Pemphigus Vulgaris Antibody Identifies Pemphaxin

Because pemphigus vulgaris (PV) IgGs adsorbed on the rDsg3-Ig-His baculoprotein induced blisters in neonatal mice, it was proposed that anti-desmoglein 3 (Dsg 3) autoantibody causes PV. However, we found that rDsg3-Ig-His absorbs autoantibodies to different antigens, including a non-Dsg 3 keratinocyte protein of 130 kDa. This prompted our search for novel targets of PV autoimmunity. The PV IgG eluted from a 75-kDa keratinocyte protein band both stained epidermis in a pemphigus-like pattern and induced acantholysis in keratinocyte monolayers. Screening of a keratinocyte λgt11 cDNA library with this antibody identified clones carrying cDNA inserts encoding a novel molecule exhibiting ∼40% similarity with annexin-2, named pemphaxin (PX). Recombinant PX (rPX-His) was produced inEscherichia coli M15 cells, and, because annexins can act as cholinergic receptors, its conformation was tested in a cholinergic radioligand binding assay. rPX-His specifically bound [3H]acetylcholine, suggesting that PX is one of the keratinocyte cholinergic receptors known to be targeted by disease-causing PV antibodies. Preabsorption of PV sera with rPX-His eliminated acantholytic activity, and eluted antibody immunoprecipitated native PX. This antibody alone did not cause skin blisters in vivo, but its addition to the preabsorbed PV IgG fraction restored acantholytic activity, indicating that acantholysis in PV results from synergistic action of antibodies to different keratinocyte self-antigens, including both acetylcholine receptors and desmosomal cadherins.

skin adhesion in which keratinocytes (KC), the stratified epithelial cells comprising the epidermis, lose their ability to adhere to one another (acantholysis) (1). Acantholysis leads to an intra-epidermal split and separation of the suprabasal epidermal layer, which is clinically manifested by blistering that denudes skin and oral mucosa. Introduction of glucocorticosteroids into the treatment of PV patients decreased mortality from 90 to 10% (reviewed in Ref. 2). Long-term corticosteroid therapy of PV patients is life-saving but causes severe side effects, including death (3,4). This urges development of nonhormonal therapy of pemphigus acantholysis. The pathophysiology of PV includes an array of IgG autoantibodies reacting with keratinocyte self-antigens with the apparent molecular mass ranging from 12 to 190 kDa (reviewed in Ref. 5), including a 130-kDa keratinocyte polypeptide (6,7). The notion that autoantibodies are the main cause of PV stems from the fact that passive transfer of pemphigus, but not normal, IgGs to neonatal mice can induce skin lesions characteristic of PV (8). Using pemphigus antibodies eluted from the 130-kDa band as a probe, Amagai et al. (9) screened the human keratinocyte gt11 cDNA library and found that two of the clones recognized by these PV antibodies represented a novel desmosomal cadherin termed desmoglein (Dsg) 3. The hypothesis that PV, a disease of skin adhesion, is caused by an antibody to Dsg 3, an adhesion molecule, prompted experiments toward elucidation of the biological effects of anti-Dsg 3 antibody. However, acantholysis could not be documented in keratinocyte monolayers treated with anti-Dsg 3 antibody. Several recombinant Dsg 3 (rDsg3) proteins were produced and used to test if adsorbed antibodies can elicit skin blistering in neonatal mice upon passive transfer (10,11). Although rDsg3 could absorb PV antibodies to Dsg 3, it failed to absorb all disease-causing antibody, and PV IgGs depleted of antibodies to Dsg 3 kept binding to KC in murine epidermis and inducing gross skin blisters (10,12). Only creation of a chimeric baculoprotein that included both the extracellular epitope of Dsg 3 and an Fc portion of human IgG 1 could fulfill both goals: elimination of all disease-causing antibodies from pemphigus serum and induction of gross skin blisters in neonatal mice injected with concentrated eluants (13,14). Explanations of this phenomenon include: 1) a possibility that the IgG portion rendered the rDsg3 with appropriate conformational epitope, which could be tested by crystallography; and 2) a possibility that the tertiary structure of the chimera mimicked non-Dsg 3 targets of pemphigus autoimmunity, which could be tested by characterizing the antigenic profile of the eluted IgG. Neither possibility was tested. Recently, it has become evident that anti-Dsg 3 antibody alone is not sufficient to cause skin blisters (15). A role for an autoantibody to another desmosomal cadherin, Dsg1, was proposed to explain skin blisters in PV patients (16). However, well-documented cases of generalized disease in PV patients lacking Dsg1 antibody (17) argued in favor of the existence of a yet unidentified disease-causing non-Dsg1/Dsg 3 antibody that could have been nonspecifically preabsorbed with rDsg3-Ig constructs. Furthermore, intraperitoneal injection of the PV IgG, which did not have anti-Dsg1 activity, into neonatal Dsg3 knockout mice (i.e. Dsg3 null mice) resulted in gross skin blisters (5). It should be mentioned that neonatal Dsg3 null mice lack the true PV phenotype, in that they do not develop spontaneous skin blisters (5,18), which has already justified their use in passive transfer experiments by different research groups studying the nature of disease-causing PV antibodies (5,15).
Recently, we have compared antibodies eluted from rDsg3 (rDsg3-His) and rDsg3-Ig (rDsg3-Ig-His), which were used in the original preabsorption experiments (10,13,14), and demonstrated that the two Dsg 3 constructs adsorb antibodies with different antigenic specificities (19). The PV IgGs eluted from rDsg3-His reacted predominantly with the 130-kDa protein band present in normal human KC in addition to a few weakly stained bands that varied among test PV sera. In marked contrast, the antibodies eluted from rDsg3-Ig-His recognized several different protein bands, including a non-Dsg 3 130-kDa band in the immunoblot of Dsg 3Ϫ/Ϫ keratinocyte proteins. Thus, crossreactivity of Dsg3-Ig-His with non-Dsg 3 antibodies explains how this chimeric baculoprotein could absorb all disease-causing PV IgG.
The vast majority of pemphigus patients develop antibodies that immunoprecipitate keratinocyte membrane proteins binding the covalent cholinergic radioligand [ 3 H]propylbenzilylcholine mustard ([ 3 H]PrBCM) (5) and compete with a cholinergic radioligand, [ 3 H]atropine, for binding to the cell membrane of intact human KC in culture (20). The nature of the acetylcholine (ACh) receptor(s) targeted by PV autoimmunity remains to be determined. Addition to either muscarinic or nicotinic antagonists to keratinocyte monolayers in both cases results in acantholysis (reviewed in Refs. 21,22), whereas cholinergic agonists stimulate cell-to-cell adhesion of KC, and can reverse, attenuate, or prevent acantholysis in keratinocyte monolayers when added to culture after, simultaneously with, or prior to PV IgG, respectively (20). The anti-acantholytic activity of cholinergic agonists suggests a novel avenue for development of non-hormonal treatment of pemphigus.
In this study, we demonstrate the nature of a novel target for non-Dsg 3 disease-causing PV IgG. Screening of the keratinocyte cDNA expression library with PV IgG immunoaffinitypurified on a 75-kDa area of the immunoblotting membrane revealed a novel human annexin-like molecule, which we named pemphaxin (PX). We produced recombinant PX (rPX-His) and demonstrated that this protein acts as a cholinergic receptor in the radioligand binding assay with [ 3 H]ACh. PV IgG specifically recognized rPX-His, and preabsorption of PV sera with rPX-His eliminated the acantholytic activity that could be restored by adding back the anti-PX antibody eluted from the affinity column. Thus, disease-causing PV antibody identified PX, a novel human annexin that acts as a keratinocyte cell surface receptor for ACh, and, therefore, may mediate known biological effects of this cytotransmitter on adhesion and motility of KC.

EXPERIMENTAL PROCEDURES
Sources of Sera and Tissue-The sera and IgG fractions were from well-established PV patients, and from healthy volunteers. This study had been approved by the University of California Davis Human Subjects Review Committee. The diagnosis of PV was made based on the results of both comprehensive clinical and histological examinations together with immunological studies, which included direct immunofluorescence (DIF), indirect immunofluorescence (IIF) on various epi-thelial substrates, immunoblotting, and immunoprecipitation, following standard protocols (23). The serum samples were stored frozen at Ϫ80°C until use in experiments. The serum IgG fractions were isolated using 40% ammonium sulfate followed by dialysis with Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline (PBS; Life Technologies, Inc., Gaithersburg, MD), lyophilized, and reconstituted in PBS as detailed elsewhere (5). The protein concentration was determined using the Micro BCA kit (Pierce). The samples of normal human neonatal foreskins that were used to start keratinocyte cell cultures were transported to the laboratory in culture medium, and the samples of normal human abdominoplasty skin that served as a source of keratinocyte membrane protein for immunoblotting were frozen immediately after harvesting.
Immunoaffinity Purification of Acantholytic Anti-keratinocyte PV Antibody-The enriched fraction of human keratinocyte membrane protein (5) was used as a substrate in immunoblotting experiments aimed at characterizing novel PV antigens. The epidermis was separated from the dermis by incubation in RPMI 1640 medium (Sigma), supplemented to contain 200 mM EDTA for 90 min at 37°C and 5% CO 2 (24), and harvested into a 50-ml polyethylene centrifuge tube filled with ice-cold Tris-buffered saline (TBS), pH 7.4, that contained the following protease inhibitors: 2 mM phenylmethylsulfonyl fluoride, 0.1 mg/ml bacitracin, 10 g/ml leupeptin, 10 g/ml soybean trypsin inhibitor, 10 g/ml pepstatin A, and 10 g/ml chymostatin (all from Sigma). The epidermis was then washed three times by centrifugation, put on ice, and homogenized with a PowerGen tissue-and-cell disrupter (Fisher Scientific, Santa Clara, CA) in the same buffer containing 20 mM Ca 2ϩ . Large organelles and epidermal debris were removed by centrifugation at 2000 ϫ g for 45 min at 4°C, and the cell membrane fraction was pelleted from the supernatant by centrifugation at 80,000 ϫ g for 1 h at 4°C. The pellet was solubilized in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) buffer containing 2% SDS and 5% ␤-mercaptoethanol, boiled for 5 min, and cleared by centrifugation at 40,000 ϫ g for 1 h at 4°C. Western blotting of SDS-PAGE-resolved proteins was performed as reported previously (5) with minor modifications. Briefly, the proteins were separated on a 7.5% SDS-PAGE gel and transferred to an Immobilon-P membrane (Millipore Corp., Bedford, MA), which was blocked, first with 5% milk in TBS for 1 h at 37°C and then with TBS containing 1% normal goat serum, 3% dried milk, and 0.05% Tween 20 (Sigma) overnight at 4°C, and cut into 4-mm wide vertical strips. Each strip was exposed to a primary antibody, i.e. PV or normal human serum, for 1 h at room temperature and then washed thoroughly. The protein bands recognized by PV and normal human IgGs were visualized by biotinylated goat anti-human IgG antibody (Pierce) and developed using a biotin/avidin system (Vectastain ABC system; Vector Laboratories, Burlingame, CA). The specificity of binding was determined in negative control experiments, in which the primary antibodies were omitted. The PV IgG fractions were isolated from the immunoblotting membrane areas that were recognized uniquely by PV IgG, but not normal human IgG, following a procedure described previously (25). Briefly, approximately 3-mm wide horizontal strips carrying a keratinocyte membrane protein with a particular molecular mass of Ϯ3 kDa were cut out from the immunoblotting membrane and incubated overnight with PV serum diluted 1:5 in TBS containing 20 mM CaCl 2 , 0.05% Tween 20 (Sigma), and 1% non-fat skim milk to allow antibody binding. The strips were then washed thoroughly, and the antibodies were eluted by a 3-min incubation at 37°C in a solution containing 500 l of 20 mM sodium citrate, 1% milk, and 0.05% Tween 20 (pH 3.2) and immediately neutralized by adjusting the pH to 7.4 with the 2 M Tris base.
Immunofluorescence Screening Experiments-The IIF experiments testing the ability of PV IgG eluted from the strips of immunoblotting membranes to specifically stain KC in the tissue samples were performed as described previously (5) with minor modifications. Briefly, 4to 8-mm cryostat sections of freshly frozen normal human skin, monkey esophagus, or murine skin were incubated overnight at 4°C with the immunoaffinity-purified PV IgG fractions, after which the tissue sections were washed and binding of primary antibody was visualized by incubating the tissue section with fluorescein isothiocyanate (FITC)labeled goat anti-human IgG antibody (Pierce) for 1 h at room temperature. The specificity of antibody binding was demonstrated by omitting the primary antibody, which abolished the staining. The immunofluorescence images were obtained using a fluorescence microscope (Axiovert 135, Carl Zeiss Inc., Thornwood, NY) with a charge-coupled device video camera (Photon Technology International, Monmouth Junction, NJ) attached.
Cell Culture Screening Experiments-Acantholytic activity of the eluted PV IgGs, which stained the stratified epithelial substrate in a pemphigus-like, "intercellular" pattern, were tested in the monolayers of normal human foreskin KC isolated from the epidermis and grown at 37°C in serum-free keratinocyte growth medium (KGM; Life Technologies, Inc.) containing 0.09 mM Ca 2ϩ in a humid 5% CO 2 incubator, as detailed elsewhere (26). To observe changes in cell morphology, second passage KC were seeded into 6-well tissue culture plates at a cell density of 1 ϫ 10 5 /well and grown to confluence (i.e. for 5-7 days) in 2 ml of KGM per well. The monolayers were then fed with equal amounts of test PV (experiment) or normal human serum (control) IgG fractions, 10 g/ml KGM, and returned to a 5% CO 2 incubator for a 12-h incubation at 37°C. After incubation, the cells were fixed with 3% glutaraldehyde and stained with the trypan blue dye solution (Sigma), and the images of the experimental and control keratinocyte monolayers were captured using a camera-adapted light microscope (Olympus Corp., Lake Success, NY) Screening of cDNA Library-Following standard procedures (27), the human keratinocyte gt11 cDNA library (CLONTECH, Palo Alto, CA) was screened with PV antibody that was immunoaffinity-purified from a 75-kDa keratinocyte membrane protein band. Briefly, the host bacteria Y1090r-were grown overnight, infected with phages from the library for 30 min, plated on Mg 2ϩ -contained agar plates, and grown overnight at 37°C. Over 3 million plaques formed on the bacterial lawns were screened by lifting isopropyl-D-thiogalactoside (IPTG; Sigma) containing nitrocellulose filters (Millipore Corp.). After blocking with 3% dry milk (Sigma) in TBS, the filters were incubate for 2 h at room temperature with the immunoaffinity-purified antibody. The plaques specifically recognized by the antibody were visualized using horseradish peroxidase-conjugated goat anti-human IgG (Bio-Rad, Hercules, CA). The positive plaques were isolated and rescreened until a single clone was isolated. The insert from isolated clones were amplified using a pair of cloning primers specific for the gt11 vector: 5Ј-gggggggtaccggatccccggtcgacggtttccatatgg-3Ј (forward) and 5Ј-cccgggatccatatggtaccaagcttatttttgacaccagacca-3Ј (reverse). The polymerase chain reaction (PCR) products were purified from the gel using the silica membrane spin-column technology (QIAquick Spin, Qiagen, Santa Clarita, CA) and sequenced in both directions with a pair of specific sequence primers: 5Ј-gactcctggagcccg-3Ј (forward) and 5Ј-ggtagcgaccggcgc-3Ј (reverse) using an automated DNA sequencing system (ABI Prism 377, Perkin-Elmer). Homology searches were run against the GenBank nucleotide and protein sequence data bases using the BLAST search program from the National Center of Biological Information web site. The amino acid multiple sequence alignment was performed using Gene Jockey III software (Biosoft, Cambridge, UK). The cDNA insert was removed from the purified gt11 phagemid and subcloned into pBluescript vector (Stratagene, La Jolla, CA) for further characterization.
PCR Experiments-PCR was performed as described by us elsewhere (5). Briefly, each reaction had a final volume of 50 l containing the DNA templates, 1ϫ PCR buffer (Promega, Madison, WI); 0.2 mM each of dATP, dCTP, dGTP, dTTP; 2 units of Taq DNA polymerase (Promega); and 1 M each of the sense and antisense primers. The reaction mixture was first heated at 95°C for 5 min and hot-started with 2 units of DNA Taq-polymerase (Life Technologies, Inc.) followed by 35 cycles (or 15 cycles for cloning experiments) of denaturing at 95°C for 60 s, annealing at an appropriate temperature (optimized for primers used in each PCR) for 60 s, and extension at 72°C for 120 s. In the final cycle, the extension was increased to 8 min. The PCR products were electrophoresed on 2% agarose gels containing 1 g/ml ethidium bromide and photographed under fluorescent UV illumination (AlphaImager 2000, Alpha Innotech Corp., San Leandro, CA). The size of the PCR product was estimated by using a 100-or a 250-bp DNA ladder standard (Life Technologies, Inc.).
Expression of rPX-His in Escherichia coli-The expression vector pQE-30 (Qiagen), which is designed to express proteins containing a 6xHis-tag at the N-terminal, was used to express rPX-His. The vector was linearized by digestion with the SphI and KpnI restriction enzymes for 1 h at 37°C, then purified from an agarose gel, and incubated at 37°C with 5 units of alkaline phosphatase (Promega) to enhance the efficiency of ligation. cDNA from the PX gt11 clone was amplified by PCR with the following primers: 5Ј-ccgcatgcgatgacgatgacaaaatgtctgtgactggcgggaagatggc-3Ј (forward) and 5Ј-cccgggatccatatggtaccaagcttatttttgacaccagacca-3Ј (reverse). The forward primer was designed to have an additional SphI restriction site, which allows the insert to be ligated in-frame with the 6xHis gene of the pQE-30 vector. The PCR product was double digested with SphI and KpnI restriction enzymes and purified. Digested product was directionally cloned into unique SphI and KpnI sites in the multiple cloning site of the pQE-30 vector. The ligated vector was used to transform E. coli expression strain M15 (Qiagen). Transformed cells were plated on a NYZ agar plate containing 25 g/ml kanamycin and 100 g/ml ampicillin and grown overnight. To verify the clone that produced the rPX-His protein, transformed bacterial colonies were blotted to a marked nitrocellulose filter and inversely placed on an IPTG-containing NYZ agar plate and grown for 4 h. The filter was then treated with denaturing buffer, neutralized, blocked with 3% non-fat milk in TBS and screened for colonies that produced rPX-His using anti-RGS-His monoclonal antibody (Qiagen). Positive clones were selected from the original plate, and their plasmids were sequenced with a specific primer to confirm that the correct PX cDNA had proper frame and orientation. A representative clone was inoculated into NYZ medium containing 25 g/ml kanamycin and 100 g/ml ampicillin. The culture was incubated, with shaking, at 37°C until an A 600 of 0.6 was reached, and IPTG was added to a final concentration of 2 mM. Culture samples (2 ml each) were collected every hour during 4 h and centrifuged, and the bacterial pellets were dissolved in sample buffer and analyzed by SDS-PAGE with Coomassie Blue staining.
Production and Purification of rPX-His-Large scale rPX-His production was performed in 1 liter of medium, as described above. The cell pellet was lysed at room temperature by stirring the pellet in a buffered solution containing 8 M urea, pH 8.0 (lysis solution). Once the solution became translucent, the cellular debris was removed by centrifugation at 40,000 ϫ g for 1 h at 4°C. The clarified supernatant was incubated with nickel-nitrilotriacetic acid-agarose resin (Ni-NTA, Qiagen) to capture the His-tagged protein. The resin was washed with several volumes of buffered 8 M urea, pH 6.3, until a A 280 of about 0.001 was achieved, and loaded into a column. The rPX-His protein was eluted from the column using either denaturing or non-denaturing condition. The denatured rPX-His was eluted with a buffer containing 8 M urea, pH 5.9 and 4.5, resolved by SDS-PAGE, and analyzed by immunoblotting with PV IgG. Or, the immobilized rPX-His was first renatured over a period of 1.5 h in a linear 6 to 1 M urea gradient in 500 mM NaCl, 20% glycerol, 20 mM Tris-Cl, pH 7.4, containing protease inhibitors, and then eluted with a non-denaturing buffer containing 50 mM NaH 2 PO 4 , 300 mM NaCl, and 250 mM imidazole, pH 8.0. The purified renatured rPX-His was used in the radioligand binding assays as well as for immunoaffinity purification of anti-PX PV antibody.
Radioligand Binding Assays with rPX-His-Nitrocellulose filters (13-mm diameter, catalog no. HAWPO1300, Millipore Corp.) with a total protein capacity of 160 g/cm 2 were placed into the bottom of each well of a bovine serum albumin-pretreated 24-well standard cell-andtissue culture plate (Nalco Nunc International, Denmark). One g of the affinity-purified rPX-His was diluted in 300 l of PBS and loaded into each filter for overnight incubation at 4°C to allow complete absorption of rPX-His by the filter (determined in a series of preliminary experiments by measuring the optical density at 280 nm of free rPX-His remaining in the solution). The membranes carrying rPX-His were blocked with 2% bovine serum albumin for 1 h at room temperature, after which the plates were put on ice, washed three times with ice-cold PBS, and exposed in triplicate for 1 h to increasing, from 0 to 1000 nM, concentrations of [ 3 H]ACh iodide (82.0 mCi/mmol, NEN Life Science Products, Boston, MA). Nonspecific binding was measured in parallel wells, in which the filters were exposed to the same increasing doses [ 3 H]ACh in the presence of 100-fold concentrations of non-labeled ACh iodide (Sigma). The filters were then washed thoroughly with ice-cold PBS, placed in 6-ml vials containing 5 ml of liquid scintillation mixture (Ecolite, ICN, Costa Mesa, CA), and their radioactivity was counted in the liquid scintillation counter (model 1409, Wallac Inc., Gaithersburg, MD). The specific binding was computed by subtracting the nonspecific binding from total binding, and the binding capacity (B max ) and dissociation constant (K d ) were calculated using the ligand binding analysis software Prism (GraphPad, San Diego, CA). In a separate set of radioligand binding experiments, we investigated the ability of the cholinergic radioligand [ 3 H]PrBCM (5 mCi/mmol of the customized [ 3 H]PrBCM; NEN Life Products) to label rPX-His and the ability of the nicotinic agonist nicotine and the muscarinic agonist muscarine (both from Sigma) to abolish rPX-His labeling with [ 3 H]PrBCM. Prior to the assay, [ 3 H]PrBCM was cyclized in 10 mM PBS at 30°C for 20 min to activate the aziridinum ions (28).
Immunoaffinity Purification and Characterization of Anti-PX PV IgG-PV sera were diluted 1:5 in Immunopure Gentle binding buffer (Pierce) and incubated overnight at 4°C with rPX-His immobilized on Ni-NTA resin. The pass-through serum fraction was collected, and the IgGs were isolated using 40% ammonium sulfate precipitation followed by dialysis against Ca 2ϩ -and Mg 2ϩ -free PBS. The rPX-His column with bound PV antibody was washed 10 times with TBS containing 300 mM NaCl, and the immunoaffinity-purified anti-PX IgG fraction was eluted from the column by Immunopure Gentle elution buffer and desalted on a D-Salt Exellulose plastic desalting column (both from Pierce). The pattern of specific binding of the eluted antibody was examined by IIF on human skin and monkey esophagus. The antigenic profile of the eluted PV IgG was identified by immunoprecipitation of metabolically labeled human keratinocyte proteins (see below), which is considered the most sensitive and specific approach to characterize the antigenic specificity pemphigus antibodies (7).
Metabolic Labeling of Cultured KC and Immunoprecipitation Assay-Second passage human foreskin KC were grown to approximately 90% confluence, washed thoroughly with prewarmed (37°C) PBS, incubated for 15 min at 37°C in methionine-free Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 15% newborn calf serum, and then exposed to 100 Ci/ml [ 35 S]methionine (1000 Ci/mmol, Amersham Pharmacia Biotech, Arlington Heights, IL) in 1.8 mM Ca 2ϩ labeling medium for 16 h in a humid, 5% CO 2 incubator at 37°C. The keratinocyte monolayers were then washed thoroughly, and the cells were scraped with a rubber policemen; pelleted by centrifugation at 300 ϫ g for 5 min at 4°C; resuspended in ice-cold 10 mM TBS containing 0.025% NaN 3 , 20 mM Ca 2ϩ , 1% Nonidet P-40 (Amersham Pharmacia Biotech) and the protease inhibitors 1 mM iodoacetamide, 2 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 5 g/ml pepstatin A, and 5 g/ml chymostatin; put on ice; and homogenized. Solubilized [ 35 S]methionine-labeled proteins were separated by centrifugation at 40,000 ϫ g for 60 min at 4°C and used as a source of naturally folded keratinocyte proteins. The radiolabeled keratinocyte protein solution was incubated with immunoaffinity-purified anti-PX PV IgG overnight at 4°C with gentle shaking. The immune complexes were precipitated with slurry protein A-Sepharose suspension, washed, and resolved by 7.5% SDS-PAGE. The gels were fixed and enhanced with 1 M sodium salicylate, and the radioactivity was analyzed using the storage phosphor autoradiography feature of the Storm system (Molecular Dynamics, Mountain View, CA).
Antibody Transfer to Neonatal Mice-The PV phenotype was induced in neonatal mice by passive transfer of PV patients' serum IgG fractions to normal Balb/c mice (8). The IgGs were injected intraperitoneally through a 30-gauge needle at a dose of 20 mg/g of body weight per day into 10-to 12-h-old pups. The neonates always received the same amounts of PV IgG (experiment) and normal human IgG (control). The latter was isolated from normal human serum purchased from Sigma Chemical Co. The mice were sacrificed when fully developed skin lesions could be seen or, if no gross lesions could be observed, approximately 24 h after the last injection. The lesional and perilesional skin samples were collected and examined by staining with hematoxylin and eosin and by DIF with FITC-conjugated goat anti-human IgG antibody (Pierce).
Statistics-The results of quantitative experiments were expressed as mean Ϯ S.D. Significance was determined using the Student's t test.

RESULTS
Selection of an Immunoaffinity-purified Acantholytic Antikeratinocyte PV IgG as a Candidate for cDNA Library Screening-In an attempt to identify the pathogenic PV antibody, we investigated the ability of different fractions of immunoaffinity-purified anti-keratinocyte PV IgGs to: 1) stain the stratified epithelial substrates in a fishnet-like, "intercellular" pattern, which is diagnostic of PV (1); and 2) induce acantholysis in keratinocyte monolayers, which has become a standard approach to test disease-causing ability of PV antibodies (29,30). Among tested PV IgG fractions, the antibody eluted from the horizontal strip excised from the 75-kDa area of the immunoblotting membrane produced intercellular epithelial staining of both normal human skin and monkey esophagus in IIF experiments (Fig. 1, A and B). Treatment of confluent monolayers of normal human KC with this immunoaffinity-purified PV IgG fraction, but not with normal human IgG, produced changes of the cell morphology characteristic of pemphigus acantholysis (Fig. 1, C and D). No acantholysis could be seen in cultures treated with equal amounts of PV IgG eluted from the 130-kDa area of immunoblots of normal human keratinocyte proteins (data not shown). Therefore, PV IgG immunoaffinity-purified on a 75-kDa band was selected to probe the gt11 human keratinocyte cDNA expression library.
Isolation of cDNA Clones Encoding PX and Sequence Anal-ysis-Approximately 3 ϫ 10 6 plaques of gt11 human keratinocyte cDNA expression library were screened with the affinity-purified antibody from three PV sera (codes: PRC-45, PRC-46, and PRC-47), which contained the anti-75-kDa band acantholytic PV IgG that stained the stratified epithelium in a pemphigus-like pattern. In the first round of screening, four plaques were found to be positive for antibody binding. However, only two clones, designated as K5 and K12, remained immunoreactive after subsequent rescreening. Because PV IgG eluted from the filter blotted with both K5 and K12 clones stained monkey esophagus in a pemphigus-like pattern (data not show), both clones were selected for further characterization. PCR amplification of the cDNA insert using a pair of gt11 cloning primer revealed that K5 and K12 clones carried the 1.3-and 1.4-kb cDNA inserts, respectively ( Fig. 2A). Unexpectedly, sequence analysis of the cDNA inserts from both clones predicted the same open reading frame of 1035 bp, encoding a full-length protein comprised of 345 amino acids (Fig. 2B) with a calculated molecular mass of 38.3 kDa. Examination of the nucleotide sequence revealed an in-frame stop codon situated upstream of the first ATG codon, which indicated that a complete coding region was identified. There were two tandem ATG potential translation initiation codons after the upstream in-frame stop codon. The first one most likely represented the initiation codon, because it was preceded with the Kozak consensus sequence (31). No poly(A) tail was detected. A BLAST search of the GenBank data base at the NCBI web site showed that the nucleotide sequence encoded a previously unknown molecule. The deduced amino acid sequence revealed a high degree of homology to the members of the Ca 2ϩ -dependent annexin protein gene family. The strong- A, PCR amplification of cDNA inserts from gt11 phages isolated from the clones K5 and K12 using specific gt11 forward and reverse cloning primers. The 1.5-and 1.6-kbp PCR products carried copies of 1.3-and 1.4-kbp cDNA inserts, respectively, from the two clones that were specifically recognized by affinity-purified PV IgG as a result of screening of 3 million plaques of a gt11 human keratinocyte cDNA expression library. Sequence analysis of both cDNA inserts revealed that both encoded for the same novel molecule, PX. B, the nucleotide sequence and the predicted amino acid sequence of PX. The in-frame upstream and downstream stop codons are underlined. The Kozak sequence that precedes the potential initiation ATG codon is double underlined. C, multiple amino acid sequence alignment of PX with the annexin-2 sequences reported for different species showing that PX shares the same amino acids in most of the conserved regions. Shaded regions indicate the identical amino acid residues among all compared sequences. The arrow denotes potential glycosylation site. The asterisks denote the potential type II Ca 2ϩ binding sites. The potential actin bundling site is underlined. Anx-2, annexin-2. est similarity, approximately 40%, was observed with annexin-2 present in chicken (GenBank accession number P17785), cow (P04272), rat (Q07936) and humans (NP004030). The amino acid sequence alignment (Fig. 2C) revealed several conserved regions, including the type II Ca 2ϩ binding sites (32,33) and the actin bundling site that plays a role in Ca 2ϩ -dependent bundling of actin microfilaments by annexins (34). Because of its homology to the members of the annexin protein gene family, we tentatively named this newly discovered PV antigen pemphaxin (i.e. pemphigus ϩ annexin ϭ pemphaxin).
Expression of the rPX-His Fusion Protein in E. coli and Its Affinity Purification-To allow experiments with immunoaffinity-purified anti-PX PV antibody, we produced a full-length recombinant PX. Because both K5 and K12 clones carried fulllength cDNAs encoding the complete open reading frame of PX, we chose to directionally clone the K5 cDNA to the pQE-30 expression vector, which was designed to express PX protein carrying a poly-His tag at its N terminus. The cloned pQE-30-PX was transformed into E. coli M15 cells, and the colonies expressing rPX-His were selected by screening with anti-RGS-His monoclonal antibody. Antibody staining revealed six strongly positive colonies that contained correct PX inserts, as confirmed by subsequent sequencing. Clone 1 was selected for a time course characterization of PX expression. As seen in Fig.  3A, the transfected bacteria began to produce rPX-His after induction with 2 mM IPTG, and the amount of this fusion protein, estimated by the time course study with the time points of 1, 2, 3, and 4 h, gradually increased and reached saturation at 4 h after induction. As expected from the deduced molecular mass of PX, the newly produced rPX-His migrated with a 40-kDa protein band on the 12% SDS-PAGE gel. No proteins were induced by IPTG in control, non-transfected E. coli M15 cells (data not shown). The rPX-His was isolated from the mixture of E. coli proteins on the Ni-NTA column via its His residues. The rPX-His fusion protein was eluted from the column, and its purity was confirmed by finding a single band in 12% SDS-PAGE-resolved eluant (Fig. 3A, lane PX). The ability of rPX-His to exhibit PX conformational epitope(s) recognized by PV antibody was confirmed by immunoblotting of affinity-purified rPX-His with the three PV sera that were used in the cDNA library screening experiments (Fig. 3B).
Cholinergic Radioligand Binding by rPX-His-Cholinergic ligand binding properties of annexins-1, -2, and -3 (35) suggested that PX also acts as a cholinergic receptor binding ACh on the cell surface of KC. To test this hypothesis, rPX-His was used in a standard radioligand binding assay. The saturable specific binding was achieved with the reversible cholinergic radioligand [ 3 H]ACh (Fig. 4A). The analysis of binding kinetics revealed the K d value of 909 nM and a B max of 176 pmol/mg of protein, indicating that, on the cell membrane of KC, PX may act as a low affinity receptor for endogenously produced and secreted ACh.
Because we demonstrated in a previous study (5) that 85% of pemphigus patients develop autoantibodies, which immunoprecipitate a keratinocyte membrane protein covalently labeled with the cholinergic radioligand [ 3 H]PrBCM, we further asked whether [ 3 H]PrBCM can specifically label rPX-His. The specificity of [ 3 H]PrBCM binding to rPX-His was demonstrated in the binding inhibition experiment using non-labeled cholinergic ligands ACh, nicotine, and muscarine as competitors (Fig.  4B). As expected, ACh as well as its nicotinic and muscarinic congeners decreased significantly (p Ͻ 0.05) the amount of [ 3 H]PrBCM bound to rPX-His, indicating that PX exhibits dual, muscarinic and nicotinic pharmacology. The dosedependent radioligand binding inhibition assay with [ 3 H]PrBCM could not be performed because of the irreversible nature of its binding to a receptor molecule, via an alkylation reaction (36).
Characterization of Immunoaffinity-purified Anti-PX PV Antibody-The anti-PX PV IgG was immunoaffinity-purified on rPX-His immobilized on the Ni-NTA column via its His tags, and the PV IgG fraction eluted from the resin was characterized by: 1) IIF assay using human skin and monkey esophagus as substrates; and 2) immunoprecipitation assay with metabolically radiolabeled keratinocyte proteins. In the IIF assays, the immunoaffinity-purified anti-PX PV IgG stained, in a distinct fishnet-like, pemphigus pattern, the stratified squamous epithelium in human skin and monkey esophagus (Fig. 5, A and   FIG. 2-continued  B). The epithelia of other types, such as those lining human bronchi, lung alveoli, small and large intestine, and renal glomeruli, did not exhibit specific staining (data not shown), indicating that the stratified epithelium is a major site of the epithelial expression of PX in human beings. We did not test non-epithelial tissues in this study.
Although addition of a 6xHis-tag to PX should not alter its conformational epitope, we sought to rule out even a remote possibility that, in addition to anti-PX, the rPX-His fusion protein absorbs antibodies of other specificities. The purity of PV IgG eluted from rPX-His was tested in a immunoprecipitation assay, which allows an antibody to recognize its antigen in the native form, to increase the sensitivity and specificity of antibody characterization. The immunoprecipitation assay showed that the affinity-purified anti-PX PV IgG precipitated keratinocyte proteins with apparent molecular masses of 40 and 80 kDa (Fig. 5C). Because the deduced molecular mass of PX is 38.3 kDa, these results suggested that PX exists as a monomer and a homodimer in KC. This hypothesis was further supported by demonstration of the predicted reciprocal changes in the relative amounts of the 40-and 80-kDa products depending on the presence or absence of the reducing agent ␤-mercaptoethanol in the SDS-PAGE buffer (Fig. 5C). Indeed, the covalent linkage of two annexins in a dimer is common for certain annexins (reviewed in Ref. 37).
Absorption of Disease-causing PV Antibodies with rPX-His-To determine the pathophysiological significance of anti-PX antibody in pemphigus, we next asked if depletion of the PV IgG fraction of anti-PX antibody could affect the ability of PV IgG to cause gross skin blisters in neonatal mice. Equal amounts of the intact whole PV IgG fraction (positive control) and the PV IgGs that either passed through the Ni-NTA column containing immobilized rPX-His or were eluted from the column were injected intraperitoneally into 10-to 12-h-old Balb/c mice at a concentration of 20 mg of IgG/g of body weight per day. Only the mice that received non-absorbed PV IgGs reproducibly developed pemphigus-like gross skin lesions between the 16th and 24th h after a single injection. The mice injected repeatedly with either the pass-through (Fig. 5D) or the immunoaffinity-purified anti-PX IgGs (not shown) did not develop any macro-or microscopic skin changes, despite deposition of injected IgGs in mouse epidermis in both cases (Fig.   FIG. 3. Expression of the rPX-His fusion protein in E. coli and its affinity purification. A, the time-course study of the expression of rPX-His. The selected E. coli M15 cells transformed with pQE30-PX were grown in NYZ medium to an optical density of 0.6. at 600 nm and induced with 2 mM IPTG. 1-ml samples of bacterial culture were collected before induction and at 1, 2, 3, and 4 h post induction. The protein extracts of these samples were analyzed on 12% SDS-PAGE gel stained with Coomassie Blue. A sample of rPX-His purified on a Ni-NTA column is designated as PX, and the wash-through fraction is designated as W. No additional proteins were produced in the control experiments in which non-transfected E. coli M15 cells were induced with IPTG (not shown). B, the conformational epitope of rPX-His allows its immunorecognition by PV IgG. Western blots of affinity-purified rPX-His were stained with sera from the three PV patients whose IgG fraction was used to screen gt11 human keratinocyte cDNA expression libraries. Binding of anti-PX PV IgG was visualized using horseradish peroxidase-conjugated goat anti-human IgG antibody. No staining could be seen in the negative control experiment in which the primary antibody was omitted (not shown). In the reference lane, denoted PX, the rPX-His fusion protein is visualized by staining with Coomassie Blue. 5E). These results indicated that, although absorption with rPX-His eliminates the acantholytic activity of PV serum, the anti-PX antibody alone is not sufficient to induce acantholysis and gross skin blisters in neonatal mice. Therefore, we hypothesized that, although anti-PX antibody is essential for acantholysis development, it is not the only one in the pool of disease-causing PV antibodies that are required to break the integrity of live epidermis.
To test this hypothesis, we sought to determine if acantho-lytic activity of preabsorbed PV IgGs could be restored by adding back the adsorbed anti-PX antibody. As seen in Fig. 5 (F  and G), the pups injected with the pass-through PV IgGs supplemented with anti-PX IgG eluted from the affinity column produced the PV phenotype that was indistinguishable from the epidermal acantholysis and gross skin blisters produced by non-adsorbed PV IgG (not shown). These results clearly indicated that, in addition to anti-PX antibody, the pool of diseasecausing PV IgG contains autoantibodies to other keratinocyte self-antigens and suggested that a cumulative effect of antikeratinocyte antibodies of different specificities is required to break up the integrity of live epidermis and induce skin blistering.

DISCUSSION
In this study we selected the PV IgG fraction that can both stain the epithelial substrates in the pemphigus-like pattern and induce acantholysis in keratinocyte monolayers to probe gt11 keratinocyte cDNA library for novel targets of diseasecausing PV antibodies. The PV antibody immunoaffinity-purified on a 75-kDa keratinocyte protein band identified a novel human annexin-like molecule, which we termed PX. Recombinant PX was produced and shown to bind specifically ACh and its nicotinic and muscarinic congeners. The obtained results indicate that PX may serve as a cell surface cholinergic receptor mediating a novel ACh signaling pathway involved in the physiological control of cell-to-cell adhesion and that autoimmunity to PX may lead to acantholysis.
Pemphigus is an autoimmune disease with a complex pathophysiology. Both humoral (38) and cellular (39) effectors of autoimmune aggression against KC are involved in the pathogenesis of this disease, and it has been demonstrated that local activation of trypsin-like serine proteases, such as plasminogen activator (40), complement (41), eicosanoids (42), and proinflammatory cytokines (29,43), all can contribute to acantholysis. The precise mechanism leading to acantholysis in PV, however, is yet to be determined. It is currently held that an autoantibody to the 130-kDa adhesion molecule Dsg 3 causes pemphigus by disrupting directly the keratinocyte cell-to-cell bridges or desmosomes (44,45). The intuitive notion that the disease of skin adhesion is caused by an antibody to the adhesion molecule, however, awaits its direct experimental confirmation. Meanwhile, Kitajima et al. (46) demonstrated that desmosome formation induced by switching the incubation medium from a low to a high Ca 2ϩ content is not inhibited by the binding of PV IgG to the cell membrane of cultured KC. In agreement with this report, we could not detect any morphological changes in the keratinocyte monolayers treated with the anti-130-kDa PV IgG for 16 h, whereas the acantholysis in cell monolayers usually develops within 12 h after addition of the whole PV IgG fraction (20, 29, 30). Fan et al. (47) attempted to create an animal model of PV by immunizing four different strains of mice, Balb/c, DBA/1, SJL/J, and HRS/J, with fulllength Dsg 3 protein, recombinant extracellular portion of Dsg 3, and the synthetic peptides spanning the entire Dsg 3. However, they found no signs of pemphigus, oral or cutaneous, in any of the animals, despite relatively high, up to 1/2560, titer of circulating anti-Dsg 3 antibodies produced by immunized animals. Furthermore, even after the immune sera were concentrated 10-fold and inoculated into neonatal mice, the mice of only one strain, Balb/c, developed the lesion. These results demonstrated that, at the serum titers that are equivalent or exceeding those found in PV patients, the anti-Dsg 3 antibody is not sufficient to cause pemphigus symptoms. The suprapharmacological doses of this antibody, however, can physically interfere with cell-to-cell adhesion, as illustrated by the occurrence of microscopic changes in the oral mucosa of immuno-  2 and 3) were used to immunoprecipitate 35 Smetabolically labeled human keratinocyte protein extract, as detailed under "Experimental Procedures." The immunoprecipitate in lanes 1 and 3 was diluted in SDS-PAGE buffer containing both 2% SDS and 5% ␤-mercaptoethanol, which allowed predominant visualization of rPX-His in the form of a 40-kDa monomer. The immunoprecipitate resolved in lane 2 was treated without the reducing agent ␤-mercaptoethanol, which produced a reciprocal staining picture, because omission of ␤-mercaptoethanol allowed predominant visualization of rPX-His in a form of a naturally assembled homodimer with an apparent molecular mass of 80 kDa. D and E, results of passive transfer of PV IgG preabsorbed with rPX-His to a neonatal Balb/c mouse. Lack of any visible alteration of skin integrity in a pup injected intraperitoneally during 2 days with the pass-through PV IgG fraction in a total dose of 40 mg/g body weight. Demonstration of the deposits of injected pass-through PV IgGs in the epidermis of this mouse by DIF (E). Scale bar, 50 m. F and G, results of the passive transfer experiment using the pass-through PV IgG fraction that was supplemented with the immunoaffinity-purified anti-rPX-His IgG. An extensive blister with a loosely attached peripheral skin (positive Nikolsky sign) in a neonate approximately 16 h after a single intraperitoneal injection of 20 mg/g PV IgG (F). The skin blister in this pup resulted from a typical PV-like suprabasilar acantholysis observed by hematoxylin and eosin examination of the perilesional skin (G). The fishnet-like deposits of injected IgG in the epidermis of this mouse were confirmed by DIF (not shown). Scale bar, 50 m. deficient Rag-2 knockout mice grafted with a spleen producing anti-Dsg 3 antibodies (48). Unfortunately, the interpretation of findings in mice with adoptively transferred anti-Dsg 3 antibodies in the study of Amagai et al. (48) is complicated by its rather controversial nature, which includes direct conflict with the existing data. For instance, according to Amagai et al. (48), lack of skin changes in Balb/c mice immunized with Dsg 3 is attributed to inability of this strain of mice to produce anti-Dsg 3 antibody, whereas Fan et al. (47) achieved high anti-Dsg 3 antibody titers in these animals, albeit without any mucocutaneous signs of PV. Furthermore, Amagai et al. (48) opine that, by analogy with the interpretation of the Dsg3 null phenotype (18), a transient hair loss accompanied by transient microscopic alterations of keratinocyte adhesion in the oral cavity, which is all that can be observed in the recipient Rag-2Ϫ/Ϫ mice, should be interpreted as the PV phenotype. However, the following facts argue against this interpretation: 1) hair loss is not a sign of PV (1); 2) true PV is a disease severe enough to kill approximately 90% of patients, if left untreated (reviewed in Ref. 2); and 3) neither recipient Rag-2Ϫ/Ϫ mice nor Dsg3 null mice develop spontaneous skin blisters (5,15,18). Nevertheless, the notion about the pathophysiological significance of Dsg 3 antibody in PV has been supported by the results of in vivo experiments in which pemphigus antibodies affinity-purified on the rDsg3-Ig chimera induced gross skin blisters in neonatal mice (13,14). Unfortunately, the profile of PV IgGs adsorbed by the rDsg3-Ig-His baculoprotein has never been shown, leaving unresolved the purity and specificity of the antibodies used in the passive transfer experiments. Therefore, we had to characterize the antigenic reactivity of PV IgG absorbed with rDsg3-Ig-His in our laboratory (19). We established that the antibodies adsorbed on rDsg3-Ig-His are directed toward several keratinocyte proteins, including an unknown 130-kDa self-antigen recognized in the Western blot of keratinocyte proteins of Dsg3 null mice.
To select the PV IgG fraction that most likely contains disease-causing antibody, we screened PV IgG fractions eluted from different areas of the immunoblotting membrane for their ability to both: 1) stain epidermis in a fishnet-like, pemphigus pattern; and 2) produce acantholysis in keratinocyte monolayers. The anti-75-kDa band PV IgG met both criteria. Failure of the antibody eluted from the 130-kDa area of the immunoblotting membrane to fulfill both criteria was not surprising, because in the past this antibody was selected for the cDNA screening experiments that identified Dsg 3 based on the first criteria only (9). In our study, anti-75-kDa band PV antibody caused acantholysis, which could be observed at 0.09 mM Ca 2ϩ in KGM. Although expression of Dsg 3 in KC requires preincubation of the cells at high, from 1.8 to 2.55 mM, extracellular Ca 2ϩ (9,49,50), we and other workers have previously demonstrated that binding of disease-causing PV IgGs to KC and acantholysis in cell monolayers both occur at as low as 0.1 mM Ca 2ϩ (20,29). This fact suggests that, in addition to blocking the "adhesive sites" of desmosomal cadherins with anti-Dsg PV IgG, binding of pemphigus antibodies to KC initiates an intracellular signaling cascade that can lead to disassembly of other types of intercellular junctions comprised of classical cadherins, such as tight junctions, adherence junctions, and gap junctions, all of which can mediate keratinocyte cell-to-cell adhesion at low Ca 2ϩ (51)(52)(53).
Screening of the gt11 keratinocyte cDNA expression library with the acantholytic anti-75-kDa band PV antibody identified PX, a novel human annexin-like molecule. It appeared that two of 3 ϫ 10 6 plaques labeled with PV IgGs carried cDNA encoding for the same previously unknown annexin-like molecule with the predicted molecular mass of the translated product of 38.3 kDa. Sequence alignment with known annexins showed that PX shares the same amino acids in most of the conserved regions and is ϳ40% similar to annexin-2. Annexin-2 may exist as a monomer, dimer, heterodimer, or heterotetramer in which two annexin-2 molecules combine with two smaller subunits, p11, that resemble the S-100 protein of the calmodulin family (54). Because the PV IgG immunoaffinity-purified on rPX-His labeled keratinocyte proteins with apparent molecular masses of 40 and 80 kDa, it can be postulated that PX forms homodimers.
Annexins comprise a unique family of Ca 2ϩ -and phospholipid-binding proteins encoded by some 20 different genes, which are ubiquitous among eukaryotic organisms, single-celled organisms, and plants and animals (reviewed in Refs. 55,56). Individual annexins have been described under the names anchorin, calcimedin, calelectrin, calpactin, calphobindin, chromobindin, endonexin, lipocortin, and synexin. Different annexins have been shown to: 1) participate in ligand-mediated cell signaling both directly, by forming Ca 2ϩ -sensitive, voltagegated Ca 2ϩ channels, and indirectly, by generating membranederived second messengers; 2) mediate anti-inflammatory action of glucocorticosteroids via inhibition of phospholipase A 2 ; 3) regulate and directly mediate cell-to-cell adhesion; 4) mediate endo-and exocytosis; 5) inhibit blood coagulation; 6) regulate Ca 2ϩ -dependent Cl Ϫ conductance; and 7) participate in the processes of cell proliferation, apoptosis, and virus infection (reviewed in Refs. 37,[57][58][59][60]. PX turned out to be a sixth protein of the annexin protein gene family identified in normal human skin to date. Annexins-1, -2, -5, -6, and -7 have been demonstrated previously (61,62). Expression of annexins in epidermis is differentiation-dependent (63). Annexin-1 immunoreactivity is found almost entirely around the perimeter of KC, especially tonofilament/ desmosome-rich prickle KC (61). It has been noted that raising intracellular Ca 2ϩ results in peripheral relocations of annexins-2, -4, -5, and -6 from the perinuclear areas (64). Annexin-2 has been shown to be directly involved in regulation of cell adhesion and migration (65). The presence in PX of the conserved sites providing for Ca 2ϩ binding and for bundling of actin filaments suggests that PX, just like annexin-2, regulates assembly and maintenance of the cytoskeletal units. This actin polymerization is now believed to play a crucial role in epithelial cell-to-cell adhesion, because disruption of this process in an animal model causes skin lesions indistinguishable from PV lesions (66).
Although annexins lack a leader sequence (and do not pass the Golgi apparatus), they are found on the keratinocyte cell surface, where they can function as receptors. Extracellular annexins have been demonstrated to bind collagen, tenascin, and plasminogen activator (65,(67)(68)(69)(70). Binding of tenascin-C to annexin-2 provokes three cellular responses: loss of adhesion, lateral migration, and enhanced cell division (71). Tenascin expression is induced in pemphigus skin as well as in the skin of other blistering dermatoses (72).
To characterize PX, we produced full-length recombinant protein using pQE-30 vector, which contained IPTG-inducible promoter transformed into the E. coli M-15 competent cells. Plasmid purified from this clone was analyzed by restriction enzyme analysis and sequencing. Both confirmed that the PX DNA insert was 100% correct. The rPX-His was affinity-purified and used in standard receptor-ligand binding assays with the cholinergic radioligand [ 3 H]ACh. The analysis of the saturable binding of [ 3 H]ACh showed that PX can function as a low affinity cholinergic receptor on the cell membrane of KC. These results were expected, because choline, which itself serves as a pharmacological agonist of cholinergic receptors (73,74), has been shown to specifically bind to annexin-1, -2, and -3 (35). Likewise, rPX-His could be specifically tagged with a covalent cholinergic radioligand [ 3 H]PrBCM, which was previously used by us to label keratinocyte membrane proteins immunoprecipitated by 85% of pemphigus patients (5).
The results of pharmacological experiments demonstrated that rPX-His exhibited conformational structure, thus allowing specific binding of cholinergic ligands. Post-translational modification is not required for ligand binding to single-unit ACh receptors, such as the muscarinic receptor (75,76). However, the affinity of ACh binding by the wild-type PX, which can forms dimers, may be different from that shown by rPX-His in vitro, because the bacterial system in which it was expressed was not capable of post-translational modification, such as glycosylation, which is known to play an important role in ligand binding by multi-subunit ACh receptors such as the nicotinic receptor (77). Thus, PX can act as a novel keratinocyte cell surface receptor for the cytotransmitter ACh, synthesized and secreted by human KC in autocrine and paracrine fashions, and mediate known effects of ACh and cholinergic drugs on keratinocyte adhesion (reviewed in Refs. 21,22). PX can also represent, at least in part, the putative keratinocyte cholinergic receptors targeted by PV IgG (5,20).
The drugs that act at keratinocyte cholinergic receptors have been shown to alter cell motility and adhesion. Exposure of suspended KC to ACh results in attachment and spreading of the cells on the dish surface and development of intercellular contacts within 20 -30 min, whereas non-stimulated cells accomplish this process within 90 -120 min. On the other hand, exposure of a confluent keratinocyte monolayer to pharmacological antagonists of ACh leads to a characteristic acantholytic response. The cells retract their cytoplasmic projections, lose cell-to-cell attachments, detach from each other, and become round in shape and non-motile-characteristics that remarkably resemble pemphigus acantholysis in vitro (20). We have previously reported that ACh and its muscarinic and nicotinic congeners can prevent and reverse acantholysis produced in keratinocyte cultures by PV IgG (20). A receptor/ligand type of interaction of disease-causing PV IgG, with its target being a keratinocyte cell membrane protein, was first proposed by Patel et al. (78) based on the results of time-course study of the fate of the PV antibody/antigen complex. A direct evidence of activation of second messenger systems in response to PV IgG binding to KC have been obtained in the studies showing changes with phospholipase C, inositol 1,4,5-trisphosphate, transmembrane flux and intracellular levels of Ca 2ϩ , intracellular cAMP/cGMP ratios, and activity and intracellular location of protein kinase C (reviewed in Refs. 5,79). Therefore, binding of anti-PX antibody to KC may lead to acantholysis by competing with the natural agonist ACh, thus interrupting physiological regulation of keratinocyte adhesion. In keeping with the notion that autoantibody-mediated ligation of PX on the cell membrane of KC can alter the cell adhesive function are the results showing that an antibody to annexin-2 inhibits cell-to-cell attachment (80).
To determine the role of anti-PX antibody in pemphigus pathophysiology, we preabsorbed PV sera with rPX-His and tested acantholytic activities of both the PV IgGs depleted of anti-PX antibody and the PV IgG eluted from rPX-His. Neither IgG fraction could induce micro-or macroscopic mucocutaneous lesions in neonatal Balb/c mice. Addition of the adsorbed anti-PX PV IgG to the preabsorbed IgG fraction restored its acantholytic activity. These findings suggested that anti-PX antibody is one of the major contributors to skin blistering in PV patients. The fact that anti-PX PV antibody alone was sufficient to cause acantholysis in vitro (Fig. 1) but could not do so in vivo was not surprising. Obviously, the cell-to-cell adhesion of KC cultured at low Ca 2ϩ is less sophisticated than that taking place in live epidermis, with regard to a variety of adhesion molecules and control mechanisms, which include local anti-acantholytic factors such as interleukin-10 (30). Needless to say, the integrity of the epidermal barrier in higher species relies on more than a single molecule. For example, inactivation of an adhesion molecule such as Dsg 3 does not lead to skin blisters and is well compatible with the normal life span of Dsg3 null mice (5,18), whereas a loss of immunological tolerance to keratinocyte self-antigens in PV is potentially lethal in 90% of patients (reviewed in Ref. 2). Therefore, to explain clinical and immunological correlations in PV, we propose a "multi-hit" hypothesis, which postulates that acantholysis in PV results from simultaneous and cumulative effects of autoantibodies directed toward different keratinocyte self-antigens, including the "structural" antigens, such as desmosomal cadherins, and "functional" antigens, such as cell surface receptors regulating function of the adhesion and cytoskeletal units.
The rationale behind our emphasis on the importance of "functional" targets of PV autoimmunity stems from recent discoveries of the genetic defects that underlie certain skin diseases. For instance, patients with genetic defects of the adhesion molecules Dsg1 and desmoplakin develop neither macroscopic nor light-or electron-microscopic alterations of keratinocyte cell-to-cell adhesion but produce instead a palmoplantar keratoderma, represented by linear and focal hyperkeratosis on palms and soles (81)(82)(83). In marked contrast, intra-epidermal split and PV-like skin lesions in patients with keratosis follicularis, or Darier-White disease, and patients with benign familial pemphigus, or Hailey-Hailey disease, result from a mutation in the genes coding for Ca 2ϩ pumps, the ATP2A2 and ATP2C1, respectively (84,85). Calcium metabolism in the epidermis of PV patients may also be altered. We have recently found that PV patients develop autoantibodies to the novel human ␣9 ACh receptor subunit that comprises ACh-gated Ca 2ϩ channels on the cell membrane of human KC (19).
In summary, in this study we identified PX, a novel annexinlike molecule, which can function as a keratinocyte cholinergic receptor mediating biological effects of ACh on KC, including regulation of cell-to-cell adhesion. PX is targeted by PV autoimmunity and may represent one of the major targets for acantholytic autoantibodies. Further studies should be directed to elucidate the biochemical mechanisms by which the anti-PX antibody alters keratinocyte adhesion in vitro and the biological effect(s) caused by cholinergic ligand binding to PX. Furthermore, because annexins are well known mediators of antiinflammatory effects of glucocorticosteroids in the skin (86), and because glucocorticosteroids can directly protect KC from the acantholytic effect of PV IgG in vitro (87), it will be important to elucidate possible relationships between the effects of glucocorticosteroids on PX and keratinocyte adhesion. Such an association may lead toward development of non-hormonal treatment of PV, because cholinergic drugs that, just like glucocorticosteroids, exhibit direct anti-acantholytic activity (20) may do so by competing with PV IgG for binding to PX on the cell membrane of KC.