complex was blotted to a Zeta probe membrane (Bio-Rad) by capillary contact transfer for 0.5-2 h. The DNA was cross-linked to the membrane with a UV cross-linker (Stratagene). Immunological detection of digoxigenin-labeled DNA was then carried out according to the manufacturer's instructions (Roche Applied Science).

Identification of the Transcriptional Initiation
Site for GABRA4 -Previously we isolated a GABRA4 cDNA clone from a mouse library (28); the 5Ј-end of this clone is located at Ϫ144 bp relative to the coding ATG. As a first step in characterizing the mouse GABRA4 promoter, we conducted RACE analysis of total RNA isolated from the adult mouse brain to map the 5Ј-end of the GABRA4 mRNA. We designated the nucleotide residue A of the coding ATG as the ϩ1 position in GABRA4. Sequence analysis of the RACE PCR products revealed multiple potential start sites for the GABRA4 mRNA clustered in the first exon within a region from Ϫ150 to Ϫ113 bp (Figs. 1 and 2B), which included the beginning of our cDNA clone at position Ϫ144. The sequences of the RACE PCR products corresponded exactly with the published mouse genomic sequence (36) without interruption, indicating that no intron was present at the 5Ј-end of GABRA4 coding exon 1. In a search of the GenBank TM data base, we found that the 5Ј-end of a mouse neocortex GABRA4 cDNA transcript AK043806 was located at position Ϫ114, which was within the initiation region that we had identified. A second GABRA4 transcript AK013727, which was isolated from the hippocampus, was reported to originate at position Ϫ499, about 300 -400 bp further upstream. We wondered whether multiple transcripts for GABRA4 might exist, or whether the short transcripts we detected might have arisen as the result of preferential amplification of short fragments. We therefore designed three additional PCR primers that annealed to the sequence upstream of position Ϫ150. These primers should have enabled the amplification of the long transcript of GABRA4 mRNA. However, we were not able to detect any long PCR products, even after two rounds of nested PCR with an additional 30 cycles. This observation suggested that the short transcripts we identified did in fact represent the major products of GABRA4 transcription in the mouse.
We performed primer extension analysis to locate the transcription start sites more precisely. The [␥-32 P]ATP-labeled anti-4 primer (complementary to the first exon of GABRA4) was used to prime RT and the resulting cDNA products were resolved on a sequencing gel. Primer extension analysis revealed that transcription start sites for GABRA4 were located at positions Ϫ184, Ϫ149, Ϫ132, and Ϫ113 ( Fig. 2A). The product corresponding to position Ϫ113 was 3-fold more abundant than the others. Based on these results, we conclude that the mouse GABRA4 is transcribed from multiple locations within exon 1, spanning the region from Ϫ184 to Ϫ113 bp, and that position Ϫ113 is most likely represents a major transcription start site.
Consensus Sequence Elements in the 5Ј-Flanking Region of GABRA4 -We searched the 5Ј-proximal region of GABRA4 for sequence motifs that were commonly found in the promoters of genes expressed in a neuron-specific manner. Sequence analysis of a 3.7-kb fragment of the 5Ј-flanking region revealed no TATA box, but the presence of two tandem direct repeats, between Ϫ1267 and Ϫ1214 bp, and between Ϫ143 and Ϫ112 bp, and one short poly-pyrimidine track at Ϫ1213 bp. The repeats located at Ϫ1267 bp contained 10 CTCTGT motifs, whereas the repeats at Ϫ143 bp contained two direct repeats of 16 nucleotides long. In many TATA-less promoters, an initiator (Inr) element with the consensus sequence YYANWYY had been identified and found to direct transcription (43). Several elements bearing the Inr consensus sequence were found within the region of Ϫ1.0 kb to ϩ1 bp but none of these putative Inr elements was located near the transcription start sites that we identified (Fig. 1). Neural-specific promoters very often contain a 21-bp neuron-restrictive silencer element (NRSE) in the proximal promoter region (44 -46). In an attempt to search for NRSE-like elements in the GABRA4 promoter, we examined a 10-kb region of the mouse gene upstream and downstream of the coding ATG. No sequence motif that shared Ͼ70% homology with the consensus NRSE was found within this region. This observation indicates that elements other than the NRSE are likely present in the GABRA4 promoter to control neuronal-specific expression.
Identification of the Minimal Promoter Region for Mouse GABRA4 -To identify the minimal promoter region, we employed the luciferase reporter assay to test the ability of promoter fragments to direct the expression of the reporter gene in cultured cortical neurons and cells derived from non-neuronal origins. GABRA4 is expressed in a limited number of brain regions including the outer layers of the cortex (14,23). We reasoned that the promoter constructs should be active in an environment in which endogenous GABRA4 was expressed. We first used RT-PCR analysis to determine whether the endogenous GABRA4 was expressed in cultured cortical neurons. With similar levels of GAPDH expression, GABRA4 mRNA was minimally detectable in cortical cells after 2 days in vitro, and then increased from 4 days in vitro, reaching its peak after 10 days in vitro (Fig. 3A, lanes 1-5). As negative controls for RT-PCR, we synthesized cDNA from 10 day in vitro cells either without the addition of reverse transcriptase or without RNA in the RT reactions. No RT-PCR signals were detected from these two control samples for either GAPDH or GABRA4 message (Fig. 3, lanes 7 and 8).
We also performed real-time PCR analysis to quantify the amount of GABRA4 expression in neurons cultured from days 2-10 in vitro. We used the arbitrary ratio of GABRA4 and GAPDH to represent the relative quantity of GABRA4 expression. Serial dilutions of linearized plasmids carrying GABRA4 and GAPDH cDNA were used to generate standard curves. Fluorescence emission of the SYBR green I dye upon binding of the double-stranded DNA was recorded in each PCR cycle after the annealing step. The quantity of the cDNA in each sample was determined by linear interpolation from the standard curve run on the same plate. A control sample with no reverse transcriptase in the RT reaction was included in each experiment for monitoring genomic DNA contamination. Basal level of GABRA4 mRNA expression was detected in 2-day in vitro cells (Fig. 3B). GABRA4 expression steadily increased with time in culture, reaching its peak at 10 days in vitro, at which expression was about 10-fold higher than at 2 days in vitro (Fig. 3B). We then employed the dual luciferase assay to test whether a 3.7-kb GABRA4 promoter fragment P1 could recapitulate the time course of endogenous GABRA4 expression. The P1 fragment was directly cloned by PCR amplification of a region from Ϫ3716 to Ϫ19 bp of GABRA4. The firefly luciferase construct that carried the P1 promoter fragment was designated as pLuc-P1. A CMV promoter driven Renilla luciferase construct pRL-CMV was co-transfected with the promoter construct as internal control for normalization in dual luciferase assay. We transfected the construct pLuc-P1 into neurons after various times in culture and assayed for luciferase activity. The luciferase activity was defined as the ratio of the light units of GABRA4 promoter driven firefly luciferase versus CMV promoter driven Renilla luciferase. The relative luciferase activity for each sample was normalized to that of the SV40 minimal promoter-driven luciferase construct P SV40 for comparison among different experiments. The pLuc-P1 construct did not elicit significant luciferase activity on 2 days in vitro, but the promoter activity of this construct increased with time in culture and reached activity of 104% relative to the control P SV40 in neurons after 10 days in vitro (Fig. 3C). The negative control pGL-BV vector, which contained the luciferase coding sequence but no eukaryotic promoter or enhancer sequence, only gave rise to basal levels of expression (Fig. 3C). This result demonstrates that a 3.7-kb fragment of the 5Ј region of GABRA4 is able to drive the expression of the luciferase reporter gene in a neuronal environment and that the time course of promoter activity in neurons is consistent with the profile of endogenous GABRA4 expression. The promoter activity of pLuc-P1 was consistently reproducible between days 7 and 9 in vitro, so neuronal cultures at this age were used for all subsequent experiments. No endogenous GABRA4 transcript was detected in primary mouse astrocytes or human embryonic kidney cells (HEK293) (data not shown). The activity of the pLuc-P1 construct was low in both astrocytes and HEK293 cells (Fig. 4, A  and B).
The core promoter of a gene is generally located in the vicinity of the transcription start site(s), whereas additional regulatory elements that direct tissue-and cell type-specific expression of the gene may lie either upstream or downstream. To further characterize the elements that regulate the expression of GABRA4, we tested a series of constructs representing fragments of the 5Ј-flanking domain of GABRA4 for their ability to direct transcription of the reporter gene in neuronal and non-neuronal cells.
Promoter constructs pLuc-P1 through pLuc-P7, which carried 3.7-, 2.6-, 2.3-, 2.2-, 1.3-, 0.9-, and 0.4-kb fragments of the 5Ј region of GABRA4, respectively, gave rise to similar levels of luciferase activity in cultured cortical neurons, but little to no activity in astrocytes or HEK293 cells (Fig. 4, A and B). When the construct was truncated to 166 bp, as in pLuc-P8, the promoter activity in neurons decreased sharply relative to all the other constructs analyzed (Fig. 4, A and B). We reasoned that the sudden change of activity between pLuc-P7 and pLuc-P8 might be because of the loss of one or more positive regulatory element(s).
We also assessed the functional significance of the putative Inr element at position Ϫ488, because it was located close to the 5Ј-end of the published long GABRA4 transcript AK013727. Two deletion constructs, pLuc-P1⌬1 and pLuc-P1⌬2, were generated by PCR, in which the start sites for the short transcripts were eliminated. These two deletion constructs showed drastically reduced promoter activity (Fig. 4, C and D), indicating the loss of critical transcriptional elements in these constructs and the inability of the putative upstream Inr element to support GABRA4 expression. Taken together, these results support the idea that the minimal promoter region for driving neuronspecific expression of GABRA4 is located within the region Ϫ444 to Ϫ19.
cis-Regulatory Elements Critical for GABRA4 Minimal Promoter Activity-To delineate the elements that were critical for GABRA4 promoter activity, we generated internal deletion mutants of the pLuc-P7 construct and assayed their activity for directing transcription in neurons. Deletion constructs pLuc-P7⌬1, -⌬2, -⌬3, -⌬7, and -⌬8 showed similar promoter activity compared with the parental pLuc-P7 construct (Fig. 5). These deletions eliminated the two repeats at Ϫ143 bp and a series of putative binding sites for transcription factors, specifically an Sp1 site located at the 5Ј-end of the promoter fragment, Ap1, c-Myb, and E-box. These findings suggest that the potential binding sites within these regions are not required for GABRA4 promoter activity in neurons in a transient transfection assay. The pLuc-P7⌬4, -⌬5, and -⌬6 constructs harbored deletions that eliminated the two Sp1 sites close to the 3Ј-end of the FIG. 4. GABRA4 promoter construct activity in cultured neurons, astrocytes, and HEK293 cells. A and C, schematic illustrations of the promoter constructs used for the transfection experiments. The 5Ј and 3Ј boundaries of the promoter for each construct are identified by numbers. The positions of the two direct repeats, transcription initiation sites, and 5Ј-end of AK013727 cDNA for GABRA4 are marked on each construct. B and D, luciferase reporter analysis of promoter constructs in cultured cortical neurons (n ϭ 6 -36), astrocytes (n ϭ 6 -9), and HEK293 cells (n ϭ 6 -12). The relative luciferase activity for each construct is calculated as the ratio of the luciferase activities of GABRA4 promoter driven construct versus the P sv40 driven construct. The results are expressed as mean relative luciferase activity Ϯ S.E. promoter and other elements (Fig. 5A). We designated the Sp1 site at position Ϫ214 as Sp1-A, and the one at position Ϫ198 as Sp1-B. In construct pLuc-P7⌬5, deletion of the c-Myb, E-box, and the Sp1-A site resulted in about 60% reduction of promoter activity compared with the pLuc-P7 construct (Fig. 5B). In construct pLuc-P7⌬4, deletion of the Sp1-A site resulted in about 50% reduction of promoter activity, suggesting that the Sp1-A site and its surrounding sequences were critical for promoter activity (Fig. 5B). Deletion of the two Sp1 sites and the two direct repeats at Ϫ143 bp (as in pLuc-P7⌬6) drastically reduced the promoter activity by Ͼ80%, whereas deletion of the repeats alone, as in pLuc-P7⌬7, did not affect the promoter activity significantly (Fig. 5B). These results suggest that the two Sp1 sites within the region Ϫ257 to Ϫ179 are critical for maximal GABRA4 promoter activity.
Sp3 and Sp4 Interact with cis-Elements in the GABRA4 Promoter-To investigate whether the putative Sp1 elements in the GABRA4 promoter can interact with members Sp/KLF family of transcription factors, we used a gel shift assay to test the ability of a recombinant human Sp1 protein (rSp1) to bind to the two Sp1 sites on the GABRA4 promoter. Two doublestranded oligonucleotides containing the Sp1-A and Sp1-B sites were used in the gel shift assay. The Sp1-A oligo contained the consensus GC-box core sequence GGGCGG flanked by the naturally occurring GABRA4 promoter sequence. The Sp1-B oligo contained the inverse complementary of the Sp1 consensus core sequence, CCCGCC, flanked by the promoter sequence. DNAprotein complexes were formed when either Sp1-A or Sp1-B oligo was incubated with rSp1 (Fig. 6, A and B, lanes 2). Competition binding studies were used to further examine the specificity of rSp1 binding to the Sp1-A and Sp1-B sites. Binding of rSp1 to either Sp1-A or Sp1-B sites was inhibited by the addition of unlabeled Sp1-A (Fig. 6A, lanes 3-5) or Sp1-B oligo (Fig. 6B, lanes 3-5). An Sp1 core oligo containing the consensus GC box was also able to compete with the Sp1-A and Sp1-B oligos for the binding of rSp1 (Fig. 6, A and B, lanes 6 -8).
Mutations were then made in the Sp1 binding core sequence on the Sp1-A and Sp1-B oligos and the resulting oligos were designated as MutA and MutB, respectively. The mutant oligos were unable to form a complex with the rSp1 protein (data not shown). Moreover, a 100-fold molar excess of either MutA or MutB was not able to compete with Sp1-A or Sp1-B binding for rSp1 (Fig. 6, A and B, lanes 9 -11). These results suggest that rSp1 can indeed interact specifically with the two Sp1 sites on the GABRA4 promoter.
To test the contributions of the two Sp1 sites to the GABRA4 promoter activity, we generated mutations at the Sp1-A and Sp1-B sites on pLuc-P7, and assayed the activity of mutant constructs in cortical neurons. Mutation of the Sp1-A site alone decreased the activity of the construct by 20%, whereas mutation at the Sp1-B site alone reduced promoter activity by 60% (Fig. 6C). When both Sp1 sites were mutated, the promoter activity was further reduced to Ͼ80% (Fig. 6C). These results indicate that both Sp1 sites contribute to the promoter activity and the Sp1-B site has a major effect on promoter function.
The Sp/KLF family of transcription factors comprises at least  (lanes 4, 7, and 10), and 2 nM (lanes 5, 8, and 11). C, luciferase reporter analysis of promoter constructs with mutations at Sp1 binding sites in cultured cortical neurons (n ϭ 7-36). The asterisks represent significant differences at the level of p Ͻ 0.01 (**) and p Ͻ 0.05 (*) when different constructs are compared with pLuc-P7 using one-way analysis of variance and post-hoc Dunnett's test.

FIG. 5. Internal deletion analysis of the pLuc-P7 promoter construct in cortical neurons.
A, schematic illustration of the deletion constructs used. The deletion boundaries for each construct are identified by numbers. The positions of the direct repeats and potential transcription factor binding sites are indicated on each construct. B, luciferase reporter analysis of mutant promoter constructs in cultured cortical neurons (n ϭ 7-36). The asterisks represent significant differences at the level of p Ͻ 0.01 when different constructs are compared with pLuc-P7 using one-way analysis of variance and post-hoc Dunnett's test. 16 mammalian gene products that share the signature zinc finger motifs for DNA binding (33)(34)(35). Among them, Sp1, Sp3, and Sp4 are closely related transcriptional activators that recognize identical DNA sequences. To investigate which Sp/KLF transcription factors interact with the GABRA4 promoter, we conducted gel shift analysis using nuclear extract isolated from adult mouse brain.
Three DNA-protein complexes were formed when Sp1-A double-stranded oligo was incubated with the brain nuclear extract (Fig. 7A, lane 2). Addition of unlabeled Sp1-A oligo was able to abolish the formation of all three complexes, whereas addition of MutA oligo did not inhibit the complex formation (Fig. 7A,  lanes 3 and 4). The Sp1 core oligo was able to abolish the formation of complex I and weakly inhibit complex III, but did not affect the formation of complex II (Fig. 7A, lane 5). Addition of 0.8 g of rabbit antibodies against c-Myc (control) and Sp1 did not alter the binding pattern of the brain nuclear extract to the Sp1-A oligo (Fig. 7A, lanes 6 and 7). However, addition of rabbit antibodies against Sp3 and Sp4 abolished the formation of complex I without affecting complexes II and III significantly (Fig. 7A, lanes 8 and 9). We noticed that no supershifted band was observed when anti-Sp3 and -Sp4 antibodies were added in the binding reactions although the formation of complex I was abolished (Fig. 7A, lanes 8 and 9). These results implied that the addition of antibodies in the reactions might prevent the formation of DNA-protein complex at the Sp1-A site. Taken together, these results indicate that multiple factors can interact with the Sp1-A oligo and that both Sp3 and Sp4 are present in the bound complexes at the Sp1-A site.
In contrast to the Sp1-A site, only a single DNA-protein complex was formed when Sp1-B oligo was incubated with brain nuclear extract (Fig. 7B, lane 2, complex II). Unlabeled Sp1-B oligo was able to abolish the formation of the bound complex, whereas MutB did not inhibit the complex formation (Fig. 8A, lanes 3 and 4). The Sp1 core oligo strongly inhibited the formation of complex II but did not totally abolish the binding (Fig. 7B, lane 5). Addition of 0.8 g of rabbit antibodies against c-Myc (control) and Sp1 did not alter the binding pattern of the brain nuclear extract to the Sp1-B oligo (Fig. 7B,  lanes 6 and 7). However, addition of rabbit antibodies against Sp3 and Sp4 resulted in the appearance of supershifted DNAprotein-antibody complex I (Fig. 7A, lanes 8 and 9). The antibodies against Sp3 were able to supershift most of bound com- FIG. 7. Identification of the Sp/KLF factors bound to the Sp1-A and Sp1-B sites using nuclear extracts derived from brain. Gel shift assay of brain nuclear extract binding to the Sp1-A and Sp1-B sites. 8 -10 g of brain nuclear extract were incubated with 10 ϫ 10 3 cpm of 32  plex II, whereas the antibodies against Sp4 supershifted the complex to a lesser extent. We also examined whether the brain-specific Sp/KLF factor BTEB1, a transcriptional activator that recognizes the GC-box sequence, was present in the bound complexes at the Sp1-A and Sp1-B sites. No supershifted band was observed in the gel shift assay using antibodies against BTEB1, suggesting the absence of the BTEB1 epitope in the bound complexes at the Sp1-A and Sp1-B sites (data not shown). Taken together, these results indicate that Sp3 and Sp4 can interact with both Sp1 sites and may potentially function to regulate GABRA4 expression.
We wondered whether a different binding pattern of the Sp/KLF factors to the Sp1-A and Sp1-B sites might exist in a non-neuronal cell type. We conducted gel shift assays using nuclear extract from HeLa cells. Two DNA-protein complexes were formed when either Sp1-A or Sp1-B oligo was incubated with the HeLa nuclear extract (Fig. 8, A and B, lanes 2, complex  II and III). The complex formation at Sp1-A and Sp1-B sites was inhibited by the addition of unlabeled Sp1-A and Sp1-B oligos, respectively (Fig. 8, A and B, lanes 3). Sp1 core oligo was able to compete with the binding complex formation at both Sp1-A and Sp1-B sites (Fig. 8, A and B, lanes 4). The addition of MutA and MutB oligos did not affect DNA-protein complex formation at the Sp1-A and Sp1-B sites, respectively (Fig. 8, A and B, lanes 5). Addition of antibodies against c-Myc (control), Sp2, and Sp4 did not alter the binding pattern at either Sp1-A or Sp1-B sites (Fig. 8, A and B, lanes 6, 8, and 10). In contrast, addition of anti-Sp1 antibodies did not affect the formation of complex III at Sp1-A and Sp1-B sites but partially abolished complex II, which resulted in the appearance of a supershifted complex I (Fig. 8, A and B, lanes 7). Addition of anti-Sp3 antibodies abolished the formation of complex III at Sp1-A and Sp1-B sites without affecting complex II significantly (Fig. 8, A  and B, lanes 9). These results suggest that both Sp1 and Sp3 factors are able to interact with the Sp1-A and Sp1-B sites on the GABRA4 promoter in a non-neuronal cell line. The unique binding patterns of Sp/KLF factors to the Sp1-A and Sp1-B sites might result in differential regulation of GABRA4 in different cell types.
The Roles of Sp3 and Sp4 in Regulating the GABRA4 Promoter-To further investigate whether Sp3 and Sp4 were functionally involved in the expression of GABRA4, we analyzed the influence of Sp3 and Sp4 on the endogenous GABRA4 expression by real-time PCR. Expression vectors carrying Sp1, Sp3, and Sp4 cDNA were transfected into day 7 in vitro neurons and the expression of endogenous GABRA4 was analyzed 2 days after transfection. Transfection of pcDNA3.1 vector that did not contain any Sp/KLF factor was used as a negative control. Overexpression of Sp1 slightly increased the GABRA4 level to about 1.3-fold over the control. In contrast, overexpression of Sp3 and Sp4 augmented GABRA4 expression to about 2.3-fold (Fig. 9A). We also examined whether expression of the Sp/KLF factors could affect the promoter activity on the plasmid construct. Co-transfection of the pLuc-P7 vector with pcDNA3.1, Sp1, Sp3, or Sp4 expression vectors was carried out in day 7 in vitro neurons and the luciferase activity was assayed 2 days after transfection. We observed that overexpression of Sp1 did not affect the pLuc-P7 promoter activity, whereas expression of Sp3 and Sp4 increased the promoter activity to 1.3-and 1.7-fold (Fig. 9B). These results indicate that both Sp3 and Sp4 are able to augment GABRA4 expression and are likely to be involved in the regulation of this gene in vivo. DISCUSSION In this study, we have analyzed the proximal promoter region of GABRA4, a gene that has been implicated in neuronal hyperexcitability. By truncating the 5Ј-promoter region of GABRA4, we identified a minimal promoter for GABRA4 that was able to drive high levels of expression of the reporter gene in cultured cortical neurons much more efficiently than in non-neuronal cells. The minimal promoters for a number of neuronal genes have been shown to direct neuronal specific expression (37,44,45,(47)(48)(49)(50)(51)(52). These neural promoters often contain an NRSE in the proximal promoter region (44,45). The neuron-restrictive silencing factor (NRSF/REST), a zinc finger repressor, which is expressed primarily in non-neuronal tissues, can interact with the 21-bp NRSE sequence to repress gene expression (44 -46, 53). However, the effect of the NRSEcontrolled neural-specific gene expression varies among different promoters. For instance, the NRSE of the SCG10 gene suppresses downstream gene expression by more than 10-fold (46,53), whereas the glutamate receptor 2 (GluR2) (52) NRSE FIG. 9. The effects of the Sp/KLF family of transcription factors on endogenous GABRA4 mRNA expression and the activity of promoter construct pLuc-P7 in cultured mouse cortical neurons. A, real-time PCR analysis of GABRA4 expression in cortical neurons transfected with expression vectors of Sp1, Sp3, and Sp4. The vectors pEVR2/Sp1, pRC/CMV-Sp3, and pRC/CMV-Sp4, which carried Sp1, Sp3, and Sp4, respectively, were transfected into day-7 in vitro neurons, and RNA was collected 48 h later. The pcDNA3.1 vector was used as a negative control. The concentrations of GABRA4 and GAPDH were measured for each sample based on the standard curves generated using linearized plasmid DNA of known concentrations. The relative concentration of GABRA4 mRNA was then expressed in terms of the mean ratio of GABRA4 versus GAPDH. The results are reported as the mean increase (relative to the negative control) Ϯ S.E. (n ϭ 4). The asterisks represent significant differences at the level of p Ͻ 0.01 (*) and p Ͻ 0.05 (**) by one-way analysis of variance and post-hoc Dunnett's test. B, the vectors pEVR2/Sp1, pRC/CMV-Sp3, and pRC/CMV-Sp4, which carried Sp1, Sp3, and Sp4, respectively, were co-transfected with pLuc-P7 construct into day 7 in vitro neurons and luciferase activity was measured 48 h later. The pcDNA3.1 vector was used as a negative control. The results are reported as the mean increase (relative to the negative control) Ϯ S.E. (n ϭ 4). The asterisks represent significant differences at the level of p Ͻ 0.01 (*) by one-way analysis of variance and post-hoc Dunnett's test. element represses promoter activity only moderately, even though the identified NRSE element shares more than 70% homology to the NRSE within the SCG10 gene. It has been proposed that the weak NRSE silencer functions as a "modulator" rather than a "dominant switch" for gene expression, and that other elements are likely to be involved in controlling neural-specific expression (52). The GABRA4 promoter does not appear to possess a prominent NRSE element to control neural-specific expression. This observation implies that elements other than NRSE must control the tissue specificity of the promoter. Among other genes encoding neurotransmitter receptors, the promoters for the N-methyl-D-aspartate receptor subunit 1 (NR1) and GABA A -R ␦ subunit (GABRD) genes utilize the brain-specific factors MEF2C (54) and BSF-1 (55), respectively, to stimulate promoter activity in neurons. It is likely that GABRA4 also utilizes neural-specific activators to regulate neural specific expression.
Sequence alignment has revealed that two conserved Sp1 binding sites, Sp1-A and Sp1-B, are critical for GABRA4 expression (Fig. 1). The Sp1 binding site is recognized by the Sp/KLF family members of transcription factors (33)(34)(35). Similar binding affinity to the classic Sp1 site has been observed in the closely related Sp1, Sp3, and Sp4 factors, as well as other subfamily members such as BTEB1, TIEG1, and TIEG2 (33-35, 40, 56). The ubiquitously expressed Sp1 and Sp3 and the relatively brain-specific Sp4 and BTEB1 are transcriptional activators that are able to stimulate transcription, possibly by recruiting TAFs, TFIID, and RNA polymerase to the initiation sites (40,57).
Gel shift analyses of brain-derived nuclear extract suggest that specific protein-DNA complexes are formed at both Sp1 sites. The Sp1-A oligo forms three distinct complexes, one of which includes both Sp3 and Sp4, whereas the identity and function of the other complexes are yet to be identified. A single DNA-protein complex is formed at the Sp1-B site that contains both Sp3 and Sp4, with Sp3 as the dominant factor present in the complex. In contrast, in nuclear extract derived from the non-neuronal HeLa cell line, the bound complexes contained Sp1 and Sp3, with Sp1 as the dominant factor in the complex.
These observations suggest that the GABRA4 promoter may be differentially regulated in neurons and non-neuronal cells, contributing to the neuronal specificity of the promoter. Sp3 has previously been shown to act as a bi-functional transcriptional activator and repressor, depending on promoter context and acetylation state of the protein (59, 60). One possible explanation of the neural specificity of the GABRA4 promoter is that, in non-neuronal cells, the presence of Sp3 at the promoter acts to suppress transcription, whereas in neurons Sp3 interacts with Sp4 to activate transcription. How the Sp/KLF factors regulate neural-specific gene expression is an interesting question and needs to be investigated further.
Using 5Ј-RACE and primer extension assays, we have mapped the transcription initiation sites of GABRA4 to the region between Ϫ184 to Ϫ113 bp relative to the coding ATG ( Figs. 1 and 2). Transcripts with different start sites for GABRA4 are reported in the GenBank TM data base, including AK043806 that originated from Ϫ114 bp and AK013727 that originated from Ϫ499 bp. Our deletion analyses suggest that the initiation region we have identified is able to support high levels of luciferase reporter expression in cultured cortical neurons, whereas the putative initiation site at Ϫ499 bp alone is not able to give rise to reporter gene expression to a comparable level (Fig. 4, C and D). These results suggest that the initiation region reported here includes the major start sites for the GABRA4 gene and is functionally important for high levels of promoter activity.
As has been described for many neuron-specific genes, including genes encoding members of the GABA A -R family and other ligand-gated ion channels, the GABRA4 promoter lacks a TATA box or other classic elements associated with the start of transcription and contain multiple start sites (37,48,49,52,61). Multiple start sites can give rise to mRNAs that differ in their 5Ј-untranslated region, which may potentially influence their post-transcriptional regulation (37,58,61). For example, GluR2 is transcribed from a region Ϫ481 to Ϫ286 bp 5Ј to the coding ATG (52). Because of the presence of a 32-34-bp imperfect GU repeat in the long GluR2 transcripts, the translation efficiency of these mRNA species is greatly inhibited possibly as the result of formation of a highly stable secondary structure within the repeats (58). In the case of GABRA4, several groups have shown that the GABRA4 transcript is present in all regions of the thalamus (23) but the ␣4 protein is only expressed in the dorsal lateral geniculate region (14). These observations suggest that post-transcriptional regulation of GABRA4 may indeed contribute to the diversity of GABRA4 protein expression. Although there are two 16-nucleotide repeats spanning the initiation region of GABRA4, we failed to find a stable stem-loop structures formed around this region. In addition, deletion of these repeats had no effect on promoter activity in vitro. However, we have not assessed the effects of different GABRA4 transcripts on translation efficiency of the gene. Whether the diversity of GABRA4 transcripts is critical for post-transcriptional regulation of GABRA4 expression remains to be further investigated.
In this study, we have analyzed the proximal promoter region of GABRA4 as the first step in unraveling the molecular mechanisms involved in the plasticity of GABRA4 in a number of physiological and pathophysiological conditions. The major conclusions of this report are that GABRA4 is transcribed from different locations within a region Ϫ184 to Ϫ113 bp 5Ј to the coding ATG, and that the transcription factors Sp3 and Sp4 are critical in regulating the activity of the GABRA4 promoter in neurons.