Comparison of molecularly cloned bullous pemphigoid antigen to desmoplakin I confirms that they define a new family of cell adhesion junction plaque proteins.

Bullous pemphigoid is a subepidermal blistering disease in which patients have autoantibodies against the plaque of the hemidesmosome. Starting with a previously isolated 2-kilobase (kb) cDNA for bullous pemphigoid antigen (BPA), we used primer extension of keratinocyte mRNA to isolate overlapping cDNAs with a combined open reading frame of 6.3 kb, encoding most (243 kDa) of the BPA, but lacking the far amino terminus. Analysis of this amino acid sequence revealed a carboxyl-terminal domain containing two regions of 174 and 176 residues with high sequence identity. Most of the amino-terminal two-thirds of BPA is predicted to be in an alpha-helical conformation in which two chains would aggregate into a coiled-coil rod structure. BPA and desmoplakin I, a desmosome plaque protein, show remarkable sequence and structural homology. In its carboxyl-terminal domain, desmoplakin I also has 176 residue repeats with 40% sequence identity to those in BPA. The repeats in both molecules have a regular linear distribution of acidic and basic residues with a period of 9.5, the same as that found in the 1B segment of keratin filaments, suggesting a means of ionic interaction between keratin and these plaque proteins. Also, desmoplakin I, like BPA, is predicted to have a rod domain, which in both proteins has similar regular charge periodicities, suggesting a means of ionic self-aggregation. These findings extend those of Green et al. (Green, K. J., Parry, D. A. D., Steinert, P. S., Virata, L. A., Wagner, R. M., Angst, B. D., and Nilles, L. A. (1990) J. Biol. Chem. 265, 2603-2612) which show that BPA and desmoplakin I represent the first members of a new family of adhesion junction plaque proteins.

Bullous pemphigoid is a subepidermal blistering disease in which patients have autoantibodies against the plaque of the hemidesmosome. Starting with a previously isolated 2-kilobase (kb) cDNA for bullous pemphigoid antigen (BPA), we used primer extension of keratinocyte mRNA to isolate overlapping cDNAs with a combined open reading frame of 6.3 kb, encoding most (243 kDa) of the BPA, but lacking the far amino terminus. Analysis of this amino acid sequence revealed a carboxyl-terminal domain containing two regions of 174 and 176 residues with high sequence identity. Most of the amino-terminal two-thirds of BPA is predicted to be in an a-helical conformation in which two chains would aggregate into a coiled-coil rod structure. BPA and desmoplakin I, a desmosome plaque protein, show remarkable sequence and structural homology. In its carboxyl-terminal domain, desmoplakin I also has 176 residue repeats with 40% sequence identity to those in BPA. The repeats in both molecules have a regular linear distribution of acidic and basic residues with a period of 9.5, the same as that found in the 1B segment of keratin filaments, suggesting a means of ionic interaction between keratin and these plaque proteins. Also, desmoplakin I, like BPA, is predicted to have a rod domain, which in both proteins has similar regular charge periodicities, suggesting a means of ionic self-aggregation. These findings extend those of Green et al. (Green, K. J., Parry, D. A. D., Steinert, P. S., Virata, L. A., Wagner, R. M., Angst, B. D., and Nilles, L. A. (1990) J. Biol. Chem. 265,2603-2612 which show that BPA and desmoplakin I represent the first members of a new family of adhesion junction plaque proteins.
Bullous pemphigoid is an autoimmune disease in which patients develop subepidermal blisters and antibodies directed against the epidermal basement membrane zone (1). The autoantibodies from these patients have been shown to bind to hemidesmosomes, structures thought to be important in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We have previously cloned a partial 2-kb cDNA encoding about 76 kDa of the carboxyl domain of BPA (8). Recent cloning of desmoplakin I (DPI) (9), a desmosome plaque protein (10-12), revealed sequence homology with BPA for about 170 amino acids in the carboxyl domains of both proteins.
In this report, we used the initially isolated BPA cDNA clone to isolate overlapping cDNA clones encoding for 243 kDa of BPA. Analysis of the deduced amino acid sequence of this 6.3-kb open reading frame revealed remarkable sequence and structure homology between BPA and DPI and suggests that these proteins define a new family of adhesion plaque proteins.

EXPERIMENTAL PROCEDURES
Construction and Screening of cDNA Libraries- Fig. 1 shows from which libraries (labeled by Roman numerals) each subclone was obtained. All cDNA libraries were constructed with poly(A)-RNA from cultured human keratinocytes (8). Xgtll library I was constructed by oligo(dT) priming, and clone 1-1, identified by screening with bullous pemphigoid patients' sera, is the one previously described (8). The other libraries were made by primer extension of keratinocyte poly(A)-RNA with reverse transcriptase, followed by RNase H and DNA polymerase I, according to the basic method of Gubler and Hoffman (13) and the specific methods of suppliers (Boehringer Mannheim cDNA Synthesis Kit, Clontech c-Clone I1 cDNA Synthesis Kit, and Invitrogen X Librarian). Primers used were 17 or 18 nucleotides and, for construction of each of the libraries indicated, were complementary to the following areas: 11-5' end of 2-kb EcoRI fragment of cDNA 1-2; 111-141 nucleotides downstream from the 5' end of cDNA 11-1; IV and V-215 nucleotides downstream from the 5' end of cDNA 111-1. EcoRI linkers (Bethesda Research Laboratories) or adaptors (Amersham Corp.) were ligated to the cDNA for cloning into Xgtll (14) or Lambda Zap I1 (Stratagene), with subsequent packaging (Gigapack I1 Gold, Stratagene). Libraries were screened by hybridization with probes labeled by random priming (15, 16). Hybridization was performed in 6 X SSC (1 X SSC = 150 mM NaC1, 15 mM Na:,citrate, pH 7), 0.5% sodium dodecyl sulfate, 12.5 X Denhardt's solution a t 68 "C for 16 h. The most stringent final wash was at 68 "C in 0.1 X SSC, 0.1% sodium dodecyl sulfate. Probes used to screen each library were as follows (see Fig. 1): I-Hind111 218-bp fragment at the 5' end of the 1.5-kb fragment of 1-1, and EcoRI 1.1-kb fragment from the 3' end of 1-2; 11-350-bp EcoRI fragment at the 5' end of cDNA 1-2; 111-216-bp  Northern analysis with isolated cDNAs was performed on 10 pg/ lane of total RNA from keratinocytes, as previously described (8).
AnalFsis of Amino Acid Sequence and Homologies-Secondary structure predictions were carried out using the methods of Chou and Fasman (17) and Gamier et al. (18). In conjunction with these methods, patterns of heptad substructure in the sequence were sought that were compatible both with an n-helical structure and a coiledcoil conformation (19,20).
University of Wisconsin Genetics Computer Group software (21) was used as follows: (i) the RPA amino acid sequence was compared for homologies against the NRRF version 25 protein database using FASTA, and (ii) a comparison of the RPA and the DPI amino acid sequences was undertaken with COMPARE (based on the evolutionary distance between amino acids, with a window of 21 amino acids and a stringency of 15) and plotted with DOTPLOT. Other computer analyses of sequence comparisons were based on the MATCH algorithm (22).
Fast Fourier transforms were used to determine significant periodicities in the linear distribution of acidic and basic residues along the sequence of RPA using the methods described by Parry ( 2 3 (24), and McLachlan and Stewart (25).

RESULTS
Cloning of BPA cDNA-A Xgtll cDNA library (called I), constructed by oligo(dT) priming of poly(A)-RNA from cultured keratinocytes, was screened by hybridization with restriction fragments from cDNA (called 1-1 in Fig. 1) previously cloned from this library (8). This hybridization resulted in finding two overlapping cDNA sequences (Fig. 1, 1-2 and  1-3) containing one long open reading frame (in frame with that previously described (8)) with a stop codon near the 3' end of cDNA 1-3. (The 3'-terminal 131 nucleotides of the previously reported BPA cDNA clone 1-1 were found to be different from the corresponding region of clones 1-2 and 1-3. Since 1-2 and 1-3 have identical sequences in this region, and because the stop codon defined by these clones corresponds exactly to a recently reported mouse partial BPA cDNA (26), we used their sequence to determine BPA amino acid sequence in this paper). Because no further clones could be found in the oligo(dT)-primed library, we used primer extension of keratinocyte poly(A)-RNA with primers near the 5' ends of previously isolated clones to sequentially generate new X cDNA libraries (called 11,111, IV, and V) which were screened with probes from the furthest 5' ends of the clones from which the primers were obtained (Fig. 1). A total of six overlapping cDNA clones (Fig. 1, 1-2  coding sequence upsteam of cDNA V-1. Therefore, in the analysis below of the amino acid sequence of BPA, the extreme amino-terminal sequence is not included. In addition to the finding of a continuous open reading frame in phase with the previously reported frame, the identity of these overlapping clones was confirmed by Northern analysis in which all hybridized to the 9-kb BPA mRNA (Fig.  2) (8, 26).
Analysis of Deduced Amino Acid Sequence of BPA-Excluding the first 153 residues of the BPA sequence, which are predicted to form part of a non-a-helical amino-terminal domain, most of the next 1,162 residues are postulated to form a discontinuous coiled-coil a-helical rod of the general type found in paramyosin, myosin, and keratin molecules (Figs. 3 and 4). This conclusion is based on the following observations. 1) According to the methods of Chou and Fasman (17) and Robson (18) the sequence is predicted to be predominantly @-helical. In particular, 80-90% of the two major rod domains (lengths 67.6 and 62.0 nm, respectively; shaded in Figs. 3, 4), as well as the shorter rod domains 1 (14.3 nm) and 2 (14.4 nm) (Fig. 4) are predicted to be ahelical. Each of the two long rod domains (in contrast to rod domains 1 and 2) have a highly significant periodic linear disposition of acidic and basic residues ( Table I). 2) The rod domains all have a heptad substructure of the form (a-b-c-de-f-g)", where a and d are usually filled by apolar residues (occupancy rate -75%) (27) and are thus capable of forming a close packed hydrophobic core in a two-chain coiled-coil molecule. (Fig. 3 shows the heptad structure of the two major rod domains that are shaded in Fig. 4). 3) The rod segments have a high charged/apolar ratio (1.2-1.5 for all the rod domains in Fig. 4), a feature generally indicative of an elongated structure. 4) When lined up in parallel and without relative axial stagger, the BPA chains are capable of forming many ionic interactions between the e and g residues of different chains, thus stabilizing the coiled-coil rod. The major rod segments and rod 1 in Fig. 4 have about 0.35 and 0.7 interchain ionic interactions/heptad pair, respectively; these are highly significant values (27). However, since rod 2 lacks such interactions, it may have an enhanced flexibility relative to the other rod segments. Note also that two short regions of lengths 58 and 28 residues in the rod domain (shown in Fig. 4 as black bars) have secondary structures that are not yet clearly defined by current methods. However, the latter 772, AA898-1315) and areas marked 1 (AA154-259) and 2 (AA801-897) indicate a-helical coiled-coil rods. Segments 1 and 2, however, lack the regular linear disposition of acidic and basic residues observed for the shaded segments. The black bars between the rod domains indicate structure that is not well-defined using predictive methods. Circles at left represent the amino-terminal domains, which in BPA is predicted to be predominantly p, turn, and coil structure. Boxes labeled A , E , and C show 176-residue amino acid repeats.
Identity indicates same amino acids in segment comparisons, whereas homology indicates chemically similar amino acids ( i e . apolar, acidic, or basic). AA, amino acids region is both apolar and non-a-helical, and could represent a "globular" region at which the rod undergoes a major change in direction.
Analysis of the carboxyl domain of BPA reveals a 174residue sequence and a 176-residue sequence ( B and C, respectively in Fig. 4) that have marked sequence identity and homology (Fig. 5, top). In addition, each of these regions consists of approximately 4.5 repeats of a 38-residue motif whose consensus sequence is shown in Table 11.
Homology of BPA and DPI-Comparison of BPA and DPI by use of an evolutionary matrix of similarity for amino acids demonstrated remarkable homology between these proteins, especially at their carboxyl-terminal domains (Fig. 6). Searching of the NBRF protein database revealed no evolutionary homologies comparable to that seen with DPI, but did reveal slight homology of BPA to myosin (e.g. FASTA score for BPA compared to DPI was 389, whereas BPA compared to several myosins was 64-78; mean score for all sequences was 23). This minor homology to myosin, also a coiled-coil ahelical rod, was entirely contributed by the rod domain of BPA.
Like BPA, DPI has in its carboxyl domain 176 residue repeats containing internal 38 residue motifs (9). There are three of these domains in DPI, called A, B, and C (Fig. 4), and these show remarkable identity and homology (both chemically, Fig. 5, and evolutionarily, Fig. 6) to the domains in BPA. In addition, areas adjacent to these repeats (X and Y in Fig. 4) show a high degree of identity and homology. BPAs B and C domains are so-named because they show most homology to the DPI B and C domains, respectively. Analysis by fast Fourier transform indicates that these repeats in both BPA and DPI show a regular periodic charge distribution that closely approximates that exhibited by the 1B segment of the rod domain of keratin intermediate filaments (Ref. 9 and  Table I), suggesting a potential means of ionic interaction between these filaments and these plaque proteins.
The rod domains of BPA and DPI also show some amino acid homology (Fig. 6) that is about of the same degree as that seen between BPA and myosin, suggesting that it is due to the shared a-helical heptad substructure. As the periodicity in the acidic and basic residues in the rod domain of both BPA and DPI is similar (Table I), a potential means of selfaggregation for each of these plaque proteins through the maximization of ionic interactions becomes self-evident (9).

DISCUSSION
Cloning and analysis of approximately 6.3 kb of cDNA encoding 243 kDa of BPA revealed that BPA consists of two major and distinct structural domains. Most of the approximately two-thirds of the molecule on the amino end is predicted to form an a-helical coiled-coil rod of length -160-170 nm. This rod domain has a regular and alternating acidic and Basic (KR): lysine (K) a n d arginine (R).  indicates chemical similarity. B P -B and BP-C show 35% identity and 57% homology (= identity + similarity). BP-C and DP-C show 43% identity and 64% homology. basic charge periodicity along its length, suggesting that, as predicted for DPI (9), it might self-aggregate through ionic interactions to form filaments or a two dimensional meshwork. The approximately one-third of the molecule on the carboxyl end contains two segments of 174 and 176 residues (called B and C, Fig. 4) that have high homology to each other and contain internal 38-residue repeating motifs (Table 11). Along both of these segments, there is a regular acidic and basic charge periodicity that corresponds to the periodicity seen in the 1B rod domain of keratin intermediate filaments, suggesting a means of ionic interaction between these molecules. Such a prediction is consistent with previous immunoelectron microscopic findings that antibodies raised to the carboxyl domain of BPA bind the inner plaque of the hemidesmosome, where keratin filaments appear to insert (28). It is important to note that sequencing data in this paper shows a difference in the 3' 131 nucleotides that we previously reported (8). Our most recent sequence of the 3' end agrees with the sequence of the mouse BPA cDNA reported by Amagai et al. (26). Whether our previous 3' sequence represents alternative splicing or a mistake in cloning is not yet clear.

ILQFLHEPSSNTRVFPNPNNKQALYYSELLRnCVFDVESQCFLF I A T K L K D Q K S W R N I I C P Q T X R K L T Y K E U E K
The structural and sequence homology of BPA and DPI is remarkable. Both have a predicted rod domain with similar charge periodicities. Also, in their carboxyl-terminal domains there is a high degree of identity and homology of the amino acid sequences, including the organization of 176 residue blocks of repeating amino acids. Even more striking is the finding of almost identical charge periodicities in these domains of both of these molecules.
The homologies noted between BPA and DPI provide one possible explanation for the controversy regarding whether antibodies against DPI, which bind very well to the desmosome plaque, also bind the hemidesmosome plaque. Some