The cytoskeleton and substratum adhesion in chick embryonic corneal epithelial cells

The majority of human tumours are carcinomas of epithelial origin, many of which are invasive and metastatic because of altered adhesive and motile properties. It is therefore of the utmost importance to study these properties in normal epithelial cells in order to develop a basis on which to understand the pathological behaviour of carcinoma cells. With these broad objectives in view, we have begun a detailed study of the adhesion and motility of chick embryonic corneal epithelial cells.

Epithelial cells have two sets of adhesions: those with the substratum or extracellular matrix, a basement membrane in vivo, and their mutual cell-cell adhesions. In the case of the multilayered chick embryo corneal epithelium, all the cells have adhesions with others but only the cells of the basal layer have adhesions with the basement membrane (Hay & Revel, 1969).

In this paper we are mainly concerned with cell-substratum adhesions. Hay & Revel (1969) have suggested that hemidesmosomes are involved in adhesion to the basement membrane in vivo. Hemidesmosomes are so called because their structure as seen by transmission electron microscopy resembles that of half an intercellular desmosome. As far as we are aware, hemidesmosomes have not been studied in vitro. On the other hand, a recent paper by Heath (1982), studying the appearance of corneal epithelial cells on glass substrata by interference reflection microscopy (IRM), reported the presence of focal contacts beneath the peripheral cells of explants. Focal contacts have been studied principally in cultured fibroblasts (Izzard & Lochner, 1976; Heath & Dunn, 1978). They appear as discrete dark areas by IRM and are believed to represent adhesions where the separation between the lower cell membrane and the substratum is 10 – 15 nm (Izzard & Lochner, 1976). In addition, we have shown in a previous study (Nicol & Garrod, 1982) that 15-day embryonic chick corneal epithelial cells in monolayer stain with antibody to the adhesive glycoprotein, fibronectin. As with focal contacts, fibronectin has been mostly studied in fibroblasts, where it has been implicated particularly in adhesion to the substratum (Hynes, 1976; Yamada, Olden & Pastan, 1978; Pear1stein & Gold, 1978; Grinnell, 1978). There have been some reports of association between fibronectin and epithelial cells (e.g., Chen, Maitland, Gallimore & McDougall, 1977). However, recent publications have suggested that fibronectin does not stimulate the attachment of some epithelial cells (see Kleinman, Klebe & Martin, 1981), but instead have implicated the basement membrane glycoprotein, laminin, in epithelial and carcinomal cell adhesion (Terranova, Rohrbach & Martin, 1980; Vladovsky & Gospoderowicz, 1981).

The situation is thus somewhat confused. Do corneal epithelial cells possess two mechanisms of cell-substratum adhesion, one involving hemidesmosomes and the other fibroblast-like focal contacts? Are these alternative mechanisms, one operating in vivo and the other in vitro? What part is played by fibronectin in cell-substratum adhesion ?

With such questions in mind, we have conducted an investigation of the adhesion of 15-day chick embryo corneal epithelial cells to a gelatin substratum. The techniques employed were electron microscopy, IRM and fluorescent antibody staining. We also examined the distribution of the cytoskeletal elements with which adhesions are believed to interact on the cytoplasmic side of the membrane. These are tono-filaments or prekeratin filaments in the case of desmosomes (Lazarides, 1980; Henderson & Weber, 1981) and actin microfilaments in the case of focal contacts (Heath & Dunn, 1978). There has been no previous thorough investigation of the relationship between cell-substratum adhesions and the cytoskeleton in epithelial cells.

Cells were isolated from 15-day chick embryos and cultured as described previously (Nicol & Garrod, 1979) except that a thin covering of gelatin solution (Sigma type I; 100 μ g/ml in water) was allowed to set on either glass coverslips or plastic tissue-culture dishes (Nunclon, 50 mm or Sterilin, 35 mm) before plating out the cells. This was found to improve cell attachment and spreading.

Clean chicken gizzards were minced and extracted overnight in 50% glycerol containing 0·067% phosphate-buffered saline (PBS) and 0·1 mM-phenylmethylsulphonyl fluoride (PMSF) at 4 °C. They were then extracted with two 15-min washes with 0·1 M-KCl to polymerize actomyosin in the tissue. Divalent cationsand tropomyosin were removed by washes with 0·05 M-NaHCO₃ containing 0 ·1 mM-PMSF, followed by 0·05 M-NaHCO₃ containing 0 ·1 mM-PMSF and 10 mM-EDTA. The material was then washed several times with distilled water followed by acetone at 0 °C, and was then blended with acetone before being dried in air overnight. Actin was extracted from the resulting acetone powder by the method of Mead (1980). The actin preparation was then solubilized by boiling for 3 min in 1% sodium dodecyl sulphate (SDS) and 1% β-mercaptoethanol, and electrophoresed on a preparative 15% Laemmli polyacrylamide gel. The 46000 molecular weight band was cut from the gel, eluted (Lazarides & Weber, 1974) and used to immunize rabbits that had been previously screened to demonstrate their freedom from anti-actin autoantibodies. Actin at about 200 μ g/ml in PBS was emulsified with an equal volume of Freund’s complete adjuvant. Each rabbit received 1 ml of the emulsion distributed between two intramuscular sites on the thighs and two subcutaneous sites in the back. Animals were given booster injections after 4 weeks and 8 weeks, and bled 2 weeks after the final injection.

Prekeratin was extracted from bovine snout epidermis using citric acid/sodium citrate (CASC) buffer (pH 2·6) and purified further by serial isoelectric precipitations pH 4·5–7·0, according to the method of Matoltsy (1965). Antibodies to prekeratin were raised in guinea pigs as rabbits often possess autoantibodies to intermediate filaments (Osborn, Franke & Weber, 1977). Prekeratin at 1 mg/ml in PBS was emulsified with an equal volume of Freund’s complete adjuvant. Each guinea pig was injected with 1 ml of the emulsion distributed between two intramuscular sites in the thighs and two subcutaneous sites on the back. Booster injections were given after 21 days and 3 months and the animals bled out 8 days after the final injection.

Fibronectin was prepared from citrated chicken plasma by affinity chromatography on a gelatin-Sepharose 4B column according to Engvall & Ruoslahti (1977). Antibodies were raised in rabbits using a similar injection schedule to that used for actin. Affinity-purified anti-fibronectin was obtained according to the method of Chiquet, Puri & Turner, (1979).

Anti-tubulin serum was a gift from Dr G. M. Mead. The tubulin for its preparation was obtained from calf brain as described by Mead, Cowin & Whitehouse (1979).

Rabbit anti-mouse embryo laminin was a kind gift from Dr B. Hogan of the Imperial Cancer Research Fund.

Cells were washed three times in phosphate-buffered saline (PBS) at room temperature, fixed in 3·5% paraformaldehyde in PBS for 1 h and then washed in 0·1 M-NH₄Cl in PBS for 30 min. They were then washed in PBS alone prior to staining. In order to stain intracellular antigens, the cells were made permeable in acetone (2 min at − 20 °C in 50% acetone; 5 min at 4 °C in 100% acetone; 2 min at − 20 °C in 50% acetone) and then washed again in PBS.

Cells were stained for 30 min with a suitable dilution of antiserum. Then, following three washes in PBS, the bound immunoglobulins were stained for 30 min with commercial fluo-rescein-conjugated antibodies. To detect the anti-fibronectin, anti-laminin, anti-actin and antitubulin, fluorescein (FITC)-conjugated sheep anti-rabbit immunoglobulin (Wellcome) was used. To detect anti-prekeratin antibodies, FITC-conjugated rabbit anti-guinea pig immunoglobulin (Wellcome) was used. Both were diluted 1:20. Following extensive washing in PBS, the coverslips were mounted in glycerol :PBS (9:1) and sealed with nail varnish.

Cells were also double-labelled for fibronectin and prekeratin, actin and prekeratin, and tubulin and prekeratin. For this, the first antibody was stained with rhodamine (TRITC)-conjugated goat anti-rabbit IgC immunoglobulin (Wellcome), while the second antibody (prekeratin) was localized as before. No cross-reactivity was found between the rabbit antiguinea pig immunoglobulin and the rabbit antibodies.

Fluorescence was observed on a Zeiss Photomicroscope III as previously described (Nicol & Garrod, 1982). Photographs were taken on HP5 or FP4 black and white film.

Monolayers for electron microscopy were washed three times with PBS, fixed in situ in 2 · 5% glutaraldehyde in PBS for 30 min and post-fixed in 2% osmium tetroxide in PBS for 20 min. Washing in distilled water was followed by staining for 20 min in 1% uranyl acetate. Dehydration through an ethanol series was followed by removal of the cell monolayers from the culture dishes using propylene oxide. Monolayers were embedded in either Spurr’s resin or Araldite. Ultrathin sections were cut using glass knives on an LKBIII Ultratome. Sections were stained on the grids with lead citrate before being observed and photographed using either a Phillips 300 or 201 transmission electron microscope, the former being fitted with a goniometer stage.

This was carried out using a Zeiss Photomicroscope III with a HB50 mercury vapour light source, a green band-pass filter (546 nm) and an × 63 Antiflex objective.

Cryostat sections (6 μ m) of intact 15-day corneas were stained with antibodies directed against fibronectin, laminin, prekeratin, actin and tubulin. The results obtained on viewing stained sections by fluorescence microscopy are shown in Figs, 1—6. The phase-contrast picture in Fig. 1 shows that the outer stratified corneal epithelium is separated from the collagenous stroma by a basement membrane. Antifibronectin stained the stroma strongly, the basement membrane weakly but not the epithelium (Fig. 2). In contrast, anti-laminin gave very bright staining of the basement membrane, weak staining of the epithelium and minimal staining of the stroma (Fig. 3). Prekeratin was located exclusively in the epithelium (Fig. 4), whereas actin and tubulin staining were apparent in both epithelium and stroma (Figs. 5, 6). Anti-actin staining was more intensive in the outer cell layer (periderm) of the epithelium than elsewhere.

Fig. 7 shows a region of contact between two corneal epithelial cells including three desmosomes and a gap junction. The desmosomes have dense submembranous plaques, associated tonofilaments and intermembrane densities with midlines. Fig. 8 shows parts of two corneal epithelial cells in contact with the basement membrane. The lower cell surface shows submembranous densities that appear to be associated with tonofilaments in the cytoplasm and extracellularly with dense material that traverses the basement membrane. Fig. 9 shows the basal region of a cell and the basement membrane at slightly higher magnification. One of the submembranous structures shows a layered composition similar to that of the classical hemidesmosome. It has a dense plaque closely apposed to the plasma membrane. Immediately internal to the plaque is a narrow region of lower electron density followed by a region of tono-filaments. Again there is some suggestion of a periodicity in the basement membrane. Wherever and whenever they are encountered in sections desmosomes and hemi-desmosomes are about 0 · 1 μ m in width. This suggests that in surface view they may be roughly circular and approximately 0 · 1 μ m in diameter.

The general appearance of a small monolayered island of corneal epithelial cells cultured on gelatin and viewed by Nomarski optics is shown in Fig 10. The individual cells can be identified by means of their nuclei. Cells at the periphery have wellspread lamellae extending outwards over the substratum. Fine processes can be seen both at the free edge and between cell boundaries.

Fig. 11 shows the periphery of an island viewed by IRM. Small dark areas can be seen associated with the lamellar regions of peripheral cells. These areas are mooted to be regions of closest approach between the lower membrane of the cell and the substratum. They have been called focal contacts by Izzard & Lochner (1976) and are believed to correspond to adhesive areas. The focal contacts of corneal epithelial cells tend to be elongated structures with their long axes roughly perpendicular to the peripheral margins of the leading lamellae. They are of the order of one to several micrometres in length. No such areas occur on cells remote from the periphery, although some more regularly shaped dark areas were encountered. IRM images obtained with fixed cells on gelatin substrata were indistinguishable from those shown by cells on glass and by living cells.

Staining with anti-fibronectin was concentrated beneath the peripheral cells of monlayered islands (Fig. 12). Characteristically it consisted of fibrils having a general orientation at right angles to the peripheral margin. There were also some peripherally located, non-fibrillar accumulations of fibronectin. A few fibrils were present under centrally located cells, but there was an absence of staining on the dorsal surface and between cells. Fibrillar patterns were also detectable on the substratum beyond the margins of the islands, as though cells had withdrawn adhesions from these areas. Areas of gelatin substratum that had apparently not been in contact with cells showed particulate or punctate staining for fibronectin. Sometimes the substratum immediately surrounding cellular islands was devoid of such staining.

The IRM image presented by cells stained with anti-fibronectin was somewhat confused because, after staining, the fibronectin fibrils often appeared black (Fig. 13). These fibrils tended to obscure focal contacts, making it difficult to decide whether there was any correspondence between the two.

The fibrillar patterns of fibronectin staining were established after cells had been in culture for 24 h and were maintained at least up to 72 h. No differences in fibronectin patterns were observed between cells cultured on glass, tissue-culture plastic or gelatin.

It is important to determine whether fibronectin plays a role in the adhesion of corneal epithelial cells. The occurrence of fibrillar fibronectin patterns mainly beneath the peripheral cells of cellular islands provides circumstantial evidence that it may be involved, because these cells are those most actively involved in locomotion of the cell sheet and are also those that possess focal contacts. However, two other experiments that we have performed present a rather different picture.

Firstly, cells were cultured on gelatin in medium containing foetal calf serum that had been depleted of fibronectin by three passages through a gelatin-Sepharose 4B column. (No fibronectin was detectable in this serum by polyacrylamide gel electrophoresis or immunoelectrophoresis.) Although the attachment of cells to the substratum was reduced under these conditions, many cells were able to attach and spread. Such cells were cultured for 48 h and then fixed and stained for fibronectin. The most common situation was to find islands of cells that were morphologically normal but completely devoid of fibronectin staining (Figs. 14, 15).

Secondly, cells were cultured in medium containing complete foetal calf serum and then treated with anti-fibronectin IgG at high concentration (4 · 5 mg / ml) in medium containing fibronectin-depleted foetal calf serum. (Anti-fibronectin has been shown to cause redistribution of fibronectin matrices and cell detachment in fibroblasts (Yamada, 1978) and chick embryo limb-bud mesenchyme cells (Garrod, unpublished).) Corneal epithelial cells thus treated remained attached to the substratum for at least 24 h in the presence of IgG and appeared morphologically normal even though staining with fluorescent antibody showed that the fibrillar fibronectin pattern had been disrupted (Figs. 16, 17).

Corneal epithelial cells in vitro exhibited no specific intra-or extracellular staining with anti-laminin antibody.

Microfilament bundles or stress fibres were clearly distinguishable in the marginal cells of large islands (Fig. 18) and throughout the cells of small, well-spread islands. The stress fibres of marginal cells were predominantly orientated in one of two directions. In the leading lamellae, the filaments extended radially almost to the extreme periphery of the cells, while the majority of other filaments were orientated parallel to the edge giving the appearance of a ring around the islands.

Figs. 19 and 20 show that the peripheral ends of many of the radially orientated filaments in the leading lamellae corresponded with focal contacts as seen by IRM. However, none of the submarginal ring filaments showed such correspondence even when the ring was very close to the periphery.

Figs. 21, 22 and 23 show the edge of a large island by phase-contrast microscopy and stained for prekeratin and actin. The prekeratin filaments are arranged in a dense reticulate network, which differs from the pattern exhibited by microfilament bundles in several important respects. Firstly, the prekeratin filaments generally did not extend to the peripheries of the well-spread marginal cells. Secondly, prekeratin filaments did not show marked orientation perpendicular or parallel to the edge of the island. Thirdly, it was often the case that prekeratin filaments of one cell were aligned with those of the neighbouring cell at the intercellular boundary. Fourthly, prekeratin staining characteristically demonstrated a perinuclear ring of filaments. Fifthly, the ends of prekeratin filaments did not coincide with focal contacts.

The pattern of microtubules appeared as a reticulate network somewhat similar to prekeratin pattern but differing in a number of respects (Figs. 24, 25). Firstly, there were fewer microtubules than prekeratin filaments and they were in general more clearly defined. Secondly, some microtubules were orientated perpendicularly to the island edge and extended to the periphery of cells. Thirdly, in some cases there appeared to be a perinuclear ring of microtubules, but this was less dense than that observed with prekeratin filaments. Fourthly, there was no alignment of microtubules at intercellular boundaries. In addition there was no correspondence between microtubules and focal contacts.

As in the intact cornea, the intercellular contacts of reaggregated corneal epithelial cells in vitro were dominated by desmosomes (see also Overton, 1977; Nicol & Garrod, 1982). If anything, desmosomes appeared to be more numerous between cells that had been in culture for 48 h, than between the cells of the original intact 15-day epithelium (Fig. 26). As far as we could tell, desmosomes in vitro had a structure identical to those in vivo and were of the same approximate size, 0 · 1 μ m in diameter.

Cell contacts with the gelatin substratum appeared to be of two types. Firstly, there were structures of medium electron density and greater than 0 · 5 μ m in lateral width (Fig. 27). These structures had bundles of fine filaments (presumably microfilaments) entering them obliquely from the cytoplasm. In this they resemble the adhesion plaques of chick heart fibroblasts described by Abercrombie, Heaysman & Pegrum (1971), which are equivalent to focal contacts (Heath & Dunn, 1978).

Secondly, there were smaller, more electron-dense sub-membranous plaques that, from their size (about 0 · 1 μ m), could be hemidesmosomes (Figs. 28, 29). We have never found their structure to be as clearly defined as that of the hemidesmosomes found in vivo. Even with the application of the goniometer stage we have never succeeded in resolving them into a dense plaque, a lighter zone and a denser region where the tonofilaments insert. Sections of the cell surface adjacent to the substratum were sometimes cut obliquely through these structures (Fig. 30). They then appeared as dense discoids about 0 · 1 μ m in diameter from which numerous cytoplasmic filaments (tonofilaments) radiate.

Corneal epithelial cells from 15-day chick embryos, after dissociation with trypsin, adhere to a gelatin surface and reaggregate to form islands (see also Nicol & Garrod, 1979; Middleton, 1973; Heath, 1982). The intercellular contacts between these reaggregated cells are characterized by numerous desmosomes that resemble those between the cells in the intact cornea (Nicol & Garrod, 1982). Our main result is that the contacts between the cells and the gelatin substratum show two types of adhesive structures, focal contacts (adhesion plaques) and hemidesmosome-like densities. The nature and significance of these cell-substratum adhesive structures will now be discussed in more detail.

As has been reported recently by Heath (1982), IRM reveals focal contacts between corneal epithelial cells and their culture substratum. In shape and orientation these resemble focal contacts described previously in fibroblasts: they are elongated and often orientated with their long axes perpendicular to the peripheral edges of the cells’ leading lamellae (Izzard & Lochner, 1976; Abercrombie & Dunn, 1975; Heath & Dunn, 1978). The focal contacts of corneal epithelial cells are also about the same size as those of fibroblasts. We have found focal contacts up to 4 μ m in length, which is approximately the same as was reported for chick heart fibroblasts by Heath & Dunn (1978).

When corneal cells were stained with anti-actin antibody and their fluorescent and IRM images compared, it was found that each focal contact corresponded with the peripheral end of an actin filament bundle. In this respect corneal epithelial cells also resemble chick heart fibroblasts (Heath & Dunn, 1978). However, throughout the islands there were many actin filament bundles that did not terminate in focal contacts.

Focal contacts of corneal epithelial cells were located almost exclusively beneath the peripheral cells of islands. Studies of the behaviour of epithelial cells in vitro have suggested that the marginal cells of epithelial sheets and islands are those principally responsible for adhesion to and spreading over the substratum (Vaughan & Trinkaus, 1966; Middleton, 1973; Garrod & Steinberg, 1975; DiPasquale, 1975). Therefore, it seems that there is correspondence between cell activity and the location of focal contacts. Moreover, the peripheral cells are the only ones that possess significantly large areas of lamellar cytoplasm: in general, focal contacts appear to be located beneath leading lamellae.

The interpretation of our results relating to fibronectin and cell-substratum adhesion is not straightforward. Two pieces of evidence suggest that fibronectin is probably not involved in adhesion. Firstly, cells were able to adhere to gelatin and to spread in medium containing fibronectin-depleted serum. Moreover, once so spread they showed no evidence of being able to synthesize fibronectin, because none could be detected by immunofluorescence after 48 h of culture. Secondly, even though it produced redistribution of the fibronectin fibrils exhibited by cells in complete medium, anti-fibronectin IgG did not cause detachment of cell islands.

Fibronectin fibrils were located predominantly, though not exclusively, beneath peripheral cells of islands. It would thus be possible to argue that their spatial distribution, like that of focal contacts, coincides with that of the most active and most adhesive cells of the islands, thus providing circumstantial evidence in favour of a role for fibronectin in cell-substratum adhesion. If we are to argue that no such role exists in the case of corneal epithelial cells, how are we to account for this pattern of fibronectin distribution? We suggest that, during outward spreading over the substratum and retraction (Heath, 1982), the peripheral cells physically reorganize the fibronectin that is adsorbed or bound to the substratum into a fibrillar pattern.

Cells placed in culture after trypinization clearly have no difficulty in organizing and assembling intercellular desmosomes that appear identical to those present in vivo’, corneal epithelial cells (Overton, 1974, 1977; Nicol & Garrod, 1982); human cervical cancer cells (Dembitzer et al. 1980); embryonic heart cells (Wiseman & Strichler, 1981). The same does not appear to be true for hemidesmosomes, however. Dodson & Hay (1974) have demonstrated the synthesis of basement membrane components by corneal epithelial cells in culture and Sugrue & Hay (1981) have described the response of the basal surface of corneal epithelial cells to matrix components. In our cultures on gelatin substrata (gelatin is a form of denatured collagen) we have found structures at the cell-substratum interface that, on account of their position, frequency, density, filament association and size, seem likely to be hemidesmosomes. Their size, about 0 · 1 μ m in width, is the same as that of the desmosomes and hemidesmosomes found in vivo and clearly distinguishes them from adhesion plaques or focal contacts that are of the order of one to several μ m in width.

Our identification of these structures as hemidesmosomes, although based on a number of criteria, must be regarded as tentative because they are not as well-organized as the hemidesmosomes found in vivo; in particular, their cytoplasmic plaques are not as clearly delineated. The reasons why they are not well formed are worthy of discussion. Firstly, it may be that the substratum is inadequate to permit or induce normal hemidesmosomal assembly. Despite the cells’ ability to synthesize basement membrane components (Dodson & Hay, 1974), it may be that during their time in culture (rarely more than 48 h in our experiments) they do not succeed in producing anything resembling a normal basement membrane. It is significant here that laminin could not be detected in corneal epithelial cell cultures by fluorescent antibody staining. Secondly, the internal cells as well as the peripheral cells of corneal epithelial islands in culture show some degree of motility (D. Billig, unpublished observations). It may be, therefore, that hemidesmosomes in culture are less permanent than those in vivo.

Hemidesmosomes are associated with tonofilaments that are probably composed of prekeratin, although this remains to be demonstrated. (Henderson & Weber (1981) have clearly established that the desmosome-associated tonofilaments of HeLa cells are composed of prekeratin.) Using fluorescence microscopy we were able to distinguish prekeratin filaments clearly in our cells and noted that there was no spatial correspondence between these filaments and focal contacts visualized by IRM. The separation between the substratum and the cell membrane in the region of hemidesmosomes was not more than 20 nm measured by transmission electron microscopy and thus hemidesmosomes should appear dark by interference reflection microscopy (Izzard & Lochner, 1976). We feel, however, that these hemidesmosomes cannot be visualized by IRM because they are only about 0 · 1 μ m in diameter and thus beyond the limit of resolution of the light microscope.

The main conclusion to be drawn from our results is that 15-day chick embryo corneal epithelial cells cultured on a gelatin substratum possess two distinct cellsubstratum adhesion mechanisms. The first is a fibroblast-like focal contact-actin microfilament bundle system. The second is a hemidesmosome-like system, probably associated with prekeratin tonofilaments. The latter system predominates in vivo at the cell-basement membrane interface. It is perhaps not surprising that the cells should go some way towards reorganizing such a system in vitro, given suitable culture conditions.

In the normal intact corneal epithelium we have found no evidence by electron microscopy for structures resembling focal contacts or adhesion plaques in contact with the basement membrane. Are the focal contacts merely produced in response to the artificial conditions pe1taining in culture or do they have some significance in relation to the adhesive and motile behaviour of cells in vivoi The most obvious difference between corneal cells in vitro and those in the intact cornea is that the former possess free edges when at the periphery of an island, whereas the latter exist as a continuous sheet in which all cells have lateral contact with others. It may be significant that actin-focal contact associations occur predominantly at the periphery of islands in vitro and are therefore associated with those cells that have a free edge. We suggest that this occurrence of focal contacts is related to wound-healing and that the free edge of an island in vitro can be equated with the edge of a wound. Such a view would be consistent with some observations made on distribution of actin in epithelial wound-healing in, for example, amphibian skin (Repesh & Oberpriller, 1980): here, cells at the wound margin became elongated and flattened with long pseudopodial projections. Microfilaments were seen extending towards the plasma membranes of these processes. It has also been reported that cytoplasmic actin increases with respect to tonofilaments and desmosome sin squamous epithelial carcinoma cells when they become invasive (Kocher, Amaudruz, Schindler & Gabbiani, 1981).

Finally, we point out that this duality of cell-substratum adhesion mechanisms supports one of the principles of cell adhesion suggested by Garrod & Nicol (1981); namely, that any given cell type possesses a number of different molecular mechanisms of adhesion.

This work was supported by the Cancer Research Campaign.