Differentiation (1 983) 23 : 189-205 Differentiation !c) Springer-Vcrlag 1983

Spatial Distribution of Proteins Specific for Desmosomes and Adhaereas Junctions in Epithelial Cells Demonstrated by Double Immunofluorescence Microscopy Benjamin Geiger’, Erika Schmid ’, and Werner W. Franke ’

Department of Chemical Immunology, The Weizmann Institute of Science, Rehovot 76100, Israel Department of Membrane Biology and Biochemistry. Institute of Cell and Tumor Biology, German Cancer Research Center,

D-6900 Heidelberg, Federal Republic of Germany

Abstract. The spatial relationships between the protein con- stituents of two junctional structures, adhaerens junctions and desmosomes, were determined by double immunofluor- escence microscopy using marker proteins specific for these structures. Adhaerens junctions were visualized by immuno- fluorescent labeling for the membrane-associated protein vinculin and by their association with actin filaments. Des- mosomal components were identified by labeling with anti- bodies to a group of minor desmosomal plaque proteins (DPl antigens) and their association with filaments stained by cytokeratin antibodies. Double immunofluorescence mi- croscopy of these components was performed in several tissues and cultured cells, including intact intestine, disso- ciated intestinal cells, and two morphologically different types of epithelial cells, cultured bovine kidney (MDBK), and mammary gland (BMGE) epithelial cells. This allowed the direct demonstration that each filament system is asso- ciated exclusively with its specific membrane-bound junc- tional protein. Vinculin and DP1-protein were found in dis- tinct sites in the subapical intercellular junctional complex of intestinal epithelium and MDBK cells. Cell-substrate focal contacts contained vinculin and actin and showed no apparent relationships to the tonofilament system whereas intercellular contacts of BMGE cells were characterized by positive staining for DPI-protein and associated cytokera- tin filaments. Immunolabeling of the cultured cells at differ- ent intervals after plating for the cytoskeletal elements and their membrane anchorage proteins was used to determine the temporal sequence of their organization. We propose that this approach may be used for the molecular definition and identification of cellular contacts and junctions as well as for studies of junction topology, dynamics of junction- cytoskeleton interactions, and junction biogenesis.

Introduction

Many studies with cultured cells as well as with intact tis- sues have indicated that intercellular contacts play a major role in the control of a variety of cellular properties and processes, including shape and growth, structural and func- tional polarity, differentiation and locomotion (for reviews see (56, 59, 603). Unfortunately, the identity of the specific types of cell contacts involved in each of these processes is in most cases not clear. Moreover, with the exception of gap junctions (for reviews see [32, SO]), most of the spe- cific contact regions such as tight junction, adhaerens (focal)

junctions, and desmosomes are defined mainly by morpho- logic criteria. Their molecular components and properties are largely unknown.

Two of these cellular junctions are regularly associated, at their cytoplasmic faces, with filamentous cytoskeletal structures. Desmosomes and hemidesmosomes are specifi- cally associated with intermediate-sized filaments of the cy- tokeratin type (tonofilaments) which bind to the membrane through a layer of densely interwoven fuzzy material, the desmosomal plaque [5, 10, 36, 561. The other type is the adhaerens junction, also referred to as “focal junction”, which is characterized by the association with actin micro- filament bundles and includes structurally diverse sites such as the zonula adhaerens of polarized epithelia, peripheral dense plaques of smooth muscle, focal contacts of cultured cells, and the fascia adhaerens of cardiac muscle [23, 561. These two distinct types of junction show certain similari- ties. They are often found closely spaced in the apical junc- tional complex of epithelial cell layers, where each of them is symmetrically organized in the two contacting cells. Both can be formed in regions of contact not only between adja- cent cells but also in sites of attachment to extracellular matrix or artificial support where they appear as hemides- mosomes or focal contacts, respectively. In both types of junction the associated filaments apparently bind to the membrane indirectly through some peripheral membrane- attached protein layer (‘plaque’).

Recently, it has been shown that an intracellular consti- tuent associated with the focal junctions is the protein vin- culin [4, 21, 23, 27, 281. In fact, the presence of vinculin at the contact area, in intimate association with the termini of actin bundles, was used for the definition of focal junc- tions [23]. Recent progress has also been achieved in the Molecular characterization of desmosomes. These studies include identification of the components which constitute the intercellular core (midline structure or ‘desmogloea’) of the desmosomes (e.g. [30]) as well as high molecular weight polypeptide components of the desmosomal plaque [18], including the two major desmosomal plaque compo- nents, the desmoplakins [19, 20, 411. The availability of specific antibody reagents to vinculin and to certain desmo- soma1 plaque proteins as well as to the protein constituents of the filamentous systems linked to these junctions, has enabled us to study, using immunofluorescence microscopy, the relationships between the constituents of the two types of junction and the cytoskeletal filaments associated with them.

190

Methods

Tissues and Cells. Bovine small intestine was freshly taken at slaughter, cut into small pieces, and directly frozen in isopentane cooled with liquid nitrogen. Before sectioning the frozen tissue pieces were stored at -70" C. For prepara- tion of dissociated intestinal epithelial cells segments of small intestine were cut open and rinsed thoroughly with phosphate buffered saline (PBS). Mucosal cells were scraped off using glass slides, resuspended in PBS, and fur- ther dissociated by pipetting. Drops of this cell suspension were transferred to coverslips, briefly airdried, and then incubated for 2 min in 100 mM MES buffer (2-(N-morpho- 1ino)ethanesulfinic acid) pH 6.0, containing 5 mM MgCl,, 3 mM EGTA, and 0.5% Triton X-100. Immediately after this treatment the adherent cells were fixed in methanol at -20" C for 5 min, followed by six brief immersions in acetone at - 20" C, and some washed in PBS at room tem- perature.

Bovine mammary gland epithelial cells were isolated from cow udder and grown in vitro as described previously [12, 541. In the present study we used an established line (BMGE +H) continually grown in Dulbecco's minimal es- sential medium (DMEM) containing 20% fetal calf serum and insulin, hydrocortisone, and prolactin (1 pg/ml each) [54]. MDBK cells from bovine kidney ([40]; ATCC CLL 22) were grown in minimal essential medium Eagle (MEM) supplemented with 10% fetal calf serum.

Antibodies. The following previously documented anti- bodies were used: (1) Antibodies against prekeratin from desmosome-attached tonofilaments of bovine muzzle raised in guinea pigs [ lq . (2) Antibodies against prekeratin from bovine hoof raised in rabbits [33]. (3) Antibodies raised in guinea pigs against an electrophoretically purified high molecular weight polypeptide fraction from bovine muzzle epidermis which react with a group of minor desmosomal plaque polypeptides (antigen 'DPl', [ls]). On bovine cells, these antibodies co-localized with antibodies to desmoplak- ins I and 11, the major polypeptide components of the des- mosomal plaque [19, 20,411. The immunoglobulin fraction was isolated by DEAE cellulose chromatography and ex- haustively preabsorbed on reconstituted filaments of bovine epidermal prekeratin [49] to avoid possible cross-reaction with prekeratin polypeptides. (4) Antibodies against vincu- lin raised in rabbits and guinea pigs [21, 27, 281. (5) Anti- bodies against actin raised in rabbits by repeated injections of chicken muscle actin [2]. (6) Antibodies to chicken gizzard a-actinin raised in rabbits and guinea pigs [24]. (7) Rabbit antibodies to pure chick brain tubulin [25].

Second antibodies were prepared as described pre- viously [24]. Briefly, antibodies against rabbit and guinea pig IgG were raised in goats and affinity-purified. Each antibody was conjugated either with dichlorotriazinyl ami- no fluorescein (DTAF) or with rhodamine-lissamine sul- fonyl chloride as described previously. These antibodies were further absorbed on the heterologous immunoadsor- bent to avoid possible cross-reactivity.

Immunofluorescent Labeling. Frozen sections of bovine in- testine were prepared at -20" C in a cryostat (Frigocut, Model 2700, Jung, Nussloch, Federal Republic of Germany) using steel blades (for 4-5 pm) or Ralph glass knives (for 1-2 pm). Indirect immunofluorescence on these

sections was performed as described previously [16, 17 . Cell monolayers cultured on glass coverslips were treated with Triton-MES buffer (pH6.0) and fixed as described above for dissociated intestinal epithelial cells.

For labeling with a single antibody the procedure was essentially as described previously [12]. For double labeling we used the following combinations of different antibodies : Prekeratin/DPl antigen; prekeratin/actin; prekeratin/ tubulin; prekeratin/vinculin; DP1-protein/actin; DPl- protein/tubulin ; DPI-protein/a-actinin ; vinculin/a-actinin ; vinculin/actin. In all these cases the specific combinations of the first as well as both of corresponding second anti- bodies were applied simultaneously. These rabbitlguinea pig double labels appeared highly specific. More complex was the double labeling for DPl-protein and vinculin. We found that our guinea pig antibodies to chick vinculin bound much better than the rabbit antibodies to mamma- lian vinculin. To allow for double labeling of the two anti- gens we used the four-stage labeling procedure: the speci- men was first reacted with one of the primary antibodies, then it was washed and stained with a suitable fluorescent antibody. After washes, the specimen was labeled with the second set of primary and secondary antibodies. This proce- dure resulted in highly specific labeling of the first antigen with its corresponding fluorophore, while the second- applied labeled antibodies also faintly stained the first anti- gen. However, we did not have any difficulty in determining the exact pattern of distribution of both DP1 protein and vinculin, and we verified our observations by changing the order of labeling of the two junctional proteins.

Electron Microscopy. Cells grown on coverslips were fixed, dehydrated, embedded, and processed for electron micros- copy as described previously [12, 13, 181.

Light Microsocpy. Immunofluorescence microscopy and in- terference-reflection microscopy were carried out in a Zeiss photomicroscope I11 (Carl Zeiss, Oberkochen, FRG) equipped with filter sets for fluorescein and rhodamine fluo- rescence. We have routinely used the following objectives: Plan Neofluar 25/03 and Planapochromat 63/1.4 for fluo- rescence observations and Antiflex 6311.25 for interference reflection microscopy [2].

Results

Junctional Components and Associated Cytoskeletal Ele- ments in Intestinal Epithelium

The ultrastructure of mammalian intestinal cells, notably the apical region of the absorptive cells including the var- ious types of junction, has been described in detail [lo, 34, 491. In particular, the close vicinity of desmosomes and the zonula adhaerens and their association with different filaments has been shown [14, 18,19,26,27]. In the present study the two junctional proteins, vinculin and desmosomal plaque (DP1) protein, and the specific associated cytoskele- tal filaments, actin and cytokeratin, were localized in frozen sections of bovine intestine, using single and double immu- nofluorescent labeling. The immunolabeling for both vincu- lin and actin was most intense in the intestinal smooth mus- cle cells of the muscularis mucosae. In these cells vinculin was organized in rather regularly distributed peripheral stripes (Fig. 1 A) which probably corresponded to the plas-

191

Fig. 1A-F. Imrnunofluorescence labeling of frozen sections of bovine intestine for vinculin (A and B), actin (C and D), DPI-protein Q, and cytokeratin 0. The staining is shown in smooth muscle cells of the muscularis mucosue (A and C) and in the epithelial cells (B, D, E, F). The antibodies used were rabbit antibodies to actin and cytokeratin, guinea pig antibodies to vinculin and DP1-proteins, as well as the respective rhodamine-labeled goat antibodies. The burs indicate 10 prn

ma membrane-bound dense plaques also known to contain a-actinin [28, 551. Actin antibodies, on the other hand, in- tensely stained the whole cytoplasm of the smooth muscle cells (Fig. 1C). By comparison, the labeling for vinculin and actin in the epithelial cells was weak, but was most

prominent in the apical region including the brush border (Fig. 1B and D). Both DP1 protein and cytokeratin ap- peared together exclusively in epithelial cells (Fig. 1 E and F). However, while spotted labeling of DPl protein was dense in the terminal web region (Fig. lE), labeling with

192

Fig. 2A-F. Double immunofluorescence labeling of frozen sections of bovine intestinal epithelium: A and B, labeling for DPI-protein (A) and vinculin (B) using guinea pig antibodies and the four-stage technique. Note the association of vinculin with the subapical region (B) and the enrichment of DP1-protein in that area as well as along the basolateral membranes. C and D Labeling of the same area for DP1-protein (C, guinea pig antibodies) and cytokeratin (D, rabbit antibodies). E and F Labeling of the same area for cytokeratin (E, guinea pig antibodies) and actin (F, rabbit antibodies). Note the enrichment of actin in smooth muscle cells (SM) and at the apical portion of the epithelial cells (arrow). Ears indicate 10 p

193

cytokeratin antibodies was also distributed throughout the entire cytoplasm (Fig. 1 F).

The spatial relationships of these junctional elements were more directly delineated by double immunofluorescent labeling. The results (Fig. 2) indicated that both junctional components, DP1 protein and vinculin, were present in the same cellular regions (Fig. 2A and B). The two labeling patterns, however, were not identical : vinculin label was concentrated predominantly in the apical zone containing the terminal web while spotted DP1-antigen staining was also seen along the basolateral membranes. This difference of labeling and additional internal specificity controls (see Methods) were critical for evaluation of the double staining for both vinculin and DP1-protein on the same specimens.

DP1-protein and cytokeratin (Fig. 2C and D) were both found in the same epithelial cells, but while the cytokeratin labeling was seen in the entire cytoplasm between the subap- ical skeletal disk (cf. [14]) and the basal cell regions, the DPl -antigen labeling was most prominent in the apical portion. Different distributions were also found when label- ings for cytokeratin (Fig. 2E) and actin (Fig. 2F) were com- pared on the same sections; like vinculin, actin was strongly enriched in the muscularis mucosa and in the most apical region (arrow).

To analyze the different components of the intestinal jucntional complex at higher resolution we also used double immunolabeling on freshly dissociated intestinal epithelial cells. The double labeling for vinculin and actin (Fig. 3A, B) indicated that actin was enriched in both the microvilli and terminal web but vinculin was present in significant amounts in the terminal web region only. Moreover, when observed in side views, vinculin labeling appeared as a con- tinuous belt at the periphery of the terminal web. Similarly, double labeling for DP1 protein and actin (Fig. 3C, D) showed the noncoincidence of these two proteins; most strikingly, the enrichment of actin in the brush border zone was located apically from the ring of desmosomes which were visualized as a ring of subapical ‘dots’. Actin was also not conspicuously associated with the desmosomes located along basolateral regions. Cytokeratin was signifi- cantly enriched in the apical region of the cells, but extended arrays of cytokeratin fibrils were also detected throughout the whole cell (Fig. 3F, H). Labeling of the same cells for DP1 antigen (Fig. 3E) showed that individual DPl-protein containing spots are found along the entire cell surface membrane surrounding cytokeratincontaining areas. How- ever, the level of resolution obtained did not allow demon- stration of direct attachments of the arrays of cytokeratin fibrils with individual DP1 protein-positive spots. It was, however, apparent from the double labeling for actin (Fig. 3 G) and cytokeratin (Fig. 3 H) that these two filament systems were preferentially associated with different cellular regions: the former was enriched in the region from the terminal web and up, whereas the keratin-type filaments extended from the terminal web toward the basal cell region.

Especially instructive was the double labeling for vincu- lin and DPI protein on the same cells (Fig. 3 I-L). in many cells the two apical rings were not resolved and seemed to overlap. We found, however, examples in which the con- tinuous ring of vinculin (Fig. 3 J and L) was located towards the cell apex, relative to the circular array of desmosomes (Fig. 31 and K, respectively), as indicated by the matching double lines in Fig. 3. The simultaneous appearance of the

two antigens here was probably due to the use of antibodies from the same species, guinea pig, for double labeling of DP1 antigen and vinculin. In these preparations, the fluo- rescent antibodies added last stained the relevant antigen most strongly but also stained the other antigen very faintly (see Methods). This did not interfere with the specific dis- tinction between the two antigens but rather provided a useful landmark for comparison.

Relationships of Focal Junctions and Desmosomes in Cultured MDBK Cells

Similar to MDCK cells [51], cultured bovine kidney epithe- lial cells of the MDBK line also retained the typical polarity found in the tissue. Electron microscopy of these cells (Fig. 4A-E) revealed an apical junctional complex with well defined tight junction, with gap junctions, with a ring (zonula) of intermediate (adhaerens) junction, and with a subapical ring of distinct desmosomes (maculae adhaer- entes). Circular arrays of microfilaments were found extend- ing parallel to the zonula adhaerens, apparently attaching to the fuzzy plaque regions characteristic of the cytoplasmic face of the intermediate junctions (for details see legend to Fig. 4A-E). Individual junctions of the fascia adhaerens type could also be seen in other, non-apical regions of adja- cent MDBK cells as indicated by the arrowheads in Fig. 4A. The different morphology of the cytoplasmic plaque struc- tures of desmosomes and adhaerens junction is evident from Fig. 4E where both plaque containing junctions were ob- served in close proximity in the same section.

At their ventral surfaces these cells form contacts with the substrate. This was found both by electron microscopy (Fig. 5A, B) and by interference reflection at the light mi- croscope level (Fig. 5C). Most of these ventral contacts had the typical appearance of focal contacts and were associated with bundles of actin microfilaments (Fig. 5A, B). Only rarely, did we recognize hemidesmosome-like formations (not shown).

Double immunofluorescent labeling for the microfila- ment-associated components using vinculin and actin anti- bodies revealed two distinct patterns of organization. By focusing the microscope down to the level of the substrate a specific pattern of organization of the microfilament sys- tem was found. Figures 6A and B show the ventral distribu- tion of vinculin and actin, respectively, in the same group of cells. Vinculin antibodies in this area extensively labeled the cell-to-substrate focal contacts while actin filament bundles seemed to terminate at these sites. In the apical region of the same cells, both vinculin (Fig. 6C) and actin (Fig. 6D) were largely concentrated into a peripheral belt, roughly delineating the zonula adhaerens. Usually, the im- munolabeled, junction-associated vinculin appeared as a thinner belt structure than that found with actin, even when the two exhibited a similar intensity of label. Likewise, a-actinin was concentrated in this ring zone (not shown; for MDCK cells see Refs. 22 and 39). Focusing through the cells did not suggest any continuity between the micro- filament networks at the bottom of the cell and the apical intercellular junctions. It should be pointed out that the junctional belt was apparently present only along the inter- cellular contacts and stopped near the ‘free’ cell margins at the periphery of cell colonies.

The intermediate filament systems present in MDBK cells displayed a distinct pattern of distribution. Bundles

194

Fig. 3A-L. Double immunofluorescence labeling of dissociated bovine intestine epithelial cells: A and B, the same intestinal cells and isolated brush borders double labeled with guinea pig antibodies for vinculin (A, fluorescein) and rabbit antibodies for actin (B, rhodamine). The urrowheucis point to the terminal web region. C and D Double labeling for DPl-protein (C, guinea pig antibodies) and for actin (as in B). E and F Labeling of the same cells for DPI-protein (E, guinea pig antibodies) and cytokeratins (F, rabbit antibodies). G and H Labeling of the same cells for actin (G, rabbit antibodies) and cytokeratins (H, guinea pig antibodies). The arrowhead points to the subapical terminal web. I and J, K and L Double labeling for DPI-protein (I and K) and vinculin (J and L), respectively. The labeling was performed by the four-stage procedure. Matched double burs point to the more apical, vinculin-positive ring and the lower DP1-protein containing ring. The bur indicates 10 pm

195

Fig. 4A-E. Electron microscopy of cultured MDBK cells. A Survey electron micrograph showing the cell monolayer in association with the substratum (S). The apex (a) of the polarized cells contains numerous microvilli. In the subapical region a ring of adhuerens-type junction (zonula adhaerens) is indicated by the arrows. The double-arrow in the lower right area points to individual intercellular udhuerens junctions of the puncta adhuerentiu type. Desmosomes are found along the basolateral membranes (urrowhends). B Higher magnification of the subapical region showing the zonula adhuerens (marked by brackets) with the associated belt of microfilaments. In the more basal area desmosomes are detected (urrowhend). C Sections grazing to the zonula adhuerens close to its lateral association with the membrane. Note periodical arrays of dense f imy material along the microfilament belt (the periodicity is indicated by the vertical bars). D Nearly horizontal sections of lateral intercellular contacts, showing desmosomes in a relatively high packing density and the attached tonofilaments. In E a similar section is shown in which a typical desmosome (arrowhead) is found in the vicinity of an udhaerens type junction (arrow). Bars represents 1 pm (A) and 0.2 pn (B-E)

196

Fig. 5A-C. Electron microscopy (A and B) and interference-reflection microscopy (C) of focal contacts of cultured MDBK cells. Contact regions with the substrate show typical end-on attachment of the filaments to the membrane (the plane of sectioning is nearly perpendicular in A and parallel in B to the direction of the microfilament bundles). The edges of the focal contacts are marked with arrows. Interference reflcction images (C) of cultured MDBK cells show typical focal contacts (arrows) and interference-gray close contacts. The bars in A and B represcnt 0.2 pm, that in C denotes 10 pm

Fig. 6A-D. Double immunofluorescence microscopy of the same group of MDBK cells stained for vinculin (guinea pig antibodies, A and C) and for actin (rabbit antibodies, B and D). The microscope was focused on the ventral focal plane, i.e., a t the level of the substrate (A and B), or on the subapical plane (C and D). Note the actin bundles near the ventral plasma membrane, with their vinculin containing termini (arrows in A and B), and the continuous subapical rings near the zonula adhaerens. The bar indicates 10 pn

197

of vimentin filaments, which are also expressed in ths cell line as in MDCK cells [is], were not specifically associated with intercellular boundaries but rather were often located near the ‘free’ cell margins (not shown; cf. [54]). By con- trast, cytokeratin (Fig. 7B) labeling was associated with fi- brillar arrays that were densely packed around the nucleus but also extended through the whole cytoplasm, often running onto the intercellular boundaries. The DP1 -protein labeling of the intercellular junctions appeared to be pre- dominant along the desmosomal complexes, especially at the apical junction complex of the cells (Fig. 7A). In this region the DP1-protein was organized in arrays of mem- brane-bound dots, each dot representing a desmosomal plaque. Comparison of DPl -protein and cytokeratin label- ing on the same cells (Fig. 7A and B) demonstrated that the cytokeratin fibrils terminate at the DP1-protein contain- ing sites.

To study the patterns and possible interrelationships between focal contacts and desmosomes we double-labeled the same MDBK cells with antibodies to vinculin (Fig. 8A and B) and DP1 protein (Fig. 8 C). As shown before, vincu- lin was associated, at the bottom of the cells, with focal contacts (Fig. 8A) while at the top of the cells (Fig. 8B) it appeared as a continuous peripheral membrane-bound belt. In top views of vertical cell boundaries the dotted arrays of desmosomes appeared to be somehow spatially related to the junction-associated vinculin (compare Figs. 8B and C). However, in many areas where the lateral membrane was viewed obliquely the two structures were clearly distinguished (see matched twin arrows in Fig. 8B and C). Moreover, DPl protein-positive dots were also found, by differential focusing, along the lateral borders of the cells, below the apical junction complex zone. It should be emphasized that whenever the two junctional ele- ments were spatially separated, we did not observe close interactions of DP1-antigen with actin or of vinculin with cytokeratin. Since in the double labeling experiments for vinculin and DPl protein we used guinea pig antibodies as primary reagents for both antigens (see Methods), we tested the spatial relationships of the two junctional systems by using different types of antibodies, namely actin anti- bodies made in rabbits in combination with the guinea pig antibodies recognizing the DPl antigen. The results (Fig. 8 D and E, respectively), were essentially identical to those obtained with vinculin and DP1-protein labeling.

Junctional Elements of Cultured Bovine Mammary Gland Epithelial (BMGE) Cells

Mammary gland epithelial cells (for derivation and mor- phology see Refs. 12 and 54) which have been continually grown in medium containing hydrocortisone, insulin, and prolactin (BMGE + H cells) are also characterized by pro- nounced polar architecture, with numerous microvilli on their upper surface (Fig. 9A; cf. [54]). However, in contrast to MDBK and MDCK cells, BMGE+H cells show, even in dense culture, only small regions of close cell-tocell ap- position. Characteristically, they produce numerous cyto- plasmic projections which either appear as microvilli-like structures or form intercellular ‘bridges’ containing one or several desmosomes (Fig. 9B-D) (cf. [18, 541). BMGE+H cells, unlike many cultured epithelial cell lines including MDBK and MDCK cells [15], contain only intermediate filaments of the cytokeratin type, which are abundant, often

Fig. 7A, B. Double immunofluorescent labeling of MDBK cells for DPI-protein (A) and cytokeratins (B) using the four stage pro- cedure with guinea pig antibodies to both antigens (antibodies to DP1 used first). The arrowhead points to a desmosome which is clearly associated with a cytokeratin filament bundle. The bur indi- cates 10 pn

associated with desmosomes (Fig. 9C, D), and do not pro- duce vimentin filaments [54]. By contrast, several BMGE cell cultures grown without hormones do produce vimentin in addition to cytokeratin filaments [15, 541. Junctions of the adhaerens type are usually sparse in BMGE+H cells and apparently do not form a subapical belt zone (zonula). Gap junctions are often seen in the immediate vicinity of desmosomes (Fig. 9 E).

BMGE + H cell cultures displayed two major types of contact. At their ventral surfaces these cells formed focal contacts with the substrate which were detected by electron microscopy (Fig. 10A, B) or by interference reflection op- tics (Fig. 1OD). Hemidesmosomal formations were also consistently found at ventral surfaces (Fig. 1OC). Immuno- fluorescent labeling for vinculin and actin (Fig. 11 A and B) suggested that the microfilament system was anchored to the membrane predominantly at the ventral region and at sites identified as focal contacts by their interference re- flection contrast images (Fig. 10D).

198

Fig.8A-E. Double immunofluorescence labeling of MDBK cells for vinculin (A and B) and DPl-protein (C). The microscope was focused on the ventral focal plane (A) or on the subapical region (B and C). The staining was performed by the four stage procedure, using vinculin antibodies in the first place. The matched double arrows in B and C point to the more apical vinculin-containing junction (kfi) and the lower ring of desmosomes (right). D and E Double labeling of the same MDBK cells for actin (D, rabbit antibodies) and DP1-protein (E). The bar indicates 10 pm

The labeling for cytokeratin was associated with elabo- rate filamentous networks (Fig. 11 C) and extensive labeling was found in many of the intercellular bridges. Double la- beling for cytokeratins and DP1-protein (Fig. 11 C and D) indicated that the central portions of these cytokeratin-rich contact regions (‘bridges’) were prominently stained as ‘dots’ with antibodies to DP1-protein (Fig. 11 D). In many instances the cytokeratin fibril arrays on either side of the central contacts in the bridges appeared discontinuous [12, 181, the gap presumably corresponding to the desmosome itself. In addition to the junctional DP1-protein, immunola- beling with DPI-antibodies was also found in numerous intracellular dots (Fig. 11 D), many of which might be related to either hemidesmosomal structures at the ventral cell membrane (Fig. 1OC) or to endocytosed desmosomes similar to those found in cells after treatment with trypsin (43,441 or Cazf-chelating agents [35].

The two types of cellular contacts and their respective associated filaments were mutually exclusive. Thus cells double-immunolabeled for cytokeratin and vinculin (Fig. 12A and B) showed no apparent relationships be- tween the locations of the two proteins. Double decoration with antibodies to DP1-protein and vinculin (Fig. 12C and D) revealed that the latter was absent from the bridges

containing DPI-antigen. Similarly no association was found between actin and DP1-potein (Fig. 12E and F). In some regions where actin bundles came close to a desmo- some, a ventral hemidesmosome, or to desmosome-contain- ing bridges, fine focusing indicated that the actin actually did not associate with the specific DPl-antigen structure (e.g. Fig. 12G and H). As previously shown in several other cells (e.g. [12]) the arrays of cytokeratin fibrils were appar- ently different and distributed independently of actin cables and microtubule arrays stained with tubulin antibodies (not shown).

Discussion Intercellular junctions and the corresponding ‘ hemi-junc- tional compelxes ’ (hemidesmosomes, focal contacts) were originally defined by morphologic criteria only [lo, 561. However, the electron microscopic appearance of the var- ious junctions is not consistent in diverse cells, and this uncertainty of definition is especially problematic in the case of desmosomes and adhaerens junctions which both show dense plaque material on their cytoplasmic side and associations with cytoplasmic filaments. So, in several cells it is difficult, if not impossible, to distinguish these two

199

Fig. 9A-E. Electron microscopy of BMGE+H cells. A Survey micrograph showing that the ells are flatly spread on the substrate (9. The cells are clearly polarized but do not contain an extended apical intercellular junction complex. Contacts occur mainly through microvillus-like projections (arrows) and desmosome-containing bridges (arrowheads). B Enlargement of part of the cell shown in A. The desmosome is identified (arrowhead) as well as the attached tonofilaments. C Desmosomes in a broader intercellular contact containing several (three) desmosomes. D Tonofilaments are seen to be laterally associated with the desmosomal plaque. E In the vicinity of desmosomes (arrowhead) gap junctions (GJ) can sometimes be detected. Bars indicate 1 pm (A, B) and 0.2 pm (C-E)

types of junction by electron microscopy. For example, as a result of this uncertainty of definition various forms of junction that resemble desmosomes, but are not identical to typical epidermal desmosomes, have been described as ‘ desmosome-like specializations’, puncta adhaerentia (cf. [ l l , 45]), ‘desmosome-like junctions’ of glioblasts [38] and glial cells of annelids [ i l l , ‘desmosome-like structures’ or ‘attachments ’ between synovial cells [29] and between Sertoli cells and germ cells [ 1 , 42, 52, 531, ‘ simplified desmo- somes’ as encountered in lactating mammary gland cells of rodents [37, 471, ‘maculae adhaerentes diminutae‘, or ‘modified desmosome’ structures in certain cultured epithe- lial cells [62]. ‘Desmosome-like structures’ have also been described by electron microscopists in a number of tumors devoid of typical desmosomes such as sarcomas, meningio- mas, neuroblastomas, and lymphomas [8, 291. This list of names illustrates the problem of identifying junctions by their electron microscopic morphology alone and points to the need for independent, molecular criteria for defini-

tion of junctions. We propose to classify the two types of junction according to constitutive proteins specifically asso- ciated with their cytoplasmic plaques. In this sense we use the term desmosome-type junctions for plasma membrane domains that are associated with characteristic desmosomal plaque proteins, such as ‘Dl-protein’ and desmoplakins [19, 20, 411, and are devoid of vinculin. In addition to typical desmosomes of epithelial cells, this includes the maculae adhaerentes of intercalated disks of cardiac muscle and Purkinje fiber cells [9, 18, 19,411 as well as hemidesmo- somes. On the other hand, we refer as ‘adhaerens type junc- tions ’ to those membrane domains that are associated with vinculin and often also with a-actinin. Among these are the zonula and fascia adhaerens of polarized epithelia and cardiac muscle, peripheral dense plaques of smooth muscle cells and cell-substrate focal contacts [3, 21, 23, 26-28, 581. It is to be expected that future studies will reveal additional desmosome-specific and adhaerens junction-specific pro- teins as well as constituents which allow the correlation

200

Fig. IOA-D. Electron microscopy of the ventral cell membrane of BMGE+H cells showing focal contacts (indicated by the arrows) at low (A) and high (B) magnification. Note the bundles of microfilaments which terminate near the focal contact. C Hemidesmosome at the cell substrate contact (indicated by the two urrowheadr). D Interference-reflection light microscopy of BMGE + H cells showing ‘dark’ interference reflection images typical for focal contacts (mrows) as well as gray areas typical of close contacts. S, substrate. Burs in A-C indicate 0.2 pm and that in D 10 pm

of morphologic diversity within each of these two classes of junction with differences in molecular composition.

In this study we have examined the spatial distributions of the molecular constituents of two types of intercellular junction, focal (or adhaerens) junctions, and desmosomes. These two types of contact formation are unique in the sense that they are usually associated, at their cytoplasmic faces, with defined cytosekeltal networks. In focal junctions such as zonula adhaerens, dense plaques, fascia adhaerens, and cell-to-substrate focal contacts, the prominently asso- ciated cytoskeletal filaments are the microfilaments. In des- mosomes or hemidesmosomes the associated fibrils are in- termediate-sized filaments of the cytokeratin type [12, 571. Despite the remarkable structural and biochemical dissimi- larities, these two classes of junction have some characteris- tics in common as, for example, the existence of membrane- bould cytoplasmic plaque structures which probably mediate the attachment of the filaments to the membrane. Recently, it has been shown that vinculin, a protein of mo- lecular wieght 130,000, can serve as a marker for focal junc- tions [4, 24, 27, 28, 581 and that antibodies recognizing the DP1-antigen, a group of large polypeptides of sizes from 164,000 to 240,000 isolated from purified epidermal desmo- somes [7, 181, can be used to localize, by light and electron microscopic immunocytochemistry [ 181, the intracellular plaques or desmosomes.

Our results indicate that the defined molecular constitu- ents of the two junctional systems and the respective cytos-

keletal filaments are independently segregated in tissues and in cultured epithelial cells. This has been manifested at sev- eral levels. In tissues such as intestine, labeling of vinculin as well as of actin and a-actinin can be detected in essen- tially every cell and is most prominent in the muscularis mucosae. On the other hand, the combination of DP1- protein and cytokeratins is found only in the epithelial cells. The compositional difference and independence of the two junctional systems can be further shown at subcellular lev- els. In the intestinal epithelium both junctional systems are found in the terminal bar region. Their locations could not be resolved in detail in frozen sections, but in dissociated epithelial cells we have often found that the apical dotted labeling for DPl-protein and desmoplakins (not shown here) is located below the vinculin- and a-actinin-positive belt. These results indicate that the different junctional plaque proteins sharply segregate into the specific ultra- structurally defined intercellular contacts [lo, 341. This, as well as our immunofluorescent microscopic finding that the DPl -protein-positive sites on basolateral membranes of iso- lated cells do not react with vinculin and a-actinin anti- bodies provides evidence that the molecular composition of the two junctions is different and mutually exclusive and that they represent different laterally-segregated biochemi- cal domains. This is in line with previous immunoelectron microscopic studies which have demonstrated that a-actinin and vinculin are associated with the zonula adhaerens of intestinal epithelium and fasciae adhaerentes of cardiac

201

Fig. 11 A-D. Double immunofluorescence labeling of BMGE +H cells for actin (A) and vinculin (B), or for cytokeratin (C) and DP1-protein (D). The labeling for actin and cytokeratin was performed with rabbit antibodies and rhodamine-labeled goat anti rabbit-IgG. For the labeling of the respective cells for vinculin and DP1-protein guinea pig antibodies were used, followed by fluorescein labeled goat anti-guinea pig IgG. Notice the association of vinculin with the termini of actin stress fibers (arrows). The bur indicates 10 pm

muscle, but show no significant association with adjacent desmosomes [28, 581.

Cultured epithelial cells have also revealed this differen- tial composition and distribution. In MDBK cells, vinculin as well as a-actinin and actin are prominently associated with the subapical zonulu udhaerens (for possible enrich- ments of certain membrane-bound proteins in this region see also Ref. 39) and with distinct ventral contact sites, whereas DP1-antigen is not detected in focal adhesion sites and is located on lateral walls in dispersed arrays in posi- tions basal to the zonula adhaerens. The independent and mutually exclusive distribution of major protein constitu- ents of adhaerens junctions and desmosomes is especially clear in the nonpolar BMGE + H cells which form intercel- lular contact bridges containing desmosomes but are devoid of detectable amounts of vinculin and a-actinin. Vice versa, cell-substrate focal contacts of BMGE + H cells as identified by interference reflection contrast microscopy and immuno-

labeling for vinculin are not labeled by antibodies to DP1- proteins. In addition to the surface membrane-bound des- mosomes, DPl-proteins can also be seen in small intracellu- lar spots which could be identified by immunoelectron mi- croscopy either as hemidesmosomes or as intracellular vesicle-bound plaque fragments ([MI; cf. [35]).

The two types of membrane-bound junctional protein are not only mutually exclusive but also show strict specific- ity with respect to the attached cytoskeletal filaments. In intestinal cells, cytokeratin filaments are apically anchored to the ring of DPI-antigen positive desmosomes (‘subapical skeletal disk’) [I41 and from that area extend basally throughout the cytoplasm. By contrast, the density of actin microfilaments is prominent in the terminal bar and apically thereof, including the microvilli.

The cultured cells provide an opportunity to analyze in greater resolution not only the spatial relationships be- tween the components specific for either type of junction

202

Fig. 12A-H. Double immunofluorescence labeling of BMGE+H cells for the following antigens: A and B Labeling for cytokeratins (A, rabbit antibodies) and vinculin (B, guinea pig antibodies); C and D, labeling for DPI-protein (C) and vinculin (D, using the four stage procedure, vinculin labeled first). The arrows point to the same areas. E and F Labeling for actin (E, rabbit antibodies) and DP1 -protein (F, guinea pig antibodies). The exclusion of DPI protein-staining from intercellular actincontaining contacts is indicated by the matching arrows. G and H Double labeling for DPI-protein (G) and actin (H) as in E and F. Desmosomes are observed along the cell membranes, close to but not coinciding with actin bundles (arrows). The bar indicates 10 prn

but also to determine the dynamics of their formation. MDBK cells display enrichment of vinculin-containing ad- haerens junctions in two distinct cellular domains. Near the ventral surface vinculin is associated with cell-substrate focal contacts. These are formed relatively soon after plat- ing (1-2 h) and appear in both single cells and continuous sheets of cells in densely plated cultures. Double labeling with actin antibodies or with fluoresceinated phalloidin (not

shown) has indicated that actin bundles terminate in these sites as described for other cells [21, 23, 31, 611. The concen- tration of actin and vinculin and their organization in the apical domain is detected in cells forming contacts with their neighbors, and it is only in these contact regions where actin, a-actinin, and vinculin apparently interact to form a continuous membrane-associated belt. The level of resolu- tion of immunofluorescence microscopy is not sufficient

203

to determine whether the association of the actin filaments in this belt with the membrane is ‘end on’ or sidewise but electron microscopy indicates that most of the actin micro- filaments in that region are parallel to the membrane and appear to be laterally attached to the plaques of the zonula adhaerens. Whether, in addition, actin filaments insert ‘end on’ as suggested by the presence of vinculin at the actin- membrane contact sites, in analogy to other vinculin-con- taining focal junctions [21, 23, 27, 28, 581, remains to be examined in detail.

The junctional chain of DP1 protein-positive desmo- somes in MDBK cells is located in the vicinity of the vincu- lin-rich belt of the zonula adhaerens but is clearly a distinct structural and biochemical entity. Unlike the continuous vinculin belt, labeling of DP1-antigen (this study) and des- moplakins [19, 20, 411 appears in the form of a chain of distinct spots which occasionally can be resolved, in some- what tilted orientations, from the belt of the zonula adhaer- ens. The labeling for cytokeratin fibrils in these cells is intense in the central cytoplasm and only thin bundles of filaments extend towards the desmosomes, apparently ter- minating at the desmosomal plaques.

The independence of the two cytoskeletal systems and their junctional anchorage points is even more prominent in BMGE+H cells. Cytokeratin filament bundles run into characteristic intercellular bridges and seem to terminate at contact points identified as desmosomes by their intense labeling with DP1 -protein antibodies. The microfilament system in BMGE+H cells form tightly packed bundles which terminate at the ventral foci of substrate adhesion plaques. The association of actin with these intercellular bridges is very limited and apparently not intercalated into the cytokeratin bundles. Within the lateral contact regions actin is detected mostly in those microvilli-like projections which do not contain desmosomes and cytokeratin bundles, and these udhaerens-type contacts are apparently very rare in BMGE + H cells. The different distribution of junctional elements in the two cultured epithelial cells, i.e., MDBK cells on the one hand and BMGE+H cells on the other, and its effect on the properties of the specific junctions may be related to differences of the functional properties in these cells in the respective tissues. The MDBK cells exhibit an intercellular contact-dependent polarity similar to that of intact intestinal and renal epithelium [6, 511. It has been shown that specific surface membrane constitu- ents of these cells segregate into either the apical or basolat- eral membrane domains [6, 39, 48, 511. Both desmosomal and adhaerens junctions are found primarily in the basolat- eral portion and are enriched along the border line between the two membrane domains. Whether these junctional con- stituents are also involved in the sorting of membrane pro- teins in polarized epithelia remains to be examined (for dissociation experiments see, e.g., Refs. 46, 63).

The present study has not been designed to analyze the causal and temporal relationships of formation of the dif- ferent junctional elements. Nevertheless, some of our obser- vations seem to illuminate certain aspects of junction bio- genesis. In cultured MDBK and BMGE+H cells, the for- mation of the ventral focal junctions is the most rapid process. Well developed cell-substrate focal contacts, deter- mined by interference reflection microscopy or by immuno- labeling for vinculin, can be detected as soon as 1-2 h after plating of the cells and usually reach the ‘mature’ extensive pattern after 5-6 h. The rate of assembly of the junctional

vinculin-actin belts at te apex of MDBK cells depends, of course, on the plating density. We have found, however, that even in densely plated MDBK cell cultures the forma- tion of the apical zonula adhaerens and its association with vinculin and actin is significantly slower than the appear- ance of vinculin- and actin-positive focal contacts. Roughly, the formation of a continuous adhaerens junction-asso- ciated microfilament belt in these cells has only been ob- served after approximately 10 h and more. The formation of the apical chain of desmosomes seems to occur almost concomitant with the advent of the zonula adhaerens. Sparse DPl -protein-positive spots are apparent early after incuba- tion of densely plated MDBK cells but the fully developed dense circular array of apical desmosomes does not seem to develop until about 8-12 h after plating of the culture. Additional detailed studies will be necessary to establish the temporal sequence of organization of the different junc- tional elements and their assembly into the structurally and functionally ‘mature’ intercellular junctional complexes and their association with the specific cytoskeletal filament structures.

Acknowledgments. B.G. is an incumbent of the Ch. Revson chair in biology. This study has been financially supported in part by the Deutsche Forschungsgemeinschaft and a special cooperative grant of the German Cancer Research Center.

References 1. Altdorfer J, Fukuda T, Hedinger C (1974) Desmosomes in

human seminiferous epithelium. Virchows Arch [Cell Pathol]

2. Avnur A, Geiger B (1981) Substrate attached membranes of cultured cells: Isolation and characterization of ventral cell membranes and the associated cytoskeleton. J Mol Biol

3. Bloch RJ, Geiger B (1980) The localization of acetylcholine receptor clusters in areas of cell-substrate contact in cultures of rat myotubes. Cell 21 :35-42

4. Burridge K, Feramisco J (1980) Microinjection and localization of a 130 K protein in living fibroblasts; a relationship to actin and fibronectin. Cell 19: 587-595

5. Campbell RD, Campbell JH (1971) Origin and continuity of desmosomes. In: Reinert J, Ursprung H (eds) Origin and con- tinuity of cell organelles, Vol2. Springer-Verlag, Berlin, Heidel- berg, New York p 261-298

6. Cereijido M, Robbins ES, Dolan WJ, Rotunno CA, Sabatini DD (1978) Polarized monolayers formed by epithelial cells in a permeable and translucent support. J Cell Biol77 : 853-880

7. Cohen SM, Gorbsky GJ, Steinberg MS (1980) Monoclonal an- tibodies to the intercellular glycoprotein components of desmo- somes. J Cell Biol87:88a

8. Erlandson RA (1981) Diagnostic transmission electron micros- copy of human tumors. Masson Publishing USA Inc, New York, p 107-116, p 169-186

9. Eriksson A, Thornell L-E (1979) Intermediate (skeletin) fila- ments in heart Purkinje fibers. J Cell Biol 80:231-247

10. Farquhar MG, Palade GE (1963) Junctional complexes in var- ious epithelia. J Cell Biol 17 : 375-41 2

11. Fawcett DW (1981) The cell, 2nd edn. Saunders Company, Philadelphia, p 1-862

12. Franke WW, Weber K, Osborn M, Schmid E, Freudenstein C (1978) Antibody to prekeratin. Decoration of tonofilament- like arrays in various cells of epithelial character. Exptl Cell Res 116:429-445

13. Franke WW, Grund C, Osborn M, Weber K (1978) The inter- mediate-sized filaments in rat kangaroo PtK, cells. I. Morphol- ogy in situ. Cytobiologie 17:365-391

16: 181-194

152: 361-379

204

14. Franke WW, Appelhans B, Schmid E, Freudcnstein C, Osborn M, Weber K (1979) The organization of cytokeratin filaments in the intestinal epithelium. Eur J Cell Biol 19:255-268

15. Franke WW, Schmid E, Winter S, Osborn M, Weber K (1979) Widespread occurrence of intermediate-sized filaments of the vimentin-type in cultured cells from diverse vertebrates. Exptl Cell Res 123 : 25-46

16. Franke WW, Schmid E, Freudenstein C. Appelhans B, Osborn M, Weber K, Keenan TW (1980) Intermediate-sized filaments of the prekeratin type in myoepithelial cells. J Cell Biol 84:633-654

17. Franke WW, Denk H, Kalt R, Schmid E (1981) Biochemical and immunological identification of cytokeratin proteins pres- ent in hepatocytes of mammalian liver tissue. Exptl Cell Res

18. Franke WW, Schmid E, Grund C, Miiller H, Engelbrecht I, Moll R, Stadler J, Jarasch E-D (1981) Antibodies to high mo- lecular weight polypeptides of desmosomes : specific localiza- tion of a class of junctional proteins in cells and tissues. Differ- entiation 20: 21 7-241

19. Franke WW, Moll R, Schiller DL, Schmid E, Kartenbeck J, Mueller H (1982) Desmoplakins of epithelial and myocardial desmosomes are immunologically and biochemically related. Differentiation 23 : 1 15-1 27

20. Franke WW, Moll R, Miiller H, Schmid E, Kuhn C, Krepler R, Artlieb U, Denk H (1983) Immunocytochemical identifica- tion of epithelium-derived human tumors using antibodies to desmosomal plaque proteins. Proc Natl Acad Sci USA 80: 543-547

21. Geiger B (1979) A 130 K protein from chicken gizzard: its localization at the termini of microfilamcnt bundles in cultured chicken cells. Cell 18: 193-205

22. Geiger B (1981) The association of rhodamine-labelled a-actinin with actin bundles in demembranated cells. Cell Biol Intern Rept 5 : 627-634

23. Geiger B (1982) Involvement of vinculin in contact-induced cytoskeletal interactions. Cold Spring Harbor Symp Quant Biol 46:671-682

24. Geiger B, Singer SJ (1979) The participation of a-actinin in the capping of cell-membrane components. Cell 16: 21 3-222

25. Geiger B, Singer SJ (1980) Association of microtubules and intermediate filaments in chicken giz7ard cells as detected by double immunofluorescence. Proc Natl Acad Sci USA 77 :4769-4773

26. Geiger B, Tokuyasu KT, Singer SJ (1 979) The immunochemical localization of a-actinin in intestinal epithelial cells. Proc Natl Acad Sci USA 76:2833-2837

27. Geiger B, Tokuyasu KT, Dutton AH, Singer SJ (1980) Vincu- lin, an intracellular protein localized at specialized sites where microfilament bundles terminate at cell membrane. Proc Natl Acad Sci USA 77:41274131

28. Geiger B, Dutton AH, Tokuyasu KT, Singer SJ (1981) Immun- oelectron microscopic studies of membrane-microfilament in- teraction. The distributions of a-actinin, tropomyosin and vin- culin in intestinal epithelial brush border and in chicken gizzard smooth buscle. J Cell Biol 91 :61&628

29. Ghadially FN (1982) Ultrastructural pathology of the cell matrix, 2nd edn. Butterworths, London, p 1-947

30. Gorbsky G. Steinberg M (1981) Isolation of the intercellular glycoproteins of desmosomes. J Cell Biol90: 243-248

31. Heath JP, Dunn GA (1978) Cell to substratum contacts of chick fibroblasts and their relation to the microfilament system. A correlated interference - reflexion and high-voltage electron- microscope study. J Cell Sci 29: 197-212

32. Hertzberg EL, Gilula NB (1979) Isolation and characterization of gap junctions from rat liver. J Biol Chem 254:2138-2147

33. Horwitz B, Kupfer H, Eshhar Z, Geiger B (1981) Reorganiza- tion of arrays of prekeratin filaments during mitosis. Exptl Cell Res 134:281-290

34. Hull BE, Staehelin LA (1979) The terminal web. A reevaluation of its structure and function. J Cell Biol81: 67-82

131 1299-318

35. Kartenbeck J, Schmid E, Franke WW, Geiger B (1982) Differ- ent modes of internalization of proteins associted with adhaer- ens junctions and desmosomes: experimental separation of lat- eral contacts induces endocytosis of desmosomal plaque materi- al. EMBO J 1 : 725-732

36. Kelly DE (1965) Fine structure of desmosomcs, hemidesmo- somes, and an adepidermal globular layer in developing new epidermis. J Cell Biol28 : 51-72

37. Krepler R, Denk H, Weirich E, Schmid E, Franke WW (1981) Keratin-like proteins in normal and neoplastic cells of human and rat mammary gland as revealed by immunofluorescence microscopy. Differentiation 20: 242-252

38. Lim R, Troy SS, Turriff DE (1977) Fine structure of cultured glioblasts before and after stimulation by a glia maturation factor. Exptl Cell Res 106:357-372

39. Louvard D (1980) Apical membrane aminopeptidase appears a t site of cellcell contact in cultured kidney epithelial cells. Proc Natl Acad Sci USA 77:41324136

40. Madin SH, Darby NB (1978) Established kidney cell line of normal adult bovinc and ovine origin. Proc Soc Exptl Biol Med 98 : 574-576

41. Mueller H, Franke WW (to be publsihed) Biochemical and immunological characterization of desmoplakins I and 11, the major polypeptides of the desmosomal plaque. J Mol Biol

42. Nagano T, Suzuki F (1978) Cell to cell relationships in the seminiferous epithelium in the mouse embryo. Cell Tissue Res 189: 389-401

43. Overton J (1968) The fate of desmosomes in trypsinized tissue. J Exptl Zoo1 168 : 203-21 4

44. Overton J, DeSalle R (1980) Control of desmosome formation in aggregating embryonic chick cells. Develop Biol75: 168-176

45. Palay SL (1967) Principles of cellular organization in the nervous system. In: Quarton GC, Melnechuk T, Schmitt FO ( 4 s ) The neurosciences, a study program. Rockefeller Univer- sity Press, New York

46. Pisam M, Ripoche P (1976) Redistribution of surface macro- molecules in dissociated epithelial cells. J Cell Biol 71 :907-920

47. Pitelka DR (1979) Cell contacts in the mammary gland. In: Larson BL (ed) Lactation, a comprehensive treatise. Academic Press, London New York, p 41-66

48. Reggio H, Coudrier E, Louvard D (1982) Surface and cytoplas- mic domains in polarized epithelial cells. In: Hoffman JF, Gieb- isch GH, Bolis L (eds) Membranes in growth and development. Alan R Liss, New York, p 89-105

49. Renner W, Franke WW, Schmid E, Geisler N, Weber K, Man- delkow E (1981) Reconstitution of intermediate-sized filaments from denatured monomeric vimentin. J Mol Biol 149:285-306

50. Revcl JP, Nicholson BJ, Yancey SB (1982) Partial sequence and turnover of rat liver gap junction protein. Cold Spring Harbor Symp Quant Biol46:633-637

51. Rindler MF, Chuman LM, Shaffer L, Saier MH (1979) Reten- tion of differentiated properties in an epithelial cell line (MDCK). J Cell Biol81: 635-648

52. Romrell LJ, Ross MH (1979) Characterization of sertoli cell- germ cell junctional specializations in dissociated testicular cells. Anat Rec 193 : 23-42

53. Russell L (1977) Desmosome-like junctions between Sertoli and germ cells in the rat testis. Anat Rec 148:301-312

54. Schmid E, Schiller DL, Grund C, Stadler J, Franke W (1983) Tissue type-specific expression of intermediate filament pro- teins in a cultured epithelial cell line from bovine mammary gland. J Cell Biol 96:37-50

55. Schollmeyer JE, Furcht LT, Goll DE, Robson RM, Stromer MH (1976) Localization of contractile proteins in smooth mus- cle cells and in normal and transformed fibroblasts. In: Goldman R, Pollard T, Rosenbaum J (eds) Cell motility. Cold Spring Harbor Laborator, New York, p 361-388

56. Staehelin LA (1974) Structure and function of intercellular junctions. Intern Rev Cytol 39: 191-283

57. Sun T-T, Green H (1978) Immunofluorescence staining of keratin fibers in cultured cells. Cell 14:469-476

205

58. Tokuyasu KT, Dutton AH, Geiger B, Singer SJ (1981) Ultra- structure of chicken cardiac muscle as studied by double im- munolabeling in electron microscopy. Proc Natl Acad Sci USA 78 : 761S7623

59. Trinkaus JP (1976) On the mechanism of metazoan cell move- ment. In: Poste G, Nicolson GL (eds) The cell surface in animal embryogenesis and development. Elseviermorth-Holland), Amsterdam, p 225-329

60. Vasiliev JM, Gelfand IM (1977) Mechanisms of morphogenesis in cell cultures. Intern Rev Cytol50: 159-273

61. Wehland J, Osborn M, Weber K (1979) Cell-to-substratum contacts in living cells. A direct correlation between interfer-

ence-reflexion and indirect immunofluorescence microscopy using antibodies against actin and a-actinin. J Cell Sci

62. Zerban H, Franke WW (1978) Modified desmosomes in cul- tured epithelial cells. Cytobiologie 18: 360-373

63. Ziomek CA, Schulman S, Edidin M (1980) Redistribution of membrane proteins in isolated mouse intestinal epithelial cells. J Cell Biol86: 849-857

37: 257-273

Received August 1982 / Accepted in revised form December 1982