A monoclonal antibody identifies a novel GPI

0 downloads 0 Views 3MB Size Report
(1988). Fibronectin and vitronectin regulate the organization of Arg-Gly-. Asp adhesion .... Cell Regul. 1, 37-. 44. Nicholson, L. J., Pei, X. F. and Watt, F. M. (1991).
1413

Journal of Cell Science 107, 1413-1428 (1994) Printed in Great Britain © The Company of Biologists Limited 1994

A monoclonal antibody identifies a novel GPI-anchored glycoprotein involved in epithelial intercellular adhesion Mauro Rabino1, Livio Trusolino1, Maria Prat1, Ottavio Cremona1, Paola Savoia1 and Pier Carlo Marchisio1,2,* 1Department of Biomedical Sciences and Human Oncology, University of Torino, 10126 Torino, 2DIBIT, Department of Biological and Technological Research, San Raffaele Scientific Institute,

Italy Via Olgettina 58, 20132 Milano,

Italy *Author for correspondence at address 2

SUMMARY BD31 mAb, raised against a line of gastric carcinoma cells, reacts with intercellular boundaries of human transformed cells originating from carcinomas or sarcomas growing in epithelial-like clusters as well as in primary cultures of epithelial and endothelial cells. BD31 also reacts with intercellular rims of normal and transformed epithelial tissues and is particularly abundant in glands and fast-growing epithelia but absent in nervous and muscle tissues as well as in blood and in mesenchyme-derived cells. Confocal analysis indicates that BD31 is located in dots at cell-cell contacts but not in basal and apical domains of cultured and in situ epithelial cells. mAb BD31 precipitates a 100 kDa protein from cells labeled with [35S]methionine or [3H]glucosamine as well as from 125I-surface-labeled cells. This glycoprotein resists to trypsin in the presence of Ca2+, releases an 80 kDa fragment in the medium and does not

react with antibodies to the conserved cytodomain of known cadherins or, specifically, to the ectodomain of Ecadherin in western blotting; moreover, the lack of cadherin cytodomain and protein removal by phosphoinositide-specific phospholipase indicate its membrane anchoring by a glycosyl-phosphatidylinositol (GPI) moiety. BD31 IgGs do not impair cell-matrix adhesion but induce inhibition of Ca2+-dependent aggregation, loss of cell-cell adhesion, scattering of confluent cells and appearance of migratory cell phenotypes in a dose- and time-dependent manner. This novel GPI-anchored glycoprotein may regulate intercellular adhesion by a mechanism involving membrane-associated phospholipases.

INTRODUCTION

All cadherins are structurally related and show remarkable tissue specificity. Human E-cadherin, also called CAM 120/80 (Damsky et al., 1983; Mansouri et al., 1988), is mainly expressed in human epithelia and is the homologue of other cadherins such as uvomorulin in mouse (Ringwald et al., 1987) and L-CAM in chicken (Gallin et al., 1983, 1987). P-cadherin was originally identified in mouse placenta but it has been shown to be distributed in many other tissues (Nose et al., 1987; Shimoyama et al., 1989), while cadherin 5 or VE-cadherin is selectively localized in endothelia (Lampugnani et al., 1992). Finally, N-cadherin is widely expressed in the nervous system, skeletal and cardiac muscle (Hatta et al., 1988). Recently, a COOH-terminal truncated form of cadherin has been found in chick embryo brain and has been named Tcadherin (Ranscht and Dours-Zimmermann, 1991; Vestal and Ranscht, 1992). T-cadherin is unusual, since it lacks the cytoplasmic domain and is therefore unable to link the cytoskeleton directly. Moreover, it is membrane-linked via a glycosylphosphatidylinositol (GPI) moiety and shares this feature with a family of GPI-anchored molecules described in many tissue types (reviewed e.g. by Ferguson and Williams, 1988). The Tcadherin mRNA has not been reported in mammalian tissues but predominantly in the avian brain and also in some other

Cell adhesion is an essential function controlling tissue organization during embryogenesis, wound repair, regeneration and maintenance of cell polarity (Ekblom et al., 1986; Fleming and Johnson, 1988) and its perturbation is crucial for the genesis and progression of tumours (for review see Ruoslahti and Giancotti, 1989). Cell-substratum and cell-cell adhesion phenomena are controlled by at least four superfamilies of membrane glycoproteins: (i) intercellular adhesion molecules with Ca2+-dependent homotypic binding (cadherins: Takeichi, 1990, 1991; Kemler, 1992); (ii) Ca2+-independent homotypic cell adhesion molecules belonging to the Ig superfamily (CAMs: Edelman and Crossin, 1991; Buck, 1992; Goridis and Brunet, 1992); (iii) integrins (for a recent review see Hynes, 1992); (iv) proteoglycans (Ruoslahti and Yamaguchi, 1991). Cadherins (Takeichi, 1990, 1991) form a molecular family involved in mutual cell association. They are transmembrane structural components of adhesion junctions (Volk and Geiger, 1984) where they link the actin microfilament network (Gumbiner and Simons, 1986; Hirano et al., 1987; Ozawa et al., 1989, 1990) via a chain of specific cytoskeletal proteins called catenins (Nagafuchi and Takeichi, 1989; Ozawa et al., 1989).

Key words: adhesion, epithelium, GPI, confocal, neoplastic cells cadherin

1414 M. Rabino and others embryonic avian tissues (Ranscht and Dours-Zimmermannn, 1991). In this paper we report the topographical, biochemical and functional characterization of a widely distributed glycoprotein, recognized by mAb BD31, originally raised against the gastric carcinoma cell line KATO III (Prat et al., 1986). This glycoprotein is found at intercellular contacts, is endowed with some cadherin-like properties, namely Ca2+-dependent trypsin resistance and direct involvement in Ca2+-dependent intercellular aggregation, and is anchored to the plasma membrane via a GPI moiety. Therefore, it may represent a novel form of truncated adhesion molecule with widespread distribution in many normal and neoplastic epithelial cells. Such GPIanchored truncated protein is enriched in the lateral adhesion domain but is not expressed in the apical domain like many other GPI-anchored proteins (reviewed by Lisanti and Rodriguez-Boulan, 1991; see, however, Zurzolo et al., 1993). Because of its membrane-anchoring it is not likely to transduce any direct signal to the cytoskeleton but, being functionally involved in intercellular adhesion, it may regulate tissue organization by a local mechanism mediated by phospholipases. MATERIALS AND METHODS Cell cultures FG-2 pancreatic carcinoma cells were a gift from V. Quaranta, Scripps Research Institute, La Jolla, CA; the keratinocyte cell line NCTC 1810 was provided by M. Del Rosso, University of Firenze, Italy. The other cell lines were obtained from American Type Culture Collection (Rockville, MD). All cell lines were cultured in RPMI 1640 medium (Biochrom KG, Berlin, Germany), supplemented with 10% fetal calf serum (Biochrom KG), 4 mM glutamine and penicillinstreptomycin (50 i.u./ml). They were maintained at 37°C in a humidified atmosphere of 5% CO2. Primary human umbilical cord vein endothelial cells were a gift from E. Dejana, Istituto Mario Negri, Milano, Italy; human skin keratinocytes and fibroblasts were kindly provided by M. De Luca, IST, Genova, Italy; human thymus epithelial cells were prepared by D. Ramarli, University of Verona, Italy; human kidney mesangial and tubular cells were provided by R. Coppo, Division of Nephrology, Ospedale Infantile Regina Margherita, Torino, Italy. Antibodies The murine mAb BD31, of IgG1k isotype, was obtained following two immunizations with the paraformaldehyde-fixed gastric carcinoma cell line KATO III, as previously described (Prat et al., 1987). Four days after the last injection, immune spleen leukocytes were fused with P3.X63.Ag8.653 myeloma cells. Two weeks later hybrid supernatants were screened for their ability to bind to glutaraldehyde-immobilized KATO III cells in a solid-state immunoenzymatic assay, using an affinity chromatography-purified goat antimouse immunoglobulin preparation labeled with horseradish peroxidase (KPL, Gaithersburg, VA). The reaction was visualized by adding orthophenylene-diamine as chromogen (Sorin Biomedica, Saluggia, Italy) and read in a Titertek Multiscan (Flow Laboratories, Irvine, Scotland) at a 492 nm wavelength. BD31 ascitic fluid was purified by ammonium sulphate precipitation and loading onto a Protein A column in high salt and pH conditions; the crude antibody preparation was adjusted to 3.3 M NaCl in the presence of 1/10 volume of 1 M sodium borate (pH 8.9); bound material was eluted with 100 mM glycine (pH 3.0) and the eluate was neutralized with 1/10 volume of 1 M Tris-HCl (pH 8.0). The polyclonal antiserum R14 was a gift from G. Levi, Ecole

Normale Superieure, Paris, France; it was raised against a synthetic peptide corresponding to a region of high homology of the cytoplasmic domain of all cadherins (Geiger et al., 1990). Other mAbs and their source were: 4B4 to the integrin chain β1 (Coulter Immunology, Hialeah, FL); 7B4 to human V-cadherin (gift from E. Dejana; Lampugnani et al., 1992); hVIN 1 to human vinculin (Sigma Immunochemicals, St Louis, MO); DP 2.15 to desmoplakins 1 and 2 (ICN ImmunoBiologicals, Lisle, IL); DO-24 to a human c-Met extracellular epitope (Prat et al., 1991); HECD-1 to human E-cadherin (Shimoyama et al., 1989; purchased from Takara Shuzo Co., LTD, Kyoto, Japan). Finally, rabbit villin and fimbrin sera were obtained from K. Weber, Max-Planck-Institut für biophysikalische Chemie, Göttingen, Germany. Cell treatments Trypsin treatment To prepare the trypsin fragment, cells were treated as previously described (Cunningham et al., 1984). Briefly, surface 125I-labeled cells were rinsed in a buffer containing 150 mM NaCl, 5 mM KCl, 0.6 mM MgSO4.7H2O, 1 mM CaCl2.2H2O, 10 mM HEPES, pH 7.4. Cells were then incubated for 30 minutes at room temperature with 100 µg of trypsin (type IX; Sigma Chemical Co., St Louis, MO) dissolved in 2 ml of buffer. The reaction was stopped by addition of 20 µl of phenylmethylsulphonyl fluoride (2 mM PMSF in ethanol). After centrifugation of the medium the fragment was immunoprecipitated from the trypsin-released material. Trypsin and Ca2+ sensitivity of BD31 antigen were also tested by incubating FG-2 cells at 37°C for 1 hour with the following treatments: 1 mM EGTA, 0.0001% trypsin plus 1 mM EGTA, 0.01% trypsin plus 1 mM EGTA, or 0.01% trypsin plus 1 mM CaCl2 in HEPES-buffered HBSS (HHBSS, pH 7.4). Cells were then harvested, washed twice with PBS (phosphate-buffered saline) containing Ca2+ and Mg2+ and extracted (see below). Trypsin-treated cells were suspended in 0.05% soybean trypsin inhibitor (Sigma Chemical Co.) in the above HHBSS buffer prior to washing. PI-PLC treatment To perform phospholipase digestion, surface 125I-labeled cells were incubated for 60 minutes at 37°C with 2 units/ml phosphatidylinositol-specific phospholipase C (PI-PLC) from Bacillus cereus (Boehringer Mannheim, Germany) in PBS containing a mixture of protease inhibitors (the same used for immunoprecipitations, see below). The culture supernatants were collected, deprived of cellular debris by centrifugation, and immunoprecipitated with mAb BD31. In some experiments, 5 mM CaCl2, known to inhibit PI-PLC activity, was included as a control during digestion. Other control samples were incubated only in PBS and protease inhibitors. Neuraminidase treatment Neuraminidase (1 unit, from Vibrio cholerae, type II; Sigma Chemical Co.) was added as delivered by the manufacturer (aqueous solution, pH 5.5) to a 6 cm diameter culture plate of confluent [35S]methionine-labeled FG-2 cells (see below), for 45 minutes at 37°C before cell processing for immunoprecipitation. Cell labeling and extraction Metabolic labeling Subconfluent FG-2 cells were washed twice with methionine-free medium (Flow Laboratories) supplemented with 10% fetal calf serum, glutamine and antibiotics and cultured for 5 hours in the same medium. This medium was then removed and replaced by 2 ml of the same methionine-free medium in the presence of 100 µCi/ml [35S]methionine (Amersham Corp., Arlington Heights, IL). In the experiments aimed at studying intracellular glycosylation, tunicamycin (Sigma Chemical Co.) was added to the unlabeled pre-incubation medium and to the [35S]methionine-labeled medium at a con-

Novel GPI-anchored adhesion protein 1415 centration of 20 µg/ml. After overnight incubation in the radioactive medium, cell layers were washed three times with PBS and extracted on ice for 20 minutes with 1 ml of 0.15 M NaCl, 50 mM Tris-HCl, pH 8.5, 1% sodium deoxycholate, 1% Triton X-100, 0.1% SDS, 0.02% sodium azide, 2 mM PMSF (phenylmethylsulphonyl fluoride), leupeptin (50 µg/ml), pepstatin (5 µg/ml), aprotinin (20 units/ml) (according to Kajiji et al., 1989). Extracts were centrifuged for 15 minutes at 15,000 g at 4°C and supernatants were used for immunoprecipitations. For [3H]glucosamine metabolic labeling, experiments were performed using the same method as for [35S]methionine. The medium used was glucose-free and the amounts of [3H]glucosamine (Amersham Corp.) were 100 µCi for each ml of glucose-free medium. 125Iodine-surface

labeling FG-2 cells that had grown to confluency were washed three times with PBS. Each cell plate was incubated with 1 ml PBS containing 20 µl lactoperoxidase (5 mg/ml, from Boehringer Mannheim), 10 µl glucose oxidase (50 units/ml, from Boehringer Mannheim), 1 mCi 125iodine (Amersham Co.), 10 µl 0.25 M glucose for 5 minutes at room temperature. Another 10 µl sample of 0.25 M glucose was added and the reaction continued for 5 more minutes. The reaction was stopped with 2 ml of RPMI medium containing 0.02% sodium azide. The cell layer was washed three more times with RPMI containing 0.02% sodium azide and extracted as described above. In order to carry out V8 protease digestion experiments, bands immunoprecipitated from surface 125I-labeled extracts were excised, solubilized with a buffer containing 125 mM Tris-HCl, pH 6.8, 0.1% SDS, 5% β-mercaptoethanol, 1 mM EDTA, 500 ng V8 protease (Boehringer Mannheim), and re-run using SDS-PAGE. Immunoprecipitation Radiolabeled lysates were pre-cleared onto Protein A-Sepharose (Pharmacia LKB Biotechnology Inc., Uppsala, Sweden) coupled to normal rabbit serum (Flow Laboratories) for 1 hour at 4°C. The supernatants were collected and incubated with monoclonal or polyclonal antibodies for 3 hours at 4°C. The immunocomplexes were absorbed onto Protein A-Sepharose, previously reacted with affinity-purified rabbit anti-mouse Igs (Dakopatts, Copenhagen, Denmark) if necessary, for 1 hour at 4°C. After several washes in PORT buffer (10 mM Tris-HCl, pH 8.5, 0.15 M NaCl, 0.5% Tween-20, 0.1% Renex 30, 2.5 mM sodium azide, 0.1% ovalbumin), samples were eluted in Laemmli buffer (Laemmli, 1970) with or without β-mercaptoethanol. Proteins were separated by SDS-PAGE, fixed, fluorographed when required, dried, and exposed to Amersham Hyperfilm for autoradiography. Protein size was estimated using markers such as myosin (200 kDa), phosphorylase b (92.5 kDa), bovine serum albumin (69 kDa), egg albumin (46 kDa), and carbonic anhydrase (30 kDa) that had been prelabeled by [14C]methylation (Amersham Co.). Immunoblotting Cells were solubilized in the boiling buffer described by Laemmli (1970), in the presence or in the absence of the reducing agent β-mercaptoethanol. Equal amounts of proteins (300 µg) were loaded into all lanes. Alternatively, immunoprecipitates prepared as described above were eluted with Laemmli buffer and loaded. After SDSPAGE, proteins were transferred to nitrocellulose filters (Hybond, Amersham Corp.) and analyzed as described (Towbin et al., 1979). Filters were then probed either with the polyclonal serum R14 or with mAb HECD-1. Specific binding was detected by the enhanced chemiluminescence system (ECL, Amersham Corp.). Immunocyto- and immunohistochemistry Indirect immunofluorescence microscopy Subconfluent cells from different cell lines (see Table 1) were plated onto 24-well Costar plates containing 1.1 cm2 round glass coverslips.

Table 1. Reactivity of mAb BD31 in different cultured cells Cells

Origin

Reactivity

Human FG CACO 2 HT 29 CALU 3 MCF 7 HT-1080 SAOS-2 HOS NCTC 18-10 WI-38

Pancreatic adenocarcinoma Colon adenocarcinoma Colon adenocarcinoma Lung adenocarcinoma Breast adenocarcinoma Fibrosarcoma Osteosarcoma Osteosarcoma Keratinocyte cell line Lung fibroblast cell line

Non human MDCK COS NIH-3T3

Dog kidney Monkey kidney Mouse fibroblasts

Human primary cultures Endothelial cells Keratinocytes Thymic epithelial cells Blood cells Fibroblasts Smooth muscle cells Kidney mesanglial cells Kidney tubular cells

Umbilical vein Skin biopsy Thymus Bone marrow Dermis Artery Kidney biopsy Kidney biopsy

+++ ++ +++ + ++ ++ ++ ++ ++ + ++ +++ ++

Table 2. Reactivity of mAb BD31 in normal and neoplastic tissues Tissue Epidermis Spinous layer Basal layer Hair follicles Sebaceous glands Sweat glands Blood vessels and placenta Endothelium Smooth muscle layer Placenta

Reactivity + ++ ++ ++ -

Glands Tubular uterine glands Mammary gland Colon glands Thyroid Liver

++ ++ + + ++

Kidney Tubular epithelia Glomeruli

+ -

Nervous system Cerebrum Cerebellum Peripheral nerves

-

Muscular tissues Smooth muscle cells Skeletal muscle fibers Cardiac muscle fibers

-

Tumors Basal cell carcinomas Squamous cell carcinomas Melanomas Colon carcinomas Thyroid papillary and follicular carcinomas Endometrium carcinomas

++ ++ +++ +++

1416 M. Rabino and others

Fig. 1. BD31 antigen detected by indirect immunofluorescence in FG-2 (b,e) and HT29 cells (d) simultaneously stained for F-actin with F-PHD (a,c): the antigen appears to be concentrated at the boundaries between closely apposed cells and is never detectable along free cell edges; the fluorescence pattern consists of a thin sharp continuous line highlighting the margins of each cell; this line is occasionally interrupted (e.g. a,b, at arrowhead). Detergent treatment removes most immunoreactivity (e). The BD31 immunoreactivity pattern is different from the pattern obtained with mAb to vinculin (f,g) or to desmoplakins (h) in FG-2 cells. In some panels the focal plane was deliberately raised from the plane of substratum adhesion to highlight intercellular adhesion structures; since the high power lens used for taking these pictures has a limited depth of focus and the cells are rather cuboidal some cellular structures are inevitably blurred. Bar, 10 µm. Coverslip-attached cells were fixed in 3% formaldehyde (from paraformaldehyde) in PBS, pH 7.6, containing 2% sucrose for 5 minutes at room temperature. After rinsing in PBS, cells were either used as such (non-permeabilized) or permeabilized by soaking coverslips for 3-5 minutes in HEPES-Triton X-100 buffer (20 mM HEPES, pH 7.4, 300 mM sucrose, 50 mM NaCl, 3 mM MgCl2, 0.5%

Fig. 2. Confocal microscopy analysis of BD31 immunoreactivity in FG-2 cells. A single optical section obtained along the z-axis across a colony of BD31-stained FG-2 cells shows that the antigen is organized in a dot-like intercellular pattern, suggesting its association with discrete molecular clusters. Confocal microscopy avoids the out-of-focus problems encountered with cuboidal FG-2 cells (see Fig. 1).

Triton X-100). This procedure of fixation and permeabilization permits immunostaining of cytoskeletal and adhesion components (Marchisio et al., 1984). In some tests fixation was carried out by soaking coverslips in methanol (5 minutes, −20°C) followed by acetone (5 seconds, −20°C). Indirect single-label and double-label immunofluorescence experiments were performed as reported (Marchisio et al., 1984; Dejana et al., 1988). Briefly, the primary antibody was layered onto fixed and permeabilized cells and incubated in a moist chamber for 30 minutes. After rinsing in 0.4% TBS-BSA (Tris-buffered saline plus 0.4% bovine serum albumin), coverslips were incubated with the appropriate rhodamine-tagged secondary antibody (Dakopatts) for 30 minutes at 37°C in the presence of 2 µg/ml of fluorescein-labeled phalloidin (F-PHD; Sigma Chemical Co.). Coverslips were mounted in Mowiol 4-88 (Hoechst AG, Frankfurt/Main, Germany). Routine observations were carried out in a Zeiss Axiophot pho-

Fig. 3. Confocal microscopy analysis of a cluster of FG-2 cells double-stained for F-actin with F-PHD (pseudocolour blue) and simultaneously with mAb BD31 coupled to a rhodamine-tagged secondary antibody (pseudocolour white). A stack of serial optical sections, every 0.7 µm along the z-axis, were obtained by scanning with the argon laser source with a combination of filters minimizing the overlapping of fluorescein and rhodamine light emission spectra. Individual z-axis digitized optical sections were stored, individually retrieved and three-dimensionally reconstructed with the VANIS™ software in a Silicon Graphics computer. The reconstructions were stepwise-rotated (10° each step) by 180° to show that the location of the BD31 antigen was restricted at the contact boundary between individual cells (frame at 90°) and was virtually excluded either from apical surface or from basal contact domains. Note that the frames at 0° and 180° are mirror images of the same object.

Novel GPI-anchored adhesion protein 1417

1418 M. Rabino and others tomicroscope equipped for epifluorescence. Fluorescence images were recorded on Kodak T-Max 400 films exposed at 1000 ISO and developed in T-Max Developer for 10 minutes at 20°C. Confocal laser scanning microscopy was performed in a Sarastro 2000 (Molecular Dynamics, Sunnyvale, CA) laser scanning system attached to a Zeiss Axioskop fluorescence microscope and using a Zeiss planapochromatic lens ×63, NA 1.4. Image processing and three-dimensional reconstruction analysis were performed using the Vanis™ program running in a Silicon Graphics Personal Iris computer (Silicon Graphics). Stored images were recorded from the screen of a computer-coupled color image recorder (Avionics Corporation Ltd., Japan) using either Kodak T-Max 100 black-and-white films or Ektachrome 100 color slide films. Indirect immunoperoxidase technique Small samples of normal human tissues or tumors obtained from surgical biopsies (see Table 2) were mounted in OCT 4583 embedding compound (Miles Scientific, Naperville, IL), frozen in liquid nitrogen-cooled isopentane and stored in a −80°C ultrafreezer. Frozen serial sections were cut at 6 µm in a Leitz cryomicrotome, transferred onto microscope slides coated with poly-L-lysine (Sigma Chemical Co.), air-dried and stored at room temperature overnight. The samples were then fixed for 10 minutes in a chloroform-acetone mixture (1:1, v/v), air dried and incubated for 10 minutes in PBS supplemented with 1% serum of the same species as the secondary antibody. Serial sections were overlaid with 50 µl of undiluted supernatant or ascitic fluid at a 1:40 dilution in 0.4% TBS-BSA and incubated at room temperature for 30 minutes in a moist chamber. After a thorough wash in PBS, the sections were incubated with the appropriate biotinylated secondary antibodies and processed for the ABC method (avidin-biotin-peroxidase complex) using the Vectastain ABC Kit (Vector Laboratories Inc., Burlingame, CA). After several washes, 100 µl of substrate were added for 5 to 10 minutes, prepared as follows: 5 mg 3-amino-9-ethylcarbazole (Sigma Chemical Co.) was dissolved in 1 ml N,N-dimethylformamide (Merck, Darmstad, Germany) supplemented with 9 ml 100 mM sodium acetate, pH 5.2, and 100 µl of 12% H2O2. All samples were counterstained with Mayer’s haemalum solution, mounted in Kaiser’s glycerol gelatin (Merck) and examined with a Zeiss Axiophot photomicroscope equipped with planapochromatic lenses. Adhesion inhibition and aggregation assays For adhesion inhibition assays a previously employed test was used (De Luca et al., 1990). Briefly, 96-well plates were coated for 1 hour at 37°C with mouse laminin (10 µg/ml, Sigma Chemical Co.) or human plasma fibronectin (10 µg/ml; purified from a gelatin column). FG-2 or HT29 cells were plated (3×104 cells per well) in serum-free RPMI medium and mAb BD31-purified IgGs, control mouse IgGs or unrelated IgGs were added. IgGs were used at the concentrations of 1.5, 3.0, 6.0 and 12.0 µg/ml. Cells were incubated for 2 hours at 37°C, washed, fixed in 3% formaldehyde, 2% sucrose in PBS for 20 minutes, stained with 0.5% crystal violet, 20% methanol for 15 minutes, washed and dried. The dry dye was eluted with 0.1 M sodium citrate in 50% ethanol, pH 4.2. Absorbance was read in a Titertek Multiscan (Flow Laboratories) at 540 nm wavelength. We also studied cell cohesion as previously described (Marchisio et al., 1991). Briefly, in a set of experiments, FG-2 cells, grown to confluency on 24-well Costar plates, were incubated for 4 hours with serum-free RPMI medium supplemented with increasing concentrations of BD31-purified IgGs or Fab fragments (1.5 to 25 µg/ml). In parallel sets of experiments cell cohesion was studied as a function of time of exposure to BD31 IgGs and monitored for times ranging from 0.5 to 24 hours with a single concentration of IgGs (3 µg/ml). In other experiments FG-2 cells grown to confluency were trypsinized and the cell suspension was immediately replated in the presence of two concentrations of BD31 IgGs (3 µg/ml and 6 µg/ml); cells were then monitored hourly for 19 hours. Control cultures received equal

amounts of mouse pre-immune IgGs. Cells were then observed in phase-contrast microscopy or fixed, stained for F-actin by F-PHD and observed in the fluorescence microscope. For quantitative aggregation analysis, experiments were performed as previously described (Urushihara et al., 1979) with minor modifications. Briefly, to obtain a suspension of dispersed cells with intact Ca2+-dependent molecules, cells were treated with 0.01% trypsin and 10 mM CaCl2 in HHBSS at 37°C for 60 minutes on a gyratory shaker at 100 rpm. Cells were then rinsed twice in HHBSS supplemented with 10 mM CaCl2 and flushed several times through a Pasteur pipette. To prepare cells without cadherins, cells were treated with 0.01% trypsin in the presence of 1 mM EGTA at 37°C for 30 minutes. A total of 105 cells suspended in 500 µl of HHBSS were placed in each well of a 24-well plastic plate and incubated at 37°C for 50 minutes in the following conditions: HHBSS; HHBSS plus 10 mM CaCl2; HHBSS plus 10 mM CaCl2 plus increasing concentrations of BD31 Fab fragments (3, 6, 12 and 24 µg/ml). After incubation, samples were removed and diluted immediately in Isoton II (Coulter Electronics Inc., Hialeah, FL), and then counted in an automated counter (model Z1, Coulter Electronics Inc.). The extent of cell aggregation was represented by the index (N0−Nt)/N0, where Nt is the total particle number after the incubation time t and N0 is the total particle number at the start of incubation. Controls of cell viability during cohesion assays were routinely carried out. All assays were repeated a minimum of three times, with consistent results. Isolation of Fab fragments was performed by papain digestion (Pierce Chemical Co., Rockford, IL) according to Parham (1986). These aggregation and cohesion tests were intended to monitor whether intercellular adhesion was affected by mAb BD31.

RESULTS Localization of BD31 antigen in different normal and transformed cells Cultured cell monolayers were examined by immunofluorescence microscopy using mAb BD31. Several cell types were studied (see Table 1) but we mostly report data obtained on FG-2 cells originating from a human pancreatic carcinoma or on HT29 colon carcinoma cells. As shown in Fig. 1a-d, mAb BD31 reacted with an antigen located at the boundaries between closely apposed cells. The antigen was detected only at contacting cell borders and never found along contact-free cell edges. In conventional immunofluorescence microscopy the staining pattern mostly consisted of a thin sharp line highlighting the contact margins of each cell but this line occasionally looked discontinuous (Fig. 1b and d, e.g. at arrowheads), suggesting that BD31 antigen might be clustered in discrete sites. Indeed, a more accurate observation performed by optical sectioning with a confocal microscope along the zaxis and thus avoiding out-of focus problems revealed that the antigen was organized in tandemly organized dots (Fig. 2), and thus supported its possible location in discrete spots or even associated to junctions. Confocal microscopy was also used to study the topography of BD31 antigen all around the cell. By using the VANIS™ program, designed for the three-dimensional reconstruction and rotation of z-axis serial optical sections collected across a small colony of BD31-stained FG-2 cells, we have been able to show that this antigen is enriched at cell-cell contacts and absent either from the cell-substratum contact surface or from the apical domain (Fig. 3; note the 90° rotation frame). This immunostaining pattern was quite labile and was

Novel GPI-anchored adhesion protein 1419 strongly reduced or abolished in cells treated with the detergent-containing buffers used to permeabilize cell membranes to Igs, thus indicating that the epitope was extractable by mild detergents (Fig. 1e). The immunoreactivity was also abolished by methanol-acetone extraction (not shown). BD31 staining pattern was different from that obtained on the same cells with antibodies to vinculin (Fig. 1f and g) or to desmoplakins (Fig. 1h). This intercellular staining pattern was observed in many transformed human cell types (Table 1) forming epithelial-like cell clusters. Most positive cells produced and assembled keratins but some, such as HT 1080, SAOS 2 and HOS, did not. WI-38 human fibroblasts were negative for both BD31 antigen and keratins. Epithelial-like assembly but not the expression of epithelial markers seems to be associated with BD31 antigen positivity. Primary cultures of keratinocytes, endothelial cells, thymic epithelial cells and kidney tubular cells, all showing an epithelial histotype, displayed intercellular positivity of variable intensity (Fig. 4). This immunoreactivity was detergent-sensitive and differently localized compared to that of other intercellular molecules. On the other hand, fibroblasts, bone marrow-derived cells, kidney mesangial cells and smooth muscle cells were all negative. All the non-human cells tested were negative as well, indicating that the BD31 epitope is species-specific. Expression of BD31 antigen in tissues Immunoperoxidase staining was used to examine the expression and localization of BD31 antigen in different normal and tumor tissues (Table 2 and Fig. 5). In the epidermis the antigen was irregulary detected along

intercellular contacts mostly in suprabasal layers (Fig. 5a). On the contrary, epidermal appendages showed a strong reactivity: hair follicles were remarkably stained in their central area (Fig. 5b), which corresponds to the hair matrix, i.e. to a region containing immature proliferating cells involved in hair growth. Sebaceous (Fig. 5c) and sweat glands (Fig. 5d) also showed intensely positive intercellular boundaries. Other secreting epithelia, e.g. tubular uterine glands (Fig. 5e), Fallopian tube glands (Fig. 5f), thyroid gland (Fig. 5g), liver (Fig. 5h), renal tubules (Fig. 5i), large intestine mucosa (Fig. 5j), and mammary alveoli and ducts (not shown), expressed intercellular BD31 antigen at variable levels. Blood vessels were apparently negative in all the tissues examined, indicating that positivity to mAb BD31 may be a feature of cultured proliferating endothelial cells (see Fig. 4f). Cardiac muscle cells, which display a well-developed adhesion complex, were negative (Fig. 5k). The expression of the antigen was also studied in some tumors. We found that basal cell carcinomas (Fig. 5l), derived from glandular cell precursors (Lever and Schaumburg-Lever, 1990), were strongly and specifically stained, whereas squamous cell carcinomas in different organs, as well as melanomas, were negative (not shown). Finally, strong to weak irregular positivity was also associated with colorectal adenocarcinomas observed at different stages of neoplastic progression (unpublished data). The strictly lateral location of BD31 antigen was checked also in colon epithelium in situ and found to be different from the apical location of fimbrin and villin, two structural markers of the apical brush border (not shown).

Fig. 4. BD31 antigen detected by indirect immunofluorescence in keratinocytes (a,b), thymic epithelial cells (e), endothelial cells (f,g) and dermal fibroblasts (h). In keratinocyte colonies the antigen appears to be enriched at the boundaries between closely apposed cells and is never detectable along free edges (a, e.g. at arrowhead); detergent treatment virtually removes immunoreactivity (b). The fluorescence pattern is similar to that obtained by immunostaining with a mAb to to the β1 integrin chain (d, e.g. at arrowhead) but is different from the desmoplakin pattern (c). Human thymic epithelial (e) and endothelial cells (f) are also stained along contact margins; the latter cells are similarly stained for comparison with a mAb to the endothelium-specific cadherin 5 or VE-cadherin (g). Fibroblasts are BD31-negative (h). Bar, 10 µm.

1420 M. Rabino and others

Fig. 5. Immunoperoxidase staining of different normal and malignant tissues: in normal epidermis the antigen was weakly and irregulary detected along intercellular contacts, particularly in suprabasal layers (a, e.g. at arrowheads). In contrast, epidermal appendages showed strong reactivity: hair follicles were remarkably stained at the hair matrix (b, indicated by arrowheads); the antigen was also detected in the typical swollen cells of sebaceous glands (c) and in cells lining eccrine sweat glands (d). Other secreting epithelia expressed BD31 antigen at high levels: e.g. tubular uterine glands (e), Fallopian tube glands (f), thyroid gland (g), liver (h), renal tubules (i), and intestinal mucosa (j). The adhesion complex of cardiac muscle cells was negative (k). Among cutaneous tumours, a strong and specific pericellular staining was observed in basal cell carcinomas (l) but not in squamous cell carcinomas. Blood vessels, peripheral nerves and stromal tissues were negative. Bar, 3 µm; for the left part of (k), 10 µm.

Biochemical characterization of BD31 antigen Immunoprecipitation of extracts obtained from [35S]methionine-labeled FG-2 cells was employed to identify the molecule recognized by mAb BD31 (Fig. 6). After extraction in lysis buffer containing a mixture of ionic and non-ionic detergents (1% sodium deoxicholate, 0.1% SDS and 1% Triton X-100) a single broad band of 100 kDa apparent molecular mass was detected under non-reducing conditions (Fig. 6A). The protein was also fully extracted by non-ionic detergents (e.g. 0.5% Triton X-100). This property confirms the total extraction of the BD31 antigen observed in immunofluorescence upon treatment with HEPES-Triton permeabilization buffer.

An identical band was also detected after metabolic labeling with [3H]glucosamine, indicating the glycoprotein nature of the BD31 antigen. Incubation of cells with tunicamycin during [35S]methionine labeling yielded a major protein band with an apparent molecular mass of 75 kDa, which may represent the approximate size of the core protein. Neuraminidase treatment upon metabolic labeling gave a broad band of slightly reduced size, indicating that the BD31 antigen bears sialic acid-containing oligosaccharide chains. To show the external exposure of the BD31 glycoprotein antigen, immunoprecipitation experiments were performed upon cell surface radioiodination. A rather sharp 100 kDa band

Novel GPI-anchored adhesion protein 1421 was observed in 125I-labeled immunoprecipitates (Fig. 6A). Running the BD31 immunoprecipitate in the presence of βmercaptoethanol under reducing conditions induces an apparent increase in molecular mass (about 15 kDa), indicating the existence of intrachain disulphide bonds (Fig. 6B). Surface iodination allows the detection of a minor 80 kDa band, which was considered to be a proteolytic product of the 100 kDa native form. In fact, V8 protease digestion of bands obtained by immunoprecipitation of either the 100 or

the 80 kDa band gave essentially the same peptide pattern (Fig. 6C). The 100 kDa band was immunoprecipitated in different relative quantities from all the cell lines that were BD31positive in immunofluorescence (Fig. 7). Addition of the antibody either to detergent lysates or to intact cells followed by extensive washing and detergent lysis during the immunoprecipitation procedure gave identical results, thus confirming the external exposure of the epitope recognized by mAb BD31 (not shown). Comparison of BD31 antigen with other cell-cell adhesion molecules The localization of BD31 antigen at cell-cell contacts and its structural properties suggest that this glycoprotein is related to other molecules involved in intercellular adhesion. The integrins found along intercellular boundaries of keratinocytes are α2β1 and α3β1 (e.g. see Larjava et al., 1990; De Luca et al., 1990; Marchisio et al., 1991). In immunofluorescence experiments performed using anti-β1 antibodies on FG2 cells we localized these integrins at cell-cell contacts as well as in other membrane domains. In immunoprecipitation experiments performed on FG-2 cell lysates with mAb 4B4 two major bands of 110 and 150 kDa, which represented the β1 chain and one of the α chains, respectively, were precipitated (Fig. 8A). We did not further investigate β1-associated α chains expressed by FG-2 cells, but just confirmed that FG-2 cells also express high amounts of α6β4 endowed with a different electrophoretic mobility (not shown; see Kajiji et al., 1989). The possibility may thus be reasonably excluded that BD31 antigen belongs to the integrin superfamily.

Fig. 6. (A) BD31 immunoprecipitation analysis of metabolically and surface-labeled FG-2 cell extracts. The molecule immunoprecipitated by mAb BD31 is a 100 kDa glycoprotein (shown by the incorporation of [35S]methionine and [3H]glucosamine) that is surface exposed, since it is labeled by lactoperoxidase-catalyzed 125I iodination; treatment with tunicamycin during metabolic labeling produces a considerable downshifting in molecular mass to about 75 kDa; neuraminidase treatment of [35S]methionine-prelabeled cells induces a slight decrease in apparent molecular mass, indicating moderate sialylation. (B) Different electrophoretic mobility of BD31 antigen under reducing and non-reducing conditions. Bands immunoprecipitated in non-reducing conditions from surface 125Ilabeled extracts were excised, solubilized with Laemmli buffer in the presence of the reducing agent β-mercaptoethanol and re-run in SDS-PAGE; this treatment increases the apparent mobility of the band, suggesting reduction of intrachain disulphide bonds. Molecular mass markers are indicated on the left. (C) The minor 80 kDa band detected in surface radioiodination experiments is a proteolytic fragment of the 100 kDa form. Bands immunoprecipitated from 125Ilabeled extracts were excised, digested with V8 protease and re-run in SDS-PAGE; the peptides derived from the enzymic digestion show comparable elecrophoretic mobility. Arrows indicate bands representing minor undigested portions of the 100 kDa native form and its 80 kDa derivative. Molecular mass markers are indicated on the right in kDa.

Fig. 7. Immunoprecipitation of BD31 antigen from [35S]methioninelabeled extracts from different tumour cell lines grown at subconfluency The extracts were normalized for incorporation of radioactivity into proteins (2×107 cpm). The protein is expressed at variable extent in epithelial and non-epithelial transformed cells displaying epithelioid organization. Molecular mass markers are indicated on the left in kDa.

1422 M. Rabino and others Next, we tested whether the BD31 antigen might be a member of the superfamily of cell adhesion molecules linked by homotypic interactions. First, we carried out immunoprecipitation with R14 serum that was raised to a conserved Cterminal peptide of cadherins (Geiger et al., 1990). A band of molecular mass higher than that recognized by mAb BD31 was immunoprecipitated (Fig. 8A), demonstrating that at least one classical cadherin is expressed by these cells. To test the Ca2+-dependent trypsin resistance of this molecule (Fig. 8B) we treated metabolically labeled FG-2 cells with the Ca2+ chelator EGTA (lane E), with trypsin plus Ca2+ (lane TC), with low trypsin plus EGTA (lane LTE), and with trypsin plus EGTA (TE). Indeed the BD31 antigen was sensitive to trypsin in a Ca2+-dependent manner, since the 100 kDa band was preserved with high Ca2+, but was strongly reduced in the presence of EGTA even at very low trypsin concentration (0.0001%). The amount of [35S]methionine-labeled protein immunoprecipitated by mAb BD31 was comparable in EGTA-treated and in control cells; immunofluorescence experiments show that EGTA treatment induces a redistribution of the protein on the cell membrane but not its disappearance (not shown). The same kind of treatment was performed on surface-iodinated cells and confirmed the Ca2+ sensitivity of the BD31 antigen (Fig. 8C). These experiments allowed the detection of a trypsin-cleaved proteolytic fragment of 80 kDa that was released in the culture medium. Such a degradation fragment was often observed in detergent extracts of cells in which Ca2+ was kept at low concentration because of the high sensitivity of this molecule to proteolytic degradation. BD31 antigen is a GPI-anchored glycoprotein BD31 glycoprotein does not share the conserved COOH-terminal domain expressed by cadherins (Fig. 9A). In fact the 100 kDa band immunoprecipitated by BD31 and transferred onto a nitrocellulose membrane was not decorated in western blotting by the rabbit antiserum R14, raised against a synthetic peptide corresponding to a region of high homology of the cytoplasmic domain of all cadherins (Geiger et al., 1990). Moreover, mAb HECD-1, raised against human Ecadherin, did not recognize the band immunoprecipitated by mAb BD31 (Fig. 9B). Conversely, probing blots of the total cell extract or immunoprecipitates obtained with R14 as well as with HECD-1 yielded bands that were detected by the same antibodies in western blotting. We also found that the BD31 epitope is sensitive to the denaturing treatment required for western blot analysis. The above biochemical features of the BD31 antigen indicated that we were dealing with a molecule devoid of the COOH-terminal domain shared by known cadherins and also devoid of reactivity to at least one epitope of E-cadherin ectodomain identified by mAb HECD-1. Its sensi-

tivity to membrane extraction conditions suggested that it could be truncated and GPI-anchored. To investigate GPIanchoring (Fig. 9C), radioiodinated cells were treated with phosphatidylinositol-specific phospholipase C (PI-PLC), an enzyme that releases GPI-linked proteins from the plasma membrane (Low and Saltiel, 1988). Treatment with PI-PLC (2 units/ml for 60 minutes at 37°C) removed a large fraction of the 100 kDa glycoprotein from the cell surface. The released labeled glycoprotein was detected by immunoprecipitation in the culture medium after PI-PLC treatment while, in untreated cells, the protein was immunoprecipitated only from the total cell extract and was not appreciably detected in the culture medium. As a negative control, the COOH-terminal truncated form of the Met/HGF receptor, which is not anchored by a GPI, was completely insensitive to PI-PLC treatment and persisted on the cell membrane after the incubation (not shown). These

Fig. 8. Immunoprecipitation of BD31 antigen from [35S]methionine-labeled FG2 cell extracts (A and B) and from surface 125I-labeled FG-2 cells (C). (A) Comparison of BD31 antigen with other molecules of cell-cell contacts. Immunoprecipitation of FG-2 cell extracts with R14, a rabbit serum directed against the conserved COOH-terminal end of cadherins, indicates that epithelial cadherins are expressed by FG-2 cells and that their electrophoretic mobilty is slower than that of BD31 antigen; mAb 4B4 against the β1 integrin chain precipitates integrin heterodimers belonging to the β1 subfamily (α2β1 and α3β1) that have been identified at intercellular contacts, yielding a band pattern different from that of BD31 antigen. (B) BD31 antigen is protected from trypsin digestion by Ca2+. Equal numbers of metabolically labeled FG-2 cells were treated with low (0.0001%) or high (0.01%) amounts of trypsin in the presence of 1 mM EGTA (LTE and TE, respectively) or with 0.01% trypsin in the presence of 1 mM CaCl2 (TC) and processed for immunoprecipitation. Control samples were untreated cells (Control) or cells treated with 1 mM EGTA (E). BD31 antigen was protected by Ca2+ from the lytic activity of trypsin (TC) and digested totally (TE) or partially (LTE) upon chelation of Ca2+. (C) Generation of the typical tryptic fragment by treating 125I-surface-iodinated FG-2 cell monolayers with trypsin in the presence of Ca2+; immunoprecipitation of cell extracts (Cells) and supernatants (SN) with mAb BD31 showed a slight decrease of the membrane protein because of the release into the medium of the 80 kDa fragment. The gel was intentionally overexposed to show the weaker tryptic fragment band. Molecular mass markers are indicated on the left in kDa.

Novel GPI-anchored adhesion protein 1423 data provide evidence to support anchoring of the BD31 antigen to the plasma membrane via a phosphatidylinositol glycan.

organization of FG-2 cells seeded in the presence of mAb BD31 was impaired and individual cells were either less spread, mantaining a rounded form, or displayed a motile fibroblastoid phenotype (Fig. 11c, f and i; see arrowheads). Reaggregation of FG-2 cells treated with trypsin in the presence of Ca2+, in order to obtain a suspension of dispersed cells bearing intact Ca2+-dependent molecules, was also assayed and found to be progressively inhibited by increasing concentrations of BD31 Fab fragments (Fig. 12). These experiments suggest that BD31 glycoprotein is not involved in cell-substratum adhesion, at least on the tested matrix molecules, but is required for maintaining epithelial cell cohesion and aggregation properties. In fact, Ig binding to the BD31 antigen is sufficient to interact with and to prevent the adhesive role of other cell adhesion molecules and to induce a marked rearrangement of the cytoskeleton of scattered cells.

BD31 antigen is functionally involved in intercellular adhesion A set of experiments was devoted to test the function-blocking properties of mAb BD31 on cell-substratum adhesion using a test previously employed with human keratinocytes (De Luca et al., 1990). By no means did BD31 IgGs affect cell-substratum adhesion on a laminin or fibronectin matrix coating in terms of number of adherent cells (not shown). Conversely, inspection of FG-2 cell colonies challenged with purified BD31 Fab fragments at concentrations ranging from 1.5 to 25 µg/ml showed that colony organization was disrupted in a concentration- and time-dependent manner. Cells mostly lost their cobblestone pattern, acquired a rounded phenotype and eventually began scattering upon loss of their typical aggregation features (Fig. 10a-j). Many of these cells, devoid of cell-to-cell contacts, acquired a nonpolarized fibroblast-like appearance (Fig. 10a-j, e.g. at arrowheads). These morphological changes were superimposable on the effects exerted on confluent FG-2 cells by equal amounts of mAb HECD-1 IgGs directed against E-cadherin (Fig. 10k-o). At high power the F-actin cytoskeleton of these scattered cells showed the typical features of cells undergoing locomotion, including front ruffles and uropodia (not shown). The effect of BD31 IgGs was also time dependent, since the above effects became progressively more marked upon exposure to the antibody (not shown). The capacity of impairing intercellular adhesion and inducing cell scattering was more evident on cells that had previously formed monolayers than on cells undergoing Fig. 9. (A) BD31 antigen does not share the conserved COOH-terminal domain of cadherins. Cell dynamic processes such as extracts of confluent FG-2 cells were immunoprecipitated under reducing conditions with anti-panattachment to the substratum cadherin serum R14 or mAb BD31. The immunoprecipitates were immunoblotted with R14. For and surface spreading; however, comparison the starting cell extract (total extract) was blotted in parallel. The 55 kDa bands represent as progressive synthesis and the rabbit Igs of serum R14 or the mouse Igs of mAb BD31, both used for immunoprecipitation, detected by the ECL system. (B) BD31 antigen does not share the ectodomain of E-cadherin. Cell positional arrangement of extracts of confluent FG-2 cells were immunoprecipitated under reducing conditions with mAb HECDadhesion molecules occurred, 1 to human E-cadherin or mAb BD31. The immunoprecipitates were immunoblotted with mAb HECDthe function-blocking properties 1. For comparison the starting cell extract (total extract) was blotted in parallel. The 55 kDa bands of mAb BD31 could be represent the mouse Igs of mAbs HECD-1 and BD31 used for immunoprecipitation, detected by the observed (Fig. 11): after ECL system. (C) BD31 antigen is attached to the cell membrane via a GPI anchor; 125I-surface-labeled carrying the cohesion assay for FG-2 cells were incubated in the absence (control) or in the presence of PI-PLC and equal amounts of 19 hours, we found that the cell-associated (cells) and released (SN) material were immunoprecipitated. As a further control, cells mouse pre-immune serum had were treated with PI-PLC in the presence of Ca2+, which are known to inhibit the enzyme. Most BD31 negligible effects on aggregaantigen was removed from cells by PI-PLC and recovered in the cell supernatant (SN). Molecular mass tion of cells whereas the colony markers are indicated on the left in kDa.

1424 M. Rabino and others

Fig. 10. Effects of purified BD31 IgGs (a-e), BD31 Fab fragments (f-j) and mAb HECD-1 IgGs (k-o) on FG-2 cell colonies. After the experiment cells were fixed and stained with F-PHD for F-actin (a-e) or observed by phase-contrast microscopy (f-o). FG-2 cell colonies, exposed for 4 hours to mouse pre-immune IgGs (a, f and k), appeared as a continuous cell monolayer; 4 hours incubation with BD31 IgGs (at different concentrations, indicated as µg/ml of incubation medium in the upper left corner of frames b-e) induced disorganization of the monolayer, cell scattering and the appearance of numerous cells with the typical cytoskeletal phenotype of fibroblast-like motile cells (indicated by arrowheads in b-d and g-j). This pattern was much more evident at higher power (not shown). Analogous effects were produced by incubating colonies with mAb HECD-1 IgGs at the same concentrations as for BD31 Fab fragments (k-o, see arrowheads in n and o). Bar, 15 µm (a-e); 30 µm (f-o).

DISCUSSION In this paper we describe some biological, biochemical and functional properties of a molecule that is observed at cell-cell contacts of normal and transformed cells organized as epithelial sheets. This molecule has been identified by a mouse mAb, denoted BD31 (Prat et al., 1986), originally raised to intact

formaldeheyde-fixed gastric carcinoma cells (KATO III) within a program aimed at identifying tumor-associated antigens expressed at the cell surface (Prat et al., 1987). Therefore, the strategy of this investigation was conditioned by the original identification and availability of this unique reagent. BD31 mAb stains intercellular boundaries both in cultured

Novel GPI-anchored adhesion protein 1425

Fig. 11. Time-course of the effects of purified BD31 IgGs on suspended FG-2 cells undergoing attachment. Cells were trypsinized and immediately replated onto 35 mm tissue culture dishes in the presence of the mouse pre-immune IgGs (a-c) or BD31 IgGs at a concentration of 3 µg/ml (d-f) and 6 µg/ml (g-i). The assay was carried for 19 hours and cells were observed hourly (the hours after the exposure to the IgGs when the pictures were taken are indicated in the upper left corner of frames a-c). After 5 hours, cohesion was moderately impaired and cell density was lower (b, e, and h). After 19 hours, a monolayer was fully organized in the control plates (c) while cells exposed to BD31 IgGs were less spread and often showed a motile fibroblastoid appearance (f and i; see arrowheads). Bar, 30 µm. 0.6

Degree of aggregation N0−N50/N0

0.5

0.4

0.3

0.2

0.1

0 A

B

C

D

E

F

G

H

Fig. 12. Aggregation of FG-2 cells under different conditions. Cells treated with 0.01% trypsin plus 1 mM EGTA were allowed to aggregate in the absence (A) or in the presence (B) of Ca2+. Cells treated with 0.01% trypsin plus 10 mM CaCl2 to preserve Ca2+dependent molecules were allowed to aggregate in the presence (C) or in the absence (D) of Ca2+ or in the presence of Ca2+ plus increasing concentrations of BD31 Fab fragments (E, 3µg/ml; F, 6 µg/ml; G, 12 µg/ml; H, 24 µg/ml). Aggregation of cells was inhibited by mAb BD31 in a dose-dependent manner. Each point was the average of triplicate experiments.

cells and in tissue specimens, and identifies a molecule that, on the basis of biochemical tests, seems to correspond to a novel human adhesion molecule sharing some biochemical features with known cadherins but devoid of cadherin homology on the basis of lack of immune cross-reactivity. In fact, it shows cadherin properties like Ca2+-dependent trypsin resistance and generation of a soluble tryptic fragment, but the conserved COOH-terminal domain common to all cadherins is missing; this information, together with the finding that it is membrane anchored via a GPI moiety, makes this molecule similar to the T-cadherin reported in chick embryonic tissues and particularly in the developing central nervous system (Ranscht and Dours-Zimmermann, 1991). It differs, though, for several important general, topographical and biochemical features that indicate that it represents a novel molecular entity. First, the BD31-defined epitope belongs to a strictly human protein and cannot be detected even in primate cells; second, it is not expressed in the nervous and muscle tissues but is highly expressed by secreting epithelia and by rapidly proliferating normal epithelial cells; third, it has a widespread distribution in contact rims of normal and tumour cells growing with epithelioid or frankly epithelial organization; fourth, it is exposed at cell-cell boundaries in tissue sections of carcinomas and particularly in those originating from secreting epithelia. Unlike T-cadherin, it is soluble in non-ionic detergents and thus does not seem to be associated with other cytoskeleton-

1426 M. Rabino and others linked membrane proteins, as suggested by Ranscht and DoursZimmermann (1991) for T-cadherin. For the above reasons we suggest that the glycoprotein we describe in this paper may be loosely related to T-cadherin but represents a different molecule because of its very different tissue distribution. The fact that it lacks the conserved cytoplasmic domain and is GPI-membrane-anchored qualifies this glycoprotein to be a novel molecule involved in intercellular adhesion. It may be either an original entity or a member of a subfamily of tissue-specific truncated adhesion molecules that has to be defined; moreover, it must not be considered as a truncated form of a known epithelial adhesion molecule: namely, P-cadherin, E-cadherin or CEA, on the basis of the following morphological and biochemical considerations: (i) BD31 differs from P-cadherin because of its very different immunohistochemical distribution. In fact, P-cadherin is stained in basal or lower layers of stratified or pseudo-stratified epithelia but is absent from simple or unstratified epithelia such as gastrointestinal epithelium or hepatocytes (Shimoyama et al., 1989); on the contrary, BD31 antigen is slightly detectable in the upper layers of epidermis but is strongly immunoreactive in glandular epithelia and hepatocytes; in addition, P-cadherin has been found to be expressed in squamous cell carcinomas (Nicholson et al., 1991), whereas our protein is missing from these neoplasms. (ii) Although BD31 displays a tissue distribution similar to E-cadherin, we specifically showed that it does not share either the ectodomain or the cytodomain with E-cadherin itself. (iii) Finally, the epithelial CAM CEA does not show Ca2+-dependent trypsin resistance and has a higher molecular mass. Eventually, the molecular cloning and sequencing of BD31 antigen should clarify this issue. BD31 glycoprotein is functionally involved in supporting cell-cell adhesion on its own or in conjunction with other cellcell adhesion molecules like classical cadherins or integrins (e.g. see Lampugnani et al., 1991, 1992). mAb BD31 was investigated for its ability to: (i) prevent substratum adhesion of FG-2 cells; (ii) disrupt FG-2 cell clusters; (iii) affect the reciprocal interactions of suspended cells undergoing attachment; (iv) inhibit re-aggregation of a suspension of dispersed cells with intact Ca2+-dependent molecules. Experimental evidence indicates that mAb BD31 does not modify cell-substratum adhesion but rather induces scattering and motility of aggregated epithelial cells. Moreover, it perturbs cell-cell interactions during the early events of attachment and spreading that precede monolayer formation, and inhibits Ca2+-dependent cell aggregation mediated by BD31 antigen. BD31 may do this by interfering with homotypic bonds like those made by classical cadherins, or it may be speculated that it interacts heterotypically with lateral β1 epithelial integrins, whose membrane counter-receptors have not yet been clearly shown (compare panels a and d of Fig. 4, which show matching locations for BD31 glycoprotein and β1 integrin in keratinocyte colonies). Antibodies directed against individual molecules involved in intercellular adhesion disrupt reciprocal bonds and impair tissue organization (Damsky et al., 1983; Marchisio et al., 1991). This means that perturbing even one component of the molecular complex controlling intercellular adhesion may alter the unstable equilibrium governing epithelial assembly. The membrane-anchoring of the BD31 antigen by means of a GPI moiety deserves some comments. GPI-anchoring confers

on membrane proteins unique biological properties that range from higher mobility on the plane of the membrane to independence from the cytoskeleton, susceptibility to external enzyme regulation, and potential generation of intracellular metabolic signals (reviewed by Low, 1987, 1989b; Ferguson and Williams, 1988). GPI-anchoring has already been reported in at least four human adhesion molecules, i.e. N-CAM120 (He et al., 1986; Hemperly et al., 1986), carcinoembryonic antigen (CEA; Benchimol et al., 1989), heparan sulphate proteoglycan (Ishihara et al., 1987) and LFA-3 (Selvaraj et al., 1987). The fact that a cell-cell adhesion molecule presumably binds a partner molecule on the adjacent cell without interacting with its cytoskeleton suggests that adhesive recognition may be regulated by cleavage of the lipid anchor. Such cleavage and the ensuing de-adhesion may be caused by different phospholipases (e.g. PLD; Davitz et al., 1987; Low and Prasad, 1988). Such phenomena occurring in a cadherin-like molecule may have several consequences in rapidly proliferating epithelia either during development or in tumor growth. One consequence may be that proliferating cells may reciprocally detach and may be relocated upon mitosis without obvious involvement of the cytoskeleton. One furher consequence is that the adhesion molecule may generate signals like 1,2-diacylglycerol (DAG; Asaoka et al., 1992), phosphatidic acid (PA; Moolenaar et al., 1986; Murayama and Uy, 1987) or inositol glycan mediators (Robinson, 1991) that are reportedly released upon the hydrolysis of the GPI group (Low, 1989a; Romero, 1991). We speculate that these signals may interact with the complex signaling network transduced by growth factor receptors (e.g. see Stefanova et al., 1991) and directly link control of cellular adhesion to control of cell proliferation. These speculations rest largely on assumptions that must be tested with appropriate experiments. It is widely accepted that, in many but not all epithelial cells (Zurzolo et al., 1993), GPI groups represent signals for the apical sorting of membrane proteins (for review see Lisanti and Rodriguez-Boulan, 1990, 1991). We have paid particular attention to this issue and performed a detailed localization analysis, using the confocal microscope, aimed at showing that the BD31 antigen is restricted to lateral membrane domains. Although we have not yet studied the dynamics and the topography of its exposure in appropriate experiments, three-dimensional reconstruction and object rotation analysis of confocal optical sections (Fig. 3) suggest that this GPI-linked protein is specifically trapped in cell-cell borders and excluded from apical and basal domains. This also occurs in the highly polarized cells of human colon mucosa studied by immunohistochemistry within their tissue organization. Overall, the data of the present paper provide evidence for a novel membrane glycoprotein that is specific to human epithelial cells and is anchored by a GPI moiety. Even if its molecular structure and the corresponding cDNA have not yet been identified, we feel justified in proposing that this molecule is involved in intercellular adhesion and may share some properties with already described GPI-linked adhesion molecules, such as CEA, N-CAM120 and T-cadherin. This work was supported by Progetto Finalizzato ‘Biotecnologie e Biostrumentazione’ and ‘Applicazioni Cliniche della Ricerca Oncologica’ Consiglio Nazionale delle Ricerche (CNR, Rome), by the Associazione Italiana per la Ricerca sul Cancro (AIRC, Milano), and

Novel GPI-anchored adhesion protein 1427 by the Ministero per l’Università e la Ricerca Scientifica e Tecnologica (MURST, Roma). We are indebted to A. Graziani and L. Naldini for stimulating discussions. Ms R. Albano provided excellent technical assistance.

REFERENCES Asaoka, Y., Nakamura, S., Yoshida, K. and Nishizuka, Y. (1992). Protein kinase C, calcium and phospholipid degradation. Trends Biochem. Sci. 17, 414-417. Benchimol, S., Fuchs, A., Jothy, S., Beauchemin, N., Shirota, K. and Stanners, C. P. (1989). Carcinoembryonic antigen, a human tumor marker, functions as an intercellular adhesion molecule. Cell 57, 327-334. Buck, C. A. (1992). Immunoglobulin superfamily: structure, function and relationship to other receptor molecules. Semin. Cell Biol. 3, 179-188. Cunningham, B. A., Leutzinger, Y., Gallin, W. J., Sorkin, B. C. and Edelman, G. M. (1984). Linear organization of the liver cell adhesion molecule L-CAM. Proc. Nat. Acad. Sci. USA 81, 5787-5791. Damsky, C. H., Richa, J., Solter, D., Knudsen, K. and Buck, C. A. (1983). Identification and purification of a cell surface glycoprotein mediating intercellular adhesion in embryonic and adult tissue. Cell 34, 455-466. Davitz, M. A., Hereld, D., Shak, S., Krakow, J., Englund, P. T. and Nussenzweig, V. (1987). A glycan-phosphatidylinositol-specific phospholipase D in human serum. Science 238, 81-84. Dejana, E., Colella, S., Abbadini, M., Gaboli, M. and Marchisio, P. C. (1988). Fibronectin and vitronectin regulate the organization of Arg-GlyAsp adhesion receptor at focal contacts of cultured human endothelial cells. J. Cell Biol. 107, 1215-1223. De Luca, M., Tamura, R. N., Kajiji, S. Bondanza, S., Rossino, P., Cancedda, R., Marchisio, P. C. and Quaranta, V. (1990). Polarized integrins mediate keratinocyte adhesion to basal lamina. Proc. Nat. Acad. Sci. USA 87, 6888-6892. Edelman, G. M. and Crossin, K. L. (1991). Cell adhesion molecules: implications for a molecular histology. Annu. Rev. Biochem. 60, 155-190. Ekblom, P., Vestweber, D. and Kemler, R. (1986). Cell-matrix interactions and cell adhesion during development. Annu. Rev. Cell Biol. 2, 27-47. Ferguson, M. A. J. and Williams, A. F. (1988). Cell-surface anchoring of proteins via glycosyl-phosphatidylinositol structures. Annu. Rev. Biochem. 57, 285-320. Fleming, T. P. and Johnson, M. H. (1988). From egg to epithelium. Annu. Rev. Cell Biol. 4, 459-485. Gallin, W. J., Edelman, G. M. and Cunningham, B. A. (1983). Characterization of L-CAM, a major cell adhesion molecule from embryonic liver cells. Proc. Nat. Acad. Sci. USA 80, 1038-1042. Gallin, W. J., Sorkin, B. C., Edelman, G. M. and Cunningham, B. A. (1987). Sequence analysis of a cDNA clone encoding the liver cell adhesion molecule L-CAM. Proc. Nat. Acad. Sci. USA 84, 2808-2812. Geiger, B., Volberg, T., Ginsberg, D., Bitzur, S., Sabanay, I. and Hynes, R. O. (1990). Broad spectrum pan-cadherin antibodies, reactive with the 24 amino acid residues of N-cadherin. J. Cell Sci. 97, 607-614. Goridis, C. and Brunet, J. F. (1992). NCAM: structural diversity, function and regulation of expression. Semin. Cell Biol. 3, 189-197. Gumbiner, B. and Simons, K. (1986). A functional assay for proteins involved in establishing an epithelial occluding barrier: identification of a uvomorulinlike polypeptide. J. Cell Biol. 102, 457-468. Hatta, K., Nose, N., Nagafuchi, M. and Takeichi, M. (1988). Cloning and expression of cDNA encoding a neural calcium-dependent cell adhesion molecule: its identity in the cadherin gene family. J. Cell Biol. 106, 873-881. He, H.-T., Barbet, J., Chaix, J. C. and Goridis, C. (1986). Phosphatidylinositol is involved in the membrane attachment of NCAM120, the smallest component of the neural cell adhesion molecule. EMBO J. 5, 2489-2494. Hemperly, J. J., Edelman, G. M. and Cunningham, B. A. (1986). cDNA clones of the neural cell adhesion molecule (N-CAM) lacking a membranespanning region consistent with evidence for membrane attachment via a phosphatidylinositol intermediate. Proc. Nat. Acad. Sci. USA 83, 9822-9826. Hirano, S., Nose, A., Hatta, K., Kawarkami, A. and Takeichi, M. (1987). Calcium-dependent cell-cell adhesion molecules (cadherins): subclass specificities and possible involvement of actin bundles. J. Cell Biol. 105, 2501-2510. Hynes, R. O. (1992). Integrins: versatility, modulation and signaling in cell adhesion. Cell 69, 11-25

Ishihara, M., Fedarko, N. S. and Conrad, H. E. (1987). Involvement of phosphatidylinositol and insulin in the coordinate regulation of protoheparan sulfate metabolism and hepatocyte growth J. Biol. Chem. 262, 4708-4716. Kajiji, S., Tamura, R. N. and Quaranta, R. N. (1989). A novel integrin (αEβ4) from human epithelial cells suggests a fourth family of integrin adhesion receptors. EMBO J. 8, 673-680. Kemler, R. (1992). Classical cadherins. Semin. Cell Biol. 3, 149-155. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 230, 680-685. Lampugnani, M. G., Resnati, M., Dejana, E. and Marchisio, P. C. (1991). The role of integrins in the maintenance of endothelial monolayer integrity. J. Cell Biol. 112, 479-490. Lampugnani, M. G., Resnati, M., Raiteri, M., Pigott, R., Pisacane, A., Houen, G., Ruco, L. P. and Dejana, E. (1992). A novel endothelial-specific membrane protein is a marker of cell-cell contacts. J. Cell Biol. 118, 15111522. Larjava, H., Peltonen, J., Akiyama, S. K., Yamada, S. S., Gralnick, H. R., Uitto, J. and Yamada, K. M. (1990). Novel function of β1 integrins in keratinocyte cell-cell adhesion. J. Cell Biol. 110, 803-815 Lever, W. F. and Schaumburg-Lever, G. (1990). Histopathology of the Skin, 7th edn. J. B. Lippincott Co., Philadelphia. Lisanti, M. P. and Rodriguez-Boulan, E. (1990). Glycophospholipid membrane anchoring provides clue to the mechanism of protein sorting in polarized epithelial cells. Trends Biochem. Sci. 15, 113-118. Lisanti, M. P. and Rodriguez-Boulan, E. (1991). Polarized sorting of GPIlinked proteins in epithelia and membrane microdomains. Cell Biol. Int. Rep. 15, 1023-1049. Low, M. G. (1987). Biochemistry of the glycosyl-phosphatidylinositol membrane protein anchors. Biochem. J. 244, 1-13. Low, M. G. (1989a). Glycosyl-phosphatidylinositol: a versatile anchor for cell surface proteins. FASEB J. 3, 1600-1608. Low, M. G. (1989b). The glycosyl-phosphatidylinositol anchor of membrane proteins. Biochim. Biophys. Acta 988, 427-454. Low, M. G. and Prasad, A. R. S. (1988). A phospholipase D specific for the phosphatidylinositol anchor of cell-surface proteins is abundant in plasma. Proc. Nat. Acad. Sci. USA 85, 980-984. Low, M. G. and Saltiel, A. R. (1988). Structural and functional roles of glycosylphoshatidylinositol in membranes. Science 239, 268-275. Mansouri, A., Spurr, N., Goodfellow, P. B. and Kemler, R. (1988). Characterization and chromosomal localization of the gene encoding the human cell adhesion molecule uvomorulin. Differentiation 38, 67-71. Marchisio, P. C., Cirillo, D., Naldini, L., Primavera, M. V., Teti, A. and Zambonin Zallone, A. (1984). Cell-substratum interaction of cultured avian osteoclasts is mediated by specific adhesion structures. J. Cell Biol. 99, 16961705. Marchisio, P. C., Bondanza, S., Cremona, O., Cancedda, R. and De Luca, M. (1991). Polarized expression of integrin receptors (α6β4, α2β1, α3β1 and αvβ5). and their relationship with the cytoskeleton and basement membrane matrix in cultured human keratinocytes. J. Cell Biol. 112, 761-773. Moolenaar, W. H., Kruijer, W., Tilly, B. C., Verlaan, I., Bierman, A. J. and de Laat, S. W. (1986). Growth factor-like action of phosphatidic acid. Nature 323, 171-173. Murayama, T. and Ui, M. (1987). Phosphatidic acid may stimulate membrane receptors mediating adenylate cyclase inhibition and phospholipid breakdown in 3T3 fibroblasts. J. Biol. Chem. 262, 5522-5529. Nagafuchi, A. and Takeichi, M. (1989). Transmembrane control of cadherinmediated cell adhesion: a 94 kDa protein functionally associated with a specific region of the cytoplasmic domain of E-cadherin. Cell Regul. 1, 3744. Nicholson, L. J., Pei, X. F. and Watt, F. M. (1991). Expression of ε-cadherin, P-cadherin and involucrin by normal and neoplastic keratinocytes in culture. Carcinogenesis 12, 1345-1349. Nose, A., Nagafuchi, A. and Takeichi, M. (1987). Isolation of placental cadherin cDNA: identification of a novel gene family for cell-cell adhesion molecules. EMBO J. 6, 3655-3661. Ozawa, M., Baribault, H. and Kemler, R. (1989). The cytoplasmic domain of the cell adhesion molecule uvomorulin associates with three independent proteins structurally related in different species. EMBO J. 8, 1711-1717. Ozawa, M., Ringwald, M. and Kemler, R. (1990). Uvomorulin-catenin complex formation is regulated by a specific domain in the cytoplasmic region of the cell adhesion molecule. Proc. Nat. Acad. Sci. USA 87, 42464250. Parham, P. (1986). Preparation and purification of active fragments from

1428 M. Rabino and others mouse monoclonal antibodies. Cellular Immunology, 4th edn. Blackwell Scientific Publications, California. Prat, M., Rossino, P., Marchisio, P. C., Comoglio, P. M. and Corbascio, G. C. (1986). Identification of a glycoprotein expressed in cell-cell contacts by means of a monoclonal antibody. Cell. Biol. Int. Rep. 10, 159. Prat, M., Rossino, P., Bussolati, G., Morra, I. and Comoglio, P. M. (1987). Production of monoclonal antibodies for the immunohistochemical detection of gastric carcinomas. Cancer Detect. Prev. 10, 293-301. Prat, M., Narsimhan, R. P., Crepaldi, T., Nicotra, M. R., Natali, P. G. and Comoglio, P. M. (1991). The receptor encoded by the human c-MET oncogene is expressed in hepatocytes, epithelial cells and solid tumors. Int. J. Cancer 49, 323-328. Ranscht, B. and Dours-Zimmermann, M. T. (1991). T-cadherin, a novel cadherin cell adhesion molecule in the nervous system lacks the conserved cytoplasmic region. Neuron 7, 391-402. Ringwald, M., Schuh, R., Vestweber, D., Eistetter, H., Lottspeich, F., Engel, J., Dolz, R., Jahnig, F., Epplen, J., Mayer, S., Müller, C. and Kemler, R. (1987). The structure of cell adhesion molecule uvomorulin. Insights into the molecular mechanism of Ca2+ dependent cell adhesion. EMBO J. 6, 3647-3653. Robinson, P. J. (1991). Phosphotidylinositol membrane anchors and T cell activation. Immunol. Today 12, 35-41. Romero, G. (1991). Inositolglycans and cellular signalling. Cell Biol. Int. Rep. 15, 827-852. Ruoslahti, E. and Giancotti, F. G. (1989). Integrins and tumor cell dissemination. Cancer Cells 1, 119-126.

Ruoslahti, E. and Yamaguchi, Y. (1991). Proteoglycans as modulators of growth factor activities. Cell 64, 867-869. Selvaraj, P., Dustin, M. L., Silber, S., Low, M. G. and Springer, T. A. (1987). Deficiency of lymphocyte function-associated antigen 3 (LFA-3) in paroxysmal nocturnal hemoglobinuria. Functional correlates and evidence for a phosphatidylinositol membrane anchor. J. Exp. Med. 166, 1011-1025. Shimoyama, Y., Hirohashi, S., Hirano, S., Noguchi, M., Shimosato, Y., Takeichi, M. and Abe, O. (1989). Cadherin cell-adhesion molecules in human epithelial tissues and carcinomas. Cancer Res. 49, 2128-2133. Stefanova, I., Horejsi, V., Ansotegui, I. J., Knapp, W. and Stockinger, H. (1991). GPI-anchored cell-surface molecules complexed to protein tyrosine kinases. Science 254, 1016-1019. Takeichi, M. (1990). Cadherins: a molecular family important in selective cellcell adhesion. Annu. Rev. Biochem. 59, 237-252. Takeichi, M. (1991). Cadherins cell adhesion receptors as a morphogenetic regulator. Science 251, 1451-1455. Towbin, H., Staehelin, T. and Gordon, J. (1979). Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications. Proc. Nat. Acad. Sci. USA 76, 4350-4353. Urushihara, H., Ozaki, H. S. and Yakeichi, M. (1979). Immunological detection of cell surface components related with aggregation of Chinese hamster and chick embryonic cells. Dev. Biol. 70, 206-216. Vestal, D. J. and Ranscht, B. (1992). Glycosylphosphatidylinositol-anchored T-cadherin mediates calcium-dependent, homophilic cell adhesion. J. Cell Biol. 119, 451-461. Volk, T. and Geiger, B. (1984). A 135-kD membrane protein of intercellular adherens junctions. EMBO J. 3, 2249-2260.