Role for RFX transcription factors in non-neuronal cell-specific inactivation of the microtubule-associated protein MAP1A promoter.

Microtubule-associated protein MAP1A is expressed abundantly in mature neurons and is necessary for maintenance of neuronal morphology and localization of some molecules in association with the microtubule-based cytoskeleton. Previous studies indicated that its complementary expression together with MAP1B during nervous system development is regulated at the transcriptional level and that the mouse Map1A gene is transcribed under the control of 5' and intronic promoters. In this study, we investigated the regulatory mechanisms that govern the neuronal cell-specific activation of the MAP1A 5' promoter. We found that two regulatory factor for X box (RFX) binding sites in exon1 of both the mouse and human genes are important for effective transcriptional repression observed only in non-neuronal cells by reporter assays. Among RFX transcription factor family members, RFX1 and 3 mainly interact with repressive elements in vitro. Cotransfection studies indicated that RFX1, which is expressed ubiquitously, down-regulated the MAP1A 5' promoter activity in non-neuronal cells. Unexpectedly, RFX3, which is abundantly expressed in neuronal cells, down-regulated the transactivity as well, when it was expressed in non-neuronal cells. Both RFX1 and 3 did not down-regulate the transactivity in neuronal cells. These results suggest that RFX1 and 3 are pivotal factors in down-regulation of the MAP1A 5' promoter in non-neuronal cells. The cell type-specific down-regulation, however, does not depend simply on which RFX interacts with the elements, but seems to depend on underlying profound mechanisms.

MAP1A is a high molecular weight MAP 1 that plays an important role in determining neuronal morphology and in determining the balance between plasticity and stability of neuritis (1,2). MAP1A and 1B are particularly abundant filamentous MAPs in the brain and possess highly similar structures and functions in their stabilization of microtubule bands.
Recent studies have demonstrated that MAP1A interacts with postsynaptic density-93, a membrane-associated guanylate kinase (3), and is involved in the general attachment of mRNA to the cytoskeleton (4). Thus, MAP1A seems to act not only in stabilizing microtubule bands but also in regulating the distribution of associated molecules as a linker between them and the microtuble-based cytoskeleton. MAP1B has been shown to link ␥-aminobutyric acid, type C receptors to microtubule (5), indicating that MAP1A and 1B anchor distinct target molecules to the cytoskeleton. The different expression patterns of MAP1A and 1B, which are transcriptionally regulated (6), also support the functional dichotomy of these two filamentous MAPs. MAP1B is the earliest MAP expressed during central nervous system development; it is most highly expressed during the postneonatal period, diminishing as the animal matures to adulthood (6 -9). In contrast, MAP1A expression increases continually during the postneonatal period and is maintained at a high level throughout adulthood (6, 8 -10).
Multiple promoters, cell type-specific cis-regulatory elements and interacting homeoproteins have been reported for the MAP1B gene (11)(12)(13); little is known, however, about the transcriptional regulation of the MAP1A gene. We previously revealed that two mRNA species differing at their 5Ј ends were transcribed from the mouse Map1a gene and that each transcript was regulated under independent promoters (14). The two alternative transcripts were expressed principally in the brain, but a short transcript under the control of an intronic promoter had a broader tissue-specific expression pattern. However, no cis-regulatory elements that confer neuronal cellspecific expression in a reporter assay were identified within about 2 kb 5Ј to the transcription initiation sites of either promoter region. In the present study, we further investigate the regulatory elements necessary for neuronal cell-specific expression of the MAP1A gene. We present here several lines of evidence that RFX proteins, which comprise a unique wingedhelix transcription family, down-regulate its 5Ј promoter activity in non-neuronal cells. RFX homologues in Drosophila and Caenorhabditis elegans are expressed specifically in subsets of sensory neurons (15,16), and the C. elegans RFX homologue is necessary for the normal function of these cells (15), suggesting that transactivation of target genes by RFX is crucial for the nervous system in these species. We propose that mammalian RFX proteins contribute to the neuronal cell-specific gene expression by their repressive activity outside the nervous system.

EXPERIMENTAL PROCEDURES
RT-PCR-For expression analysis of the human long isoform of MAP1A mRNA, 2 g of RNA isolated from culture cells with an RNAeasy mini-kit (Qiagen, Hilden, Germany), were primed with random * This work was supported in part by a grant for COE research from the Ministry of Education, Science, Culture and Sports in Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. hexamers and reverse-transcribed with 200 units of SuperScript II reverse transcriptase (Invitrogen) at 42°C. PCR was carried out using the primer pair hU94 (5Ј-GGCTGGGGCTCCGAAGTCC-3Ј) and hL456 (5Ј-GCTGCTAAGAAGATGATGAACCTC-3Ј). PCR with a primer pair for ␤-actin (forward: 5Ј-TCAGAAGGACTCCTATGTGG-3Ј, reverse: 5Ј-TCTCTTTGATGTCACGCACG-3Ј) was performed in parallel as a control. Each PCR was carried out using Taq DNA polymerase (Takara Biomedicals, Tokyo, Japan) for 30 cycles in a GeneAmp 9700 (Applied Biosystems, Foster City, CA) under the following conditions: denaturing at 96°C for 10 s, annealing at 58°C for 10 s, extension at 72°C for 20 s. Some of the amplified products were subcloned into the pGEM T easy vector (Promega, Madison, WI) and sequenced in both directions with either a Prism Dye primer Sequenase kit or a BigDye Sequenase kit (Applied Biosystems). The sequence products were analyzed with a Geneanalyzer 373 and 310 (Applied Biosystems).
Cloning of Human MAP1A Promoter-Based on the similarity between the mouse Map1a gene and sequence in the human chromosome 15 working draft sequence segment (GenBank accession no. NT010194), we designed primer sets (5Ј primer: 5Ј-GGAATGGCTG-GAATACTGTT-3Ј, 3Ј primer: 5Ј-ACTGCGGTGGAGGGTGTGAC-3Ј) to amplify putative exon 1 and the 5Ј-untranslated region of the human MAP1A gene. Fragments amplified from HeLa cell genomic DNA were subcloned into the pGEM-T vector (Promega) and sequenced.
Luciferase Reporter Constructs-The mouse Map1a promoter constructs p1(ϩ23), p1(ϩ57), and p1(ϩ104) were made by inserting PCR fragments into the pGL3 basic vector (Promega). The PCR fragments were amplified from a subclone of the mouse Map1a gene derived from the phage clone N4 described previously (14). A 5Ј primer used for all three constructs was designed to add an XhoI site at the 5Ј end, and 3Ј primers specific for p1(ϩ23), p1(ϩ57), and p1(ϩ104) were designed to add NcoI, HindIII, and XhoI sites, respectively, at the 3Ј end. Each amplified fragment was digested with appropriate restriction enzymes and inserted into the respective sites in the pGL3 basic vector.
The human MAP1A promoter constructs hMP1-2, hMP1-7, and hMP1-8 were made in a similar way. PCR fragments were amplified from the human MAP1A gene subclone described above and were engineered to contain an XhoI site at the 5Ј end and a HindIII site at the 3Ј end. The fragments were introduced between the XhoI-HindIII sites of the pGL3 basic vector after restriction enzyme digestion.
The reporter construct, hMP1-2 G4BS, was made by introducing a PCR-amplified fragment corresponding to 5 tandem repeats of the GAL4 binding site into hMP1-2. The original fragment containing the 5ϫ GAL4-binding sites was derived from the pFR-Luc plasmid supplied as a component of the PathDetect ® system (Stratagene) and engineered to contain appropriate restriction sites at both ends for subcloning.
RFX Expression Constructs-The human RFX1 and RFX3 expression constructs, pFC R1 and pFC R3, were generated by subcloning fulllength human RFX1 and RFX3 cDNAs into the NotI and BamHI sites of the pFLAG-CMV-5a vector (Sigma). Full-length RFX1 and RFX3 cDNA (GenBank accession nos. NM002981 and X76092) were obtained by PCR from a MARATHON human brain cDNA library (Clontech) and engineered to contain a NotI site at the 5Ј end and a BamHI site at the 3Ј end. The pFC R1DBD was constructed by inserting a PCR fragment encoding residues 420 -512 of RFX1 into the pFLAG-CMV-5a.
pFC R1-G and pFC R3-G were obtained by inserting a cDNA fragment encoding the GAL4 DNA-binding domain (DBD) in front of the RFX cDNA in pFC R1 and pFC R3. The fragment was derived from pAS2-1, the bait vector in the yeast two-hybrid system (Clontech). Expected chimeric proteins have PGGRVG residues between GAL4 DBD and RFX1, and PGGRPGRSPPSPSPTNNTTIIHSQETI between GAL4 DBD and RFX3. pFC E-G was constructed by introducing the cDNA fragment encoding GAL4 DBD alone into the pFLAG-CMV-5a vector.
All of the PCR fragments described above were sequenced and confirmed as not harboring any PCR errors. The constructs were introduced into HeLa cells, and the expression of proteins of the expected sizes was verified by Western blotting.
Transfection and Luciferase Assay-Transfection and luciferase assays were performed as described in our previous report with minor modification (14). Briefly, each culture cell was plated on Falcon 12-well tissue culture plates at appropriate densities 16 -18 h before transfection. Transfection was performed using LipofectAMINE and PLUS Reagent (Invitrogen) according to the manufacturer's instructions. 0.4 ml of Opti-MEM (Invitrogen) containing 0.3 g of the reporter DNA, 0.03 g of the pRL-TK vector as internal standard for transfection efficiency, 1.5 l of LipofectAMINE, and 3 l of Plus reagent was applied to one well, in which cells were incubated at 37°C for 3 h. In cotransfection experiments, 0.2 g of the reporter and 0.2 g of the expression construct, instead of 0.3 g of the reporter alone, were applied. After a 3-h incubation, 0.4 ml of RPMI 1640 medium or Dulbecco's modified Eagle's medium with 20% fetal bovine serum was added to each well. The medium was replaced with fresh medium containing 10% fetal bovine serum 24 h after transfection. Transfected cells were harvested by lysis after an additional 24 h culture. Cell lysis buffer and reagents used in the luciferase assay were supplied in a dual luciferase assay kit (Promega). Luciferase activity was measured by Lumicounter (MicroTech Niti-on, Saitama, Japan). For each measurement, luciferase activity was normalized by the activity of sea-pansy luciferase derived from the pRL-TK. Each transfection was performed in triplicate for one experiment, and at least three independent experiments were conducted.
DNase I Footprinting and Electrophoretic Mobility-shift Assay (EMSA)-Nuclear proteins used for DNase I footprinting and EMSA were prepared as described previously (14). To prepare the probe for DNase I footprinting, the DNA fragment from Ϫ141 to ϩ116 relative to the transcription initiation nucleotide of exon 1 of the mouse Map1a gene was subcloned into pBlueScript II (Stratagene). The construct was digested with XhoI and filled in with Klenow fragment and [␣-32 P]dCTP to label the 3Ј end of the coding strand. The end-labeled DNA was digested with BamHI, and the excised fragment was gel-purified. Nuclear proteins and 2000 -3000 cpm of the labeled probe were incubated in a final volume of 100 l (20 mM Tris, 1 mM EDTA, 10% glycerol, 0.1% Nonidet P-40, 2 mM MgCl 2 , 1 mM dithiothreitol, 2% polyvinyl alcohol, and 1 g of poly(dI-dC)) at room temperature for 20 min. After the addition of 100 l of salt mix (10 mM MgCl 2 , 5 mM CaCl 2 ), DNase I digestion was performed on ice for 3 min by adding 10 l of diluted DNase I (Takara Biomedicals). Digestion was stopped by adding 200 l of stop mix (200 mM NaCl, 40 mM EDTA, 1% SDS, 125 g/ml tRNA, and 100 g/ml proteinase K), followed by incubation at 37°C for 15 min. DNA was extracted, precipitated, resuspended in denaturing loading dye, and separated on a sequencing gel. The Maxam-Gilbert sequencing G ladder of the probe was run in parallel as a molecular size marker.

Exon1 of the MAP1A Gene Contains Regulatory Elements That Repress Reporter Gene Expression in Non-neuronal
Cells-Of two 5Ј alternative transcripts driven by independent promoters, the expression of the shorter transcript, regulated by the intronic promoter, was not highly restricted within neuronal cells (14). Therefore, we focused our investigation on the 5Ј promoter that confers brain-specific expression of the long transcript in mouse tissues and neuronal cell-specific ex-pression among culture cells. Because we did not find significant neuronal cell-specific cis elements within a 2.4-kb region 5Ј to exon 1 (14), we examined a region 3Ј to the transcription initiation site. We made several constructs differing in their 3Ј ends, but all constructs contained in common the 5Ј basal promoter consisting of three Sp1-binding sites and one Y/CCAAT box (Fig. 1A). 3Ј extension of the inserted genomic fragment to a part of exon 1 resulted in repression of promoter activity in non-neuronal cells, especially in HeLa cells, whereas it did not significantly affect transactivity in neuronal cells (Fig. 1B). Further extension of the 3Ј end had no significant effects on promoter activity in either neuronal or non-neuronal cells (data not shown).
To determine whether the region necessary for the transcriptional repression in non-neuronal cells was occupied by nuclear proteins, and to determine the precise nucleotide regions that interact with nuclear proteins, we performed DNase I footprinting. The result shown in Fig. 2 demonstrates that protected patterns differed slightly from each other depending on the cell type from which the nuclear proteins were prepared. Two regions spanning from ϩ24 to ϩ46 and from ϩ68 to ϩ89 relative to the transcription start site of exon 1, however, were protected in all cases. In addition, multiple regions of the basal promoter were also protected, confirming our previous report (14). Data base search by MatInspector V2.2 at transfac.gbf.de (17) revealed that both of the protected regions in exon 1 include nucleotide sequences highly similar to the RFX1 binding consensus sequence. Thus, the results presented in Figs. 1 and 2 suggest that exon 1 of the mouse Map1a gene contains repressive elements that confer neuronal cell-specific expression of the long transcript and that the elements are most likely RFX1 binding sites, yet they are occupied by nuclear proteins prepared from both non-neuronal and neuronal cells.
RFX-binding Motifs Are Conserved in the Human MAP1A Gene-To further investigate the role of the RFX1 binding sites in MAP1A transcriptional regulation, we searched for con- Basic. The basal promoter consists of three Sp1 sites and one Y/CCAAT box. Three constructs-p1(ϩ23), p1(ϩ57), and p1(ϩ104)-contain genomic fragments represented by arrows. The nucleotide number is relative to the transcription initiation site of exon 1. B, three nonneuronal cells, NIH3T3, L cell, and HeLa, and three neuronal cells, Neuro2A, C1300, and TGW, were transfected with the indicated constructs together with pRL-TK as an internal control for transfection efficiency. The cells were harvested and assayed for luciferase activity 48 h after transfection. Promoter activity was normalized by the seapansy luciferase activity derived from pRL-TK, and is given as -fold activity above the promoterless pGL3 Basic. The values are averages Ϯ S.E. of at least nine data points. Transfection was performed in triplicate, and three or more independent transfection experiments were performed.
FIG. 2. Nuclear proteins from neuronal and non-neuronal cells bind to RFX sites in exon 1 of the MAP1A gene. DNase I footprint analysis of a DNA fragment spanning the region Ϫ141 to ϩ116 relative to the transcription initiation site was performed. The fragment, labeled at the 3Ј end of the coding strand, was incubated with either 10 g or 3 g of nuclear proteins from Neuro2a, NIH3T3, TGW, and HepG2 cells. Ten g of bovine serum albumin was used as a control. The Maxam-Gilbert sequencing G ladder of the fragment was run as a molecular size marker (G). served elements in the human MAP1A gene, for which the long transcript has not yet been identified. We retrieved 20 kb in sequence 5Ј to, and 5Ј of the human MAP1A gene from a Homo sapiens chromosome 15 working draft sequence segment (Gen-Bank accession no. NT010194). Comparison of the region with the 5Ј part of the mouse Map1a gene (Ref. 14, GenBank accession no. AF182208 -11) revealed that the 5Ј exons are conserved in the human gene. Segments highly similar to exons 1, 2, and 3 of the mouse gene were identified about 9.7 kb, 8.4 kb, and 3.0 kb, respectively, upstream from the previously identified human exon 1 (18). We confirmed that the segments homologous to mouse exons 1, 2, and 3 were transcribed as parts of MAP1A mRNA in human brain by PCR from a human brain cDNA library (data not shown); exon numbers of the human gene will be presented as they occur in the mouse gene hereafter in this study (Fig. 3A). We designed primer sets to obtain a human genomic fragment containing the MAP1A 5Ј promoter and exon 1 and amplified a fragment from HeLa cell genomic DNA. Its sequence was completely identical with the sequence reported in GenBank, and a portion is shown in Fig. 3B aligned with the corresponding mouse genomic sequence. The basal promoter region containing three Sp1 sites and one Y/CCAAT box is conserved between human and mouse. Both RFX binding motifs in exon1 are also conserved in the human gene, with minor differences in the 5Ј RFX binding motif. Next, we examined by RT-PCR whether the long isoform of MAP1A mRNA, the 5Ј part of which consists of exons 1, 2 and 4, is expressed specifically in human neuronal cells. A 386-bp DNA fragment was amplified with a set of specific primers from RT products of neuroblastoma mRNA (Fig. 3C, TGW and LA-N-5). The fragment was confirmed to be an expected product by sequencing. The long transcript was also expressed at low levels in A431 epidermoid carcinoma cells and HeLa adenocarcinoma cells (Fig. 3C). In contrast, amplified products were hardly observed from HT29 colorectal adenocarcinoma cells, HepG2 hepatocellular carcinoma cells, and Saos-2 osteosarcoma cells (Fig. 3C). These results demonstrate that the genomic structure and RFX-binding motifs in exon 1 are conserved in human, and the long transcript is expressed preferentially in human neuronal cells.
Identification of Two RFX-binding Sites as Repressors of MAP1A Gene Transcription-On confirmation that the exonintron structure and 5Ј promoter region were conserved in the human gene, further reporter assays were conducted using human genomic fragments to take advantage of their ability to be examined in a variety of homologous cell lines. The human MAP1A basal promoter construct, hMP1-2, efficiently transactivated the reporter gene up to 70-to 300-fold above the promoterless reporter, pGL3 basic in all cell lines tested except for Saos-2 osteosarcoma cells, in which the activity was 14-to 20-fold above the pGL3 basic. An additional potential Sp1 binding site found only in the human gene (Fig. 3B, Sp1(H)) was not required for promoter activity (data not shown). Like the mouse promoter, the 5Ј end of exon 1, which contains the two RFX-binding sites, has a remarkable repressive effect in non-neuronal cells (Fig. 4, A and B). Promoter activity of a reporter hMP1-8 construct containing the two RFX sites was 5% or less of that of a basal promoter construct, hMP1-2 in HeLa, HT29, and A431 cells. Repression was also seen in Saos-2 osteosarcoma and HEK293 embryonic kidney cells. In contrast to the mouse promoter, however, the two RFX sites acted as weak repressors in the neuronal cell lines TGW and LA-N-5. To confirm that the RFX-binding sites are critical for repression, we created reporters with mutations in the RFX sites. Because an RFX protein can bind to a half-consensus site as a monomer, albeit less effectively than as part of a homo-or heterodimers binding to a full palindromic consensus site (19,20), mutations were introduced in each of the four half-sites of the two full palindromic consensus sequences in the MAP1A gene (Fig. 4C). As shown in Fig. 4D, all mutations resulted in up-regulation of transactivity compared with the parental reporter hMP1-8 in HeLa, HT29, and A431 cells, indicating that all the mutations disabled repression. Among the four RFXbinding half-sites, the 3Ј half of the 3Ј RFX, mutated in hMP1-8mu1, which was 8-to 10-fold more active than the wild-type in epithelial cells, appears to be most crucial for repressive activity. This site also seemed responsible for the weak repression in TGW cells. The 5Ј half of the 3Ј RFX-binding site and either halves of the 5Ј RFX-binding site were considered necessary for full repression in epithelial cells. These three mutations had no significant effects on promoter activity in TGW.
RFX1 and RFX3 Bind to Repressors in the MAP1A Gene-The above reporter assays strongly suggested that the two RFX sites are repressors acting preferentially in non-neuronal cells. DNase I footprinting, however, indicated that the RFX sites in the MAP1A gene were occupied by nuclear proteins from all cell types. Identification of the binding proteins would give an insight into the mechanism by which transcription is regulated differently among cell types. Gel-shift experiments revealed that multiple complexes that differ among cell types interact with the 3Ј RFX sequence of the MAP1A gene (Fig. 5A). Five retarded bands, A, B, and C, representing relatively slow moving complexes, and D and E, representing fast moving ones, were identified using TGW nuclear extract. Bands A, B, and C were obtained using nuclear extracts from all cells examined, but their relative intensities were variable. Band E was also observed very faintly in experiments conducted with LA-N-5 and Neuro2A nuclear extracts. These retarded band patterns were observed as well when the 5Ј RFX sites in the mouse and human MAP1A genes, an RFX site in the interleukin 5 receptor ␣ gene, and a polyoma virus EP site, were used as probes (data not shown). The specificity of complex formation was examined further by competition experiments. Complexes formed with both HeLa and TGW nuclear extracts were competed apart by an excess amount of non-radiolabeled double-stranded oligonucleotide containing the 3Ј RFX sequence (Fig. 5B, lane self). Competing oligonucleotides containing the MAP1A 5Ј RFX sequence or the polyoma virus EP sequence diminished complex formation as well, but less effectively than those encoding the MAP1A 3Ј RFX (compare lanes 5ЈRFX and PyEP with lane self in Fig. 5B). Oligonucleotides containing the MAP1A 3Ј RFX site harboring the same mutations as were introduced in the reporter assay did not affect the band patterns, confirming that complex formation depends critically on the entire core sequence of the RFX site (Fig. 5B, mu1 and mu2).
Next, supershift experiments with specific antibodies were performed to identify which RFX proteins are involved in the complexes. Among mammalian RFX proteins, RFX1-RFX3 comprise a subfamily because of their highly similar target DNA specificities (20). As shown in Fig. 5C, all bands except for band E were supershifted by an anti-RFX1 antibody. The supershifted band ss1 was not thought to contain a complex derived from band A and B, because it was not observed when EMSA was conducted using the HeLa cell nuclear extract, which shifts band A and B intensely. The complex derived by the anti-RFX1 antibody and band A and B did not seem to enter FIG. 4. RFX sites are repressive elements highly effective in non-neuronal cells. A, schematic of human genomic fragments examined by luciferase assay is shown. Some cis-elements and RFX sites in exon 1 are indicated. Arrows represent the region examined by luciferase assay. Fig. 3B shows precise nucleotide positions. B, five human non-neuronal cell lines (HeLa, HT-29, A431, HEK293, and Saos-2) and two human neuronal cell lines (TGW and LA-N-5) were transfected with the indicated constructs, together with pRL-TK. The cells were harvested and assayed for luciferase activity 48 h after transfection. Promoter activity was normalized by sea-pansy luciferase activity. Average activities of hMP1-2 were set at 100%, and promoter activity of hMP1-7 and hMP1-8 is presented as percent activity relative to that of hMP1-2. The values are averages Ϯ S.E. of at least nine data points. C, substituted nucleotides in the RFX sites of the parental clone hMP1-8 are underlined and in lowercase. One of the four halves of the two palindromic sites was mutated in each of the following constructs: hMP1-8 mu1, mu2, mu3, and mu4. D, three non-neuronal cell lines (HeLa, HT-29, and A431) and one neuronal cell line (TGW) were transfected with the indicated constructs together with pRL-TK. The cells were harvested and assayed for luciferase activity 48 h after transfection. Promoter activity was normalized by sea-pansy luciferase activity. Promoter activity of mutants is given as an -fold activity above the average activity of hMP1-8. The values are averages Ϯ S.E. of at least nine data points. the gel efficiently, because high amounts of radioactive complexes were observed in the loading wells above the gel in the lanes where anti-RFX1 antibody had been applied (data not shown). Thus, the supershifted band ss1 was most likely derived from band C. The complex E band was supershifted by an anti-RFX3 antibody, but the supershifted band could not be separated clearly from the other bands, as was reported previously (21). The complex C band was also supershifted by anti-RFX3 antibody, which was observed most clearly in experiments using HeLa nuclear extracts. Incubation with anti-RFX2 antibody did not change the band patterns. These results and previous reports (21,22) suggest that complexes A, C, and E contain RFX 1/1, RFX 1/3, and RFX 3/3 homodimers, respectively. The remaining complexes B and D, which have not been reported previously, appear to contain RFX1, but not RFX3 or 2. Differences between bands A, B, and D are possibly due to various other cofactors included in the complexes.
Both RFX1 and RFX3 Down-regulate the MAP1A Promoter in Non-neuronal Cells-Because EMSA experiments suggested that RFX1 and 3 are the major proteins binding to the RFX sites in the MAP1A gene, we examined their functional effects on the promoter activity in reporter assay. We made RFX1 and RFX3 expression constructs driven by the CMV promoter. The expression of proteins in expected sizes and appropriate nuclear localizations of EGFP-tagged proteins were confirmed by Western blotting and confocal microscopy, respectively (data not shown). We then examined their effects on the hMP1-8 reporter activity in non-neuronal and neuronal cells (Fig. 6A). Overexpression of RFX1 resulted in down-regulation of the reporter compared with control in all non-neuronal cells examined (Fig. 6B, pFC R1). Reporter activity was also down-regulated when RFX 3 was overexpressed in HT29 and HEK293 cells (Fig. 6B, pFC R3). However, overexpression of RFX3 did not result in down-regulation, but slight up-regulation of the reporter in HeLa cells. Forced expression of the RFX1 DNA binding domain, which was expected to act as a dominantnegative protein by occupying the RFX sites, up-regulated transcription 2-fold compared with control in non-neuronal cells (Fig. 6B, pFC R1DBD). This further confirmed that the RFX sites in the MAP1A gene function as repressors in nonneuronal cells. Neither RFX1 nor RFX3 down-regulated transactivity in TGW cells. Although these overexpression experiments clearly indicate that ubiquitously expressed RFX1 downregulates the MAP1A 5Ј promoter in non-neuronal cells and not in neuronal cells, the results are ambiguous in terms of RFX3 function. Therefore, we performed additional cotransfection studies in which cDNAs encoding GAL4 DBD-RFX chimeric proteins were introduced (Fig. 6C) in an attempt to minimize the effect of endogenous RFX proteins. In this system, both GAL4 DBD-RFX1 and -RFX3 proteins down-regulated reporter activity in non-neuronal cells (Fig. 6D); the effects were more obvious than in the simple overexpression study shown in Fig.  6B. The most effective down-regulation was observed in HeLa cells, in which the introduction of GAL4 DBD-RFX1 resulted in down-regulation of reporter activity to 13% of the control. In contrast to the overexpression experiments, GAL4 DBD-RFX3 also down-regulated reporter activity to 35% of the control in HeLa cells. In two other non-neuronal cell lines, the rates of down-regulation by RFX1 and 3 were not significantly different. Again, reporter gene down-regulation was not observed upon expression of chimeric RFX proteins in TGW. Based on these experiments, we conclude that both RFX1 and RFX3 down-regulate the MAP1A promoter activity in non-neuronal cells but not in neuronal cells. DISCUSSION In the present study, we have shown that the two RFXbinding sites in the MAP1A gene act as repressors specifically in non-neuronal cells. We identified several distinct complexes bound to the RFX sites sequences in the MAP1A gene and confirmed that they contained RFX1, RFX3, or both. Forced expression of dominant-negative RFX protein attenuated the repression in non-neuronal cells, confirming that they are involved in the in vivo process. Overexpression of RFX1 and RFX3 and forced expression of Gal4 DBD-RFX chimeric proteins indicated that both RFX1 and RFX3 down-regulate reporter gene expression specifically in non-neuronal cells. These data strongly suggest that the 5Ј promoter of the MAP1A gene is under the control of RFX transcription proteins and that RFX1 and 3 contribute to the neuronal cell-specific expression of the gene by their repressive activity in non-neuronal cells.
RFX1, a prototypic mammalian RFX protein, was first cloned FIG. 5. Each cell expresses unique nuclear protein complexes that bind to the RFX site of the MAP1A gene. A, gel shift experiments were performed with nuclear extracts from the indicated cells and a 32 P-labeled 3Ј-RFX probe. Amounts of nuclear proteins extracted from each cell line and applied were as follows: 2.5 mg (TGW), 10 mg (LA-N-5), 2 mg (Nauro2A), 2.5 mg (HeLa), 5 mg (HT-29), 10 mg (A431), 2 mg (HEK293), and 3.5 mg (NIH3T3). The reaction was performed without nuclear protein (Ϫ) as a negative control. The binding reaction was electrophoresed on a 4% polyacrylamide Ϫ0.25% TBE gel. B, 10fold molar excess of unlabeled competitor oligonucleotide was added to the binding reaction before the radiolabeled probes were applied. The competing oligonucleotides used were as follows: unlabeled 3Ј-RFX probe (self), 3Ј-RFX mutant1 (mu1), 3Ј-RFX mutant2 (mu2), 5Ј-RFX, and Polyoma virus EP site (PyEP). The nucleotide sequences of the competitors are described under "Experimental Procedures." An experiment without added competitor was performed as a negative control (Ϫ). Two g of either HeLa or TGW nuclear extract was applied. C, antiserum against RFX1 (1), RFX2 (2), or RFX3 (3) was added to the reaction after the DNA-protein complex was formed. Preimmune rabbit serum was added in each control lane (Ϫ). Other conditions, including the volumes of nuclear extract, were as in A.
as a transcription factor that interacts with the X-box in the class II MHC gene (19,23). Subsequent studies revealed that RFX family members, of which there are five in human (RFX1-5) and four in mice (RFX1-3 and RFX5), share a unique winged-helix type DNA-binding domain (24 -26). In addition, one family member in each of several other eukaryotes-dRFX in Drosophila, DAF-19 in C. elegans, sak1 in Saccharomyces pombe, and Crt1 in Saccharomyces cerevisiae -has been identified (15,25,(27)(28)(29). Thus, the RFX proteins appear to have been conserved throughout evolution in a wide variety of species, and their functions elucidated to date are diverse. Crt1 acts as a repressor in the DNA damage and replication checkpoint pathways (28). dRFX in Drosophila and DAF-19 in C. elegans are expressed in type I sensory neurons and all ciliated sensory neurons, respectively (15,16). Loss of daf-19 function causes an absence of cilia, resulting in severe sensory defects (15). In mammals, the function of RFX5 has been studied most intensively, because mutations of the gene coding RFX5 cause bare lymphocyte syndrome, a type of primary immunodeficiency (26). RFX5 is essential for the activation of class II MHC transcription as a DNA-binding component of RFX, a complex binding to the X box (30). Although RFX1 was originally cloned by virtue of its in vitro interaction with the X box in the class II MHC promoter, its role in the regulation of class II MHC expression has been unproven. Aside from the class II MHC gene, several potential target genes that may be under the control of RFX1 and/or the closely related RFX2 and 3 have been reported. EF-C or EP sites, partially palindromic sequences to which RFX1-3 bind, have been found in the enhancers of hepatitis B virus, polyomavirus, cytomegalovirus, and Epstein-Barr virus (31)(32)(33)(34). RFX-binding sites have been also recognized in some cellular genes, including the genes for c-Myc (35), proliferating cell nuclear antigen (36), the ribosomal protein rpL30 (37), collagen ␣2(I) (38), and the interleukin 5 receptor ␣ (22). RFX1, with or without RFX2 and/or 3, binds to these elements in vitro and seems to regulate their transcription. The physiological significance of RFX1-3, however, has yet to be determined by knockout analysis. According to the phylogenetic tree of the RFX DNA-binding domain, the RFX1-3 subgroup is most closely related to dRFX and DAF-19 (29). X-box sequences for the RFX1-3 subgroup are very similar to that for DAF-19 (15), whereas they differ from those for RFX4 and RFX5 (20). Thus, the mammalian RFX1-3 subgroup is expected to play an important role in the nervous system, like its Drosophila and C. elegans homologues. The present results are the first evidence that RFX1 and 3 indeed are involved in the cell type-specific transcription of a gene important for the neuronal function; however, it seems to be accomplished by their repressive activity in non-neuronal cells and not by their activity in neuronal cells.
RFX-binding sites in the genes for c-Myc (35), proliferating cell nuclear antigen (36,39), and collagen ␣2(I) (38) are nega- , and TGW cells were transfected with a hMP1-8 reporter construct and the indicated expression constructs. An empty CMV expression vector, pFC E, was used as a control. The cells were harvested and assayed for luciferase activity 48 h after transfection. Promoter activities are presented in relative luciferase units without normalization, and the values are averages Ϯ S.E. of at least nine data points. C, schematic of a reporter construct and expression constructs for GAL4 DNA-binding domain (GAL4DBD)-RFX chimeras. The reporter construct has 5 tandem repeats of the GAL4-binding site in front of the MAP1A 5Ј basal promoter, followed by the luciferase gene. The expression constructs contain the fragments for the GAL4DBD-RFX1 chimera (pFC R1-G) and GAL4DBD-RFX3 chimera (pFC R3-G), and the expected proteins were confirmed as translated in HeLa cells. D, HeLa, HT-29, HEK293, and TGW cells were transfected with a hMP1-2 G4B reporter construct and the indicated expression constructs. A vector, pFC E-G, containing the fragment for GAL4DBD alone was used as a control. The cells were harvested and assayed for luciferase activity 48 h after transfection. Promoter activities are presented in relative luciferase units without normalization, and the values are averages Ϯ S.E. of at least nine data points.
tively acting regulatory elements, as in the MAP1A gene, whereas those in most of the viral promoters (34,40) and the genes for ribosomal protein rpL30 (37) and interleukin 5 receptor ␣ (22) are positively acting regulatory elements. Thus, RFX-binding sites function differently in a context-dependent manner, possibly reflecting various interactions between transcriptional activation and repression domains in RFX proteins and other regulatory factors recruited by adjacent elements (22,41). Besides the context of the regulatory unit, cofactors recruited by the DNA-binding protein are now considered to be important modulators of the transcriptional control and often become the determinant of cell and tissue specificity. In fact, such a cofactor is identified and well characterized in association with another member of RFX proteins. The class II transactivator, CIITA, which itself does not bind DNA and is recruited by RFX5 to the X box, is the determinant for cell-type specificity and interferon-␥ inducibility of MHC class II expression (42)(43)(44). Thus, it is tempting to speculate that non-neuronal cell-specific repression by RFX1 and 3 depends on a corepressor only available in non-neuronal cells. Unfortunately, little is known about the precise mechanisms of repression and activation by the RFX1-3 subfamily and the cofactors associated with RFX1-3. RFX1 has been reported to coexist in a binding complex with p107, a retinoblastoma-related protein, to the proliferating cell nuclear antigen promoter, but a direct interaction between RFX1 and p107 has not been ascertained (36). p107, best known as the corepressor for E2F-dependent transcription, has also been reported to associate with Sp1 and repress Sp1-dependent transcription in reporter assays (45). Because the promoter activity of the MAP1A gene highly depends on Sp1-binding sites (14), p107 recruitment via RFX proteins may be involved in the down-regulation of the MAP1A promoter. However, our preliminary experiments failed to detect p107 or other pocket protein family members in HeLa cell nuclear protein complexes bound to RFX sites in the MAP1A gene (data not shown). Further study of cofactors involved in the DNA-binding complexes mediated by RFX is necessary to understand the mechanisms underlying its repressor activity, and such studies may provide insight about cell specificity as well.
In conclusion, we propose that RFX1 and RFX3 participate in transcriptional regulation of a gene important for the nervous system. The MAP1A 5Ј promoter, which is under their control, is a robust model system in which non-neuronal cell-specific repressor mechanisms RFX1 and RFX3 can be explored.