Repressor Element Silencing Transcription Factor/Neuron-restrictive Silencing Factor (REST/NRSF) Can Act as an Enhancer as Well as a Repressor of Corticotropin-releasing Hormone Gene Transcription*

The repressor element-1/neuron-restrictive silencing element (RE-1/NRSE) mediates transcriptional repression by the repressor element silencing transcription factor/neuron-restrictive silencing factor (REST/NRSF) in many neuron-specific genes. REST/NRSF is expressed most highly in non-neural tissues, where it is thought to repress gene transcription, but is also found in developing neurons and at low levels in the brain. Its null mutation in vivoresults in embryonic lethality in mice. While the RE-1/NRSE-mediated repressive influence of REST/NRSF is well established, results in transgenic studies have suggested that the action of the system is more complex. Here, we report that transcription of the corticotropin releasing hormone (CRH) gene is regulated by REST/NRSF, in part through the RE-1/NRSE. Expression of transfected Crh-luciferase constructs was down-regulated by REST/NRSF in a RE-1/NRSE-dependent fashion in both muscle-derived L6 and REST/NRSF co-transfected neuronal PC12 cells. Treatment of L6 cells with trichostatin A revealed that REST/NRSF repression depends, in part, on histone deacetylase activity in these cells. In another neuronal cell line, NG108, REST/NRSF also repressed expression from constructs containing an intact RE-1/NRSE. However, unexpectedly, REST/NRSF up-regulated expression levels of constructs lacking an intact RE-1/NRSE. These results suggest that REST/NRSF can act as both a repressor of Crh transcription, via the CrhRE-1/NRSE, and an enhancer of Crh transcription, via a mechanism independent of the Crh RE-1/NRSE.

identified in over 30 genes, most of which are expressed in neurons. The repressor element silencing transcription factor (REST/NRSF) has been shown in a variety of genetic contexts to repress transcriptional activity via binding to this element (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18). It was originally thought that the REST/NRSF-RE-1/ NRSE system served as a molecular switch that helped distinguish neural from non-neural cell types, as the repression was thought to occur in non-neural cells, which contain REST/ NRSF, but not in neural cells, which either lack or contain only relatively low levels of REST/NRSF (6,10,13,16). It has since become clear that REST/NRSF-RE-1/NRSE function is more complex. For example, multiple splice variants of REST/NRSF have been identified, including at least one neural-specific form (19). Additionally, several groups have suggested that REST/ NRSF and/or the RE-1/NRSE may serve a dual function, as either repressor or activator, depending on the spatial and temporal context of its expression (1,2,19,20).
REST/NRSF is thought to contain two main repressor domains (21). Recently, several reports have suggested that REST/NRSF can repress, in part, through association of a domain in its N-terminal with the Sin3A-histone deacetylase 2 (HDAC2) complex (22)(23)(24). Sin3A⅐HDAC2 is one of several histone deacetylase complexes thought to down-regulate gene expression. Histone deacetylation is thought to lead to a "closed" conformation of the chromatin surrounding a gene, thereby restricting access of the RNA polymerase II complex and, thus, repressing transcription (25,26). Using trichostatin A (TSA), a HDAC-specific inhibitor that has been shown to disrupt histone deacetylation activity in cell cultures (27), several groups have demonstrated the importance of HDAC activity to REST/NRSF's repressive function (22)(23)(24)28). The nuclear receptor corepressor (N-CoR), which can interact with Sin3A and other HDACs, including HDAC 3, 4, 5, and 7, has also been shown to mediate REST/NRSF repression in certain genetic contexts (29). REST/NRSF's C-terminal zinc finger has been shown to associate with at least one other factor, CoREST, to mediate a histone deacetylase-independent repressor activity (22,23,30).
In this study, we evaluate the possible function of the REST/ NRSF-RE-1/NRSE system in the transcriptional regulation of the corticotropin releasing hormone (CRH) gene. CRH is a 41-amino acid hypothalamic neuropeptide that plays a central, well characterized role in development and homeostatic maintenance, modulating the hypothalamic-pituitary-adrenal axis in response to stress (31)(32)(33). In this role, secreted CRH stimulates the production of adrenocorticotropic hormone (ACTH) in the anterior pituitary gland, which, in turn, stimulates the production of glucocorticoids in the adrenal cortex. Glucocorti-coids mediate a range of physiological responses to stress (34). Abnormalities of CRH expression or function have been suggested to play a role in many disorders, including Cushing disease, depression, and anorexia nervosa (35)(36)(37). It also may be involved in modulating immune responses (38).
The RE-1/NRSE sequence found in the first intron of the CRH gene corresponds very closely to the RE-1/NRSE element previously shown to mediate transcriptional repression in other genes (14). Finding that the RE-1/NRSE mediates an important regulatory function for Crh transcription would be particularly interesting, as most in vitro studies to date on Crh transcriptional regulation have relied on constructs of the Crh promoter lacking the intron and, therefore, lacking the RE-1/ NRSE element (39 -44).
In order to examine this possible function, we first show that REST/NRSF binds to the Crh RE-1/NRSE and that this binding is both specific and saturable. We then demonstrate that transcriptional repression of the CRH gene by REST/NRSF requires an intact RE-1/NRSE, both in non-neuronal L6 rat myoblast cells, as well as in the neuronal PC12 rat pheochromocytoma and NG108-15 rat neuroblastoma-glioma cell types. Consistent with previous reports, we show that the Crh transcriptional repression observed in L6 cells is TSA-sensitive, and therefore likely to be linked to HDAC activity, but that this mechanism does not account for the full degree of repression (22)(23)(24)30). We also report the novel finding that REST/NRSF can function as a RE-1/NRSE-independent enhancer of Crh transcription in NG108 cells. Together, our results suggest that REST/NRSF-RE-1/NRSE activity is quite complex, and both genetic and cellular contexts appear to be critical for determining the net effect on Crh transcriptional activity.

EXPERIMENTAL PROCEDURES
Cell Culture-Rat myoblast L6 cells (American Type Culture Collection (ATCC), Rockville, MD) were maintained in Dulbecco's modified Eagle's medium (Sigma) containing 10% fetal calf serum (Life Technologies, Inc.), and 1 mM sodium pyruvate (Sigma), rat PC12 cells (kind gift of M. Greenberg, Harvard Medical School and Children's Hospital, Boston, MA) in Dulbecco's modified Eagle's medium containing 10% heat-inactivated horse serum (Life Technologies, Inc.), 5% fetal calf serum, and 1% filter-sterilized L-glutamine (Sigma), and rat NG108 cells (kind gift of Ian C. Wood, Wellcome Laboratory for Molecular Pharmacology, University College London and School of Biochemistry and Molecular Biology, University of Leeds, United Kingdom) in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. All cell cultures also contained 1% penicillin/streptomycin/amphotericin B (Life Technologies, Inc.). PC12 cells were grown on poly-L-lysine (Sigma)-treated plates and were passaged a maximum of five times before discarding. Each experiment was conducted using cells of the same passage.
RT-PCR-Total RNA was prepared from cultured L6 cells, PC12 cells treated with 50 ng/ml NGF (Sigma) for 48 h, and NG108-15 cells treated with 10 M forskolin, 100 M 3-isobutylmethylxanthine (IBMX) (Sigma) for 6 h, using TRI Reagent (Sigma) according to manufacturer's protocol. The recovery of RNA was quantified spectrophotometrically. RNA was digested using DNA-free TM (Ambion), according to the manufacturer's protocol, to eliminate contaminating endogenous DNA. Single-stranded cDNA was synthesized using Superscript II TM reverse transcriptase (RT) (Life Technologies, Inc.) according to the manufacturer's protocol, using random hexamers [(pd(N) 6 ] (Roche Molecular Biochemicals) as primers. Following an initial 3 min at 96°C, each PCR cycle (MJ Research) consisted of 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min, followed by a 5-min extension at 72°C. PCR was also conducted on identical samples lacking RT treatment, to control for any remaining endogenous DNA contamination. PCR reactions were performed on 1/10 of the RT reaction, in 1 ϫ PCR buffer (Roche Molecular Biochemicals) using 2 units of Taq enzyme (Roche Molecular Biochemicals), 0.4 mM dNTP, and 0.2 M each primer. The PCR products were separated and visualized in an ethidium bromide containing 1.5% TBE-agarose gel, using a 100-bp DNA ladder (Roche Molecular Biochemicals). The sense (S) and antisense (AS) REST/NRSF-specific and actin-specific (control) primers used were as follows: rNRSF-S, GACAGGTTCACAACGGGCC; rNRSF-AS, CCCTTCGGCACTTCGC-CGCT; actin-S, ATTCCTATGTGGGCGACGAG; and actin-AS, TGGAT-AGCAACGTACATGGC (45). The primers, rNRSF-S and rNRSF-AS, were intron-spanning, such that RT-PCR utilization of mRNA versus endogenous DNA templates was distinguishable. The primers were specific for full-length REST/NRSF, as distinguished from splice variants such as REST4.
Electrophoretic Mobility Shift Assays (EMSAs)-Nuclear extracts from L6 cells were prepared according to Schreiber et al. (46) from ϳ10 6 cells per extract; Buffer A contained a final concentration of 1.0 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotenin, 0.5 g/ml leupeptin, and 0.7 g/ml pepstatin, and Buffer C was the same except for phenylmethylsulfonyl fluoride, which was at 1.0 mM. Final nuclear extract protein concentration was quantitated by spectrophotometry using a standard curve (BCA Protein Assay Reagent, Pierce). All oligonucleotide probes used were double stranded. They were synthesized as complementary single strands (BIOSOURC-E/Keystone) which were then annealed by equimolar mixing, boiling for 5 min and slow cooling to room temperature. Radiolabeled probe was made by 5Ј phosphorylation of 1 pmol of the double-stranded (ds) RE-1/NRSE probe with 50 pmol of [␥-32 P]ATP (product number BLU-2A, PerkinElmer Life Sciences) by the T4 polynucleotide kinase forward reaction lasting 30 min at 37°C, in a total volume of 50 l. Labeled RE-1/NRSE probe was purified by centrifugation through a Sephadex G-50 QuickSpin column (or G-25 column, for the CRE probe) (Roche Molecular Biochemicals). Final counts for different experiments were in the range of 6 ϫ 10 4 to 3 ϫ 10 5 cpm/l, and between 11 and 14 fmol of labeled probe were used per lane. Both the RE-1/NRSE probe and the various competitor probes (except for the unrelated CRE probe) contained EcoRI 5Ј overhangs for possible use in subcloning. In addition to the 21-bp core of the RE-1/NRSE, we included 9 bp of 5Ј-flanking DNA and 7 bp of 3Ј-flanking DNA; this 37-bp sequence was doubled in tandem. For the imperfectly conserved potential RE-1/NRSE (p(RE-1)) sequences, we noted the outside limits of conservation (i.e. the most 3Ј and most 5Ј conserved bases of the RE-1/NRSE) and counted out both 5Ј and 3Ј to include an equivalent amount of flanking DNA; some of these probes are therefore longer and some are shorter than the RE-1/NRSE probe. The sequences of the probes used were as follows (only sense strand shown): [␥-32 P]RE-1 probe, AATTCGTACCTAGCTTCAGCACC-GCGGACAGCGTCACCGAAGTACCTAGCTTCAGCACCGCGGACAG-AGTCACCGAAG; RE-1 MUT (with mutation underlined in bold), AATT-CGTACCTAGCTTCAGCACCGCTTACAGCGTCACCGAAGTACCTA-GCTTCAGCACCGCTTACAGAGTCACCGAAG; p(RE-1) A , AATTCGA-GCAGAGGCAGCACGCAATCGAGCTGTCAAGAGAGCGTCAGCTTA-GAGCAGAGGCAGCACGCAATCGAGCTGTCAAGAGAGCGTCAGCT-TAG; p(RE-1) B , AATTCGCTGCTGTGGTGAGCCCCGGAGCCAGCTG-CCCATGTGCTGCTGTGGTGAGCCCCGGAGCCAGCTGCCCATGTG; p(RE-1)C, AATTCTGCAGATGCCTCAGCGCTCGCTCGACAGCCGCG-CGGAGTGCAGATGCCTCAGCGCTCGCTCGACAGCCGCGCGGAG; p(RE-1) D , AATTCTGTATTTCTGTGTCGTAACAAAACAGCGTTATTT-GTTGTATTTCTGTGTCGTAACAAAACAGCGTTATTTGTG; Crh CRE, GCTCGTTGACGTCACCAAGA. Binding buffer (1 ϫ) was: 20 mM HEPES (pH 8), 0.1% Nonidet P-40, 10% glycerol, 1 mM dithiothreitol, 2.5 mM MgCl 2 , 250 mM KCl. Poly(dI-dC) (Sigma) was used in binding reaction at a concentration of 0.125 g/ml. Polyclonal REST antibody (kind gift of Gail Mandel, Howard Hughes Medical Institute, Department of Biochemistry and Cell Biology, State University of New York, Stony Brook, NY) was used at a 1:100 dilution, ϳ416 ng/reaction. The control cAMP-response element-binding protein antibody is described in Waeber et al. (47). Binding buffer, poly(dI-dC), and double-distilled (dd) H 2 O were combined with nuclear extracts in 1.5-ml Eppendorf tubes and put on ice for 5 min. Competitor oligonucleotides and/or antibodies were added, and the tubes were placed at 4°C for 10 min. ␥-32 P-Labeled RE-1/NRSE probe was then added and the reaction was incubated for 15 min at 4°C. The reactions were then loaded onto a 5% polyacrylamide gel (Proto-Gel, Minneapolis, MN) and run at 175 mV for roughly 3 h at 4°C. The gels were then vacuum-heat dried and bands were visualized by exposure to autoradiographic film (Kodak) for 3 to 6 h at Ϫ80°C.
Plasmids-All primary reporter plasmids were cloned into the pXP2 luciferase backbone (ATTC) (48). All cloning transformations were performed using Subcloning Efficiency DH5␣ cells according to manufacturer's protocol (Life Technologies, Inc.). Source constructs for Crh sequence were: CRH5.0, containing 5 kb of 5Ј promoter sequence (B57 mouse) cloned into Bluescript (SKII ϩ ) (Stratagene); Xho-Kpn-Kpn, containing the full-length CRH gene plus ϳ6-kb flanking sequence (B57 mouse) (49). Full-length constructs (Intron(ϩ) WT , Intron(ϩ) MUT ) were made as follows: we excised a fragment containing roughly 5 kb of 5Ј Crh sequence from CRH 5.0 by cutting at a BamHI site 5 kb upstream from the promoter and at the unique SalI site at position 409 in the promoter. A 3Ј contiguous 712-bp SalI-KpnI fragment was then excised from the Xho-Kpn-Kpn construct, cutting at the same SalI site and a KpnI site at position 1121. A third contiguous fragment was created by using forward primer starting at (5Ј position 1097) CGCACCAGTTG-GAGCTTTGCAGGT (3Ј), and reverse primer, (5Ј position 1292) TTTT-TTTTAGATCTTAGGGGCGCTCTCTGAAGC (3Ј), to amplify the sequence flanking the KpnI (1117) site (see underlined, bold below) and up to the translational start site (1292) by PCR (same reagents as above, except using 30 pmol of each primer on 616 ng of template DNA), starting with 96°C for 2 min and then 24 cycles of: 60°C, 1 min; 72°C, 2 min; 96°C, 30 s; followed by a final extension at 72°C for 4 min. Another forward primer was used to introduce the GG/TT mutation (italics, bold) in the RE-1/NRSE at position 1138 -39: (5Ј position 1110) CTTTGCAGG TAC CTAGCTTCAGCACCGCTTACAGCGTCACCG (3Ј). We attached a BglII site at the 5Ј end of the reverse primer, with 8 Ts extending to allow for eventual cutting by BglII and cloning into the BglII site of pXP2. The resulting PCR fragments were digested with KpnI and BglII. We then performed a double-ligation of the 712-bp SalI-Kpn and 171-bp KpnI-BglII fragments into SalI/BglII-digested pXP2. The process was repeated, substituting the GG/TT KpnI-BglII fragment. Finally, the resulting two vectors were digested with BamHI and SalI to allow for insertion of the 5-kb BamHI-SalI fragment to result in the final full-length constructs, Intron(ϩ) WT and Intron(ϩ) MUT . Intron-less constructs were made by cutting the Intron(ϩ) WT construct at the SbfI site at position 541 and the BglII site at position 1310 to excise the intron. Forward and reverse strands of a linker containing the Crh RE-1/NRSE (including flanking sequence, as used in the EMSAs) element and containing a 5Ј SbfI half-site were synthesized (BIOSOURCE/Keystone): (forward) GGAGGCATCCTGA-GATCTGTACCTAGCTTCAGCACCGCGGACAGCGTCACCGAA; (reverse) GATCTTCGGTGACGCTGTCCGCGGTGCTGAAGCTAGGTAC-AGATCTCAGGATGCCTCCTGCA. As before, the strands were combined, boiled, and slow-annealed (see EMSA methods). The resulting linker thus contained a Crh RE-1/NRSE ϩ flanker DNA cassette (in italics), preceded by a BglII site (5Ј) and followed by a BglII half-site (3Ј). The complete SbfI-BglII linker was cloned into the SbfI/BglII-digested Intron(ϩ) WT construct (intron thus removed), thus replacing the intron with the SbfI-BglII, RE-1/NRSE-containing, linker to make the Intron(Ϫ)RE1(ϩ) construct. We made the Intron(Ϫ)RE1(Ϫ) construct by digesting this new vector with BglII to excise the RE-1/NRSE element, and then religating. The Intron(Ϫ)RE1 MUT construct was created by following the same procedure used to make Intron(Ϫ)RE1(ϩ), simply substituting a GG/TT mutated linker for the one detailed above. All constructs were purified using well established miniprep (50) or maxiprep (Qiagen) procedures and were sequenced across the full promoter and transcribed regions, including all cloning junctions, to verify correct sequence. DNA fragments were isolated from agarose gels using the Qiaex II and QIAquick gel extraction kits (Qiagen).
Luciferase Assays-We transfected all cell lines in 24-well plates, using FuGene6 reagent according to the manufacturer's instructions (Roche Molecular Biochemicals). FuGene:DNA ratio was kept at 3:1 in all experiments. 48 h after transfection, cells were lysed using 1 ϫ Passive Lysis Buffer (Promega). We controlled for total DNA in all co-transfection experiments, replacing REEX-1 with an empty vector, and also controlled for transfection efficiency by co-transfecting a second, thymidine kinase-or cytomegalovirus-driven, reporter in all experiments. Null vector (pXP2) values were subtracted from all measurements. In some experiments, 100 nM TSA was applied for 24 h prior to lysis. Each well in the L6 transfections contained 0.5 g of pRL-TK (Promega) with 0.5 g of firefly reporter vector (full-length constructs) or 0.2 g of pRL-TK with 0.2 g of firefly reporter (intron-less constructs). Signal measurement was done using the Dual Luciferase Assay kit, according to manufacturer's instructions (Promega). PC12 and NG108 transfections were also conducted in 24-well plates. PC12 cells were transfected with 10 ng of cytomegalovirus-CAT (Invitrogen) and 0.2 g of firefly reporter, along with 0.2 g of either pcDNA1.1/amp (Invitrogen) or REEX-1 (REST expression cassette in pcDNA1/amp backbone, kindly provided by Gail Mandel). NGF was applied to PC12 cells at a final concentration of 50 ng/ml, for 48 h prior to lysis by 1 ϫ Passive Lysis Buffer. CAT assays were performed using the Promega kit, according to manufacturer's protocol ( 14 C-labeled chloramphenicol from PerkinElmer Life Sciences). NG108 cells were transfected identically to the PC12 cells, both initially using CAT and then using ␤-galactosidase, to normalize transfection efficiency. Results reported herein used ␤-galactosidase, but no difference was seen when CAT was used. In individual experiments, we used either 0.1 g or 0.01 g of cytomegalovirus-␤-galactosidase (CLONTECH). After confirming that results were identical to using pcDNA1.1/amp, in some experiments we used pBluescript to equalize total DNA in all wells. NG108 cells were treated with 10 M forskolin, 100 M IBMX for 6 h prior to lysis by 1 ϫ Passive Lysis Buffer. ␤-Galactosidase was measured according to manufacturer's instructions, using the Luminescent ␤-galactosidase Detection Kit II (CLONTECH). All firefly luciferase, Renilla luciferase, and ␤-galactosidase readouts were conducted on an signal integrating Berthold LB 9501 luminometer.
Data Analysis-Luciferase expression data are displayed as the means (after normalization to CAT or ␤-galactosidase expression), with standard errors shown in the positive direction. Statistical significance was evaluated using the Student's t test, with the Bon Feronni correction for multiple comparisons, where appropriate.

A Highly Conserved Intronic Element Sharing Close Similarity to an Established Regulatory Element-
The RE-1/NRSE noted by Shoenherr et al. (14) in the first intron of the CRH gene shares considerable similarity, varying by only one nucleotide in human, to a consensus sequence derived from the comparison of RE-1/NRSE elements found in 19 different genes, many of which have been shown to be regulated by REST/NRSF in vitro (14). The RE-1/NRSE element is also highly conserved among the CRH genes of several species, including mouse, the focus of the present study (Fig. 1). In all mammalian species, the Crh RE-1/NRSE is located in the gene's single intron (position 1127 in mouse), ϳ140 bp upstream from the beginning of the second exon, which encodes prepro-CRH. Moreover, the 21-bp element is more highly conserved than the rest of the intron, as well. Outside of the RE-1/NRSE, the Crh intron is roughly 75% conserved (90% rat/mouse; ϳ73% human/rat, mouse; 76% human/sheep; ϳ67% sheep/rat, mouse). The RE-1/NRSE element is 90 -95% identical among all species examined to date, including Xenopus (14). This high degree of conservation, combined with the non-coding, intronic position of the element suggests that it might mediate an important regulatory function in Crh transcription.
REST/NRSF Is Expressed in Non-neuronal L6 Muscle Cells but Not in More Neuronal PC12 or NG108-15 Cells-We performed RT-PCR on L6, NGF-treated PC12, and forskolin/ IBMX-treated NG108 cells to determine their REST/NRSF expression profiles. REST/NRSF expression was only observed in the L6 cells (Fig. 2). We chose these cell types because they have been used extensively in previous studies of REST/NRSF-RE-1/NRSE function and are thus relatively well characterized for this system, with non-neural L6 cells containing high levels of REST/NRSF, and neuronal PC12 and NG108 cells having virtually none (6,16,51). Consistent with the hypothesis that REST/NRSF may repress Crh transcription, we confirmed by Northern blot that Crh expression and REST/NRSF levels are reciprocal, with Crh expression absent in L6 cells and present in PC12 cells (data not shown). The latter finding confirms a prior report of CRH expression in PC12 cells (52).
REST/NRSF Binds Specifically to the Crh RE-1/NRSE-To establish REST/NRSF binding to the Crh RE-1/NRSE, to evaluate the specificity of this binding, and to test its sensitivity to sequence perturbation, we conducted a series of EMSAs. We obtained nuclear extracts from L6 rat myoblast cells and used several different oligonucleotide probes ("Experimental Proce-

FIG. 1. The murine Crh RE-1/NRSE is very similar to a consensus RE-1/NRSE (14)
, and it is highly conserved among at least five species. The high level of similarity and conservation, along with its intronic position, suggests that the Crh RE-1/NRSE is functionally important for Crh transcriptional regulation (modified from Schoenherr et al. (14)).
dures"). Using the Crh RE-1/NRSE probe, we observed binding activity (Fig. 3A, lane 2). This shift was specifically eliminated by a REST-specific polyclonal antibody, but not by a cAMPresponse element-binding protein-specific antibody (Fig. 3A,  compare lanes 3 and 4), and was competed away by a 33-fold molar excess of cold RE-1/NRSE oligonucleotide (Fig. 3A, lane 5). Along with the highly conserved RE-1/NRSE sequence, we also tested a mutant form of the element (RE1 MUT ), containing a GG-to -TT (GG/TT) mutation at its center that has previously been shown to disrupt REST/NRSF binding (10), as an unlabeled competitor (Fig. 3A, lane 11). Additionally, we tested several other potential RE-1/NRSE elements, p(RE-1) A-D , which were found elsewhere in the murine CRH gene and showed some sequence similarity to the RE-1/NRSE (Fig. 3B), for their ability to compete with the conserved RE-1/NRSE for REST/NRSF binding (Fig. 3A, lanes 6 -9). In multiple experiments (n ϭ 3), only the intact RE-1/NRSE was observed to compete for REST/NRSF binding, indicating both that REST/ NRSF binding to the Crh RE-1/NRSE is greatly disrupted by the GG/TT mutation and that the several p(RE-1)s are not likely to be important for REST/NRSF function.
The Crh RE-1/NRSE Is Functionally Important for Crh Transcriptional Regulation-Having demonstrated sequencespecific, saturable binding of REST/NRSF to the Crh RE-1/ NRSE, we next wished to test the possible functional significance of the observed REST/NRSF-RE-1/NRSE association. We made five different luciferase reporter constructs to establish the importance of the Crh RE-1/NRSE and its flanking DNA in Crh transcriptional regulation (Fig. 4). All constructs contain 5 kb of upstream 5Ј-flanking sequence, in addition to the promoter and first, non-coding, exon of the mouse CRH gene. Two of the constructs contain the intron as well, with either an intact (Intron(ϩ) WT ) or a GG/TT-mutated (Intron(ϩ) MUT ) RE-1/NRSE site. An additional three constructs contain no intron and either an intact RE-1/NRSE (Intron(Ϫ)RE1(ϩ)), a GG/TTmutated RE-1/NRSE (Intron(-)RE1 MUT ), or no RE-1/NRSE at all (Intron(Ϫ)RE1(-)) (Fig. 4). All constructs were cloned into a pXP2 firefly luciferase reporter backbone. We transiently transfected L6, PC12, and NG108 cells with these constructs, either alone or with a co-transfected, cytomegalovirus-driven, REST/NRSF expression vector, REEX-1 (21). Each condition was tested multiple times, with either three or, more commonly, four wells per condition.
We first tested the importance of the RE-1/NRSE element, itself, in regulating Crh transcriptional activity. If the RE-1/ NRSE were mediating transcriptional repression by REST/ NRSF, disrupting the RE-1/NRSE in L6 cells would be expected to lead to transcriptional de-repression, since these cells contain high levels of endogenous RE-1/NRSE-binding REST/NRSF activity (6,16). A marked de-repression was indeed observed in these cells upon disruption of the RE-1/NRSE sequence (Fig. 5A). With an intact RE-1/NRSE, both the intron-containing (Fig. 5A, left panel) and the intron-lacking (Fig. 5A, right panel) constructs were essentially silent. In all cases, with the RE-1/NRSE either mutated or deleted altogether, the expression levels were markedly up-regulated. This de-repression ranged from 18-to over 60-fold. In addition, intron-lacking constructs in general were more active (Fig. 5A, note different ordinate scales).
REST/NRSF represses Crh transcription via the RE-1/ NRSE-In order to test for the function of the REST/NRSF protein itself, we transiently transfected PC12 and NG108 cells (which lack REST/NRSF, see Fig. 2) with the above constructs, both with and without co-transfection of the REST expression vector. Undifferentiated PC12 cells did not express any of the Crh constructs used (data not shown). NGF treatment has previously been shown to enhance endogenous CRH production in these cells and to result in a more neuronal phenotype (52,53). We therefore tested NGF-treated PC12 cells and found that they did express the Crh constructs. All subsequent PC12 transfection experiments were consequently conducted using NGF-treated cells. When transfected alone (without REST/ NRSF), expression of intron-containing constructs (Fig. 5B, left   FIG. 2. REST/NRSF mRNA is expressed in L6 muscle cells but not in neuronal PC12 or NG108-15 cells. RT-PCR of RNA extracted from L6, NGF-treated PC12, and forskolin/IBMX-treated NG108-15 cells was performed using an intron-spanning primer pair specific for full-length REST/NRSF, as described under "Experimental Proce- panel) tended to be less than that of constructs lacking the intron (Fig. 5B, right panel; note different ordinate scales). In the absence of added REST/NRSF, disruption of the RE-1/ NRSE resulted in a minimal increase in expression levels (Fig.  5B) compared with the over 20-fold increase observed in L6 cells, consistent with the undetectable amount of REST/NRSF mRNA we observed in the NGF-treated PC12 cells. In the setting of an intact RE-1/NRSE sequence, co-transfection of REST/NRSF appeared to lead to greater suppression in the absence of the intron (86% repression, p Ͻ .00005, for the Intron(Ϫ)RE1(ϩ) construct) than in the presence of the intron (35% repression, p Ͻ .047, for the Intron(ϩ) WT construct) (Fig.  5B). In the presence of the intron, disruption of the RE-1/NRSE element (Intron(ϩ) MUT ) eliminated REST/NRSF-induced transcriptional repression. Disruption or deletion of the RE-1/NRSE element in the intron-lacking constructs (Intron(Ϫ)RE1(Ϫ) and Intron(Ϫ)RE1 MUT ) (Fig. 5B, bars 7-10) also eliminated the significant REST/NRSF-mediated repression observed with an intact RE-1/NRSE (Fig. 5B, bars 5 and 6). These results suggest that REST/NRSF itself functions as a repressor of Crh transcription and, furthermore, that this repression is mediated by the intronic RE-1/NRSE. As noted above, the presence of the intron, independent of the RE-1/NRSE, seemed to reduce both the basal levels of construct activity and the amount of repression seen upon REST/NRSF introduction in PC12 cells. To confirm that this result was not peculiar to one particular line of PC12 cells, which are known to drift genetically upon passaging, two other independent PC12 lines were also similarly tested and yielded essentially identical results (data not shown).
REST/NRSF-RE-1/NRSE-mediated Repression Occurs in Part through a HDAC-dependent Mechanism-In order to evaluate deacetylation as a possible mechanism for REST/NRSF repression of the CRH gene, we treated L6 cells with 100 nM TSA for 24 h and looked for disruption of the previously observed RE-1/NRSE-mediated repression. As shown in Fig. 5C, TSA treatment led to significant de-repression only for the constructs containing an intact RE-1/NRSE (bars 1 and 2, 5  and 6). Intron(ϩ) WT expression was de-repressed by roughly 14-fold while Intron(ϩ) MUT expression remained unchanged.
Similarly, Intron(Ϫ)RE1(ϩ) expression was boosted by almost 4-fold while Intron(Ϫ)RE1(Ϫ) expression was not de-repressed, and actually fell slightly. These results suggest that the HDACmediated repression relieved by TSA treatment depends on the presence of an intact RE-1/NRSE, irrespective of the presence FIG. 5. REST/NRSF represses Crh transcription through a RE-1/NRSE-dependent mechanism that is partially HDAC dependent. A, in L6 cells, disruption or deletion of the RE-1/NRSE results in a significant de-repression (p Ͻ .0001, using the Bon Feronni t test correction for multiple comparisons), both in the presence (left panel) and absence (right panel) of the intron. The intron appears to lower the overall level of expression in these cells (note different ordinate scales of the left and right panels) (n ϭ 4 replicates per condition). B, in PC12 cells treated with 50 ng/ml NGF, expression is significantly repressed by the introduction of REST/NRSF when the RE-1/NRSE is intact (bar 2, p Ͻ .05; bar 6, p Ͻ .00005) but not when the RE-1/NRSE is either disrupted (bars 4 and 8) or deleted (bar 10). As with the L6 cells, the presence of the intron appears to result in an overall lowering of expression levels (note different ordinate scales of the left and right panels). The intron also appears to limit the degree of RE-1/NRSEmediated transcriptional repression by REST/NRSF (n ϭ 4 replicates per condition). C, 100 nM TSA treatment partially relieves RE-1/NRSEmediated repression in L6 cells, regardless of intron context. TSA treatment led to significant partial de-repression of Crh-luciferase construct expression when the RE-1 was intact (bars 1 and 2, p Ͻ .004; and bars 5 and 6, p Ͻ .002) but not when the RE-1 was disrupted (bars 3 and 4 and bars 7 and 8). This RE-1 dependence was observed both with and without the intron present and suggests that RE-1 mediated REST repression in L6 cells occurs, in part, through a HDAC-dependent mechanism (n ϭ 3 replicates per condition). or absence of the intron. This, in turn, suggests that REST/ NRSF represses CRH gene expression, in part, through a HDACdependent mechanism. The TSA-dependent de-repression, however, does not account for the full degree of de-repression observed when the RE-1/NRSE is disrupted. For example, TSA treatment of Intron(ϩ) WT does not lead to the fully de-repressed level resulting from disruption of the RE-1/NRSE in Intron(ϩ) MUT (Fig. 5C, compare the de-repression between bars 1 and 2 with that between bars 1 and 3). This is not surprising, as REST/NRSF has been previously shown to associate through its C-terminal zinc finger, with a protein called Co-REST to mediate repression by an HDAC-independent mechanism (22,23,30).
REST/NRSF Can Act as an Enhancer of Crh Transcription-NG108 cells can be differentiated to a more neuronal phenotype by treating with 10 M forskolin and 100 M IBMX (54,55). We therefore used forskolin/IBMX-treated NG108 cells to study the function of REST/NRSF-RE-1/NRSE in the setting of another neuronal phenotype (Fig. 6). As in PC12 cells, co-transfected REST/NRSF significantly repressed expression only when an intact RE-1/NRSE was in place, regardless of whether the remainder of the intron was present (Fig. 6, A and B). As with L6 and PC12 cells, repression appeared to be greater (70% repression, p Ͻ .005) in the RE-1/NRSE-containing construct lacking the intron (Fig. 6B, bars 1 and 2) than in the construct containing the intron (28% repression, p Ͻ .004) (Fig. 6A, bars 1 and 2). However, when the RE-1/NRSE was either disrupted (Intron(ϩ) MUT and Intron(Ϫ)RE1 MUT ) or deleted altogether (Intron(Ϫ) RE1(Ϫ)), rather than a lack of repression, we observed a significant 1.2-2.5-fold up-regulation of reporter activity over baseline (compare bars 3 and 4 in Fig. 6A; bars 3 and 4 in Fig. 6B; bars 5 and 6 in Fig. 6B). The presence of the intron appeared to facilitate, rather than mute, this enhancer effect of REST/NRSF on transcription (note different ordinate scales in Fig. 6, A and B), although it is not clear whether this was a specific or nonspecific (spacer or conformational) effect. The up-regulation of transcriptional activity was seen consistently in three independent experiments (data not shown), with both the intron-containing and the intron-lacking constructs. Similar enhancer and repressor effects were also seen in cells untreated with forskolin/IBMX, indicating that PKA stimulation in this context does not change the fundamental properties of the system (data not shown). It appears that REST/NRSF can act as an activator of Crh transcription, independent of the Crh RE-1/NRSE, in addition to acting as a RE-1/ NRSE-dependent repressor of Crh transcription. This enhancer function is unmasked by the disruption of the repressionmediating RE-1/NRSE element. A dual role for REST/NRSF, as an enhancer as well as repressor of gene transcription, has been previously suggested (1,2,20). In this case, the observed effect on Crh transcription could be a direct effect of REST/NRSF or an indirect effect of REST/NRSF, if, for example, another REST/ NRSF target is involved in regulating Crh transcription. The enhancer effect of REST/NRSF seen in NG108 cells is not simply due to their being transfected with this factor, as Crh transcription is not enhanced in L6 cells following similar transfection with REST/NRSF (data not shown). DISCUSSION In L6 and PC12 cells, the REST/NRSF-RE-1/NRSE system appears to function in Crh regulation much as it has been previously shown to function in other genetic contexts. NG108 cells reveal another potential mode of CRH gene regulation by REST/ NRSF. It has become clear, from previous studies, that genetic context is extremely important in determining the function of the REST/NRSF-RE-1/NRSE system (1, 4 -10, 12, 17, 18, 20). A possible explanation for at least a part of this variability is that REST/NRSF may act through independent mechanisms, associating with different co-factors, such as the Sin3A complex, the nuclear receptor corepressor (N-CoR), and CoREST, to repress expression of different REST/NRSF-dependent genes (29). Results from the knockout of REST/NRSF in mouse strongly suggest that tissue/cell-type, as well as genetic context, is important for determining the effect of REST/NRSF on gene expression (2). Our results underscore the fact that the cellular context is critical for understanding and defining the function(s) of the system. Not only is the distinction between neuronal and non-neuronal cell types important, but perhaps equally important is the distinction between different neuronal cell types. Furthermore, various forms of REST/NRSF are expressed in many neurons within the brain (19), suggesting more than a repressive function for REST/ NRSF. This is particularly relevant for a gene such as Crh, which is differentially regulated, at times in opposite directions, in different brain regions and cell types (56,57).
The intron appears to be important for Crh transcriptional regulation in more than one way. The 21-bp RE-1/NRSE element, itself, exerts a clear influence on CRH gene transcription. In L6 cells, which contain a high level of REST/NRSF protein and do not express CRH, the element mediates a potent transcriptional repression of the gene. This repression is substantially relieved by either mutating or deleting the RE-1/ NRSE. In NGF-treated PC12 cells, which express the CRH gene, addition of REST/NRSF results in transcriptional repression of Crh construct expression when an intact RE-1/NRSE is present, but not when the RE-1/NRSE has been mutated or deleted. Treatment of L6 cells with 100 nM TSA confirms that REST/NRSF may repress partially through a HDAC-dependent mechanism. In both cell types, the remainder of the intron (i.e. excluding the RE-1/NRSE) appears to confer a generalized down-regulation of basal expression levels. It is possible that this is due to the presence of an as yet unidentified regulatory element in the intron or simply to a nonspecific, possibly conformational, effect of the intron.
NG108 cells provide another opportunity to study REST/ NRSF action in a neuronal cell type. Consistent with the results obtained in L6 and PC12 cells, the addition of REST/ NRSF to these cells leads to a repression of Crh construct expression when an intact RE-1/NRSE is present, but not when it is either disrupted or deleted. Unlike in PC12 cells, however, disruption or deletion of the RE-1/NRSE element in NG108 cells not only restores Crh construct expression to basal levels but actually leads to a significant enhancement of expression. The intron appears to facilitate this enhancement in NG108 cells, while reducing the amount of RE-1/NRSE-mediated repression by REST/NRSF in both PC12 and NG108 cells.
Our results suggest that, in the context of the CRH gene, REST/NRSF may exert both a RE-1/NRSE-dependent repressive effect and a RE-1/NRSE-independent enhancer effect simultaneously, and it is the balance of these effects that determines REST/NRSF's final impact on transcriptional activity. It is not yet clear whether the RE-1/NRSE-independent activity of REST/NRSF on Crh transcription is direct, acting on some novel element within the gene, or indirect, acting on another gene or factor involved in Crh transcriptional regulation. It is strictly possible that the novel enhancer effect of REST/NRSF is limited to NG108-15 cells. However, it is also possible that it extends to other neuronal cell types and possibly even to nonneuronal cell types. Whether any such observations have physiological correlates can ultimately only be evaluated satisfactorily in vivo. However, these results suggest that, depending on the genetic and cellular contexts, the net effect of REST/ NRSF function can be very different from what its RE-1/NRSEmediated repressor function alone would suggest.