Structural Analysis of the Predicted Coiled-coil Rod Domain of the Cytoplasmic Bullous Pemphigoid Antigen (BPAG1) EMPIRICAL LOCALIZATION OF THE N-TERMINAL GLOBULAR DOMAIN-ROD BOUNDARY*

The bullous pemphigoid antigen BPAG1 is required for keratin filament linkage to the hemidesmosome, an adhesion complex in epithelial basal cells. BPAG1 struc- tural organization is similar to the intermediate fila-ment-associated proteins desmoplakin I (DPI) and plec- tin. All three proteins have predicted dumbbell-like structure with central (cid:97) -helical coiled-coil rod and regions of N- and C-terminal homology. To characterize the size of the N-terminal globular domain in BPAG1, two polypeptides spanning possible boundaries with the coiled-coil rod domain of BPAG1 were expressed in Escherichia coli. BP-1 ( M r (cid:53) 111,000), containing amino acids 663-1581 of BPAG1 (Sawamura, D., Li, K., Chu, M.-L., and Uitto, J. (1991) J. Biol. Chem. 266, 17784– 17790), and BP-1A, with a 186 amino acid N-terminal deletion, were purified. BP-1 and BP-1A behave as highly asymmetric dimers in aqueous solution according to velocity sedimentation and gel filtration. Both have globular heads with rod-like tails of roughly equal length, 55–60 nm, upon rotary shadowing. BP-1A con- tent of (cid:97) -helix, determined by circular dichroism, is (cid:59) 90%, consistent with (cid:97) -helical

The intracellular bullous pemphigoid antigen (BPAG1) 1 is a part of the cytoplasmic plaque of the hemidesmosome, a supramolecular structure that links keratin intermediate filaments to extracellular matrix in a number of epithelial cell types, including epidermal keratinocytes (1,2). BPAG1 (or the 230-kDa bullous pemphigoid antigen) was identified by autoimmune antibodies of patients with the dermal-epidermal blistering disease, bullous pemphigoid (3,4). BPAG1 is rapidly assembled into a stable anchoring contact, or prehemidesmosome, at the ventral surface of freshly plated keratinocytes in culture (5)(6)(7). Disruption of the BPAG1 gene in mice by homologous recombination prevents keratin filament attachment to the hemidesmosome and consequently weakens dermal-epidermal adhesion but does not affect formation of the membraneassociated dense plaque of the hemidesmosome (8). The hemidesmosome also contains the high molecular weight cytoplasmic component HD-1 (9) and two transmembrane proteins, BPAG2, which contains an extracellular collagenous domain (10 -13), and the integrin ␣ 6 ␤ 4 (5,14).
BPAG1 is homologous to two proteins which associate with intermediate filaments (IF): desmoplakin I (DPI), which is part of the desmosome, a cell-cell junction; and plectin, a ubiquitous cytoskeletal protein that binds IF subunits in vitro (15)(16)(17)(18)(19)(20). The predicted molecular masses of the three proteins are Ͼ300 kDa, and all three have homologous N-and C-terminal domains. Cell transfection and molecular studies have demonstrated that the C-terminal domains for plectin and DPI specifically interact with several IF types (21)(22)(23). The central regions of all three proteins, which lack homology to one another, nonetheless contain highly significant heptad repeats characteristic of ␣-helical coiled-coil rods, similar to the rod domain of myosin or to the core of IF heterodimers (20,24). Purified DPI dimerizes and has an extended dumbbell-like shape in which terminal globular regions are separated by a central rod (25). Plectin and BPAG1 appear to take similar but less well-characterized conformations (26,27).
The actual size of the N-terminal globular domains in BPAG1, DPI, or plectin have not been measured directly. In the case of BPAG1, this is of particular interest since the presence of heptad repeats in its N-terminal homology domain has suggested that coiled-coil rod formation may extend as far as 450 amino acids into this domain from the central, nonhomologous rod region (16,18,20). In plectin and DPI, the N-terminal homology region lacks certain characteristics generally associated with coiled-coil structures, such as a high (Ͼ1.0) ratio of charged/apolar amino acids, and, thus, rod formation in these domains is considered unlikely (19,20). In N-terminal conserved regions of BPAG1, however, the presence of heptad repeats along with a high level of predicted interchain charge interactions has led to predictions that ␣-helical coiled-coil formation initiates at residue 708 or 875 (16,18) according to the numbering of Sawamura et al. (18) rather than near the boundary of the N-terminal homology domain (residue 1145) as is the case with plectin and DPI (20). In order to directly establish the boundary of the N-terminal globular and central rod domains in BPAG1, two BPAG1-derived polypeptides containing the predicted N-terminal rod domain transitions were expressed in Escherichia coli and purified. The renatured polypeptides dimerize and, by rotary shadowing, are shown to have highly asymmetric structures in which a globular Nterminal conincides with the N-terminal homology region defined by comparison with plectin and DPI.

EXPERIMENTAL PROCEDURES
Materials-cDNA clones II-1 and III-1, representing a portion of the BPAG1 sequence, were obtained from J. R. Stanley (NIH) and are described in Fig. 1 of Tanaka et al. (16). The pET series of protein expression vectors (28) were supplied by Novagen (Madison, WI). Purified human fibrinogen was a kind gift of S. I. Chung (NIH), while rabbit tropomyosin was donated by F. Whitby and G. N. Phillips (Rice University). Zwittergent 3-16 (n-hexadecyl-N,N-dimethyl-3-ammonio-1-propanesulfonate) was purchased from Calbiochem.
Construction of Protein Expression Vectors-Plasmids were carried in JM101 (Stratagene) except where indicated. BP-1 was assembled as follows. Clone III-1 was cut at SalI (from the plasmid multiple cloning site) and AsnI (which is in the region of overlap with clone II-1). Similarly, clone II-1 was excised at the AsnI site overlapping clone III-1 and at PstI in the polylinker region of the vector. The fragments from clone II-1 and III-1 were ligated simultaneously into pEX-2 (29) to create pEX-BP-1, a ␤-galactosidase fusion protein. The 3Ј and 5Ј ends of BP-1 were verified as correct by dideoxy sequencing using specific primers for pEX-2, but protein could not be expressed in significant quantities. The entire BP-1 coding sequence, contained as an EcoRI fragment in pEX-BP1, was ligated into pET-5c (28), which contains an upstream promoter for bacteriophage T7 RNA polymerase. Protein was expressed in the bacterial strain BL21(DE3) following isopropyl-1-thio-␤-D-galactopyranoside-induced synthesis of T7 RNA polymerase (28). BP-1 spans amino acids 663-1581 of the predicted BPAG1 amino acid sequence (18) and includes N-and C-terminal peptides derived from the cloning vector itself (see Table I). BP-1A was created by insertion of the 2.2-kilobase HincII/EcoRI fragment of BP-1 into pET-5c, such that the 186 N-terminal amino acids derived from BPAG1 were deleted. pET-5c was first linearized at BamHI, then blunt-ended by the Klenow enzyme, and finally cut with EcoRI prior to ligation to the HincII/EcoRI fragment of BP-1. The exogenous sequences present at the C-terminal of BP-1A are identical with those in BP-1, so that the proteins differ only at their N termini.
Conditions for Expression of BPAG1-derived Polypeptides-BL21(DE3) transformed with pET vectors was selected in 50 g/ml ampicillin in LB medium on agar at 37°C, and transformants best expressing protein were stored at Ϫ70°C in 50% glycerol as stock. Overnight cultures were diluted 1:50 with 500 ml of fresh LB containing ampicillin and cultured at 37°C. At an A 600 of 0.6 -0.8, isopropyl-1-thio-␤-D-galactopyranoside (0.4 mM) was added, and cells were harvested 2-4 h later. Cell pellets were resuspended in a small volume of 5% sucrose, 25 mM Tris, pH 8.0, 1 mM DTT, and 1 mM EDTA and lysed in a French press (16,000 p.s.i.) (SLM Instruments, Inc., Urbana). For extraction of BP-1, 0.2 M NaCl, 1% deoxycholic acid, 1% Nonidet P-40, 20 mM Tris, pH 7.6, 2 mM EDTA (30) was added to the lysate which was then centrifuged at 5,000 ϫ g for 10 min at 4°C. The pellet was resuspended in 0.5% Triton X-100 and 1 mM EDTA and washed again. This pellet is thought to contain primarily bacterial inclusion bodies (30) and was dissolved in 8 M urea, 30 mM Tris-HCl, pH 7.6, and 1 mM DTT and clarified by centrifugation. BP-1A was present in soluble form in a French Press lysate, prepared as above following centrifugation at 200,000 ϫ g for 30 min at 4°C.
Purification of BP-1-The BP-1 extract was run on DEAE-cellulose in the presence of 8 M urea, 30 mM Tris-HCl, pH 7.6, and 1 mM DTT, and eluted at ϳ50 mM NaCl in a 0 -0.5 M NaCl gradient in the same buffer. The partially purified protein was renatured by dialysis into 20 mM Hepes-Na, pH 7.0, and 1 mM DTT in a microdialyzer (Life Technologies, Inc.) for 1-2 h at room temperature and shifted to 4°C overnight. Dialyzed protein was centrifuged for 30 min at 200,000 ϫ g in a Beckman TL-100 ultracentrifuge to remove aggregated material. resolution of BP-1A from contaminants was greatly improved by slowing the rate of elution, as would be expected for a highly asymmetric protein. The purest material was obtained by pooling the earliest fractions of the BP-1A peak. In some cases, the peak fractions from S-500 were loaded on a hydroxylapatite column equilibrated with the starting buffer 0.1 mM phosphate, pH 7.4, 0.5 mM DTT, and 0.2 mM EDTA. After washing, BP-1A was eluted with a 0.1-0.4 M phosphate gradient in 0.5 mM DTT and 0.2 mM EDTA. BP-1A was further concentrated by stirring ultrafiltration (Amicon) if necessary for circular dichroism measurements.
Physical Studies-Sucrose gradient centrifugation was performed according to Martin (34), and erythrocyte spectrin dimer (123 Å) (35). Spectrin dimer was obtained from human red blood cell ghost spectrin tetramer as described (36): red blood cell ghosts were dialyzed overnight at 4°C in 0.1 mM EDTA, 0.01 mM DTT at a pH of 8 -9.5 adjusted with diluted NH 4 OH. After dialysis, the ghosts were centrifuged at 200,000 ϫ g for 1 h at 4°C. The supernatant was incubated at 37°C for 15 min, spun again on an Eppendorf centrifuge at top speed, and stored at Ϫ20°C prior to gel filtration. The elution positions for the marker proteins were determined on Coomassie Blue-stained SDS-PAGE gels according to distinctive subunit molecular masses: thyroglobulin (670 kDa), which gives rise to a family of 10-to 330-kDa polypeptides upon reduction for SDS-PAGE, fibrinogen (340 kDa) with 68 kDa, 55-kDa and 46-kDa subunit polypeptides, and spectrin dimer, comprised of a 220-and a 240-kDa polypeptide. BP-1 and BP-1A were identified by a rabbit polyclonal antibody raised to BP-1 in this laboratory.
The Stokes radii of the standards were plotted against erf Ϫ1 (1 Ϫ K d ), which gives a linear plot with standard proteins (37). erf Ϫ1 is the inverse error function, and K d is a partition coefficient or measure of relative elution volume. 1 Ϫ K d is defined as: (38), where the elution volume of the protein is V e , the void volume (8 ml on Superose 6) is V o , and the bed volume (24 ml on Superose 6) is V H 2 O .
Estimated protein molecular weights were calculated as: where all values are under standard conditions (20°C in water): (viscosity) ϭ 1.004; N ϭ Avogadro's number; ϭ protein partial specific volume (estimated 0.725 cm 3 /g); ϭ density of water (0.998 g/cm 3 ). The ratio of actual frictional coefficient (f) to the coefficient for a spherical protein of the same molecular weight (f 0 ) can be estimated as follows (34): Rotary Shadowing-Partially purified BP-1 or BP-1A, about 20 -30 g/ml, was dialyzed at 4°C against 0.155 M ammonium acetate, pH 7.5 overnight, as described (36,39). After dialysis, glycerol was added to a 50% final concentration. A fine mist from the mixture, sprayed by an airbrush (Paasche Type H), was sampled on freshly cleaved mica and shadowed under vacuum with platinum at an angle of about 6°C in a Balzer's Freeze Fracture Apparatus equipped with a rotating stage. Carbon-coated replicas were examined under an Hitachi H-500 electron microscope. Photographic negatives (ϫ 30,000 -60,000) were projected onto a table with a measuring cursor linked to a digitizer. Corrections were made at each magnification according to measurements of a standard diffraction grating. To further validate the method, rabbit tropomyosin was analyzed in parallel in several experiments.
Determination of Protein Concentration and Analysis of Circular Dichroism-Samples for CD were dialyzed three times in 20 mM sodium phosphate, pH 7.4, and 0.2 mM DTT, and centrifuged at 200,000 ϫ g for 30 min at 4°C. We observed that nondenatured BP-1A which had been subjected to repetitive freezing and thawing showed substantial absorbance above 310 nm (data not shown) which could be ascribed to light scattering as a result of protein aggregation (40). Thus, CD was per-formed only on protein samples not previously frozen. The concentration of BP-1A was determined following the addition of 6 M guanidine HCl (41), based on a predicted amino acid composition of 17 tyrosines and 4 tryptophans/molecule. Values for protein concentration were also checked by amino acid analysis (Institute of Biosciences and Technology, Texas A & M University). A quartz cell (Hellma, Mulheim, FRG) of 1-mm path length was used for CD, and the spectrum was recorded on a J-600 Jasco Spectropolarimeter (Japan Spectroscopic Co. LTD). Spectra represent the average of four accumulations at 0.2 nm/step.
Secondary Structure Analysis-Coiled-coil predictions for BPAG1, DPI, and plectin were obtained by the method of Lupas et al. (42). Protein sequences were retrieved from GenBank or EMBL.

RESULTS
Protein Purification-Two polypeptides encompassing the predicted structural transitions from the N-terminal homology domain to the central ␣-helical coiled-coil rod portion of BPAG1 were expressed in E. coli (see Table I). One of these polypeptides, BP-1, could be renatured in a largely soluble form with a small number of lower molecular weight impurities (Fig. 1A) after extraction in 8 M urea from bacterial inclusion bodies (lane a), purification on DEAE-cellulose (lane b), and dialysis from urea (lane c). The migration of BP-1 on SDS gels is consistent with a predicted molecular mass of 111 kDa (Fig. 1A,  lane c). A second purified polypeptide, BP-1A, was constructed by an N-terminal deletion of 186 amino acids from BP-1 (see Table I). In contrast to BP-1, BP-1A was expressed in the cytosol of lysed E. coli. BP-1A isolated following ion exchange and gel filtration appeared to be Ͼ95% pure (Fig. 1B, lane c), and this was confirmed by laser scanning densitometry of Coomassie Blue-stained gels. A high molecular weight contaminant, which represents Ͻ5% of total protein (Fig. 1B, lane c), could not be removed by ammonium sulfate precipitation or hydroxylapatite chromatography. Other polypeptides from this region of BPAG1 were expressed at high levels in E. coli, but none could be satisfactorily purified and characterized (55).
Further purification of BP-1 and BP-1A by heat treatment was attempted since many water-soluble coiled-coil proteins, such as tropomyosin and light meromyosin, renature spontaneously when cooled following heating to 95°C (43,44). Interestingly, both BP-1 and BP-1A precipitated out of solution when heat-denatured (Fig. 2). Horowitz and co-workers (45) have shown that a detergent, such as Zwittergent 3-16, when near its critical micelle concentration, can facilitate native protein folding during rapid renaturation by binding to exposed hydrophobic surfaces that otherwise induce aggregation. In the presence of 0.1 mg/ml Zwittergent 3-16, BP-1 and BP-1A renature primarily to soluble forms (Fig. 2) following heat treatment. While overall protein purity is not improved by this procedure, the detergent requirement for renaturation of BP-1 and BP-1A following heat treatment is consistent with the presence of complex globular folding domains in addition to the predicted ␣-helical coiled-coil.
Physical Characterization-BP-1A comigrates with bovine serum albumin (s 20,w 0 ϭ 4.4, mass ϭ 68 kDa) on velocity sucrose gradients as a single homogeneous peak (Fig. 3). BP-1 has a slightly higher estimated s 20,w 0 of 4.85 than BP-1A. Traces of the two proteins were found at the bottom of both of the sucrose gradients, but at such low levels (Ͻ5%) to suggest that aggregation was not a major factor influencing their behavior. The two proteins also eluted before much higher molecular weight globular standards on Superose 6 gel filtration, suggesting that BP-1 and BP-1A have highly asymmetric or elongated structures. The Stokes radii of BP-1 and BP-1A are greater than both thyroglobulin and fibrinogen but smaller than the spec-  ) GILEDERAS-COOH a Range of amino acid residues incorporated according to the predicted BP230 primary amino acid sequence (18). b Based on resistance to extraction by nonionic detergent and deoxycholic acid. trin dimer (Fig. 4) and were estimated to be 115-121 Å and 111 Å, respectively. Solution molecular weights of BP-1 and BP-1A were inferred from the measured sedimentation coefficients and Stokes radii as 235 kDa and 195 kDa. Both proteins thus appear to associate in solution as dimers, consistent with formation of ␣-helical coiled-coil. In addition, each is predicted to have an axial ratio of Ͼ20:1 (Table II). Gel filtration of highly asymmetric proteins such as these on Sepharose 4B has been reported to underestimate actual Stokes radii when globular standards are used for calibration (46). However, the elution of the asymmetric protein standards fibrinogen and dimeric human erythrocyte spectrin correlated well with the three globular protein standards on Superose 6 ( Fig. 4), and, therefore, the estimates of Stokes radius for BP-1 and BP-1A on Superose 6 appear to be valid.
Rotary Shadowing-To evaluate both the existence and the extent of the predicted ␣-helical coiled-coil domain in BPAG1, we performed rotary shadowing on both BP-1 and BP-1A. Rotary-shadowed images of BP-1 and BP-1A were rod-like with an apparent globular structure at one end; a sample field of BP-1A is shown in Fig. 5a. The knob-like or globular end is more readily visible on BP-1 (Fig. 5, b and c) than on BP-1A (Fig. 5, d, e, and f) as expected since BP-1 is larger and contains a greater proportion of the N-terminal homology region. The lengths of the rod-like regions of both proteins were measured from images having a clearly defined knob at one end and are distributed in a Gaussian fashion: 60 Ϯ 9 nm (N ϭ 185) for BP-1 and 55 Ϯ 8 nm (N ϭ 159) for BP-1A (Fig. 6). The length of rabbit tropomyosin was measured to be 44 Ϯ 4 nm (N ϭ 111), which gives a predicted length that is within the range determined recently by x-ray diffraction (47) and confirms the accuracy of our technique. The knob-like regions were variable in size, however, due perhaps to variability in spreading of the proteins, and their size could not be quantitated directly. A prominent kink was also observed in the tail region of both proteins (Fig. 5, c and f). The length from the end of the tail to the kink in BP-1A was measured to be 26 Ϯ 5 nm (n ϭ 61).
Circular Dichroism-The content of ␣-helical secondary structure in BP-1A was directly estimated from its CD spectrum in several preparations. As shown in Fig. 7, a prominent positive peak at 193 nm, a pair of peaks of negative ellipticity at 208 and 222 nm, and a 2:1 ratio of [] 193 to [] 222 could be observed, all of which are characteristics of ␣-helices (48). The value of [] 222 was Ϫ40,000, based on estimation of protein concentration in 6 M guanidine HCl, and is consistent with 100% ␣-helix content. According to the fitting method of Yang et al. (48), BP-1A is estimated to contain 90% ␣-helix and 10% random coil.

Location of N-terminal to Rod Boundary in BPAG1-
The BPAG1-derived polypeptides BP-1 and BP-1A were designed to span a junction between the predicted globular N-terminal and the central ␣-helical coiled-coil rod domains of BPAG1. The molecular weights of both polypeptides in solution, based on velocity sedimentation and gel filtration, are consistent with dimer formation. The very high percentage of ␣-helix in BP-1A, as determined by circular dichroism, confirms predictions that the central region of BPAG1 will form an ␣-helical coiled-coil dimeric rod, characteristic of myosin, troposyosin, and other members of the intermediate filament family (16,18,20). The calculated axial ratios of BP-1 and BP-1A in solution are greater than 20 (Table II), and a substantial rod-like extension is observed in both molecules by rotary shadowing electron The Stokes radii of the maker proteins are plotted against the erf Ϫ1 (1 Ϫ K d ), which is related to the elution volume. BP-1 and BP-1A eluted close together between fibrinogen and the spectrin dimer. The Stokes radii for BP-1 (118 Å) and BP-1A (111 Å) were predicted by interpolation. Some experimental variation in BP-1 was observed (arrows) and an average Stokes radius (R S ) is used. The markers were: 1, catalase; 2, ␤-galactosidase; 3, thyroglobulin; 4, fibrinogen; and 5, spectrin dimer. microscopy. Significantly, the measured tail lengths of BP-1 and BP-1A are nearly identical (60 and 55 nm, respectively), and, if forming ␣-helical coiled-coil, are predicted to be 383 Ϯ 57 amino acids in length, based on the average of the measured tail lengths for BP-1 and BP-1A, 57.5 Ϯ 8.5 nm, and a 1.5 Å rise per amino acid for ␣-helical coiled-coils (47). The observed junction of the rod domain with adjacent globular structures in BP-1 and BP-1A corresponds to amino acid residues 1197 Ϯ 57 of the complete BPAG1 molecule. Earlier predictions for the boundary of the coiled-coil, at residue 708 (16,27) or at residue 875 (18), are not supported by these data, and would imply much longer tails as imaged by rotary shadowing. The hypothesis (20) that the coiled-coil rod begins at the end of the Nterminal homology with DPI and plectin, approximately at residue 1145, appears to be correct. The kink observed in the tail region of some molecules of BP-1A was estimated to be 26 Ϯ 5 nm from the end of the rod, corresponding to residues 1408 Ϯ 33. This is in rough agreement with a predicted interruption within the BPAG1 rod domain that occurs from residue 1327 to 1354 (20). Our estimate of the N-terminal rod boundary is also supported by the predictive algorithm of Lupas et al. (42), which is based on a statistical analysis of amino acid distribution in each position of the heptad repeat in known coiled-coil proteins. When BPAG1 is analyzed by this algorithm, residues from 1137 to 1870 form a continuous coiled-coil with a short interruption at residues 1325 to 1360 (Fig. 8). Similar analysis of DPI and plectin (not shown) also aligns the start of rod formation with the end of N-terminal homology, as predicted previously (20). Interestingly, two short regions in the N-terminal homology domain, none more than four heptad repeats in length, show greater than 99% probability of coiled-coil formation, but these are shorter than required for coiled-coil formation in solution and, in the context of other secondary structure motifs, are not likely to assemble as rods (49). Previous estimates that coiled-coil rod extends into the Nterminal homology domain of BPAG1 were based upon the high frequency (ϳ70%) of hydrophobic amino acids in the a and d positions of heptad repeats (a, b, c, d, e, and f) found in this domain as well as higher than average intrachain charge interactions consistent with coiled-coil formation (17,20). These characteristics may instead give rise to non-rod ␣-helix (such as ␣-helical bundles) in the BP-1A head domain instead of coiledcoil rod (20). This conclusion is also consistent with the very high content of ␣-helix in BP-1A as estimated by circular dichroism.
BPAG1 Structure and Function-A model for BPAG1 structure based on our results is shown in Fig. 9. The N-terminal globular domain occupies a position homologous to that predicted for both plectin and DPI. Specific C-terminal repeats also found in plectin and DPI (16,18,20) may mediate IF interactions (21)(22)(23), and these repeats would demarcate a central rod of 107 nm. The measured rod length of purified DPI, 130 nm (25), is consistent a rod bounded by the same N-and C-terminal homology domains as in BPAG1 (20). A rotary shadowing study of purified BPAG1 from bovine tongue reported an overall rod length of more than 140 nm (27). This larger estimate was based on a very small number of images of BPAG1 renatured from 9.5 M urea, while the majority of BPAG1 was in aggregates. Since formation of additional rod in the C-terminal homology domain is very unlikely, the reported dumbbell structures with rod length of 140 nm probably represent partially denatured structures.
Although analysis of soluble BPAG1 polypeptides was a necessity in this study, the approach can be justified on the grounds that BPAG1 is expressed in cultured keratinocytes in a soluble or cytosolic form which may act as a precursor for incorporation into the detergent-resistant plaques found in culture (3,6,50). DPI and DPII are also soluble once purified (25), and they are present largely in soluble form in epithelial cells cultured in low (Ͻ0.05 mM) Ca 2ϩ ; induction of desmosome formation by raising extracellular Ca 2ϩ causes a shift from cytosolic to desmosome-associated forms, suggesting a precursor-product relationship for the soluble and insoluble forms of the desmoplakins as well (51,52). We presume that once BPAG1 has been incorporated into the hemidesmosome, major tertiary folding patterns are unchanged. Ultrastructural and genetic evidence suggest that the BPAG1 C-terminal is associated with the cytoplasmic plate of the hemidesmosome, which is a junction for keratin filament attachment. In BPAG1 knockout mice, not only is the plate absent but keratin attachment to the hemidesmosome is also abrogated (8). The cytoplasmic plate is found at a distance of about 0.1 m from what is the most prominent feature of the hemidesmosome, the plasma membrane-associated dense plaque (2,8,53). Labeling with bullous pemphigoid patient autoantibodies directed to BPAG1 shows predominance of staining at an average of 90 nm from the plasma membrane, and two rabbit antisera to the BPAG1 C-terminal also bind at the approximate location of the keratin attachment plate (6,12,54). Ultrastructural localization of the BPAG1 N-terminal has not been addressed experimentally, although studies in cultured normal human keratinocytes (6) and SCC-12 cells (13) strongly suggest that BPAG1 is exclu-sively intracellular, based on its resistance to degradation by extracellular protease. The central discontinuous rod of BPAG1 may provide a flexible link to another cytoplasmic structure such as the membrane-associated dense plaque, where the N-terminal domain itself could in turn bind to transmembrane hemidesmosome components such as BPAG2 or the integrin ␣ 6 ␤ 4 .
In summary, we have carried out high resolution mapping of the N-terminal homology region of BPAG1, empirically localizing a boundary between the N-terminal globular and central rod domain. Comparison of two polypeptides differing only in an N-terminal deletion has enabled us to specify the absolute orientation of the rotary shadowing images of these molecules unambiguously. These data on BPAG1 domain structure should facilitate more rational design of polypeptides for in vitro analysis and cell transfection studies of N-terminal globular domain function in all members of the BPAG1/DPI/plectin protein family.