Epidermal cell migration on collagen and collagen-derived peptides

The way in which epithelial cells interact with adjacent connective tissue components is of major importance in epithelial function. For example, epithelial sheets are normally underlain by a basement membrane that contains type IV collagen (Kefalides, 1973). The structural integrity of the sheet appears to be dependent upon certain basement membrane proteins that bind the basal epithelial cells to the type IV collagen (Stanley et al. 1981; Terranova, Rohrbach & Martin, 1980). In addition, collagen substrates can influence the differentiation of certain types of epithelial cells (Michalopoulos & Pitot, 1975; Meier & Hay, 1975; Murray etal. 1979). Finally, when neoplastic epithelial cells invade deeper tissues or when normal epithelial cells migrate to close a wound, the mobility of the cells is dependent upon the availability of suitable migration-supporting macromolecules in the connective tissue compartment.

Recently (Donaldson & Dunlap, 1981), we found that untreated nucleopore filters inserted under one edge of a fresh skin wound on adult newts were a poor substrate for epidermal cells migrating from the wound. However, if the filters were coated with gelatin, their ability to support migration increased dramatically. Using scanning electron microscopy, we learned that as the epithelial sheet advances on the filter the cells adjacent to the leading edge extend broad, thin, fan-shaped processes onto the substrate in the direction of movement. These processes possess specialized attachments on their undersurface that bind the cell to the substrate. This ability of gelatin to convert a poor migration substrate into a good one and the fact that collagen is a major component of most extracellular matrices stimulated the present investigation into the response of migrating epidermal cells to various types of collagen and certain collagen-derived peptides.

Adult male newts (Notophthalmus viridescens) obtained from Connecticut Valley Biological were stored at 4 °C in 1/10 strength operating solution (Rose & Rose, 1965) containing 0·14 ml/1 of Wardley’s Aquatonic (a commercial product commonly used for disease prevention in tropical fish). At least a week prior to use, the animals were moved to room temperature and kept in the same solution under natural illumination.

Rectangular wounds with their long axis extending proximo-distally were made by removing a piece of skin from the dorsal surface of each hind limb between the knee and ankle. Wounded limbs were amputated through the thigh and explanted into 5 ml of full-strength Holtfreter solution (HS) containing 0·05 g/1 of streptomycin in 35 mm × 10 mm plastic dishes. The clotted blood was then cleaned from each wound, the limbs were transferred to 5 ml of fresh HS and a triangular piece of nucleopore filter was then inserted, sharp end first, under the anterior wound margin (Fig. 1A). Eight hours later, limbs were fixed overnight in 10% formalin, and then the migrating cells were stained by immersing the entire limb briefly in 0·1% crystal violet. After the filter was dissected free from the limb (Fig. 1B), it was placed under a compound microscope equipped with a Leitz drawing tube. The magnified image of the filter and wound epithelium was drawn onto a sheet of paper, and the area occupied by a standardized width of wound epithelium (distance migrated) was determined using a polar planimeter. These values were used to compare the effectiveness of various substrates. Experiments were designed so that one hind limb of an animal served as a control, and the other as an experimental. This allowed us to analyse the results with a paired t-test, thereby reducing the impact of animal variability. P values less than 0·05 were judged significant.

Nucleopore filters (0·2 μm pore size, 25 mm diameter) were immersed for 15 min in 0·5% acetic acid and were then washed briefly with distilled water. Subsequently, 0·5 ml of 0·1 M-acetic acid, or the same vehicle containing, in solution, one of the following genetic types of collagen (human I, newt I, bovine II, bovine IV) or one of three different cyanogen bromide (CNBr) peptides (α₁ (I)CB3, CB7, or CB8) was pipetted onto the dull side of the filter, spread evenly, and placed at 23 °C overnight to dry. In one experiment, filters were coated with glutaraldehyde-linked collagen or bovine serum albumin (BSA). A solution of type I collagen from chick embryo skin (8·5 mg/ml in 0·1 M-acetic acid) was mixed in equal amounts with 2% glutaraldehyde, and 200 μlof the resulting mixture was spread evenly over a filter and allowed to polymerize overnight in a moist chamber. Other filters were coated similarly with BSA. To make the viscosity of the BSA solution similar to the collagen solution, 57 mg/ml of BSA was required. The polymers were then treated as follows (modified from the method of Macieira-Coelho & Avrameas, 1972): 11 h in 2% glutaraldehyde, 60 h in Na borate/HCl buffer (0·02 M, pH 8·1), 20 h in borate buffer containing 0·2 M-glycine (to block free aldehydes), 48 h in phosphate-buffered saline/EDTA at 4 °C (0·007 M-phosphate (pH 7·4), 2 mg EDTA/1) and finally, overnight in HS.

Type I collagen and α₁ α₁ (I) CB3, CB7, and CB8 were gifts from Dr Jerome M. Seyer. These collagens were extracted from either human skin or placenta and purified as described previously (Seyer, Hutchenson & Kang, 1976). Bovine type II collagen, prepared from foetal bovine skeleton (Stuart et al. 1979) was a gift from Dr Michael A. Cremer. Chick type I (Dixit, Seyer & Kang, 1977) and type IV collagen from bovine anterior lens capsule (Dixit, 1978) were contributed by Dr Saryu N. Dixit. The newt type I collagen was prepared by digestion of newt tails with pepsin and purified by fractional precipitation with salt (Smith & Linsenmayer, 1982).

Preliminary experiments in which we coated filters with 125 μg/ml of type I collagen showed that this would greatly improve migration compared to untreated filters. We then conducted a dose-response study, in which one hind limb of each animal received a filter implant treated with 125 μg/ml of type I collagen (hereafter referred to as collagen controls) and the other limb received an implant treated with less type I. collagen, the exact amount depending upon the group to which it belonged. Fig. 2 shows that the dose-response curve breaks at 1·25 μg/ml. At this concentration, migration was still as good as that on the collagen control filters. However, lowering the amount of collagen to 0·25 μg/ml appeared to lessen its ability to support migration (although the decline was not quite significant). Migration on filters receiving 0·125 μg/ml was significantly retarded, being only 38% as good as the collagen controls. Migration on untreated filters was only 11% as good as the collagen controls. Heat-denaturing a sample of type I collagen for 55 min had no effect on its ability to serve as a migration substrate when tested at 125 μg/ml or 1·25 μg/ml (Table 1).

Thus the epidermal binding sites on type 1 collagen do not seem to be related to the tertiary structure of the molecule but rather to the covalent structure of the component alpha chains.

We next compared several types of collagen as migration substrates. Fig. 3A shows the results of experiments comparing 1·25 μg/ml of human type 1 (the left bar in each pair) to that same concentration of newt type I, bovine type II, or bovine type IV collagens. In each experiment, the test collagen was as effective as the type I reference. Fig. 3B underscores the similarity in the various collagens by showing that when the three test collagens were reduced to one-tenth of the amount in Fig. 3A they lost effectiveness, just as did human type I in the dose-response curve.

Finding that the various collagens all improved migration equally well, we wondered if this was simply a non-specific protein effect. Filters coated with BSA were no better than untreated filters (data not shown). However, without proof that BSA actually remained on the filter, no conclusion was possible. Therefore we placed glutaraldehyde-linked polymers of BSA onto filters in films thick enough to be seen. Chick embryo type I collagen films were similarly produced for comparison. Of seven limbs receiving the collagen film, five showed migration, ranging from 126 units to 506 units, with a mean of 310. No migration occurred on any BSA filter. Since the BSA was applied to a filter previously coated with 1·25 μg/ml of type II collagen, the BSA had converted a good substrate into a poor one. Thus, while the collagen polymer was inferior to airdried collagen as a substrate, it was clearly superior to the BSA polymer.

Table 2 shows that at 1·25 μg/ml, all three of the α₁ (I) cyanogen bromide peptides tested were somewhat inferior to native type I collagen as a substrate. Of greater significance than this inferiority is the fact that when each of the peptide means in Table 2 was compared to the mean (132 ±26) for 20 untreated filters from various other experiments, the peptides were clearly superior (P values less than 0·0002 in each case). Thus, the α₁ chain has at least three, and probably many, epidermal cell binding sites.

Attachment assays have been used by several investigators to explore epithelial— collagen interactions. Thus, freshly isolated guinea pig epidermal cells (Murray et al. 1979) and a cell line (PAM212) derived from mouse epidermis (Terranova et al. 1980) both show a distinct preference for type IV collagen. In contrast, adherence of freshly prepared rat hepatocytes to plastic dishes is improved equally well by a coating of any of the five collagen types (Rubin, Hook, Obrink & Timpl, 1981). Since newt epidermal cells migrated equally well over interstitial collagen (type I), cartilage collagen (type II) and basement membrane collagen (type IV) it might appear that these cells resemble rat hepatocytes more than they do the epidermal cells described above. Such a conclusion assumes that migration studies and adhesion assays measure the same thing. While migration involves adhesion to the substrate, it includes cellular events subsequent to attachment as well. Thus, it is important to know how different levels of cell-substrate adhesion affect mobility. The available evidence suggests that given a choice, cells will preferentially choose the more adhesive of two substrates (Harris, 1973) and this choice will then lead to a decrease in mobility. For example, Gail & Boone (1972) found that normal and transformed fibroblasts seeded onto Pyrex are less motile and more adherent to the substrate than are cells on cellulose acetate. Similarly, the more negative the substrate charge (a condition tending to decrease the affinity of normally negative cell surfaces for the substrate) the more motile cells become (Sugimoto & Hagiwara, 1979). Finally, lanthanum treatment decreases the motility of glial cells and increases their adherence to plastic culture dishes (Letourneau & Wessells, 1974). (This last example, however, may have been due in part to lanthanum effects on calcium control of the glial contractile machinery.) If we accept this indirect relationship between mobility and adhesion as a general principle for all cells (it has been shown convincingly only for fibroblasts) we must conclude that newt epidermal cells adhere equally well to all the native collagens tested. How can this conclusion be reconciled with adhesion assays that show a preference of epidermal cells for type IV collagen ?

Attachment assays are conducted after cells in culture, or tissue cells, are disaggregated by enzymes. If protein synthesis in such cells is shut off, they show a great reduction in the ability to bind collagen (Terranova et al. 1980). These assays therefore actually measure the ability of enzymically altered cell surfaces to regenerate attachment factors for collagen. Consequently, when a particular collagen preference is found in attachment assays this may simply reflect a difference in the rate of regeneration of separate classes of collagen-binding proteins. By contrast, in migration studies such as ours, where epithelial cells are not exposed to enzymes, the interaction with collagen may involve a wider spectrum of collagen receptors. While not designed to address this question, the work of Stenn, Madri & Roll (1979), in which mouse skin explants were made into plastic dishes coated with collagen types I, III, IV and V, suggests this is the case. The authors state that after 3 days of culture, migration on collagen was 59% greater than on uncoated dishes. No preference for any of the collagens was mentioned. In addition, their figure 2, which shows the ability of a proline analogue to inhibit migration on the various collagens when added after 2 days in culture, shows similar end-points at day 3 for all collagen types. Thus, in mouse epidermis a preference for type IV in attachment assays is apparently not reflected in migration behaviour.

There has been considerable interest recently in collagen attachment factors. From studies in tissue culture, two major adhesive proteins have been implicated, fibronectin and laminin. Fibronectin is a fibroblast cell-surface and blood-borne glyco-protein found in many species (Yamada & Olden, 1978), including salamanders (Repesh, Furcht & Smith, 1981). In attachment assays, fibronectin improves the ability of fibroblasts to bind to type I and type IV collagen, but has no effect on the binding of freshly isolated guinea-pig epidermis to these collagens (Murray et al. 1980). Since in the α₁ (I) chains, the major fibronectin binding site is found in CB7 (Kleinman, McGoodwin & Klebe, 1976), the fact that migration was similar on all three peptides in our experiments suggests that fibronectin is not necessary in this system for epidermal-collagen interactions. However, because we have not actually measured the affinity of newt fibronectin for the three human peptides, this conclusion must be considered tentative.

Laminin is another cell surface glycoprotein, but is secreted primarily by epithelial cells (Foidart et al. 1980). Unlike fibronectin, it is not found in serum (Terranova et al. 1980). It is a component of basement membranes where it is localized in the lamina lucida between the epithelial cells and the type IV collagen making up the lamina densa (Foidart et al. 1980). The preferential adherence of epidermal cells to type IV collagen, appears to be mediated by laminin (Terranova et al. 1980). This specific affinity for type IV collagen plus its location in basement membranes suggest that laminin is the ‘glue’ that binds epithelial cells to their basement membranes. At this point it is not possible to comment on the role of laminin in migrating newt epidermis, except to say that since this system shows no preference for type IV collagen, any type IV-specific ligand must be accompanied by other collagen binding proteins as tvell.