Differential Transcriptional Control of the Superoxide Dismutase-2 κB Element in Neurons and Astrocytes*

In addition to their conventional G-C/T target sequences, Sp1 family transcription factors (Sp-factors) can interact with a subset of the target sequences for NFκB. Due to the low level of bona fide NFκB activity in most resting cells, this interaction between Sp-factors and κB-sites could play important roles in cell function. Here we used mutagenesis of a canonical κB element from the immunoglobulin and HIV promoters to identify the GC-rich sequences at each end required for Sp-factor targeting. Through screening of multiple κB elements, a sequence element located in the second intron of superoxide dismutase-2 (SOD2) was identified as a good candidate for both NFκB and Sp-factor binding. In neurons, the prominent proteins interacting with this site were Sp3 and Sp4, whereas Sp1, Sp3, and NFκB were associated with this site in astroglia. The neuronal Sp-factors repressed transcriptional activity through this κB-site. In contrast, astroglial Sp-factors activated promoter activity through the same element. NFκB contributed to control of the SOD2 κB element only in astrocytes. These findings imply that cell-type specificity of transcription in the central nervous system, particularly with regard to κB elements, may include two unique aspects of neurons: 1) a recalcitrant NFκB and 2) the substitution of Sp4 for Sp1.

Transcription factors belonging to the family typified by Sp1 are ubiquitously expressed in mammalian cell types. Sp-factors are involved in the expression of a large number of genes, including most of those known as housekeeping genes; therefore, these factors might participate in every aspect of cellular activity. Indeed, null mutation of Sp1 in mouse is embryonic lethal (1); similarly, mice genetically ablated for Sp3 die soon after birth (2). The abnormalities arising from Sp4 ablation are most apparent in the nervous system (3,4), consistent with the finding that Sp4 expression is highly enriched in neurons (4 -6). Other zinc-finger transcription factors with homology to Sp1, including Sp2, have DNA-binding preferences quite distinct from those for Sp1, -3, and -4 (7). Collectively, these findings indicate that Sp-factors are not redundant in their roles and that none is dispensable. Abundant evidence shows that each Sp-factor possesses discrete functional properties (8). Sp1 is generally considered a transcriptional activator, although an inhibitory region has recently been mapped in its extreme amino terminus (7,9). Sp3 can be an activator or inhibitor dependent on its translation initiation sites, post-translational modifications, and the sequence context of a given cis element (10). Sp4 also exhibits some flexibility with regard to transactivation, but the determinants of its activity remain obscure. The importance of Sp-factors has also become manifest in two forms of human disease. Leaching Sp1 from chromatin by a CUG expansion appears to be a key step for the development of myotonic dystrophy type I (11). Similarly, sequestering Sp-factor activity by mutant huntington protein might be one of the primary cellular events in the pathogenesis of Huntington disease (12,13).
Apart from binding to their consensus GC-or GT-rich sequences, Sp-factors are able to bind to non-canonical sequences, in particular to some B elements (14 -18). In neurons, the interaction between Sp-factors and B elements could be especially important for cell function because of the paucity and/or recalcitrance of NFB activity in neurons (5, 17, 19 -22). The prominent proteins binding to B elements in neurons are Spfactors, and the activity of Sp-factors can be diminished by toxic levels of glutamate (17,21). In mixed neuron-glia cultures, glutamate induces NFB activity in the glia but not in the neurons; NFB is unresponsive to glutamate in pure cultures of glia or neurons (19). For these reasons, and to further elucidate the disparate gene-regulatory mechanisms utilized by neurons and glia, it is important to characterize in neurons the influence of Sp-factors on the transcription of genes that are responsive to NFB in other cell types.
One gene typically controlled by NFB factors and playing an important role for cell survival is superoxide dismutase-2 (SOD2 2 ; also known as manganese SOD) (23). SODs are a group of enzymes engaged to fend off cellular stress initiated by reactive oxygen species. SOD2 is located in the mitochondrial matrix and plays an indispensable role in protecting cells from a * This work was supported by NINDS, National Institutes of Health Grant R01NS046439. 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. 1 To whom correspondence should be addressed: Reynolds Institute on Aging, 629 Jack Stephens Dr., #807, Little Rock, AR 72205. Tel.: 501-526-5811; Fax: 501-526-5830; E-mail: bargerstevenw@uams.edu. 2 The abbreviations used are: SOD2, superoxide dismutase-2; APP, amyloid precursor protein; CMV, cytomegalovirus; COX2, cyclooxygenase-2; EGFP, enhanced green fluorescent protein; EMSA, electrophoretic mobility shift assay; FBS, fetal bovine serum; HIV-LTR, human immunodeficiency virus long terminal repeat; LPS, lipopolysaccharide; MEM, minimal essential medium (Earle's salts); MHC, major histocompatibility complex; ROS, reactive oxygen species; RSI, rat SOD2 intron; TK, thymidine kinase; TNF, tumor necrosis factor; C/EBP, CAAT/enhancer-binding protein.
myriad of insults (24 -27). In many types of cells, reactive oxygen species can activate NFB, which in turn up-regulates the compensatory expression of prosurvival genes. These may include SOD2, because its promoter contains a functional B cis element. Interestingly, the functional B site has been mapped into the second intron of both human and mouse SOD2 genes (23,28,29). This intron also harbors other essential enhancer elements (C/EBP-1, C/EBP-2, and C/EBP-x) besides the B site. Similar gene organization is found in the rat SOD2 gene, where the B site is conserved with a slight variation from human and mouse counterparts (Table 1). Based on our analysis of B sequences required for Sp-factor binding, we speculated that the SOD2-B sequence could be an efficient site for both NFB and Sp-factor binding. We found that the SOD2-B site could be bound by both Sp-factors and NFB in astrocytes, whereas in neurons the prominent binding factors were Sp3 and Sp4. Reporter assays showed that the intron was inhibitory in neurons and the B site was solely responsible for this effect. In astrocytes, the intron acted as an activator driven by NFB and Sp-factors (Sp1 and Sp3).
Cell Cultures-Primary neuronal cultures were established from the neocortices of 18-day Sprague-Dawley rat embryos as described previously (21). Neurons were maintained in Neurobasal medium containing B27 supplement (both from Invitrogen), 0.5 mM L-glutamine, and 10 g/ml gentamycin sulfate. This serum-free medium combined with a transient (first 5 days) exposure to the mitotic inhibitor cytosine arabinoside (3-10 M) achieved cultures that were at least 99% neurons (19). All neuronal cultures used in this study were 8 -10 days in vitro. Astrocyte cultures were established from Sprague-Dawley rat postnatal day 2 and maintained in minimal essential medium with Earle's salts (MEM, Invitrogen) supplemented to 10% with fetal bovine serum (FBS, Invitrogen), as described previously (19). After reaching confluency, the astrocytes were denuded of microglia by shaking and vigorous lavage; the remaining cells had a morphology consistent with type I astro-cytes and were subcultured for experiments by trypsinization. The NT2 (NTera2/D1) human neural-lineage teratocarcinoma cell line was obtained from Stratagene (La Jolla, CA). The T98G human astroglioma cell line was obtained from American Type Culture Collection (Manassas, VA); the N9 mouse microglial cell line was described previously (30). Mammalian cell lines were maintained in MEM plus 10% FBS. NT2 cells were differentiated according to the rapid, neurosphere protocol developed by Cheung et al. (31). Specifically, the NT2 cells were grown in suspension and treated with retinoic acid (10 M) for 14 days, followed by another 5 days of enrichment with arabinoside (1 M) and uridine (10 m) in Neurobasal/B27 medium; most cells were morphologically consistent with mature neurons after enrichment. Drosophila melanogaster SL2 cells were maintained at room temperature in Schneider's insect medium (Sigma) with 10% FBS and 10 g/ml gentamycin sulfate.
Plasmids-pGL3-basic was purchased from Promega. pGL3-RSI was constructed by PCR-cloning rat SOD2 intron 2 (RSI) into MluI and BglII sites of pGL3-basic. Relative to the B site, the cloned sequence included 400 bp in the 5Ј direction (primer: GCC GAC GCG TGC CAA CCA CAA CTT CTG GG; MluI site underlined) and 100 bp in the 3Ј direction (primer: CCG GAG ATC TGT CTC CAC GGA AGG G; BglII site underlined) to the B site, and the clone was verified by sequencing. This sequence contains most of the elements responsive to TNF stimulus (23). GeneEditor TM in vitro Site-directed Mutagenesis System (Promega) was used to mutate the B site (GGG-GAAAAGCCC 3 GGGGAAAAGatC), and the mutation was verified by sequencing. A truncated form of the SV40-promoter (SV40-prom ⌬ (17)), in which six tandem-repeat Sp1 sites were deleted, was derived from pGL3-prom (Promega). The TK promoter was obtained from pRL-TK (Promega). pGL3-RSIprom ⌬ and pGL3-RSImut-prom ⌬ were constructed by inserting SV40-prom ⌬ into the multiple cloning sites (BglII and HindIII sites) of pGL3-RSI and pGL3-RSImut, respectively. pGL3-RSI-TK and pGL3-RSImut-TK were constructed by the same strategy. The p65 and IB␣ coding regions were amplified from NT2 cells with a high fidelity PCR system (cat. no. 1732 641, Roche Applied Science) and cloned in-frame into pEFGP-N1 (p65) or pDsRed2-N1 (IB␣) (BD Biosciences-Clontech, Mountain View CA), creating p65-EGFP and IB-DsRed2 fusion proteins. Clones were verified by sequencing, and p65-EGFP proteins were located in nuclei while IB␣ cotransfection retained all p65-EGFP in cytosol (32). The pRL-CMV (Promega) plasmid expressing Renilla luciferase constitutively was used as a control for transfection efficiency and viability. All plasmids were prepared with Qiagen Midiprep kits, and DNA quality and quantity were determined by both spectrophotometry and visual inspection in agarose gels.
Transient Transfection and Luciferase Activity Assay-Mammalian cells were cultured in 24-well plates and transfected using Lipofectamine 2000 (Invitrogen). Astrocytes were transfected when the cultures reached 60 -70% confluency, whereas neurons were plated and transfected at a density of 1 ϫ 10 5 / cm 2 . One microgram of total DNA per well was used; neuron transfections: 0.5 g of pGL3-basic or pGL3-prom ⌬ series, 0.1 g of pGL3-TK series, and 0.1 g of pRL-CMV; astrocyte transfections: 0.6 g of pGL3-TK series and 0.1 g of pRL-CMV; pBluescript II (Stratagene) was used to adjust the total to 1 g. When decoy oligonucleotides (50 nM final concentration) were used, pBluescript II was omitted. For p65 and IB␣ co-transfection, 0.2 g (per well) of plasmid encoding these proteins was used. Transfections took place in the existing media (for neurons, Neurobasal with B27 supplement; for astrocytes, MEM plus 10% FBS). After 60 -90 min of incubation, DNA and Lipofectamine were washed off, and fresh maintenance medium was applied. Cells were harvested ϳ24 h after transfection for luciferase assay. Dual-luciferase reporter assay system (Promega) was used to detect both firefly and Renilla luciferase activities through a Veritas luminometer (Promega). All experiments were repeated at least five times. SL2 cells were transfected by pPACSp1 and pPACSp3 with a calcium phosphate method, which has been described previously (17).
Nuclear Protein Extraction and EMSA-The detailed protocols have been described previously (17,19). N9 cells were activated with 10 nM LPS for 60 min, and primary rat astrocytes were treated with TNF (100 ng/ml) for 60 min prior to nuclear protein extraction.
Chromatin Immunoprecipitation Assay-Cells were fixed with 1% formaldehyde in MEM for 10 min at room temperature, followed by two washes with phosphate-buffered saline. Cells were scraped and harvested in phosphate-buffered saline; this phosphate-buffered saline and all subsequent buffers contained the following protease inhibitors: phenylmethylsulfonyl fluoride (phenylmethylsulfonyl fluoride, 0.5 mM), leupeptin (10 g/ml), aprotinin (10 g/ml), and calpain inhibitor N-acetyl-Leu-Leu-norleucinal (1 M). After centrifugation, cell pellets were dissolved in SDS lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1) and sonicated. The sonication conditions were optimized for each of the two cell types used. Chromatin was reproducibly reduced to fragments of 200 -1000 bp by 8 and 12 sonication pulses in NT2 and T98G cells, respectively. Each pulse lasted 10 -15 s, and the energy output was set at 25% of sonicator maximum (F60 Sonic Dismembrator, Fisher Scientific). Cell debris was removed by centrifugation at 16,000 ϫ g for 10 min at 4°C. Subsequent steps utilized buffers supplied in a kit by Upstate Biotech (Lake Placid, NY), essentially following the manufacturer's instructions. The cleared supernatants were diluted 10-fold with chromatin immunoprecipitation dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl, 16.7 mM Tris-HCl, pH 8.1). The solution was precleared with a slurry of protein A/G-agarose (Invitrogen) in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) containing 200 g/ml sheared salmon-sperm DNA (Roche Molecular Biochemicals GmbH, Germany) and 1 mg/ml bovine serum albumin, incubating for 60 min at 4°C with rotation, followed by removal of the A/G-agarose by centrifugation. The precleared solution was incubated with 2 g of primary antibody at 4°C overnight. Antibody complexes were recovered with protein A/G-agarose in salmon sperm DNA and bovine serum albumin; the agarose pellets were then washed with 1 ml of the following buffers in order: once with low salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, 150 mM NaCl, pH 8.1); once with high salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, 500 mM NaCl, pH 8.1); once with LiCl wash buffer (250 mM LiCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, pH 8.1); three times with TE buffer, pH 8.0. DNA-protein complexes were eluted from the agarose pellets with two sequential aliquots of 250 l of elution buffer (1% SDS, 0.1 M NaHCO 3 ). The eluted samples were combined, and protein-DNA crosslinking was reversed by incubation in 200 mM NaCl at 65°C for 4 h, followed by proteinase K (10 g/ml) digestion at 45°C for 1 h. The chromatin immunoprecipitation DNA was extracted once with phenol/chloroform and once with chloroform and then precipitated in ethanol. Purified DNA was dissolved in 30 l of H 2 O. One microliter of DNA was amplified (56°C for annealing) in a 25-l PCR reaction (forward primer: 5Ј-CGA ACC TTG AAT TAC GGG AAA; reverse primer: 5Ј-CCT GGT GTC AGA TGT TGC CT). PCR products (130 bp) were fractionated and visualized on a 1.2% agarose gel.
Reverse Transcription-PCR-The general protocol has been described elsewhere (17). Specific primers and PCR conditions were as follows: SOD2 upstream primer: 5Ј-CTG GCC AAG GGA GAT GT-3Ј; downstream primer: 5Ј-GGC CTG TGG TTC CTT GC-3Ј (27 cycles, producing a 309-bp product). ␤-Actin upstream primer: 5Ј-GTC CTC TGC CAT GTG GTT TTC-3Ј; downstream primer: 5Ј-GCT GCG CTC TCG TAA TTG TG-3Ј (21 cycles, producing a 439-bp product). Annealing temperature for all reactions was 60°C. The cycle numbers and other conditions were optimized to achieve linear rates of amplification. All data depicted are representative of at least three experiments.

Sequence Requirements for Sp-factors Binding to B Sites-
We previously determined that three different classes of transcription factor (Sp-factors, NFB, and RBP-J) can bind to the Ig/HIV-B site (17), a sequence widely used as a probe in "NFB" EMSA analyses. RBP-J binding is conferred by a sequence located at the 3Ј half of this site (5Ј-CCTGGGAAA-3Ј in the complementary strand) and requires additional 3Ј flanking sequences outside the core B element. All these interactions between transcription factors and the Ig/HIV-B sequence appeared to be sequence-specific (17). We extended this investigation to establish the efficiency of the interaction between Sp-factors and this B site. Nuclear extract containing Sp-factors, NFB, and RBP-J was prepared from LPS-activated N9 microglial cells; the Sp-factors in these cells consist primarily of Sp1 and Sp3 (5, 6). EMSA analysis was performed using the Ig/HIV-B sequence as a 32 P-labeled probe and various sequences as unlabeled ("cold") oligonucleotides in competition assays. When Ig/HIV-B was used as an autologous competitor, binding was diminished for each of the three classes of transcription factors; but as a fraction of its baseline value, binding of the Sp-factors was the least sensitive among the three (Fig. 1A). In contrast, a canonical Sp1 target sequence competed much more efficiently with the B sequence for binding to Sp-factors; this sequence had no effect on the binding activity of NFB or RBP-J. No appreciable competition for any of the three transcription factors was observed when unrelated target sequences for transcription factors AP-1 or Oct1 was used. A similar competition pattern was observed when the Sp1 target sequence was used as a radiolabeled probe (data not shown). These results suggest that the affinity of Sp-factors for the Sp1 target sequence is ϳ30-fold greater than that for the Ig/HIV-B sequence. However, a precise quantitation of this difference is confounded by a third-order competition for probe between the Sp-factors and the other transcription factors in these extracts, the quantities of which undoubtedly vary with cellular conditions. Sp-factors bind to a subset of B elements (14), so it is important to determine the sequence requirements for this interaction. We made such determinations within the context of the widely used Ig/HIV-B sequence. The 5Ј-end of the sequence contains four adjacent guanines (G4 motif). Each guanine in the G4 motif was individually mutated to cytosine (G 3 C) (B-mut1-4, Table 1). To evaluate the binding efficiency, the mutated oligonucleotides were used as unlabeled competitors for Sp-factor binding to the 32 P-labeled Ig/HIV-B sequence. Again, activated N9 nuclear extracts were used to assay Spfactors and NFB in the same EMSA analysis. Mutation of any of the four guanines abolished Sp-factor binding, because substantial competition was not observed from any of the mutated sequences (Fig. 1B). In contrast, none of the mutations interfered with RBP-J binding, because the 3Ј-half of Ig/HIV-B sequence was unchanged. The third and fourth guanines were required for NFB binding, as expected from their location within the B core sequence. But mutation of the first two guanines did not affect NFB binding. The first guanine was predicted to be dispensable, because it lies outside the B core sequence. NFB binding after mutation of the second guanine can be explained by the retention of a B site (GGGAAAGTCC) in the complementary strand.
To test the sequence requirements of the 3Ј-end of the Ig/HIV-B site for Sp1 binding, we made another mutation (B ⌬C , Table 1) removing the distal cytosine from the three cytosines (C3 motif) present there. When B ⌬C was used as a labeled probe, NFB was the only prominent band, and little Sp1 binding could be detected (Fig. 1B). In competition assays, B ⌬C could compete for NFB binding with the same efficiency as the intact Ig/HIV-B sequence (Fig. 1B), whereas B ⌬C was not an efficient competitor for RBP-J or Sp-factors binding to the original Ig/HIV-B probe. These data indicated that both the G4 motif and the C3 motif were essential for Sp-factors binding to the Ig/HIV-B probe.
Together with our earlier work (19), the experiments presented here indicate that Sp-factor binding to the Ig/HIV-B site requires four guanines at the 5Ј-end and three cytosines at the 3Ј-end of the same strand. For example, the IL6-B site has three guanines at 5Ј, which makes it a poor site for Sp1 binding (17). Generally, the shorter the spacing between 5Ј guanines and 3Ј cytosines the more favorable is Sp1 binding; a heavy GC content in the spacer and flanking regions increased Sp1 binding efficiency to a B site (data not shown).
Assessment of Sp-factor Binding to Gene-specific B Sites-Based on the sequence requirements noted above, we identified promising B sites from other genes, focusing on genes with neuronal expression. These B enhancer sequences were tested for binding by Sp-factors. Sequences were selected based on a high content of guanines and cytosines at respective ends with a low GC content between (the spacer region). The human amyloid precursor protein (APP) promoter contains two elements consisting of the same B sequence (APPB1) repeated; induction of APP by stimuli such as interleukin-1 reportedly depends on one or both of these elements (33). We found another can-  Table 1) were applied as cold competitors for the binding to radiolabeled Ig/HIV-B probe. In the right panel, Ig/HIV-B ⌬ C was used as the radiolabeled probe for tests of direct binding (lane 2) or it was applied as a cold competitor of radiolabeled Ig/HIV-B (lanes 3-5). In both panels, cold competitors were applied at 3-, 10-, or 30-fold molar excess.

TABLE 1 B-oligonucleotides used in EMSA studies
The core target for NFB is underlined.
Only Sp-factors could be detected when APPB2 site was used as a labeled probe in EMSA (data not shown). Of the remaining B oligonucleotides, IL6-B was unable to compete efficiently for Sp-factor binding, whereas APP-B1 interacted with Spfactors inefficiently (17). These data were consistent with our Ig/HIV-B mutation results; specifically, the IL6-B sequence contains only three guanines on the 5Ј-end, and the APP-B1 sequence contains only two cytosines on the 3Ј-end. The COX2 and MHC class I gene sequences are identical in their core B sequence, and neither could compete for Sp-factors efficiently.
On the other hand, Bcl-x and rat SOD2 sequences indicated substantial Sp-factor binding. We labeled some of the oligonucleotides with 32 P to test them in direct binding assays (Fig. 2B). Interestingly, although the Bcl-x-B sequence could compete with Ig/HIV-B for both Sp-factors and NFB, its prominent binding factors were of the Sp-family (Fig. 2B), presumably because of the high GC content of the Bcl-x sequence. In contrast, COX2-B and MHC I-B were effective targets for NFB but showed very poor Sp-factor binding. As competitors, the Sp-binding sequence 50-14 (19) and the NFB-specific IL6-B sequence had essentially reciprocal effects on active binding. It is noteworthy that the Bclx-B sequence also showed substantial binding by Sp3 ⌬ , an alternative translation product of the Sp3 mRNA that can be detected on a Sp1 consensus probe (e.g. Ref. 17).
We also tested recombinant Sp1 or Sp3 produced in SL2 cells. In a direct-binding EMSA, both proteins bound the Bclx-B sequence but not COX2-B or MHC I-B sequences (Fig.  2C), confirming that Sp-factors could not interact with COX2-B and MHC I-B efficiently. These data indicated that Bcl-x-B was an Sp-favored site, whereas COX2-B and MHC I-B were mainly NFB binding sites.
Factors Binding to SOD2-B Site in Neurons and Glia-The SOD2-B site resides in the second intron of the gene and is reportedly essential for induction of SOD2 by multiple stimuli in human and mouse (23,42). The sequence of this site is highly conserved in human, rat, and mouse (Table 1). Although these differ slightly in the spacer region, all species contain a G4 motif at the 5Ј-end and a C3 motif at the 3Ј-end. Also, the low GC content in the spacer region suggests that the SOD2-B site could be a target sequence for both Sp-factors and NFB. Indeed, consistent with the competition assays, both Sp-factors and NFB were able to bind to the SOD2-B site (Fig. 3A). Two SOD2-B-binding components could be detected in astrocytes that were not present in neurons, and these astrocytic complexes were further enhanced by TNF treatment (Fig. 3B), suggesting these two bands were derived from Rel-family proteins related to NFB. Consistent with our earlier results (17), the prominent neuronal B-binding factors were Sp-factors that had activity diminished by toxic levels of glutamate when SOD2-B, Ig/HIV-B, and Sp1 probes were surveyed (Fig. 3B).
We expanded our investigation of the SOD2 B site to determine the effect of Sp-factors on the activity of the enhancer. First, we attempted to identify those factors binding to the SOD2-B site in neurons and astrocytes. We focused on those bands with migration rates similar to known Sp-factors and NFB (Complexes I-III in Figs. 3 and 4). We recently determined that Sp-factors are differentially expressed in neurons and glia: Sp3 is expressed in both glia and neurons, Sp1 is only detectable in glia, and Sp4 is highly enriched in neurons (5, 6). These distinct expression patterns might have important consequences for cell function. Therefore, it is of interest to determine which of these Sp-factors and NFB dimers bind to the SOD2-B site in various cell types.
Nuclear proteins from TNF-activated astrocytes or untreated cortical neurons were used in supershift assays. The top  Table 1. B, BclX-B, COX2-B, and MHCI-B sequences were radiolabeled to directly probe the binding activities in activated N9 nuclear extracts. The indicated cold oligonucleotides were added into some reactions to compete with these probes as a test of binding specificity. Cold oligonucleotides were at a 30-fold molar excess compared with probes. In the last lane, an Ig/HIV-B probe was used as a positive control. C, Sp1 and Sp3 expressed in SL2 cells were used to show the interaction with different B probes.
band (Complex I, Fig. 4) was sensitive to Sp1 and Sp3 antibodies in astrocytes, whereas it was removed by Sp3 and Sp4 antibodies in neurons. Complex II, only present in astrocytes, was removed by antibodies against p50 or p65, indicating a p50/p65 heterodimer. Another astrocyte-specific species, Complex III, was sensitive to p50 antibodies but unaffected by p65 antibody, suggesting that it comprised the p50 homodimer. Both Relfamily dimers were expressed in astrocytes under basal condition, and the level of p50/p50 was substantially higher than that of p50/p65 (Fig. 3B). After TNF stimulation, the p50/p65 level surpassed that of p50/p50 (Fig. 3B). Both astrocytes and neurons presented somewhat abundant, fast-migrating complexes (arrowheads, Fig. 4), possibly produced by proteins binding to the flanking sequences. The levels of these uncharacterized complexes in astrocytes were diminished after activation of NFB by TNF, suggesting that binding of the responsible proteins to the probe was competed by active NFB, as seen with RBP-J at the B element in the interleukin-6 promoter (43).
To confirm that Sp-factors occupy the B element from SOD2 intron 2 in situ, we performed chromatin immunoprecipitation assays. Because the available Sp-factor antibodies were generated against human proteins, it was necessary to utilize human cells for efficient immunoprecipitation. Cells of the NT2 line were differentiated into neurons (31); the T98G astroglioma line was used as an astrocyte model and was treated with TNF to activate NFB. NT2 cells showed prominent localization of Sp3 and Sp4 to a 130-bp sequence surrounding the B element (Fig. 5). In T98G cells, by contrast, the p65 and p50 components of NFB were associated with this region more prominently. T98G cells also showed a paucity of Sp4 binding, consistent with data from primary astrocytes.
The SOD2 Intronic B/Sp Site Is Inhibitory in Neurons-The B site in the RSI region was explored further for modulation by Sp-factors. The rat SOD2 intron (RSI), including ϳ400 bp upstream and 100 bp downstream from the B site, was cloned into two different reporter vectors with different promoter activities (see "Experimental Procedures"); a version in which the B site was mutated was inserted in the same vectors. Reporter pGL3-prom ⌬ has a very weak promoter, whereas pGL3-TK has a strong thymidine kinase promoter. When these reporters were transfected into primary neurons, their basal activities differed (Fig. 6, A and B). However, they both showed a similar pattern of inhibition by the RSI sequence and release from this inhibition by mutation of the B site, indicating that the effect of the RIS was not dependent upon the basal promoter with which it was coordinated. Similar results were observed in the NT2 line (Fig. 6C). Although this repressive activity of the RSI in neuronal cells was surprising, the sequence had its expected stimulatory effect in rat primary astrocytes   Cultures were then fixed in formaldehyde and processed for immunoprecipitation with antibody to p50, p65, Sp1, Sp3, or Sp4. A negative-control precipitation was also performed with protein A/G-agarose alone. Co-precipitating DNA was subjected to PCR with primers directed against the B region in the SOD2 second intron. From the blank precipitation, the extract from the boiled pellet (no Ab) and the initial supernatant (input) were also subjected to PCR as negative and positive controls, respectively.
( Fig. 6D). Mutation of the B site dramatically reduced the activation, although the remaining activity was significantly higher than promoter alone. This result is consistent with multiple sequences in the intron enhancing promoter activity in astrocytes, a substantial portion of which is mediated by the B site. Together, these data indicate differential regulation of the SOD2 intronic element in neurons versus non-neuronal cells.
It was important to confirm that the effects seen in reporter constructs were relevant to the actual SOD2 gene. Therefore, we performed reverse transcription-PCR to measure SOD2 mRNA levels in neurons treated with a double-stranded oligonucleotide containing a Sp-factor target sequence ("decoy"); an oligonucleotide of a scrambled sequence ("scram-con") was applied as a control for the Sp decoy. When cortical neurons were exposed to Sp decoy for 20 h, steady-state levels of SOD2 mRNA appeared to rise (Fig. 7). No such change occurred with application of the scrambled control oligonucleotide.
Rel-and Sp-factors Differentially Regulate the SOD2 Intronic B Site-To explore the role of individual transcription factors in regulating the activity of the SOD2 intronic site, we used the Sp-factor decoy oligonucleotide and a plasmid encoding IB␣ to repress the activities of Sp-factors and NFB, respectively. In primary neurons, Sp1 decoy completely alleviated the inhibitory effect of RSI (Fig. 8A). In contrast, scram-con appeared to be ineffective (p Ͻ 0.01 for RSI-TK versus RSI-TK/decoy; p Ͻ 0.01 for RSI-TK/decoy versus RSI-TK/scram-con; p Ͼ 0.1 for comparisons between the results in any two decoy-treated conditions). Co-transfection of a plasmid encoding IB␣ had no effect on reporter activities (p Ͼ 0.1, RSI-TK versus RSI-TK/ IB␣; p Ͼ 0.1, RSImut-TK versus RSImut-TK/IB␣). We also co-transfected a plasmid encoding p65 with reporters. Such overexpression of p65 significantly enhanced RSI-TK promoter activity in neurons (p Ͻ 0.001, RSI-TK versus RSI-TK/p65).
Surprisingly, p65 still increased expression moderately when the RSI was mutated (p Ͻ 0.01, RSImut-TK versus RSImut-TK/ p65), suggesting that p65 might contribute to SOD2 intron-dependent transcription through a cryptic B site or by activating other factors. The influence of p65 was completely blocked by IB␣ co-transfection, consistent with our previous observation (32). Application of Sp1 decoy oligonucleotides and IB␣ to primary astrocytes yielded results very different from those observed in neurons (Fig. 8B). Sp1 decoy significantly reduced the activity of RSI-TK (p Ͻ 0.01 for RSI-TK versus RSI-TK/ decoy); the scram-con was ineffective (p Ͼ 0.1, RSI-TK versus RSI-TK/scram-con). This result indicates that the Sp-factors in astrocytes (Sp1 and Sp3) act as activators on the B site. Cotransfection of IB␣ suppressed RSI-dependent transcription more effectively than Sp1 decoy (p Ͻ 0.005 for RSI-TK versus RSI-TK/IB␣; p Ͻ 0.01 for RSI-TK/decoy versus RSI-TK/ IB␣). Interestingly, IB␣ also significantly reduced expression when the RSI was mutated (p Ͻ 0.01 for RSImut-TK versus RSImut-TK/IB␣), consistent with the ability of p65 to activate the mutated intron in neurons. A combination of Sp1 decoy and IB␣ co-transfection completely removed the activity derived from the SOD2 intron (p Ͻ 0.005, RSI-TK/IB␣ versus RSI-TK/ decoy/IB␣; p Ͼ 0.05 for RSI-TK/decoy/IB␣ versus RSImut-TK/decoy/IB␣). Application of scram-con along with IB␣ did not significantly alter the activity from transfection with IB␣ alone. These data indicate that the dominant activator for the SOD2-B in astrocytes was NFB, whereas Sp1 and Sp3 also significantly enhanced the intron activity through the same site.

DISCUSSION
The transcriptional induction of SOD2 by NFB in some cell types has been invoked in support of the broader hypothesis that this transcription factor participates in responses to oxidative stress. Our results indicate that the cis element most important to NFB-dependent inductions of SOD2 is capable of interacting with either NFB or Sp-factors, to a degree dependent on cell type-specific determinants. More surprisingly, this B element participates in the suppression of SOD2 in CNS neurons, as a result of the specific Sp-factors they express and the recalcitrance of their NFB. We have further delineated the sequence requirements for Sp-factors binding to B-sites; low GC content in the spacer sequences seems to be required. However, a very high overall GC content in the sequence will favor Sp-factor over NFB, as seen in the APP2 sequence (and less dramatically in the Bcl-x-B sequence), effectively creating a Sp-dependent element from a sequence that might otherwise share binding and regulation by NFB. Conversely, low GC content discourages the interaction with Sp-factors, as seen in the IL-6 sequence. Hence, DNA sequences exist with overlapping binding affinities for NFB and Sp-factors, and this relationship can lead to surprising regulatory consequences.
Sp-factors are abundantly expressed and constitutively active in most cell types, and NFB activity is undetectable in many cells under basal conditions. In addition, Sp1 is not sensitive to CpG methylation but NFB is. Therefore, the interaction between Sp-factors and B sites could play an important role in homeostatic expression of certain genes through their B element. The situation in CNS neurons appears to be an extreme example. Evidence indicates that DNA-binding activity from bona fide NFB is severely restricted in CNS neurons (17,19,20,32) (Fig. 3B), making the role of Sp-factors in these cells much more critical. In contrast to neurons, astrocytes were found to exhibit varying degrees of NFB activity under basal and TNF-stimulated conditions. Adding to this contrast between cell types, we have recently determined that Sp-factors are differentially expressed in neurons and astrocytes (5,6). Sp1 is much more highly expressed in glial cells than in neurons, whereas Sp4 is conversely enriched in neurons.
The above relationships of transcription factors to specific cell types appear to create remarkable consequences for expression of SOD2. In neurons, the prominent proteins interacting with SOD2-B site were Sp3 and Sp4. Reporter assays indicated these neuronal Sp-factors repressed promoter activity through the intronic B site, because the inhibition was relieved by either mutation of the B site or by application of Sp-decoy. Co-transfection of IB␣ suppressed the RSI promoter activity in astrocytes (but not in neurons), permitting detection and manipulation of residual activity that resulted from the Sp1 and/or Sp3 expressed abundantly in astrocytes. The role of Sp3 notwithstanding, this contrast is consistent with prior observations made by ourselves (17) and others (44 -46) with regard to the transactivating potential of Sp1 and Sp4. The latter appears to act more commonly as a transrepressor, although its behavior is not completely understood and may depend on the context of the binding sequence. In a separate but analogous situation, mutation of the B element in the RSI left a residual promoter activity in astrocytes that was still sensitive to IB␣. This is consistent with earlier reports analyzing promoter activity from intron-2 of the mouse and human SOD2 genes, wherein the identified B site was shown to be only partially responsible for inductions by TNF and interleukin-1 (23,28,42).
Several Sp-binding B sites have been reported so far, and consistent with our data, all these sequences have high GC content. For the classic Ig/HIV-B site, the affinity for Sp-factors was ϳ30-fold lower than that of NFB (p50/p65). The four Gs at the 5Ј-end and three Cs at the 3Ј-end were all required for Sp-factors to interact with this site. Our data are largely consistent with an earlier report using HeLa cell extracts (14). However, that study found the IL6-B sequence to be a good target for Sp1, but we were unable to detect a substantial inter- FIGURE 8. Sp-factors are inhibitory for RSI activity in neurons but stimulatory in astrocytes. pGL3-TK, pGL3-RSI-TK, and pGL3-RSImut-TK were transfected in primary rat neurons (A) and astrocytes (B). Sp-factor decoy oligonucleotides (50 nM, final concentration) and IB␣ were used to suppress the activity of Sp-factors and NFB, respectively. In each set of transfection, the activity (ratio between firefly luciferase activity and Renilla luciferase activity) from pGL3-TK was arbitrarily set as 1.0. pCMV-RL was co-transfected to monitor the transfection efficiency. See text for statistical data. action between Sp1 and the IL6-B in either direct binding assays or competition assays; it is unclear what causes this discrepancy. Sp1 interacts with a B site (GGGGGTGACCCC) located in the P-selectin promoter, where it confers constitutive basal activity (14). In that system, Sp-binding activity can be replaced by higher affinity NFB (p50/p50) factors to turn off the promoter activity. A similar competitive situation has been observed in the human Fas promoter (15), where Sp1 is bound to a B site (GGGCGTTCCC) under unstimulated conditions then replaced by NFB after activation. Our results from primary neurons are analogous to these systems in that transfected p65 was able to relieve repression caused by Sp-factors.
By contrast to these instances of physical and functional competition between NFB and Sp-factors, there are other systems in which the two classes of factors appear to function simultaneously in the same cell or population of cells. For instance, a body of data assembled by Liu et al. (18) document Sp-binding at a B site (GGGACTGGCC) in the mouse N-methyl-D-aspartic acid receptor 1 promoter in the P19 neuroblast cell line. Although NFB activity is present in those cells, Sp-factors drive most of the activity of the N-methyl-Daspartic acid receptor 1 B element therein. Interestingly, P19 cells express the Sp-factors (Sp1/Sp3) we observed in astrocytes. And it was in astrocytes that we found NFB and Spfactors to co-exist and both transactivate the RSI. Consistent with those functional data, EMSA indicated that TNF-mediated activation of NFB in astrocytes did not alter Sp-factor binding to the SOD2-B.
We found the repressive effect of neuronal Sp-factors to be promoter-specific (Fig. 7). This is consistent with the identification of the neuronal Sp-factors as Sp3 and Sp4, because these factors are more equivocal in their tendency to activate or repress than is Sp1. For instance, Ahlgren et al. (47) found that Sp3 inhibits transcription from a G-C/T box in an artificial reporter construct but not one in the promoter for cytochrome P450. We previously showed that two B sites in the human APP promoter can interact with Sp-factors (17). No effect on reporter gene expression was detected when these two sites were mutated within the context of the entire promoter. 3 Also, when APP-B2 sites were cloned into pGL3-prom ⌬ , these sites did not repress expression in primary neurons. Thus, binding of an Sp-factor to a DNA sequence is not tantamount to a transcriptional effect, and this is true even for repression.
Oxidative stress typically induces SOD2 expression, and excitotoxic levels of glutamate produce superoxide and other reactive oxygen species (48,49). We previously documented a rapid attenuation of the DNA-binding and transactivation activity of neuronal Sp-factors upon challenge of cerebral neurons with glutamate (17,21). Severe cellular oxidation after initiation of the Fenton reaction in neurons attenuates Sp-factor activity, as well (21). Thus, under conditions of stress, the tonic repression of the SOD2 gene by Sp3 and Sp4 may be alleviated by removal of these factors. We recently found that most of the effect of glutamate is explained by the activation of calpain, which converts Sp3 and Sp4 into fragments that retain DNA-binding activity (6). Although these fragments lack the transactivation domain, we and others (50) have determined that the zinc-finger domain that remains after proteolysis of Sp1 can serve as a transcriptional inducer, perhaps functioning as a cofactor of sorts. Therefore, it cannot be presumed that transcriptional repression by Sp-factors is necessarily attenuated by this limited proteolysis. Nevertheless, glutamate causes a substantial reduction of total DNA binding by Sp-factors and their fragments (as does oxidation), suggesting that the repression of SOD2 expression by Sp-factors is indeed relieved under these conditions.
In conclusion, we have demonstrated for the first time an inhibitory effect of Sp-factors mediated through a B site. This effect is a direct result of the transcription factor milieu in neurons, particularly the relative dominance of Sp4 over Sp1 and the paucity of NFB activity. The same B site was regulated in an opposite fashion in astrocytes, which contained Sp1 and active NFB. Transfection of Sp1 was unable to reverse the repressive effect of the RSI in neurons. 3 This is consistent with the dominant-repressive effect that Sp4 had over Sp1 in our earlier experiments (17). Further, other laboratories have found that Sp4 and/or Sp3 can exert a dominant-repressive effect even when Sp1 is present (44,51). In addition to providing insights about the regulation of SOD2, these data bolster the conclusion supported by several other approaches (5, 32) that NFB is relatively unimportant for transcription in cerebral neurons.