Silencer elements modulate the expression of the gene for the neuron-glia cell adhesion molecule, Ng-CAM.

The combined factors that regulate the expression of cell adhesion molecules (CAMs) during development of the nervous system are largely unknown. To identify such factors for Ng-CAM, the neuron-glia CAM, constructs containing portions of the 5′ end of the Ng-CAM gene were examined for activity after transfection into N2A neuroblastoma and NIH3T3 cells. Positive regulatory elements active in both cell types included an Ng-CAM proximal promoter with SP1 and cAMP response element motifs extending 447 base pairs upstream of a single RNA start site and a region within the first exon corresponding to 5′-untranslated sequences. Negative regulatory elements included five neuron-restrictive silencer elements (NRSEs) and a binding site for Pax gene products in a 305-base pair segment of the first intron. Constructs containing the promoter together with the entire first intron were active in N2A cells but were silenced in NIH3T3 cells. This silencer activity was mapped to the NRSEs. In contrast, the Pax motif inhibited activity of Ng-CAM constructs in both cell types. The DNA elements defined in these transfection experiments were examined for their ability to bind nuclear factors. The region within the first exon formed a DNA-protein complex after exposure to nuclear extracts prepared from both NIH3T3 and N2A cells. The NRSE region formed a more prominent complex with proteins prepared from NIH3T3 cells than it did with extracts from N2A cells. A member of the Pax protein family, Pax-3 bound to the Pax motif. Mutations introduced within the Pax motif in its ATTA sequence eliminated this binding whereas mutations in its GTTCC sequence did not, suggesting that paired homeodomain interactions are important for the recognition of Pax-3 by this DNA target sequence. The combined data suggest that negative regulation by NRSEs and Pax proteins may play a key role in the place-dependent expression patterns of Ng-CAM during development.

Cell adhesion molecules (CAMs) 1 are essential for guiding tissue formation (1) and play key roles in the development of the nervous system. CAMs important in the nervous system include N-CAM-related molecules such as Ng-CAM (2, 3), Nr-CAM (4,5), L1 (NILE) (6,7), neurofascin/ABGP (8,9), TAG-1/ axonin-1/F3 (10 -12), and contactin (13). Ng-CAM, L1, Nr-CAM, neurofascin/ABGP from vertebrates and neuroglian from Drosophila (14) comprise a subfamily of neural CAMs containing six immunoglobulin domains and five fibronectin type III repeats. Molecules of this Ng-CAM subfamily are expressed prominently in axonal pathways in both the central nervous system and peripheral nervous system and are involved in neurite fasciculation and outgrowth. Each of these neural CAMs has a characteristic spatial and temporal expression pattern during neural morphogenesis but the factors that restrict expression of Ng-CAM and other neural CAM genes to particular populations of neural cells are not well understood.
To define the sequences of DNA responsible for place-dependent expression of CAMs, we have focused on signals from homeobox and Pax gene products (15)(16)(17)(18). An attractive hypothesis is that Ng-CAM and other neural CAMs are targets of homeodomain and Pax proteins (19). During neural development, a number of transcriptional regulators encoded by the homeobox and Pax gene families appear in defined expression patterns along the anterior-posterior and dorsal-ventral axes of the embryo that correlate with the patterns of a variety of CAMs. Moreover, mutations in Pax genes are known to alter the programs of neural differentiation and migration, processes that are influenced by the activity of neural CAMs. For example, mutations in Pax-6 and Pax-3 genes lead to developmental defects of the nervous system as shown, respectively, in small eye and splotch mutant mice (20). Some of these defects appear to be caused by aberrant neuronal migration (21,22), a process normally modulated by neural CAMs.
To examine the factors regulating Ng-CAM expression, we have isolated genomic clones containing the 5Ј end of the Ng-CAM gene, characterized its proximal promoter, and located two regulatory regions within a 305-base pair segment of the first intron. One region of the first intron was found to contain five neuron-restrictive silencer elements (NRSEs) which extinguished expression of the Ng-CAM gene in a fibroblast cell line, but not a neuroepithelial cell line. Another region contained GTTCC and ATTA sequences characteristic of binding sites for Pax proteins. This Pax motif was found to bind Pax-3 in gel mobility shift experiments; such binding was disrupted when specific mutations were introduced in the ATTA sequence within the Pax motif. In transfection experiments with both NIH3T3 and N2A cells, this Pax motif was found to be a negative regulator of Ng-CAM gene expression independent of the silencing imposed by NRSEs. Our studies suggest that the NRSEs and Pax motif may play critical roles in the place-dependent expression of Ng-CAM in the nervous system.

MATERIALS AND METHODS
Analysis of Ng-CAM Genomic DNA-A chicken genomic DNA library in the cosmid vector pWE15 (Clontech) was screened using a 400-base pair 32 P-labeled EcoRI-BstXI fragment derived from the 5Ј end of the Ng-CAM cDNA sequence (3). A single cosmid clone (Cos-Ng) was isolated. Sequencing of the 2536-base pair segment of Ng-CAM genomic DNA was performed by the dideoxy chain termination method using * This work was supported by a United States Public Health Service Grant HD09635 and a grant from the G. Harold and Leila Y. Mathers Charitable Foundation (to G. M. E.). 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) U31086. 1 The abbreviations used are: CAM, cell adhesion molecule; NRSE, neuron-restrictive silencer elements; CAT, chloramphenicol acetyltransferase; GST, glutathione S-transferase; CRE, cAMP response element. Sequenase (U. S. Biochemical Corp.) (23) and the data were compiled using the GCG package (University of Wisconsin). The NRSEs and the Pax motif were not revealed in conventional data base searches, but were found after analysis of Ng-CAM constructs in transfection experiments. Ng-CAM NRSEs were aligned with the consensus NRSE sequence (24) using the fasta algorithm (25). Oligonucleotides were made using a 392 DNA synthesizer (Applied Biosystems, Chatsworth, CA).
Determination of the Transcription Initiation Site-The transcription start site was determined through a combination of primer extension and RNase protection analyses. Poly(A) ϩ RNA was isolated from 12 day embryonic chick brains using the Fast Track mRNA isolation kit (Invitrogen, San Diego, CA). For primer extension, 4 g of poly(A) ϩ RNA was hybridized to a 32 P-labeled oligonucleotide designated oNg-7, derived from the antisense strand of the first exon of the Ng-CAM gene (Fig. 2, from position ϩ185 to ϩ166). Primer extension was performed using Superscript reverse transcriptase (Life Technologies Inc.) at 42°C for 1 h.
RNase protection analysis was performed using the RPA II kit (Ambion). The template for the RNA probe was made by polymerase chain reaction from the genomic DNA. The downstream primer used for the amplification was Ng-7/T3-, an oligonucleotide identical in sequence to oNg-7, but also containing the T3 RNA polymerase promoter at the 5Ј end. The upstream primer used in the polymerase chain reaction was an oligonucleotide designated oNg-6, derived from the sense strand of the Ng-CAM genomic sequence (Fig. 3, position Ϫ172 to Ϫ156). A labeled RNA probe was synthesized from the template using T3 RNA polymerase and [ 32 P]UTP. Template DNA was then digested with RNase-free DNase. Two fmol of probe in elution buffer was mixed with 1 g of poly(A) ϩ RNA from 12 day embryonic chick brains and hybridized at 45°C for 16 h. RNase protection analysis was performed using different dilutions of RNase to determine the optimal conditions for cleavage. The products of both primer extension and RNase protection analysis were resolved on an 8% polyacrylamide sequencing gel.
Construction of Reporter Plasmids-Reporter gene plasmids were constructed in the promoterless chloramphenicol acetyltransferase (CAT) gene vector pCAT-basic (Promega, Madison, WI) and pBluescript-II (Stratagene, La Jolla, CA) and are shown schematically in Fig.  3. Ng447 and Ng447/182 plasmids were constructed by polymerase chain reaction amplification of DNA segments spanning the regions between Ϫ447 and ϩ61 and Ϫ447 and ϩ182, respectively, and were inserted into the pCAT basic vector (Promega). The Ng4, Ng1.5, and Ng200 plasmids containing 4 kilobases, 1.5 kilobases, and 200 base pairs of Ng-CAM gene 5Ј-flanking sequence, respectively, were made by insertion of appropriate restriction fragments into the pCAT-basic vector. The Ng447ϩI construct was made by insertion of a 2536-base pair BamHI-PvuII fragment into pBluescript and the CAT gene was inserted into the KpnI site immediately downstream. The constructs SgBsp, EE, EA, ES, and AS constructs containing internal deletions of the first intron were prepared by excision of SgrAI-BspMI, EcoRV-EcoRV, EcoRV-AatII, EcoRV-SmaI, and AatII-SmaI restriction fragments, respectively, from the Ng447ϩI construct followed by religation. Synthetic NRSE/Ng-CAM promoter constructs SS and SAS were prepared by insertion of the Ng-CAM NRSEs (the region between ϩ1388 and ϩ1566) in either orientation upstream of the promoter in the Ng447 construct.
Cell Culture, Transfection Experiments, and Promoter Assays-Mouse neuroblastoma N2A cells were passaged in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Mouse NIH3T3 fibroblasts were passaged in Dulbecco's modified Eagle's medium with 10% newborn calf serum. 2 ϫ 10 5 cells were seeded onto 6-well 60-mm tissue culture plates (Corning) and grown until 50% confluency was attained. Each well of cells was transfected with 2.5 g of plasmid DNA, in a mixture containing 6 l of lipofectamine reagent in 1 ml of Opti-MEM media (Life Technologies, Inc.). The plasmid RSV-␤gal (0.5 g) was co-transfected in all experiments and served as an internal control to normalize the transfection efficiencies between different cell types and experiments. The Opti-MEM media was replaced after 6 h with Dulbecco's modified Eagle's medium containing the appropriate serum. Cells were incubated for an additional 48 -72 h, collected, and resuspended in 150 l of 0.25 M Tris, pH 7.5. Cell extracts were prepared by freezing and thawing. ␤-Galactosidase activity was assayed using the Fluoreporter lacZ/galactosidase quantitation kit (Molecular Probes, Eugene, OR). Levels of activity were measured using the Cytofluor 2350 fluorescence measurement system (Millipore). Cell extracts were assayed for CAT activity as described (26). Levels of CAT activity were quantitated on a PhosphorImager (Molecular Dynamics, Sunnyvale, CA). The values for CAT activities reported for various Ng-CAM constructs were derived from at least four separate experi-ments performed in duplicate. Within such experiments, the activity levels for a given construct varied no more than 5%.
Preparation of Nuclear Extracts, the Pax-3 Fusion Protein, and Gel Mobility Shift Assays-Nuclear extracts were prepared from NIH3T3 and N2A cells as described previously (26). The mouse Pax-3 cDNA (kindly provided by G. Chalepakis and P. Gruss of the Max-Planck Institute, Gottingen, Germany) was inserted into the pGEX-2T vector. GST and GST/Pax-3 fusion proteins were produced in Escherichia coli NM522 cells and purified as described previously (17). The DNA fragment corresponding to the ϩ82 to ϩ182 region of the Ng-CAM gene was prepared by restriction endonuclease cleavage with SgrAI and BspMI. A 178-base pair DNA fragment containing the five NRSEs was prepared by excision of a HindIII fragment from the SS construct. The Pax motif probe was prepared by annealing 5 pmol of complementary oligonucleotides derived from the Pax motif of the Ng-CAM gene (ϩ1679 to ϩ1708). All probes were labeled using the Klenow fragment of E. coli DNA polymerase I [ 32 P]dCTP (6000 Ci/mmol) (DuPont NEN) (27), and purified by electrophoresis on a 10% polyacrylamide gel.
For binding reactions involving nuclear extracts, 5 fmol (approximately 20,000 cpm) of either ϩ82/ϩ182 or NRSE probe was mixed with 10 g of protein from either NIH3T3 or N2A cells in a buffer containing 10 mM Hepes, 200 mM KCl, 0.5 mM dithiothreitol, 0.1% Nonidet P-40, 100 ng/ml poly(dI-dC) with 10 g of bovine serum albumin. Competitors (either 10-or 100-fold excess of cold ϩ82/ϩ182 or NRSE DNAs) were included in some binding reactions. Binding components were incubated for 15 min at room temperature and subjected to electrophoresis on a 4% polyacrylamide gel in 0.25 ϫ TBE buffer at 400 V for 2 h at 4°C.
For binding reactions involving the Pax-3 fusion protein, 200,000 cpm of probe was incubated with 0.5, 2.5, or 5 g of GST/Pax-3 fusion protein. Binding reactions were performed in a volume of 20 l containing 15 mM Tris-HCl, pH 7.5, 6.5% glycerol, 90 mM KCl, and 0.2 mM dithiothreitol. Bovine serum albumin (0.5 mg/ml) and sheared salmon sperm DNA (100 ng) were added to each binding reaction to reduce nonspecific binding. Reactions were incubated at room temperature for 30 min and subjected to electrophoresis on a 5% polyacrylamide gel in 0.5 ϫ TBE buffer at 200 V at 4°C. Gels were dried and exposed to film at Ϫ70°C.

Structure of the 5Ј End of the Ng-CAM Gene: Identification of the Proximal Promoter and Regulatory Sequences in the First
Intron-A cosmid clone (designated Cos-Ng) was isolated after screening with a 32 P-labeled 400-base pair EcoRI-BstXI fragment derived from the 5Ј end of the Ng-CAM cDNA sequence (3). Cos-Ng contained an insert approximately 20 kilobases in length. The borders for the first seven exons and introns were located by restriction mapping and Southern blotting and their sequences were determined using oligonucleotide primers derived from the Ng-CAM cDNA sequence ( Fig. 1).
To determine the site of transcription initiation within the Ng-CAM gene, we performed primer extension and RNase protection analyses of poly(A) ϩ RNA isolated from chick brain tissue at embryonic day 12. Extension of an antisense oligonucleotide primer (oNg-7) yielded a single radiolabeled band of 185 nucleotides in length (data not shown). Two prominent bands were observed in an RNase protection experiment using a radiolabeled RNA probe corresponding to the region from Ϫ300 to ϩ185. Because the primer extension and the largest RNase protection products had termini at exactly the same upstream base pair in the genomic sequence, the nucleotide at this position was designated ϩ1, the start of transcription initiation (see Fig. 2).
The DNA sequence of a 2536-base pair segment of the 5Ј end of the Ng-CAM gene including the 5Ј-flanking sequence, the first exon, the first intron, and the second exon was determined (Fig. 2). This sequence has been deposited in GenBank under accession number U31086. Several potential regulatory motifs were found in the 5Ј-flanking region immediately adjacent to and upstream of the RNA start site. They included two consensus binding sites for the SP1 transcription factor (28) located at Ϫ124 and Ϫ15 as well as a consensus cyclic AMP response element (CRE), located at Ϫ255, which is known to bind mem-bers of the CREB family of transcription factors (29). After analyses of deletion constructs in transfection experiments, additional elements were located in a 178-base pair segment of the first intron (ϩ1388 to ϩ1566) that included five copies of a sequence similar to that of the NRSE. Further analysis also uncovered a 30-base pair element (ϩ1679 to ϩ1708) similar to that of a binding site for Pax gene products (20,30,31). This Ng-CAM Pax motif contains GTTCC and ATTA core sequences characteristic of a target sequence recognized by those Pax proteins that contain two independent DNA binding domains, paired domain and a homeodomain (20,32).
Examination of Regulatory Elements in the Ng-CAM Gene by Cellular Transfection Experiments-To locate the regions important for regulation of the Ng-CAM gene, 14 different reporter constructs were prepared and tested in transient transfection experiments in two mouse cell lines, N2A neuroblastoma cells and NIH3T3 fibroblasts. N2A cells express a number of neural CAMs including N-CAM and L1, a property suggesting that they may be well suited for assaying the expression of Ng-CAM constructs. NIH3T3 fibroblasts were selected as cells in which Ng-CAM constructs were expected to show little or no activity.
The various Ng-CAM reporter constructs and observed CAT activities in N2A and NIH3T3 cells are summarized in Fig. 3. Five reporter constructs containing different lengths of 5Јflanking sequence (see Fig. 3, constructs Ng4, Ng1.5, Ng447, Ng200, Ng447/182) were constitutively active in both NIH3T3 and N2A cells. The Ng447 construct, containing 447 base pairs of 5Ј-flanking sequence and 61 base pairs of the first exon, was approximately 2-fold more active than Ng4 and Ng1.5 in N2A cells, but not in NIH3T3 cells. The proximal promoter region within this construct contained two consensus binding sites for the SP1 transcription factor and a CRE. The Ng200 construct, in which the promoter region between Ϫ447 and Ϫ200 containing the CRE sequence was deleted, showed 3-fold less activity in both NIH3T3 and N2A cells than the Ng447 construct. This result indicated that regulatory elements between Ϫ200 and Ϫ447 are required for basal promoter activity. The promoterless CAT vector, pCAT-Basic (Bas) showed little, if any, activity in both cell types.
A construct designated Ng447/182, containing the entire first exon was approximately 2-fold more active than the Ng447 construct in both NIH3T3 and N2A cells. These data suggested that the segment between ϩ61 and ϩ182 is a positive regulatory region of the Ng-CAM gene. A construct designated Ng447ϩI, containing the Ng-CAM proximal promoter together with the entire first exon, first intron, and a portion of the second exon was active in N2A cells but was completely silent in NIH3T3 cells. This finding suggested that sequences within the first intron of the Ng-CAM gene were capable of silencing the gene in non-neuronal cells.
NRSEs Within the First Intron Silence Ng-CAM Expression in NIH3T3 Cells, but Not in N2A Cells-To pinpoint the sequences that silenced Ng-CAM gene expression in NIH3T3 cells, we examined various constructs having deletions of segments of the first intron of the Ng-CAM gene for loss of silencing in transfection experiments. Constructs with deletions of the NRSEs were tested in both NIH3T3 and N2A cells (Fig. 3). The Ng-CAM NRSEs show a high degree of similarity with the NRSE consensus sequence, but they are all in the orientation of the reverse complement from that found in the other genes (Fig. 4). Two constructs, designated EE and EA, which eliminated NRSEs1-4 and NRSEs1-5, respectively, showed an increase in activity (i.e. loss of silencer activity) as compared to the 447ϩI construct. The same deletion constructs showed only a slight increase in activity in N2A cells.
To examine whether the NRSEs were sufficient for silencing of the Ng-CAM proximal promoter, constructs were prepared in which the five NRSEs were placed upstream of the Ng447 region of the Ng-CAM gene linked to a CAT reporter. Two constructs were prepared in which the NRSEs were present either in the orientation normally found in the gene (SAS) or in the opposite orientation (SS). The SAS and SS constructs were both silent in NIH3T3 cells, showing CAT activities that were comparable to the Ng447ϩI and pCAT-Basic constructs (Fig.  3). However, in N2A cells, the SAS and SS constructs were highly active and showed no reduction of Ng-CAM proximal promoter activity. These studies suggest that, while the NRSEs are sufficient to silence Ng-CAM promoter activity in NIH3T3 cells, they do not show silencing in N2A cells. The Pax Motif Negatively Regulates Ng-CAM Gene Expression in Both NIH3T3 and N2A Cells-To test whether the Pax motif had an effect on Ng-CAM expression, three constructs were prepared and examined for activity (Fig. 3). A construct designated AS, having a 145-base pair deletion which removed the Pax motif, showed increased activity (i.e. a release from silencing) as compared to the Ng447ϩI construct. The AS deletion showed a release from silencing in both NIH3T3 and N2A cells (Fig. 3, compare AS to Ng447ϩI). The same effects were observed in a construct in which the GTTCC and ATTA sequences within the Ng-CAM Pax motif were deleted (see the construct designated Pax, Fig. 3). A construct designated ES, in which both the NRSEs and the Pax motif were removed was more active than all other deletion constructs in NIH3T3 cells. The effects of the NRSEs and the Pax region were additive, i.e. the level of activity from the ES construct was approximately equal to the sum of activities from the EA and AS constructs. The combined data suggest that, in some contexts, the NRSEs and the Pax motif may act together to regulate expression of the Ng-CAM gene.
Binding of Nuclear Proteins to Ng-CAM Regulatory Elements-To examine whether nuclear protein factors could bind to the various regulatory regions of the Ng-CAM gene, we conducted a series of gel mobility shift experiments using two segments of the Ng-CAM gene shown in these studies to be important for regulation in cellular co-transfection experi- ments: 1) the ϩ82/ϩ182 region, and 2) the NRSE region. Using the ϩ82/ϩ182 probe, a DNA-protein complex was detected using nuclear extracts prepared from both NIH3T3 and N2A cells (Fig. 5, panels A and B, lane 2). Formation of these complexes was inhibited when an excess of cold ϩ82/ϩ182 competitor was added (Fig. 5, panels A and B, lanes 3 and 4), but not when an excess of cold NRSE competitor was added (Fig. 5,  panels A and B, lanes 5 and 6) to binding reactions.
Binding experiments using the NRSE region as probe detected a clear cut DNA-protein complex with nuclear extracts prepared from NIH3T3 cells and a slight, but detectable complex with proteins from N2A cells (Fig. 5, panels C and D, lane  2). Formation of the NRSE-protein complex (particularly evident in NIH3T3 cells) was inhibited with an excess of cold NRSE competitor (Fig. 5C, lanes 3 and 4). Moreover, the ϩ82/ ϩ182 segment did not compete for the formation of the NRSE-protein complex (Fig. 5, panels A-D, lanes 5 and 6). Similar competition experiments performed in N2A cells showed no diminution of the faint signal, regardless of the competitor used (Fig. 5D, compare the band intensities in lanes 5 and 6 with those in lanes 2, 3, and 4). These results suggested that proteins that bind to the NRSE are much more abundant in NIH3T3 cells than in N2A cells. Furthermore, the nuclear proteins bound by the ϩ82/ϩ182 and the NRSE regions of the Ng-CAM gene appear to be different.
Binding of Pax-3 to the Pax Motif-To date, no target gene containing a binding sequence for Pax-3 has been identified in vertebrates. To determine whether the Pax motif in the first intron of the Ng-CAM gene could bind directly to Pax-3, we prepared a GST/Pax-3 fusion protein in bacteria. We tested this protein in gel mobility shift experiments for binding to the Ng-CAM Pax motif and to variants of this motif containing substitutions in the base pairs important for recognition of the DNA binding moieties of Pax proteins. Pax-3 was chosen for these experiments for two reasons. First, Pax-3 is known to be expressed during periods of neuronal differentiation and migration that coincide with Ng-CAM expression by postmitotic neurons (20,33). Second, the Pax motif in the first intron of the Ng-CAM gene resembled a binding sequence for a particular subset of Pax proteins that include Pax-3, namely those proteins that contain both a paired domain and a homeodomain. The Ng-CAM Pax motif contains a GTTCC sequence that, in other genes, has been shown to interact with the 128-amino acid paired domain (30) and an ATTA sequence that has been shown to interact with the paired-type homeodomain (20,32,34).
Four different probes were tested for binding to Pax-3 ( Either the ϩ82/ϩ182 (panels A and B) or the NRSE (panels C and D) probe was mixed with nuclear extracts prepared from NIH3T3 cells (panels A and C) or from N2A cells (panels B and D). For binding reactions, probes were incubated without extract (all panels, lane 1) or with the appropriate nuclear extract (lanes 2-6). DNA-protein complexes formed between the ϩ82/ϩ182 and NRSE regions of the Ng-CAM gene with nuclear extracts are indicated with brackets labeled EX and NRSE, respectively. Competitor DNAs were added to some binding reactions (all panels, lanes [3][4][5][6]. Competitors included either a 10-or 100-fold excess of the unlabeled DNA probe used in the specific binding reactions (all panels, lanes 3 and 4, respectively). To test whether the NRSE sequence could compete for binding of the ϩ82/ϩ182 probe to nuclear extracts, either a 10-fold or a 100-fold excess of cold NRSE DNA was added to the ϩ82/ϩ182 binding reactions (panels A and B, lanes 5 and 6, respectively). A similar test for cross-competition was performed by adding either a 10-or 100-fold excess of cold ϩ82/ϩ182 DNA to the NRSE binding reactions (panels C and D, lanes 5 and 6, respectively). Reactions were subjected to electrophoresis on a 4% polyacrylamide gel in 0.25 TBE buffer at 400 V for 2 h at 4°C. 6). The Ng-wt probe contained the wild-type Pax motif (Fig. 3, the region between 1679 and 1708). The Ng-H variant contained 3-base pair substitutions that altered the ATTA motif. The Ng-P variant contained 11 base pair substitutions. Four substitutions destroyed the GTTCC motif which has been shown to be essential for the binding of the e5 DNA sequence to the paired domains of Pax-1 and Pax-3 (30,32,35). The seven other substitutions made in Ng-P were introduced after comparing the sequence of the Ng-CAM gene Pax motif to the consensus binding sequences for Pax proteins (31,36,37) and took account of critical base pair substitutions that may disrupt paired domain interactions with DNA. The third variant, Ng-HP (Fig. 6), contained the combination of mutations made in both Ng-H and Ng-P variants.
The Ng-wt probe showed binding to the GST/Pax-3 protein, but no binding to the GST control protein. Similar binding experiments using the Ng-HP and Ng-H variants showed little or no binding to Pax-3. The Ng-P variant showed no detectable decrease in binding to Pax-3 as compared to that of the Ng-wt. This result suggested that the GTTCC motif and other base pairs important for paired domain interactions were not necessary for Pax-3 binding. Rather, these experiments suggested that it was the ATTA motif that was important for Pax-3 binding to the Ng-CAM Pax motif. DISCUSSION In the chicken, Ng-CAM is first detected in the central nervous system at embryonic day 3 in cells of the ventral neural tube that are the precursors of motor neurons (38). At later stages, Ng-CAM appears on a number of neurons and is distributed mainly on axons rather than cell bodies. In the cerebellum, Ng-CAM appears in Purkinje cells, is expressed by a number of different fibers, and plays a role in the migration of granule cell neurons on radial glia. The synthesis of Ng-CAM also shows dynamic changes during both myelination and dur-ing nerve regeneration. At the onset of myelination, Ng-CAM expression decreases in the central nervous system but not in the peripheral nervous system. Ng-CAM expression is increased after peripheral nerve injury in the spinal cord and in the nerve at the lesion site, but is decreased in dorsal root ganglia (39). These dynamic patterns of Ng-CAM synthesis in the nervous system prompted us to identify factors that regulate Ng-CAM gene expression.
Structure of the 5Ј End of the Ng-CAM Gene-We have isolated the 5Ј end of the chicken Ng-CAM gene (Fig. 7). The first seven exons encode Ng-CAM mRNA sequences which include the 5Ј-untranslated region, the initiator methionine, the signal peptide, the amino terminus, and the first two immunoglobulin domains of Ng-CAM (Fig. 1). Comparison of the partial Ng-CAM gene structure to the gene structures for other closely related neural CAMs such as L1 (40) and Nr-CAM 2 reveals a close correspondence in how untranslated and coding sequences are partitioned into exons. The immunoglobulin domains typical for these molecules are all encoded by two exons, a feature that sets them apart from immunoglobulin genes, consistent with an evolutionary divergence (41). Chicken Ng-CAM, mouse L1, and chicken Nr-CAM, and chicken neurofascin proteins share approximately 40% identity in their amino acid sequences (3,4,6). It has been long debated whether Ng-CAM and L1 are functional equivalents in the chicken and mouse. Based on the comparison of the 5Ј gene structures of these neural CAMs, it is still not possible to establish whether the Ng-CAM and L1 genes are in the same evolutionary lineage. Further comparison of the cis control elements of these neural CAM genes may resolve the issue and may also reveal similarities and differences in regulatory regions affecting expression of these genes for neural CAMs.
The Proximal Promoter-The Ng-CAM gene was determined to have a single RNA start site, yet the Ng-CAM promoter was found not to contain a TATA box, a sequence typically contained in promoters for genes having a single RNA start site. However, the sequence GGGGCGGGGC, which matches a consensus binding site for the SP1 transcription factor (28) was found immediately upstream of the RNA start site at Ϫ15. Thus, it is possible that SP1 and associated proteins may be part of a mechanism controlling transcription of the Ng-CAM gene (42).
A region of 447 base pairs of 5Ј-flanking upstream sequence was sufficient for basal Ng-CAM promoter activity in cells; addition of up to 4 kilobases of Ng-CAM 5Ј-flanking sequence showed no further increases in this activity. Potential regulatory sequences located in the proximal promoter included an additional SP1 motif at Ϫ134 and a consensus CRE (TGACGTCA) at Ϫ255 (Fig. 7). In both NIH3T3 and N2A cells, a 3-fold decrease in expression was observed for constructs that had a deletion in the Ϫ200 to Ϫ447 region of the Ng-CAM proximal promoter. This region contained the CRE and thus, trans-factors of the CREB family (43,44) may control Ng-CAM gene expression.
Inclusion of the region between ϩ82 and ϩ182 in the first exon of the Ng-CAM gene in constructs led to a 2-fold stimulation of Ng-CAM promoter activity in both NIH3T3 and N2A cells. It is likely that this sequence may impart either additional stability of Ng-CAM mRNAs or binds a transactivator important for transcription of the Ng-CAM gene. We found that the ϩ82 to ϩ182 region bound to nuclear proteins from both NIH3T3 and N2A cells.
Silencer Activity of NRSEs Located in the First Intron of the Ng-CAM Gene-We have located five NRSEs within the first intron of the Ng-CAM gene (Fig. 4). Single copies of the NRSE have been found in several neurally expressed genes including in the SCG10 gene (24), the gene for the rat type II sodium channel (45), and the gene encoding synapsin I (46). Five NRSEs appear in tandem within the first intron of the Ng-CAM gene and a tandem arrangement of NRSEs has been observed, for example, in the first intron of the rat brain derived neurotrophic factor gene (47), although so far no experiments have addressed whether this sequence has silencer activity.
In cellular transfection experiments, we found that the Ng-CAM NRSEs silenced the Ng-CAM proximal promoter in NIH3T3 cells. In contrast, NRSEs did not silence the promoter in N2A cells. Thus, the NRSEs silenced Ng-CAM gene expression in non-neuronal cells. The multiple copies of NRSEs in the Ng-CAM first intron suggest that they may bind proteins cooperatively. Recently, a neuron-restrictive silencer factor (NRSF, also called REST) has been identified which binds to the NRSE (48,49). The protein contains eight zinc fingers which are related to those found in proteins of the GLI-Krü ppel family.
Consistent with their activity patterns, we found that the Ng-CAM NRSEs formed a more prominent DNA-protein complex with nuclear proteins from NIH3T3 cells than they did with proteins from N2A cells. It is therefore likely that the protein enriched in NIH3T3 cells which binds to the NRSEs and silences Ng-CAM gene expression is NRSF/REST or a related protein. Interestingly, constructs with deletions in the NRSEs also show slightly elevated expression in N2A cells when compared to constructs containing the NRSEs (see Fig.  3). Thus, in N2A cells a minor amount of NRSF/REST may contribute to silencing, but the predominance of positive factors may greatly override this activity. In a preliminary search of the published L1 gene sequence (40) we have located an NRSE with the sequence TCTGCTGTCCGTGGTGCTGGA within the first intron at position 277-297. The possibility must therefore be considered that NRSEs may be used in the negative regulatory programs of other neural CAM genes in the Ng-CAM family.
Function of the Pax Motif and Binding to Pax-3-We have previously suggested that cell adhesion molecules are likely to be a major class of target genes regulated by homeobox and Pax transcription factors during embryonic development (15)(16)(17)(18)(19). In the present study, we were particularly interested in whether the regulatory sequences of the Ng-CAM gene, encoding a CAM that is restricted to the nervous system, are potential targets for Pax proteins. The cellular transfection experi-ments showed that constructs having deletions in either the region containing the Pax motif, or site-directed mutations within the GTTCC and ATTA sequences in the Pax motif, were more active than similar constructs containing the unmutated Pax motif. Furthermore, in NIH3T3 cells, the combination of NRSEs and the Pax motif was more effective in silencing than was the NRSE region alone. We found using gel mobility shift experiments that the Pax motif bound to Pax-3. This binding was disrupted by mutations in the ATTA sequence, but not by mutations in the GTTCC sequence and adjacent base pairs. These data demonstrate that Pax-3 binds directly to the Pax motif in the Ng-CAM gene and that this interaction is mediated primarily by the homeodomain and not by the paired domain of Pax-3. Previous studies showing specific Pax-3 binding to DNA have employed synthetic binding sequences derived from the even-skipped gene promoter from Drosophila (32,35). The Pax motif in the Ng-CAM gene described in the present study represents the first naturally occurring vertebrate target sequence found for the Pax-3 gene product.
Recently, a sequence containing an ATTA motif has been located at Ϫ170 in the promoter of the gene for the neural CAM L1 (40). While this motif has some similarities with the Ng-CAM Pax motif, it also shows some important differences. The two Pax motifs are similar in that they both contain ATTA motifs and share a few identities in the base pairs flanking this sequence. However, in contrast to the Ng-CAM Pax motif, the L1(Ϫ170) motif closely resembles the consensus binding site for Pax-6 (31, 37). L1(Ϫ170) has been shown to bind to Pax-6, but does not bind to Pax-3. Unlike the binding of the Ng-CAM Pax motif to Pax-3 which is eliminated when mutations are introduced into the ATTA sequence, the binding of L1(Ϫ170) to Pax-6 was eliminated by mutations disrupting sequences which interact with the paired domain (50). The binding was unaffected by mutations in the ATTA motif that disrupt homeodomain interactions. It will be revealing to determine whether other genes encoding neural CAMs that are related to Ng-CAM and L1 at the amino acid level, such as Nr-CAM and neurofascin, also contain Pax motifs. Genes encoding this family of neural CAMs may all contain Pax motifs with subtle variations in sequence composition that may determine binding preferences and selective CAM gene control by different Pax proteins.
It will be necessary to determine in vivo what roles the DNA control elements described here play in the developmental expression pattern of Ng-CAM in the nervous system. It is likely that particular combinations of Ng-CAM regulatory elements identified in this study (see Fig. 7) are utilized to control specific contexts of Ng-CAM expression during neural development and regeneration. For example, the NRSEs and Pax motif and their bound proteins may act combinatorially to restrict Ng-CAM expression to particular classes of neural cells during development. This possibility may be explored in chicken embryos by using retroviral vectors (51). Chicken Ng-CAM constructs can also be tested in transgenic mice, an approach has been used successfully across species to analyze the brainspecific expression directed by regulatory sequences from the chicken gene encoding the a2 neuronal acetylcholine receptor (52). Such animal studies will be particularly useful in determining how the regulatory regions identified in the present study function to determine place-dependent expression of Ng-CAM.