Collagen XVII Is Destabilized by a Glycine Substitution Mutation in the Cell Adhesion Domain Col15*

Collagen XVII is a hemidesmosomal transmembrane molecule important for epithelial adhesion in the skin. It exists in two forms, as a full-length protein and as a soluble ectodomain that is shed from the keratinocyte surface by furin-mediated proteolysis. To obtain information on the conformation and the functions of this unusual collagen, its largest collagenous domain, Col15, was expressed in a eukaryotic episomal expression system and purified by DEAE and fast protein liquid- Mono S chromatography. The protein was triple-helical ( T m of 26.5 °C) when produced in cultures containing ascorbic acid. When the vitamin supply was limited, the 4-hy-droxyproline content was reduced from 74 to 9%, which, in turn, resulted in a drastic reduction of the stability of the triple helix. The glycine substitution mutation G627V associated with junctional epidermolysis bullosa, a human blistering skin disease, also had a striking effect on thermal stability of rCol15 causing partial unfolding already at 4 °C. Col15 promoted cell adhesion of epithelial and fibroblastic cell lines with a b 1 integrin-mediated mechanism. In concert with this, in acquired autoimmune blistering skin

Collagens are a family of closely related, although genetically distinct, extracellular matrix proteins. Each collagen consists of three polypeptide chains, ␣ chains, which contain a characteristic repeating "collagenous" triplet amino acid sequence -Gly-Xaa-Yaa-, where Xaa and Yaa denote amino acids other than glycine. In all collagen types, the ␣ chains also have non-collagenous domains of varying sizes (1). Typically, the three polypeptide chains are twisted around each other into a collagen triple helix; however, only suggestive data for this exist in the recently characterized transmembrane collagens, types XIII and XVII (for review, see Ref. 2). Collagen XVII, also known as the 180-kDa bullous pemphigoid antigen, or BP180, is a structural component of hemidesmosomes, multiprotein complexes that mediate the adhesion of epidermal keratinocytes to the underlying basement membrane (3,4). The cDNA sequence codes for a type II integral transmembrane protein of 1497 amino acids, with an intracellular domain of 560 amino acids, a short transmembrane stretch, and an extracellular collagenous domain of 914 amino acids with multiple interruptions. The length of the individual collagenous subdomains varies from 14 to 242 amino acid residues (5). Recently, it was established that collagen XVII exists as two molecular forms, i.e. as a full-length transmembrane homotrimer of three 180-kDa ␣1(XVII) chains and as a 120-kDa soluble form that corresponds to the extracellular domain and is presumably released from the cell surface through furin-mediated proteolytic processing (6,7). In some situations also a shorter, approximately 90 -100-kDa fragment, has been observed (3,4,6). Very little is known about the molecular shape of collagen XVII under physiological conditions. Rotary shadowing electron microscopy of collagen XVII from bovine cells lines or of recombinant fragments expressed in COS-1 cells revealed asymmetric molecules with an elongated shape (8,9). A 90-kDa pepsin/ trypsin fragment of collagen XVII in detergent extracts of keratinocytes was resistant to further trypsin digestion at physiological temperatures, suggesting that it was triple-helical (6).
Based on its localization at the basolateral surface of basal keratinocytes, collagen XVII is likely to link the cytoskeleton to the extracellular matrix. Indeed, its intracellular domain interacts with the ␤4 integrin subunit and is essential for its incorporation into hemidesmosomes (10 -13). The ␣6 integrin subunit interacts with the extracellular NC16a domain adjacent to the transmembrane stretch (14). The role of collagen XVII in maintaining adhesion is supported by studies on pathological skin conditions. Mutations in the collagen XVII gene, COL17A1, lead to junctional epidermolysis bullosa, a hereditary blistering skin disease with epidermal detachment from the basement membrane (15)(16)(17)(18)(19)(20)(21). Tissue-bound and circulating IgG and IgA autoantibodies to collagen XVII are also associated with skin blistering in autoimmune bullous skin diseases (22)(23)(24)(25)(26)(27)(28). Nevertheless, the conformation, the functions, and most ligands of the collagenous ectodomain have remained elusive.
Low abundance of collagen XVII in tissues has impeded purification of sufficient amounts of native collagen for functional studies. Therefore, we produced the largest collagenous domain, Col15, as a eukaryotic recombinant fragment. Under physiological conditions, it contains adequate amounts of 4-hydroxyproline and forms a stable triple helix which, however, is destabilized by lack of 4-hydroxyproline or by substitution of a particular glycine residue. Further, we demonstrate that the recombinant domain can promote cell adhesion and is the target of autoantibodies in patients with blistering autoimmune diseases.

Production of a Bacterial Fusion Protein and Domain-specific Anti-
bodies-For production of domain-specific antibodies against collagen XVII, a bacterial fusion protein, GST 1 -Col15-2 (see Fig. 1A), spanning amino acids 774 -807 (5,17), was generated using the GST Gene Fusion System (Amersham Pharmacia Biotech, Uppsala, Sweden). The corresponding human cDNA (GenBank accession number M911669) was amplified from keratinocyte mRNA by reverse transcriptase PCR (Titan reverse transcriptase PCR, Roche Molecular Biochemicals, Mannheim, Germany) with the sense primer 5Ј-GCGCGGATCCGACCCAG-GAAAGCCAGGT -3Ј (nucleotides 2425-2442) and the antisense primer 5Ј-GCGCGGAATTCACTTGCCTGGAGCTCC Х3 (nucleotides 2512-2526). Following digestions with BamHI and EcoRI (restriction sites underlined) the fragment was cloned into the 3Ј-end of the GST gene in the expression vector pGEX-2T (Amersham Pharmacia Biotech), and dideoxynucleotide sequence analysis was performed. The construct was expressed in Escherichia coli DH5␣, and the fusion protein was purified using the GST-glutathione affinity system (Amersham Pharmacia Biotech). For production of the antibody Ab-Col15-2, rabbits were immunized using standard conditions (Eurogentec, Ougrée, Belgium).
Construction and Transfection of Wild-type and Mutated Eukaryotic Expression Vectors-A cDNA fragment corresponding to the Col15 domain of human collagen XVII cDNA was amplified by Titan reverse transcriptase PCR of keratinocyte mRNA according to the manufacturer's instructions. The sense primer was 5Ј-GCGCGCTAGCAGGAAGC-CCTGGCCCTAAA-3Ј (nucleotides 1804 -1821) and the antisense primer 5Ј-ATTAGCGGCCGCTCACTTGCCTGGAGCTCC-3Ј (nucleotides 2512-2526), the underlined sequences representing the NheI and NotI restriction sites, respectively. After NheI and NotI digestions, the fragment was cloned into a modified episomal expression vector pCEP-Pu containing the signal peptide sequence of BM-40 and a puromycin resistance gene (29). The correct sequence of the clone pCEP-Col15 was verified by dideoxynucleotide sequencing. It corresponded to amino acids 567-807 of collagen XVII.
Human kidney 293-EBNA cells constitutively expressing the EBNA-1 protein of Epstein-Barr virus to enhance transfection efficiency were used according to the manufacturer's instructions (Invitrogen, Groningen, The Netherlands). The cells were grown in Dulbecco's modified Eagle's medium and nutrient mix F-12 medium (Life Technologies, Inc.) containing 10% fetal calf serum and 0.35 mg/ml G418 (Invitrogen). One million cells/10-cm culture dish were transfected with 25 g of pCEP-Col15 or pCEP-Col15-G627V DNA using the calcium phosphate method. Following a selection with 0.5 g/ml puromycin (Sigma, Deisenhofen, Germany), the transfected cells were grown to confluence, washed twice with phosphate-buffered saline, and switched to serum-free medium containing 50 g/ml ascorbic acid (Fluka, Deisenhofen, Germany). Freshly made ascorbic acid was added every 24 h or, in some experiments, only every 48 h, or omitted totally. The media were collected every 48 h, cooled, centrifuged to remove cellular debris, and supplemented with 1 mM Pefablock (Merck, Darmstadt, Germany) and 1 mM N-ethylmaleimide (Sigma).
Purification and Amino Acid Analysis of Recombinant Col15 Domain-Medium containing the recombinant Col15 domain, called rCol15 hereafter, was dialyzed against 0.05 M Tris-HCl, pH 8.6, at 4°C and chromatographed on a DEAE-cellulose column (Whatman, Maidstone, U. K.) equilibrated in the same buffer. rCol15 did not bind to DEAE-cellulose, but a significant amount of contaminating proteins was removed with this step. For further purification, rCol15 fractions were dialyzed against 0.02 M citrate, pH 5, at room temperature and passed onto an FPLC Mono S column (Amersham Pharmacia Biotech) equilibrated in the same buffer and subsequently eluted with a linear 0 -0.5 M NaCl gradient.
For amino acid analysis, Mono S-purified rCol15 was blotted to polyvinylidene difluoride membrane (Problott, Perkin-Elmer Applied Biosystems) and hydrolyzed under vapor-phase condition using 6 N HCl for 75 min at 150°C in a nitrogen atmosphere. The amino acid composition was analyzed with the 421-amino acid analyzer (Perkin-Elmer) according to the manufacturer's instructions, with a slight modification for optimal collagen sample analysis. Proteolytic Digestions of rCol15 Domain-Digestion of DEAE-purified rCol15 with highly purified bacterial collagenase (Advanced Biofactures Inc., Lynbrook, NY) was performed with 40 units/ml enzyme in 0.05 M Tris-HCl, 0.01 M CaCl 2 , pH 7.7, at 37°C for 2 h. For testing the triple-helical conformation with trypsin as a probe (30), 10 g/ml DEAE-purified rCol15 or rCol15-G627V was treated with 10 g/ml trypsin (Boehringer Ingelheim) at increasing temperatures between 4 and 40°C for 2 min, and the reactions were stopped by adding 10 g/ml soybean trypsin inhibitor (Sigma). For immunoblotting with Ab-Col15-2 using standard techniques, SDS-polyacrylamide gel electrophoresis with 6 -22% polyacrylamide gradient gels was used. The scanning of the immunoblot signals and quantitation of digestion products were performed with Image Master VDS for gel electrophoresis and Image Master software (Amersham Pharmacia Biotech).
Circular Dichroism Analysis of rCol15 Domain-Purified rCol15 was dissolved in 0.02 mM sodium phosphate, 0.05 M NaCl, pH 7.4, at 0.15 mg/ml. Circular dichroism spectra at 4°C and 50°C were recorded in a DC6 Jobin Yvon spectropolarimeter equipped with a quartz cell (0.1-mm path length) thermostatted with a water jacket. Melting curves were measured by monitoring ellipticities at 221 nm as a function of temperature in the water jacket of the CD cell. The temperature was raised linearly at a rate of 12°C/h in a programmable water bath. The degree of helicity, F(T), of the peptide was calculated according to where T is the ellipticity value measured at temperature T. n , the ellipticity value of the triple-helical polypeptide, does not depend on T and was measured at T ϭ 4°C. d,T is the ellipticity value for the denatured polypeptide and was linearly dependent on T. d,T at temperatures below the melting temperature T m was determined by linear extrapolation from T values above T m . Cell Adhesion Assays-Skin epithelial cells (A431, HaCaT) and lung fibroblasts (L132, Wi26) were grown as described (6,31). Multiwell tissue culture plates (96 wells, Costar Corp., Faust, Germany) were coated with serial dilutions of collagen IV (5 g/ml, kindly provided by Dr. Klaus Kü hn, Max Planck Institute for Biochemistry, Martinsried, Germany) or of DEAE-purified rCol15 (100 g/ml) overnight at 4°C. For some experiments, rCol15 was denatured at 56°C for 20 min before coating. After saturation with 1% bovine serum albumin (fraction V, Sigma), the plates were used immediately for short term (30,45, and 60 min) adhesion assays (31). For cell adhesion inhibition assays, the monoclonal antibody P4C10 to human ␤1 integrin (Life Technologies; diluted 1:500 and 1:2000) was added to the cells before plating. Adherent cells were photographed using an Axiovert phase contrast microscope (Zeiss, Oberkochen, Germany). All assays were done as triplicates.
Detection and Immunoadsorption of Circulating Autoantibodies with rCol15-For testing IgG and IgA reactivity with rCol15, bullous pemphigoid and linear IgA disease patient sera were used, respectively. The diagnoses were based on clinical findings, characteristic histology, and demonstration of linear autoantibody binding at the skin basement membrane zone with immunofluorescence techniques (32). For immunoblotting with rCol15 as antigen, the sera (diluted 1:20) were incubated overnight, and alkaline phosphatase-labeled chain-specific anti-human IgG and IgA (Sigma) second antibodies were incubated for 1.5 h.
For affinity adsorption of patient autoantibodies, a nitrocellulose strip containing rCol15 was prepared as described (33). 50 l of serum diluted 1:10 in blotting buffer was reacted with the strip overnight, and the antibodies were eluted in 200 l of 0.1 M glycine HCl buffer, pH 2.5. After neutralization, the immunoglobulins were used for immunofluorescence staining of intact or sodium chloride split (34) human skin with fluorescein isothiocyanate-labeled anti-human IgG or IgA (Dako) second antibodies. Controls included sera from healthy volunteers or persons with unrelated skin diseases.

RESULTS
Expression and Purification of the rCol15 Domain of Collagen XVII-The Col15 domain of collagen XVII consists of 242 amino acids (residues 567-808 of collagen XVII) and is thus remarkably larger than any of the 14 other collagenous domains, which vary from 14 to 45 amino acids in length (Fig.  1A). Its calculated molecular mass deduced from the cDNA sequence is 2245 kDa, and its predicted amino acid sequence is characterized by 80 perfect Gly-Xaa-Yaa triplets with one Gly-Xaa-Gly (Gly-Ser-Gly) interruption in positions 639 -641.
A cDNA fragment of 723 base pairs corresponding to Col15 was generated with reverse transcriptase PCR from human keratinocyte mRNA and cloned into a modified episomal expression vector pCEP-Pu (29). The resulting clone, pCEP-Col15, was transfected into human embryonic kidney 293-EBNA cells. Compared with nontransfected controls, the transfected cells secreted approximately 10 g/ml of an additional polypeptide with an apparent molecular mass of 34 kDa (Fig. 1B). This recombinant fragment, designated rCol15, was identified in immunoblots with the antibody Ab-Col15-2 (Fig.  1C) against a carboxyl-terminal segment of the Col15 domain. In a first purification step by DEAE-cellulose chromatography, rCol15 did not bind but was partially purified by elimination of

Col15 Domain of Collagen XVII
other medium proteins (Fig. 1B). Further purification was achieved using FPLC Mono S chromatography and elution with a linear NaCl gradient (Fig. 1B). rCol15 was sensitive to highly purified bacterial collagenase (Fig. 1D), confirming its collagenous nature.
4-Hydroxyproline Content of rCol15 Domain-To determine the extent of prolyl hydroxylation and to analyze the effect of 4-Hyp content on triple helix stability, rCol15 was purified and subjected to amino acid analysis. When produced in the presence of 50 g/ml ascorbic acid (added to the media every 24 h), rCol15 contained 20 4-Hyp residues/polypeptide chain. The maximum number of Pro residues in the Yaa position of the Gly-Xaa-Yaa triplets which serve as substrate for prolyl-4hydroxylase (35) is 27. Thus, 74% of the eligible prolyl residues were hydroxylated. Because the total number of proline residues in rCol15 is 53, the 4-Hyp:Pro ϩ 4-Hyp ratio was 37.7%. When ascorbic acid was added only every 48 h, the number of 4-Hyp residues/polypeptide chain was only 2.5.
Conformation and Thermal Stability of rCol15 Domain-The triple-helical conformation and the thermal stability of rCol15 were assessed by CD spectra. At 4°C, purified fully hydroxylated rCol15 produced a spectrum characteristic for triple-helical collagens ( Fig. 2A). The extreme values of mean residue ellipticity (4,700 at 221 nm and Ϫ52,500 deg ϫ cm 2 ϫ dmol Ϫ1 at 197 nm) were comparable to those produced by triple-helical collagen I (36). At 50°C, rCol15 had a CD spectrum of randomly coiled polypeptides ( Fig. 2A). The thermal stability was monitored by measuring the signal at 221 nm as a function of temperature. The transition curve of rCol15 revealed a melting temperature of 26.5°C (Fig. 2B).
Limited trypsin digestion (30) was also used to probe the triple-helical structure of rCol15. The digestions were performed by reacting 10 g/ml rCol15 with 10 g/ml trypsin for 2 min at increasing temperatures between 4 and 40°C. With this assay, fully hydroxylated rCol15 lost its stability to about 50% at 25°C (Fig. 3A) and rCol15 produced without ascorbic acid at a significantly lower temperature of 18°C (Fig. 3B).
Mutated Col15 Domain Is Inadequately Folded-Site-directed mutagenesis was used to generate a Gly to Val substitution at amino acid residue 627 within the amino-terminal third of the Col15 domain. This gene defect has been reported in a patient with junctional epidermolysis bullosa who was compound heterozygous for a nonsense mutation on the other allele and, therefore, functionally homozygous (hemizygous) for the glycine substitution (16). The clone containing the mutation pCEP-Col15-G627V was expressed in 293-EBNA cells, and the corresponding recombinant fragment, rCol15-G627V, was partially purified with DEAE-cellulose chromatography. Thermal stability of rCol15-G627V was probed by trypsin digestion at increasing temperatures between 4 and 30°C. The 34-kDa rCol15-G627V fragment was partially degraded into an intermediate digestion product of about 26 kDa already at 4°C, and the 34-kDa band disappeared entirely at 16°C, indicating complete unfolding (Fig. 3C). The intermediate 26-kDa product resisted trypsin between 4 and 16°C but gradually lost this resistance between 18 and 26°C; its melting temperature was about 21°C (Fig. 3C). These experiments demonstrated that the glycine substitution significantly reduced the conformational stability of the Col15 domain.
The major 26-kDa intermediate product was detectable with the antibody Ab-Col15-2, which recognizes the 34 most COOHterminal amino acids of Col15 (Fig. 1A), indicating that this product contained the COOH terminus of Col15 domain. Therefore, it is likely that the stretch of 60 amino acids located amino-terminally from the G627V substitution was not folded correctly and thus sensitive to trypsin digestion. The formation of an intermediate digestion product of 26 kDa with a melting temperature only slightly lower than that of nonmutated rCol15 suggests that the region of Col15 located carboxyl-terminally of the mutation was folded correctly.
Col15 Domain Promotes Adhesion of Epithelial Cells and Fibroblasts-To obtain more information on the structurefunction relationships of collagen XVII, rCol15 was tested for its ability to induce adhesion of HaCaT, A431, L132, and Wi26 cells. rCol15 induced cell adhesion in a concentration-dependent manner (Fig. 4A). Interestingly, the cell adhesion-promoting activity of denatured rCol15 was higher than of native rCol15. In both cases, the maximal adhesion plateaus were lower than those observed with collagen IV, which was used as a positive control (Fig. 4B). Maximum cell adhesion to rCol15 was obtained at 50 -100 g/ml, the corresponding value for collagen IV being 2-5 g/ml. Spreading of L132 fibroblasts and of Wi26 epithelial cells was induced by rCol15, but to a lesser extent than by collagen IV (Fig. 4C). HaCaT cells (Fig. 4C) do not usually display a distinct spread morphology, and there was no major difference between these cells adhering to rCol15 or collagen IV. However, on native rCol15, HaCaT cells formed an alignment like "necklace of pearls" (Fig. 4C). Cell adhesion inhibition experiments with function-blocking anti-integrin antibodies indicated that integrins of the ␤1 family mediated cellular interactions with rCol15 (Fig. 5).
Col15 Domain Is Targeted by Autoantibodies in Bullous Diseases-To investigate whether Col15 contained epitopes relevant for human autoimmune diseases, rCol15 was used as antigen in immunoblots with sera from patients with bullous pemphigoid or linear IgA disease. In both diseases, circulating and tissue-bound autoantibodies to collagen XVII have been reported (23-28). Sixty-six patient sera that contained IgG or IgA reactive with collagen XVII and its soluble ectodomain (32) FIG. 3. Trypsin digestions of rCol15. DEAE-purified rCol15 was treated with 10 g/ml trypsin for 2 min at increasing temperatures between 4 and 40°C, and the reactions were stopped by adding soybean trypsin inhibitor. Scanning and quantitation of the immunoblot signals obtained with the antibody Ab-Col15-2 showed that the stability of rCol15 was reduced by omission of ascorbate or by substitution of the Gly residue at position 627. Panel A, rCol15 produced in the presence of 50 g/ml ascorbic acid lost its stability to about 50% at 25°C. Panel B, rCol15 produced without ascorbic acid lost its stability to about 50% at 18°C. Panel C, rCol15-G627V was produced by converting the codon for glycine at amino acid position 627 to a codon for valine. Compared with the 34-kDa undigested sample (U), trypsin treatment already at the lowest temperature, 4°C, resulted in an appearance of intermediate digestion product of 26 kDa. The 34-kDa rCol15-G627V band disappeared at 16°C, and the 26-kDa intermediate fragment was lost to about 50% at 21°C.
were assessed for reactivity with rCol15. Of these, 23 (35%) contained either IgG or IgA recognizing rCol15 (Fig. 6A). As controls, 60 normal human sera or sera of patients with unrelated skin diseases were used. None of these showed IgG or IgA reactivity with rCol15. To evaluate the pathophysiological relevance of the autoantibodies reactive with rCol15 on immunoblots, IgG from pemphigoid sera was affinity-purified with rCol15 bound to nitrocellulose. After elution from the nitrocellulose matrix, the immunoglobulins were used for indirect immunofluorescence staining of normal human skin. IgG from bullous pemphigoid sera stained the basement membrane zone of normal human skin (Fig. 6B), indicating that the circulating autoantibodies that recognized rCol15 in immunoblots reacted with native collagen XVII in situ.

DISCUSSION
The primary structure of collagen XVII deduced from the cDNA sequence suggests that its extracellular domain may form a collagen triple helix with multiple interruptions. Because the authentic protein is not available in amounts allowing conformational analysis by CD spectroscopy, the largest collagenous domain, Col15, was produced as a eukaryotic recombinant fragment that was fully triple-helical. Of pivotal importance for this tertiary structure was an adequate supply FIG. 4. Cell adhesion to rCol15. A, freshly suspended L132 fibroblasts (left panel) or HaCaT cells (right panel) were seeded on wells coated with different concentrations of native or heat-denatured rCol15. The extent of cell adhesion was determined by a colorimetric assay (31). B, the maximal adhesion to native or heat-denatured rCol15 recorded in the dose-response curves shown in panel A were compared with that observed on collagen IV. The histograms show adhesion of Wi26, A431, L132, and HaCaT cells. C, morphology of L132 fibroblasts (panels a-c) and HaCaT epithelial cells (panels d-f) attached to collagen IV (panels a and d), native rCol15 (panels b and e), and to denatured rCol15 (panels c and f). Photographs were taken at the end of the 45min adhesion period, after washing off unattached cells. Attached cells were fixed and stained as described (31). of ascorbic acid in the cell culture media. This vitamin is an essential cofactor for the hydroxylation of prolyl residues to 4-Hyp, which in turn is needed for stabilization of the collagen triple helix (35). This is an important aspect because in previous studies addressing the molecular shape and conformation of collagen XVII, a protein produced without ascorbic acid was analyzed (7,8,37).
The midpoint of the helix-to-coil transition, T m , of 26.5°C for the rCol15 was relatively low. This was not the result of underhydroxylation because the 4-Hyp content of rCol15 was 83/ 1000 amino acid residues, which agrees well with that of other collagens (54 -126/1000 residues; Ref. 34). The reduced thermal stability thus is a consequence either of the shorter length of rCol15 or of its amino acid sequence that allows the formation of a less stable triple helix (38,39). In the full-length protein, the triple helix of Col15 is more stable because it is presumably stabilized by cooperative interactions with neighboring triplehelical domains (6).
Unlike many other glycine substitution mutations in the genes for skin basement membrane collagens (20,21,40), the mutation G627V causes a drastic reduction of the conformational stability of the Col15 domain. The mutated fragment was partially digested by trypsin already at 4°C and completely abolished at 16°C. A carboxyl-terminal intermediate digestion product had a T m of about 21°C. These findings suggest that the G627V substitution changes the folding and decreases the thermal stability of collagen XVII. In concert with this, G627V has been shown to be associated with junctional epidermolysis bullosa in a patient who was functionally homozygous for the glycine substitution (16). In the skin of the proband, the extracellular domain of collagen XVII was absent, whereas some intracellular collagen XVII epitopes were discerned with immunofluorescence staining. This staining pattern, together with our finding that mutated rCol15-G627V was easily degraded in vitro, suggests that the mutated ectodomain of collagen XVII in the skin of the proband was not folded correctly and was consequently sensitive to degradation. Thus, the absence of functional Col15 domain in the junctional epidermolysis bullosa patient's skin led to deficient epidermal adhesion and to the phenotype of skin fragility.
In concert with the genetic evidence for the role of collagen XVII in dermal-epidermal integrity, rCol15 promoted cell adhesion in vitro in a ␤1 integrin-dependent manner. The native triple-helical rod of fibrillar collagens I, II, III, and V, basement membrane collagen IV, and of microfibrillar collagen VI, has been found to induce cell adhesion (for review, see Ref. 41), and this may represent a general property of collagenous helices. Even though many collagens upon denaturation show diminished cell adhesion activity (42)(43)(44)(45), heat-denatured rCol15 bound cells efficiently. In some similar cases, cell adhesion to denatured collagens involves RGD motifs in the primary sequence of the ␣ chains which become exposed upon unfolding of the triple-helices (46,47). Instead of an RGD, rCol15 contains a GER sequence that is recognized by integrins in collagen I (48). Induction of cell adhesion by denatured rCol15 is also interesting in the context of shedding of the ectodomain, which may generate new cell binding sites. However, the exact nature of the cell adhesion and integrin binding sites in rCol15 remains to be determined.
The functional significance of Col15 for epidermal adhesion is underscored further by the fact that in acquired blistering skin disorders autoantibodies target this domain. Using rCol15 as antigen, we showed that approximately one-third of the tested autoimmune sera contained either IgG or IgA reactive with it. This is in contrast to previous studies with mainly bacterial recombinant fragments, which found most immunodominant epitopes in the NC16a domain (25,49,50), some in the distal amino-and carboxyl-terminal regions of the molecule, but none in the Col15 domain (23,28,51). By affinity adsorbing the rCol15-reactive immunoglobulins from the patient sera, we also showed that the circulating autoantibodies against rCol15 target collagen XVII at the epidermal basement membrane zone in situ. This suggests that tissue-bound autoantibodies to collagen XVII perturb Col15 functions by interfering with its cell adhesion activity and thereby contribute to blister formation.
In conclusion, recombinant expression of human transmembrane collagen XVII in eukaryotic cells opens new possibilities for assessing the structure-function relationships of this unusual collagen. Furthermore, the recombinant expression system should be useful in elucidating the effects of human mutations or autoantibody binding on the stability and ligand binding of collagen XVII. FIG. 5. Inhibition of cell adhesion to rCol15 by a ␤1 integrin antibody. Cells were seeded on substrate-coated wells in the presence or absence of a monoclonal antibody to ␤1 integrin (P4C10) diluted 1:500 or 1:2,000 as indicated. The extent of cell adhesion was measured after 45 min by a colorimetric assay (31).
FIG. 6. IgG and IgA autoantibodies in bullous skin diseases target rCol15. Panel A, immunoblotting of rCol15 with patient autoimmune sera. As a positive control, the antibody Ab-Col15-2 was used (lane C). rCol15 was recognized by IgG autoantibodies from a pemphigoid serum (lane 1) and by IgA autoantibodies from a linear IgA dermatosis serum (lane 2). Panel B, autoantibodies from a pemphigoid serum immunoadsorbed with rCol15 were used for indirect immunofluorescence staining of normal human skin. The affinity-purified IgG to rCol15 stained the dermo-epidermal basement membrane zone in a linear fashion (arrow).
concerning the 293-EBNA expression system; Dr. Klaus Kü hn for collagen IV and Dr. Norbert Fusenig for the HaCaT cells. We also acknowledge the excellent technical assistance of Margit Schubert, Michaela Floeth, Monika Pesch, and Anja Mattila and the expert help of Mika Kihlström with the illustrations.