Cloning of the gene for human pemphigus vulgaris antigen (desmoglein 3), a desmosomal cadherin. Characterization of the promoter region and identification of a keratinocyte-specific cis-element.

Pemphigus vulgaris antigen is a cadherin-like desmosomal cell adhesion molecule expressed primarily in suprabasal keratinocytes within the epidermis. Previously characterized structural features have defined this molecule as a desmoglein, DSG3. In this study, we have cloned the human DSG3 gene and examined the transcriptional regulation of its expression. The total gene consisted of 15 exons and was estimated to span >23 kilobases. Comparison of exon-intron organization of DSG3 with bovine DSG1 and several classical cadherin genes revealed striking conservation of the structure. Up to 2.8 kilobases of the upstream genomic sequences were sequenced and found to contain several putative cis-regulatory elements. The promoter region was GC-rich and TATA-less, similar to previously characterized mammalian cadherin promoters. The putative promoter region was subcloned into a vector containing chloramphenicol acetyl transferase reporter gene. Transient transfections with a series of deletion clones indicated that the DSG3 promoter demonstrated keratinocyte-specific expression, as compared with dermal fibroblasts examined in parallel, and fine mapping identified a 30-base pair segment at −200 to −170 capable of conferring epidermal specific expression. The results provide evidence for the transcriptional regulation of the pemphigus vulgaris antigen gene, potentially critical for development of the epidermis and physiologic terminal differentiation of keratinocytes.

Pemphigus vulgaris antigen is a cadherin-like desmosomal cell adhesion molecule expressed primarily in suprabasal keratinocytes within the epidermis. Previously characterized structural features have defined this molecule as a desmoglein, DSG3. In this study, we have cloned the human DSG3 gene and examined the transcriptional regulation of its expression. The total gene consisted of 15 exons and was estimated to span >23 kilobases. Comparison of exon-intron organization of DSG3 with bovine DSG1 and several classical cadherin genes revealed striking conservation of the structure. Up to 2.8 kilobases of the upstream genomic sequences were sequenced and found to contain several putative cis-regulatory elements. The promoter region was GCrich and TATA-less, similar to previously characterized mammalian cadherin promoters. The putative promoter region was subcloned into a vector containing chloramphenicol acetyl transferase reporter gene. Transient transfections with a series of deletion clones indicated that the DSG3 promoter demonstrated keratinocytespecific expression, as compared with dermal fibroblasts examined in parallel, and fine mapping identified a 30-base pair segment at ؊200 to ؊170 capable of conferring epidermal specific expression. The results provide evidence for the transcriptional regulation of the pemphigus vulgaris antigen gene, potentially critical for development of the epidermis and physiologic terminal differentiation of keratinocytes.
Pemphigus vulgaris is an acquired blistering skin disease characterized by circulating IgG autoantibodies in the serum of affected individuals (1). These circulating IgG antibodies have been shown to be pathogenetic by intracutaneous injection of the IgG fraction into the skin of neonatal mice with resultant intraepidermal blistering (2). Similarly, dysadhesion of the epidermal keratinocytes has been demonstrated in a skin organ culture system by incubation with pemphigus vulgaris sera (3,4). Therefore, the patients' sera were used to characterize the antigen responsible for epidermal cell adhesion. Early immunofluorescence studies demonstrated that these antibodies recognize epitopes in the intercellular space within stratifying squamous epithelia, such as the human epidermis (5). Immunoblot analyses identified a 130-kDa glycoprotein, which was designated as pemphigus vulgaris antigen (PVA) (6). 1 Immunoelectron microscopic studies localized the PVA epitopes to the extracellular portion of keratinocyte desmosomes within the epidermis (7). Understanding of the molecular structure of PVA was furthered by isolation of cDNAs corresponding to human PVA sequences (8). A partial PVA cDNA clone was isolated from a gt11 expression library by immunoscreening using serum from a patient with pemphigus vulgaris. This cDNA was used to screen human epidermal keratinocyte cDNA libraries, resulting in the isolation of two additional overlapping cDNAs, which delineated an open reading frame of 2,997 bp of the mRNA (8). Examination of the deduced amino acid sequences, without post-translational modification, suggested that PVA mRNA encodes a 115-kDa polypeptide with homology to the desmoglein family of desmosomal transmembrane adhesion proteins and was subsequently designated as desmoglein 3 (DSG3) (8,9).
The desmogleins are members of the cadherin supergene family, sharing several features with classical cadherins (10). Thus, PVA is a cadherin-like desmosomal cell adhesion molecule that demonstrates tissue-specific expression, as determined at the protein and mRNA levels. However, the mechanisms responsible for the tissue-specific expression of this gene or the cis-elements regulating its expression have not been delineated. In this study, we have cloned human PVA genomic sequences, delineated the entire exon-intron organization, and characterized the promoter region. We have also identified a 30-bp cis-element at Ϫ200 to Ϫ170 that contributes to the keratinocyte-specific expression of PVA.

MATERIALS AND METHODS
Isolation and Characterization of PVA Genomic Clones-To isolate human PVA genomic clones, a 468-bp cDNA was generated by RT-PCR using a commercial kit (Life Technologies, Inc.). RNA was isolated from * Supported in part by the U.S. Public Health Service, National Institutes of Health, Grants P01-AR38923 and T32-AR07561. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM (11), and a primer D3 (see Fig. 5) corresponding to positions 529 -509 was used for reverse transcription using avian myeloblastosis virus reverse transcriptase. Two primers, PUI at Ϫ73 to Ϫ53 (5Ј-TAGAGCCCGACATGTG-3Ј) and D4 at 395-375 (5Ј-CATTAACTGCAGACGGCTGC-3Ј) were used to PCR amplify the cDNA. The corresponding 468-bp product was subcloned into the PCR vector, PCRII (Invitrogen, San Diego, CA), and sequenced (12). This insert was radioactively labeled by nick translation and used to screen a FIXII genomic library (Stratagene, La Jolla, Ca). Eight positive genomic clones were identified and purified (Quiagen, Chatsworth, CA). Subsequent characterization of these clones by restriction enzyme digestions revealed that they represented four unique clones. These clones varied from 15.7 to 18.3 kb in size, as determined by 0.8% agarose gel electrophoresis of the inserts. The relative alignment of the genomic clones was determined by restriction enzyme digestions with BamHI, HindIII, XbaI, EcoRI, SacI, and SalI, under conditions recommended by the manufacturer (IBI, New Haven, CT). The restriction enzyme digests were analyzed on 0.8% agarose gel electrophoresis and subsequently by Southern hybridization utilizing PVA cDNA-specific oligomer primers (12). A 3.4-kb XbaI fragment containing the promoter (see "Results") was subcloned into the Bluescript (Stratagene) vector and sequenced in both directions using standard techniques (13) (U.S. Biochemical Corp.). The nucleotide sequence data were analyzed for eukaryotic promoter consensus sequences using a computer program, SIGNAL (pcGene).
Elucidation of the Exon-Intron Organization-Direct cycle sequencing (Amplitaq, Perkin-Elmer, Norwalk, CT) of genomic clones with cDNA-specific primers permitted identification of the exon-intron borders. These analyses identified 11 exons, which corresponded to ϳ50% of the 5Ј-half of the coding region of PVA cDNA (8). To identify the remaining exons, total genomic DNA, isolated from peripheral blood leukocytes, was used as template for PCR amplification using primers designed on the basis of cDNA sequences. The PCR products were subjected to cycle sequencing using the same primers in an automated nucleotide sequencer (ABI).
Northern Analysis-Total RNA for Northern blot analysis was isolated from cultured normal human keratinocytes (NHKs) (11) and hybridized with a 468-bp PVA cDNA probe under standard conditions (12). For comparison, parallel filters were hybridized with BPAG1 and BPAG2 cDNAs (14).
Determination of Transcription Initiation Sites-Two complementary approaches were taken to determine the 5Ј-end of the PVA mRNA. First, primer extension analyses were performed (Promega PE kit) using an antisense 20-bp oligomer corresponding to nucleotide positions ϩ3 to Ϫ18, 5Ј-end-radiolabeled with [␥ 32 P]dATP and annealed to keratinocyte RNA. The primer was extended to the 5Ј-end of the mRNA using avian myeloblastosis virus reverse transcriptase. The radioactive primer extension products were fractionated on 8.0% polyacrylamide gels, and autoradiograms were developed by exposure of the gel to x-ray film. The sizes of the bands were estimated by comparison with the radioactive molecular weight markers ⌽X174 DNA HindI included in the kit.
The second approach to determine the transcription initiation site was to develop a first-strand cDNA from keratinocyte RNA using an antisense 21-bp oligomer, D3 (see Fig. 5), in the position extending from 529 to 509. The newly synthesized cDNA was then used as a template for PCR amplification with a series of primer pairs that all consisted of a common downstream primer, Dl, in positions 89 -70, and an upstream primer that was placed in different locations in the 5Ј-flanking region. To delineate the transcription initiation site, parallel amplification of genomic sequences was performed utilizing the same set of the upstream primers as used for amplification of the cDNA. However, since the first intron was found to be relatively large (Ͼ6.5 kb), the genomic downstream primer was placed at the 5Ј-end of the intron 1 (the sequence of this primer, D2, was 5Ј-TCGCAATCCTCCATGAACCA-3Ј). The conditions for PCR amplification were as follows: 94°C for 4 min followed by 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min for 38 cycles utilizing Taq polymerase (Perkin-Elmer) and a thermal cycler (OmniGene, Marsh Biomedical, Rochester, NY). The amplification products from cDNA and genomic DNA templates were examined by 1.5% agarose gel electrophoresis.
Development of PVA Promoter/CAT Constructs-Two of the genomic phage clones (see Fig. 1, numbers 1 and 2) were shown to contain the 5Ј-end of the coding region, as well as flanking DNA sequences (see "Results"). To examine the functionality of the putative promoter region within the 5Ј-end of the PVA gene, a series of promoter/CAT reporter gene constructs were developed. Specifically, PCR amplification products were synthesized using clones 1 or 2 as templates and primers that extended from position Ϫ1 upstream to positions Ϫ2,300, Ϫ1,200, Ϫ600, Ϫ500, Ϫ400, Ϫ300, Ϫ260, Ϫ230, Ϫ200, and Ϫ160, respectively. The upstream primers were designed to contain a restriction enzyme site for XbaI, and the downstream primer was designed with a restriction enzyme site for HindIII, which allowed direct insertion of the PCR products into the promoterless CAT construct, AceB (U.S. Biochemical Corp.). These constructs were sequenced for verification by standard techniques (U.S. Biochemical Corp.). A commercially available construct, AceC, in which the SV40 promoter was cloned into the AceB plasmid, was also used for transfections as a positive control. To identify keratinocyte-specific cis-elements (see "Results"), 36-bp oligonucleotides containing 30-bp segments within the segments spanning from Ϫ300 to Ϫ170 and flanked by restriction enzyme sites for HindIII and BamHI were cloned in front of a TK-CAT construct (pBLCAT2). The keratinocyte-specific region was confined to Ϫ200 to Ϫ170, and subsequently, four additional constructs containing 2-bp substitutions in the putative Oct binding site in this region were developed.
Transient Cell Transfections and CAT Assays-The newly synthesized PVA promoter/CAT constructs were used for transient transfections of primary cultures of NHKs or normal human fibroblasts (NHFs) as described elsewhere (14).
Following transfection, the cells were incubated for 36 h at 37°C, and cellular proteins were isolated and used for assay of CAT activity (14). CAT activity was determined as the percentage of the acetylated form of chloramphenicol relative to total radioactive chloramphenicol in the sample. The CAT activities were corrected for the amount of protein in each cell extract.

RESULTS
Cloning Strategy of Human PVA Genomic DNA-PVA genomic clones were isolated using a 468-bp cDNA probe corresponding to the 5Ј-end of mRNA. The probe was used to screen a human lung fibroblast FIXII genomic DNA library. Initial hybridization identified four unique clones (clones 1-4), which contained inserts of 18.5, 15.7, 16.1, and 16.0 kb in size, respectively (Fig. 1). These clones were purified and subjected to further analysis by restriction enzyme digestion and nucleotide sequencing. Enzyme digestions with a variety of restriction endonucleases (see "Materials and Methods") allowed alignment of these four clones (Fig. 1). The results indicated that clones 1 and 2 were overlapping with each other, while clones 3 and 4 contained overlapping sequences. On the other hand, there was no overlap between clones 1 and 2 with either clone 3 or 4, as determined by restriction enzyme site mapping (Fig. 1). The total genomic DNA distance contained within these clones is approximately 40 kb.
Exon-Intron Organization of the Gene-The -phage clone 1 was sequenced with primers corresponding to the 5Ј-end of the open reading frame as well as to the 5Ј-untranslated region of the cDNA sequence, previously published (8). An exon-intron border was observed 46 bp downstream from the translation initiation site, where the genomic clone sequence diverged from the cDNA sequence. This border was subsequently shown to be the exon 1/intron 1 border (see below). Similar sequencing of clones 3 and 4 with primers that corresponded to the cDNA sequences downstream from exon 1 identified 10 additional exons (numbers 2-11) (Fig. 1). Comparison of the sequences encoded by the 11 exons revealed that they corresponded to approximately 50% of the coding region of PVA cDNA. To identify the remaining exons, total human genomic DNA was used as template for PCR amplification utilizing primers synthesized on the basis of coding sequences in the 3Ј-half of the gene. Sequencing of the PCR products identified four additional exons, and therefore DSG3 is composed of a total of 15 exons (Fig. 1). The sites of the exons, as determined by nucleotide sequencing and by comparison with the open reading frame in the cDNA (8), varied from 33 to 864 bp (Table I). All exons were flanked by consensus 5Ј-donor and 3Ј-acceptor splice sites (Table I).
The sizes of the introns were estimated from the size of PCR products generated by amplification of genomic DNA by prim-ers placed on adjacent exons. Using this methodology, the sizes of introns 2-14 were estimated to be ϳ0.3 to ϳ2.8 kb (Table I).
No PCR product was obtained when attempts were made to amplify intron 1 sequences. On the basis of sequence information on clones 2 and 3, the size of intron 1 was determined to be Ͼ6.5 kb (see "Discussion"). Based on combined sizes of the introns and exons, the entire human PVA gene is estimated to be Ͼ23 kb.
Characterization of the 5Ј-End and the Flanking Region of the Gene-To identify putative cis-regulatory elements within the 5Ј-flanking DNA, 2.8 kb of genomic DNA upstream from exon 1 was sequenced (Fig. 2). The results were analyzed using a computer data base in the program SIGNAL, which compared this sequence with known eukaryotic cis-elements. Several putative cis-elements, which have been shown to be important in the expression of other epidermal genes were identified (Fig. 2). These regulatory cis-elements included four AP-1, three AP-2, and one SP-1 binding site (15,16). In addition, a putative glucocorticoid-responsive element (TGTCCT) (17) was found in positions Ϫ2205 to Ϫ2200 (in relation to the translation initiation site). A CK-8mer, a sequence suggested to confer keratinocyte-specific expression to certain keratins was found in positions Ϫ2395 to Ϫ2388 (18). Three segments conforming to the consensus sequence, T(G/T)NNG(C/T)AA(G/T), for C/EBP ciselement, which has been implicated in terminal differentiation processes (19,20), were identified. Finally, an Oct-like motif, ATGCAAGC (at Ϫ190 to Ϫ182), was identified. The functionality of these putative cis-elements will be determined in future studies. No canonical TATA or CAAT sequence was identified within the first 300 bp upstream from the translation initiation site. However, a 17-bp region at Ϫ154 to Ϫ137 upstream from the translation initiation site was found to be GC-rich (82.3%).
Demonstration of PVA mRNA Expression in Cultured Keratinocytes-Northern analysis of keratinocyte mRNA isolated from cells grown under conditions described under "Materials and Methods" revealed the expression of specific transcripts for the bullous pemphigoid antigen-1 (BPAG1) and bullous pemphigoid antigen-2 (BPAG2) genes of 9.0 and 6.0 kb, respectively (21). Hybridization of the same filters with the 468-bp cDNA revealed the presence of characteristic 6.0-kb PVA mRNA (Fig.  3), indicating that the NHK cells used in this study express the PVA gene.
Identification of Transcription Initiation Site(s)-To determine the transcription initiation site(s) within the 5Ј-flanking DNA, primer extension utilizing human keratinocyte RNA as a template was performed. A major primer extension product with an apparent size of 112 nucleotides was detected in two separate experiments. This extension product would place the major transcription initiation site at position Ϫ109 upstream of the translation initiation site (Fig. 4). Extended exposure of the autoradiograms revealed two minor bands, approximately 143 and 91 bp in size (see Fig. 4). The transcription initiation sites for these two extension products correspond to positions of about Ϫ140 and Ϫ88, respectively.
To confirm the results of the primer extension experiments, an RT-PCR analysis of PVA mRNA was performed (Fig. 5). For   FIG. 1. Isolation of genomic DNA clones corresponding to the human PVA gene and its exon-intron organization. A 468-bp cDNA corresponding to the 5Ј-end of the published (8) partial PVA cDNA was used to screen a FIXII human lung fibroblast genomic DNA library. Four unique genomic clones (numbers 1-4), varying from 15 to 18 kb in size, were isolated. Restriction enzyme digestions with EcoRI, SacI, HindIII, and BamHI endonucleases and Southern blot hybridizations with exon-specific oligonucleotides allowed precise alignment of the clones. As a result of characterization of these clones, the first 11 exons of the PVA gene were localized using cycle sequencing. Subsequent sequencing of downstream human genomic DNA identified four additional exons. Thus, the entire gene contains 15 exons, which span Ͼ23 kb.  3  132  1300  gtaagt  atttggggctgtag  4  154  450  gtaagt  ccctctgttcctag  5  146  600  gtaagt  tcctgtattcctag  6  166  1000  gtacag  ttatccaaaattag  7  129  300  gtacac  tctcgccttttcag  8  186  2800  gtaagg  tattttttctccag  9  270  740  gtaaga  atattatgaaacag  10  140  1300  gtaaga  cattgttcttacag  11  225  1800  gtgagt  ttctttctctacag  12  401  300  gtaagc  cacatggtttgcag  13  63  1700  gtaagt  tcccttgtttttag  14  283  1300  gtaatt  tttctctgtcttag  15  864 this purpose, the first strand cDNA was generated using an antisense oligonucleotide, D3 (Fig. 5), which extended the reaction from 529 to 509 toward the 5Ј-end of the mRNA. The newly synthesized cDNA was then used as template for PCR amplification using a downstream primer (D1, see Fig. 5) and a set of upstream primers (numbers 1-8, and a), which correspond to different segments of the genomic DNA. In parallel, PCR amplification of genomic DNA using the same set of upstream primers and a downstream primer in intron 1 (D2 in Fig. 5) was performed. PCR analysis of cDNA showed that primer pairs containing the upstream primers 1-4 clearly resulted in PCR products of the expected size, indicating that these sequences were present in the cDNA synthesized from the keratinocyte mRNA (Fig. 5C). In contrast, primer pairs containing the upstream primer 5, 6, 7, or 8 failed to amplify the cDNA, indicating that these sequences were not present in the cDNA. An additional primer, a, was then designed to define the sequence between primers 4 and 5, and was used for similar PCR analysis. The primer a did not yield a product when cDNA was used as a template (data not shown). When genomic DNA was used as a template, primers 1-8 (Fig. 5C) and a (not shown) resulted in PCR products of the expected sizes. These data narrow the region of the transcription initiation site to the sequences between primers 4 and a and are consistent with the site at about Ϫ140, thus confirming the primer extension analyses. Collectively, the primer extension and RT-PCR analyses indicated that there are three potential transcription initiation . Lane 2 shows a negative control (kit mRNA omitted). Lane 3 reveals the results of primer extension reaction with keratinocyte RNA (10 g) using a PVA-specific oligomer P18 as primer (see text). A major band of about 112 bp could be seen after overnight exposure (thick arrow). Upon a longer exposure (not shown), two additional minor bands, 142 and 90 bp, could be visualized (thin arrows). Lane 4 is the negative control primer extension reaction using P18 oligomer without RNA. Lane M contains radiolabeled molecular weight markers ⌽X174 DNA/HinfI. sites, the major one residing at the approximate position Ϫ109 and two minor transcription initiation sites at positions Ϫ140 and Ϫ88. These sites reside just downstream of the GC-rich region, suggesting that the PVA promoter is GC-rich and TATA-less, similar to previously characterized mammalian cadherin promoters (22)(23)(24).
Demonstration of Functional PVA Promoter Activity-A 2.3-kb PVA promoter/CAT construct (phPVA2.3), the positive control plasmid containing the SV40 promoter (AceC), and control plasmid (AceB) were transiently transfected into NHK cells (Fig. 6). The resulting CAT activity of the PVA promoter construct was 75.8% of the AceC control, whereas the AceB control showed negligible CAT activity. These results clearly demonstrate the functional promoter activity of the PVA gene.
To identify putative cis-regulatory elements within the 2.3-kb region, deletion constructs were developed by PCR amplification of segments extending upstream from position Ϫ1 (Fig. 6). Parallel transfection of keratinocytes with these deletion constructs demonstrated reproducible variability in CAT expression, suggesting the presence of up-regulatory (enhancer-like) and down-regulatory (silencer-like) elements (Fig. 6).
Identification of a Keratinocyte-Specific cis-Element-To examine the keratinocyte-specific expression of the PVA promoter, the same series of deletion constructs extending from Ϫ2300 to Ϫ160 was used for transient transfections of NHK and NHF cultures in parallel. The five longest clones with their 5Ј-ends extending from Ϫ2,300 to Ϫ400 demonstrated little, if any, expression in fibroblasts, in comparison with the SV40/ CAT construct (AceC) (Fig. 6). Quantitation of the expression revealed that the activity in NHFs was consistently less than 4% of the corresponding activity noted with the same constructs in NHK cultures.
Further downstream 5Ј-deletion constructs revealed evidence for cell type-specific expression of the promoter. Specifically, while transfection within constructs extending from Ϫ300 to Ϫ200 showed predominant expression in keratinocytes, transfections with the construct Ϫ160 phPVA0.16 revealed ϳ2-fold higher CAT activity in fibroblasts than in keratinocytes (Fig. 6). These results suggested the existence of keratinocyte-specific cis-elements between Ϫ200 and Ϫ160.
To fine map the putative cis-elements conferring keratinocyte-specific expression to the PVA gene, 30-bp fragments covering 120 bp between Ϫ290 and Ϫ170 of the promoter were cloned in front of a TK-CAT construct and used for similar transfection in keratinocyte and fibroblast cultures. The results, shown in Fig. 7, revealed that the relative activity was markedly higher in keratinocytes with a construct containing the segment from Ϫ200 to Ϫ170 in front of the TK promoter. Careful examination of the corresponding sequence revealed the presence of the ATGCAAGC motif, which has partial homology to the Oct consensus sequence, ATGCAAAT (Fig. 7). However, development of TK-CAT constructs containing the Ϫ200 to Ϫ170 segment, in which 2-bp nucleotide substitutions in the putative Oct domain were introduced, revealed that this motif was not functional in providing keratinocyte-specific expression.

DISCUSSION
The cadherins constitute a supergene family of cell adhesion proteins, which can be further divided into two subfamilies, namely the classical cadherins and desmosomal cadherins (10). Within the desmosomal cadherins, there are two classes of adhesion molecules, the desmogleins and the desmocollins (25). In this study, we cloned and analyzed the gene for pemphigus vulgaris antigen, a desmoglein type cell adhesion molecule. Specifically, we have characterized the human PVA gene, DSG3, and the upstream regulatory sequences. The entire human DSG3 gene was found to consist of 15 exons, and the gene has been mapped to human chromosome 18q (26).
To examine the evolutionary divergence of the gene structure between different cadherins, the sizes of the DSG3 exons were compared with those in previously published cadherin genes (Table II). The exon sizes, in general, demonstrated striking conservation in exons 3-11 between human DSG3, bovine DSG1, and mouse P cadherin (Table II). In particular, the conservation of exon sizes between DSG1 and DSG3 was remarkable, with the exception of exon 12. Careful alignment of the genes encoding classical cadherins of different species with the desmoglein genes indicated that the mouse E-cadherin, the human and mouse N-cadherin, and the chicken L-cadherin genes consisted of 16 exons, the correspondence of DSG3 exons 3-11 being highest with cadherin gene exons 4 -12 (Table II). Thus, the cadherin genes may have an evolutionary insertion of a 5Ј-exon, or alternatively, DSG1, DSG3, and P-cadherin genes may have an evolutionary deletion of a 5Ј-exon (Table II).
In contrast to the conservation of the exon sizes, the introns varied widely in their sizes (22,(27)(28)(29)(30)(31). It has been suggested that the variation in the intron size prevents intergenic crossover events within the cadherin gene family (29). The exact size of the first intron of DSG3 could not be determined, since exon 1 and exon 2 resided in different phage clones, which were apparently not overlapping, as judged by the lack of common restriction enzyme fragments. Attempts to PCR amplify intron 1 sequences from genomic DNA were also unsuccessful. However, the sequence downstream from exon 1 in clone 1 was 2 kb in size, and the sequence upstream from exon 2 in clone 3 was 4.5 kb in size. Thus, the first intron appears to be at least 6.5 kb in size. This observation is consistent with the exon-intron organization of the 5Ј-end of other cadherin genes, which also FIG. 5. RT-PCR strategy for identification of the 5-end of the human PVA mRNA. A, nucleotide sequences of the PVA gene spanning from Ϫ1 to Ϫ300 upstream from the translation initiation site and the positions of upstream oligomer primers 1-8 and a used for PCR amplifications are indicated (see Fig. 2 and Table II). B, position of antisense primer D3 used for the first strand cDNA synthesis from keratinocyte mRNA using avian myeloblastosis virus reverse transcriptase. Primer D1 was used for cDNA PCR amplification with the upstream primers shown in A. Primer D2, placed at the 5Ј-end of intron 1 (not shown) was used as the downstream primer for amplification of genomic DNA (see text). C, amplification products using cDNA (left panel) or genomic DNA (right panel) as template. Note the clearly detectable PCR amplification products with upstream primers 1-4 and no products with primers 5-8 when cDNA was used as a template. All eight upstream primers yielded a PCR amplification product when genomic DNA was used as a template. MW lanes are the ⌽X174 DNA HaeIII molecular weight markers.
demonstrate the presence of large first introns, up to 100 kb in size (28 -30, 32). Thus, the 5Ј-end of the human PVA promoter gene has similarities with other previously characterized nonhuman cadherins.
Genomic cloning of the PVA sequences allowed identification of three transcription initiation sites, with an apparent major site residing at the approximate position of Ϫ109 and two minor transcription initiation sites at positions Ϫ140 and Ϫ88, in relation to the putative translation initiation site. In this context, it should be noted that original human PVA cDNA cloning (8) identified two methionine residues at the potential translation initiation site. These investigators assigned the first methionine residue as the most amino-terminal amino acid of the putative polypeptide. We have recently cloned the corresponding mouse cDNA sequences (33) and demonstrated that the first methionine residue within the human sequence is conserved in the mouse gene, while the second methionine residue is substituted by a threonine. Thus, the first methionine residue in the human sequence is also likely to be the first amino acid of the putative polypeptide and is consequently assigned as ϩ1.
Examination of the upstream sequences identified features characteristic of eukaryotic promoters, including previously cloned mammalian cadherin promoters (22)(23)(24). Specifically, the PVA promoter had a GC-rich region around Ϫ155 to Ϫ137 upstream from the translation initiation site. Furthermore, similar to mouse E-cadherin and P-cadherin, as well as chicken L-cadherin sequences, there was no canonical TATA box. These features are consistent with identification of multiple transcription initiation sites within the promoter region (34).
Pemphigus vulgaris antigen has previously been shown to be expressed primarily in stratifying squamous epithelia, such as the epidermis of the skin. In this study, we developed a series of functional PVA promoter/CAT reporter gene constructs and tested the tissue-specific expression by parallel transfections of cultured keratinocytes and fibroblasts. The results clearly demonstrated that the PVA promoter is active in normal human keratinocytes, while fibroblasts expressed the promoter activity at an extremely low level. These observations indicate that the 5Ј-flanking sequences of the PVA gene contain cis-elements that confer tissue-specific expression to the gene. This observation is also consistent with the presence of tissue-specific cis-elements found in the 5Ј-flanking regions of other epidermal specific genes, such as BPAG1, and keratins 5 and 14 (14,35). This does not, however, exclude the role of possible regulatory elements located within introns or the 3Ј-flanking DNA in tissue-specific expression. For example, the keratin 1 gene contains a negative tissue-specific element in the 3Ј-flanking DNA (36).
Transient transfections of 5Ј-deletion constructs of the PVA FIG. 7. Identification of a keratinocyte-specific segment within the PVA promoter. Oligomers, corresponding to 30-bp segments within the PVA promoter in the positions between Ϫ290 and Ϫ170, were cloned in front of TK-CAT and used in transient transfections of keratinocytes and fibroblasts in culture. Schematic structures of the constructs are shown on the left, and the relative activity of each construct is indicated on the right. Note the nucleotide sequence within the Ϫ200 to Ϫ170 segment, which contains an 8-bp segment (boldface type) with partial homology (*) to the Oct consensus motif.
FIG. 6. Development of human PVA promoter/CAT constructs and their relative activities. A 3.4-kb XbaI fragment of phage clone number 1 was used as a template for PCR amplification of the promoter region of the PVA gene. The PCR products extending from Ϫ1 to Ϫ160 through Ϫ2,300 bp in size were subcloned into promoterless CAT construct, AceB (left panel). These deletion constructs were used for transient transfections of cultured keratinocytes ( ), and fibroblasts (f), as described under "Materials and Methods." The promoter activity was determined by measuring radioactive counts in the acetylated and nonacetylated forms of [ 14 C]chloramphenicol and is expressed relative to AceC activity (1.0). AceC is a positive control with the SV40 promoter subcloned into the AceB, a promoterless CAT construct; the latter one was used as the control and showed negligible activity (right panel).
gene suggested that the tissue-specific elements reside within 0.3 kb of the 5Ј-upstream sequences. Careful examination of putative cis-elements within this region demonstrated that there is an AP2 site at Ϫ149 to Ϫ142. This site is at the border of the most upstream transcription initiation site, and it is conceivable that this AP2 element plays a role in keratinocytespecific expression of the PVA gene. This suggestion is supported by previous findings on the role of AP2 in expression of various keratins, including KRT5 and KRT14 (16). Search for other putative cis-elements in the PVA promoter region identified several AP1 binding sites, two additional AP2 sites, a putative glucocorticoid-responsive element, and one Sp1 site, but the functional significance of these cis-elements is currently unknown. It was of interest that a sequence previously designated as CK-8mer (18), which has been suggested to confer tissue-specific expression to human keratin 14, was found in the position Ϫ2,395 to Ϫ2,388. It should be noted that a similar CK-8mer sequence has previously been detected in the 230-kDa bullous pemphigoid antigen gene promoter region without apparent functional significance (14). It appears, therefore, that this particular sequence may not play a significant role in conferring keratinocyte-specific expression to the PVA gene. Instead, transient transfections with the 5Ј-deletion constructs of the PVA promoter used in this study were able to identify a region extending from Ϫ200 to Ϫ170 that conferred keratinocyte-specific expression either when tested in the context of either its own promoter or a heterologous TK promoter linked to the CAT reporter gene. This segment was found to contain a sequence homologous to the Oct motif (37,38). The Oct transcription factors belong to the POU domain family of proteins known to function as developmental regulators, both in early embryogenesis and in cell type-specific terminal differentiation events (37,38). Three of these factors Skn1a, Oct 6, and Oct11, have been shown to be enriched within the epidermis, suggesting a role in the keratinocyte differentiation process. Thus, the putative Oct cis-element identified within the proximal region of the PVA promoter was suggested to contribute to the epidermal specific expression of this gene. However, nucleotide substitutions in this motif disproved this suggestion. Thus, dynamic interplay between a repertoire of distinct regulatory elements is likely to be required for proper tissue and differen-tiation-specific expression of the PVA gene (39).
In summary, the results of this study, which characterize the exon-intron organization of the human PVA gene and demonstrate tissue-specific expression of the gene at the transcriptional level, provide novel insight into understanding the normal physiology of human epidermis.