Pemphigus Foliaceus in Vitro and in Vivo of Patients and Prevents Pathomechanisms of Autoantibodies to Desmoglein 1 in a Majorit

The Journal of Immunology

The Plant Lectin Wheat Germ Agglutinin Inhibits the Binding of Pemphigus Foliaceus Autoantibodies to Desmoglein 1 in a Majority of Patients and Prevents Pathomechanisms of Pemphigus Foliaceus In Vitro and In Vivo

Susana Ortiz-Urda,1*‡ Adelheid Elbe-Bu¨rger,† Josef Smolle,§ Yvonne Marquart,* Yakov Chudnovsky,‡ Todd W. Ridky,‡ Pamela Bernstein,‡ Klaus Wolff,* and Klemens Rappersberger*¶

Pemphigus foliaceus (PF) is a life-threatening autoimmune blistering skin disease caused by pathogenic IgG autoantibodies against desmoglein 1 (dg1), a desmosomal cadherin-type adhesion glycoprotein. Using lectins and glycosidases, we have shown that dg1 displays an N-glycosylation pattern of the complex triantennary type. We have found that lectins and glycosidases interfere with N-bound sugar residues on the amino-terminal ectodomain of dg1 and completely abolish, in vitro, the antigenicity of dg1 in most of the patients’ sera. Moreover, in an ex vivo model using punch biopsies from normal human skin, we demonstrate that preincubation of the epidermis in wheat germ agglutinin (WGA) prevents PF autoantibody binding, acantholysis, and subcorneal blistering. In addition, we show that topical treatment with WGA inhibits PF autoantibody binding to keratinocytes in both newborn BALB/c mice and in organotypic human epidermis grafted onto the back of SCID mice. The epidermis of these pre- treated animals displays a regular morphology, whereas control animals develop the immunopathologic phenotype of PF. These ﬁndings suggest that WGA may interfere with autoantibody binding to dg1, preventing experimental PF without affecting the adhesive function of dg1. Our observations may provide a new approach to the therapy of PF. The Journal of Immunology, 2003, 171: 6244–6250.

emphigus foliaceus (PF)2 is a disabling autoimmune bul- motif, localized on the outermost amino-terminal ectodomain, E I lous disease characterized by pathogenic autoantibodies (11). Substantial experimental evidence suggests that the expres- P directed against the extracellular domain of desmoglein 1 sion and adhesive function of dg1 also depends on the presence of (dg1) (1, 2). Binding of these autoantibodies to dg1 is followed by calcium ions and protein kinase C activity (14–16). Moreover, the loss of cell-to-cell adhesion, acantholysis, and blister formation proper formation of desmosomes, as well as regular epidermal (3). dg1 is a 160-kDa desmosomal cadherin-type cell adhesion stratification and pemphigus Ag expression, depends on N-aspar- glycoprotein organized into four extracellular domains (E I to E aginyl glycosylation (15–17). IV), which are connected by extra- and intracellular anchoring The molecular mechanisms by which PF autoantibody binding domains to four intracellular domains (4–7). Within the extracel- to dg1 impairs cell adhesion and causes blistering, are enigmatic. lular portion of desmosomes, dg1 establishes cell-cell adhesion by The claim that proteases induce blister formation has recently been homotypic and heterotypic protein binding with desmocollins, refuted (18), and it has been shown that neither complement acti- other molecules of the family of desmosomal cadherins (8–10). vation nor surface cross-linking of dg1 by PF autoantibodies is Pemphigus sera recognize conformationally sensitive epitopes in required for induction of acantholysis (19, 20). There still exists the amino-terminal region of dg1 (11–13). Previous investigations the concept that PF autoantibody binding to dg1 transduces signals have shown that homotypic protein binding depends on the pres- via Ca2ϩ and inositol triphosphate, resulting in a loss of cell-ad- ence of an arginine (arg/R)-alanine (ala/A)-leucine (leu/L) (RAL) hesive properties of dg1 (21, 22). The RAL-homotypic protein-binding motif is localized within Divisions of *General Dermatology and †Immunodermatology, Department of Der- the most antigenic epitopes of dg1 (11), and PF autoantibody bind- matology, University of Vienna, Vienna, Austria; ‡Department of Dermatology, Stan- ing may directly affect the cell-adhesive function of dg1. The RAL ford University School of Medicine, Stanford, CA 94305; §Division of Analytic Mor- phologic Dermatology, Department of Dermatology, University of Graz, Graz, sequence, represented by residues 79–81, is in close proximity to Austria; and ¶Department of Dermatology, Hospital Rudolfstiftung, Vienna, Austria two calcium-binding motifs, residues 95–104 and 129–138. Also, Received for publication October 31, 2002. Accepted for publication September two N-glycosylation sites on positions 61 and 131 flank the RAL 26, 2003. motif. These molecular features suggested that Ca2ϩ and sugar The costs of publication of this article were defrayed in part by the payment of page residues (6, 12, 23) could play a role for the antigenicity of dg1, charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. most likely by determining the confirmation of the extracellular domain. 1 Address correspondence and reprint requests to Dr. Susana Ortiz-Urda, Program in Epithelial Biology, Stanford University School of Medicine, Center for Clinical Sci- In this study, we first defined the composition of sugar residues ences and Research Building, Room 2145, Stanford, CA 94305. E-mail address: of dg1. Wheat germ agglutinin (WGA), a plant lectin with a high [email protected] affinity for N-glycosylated sugar residues that also interferes with 2 Abbreviations used in this paper: PF, pemphigus foliaceus; dg1, desmoglein 1; WGA, wheat germ agglutinin; KGM, keratinocyte growth medium; IF, immunoflu- the extracellular calcium concentration (24, 25), displayed high orescence. affinity for dg1. Subsequently, we demonstrated that targeting of

Copyright © 2003 by The American Association of Immunologists, Inc. 0022-1767/03/$02.00 The Journal of Immunology 6245 carbohydrate moieties with WGA could abolish the binding of PF cubated with either PF sera or anti-dg1 mAb, and assayed for single and autoantibodies to dg1, an effect that was also observed after deg- double IF microscopy. lycosylation of dg1 in vitro. Using the BALB/c mouse (19, 26) and Ex vivo model and animal model SCID/human-xenograft models (27), we then provide evidence For ex vivo studies, 4-mm punch biopsies from normal human skin were that topically applied WGA penetrates through the epidermis, incubated with either PF sera or with normal human serum in KGM for 3 binds to the surface of keratinocytes, and inhibits PF autoantibody days in a tissue incubator and then snap frozen in liquid nitrogen. binding, preventing acantholysis and blister formation. Animal model To investigate whether lectins topically applied to the skin may also inhibit Materials and Methods PF autoantibody binding in vivo, we took advantage of the pemphigus Synthesis of reagents mouse model (19, 26). In the first step, we partially removed the cornified cell layer of 96 Tissues. Normal human skin from patients that underwent breast reduc- BALB/c mice with 3% urea in Vaseline and occlufol (Tegaderm, Borken, tion was provided by plastic surgeons. Keratomed strips of skin and 4-mm Germany) for 6 h. Subsequently, 36 newborn BALB/c mice were treated punch biopsies were used for biochemical studies or immediately incu- with WGA and 12 with Con A at concentrations ranging from 0.5 to 2.5 bated in keratinocyte growth medium (KGM) (Clonetics, San Diego, CA) mg/ml in propylene glycol under occlusive plastic dressing. As a control, and further processed for in vitro, ex vivo, and in vivo studies. 27 animals were treated with vehicle alone, and 21 animals were treated Sera. Sera were obtained from 12 PF patients. Selection criteria were as with U. europeaus 1. Twelve hours later, the animals were injected i.p. follows: clinical symptoms characteristic of PF, presence of acantholysis with PF sera of 11 different PF patients. We injected a total of 63 animals within the granular cell layer upon histological examination, positive direct with PF sera and 33 animals with normal human sera through a 30-gauge and indirect immunofluorescence (IF), and exclusive reaction with dg1 by needle in doses of 12 mg of protein per gram of body weight per day. The immunoblotting, immunoprecipitation, and ELISA. ELISA index of posi- injections were given twice daily on 3 consecutive days. Animals were tive PF patients’ sera was between 140 and 200 (28). For control, we used examined daily and photodocumented and sacrificed 12 h after the last sera from healthy human volunteers. injection. Skin samples were taken and further processed for immunomor- Antibodies. Mouse mAb to dg1 (DG 3.10 and Dsg1 P 23) were purchased phological studies. At the time of the biopsy, sera were obtained and as- from Progen (Heidelberg, Germany). Rat Abs specific for human 6 integrin sayed by indirect IF microscopy for titers of circulating human IgG. (MAB1972) were purchased from Chemicon International (Temecula, Cell cultures and organotypic skin CA). As second-step reagents, we used affinity-purified FITC-labeled goat anti-human-IgG, FITC-labeled rabbit anti-mouse-IgG, tetramethylrhodam- Keratinocytes and fibroblasts from sterile skin samples were cultured sep- ine isothiocyanate-labeled goat anti-human IgG (all from Axell Laborato- arately under regular cell culture conditions until near confluence. Bovine ries, Westbury, NY), FITC-labeled goat anti-rat IgG, and FITC-labeled collagen gels were polymerized, supplemented with confluent human der- rabbit anti-human IgG (both from Chemicon International). For immuno- mal fibroblasts, and grown over a period of 7 days. Subsequently, a ker- blotting, we used alkaline-phosphatase-conjugated goat anti-human and atinocyte suspension was seeded onto the gel surface and further cultured goat anti-mouse IgG as detector Abs (Promega, Madison, WI). for 2–4 days. Then, the collagen gel and keratinocytes were transferred Lectins. Different lectins were used as FITC- or digoxigenin-labeled or onto the membrane of Transwell plates (Costar, Cambridge, MA) and fur- ϩ unlabeled reagents: Con A type 4 (Sigma-Aldrich, St. Louis, MO), WGA, ther cultured in KGM supplemented with 1.5 mmol of Ca2 . Complete soybean agglutinin, peanut agglutinin, Sophora japonica agglutinin, Ban- growth of a mature, multilayered, and differentiated epithelium was deiraea simplicifolia 2, Ulex europeaus agglutinin 1, Maackia amurensis, achieved within 24–96 h. Then, the tissue was harvested and further pro- Datura stramonium, and Aleuria aurantia (all from Vector Laboratories, cessed (31). Burlingame, CA). For the detection of tissue-bound unlabeled lectins we used FITC-conjugated rabbit Igs to WGA, Con A, U. europeaus agglutinin SCID/human-xenograft model 1, and soybean agglutinin (Dakopatts, Copenhagen, Denmark). Inhibiting Six-week-old SCID mice were obtained from a colony maintained in our sugars used were as follows: 200 mM ␣-methyl mannoside/200 mM laboratory. Human skin equivalents were grafted on the backs of 12 mice, ␣-methyl glucoside mixture for Con A; and 500 mM N-acetylglucosamine, secured with sutures, and dressed with gauze. Three weeks later, bandages chitin hydrolysate, or 100 mM acetic acid for WGA. were removed. First, we partially removed the cornified cell layer with 3% Glycosidases. To hydrolyze glycoconjugates, we used the following panel urea in Vaseline and occlufol for 6 h. Subsequently, 9 mice were treated of glycosidases: N-glycosidase F, endoglycosidase-F2/F1, peptide-N4-(N- overnight with WGA, and 3 were treated with U. europeaus 1 under similar acetyl-␤-D-glucosaminyl) asparagine amidase F of Flavobacterium menin- conditions. Mice were then injected with PF sera dissolved in 300 ␮lof gosepticum, recombinant from Escherichia coli; O-glycosidase from Dip- PBS. Injections were repeated after 24 h. Mice were examined every 12 h loccocus pneumoniae; endoglycosidase-H from Streptomyces plicatus; and killed 48 h after the first injection. Skin samples were taken and further ␣-mannosidase from jack beans (Canavalia ensiformis); ␤-galactosidase processed for immunomorphological studies. At the time of the biopsy, from bovine testes (all from Boehringer Mannheim, Mannheim, Germany). sera were obtained and assayed by indirect IF microscopy for titers of circulating human IgG. Light- and IF microscopy, immunoblotting, and immunoprecipitation of epidermal protein extracts Results Dg1 displays a N-glycosylation pattern of the triantennary These studies were performed according to routine methods (29). complex type Deglycosylation of protein extracts and structural In the first set of experiments, we analyzed the carbohydrate struc- characterization of sugar residues ture of dg1. Digestion of dg1 with N-glycosidase F, which hydro- Protein extracts were deglycosylated according to routine methods (30). The lyzes all types of N-glycan chains, reduced the molecular mass by carbohydrate structures of blotted glycoproteins were analyzed with digoxige- 10–15 kDa (Fig. 1a, lanes 4 and 5). In contrast, treatment of dg1 nin-labeled lectins and an alkaline-phosphatase-labeled anti-digoxigenin Ab. with endoglycosidase F1/F2, which is selective for N-glycosyla- Lectin binding studies tion of biantennary complex- and high mannose-type sugars, or For the detection of sugar residues, cryosections of normal human skin with endoglycosidase H, which is selective for N-glycosylation of were incubated for 20 min at 20°C with FITC-conjugated and unconju- high mannose-type sugars, did not affect the molecular mass of the gated lectins, diluted to a concentration of 0.1 mg/ml in 0.1 M PBS (pH protein (Fig. 1b, lanes 3 and 4). Deglycosylation of O-linked sugar 6.8). Sections labeled with unconjugated lectins were subsequently incu- moieties resulted in a slight reduction of the molecular mass of dg1 bated with PF sera as well as with the two different anti-dg1 mAb. (Fig. 1b, lane 5). The degree of deglycosylation was monitored Deglycosylation of tissue sections with digoxigenin-labeled lectins (Fig. 1a, lane 9). Five-micrometer cryosections were incubated with 0.4 U of glycosidase in To specify the carbohydrate moieties of dg1, we then precipi- PBSfor2hat20°C. After thoroughly rinsing in PBS, sections were in- tated protein lysates with the mAb Dsg1 P 23 and subsequently 6246 WGA PREVENTS PF PATHOMECHANISMS

In certain patients with PF, the autoantigenicity of dg1 is determined by carbohydrate moieties, in in vitro and ex vivo organ cultures In immunoprecipitation experiments, we confirmed that PF autoantibodies react with a protein that migrates with dg1 at 160 kDa, as detected by labeling with mAb DG 3.10 directed to intracellular epitopes of dg1 and PF sera (Fig. 1, lanes 2 and 3). After deglycosylation with N-glycosidase F, dg1 migrated to a molecular mass of 140–145 kDa as shown by labeling with both mAb, Dsg1 P 23, and DG 3.10 (Fig. 1, lanes 4 and 5). However, labeling of the deglycosylated protein with PF sera was reduced (Fig. 1, lane 6)or FIGURE 1. Immunoprecipitation of epidermal protein lysates with negative (lanes 7 and 8). Next, we investigated the sera of 12 PF mAb to dg1 before and after deglycosylation with N-, F1/F2-, H-, and O-glycosidases. a, In consecutive immunoblotting experiments, dg1 in its patients and showed that 8 sera (including the two patients shown native form appears at a molecular mass of 160 kDa, as indicated by la- in Figs. 1a, lanes 7 and 8, and 2b, lanes 3 and 4) did not react with beling with mAb Dsg1 P 23 and with PF serum (lanes 2 and 3, arrow). deglycosylated dg1, whereas 4 sera labeled deglycosylated protein After deglycosylation, dg1 appears at a reduced molecular mass of 140– (Fig. 2b, lanes 10–13). One of the sera reacted only weakly with 145 kDa, as detected by labeling with mAb DG 3.10 and Dsg1 P 23 (lanes the deglycosylated protein (Fig. 1a, lane 6). 4 and 5). The deglycosylated protein is recognized only weakly or not at all These findings were consistent with deglycosylation experi- by different PF sera (lanes 6–8) or WGA (lane 9). Molecular mass markers ments using N-glycosidase F on cryostat sections of normal human are shown in lane 1. b, The molecular mass of dg1 was not affected by skin. Whereas binding of mAb Dsg1 P23 was not affected, deg- digestion with endoglycosidase F1/F2 or endoglycosidase H (lanes 3 and lycosylation completely abolished the consecutive binding of those 4). After O-deglycosylation, there is a slight reduction in the molecular PF sera that did not react with deglycosylated dg1 in immunopre- mass of dg1 (lane 5). Lane 2, Labeling of dg1 with mAb Dsg1 P 23 is indicated. cipitation experiments (Fig. 3, a and b). Identical results were achieved when skin sections were preincubated with WGA (2.5 mg/ml sodium chloride). Immunobinding of mAb Dsg1 P23 was stained the electrophoretically separated protein with various not affected; however, the binding of the seven PF sera was com- digoxigenin-labeled lectins. dg1 clearly reacts with M. amurensis, pletely abolished (Fig. 3, c and d). In contrast, the binding prop- WGA, and Con A, indicating a glycosylation pattern that consists erties of the four PF sera that were reactive with deglycosylated of sialic acid terminally linked ␣(2,3) to galactose, several mole- dg1 in immunoprecipitation, was not affected and showed typical cules of N-acetyl glucosamine and ␣-mannose D (ϩ) residues (Fig. PF staining (data not shown). In control experiments, preincuba- 2a, lanes 3–5). There was no reaction with D. stramonium, A. tion of lectins with their specific sugars abolished their ability to aurantia, and U. europeaus agglutinin, indicating absence of ter- block PF sera binding on cryostat sections. Identical results were minally linked galactose ␤(1–4)-N-acetylglucosamine without achieved when cryostat sections were incubated with lectins with- sialic acids and fucose (Fig. 2, lane 7). However, there was a weak out affinity to dg1, such as soybean agglutinin (Fig. 3, e and f). reactivity with peanut agglutinin, indicating an O-glycosylation Preincubation of PF autoantibodies with WGA did not affect composed of galactose ␤(1–3) linked to N-acetylgalactosamine subsequent binding of PF autoantibodies to the surface of keratin- (Fig. 2, lane 6). Taken together, these findings suggest that dg1 ocytes of cryostat sections from normal human skin (data not displays a N-glycosylation pattern of the complex triantennary shown). In the next set of experiments, we used the PF ex vivo type and a certain degree of O-bound sugar residues. organ culture model in which 4-mm punch biopsies of normal human skin were incubated with PF sera. H&E-stained sections revealed subcorneal split with acantholytic keratinocytes within the blister fluid. IF studies revealed PF autoantibodies bound to the cell membranes of keratinocytes (Fig. 4, a and b). Strikingly different results were obtained when these experiments were performed after overnight preincubation of organ cultures in WGA. We found that WGA strongly labeled the surface of the keratinocytes in a pemphigus-like pattern, but we did not observe binding of PF autoantibodies from sera that were unreactive with deglycosylated dg1 (Fig. 4, c and d). In H&E-stained sections, acantholysis and intraepidermal split formation were absent (Fig. 4f). However, staining with the four PF sera that did react with deglycosylated dg1 in vitro (see above) as well as with mAb Dsg1 P 23 was unaffected by WGA treatment of epidermis (Fig. 4e). In- cubation in normal human sera did not induce any immunopatho- logical changes. FIGURE 2. Immunoprecipitation of epidermal protein lysates with dg1 and subsequent labeling with digoxigenin-labeled lectins. a, dg1 (arrow) WGA, topically applied, prevents pathobiologic effects of certain displays reactivity with M. amurensis (lane 3), WGA (lane 4), Con A (lane PF sera in vivo, in a mouse model (Table I) 5), and peanut agglutinin (lane 6). dg1 does not react with Ulex 1(lane 7). Labeling with mAb Dsg1 P 23 is indicated (lane 2). The molecular mass Twelve hours after the last injection of PF sera, all animals that standards are shown (lane 1). a and b, After deglycosylation of dg1 with were topically treated either with vehicle alone or U. europeaus 1, N-glycosidase F, seven sera (a, lanes 7 and 8, and b, lanes 3 and 4) did not a lectin without affinity to dg1, developed a PF-like disease, with react with deglycosylated dg1, whereas four sera labeled deglycosylated generalized erythema and blisters (Fig. 5a, right mouse). Exami- protein (b, lanes 10–13). nation of cryostat sections of the skin of these animals showed The Journal of Immunology 6247

FIGURE 3. a and b, Deglycosylation of cryostat sections of normal human skin with N-glycosidase F did not affect binding of mAb Dsg1 P23 (a) but completely abolished the subsequent binding of glycosylation-dependent PF sera (b). c and d, Also, preincubation of cryostat sections of normal human skin with WGA does not affect labeling with mAb Dsg1 P 23 (c) but completely prevents labeling of keratinocytes with PF autoantibodies (d). e and f, Preincubation of cryostat sections of normal human skin with soybean agglutinin, a lectin without afﬁnity for dg1 that binds to the cell surface of keratinocytes, does not affect the binding of mAb Dsg1 P 23 (e) or PF autoantibodies (f).

PF-like autoantibody binding to the surface of keratinocytes and Table I). Sera obtained from the mice, as assayed by indirect IF subcorneal blisters with numerous acantholytic keratinocytes (Fig. microscopy, contained circulating human IgG (data not shown). 5, b and c). Also, two of six animals treated with Con A displayed a PF-like clinical and immunomorphological phenotype. One an- WGA prevents pathologic effects of certain PF sera in a SCID/ imal died early after the first injection and was excluded from the human-xenograft disease model study. However, clinical and immunomorphological examination Human epidermal equivalents were grafted onto the backs of of the remaining three animals did not show any pathological fea- SCID mice. One portion of the animals was topically treated with tures reminiscent of PF. Moreover, 19 of 21 WGA-treated mice WGA and the other group with U. europeaus 1. Twelve to 18 h that were injected with sera that did not react with deglycosylated after the last injection, all animals injected with glycosylation- dg1 displayed no signs of a PF-like disorder. The morphology of independent PF sera, as well as all animals that were topically the epidermis appeared completely normal in H&E staining (Fig. treated with U. europeaus 1, developed PF-like clinicopathological 5d). IF microscopy revealed WGA staining throughout all layers of features (Fig. 6, a and b). In contrast, seven of nine WGA-treated the epidermis, whereas no in vivo binding of human IgG was de- mice that were injected with sera unreactive with deglycosylated tected (Fig. 5, e and f). Two animals in this group died; investi- dg1 were unaffected (Fig. 6, c and d). WGA labeled the upper gation of their skin did not display any conclusive findings. In layers of the epidermis (Fig. 6e). The human origin of the grafts contrast, all four animals injected with glycosylation-independent was verified with human-specific mAb to ␣ integrin (Fig. 6f). sera developed PF-like clinicopathological features indistinguish- 6 able from those shown in Fig. 5, a–c. The skin of mice injected with normal human sera did not dis- Discussion play any pathological changes regardless of whether they were PF is a chronic disabling bullous autoimmune disease mediated by treated with either WGA, Con A, or other lectin (for details, see Abs that impair adhesion of keratinocytes. Therefore, treatment of

FIGURE 4. Ex vivo model for PF. a, Incubation of 4-mm punch biopsies from normal human skin in PF serum induces subcorneal blisters with acantholytic keratinocytes (H&E). b, The cell membranes of keratinocytes that form the blister roof and of acantholytic cells are labeled with PF autoantibodies. c, Preincuba- tion of these punch biopsies in WGA overnight before incubation in PF sera shows that WGA labels the surface of all keratinocytes in a pemphigus-like pattern. d and e, There is essentially no labeling with PF autoantibodies (d), but staining with mAb Dsg1 P 23 is not affected (e). f, Histologically, these skin sections display a normal morphology. 6248 WGA PREVENTS PF PATHOMECHANISMS

Table I. In vivo testing of lectins in a mouse model

Ulex europeaus- Total No. Con A-Treated Mice WGA-Treated Mice Vehicle-Treated Mice Treated Mice of Animals

PF sera nonreactive with 6 sera tested 7 sera tested 7 sera tested 7 sera tested deglycosylated dg1 (7) 6 animals 21 animals 14 animals 14 animals 55

3 animals remained healthy. 19 animals remained healthy. All developed disease. All developed disease. 2 animals developed 2 died. disease. 1 died within 24 h. PF sera reactive with 4 sera tested 4 sera tested deglycosylated dg1 (4) 0 4 animals 4 animals 0 8

All developed disease. All developed disease. Normal human sera 6 animals remained healthy. 11 animals remained healthy. 9 animals remained 7 animals remained 33 healthy. healthy. Total no. of animals 12 36 27 21 96

PF is performed with various aggressive immunosuppressive reg- Using epidermal protein lysates from normal human skin, we imens, with their associated deleterious side effects. In this study, ﬁrst demonstrated that dg1 displays a triantennary complex type we asked whether the autoantigen of PF, dg1, might be addressed glycosylation pattern that predominantly consists of galactose, as a possible therapeutic target. mannose, N-acetylglucosamine, and, on the outermost position, Most PF autoantibody binding sequences are localized between several molecules of sialic acid. We then examined by immuno- dg1 residues 50 and 219. As shown by Sekiguchi et al. (13), re- precipitation and immunoblotting the binding properties of PF au- moval of autoantibodies against the 161 N-terminal residues of toantibodies to native and deglycosylated dg1, in vitro, and dg1 by immunoadsorption eliminates the ability of PF sera to in- showed that N-deglycosylation of protein lysates reduced or induce cutaneous blisters in neonatal mice. This portion of dg1 in- hibited the consecutive binding of 75%. However, the binding of cludes two potential N-glycosylation sites on positions 61 and 131, four other serum probes to dg1 was clearly shown to be indepen- two calcium-binding motifs, represented by residues 95–104 and dent of the presence of carbohydrate moieties. 129–138, and the RAL sequence (aa 79–81). These peculiar bio- These observations were further conﬁrmed in several in vitro, ex chemical features, and the fact that PF autoantibody binding sites vivo, and in vivo experiments. We showed that either deglycosy- are conformational epitopes, led us to speculate whether targeting lation of the epidermis or labeling of epidermal carbohydrates with of N-linked carbohydrate moieties in the antigenic epitopes of dg1 WGA abolished the subsequent autoantibody binding of 7 of 12 PF may affect PF autoantibody binding, in a manner similar to that sera with associated prevention of PF pathology. In contrast, the shown with pemphigus vulgaris (32). pathobiologic effects mediated by 4 glycosylation-independent PF

FIGURE 5. Experimental PF induced in neonatal BALB/c mice. a, After the injection of PF sera, the animals develop generalized erythema and blisters (right). In contrast, animals that have been topically treated with WGA before the injection of PF sera have a normal appear- ance (left). Arrows indicate blisters. b and c, Un- treated animals show PF autoantibodies bound in vivo in a pemphigus-like pattern (b) and display histopathological features of PF (c). d, WGA-treated animals show a completely normal morphology of the epidermis. e and f, The keratinocytes are labeled with WGA (e) but do not display binding of PF autoantibodies (f). The Journal of Immunology 6249

FIGURE 6. Experimental PF induced in organotypic human epidermal equivalents grafted onto SCID mice. a, After three injections of PF sera, a subcorneal blister has developed. b, Direct IF displays a typical pemphigus-type labeling pattern. c, In contrast, organotypic skin grafts topically treated with WGA before the injection of PF sera do not display any morphological features of PF. d, PF autoantibodies are not detected in these skin sections. e and f, WGA labeling of the surface of keratinocytes is shown in e, and the human origin of these skin sections is verified with human-specific mAb to ␣ 6 integrin (f). sera remained unaffected by the topical application of WGA. Iden- bohydrate moieties. To address this issue, it would be of interest to tical PF pathology was seen in experiments with BALB/c mice clarify to which dg1 epitopes the various Ca2ϩ- and carbohydrate- treated with U. europeaus agglutinin 1, a lectin without affinity to dependent or -independent PF autoantibodies bind, and whether dg1, where all animals developed a PF-like pathology. Taken to- there is any association with disease severity. gether, these findings strongly indicate that WGA has a protective In addition, there exists the possibility that WGA binding to dg1 function against the pathobiologic effects of certain PF autoanti- indirectly exerts its effects by interference with Ca2ϩ homeostasis bodies, in vitro and in vivo. (24, 25). However, this hypothesis seems rather unlikely, because This phenomenon can be explained by the peculiar localization we added Ca2ϩ in our in vitro experiments. Moreover, we never of N-linked sugar residues flanking the RAL motif in close prox- observed morphological changes in keratinocytes during in vivo 2ϩ imity to the Ca binding sites. Bound lectins may spatially block experiments with WGA over a period of 3 days. If topically ap- the consecutive targeting of antigenic epitopes by PF autoantibod- plied WGA exerts influences on Ca2ϩ homeostasis in the epider- ies. Similar effects have been demonstrated in leukemic mast cells, mis, one would expect certain changes in keratinocyte differenti- where WGA binding to the Fc⑀ receptor RI inhibits the subsequent ation and morphology, at least at the light-microscopic level (35). binding of a specific monoclonal reagent (33). Alternatively, there In addition, because it is known that both heterophilic and to a also exists the possibility that lectin binding alters the conforma- lesser degree homophilic desmoglein protein binding is Ca2ϩ sen- tion of dg1 with reduction in its antigenicity. This hypothesis is sitive, disturbance of Ca2ϩ concentrations in the murine and hu- supported by experiments showing that the PF autoantigen is conformation dependent (11, 12). However, Amagai et al. (12) have man epidermis by topically applied WGA would influence the ad- demonstrated that Ca2ϩ exclusively, but not carbohydrates, may hesive properties of dg1 (10, 36). play the decisive role in the conformation of dg1 as PF autoanti- Taken together, our observations in a human autoimmune skin gen. This obvious contradiction to our observations might be ex- disease suggest that it is possible to prevent autoantibody binding plained by the heterogeneity of certain PF sera in their binding to to a defined autoantigen by the topical application of WGA and selected epitopes, as shown in previous investigations (11) and thus interrupt pathobiologic mechanisms. demonstrated in this study. Moreover, a recent study has shown a disease-dependent intramolecular epitope spreading phenomenon in the course of endemic PF, a South American variant of the Acknowledgments ϩ disease (34). Our findings suggest that, in addition to Ca2 -depen- We are grateful to P. A. Khavari for his valuable help preparing this dent epitopes of dg1, certain dg1 epitopes are determined by car- manuscript. 6250 WGA PREVENTS PF PATHOMECHANISMS

References 20. Espana, A., L. A. Diaz, J. M. Mascaro, Jr., G. J. Giudice, J. A. Fairley, G. O. Till, and Z. Liu. 1997. Mechanisms of acantholysis in pemphigus foliaceus. Clin. 1. Amagai, M. 2003. Desmoglein as a target in autoimmunity and infection. J. Am. Immunol. Immunopathol. 85:83. Acad. Dermatol. 48:244. 21. Seishima, M., C. Esaki, K. Osada, S. Mori, T. Hashimoto, and Y. Kitajima. 1995. 2. Emery, D. J., L. A. Diaz, J. A. Fairley, A. Lopez, A. F. Taylor, and G. J. Giudice. Pemphigus IgG, but not bullous pemphigoid IgG, causes a transient increase in 1995. Pemphigus foliaceus and pemphigus vulgaris autoantibodies react with the intracellular calcium and inositol 1,4,5-triphosphate in DJM-1 cells, a squamous extracellular domain of desmoglein-1. J. Invest. Dermatol. 104:323. cell carcinoma line. J. Invest. Dermatol. 104:33. 3. Mahoney, M. G., Z. Wang, K. Rothenberger, P. J. Koch, M. Amagai, and J. R. Stanley. 1999. Explanations for the clinical and microscopic localization of 22. Seishima, M., Y. Iwasaki-Bessho, Y. Itoh, Y. Nozawa, M. Amagai, and lesions in pemphigus foliaceus and vulgaris. J. Clin. Invest. 103:461. Y. Kitajima. 1999. Phosphatidylcholine-specific phospholipase C, but not phos- 4. Schmelz, M., R. Duden, P. Cowin, and W. W. Franke. 1986. A constitutive pholipase D, is involved in pemphigus IgG-induced signal transduction. Arch. Dermatol. Res. 291:606. transmembrane glycoprotein of Mr 165,000 (desmoglein) in epidermal and non- epidermal desmosomes. I. Biochemical identification of the polypeptide. Eur. 23. Kapprell, H. P., P. Cowin, W. W. Franke, H. Ponstingl, and H. J. Opferkuch. J. Cell Biol. 42:177. 1985. Biochemical characterization of desmosomal proteins isolated from bovine 5. Koch, P. J., M. J. Walsh, M. Schmelz, M. D. Goldschmidt, R. Zimbelmann, and muzzle epidermis: amino acid and carbohydrate composition. Eur. J. Cell Biol. W. W. Franke. 1990. Identification of desmoglein, a constitutive desmosomal 36:217. glycoprotein, as a member of the cadherin family of cell adhesion molecules. Eur. 24. Nagata, Y., and M. M. Burger. 1974. Wheat germ agglutinin. Molecular char- J. Cell Biol. 53:1. acteristics and specificity for sugar binding. J. Biol. Chem. 249:3116. 6. Nilles, L. A., D. A. Parry, E. E. Powers, B. D. Angst, R. M. Wagner, and 25. Sjolander, A. 1988. Direct effects of wheat germ agglutinin on inositol phosphate K. J. Green. 1991. Structural analysis and expression of human desmoglein: a formation and cytosolic-free calcium level in intestine 407 cells. J. Cell. Physiol. cadherin-like component of the desmosome. J. Cell Sci. 99:809. 134:473. 7. Garrod, D. R., A. J. Merritt, and Z. Nie. 2002. Desmosomal adhesion: structural 26. Roscoe, J. T., L. Diaz, S. A. Sampaio, R. M. Castro, R. S. Labib, Y. Takahashi, basis, molecular mechanism and regulation (review). Mol. Membr. Biol. 19:81. H. Patel, and G. J. Anhalt. 1985. Brazilian pemphigus foliaceus autoantibodies 8. Garrod, D. R., A. J. Merritt, and Z. Nie. 2002. Desmosomal cadherins. Curr. are pathogenic to BALB/c mice by passive transfer. J. Invest. Dermatol. 85:538. Opin. Cell Biol. 14:537. 27. Zillikens, D., E. Schmidt, S. Reimer, I. Chimanovitch, K. Hardt-Weinelt, 9. Marcozzi, C., I. D. Burdett, R. S. Buxton, and A. I. Magee. 1998. Coexpression C. Rose, E. B. Brocker, M. Kock, and W. H. Boehncke. 2001. Antibodies to of both types of desmosomal cadherin and plakoglobin confers strong intercel- desmogleins 1 and 3, but not to BP180, induce blisters in human skin grafted onto lular adhesion. J. Cell Sci. 111:495. SCID mice. J. Pathol. 193:117. 10. Syed, S. E., B. Trinnaman, S. Martin, S. Major, J. Hutchinson, and A. I. Magee. 2002. Molecular interactions between desmosomal cadherins. Biochem. J. 28. Lenz, P., M. Amagai, B. Volc-Platzer, G. Stingl, and R. Kirnbauer. 1999. Des- 362:317. moglein 3-ELISA: a pemphigus vulgaris-specific diagnostic tool. Arch. Derma- 11. Kowalczyk, A. P., J. E. Anderson, J. E. Borgwardt, T. Hashimoto, J. R. Stanley, tol. 135:143. and K. J. Green. 1995. Pemphigus sera recognize conformationally sensitive 29. Rappersberger, K., N. Roos, and J. R. Stanley. 1992. Immunomorphologic and epitopes in the amino-terminal region of desmoglein-1. J. Invest. Dermatol. biochemical identification of the pemphigus foliaceous autoantigen within des- 105:147. mosomes. J. Invest. Dermatol. 99:323. 12. Amagai, M., K. Ishii, T. Hashimoto, S. Gamou, N. Shimizu, and T. Nishikawa. 30. Haselbeck, A., E. Schickaneder, H. von der Eltz, and W. Hosel. 1990. Structural 1995. Conformational epitopes of pemphigus antigens (Dsg1 and Dsg3) are cal- characterization of glycoprotein carbohydrate chains by using digoxigenin-la- cium dependent and glycosylation independent. J. Invest. Dermatol. 105:243. beled lectins on blots. Anal. Biochem. 191:25. 13. Sekiguchi, M., Y. Futei, Y. Fujii, T. Iwasaki, T. Nishikawa, and M. Amagai. 31. Hinterhuber, G., Y. Marquardt, E. Diem, K. Rappersberger, K. Wolff, and 2001. Dominant autoimmune epitopes recognized by pemphigus antibodies map D. Foedinger. 2002. Organotypic keratinocyte coculture using normal human to the N-terminal adhesive region of desmogleins. J. Immunol. 167:5439. serum: an immunomorphological study at light and electron microscopic levels. 14. Denning, M. F., S. G. Guy, S. M. Ellerbroek, S. M. Norvell, A. P. Kowalczyk, Exp. Dermatol. 11:413. and K. J. Green. 1998. The expression of desmoglein isoforms in cultured human 32. Konohana, I., A. Konohana, P. Z. Xia, R. E. Jordon, W. D. Geoghegan, and keratinocytes is regulated by calcium, serum, and protein kinase C. Exp. Cell Res. M. Duvic. 1988. Expression of pemphigus vulgaris antigen in cultured human 239:50. keratinocytes: effect of inhibitors, tunicamycin, and lectins. J. Invest. Dermatol. 15. Pasdar, M., Z. Li, and H. Chan. 1995. Desmosome assembly and disassembly are 90:708. regulated by reversible protein phosphorylation in cultured epithelial cells. Cell 33. McCloskey, M. A. 1993. Immobilization of Fc⑀ receptors by wheat germ agglu- Motil. Cytoskeleton 30:108. tinin: receptor dynamics in IgE-mediated signal transduction. J. Immunol. 16. Penn, E. J., I. D. Burdett, C. Hobson, A. I. Magee, and D. A. Rees. 1987. Struc- 151:3237. ture and assembly of desmosome junctions: biosynthesis and turnover of the major desmosome components of Madin-Darby canine kidney cells in low cal- 34. Li, N., V. Aoki, G. Hans-Filho, E. A. Rivitti, and L. A. Diaz. 2003. The role of cium medium. J. Cell Biol. 105:2327. intramolecular epitope spreading in the pathogenesis of endemic pemphigus fo- 17. Overton, J. 1982. Inhibition of desmosome formation with tunicamycin and with liaceus (fogo selvagem). J. Exp. Med. 197:1501. lectin in corneal cell aggregates. Dev. Biol. 92:66. 35. Komuves, L., Y. Oda, C. L. Tu, W. H. Chang, C. L. Ho-Pao, T. Mauro, and 18. Mahoney, M. G., Z. H. Wang, and J. R. Stanley. 1999. Pemphigus vulgaris and D. D. Bikle. 2002. Epidermal expression of the full-length extracellular calcium- pemphigus foliaceus antibodies are pathogenic in plasminogen activator knock- sensing receptor is required for normal keratinocyte differentiation. J. Cell. out mice. J. Invest. Dermatol. 113:22. Physiol. 192:45. 19. Rock, B., R. S. Labib, and L. A. Diaz. 1990. Monovalent FabЈ immunoglobulin 36. Chitaev, N. A., and S. M. Troyanovsky. 1997. Direct Ca2ϩ-dependent hetero- fragments from endemic pemphigus foliaceus autoantibodies reproduce the hu- philic interaction between desmosomal cadherins, desmoglein and desmocollin, man disease in neonatal BALB/c mice. J. Clin. Invest. 85:296. contributes to cell-cell adhesion. J. Cell Biol. 138:193.