Regulation of Tissue-specific Expression of Alternative Peripheral Myelin Protein-22 ( P M . 2 2 ) Gene Transcripts by Two Promoters*

Mutations affecting the peripheral myelin protein-22 (PMP22) gene have been shown to be associated with inherited peripheral neuropathies. To provide the mo- lecular basis for the analysis of such mutations, we have cloned and characterized the human PMP22 gene. It spans approximately 40 kilobases and contains four coding exons. Detailed analysis of its 5"flanking region sug- gested the presence of two alternatively transcribed, but untranslated exons. Mapping of separate PMP22 mRNA transcription initiation sites to each of these exons indicates that PMP22 expression is regulated by two alternatively used promoters. In support of this hypothesis, both putative promoter sequences demonstrated the ability to drive expression of reporter genes in transfection experiments. Furthermore, the structures of the 5'-portions of the PMP22 genes appear to be identical in rat

Mutations affecting the peripheral myelin protein-22 (PMP22) gene have been shown to be associated with inherited peripheral neuropathies. To provide the molecular basis for the analysis of such mutations, we have cloned and characterized the human PMP22 gene. It spans approximately 40 kilobases and contains four coding exons. Detailed analysis of its 5"flanking region suggested the presence of two alternatively transcribed, but untranslated exons. Mapping of separate PMP22 mRNA transcription initiation sites to each of these exons indicates that PMP22 expression is regulated by two alternatively used promoters. In support of this hypothesis, both putative promoter sequences demonstrated the ability to drive expression of reporter genes in transfection experiments. Furthermore, the structures of the 5'-portions of the PMP22 genes appear to be identical in rat and human, supporting the biological significance of the observed arrangement of regulatory regions. The relative expression of the alternative PMP22 transcripts is tissue-specific, and high levels of the exon 1A-containing transcript are tightly coupled to myelin formation. In contrast, exon 1B-containing transcripts are predominant in non-neural tissues and in growth-arrested primary fibroblasts. Interestingly, although a strong upregulation of PMP22 mRNA was observed in cultured Schwann cells in the presence of the adenylate cyclase activator forskolin under various culture conditions, the regulation of the different PMP22 mRNA species did not mimic the regulation that occurs during myelin formation in vivo. The observed regulation of the PMP22 gene by a complex molecular mechanism is consistent with the proposed dual role of PMP22 in neural and non-neural tissue.
The 22-kDa peripheral myelin protein PMP22 has recently been characterized as a major protein component of peripheral * Contributed equally to this work. e To whom correspondence should be addressed. nerve myelin (reviewed by Suter et al. (199311, where it comprises about 2 4 % of the total protein found in isolated myelin preparations (Kitamura et al., 1976;Snipes et al., 1992;Pareek et al., 1993). PMP22 protein is predominantly expressed throughout the full thickness of the compact myelin sheath , although expression of PMP22 mRNA has also been found in extra-neural tissues like lung, gut, and heart (Manfioletti et al., 1990;Spreyer et al., 1991;Welcher et al., 1991;Patel et al., 1992). Interestingly, the molecular cloning of PMP22 cDNA by differential hybridization Welcher et al., 1991;Spreyer et al., 1991) revealed that PMP22 is identical to the growth arrest-specific mRNA 3 (gas-31, a molecule whose expression is negatively correlated with cellular growth in cultured fibroblasts (Manfioletti et al., 1990;Suter et al., 199213). The significance of the identity between PMP22 and gas-3 and its implication for the function of PMP22 in and outside of the peripheral nervous system remain to be determined.
Point mutations in the mouse pmp22 gene are most likely responsible for the severe myelin deficiencies and the excessive proliferation of Schwann cells that underlie the profound peripheral neuropathies in the neurological mouse mutants, trembler and trembler-J (Suter et al., 1992a(Suter et al., , 1992b. Based on these findings and its location in a region of conserved synteny, PMP22 has been proposed as a candidate gene for the most common inherited peripheral neuropathy in humans, Charcot-Marie-Tooth disease type 1A (CMTlA),' and the rarely encountered Dejerine-Sottas syndrome (Suter et al., 1992aSnipes et al., 1993). In support of this hypothesis, the human PMP22 gene was mapped within a large duplication on chromosome 17 Timmerman et al., 1992;Valentijn et al., 1992a;Matsunami et al., 1992), a genetic abnormality associated with the majority of familial and spontaneous CMTlA cases (Lupski et al., 1991;Raeymaekers et al., 1991;Hoogendijk et al., 1992;Wise et al., 1993;Ionasescu et al., 1993). Confirmatory evidence for the crucial role of PMP22 in the dysmyelinating neuropathies CMTlA and Dejerine-Sottas syndrome has come from the finding of point mutations within the PMP22 gene in some patients who do not carry the CMTlA duplication (Valentijn et al., 1992b;Roa et al., 1993aRoa et al., , 1993b. Furthermore, PMP22 has also been implicated in an-7ho Promoters for Alternative PMP22 Transcripts other peripheral neuropathy, hereditary neuropathy with liability to pressure palsies (HNPP; tomaculous neuropathy), which is associated with a deletion of the same chromosomal region that is duplicated in CMTlA patients (Chance et al., 1993;Nicholson et al., 1994). Thus, a variety of genetic mechanisms involving PMP22, including point mutations, duplications, and deletions, may give rise to distinct hereditary peripheral neuropathies. Based on these results, the copy number of the PMP22 gene has been hypothesized t o be a major determining factor in the molecular pathology of hereditary peripheral nerve myelin disorders (reviewed by Snipes et al. (1993), Lupski et al. (1993), Suter and Patel (1994), and Patel and Lupski (1994)). According to this hypothesis, inheritance of only one copy of the PMP22 gene results in HNPP, two copies (the diploid number) results in normally myelinated nerves, three copies (dup(l7p)) leads to CMTlA, and four copies, as seen in homozygous CMTlA patients, results in a very severe CMT1-type peripheral neuropathy. Thus, it seems reasonable to assume that the correct regulation of PMP22 is critically important for myelin function. Furthermore, based on teleological considerations, one would expect that the regulation of PMP22 expression should be exceptionally tightly controlled.
The myelin genes offer an intriguing experimental system to study coordinate transcriptional regulation since the expression of structural myelin proteins appears to be up-regulated with the onset of myelination during development and regeneration of the peripheral nervous system (PNS) (reviewed by Snipes et al. (1992Snipes et al. ( , 1993). In that context, it has been suggested that a master regulator might exist in glial cells, comparable to MyoD and its role in muscle development (Berndt et al., 1992). PMP22 expression is regulated similarly to other structural proteins found in compact myelin-like myelin basic protein (MBP), protein zero (PO) (Spreyer et al., 1991;Kuhn et al., 1993). Thus, a detailed comparison of the individual promoters regulating the major PNS myelin protein genes is likely to yield information to support or refute the hypothesis of coordinate gene regulation by common transcription factors in Schwann cells. The molecular mechanism governing the expression of the PMP22 gene appears particularly interesting since PMP22 is also expressed in non-neural tissues. It is anticipated that some of this PMP22 expression is due to innervation by peripheral nerves. In addition, however, PMP22 may subserve a function that is not restricted t o its role in PNS myelin.
In this paper, we describe the detailed characterization of the structure of the human PMP22 gene, providing the basis for a genomic DNA-based test system to identify point mutations in PMP22 that co-segregate with the neurodegenerative phenotype in patients affected with hereditary motor and sensory neuropathies (Roa et al., 1993a(Roa et al., , 1993b. Our results also indicate that the PMP22 gene is regulated by two alternative promoters that are located immediately upstream of two alternative 5"noncoding exons (exons 1A and 1B). A similar organization is conserved in the rat PMP22 gene, a finding that strongly supports the biological significance of this arrangement of potential regulatory sequences. Using the experimental advantages of the rat system, we have been able t o characterize the expression of these two alternative transcripts under a variety of conditions in which PMP22 had been shown to be regulated. We demonstrate that, while both transcripts are coexpressed in iissues and cell lines, the transcripts containing exon 1A are preferentially expressed in myelinating Schwann cells, while transcripts containing exon 1B are preferentially expressed in tissues that do not form peripheral myelin. In agreement with these findings, exon 1B expression predominates in cultured rat embryonic fibroblasts, although both exon 1A and 1B expression are up-regulated if the cells are arrested in the Go phase of the cell cycle. Surprisingly, a similar regulation is also observed after forskolin treatment of primary rat Schwann cells under conditions where myelin gene expression is induced.

EXPERIMENTAL PROCEDURES
Identification a n d Characterization of Rat Exons IA and IE--Two nonoverlapping clones containing rat exons 1A and 1B were isolated by screening a r a t A phage genomic library (Stratagene) with synthetic DNA oligomers. Positive plaques were purified, and the inserts were subcloned into plasmid vectors (pSP72). The presence of exons 1A and 1B has been verified by Southern blotting and characterized by automated dideoxy sequencing.
RNA Isolation--RNA was isolated by the single step guanidine thiocyanate method (Chomczynski and Sacchi, 1987). Adult human tissues were obtained at post-mortem examination within 6 h of the time of death and were frozen immediately on dry ice. Isolation of total RNA from rat tissues including rat sciatic nerves taken during postnatal development and following focal sciatic nerve injury has been described .
Northern Blot Analysis-Northern blot analysis of total cellular RNA was performed as described . Either a 1.5-kb rat PMP22 cDNA insert or a 0.6-kb human PMP22 cDNA insert, both containing the entire PMP22 coding region, was used as a probe .
Reverse Panscriptuse 5'-RACE-PCR"cDNA was prepared from 1 pg of total RNA isolated from human sciatic nerve using the oligonucleotide PMP-8 (5'-GGGAGGCGAAACCGTAG-3') to prime the synthesis . The 5'-cDNA sequence was amplified using the 5'-RACE kit (Life Technologies, Inc.) according to the manufacturer's conditions. The first round of PCR amplification was performed with the anchor primer included in the kit as well as primer PMP-1.3 (5'-GTGGTGGACATTTCCTGAGGAAG-3'). The second round of amplification used the anchor primer and primer PMP-1.2 (5"CTGAC-GATCGTGGAGAC-3'). DNA fragments of 200-300 bp were isolated, cloned into the pCRlOOO vector (Invitrogen), and sequenced. been used: 5'-CGGCCAAACAGCGTAACCCC"TCCAAG-3' (exon Primer Extension Analysis-The following oligonucleotides have lA, positions +171 to +142 in Fig. 4A) and 5'-GCCCAAAGGAACAGCT-GTCCCGATCCTCAG-3' (exon 1B; positions +143 to +114 in Fig. 4E). Oligonucleotides (20 PM) were labeled using the 5"terminal DNA labeling system (Life Technologies, Inc.) according to the manufacturer's instructions. Free nucleotides were removed by two consecutive rounds of ethanol precipitations in the presence of ammonium acetate, and the labeled oligonucleotides were collected in 20 p1 of nuclease-free H,O. 2.5 PM labeled primer was coprecipitated with 20 pg of total femoral nerve RNA or 20 pg of yeast tRNA (as control) using ethanol and sodium acetate. The dried pellet was directly dissolved in hybridization buffer containing 800 pl of formamide, 66 p1 of H,O, and 132 pl of 20 x PIPES (3 M NaCl, 0.1 M PIPES, 0.1 M EDTA, pH 6.8). Hybridization was carried out at 30 "C overnight. After ethanol precipitation, primer extension was performed in a 20-1.11 reaction volume using the Superscript RNase H-reverse transcriptase system (Life Technologies, Inc.) according to the manufacturer's protocol. 16 pl of 0.5 N NaOH was added, and the solution was incubated for 2 min a t 95 "C. The pH of the solution was neutralized by adding 8 pl of 1 M HCl, and the reaction mixture was ethanol-precipitated in the presence of ammonium acetate. The pellet was dissolved in 4 p1 of H,O and 6 pl of DNA loading buffer (Sequenase kit, Stratagene). Samples were heated to 95 "C for 10 min just prior to analysis on a standard 8% sequencing gel. The primer extension products were sized by parallel sequencing reactions using the primer extension primers on plasmids containing appropriate DNA fragments.
Riboprobes-Templates for the preparation of riboprobes were constructed by reverse transcriptase PCR using total RNAisolated from rat or human peripheral nerve. The PCR products were subcloned using the TA cloning system (Invitrogen; pCR2000 vector) according to the manufacturer's instructions. The following PCR primers were synthesized on Applied Biosystems model 391 PCR Mate DNAsynthesizer: lA, human exon 1Atranscript (transcription initiation; Fig. 3E
DNA sequences of all plasmids used were verified by dideoxynucleotide sequencing.
RNase Protection Analysis-RNase protection analysis was performed essentially as described previously (Pareek et al., 19931, except that the protected fragments were resolved on an 8% acrylamide, 19% formamide, 1 x Tris borateiEDTA gel. Briefly, template plasmids were linearized with appropriate restriction endonucleases (BamHI or XbaI) at the 3'-end of the insert and purified on a 1% Tris borateEDTAagarose gel using Geneclean I1 (BIO 101, Inc.). I n vitro transcription was performed in the presence of [a-32PldCTP using T7 or SP6 RNA polymerase (Promega). After treatment with RQ1 DNase (Promega), the samples were extracted with phenolkhloroform and precipitated with ethanol in the presence of sodium acetate after the addition of 20 pg of camer yeast tRNA. Riboprobes were hybridized with 2-5 pg of total RNA overnight at 55 "C. RNA:DNA duplexes were sequentially treated with RNase A and proteinase K, followed by phenolkhloroform extraction. Protected fragments were precipitated with isopropyl alcohol in the presence of 20 pg of yeast tRNA and resuspended in RNA sample buffer. Finally, the samples were boiled for 2 min just prior to the analysis on 8% acrylamide gels. The protected fragments were sized by comparison to 4x174 DNA standards end-labeled with [y-32PlATP using T4 polynucleotide kinase and RNAmarkers generated by in vitro transcription using a RNA Century marker template set (Ambion Inc.) according to the manufacturer's instructions.
Construction of Chloramphenicol Acetyltransferase Expression Vectors-In order to generate chloramphenicol acetyltransferase expression constructs, two regions of cosmid c132-G8  were first subcloned into pBSKS-: a 4.0-kb PuuII fragment (~1 3 2 -G8P4a) and a 4.5-kb EcoRI fragment (p132-GSR5a). These fragments were subsequently subcloned into the chloramphenicol acetyltransferase expression vector pBLCAT3 (Luckow and Schutz, 1987) as described below. To test promoter activity of sequences upstream of exon lA, a three-way ligation was performed. A 3.2-kb EcoRV/AuaI fragment from p132-G8P4a was allowed to ligate to a 140-bp PCR fragment (amplified from positions -268 to +170 in Fig. 4A and then digested with AvaI) that terminates within exon 1A and upstream of any ATG sequences that might interfere with the translation of the reporter gene. The resulting 3.34-kb fragment was cloned into the BamHI site of pBLCAT3, after filling i n free BamHI ends with Klenow polymerase, to obtain pPMPlCATF and pPMPlCATR with inserts in the forward and reverse orientations, respectively.
To test promoter activity of sequences upstream of exon lB, we engineered the plasmid p132-G8R5a such that its insert terminated within exon 1B upstream of any ATG sequences. This was done by ligating a 2.9-kb KpnIISacII fragment from p132-G8R5a to a 320-bp PCR-derived fragment (amplified from positions -224 and +143 in Fig. 4B followed by digestion with SacII). The entire 3.2-kb fragment was cloned into pBSKS-at the EcoRV site. The resulting clone, p12-1, was linearized with HindIII, filled in using Klenow polymerase, and digested with EcoRV to allow for unidirectional deletion using exonuclease 111. The insert from one of the resulting deleted derivatives, which included 3 kb of sequence upstream of the 5'-boundary of exon lB, released by digestion with EcoRI, was ligated into the BamHI site of pBLCAT3, after appropriate repair of the ends, in forward and reverse orientation to yield pPMP2CATF and pPMPBCATR, respectively.
Tissue Culture a n d Dansfection-The hamster fibroblast cell line FLJK88 (Fuscoe et al., 1983) was used for DNA transfection as a representative fibroblast cell line. Cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% (v/v) fetal calf serum, along with 100 pg of streptomycin and 100 units of penicillidml, under an atmosphere of 5% CO,. Approximately 24 h before transfection, RJK88 cells were plated at a density of 5 x lo5 cells/lOO-mm culture plate. 10 pg of DNA prepared using QIAGEN columns were transfected into cells in 100-mm plates by the calcium phosphate precipitation technique (Graham and van der Eb, 1973).
Primary rat Schwann cells were isolated from sciatic nerves of 2-3day-old Sprague-Dawley rats and expanded as described previously (Pareek et al., 1993). After three passages, the cells were frozen in aliquots. For transfections, the frozen cells were regrown and passaged one additional time before use. As judged by SlOO protein immunohistochemistry (rabbit anti-cow, Dako Corp.), more than 95% of the cells exhibited a Schwann cell phenotype after this procedure. 60-80% confluent 100-mm culture dishes were fed 6 h before transfection. Since primary cells are often variable in their transfection efficiencies, 10 pg of the PMP22 promoter constructs was cotransfected by the calcium phosphate precipitation method with 500 ng of a CMV-LacZ control construct for normalization purposes .
Chloramphenicol Acetyltransferase Assay-For the analysis in fibroblasts, cells were harvested 48 h after transfection and lysed by three cycles of freezing and thawing. Cellular debris was removed by centrifugation, and equal amounts of protein, as determined by the Bradford assay (Bio-Rad), were used for standard chloramphenicol acetyltransferase assays (Gorman et al., 1982). The percentage of ['4C]chloramphenicol converted to acetylated product (percent conversion) was determined by quantitating autoradiographic analysis of the TLC plate in a Betascope analyzer, by quantitating the radioactivity on TLC plates with a liquid scintillation counter, or by densitometric analysis ofthe x-ray films. All assays were performed within the linear range of chloramphenicol acetyltransferase activities with respect to incubation time and sample concentration.
Schwann cells were harvested 60 h after transfection, and cellular extracts were prepared by freezing and thawing as described above followed by 10 min of ultrasonification in an ultrasonification water bath. Samples were standardized by determining the protein concentration and cotransfected p-galactosidase activity. Equal amounts of p-galactosidase activity were analyzed in chloramphenicol acetyltransferase assays  as described above.
Growth-arrested Fibroblasts-Primary rat embryonic fibroblasts (Woods and Couchman, 1992) (used at passages 11-14; kindly provided by Dr. Gerald Fuller, University of Alabama a t Birmingham) were growth-arrested essentially as described for NIH 3T3 cells (Manfioletti et al., 1990). Briefly, fully growing fibroblasts were cultured in a-minimal essential medium (Life Technologies, Inc.) supplemented with 10% bovine calf serum (Hyclone Laboratories) to approximately 70% confluency (full growth) before harvesting for the isolation of total cellular RNA. Parallel cultures were allowed to become confluent (density growth arrest) and were fed every 2 days for a total of 8 days. RNAfrom these cells were harvested every 48 h. A second set of cultures were plated a t a density of 2 x lo5 celld100-mm2 plate and grown overnight and harvested (full growth) or switched to medium without serum (serum-deprived) before RNAwas isolated at 0,3,12,24,48, and 72 h after addition of serum-free medium. RNA was prepared as described above and quantitated by optical density (260 nm) and ethidium bromide staining on formaldehyde, 1.2% agarose gels. Growth kinetics were determined by bromodeoxyuridine (BrdUrd) incorporation (see below). Samples were analyzed by Northern blot analysis for the induction of PMP22 mRNA and by RNase protection using the rat R2 riboprobe to determine the relative amount of the two major PMP22 transcripts.
Bromodeoxyuridine Incorporation in Fibroblasts-At the time of harvest of fibroblasts for RNA, parallel cultures were treated with BrdUrd (50 final concentration) for 2 h a t 37 "C. The cells were then rinsed in phosphate-buffered saline and fixed in cold (-20 "C) 50 mM glycine, pH 2.0, in 70% ethanol overnight. The futed cultures were rinsed with ethanol and air-dried. The cultures were then blocked with 10% normal goat serum in phosphate-buffered saline, reacted for 1 h with undiluted monoclonal anti-BrdUrd (Amersham Corp.), and detected with biotin-%o Promoters for Alternative PMP22 Transcripts cDNA-A : . . GCTGTTTGGCCGG/GCAGAAACTCCGCTGAGCAGAACTTGCCGCCAGAAi (Hayasaka et al., 1992) c D N A -B : . . GTTCCTTTGGGCT/GCAGAAACTCCGCTGAGCAGAACTTGCCGCCAGmX (Edomi et al., 1993) I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I   I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I CDNA-C : . . GCTCTCCTCGCAG/GCAGAAACTCCGCTGAGCAGAACTTGCCGCCAGmX  FIG. 1. Comparison of the 5'-ends of human PMP22 cDNAs. Identical nucleotides are marked by vertical bars. The putative position of an intron is indicated with a slash. The translation start is underlined. References are given on the right. conjugated goat anti-rabbit IgG (1:400; Jackson ImmunoResearch Laboratories, Inc.). Streptavidin-conjugated horseradish peroxidase (1:400; Jackson ImmunoResearch Laboratories, Inc.) was added, and diaminobenzidine was used as substrate for visualization. Unlabeled nuclei were counterstained with 0.05% methylene blue.

RESULTS
Identification of Human PMP22 cDNAs with Different 5'-Sequences-We have identified several PMP22 cDNA clones by a combination of reverse transcriptase PCR and 5"RACE-PCR using RNA derived from human sciatic nerve as a template (data not shown). The alignment of the resulting DNA sequences with published human PMP22 cDNAs Hayasaka et al., 1992;Edomi et al., 1993) revealed the presence of three different 5"flanking sequences that diverge from the common sequence at the same nucleotide position (Fig. 1). The sequence of cDNA-C displays the classical features of an aberrantly spliced mRNA that has retained the 3'-border of an intron (conserved splice acceptor site adjacent to a polypyrimidine stretch). This suspicion was confirmed by DNA sequence analysis of appropriate genomic clones (Table I), and RACE-PCR and primer extension analyses gave no indication that cDNA-C represents a frequent PMP22 mRNA species (data not shown). Consequently, cDNA-C was not analyzed further. Based on the different sequences of cDNA-A and cDNA-B, we hypothesized that these cDNAs represent alternatively spliced 5'-untranslated exons in the PMP22 gene that have been joined to a shared exon 2 to generate the observed mRNA (cDNA) diversity.
Structure of the Human PMP22 Gene-In order to study the genomic organization of the human PMP22 gene and to confirm the presence of alternatively spliced exons in its 5"flanking region, we have isolated genomic cosmid clones spanning the PMP22 locus from a cosmid library constructed from flowsorted human chromosome 17 . We previously reported two overlapping cosmid clones, c132-G8 and c103-Bl1, which encompass the PMP22 gene. In the present study, cosmid c49-E4 (which is similar to c132-G8) was used for detailed characterization. The overlapping cosmids c103-Bll and c49-E4 were found to contain the entire PMP22 gene, which spans about 40 kilobases (Fig. 2). Restriction enzyme and Southern blot analyses revealed that the human PMP22 gene contains four coding exons (exons 2-5). Exon 2 (putative first transmembrane region) and exon 3 (extracellular loop including the N-linked glycosylation site) appear to encode specific domains of the PMP22 protein. In contrast, exon 4 covers the second putative transmembrane domain as well as half of the third, and exon 5 contains the information that determines the second half of the third transmembrane region, the putative fourth transmembrane region, and the 3"untranslated region. Furthermore, as predicted by the variable cDNAs, two alternative exons 1A and 1B are located in the 5"untranslated region of the gene (Fig. 2), corresponding to cDNA-A and cDNA-B (Fig. 1). All exodintron boundaries are in good agreement with known consensus sequences (Table I) (Mount, 1982).
Mapping of the 5I-End.s of Human PMP22 mRNAs-Extensive reverse transcriptase 5"RACE-PCR analysis suggested that the presence of exons 1A and 1B in human PMP22 RNAs is mutually exclusive, and no evidence for an additional exon in the human PMP22 gene upstream of exon 1A was observed (data not shown). Thus, we hypothesized that PMP22 expression is regulated by two different promoters that are located upstream of each of the alternative exons 1A and 1B (Fig. 2).
Based on the genomic human PMP22 DNA sequence and appropriate cDNA sequences, oligonucleotide primers specific for either exon 1A or 1B were synthesized and used in separate primer extension experiments to determine the transcription initiation sites of the two observed mRNA species. In both experiments, several extended cDNA products were obtained (Fig. 3, A and C ) . In order t o eliminate signals due to premature terminations of the reverse transcriptase reactions and potential nonspecific priming of the oligonucleotides used, the results of the primer extension experiments were confirmed by appropriate RNase protection analysis (Fig. 3B). Specific 32P-labeled cRNA probes were synthesized in vitro from template plasmids that contained PCR-amplified DNA fragments spanning the different putative mRNA ends as tentatively determined by results of the primer extension analysis and DNA sequencing of the putative promoter regions (see "Experimental Procedures" for description of probes). The combined results of both types of analyses suggest the usage of one major transcription initiation site for each mRNA species (Fig. 4).
According t o the mapping results of exon lA, the longest isolated human PMP22 cDNA-A clone is missing 19 nucleotides at its 5'-end ( Fig. 5 A ) (Hayasaka et al., 1992). Interestingly, a sequence comparison of the longest corresponding rat PMP22 cDNA-A (Spreyer et al., 1991) with the appropriate genomic human PMP22 sequence reveals a nucleotide identity of 92% in the 50 nucleotides immediately following the transcription initiation site and 63% within the entire exon lA (Fig. 5 A ) . The 19 nucleotides putatively missing in the human cDNA-A clone can be found with only one nucleotide mismatch in the corresponding rat clone whose 5'-end coincides with the experimentally determined human mRNA-A start site. Since the arrangement of two alternative exons in the 5'-untranslated region is conserved between PMP22 genes in human and rat (see below), these results further support the correct mapping of the cap site of the human mRNA-A and make the presence of additional untranslated exons upstream of exon 1A unlikely.
The longest human PMP22 cDNA-B clone is 42 nucleotides short of being full-length (Fig. 5B). A comparison of the latter sequence with the longest mouse PMP22 clone reveals that the 5"sequence of exon 1B is highly conserved between human and mouse (Fig. 5B). The mouse clone stops 5 nucleotides short of the corresponding human transcription initiation site, and its first 37 nucleotides are 95% identical to the human genomic sequence (nucleotide identity over the entire exon 1B is 70%). Although these findings indirectly support our mapping results of human exon lB, the significance of this observation remains elusive due to the current lack of information on the genomic organization of the mouse pmp22 gene.
Sequence of the Putative Human PMP22 Promoters-We have previously shown that a cloned 4.5-kb EcoRI fragment derived from the cosmid c132-G8 (p132-G8R5) contained a portion of the CpG island associated with the PMP22 gene . Restriction mapping and Southern blot analysis Exon sequences are shown in upper-case letters, and intron sequences in lower-case letters. The numbers directly below the sequence refer to 5'and 3'-nucleotide positions in the PMP22 cDNA flanking each intron (Pate1 et al., 1992). Below these numbers, amino acids at the exodintron junctions are indicated.  with probes specific for exons 1A and 1B confirmed that the regulatory regions of the PMP22 gene were likely located within this fragment (data not shown). The results of the combined efforts of DNA sequencing, primer extension, and RNase protection assays used to map exons 1A and lB, as well as their respective 5"flanking regions, are shown in Fig. 4. The putative promoter P1 contains a TATA box-like sequence 30 bp upstream of the corresponding mRNA cap site, a typical feature for most eukaryotic genes. In addition, an inverted CCAAT box, an element that is often found in RNA polymerase I1 promoters (McKnight and Kingsbury, 19821, is also present at positions -43 to -47. In contrast, neither TATA-like nor CCAAT sequences can be found in the putative promoter P2. Agents that increase the level of cellular CAMP are known inducers of PMP22 expression in cultured Schwann cells (Spreyer et al., 1991;Pareek et al., 1993). Thus, we analyzed both promoters for putative consensus sequences for the binding of specific transcription factors that have been implicated in gene regulation by CAMP. Two potential binding sites for the transcription factor NF1 at positions -217 to -221 and -241 to -245 in promoter P1 (Zhang and Miskimins, 1993) and overlapping consensus sequences for the binding of the transcription factors AP-2 and Sp-1 at positions -67 to -79 in promoter P2 (Imagawa et al., 1987) might be significant in this context (Fig. 4).
Homologies to Myelin Gene Promoters-A computer search aimed at the identification of shared sequence elements be- tween the two putative PMP22 promoters and the 5"flanking regions of other myelin genes was performed. Since no footprinting data of the PMP22 promoter sequences are available at this time, we concentrated our efforts on a comparative analysis of sequence elements that have been previously identified by functional assays in the proteolipid protein and MBP gene promoters. Using this approach, several potentially important sequence motifs in the two putative PMP22 promoters were identified. First, a striking similarity was observed between the mouse MBP enhancer (Fors et al., 1993) and a sequence motif in the putative PMP22 promoter P1 just downstream of the TATA-like sequence (Fig. 6A) Fig. 4. References to other sequences are given by Nave and Lemke (1991) and Fors et al. (1993). PLP, proteolipid protein.
putative general transcription factor, M1, in the mouse MBP promoter; a similar sequence motif is also found in the putative P2 promoter (Fig.  6B). Finally, the P1 promoter contains a sequence element that is similar to a footprinted sequence (FP330; Fig. 6 C ) in the proteolipid protein promoter, which is also present in the PO promoter of various species (rat, mouse, human), as well as in the mouse MBP promoter (MOTIF B) (Nave and Lemke, 1991). This finding is of particular interest since the transacting regulator(s) binding at FP330 are poten- TGTGTTTGAGGCCACCCTGAGGATCGGGACAGCTG~TCCTTTGGGCT ****CC******T*AT**--*CTCT*A**T******C******AA** ****CC******T**T*C--*CTCT*A*C*G*****C*G*****C******AA** tial candidates for common transcriptional regulators of myelin genes (Nave and Lemke, 1991). Functional Analysis of the Putative Human PMP22 Promoters 1 and 2-In order to analyze putative promoter activity within sequences upstream of exons 1A and lB, reporter plasmids were constructed in the vector pBLCAT3, which contains the gene encoding bacterial chloramphenicol acetyltransferase. The resulting plasmids, pPMPlCATF and pPMPlCATR, contain approximately 3.3 kb of sequence upstream of exon 1A in the forward and reverse orientations, respectively, and the plasmids pPMP2CATF and pPMP2CATR contain approximately 3.2 kb of sequence upstream of exon 1B in the forward and reverse orientations, respectively. These plasmids along with a positive and negative control were transfected into the hamster fibroblast line RJK88 (Fig. 7A) and primary rat Schwann cells (Fig. 7B with virtually identical assay results. Overall, the sequences upstream of exon 1A as well as exon 1B display significant promoter activity, with the sequence upstream of exon 1B being a more potent promoter in both cell types than the sequence upstream of exon 1A. The construct containing the putative promoter 1 sequence in the forward orientation (pPMP1CATF) consistently yielded at least 10-fold higher chloramphenicol acetyltransferase expression compared with the construct containing the promoter 1 sequence in the reverse orientation (pPMP1CATR). The promoterless vector pBLCAT3 showed a variable background, often approaching that of the construct pPMP1CATF. Nevertheless, the apparent differences in activities between the constructs containing the promoter 1 sequences in the forward versus reverse orientation, the position of promoter 1 upstream of the highly active promoter 2, the perfectly positioned (T)ATA box, and the mRNA mapping results strongly suggest that the sequences upstream of exon 1A function as a PMP22 gene promoter.
Conservation of the 5'-Regulatory Regions in the Rat and Human PMP22 Genes-In order to establish the rat as an experimental system in which to study the regulation of the two alternative PMP22 transcripts in detail, we cloned and %o Promoters for Alternative PMP22 Banscripts analyzed the putative regulatory region of the rat PMPZZ gene.
The results demonstrate that, as anticipated from rat PMP22 cDNA comparisons (Fig.   11, the genomic structure of the PMPZZ gene with regard to exons 1A and 1B is, indeed, evolutionarily conserved in rat and human (see schematic in Fig. 2 for the region analyzed). Table I provides the sequence of the exodintron boundaries between exons 1 N l B and 2 of the rat PMP22 gene and details the sites of RNA splicing, which gives rise to the two alternative rat PMP22 mRNA transcripts. These combined findings support an important role for the observed use of alternative exons in the regulation of the PMP22 gene and prompted an examination of the quantitative expression of alternative PMP22 transcripts in various experimental paradigms.
The Expression of Alternative PMPZZ Danscripts Displays Marked Tissue Specificity-To determine the presence and relative abundance of both transcripts simultaneously, we designed human-and rat-specific riboprobes for RNase protection analysis (Fig. €24). Two riboprobes were designed for each species. Riboprobe R1 was designed to be fully complementary for -300 bp to all transcripts containing exon 1A contiguous to exon 2 and, in addition, to be complementary over -250 bp to all other PMP22 transcripts that carry exon 2 but not exon 1A (Fig. 8B). A reciprocal riboprobe, designated R2, was designed Shown is a diagram of idealized riboprobes used in RNase protection assays to detect and quantitate the two alternatively spliced forms of PMP22 mRNA, the actual riboprobes used vary slightly with regard to the size of the fragments that they protect from RNase digestion. A shows that the differences between riboprobes R1 and R2 reside in the specificity of the tively. B shows how each riboprobe individually (in this case, probe R1) approximately 50-bp 5"extension specific for exons la and lb, respeccan detect and quantitate both exon la-and lb-containing transcripts that can be identified on a polyacrylamide gel as shown in C .
to be fully complementary for -300 bp to exon 1B-containing transcripts contiguous to exon 2 and complementary for -250 bp to all other PMP22 transcripts that carry exon 2 (schematized in Fig. 8C). RNase protection analysis for the expression of human and rat alternative PMP22 transcripts in various tissues is shown in Fig. 9. This experiment was designed to detect the ratios of the two PMP22 transcripts in tissues, no effort was made to control for tissue-specific autolysis rates in the human post-mortem samples. All tissues expressed detectable amounts of both transcripts ( Fig. 9 and data not shown). The high levels of PMP22 mRNA in rodent heart, lung, and gut have been reported previously (Manfioletti et al., 1990;Welcher et al., 1991). Differences between human and rat lung and gut PMP22 mRNA levels may reflect species differences, age differences, sampling differences, or autolysis artifact. Significantly, however, the point is that transcripts containing exon 1B are preferentially expressed in all human and rat tissues examined, except for peripheral nerve. In the rat sciatic nerve, the levels of both PMP22 transcripts are much higher than in any other tissue examined, and in contrast to non-neural tissues, the exon 1A-containing transcript is predominant, comprising approximately 80% of total PMP22 mRNA (see quantification in Figs. 1OB and 11B).
Based on the observed reciprocal pattern of expression using riboprobes R1 and R2, we inferred that alternative transcripts containing exon 1A or 1B account for most, if not all, of the PMP22 mRNA species in human and rat. To test this hypothesis more directly, we performed a quantitative analysis of five independent mRNA samples from the sciatic nerves of 60-dayold rats by RNase protection using both rat riboprobes R1 and R2 separately. Taking into account the differences in radiolabel incorporated as a function of probe length, we calculated the absolute value of the maximum contribution of exon 2-contain- ing transcripts that were not detected as -300-bp fragments by either of the riboprobes to be 0.5 2 2%, indicating that the vast majority of PMP22 transcripts detected contain either exon 1A or 1B. These findings validate the capability of each probe alone to identify the relative proportion of each transcript. Furthermore, the identical tissue specificity of the expression of alternative PMP22 transcripts in rat and human as shown in Fig. 9 reiterates the similarity of the two species in the regulation of PMP22 gene expression as suggested by the similar structures of the rat and human PMP22 regulatory regions. Regulation of the Exon 1A-containing PMP22 Dunscript Is llghtly Associated with Myelin Formation during Sciatic Nerve Development-To further investigate the regulation of alternative PMP22 transcripts, we performed an RNase protection analysis on peripheral nerve samples taken from the period of myelination that occurs during early postnatal development in the rat. These studies are complementary to Northern and Western blot studies performed previously in our laboratory . We have repeated those Northern blots in this study since it is the method of choice for the quantitation of changes in total PMP22 mRNA, while the RNase protection analysis allows accurate determination of the relative amounts of the two alternative transcripts. By combining the results of both types of analyses, the total expression of each transcript at each stage of development could be determined. Fig. 1OA (middle panel) demonstrates the dramatic increase in the expression of total PMP22 mRNA during postnatal PNS development of the rat, and Fig. 1OA (lower panel) shows that both the exon 1A-containing transcript and the exon 1B-containing transcript are regulated, at first approximation, in a time course that parallels myelination. For quantitation, the relative expression of the two transcripts was determined by two-dimensional scintillation scanning using an AMBIS gasphase detection scintillation scanner and is shown graphically along with the autoradiographs from the RNase protections in Fig. 1OB. The results demonstrate that the exon 1A-containing transcript converts from being the minor mRNA species, accounting for approximately 30% of the total PMP22 mRNA at birth, to the major transcript, accounting for approximately 80% of the PMP22 mRNA in adult rat sciatic nerves. Furthermore, from the combined quantitative analysis of the Northern blot and RNase protection data, we can estimate that the exon 1A-containing transcript is induced approximately 25-fold, while the exon 1B-containing transcript is induced only approximately 7-fold during the postnatal myelination period. Detailed analysis reveals also that the expression of the exon 1A-containing transcript seems to be more closely correlated with the process of myelination since the highest levels of the exon 1A-containing transcript are detected several weeks postnatally, while the expression of the exon 1B-containing transcript peaks at postnatal day 7.
Regulation of the Exon 1A-containing PMP22 Dunscript Is Associated with Myelin Formation during Sciatic Nerve Regeneration-In order to confirm the association between the regulation of expression of exon 1A-containing PMP22 transcripts and myelin formation, we exploited the tight regulation of myelin gene expression that occurs in the portion of a nerve distal to the site of focal nerve crush injury as the denervated nerve stump undergoes Wallerian degeneration followed by axon and myelin regeneration. RNase protection analysis of mRNA from the distal sciatic nerve of crush-lesioned rats shows a dramatic repression in the levels of the exon 1A-containing transcript to almost undetectable levels 3 days postcrush, followed by its re-expression to become the major PMP22 transcript by 21 days after injury (Fig. 1lA). This observed time course is compatible with a model in which exon 1A expression is tightly correlated with the formation of myelin and inversely correlated with the breakdown of myelin, as expected for a myelin-specific transcript. The exon 1B-containing transcript also shows a small down-regulation in the first 3-7 days following nerve injury, but not nearly as marked as that observed for the exon 1A-containing transcript. When the percentage of each transcript at each time point is plotted versus days following nerve crush injury, it is apparent that the regulation of the exon 1A-containing transcript more tightly correlates with the profound initial inhibition and later reinduction of myelin synthesis than does the regulation of the exon 1Bcontaining transcript (Fig. llB, inset 1.  In summary, we conclude from the studies of PMP22 expression during development and following nerve injury that expression of the exon 1A-containing PMP22 transcript is tightly correlated with the formation of myelin. The exon 1B-containing transcript is also correlated with myelin formation, albeit to a much lesser extent.

Exon 1B-containing Dunscripts Predominate in Forskolininduced Primary Schwann
Cells-PMP22 protein is well established as a component of myelin. In the previous set of experiments, we have shown that the transcripts containing exon 1A are predominantly associated with myelin formation. In primary Schwann cells, myelin gene expression can be induced by the adenylate cyclase activator forskolin (Lemke and Chao, 1988). It has been proposed that forskolin induces myelin gene expression in the absence of axons by acting as a partial substitute for the axon-Schwann cell interactions that are responsible for the initiation of myelin formation (Lemke and Chao, 1988; but see also Morgan et al., 1991). Indeed, we and others have previously shown that PMP22 and PO mRNAs are induced by forskolin in cultured Schwann cells (Spreyer et al., 1991;Pareek et al., 1993). We have reproduced our original Schwann cell growth conditions and have examined the specificity of forskolin action on the expression of the alternative PMP22 transcripts in order to compare the results with the observed PMP22 gene regulation in vivo during the formation of myelin. Surprisingly, in triplicate experiments, both exon 1A and 1B expression appear to be induced by forskolin treatment to the same extent; the exon 1B-containing transcript clearly remains the predominant PMP22 mRNA species (data not shown). Since it has been reported that culturing Schwann cells in defined medium without serum results in both a reduced fraction of dividing cells and an enhanced induction of PO mRNA following the addition of forskolin (Morgan et al., 1991), we have repeated and extended our studies by culturing the Schwann cells in serum-containing or defined medium with and without forskolin. Fig. 12 (lower panel) shows that forskolin addition up-regulated both the exon 1A-and exon 1B-containing transcripts, with the exon 1B transcript predominating. Quantitatively, the addition of forskolin to the Schwann lluo Promoters for Alternative PMP22 Danscripts PN -* -+ Ethidium bromide Agarose gel cells grown in defined medium showed the expected enhancement in PO and PMP22 mRNA levels by Northern blot analysis when compared with the forskolin-stimulated Schwann cells that had been grown in serum-containing medium (Fig. 12). Since myelination in vivo is associated with an up-regulation of the exon 1A-containing transcript out of proportion to the upregulation of the exon 1B-containing transcript, we conclude that the induction of PMP22 mRNA by forskolin treatment of primary Schwann cells does not closely resemble the gene regulation observed in in vivo myelination. Additional studies will be necessary to determine if there are any in vitro conditions that mimic the in vivo situation where the exon 1A-containing transcript is up-regulated out of proportion to the exon 1Bcontaining transcript.
Exon 1B-containing Danscripts Predominate in Growtharrested Rat Fibroblasts-Steady-state PMP22Igas-3 mRNA levels have been correlated with cellular growth arrest induced by serum deprivation or contact inhibition of NIH 3T3 fibroblasts (Manfioletti et al., 1990). We have isolated RNA from rat embryonic fibroblasts grown under favorable growth conditions (full growth) or following growth inhibition when a greater percentage of cells are arrested in the Go phase because of serum deprivation or contact inhibition. The growth kinetics of the cultures used in this set of experiments were monitored by BrdUrd incorporation (Fig. 13B). Northern blot analysis using the rat PMP22 cDNA probe shows the expected correlation of induction of PMP22 in growth-inhibited as compared with fully growing cells (Fig. 13, A (middle panel) and B). RNase protection analysis reveals that both transcripts are up-regulated in growth-arrested cells; however, the expression of the exon 1Bcontaining PMP22 transcript continues to dominate in growtharrested cells (Fig. 13A, lower panel). There is no evidence that the ratio of the two alternative PMP22 transcripts is altered

DISCUSSION
In this paper, we report the cloning and characterization of the human PMP22 gene. This analysis provides the basis for the analysis of mutations affecting the PMP22 gene in hereditary motor and sensory neuropathies, in particular Charcot-Marie-Tooth disease and Dejerine-Sottas disease. Since the underlying anomaly in the majority of hereditary motor and sensory neuropathy and HNPP patients appears to be altered levels of PMP22 due to a gene dosage effect (Lupski et al., 1992;Chance et al., 1993), the identification and preliminary characterization of regulatory regions of the PMP22 gene may also constitute the first step toward potential novel treatment strategies for these particular neuromuscular disorders .
We hypothesized from the isolation of several human and rat PMP22 cDNA species with variable 5'-ends that the PMP22 gene might be regulated by alternative promoters. The characterization of the human PMP22 gene and partial analysis of the rat PMP22 gene provided evidence to support this hypothesis with the identification of two upstream noncoding exons, 1A and lB, which display remarkable interspecies sequence similarity. Primer extension and RNase protection studies allowed mapping of the transcription initiation sites of each of the two hypothesized human PMP22 mRNA species and suggested the presence of the two alternative promoter regions. Functional promoter analysis by transfection into a fibroblast cell line and primary mitogen-expanded Schwann cells demonstrated that DNA sequences upstream of exon 1B (promoter 2) directed high levels of reporter gene expression in both cell culture systems, while only weak activity could be detected for sequences up-  Several reasons can be envisaged for the weak activity of promoter 1 in transfection assays. A trivial explanation would be that regulatory regions important for the expression of promoter 1 are missing in the transfection constructs. Alternatively, the presence of a (T)ATA-like sequence located at position -30 indicates a well defined and probably tightly regulated, tissue-specific promoter that might not be efficiently activated in cultured fibroblasts or Schwann cells. To address this question, we examined the activity of the two PMP22 gene promoters in different cell types by characterizing the endogenous expression patterns of the two PMP22 transcripts in vitro and in vivo. I t should be emphasized that our approach was focused on the measurement of steady-state mRNA levels and does not exclude the contribution of differential mRNA stability, potentially conferred by differences in the evolutionarily conserved 5'4eader sequences of the two alternative PMP22 transcripts. Our results show that the alternative PMP22 transcripts are expressed in a tissue and cell type-specific manner and that they are regulated during the development and regeneration of the sciatic nerve. Schwann cells that elaborate PNS myelin express high levels of the exon lbcontaining PMP22 transcript, while exon 1B-containing mRNA is predominant in other cell types. During postnatal PNS development, the expression of the exon 1A-carrying transcript is induced approximately 25-fold and correlates tightly with the formation of myelin, while the exon 1B-carrying transcript is induced to a much lesser extent during the same time period.
Numerous studies have established that myelin gene expression by Schwann cells is critically dependent on the presence of axons (Politis et al., 1982;Lemke and Chao, 1988). In pure Schwann cell cultures, forskolin has been shown to up-regulate the expression of the PO gene that encodes for the major protein component of peripheral nerve myelin (Lemke and Chao, 1988). These and similar findings have led to the hypothesis that forskolin can replace the requirement of axonal contact for myelin gene expression in Schwann cells and indirectly implicated the CAMP pathway in the signal transduction cascade leading to myelin gene expression (Sobue et al., 1986;Lemke and Chao, 1988). Our results indicate, however, that the regulation of the PMP22 gene during myelin formation in vivo occurs by preferential induction of the exon 1A-containing transcripts compared with the exon 1B-containing transcripts. In contrast, forskolin stimulation of primary Schwann cells el-evates the levels of both transcripts essentially equally. Although we have found potential CAMP-responsive elements in the promoter regions of PMP22, consistent with a model of transcriptional regulation of PMP22 by CAMP, it is also possible that forskolin up-regulates PMP22 mRNA in Schwann cells by selectively stabilizing PMP22 mRNA in a fashion analogous to the CAMP-dependent regulation of renin mRNA stability (Chen et al., 1993). In either case, our finding suggests that forskolin can only partially mimic the effects of axonal contact on PMP22 induction, although extensive comparative studies on the regulation of the expression of alternative PMP22 exons and other myelin proteins (e.g. PO) under various culture conditions will be required to gain further insights into the significance of myelin gene regulation by forskolin in

vitro.
The exon 1B-containing transcript is expressed at higher levels in cultured Schwann cells than the exon 1A-containing transcript, consistent with the much stronger activity of promoter 2 compared with promoter 1 in transfected forskolintreated Schwann cells. It may be possible to increase the activity of PMP22 promoter 1 in vitro by optimizing Schwann cell culture conditions for the expression of the major PNS myelin protein PO (Morgan et al., 1991) or by developing myelinating co-cultures of Schwann cells and appropriate neurons (Owens et al., 1990). However, due to the relatively poor efficiency of myelination in vitro and the complex mechanisms involved in myelin gene regulation, it appears likely that in vivo models using appropriate reporter constructs in transgenic mice will ultimately be required to define the regulatory elements of promoter 1 and to resolve the relative contributions of the two promoters during myelin formation.
Mouse PMP22 has been shown to be identical to the growtharrest specific gene, gas-3 (Suter et al., 1992131, whose mRNA is up-regulated in quiescent NIH 3T3 cells by contact inhibition and serum deprivation (Schneider et al., 1988). Our results indicate that this regulation of PMP22 mRNA in rat embryonic fibroblasts is not as striking as has been reported in NIH 3T3 cells. Comparison of the growth kinetics of the two cell types reveals two differences. First, more NIH 3T3 cells are labeled in the fully growing phase, and second, a greater degree of growth arrest is observed in the NIH 3T3 cells. We believe that the induction reported here is still significant because it correlates with the labeling index of the rat embryonic fibroblasts, it is qualitatively similar to that reported for the NIH 3T3 cells, and we have observed similar changes in a variety of cell types including C6 glioma cells, human and rat dermal fibroblasts, and rat 2 cells.' Thus, this type of PMP22 regulation appears to be a widespread phenomenon, although its specificity and significance remain uncertain. When we examined the induction of PMP22 mRNA in the rat embryonic fibroblasts by RNase protection, we were unable to detect a major shift in the ratio of the alternative 5'-exons that might indicate transcriptional regulation. This is consistent with the previous findings of Manfioletti et al. (19901, who examined the induction of gas-3 (PMP22) in fibroblasts by nuclear run-off assays and observed no change in the rate of transcription of PMP22 in growtharrested cells. It appears that up-regulation of the steady-state levels of PMP22/gas-3 mRNA in cultured NIH 3T3 cells by cellular growth arrest is mainly due to alterations in total PMP22 mRNA stability, which agrees with our finding that the ratio of the two transcripts is apparently not altered during cellular growth arrest. Thus, there seems to be post-transcriptional regulation of PMP22 mRNA in fibroblasts in vitro in addition to the apparent transcriptional regulation of PMP22 mRNA in myelinating Schwann cells in vivo.
U. Suter and G. J. Snipes, unpublished observations. The validity of the two-promoter model for the human PMPZZ gene is strongly supported by the complete conservation of this genomic arrangement between the human and rat PMP22 genes, but the exact biological importance of this specific gene organization remains to be determined. Gene regulation by multiple promoters is commonly associated with tissue-specific, or inducible, or ubiquitous gene expression. The regulation of the PMPZZ gene may reflect several of these mechanisms since PMP22 expression is inducible and developmentally regulated, but also quite widespread. Curiously, these different regulatory modes of gene regulation seem to parallel a previously hypothesized dual function of PMP22. On one hand, a major function for PMP22 in the formation and maintenance of the myelin sheath is inferred from studies of the phenotypes of the various mutated forms of PMP22. The Schwann cell-specific function in vivo is generally associated with the regulation of both exon 1A-and 1B-containing transcripts, with a much more pronounced regulation of the exon 1A-containing transcript during myelination. On the other hand, PMP22 mRNA can also be found in many tissues outside of the peripheral nerve. We originally hypothesized that PMP22 transcripts detected in these non-neural tissues might be due to contamination by peripheral nerve innervation. The findings presented in this report do not support this view since the exon 1B-containing transcript is predominantly expressed in non-neural PMP22-expressing tissues. Thus, our results have uncoupled the regulation of PMP22 in myelin-forming Schwann cells and in other cell types. We hypothesize that the differential regulation of the PMP22 gene by different promoters might directly correlate with potential multiple functions of the PMP22 protein, one in myelinating Schwann cells and the other in non-myelin-forming cells, as suggested by indirect evidence for other myelin proteins (Schneider et al., 1992).
The extremely high sequence conservation observed between the respective exons 1A and 1B of the human and rat genes suggests that the 5'-untranslated regions might subserve an important function besides their role in facilitating alternative promoter regulation. It has been demonstrated that the use of alternate 5'-untranslated regions can dictate the translational discrimination of mRNAs coding for human insulin-like growth factor I1 (Nielsen et al., 1990) and can also lead to differential regulation at the translational level of the complement protein C2 (Horiuchi et al., 1990). Although differential translatability remains a formal possibility, so far we have observed that the total level of PMP22 mRNA, irrespective of which transcript predominates, correlates with protein expression in all circumstances examined, which include during myelin formation during development and during Wallerian degeneration (Snipes et al., 19921, in forskolin-stimulated Schwann cells (Pareek et al., 1993), and in growth-arrested fibroblast^.^ So far, it has not been possible t o detect or quantitate PMP22 protein in nonneural tissues. Alternatively, these conserved sequences could be involved in the regulation of PMP22 mRNA stability given the finding that PMP22/gas-3 mRNA expression in growtharrested NIH 3T3 fibroblasts is strongly regulated at the posttranscriptional level. Again, however, there is no evidence that either transcript is stabilized over the other during growth inhibition.
The analysis of the primary sequence of the two alternative PMP22 promoters provides some hints toward their specific regulation. The promoter 1 structure is similar to tissue-specific promoters with a non-canonical (T)ATA box at the optimal distance of 30 bp upstream of the major transcription initiation site. In addition, an inverted CCAAT box is found upstream of the (TIATA box. Interestingly, the sequence around the trans. Pareek et al., unpublished observations. l b o Promoters for Alternative PMP22 Dunscripts scription initiation site, TCAG is also found at the same or a similar position in several other myelin protein genes including the PO gene , the myelin basic protein gene (Gow et al., 1992), and the proteolipid protein gene (MacMin et al., 1987). It is tantalizing to speculate that coordinate myelin gene transcription might involve a regulatory step using a common "initiator site" (Cherbas and Cherbas, 1993). Promoter 2 displays several features of a "housekeeping" gene promoter. No TATA-like sequence is obvious, and a generally high GIC content (62% compared to 50% in promoter 1) can be found in the first 352 bp upstream of the transcription initiation site. In addition, there is a consensus sequence for an Sp-1-binding site, a transcription factor that is often involved in the regulation of housekeeping genes.
One of the goals of this study was to begin to dissect the regulation of the PMP22 gene in order to gain a better appreciation for the hypothesized dosage effect of PMP22 in CMTl and HNPP. Toward this end we have determined that the regulation of the exon Id-containing transcript is the major site of control for PMP22 mRNA expression in peripheral nerves, the tissue most affected by known mutations of the PMP.22 gene. If altered dosage of PMP22 mRNA truly leads to both CMTl and HNPP, the findings presented here lead to the prediction that there might be a subset of patients who, clinically, will have CMTl by virtue of mutations in the regulatory regions governing the expression of exon 1A. Further studies will be necessary to identify such regulatory factors in order to confirm this hypothesis.