Novel alternative splicings of BPAG1 (bullous pemphigoid antigen 1) including the domain structure closely related to MACF (microtubule actin cross-linking factor).

BPAG1 (bullous pemphigoid antigen 1) was originally identified as a 230-kDa hemidesmosomal protein and belongs to the plakin family, because it consists of a plakin domain, a coiled-coil rod domain and a COOH-terminal intermediate filament binding domain. To date, alternatively spliced products of BPAG1, BPAG1e, and BPAG1n are known. BPAG1e is expressed in epithelial tissues and localized to hemidesmosomes, on the other hand, BPAG1n is expressed in neural tissues and muscles and has an actin binding domain at the NH(2)-terminal of BPAG1e. BPAG1 is also known as a gene responsible for Dystonia musculorum (dt) neurodegeneration syndrome of the mouse. Another plakin family protein MACF (microtubule actin cross-linking factor) has also an actin binding domain and the plakin domain at the NH(2)-terminal. However, in contrast to its high homology with BPAG1 at the NH(2)-terminal, the COOH-terminal structure of MACF, including a microtubule binding domain, resembles dystrophin rather than plakins. Here, we investigated RNAs and proteins expressed from the BPAG1 locus and suggest novel alternative splicing variants, which include one consisting of the COOH-terminal domain structure homologous to MACF. The results indicate that BPAG1 has three kinds of cytoskeletal binding domains and seems to play an important role in linking the different types of cytoskeletons.

BPAG1 1 and BPAG2 are the autoantigens of bullous pemphigoid (BP), an autoimmune subepidermal skin blistering disease (1). Both are components of hemidesmosomes, and BPAG1 is a 230-kDa cytoplasmic protein and anchors keratin-containing intermediate filaments (IFs) to the inner plaque of hemidesmosomes (1)(2)(3). BPAG2 is a 180-kDa transmembrane protein with interrupted collagen domains in its extracellular part (1)(2)(3). BPAG1 belongs to the plakin family, including plec-tin, desmoplakin, envoplakin, and periplakin, which associate with IFs in various tissues (4,5). Plakins have a common structure consisting of an amino (NH 2 )-terminal plakin domain, a central coiled-coil domain, and a carboxyl (COOH)terminal domain containing the IF binding domain (4,5). In addition, BPAG1 and plectin give rise to alternative splicing variants, which have an actin binding domain (ABD) at the NH 2 terminus (6 -11). BPAG1 with ABD expressed in nervous tissues and muscles are distinguished as BPAG1n from BPAG1e that has no ABD and localizes to the hemidesmosome in the epithelial tissue (11)(12)(13). Consistent with the expression and localization, BPAG1-deficient mice show skin blistering and neurodegeneration (12). Also, the mouse mutant, Dystonia musculorum (dt), which shows neurodegeneration but normal skin, is known to result from loss of BPAG1n but normal expression of BPAG1e (12).
Mammalian protein MACF (microtubule actin cross-linking factor) and Drosophila protein Kakapo are classified as members of the plakin family (13)(14)(15). Human MACF was also reported as ABP620 or trabeculin-␣ (16,17). They are a novel type of plakin proteins, because their COOH-terminal structures containing microtubule binding domain are completely different from the classical plakins. They consist of long spectrin repeats, following two EF-hand calcium binding motifs and a growth arrest-specific 2 protein (GAS2)-related region, which resembles dystrophin rather than plakin (15)(16)(17). However, their NH 2 -terminal ABD and plakin domain are very similar to that of typical plakins. In particular, the ABD and the plakin domain of MACF are closely related to those of BPAG1n not only by sequence but also by their alternative splicing pattern of the NH 2 -terminal portion of the ABD (11,13,18).
Fusion Proteins and Immunoblotting-The PCR products corresponding to each domain of BPAG1e, plectin, and MACF were digested by several restriction enzymes and cloned into pET-32a-c(ϩ) vectors Detergent-soluble and -insoluble fractions from cultured cells and tissues were prepared as described previously (20). SDS-polyacrylamide gel electrophoresis was performed using a 2-15% gradient gel (Daiichi Pure Chemicals Co., Ltd.) and transferred onto a polyvinylidene difluoride or nitrocellulose membrane sheet.
RNA Blot Analysis-Total RNA from DJM-1 cells was prepared as described above and analyzed by RNA blotting as described previously (21,22). Three g of the total RNA was applied to one lane. Human MTN Blot I and II were purchased from CLONTECH. The probes (Fig.  1B) used for detection were prepared from DJM-1 cells by RT-PCR and restriction endonuclease digestion. The DNA fragments purified by agarose gel electrophoresis were labeled with [␣-32 P]dCTP using a random primer labeling kit (RadPrime DNA Labeling System; Invitrogen). The signals were recorded and analyzed by an imaging plate system (BAS2000; Fuji Film).
Antibodies-Monoclonal and polyclonal antibodies against each domain of BPAG1 were prepared by immunizing mice with the purified pET-BPN, -BPC, -BPB, or -BPA fusion protein as described previously (23). Monoclonal antibodies to the rod domain of BPAG1e (mAb-R815 and -1A16) were prepared using the hemidesmosomal fraction as the antigen (23).
Immunofluorescence Microscopy-Freshly prepared tissues were snap-frozen in isopentane precooled in liquid nitrogen. Frozen specimens were cut at 5-6 m thick with a cryostat, mounted on glass slides, air dried, and fixed with acetone at Ϫ20°C for 10 min. DJM-1 cells grown on glass coverslips were fixed with methanol at Ϫ20°C. The fixed specimens were processed for immunofluorescence staining with primary antibodies followed by fluorescein-conjugated secondary antibody as described previously (20).

Novel Alternative Splicing Variants Detected by RT-PCR
Rodless Variant of BPAG1e-When RT-PCR was performed using total RNA from DJM-1 cells with S1 and A1 primers, which bind to 5Ј and 3Ј portions of the BPAG1e (GenBank TM accession number M69225) coding region, respectively, two fragments with ϳ8 and ϳ5 kb were amplified. The size and sequence of the larger one were identical with the known BPAG1e. On the other hand, the smaller one showed also the identical sequence to BPAG1e but lacked the exon encoding the central rod domain as shown in Fig. 1B (24,25).
Variant Expected from the Genomic Sequences-Human BPAG1 is encoded at the chromosomal locus 6p11-12 (26), and the genomic sequence of this region is registered in the Gen-Bank TM data base (GenBank TM accession number AL096710, Fig. 1A). AL096710 contains the large downstream region following the reported BPAG1 gene (25), and the sequence analysis revealed that there is an extraordinary large, more than 5 kb, open reading frame (ORF) in this region (Fig. 1A). We supposed that this ORF might be expressed as another COOHterminal domain of BPAG1 by alternative splicing and designed several primers to bind to BPAG1e or this large ORF region. Indeed, expected cDNA fragments were amplified by RT-PCR with total RNA prepared from DJM-1 cells, and the variant combining these cDNA fragments is shown as BPAG1eB in Fig. 1B  large ORF described above, showed a high degree of sequence identity to the spectrin repeat domain of MACF (GenBank TM accession number AF141968 or AB029290). Furthermore, we analyzed a cDNA clone KIAA0728 in the HUGE data base of Kazusa DNA Research Institute, which consists of a spectrin repeat domain, two calcium binding EF-hand motifs, and a GAS2 domain. GenBank TM accession number KIAA0728 is very similar to the COOH-terminal region of MACF and maps on the same chromosome 6 as BPAG1 (26), whereas MACF maps on chromosome 1 (16). Considering these facts, we designed several primers that bind to BPAG1e, spectrin repeat region of GenBank TM accession number AL096710 or KIAA0728, and performed RT-PCR with total RNA of DJM-1 cells. Actually, expected cDNA fragments were amplified, and the variant combining these cDNA fragments is shown as BPAG1eA in Fig. 1B. The entire domain structure is nearly the same as those of MACF and Kakapo, BPAG1eA; however, it does not have an ABD. The sequence from the region encoding spectrin repeats to that coding for COOH-terminal domains showed 72% similarity to that of MACF and 51% to the Kakapo protein at the amino acid level, respectively.

RNA Blot Analysis
To investigate the tissue distribution of each variant, RNA blot analyses of human tissues with four probes were performed. The probe-N, corresponding to the plakin domain, showed a ϳ15-kb band in some tissues, such as heart, placenta, liver, and most strongly in skeletal muscle, and weakly in kidney, and a slightly smaller band in brain, liver, kidney, and pancreas. In liver and kidney, the two bands were detected. No clear band, but a smear signal, was detected in other tissues, for example in lung, spleen, thymus, prostate, and testis ( Fig.  2A). The probe-A corresponding to the spectrin repeat region also showed a very strong but smear 7ϳ15-kb signal in skeletal muscle and a smear signal in several tissues that we examined. In addition, a 10ϳ11-kb band was weakly detected in testis, ovary, and weakly in small intestine (Fig. 2B). The probe-B corresponding to the large ORF, the second COOH-terminal tail, also showed a strong smear signal in skeletal muscle and a smear signal in heart (Fig. 2C). The probe-C corresponding to the COOH-globular domain gave no signal in any tissues tested (data not shown), but a strong ϳ9-kb band, the size of known BPAG1e (24), was detected in total RNA of DJM-1 cells (Fig.  2D). This ϳ9-kb band was also recognized by the probe-N. The probe-A showed two bands, ϳ10 and ϳ15 kb, and the probe-B detected no signal in DJM-1 RNA (Fig. 2D). These results indicate that the BPAG1 gene is most active in skeletal muscles and cultured keratinocytes. To determine whether the smear signal indicates ubiquitous expression following rapid turnover of mRNA or bad condition of RNA preparations needs further analyses.
The alternative splicing variants suggested here cannot explain the range of mRNAs detected in tissues examined. For example, the probe-N could not detect all variants shown by the other three probes (Fig. 2, A and D), but detected other smaller mRNA. By RT-PCR, a fragment consisting of a part of a rod domain followed by spectrin repeats was also amplified (data not shown). It has been reported that there are alternative splicing variants with and without 300 amino acids at the similar site in Kakapo (27). These data indicated the existence

Immunofluorescence Microscopy and Immunoblot Analysis
To investigate the expression of these variants at protein level, we used several antibodies against each domain of BPAG1. mAb-N46 recognizes the plakin domain of BPAG1. mAb-N619 also recognizes that domain, but cross-reacts with the corresponding domain of MACF. mAb-R815 and mAb-C319 recognize the rod domain and the COOH-terminal globular domain of BPAG1e, respectively (Fig. 3). mAb-1A16 also reacted with the rod domain of BPAG1e (data not shown). pAb-BPB and pAb-BPA are the mouse antisera, which recognize the large ORF domain and the spectrin repeat domain of BPAG1, respectively. However, pAb-BPA also recognized the corresponding domain of MACF weakly (Fig. 3). All the antibodies did not react with the plakin domain or the COOH-terminal globular domain of human plectin (data not shown).
By immunofluorescence microscopy, basement membrane zone of human skin, where hemidesmosomes localize to the basement membrane zone, was clearly stained with mAb-N46, -R815, and -C319, but pAb-BPB showed no clear staining, and the pAb-BPA stained the entire living cell layer weakly instead of the basement membrane zone (Fig. 4). The epithelial cells were also stained with the concentrated N46 (Fig. 4AЈ).
DJM-1 cells show very characteristic distribution of hemidesmosomes (28). When DJM-1 cells were stained with these antibodies, mAb-N46, -R815, and -C319 showed hemidesmosomal pattern clearly, whereas pAb-BPB evidenced no clear staining and pAb-BPA stained the entire cytoplasm weakly (Fig. 5). By immunoblot analysis of DJM-1 cell extract, mAb-N46, -1A16, and -C319 reacted with the 230-kDa band of BPAG1e, but pAb-BPB and pAb-BPA showed no reactivity (Fig. 6). mAb-N619, which cross-reacts with MACF, detected not only the band of ϳ230 kDa but also higher molecular mass polypeptides (Fig. 6). These bands could not be detected by mAb-N46 and pAb-BPA and are expected to be MACF or its fragments.
By immunofluorescence microscopy and immunoblotting, mAb-N46, mAb-R815, mAb-1A16, and mAb-C319 reacted with rabbit BPAG1e, but none of the antibody showed clear staining or recognized any band in rabbit skeletal muscle and brain (data not shown), despite the strong signal detected in RNA blotting (Fig. 2). DISCUSSION Both BPAG1n and MACF belong to the plakin family. They are very similar in the NH 2 -terminal ABD and plakin domain, with 48% identity in amino acid level, but their downstream structures are completely different. BPAG1n has a coiled-coil rod domain and a COOH-terminal globular domain, like other plakins (4,5). MACF, on the other hand, has long spectrin repeats, two EF-hand motifs and a GAS2 region, resembling dystrophin rather than plakins (15)(16)(17).
In light of the diversity of other plakin family members, we wondered whether BPAG1 genomic and cDNA sequences represented novel alternatively spliced products of BPAG1. In fact, we could detect their mRNAs by RT-PCR and RNA blotting. In certain tissues, the probe-N, corresponding to the plakin domain, detected mRNAs of larger size than ϳ9 kbp of BPAG1e mRNA. Since they were also detected by the probe-A and -B, but not by the probe-C, they are thought to represent BPAG1-eA and -eB. However, the signals obtained by probe-A and -B appeared as a smear. This could indicate rapid turnover of these mRNAs or may be depend on the quality of mRNA samples. In total RNA of the keratinocyte, DJM-1 cell, probe-N and -C clearly detected a ϳ9-kbp mRNA, coding the BPAG1e. The probe-N also detected some other smaller mRNAs, although the probe-A detected two bands of large size mRNA. These results do not coincide with our suggestion (Fig. 1). In addition, Leung et al. (29) have reported BPAG1-b of mouse recently, which has the BPAG1eA-like form with the ABD at the NH 2 terminus and the large ORF domain between the linker domain and the spectrin repeat region, although our RT-PCR system did not detect such a product. These differences may depend on a difference between species or the specificity and sensitivity of each probe or more likely suggest more alternatively spliced variants. In fact, many alternative splicing variants have been reported on plakins. BPAG1, MACF, Kakapo, and plectin have splicing variety at the 5Ј-region of ABD (8 -11, 13, 18, 27). Desmoplakin and plectin have rodless forms like the BPAG1eS (7,30,31). In particular, Kakapo have many alternative splicing patterns at both the NH 2 -and COOH-terminus, like the BPAG1 suggested here (27). Therefore, our RNA bolt data could indicate the existence of additional alternative splicing products of BPAG1. BPAG1 expressed in neural tissue is reported to have the ABD (9, 11), but it was not detected by the probe-C. Therefore, the variants expressed in neural tissue most likely include BPAG1eA with the ABD, like BPAG1-a reported in mouse (29).
Immunofluorescence microscopy of DJM-1 cells and skin sections with mAb-N46, -R815, and -C319 clearly showed the typical hemidesmosomal staining pattern (28). In addition, mAb-N46 and pAb-BPA stained the entire cytoplasm of keratinocytes weakly, suggesting the presence of BPAG1eA, which dose not localize to hemidesmosomes. However, the protein product of BPAG1eA was not detected by immunoblotting of DJM-1 cell, in contrast to the clear signal of 230-kDa BPAG1e by mAb-N46, -1A16, and -C319. Surprisingly, despite the strong signal obtained by RNA blotting, the corresponding protein was not detected in brain and skeletal muscle by both immunofluorescence microscopy and immunoblotting. Although the mRNA may not be translated, we could think that the protein is either expressed at low level in these tissues or shows high degree of turnover for the following reasons. First, BPAG1 knockout mice show neurodegeneration in addition to skin blistering (12). Neurodegeneration of Dystonia musculorum (dt) mice is caused by loss of BPAG1n (9,11). Hence the blistering in the skin is caused by loss of BPAG1e from hemidesmosomes, and the neurodegeneration is thought to be caused by loss of BPAG1n. This indicates that BPAG1 must play some important roles in neuron and/or muscle as well as in the skin. Second, BPAG1e bearing hemidesmosomes must exist in epithelial cells in certain tissues we examined in this study (32), but neither the 9-kb band of RNA nor the 230-kDa protein were detected in any tissue. This would be due to the small population of cells having hemidesmosomes and the small amounts of BPAG1e in total tissue extracts. The other variants may also localize to very restricted sites.
Except for the skin blistering and neurodegeneration, no other tissues seem to be affected in BPAG1 null mice despite the mRNA signals in many tissues detected by the probe-N and -A (12). Since BPAG1A and MACF are very similar in structure and sequence, even if BPAG1 is deficient, there might be little effect in other tissues because of functional redundancy with MACF. To analyze these molecules separately, specific antibodies would be very useful, although difficult to generate, given that the similarity between BPAG1 and MACF might produce a cross-reacting antibody. In fact, several mAbs including mAb-N619, which we produced against the plakin domain of BPAG1, reacted very strongly with the corresponding domain of MACF (Fig. 3). Therefore, it is especially important to use highly specific monoclonal antibodies for further investigation.
Plakins have a conserved exon organization. The ABD and the plakin domain are distributed to many exons, in contrast to the rod domain and the COOH-globular domain encoded in one exon, respectively (6,25,(33)(34)(35)(36). MACF does not have the rod and COOH-globular domain but has a spectrin repeat region, GAS2 domain, and EF-hand motif (15)(16)(17). The organization of corresponding exons is very similar among the genes encoding BPAG1 and MACF (data not shown). So, the alternative splicing pattern may also be conserved. Actually, mAb-N619 detected at least three large polypeptides in immunoblotting, which were not recognized by mAb-N46 and expected to be MACF (Fig. 6). They might be alternative splicing variants of MACF.
Recently, GenBank TM accessoin number KIAA0728, which corresponds to the COOH-terminal fragment cDNA clone of BPAG1eA, was reported as MACF2, and its mouse homologue has a functional microtubule binding domain (37). This means that BPAG1 has three cytoskeleton binding domains, i.e. for actin, Ifs, and microtubules, and all of them are alternatively spliced and expressed in a complex splicing pattern. BPAG1 would thus be a multifunctional cytoskeletal linking factor, in addition to its function at the hemidesmosome.