Multiple Regulatory Elements Control Transcription of the Peripheral Myelin Protein Zero Gene*

The gene encoding protein zero (P0), the most abundant protein of peripheral nervous system myelin, is expressed uniquely in Schwann cells. Previous studies have demonstrated that much of the cell type specificity of this expression is due to transcriptional control elements in the 1.1-kilobase pair 5′-regulatory region of the gene. We have now analyzed this region and have identified a set of functional elements in the 500 base pairs proximal to the transcription start site. DNA sequence conservation within the 5′ regions of the human, mouse, and rat P0 genes correlates closely with the results of promoter deletion analysis of the 1.1-kilobase pair region assayed in Schwann cell cultures and reveals a potent proximal region from position −350 to +45. Sites of protein/DNA interaction within the proximal 500 base pairs of the promoter were identified by footprinting assays. Functional transcriptional elements were identified within the protected regions in the proximal promoter by mutation and transient transfection analysis in P0-expressing cell lines. The core (or basal) P0 promoter is identified as two regulatory elements, a G/C-rich element that binds nuclear factor Sp1 and a CAAT box that binds NF-Y. These core elements are essential for the transcription observed from the transfected promoter in cultured Schwann cells.

The myelin sheath is a specialized membranous organelle of the vertebrate nervous system. Elaborated by oligodendrocytes in the central nervous system and by Schwann cells in the peripheral nervous system (PNS), 1 this organelle consists of a large sheet of plasma membrane that is repeatedly wrapped and very tightly compacted around axons (1). Myelin is essential for the rapid conduction of action potentials in vertebrates, and human diseases that compromise the integrity of the sheath (e.g. peripheral neuropathies in the PNS and multiple sclerosis in the central nervous system) are generally debilitating (2). The elaboration of myelin requires a substantial biosynthetic up-regulation of plasma membrane lipids and proteins, several of which are unique to the myelin sheath (3). In the PNS, the most abundant of these is protein zero (P 0 ), a 31-kilodalton, immunoglobulin-related, transmembrane glycoprotein that accounts for greater than 50% of the protein in mammalian PNS myelin (4). The mRNA encoding this protein is estimated to account for nearly 8% of the poly(A) RNA in actively myelinating Schwann cells (5).
A variety of studies have demonstrated that P 0 is essential for peripheral nerve development and function. Mice in which the P 0 gene has been deleted exhibit a severe peripheral neuropathy that includes hypomyelination, loss of motor control, tremors, and grossly abnormal myelin sheaths (6,7). Furthermore, mutations in the P 0 gene account for a variety of inherited human peripheral neuropathies, including Charcot-Marie-Tooth disease type 1B, Dejerine-Sottas syndrome, and congenital hypomyelination (8 -12).
High level P 0 expression is a distinctive feature of terminal differentiation in myelinating Schwann cells. Expression of P 0 mRNA peaks during the period of active peripheral myelination (the first 3 postnatal weeks in rodents) and is maintained at lower steady-state levels into adulthood (5,13). During Schwann cell development, the P 0 gene is coordinately induced together with genes encoding other myelin-specific proteins, such as myelin basic protein, PMP-22, and myelin-associated glycoprotein (14). In contrast to these proteins, however, P 0 is not expressed by oligodendrocytes in the central nervous system. The pronounced up-regulation of P 0 biosynthesis observed at the onset of overt myelination appears to be triggered by a signal associated with the surface of axons (15,16). This signal may be transduced intracellularly through elevation of cyclic AMP: in cultured Schwann cells, cAMP-elevating agents such as forskolin potentiate P 0 gene expression and strongly potentiate expression of less abundantly expressed myelin-specific genes (16,17). P 0 regulation during Schwann cell development reflects a combination of transcriptional and translational controls, although the former appears to predominate. Cell culture and transgenic mouse studies have demonstrated that major features of this regulation are controlled by a 1.1-kilobase pair region flanking the transcription start site of the P 0 gene. When linked to heterologous reporter genes, this 1.1-kilobase pair DNA fragment drives Schwann cell-specific, forskolin-inducible transcription upon transient transfection into cultured Schwann cells (18) and has been shown to target a number of transgenes exclusively to myelinating Schwann cells, on an appropriate developmental schedule, in transgenic mice (19,20). The 5Ј-DNA region therefore contains many if not most of the elements required for tissue-specific and developmentally accurate P 0 expression. To determine the cis-acting regulatory elements and transcription factors that control this expression, we examined protein/DNA interactions and promoter function of the P 0 5Ј-regulatory region in rat Schwann cells. An array of protected regions was detected on the proximal promoter, and among these a region of cell type-specific differences between P 0 -expressing and nonexpressing cell lines was noted. Muta-tional analysis of sequences within DNase I-protected regions revealed multiple functional elements, each of which is responsible for a different level of transcription. The core promoter is defined as the two proximal elements located within the protections at positions Ϫ48 to Ϫ59 and Ϫ66 to Ϫ79, which bind nuclear factors Sp1 and NF-Y, respectively.

EXPERIMENTAL PROCEDURES
Plasmids-The P 0 gene, 5Ј-regulatory sequences, and the P 0 CAT plasmids pPCATXA6 and pPCATHA16 have been previously described (18). The 5Ј truncation plasmids from Ϫ915 to ϩ45 were generated from the pPCATXA6 plasmid by Bal31 digestion, and the 5Ј termini were confirmed by sequencing. The P 0 -luciferase plasmid, Ϫ915P 0 LUC, was generated by cloning the Ϫ915 to ϩ45 P 0 promoter fragment (obtained by digesting pPCATHA16 (Ref. 18) with XbaI and HindIII) into the NheI and HindIII sites of the promoterless luciferase reporter vector, pGL2-B (Promega).
Cell Cultures-Rat Schwann cell cultures were prepared and maintained in media with forskolin and glial growth factor as described previously (21). Briefly, Schwann cells were prepared from 3-4-day-old Sprague-Dawley rat sciatic nerves. Cells were treated by anti-thy 1 and complement lysis to remove contaminating fibroblasts. Schwann cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, penicillin, and streptomycin; 2 M forskolin; and either 13 g/ml semipurified bovine pituitary extract containing glial growth factor or 10 ng/ml recombinant neuregulin (kind gift of Dr. Kuo-Fen Lee). Bovine pituitary extracts were prepared by methods previously described (22). The rat cell line B103 (23) is a P 0 -positive transformed line with Schwann cell properties (24). B103, HeLa, and the fibroblast cell line Rat2 were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and antibiotics.
Footprinting-DNase I footprinting assays were performed using Ϫ915P 0 LUC plasmid to generate radiolabeled probes of the proximal P 0 promoter (Ϫ550 to ϩ45) using published protocols (25). The P 0 promoter was first digested with HindIII and used to produce footprints most proximal to the transcription start site (proximal probes). The proximal, noncoding strand probe was made by end labeling with kinase and [␥-32 P]ATP, and the proximal, coding strand probe was made by fill in, using Klenow and [␣-32 P]dATP. The DNA was then digested with BglII, and the 600-bp fragment was purified by polyacrylamide gel electrophoresis. The distal P 0 probes were generated by digestion with BglII to generate the 5Ј-end of the probe and radiolabeling as above and then digestion with HindIII and purification by gel electrophoresis. Nuclear extracts from cultured cell lines were produced by standard methods (26). Rat Schwann cell cultures were induced to high levels of P 0 expression with 20 M forskolin for 2 days before extracts were made. The footprinting assay was performed by incubating radiolabeled probe with nuclear extracts for 10 min and digesting for 1 min on ice with DNase I. Appropriate dilutions of DNase I were predetermined for both probe alone and probe with various nuclear extracts. The reaction products were separated on 6% polyacrylamide sequencing gels using wedge spacers. The results were visualized by autoradiography.
In vivo DMS footprinting assays were performed in Schwann cell cultures treated for 2 days with 20 M forskolin utilizing the ligationmediated polymerase chain reaction method as described previously (27,28). Briefly, the procedure was as follows. Schwann cell cultures were treated with dimethyl sulfate (DMS), in parallel with untreated cultures, and the methylated genomic DNA was isolated and purified of protein. The untreated DNA was DMS-treated in vitro, and both samples were piperidine-cleaved. These DNA fragments were amplified by ligation-mediated polymerase chain reaction and radiolabeled with [␥-32 P]ATP using sets of three nested P 0 -specific primers. DMS-treated samples were electrophoresed on 6% sequencing gels. The primers for the coding strand were as follows: primer 1, 5Ј-CTGGGGTAGGGGCA-AGGG; primer 2, 5Ј-GTGGAGAGAGCGGGGGACAAG; primer 3, 5Ј-G-GGGTGGAGAGAGCGGGGGACAAGGAAC. The primers for the noncoding strand were as follows: primer 1, 5Ј-ACAATGCCCCTTCTGCTC; primer 2, 5Ј-CTGCCACCCTCCCCACCAC; primer 3, 5Ј-CCACCCTCC-CCACCACCTCTC.
Mutagenesis-Specific substitution and deletion mutations were generated in the Ϫ915P 0 LUC plasmid using two mutagenic oligonucleotides of the coding strand sequence, T4 DNA polymerase and the Transformer mutagenesis kit (CLONTECH). The specific sequence changes are described in Fig. 4A. Although the mutations in these plasmids were generated by a single round of synthesis and not by polymerase chain reaction, we observed that in more than one case two different mutagenized colonies targeted by the same mutagenic oligo yielded significantly different luciferase activities. Consequently, 500-bp fragments containing only the desired mutation were excised from all of the mutated plasmids and cloned into the corresponding site in Ϫ915P 0 LUC to eliminate the possibility of secondary mutations.
Transient Transfection and Reporter Gene Assay-Transfection with CAT plasmids was performed using a standard calcium phosphate procedure (18,29). Schwann cells were grown in medium with serum (2 M forskolin and 13 g/ml glial growth factor) and transfected in media alone, and after transfection, cells were incubated in medium with 2 M forskolin and without glial growth factor. Ten-g P 0 CAT 5Ј deletion plasmids were co-transfected with 2 g of RSV-luciferase plasmid as a measure of transfection efficiency. Cell extracts were made, and samples with equivalent luciferase activity were assayed for CAT activity using [ 14 C]chloramphenicol and thin layer chromatography as described previously (18,30). Results are shown relative to the activity of the full-length plasmid, pPCATXA6.
Transfection of cell lines with luciferase reporter plasmids was performed by the polyethylenimine (Fluka) complex formation method as described previously (31). In brief, DNA and polyethylenimine complexes were formed in 0.15 M NaCl for 30 min and applied to cell cultures grown to approximately 50% confluency. Complexes and cells were incubated for 3-4 h, the medium was removed, and fresh medium was applied. P 0 -luciferase plasmids were co-transfected with RSV/␤galactosidase plasmid used to normalize for transfection efficiency. Before transfection, Schwann cells were grown in medium containing 2 M forskolin and 10 ng/ml recombinant neuregulin, and after transfection cells were incubated in medium with 10 M forskolin and without recombinant neuregulin. Schwann cell lysates were made approximately 25 h later, and B103 cell lysates were prepared 42-44 h later. Luciferase assays and ␤-galactosidase assays were performed as described previously (18), or alternatively the Dual Light assay for both tests was used (Tropix, Inc.). Light units were normalized to ␤-galactosidase units, and the results with the mutant plasmids were expressed relative to Ϫ915P 0 LUC, which was set to 100. Shown are the results of five experiments, each performed in duplicate.

Sequence Conservation and Promoter Deletion Analyses Reveal a Potent Proximal
Promoter-We first examined sequence conservation between the aligned 5Ј-flanking regions of the human, rat, and mouse P 0 genes. These regions, including exon 1 and extending into intron 1, were pairwise aligned mouse to human and rat to human. Sequence homologies were determined by calculating the percentage of nucleotide identity within a 40-bp window whose 3Ј-end was located at the position relative to the transcription start site that is shown on the x axis of Fig. 1A. Comparisons were made by repositioning the 40-bp window both upstream and downstream from the transcription start site in increments of five nucleotides and then recalculating the percentage of nucleotide identity. The results of these comparisons reveal a strikingly high level of sequence identity between both the rat and human and the mouse and human genes, from position Ϫ350 to the transcription start site, and in the region of the first exon (Fig. 1A). The rat to human and mouse to human comparisons coincide remarkably along the entire region analyzed, including at the sharp decrease in sequence identity from position Ϫ350 to Ϫ400. In addition, two other peaks of sequence conservation are observed, one at positions Ϫ480 to Ϫ500 and a second less pronounced peak around Ϫ700. The regions from Ϫ800 to Ϫ900 and from ϩ200 to ϩ400 (first intron) are below the levels of significant similarity for a 40-bp window and are not considered to be evolutionarily conserved. Given that nearly the entire proximal 5Ј-regulatory region is strongly conserved among mammalian P 0 genes, it is difficult to identify potential discrete transcription factor binding sites using sequence conservation alone. However, the sequence of the atypical TATAA element, TTTTAA (Ϫ28 to Ϫ23), is conserved in both comparisons; pre-vious work has suggested that this sequence does indeed function as a TATAA element (18). When the three CAAT boxes at bp Ϫ78, Ϫ98, and Ϫ145 are compared, the sequences at bp Ϫ78 and at bp Ϫ145 are entirely conserved. It should be noted that the high sequence conservation observed over the entire proximal P 0 promoter, a conservation as high as that for the coding sequences in exon 1, is unusual.
To correlate the regions of high sequence identity with transcriptional function, 5Ј truncation mutants were made of the P 0 promoter from position Ϫ915 to Ϫ113 and linked to the CAT reporter gene. The 3Ј-end of the series was fixed at position ϩ45, within the P 0 5Ј-untranslated region. The regulatory activity of the 5Ј deletion mutants was analyzed by transient transfection and CAT assay in primary cultures of rat Schwann cells. The results are shown in Fig. 1B. The x axis of Fig. 1 denotes the 5Ј-end of the P 0 promoter sequence in each deletion plasmid, and the y axis denotes percentage of CAT activity relative to the full-length plasmid (Ϫ915/ϩ45), which was set to 100%. Note that the x axis of Fig. 1B (transcriptional activity) is aligned with the x axis of Fig. 1A (sequence conservation). Sequential 5Ј deletions from Ϫ915 to Ϫ400 resulted in a progressive, although modest, decrease in promoter activity up to the Ϫ400 region. At position Ϫ400, about 60% of the promoter activity remained. Within the Ϫ400 to Ϫ200 region, however, each deletion resulted in a substantial loss of CAT activity, indicating that important regulatory elements were progressively deleted (Fig. 1B). Deletions from Ϫ200 to Ϫ100 reduced activity to near background levels. The presence of negative regulatory elements, as indicated by an increase in reporter activity upon deletion, was not observed. The relative promoter activities of the 5Ј deletion mutants correlate well with the amount of sequence identity within the P 0 promoter (Fig. 1,  compare A and B), in that the region between Ϫ350 and Ϫ300 where sequence conservation rises to Ͼ80% is the same region in which 5Ј deletions begin to remove strong regulatory elements. A 3Ј deletion series starting at position ϩ45, (whose 5Ј-end was fixed at ϩ915) revealed a modest diminution of activity until the deletions extended through the putative TATAA element at Ϫ23 to Ϫ28, at which point all transcriptional activity was lost (data not shown).
DNase I Footprint Analysis Reveals Sites of Protein/DNA Interactions-Given the results of the deletion survey, we performed in vitro binding studies on the proximal P 0 promoter from Ϫ550 to ϩ45 to delineate sites of protein/DNA interaction. In these experiments, nuclear extracts were prepared from two P 0 -expressing cell lines: 1) primary cultures of rat Schwann cells grown in 20 M forskolin and 2) B103 cells, a P 0 -positive transformed rat cell line. Extracts were also prepared from two nonexpressing cell lines, Rat-2 fibroblasts and HeLa cells. Extracts from B103 cells produced a number of distinct protections on the proximal 600-bp coding strand probe ( Fig. 2A shows several of these). Protected regions labeled 7, 9, and 10 were consistently strong and distinct; however, some sites, such as sites 8 and 6, were weak in comparison. Similar contacts were observed in nuclear extracts from cultures of rat Schwann cells (Fig. 2B), and labeling the noncoding strand of the proximal probe revealed protected regions that confirmed the boundaries of the protections observed with the coding strand probe (Fig. 2B). Footprint analysis using the distal probes detected (in addition to protection labeled 10) three additional protections on the promoter from Ϫ393 to Ϫ493, which were designated 11, 12, and 13 (summarized in Fig. 3). Hypersensitive bands were noted flanking regions 5, 9, 10, and 11 (Fig. 2, arrows). Thus, 13 sites of protein/DNA contact on the proximal promoter from the start site to Ϫ500 were delineated. Neighboring protections, such as sites 7 and 8, and overlapping FIG. 1. Sequence conservation correlates with transcriptional function. A, comparison of P 0 gene sequence between aligned human and rodent species. Percentages of nucleotide identity were determined within 40 nucleotide windows analyzed at 5-nucleotide intervals from Ϫ900 to ϩ400. The dotted line is comparison between rat and human, and the solid line is the comparison between mouse and human. The y axis shows the percentage of nucleotide identity, and the x axis shows the position of the 3Ј-nucleotide of each 40-bp window analyzed, relative to the transcription start site. B, transient transfection analysis of 5Ј deletion mutants of the P 0 promoter. Ten g of each P 0 CAT plasmid and 2 g of RSV-luciferase plasmid were co-transfected into rat Schwann cell cultures. Promoter activity is expressed as a percentage relative to CAT activity obtained with the full-length pPCATXA6 construct. The 5Ј-end of each deletion construct analyzed is indicated on the x axis. Values represent the average of four separate transfections.
sites, such as sites 3-5, may be the result of binding by one protein or by multiple proteins.
The P 0 promoter mediates strong cell type-specific expres-sion. Protein/DNA contacts were compared using nuclear extracts from P 0 -expressing and nonexpressing cell lines. Extracts from the fibroblast cell line Rat2 produced similar protections over several regions (Fig. 2B). However, the protections created by Rat2 extracts appeared to be considerably weaker than the Schwann cell and B103 protections on site 4 and, even more so, on site 5. HeLa cell extracts were used as an additional nonexpressing cell type in footprint experiments utilizing both the distal and proximal end probes. Distinct differences in the boundaries for protections 3-5 were again observed with these extracts (Fig. 2C). Site 3 in B103 extracts is only weakly protected, while in HeLa it is quite strong. Also, a boundary is not evident between sites 3 and 4 with HeLa extracts, but it is present with B103 cell extracts. Similarly, a boundary of unprotected sequence appears between sites 4 and 5 with HeLa extracts but not with B103 extracts, and the protection of site 5 itself is again much weaker with HeLa extracts. These results show that while several sites of protein/ DNA contact observed on the P 0 proximal promoter are the same for extracts prepared from expressing and nonexpressing cells, cell-specific differences in binding intensity and boundaries are detected, particularly at sites 3-5. A summary of protected sites created by expressing and nonexpressing cell types is included in graphic form in Fig. 3. The importance of the DNase I-protected sequences was confirmed in a series of in vivo footprinting experiments using primary cultures of rat Schwann cells. The in vivo assay has the advantage of detecting protein/DNA contacts on the promoter in the nucleus of living cells that are actively expressing the gene. The DMS treatment allows subsequent cleavage by piperidine at guanine, and sometimes adenine, residues. The region from Ϫ250 to ϩ45 was analyzed, and the protections observed in vitro within this region overlapped with the protected sequence noted by genomic footprinting. A selected ex-  ample of the protections observed in vivo on sites 3-5 is shown (Fig. 2). The two G residues in 5Ј-CAATTGG (site 4) are strongly reduced, and two G residues in the 3Ј-end of site 5 are also protected on the coding strand (Fig. 2D). In addition, the run of five G residues of site 3 is strongly protected on the noncoding strand (Fig. 2E). The protections observed in vivo in Schwann cells suggest that interactions detected in vitro are biologically relevant in the context of living cells.
Since the protected regions revealed by DNase I and genomic footprinting are probably binding sites for transcription factors, the 600-bp proximal promoter was scanned for known transcription factor binding sites within the protected sequences. This search again identified potential CAAT box binding sites located in protected sites 4 and 5 for known CAATbinding proteins including NF-Y (see Fig. 3). In addition, immediately downstream of the CAAT sequence in site 5 is a CACATG motif, a match for the E box consensus sequence, CANNTG. Protected sites 3, 6, 7, and 9 are G/C-rich and were identified as potential AP-2, T-antigen, or PER2 consensus binding sites. In addition, a sequence homology within site 7 was also identified as a possible binding site for sterol response element-binding protein. Footprint region 10 contains a 34-bp palindromic sequence; however, potential DNA-binding factors for this site were not identified. This site and other P 0 promoter binding sites may therefore correspond to recognition sequences for novel transcription factors. It is interesting to note that site 10 displays significant sequence similarity to a footprint in another myelin gene promoter, the proteolipid protein gene (26).
Transcriptional Activities of Protein Binding Sites-We measured the contribution of each of the 13 protected regions to transcriptional activity of the promoter by performing a mutational analysis of individual sites in the context of the fulllength P 0 promoter linked to a luciferase reporter gene (Fig. 4). Most sites were altered by substitution mutagenesis, and four large protections were completely deleted (sites 7, 9, 10, and 13). Site 5 was divided into two separate mutants corresponding to sequence homologies in the 3Ј (CANNTG) and 5Ј (CAAT) regions of the protection and are referred to as 5A and 5B, respectively (see Fig. 4A). Site 1 overlaps the putative cap site, and site 2 overlaps the putative TATAA element; these sites are not shown in this study. Each construct was introduced by transient transfection into rat Schwann cells, and cell extracts were subsequently assayed for luciferase activity and normalized for transfection efficiency (see "Experimental Procedures"). This analysis revealed varying levels of transcriptional activity associated with individual sequence mutations (Fig. 4,  black bars). Substitution of four nucleotides within each of two FIG. 4. Regulatory function of protein-binding sites is determined by transient transfection analysis. A, design of the P 0 -LUC mutant plasmids shows the P 0 proximal promoter region from Ϫ492 to the transcription start site. The upper sequence is the wild type P 0 promoter sequence, and the lower sequence is the mutated sequence. The regions mutated by deletion are indicated by ⌬. B, P 0 LUC plasmids containing individual mutations in the protein binding sites were transfected into rat Schwann cell cultures (black columns) and into B103 cells (gray columns). Results are normalized for transfection efficiency using ␤-galactosidase activity and reported as percentages of the wild type activity obtained with the parental plasmid, Ϫ915P 0 LUC. sites, site 3 and site 4, had the strongest effects on promoter activity. Mutation of site 3, a potential G/C box, and of site 4, a potential CAAT box, reduced activity to 21 and 6% of the wild type levels, respectively. Mutation of elements 7, 8, and 9 also had a substantial effect on luciferase expression and reduced activity to 52, 55, and 67% of the wild type activity, respectively. Mutation of sites 5A, 5B, 6, 11, and 13 individually reduced activity minimally, resulting in activities between 70 and 80% of the wild type. Mutation of sites 10 and 12 had little or no effect, reducing activity to 88 and 100% of wild type, respectively. In summary, five individual mutations resulted in a significant loss in promoter activity (mutations 3, 4, 7, 8, and 9), while six of the other mutations had a more modest effect on transcription (mutations 5A, 5B, 6, 10, 11, and 13).
To confirm and extend these results, the mutant promoter constructs were also analyzed by transient transfection in B103 cells (Fig. 4, gray bars). The effects of mutations in sites 3, 4, 5A, 6, 7, 8, 10, 11, and 12 were similar (Ϯ20%) to the results in rat Schwann cells, and mutations in sites 3 and 4 again had the greatest effect on luciferase activity. Significant differences in promoter activity were observed with sites 5B, 9, and 13. These differences may reflect biological differences between the B103 tumor cell line and primary Schwann cells. In this regard, it should be noted that 100% wild type activity of Ϫ915P 0 LUC in Schwann cells corresponds to ϳ60,000 light units/20-l sample, while B103 cells yield ϳ4,000 light units/20-l sample for the same construct. Maximal promoter usage reflected by absolute light units measured could be dependent on the activity of elements 5B, 9, and 13, whose functions appear to be altered in B103 cells.
Differences in the boundaries of binding sites 3-5 were observed when comparing nuclear extracts from P 0 -expressing cells and nonexpressing cells in DNase I footprinting experiments. We further examined the cell type specificity of these promoter elements by comparing their activity in transient transfections in nonexpressing Rat2 cells and in Schwann cells. The region from Ϫ113 to ϩ45 (including sites 3-5) was cloned upstream of the luciferase reporter gene and tested as described in Fig. 4. When luciferase activities in both cell types were normalized to the background vector alone (pGL2-B), transcriptional activity from this minimal promoter construct was 5.0 Ϯ 1.0 relative light units in Schwann cells compared with 4.5 Ϯ 0.5 units in Rat2 cells. In contrast to the full-length 1.1-kilobase pair P 0 regulatory region (18), the P 0 promoter from Ϫ113 to ϩ45 appeared to have very similar activities in expressing and nonexpressing cell lines, indicating that strong, independent Schwann cell-specific elements were not included in this construct. In addition, when plasmids Mut 3 and Mut 4 were assayed in Rat2 cells luciferase activity dropped to 34 and 29% of wild type promoter activity, respectively, indicating that these elements were necessary for the low levels of transcription detected in nonexpressing cells. This confirms that most of the exuberant cell type specificity of the 5Ј P 0 regulatory region lies upstream of the core promoter.
Electrophoretic Mobility Shift Assays-Binding activities of the functional P 0 promoter elements were further examined by EMSAs, in which radiolabeled oligonucleotide probes of the protected sites were used to identify protein complexes in Schwann cell nuclear extracts. P 0 probes detected strong DNA binding activity in Schwann cell extracts (Fig. 5), with the exception of probes for sites 5 and 8; weak complexes were detected with site 5 only upon 10 times longer exposure of the gel. Since cAMP regulators, such as forskolin, modestly upregulate P 0 expression in cultured Schwann cells, we used EMSA to test whether forskolin-inducible complexes could be detected on the P 0 promoter elements exhibiting in vivo func-tion. Nuclear extracts from Schwann cells treated with 20 M forskolin for 2 days were compared with extracts from untreated cells. Complexes with similar mobility formed using induced and uninduced Schwann cell extracts on each probe (Fig. 5). However, a second induced complex with slower mobility appeared in forskolin-treated extracts at site 7, and a modest and reproducible increase in band intensity was detected on site 11 in extracts from induced cells. As a control for the forskolin response, an Ig enhancer probe containing the octamer sequence detected Oct-1 (upper band) in nontreated cells and an additional band corresponding to the cAMP-inducible SCIP protein (lower band) in forskolin-treated cells, as described previously (33,34). P 0 site 3 is responsible for a large portion of the promoter activity assayed in the transient transfection experiments (see Fig. 4), and a sequence-specific complex was formed on probe 3 in Schwann cell nuclear extracts (Fig. 6). Since site 3 is a G/C-rich regulatory element, as are the Sp1 and Krox-20 consensus binding sites, we performed oligonucleotide competition assays with cold competitors for these sites. Complex formation was inhibited by unlabeled Sp1 competitor but not by a cold Krox-20 binding site (Fig. 6, lanes 1-4). To confirm the identity of the complex on site 3, we performed antibody supershift experiments with antibody specific for Krox-20 and antibody that reacts with Sp1 but not Sp2, Sp3, or Sp4 (Santa Cruz Biotechnology, Inc.). The complex on probe 3 was not effected by Krox-20-specific antibody; however, Sp1 antibody caused a marked reduction of complex formation and the creation of a supershifted band (lanes 5-7). The presence of Sp1 protein in Schwann cells was confirmed with a consensus binding site of Sp1 and the Sp1 antibody (lanes 8 and 9).
Mutation of site 4 had the most dramatic effect on P 0 transcription. In EMSA with Schwann cell extracts, a site 4 probe bound a doublet of fast mobility referred to as C1 (Fig. 7A, lane  1). Competitions were performed with the site 4 probe using cold competitor oligonucleotides of binding sites for known CAAT box-binding proteins. The doublet C1 was specifically competed with the competitor site 4; however, binding sites for C/EBP and NF-Y did not alter the formation of the P 0 site 4-specific complexes (lanes 2-4). This suggests that C1 is not the same as or similar to either of these CAAT box-binding proteins. We further pursued binding activities in this region by synthesizing a probe that included protection 5 as well as protection 4, referred to as probe 4 ϩ 5. In addition to the C1 doublet, the 4 ϩ 5 probe bound an additional complex with a slower mobility in the gel, designated C2. Complex C2 was eliminated by cold 4 ϩ 5 competitor and was weakly competed by cold site 4 alone but not by an irrelevant oligo competitor, Sp1 (Fig. 7B, lanes 5-7 and 9). Cold competitor composed of site 5 alone did not effect C2 formation, indicating that the complex does not bind to the upstream CAAT sequence in site 5 (lane 8). When competition assays were performed with the 4 ϩ 5 probe using known CAAT box binding sites as competitors, C2 was competed by the NF-Y binding site (lanes 10 and 11), but competitors of NF-1 and C/EBP binding sites did not compete (data not shown). NF-Y is a widely expressed transcription factor composed of three subunits, NF-YA, NF-YB, and NF-YC, and was detected in cultured Schwann cell nuclei by immunofluorescent staining (data not shown). This prompted us to probe the C1 and C2 complexes in supershift experiments with a well characterized antibody specific for the NF-YA subunit (35). While normal rabbit serum and AP-2-specific antibody did not react with either C1 or C2, NF-YA antibody supershifted the C2 complex, indicating that this complex in Schwann cell extracts contains NF-Y (lanes 12-16). The NF-Y control probe (major histocompatibility complex class II Y box sequence) bound a complex of the same mobility in Schwann cell extracts and was also supershifted by the NF-Y antibody (Fig. 7C, lanes [17][18][19][20]. Notably, this antibody did not alter the probe 4-specific doublet, C1 (lane 14) and did not supershift C1 on the probe containing site 4 alone (data not shown). Whereas C1 reflects the binding of an as yet unidentified protein, these results suggest that C2 contains transcription factor NF-Y in Schwann cell nuclear extracts. DISCUSSION A variety of previous studies have demonstrated that most of the specificity of P 0 gene expression is due to transcriptional control and that the kilobase of DNA flanking the 5Ј-end of P 0 carries the requisite elements for Schwann cell-specific expression and appropriate developmental regulation of this gene. In particular, mouse transgenes that are placed under the control of this regulatory region are expressed exclusively in Schwann cells and are strongly up-regulated at the onset of peripheral myelination (19,20). Similarly, reporter genes linked to the same 5Ј-flanking region are transcribed at exceptionally high levels when transfected into Schwann cells but not when introduced into non-Schwann cells (18). The ability to grow and transfect rat Schwann cells has afforded us the opportunity to examine transcriptional control elements of the P 0 gene in the normal, untransformed cell type in which it is expressed, using in vitro systems that are amenable to manipulation.
Our results have revealed several interesting features of the structure and function of this upstream regulatory region. First, nearly the entire 350-bp region proximal to the transcription start site is highly conserved between three mammalian species. This nucleotide conservation is typically in the range of 90% for a 40-bp window, slightly higher than the sequence conservation observed between rat (or mouse) and human for several of the coding regions of the P 0 gene. The 5Ј deletion analysis described above demonstrates that this unusual sequence conservation corresponds to what is functionally the most significant region of the promoter. Second, the activity of the P 0 regulatory region seems to reflect the cumulative and concerted action of a large number of transcription factors, most of which appear to be activators. This is reflected in both the sequential and progressive (as opposed to large stepwise) loss in promoter activity in the 5Ј deletion series and in the large number of sites identified by footprinting analyses. Although a number of neuronal genes appear to be regulated in a cell type-specific manner by silencer elements (36), such elements were not detected in the promoter truncation experiments or in the functional assays of identified footprint sites.
Substantial transcriptional activity was localized to the region from Ϫ500 to ϩ45. Both this region and regions upstream were analyzed by in vitro DNase I footprinting, and 13 protected sites were detected using nuclear extracts from cultured Schwann cells and B103 cells. The protected regions may contain individual transcription factor binding sites or may be composite sites containing recognition sequences for more than one factor. In addition, neighboring sites such as 7 and 8 may be the result of the binding of one protein. The region most proximal to the transcription start, sites 3-5, contains a G/Crich sequence, two CAAT boxes, and an E box homologue. Mutations in site 3, the most proximal G/C box, and site 4, the most proximal CAAT box, had the most significant effects on promoter activity, each reducing activity to less than 20%, and were analyzed further by in vitro binding assays. Based on their position relative to the transcription start site, their essential functional activity, and their sequence, we have assigned sites 3 and 4 as the core or basal promoter of the P 0 gene.
While the 1.1-kilobase pair P 0 promoter has been shown to be a powerful and compact regulatory region when used in a number of transgenic mouse experiments, our knowledge is limited concerning in vivo regulation of P 0 transcription. Recent studies suggest that the steroid hormone progesterone may play a role in myelination (37). It will also be of interest to determine if the sequence at site 7 with similarity to the sterol response element is bound by sterol response element-binding protein, since this transcription factor positively regulates a number of genes required for lipid and fatty acid biosynthesis (23). While myelin gene expression is strongly up-regulated at the onset of myelination in Schwann cells, the production of membrane components is even more strongly increased (1), and sterol response element-binding protein could be an efficient means to coordinately regulate expression of myelin membrane proteins and lipids. Since agents that elevate intracellular cAMP (e.g. cholera toxin, dibutyryl cAMP, and forskolin) upregulate P 0 mRNA and protein in cultured Schwann cells with a delayed time course (33,38), we examined changes in protein binding on the P 0 promoter elements upon forskolin treatment and saw modest changes with two P 0 probes. An induced DNAbinding activity was formed on site 7, and an increase in band intensity was observed on site 11 that parallels the amount of P 0 up-regulation observed at the mRNA level. The proteins that recognize these sites may therefore be subject to indirect cAMP regulation. Other changes in promoter activity may not be detected at the level of protein/DNA interaction because they may involve either the modification of existing factors or the interaction of proteins off the DNA.
We propose that as with other tissue-specific promoter regions, specificity is influenced by the binding of a particular set of factors and the precise orientation of these factors on the DNA. In this respect, the footprinting results suggested sites 3-5 as a region of cell type-specific differences within the P 0 promoter. However, further transfection analysis reveals that the core elements, sites 3-5, are functional elements in a nonexpressing cell line. The differences in binding properties observed may reflect the binding of related but distinct factors in different cell types or interactions between factors on the core promoter and those on upstream elements. Alternatively, the footprint differences may reflect post-translational modifications of the same factors or the presence of tissue-specific protein partners. It is also possible that a Schwann cell-specific enhancer may lie in the upstream regulatory region from Ϫ500 to Ϫ1100.
The mutation with the most effect on promoter function is a 4-base pair change in the GCAATT sequence at protected site 4 (Ϫ66 to Ϫ79). This alteration also changes three nucleotides of the overlapping sequence CCAATT, on the opposite DNA strand and in the opposite orientation. Although according to our data either of the CAAT homologues may be the functional binding site, the CCAAT on the noncoding strand (Ϫ64 to Ϫ77) is more similar in sequence to the consensus NF-Y element, (A/G)(G/A)CCAAT(C/G)(A/G)G(C/A) (39), than is the GCAAT element on the coding strand (9/11 versus 6/11 nucleotide similarity). A number of CAAT box binding factors and their isoforms have been identified and cloned, among them C/EBP, NF-Y, and NF-I (40 -43). Our results demonstrate that the site 4 ϩ 5 probe binds NF-YA or a closely related protein complex in Schwann cell extracts. NF-Y is a ubiquitously expressed heteromeric complex of three subunits, all of which are required for binding to DNA (44). NF-Y binding sites are found in a number of promoters, including those of tissue-specific genes, and are frequently positioned between Ϫ60 and Ϫ130, as is the case for P 0 site 4 (45). NF-Y functions in several promoters through protein/protein interactions with transcription factors positioned on neighboring sites and has been shown to interact with factors in a cooperative manner (35,39,46,47).
The G/C-rich protection at site 3 (Ϫ48 to Ϫ59) is the other important element identified in the core promoter. Mutation of four of the cytosines reduced promoter activity to one-fifth of the wild type promoter activity. While Sp1 is the most common factor binding the G/C box in proximal promoters of genes analyzed, Krox-20 recognizes G/C-rich sequences quite similar to the Sp1 site (48). Krox-20 has previously been shown to have an important role in Schwann cell development demonstrated by the Krox-20 knockout mice, which have a severe defect in PNS myelination and barely detectable P 0 expression (49). However, competition experiments with consensus binding sites as well as antibody supershift experiments with specific antisera demonstrate that Sp1 or an Sp1-related protein binds site 3. As is the case for similarly positioned Sp1 sites in other genes, Sp1 may play a role in regulating basal transcription of the P 0 gene. In addition, Sp1 controls basal transcription via interaction of its glutamine-rich activation domain with the TATAA-binding protein-associated factors (50). In this regard, it would be interesting to determine if P 0 transcription is mediated by an Sp1/NF-Y interaction or alternatively an Sp1/TAF interaction.
In summary, a composite of regulatory elements are found within the 500-bp proximal P 0 promoter, with various elements exhibiting a range of transcriptional and protein binding activities. The P 0 basal or core promoter is composed of a CAAT element that binds NF-Y and a G/C element that binds Sp1. We now need to identify the factors binding the P 0 regulatory elements, to assess the extent to which functional interactions occur between these proteins, and to define their precise roles in regulating P 0 transcription. Since P 0 expression is increased substantially in myelinating peripheral nerves compared with Schwann cells maintained in culture, our future plans are to test the significance of these factors for P 0 production during myelination. The identification of the DNA binding proteins and the factors that activate P 0 will allow us to begin to understand the cell type specificity and developmentally regulated expression of this major myelin protein.