cAMP Response Element-binding Protein (CREB) and Nuclear Factor κB Mediate the Tamoxifen-induced Up-regulation of Glutamate Transporter 1 (GLT-1) in Rat Astrocytes*

Background: Tamoxifen (TX), a selective estrogen receptor modulator, enhances glutamate transporter (GLT-1) expression in astrocytes. Results: TX up-regulated GLT-1 expression via the CREB and NF-κB pathways. Conclusion: TX enhanced GLT-1 expression at the transcriptional level. Significance: Understanding the mechanisms of TX action on GLT-1 will contribute to the development of neuroprotectants against excitotoxicity. Tamoxifen (TX), a selective estrogen receptor modulator, exerts antagonistic effects on breast tissue and is used to treat breast cancer. Recent evidence also suggests that it may act as an agonist in brain tissue. We reported previously that TX enhanced the expression and function of glutamate transporter 1 (GLT-1) in rat astrocytes, an effect that was mediated by TGF-α. To gain further insight into the mechanisms that mediate TX-induced up-regulation of GLT-1 (EAAT2 in humans), we investigated its effect on GLT-1 at the transcriptional level. TX phosphorylated the cAMP response element-binding protein (CREB) and recruited CREB to the GLT-1 promoter consensus site. The effect of TX on astrocytic GLT-1 was attenuated by the inhibition of PKA, the upstream activator of the CREB pathway. In addition, the effect of TX on GLT-1 promoter activity was abolished by the inhibition of the NF-κB pathway. Furthermore, TX recruited the NF-κB subunits p65 and p50 to the NF-κB binding domain of the GLT-1 promoter. Mutation of NF-κB (triple, −583/-282/-251) or CRE (-308) sites on the GLT-1 promoter led to significant repression of the promoter activity, but neither mutant completely abolished the TX-induced GLT-1 promoter activity. Mutation of both the NF-κB (-583/-282/-251) and CRE (-308) sites led to a complete abrogation of the effect of TX on GLT-1 promoter activity. Taken together, our findings establish that TX regulates GLT-1 via the CREB and NF-κB pathways.

Growth factors are known modulators of the expression and function of glutamate transporters (21). TGF-␤, which is released from astrocytes in response to estrogen or TX, mediates the neuroprotective effects of both estrogen and TX (22). TGF-␣ has also been shown to afford protection in cerebral artery occlusion and cerebral ischemia models (23,24). Both EGF and TGF-␣ induce GLT-1 expression via NF-B activation in astrocytes (25). In our previous study, we showed that estrogen and TX induced up-regulation of GLT-1 via TGF-␣ (26). To further characterize the signaling pathways and molecular mechanisms involved in TX-induced up-regulation of GLT-1, we investigated the role of cAMP response element-binding protein (CREB) and NF-B in TX-induced GLT-1 up-regulation. Our results demonstrate that both the CREB and NF-B pathways are critical for TX-induced enhancement of GLT-1 expression.

EXPERIMENTAL PROCEDURES
Materials-Cell culture media and reagents were purchased from Invitrogen. TGF-␣ was from Peprotech (Rocky Hill, NJ). PP2, pyrrolidine dithiocarbamate, H89, G15, and G1 were obtained from Tocris Bioscience (Ellisville, MO). Estrogen, dibutyryl cAMP (dbcAMP), protease inhibitor mixture, and poly-D-ornithine were purchased from Sigma-Aldrich (St. Louis, MO). GLT-1, TGF-␣, CREB, NF-B, ␤-actin, and Src antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA) or Cell Signaling Technology (Danvers, MA). The RNA isolation kit was purchased from Qiagen (Valencia, CA). The luciferase reporter assay kit was obtained from Promega (Madison, WI). All chemicals were prepared in Hanks' buffered salt solution, 95% ethanol, or dimethyl sulfoxide according to the instructions of the manufacturer and diluted to final working concentrations in Opti-MEM prior to use.
Primary Cultures of Astrocytes-Astrocyte cultures were prepared as described previously (27). Briefly, after carefully removing the meninges, cerebral cortices of newborn (1-dayold) Sprague-Dawley rats were digested with dispase (Invitrogen). Astrocytes were then recovered by the repeated removal of dissociated cells and plated at a density of 1 ϫ 10 5 cells/ml. Twenty-four hours after the initial plating, the medium was changed to preserve the adhering astrocytes and to remove neurons, microglia, and oligodendrocytes. The cultures were maintained at 37°C in a 95% air, 5% CO 2 incubator for 3 weeks in minimal essential medium supplemented with 10% horse serum, 100 units/ml of penicillin, and 100 g/ml of streptomycin. These cultures showed Ͼ 95% positive staining for the astrocyte-specific marker glial fibrillary acidic protein. All experiments were performed 3 weeks post-isolation.
Western Blot Analysis-Whole cell lysates for SDS-PAGE were prepared as described previously (28). Astrocytes were treated with the designated compounds for the indicated time periods, followed by two washes with cold Hanks' buffered salt solution. Subsequently, the cells were lysed with radioimmune precipitation assay buffer containing a protease inhibitor mixture. The protein concentration in the lysate was determined with BCA. Next, 30 g of protein samples were mixed with Laemmli sample buffer containing 5% ␤-mercaptoethanol and heated at 95°C for 5 min, followed by 10% SDS-PAGE under reducing conditions. The proteins were then electrophoretically transferred to a nitrocellulose membrane. Western blot analysis was performed using primary antibodies (GLT-1, 1:1000, Millipore or Santa Cruz Biotechnology; ␤-actin, 1:5000, Sigma-Aldrich), phospho-CREB, CREB (1:1000, Santa Cruz Biotechnology) followed by secondary antibodies (1:3000, antirabbit, anti-mouse IgG peroxidase conjugates; Promega). The immunoreactive proteins were detected by an enhanced chemiluminescence Western blotting detection kit (Pierce).
Mutagenesis-The GLT-1 (EAAT2) promoter containing Ϫ954 to ϩ44 bp (954) sequences subcloned into the pGL3 basic plasmid vector (9) was used as the original template for mutations. The NF-B triple mutant (-583/-272/-251) was a gift from Dr. Sitcheran (Texas A&M Health Sciences Center). To generate a mutation at the Ϫ308 CREB binding site, the sequence 5Ј-GCC CGG CGC GGG TGA TGT CAG CTC TCG ACG AAA-3Ј was changed to 5Ј-GCC CGG CGC GGG TAA TGA CAG CTC TCG ACG AAA-3Ј by PCR-based mutagenesis using a site-directed mutagenesis kit (Stratagene). Clones were sequenced to confirm the mutations.
Construction of the Deletion Mutants of the EAAT2 Promoter-Deletion constructs were generated from the EAAT2-954 (-954/ϩ44) pGL3-basic vector. PCR was performed for 35 cycles at 94°C for 1 min, at 58°C for 1 min, and at 74°C for 1 min, with a final extension at 74°C for 10 min using the following primers attached with the XhoI and HindIII sites: 5Ј-ACT ATT CTC GAG GGA GAC GGA GGG GCA TCC CGG-3Ј for the Ϫ590/ϩ44 deleted construct (EAAT2-590), 5Ј-ATT ACC CTC GAG CGG CGC GGG TGA TGT CAG CTC-3Ј for the Ϫ315/ϩ44 deleted construct (EAAT2-315), 5Ј-GCC CTC GAG CGA AAA TAG AGA GGG ATC GCC-3Ј for the Ϫ290/ ϩ44 deleted construct (EAAT2-290), 5Ј-ACC CTC GAG CTG CAA ATC CCC AGC TCC GGC-3Ј for the Ϫ270/ϩ44 deleted construct (EAAT2-270), and 5Ј-5-GCC GCC AAG CTT TGG CTT CCC CGA GAG AGC GAT-3Ј for a reverse primer. The amplified EAAT2 DNA fragments were subcloned into the XhoI and HindIII sites of the pGL3-basic vector. The sequences of the new constructs were confirmed by DNA sequencing.
Cloning the TGF-␣ Promoter-Luciferase Vector-A 990-bp DNA fragment from Ϫ950 to ϩ40 of the human TGF-␣ gene was generated by PCR using human genomic DNA. Primer sets were designed as follows: 5Ј-ATT CTC GAG GCA CCA GGG GCC ACC TCA GAG-3Ј for sense containing the XhoI site and 5Ј-GCC AAG CTT GCT CTC CAG CCT CCT GCC CTA-3Ј for antisense containing the HindIII site. The amplified TGF-␣ DNA fragment was digested with XhoI and HindIII, and the fragment was purified according to the instructions of the manufacturer (gel extraction system, Qiagen, Valencia, CA). The purified TGF␣ promoter was subcloned into the XhoI and HindIII sites of the pGL3-basic vector (Promega). The sequence of the TGF-␣ promoter construct was confirmed by DNA sequencing.
Transient Transfections and Luciferase Assay-Astrocytes cultured in 500 l of growth medium in 24-well plates for 2-3 days were transfected overnight with 0.5 g of the pGL3 basic luciferase reporter vector containing EAAT2Pro-954 (EAAT2, provided by Dr. Fisher, Virginia Commonwealth University), various mutants of EAAT2, or TGF-␣ with Lipofectamine 2000 transfection reagent (Invitrogen) in minimal essential medium containing 5% FBS. For some experiments, astrocytes were cotransfected with the luciferase vector and expression vectors (100 ng) such as pRC/RSV-CREB, pCMV-A-CREB, pcDNA-p50, pRC/RSV-p65, and pCMV-IB-␣. Empty vectors of the corresponding expression vectors were used as controls. Luciferase reporter vectors for NF-B and CRE were purchased from Clontech. After transfection, astrocytes were treated with the designated compounds in Opti-MEM for 24 h. Luciferase activity was measured with the Bright-Glo luciferase kit (Promega) according to the instructions of the manufacturer and normalized to the protein content (Bradford protein assay, Bio-Rad). Normalization with protein content was verified by cotransfection of astrocytes with luciferase vectors and firefly reporter pGL4.75 plasmids (Promega) carrying the Renilla luciferase reporter gene.
Real-time PCR-The total RNA extraction from astrocytes was performed with the RNeasy mini RNA isolation kit (Qiagen). 2 g of total RNA was transcribed with poly(dT) oligonucleotides. The primers for GLT-1 and TGF-␣ were designed as follows: 5Ј-CCT CAT GAG GAT GCT GAA GA-3Ј (GLT-1 forward) and 5Ј-TCC AGG AAG GCA TCC AGG CTG-3Ј (GLT-1 reverse); 5Ј-ACG GTC ACT GCT GTC ATT G-3Ј (TGF-␣ forward) and 5Ј-GCG CTG GGC TTC TCG TG-3Ј (TGF-␣ reverse); and 5Ј-TCC CTC AAG ATT GTC AGC AA-3Ј (GAPDH forward) and 5Ј-AGA TCC ACA ACG GAT ACA TT-3Ј (GAPDH reverse). The CFX96 real-time PCR detection system (Bio-Rad) was used to amplify TGF-␣. The reactions were carried out in a total volume of 25 l containing 1 g of cDNA template of each sample, 0.4 M of the appropriate primers, and RT2 SYBR Green qPCR Master Mix (SA-Biosciences/Qiagen). The PCR protocol consisted of one cycle at 95°C for 10 min and 40 cycles at 95°C for 15 s and at 60°C for 1 min. All samples were normalized relative to GAPDH. A webbased PCR array data analysis (SA Biosciences/Qiagen) was used for data analysis.
ChIP Assays-ChIP analysis was performed with the EZ-ChIP chromatin immunoprecipitation kit (Millipore) following the instructions of the manufacturer. Briefly, protein-DNA complexes were cross-linked with formaldehyde (10 min at room temperature) and sonicated to lengths of 100 -500 bp. Next, 100 l of supernatant was mixed with 900 l of ChIP dilution buffer. After preclearing, 1% of the reaction was saved for PCR amplification as input. The remainder was incubated overnight at 4°C with the CREB antibody (Cell Signaling), p50 (Cell Signaling Technology), p65 (Santa Cruz Biotechnology), or control rabbit IgG (Millipore). After isolation and washing of antibody-containing complexes, DNA was extracted. PCRs were performed using the following primers: CREB, 5Ј-GGG ACA ACA GAA GAG GGA CA-3Ј (forward) and 5Ј-AGG GAT TGC AAG GTT TAG CC-3Ј (reverse); NF-B, 5Ј-CTG TGG GAC TCC CCA TCC CAG-3Ј (forward) and 5Ј-TGG CAG GAT CCC AGG GTC TAA-3Ј (reverse). PCR products were resolved on 1% agarose gel and visualized under UV light.
EMSA-EMSA was performed using the LightShift chemiluminescent kit from Thermo Scientific (Rockford, IL) following the instructions of the manufacturer. Briefly, 5 g of nuclear extract from control or treated astrocytes was incubated for 20 min on ice with biotin-labeled oligos containing NF-B or CREB binding sites of the EAAT2 promoter. In some reactions, 10 g of antibody was added prior to the reaction. Next, DNAprotein complexes were resolved in 8% non-denaturing DNA polyacrylamide gels and transferred to a nylon membrane. The complexes were detected with the chemiluminescent nucleic acid detection module from Thermo Scientific. The primer pairs used for NF-B (EAAT2 Ϫ583) were as follows: 5Ј-GGA GAC GGA GGG GCA TCC CGG GTC TCC GC-3Ј and 5Ј-GCG GAG ACC CGG GAT GCC CCT CCG TCT CC-3Ј. The primer pairs used for CRE (EAAT2 Ϫ308) were as follows: 5Ј-CGG CGC GGG TGA TGT CAG CTC TCG ACG AAA-3Ј and 5Ј-TTT CGT CGA GAG CTG ACA TCA CCC GCG CCG-3Ј. All oligos were HPLC-purified and biotin-labeled (Operon Technologies).
DNA Affinity Purification Assay-A DNA affinity purification assay was performed with a Factor Finder kit from Miltenyi Biotec, Inc. (Auburn, CA). Briefly, 1.5 g of biotinylated oligos (the same sequences as for EMSA) were incubated with 50 g of nuclear extract in binding buffer for 20 min. Next, 100 l of streptavidin microbeads were added and further incubated for 10 min. The reaction mixture was applied onto the microcolumn, which was equilibrated with binding buffer and set in the magnetic field. After washing four times with 100 l of low-salt and high-salt buffers each, bound proteins were eluted with 30 l of elution buffer and analyzed by Western blotting.
Statistical Analysis-The mean and S.E. were determined for each data set. One-way analysis of variance (ANOVA) followed by Tukey post hoc test was used to determine significance. Student's t test was used for two sets of data to determine significance. p Ͻ 0.05 was considered to represent statistically significant differences.

RESULTS
Tamoxifen Increases Astrocytic GLT-1 and TGF-␣ Expression-Several growth factors, including EGF and basic fibroblast growth factor, increase GLT-1 expression in astrocytes (29,30). Similarly, estrogen increases GLT-1 and glutamate aspartate transporter expression, thereby enhancing astrocytic glutamate uptake (31,32). Our previous study established that both estrogen-and TX-induced astrocytic GLT-1 expression was mediated by TGF-␣ signaling upon 24 h exposure (26). Given that EGF or neuron-dependent activation require a prolonged period of time to induce GLT-1 expression (33), this study assessed the effects of TX on astrocytic GLT-1 protein levels after 6-day treatment. As shown in Fig. 1A, both TX and TGF-␣ induced a significant increase of astrocytic GLT-1 protein expression at this time point. dbcAMP was used as a positive control. The time course study revealed that TX induced GLT-1 protein expression as early as 30 min posttreatment and continuously for 6 days (data not shown). TX also increased GLT-1 mRNA (Fig. 1B) and TGF-␣ mRNA ( Fig.  1C) levels, showing a significant increase within 4 h of incubation, in agreement with our previous findings (26). Analogous to the TX-induced GLT-1 protein increase, GLT-1 mRNA was also continuously increased up to 6 days post-treatment (data not shown). TX also significantly increased TGF-␣ promoter activity (p Ͻ 0.05, Fig. 1D). Moreover, blocking the TGF-␣ receptor, EGF receptor (EGFR), with AG1478 completely abrogated the TX-induced GLT-1 promoter activity, suggesting a possible role of TGF-␣ in this process (Fig. 1E).
CREB Is Involved in TX-induced GLT-1 Up-regulation-The CREB-regulated transcriptional machinery provides neuroprotection against ischemia (34). Interestingly, both the GLT-1 and TGF-␣ promoters contain one CRE consensus site, and CREmutation on the GLT-1 promoter decreases its activity (28). On the basis of these observations, we tested whether the CREB pathway plays a role in TX-induced regulation of GLT-1. As shown in Fig. 2A, both TX and TGF-␣ increased CRE reporter activity to levels comparable with dbcAMP (a positive control). The activation of CREB is triggered by PKA, and estrogen has been shown to induce this pathway (35). Accordingly, we tested whether the PKA pathway is involved in TX-induced CREB activation using H89, a PKA inhibitor. Inhibition of PKA signif-  ). B and C, GLT-1 mRNA and TGF-␣ mRNA levels were measured after TX treatment (0 -4 h) by real-time PCR (normalized to GAPDH). D, astrocytes were transfected overnight with the TGF-␣ promoter vector and treated with TX for 24 h. TGF-␣ promoter activity was measured by luciferase assay. E, astrocytes were treated with 1 M TX, 10 M AG1478 (AG), or AG1478 plus TX for 24 h, followed by the luciferase assay. *, p Ͻ 0.05; ***, p Ͻ 0.001, compared with the control; ANOVA followed by Tukey's post hoc test; n ϭ 3.

FIGURE 2. The CREB pathway is involved in TX-induced GLT-1 regulation.
A, astrocytes were transfected overnight with the CRE reporter vector, followed by 24 h treatment with 50 ng/ml TGF-␣, 250 M dbcAMP, or 1 M TX and subsequent assayed for EAAT2 promoter activity by luciferase assay (C, control). B and C, post-transfected astrocytes were incubated for 1 h with 10 M H89 prior to 24-h treatment with TX (B) or TGF-␣ (C), followed by luciferase assay. D, astrocytes were cotransfected for 24 h with the wild-type (954, containing Ϫ954 to ϩ44 bp of the EAAT2 promoter) or triple NF-B mutant (T-M) of EAAT2 promoter vectors with empty vector (EV) pRcRSV or CREB expression vector (CREB), followed by luciferase assay. ***, p Ͻ 0.001; #, p Ͻ 0.05, compared with the control; @@@, p Ͻ 0.001; ANOVA followed by Tukey's post hoc test; n ϭ 3.
icantly attenuated the effect of TX or TGF-␣ on EAAT2 promoter activity (Fig. 2, B and C). Overexpression of CREB significantly increased EAAT2 promoter activity (Fig. 2D). Previous studies reported that NF-B played a critical role in the positive regulation of EAAT2 promoter activity (36). Thus, we tested whether CREB can still activate the EAAT2 promoter when all three NF-B sites were mutated and determine its role in EAAT2 regulation independently of NF-B. The results showed that CREB overexpression significantly increased EAAT2 promoter activity in the NF-B triple mutant of the EAAT2 promoter (Fig. 2D), indicating that, in addition to NF-B, the CREB pathway also independently plays a critical role in EAAT2 regulation. CREB overexpression by the transient transfection of the CREB expression plasmid vector (pRcRSV served as a control empty vector) was confirmed by Western blot analysis (Fig. 2D).
TX Induces CREB Phosphorylation and CREB Binding to the EAAT2 Promoter in Cultured Astrocytes (in Vivo)-The phosphorylation of CREB is essential for its role as a transcriptional regulator (37). Phosphorylation of astrocytic CREB was noted as early as 5 min after TX treatment and peaked at 15 min, followed by a gradual decline to basal levels (Fig. 3A). Estrogen also increased phospho-CREB but showed a delayed effect, increasing CREB phosphorylation 1 h after treatment (Fig. 3B). The total CREB proteins were reprobed from the same blot to ensure the presence of an equal amount of unphosphorylated CREB proteins in the control and all treated groups. EAAT2 contains a CRE consensus site in the region of Ϫ308 of its pro-  (C, control). C and D, astrocytes were treated with TX or estrogen for 24 h, followed by the analysis of CREB binding to its consensus site in the EAAT2 promoter by ChIP assay. *, p Ͻ 0.05; ***, p Ͻ 0.001, compared with the control; ANOVA followed by Tukey's post hoc test; n ϭ 3. Control groups were treated with vehicle (ethanol) and IgG groups were treated with TX or E2 but immunoprecipitated with IgG instead of CREB antibody for negative control.  ). B, after overnight transfection with the NF-B reporter vector, astrocytes were treated with TX for 24 h, followed by luciferase assay. C and D, astrocytes were incubated with estrogen (E2) (C) or TX (D), and binding of NF-B p50 or p65 to its consensus sites in the EAAT2 promoter was determined by ChIP assay. Control groups were treated with vehicle (ethanol) and IgG groups were treated with TX or E2 but immunoprecipitated with IgG instead of CREB antibody for negative control. E, the intensity of TX-induced binding was quantified. ***, p Ͻ 0.001; compared with the control; @, p Ͻ 0.05; ANOVA followed by Tukey's post hoc test; n ϭ 3.

Mechanism of Tamoxifen-induced GLT-1 Expression
OCTOBER 4, 2013 • VOLUME 288 • NUMBER 40 moter. Therefore, we performed a ChIP assay to examine whether TX modulates CREB binding to the EAAT2 promoter in vivo using cultured primary astrocytes. As shown in Fig. 3, C and D, TX induced a significant increase in CREB binding to its consensus site on the EAAT2 promoter. Estrogen also significantly increased CREB binding to the EAAT2 promoter, albeit to a lesser extent than TX (Fig. 3, C and D).

TX Induces NF-B Binding to the EAAT2 Promoter in a ChIP
Assay-The NF-B signaling pathway plays a critical role in EGF-and ceftriaxone-induced GLT-1 expression (38,39). Accordingly, we tested whether the NF-B pathway also mediates the effect of TX on EAAT2 expression. As shown in Fig. 4A, TX significantly increased NF-B luciferase reporter activity, and inhibition of the NF-B pathway with pyrrolidine dithio- FIGURE 5. TX induces binding of NF-B and CREB to the EAAT2 promoter. A and B, astrocytes were treated with TX for 24 h, followed by nuclear extraction. Next, nuclear extracts were incubated with biotinylated oligos containing NF-B and CRE binding sequences of EAAT2 promoter for 20 min. EMSA was carried out to determine complex formation. An excess (100 ϫ) of non-biotinylated oligos was used to determine whether the protein-DNA complex was specific for the designated oligos. Antibodies for p65 or CREB were also added to some reactions to determine the supershift of the complexes. Two sets of TX groups are shown in A, and dbcAMP was used as a positive control. C and D, astrocytes were treated with TX for 24 h, and isolated nuclear extracts were incubated with biotinylated oligos of the NF-B and CRE binding sequences of the EAAT2 promoter, followed by elution from the magnetic beads in the microcolumn. The eluents were subjected to Western blot analysis to detect p65 and p50 (C) or CREB (D). In both groups, 10 g of nuclear extracts was loaded as input (C, control).

FIGURE 6. The role of the CRE and NF-B binding sites in TX regulation on EAAT2 promoter activity.
A and B, after overnight transfection with the wild type, the CRE mutant (-308), or the triple NF-B mutant (-583/-272/-251, TM) of the EAAT2 promoter vector, astrocytes were treated with 1 M TX (A) or 50 ng/ml TGF-␣ (B) for 24 h. EAAT2 promoter activity was measured by luciferase assay (C, control). C, after overnight transfection with the wild type (EAAT2-954, Ϫ954 to ϩ 44 bp) or various deletion mutants (270, Ϫ270 to ϩ44; 290, Ϫ290 to ϩ44; 315, Ϫ315 to ϩ44, and 590, Ϫ590 to ϩ 44 bp of the EAAT2 promoter), astrocytes were treated with TX for 24 h, followed by luciferase assay. D, after overnight transfection with various mutant constructs of the EAAT2 promoter vectors and single or multiple site mutants of CRE or NF-B, astrocytes were treated with TX for 24 h, followed by luciferase assay. *, p Ͻ 0.05; **, p Ͻ 0.01; ###, p Ͻ 0.001; compared with the control; @, p Ͻ 0.05; ANOVA followed by Tukey's post hoc test; n ϭ 3.
carbamate abrogated the TX-induced EAAT2 promoter activity (Fig. 4B). Next, we performed a ChIP assay to determine whether TX induced the binding of NF-B to the EAAT2 promoter. Estrogen was also tested as an endogenous molecule related to TX. As shown in Fig. 4C, estrogen increased the binding of the NF-B subunits p65 and p50 to the EAAT2 promoter. TX also induced significant recruitment of both p65 and p50 (Fig. 4, D and E), although the preference of NF-B subunit recruitment to the EAAT2 promoter by TX is unclear in this study.
TX Induces NF-B and CRE Binding to the EAAT2 Promoter in Nuclear Extracts (in Vitro)-Because TX induced both CREB and NF-B binding to the EAAT2 promoter in vivo in cultured astrocytes, as shown in the ChIP assay, we next performed both EMSA and DNA affinity purification assay experiments to confirm that TX induces their binding to the EAAT2 promoter in vitro using nuclear extracts. As shown in Fig. 5, A and B, TX enhanced the binding of both the p65 and p50 subunits of NF-B and CRE to the EAAT2 promoter. The EMSA results show that TX induced specific NF-B p65 binding to its consensus oligos because an excess amount of unlabeled oligos completely blocked the binding. Interestingly, the NF-B p65 antibody did not cause a supershift of these complexes. Nonetheless, the experiments corroborate that TX induced specific NF-B p65 binding to its consensus oligos (Fig. 5A). The TXinduced binding of CRE consensus oligos to the CREB protein was also specific because excess unlabeled CRE oligos competed with biotinylated CRE oligos, inhibiting the formation of a protein-DNA complex, and coincubation with CREB antibody induced the supershift of the complexes (Fig. 5B). Moreover, DNA affinity purification assay experiments showed that TX induced strong and specific bindings of NF-B and CREB to the EAAT2 promoter (Fig. 5, C and D).
Both the CREB and NF-B Binding Sites Are Critical for TX Regulation on EAAT2-Activation of CREB signaling mediates the effect of growth factors, such as EGF and TGF-␣, on GLT-1   ). B, after overnight transfection with the EAAT2 promoter vector, astrocytes were pre-treated with 1 M of G15 for 1 h followed by 24-h incubation 1 M TX. EAAT2 promoter activity was measured by luciferase assay. C and D, after overnight cotransfection with the EAAT2 promoter vector and expression vectors, including control vector eGFP ER-␣, or ER-␤, astrocytes were treated for 24 h with 1 M TX. EAAT2 promoter activity was determined by luciferase assay. **, p Ͻ 0.01; ***, p Ͻ 0.001; compared with the control; @, p Ͻ 0.05; ANOVA followed by Tukey's post hoc test; n ϭ 3.
protein expression (40). Nevertheless, there is a dearth of information regarding the mechanism by which CREB regulates GLT-1 (EAAT2) at the transcriptional levels. Because, in addition to the significant role of CREB through the CRE consensus site at Ϫ308 (28), NF-B also serves as a critical regulator of the EAAT2 promoter, we used the mutant construct of all three NF-B sites (-251/-272/-583) of the EAAT2 promoter. As shown in Fig. 6, A and B, either CRE or the triple NF-B mutation (TM) significantly decreased the EAAT2 promoter activity, confirming that both the CREB and NF-B pathways are critical in EAAT2 regulation. Moreover, TX and TGF-␣ still significantly increased EAAT2 promoter activity in these mutants (Fig. 6, A and B), indicating that inactivation of one of the CRE or NF-B sites is insufficient to completely block the effect of TX/TGF-␣ on EAAT2. We also generated deletion constructs of the EAAT2 promoter to determine the sites responsible for the effects of TX. As shown in Fig. 6C, each deletion construct containing at least one of the NF-B (Ϫ251, Ϫ272, or Ϫ583) or CRE (-308) sites significantly decreased promoter activity compared with the wild-type EAAT2, but the TX-induced stimulatory effect remained significant. Indeed, the EAAT2 mutant construct that was inactivated for both the CRE and triple NF-B sites completely abrogated the effect of TX on EAAT2 promoter activity (Fig. 6D), indicating that these two pathways are critical for TX-induced EAAT2 regulation.
CREB and NF-B Also Play a Critical Role in TGF-␣ Regulation-Because the EAAT2 (or GLT-1) promoter does not contain an estrogen response element consensus sequence, the regulation of GLT-1 expression by TX is likely mediated via TGF-␣ (Fig. 1). Interestingly, both the TGF-␣ and EAAT2 promoters contain one CRE and three NF-B consensus sites. Our results show that overexpression of CREB increased, whereas repression of CREB with mutant CREB (ACREB), the dominant negative CREB inhibitor, decreased TGF-␣ promoter activity (Fig. 7, A and B, respectively). Moreover, activation of NF-B p65 or p50 by overexpression also significantly increased TGF-␣ promoter activity, but repression of NF-B by overexpression of IB-␣ significantly decreased TGF-␣ promoter activity (Fig. 7, C-E, respectively). These results indicate that both the CREB and NF-B pathways play a critical role in regulating TGF-␣ promoter activity.

TX Enhances GLT-1 via Estrogen Receptor (ER)-␣, ER-␤, and G Protein-coupled Receptor 30 (GPR30)-
In addition to the conventional nuclear ERs (ER-␣ and ER-␤), GPR30 also mediates the biological effects of estrogen via transactivation of growth factor receptors (41). We used several modulators of the ERs to gain better understanding on the ER mechanism/s associated with TX-induced GLT-1 regulation. Our results show that G1, an agonist of GPR30, increased GLT-1 protein expression to a similar extent as TX (Fig. 8A). ICI 182,780, a nonselective antagonist of ER-␣ and ER-␤, did not block the TX-induced increase in GLT-1 protein expression, whereas G15, a GPR30 antagonist, partially, but significantly reduced the TXinduced enhancement of GLT-1 protein expression. These results indicate that GPR30 may be, at least in part, responsible for TX-induced GLT-1 regulation. Interestingly, ICI 182,780 alone also increased GLT-1 expression, possibly acting as an agonist of GPR30 (Fig. 8A). To further confirm the role of GPR30 in TX-induced GLT-1 regulation, we tested whether G15, a GPR30 antagonist, blocks the TX-induced EAAT2 promoter activity. G15 significantly, but not completely, abolished the effect of TX on EAAT2 promoter activity (Fig. 8B). Next, we examined whether ER-␣ or ER-␤ play a role in TX-induced EAAT2 promoter activity by overexpression of ER-␣ or ER-␤. As shown in Fig. 8, C and D, TX activated both ER-␣ and ER-␤, leading to a significant increase in EAAT2 promoter activity. These results indicate that TX regulates GLT-1 function via all ERs, including ER-␣, ER-␤, and GPR30.
Inhibition of Src Family Kinases Partially Blocks TX-induced GLT-1 Expression-As shown in Fig. 9, A and B, inhibition of Src family tyrosine kinases with 10 M PP2 abrogated both the TX-induced effect on GLT-1 protein expression and EAAT2 promoter activity. PP2 completely abrogated G1-induced GLT-1 protein expression but only partially (p Ͻ 0.01) suppressed the effect of TX on GLT-1.

DISCUSSION
We investigated the mechanism of TX-induced transcriptional regulation of GLT-1 in rat primary astrocytes. Our findings established that TX enhanced GLT-1 expression by activating both the CREB and NF-B pathways. Finally, the effect of TX was mediated via GPR30 as well as ER-␣ and ER-␤ (Fig. 10).
Excessive glutamate-induced excitotoxicity has been linked to various neurological disorders (38). The astrocytic glutamate transporter GLT-1 (EAAT2 in humans) is primarily responsible for synaptic glutamate uptake (ϳ90%) (42). Thus, delineating the molecular regulatory mechanisms of GLT-1 is critical for the development of efficacious therapeutics to increase GLT-1 expression and reduce synaptic glutamate concentrations. Several studies have reported that increased EAAT2 protein levels effectively reduce excitotoxicity, leading to neuroprotection in a number of neurodegenerative disease models (9,10,12). TX is routinely used to treat breast cancer because of its antagonistic effect on breast tissue (43). In recent years, it has also gained attention for its agonistic effects in the CNS as a neuroprotective agent (44). TX is neuroprotective in several experimental models by inducing antioxidative and antiapoptotic functions (45,46). Our finding that TX enhances GLT-1 expression provides novel information regarding potential molecular targets to mitigate excitotoxic injury.
Because identification of molecular targets for TX-induced GLT-1 expression could be vital for neuroprotection, our finding that both CREB and NF-B signaling play a critical role in the TX-induced enhancement of GLT-1 expression provides valuable information. We have shown previously that TX enhanced expression of GLT-1 mRNA and protein in astrocytes that was mediated by TGF-␣ (26). Given the mediating role of TGF-␣ in TX-induced GLT-1 expression, we observed that TX increased GLT-1 mRNA levels before the TGF-␣ mRNA increase (30 min versus 1 h). TX-induced GLT-1 increase prior to TGF-␣ might be due to TX-induced transactivation on EGFR via GPR30. It has been reported that estrogen induced transactivation of EGFR through GPR30 activation (47). The estrogen binding to GPR30 induced transactivation of the EGFR, leading to the rapid activation of intracellular signaling cascades within 5 min (48). This could be the case for a TX-induced rapid increase of GLT-1 prior to TGF-␣ production. This notion is also supported by our earlier study demonstrating that G1, a selective agonist of GPR30, rapidly increased GLT-1 expression within 30 min (28). In addition, the result that inhibition of the EGFR or TGF-␣ knockdown with siRNA led to abrogation of the effect of TX on GLT-1 expression provides evidence that TGF-␣ plays a major role in mediating the effect of TX on GLT-1 expression.
The GLT-1 promoter contains one CRE and three NF-B consensus sites (Fig. 10) (33). Regulation of the CREB pathway may be critical in affording neuroprotection because CREB activation leads to the expression of genes that encode neuroprotective molecules (49,50). In the rat hippocampus, TX regulates the expression of synaptophysin via CREB (51). TX-induced CREB activation is also involved in improving acute mania (52). Our finding shows that the NF-B pathway plays a critical role in TX-induced GLT-1 expression (Figs. 4 and 5). The role of NF-B in TX-induced neuroprotection is supported by observations that TX protected human coronary artery endothelial cells from hypoxic injury via NF-B activation (53). NF-B signaling is also crucial for the enhancement of GLT-1 expression induced by growth factors such as EGF and TGF-␣ (36,39). There are three NF-B consensus sites in the GLT-1 promoter, Ϫ583, Ϫ272, and Ϫ251. Several compounds are known to activate the NF-B sites specific to the GLT-1 promoter. EGF enhances GLT-1 expression mainly by activation of the Ϫ583 NF-B site of the EAAT2 promoter (39). The ␤-lactam antibiotic ceftriaxone has been shown to enhance GLT-1 expression via NF-B activation (27) exclusively by activation of the Ϫ272 NF-B site of the GLT-1 promoter. Neuronal regulation of GLT-1 expression specifically targets the Ϫ583 and Ϫ251 NF-B sites of the GLT-1 promoter (14). Our results indicate that TX-induced NF-B activation was not specific to the Ϫ583 or Ϫ272 sites because mutation of these sites on the GLT-1 promoter failed to completely abrogate the TX-enhancing GLT-1 promoter activity (Fig. 6D). This likely reflects the ability of TX to activate all of the NF-B sites and the CRE site, indicating that both the CREB and NF-B pathways are involved in TX-induced enhancement of GLT-1 expression.
The results from this study reveal that the TX-induced transcriptional regulation of TGF-␣ expression also involves CREB and NF-B. The transcription factors p53 and AP-2 are known to positively regulate TGF-␣ promoter activity (54). Our findings suggest that both NF-B and CREB activation are critical for the enhancement of TGF-␣ promoter activity. On the basis of our results, TX may directly activate the NF-B and CREB pathways to induce GLT-1 expression. However, our observation that inhibition of the TGF-␣ receptor EGFR or knockdown of TGF-␣ with siRNA abolished the stimulatory effect of TX on GLT-1 suggests that TGF-␣ is an essential mediator of the effects of TX on GLT (26). The most likely scenario is that TX may utilize NF-B and CREB signaling to generate TGF-␣, in turn activating EGFR in an autocrine mode and, thus, leading to increased GLT-1 expression by NF-B and CREB (Fig. 10). Further investigation is warranted to determine the exact role of TGF-␣ in TX-induced GLT-1 up-regulation.
TX-induced neuroprotection involves multiple mechanisms, including the activation of intracellular signaling proteins such as MAPK/ERK and PI3K/Akt (31,55). We reported previously that activation of MAPK/ERK, PI3K/Akt, and EGFR is involved in TX regulation of GLT-1 (31). Src signaling is involved in estrogen-induced neuroprotection against glutamate toxicity (56). Estrogen stimulates the c-fos gene via GPR30 through activation of both Src and EGFR signaling in breast cancer cells (57). TX activates GPR30 to stimulate cell growth activation via Src and EGFR signaling in TX-resistant breast cancer MCF-7 cells (58). Likewise, TX may activate the Src family tyrosine kinases via GPR30 to induce GLT-1 expression in primary astrocytes (Fig. 9).
Estrogen exerts its cellular effects via the nuclear receptors ER-␣ and ER-␤ as well as GPR30, a G protein-coupled ER (59,60). Our finding that TX-induced GLT-1 expression is partially inhibited by G15, an antagonist of GPR30, indicates that GPR30 is not the sole mechanism that mediates the effects of TX on GLT-1 expression. Indeed, ER-␣ and ER-␤ also contributed to the TX regulation of GLT-1 expression (Fig. 8, C and D). ER-␣ and ER-␤, localized to the plasma membrane, are known to induce rapid estrogen signaling (61). These ERs may also trigger cross-talk with membrane-associated proteins, such as EGFR. It is unclear at present whether TX-induced activation of ER-␣ and ER-␤ is via membrane-bound ER-␣/␤ or nuclear ER-␣/␤. Taken together, the results presented here support the following model for the mechanism of TX-induced GLT-1 regulation (Fig. 10). TX activates GPR30 as well as ER-␣ and ER-␤, resulting in up-regulation of TGF-␣, which is released into the extracellular milieu in an autocrine fashion. TGF-␣, in turn, activates EGFR and increases GLT-1 expression via the CREB and NF-B pathways. Activated GPR30 may also directly transactivate EGFR, leading to CREB and NF-B activation as well as increased GLT-1 expression. Because the up-regulation of GLT-1 attenuates excitotoxicity, pharmacological agents that increase GLT-1 expression and glutamate transporter function may be beneficial for the treatment of neurodegenerative diseases with an inherent excitotoxic etiology. Therefore, targeting specific sites along the scheme (Fig. 10) may lead to the development of novel therapeutic modalities to mitigate neurological disorders associated with impairment of glutamate transporters.