The effects of fibronectin on the migration of human foreskin fibroblasts and syrian hamster melanoma cells into three-dimensional gels of native collagen fibres

The biochemical composition and structural orientation of the extracellular matrix play an important role in the control of cell migration under both normal and pathological conditions in vivo (Leighton, Kalla, Kline & Belkin, 1959; Trinkaus, 1969; Willis, 1973; Strauli & Weiss, 1971; Hay, 1978). Collagen is a major constituent of the extracellular matrix (Miller, 1977) and has been used as a substratum for the culture of a number of cell types in vitro (Ehrmann & Gey, 1956; Michalopoulos & Pitot, 1975; Emermann & Pitelka, 1977). Techniques for obtaining quantitative data regarding cell migration into 3-dimensional gels of native collagen fibres have been described in a previous communication (Schor, 1980); cell migration into the collagen gel is measured by plating cells on the gel surface and then determining the percentage of cells within the 3-dimensional collagen matrix at various times thereafter. In view of the important role played by the extracellular matrix in the control of cell migration in vivo, it is desirable that cell migratory behaviour in vitro be monitored on biologically relevant macromolecular matrices, such as the 3-dimensional collagen gel, rather than on 2-dimensional artificial substrata. Our long-term objective is to study cell migration, especially in relation to the process of tumour cell invasion, on progressively more complex macromolecular matrices prepared by the stepwise addition of other matrix components to the collagen gel. We have chosen to begin with fibronectin because of its common association with collagen in the extracellular matrix (Linder, Stenman, Lehto & Vaheri, 1978) and at the cell surface (Bornstein & Ash, 1977).

Fibronectin is a high-molecular-weight glycoprotein found in an insoluble form at the surface of normal fibroblasts and other cell types (Mosher, Saksela, Keski-Oja & Vaheri, 1977; Hynes, Destree, Perkins & Wagner, 1979; Smith, Riggs & Mosesson, 1979) and in a soluble form in serum, where it has been referred to as cold insoluble globulin (Grinnell & Hays, 1978). Further information concerning the biochemistry of fibronectin and its role in mediating many aspects of cell-cell and cell-matrix interactions may be found in a number of excellent reviews (Vaheri & Mosher, 1978; Yamada & Olden, 1978; Hynes et al. 1979).

In this communication we report the effects of serum-derived fibronectin on cell migration into 3-dimensional gels of native collagen fibres, using normal human skin fibroblasts and the highly tumourigenic Syrian hamster melanoma cell line, RPMI-3460 (Moore, 1964). Data are presented indicating that the presence of fibronectin in the collagen substratum stimulates the migration of the melanoma cells, but inhibits the migration of fibroblasts into the collagen gel.

RPMI-3460 Syrian hamster melanoma cells were originally obtained from Dr M. Steinberg (Department of Pathology, New York University Medical Center) and human fibroblasts were isolated in this laboratory from foreskin specimens obtained from St Mary’s Hospital, Manchester. Stock cultures of both cell types were grown in plastic tissue culture dishes in Eagle’s MEM supplemented with 10 % foetal calf serum, 2 rriM glutamine, 1 mM sodium pyruvate, non-essential amino acids (Gibco-Biocult) and too units/ml of penicillin and streptomycin. Stock cultures were subcultured once a week and the medium changed 3 times a week. Cells to be used in the experiments were brought into suspension from stock cultures by exposure to 0·05 % trypsin (Sigma, Ltd, Cat. No. T-8253) in phosphate-buffered saline for 5 min at 37 °C, followed by the addition of equal volume of growth medium containing 10% foetal calf serum and collecting the cells by centrifugation for 5 min at 800 g.

Type I collagen was extracted from rat tail tendons as previously described (Schor, 1980). The concentration of collagen in the aqueous stock solution was adjusted to 2·3 mg/ml. Threedimensional gels of native collagen fibres were prepared in 35-mm plastic tissue culture dishes (Gibco-Biocult, Ltd, Uxbridge, Cat. No. 53066) by rapidly mixing 8·5 ml of the collagen solution with 1 ml of 10 × concentrate MEM and 0·5 ml of 4·4% sodium bicarbonate and pipetting 2-ml aliquots into the dishes. Gels set within 5 min and were incubated at 37 °C for 24 h in a humidified CO₂ incubator before use. These gels consist of a hydrated meshwork of native collagen fibres (Elsdale & Bard, 1972).

The total number of cells growing on the 3-dimensional collagen gels was determined as previously described (Schor, 1980) by dissolving the gel with bacterial collagenase (Sigma Ltd, Cat. No. C2139). The percentage of total cells within the collagen gel matrix was determined using the ‘microscopic method’ previously described (Schor, 1980). Accordingly, cultures were examined with phase-contrast optics using a Leitz Diavert microscope fitted with an SY2 photographic graticule defining an area of 0·9 × 0·65 cm’. The number of cells both on the gel surface and within the 3-dimensional collagen matrix was determined in approximately 20 regions of the gel surface selected at random moving across the diameter of the gel. Data collected from at least 3 such gels were used to calculate the mean ±S.D. of the percent of cells within the gel matrix for each result presented.

Fibronectin was prepared from human serum by affinity chromatography on gelatin-Sepharose columns as described by Engvall & Rouslahti (1977). After application of the serum, the column (12 × 1·5 cm) was washed with 0·15 M NaCl in 0·01 M potassium phosphate buffer, pH 7·4 (PBS), then with PBS containing 1·6 M urea. Fibronectin was eluted with 6 M urea in PBS and the solution was dialysed overnight against 50 vol. of buffer containing 10 mM cyclohexylaminopropane sulphate, 0·15 M NaCl and 1 mM CaCl_s (CAPS buffer) (Yamada & Kennedy, 1979). Fibronectin was concentrated by precipitation with ammonium sulphate at 40% saturation. The precipitate was dissolved in CAPS buffer and dialysed against this buffer to remove ammonium sulphate. The protein concentration was adjusted to 2–4 mg/ml (depending on the experiment) and SDS electrophoresis showed a major band (doublet) at mol.wt 220000 and minor bands (< 10 %) of low-molecular-weight components. Growth medium containing 15 % fibronectin-depleted foetal calf serum was prepared by applying foetal calf serum to the gelatin-Sepharose column and washing with growth medium (Eagle’s MEM containing 2 mM glutamine, 1 mM sodium pyruvate, non-essential amino acids (Gibco-Biocult) and 100 units/ml of penicillin and streptomycin).

Fibronectin binds avidly to a specific region of the ai chain of type I collagen between amino acid residues 568 and 835 (Kleinman, McGoodwin & Klebe, 1976). In order to examine the effects of fibronectin on the migration of RPMI-3460 melanoma cells, collagen gels containing bound fibronectin were prepared by incubating gels with 1 ml of serum-free growth medium containing either 0, 1, 5 or 50 µg of fibronectin for 1 h at 37 °C and then washing the gels 5 times with serum-free medium to remove unbound material. The melanoma cells were detached from stock cultures by exposure to trypsin, resuspended in serum-free growth medium containing 1 mg/ml bovine serum albumin (Gibco-Biocult, Cat. No. 164) and 10% (v/v) lactalbumin hydrolysate (Sigma Ltd, Cat. No. A-4503) at 4×10⁴ cells/ml and 1 ml of this cell suspension plated on control and fibronectin-preincubated gels already overlaid with 1 ml of serum-free medium. The presence of bovine serum albumin and lactalbumin hydrolysate resulted in improved cell survival in serum-free medium during the course of the experiment. Cultures were then incubated at 37 °C and both total cell number and the percentage of cells within the gel were measured daily as previously described (Schor, 1980). As can be seen in Fig. 1A, there was no increase in total cell number in serum-free medium during the 3-day incubation period on either control gels or gels preincubated with fibronectin. Cell viability was estimated at the end of the experiment by trypan blue exclusion and judged to be greater than 90% in all cases. The effects of fibronectin on cell migration into the gel are shown in Fig. 1 B. Cell migration into the gels proceeded in an approximately linear fashion on all substrata and fibronectin was observed to have a dose-dependent stimulatory effect on cell migration, with 28% of the cells present within gels preincubated with 50 µg fibronectin after 3 days in culture compared to only 10% within control gels at this time. These results also indicate that the migration of melanoma cells into the collagen gels (either with or without fibronectin) may occur in serum-free medium and is thus not dependent on factors supplied by serum.

The effects of fibronectin on cell proliferation and migration in the presence of serum containing media are shown in Fig. 2. Gels preincubated with 50 µg of fibronectin in serum-free medium were prepared as described above and then both control and fibronectin-preincubated gels were overlaid with 1 ml of growth medium containing 10% fibronectin-depleted foetal calf serum. Control gels overlaid with 1 ml of growth medium containing 10% whole foetal calf serum (i.e. containing fibronectin) were prepared at the same time. Melanoma cells were suspended at 4×10⁴ cells/ml in growth medium containing 10% of either fibronectin-depleted or whole foetal calf serum and 1 ml of these cell suspensions plated onto gels overlaid with the corresponding medium. Total cell number and the percentage of cells within the gel were measured daily for the next three days. As shown in Fig. 2, cell proliferation occurred at the same exponential rate in the presence of both 10% fibronectin-depleted serum (on control and fibronectin-preincubated gels) and 10% whole serum (on control gels). In contrast, there was significantly more cell migration into those gels preincubated with fibronectin compared to control gels, with approximately 28% of the cells found within the fibronectin-preincubated gels after 3 days of growth compared to only 17·5% within control gels. These data indicate that melanoma cells migrate to a greater extent into a collagen substratum containing fibronectin compared to a simple collagen substratum under conditions which support cell proliferation (i.e. in the presence of serum). The data presented in Fig. 2 also indicate that the amount of fibronectin in medium containing 10% whole foetal calf serum is not sufficient to influence cell migration; this result was not due to possible differences in the biological activity of human and bovine fibronectins, since the migratory behaviour of melanoma cells on gels preincubated with 50 µg of bovine fibronectin was indistinguishable from that shown in Figs. 1 and 2 for gels preincubated with human fibronectin (data not shown).

Results identical to those shown in Figs. 1 and 2 were obtained when fibronectin was present continuously in the medium during the 3-day duration of the experiment rather than preincubated with the gels. Melanoma cells were suspended at a concentration of 4×10⁴ cells/ml in either serum-free medium (containing bovine serum albumin and lactalbumin hydrolysate) or medium containing 10% fibronectin-depleted foetal calf serum and 1-ml aliquots were then plated onto collagen gels overlaid with 1 ml of the corresponding medium (control) or 1 ml of medium containing 50 µg fibronectin. The percentage of cells attached two hours after plating was determined as previously described (Schor & Court, 1979) and the remaining gels were incubated for 3 days at 37 °C, at which time both total cell number and the percentage of cells within the gel were measured. As can be seen in Table 1, melanoma cells migrated into the collagen gels to a greater extent in the presence of medium containing fibronectin compared to controls. These data also indicate that the initial attachment of melanoma cells to the collagen gel (2 h after plating) was not significantly higher in the presence of fibronectin, a finding consistent with our previous results (Schor and Court, 1979; Schor, 1979).

All concentrations of fibronectin examined (5–400 µg/ml) stimulated the migration of melanoma cells into the collagen gel, although in certain experiments concentrations of fibronectin greater than 100 µg/ml resulted in a lower stimulation than obtained with 50 µg/ml (data not shown).

The morphology of melanoma cells after 3 days of culture on collagen was not affected by the presence of fibronectin (Fig. 3). Cells in the presence or absence of fibronectin were spherical in appearance, both on the gel surface and within the 3-dimensional collagen matrix.

The effects of fibronectin on the proliferation and migration of human foreskin fibroblasts in serum-containing medium are shown in Fig. 4. Cells were suspended at a concentration of 2 × 10⁴ cells/ml in foetal calf serum and 1 ml of this suspension was then plated onto collagen gels overlaid with 1 ml of this medium (control) or medium containing 100 µg fibronectin. Cultures were incubated for 3 days at 37 °C. Fibronectin had no effect on cell proliferation during this period (Fig. 4). In contrast, fibronectin did have a significant inhibitory effect on fibroblast migration into the gel, with 5’9% of the cells within the gel after 3 days of growth in the presence of fibronectin compared to 15’5% in the absence of fibronectin. Fibronectin had no observable effect on cell morphology during this 3-day period (Fig. 5).

A similar inhibitory effect of fibronectin on fibroblast migration was obtained when fibronectin-preincubated collagen gels were used. Collagen gels were incubated with 1 ml of either serum-free medium (control) or serum-free medium containing 200 µg fibronectin. After washing 5 times with serum-free medium, gels were overlaid with the presence of both serum-free and serum-containing media, indicating that fibroblast migration into the collagen gel (as is the case with melanoma cells) does not depend on factors supplied by serum. These results are in contrast with the reported serum requirement for the migration of 3T3 cells on plastic tissue culture dishes (Lipton, Klinger, Paul & Holley, 1971).

As can also be seen in Table 2, there was an approximate doubling in total cell number during the 3-day period in serum-containing medium and no increase in cell number in serum-free medium, on both control and fibronectin-preincubation gels. Furthermore, the initial attachment of fibroblasts (within 2 h) to the native collagen gel was greater than 90% under all conditions examined and thus not dependent on the presence of exogeneous fibronectin.

We did not observe differences in the migratory behaviour of fibroblasts in medium containing 15% fibronectin-depleted foetal calf serum compared to whole foetal calf serum (data not shown); again, this is presumably due to an insufficient quantity of fibronectin in medium containing 15% whole foetal calf serum to affect cell migration. Concentrations of fibronectin in the growth medium as low as 50 µg/ml resulted in a significant inhibition of fibroblast migration (data not shown) and a stimulation of fibroblast migration was never observed with any concentration of fibronectin examined (5-400 µg/ml).

Both the melanoma cells and fibroblasts used in this study were examined for the presence of endogenous, surface-associated fibronectin using anti-fibronectin antibody ; the fibroblasts were found to have considerable surface fibronectin organized in a similar fashion as previously described (Hynes et al. 1978), while the melanoma cells did not have fibronectin which could be detected by this means (N. Tolson, personal communication).

The collagen gels used in this study consist of a hydrated, random meshwork of native type I collagen fibres (Elsdale & Bard, 1972; Schor, Allen & Harrison, 1980). Our previous studies have indicated that cells vary in their capacity to migrate from the gel surface into the 3-dimensional collagen matrix; all epithelial cells examined (both normal and tumour) do not migrate into the gel, while all fibroblasts (both normal and virally transformed) and certain tumour cells of non-epithelial origin (e.g. melanoma) do migrate into the collagen gel at rates which vary considerably according to cell type (Schor, 1980). The migration of the R3460 melanoma cells and human skin fibroblasts into the collagen gel has previously been shown not to depend on the enzymatic digestion of the collagen substratum (Schor et al. 1980).

In this study, quantitative data regarding cell migration from the surface of the collagen gel into the 3-dimensional gel matrix (previously referred to as ‘infiltration’; Schor, 1980) have been obtained by first plating cells on the gel surface and then measuring the percentage of the total cell number within the collagen matrix at various times thereafter. Two distinct parameters of cell migratory behaviour are measured by this experimental approach; these are (1) the ability of cells to move from an essentially 2-dimensional substratum (i.e. the gel surface), where cells make contact with the collagen fibres only at their basal surface, into an isotropic, 3dimensional environment (i.e. the gel matrix), where cells are potentially able to establish adhesive interactions with collagen fibres over their entire surface, and (2) the ability of cells to migrate along the collagen fibres once they are within the 3-dimensional collagen matrix. These 2 parameters appear to be capable of independent expression, since certain cell types (e.g. HeLa) are able to migrate on the gel surface or within the 3-dimensional collagen matrix (when initially placed there as a single cell suspension when the gel is formed), but are not able to migrate from the gel surface into the gel interior (Schor, 1980).

In this communication we present evidence that fibronectin isolated from serum stimulates the migration of R3460 Syrian hamster melanoma cells into 3-dimensional gels of native collagen fibres, but inhibits the similar migration of normal human skin fibroblasts. This is the first study in which quantitative data have been presented regarding the effects of fibronectin on cell migration into a 3-dimensional macromolecular matrix. Fibronectin did not affect cell proliferation or cell morphology under the experimental conditions used in this study.

As discussed above, fibronectin may influence the migration of cells into the collagen gel by altering the ability of cells to translocate from the gel surface into the collagen matrix as well as by an effect on the rate of cell migration once within the 3-dimensional gel environment. At this point we are unable to conclude with certainty by which means fibronectin produces the observed effects on cell migration into the gel. As previously described (Schor, 1980), a qualitative assessment of cell migration within the collagen gel matrix (as well as on the gel surface) may be obtained from the relative position of daughter cells within individual colonies. However, in this particular case, both the hamster melanoma cells and human skin fibroblasts produce rather diffuse colonies on and within the collagen gel (indicating that daughter cells are capable of migration under these conditions) and the presence of fibronectin did not result in apparent changes in colony morphology (data not shown). Quantitative data regarding cell migration on and within the gel may be obtained using time-lapse cinematography and this information will be presented in a subsequent communication.

Previous studies have indicated that fibronectin stimulates the migration of several transformed cell lines on plastic tissue culture dishes, as assessed by the proximity of daughter cells in individual colonies (Ali & Hynes, 1978), the distance of cell migration from cell aggregates (Yamada, Olden & Pastan, 1978) and the length of individual cell tracks recorded by time-lapse cinematography (Poussegur, Willingham & Paston, 1977). An inhibitory effect of fibronectin on cell migration was suggested by Couchman & Rees (1979) who observed that the cessation of fibroblast migration from chick heart explants was temporally correlated with the appearance of fibronectin at the cell surface.

The mechanism by which fibronectin influences cell migration is not known. Cell surface fibronectin has been observed to be in close anatomical proximity to the filamentous elements of the cytoskeleton, possibly in continuity with these elements via transmembrane linkages (Hynes & Destree, 1978; Singer, 1979), and fibronectin may therefore be expected to influence cell migratory behaviour by virtue of this association. Albrecht-Buehler (1977) has reported that the rate of centripetal membrane flow at the apical surface of migrating 3T3 cells is reduced in regions of the membrane containing associated fibronectin; since centripedal membrane flow has been postulated by Harris (1973) to result from tension applied to the surface membrane due to interaction with underlying elements of the cytoskeletal system, the results reported by Albrecht-Buehler provide further indirect evidence that fibronectin may influence cell migration by modulating the functional relationship between surface membrane and cytoskeleton.

Fibronectin binds to native type I collagen fibres (Kleinman et al. 1976). Our data obtained using gels preincubated with fibronectin indicate that fibronectin bound to collagen influences cell migration. Previous studies have suggested that fibronectin mediates the attachment of a number of cell types to collagen-coated dishes and that this is a 2-step process in which fibronectin first binds to the collagen substratum, followed by cell attachment to the fibronectin-collagen complex (Klebe, 1974; Pearl-stein, 1978). The molecular organization of collagen on these coated dishes is not known, but since they were treated with 8 M urea during the course of their preparation, it is likely that a majority of the collagen molecules are denatured (Schor & Court, 1979). Other studies have indicated that exogenous fibronectin is not required for the attachment of a variety of cell types to native type I collagen fibres (Grinnell & Minter, 1978; Schor & Court, 1979; Schor, 1979), a conclusion consistent with the results presented here in Tables 1 and 2. It is, however, possible that although fibronectin is not required for the initial attachment of certain cell types to native type I collagen fibres (as measured by the kinetics of cell attachment during the first few hours after plating), it may nevertheless increase the strength of cell adhesion to collagen; this possibility has not been experimentally verified since the techniques available for measuring the strength of cell adhesion to a substratum (Gail & Boone, 1972 cannot easily be applied to a fragile structure such as the collagen gel. If fibronectin does indeed increase the strength of cell adhesion to the collagen substratum, then the effects of fibronectin on cell migration reported here may be viewed as an example of haptotaxis, as defined by Carter (1965). According to this view, the stimulatory effect of fibronectin on the migration of melanoma cells (which lack endogenous cell surface fibronectin) may be due to an increase in the strength of cell adhesion towards a value which is optimal for cell migration, while the inhibitory effect of fibronectin on the migration of fibroblasts (which possess abundant cell surface fibronectin) may result from an increase in cell adhesion beyond this optimal value. Indeed, Hynes et al. (1978) have suggested that fibronectin at sufficiently high concentration might be expected to inhibit cell migration by virtue of its enhancement of cell adhesion, although experimental evidence indicating such an inhibitory effect was not obtained, possibly because plastic tissue culture dishes were used as a substratum for cell migration. This conclusion is consistent with the results reported by Yamada et al. (1978) who observed that the effect of exogenous fibronectin on the alignment of transformed cells is more pronounced when the cells are cultured on a collagen substratum compared to plastic, which again serves to emphasize the important role played by the substratum in the control of various aspects of cell behaviour.

Whatever the mechanism by which fibronectin acts, our data suggest that the migration of different cell types in tissues containing type I collagen fibres (e.g. dermis) may be affected in diverse ways by the presence of fibronectin in the surrounding extracellular matrix. This differential response to fibronectin may play an important role in maintaining normal tissue structure as well as contribute to the pattern of tumour cell invasion.

We wish to thank Mr G. Rushton and Mr B. Winn for excellent technical assistance during the course of this work.