Transcriptional Regulation of Brain-derived Neurotrophic Factor Coding Exon IX

Brain-derived neurotrophic factor (BDNF) is an active neurotrophin abundantly expressed throughout the nervous system. It plays an important role in synaptic transmission, plasticity, neuronal proliferation, differentiation, survival, and death. The Bdnf gene in rodents has eight non-coding exons and only a single coding exon (IX). Despite its recognized regulation by neuronal activity, relatively little is known about its transcriptional regulation, and even less about the transcription factor candidates that may play such a role. The goal of the present study was to probe for such a candidate that may regulate exon IX in the rat Bdnf gene. Our in silico analysis revealed tandem binding sites for nuclear respiratory factor 2 (NRF-2) on the promoter of exon IX. NRF-2 is of special significance because it co-regulates the expressions of mediators of energy metabolism (cytochrome c oxidase) and mediators of neuronal activity (glutamatergic receptors). To test our hypothesis that NRF-2 also regulates the Bdnf gene, we performed electrophoretic mobility shift assay (EMSA), chromatin immunoprecipitation (ChIP), promoter cloning, and site-directed mutagenesis, real-time quantitative PCR (RT-qPCR), and Western blotting analysis. Results indicate that NRF-2 functionally regulates exon IX of the rat Bdnf gene. The binding sites of NRF-2 are conserved between rats and mice. Overexpressing NRF-2 up-regulated the expression of Bdnf exon IX, whereas knocking down NRF-2 down-regulated such expression. These findings are consistent with our hypothesis that NRF-2, in addition to regulating the coupling between neuronal activity and energy metabolism, also regulates the expression of BDNF, which is intimately associated with energy-demanding neuronal activity.

By means of a combination of bioinformatics, biochemical, and molecular biological approaches, including in silico analysis, electrophoretic mobility shift and supershift assays, realtime quantitative PCR (RT-qPCR), chromatin immunoprecipitation, promoter mutational analysis, overexpression, shRNA, functional assays, and Western blot analysis, we have documented that NRF-2 functionally regulates the transcription of exon IX of the Bdnf gene in rat cortical neurons.

Results
Bdnf Exons' Expression in Visual Cortex-Using PCR primers specific to each of exons I to IX (Table 1) and RT-qPCR, we detected transcripts for each Bdnf exon in the rat visual cortex. Fig. 1 illustrates that, in comparison to exon IX (expressed as 100%), all of the other exons, including exon IXA that contributes to the pro-BDNF, were expressed in much lower quantities. The almost non-detectable ones were exons II, V, and VII. Clearly, the expression of exon IX, being at least 10-fold higher than any of the other exons, dominated over all of the other exons. As this is the only coding exon and very little was known about the cis elements in its promoter region that may regulate its expression, we chose to focus on the promoter of exon IX.
In Silico Analysis of the Promoter Region of Bdnf Exon IX- Fig. 2 shows a schematic diagram of the rat Bdnf gene, with eight non-coding exons and a single coding exon, exon IX (modified from 28). Proximal promoters of mouse and rat Bdnf exon IX with DNA sequence 1 kb 5Ј upstream and 1 kb 3Ј beyond the transcription start point (TSP) were analyzed in silico. Megalign analysis was used to find tandem repeats of GGAA or TTCC typical of NRF-2 binding motifs in the promoter region upstream of the coding portion of exon IX. We chose four putative binding sites that showed high homology between rats and mice, and partial homology with humans. The positions of these four sites upstream of the ATG codon (translation start site) in exon IX are: site I (Ϫ42 to Ϫ15), site II (Ϫ331 to Ϫ301), site III (Ϫ444 to Ϫ420), and site IV (Ϫ802 to Ϫ772). Sequences for sites I to IV are shown in Fig. 2.
In Vitro Binding of NRF-2 to the Promoter of Bdnf Exon IX-To determine the binding capacity of NRF-2 to its candidate sites, we performed EMSA. In agreement with our previous findings (26), the promoter of cytochrome c oxidase subunit 6b (COX6b) formed a specific DNA/NRF-2 shift and a DNA/NRF-2/NRF-2 antibody supershift complex when incubated with visual cortical nuclear extract (Fig. 3, lanes 1 and 3, respectively). The shift band was competed out with the addition of an excess of unlabeled COX6b (Fig. 3, lane 2).
When 32 P-labeled site II of the promoter region of Bdnf exon IX was subjected to EMSA, it showed a shift complex with the rat visual cortical nuclear extract (Fig. 3, lane 4). The complex was competed out with the addition of unlabeled competitors (Fig. 3, lane 5). A supershift band was evident when NRF-2 antibody was added to the Bdnf sequence (Fig. 3, lane 6). The addition of unlabeled probes with mutated NRF-2 binding sites was not able to compete out the shift band (Fig. 3, lane 7). Labeled Bdnf oligo with NRF-2 antibody in the absence of any brain extract did not show any shift or supershift band, indicating that there was no direct antibody-oligo interaction (Fig. 3,  lanes 8). Shift, competitor, and supershift reactions with the mutated NRF-2 site on the 32 P-labeled Bdnf oligo also did not show any distinct binding to the NRF-2 site (Fig. 3, lanes 9 -11). Site I was also tested with EMSA, and it yielded a specific shift and a supershift complex for NRF-2 that was competed out with an excess of cold competitor (Fig. 3, lanes 12, 14, and 13, respectively), but not with an excess of unlabeled probe with mutated NRF-2 sites (Fig. 3, lane 15). Again, there was no antibody-oligo interaction without the brain extract (Fig. 3, lane  16). Mutating the binding motif for NRF-2 in site I also did not produce any shift or supershift bands (Fig. 3, lanes 17-19). We have also tested sites III and IV with EMSA, but neither of them gave a shift or a supershift band (data not shown).  In Vivo Interaction via Chromatin Immunoprecipitation Assay (ChIP)-To provide further corroboration of the physical association of NRF-2 with its binding sites on the promoter of Bdnf exon IX gene, we performed a ChIP assay using rat visual cortical tissue (Fig. 4). A 0.5% dilution of input chromatin was used as a standard to indicate the efficiency of the PCR. As with EMSA, COX6b was used as a positive control. Exon 8 of NRF-1, which does not contain a NRF-2 binding motif, served as a negative control.
As shown in Fig. 4, the anti-NRF-2 antibody specifically immunoprecipitated the Bdnf exon IX promoter fragments. The appearance of a distinct band with NRF-2 antibody verified an in vivo interaction of the Bdnf gene with NRF-2. This was true for both sites I and II of the promoter region for exon IX (Fig. 4). Nerve growth factor receptor (NGFR) antibody was used as a control for the immunoprecipitation reaction, and it did not yield any PCR product. An additional "no antibody" control was used to rule out the possibility of a bead-DNA The enclosed portion at the top right for exons IXA and IX is enlarged to indicate the four putative NRF-2 binding sites tested. Each binding site contains a tandem repeat of GGAA or TTCC. The translation start site (ATG) of exon IX is located 21 nucleotides downstream of the transcription start point of exon IX. The sequences for sites I and II used for EMSA are indicated below, with the NRF-2 binding motifs underlined. There is high homology between rat and mouse sequences at these two sites. Nucleotide sequences for sites III and IV are also given. FIGURE 3. In vitro EMSA and supershift assays to show the binding of NRF-2 to putative sites II and I on the promoter of exon IX in the Bdnf gene. All lanes contain 32 P-labeled oligonucleotides containing either site II or site I, and they are indicated by a "ϩ" or a "Ϫ" sign depending on whether they also contain rat brain nuclear extract, excess unlabeled oligos, unlabeled mutant oligos, or NRF-2 antibody. NRF-2 shift, supershift, and nonspecific complexes are indicated by arrows. The positive control for NRF-2 binding is COX6b. Specific NRF-2 shift bands are revealed upon incubation with cortical nuclear extract (lanes 1 and 3). Excess unlabeled competitor competed out the shift band (lane 2). The addition of NRF-2 antibody yielded a strong supershift band (lane 3). Incubation of cortical nuclear extract with the Bdnf probes yielded a specific NRF-2 shift band (lanes 4 and 6 for site II, lanes 12 and 14 for site I) that was competed out with the addition of excess cold probe (lanes 5 and 13). The addition of NRF-2 antibody yielded specific supershift band (lanes 6 and 14). The addition of excess unlabeled mutant Bdnf probes did not compete out the shift reaction (lanes 7 and 15). Labeled Bdnf probe with NRF-2 antibody without any nuclear extract was used as a control to check for any antibody-oligo interaction, and no shift or supershift bands are evident (lanes 8 and 16). Labeled Bdnf probes with mutant NRF-2 sites did not yield prominent specific NRF-2 shift or supershift bands (lanes 9 -11 and 17-19).
interaction. Indeed, it also did not yield any PCR product. Sites III and IV were also tested with ChIP on visual cortical tissue immunoprecipitated with NRF-2 antibodies, but both sites did not reveal any interactions (Fig. 4).
Effect of Mutating NRF-2 Binding Sites on the Promoter Activity of Exon IX-Luciferase reporter assays were done to verify the significance of NRF-2 binding. Site-directed mutations of the putative NRF-2 binding sites on the promoter of exon IX were constructed. Wild type controls or mutants were transfected into cultured N2a cells. COX6b wild type and mutated NRF-2 binding sites served as controls. Fig. 5 shows a significant reduction in COX6b promoter activity in mutants to 40% of controls, and a significant fall to 62.44 and 52% in the Bdnf site I and II mutants, respectively, as compared with controls (*, p Ͻ 0.05, **, p Ͻ 0.01 when compared with respective controls).

Knockdown of NRF-2␣ Decreased mRNA Levels of Bdnf Exon IX and the Effect of KCl Depolarization in Primary Neurons-A
knockdown of NRF-2␣ was achieved in cultured primary neurons with a NRF-2␣ shRNA construct. For transfection controls, neurons received scrambled shRNA, which failed to bind to any known mRNA. In NRF-2␣-silenced neurons, NRF-2␣ protein was significantly reduced to 48%, whereas BDNF protein fell to 53% of control values (p Ͻ 0.001 for both) (Fig. 6, A  and B). Silencing of NRF-2␣ also resulted in a significant decrease in NRF-2␣ mRNA levels to 34.65% of controls (Fig. 6C, p Ͻ 0.001). Likewise, the transcripts for COX7c (positive control) was reduced to 47% and those for Bdnf exon IX to 43% when compared with controls (p Ͻ 0.001 for both, Fig. 6D).
We have shown previously that KCl depolarization up-regulated COX and specific excitatory glutamatergic neurotransmitter receptor subunits, and that this involved NRF-2 (25,26,29). Indeed, 20 mM KCl for 5 h induced an up-regulation of COX7c (to 152%) as well as Bdnf exon IX (to 142%) in cultured visual cortical neurons (p Ͻ 0.05 and p Ͻ 0.01, respectively) ( Fig. 6D). However, if neurons were transfected with NRF-2␣ shRNA, then the same KCl regimen could no longer increase transcripts levels of COX7c and Bdnf to those of control-plus KCl (p Ͻ 0.001 and p Ͻ 0.01, respectively, when compared with control-plus KCl).
Overexpressing NRF-2 Increased mRNA Levels of Bdnf Exon IX and the Effect of TTX Blockade in Primary Neurons-NRF-2␣ and NRF-2␤ overexpression plasmids were transfected into cultured primary visual cortical neurons. ␤-Actin was the internal control. Both subunits were used to form the functional NRF-2␣␤ complex. NRF-2 overexpression increased the protein levels of NRF-2␣ to 223%, NRF-2␤ to 181.67%, and those of BDNF to 167% (p Ͻ 0.001 for all, Fig. 7
Previous studies in our laboratory showed that TTX-induced impulse blockade decreased the transcript levels of COX subunits as well as those of NMDA, AMPA, and GABA A receptor subunit genes (25)(26)(27)30). To determine whether NRF-2 overexpression could rescue Bdnf transcripts from being down-regulated by TTX, primary cultured neurons that were transfected with empty vector controls or with NRF-2␣/␤ overexpression vectors were subjected to 0.4 M TTX treatments for 3 days. Control neurons showed a significant down-regulation of COX7c and Bdnf transcripts to 50 and 42%, respectively (p Ͻ 0.01 and p Ͻ 0.001, respectively; Fig. 7D). However, neurons that were transfected with NRF-2 overexpression vectors did not exhibit significant down-regulation of these transcripts with TTX. Rather, the level of COX7c was 208.28% and that of Bdnf was 195.99% when compared with controls (p Ͻ 0.01 for both; and p Ͻ 0.001 when compared with control ϩ TTX; Fig.  7D). There was no statistically significant difference between NRF-2 overexpression and NRF-2 overexpression ϩ TTX for Cox7c (p ϭ 0.061) or Bdnf (p ϭ 0.063).  . Promoter mutational analysis shows relative luciferase activity of wild type (wt) and mutant (mut) promoters of Bdnf exon IX. COX6b served as the positive control. Mutating the NRF-2 binding site in COX6b, and sites I and II in Bdnf resulted in a significant decrease in luciferase activity as compared with the wild type controls (down to 40% for COX6b mutant and to 62.44 and 52% for Bdnf sites I and II mutants, respectively). n ϭ 3 for each construct. **, p Ͻ 0.01; *, p Ͻ 0.05 as compared with the respective wild type controls.
Homology-The functional NRF-2 binding sites (sites I and site II) are conserved between rats and mice ( Fig. 2) and only partially with humans (data not shown). Although homology exists for sites III and IV, they were not functional for NRF-2.

Discussion
Using multiple techniques, including bioinformatics, biochemical, and molecular biological approaches, the present study documented the functional role of NRF-2 in the transcriptional regulation of the single coding exon, exon IX, of the Bdnf gene in rat primary visual cortical neurons. The expression of Bdnf exon IX transcripts was at least 10-fold higher than all of the other exons. Knocking down NRF-2 significantly down-regulated BDNF protein and mRNA levels, whereas overexpressing NRF-2 markedly up-regulated these levels. Moreover, activity-dependent Bdnf expression was significantly altered by the prevalence of NRF-2. NRF-2 tandem binding sites on the promoter of exon IX are conserved between mice and rats.
BDNF is the second neurotrophin discovered (31) almost three decades after the discovery of nerve growth factor (32). It plays an important role in both developing and the adult central nervous system, including neurogenesis, survival, differentiation, synaptic transmission, synaptic plasticity, and neural degeneration (33)(34)(35)(36)(37)(38)(39)(40). Decreased BDNF levels in humans have been associated with bipolar disorders, severe depression, schizophrenia, and neurodegenerative diseases, such as Alzheimer's and Huntington's diseases (15,41,42). Given the critical role of BDNF in neuronal development, plasticity, and regeneration, the transcriptional regulation of the Bdnf gene should be of significant importance (43).
The Bdnf gene is unique in that it contains eight non-coding exons (I to VIII), each reportedly has its own promoter and, with tissue-and age-specific expression, may be spliced upstream of a common coding exon, exon IX (15,16). Being the only coding exon, exon IX obviously commands a special dominance in this structural arrangement. However, very little was known about the transcriptional regulation of this exon.
Much of the attention has been paid to the regulation of exon IV (previously known as exon III). The cAMP signaling system phosphorylates and activates the transcription factor, cAMP responsive element-binding protein, and stimulates the transcription of exon IV (15,45,46). Other transcription factors, such as basic HLHB2 and NFB, have also been implicated in the expression of exon IV (47,48). Little attention has been paid to the regulation of exon IX, except for a reported epigenetic regulation by Gadd45b, a neural immediate early gene, which promotes activity-dependent demethylation of the exon IX promoter (49). Virtually nothing is known about tran- Previous reports on the expression levels of various exons have noted the presence of exons I, IV, and IXA in the rat brain (28). However, they did not analyze the expression of exon IX. Our quantitative analysis of transcript levels of all exons in the rat visual cortex indicates that exon IX, by far, has the highest expression, and all others, including exon IXA, are expressed at a much lower level (Fig. 1). This is consistent with findings by Barde's group (50) that the mature BDNF is 10-fold more abundant than the pro-BDNF, and that the pro-BDNF is a transient biosynthetic intermediate that is rapidly converted to mature BDNF intracellularly in neurons. The pro-BDNF is also an apoptotic ligand that binds to the nonspecific p75 NTR (51), whereas the mature BDNF is known to bind to the high-affinity TrkB receptors and is pro-survival (52). The greater prevalence of exon IX transcript underscores its functional significance, and yet so little was known about its transcriptional regulation.
Our in silico analysis of the promoter region of exon IX indicated multiple GGAA motifs for the binding of NRF-2. Based on our previous experience, we focused only on tandem repeats of GGAA or TTCC. We concentrated on four such repeat sites in the more proximal promoter region of exon IX, and found that only two of these sites are functional. Site I is located in the region between Ϫ15 to Ϫ42, whereas site II is located in the region between Ϫ301 to Ϫ331, both upstream of the translation start site of exon IX (Fig. 2). The other two sites are much more upstream to the first two sites, and they did not exhibit any functional binding. Thus, sites I and II may be considered as part of the basal promoter region of exon IX, and they are both located in exon IXA. Both sites I and II bound to NRF-2 in vitro (EMSA) and functionally in vivo (ChIP). Promoter mutational analysis also confirmed their functionality.
NRF-2 is an ETS-related transcription factor, and is a human homologue of the murine GA-binding protein, GABP (21). NRF-2 has two distinct subunits, NRF-2␣ and NRF-2␤. The ␣ subunit contains the DNA binding domain, whereas the ␤ subunit has the transactivation domain (reviewed in Ref. 53). Functional NRF-2 is a tetramer of ␣ 2 ␤ 2 . NRF-2 regulates the expression of many genes in various tissues, including genes for mitochondrial biogenesis (21,22,25), lineage-restricted myeloid and lymphoid genes required for innate immunity (23), and genes for cell cycle control in fibroblasts (24). The brain is a highly metabolic organ that needs a constant supply of energy to perform critical steps of neuronal activity. Disruptions of metabolic homeostasis can have dramatic consequences for information processing and cognitive function (54). Thus, the effective control of mitochondrial biogenesis and turnover is critical for the maintenance of energy homeostasis in the brain. Our laboratory has shown that NRF-2 regulates all 10 nuclear-encoded and all three mitochondrial-encoded subunit genes of cytochrome c oxidase, an important energy generating enzyme in the mitochondria (25). Moreover, we found that NRF-2 links energy metabolism and neuronal activity at the transcriptional level by also regulating specific subunits of glutamatergic receptors (26,27). The present study adds a third component to this coupling, i.e. BDNF (Fig. 8). BDNF functions at the interface of metabolism and synaptic plasticity. On the one hand, it interacts with metabolic molecules (55), influences mitochondrial energy management efficiency (56,57), and alters glucose utilization in neurons (58). On the other hand, it enhances glutamate release and increases the abundance of NMDA and AMPA receptors and their delivery to the plasma membrane, thereby up-regulating receptor activity in hippocampal and cortical neurons (59 -61). Previously, we have found that NRF-2 functionally regulates key NMDA and AMPA receptor subunit genes (26,27). With our current discovery of direct regulation of BDNF by NRF-2 in visual cortical neurons, it is highly possible that the actions of BDNF on energy metabolism and glutamatergic synapses result from upstream activation by NRF-2.
It has been reported that BDNF-TrkB signaling plays a pivotal role on mitochondrial bioenergetics through metabolic activators like PGC-1␣ (peroxisome proliferator-activated receptor-␥ co-activator 1␣) (54), a master regulator of mitochondrial biogenesis (62). PGC-1␣ is known to powerfully induce NRF-2 (63). We have also found that PGC-1␣ responds to changes in neuronal activity earlier than that of NRF-2 (64). Thus, the coupling of neuronal activity, energy metabolism, and BDNF at the transcriptional level by NRF-2, with upstream induction likely by PGC-1␣, is an efficient mechanism adopted by neurons for proper functioning, at least in the rat visual cortex.

Experimental Procedures
Experiments involving animals were approved by and conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) of the Medical College of Wisconsin (Milwaukee, WI). All efforts were made to reduce the number of animals used and their suffering.
RNA Isolation and RT-qPCR of Visual Cortical Tissue-Visual cortices dissected from P1 rat pup brains were collected in microcentrifuge tubes kept in dry ice, homogenized immediately, and followed by the extraction of total RNA by the TRIzol method according to the manufacturer's instructions (Life Technologies). Any residual genomic DNA was digested with DNase I on total RNA, and cDNA synthesis was carried out using the iScript cDNA synthesis kit (170-8891, Bio-Rad).
Quantification of Gene Expression by RT-qPCR-mRNA levels of Bdnf exons I to IX were determined in a Bio-Rad iCycler using IQ SYBR Green Supermix and/or a Cepheid Smart Cycler Detection system (Cepheid, Sunnyvale, CA) following the manufacturer's protocols. Primers for real-time are shown in Table  1. PCR runs were: hot start 2 min at 95°C, denaturation 10 s at 95°C, annealing 15 s according to the T m of each primer, and extension 10 s at 72°C for 15-30 cycles. Melt curve analyses verified the formation of a single desired PCR product. Mouse/ rat compatible Actb (␤-actin) and Gapdh were used as internal controls and the 2 Ϫ⌬⌬CT method (65) was performed for the relative amount of transcripts.
Primary Neuronal Cultures-Primary visual cortical neuronal cultures were done as described previously (66). Briefly, P0 to P1 rat pups were euthanized by decapitation. After clearing the meningeal layers from the brain, visual cortices were dissected out and chopped into small pieces. The tissues were treated with trypsin and suspended by pipetting. Neurons were then dissociated by trituration, and cells were seeded in a sixwell plate (35 mm; pre-coated with poly-L-lysine) at a density of 1 ϫ 10 6 cells/well. Cells were grown in Neurobasal-A media containing L-glutamine and B27 supplement (Life Technologies) and maintained in a humidified incubator with 5% CO 2 at 37°C. The addition of cytosine arabinoside (Ara-C) (Sigma) to the culture media helped to suppress the proliferation of glial/ non-neuronal cells.
In Silico Analysis of Bdnf Exon IX-In silico analysis of Bdnf exon IX promoter was done using sequence builder and Genequest software (DNAStar Lasergene 8 Suite). Sequences including 1 kb upstream and 1 kb downstream of the TSPs of mouse, rat, and human BDNF were derived from the genome database in GenBank TM . Alignment was done to find the conserved sites among mice, rats, and humans using Megalign. Computer-assisted search for the typical NRF-2 tandem repeats GGAA and TTCC was conducted to reveal putative NRF-2 binding sites.
Preparation of Nuclear Extracts and Electrophoretic Mobility Shift and Supershift Assays-Nuclear extraction protocol was as reported previously (67). In brief, the frozen brain tissue was rinsed in the PBS buffer, spun, and transferred to a Dounce tissue homogenizer. Buffer A was added at 2.5 times/g of tissue weight with a minimum volume of 2 ml. Five strokes of a pestle were used to homogenize the tissue to a liquid mass. After induction with Nonidet P-40 (0.5%), five additional strokes were applied and the tissue was incubated for 10 min to lyse the cells with the detergent. The homogenate was transferred equally into fresh tubes and centrifuged at 13,000 rpm for 30 s. The supernatant containing the cytoplasmic constituents was removed, and buffer C was added to the nuclear pellet. After thorough mixing of the content, the tubes were placed on a rotary shaker for 15 min, then centrifuged. The supernatant containing the protein from the nuclear extract was carefully removed and transferred to a fresh tube. The protein was measured, aliquoted, and stored at Ϫ80°C until use.

JOURNAL OF BIOLOGICAL CHEMISTRY 22589
The protocols for performing EMSA and supershift assays were as described previously (25). Based on the in silico analysis, oligonucleotide probes with NRF-2 tandem binding sites on each promoter were synthesized (Table 2), annealed, and labeled by a fill-in reaction with Klenow fragment and [ 32 P]dATP (50 Ci/200 ng; PerkinElmer Life Sciences). Each labeled probe was incubated with 2 g of calf thymus DNA and 5 g of brain nuclear extract. Supershift assays were conducted using GABP␣ (NRF-2␣) polyclonal antibodies (GABP␣, H-180, SC-22810, Santa Cruz Biotechnology). The reactions were subjected to an incubation of 20 min with the antibody at room temperature. For competition, a 100-fold excess of unlabeled oligonucleotides were incubated with brain nuclear extract before adding labeled or nonspecific oligonucleotides. Shift reactions were loaded onto 4% polyacrylamide gel and run at 200 V for 3 h in 0.25ϫ TBE buffer. Gels were dried and exposed for 4 -24 h at Ϫ80°C. COX6b with a known NRF-2 binding site was designed as previously described (25) and was used as a positive control. NRF-2 mutants with mutated sequences were used as negative controls.
ChIP Assays-The protocol was as described previously (26) with minor modifications. The procedure was done with visual cortical tissues kept on ice at all times to avoid protein degradation. Tissue was cut in a Petri dish resting on a block of dry ice and then chopped into tiny pieces, fixed in 1% formaldehyde, and resuspended in the swelling buffer (85 mM KCl, 5 mM PIPES, pH 8.0, 1% Nonidet P-40, and protease inhibitors). The tissue was homogenized using a Dounce homogenizer and centrifuged to isolate the nuclei. The nuclei were resuspended and sonicated in SDS lysis buffer (1% SDS, 50 mM Tris-HCl, pH 8.1, 10 mM EDTA). Immunoprecipitation was performed with 2 g of anti-GABP␣ (NRF-2␣) polyclonal antibodies. Two g of anti-NGFR antibodies (sc-6188, Santa Cruz Biotechnology) or no antibody blanks were used as negative controls. Based on our in silico analysis, putative NRF-2 binding sites were identified and semi-quantitative PCR was performed by utilizing primers surrounding the NRF-2 binding sites. Table 3 shows the PCR primers used for sites I-IV (see Fig. 2) of Bdnf exon IX promoter, as well as those for the positive control COX6b and the negative control exon 8 of NRF-1. PCR additives (betaine) and cycling parameters were used to improve the reproducibility and quality of ChIP. PCR products were run on 2% agarose gels to visualize the efficiency.
Luciferase Reporter Assay-Luciferase reporter constructs of the promoter of Bdnf exon IX were made by PCR cloning the proximal promoter sequences, using genomic DNA prepared from N2a cells as template. Digestion with restriction enzymes MluI and BglII was performed, followed by ligation of the product directionally into pGL3 basic vector (Promega, Madison, WI). Sequences of primers used for PCR cloning are provided (Tables 4). Site-directed mutations of NRF-2 tandem binding sites in the promoter of IX were generated using QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Wild type and mutant COX6b were used as positive controls. Primers used for mutagenesis are listed in Table 5. All constructs were verified by sequencing.
Neuroblastoma cells were plated into 24-well plates before transfection. Each promoter construct was transfected into N2a cells using Lipofectamine 2000 (Invitrogen). Each well received 0.6 g of reporter construct and 0.06 g of pRL-TK Renilla luciferase vector (68). Firefly luciferase and Renilla luciferase activities were measured sequentially using the Dual Luciferase Reporter Assay System (Promega). The firefly luciferase activity was normalized according to Renilla and expressed as relative luciferase units to reflect the promoter activity.   Plasmid Construction for NRF-2␣ Knockdown, Transfection, and KCl Treatment-To knockdown NRF-2␣ expression, two target sequences of a vector-based shRNA (5Ј-ATTGC-CCAGCCAGTCACG-3Ј and 5Ј-AGAAGACAGAAGTT-CACCG-3Ј) were cloned into the pBS/U6 parent vector. The PLKO.1-puro-CMV-TurboGFP positive control vector (SHC003, Sigma) containing TurboGFP and puromycin resistance was used to visualize transfection efficiency and selection. The pLK0.1 non-mammalian shRNA vector, which has a scrambled shRNA sequence with no known mammalian gene targets, was used as the negative control (SHC002, Sigma).
Transfection of primary neurons was carried out 4 days postplating with both NRF-2␣ shRNA constructs (2 g of each construct) or the pLKO.1 non-mammalian control (2 g), using Neurofect transfection reagent per 6-well plate according to the manufacturer's instructions (Genlantis). TurboGFP (0.5 g) vector was added to each well for transfection visualization and selection efficiency. Transfection efficiency was around 50 -60% before selection. Puromycin selection, however, effectively yielded 100% transfected cells. Transfection efficiency was observed using green fluorescence. Primary neurons transfected with shRNA against NRF-2␣ were further stimulated with KCl. A final concentration of 20 mM KCl was added to the culture medium for 5 h according to our published method (68). After this period, cells were harvested for RNA and protein isolation.
NRF-2 Overexpression and TTX Treatment-NRF-2␣ and NRF-2␤ subunit expressing vectors were constructed by PCR cloning the human NRF-2␣ and NRF-2␤ from HeLa cell cDNA and human skeletal muscle cDNA library, respectively, as described earlier (69) ( Table 6). The primer pairs used to amplify NRF-2␣ and NRF-2␤ had added HindIII and KpnI and NotI/BamHI restriction sites to their products, respectively. Amplification was done with Taq polymerase and products were cloned into pGEMT-EZ using TA cloning kit (Promega).
As mentioned above, the transfection protocol for overexpressing NRF-2 in primary cultures was similar to that described for shRNA. Either 2 g of NRF-2 overexpression vector, or 1.5 g of the pcDNA3.1 empty vector and 0.5 g of TurboGFP vector for green fluorescence were used for primary neuronal cultures. Transfected primary neurons were blocked for 3 days with TTX at a final concentration of 0.4 M and starting on the day after transfection as previously described (68). Two to 4 days after transfection, neurons were harvested for RNA and protein isolation.
Transfected neurons were washed with PBS, followed by the extraction of total RNA by the TRIzol method according to the manufacturer's instructions and as mentioned above. Any residual genomic DNA was digested with DNase I on total RNA, and cDNA synthesis was carried out using an iScript cDNA synthesis kit. mRNA levels of various genes were determined in a Bio-Rad iCycler using IQ SYBR Green Supermix and/or a Cepheid Smart Cycler Detection system as described above. Primers for real-time are shown in Table 1.
Statistical Analysis-Data were analyzed using analysis of variance. Significance between two groups was analyzed by Student's t test. p values of 0.05 or less were considered significant.
Author Contributions-B. N. conducted all of the experiments, analyzed the results, and wrote most of the paper. M. W. R. conceived the idea for the project, oversaw the experiments, helped with the analysis of the data, and wrote the paper with B. N.