Formation of hemidesmosomes in cells of a transformed murine mammary tumor cell line and mechanisms involved in adherence of these cells to laminin and kalinin

Hemidesmosomes are specialized domains of the plasma membrane of epithelial cells that mediate cell-substrate adhesion and serve as sites for attachment of keratin fila-ments to the cell surface (Schwarz et al., 1990; Garrod, 1993). They are most prominent in the basal layer of squa-mous and transitional epithelia but are found in some simple epithelial tissues as well. Moreover, hemidesmosome-like structures have been described in certain epithelial cell cul-tures (Carter et al., 1990; Owaribe et al., 1990; Ridelle et al., 1991; Hieda et al., 1992). The integrin α₆β₄ has been localized in hemidesmosomes of skin (Sonnenberg et al., 1991b; Jones et al., 1991) and cornea (Stepp et al., 1990) and may interact with one or more components of their basement membranes. There is evidence that α₆β₄ on colon carcinoma cells binds to laminin (Lotz et al., 1990; Lee et al., 1992). Although laminin is a major constituent of epi-dermal basement membranes, there is as yet no evidence that this component is also recognized by α₆β₄ of basal ker-atinocytes (DeLuca et al., 1990). In fact, several studies suggest that α₆β₄ may bind a ligand that is different from classical laminin (Sonnenberg et al., 1990; Carter et al., 1990; Kurpakus et al., 1991a).

Recently, a complex of three proteins, named kalinin, has been described in matrices of keratinocytes (Rouselle et al., 1991). This protein complex is considered to be a laminin isotype, because its three polypeptides showed strong homology to the A, B1 and B2 chains of laminin (Verrando et al., 1993; Kallunki et al., 1992). Kalinin is a component of the anchoring filaments of hemidesmosomes and is a highly adhesive substrate for keratinocytes. Another trimeric component of epidermal basement membranes, named epiligrin has been identified as a ligand for the integrin α₃β₁ (Carter et al., 1991). It is now apparent that kalinin is identical to epiligrin and to a previously described trimeric complex called BM600, absent in patients with lethal junctional epidermolysis bullosa (Verrando et al., 1987; Domloge-Hultsch et al., 1992). More recently, basement membranes were shown to contain a novel laminin isoform, named k-laminin (Marinkovich et al., 1992b). K-laminin shares the B1 and B2 chains of Englebreth-Holm-Swarm (EHS) tumor laminin, but its A chain, which is structurally and immunologically related to the A chain of kalinin, is different. Both kalinin and k-laminin are possible ligands for α₆β₄.

In addition to mediating adhesion of keratinocytes to the basement membrane, α₆β₄ may play a crucial role in anchoring keratin filaments to the plasma membrane. The cytoplasmic domain of β₄ may be specialized for this type of interaction. It is about 1000 amino acids long as com-pared to about 50-60 in other β subunits of integrins and contains, as an additional unique feature, four type III fibronectin repeats (Suzuki and Naitoh, 1990; Hogervorst et al., 1990). These type III repeats are present in two pairs connected by a region encoded by mRNA that is subject to alternative splicing (Tamura et al., 1990). In analogy to inte-grins connecting with the actin-containing cytoskeleton, the interaction of α₆β₄ with keratin filaments may be indirect and may involve several associated proteins. A hemidesmo-some-associated protein that is recognized by autoantibod-ies in sera from patients with the disease bullous pem-phigoid and named BP230 (Stanley et al., 1981), is located in the cytoplasmic plaque of hemidesmosomes (Westgate et al., 1985; Tanaka et al., 1990). Two other proteins that have recently been identified and are located in cytoplas-mic regions of hemidesmosomes are components of 200 (Kurpakus et al., 1991b) and 500 kDa (HD1) (Owaribe et al., 1991; Hieda et al., 1992). Each of these proteins may be involved in linking keratin filaments to the plasma membrane.

In this paper we describe the generation of a squamous cell line from a mouse mammary tumor cell culture. We show that these cells have the capacity to differentiate nor-mally, i.e. they stratify and form cornified cells and they develop hemidesmosomes in association with the substra-tum. These hemidesmosomes are similar to those produced by basal keratinocytes in vivo in that they contain the inte-grin α₆β₄ as well as the hemidesmosomal plaque proteins BP230 and HD1. In addition, kalinin and laminin were con-centrated underneath these hemidesmosomes. The role of α₆β₄ in mediating adhesion to these substrates was inves-tigated in antibody inhibition assays using a monoclonal antibody to α₆. This antibody inhibited adhesion to these matrix components, but only in the early phase of cell adhe-sion, whereas inhibition was lost during longer periods of incubation. A polyclonal antiserum against β₁ had no sig-nificant effect on adhesion. The loss of inhibition was asso-ciated with the redistribution of α₆β₄ to the basal surface. Anti-α₆ also inhibited adhesion of cells treated with cytochalasin B or deoxyglucose/azide, which was either not or less dependent on the length of incubation. Since redis-tribution of α₆β₄ on these treated cells was disturbed or non-detectable, we suggest that the anti-α₆ antibody hinders the binding of the α₆β₄ integrins to its ligands, but cannot prevent adhesion when the integrin is clustered.

The RAC-11P and RAC-11P/SD cell lines were cultured in Dul-becco’s modified Eagle’s medium (Gibco BRL, Paisley, UK) sup-plemented with 10% (v/v) bovine fetal calf serum. The RAC-11P cell line was originally isolated from a mouse mammary tumor and has been described previously (Sonnenberg et al., 1986). The RAC-11P/SD cells are a clonal derative of RAC-11P cells, which show extensive squamous formation. The human squamous cell carcinoma cell line UMSCC-22B was obtained from Dr T. E. Carey (University of Michigan, Ann Arbor, Michigan). All cells were grown at 37°C in a humidified, 5% CO₂ atmosphere.

Antiserum to the murine fibronectin receptor was raised in rab-bits against a fibronectin receptor preparation purified from murine Friends Erythroleukemia cells by affinity chromatograhy on a 115 kDa chymotrypsin fragment from fibronectin. The bound fibronectin receptor was eluted with RGD-containing peptides. It was kindly provided by Dr P. Bernardi (Unversita degli Studi di Padova, Padova, Italy). Polyclonal antisera against synthetic pep-tides derived from the cytoplasmic domains of human α₁, α₂, α₃, α₄, α₅, α_v and β₁ (Defilippi et al., 1991) were kindly donated by Dr G. Tarone (Universita di Torino, Torino, Italy). Rabbit antis-era to synthetic peptides corresponding to the 39 COOH-terminal amino acids from β₁ (Marcantonio and Hynes, 1988) was obtained from Dr R. O. Hynes (MIT, Boston, MA). Rabbit anti-cytoplas-mic polyclonal antibodies against α₈ (Bossy et al., 1991) and α_6B were kindly provided by Dr L. F. Reichardt (Howard Hughes Medical Institute, San Francisco, CA). Rabbit antisera corre-sponding to the cytoplasmic domains of the β₅ and α_7B subunits were kindly supplied by Dr E. Ruoslahti (La Jolla Cancer Foun-dation, La Jolla, CA) and Dr R. H. Kramer (University of Cali-fornia San Francisco, San Francisco, CA), respectively. Rabbit anti-vinculin (Geiger, 1979) was generously provided by Dr B. Geiger (The Weizmann Institute of Science, Rehovat, Israel). Rabbit anti-serum to human kalinin (Marinkovich et al., 1992a) and to human collagen type VII (Lunstrum et al., 1986) were kind gifts from Dr R. Burgeson (Cutaneous Biology Research Center, Charlestown, MA). The mAb 4C7, which reacts with human kalinin (Jaspars et al., 1993) was obtained from Dr L. H. Jaspars (Free University, Amsterdam, the Netherlands). Rabbit anti-col-lagen type IV (Engvall et al., 1982) was provided by Dr E. Eng-vall (La Jolla Cancer Foundation, La Jolla, CA). Rabbit antiserum against E-cadherin (Vestweber and Kemler, 1984) was obtained from Dr R. Kemler (Friedrich-Miescher-Laboratorium der Max-Planck-Gesellshaft, Tubingen, Germany). Serum samples from patients with bullous pemphigoid were gifts from Dr D. Boorsma (Free University, Amsterdam, The Netherlands). Monoclonal anti-body 11-5F, directed against desmoplakins I and II (Parrish et al., 1987), was a gift from Dr Garrod (University of Manchester, Man-chester, UK); mAb HD-1, against a 500 kDa component of the hemidesmosomal plaque (Hieda et al., 1992), was a gift from Dr K. Owaribe (Nagoya University, Nagoya, Japan); mAb 346-11A to murine β₄ (Kennel et al., 1986) was kindly provided by Dr S. J. Kennel (Oak Ridge National Laboratory, Oak Ridge, TN); mAb RCK 107 to cytokeratin 14 (Smedts et al., 1992) and rabbit anti-keratin, prepared against keratins isolated from skin and affinity purified (van Broers et al., 1987) were obtained from Dr F. C. S. Ramaekers (University of Limburg, Maastricht, the Netherlands); mAb GoH3 recognizing an extracellular epitope on the integrin α₆ subunit and mAbs 1A10 and 6B4 to the cytoplasmic domains of the α_6A and α_6B subunits, respectively, have been described previously (Sonnenberg et al., 1987; Hogervorst et al., 1993). Rabbit anti-rat fibronectin was purchased from Telios Pharma-ceuticals (San Diego, CA). Rabbit antiserum against recombinant murine nidogen (Fox et al., 1991), murine tumor EHS laminin and its proteolytic fragments P1, E3, E4 and E8 were kindly provided by Dr R. Timpl (Max-Planck Institute, Martinsried, Germany). Human fibronectin, mAb VIN-11-5 to vinculin, and FITC- and rhodamine-conjugated phalloidin were purchased from Sigma Chemical Co. (St. Louis, MO). FITC-conjugated antibodies to mouse IgG were from Zymed Laboratories Inc. (San Francisco, CA) and to rat, rabbit and human IgG were from Nordic Immuno-chemicals Laboratory (Tilburg, the Netherlands). Texas Red-con-jugated antibodies to mouse and rat IgG were from Molecular Probes (Eugene, OR) and to rabbit IgG were from Amersham International (Buckinghamshire, UK)

RAC-11P/SD cells were detached with trypsin/EDTA, plated on glass coverslips in culture medium containing 10% FCS and then processed for immunofluorescence staining after 1 or 2 days. Alternatively, cells were suspended and seeded onto coverslips coated with various matrix components, in serum-free culture medium containing 0.35% BSA for periods of 5 to 120 minutes. Cells were fixed with 1% formaldehyde in PBS for 10 minutes and permeabilized with 0.5% Triton X-100 for 5 minutes at room temperature. After rinsing and blocking with 1% BSA in PBS for 20 minutes, the permeabilized cells were incubated with primary antibody for 30 minutes at 37°C. The cells were stained with flu-orescein-or Texas Red-labeled antimouse, rat, rabbit or human IgG for 30 minutes. The coverslips were then washed again and mounted with Vectashield (Vector, Burlinghame, CA). The edges of the coverslips were sealed. The coverslips were viewed under a Zeiss microscope (Carl Zeiss, Inc., Thornwood, NY) equipped with a confocal scanning laser microscope (Bio-Rad, Richmond, CA).

Confluent cell cultures of RAC-11P/SD cells grown on thermanox coverslips (Miles Scientific, Naperville, IL) were fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer, followed by post-fix-ation in 1% osmium tetroxide in the same buffer. Cells were stained with 1% uranyl acetate and embedded in a mixture of LX112 and Araldite. Thin sections were cut perpendicularly to the contact surface of the cells.

Confluent cell cultures were fixed for 2 hours in graded (2 to 8%) paraformaldehyde dilutions in PBS. The cells were scraped from the dishes with a rubber policeman and pelleted in 10% gelatin in PBS. Ultrathin frozen sections were incubated with specific antibody followed by incubation with the appropriate anti-species IgG conjugated to 10 nm gold particles (Amersham Nederland, ‘s-Hertogenbosch, The Netherlands). Both incubations were for 1 hour at room temperature. After immunolabeling, the cryosections were embedded in a mixture of methylcellulose and uranyl acetate.

The cells were detached from the culture dishes by overnight incu-bation at 4°C with 20 mM EDTA in PBS. The cells were divided into two portions. One portion was incubated with anti-α₆ mAb (GoH3) and goat anti-rat IgG conjugated to 5 nm gold particles as described above. The other was fixed in 1% paraformaldehyde in PBS for 10 minutes and the cells were permeabilized for 5 min-utes in 0.5% Triton X-100 in PBS. The cells were then incubated with mAb 121 to HD1 and goat antimouse IgG conjugated to 5 nm gold particles. All cells were post-fixed and embedded in a mixture of LX1122 and Araldite as described above. Sections were viewed and photographed in a Philips CM 10 electron microscope.

Matrices of RAC-11P/SD were prepared as previously described (Delwel et al., 1993). Briefly, RAC-11P/SD cells were grown to confluency in 96-well polystyrene microtiter plates, washed three times with PBS and incubated in PBS containing 20 mM EDTA, leupeptin (10 μg/ml), phenylmethanesulfonyl fluoride (1 mM) and soybean trypsin inhibitor (10 μg/ml) overnight at 4°C. After incu-bation, the cells were removed as a single continuous sheet from their matrix by brief, but forceful pipetting. Matrices were stored in PBS at 4°C before use in cell adhesion assays. For western blot analysis, matrices of three wells were dissolved in SDS-sample buffer and heated at 60°C for 15 minutes.

Proteins in the culture medium of UMSCC-22B cells were pre-cipitated with 50% ammonium sulfate (2 hours, 4°C). The pre-cipitate was dissolved, dialyzed against PBS and applied to immunoaffinity columns of anti-kalinin 4C7 mAb, coupled to Sepharose. The immunoaffinity column was extensively washed with 10 mM phosphate buffer, pH 8, containing 1 M NaCl. Bound material was eluted with 100 mM triethylamine, pH 11.5, and immediately neutralized with one-sixth of a volume of 1 M phos-phate buffer, pH 6.8. Fractions containing human kalinin were identified by SDS-PAGE, pooled, dialyzed against PBS and stored at −70°C.

Matrix proteins were separated by SDS-PAGE and electrophoret-ically transferred to nitrocellulose paper sheets for 1 hour at 240 mA or overnight at 40 mA in 25 mM Tris-HCl, pH 8.3, 192 mM glycine, 20% (v/v) methanol and 0.0375% SDS, as described by Towbin et al. (1979). The filters were preincubated in TBS (10 mM Tris-HCl, pH 8, and 150 mM NaCl) containing 0.05% Tween-20 and 1% (w/v) BSA for 30 minutes at room temperature, washed and then incubated with polyclonal antisera diluted 1:100 or 1:250 for 30 minutes. Labeling with a second antibody, alkaline phos-phatase-conjugated goat anti-rabbit IgG (Promega Corp., Madi-son, WI) diluted 1:7500 in TBS/Tween-20 was for 30 minutes at room temperature and the reaction was developed with nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate in 100 mM Tris-HCl, pH 9.5, 100 mM NaCl, 5 mM MgCl₂.

Microtiter plates (96 well; Greiner) were coated by overnight incu-bation at 4°C with different substrate proteins (20 μg/ml) diluted in PBS. After three washes with PBS, the plates were blocked by incubation for 1 hour with 1% bovine serum albumin (BSA; Sigma Chemical Co.) in PBS. Subconfluent cultures of RAC-11P/SD cells were washed with PBS and cells were detached with 0.25% trypsin/0.02% EDTA, then washed three times with PBS and labeled with 400 μCi Na₂⁵¹CrO₄ at room temperature. After wash-ing, the labeled cells were suspended to a concentration of 1×10⁶ cells/ml in Dulbecco’s modified Eagle’s medium containing 0.35% BSA (DMEM/BSA) and added to the substrate-coated plates (100 μl per well). Cells were allowed to adhere for vary-ing periods of time at 37°C. After five washes with DMEM/BSA, the cells that had remained attached to the wells were lysed in 1% SDS and radioactivity was determined in a gamma counter. When evaluating the effect of inhibition of protein synthesis on cell adhe-sion, the cultures of RAC-11P/SD cells were preincubated with cycloheximide (25 μg/ml) for 2 hours before they were removed from the dishes. Cycloheximide remained present during further preparation of the cells and was also included in the buffer used in the adhesion assays. Cell metabolism was inhibited by prein-cubating cells with sodium azide (0.5% w/v) and deoxyglucose (50 mM; Sigma Chemical Co.). For cytoskeletal disruption, the cells were preincubated with cytochalasin B (20 μM; Sigma Chemical Co.) for 30 minutes at 37°C. In antibody inhibition experiments, RAC-11P/SD cells were preincubated for 10 min-utes in the presence of mAbs before plating.

For metabolic labeling, nearly confluent cultures of RAC-11P cells were preincubated in methionine-free minimal essential medium (Gibco) and then labeled with 50 μCi/ml [³⁵S]methionine for 4 hours. Cell were surface labeled with ¹²⁵I using lactoperoxidase as described previously (Sonnenberg et al., 1987). Metabolically and surface iodinated cells were washed, lysed in 1% Nonidet P40 in 25 mM Tris-HCl (pH 7.5), 4 mM EDTA, 100 mM NaCl, 1 mM phenylmethanesulfonyl fluoride, 10 μg/ml leupeptin and 10 μg/ml soybean trypsin inhibitor and the lysates were clarified by centrifugation and precleared with Protein A-Sepharose (Pharma-cia LKB Biotechnology Inc., Uppsala, Sweden). Samples of the precleared lysates were immunoprecipitated with antibodies pre-viously bound to Protein A-Sepharose or to Protein A-Sepharose to which rabbit anti-rat IgG or rabbit antimouse IgG was attached. The immunoprecipitates were analyzed on polyacrylamide gels according to the method of Laemmli (1970).

In experiments examining the secretion of matrix components, the cells were cultured in a medium containing 10% of the usual levels of methionine in the presence of 50 μCi/ml [³⁵S]methion-ine. The culture medium was centrifuged to remove cellular mate-rial and thereafter detergents (1% Nonidet P40) and protease inhibitors (1 mM phenylmethanesulfonyl fluoride, 10 μg/ml leu-peptin and 10 μg/ml soybean trypsin inhibitor) were added to the medium as concentrated solutions.

The RAC-11P cell line is a clonal culture derived from the RAC-10P line, which was originally isolated from a murine mammary tumor (Sonnenberg et al., 1986). Although both cell lines are of epithelial origin as indicated by their reac-tion with antibodies against keratins, during culture they develop entirely different properties. RAC-10P cells can form domes, a differentiation characteristic of glandular epithelial cells, whereas RAC-11P cannot. Instead, the RAC-11P cells form colonies of stratified cells. These colonies consist of a variable number of cell layers with, at the surface, a continuous layer of flattened cells with pyc-notic nuclei (Fig. 1).

The ultrastructural organization of these colonies closely resembles that of the normal epidermis: a basal layer of cuboidal cells, flattened suprabasilar cells and cornified superficial cells (Fig. 2a). Desmosomes were found between cells of the different layers while hemidesmosomes were seen in great abundance in the basal plasma membrane of the basal cells. These hemidesmosomes consist of a cyto-plasmic plaque to which intermediate filaments are anchored (Fig. 2b). Electron-dense patches, possibly anchoring filaments, can often be seen underneath the hemidesmosomes in association with the substratum. In the subclone RAC-11P/SD, a larger percentage of cells differ-entiate to squamous cells and these cells also contain a rel-atively larger number of hemidesmosomes. For our further studies we used these latter cells.

Double-label confocal immunofluorescence experiments were performed to study the expression of different com-ponents (BP230, HD1 and α₆β₄) of hemidesmosomes in densely grown monolayers of RAC-11P/SD cells. As shown in Fig. 3 antibodies against α₆ (mAb GoH3 or polyclonal anti-α_6A) produced a characteristic pattern of spots in close contact with the substratum. These spots are arranged in several definite lines separated from each other by unstained lines. No staining was found along the contacting edges of RAC-11P/SD cells. An identical staining pattern was seen with an antibody against β₄ (not shown) or an antibody against a 500 kDa (HD1) component of the hemidesomo-somal plaque. The staining pattern of BP230 appeared to be slightly different. Although there was complete overlap of staining by anti-BP230 antiserum with that produced by antibodies specific for α₆β₄ and HD1, the areas stained by anti-BP230 antiserum were somewhat smaller, and in the peripheral regions of cells, where staining for α₆β₄ and HD1 in general was less intense, no staining with anti-BP230 antiserum was seen.

We also compared the expression of α₆β₄ with that of vinculin, actin and keratin. As shown in Fig. 4b,c and f staining for vinculin and actin filaments is concentrated in the periphery of the cells, while α₆β₄ and keratin filaments were primarily localized in the central parts of the cell. In those cases in which keratin filaments were seen to extend into the peripheral regions of the cell and to be in contact with the cell-substratum surface, the integrin α₆β₄ could be detected at the ends of the keratin filament bundles (Fig. 4d). Similarly, we could detect vinculin-containing sites (focal contacts) at the ends of actin stress fibers (Fig. 4a). Double immunofluorescence labeling of RAC-11P/SD cells with a monoclonal murine antibody to vinculin and rabbit antisera to cytoplasmic domains of integrin α and β sub-units, revealed that the α₃, α_v and β₁ subunits are localized in these vinculin-containing focal contacts (Fig. 5). These localizations are consistent with the presence of these inte-grins in focal contacts of cultured keratinocytes (Marchisio et al., 1991). We did not observe colocalization of α₆β₄ with vinculin (Fig. 4e) nor of vinculin with keratin (Fig. 4b).

To characterize further the hemidesmosomal structures in RAC-11P/SD cells, we compared the ultrastructural local-ization of the hemidesmosomal components in RAC-11P/SD cells with their known localization in hemidesmo-somes. Ultrathin cryosections processed for immunoelec-tron microscopy of α₆β₄ by immunogold labeling showed a patchy distribution of this integrin associated with hemidesmosomal structures at the basal surface of RAC-11P/SD cells (Fig. 6a). Keratin filaments labeled with an affinity-purified antibody preparation against human epi-dermal keratins appeared to anchor into the cytoplasmic plaque of these hemidesmosomal structures (Fig. 6d). BP230 was present in the dense hemidesmosomal plaque abutting the plasma membrane (Fig. 6b), while HD1 was found in a second dense plaque located at some distance from the plasma membrane (the inner hemidesmosomal plaque) and in apparent association with intermediate fila-ments (Fig. 6e). These localizations of HD1 and BP230 are consistent with those reported in hemidesmosomes in skin (Westgate et al., 1985; Hieda et al., 1992). No anti-HD1 or anti-BP230 was bound in the cytoplasmic plaque of desmo-somes, but staining of desmoplakins I and II could be demonstrated (Fig. 6c). Thus, the hemidesomosmal struc-tures in RAC-11P/SD cells are comparable to those in skin.

To confirm that all of the components detected in the above IF and EM studies were indeed synthesized by RAC-11P/SD cells, we performed immunoprecipitations of lysates of metabolically labeled RAC-11P/SD cells. As shown in Fig. 7, the α₆ subunit of 120 kDa and the β₄ sub-unit of 200 kDa were immunoprecipitated by both anti-α₆ (mAb GoH3) and anti-β₄ (mAb 346-11A) as these two com-ponents are complexed to each other. These two subunits could also be immunoprecipitated with the mAb 1A10 specific for the cytoplasmic domains of the α_6A variant, but less efficiently than with antibodies against the extracellu-lar domains of α₆ and β₄. An antibody specific for the cyto-plasmic domain of human α_6B was also included in this analysis. However, immunoprecipitation by this antibody was not expected because the corresponding epitope is not conserved in murine α_6B (Hierck et al., 1993). Moreover, reverse transcription-PCR (RT-PCR) analysis indicated that RAC-11P/SD cells do not express mRNA for the α_6B vari-ant (not shown). Indeed, no antigens were recognized by this antibody in lysates of RAC-11P/SD cells. Auto-immune serum from a patient with bullous pemphigoid and the murine mAb 121 directed against the hemidesmosomal plaque proteins BP230 and HD1, respectively, immuno-precipitated components of 230 and 500 kDa, consistent with previous reports (Stanley et al., 1981; Hieda et al., 1992). Furthermore, RAC-11P/SD cells synthesized desmo-plakins I and II, immunoprecipitated by mAb 11-5F, and vinculin, immunoprecipitated by a polyclonal anti-vinculin antiserum. E-cadherin associated with catenins α, β and γ (plakoglobin) were precipitated by a polyclonal anti-E-cad-herin serum.

The expression of integrins on RAC-11P/SD cells as shown by immunofluorescence was further studied by immuno-precipitation experiments using ¹²⁵I surface-labeled cells and a panel of polyclonal antisera against cytoplasmic domains of integrin subunits and a rabbit antiserum against purified murine fibronectin receptor. As shown in Fig. 8, the polyclonal antisera against α_3A immunoprecipitated α_3A and β₁, which on SDS-PAGE migrated as bands of 120 and 130 kDa under reducing conditions and of 150 and 120 kDa under non-reducing conditions. The same proteins were also recognized by the various antisera against β₁, but in addition these sera immunoprecipitated a minor band of 160 kDa under reducing conditions, which matched the size of α₂. However, a polyclonal antiserum against α₂ did not immunoprecipitate α₂β₁. Further, no α_v subunit was immunoprecipitated from RAC-11P/SD cells by the poly-clonal antiserum to α_v, although it reacted with these cells in immunofluorescence. The inability of the anti-α_v and anti-α₂ sera to immunoprecipitate their corresponding inte-grin subunits may be due to a conformational change in the cytoplasmic domain, induced by solubilization of the recep-tors. Antibodies against α₆ (mAb GoH3) and α_6A (poly-clonal antisera) immunoprecipitated α_6A and β₄ subunits. No integrin subunits were immunoprecipitated with antis-era to α₁, α₄, α₅, α_7B, α₈ and β₅ subunits.

We then performed double immunofluorescence microscopy to see whether matrix components that might induce the assembly of hemidesmosomes and be responsi-ble for the localization of α₆β₄ in hemidemosomes, could be identified. Figs 9 and 10 show that the matrix of RAC-11P/SD cells reacted with antisera against kalinin and intact EHS laminin as well as against various fragments of laminin (E3, E4 and E8, not shown) and with anti-fibronectin, but not with antibodies against collagen types IV and VII. The pattern of staining by antibodies to kalinin and laminin was similar to that seen by anti-α₆β₄, sug-gesting that both matrix components are concentrated underneath the α₆β₄-positive hemidesmosomes (Fig. 9). In contrast, staining with anti-fibronectin produced a fibrillar pattern. These fibronectin-positive fibrils run parallel to microfilament bundles and are associated with vinculin-pos-itive focal contacts (Fig. 10).

Immunogold labeling, using ultrathin cryosections of RAC-11P/SD cells that were detached from the culture dish by scraping in PBS, confirmed the restricted localization of kalinin underneath the hemidesmosomes (Fig. 11a). Anti-kalinin antiserum did not react with hemidesmosomes in cells that had been treated with EDTA in PBS and further dissociated by pipetting (not shown). It is likely that in this procedure kalinin remains bound to the substrate surface. By contrast, using the same technique the integrin α₆β₄ was found to be strongly associated with hemidesmosomes (Fig. 11b).

To further demonstrate that kalinin is a major component of matrices of RAC-11P/SD cells, cell-free attached matri-ces were dissolved in SDS-sample buffer and used for analysis by western immunoblotting. Several high molecu-lar mass bands (400, 200, 190, 150, 145, 130 and 105 kDa) as well as some low molecular mass proteins, possibly from serum in the culture medium, were detectable by silver staining (Fig. 12). Three of the high molecular mass pro-teins, namely those of 145, 130 and 105 kDa were found to react with an antiserum to kalinin, consistent with pre-vious reports (Delwel et al., 1993). The 145 kDa band was stained more strongly than the 130 and 105 kDa bands. Silver stained bands of 400 and 200 kDa, which, however, were not always found and two weakly stained bands of 190 and 150 kDa did not react with anti-kalinin serum. The four bands of 150, 145, 130 and 105 kDa are those expected for kalinin and represent the A, B2, B1 chains and a pro-teolytical degradation product of the 145 kDa B2 kalinin chain, respectively. It is possible that the 190 kDa band rep-resents the precursor of the 150 kDa kalinin A chain which, during biosynthesis, becomes proteolytically processed (Marinkovich et al., 1992a). The non-reactivity of the 190 kDa precursor and 150 kDa mature kalinin A chain bands with the anti-kalinin serum is probably due to the fact that this antiserum is directed against human kalinin, which apparently does not cross-react with the murine kalinin A chain.

Because in the immunofluorescence studies a positive reaction was found with antisera to laminin on matrices of RAC-11P/SD cells, we also expected to detect this com-ponent in the isolated matrices, but surprisingly no reaction was seen using an anti-laminin serum (Fig. 12) or sera against various fragments of laminin (not shown). A likely explanation is that laminin is not tightly bound to the poly-styrene surfaces on which the RAC-11P/SD cells are grown and is dissolved in the buffer used to detach RAC-11P/SD cells and then removed by washing.

When anti-fibronectin serum was used in immunoblot-ting of RAC-11P/SD matrix, a weak band was observed at 230 kDa, which is the molecular mass of fibronectin. Since we did not see a band at this position in the silver-stained gel, the amount of fibronectin in the matrix of RAC-11P/SD cells is obviously very small.

To investigate whether the three components, fibronectin, kalinin and laminin, which we detected in matrices, are also secreted by RAC-11P/SD cells, [³⁵S]methionine-labeled culture media were subjected to immunoprecipitation with polyclonal antiserum against fibronectin, kalinin, laminin and laminin A chain, as well as antisera against nidogen and collagen IV. In Fig. 13, it can be seen that fibronectin, kalinin and laminin were indeed detectable in the culture medium, in addition to collagen type IV, the latter in rela-tively low amounts. No nidogen was immunoprecipitated from RAC-11P/SD cells. The pattern of bands, immuno-precipitated with anti-kalinin serum from the culture medium, resembled that of matrices detected by silver stain-ing, i.e. three bands of approximately 150, 140 and 130 kDa, representing the A, B2 and B1 chains, respectively, and two bands of 190 and 105 kDa, representing the A chain precursor band and the B2 chain proteolytically processed band. The 150 kDa band is in fact not a single band but a closely spaced doublet, which may have arisen from alternative proteolytic processing.

The presence of α₆β₄ and the matrix components kalinin and laminin in hemidesmosomes in RAC-11P/SD cells, suggests that binding to these components is mediated by α₆β₄. We, therefore, studied in detail the adhesion of RAC-11P/SD cells to kalinin and laminin as well as to fragments of laminin and to fibronectin. Two sources of kalinin were used; cell-free attached matrices, which appeared to con-tain kalinin but not laminin (Fig. 12) and kalinin that was purified by antibody affinity chromatography from the cul-ture medium of human UMSCC-22B cells. The purity of this latter kalinin preparation is shown in Fig. 14. The four major bands of 150, 145, 130 and 105 kDa, which appeared by electrophoresis under reducing conditions, all reacted in immunoblotting with antiserum to kalinin. No laminin A or B chains or nidogen could be detected in the immunopuri-fied kalinin preparation, which therefore can be presumed to be absent from the preparation.

Studying the kinetics of adhesion, it appeared that adhe-sion of RAC-11P/SD cells to kalinin occurred very rapidly, to laminin a little more slowly and rather slowly to fibronectin (Fig. 15). The kinetic profiles of adhesion to the isolated kalinin-matrices and to immunopurified human kalinin (Fig. 15A) were virtually the same, as were those to fragment E8 and intact laminin (Fig. 15B). RAC-11P/SD cells also showed good binding to fragment E3. Maximum adhesion to this fragment was similar to that to intact laminin, but it was reached slightly later (90 minutes versus 60 minutes). Binding to fragments P1 and E4 was also observed but this was a relatively late event and occurred after about 90 minutes when the cells began to bind to BSA.

The adhesion kinetics of RAC-11P/SD cells that had been pretreated with cycloheximide to block protein syn-thesis were similar to those of untreated cells, except that binding to E4-, P1- and BSA-coated substrates was no longer seen (not shown). This adhesion was therefore prob-ably due to the deposition of matrix components (e.g. kalinin and laminin) produced endogenously during the adhe-sion period. In further experiments we always used cyclo-heximide-pretreated RAC-11P/SD cells to ascertain that adhesion indeed occurred to the exogenous substrate. As a control for the specificity of interaction with laminin and to exclude the possibility that adhesion to the isolated kalinin-matrices is due to contamination with laminin, we also tested the effect of an antiserum to fragment E8 of laminin on adhe-sion of RAC-11P/SD cells to these matrix components. As shown in Fig. 16, adhesion to laminin could be blocked by anti-E8 at all times, while no blocking of adhesion was seen to kalinin-matrices. The absence of blocking of adhesion to kalinin-matrices by anti-E8, is consistent with the biochem-ical data that also indicated that the isolated kalinin-matrices were not contaminated with EHS laminin.

To determine whether α₆β₄ or β₁-containing integrins are involved in the binding of RAC-11P/SD cells to matrix components, cell adhesion assays were carried out in the presence of the mAb GoH3 against α₆ and a polyclonal antiserum against α₅β₁. As shown in Fig. 16, adhesion to kalinin (immunopurified and isolated matrices) was not affected by mAb GoH3 or a polyclonal antiserum to α₅β₁; combination of the two antibodies also caused no inhibi-tion (not shown). Adhesion to laminin or fragment E8 was partially inhibited by anti-α₆ mAb during the first minutes of incubation but after 60-90 minutes this was no longer noticeable. No inhibition was induced by the polyclonal antiserum to α₅β₁ added either alone or in combination with mAb GoH3 (not shown). Binding to E3 was slightly affected by anti-α₆ and anti-β₁ only in the first 15 minutes. Adhesion to fibronectin was strongly inhibited by the anti-β₁ antiserum but inhibition diminished in time similar to that induced by the mAb GoH3 on intact laminin and its fragment E8. As expected, the anti-α₆ mAb did not affect the adhesion of RAC-11P/SD cells to fibronectin. Thus, in antibody inhibition experiments the involvement of α₆β₄ in the binding to laminin and fragment E8, and of α₅β₁ to fibronectin could only be demonstrated in the initial phase of adhesion. An explanation may be that subsequent changes occur in the microenvironment or structural changes in the receptor molecules themselves, which might lead to an increase in the affinity of the receptors for their ligands in such a way that antibodies can no longer effec-tively exert their inhibitory function. There was no evidence for a role of α₆β₄ in adhesion to kalinin.

When RAC-11P/SD cells had been treated with a combi-nation of deoxyglucose and azide, adhesion to the isolated kalinin-matrices could still be observed but not to laminin or fibronectin (Fig. 17A). Remarkably, adhesion to the kalinin-matrices was now completely inhibitable by mAb GoH3 in the first 60 minutes and still strongly inhibited after that period of time. As expected, adhesion of the treated RAC-11P/SD cells to kalinin was not inhibited by the polyclonal antiserum to α₅β₁. Incubation with a com-bination of anti-α₆ and anti-α₅β₁ did not reduce adhesion to kalinin further than was already seen with anti-α₆ alone. Although pretreatment of RAC-11P/SD cells with cytocha-lasin B also made the cells more sensitive to inhibition by GoH3, particularly in the first periods of adhesion, large percentages of cells were still able to bind to the kalinin-matrices. Inhibition experiments with immunopurified human kalinin were performed to confirm these results. Because the amount of purified kalinin was limited, we only tested the effects of antibodies after 30 and 60 minutes incu-bation. At these points of time the results were identical to those of the isolated kalinin matrices (Fig. 17B).

Adhesion to laminin still occurred when cells had been treated with cytochalasin B. However, because adhesion was now more efficiently inhibited by GoH3, the treatment must have reduced the strength of adhesion. No adhesion to fibronectin was observed when cells had been pretreated with cytochalasin B. From these results we conclude that α₆β₄ is a specific receptor for both kalinin and laminin, and that strong adhesion to these matrix components defined as adhesion that cannot be blocked by GoH3 requires meta-bolic energy and is in part dependent on an intact actin-containing cytoskeleton. Furthermore, the results indicate that the adhesion to the isolated matrices is mainly, if not exclusively, mediated by kalinin.

To determine whether the diminishing inhibitory effect in time of the mAb GoH3 on adhesion of cycloheximide-treated RAC-11P/SD cells to laminin and kalinin (isolated matrices) is correlated with a redistribution of α₆β₄ on the cell surface, RAC-11P/SD cells were allowed to attach for various periods of time to coverslips coated with these sub-strates. These were then fixed and incubated with mAb GoH3. We also pretreated cells with mAb GoH3, before plating them for various periods of time. On cells adhered to laminin, α₆β₄ initially seems to be polarized in small spots and some larger spots at surfaces attached to the sub-stratum (Fig. 18). During further incubation these spots increased in size and were arranged into ring-like structures. After 30-60 minutes the majority of cells contained such α₆β₄-positive rings. Some spreading of cells on laminin is seen after 60 minutes and then the ring-like structures are lost and α₆β₄ is found in a spotty pattern underneath the cells, although more concentrated under certain parts of the cell. A similar development was seen in cells pretreated with anti-α₆ but the formation of the rings occurred some-what later (not shown). On kalinin, a similar spotty initial distribution pattern was seen, but the stage in which rings were formed during adhesion to laminin does not occur.

Instead, the cells spread and a hemidesmosomal staining pattern became apparent. Spreading of cells that had been pretreated with GoH3 was slightly retarded and then α₆β₄ was found to be more concentrated under the central parts of the cell (not shown). After 120 minutes incubation, there was no longer any difference in the distribution of α₆β₄ in treated and untreated cells. From these results we may con-clude that inhibition of the adherence of RAC-11P/SD cells to laminin and kalinin decreases when α₆β₄ is clustered at the basal cell surface. Complete loss of inhibition by GoH3 may occur when cells have formed rings (on laminin) or when they have spread (on kalinin).

Since the efficacy of GoH3 to inhibit cell adhesion is enhanced when cells are treated with cytochalasin B or deoxyglucose/azide, we also examined whether α₆β₄ was redistributed in these treated cells. When cells were incu-bated for 30 minutes on a laminin coated surface, only cytochalasin B-treated cells adhered and the adherent cells had not formed ring-like structures. Concentration of α₆β₄, however, still occured at the basal surface of cells, as evi-denced by a spotty staining with anti-α₆. In cytochalasin B-treated cells adhered for 30 minutes to kalinin, α₆β₄ was localized in fine spots at the cell base and in several cell projections extending from the cell body. A hemidesmoso-mal pattern of staining was no longer seen and also spread-ing of the cells was not observed. In cells treated with deoxyglucose and azide, redistribution of α₆β₄ was pre-vented more efficiently than in cytochalasin B-treated cells. Furthermore, spreading of the cells was inhibited. Thus, there is a correlation between inhibition of adherence by anti-α₆ and clustering of α₆β₄.

The morphology of cells attached to kalinin-matrices is essentially different from that of those attached to laminin (Fig. 19). Cells adhered to kalinin have spread within 30 minutes, whereas on laminin cells retain their rounded mor-phology until they have produced their own matrix. When laminin was coated on top of kalinin, adhesion of RAC-11P/SD cells to this combined substrate appeared to be even more efficient than on laminin or kalinin alone and cells had fully spread within 30 minutes. Thus, the combination of laminin and kalinin, a situation that resembles adherence to matrices produced endogenously, is a more efficient sub-strate than one of the two proteins separately. The fact that the RAC-11P/SD cells do not spread on laminin alone may reflect their inability to organize their actin cytoskeleton when adhered to this substrate, presumably because these cells do not express a laminin receptor that, via its cyto-plasmic domain, can interact with actin-associated proteins. However, there might, in addition to α₆β₄, be a separate receptor for kalinin, which associates with the actin-con-taining cytoskeleton and induces cell spreading.

In this paper, we describe the characterization of a cell line that differentiates in culture into squamous epithelium and is capable of forming well-defined hemidesmosomes. The cell line was originally established from a mouse mam-mary tumor and was selected because it characteristically produces foci of flattened cells. A subclone, RAC-11P/SD, undergoes even more extensive differentiation. We show that the differentiation of these cells simulates the differ-entiation in skin in vivo, in terms of the production of mul-tilayered cells with increasing stratification and terminally differentiated cornified cells. The cell line is also impor-tant for its ability to produce hemidesmosomes, which con-tain the hemidesmosomal components HD1, BP230 and α_6Aβ₄. The presence of two other components of hemidesmosomes, the cell surface protein BP180 (Klatte et al., 1989; Diaz et al., 1990) and a 200 kDa hemidesmo-somal plaque protein (Kurpakus et al., 1991b) have not been tested because no antibodies reacting with these murine antigens were available. To date, three other cell lines have been reported to produce hemidesmosomes (Owaribe et al., 1990; Ridelle et al., 1991; Hieda et al., 1992). One of these, the 804G cell line, was established from a rat bladder carcinoma and has been used to study the assembly of hemidesmosomes (Jones et al., 1991; Ridelle et al., 1992).

RAC-11P/SD cells synthesize various extracellular matrix proteins, including kalinin, laminin, fibronectin and collagen type IV and, except for collagen type IV, they all can be detected in the matrices of these cells by immuno-fluorescence. Their localization, however, is different. While laminin and kalinin appeared to be concentrated under α₆β₄-positive hemidesmosomes, fibronectin was found in association with microfilaments and focal con-tacts.

Treatment of RAC-11P/SD cells with EDTA in PBS results in detachment of the cells from the matrix. Bio-chemical analysis of these cell-free matrices confirmed the presence of kalinin and fibronectin. Laminin, however, could not be detected, possibly because it may only be loosely bound to the polystyrene surface and therefore be removed during the preparation of the matrices.

Binding of RAC-11P/SD cells to their own isolated matrices is probably exclusively mediated by kalinin. This is suggested by the facts that adhesion to these cell-free attached matrices followed the same kinetics as to immunopurified kalinin and that the effects of drugs and antibodies on adhesion were similar. The concentration of fibronectin in the isolated kalinin-matrices is probably too small to significantly contribute to their adhesive properties (fibronectin can only be detected by immunoblotting, but not by silver staining). In support of this, antiserum to fibronectin did not interfere with the adhesion of RAC-11P/SD cells to their own isolated matrices (not shown). Furthermore, (stimulated) K562 cells, which express the fibronectin receptor, α₅β₁ (Delwel et al., 1993), did not adhere to isolated matrices of RAC-11P/SD cells. However, fibronectin, when coated to surfaces at high concentrations, as used in cell attachment assays, supports adhesion of RAC-11P/SD cells, and so does laminin. The adhesion to kalinin and laminin occurs relatively more quickly than to fibronectin and the integrin α₆β₄ appeared to be involved in adhesion to both laminin and kalinin, whereas adhesion to fibronectin is mediated by β₁-containing integrins.

Binding of RAC-11P/SD cells to laminin is only partially blocked by mAb GoH3 in the initial phase of adhesion, which might be explained by the involvement during adhe-sion of other receptors for laminin not containing the α₆ subunit. One such receptor might be the integrin α₃β₁, which has been shown previously to be a receptor for laminin on certain cells (Wayner and Carter, 1987; Elices et al., 1991; Gehlsen et al., 1992). However, a polyclonal anti-β₁ antiserum had no effect on the adhesion to laminin and if added in combination with mAb GoH3, blocking was not stronger than by GoH3 alone. In this context, it is per-haps important to mention that RAC-11P/SD cells do not spread on laminin, which also might be due to absence of a β₁ integrin that can bind to laminin and establish the con-nection of the actin cytoskeleton to the plasma membrane. Thus, although α₃β₁ is expressed on RAC-11P/SD cells, there is no evidence that it plays a role, in addition to α₆β₄, in adhesion of the cells to laminin. In fact, neither in anti-body blocking assays (Sonnenberg et al., 1990) nor in affin-ity chromatography experiments (Sonnenberg et al., 1991a,b), have we been able to demonstrate that α₃β₁ medi-ates adhesion to murine EHS laminin. These findings are further supported by the results of a recent study in which cDNA for α₃ was transfected into K562 cells and the bind-ing properties of the transfected cells were examined (Weitzman et al., 1993; Delwel et al., unpublished data).

Experiments using fragments of laminin indicated that RAC-11P/SD cells express at least two distinct laminin receptors, one which binds to fragment E8 and another to fragment E3. Binding of RAC-11P/SD cells to fragment E8 followed the same kinetics as binding to intact laminin and was also similarly partially blocked by mAb GoH3 in the initial phases of adhesion. Moreover, adhesion to intact laminin could be inhibited with an antiserum to fragment E8. Therefore, a major binding site of laminin is located on fragment E8 and adhesion of RAC-11P/SD cells to it is mediated by α₆β₄. Binding to E3 was not significantly blocked by GoH3, suggesting that cell binding to this frag-ment is mediated by a different receptor(s), possibly a pro-teoglycan(s) (Sonnenberg et al., 1989; Sorokin et al., 1992). At first sight, as mentioned, the presence of two laminin receptors on RAC-11P/SD cells might explain why block-ing by GoH3 to intact laminin was not complete. However, blocking of adhesion to fragment E8 was also incomplete. Therefore, this explanation is unlikely unless the second receptor, which interacts with fragment E3 of laminin, also reacts with fragment E8. Another explanation for the par-tial blocking could be redistribution of α₆β₄ during adhe-sion to laminin. This notion is supported by the fact that after a longer period of adhesion blocking becomes com-pletely negative. This could be due to a completion of the redistribution of α₆β₄. That redistribution does indeed occur was shown by immunofluorescence analysis with anti-α₆ mAb. During adhesion to laminin, α₆β₄ first becomes polar-ized at the basal surface and then is organized into ring-like structures. When cells have been treated with cytocha-lasin B, the formation of ring-like structures is no longer noticeable and adhesion to laminin can now be completely blocked by GoH3. Apparently, the organization of α₆β₄ into ring-like structures requires an intact actin cytoskeleton and is associated with the incapacity of GoH3 to block adhe-sion. Polarization of α₆β₄ to the basal surface, however, still occurs as is expected for an integrin that interacts with the intermediate filament rather than with the actin-con-taining cytoskeleton. It is likely that in cells treated with deoxyglucose/azide, interaction with the keratin filaments cannot take place and, therefore, α₆β₄ cannot be redistrib-uted. Hence, binding is not stabilized. A final possibility is that activation of α₆β₄ during the adherence is responsible for the diminished blocking effects of the anti-α₆ mAb.

While metabolic energy and clustering of α₆β₄ seems to be essential for adhesion to laminin, this is not the case for adhesion to kalinin. Cells that have been treated with deoxyglucose/azide can still bind to kalinin, although less efficiently than untreated cells. Binding is mediated by the integrin α₆β₄ because GoH3 inhibits adhesion completely. A lower affinity of α₆β₄ for laminin than for kalinin could be responsible for the fact that deoxyglucose/azide-treated cells do not bind to laminin. The partial blocking by GoH3 of the adherence of cytochalasin B-treated cells to kalinin in the initial phase, but the complete blockage of adherence to laminin supports this notion.

The inability of GoH3 to block adhesion of untreated cells to kalinin in initial phases of adhesion, when cluster-ing is assumed to be not yet complete, might be due to the presence of a second receptor with high affinity for kalinin. We propose that the integrin α₃β₁ is this additional recep-tor for kalinin. It has been shown in antibody inhibiton assays (Carter et al., 1991) as well as in transfection studies using K562 cells (Weitzman et al., 1993; Delwel et al., unpublished), that α₃β₁ interacts with epiligrin, which is similar or identical to kalinin. In support of this we found that α₃β₁ localizes in focal contacts of cycloheximide-treated RAC-11P/SD cells when they adhere to kalinin (not shown). Furthermore, the spreading of RAC-11P/SD cells on kalinin is consistent with the presence of a β₁-contain-ing integrin interacting with this substrate and being respon-sible for attachment of actin filaments to the plasma mem-brane, ultimately leading to a reorganization of the cytoskeleleton. Moreover, treatment of cells with cytocha-lasin B prevents spreading on kalinin. Why we found no blocking by the anti-β₁ antibody of adhesion to kalinin is not clear. It could be due either to a poor quality of the anti-β₁ antibody or to a minor role of α₃β₁ in adhesion to kalinin compared to that of α₆β₄. If the latter were true, this would be essentially different from the relative role of these two integrins in keratincoyte adhesion (Carter et al., 1991). Finally, a third receptor might be involved that could be the hemidesmosomal BP180 antigen.

Beause α₃β₁ integrins can no longer be clustered in cytochalasin B-treated cells, adhesion probably depends only on α₆β₄ and consequently GoH3 can more efficiently block the adhesion of these treated cells than of untreated cells. GoH3 does not block adhesion completely, but this might be explained as mentioned above for adhesion to laminin, by redistribution of α₆β₄ on the cell surface. The possibility that α₃β₁ mediates adhesion to kalinin without the requirement of clustering seems less likely because binding to kalinin was not blocked by anti-β₁ antibodies.

In previous investigations we found that RAC-11P cells did not bind to E8 (Sonnenberg et al., 1990), whereas in this study using RAC-11P/SD cells we did see binding to E8. We have no good explanation for this discrepancy except that the two cell lines, although derived from each other, may not be identical. They may differ in the expression or the degree of activation of α₆β₄. It is also possible that the observed differences are related to differ-ences in the activity of the various batches of E8 used. How-ever, there is no difference in the two investigations in bind-ing to E3 and both cell lines also do not bind to E4 and P1.

In summary, we show that RAC-11P/SD cells, which share many properties with normal keratinocytes, utilize the integrin α₆β₄ for adhesion to both laminin and kalinin.

We thank Drs P. Bernardi, G. Tarone, R. O. Hynes, L. F. Reichardt, E. Ruoslahti, R. H. Kramer, B. Geiger, R. Burgeson, E. Engvall, R. Kemler, D. Boorsma, D. Garrod, K. Owaribe, S. J. Kennel, L. H. Jaspars and F. C. S. Ramaekers for kindly pro-viding antibodies, R. Timpl for antibodies, laminin and its prote-olytic fragments, and E. Roos and C. P. Engelfriet for critical read-ing of the manuscript. L. Oomen is thanked for help with the confocal laser scanning microscope and N. Ong for photographic work. This work was supported by grants from the Dutch Cancer Society (NKI 91-260) and the Nierstichting (The Dutch Kidney Foundation; C91.1179).