A Silencer Element in the Regulatory Region of Glutamine Synthetase Controls Cell Type-specific Repression of Gene Induction by Glucocorticoids*

Glutamine synthetase is a key enzyme in the recycling of the neurotransmitter glutamate. Expression of this enzyme is regulated by glucocorticoids, which induce a high level of glutamine synthetase in neural but not in various non-neural tissues. This is despite the fact that non-neural cells express functional glucocorticoid receptor molecules capable of inducing other target genes. Sequencing and functional analysis of the upstream region of the glutamine synthetase gene identified, 5′ to the glucocorticoid response element (GRE), a 21-base pair glutamine synthetase silencer element (GSSE), which showed considerable homology with the neural restrictive silencer element NRSE. The GSSE was able to markedly repress the induction of gene transcription by glucocorticoids in non-neural cells and in embryonic neural retina. The repressive activity of the GSSE could be conferred on a heterologous GRE promoter and was orientation- and position-independent with respect to the transcriptional start site, but appeared to depend on a location proximal to the GRE. Gel-shift assays revealed that non-neural cells and cells of early embryonic retina contain a high level of GSSE binding activity and that this level declines progressively with age. Our results suggest that the GSSE might be involved in the restriction of glutamine synthetase induction by glucocorticoids to differentiated neural tissues.

Studies in the chicken neural retina showed that GS expression is regulated by systemic glucocorticoids, which induce in this tissue a very high level of GS by directly stimulating the transcription of the gene (8 -10). The ability of glucocorticoids to induce GS expression is developmentally controlled. Glucocorticoids can induce a high level of GS expression in neural retina at late embryonic ages, but not in early embryonic retina (2,(11)(12)(13). The direct involvement of glucocorticoids in the control of GS gene transcription is evidenced by the nuclear run-on transcription assay, as well as by the finding that the upstream region of the GS gene contains a glucocorticoid response element (GRE) that can bind the glucocorticoid receptor protein and confer responsiveness to glucocorticoids on an attached reporter gene (8 -10, 14, 15). This mode of regulation, which is dependent on the presence of active glucocorticoid receptor molecules, explains the cell type specificity of GS expression in the neural retina; GS expression is always restricted to Mü ller glial cells, which are the only cells in the tissue that express a detectable level of the glucocorticoid receptor protein (1,16). It does not, however, explain why various non-neural cells, which contain functional glucocorticoid receptor molecules capable of inducing other target genes, are not responsive to the hormone and express only low levels of GS.
Neural-specific gene expression is often achieved by the action of negative transcriptional regulation. A prominent example is the neural restrictive silencer element (NRSE/RE-1) found in the regulatory regions of a number of genes expressed exclusively in the nervous system, including SCG10 (17), type II sodium channel (NaII) (18), synapsin I (19), the ␤II subunit of the nicotinic acetylcholine receptor (20), the muscarinic M4 receptor (21,22), and the neural-glial cell adhesion molecule (23). This element binds to a zinc-finger transcriptional repressor protein known as neural restrictive silencer factor (NRSF; Ref. 24) or RE-1-silencing transcription factor (REST; Ref. 25), which is ubiquitously expressed in non-neural tissues and in neuronal precursors of the central nervous system. Experiments involving targeted mutation of the NRSF-encoding gene or inhibition of NRSF by a dominant negative form of the protein indicated that NRSF is required in vivo to repress neural genes in non-neural or in undifferentiated neural tissues (26).
In the present study, we demonstrate that the regulatory region of the GS gene contains a NRSE-like element, and that this element can repress the induction of gene transcription by glucocorticoids. Our results suggest that a neural restrictive silencing activity may be directly involved in the repression of GS induction in non-neural tissues and in the prevention of precocious expression of GS in early embryonic neural retina.

EXPERIMENTAL PROCEDURES
DNA Sequencing and Plasmid Constructions-A set of plasmids containing complementary regions of the GS upstream sequence from ϩ13 to Ϫ2899 was prepared by subcloning the genomic GS clone GS113 * This work was supported by the Tel Aviv University Basic Research Fund and by the Israeli Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF105022.
‡ To whom correspondence and reprint requests should be addressed. Tel.: 972-3-640-7019; Fax: 972-3-640-6834; E-mail: vardi@post.tau.ac.il. 1 The abbreviations used are: GS, glutamine synthetase; GRE, glucocorticoid response element; GSSE, glutamine synthetase silencer element; NRSE, neural restrictive silencer element; NRSF, neural restrictive silencer factor; CAT, chloramphenicol acetyltransferase; PCR, polymerase chain reaction; PMSF, phenylmethylsulfonyl fluoride; DTT, dithiothreitol; bp, base pair()s); kb, kilobase pair(s); DMEM, Dulbecco's modified Eagle's medium; TK, thymidine kinase; RSV, Rous sarcoma virus; REST, RE-1-silencing transcription factor. (11) into the pBluescript SK ϩ vector (Stratagene). The sequence was determined by the chain termination method (27) using the Sequenase II kit (U. S. Biochemical Corp.). Both strands were sequenced using a series of oligonucleotide primers. To construct chloramphenicol acetyltransferase (CAT) vectors that are controlled by the upstream sequence of the GS gene, the coding region of CAT was excised with ClaI from pSV O -CAT (28) and ligated into the ClaI site in the linker region of Bluescript clones containing the following GS upstream sequences: ϩ13 to Ϫ1725 (1.7 GS); ϩ13 to Ϫ2117 (2.1 GS); ϩ13 to Ϫ2375 (2.3 GS); ϩ13 to Ϫ2821 (2.8 GS). The constructs 2.4 GS and 2.5 GS were generated by deletion using the Erase-a-Base System (Promega) as follows; the 2.8 GS construct was first linearized with KpnI, which generates a 3Ј overhang, and with SalI, which generates a 5Ј overhang, both cleaving in the linker region of the vector, 5Ј to the GS sequence, and thus ensuring the correct direction of the deletion process. Linearized DNA was digested with exonuclease III (30 units/g DNA) for up to 6 min. Aliquots were taken at 30-s intervals, added to solution containing S1 nuclease (6 units/g DNA), and incubated for 30 min at room temperature. The series of nuclease-digested DNAs were ligated by T4 ligase and transfected into bacterial cells. Colonies were analyzed by the colony PCR procedure (29), which allows determination of the size of the cloned fragment. The 21-bp sequence in the 2.3 GS construct (2.3 ⌬ 21GS) was deleted by PCR amplification, using oligonucleotides that correspond to the appropriate 5Ј and 3Ј sequences and contain restriction sites for recloning into 2.3 GS. The deletion was confirmed by DNA sequence analysis. The construct GRE-TK was derived from pG46TCO (14). It contains two copies of the GRE sequence linked to the TK-CAT fusion gene. To generate the 258-GRE-TK construct, the 258 fragment (Ϫ2375 to Ϫ2117) was excised from 2.3 GS with ApaI and SacI, bluntended with the Klenow fragment of DNA polymerase I, and ligated into the similarly blunted HindIII and AccI sites of GRE-TK. Orientation was confirmed by restriction enzyme analysis. Complementary oligonucleotides were synthesized (sense strand shown), containing one (AGCTTGAGCACCGCGGTCCTCCAG) or two (GAGCTCTGCACTTG-AGCACCGCGGTCCTCCAGCTTCTGCTCTGCACTTGAGCACCGCG-GTCCTCCAGCTTCT) copies of the 21-bp GSSE sequence or two copies of the NRSE sequence of the SCG10 gene (CAAAGCCATTTCAGCAC-CACGGAGAGTGCCTCTGCCAAAGCCATTTCAGCACCACGGAGAG-TGCCTCTGCAG) (17). Oligonucleotides containing a single copy of the GSSE were constructed such that, when annealed, they contained AccI and HindIII cohesive ends, while oligonucleotides containing two copies of the GSSE or NRSE were constructed such that, when annealed, they contained blunt ends. The construct GSSE-GRE-TK was generated by insertion of the oligonucleotides containing a single copy of the GSSE into HindIII-and AccI-restricted GRE-TK. The oligonucleotides containing two copies of the GSSE were ligated into the blunted HindIII site of GRE-TK to generate GSSE2-GRE-TK and ESSG2-GRE-TK or into SmaI-restricted GRE-TK to generate FGSSE2-GRE-TK. The oligonucleotides containing two copies of the NRSE were ligated into the blunted HindIII site of GRE-TK to generate NRSE2-GRE-TK. Orientation was confirmed by restriction enzymes or DNA sequence analysis. The luciferase reporter gene RSVL(SEL) (30) and the GR expression vector p6RGR (31) are under the transcriptional control of the Rous sarcoma virus (RSV). Plasmid DNA was prepared using the Qiagen plasmid preparation kit.
Tissue and Cell Culture and Transfection Procedures-Retinal tissue was isolated under sterile conditions from eyes of chicken embryos (White Leghorn) at day 10 of embryonic development (E10). The tissue was organ-cultured in Erlenmeyer flasks in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal calf serum on a gyratory shaker at 38°C. Plasmid DNA was transfected into pieces of intact retinal tissue by electroporation using a Bio-Rad Gene Pulser with voltage and capacitance settings of 400 V and 960 microfarads, respectively, as described (32). Following electroporation, retinas transfected in the same cuvette were placed in two Erlenmeyer flasks and cultured for 24 or 48 h in the absence or presence of 0.33 g/ml cortisol (Sigma). HeLa and C6 cells were grown in DMEM supplemented with 10% fetal calf serum. PC-12 cells were grown in DMEM with 8% fetal calf serum and 8% horse serum. The cells were grown in six-well plates and transfected with DNA using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. Following transfection, the cells were incubated for 24 h with or without 0.33 g/ml cortisol.
CAT and Luciferase Assays-CAT and luciferase activities were determined in tissue sonicates. CAT activity was determined as described (28). In all experiments, the CAT assay was adjusted to include an equal amount of luciferase activity originating from co-transfected RSV-L(SEL). CAT activity was determined by scanning of the TLC plates with a PhosphorImager Instrument and TINA version 2.07d software, and calculation of the percentage of substrate (chloramphenicol) converted to the acetylated products. Luciferase activity was assayed as described (30) and recorded by a LKB luminometer (LKB, Rockville, MD).
Preparation of Nuclear Extracts and Gel Mobility Shift Assays-Nuclear extracts were prepared from HeLa, C6, and PC12 monolayer cells at 80% confluence according to Lee et al. (33) with minor modifications. Briefly, cells were harvested using a rubber policeman and washed in 30 volumes of ice-cold phosphate-buffered saline. The packed cell volume was determined by pelleting for 5 min at 1200 ϫ g. Packed cells were resuspended in the equivalent of two packed cell volumes of buffer A (10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM phenylmethylsulfonyl fluoride (PMSF), and 1 mM dithiothreitol (DTT)) and allowed to swell on ice for 15 min. Cells were then lysed by five strokes through a 25-gauge needle or by five strokes of pestle A and five strokes of pestle B in a Dounce homogenizer. The cell homogenate was then centrifuged for 20 s in an Eppendorf microcentrifuge (12,000 ϫ g), and the crude nuclear pellet was resuspended in the equivalent of one packed cell volume of buffer C (20 mM HEPES pH 7.9, 25% glycerol, 0.42 M NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM PMSF, and 1 mM DTT), followed by incubation on ice with stirring for 30 min. The nuclear debris was then pelleted by spinning for 5 min at 12,000 ϫ g, and the supernatant (nuclear extract) was divided into aliquots and quickfrozen by dry ice with ethanol. The extracts were stored at Ϫ70°C. Retinal cell extracts were similarly prepared, except for the following differences. The tissue was lysed by homogenization by means of five strokes of pestle A and five strokes of pestle B in a Dounce homogenizer. The nuclear extract was prepared from the nuclear crude pellet in two steps: the nuclei were suspended in the equivalent of one packed cell volume of low salt buffer C (20 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 25% glycerol, 0.5 mM PMSF, and 1 mM DTT), followed by slow addition of the equivalent of a half packed cell volume of high salt buffer C (20 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 0.6 M KCl, 25% glycerol, 0.5 mM PMSF, and 1 mM DTT). Protein concentration in the nuclear extracts was determined using the Bradford assay (Bio-Rad). Gel mobility shift assays were performed in a 20-l final volume of reaction buffer containing 20 mM HEPES, pH 7.6, 0.1% Nonidet P-40, 10% glycerol, 1 mM DTT, 2.5 mM MgCl 2 , and 1 mg/ml poly(dI-dC). Equal amounts of nuclear extracts (5-10 g) were added to the reaction buffer and preincubated with or without DNA competitors for 10 min on ice. Endlabeled DNA probes (about 2 ϫ 10 4 cpm/reaction) were added, followed by incubation for 15 min at room temperature. The samples were resolved on a 5% nondenaturating acrylamide gel in 0.25 ϫ TBE for 2 h at 70 V. The gels were dried and exposed for 18 h at Ϫ70°C with an intensifying screen. Oligonucleotides used for competition or end-labeled and used as DNA probes were those used for cloning (see above): double-stranded oligonucleotide containing two copies of the GSSE, double-stranded oligonucleotide containing two copies of the NRSE, the sense strand of the oligonucleotide containing two copies of the NRSE (single-stranded DNA probe), and nonspecific double-stranded oligonucleotide derived from a mutated octamer-binding sequence (34).

RESULTS
The 5Ј-Flanking Region of the GS Gene Is Subject to Differential Selective Constraints-The nucleotide sequence of the 5Ј-flanking region of the GS gene (Ϫ1 to Ϫ2657) has been determined by others (15) and by us (Fig. 1A) using genomic clones isolated from two independent libraries of White Leghorn chicken. Comparison of the two sequences revealed a considerable number of sequence differences, including substitutions, insertions, and deletions. The most striking phenomenon, however, was the fact that these sequence changes were not evenly distributed along the DNA, but rather were concentrated in two distinct regions. Of 38 sequence changes, 19 were located in the region between Ϫ868 and Ϫ1700 (region 2 in Fig.  1B) and 17 in the region between Ϫ2287 and Ϫ2657 (region 4 in Fig. 1B). The region proximal to the transcriptional start site, from Ϫ1 to Ϫ868 (region 1 in Fig. 1B), contained only two sequence differences, while in the more distal region, between Ϫ1700 and Ϫ2287, no change was detected (region 3 in Fig.  1B). This dramatic variation in the rate of mutational events among the different DNA regions correlated with two other sequence parameters, attributed to differential selective constraints: GϩC content and the ratio (observed to expected) of the dinucleotide CG (35). As shown in Table I, the DNA regions 2 and 4, which contain a high ratio of sequence changes, were found to have a significantly lower GϩC content and CG dinucleotide ratio than the DNA regions 1 and 3, which contain few sequence changes or none at all. In light of the fact that differential selective constraints can often distinguish between functional and functionless DNA (36,37), this finding raised the possibility that the regulatory elements of the GS gene are predominantly located in regions 1 and 3 and not in regions 2 and 4.
Functional Analysis of the 5Ј-Flanking Region of the GS Gene-In order to learn whether differential selective constraints reflect the location of regulatory elements, we decided to focus on regions 3 and 4, which represent a highly conserved and a highly mutated DNA region, respectively (Fig. 1B). Region 3 has been already shown to contain two functional elements that are directly involved in the responsiveness of the GS gene to glucocorticoids: a glucocorticoid response element (GRE) at Ϫ2086 to Ϫ2100 (14,15) and an AP1/ATF/CRE-like site at Ϫ2110 to Ϫ2117 (15,38). To determine whether the sequence located in region 4 might also contain regulatory functions, we generated a construct containing 2.8 kb (2.8 GS) FIG. 1. Distribution of mutational events along the upstream sequence of the GS gene. A, nucleotide sequence of the 5Ј upstream region of the GS gene. Numbers appearing in the left margin refer to the first nucleotide listed in each line. The first nucleotide upstream of the transcription start site was given the number Ϫ1. The upper line presents the sequence determined in this study. The lower line indicates the differences between this sequence and the sequence of a GS clone isolated from an independent White Leghorn chicken library (43). B, schematic representation of the upstream sequence of the GS gene. The short vertical lines indicate the relative locations of the mutational events. The sequence was divided into four regions according to the density of the mutational events. of 5Ј-flanking region of the GS gene fused to the reporter gene CAT. Constructs of the GS upstream region, which contain (2.1 GS) or lack (1.7 GS) the GRE sequence, were used as a control. The various constructs were transfected into E10 retina and the ability of cortisol to induce CAT expression was assayed (Fig. 2). In all experiments, the measured CAT activity was normalized to the luciferase activity derived from a co-transfected internal control plasmid RSVL(SEL) to correct for differences in transfection efficiency. As expected, the 1.7 GS construct, which does not contain the GRE sequence, was not inducible by cortisol, whereas the 2.1 GS construct, which contains the GRE, was highly inducible. Analysis of the 2.8 GS construct revealed that the sequence further upstream of the GRE contains a putative regulatory element that can markedly repress the hormonal response. To identify this apparent silencing element, we generated a series of constructs containing varying lengths of the upstream sequence and tested their expression. Partial or complete deletion of the 5Ј-most 0.6 kb did not diminish the repressive effect. These data therefore localized the repressive activity to a 258-bp domain lying immediately upstream of the GRE site.
The repressive activity of this 258-bp domain was not restricted to the retinal tissue, but was also exerted in HeLa cells. In these cells, however, inducibility of the GS promoter constructs was dependent on co-transfection of a glucocorticoid receptor expression vector. As in the retina, cortisol induced a marked increase in CAT activity in HeLa cells transfected with the 2.1 GS construct, but not in those transfected with either the 2.8 or the 2.3 GS construct (Fig. 3). By visual inspection of the 258-bp domain, we identified a sequence of 21 nucleotides with considerable homology (13/21) to the consensus sequence of the NRSE (39). This element is found in the regulatory region of several neural-specific genes and represses their expression in non-neural cells. To determine whether this sequence is responsible for the observed repression of the GS promoter activity, we examined whether an internal deletion of the 21-bp sequence, in the context of the 2.3 GS construct, eliminates the silencing activity. Our results clearly demonstrated that deletion of the 21-bp sequence from the 2.3 GS construct (2.3 ⌬ 21GS) resulted in almost complete elimination of the silencing activity in both HeLa (Fig. 3) and embryonic retinal cells (data not shown), indicating that this sequence represents a functional silencer element.
The location of this 21-bp sequence, between Ϫ2176 and Ϫ2196, is within region 3 of the GS upstream sequence, the region which also contains other regulatory elements and in which no mutational events were identified. In contrast, region 4, which contains a high ratio of mutational events, did not show any significant regulatory activity. These findings suggest that the distribution of mutational events along the upstream sequence of the GS gene reflects the location of DNA elements involved in gene control. They also indicate that the regulatory region of the GS gene includes, in addition to positive regulatory elements, a silencing element that can repress the hormonal response.
Effect of the Silencer Element on a Heterologous GRE Promoter-To further characterize the functional properties of the 21-bp GSSE, we examined its ability to confer negative control upon a heterologous glucocorticoid-regulated promoter. The 258-bp domain (258-GRE-TK) or the 21-bp sequence (GSSE-GRE-TK) was fused immediately upstream of a "minimal" GRE-TK promoter, which contains two copies of a synthetic GRE sequence and is linked to the reporter gene CAT (14). The constructs were transfected into HeLa cells, and their activity in the presence or absence of cortisol was assessed. The 258-bp domain and the 21-bp sequence showed little or no effect on basal activity, but both of them greatly repressed the hormonal response. Addition of one more copy of the GSSE sequence (GSSE2-GRE-TK) did not increase the repressive effect. The repressive activity of the GSSE sequences was orientationindependent (ESSG2-GRE-TK), but relocation of the two GSSE copies to a remote site, 2433 nucleotides upstream of the GRE, almost completely eliminated the silencing of the hormonal response.

FIG. 2. Inducibility of the GS constructs in embryonic retinal tissue.
Retinal tissue was transfected with the various GS-CAT constructs (5 g of DNA/ 8 ϫ 10 6 cells) together with the luciferase reporter construct RSVL(SEL) (0.6 g of DNA/8 ϫ 10 6 cells). The transfected cultures were maintained for 48 h in the presence (black bars) or absence (striped bars) of cortisol. CAT and luciferase activities were then examined. The CAT assays were adjusted to include an equal amount of luciferase activity. The percentage of CAT conversion was calculated by scanning of the TLC plates with a PhosphorImager instrument. In each experiment, the value of CAT conversion in the retina transfected with the 2.1 GS construct and induced by cortisol was given the arbitrary number of 100 and used to normalize all other results. The data shown are the means Ϯ S.E. of at least three separate experiments.

TABLE I Number of mutational events, GϩC content, and CG dinucleotide ratio in the different regions of the GS upstream sequence
The number of mutational events in each DNA region (see Fig. 1) was calculated by considering changes of several adjacent nucleotides as one mutational event. GϩC content is the percentage of GϩC bases in the sequence of each DNA region. Observed (obs) CG is the number of the dinucleotide CG in each region. Expected (exp) CG is the expected number of CG in each region, calculated according to Pupko and Grauer (44). The percentage of deviation from expected is calculated according to the formula: 100 ϫ (observed Ϫ expected)/expected. As mentioned above, the GSSE sequence shows considerable homology with the consensus sequence of the NRSE (Fig. 3). However, the 4 -5-fold activation observed upon removal of the GSSE from the regulatory region of the GS gene was relatively weak compared with that observed with NaII and SCG10 genes, in which deletion or mutation of the NRSE resulted in a 16 -40-fold increase of their promoter activities in HeLa cells (17,18). To compare the silencing activity of the GSSE with that of the NRSE, we constructed a NRSE2-GRE-TK plasmid in which two copies of SCG10-NRSE replaced those of the GSSE immediately upstream of the GRE-TK promoter. The results of transfection experiments clearly indicated that, in the context of the GRE-TK promoter, the silencing activities of the NRSE and the GSSE were similar (Fig. 3). These findings suggest that the GSSE is a NRSE-like element, which, when located proximally to the GRE, has little or no effect on basal transcription, but can repress induction by glucocorticoids.
Identification of a GSSE-binding Protein-The silencing activity of the NRSE is mediated by a NRSE-binding protein, known as NRSF (24) or REST (25). This protein is expressed in non-neural cells and silences the expression of NRSE-containing genes. To learn whether the GSSE also interacts with a cellular protein(s), we performed a series of electrophoretic mobility shift assays on nuclear extracts from HeLa cells using the oligonucleotide that contains two copies of the GSSE sequence as a probe (Fig. 4). The specificity of the gel shift was established by a series of competition experiments. Specific competition was obtained with a 50-fold excess of the GSSE, but not with a similar excess of an unrelated oligonucleotide derived from a mutated octamer-binding sequence (34). As expected from the sequence homology and functional similarity, the oligonucleotide that contains two copies of the NRSE sequence (17) could also competitively inhibit the formation of the GSSE complex, although the NRSE competition was less efficacious than that observed with the homologous GSSE. As in the case of the NRSE (17,18), the GSSE binding activity was cell type-specific. This was demonstrated by gel-shift analysis of nuclear extracts prepared from the non-neural cell line HeLa, the neuronal cell line PC12 and the glial cell line C6. GSSE binding activity was detected in HeLa extracts and was competed by excess of GSSE, but not by a similar excess of an

FIG. 4. Nuclear extracts from HeLa cells contain GSSE binding activity.
Gel-shift assays were performed using nuclear extracts from HeLa cells and 32 P-labeled GSSE-containing oligonucleotide as the probe. The GSSE-specific binding complex is indicated by an arrow. Competitors used (in the molar excess indicated) were unlabeled GSSEcontaining oligonucleotide in lanes 2-4, nonspecific (NS) oligonucleotide in lanes 5-7 and NRSE-containing oligonucleotide in lanes 8 to 10. Lane 1 shows binding in the absence of competitor. unrelated oligonucleotide. By contrast, no GSSE binding activity was detected in extracts prepared from PC12 or C6 cells (Fig. 5A). These results were consistent with the repressive activity of the GSSE in these cells. In the context of the GRE-TK construct, the GSSE caused a 6-fold decline in hormonal induction in HeLa cells but only a 1.4-and a 1.7-fold decline in PC12 and C6 cells, respectively (Fig. 5B). The repressive activity in PC12 and C6 cells might be attributable to low levels of NRSF, which are hardly detectable by gel-shift analysis, but whose existence has been demonstrated by Northern blot analysis or upon long exposure of Western blots (18,24,25). Thus, it is possible that the GSSE represses hormonal induction of gene transcription by binding to the silencing factor NRSF. The fact that NRSF is expressed predominantly in non-neural cells may explain why glucocorticoids induce a high level of GS expression in neural tissues only.

GSSE Binding Activity Is Inversely Correlated with GS Expression during Embryonic Development of the Neural Retina-
Expression of GS in the neural retina is developmentally regulated. It is low at early embryonic ages and increases progressively with age. To assess the possible involvement of the GSSE-mediated silencing activity in the developmental control of GS expression, we identified the protein(s) that interact with the GSSE in the chick neural retina and examined its binding activity at different developmental ages. We performed a series of electrophoretic mobility shift assays on nuclear extracts from retinal tissue at day 7 of embryonic development, a stage at which GS expression is very low (Fig. 6A). Specific competition was obtained with a 100-fold excess of the GSSE, but not with a similar excess of an unrelated oligonucleotide. As with HeLa cell extracts, the NRSE sequence could competitively inhibit the formation of the GSSE complex and vice versa (Fig. 6B). The complexes formed by the GSSE or NRSE were similar in size, but had in retinal extracts more retarded mobility than those formed in HeLa extracts (Fig. 6C). These findings support the possibility that GSSE silencing activity is mediated by NRSF and that human and chicken NRSF differ in size. Next we performed mobility shift assays on nuclear extracts from retinal tissues at different developmental ages. As shown in Fig. 7A, GSSE binding activity was high in early embryonic retina and declined progressively with age. No binding activity was detected in adult retina. This is in contrast to single-stranded DNA binding activity, which did not alter with age (Fig. 7B). The pattern of GSSE binding activity correlates inversely with the level of GS expression as evidenced both by GS mRNA accumulation and by GS activity (Fig. 7C). These findings suggest that down-regulation of NRSF might be functionally related to the developmentally dependent increase in GS expression. DISCUSSION GS is a key enzyme in the recycling of the neurotransmitter glutamate, a task that requires a high level of GS expression in neural tissues. However, unlike many neural-specific genes that are expressed exclusively in neural tissues, GS is also a housekeeping enzyme that is expressed at a low basal level in non-neural tissues. Since there appears to be only one copy of the GS gene (2,40), it is reasonable to expect that GS is controlled by a modular regulatory mechanism that allows constitutive expression in most cell types and specific induction in neural tissues.
In an attempt to understand the molecular mechanism underlying GS regulation, we sequenced and functionally analyzed the upstream region of the gene. By comparing the sequences of two independent genomic clones, we could identify a considerable number of mutational events along the DNA sequence. These events were not evenly distributed but rather were concentrated in distinct regions that also differed in their GϩC content and CG dinucleotide ratio (observed to expected). Differences of this type are often found between exons and introns or between coding regions and pseudogenes or intergenic spacer regions, and are attributed to differential selective constraints (35)(36)(37). Functional analysis of the 5Ј upstream region of the GS gene revealed that the DNA region between Ϫ2287 and Ϫ2657, which contained a high ratio of mutational events, did not show any significant regulatory activity, whereas the region between Ϫ1700 and Ϫ2287, in which no mutational events were identified, contained several regulatory elements. These findings suggest that DNA regions involved in the regulation of gene transcription might exert a counteracting effect of selection pressure similar to that of coding regions in the gene. They also suggest that a comparative analysis of sequencing data might allow prediction of the location of putative regulatory elements in the upstream sequence of a given gene.
The regulatory region of the GS gene was found to contain, in addition to positive regulatory elements that confer responsiveness to glucocorticoid, a silencer element that can repress the hormonal response. By performing deletion and addition experiments, we narrowed the location of the silencer to a 21-bp region located 76 bp upstream of the GRE and 2176 bp upstream of the transcription start site. This GSSE sequence is closely related to the silencer element NRSE/RE1, which is present in several neural-specific genes, and represses their expression in undifferentiated neural tissues and in non-neural cells. The following lines of evidence suggested that the GSSE is a putative NRSE sequence that can bind the silencing factor NRSF and mediate repression of gene induction in a cell typespecific manner. 1) The GSSE was found to be capable of repressing hormonal induction of gene transcription both in embryonic retinal tissue and in the non-neural HeLa cells. 2) The repressive activity of the GSSE was not restricted to the GS promoter but could also be conferred on a heterologous GRE promoter (GRE-TK). Replacement of the GSSE sequence, in the context of the GRE-TK promoter, with the sequence of the NRSE of the SCG10 gene resulted in a comparable repression of the hormonal response. 3) Using a gel-shift assay, we were able to identify in both HeLa and embryonic retinal cells a factor that interacts with the GSSE, and to demonstrate that this factor also interacts with the NRSE. 4. Formation of a GSSE-protein complex was cell type-specific; it was detected in the non-neural HeLa cells, but hardly or not at all in the neuronal cell line PC12 or the glioma cell line C6. Consistent with this, GSSE repressive activity in HeLa cells was significantly higher than in PC12 or C6 cells. A similar cell type specificity has also been demonstrated for several other NRSEcontaining promoters (17-19, 21, 25). The ability of the GSSE to repress the hormonal induction of gene transcription and the ubiquitous expression of NRSF in non-neural cell types suggests that the GSSE might be directly involved in the restric-tion of GS induction to neural tissues. It is important to note that the cell type specificity of GS induction can also be controlled by the glucocorticoid receptor itself. Induction of gene expression requires a threshold level of active glucocorticoid receptor molecules; below that level, cells are not responsive to the hormone (41). This is the case in retinal neurons that do not express the receptor molecules, or express them only at low levels (1,16). In these cells GS is not inducible by glucocorticoids, despite the fact that NRSF is not expressed. Glucocorticoid receptor molecules are, however, expressed in many nonneural cell types, and in these cells the repressive activity of the GSSE might be crucial for maintaining a low basal level of GS expression.
The GSSE might also be involved in prevention of the premature expression of GS in early embryonic retina. This is suggested by the finding that the levels of GSSE binding activity and GS expression during embryonic development are inversely correlated. GS expression increases with age, whereas GSSE binding activity declines. While the repressive activity of the GSSE might play an important role in the developmental control of GS expression, it is evidently not the only mechanism involved in this process. We and others have demonstrated that changes in glucocorticoid receptor transcription activity play a role in the developmental control of GS expression (14,32,42). Transcription activity of the glucocorticoid receptor, which is modulated by a mechanism involving the c-Jun protein (32), is low in early embryonic retina and increases progressively with age. It is therefore possible that a low level of transcriptionally active receptor molecules, in conjunction with a high level of NRSF in early embryonic retina, ensures that GS would not be precociously induced.
Several models can be envisaged to explain how NRSF protein might silence the hormonal response. Two general categories of transcriptional repression can be considered. NRSF might silence by binding directly to the initiation complex or indirectly through co-repressor proteins. Alternatively, NRSF might block induction by interacting directly or indirectly with FIG. 6. Nuclear extracts from retinal tissue contain GSSE binding activity. A, gel-shift assays were performed using nuclear extracts from E7 retina and 32 P-labeled GSSE containing oligonucleotide as the probe. Competitors used (in the molar excess indicated) were unlabeled GSSE-containing oligonucleotide (lanes 3-6) and nonspecific (NS) oligonucleotide (lanes 7-10). Lane 1 shows the probe without nuclear extract, and lane 2 shows binding in the absence of competitor. B, gel-shift assays were performed using nuclear extracts from E7 retina and 32 P-labeled GSSE-containing oligonucleotide (lanes 1-9) or 32 P-labeled NRSE-containing oligonucleotide (lanes 10 -18) as the probe. Competitors used were unlabeled GSSE-containing oligonucleotide (lanes 2-4 and 16 -18), nonspecific (NS) oligonucleotide ( lanes 5, 6, 14, and 15) and NRSE-containing oligonucleotide (lanes 7-9 and 11-13). Lanes 1 and 10 show binding in the absence of competitor. C, gel-shift assays were performed using nuclear extracts from E7 retina (lanes 1 and 2) or HeLa cells (lanes 3 and 4) and 32 Plabeled GSSE-containing oligonucleotide (lanes 1 and 3) or 32 P-labeled NRSE-containing oligonucleotide (lanes 2 and 4) as the probe. The GSSE-or NRSE-specific complex is indicated by an arrow. the glucocorticoid receptor protein or by competing for limiting amounts of a co-activator in the cell. In this study we examined the repressive activity of the GSSE in the context of two different promoter constructs. The distance of the GSSE sequence from the GRE site in the GS promoter construct and in the GRE-TK construct was 76 and 34 nucleotides, respectively, while its distance from the transcription start site was 2176 and 236 nucleotides, respectively. Nevertheless, in both constructs the GSSE caused a 5-fold decline in gene induction. This might suggest that the GSSE confers repression of gene induction in a position-independent manner. However, relocation of the GSSE, in the context of the GRE-TK construct, to a position that is 2433 nucleotides upstream of the GRE and 2635 nucleotides upstream of the transcription start site almost completely eliminated the silencing of the hormonal response. Thus, it is possible that repression of the hormonal response is position-independent with respect to the transcription start site, but requires a location proximal to the GRE. Our findings also indicate that in the context of a GRE promoter, the GSSE has little or no effect on basal activity. Under these conditions the repressive activity of the GSSE is restricted to the hormonal response, allowing the basal transcription ma-chinery to express a low level of GS.
The regulatory system of the GS gene might have evolved to meet the specific requirements for a particularly high level of GS expression in glial cells in order to allow recycling of the neurotransmitter glutamate, and a low basal level in most other cells for housekeeping duties. A plausible mode for cell type-specific control of GS expression is presented in Fig. 8. In non-neural cells, which express active glucocorticoid receptor molecules, NRSF represses hormonal induction and thus maintains a low basal level of GS expression. Glial cells express a high level of active glucocorticoid receptor molecules and do not express NRSF; therefore, GS expression in these cells is high. Neurons do not express NRSF, nor do they express the glucocorticoid receptor; therefore, GS expression in neurons is low. Thus, cell type specificity of GS expression might be achieved by utilizing the positive and negative regulatory elements, the GRE and the GSSE, which are not glial-specific by themselves, but which may establish a glial-specific pattern of GS expression through their mutual activity.  32 P-Labeled GSSE-containing oligonucleotide (A) or 32 P-labeled single-stranded oligonucleotide, which contains two copies of the NRSE (B), were used as the probe. C, to quantitate the level of GSSE binding activity at the different developmental ages, autoradiograms were scanned with a PhosphorImager instrument and the relative labeling of the bands was calculated. The level of binding activity in E7 retina was given the arbitrary number of 100. Relative levels of GS mRNA accumulation (9) and GS activity (9,12) at the different developmental ages are presented.

FIG. 8. A schematic model for cell type-specific control of GS expression.
A, in non-neural cells, which express active glucocorticoid receptor molecules, NRSF represses hormonal induction and thus maintains a low basal level of GS expression. B, glial cells express a high level of active glucocorticoid receptor molecules and do not express NRSF; in these cells, GS expression is high. C, neurons that do not express either NRSF or the glucocorticoid receptor express a low level of GS.