Modulation of scatter factor/hepatocyte growth factor activity by cell-substratum adhesion

It is becoming increasingly apparent that the loss of epithelial properties and the acquisition of more mesenchymal phenotypes by epithelia, is central to development and regeneration, and is involved in pathogenic progression, particularly with regard to the invasiveness of tumour cells. One approach to understanding the factors and processes that initiate and control such epithelial/mesenchymal transformations is to examine the actions of motility factors, such as scatter factor/hepatocyte growth factor (SF/HGF).

SF/HGF is a multifunctional polypeptide growth and motility factor whose receptor is a transmembrane tyrosine kinase, the c-metproto-oncogene product (see reviews by Gherardi and Stoker (1991), Nakamura (1991)and Furlong (1992)). The information derived from three main experimental approaches, each of which used different functional assays to determine activity, has in recent years converged to confirm that a single polypeptide species is responsible for very different effects in different target cells. Scatter factor, originally described as a fibroblast-secreted activity, was found to cause the dissociation of epithelial cell colonies and increase the motility of the individual cells (Stoker and Perryman, 1985; Stoker et al., 1987; Stoker, 1989), and increase epithelial cell invasiveness into three-dimensional collagen gels (Weidner et al., 1990). Hepatocyte growth factor (also known as Hepatopoietin A), an activity isolated from the plasma of partially hepatectomised mice, was found to stimulate the proliferation of primary cultured hepatocytes (Nakamura et al., 1984; Thaler and Michalopoulos, 1985). More recently, fibroblast conditioned medium was found to induce the production of branching tubules of MDCK cells cultured in 3-D collagen gels (Montesano et al., 1991a). Structural analysis, interchangability of activities and receptor identity, all contributed to the confirmation that SF/HGF is responsible for the diverse effects (Gherardi et al., 1989; Nakamura et al., 1989; Weidner et al., 1990; Konishi et al., 1991; Naldini et al., 1991; Furlong et al., 1991; Montesano et al., 1991b; Bhargava et al., 1992).

SF/HGF is secreted by mesenchymal cells and acts in a paracrine manner on epithelial and endothelial cells (Gherardi and Stoker, 1991; Furlong, 1992), though autocrine activity of SF/HGF by a non-differentiating keratinocyte cell strain is believed to contribute to its lack of differentiation (Adams et al, 1991). Transfection of MDCK cells with human SF/HGF resulted in clones producing SF/HGF that had altered morphology, increased cell motility and altered growth properties, suggesting autocrine activity (Uehara and Kitamura, 1992), whereas transfection of NIH 3T3 fibroblasts with the SF/HGF met receptor conferred increased motility and invasiveness in the presence of SF/HGF, but not increased growth, of transfected cells over wild type (Giordano et al., 1993). SF/HGF has been shown to be involved in morphogenesis (Montesano et al., 1991a,b) and differentiation (particularly in inducing neural tissue) in vitro (Yang and Park, 1993; Stern and Ireland, 1993), and evidence is now emerging that strongly implicates SF/HGF activity as being important during embryonic development in vivo (Stern et al., 1990; Defrances et al., 1992; Sonnenberg et al., 1993).

A number of changes in components of intracellular signalling pathways have been reported as a result of SF/HGF binding to the met receptor. These include the recruitment and activation of phosphatidylinositol 3-kinase (PI 3-kinase), ras GAP, phospholipase C gamma (PLC), phosphatidyl choline-specific phospholipase C (PC-PLC), and src-related tyrosine kinases (Graziani et al., 1991; Osada et al., 1992; Comoglio, 1993; Faletto et al., 1993; Halaban et al., 1993). The consequences of these signals include MAP kinase activation, PI turnover, increased intracellular Ca²⁺concentration ([Ca²⁺]_i), increased inositol 1,3,4,5-tetrakisphosphate (IP₄) concentration, and protein kinase C (PKC) activation. Separation of the effects of SF/HGF on growth and motility has suggested that diverging downstream signalling pathways are responsible for these different effects (Hartmann et al., 1992; Tajima et al., 1992; Giordano et al., 1993), though it is not as yet clear which pathways are involved in which effects. Though biological activity of SF/HGF correlated with the presence of the met receptor in target cells, low-affinity SF/HGF binding sites, believed to be heparan sulphate proteoglycans, are found on many cell types examined (Naldini et al., 1991; Komada et al., 1992; Tajima et al., 1992); therefore, the possibility that the different effects on different cells were initiated through different receptors could not be dismissed until the recent work of Weidner et al. (1993)showed, using MDCK cells transfected with a hybrid NGF/met receptor, that scattering, increased invasiveness, tubulogenesis and mitogenesis are all mediated by the met receptor.

The present study examines the influence of cell-substratum adhesion on the effects of SF/HGF on MDCK cells. This was prompted by the finding that MDCK cells cultured on glass coverslips, for the purpose of immunofluorescence staining, scattered poorly, or not at all, in response to MRC-5 fibroblast conditioned medium, whereas, on tissue culture plastic (TCP), pronounced scattering was observed. Here, MDCK cells were cultured on variously coated TCP or glass surfaces, and the effects of fibroblast conditioned medium, or medium containing purified recombinant mouse SF/HGF (rmSF/HGF), on the degree of scattering and DNA synthesis were determined. Scattering was found to be dependent on the substratum, but enhanced DNA synthesis was not. This demonstrates a clear separation of the mitogenic and motogenic effects of SF/HGF in the same cells. The data obtained also suggest a separation of the motility and colony dissociation components of scattering, and that a major difference in the signalling pathways leading to these different components is the modulation of the latter by the activity of ECM receptors.

MDCK cells were maintained in culture and serially passaged in Dulbecco’s modified Minimal Essential Medium (Gibco, UK) supplemented with 10% (v/v) fetal bovine serum (FBS) (Gibco, UK) and antibiotics, and harvested from sub-confluent cultures using trypsin/EDTA as described previously (Clark et al., 1992). All cells were cultured at 37°C in a humidified 5% CO₂atmosphere.

Individual wells of 24-well plates (Corning, UK), either with or without 13 mm diameter glass coverslips, were variously treated with attachment factors, to provide coated tissue culture plastic (TCP) or glass surfaces for culture of MDCK cells. To each well, either 1 ml of DMEM containing 10% FBS, 0.5 ml bovine fibronectin (20 μg ml⁻¹in PBS; Sigma, UK) or 0.5 ml human vitronectin (1 μg ml⁻¹in PBS; Sigma, UK) was added and incubated at 37°C for 90 minutes. Laminin-coated surfaces were obtained by firstly adding 0.5 ml poly-L-lysine (10 μg ml⁻¹; Sigma, UK) to wells for 30 minutes, removing excess, washing and incubating with 0.5 ml of 5 μg ml⁻¹laminin in PBS (Sigma, UK) for 60 minutes. After these incubations, excess solutions were removed and 0.5 ml of 0.1% (w/v) bovine serum albumin (Sigma, UK) in PBS was added to each well and incubated at 37°C for 30 minutes to block any remaining unoccupied sites.

In some experiments, attachment factors were covalently coupled to glass coverslips using a modification of the method of Aplin and Hughes (1981)for cross-linking proteins to aminosilane-coupled glass with glutaraldehyde. Briefly, clean glass coverslips were exposed to 1% 3-aminopropyltrimethoxy silane (Sigma, UK) in 95% ethanol for approx. 30 seconds at room temperature, and washed three times in 95% ethanol, then rinsed in water. The coverslips were then reacted with 2% glutaraldehyde in PBS, also containing 40 mM sodium cyanoborohydride (Sigma, UK) for 1 hour. After rinsing twice in PBS, coverslips were then reacted with a solution of the desired protein in PBS containing 40 mM sodium cyanoborohydride for 1 hour. Following this period, blockage of unreacted sites was acheived by incubating the coverslips in a solution of 100 μg ml⁻¹BSA in PBS containing 40 mM sodium cyanoborohydride for a further hour, then rinsed in PBS before cells were seeded onto the surfaces as below.

The degree of scattering was determined in a modification of the standard scatter titre assay (Stoker and Perryman, 1985). MDCK cells were seeded at low density into 24-well plates (12,000 cells per well), which had been previously coated with various attachment factors, as above. The cells were cultured in 1 ml of control growth medium (DMEM + 10% FBS), which was removed after 24 hours and replaced with either fresh control medium or the same medium containing purified mouse recombinant SF/HGF (a gift from Dr E. Gherardi, ICRF, Cambridge, UK) at a concentration of 1 or 5 ng ml⁻¹(DMEM + 10% FBS + SF/HGF). After a further 24 hours of culture, medium was removed, each well rinsed with PBS, and the cells fixed in 4% paraformaldehyde in PBS. After at least one hour, fixative was removed and cells were stained with 0.1% Coomassie Blue in methanol:water:acetic acid (50:50:7, by vol.) for approximately 10 minutes, then rinsed twice with distilled water, and allowed to air dry. In order to test if time in culture before exposure to SF/HGF, or a longer period of exposure to SF/HGF, had any effect on the responses of MDCK cells, two further experiments were undertaken. Firstly, cells were cultured for 48 hours in control medium, then cultured for a further 24 hours in fresh control medium with or without 1 ng ml⁻¹SF/HGF (48+24SF) before fixing and staining. Secondly, cells were cultured for 24 hours in control medium followed by a further period of 48 hours in fresh medium with or without 1 ng ml⁻¹SF/HGF (24+48SF).

In order to quantify the dissociation, or scattering, of colonies of MDCK cells, the average distance between nearest neighbouring nuclei was determined. Stained cultures were photographed under bright-field optics using 35 mm black and white negative film, and the negatives mounted into 35 mm projection slide mounts. Two fields from the centre of three separate wells were photographed for each treatment. Each mounted image was projected onto paper and the outlines of all nuclei were traced. The approximate centroid of each nucleus was marked, the distance between each nucleus and its nearest neighbour measured, and the average for each treatment calculated. Comparisons were made between average internuclear distances for different treatments.

Cells were seeded at low density (6,000 cells per well) into uncoated TCP 24-well plates. After culture for 24 hours in DMEM + 10% FBS, the medium was removed and replaced by either fresh control medium or fresh medium with 5 ng ml⁻¹mouse recombinant SF/HGF. Cultures were maintained for 1, 2 and 3 days before fixation and staining as above, and the number of cells in 10 fields (randomly from a track across the centre of each well) was determined microscopically. The mean values of cells per field from three wells per data point were averaged and expressed as cells per mm².

To determine the level of DNA synthesis, MDCK cells were seeded into variously coated wells, as above, and cultured for 24 hours in control growth medium (DMEM+10% FBS), after which time the medium was removed and replaced by either fresh control medium or fresh medium with 5 ng ml⁻¹mouse recombinant SF/HGF. Cells were cultured for a further 24 hours, before these media were replaced with 1 ml per well fresh medium (with or without SF/HGF as appropriate) containing 0.5 μCi ml⁻¹[³H]thymidine (specific activity 82 Ci mmol⁻¹; Amersham Life Science, UK) and cultured for 3 hours. Each well was then washed twice with PBS and three times with 10% (w/v) trichloroacetic acid. The residue in each well was solubilised in 0.6 ml 0.1% SDS in 0.1 M NaOH overnight. A 0.3 ml sample from each well was liquid-scintillation counted, and the protein content was determined in a corresponding 0.2 ml sample using the Bradford assay (Bio-Rad kit, Bio-Rad, UK). Six wells were set up for each medium/substratum combination. The number of dpm per well was in the ranges 7,000-12,000 in control cultures, and 13,000-20,000 in SF/HGF cultures. Protein content was found to be between 110 and 240 μg per well. Thymidine incorporation is expressed as dpm μg protein⁻¹.

Statistical comparisons between means, in the Scatter quantification, cell attachment, cell numbers and DNA synthesis experiments, were made using Student’s t-test (two-tailed). Differences were said to be significant if the Pvalue obtained was ≤0.05.

On planar culture surfaces, sub-confluent MDCK cells normally form coherent islands of relatively flattened cells. SF/HGF has been shown to alter this morphology by promoting the initial expansion of colonies as the result of further flattening of individual cells, followed by dispersion of the cells that make up these colonies, and an increase in their motility (Stoker and Perryman, 1985; Stoker et al., 1987; Stoker, 1989). On fibronectin-coated surfaces, colonies often had a more expanded irregular outline than on the other surfaces, and cells with more axially polarized morphology were seen more frequently (Fig. 1I). On serum- or vitronectin-coated glass, cells and colonies were less flattened than on fibronectin-coated surfaces, but soon after the addition of 1 ng ml⁻¹SF/HGF cells on all surfaces spread and colonies expanded such that cultures looked similar on all surfaces. This spreading became apparent after approximately 1 hour and appeared maximal after 4-5 hours (not shown). By 24 hours, MDCK cells formed typical colonies on all of the surfaces tested in control medium (DMEM + 10% FBS) (Fig. 1). Single cells, when present, had either a rounded or radially flattened morphology, though occasionally, single, axially polarized cells could be seen (Fig. 1).

The patterns of the scattering response found after exposure for 24 hours to either rmSF/HGF (1 or 5 ng ml⁻¹) or conditioned medium (not shown) were essentially identical. When cultured in the presence of 25% MRC-5 conditioned medium (not shown) or 1 ng. ml⁻¹rmSF/HGF (each also containing 10% FBS) for 24 hours (Fig. 1), MDCK cells had an obviously scattered morphology on fibronectin-coated surfaces, and on serum-coated TCP, but not when cultured on serum-coated glass, or vitronectin-coated TCP or glass, where colonies remained intact (Fig. 1). Though the cells were not separated to any degree, these non-scattered MDCK colonies appeared to be markedly expanded, with colony outlines being more irregular (Fig. 1J,K). Cells cultured on laminin-coated surfaces were markedly scattered to a similar, or greater, degree than on fibronectin (not shown).

The contrast between the degree of scattering on serumcoated glass and TCP surfaces is most clearly illustrated in Fig. 2, where, in the same well, MDCK cells cultured in the presence of 1 ng. ml⁻¹rmSF/HGF can be seen at different planes of focus (compare Fig. 2Aand B). The cells on the upper surface of the glass coverslip have remained as epithelial islands with few, if any, single cells, whereas on the adjacent TCP the cells are highly scattered, many of them having fibroblastic morphology (Fig. 2B).

When cells were cultured for 48 hours before being exposed to 1 ng ml⁻¹rmSF/HGF for 24 hours (48+24SF), those on serum-coated glass were poorly, if at all, scattered, and those on fibronectin-coated glass were markedly dispersed (Fig. 3C,D). When cells were cultured for 24 hours in control conditions followed by 48 hours in medium containing 1 ng ml⁻¹rmSF/HGF (24+48SF) (Fig. 3E), a small amount of scattering was seen on vitronectin-coated glass, but this was still markedly less than for cells cultured on fibronectin-coated glass (Fig. 3F). On serum and vitronectin-coated glass, cultures exposed to SF/HGF for 48 hours, many colonies had highly irregular margins (Fig. 3E,G,H). In many instances, cells at colony margins possessed extremely elongated processes, giving colonies a highly stellate or spikey outline (Fig. 3E,G,H).

In an attempt to obtain a more objective and quantitative measure of scattering, the average distance between nearest neighbouring nuclei were determined. The values obtained for the various conditions are expressed here as internuclear distance (IND) (Fig. 4). In the presence of 5 ng ml⁻¹SF/HGF, the pattern of values of IND objectively confirms that seen in Fig. 1. A second experiment using MRC-5 conditioned medium produced a series of values of identical pattern (not shown). Values on serum-coated glass, vitronectin-coated glass and TCP are not significantly different from their respective control values, whereas serum-coated TCP and fibronectin-coated surfaces have values that are substantially higher than their controls. However, this method does not appear to be sensitive enough to clearly detect colony expansion in the absence of scatter.

Because fibronectin is displaced from glass surfaces in the presence of serum (Grinnell, 1986), immobilisation on glass coverslips of fibronectin, vitronectin or a mixture of these two, was used to test if differences in substratum adhesiveness due to the different attachment factors were themselves responsible for the differences in scattering, or if a positive signal from fibronectin or a negative signal from vitronectin could account for the differences.

Cells cultured on surfaces of immobilised fibronectin or vitronectin at similar concentrations (10 μg ml⁻¹concentration during immobilisation) had a similar degree of attachment, but whereas SF/HGF-induced colony dissociation was markedly reduced or absent on vitronectin coverslips, cells cultured on immobilised fibronectin were clearly dissociated (Table 1). When cultured on surfaces containing both fibronectin and vitronectin, cell attachment was significantly increased, and colonies were dissociated after exposure to SF/HGF (Table 1).

Though previous studies have suggested that in MDCK cell monolayer cultures proliferation is unaffected by SF/HGF (Gerhardi et al., 1989; Tajima et al., 1992), experiments were carried out to assess the possible effects of SF/HGF (and the possible effects of the substratum) on cell numbers and the level of DNA synthesis. In Fig. 5Ait can be seen that SF/HGF stimulates proliferation on TCP as compared to standard culture conditions (i.e. DMEM + 10% FBS). Significant increases in cell numbers were found at days 1, 2 and 3.

The level of DNA synthesis, as determined by the incorporation of [³H]thymidine, was influenced by two factors: in the absence of SF/HGF, the substratum composition had a small effect; in the presence of SF/HGF, DNA synthesis was increased to similar levels on all substrata. In control cultures (DMEM + 10% FBS), the level of DNA synthesis was found to be influenced by substratum composition. Thymidine incorporation of MDCK cells cultured on vitronectin-coated surfaces was significantly greater (approximately 33%) than on corresponding fibronectin-coated surfaces (Fig. 5B). On serum-coated glass, thymidine incorporation was greater by approximately 40% than on fibronectin-coated glass, whereas no significant difference was found between serum- and fibronectin-coated TCP. This increased DNA synthesis in the absence of SF/HGF occurred on surfaces on which cells did not scatter in response to SF/HGF.

In SF/HGF-containing medium (DMEM + 10% FBS + 5 ng ml⁻¹rmSF/HGF), thymidine incorporation was significantly elevated, as compared to controls, in MDCK cells cultured on all surfaces tested (Fig. 5B). These increases range from 30% (on serum-coated glass) to 97% (serum-coated TCP). The differences in the degree of enhancement by SF/HGF reflect the differences in the values in the absence of SF/HGF, since the enhanced level of thymidine incorporation reached as a result of exposure to SF/HGF was similar on all surfaces. No significant differences were found between these increased levels of thymidine incorporation in SF/HGF-containing medium for cultures on different surfaces (Fig. 5B).

The importance of the substratum in modulating the scattering of MDCK cells was underlined, when a population of high-passage-number MDCK cells (designated MDCK-fs) spontaneously developed an altered scattering-response to substrata and to SF/HGF. On serum- and vitronectin-coated TCP, these cells appeared normal in control medium, but were visibly more scattered by SF/HGF than low-passage-number MDCK cells cultured in parallel (compare Fig. 6and Fig. 1). These cells were moderately scattered on fibronectin-coated TCP and markedly dissociated on fibronectin-coated glass in the absence of SF/HGF (Fig. 6C,I). On all glass-coated surfaces, the cells were relatively scattered, particularly on fibronectincoated glass, (Fig. 6G-I) and, as on TCP, their scattering response to SF/HGF was markedly enhanced in comparison to low-passage-number cells (Fig. 6J-L). Low-passage-number cells did not scatter on serum-coated glass or vitronectincoated surfaces, whereas the high-pass strain cells were highly scattered by SF/HGF on all surfaces (Fig. 9).

The morphology of MDCK cell cultures can be profoundly altered by SF/HGF. In monolayer culture, the cells lose cell-cell contacts and aquire a fibroblastic morphology (Stoker and Perryman, 1985; Stoker et al., 1987). When added to MDCK cell cysts growing in 3-D collagen gels, SF/HGF promotes the formation of branching tubular structures (Montesano et al., 1991a,b; Weidner et al., 1993). The scattered morphology induced in MDCK monolayer cultures by SF/HGF has been reported to occur in two stages. During the first 5-6 hours after exposure to SF/HGF, colonies expand as a result of an increase in spread area of individual cells (Stoker and Perryman, 1985; Dowrick et al., 1991; Li et al., 1992). After this time, breakdown of cell-cell contacts and increase in cell motility lead to the dissociation of colonies (Stoker and Perryman, 1985). An earlier morphological response to SF/HGF, the formation of a large apical circular ruffle that closes to form a macropinocytotic vesicle, does not appear to be related directly to scattering (Dowrick et al., 1993). Nearly all previous studies of scatter activity, including the standard scatter titre assay (Stoker et al., 1985), have been carried out on tissue culture plastic in the presence of serum. The studies of MDCK cells in 3-D collagen matrices have shown that scattering does not occur or is an early, transient response, as branching tubules are formed from cysts on exposure to SF/HGF (Montesano et al., 1991a,b; Weidner et al., 1993). The data presented here suggest that scattering of monolayer cultures is strongly influenced by the composition of the substratum. When MDCK cells are cultured on serum- or vitronectin-coated glass (and to a lesser extent vitronectin-coated TCP), SF/HGF will enhance DNA synthesis and induce initial colony expansion, but the cells fail to separate. When cultured on fibronectin-coated surfaces, colonies appear irregular (similar to the initial colony expansion seen early after SF/HGF addition to cultures on other surfaces) in the absence of SF/HGF, and in the presence of SF/HGF scattering is marked. It would seem therefore that the initial phase of the response to SF/HGF, i.e. colony expansion, is independent of the substratum composition, but the breakdown of cell-cell contacts and dispersion of the colonies is modulated by the substratum composition.

In this study, SF/HGF significantly elevated the level of DNA synthesis of MDCK cells on all of the surfaces tested. Previous reports on the effect of SF/HGF on MDCK cell growth have suggested that these cells are not responsive to the growth-stimulatory effects of SF/HGF (Gherardi et al., 1989; Tajima et al., 1992). Recently, it was shown that MDCK cells grown in collagen gels are growth-suppressed, and that SF/HGF will stimulate growth in these conditions (Weidner et al.,1993). These discrepancies are likely to reflect differences in culture conditions. Gherardi et al. (1989)assayed thymidine incorporation in serum-free conditions and found no effect of SF/HGF. Tajima et al. (1992)cultured cells in the presence of serum, but at seeding densities and for times that are likely to result in high density or confluence, with or without SF/HGF. They found no effect of SF/HGF on DNA synthesis. In the present study, the determination of thymidine incorporation was carried out in semi-confluent cultures in the presence of serum. It would seem that the mitogenicity of SF/HGF for MDCK cells is only apparent in the presence of serum factors, and that SF/HGF will not stimulate DNA synthesis in density-inhibited cultures.

The original observation prompting this study, where MDCK cell scattering was found on TCP but not on glass coverslips in the same dish, suggested differences in the adsorption of cell attachment factors from serum onto the different surfaces. The pattern of differences in colony morphology suggests that MDCK cell attachment and spreading is dominated by fibronectin on TCP, and by vitronectin on glass. Fibronectin (approx. 40 μg ml⁻¹) and vitronectin (200-400 μg ml⁻¹) are both present in serum (Underwood and Bennett, 1989), but in the presence of serum it has been shown that fibronectin, but not vitronectin, is competed off glass surfaces (Grinnell, 1986). It has been suggested that the avidity of vitronectin for glass is such that, in the presence of serum, glass surfaces are essentially vitronectin-coated (Edwards et al., 1987; Burridge et al., 1988). This is supported in the present study by the responses of MDCK cells to SF/HGF, which is identical on serum- and vitronectin-coated glass. SF/HGF will scatter MDCK cells cultured on glass when the surface is pre-coated with fibronectin (or laminin). Differences in the responses of the MDCK-fs strain suggest that in these cells a signalling pathway is constitutively activated such that the activity of SF/HGF can be replaced by substratum-initiated signals. In the absence of SF/HGF, MDCK-fs cells scattered on fibronectin-coated glass, but less so on serum- and vit-ronectin-coated glass (or fibronectin-coated TCP), which indicates that a combination of signals from both fibronectin and vitronectin receptors is sufficient to scatter cells of this strain.

One possible reason for the difference in scattering on fibronectin and vitronectin could be inherrent differences in the adhesion and motility of MDCK cells on these different surfaces. Preliminary observations of adhesion to these surfaces do indeed indicate that early attachment and spreading of MDCK cells is reduced on serum- and vitronectin-coated glass, compared to fibronectin, at the concentrations used for passive adsorption (unpublished observations), and therefore this may account for scattering differences. However, a number of observations suggest that this is not the case. After 24 hours of control culture, the point at which cultures are exposed to SF/HGF, cells and colonies are well spread on serum and vit-ronectin. Addition of SF/HGF-containing medium promotes the further spreading of the cells on all surfaces such that colony expansion is evident; cells on all surfaces appear spread to the same degree (unpublished observations). After 24 hours of exposure to SF/HGF, the margins of unscattered colonies are irregular in outline, suggesting motile activity. After a further 24 hours of culture in SF/HGF, many cells at colony margins have formed extremely extended protrusions, giving the colonies a bizarre, stellate appearance (this is also evident in colonies on fibronectin that have not scattered). This morphology is here interpreted as being indicative of ‘frustrated’ locomotion, i.e. attempted locomotion by marginal cells is thwarted by the persistence of cell-cell adhesion. These findings indicate that stimulation of MDCK cell motility and breakdown of cell-cell adhesion are separate events.

The effect of adhesiveness on scattering was directly tested by culturing the cells on surfaces of immobilised fibronectin or vitronectin. Although the degrees of initial attachment to these surfaces were essentially identical, significant SF/HGF-induced dissociation of colonies was not seen on vitronectin surfaces. Similarly, when higher concentrations of vitronectin (10 μg ml⁻¹) were used to non-covalently coat glass, initial cell attachment was considerably increased over that in the earlier experiments, but SF/HGF-induced colony dissociation was not enhanced (unpublished results).

The conversion by SF/HGF of phenotypically epithelial cells to a more mesenchymal phenotype, has previously been shown to be separable from the effects on cell growth. Certain cell types (including MDCK cells) are reported to scatter, but their growth is unaffected by SF/HGF; others are growth-stimulated but not scattered and, in yet others, both effects have been found (Hartmann et al., 1992; Tajima et al., 1992; Giordano et al., 1993). A naturally occurring SF/HGF splice varient and an artificially truncated form lacking the light chain both bind to the met receptor (which becomes autophosphorylated) and are reported to scatter MDCK cells but not stimulate hepatocyte mitogenesis (Hartmann et al., 1992). The present study has examined the effects of SF/HGF on the morphology and level of DNA synthesis of MDCK monolayer cultures, and how these parameters are influenced by cell-substratum adhesion. SF/HGF-induced scattering of MDCK colonies was here found to be dependent on the composition of the substratum. SF/HGF was found to enhance DNA synthesis of MDCKs growing in the presence of serum, independently of substratum composition. These findings may provide a means of clearly separating the scattering and growth-factor effects of SF/HGF in the same cell type. This obvious separation of growth stimulation from scattering, found in this study, provides further evidence of a divergence of signalling downstream from the met receptor, the branch leading to scattering requiring direct or indirect interaction with ECM receptors. Weidner et al. (1993)pointed out that this separation downstream of met has parallels in other growth factors; for example, mitogenesis has been shown to be stimulated by FGF receptors lacking a phosphorylation site required for the activation and recruitment of PLC gamma, i.e. PI turnover is abolished (Mohammadi et al., 1992; Peters et al., 1992).

The mechanism by which the substratum modulates SF/HGF activities is not clear. Two main possibilites exist: ‘outside-in’ signalling via ECM receptors may be required to provide upstream signals for the pathway(s) leading to cell-cell separation; or the ECM receptors themselves are downstream substrates whose ligand binding affinty is modulated by SF/HGF-induced intracellular signals, i.e. ‘inside-out’ signalling (for reviews see Damsky and Werb, 1992; Ginsberg et al., 1992; Hynes, 1992; Schwartz, 1992). It is probable that both mechanisms are involved, since both the breakdown of lateral contacts and the modulation of cell-substratum adhesion (integrins may be the substrates in this instance) are likely to be required to scatter epithelial cells. Signals generated as a result of integrin/ligand binding have been shown to act upstream of PLC gamma and be required for PDGF activity in anchorage-dependent cells (Schwartz, 1992). Conversely, integrin-mediated migration of endothelial cells triggers and requires distinct signalling pathways on vitronectin and collagen (Leavesly et al., 1993), suggesting that ECM receptors are both transducers of and substrates for motilitypromoting signals. Here, experiments in monolayer cultures using substrata of mixed ECM composition immobilised onto solid surfaces have helped to elucidate whether an inhibitory or stimulatory signal is acting in the breakdown of cell-cell contacts. SF/HGF-induced colony dissociation is marked on surfaces comprising a mixture of immobilised fibronectin and vitronectin, or fibronectin alone, but not on vitronectin, suggesting that fibronectin receptors provide positive signals for the dissociation of MDCK cell-cell contacts.

Potential downstream targets of scatter-inducing signalling pathways have recently been suggested by experiments showing that activation of a temperature-sensitive mutant of v-src expressed in MDCK cells, increases the phosphorylation of the E-cadherin/β-catenin complex, which correlates with the rapid breakdown (within 60 minutes of switch to the permissive temperature) of cell-cell contacts and the conversion of the cells to a fibroblastic, invasive phenotype (Behrens et al., 1993). In a report of preliminary work (Takeichi, 1993), both SF/HGF and EGF are said to result in the phosphorylation of β-catenin and plakoglobin, suggesting that both adherens and desmosomal juntions could be affected. Other potential intermediates include the small GTP-binding proteins, rac and rho, shown to be involved in modulating cytoskeletal organisation in fibroblasts (Ridley and Hall, 1992; Ridley et al., 1992). The role of c-fosproto-oncogene product has been shown to be important in epithelial-fibroblastic conversion. Activation of c-fos is sufficient to disrupt cell-cell contacts and convert cultured mammary epithelial cells (irreversibly, if activation is prolonged) to a fibroblastic phenotype of increased invasiveness (Reichmann et al., 1992). Cytoskeletal organisation is an obvious target in the process of scattering (Dowrick et al., 1993), but the role of SF/HGF in this is far from clear.

A report that appeared during the revision of this paper gives further significance to the data presented here. Santos and Nigam (1993)found that when MDCK cells were cultured in 3-D basement membrane gels (as opposed to type I collagen gels), SF/HGF-induced tubulogenesis was abolished. They examined the question of which components of basement membrane were responsible by culturing cells in type I collagen gels to which these various components were added individually. It was found that type IV collagen and heparan sulphate proteoglycan strongly inhibited tubulogenesis, whereas vitronectin or TGF-β only reduced SF/HGF-induced tubule formation to a small degree (approx. 10%) but allowed the formation of long, unbranched tubules as opposed to the highly branched tubules formed in their absence.

It has previously been suggested that tubulogenesis is, in 3-D culture, the equivalent of scatter in monolayer culture (Montesano et al., 1991b; Gumbiner, 1992). However, the results of the present study indicate that this may not be the case. Fig. 7summarises and compares the findings of Santos and Nigam (1993)with those of the present study. Although it is not possible to be certain that the activities of SF/HGF in monolayer cultures can be simply extrapolated to those in 3-D cultures, together these data suggest that the morphogenetic activity of SF/HGF is profoundly modulated the ECM, and provide some insight into possible cellular mechanisms involved in tissue remodelling during tubulogenesis. It is possible to form the hypothesis, on the basis of these comparisons, that tubule formation can result from increased motility of MDCK cells, but that branched tubular morphology requires breakdown of cell-cell contacts. These models of SF/HGF-induced morphogenetic changes are likely to provide a means of elucidating the mechanisms of some important cellular developmental processes.

Thanks are due to Dr Ermano Gherardi and his colleagues for the kind gift of purified recombinant SF/HGF, and to Yash Bhasin and Barry Crook for their excellent technical assistance. I also thank Sir Michael Stoker, Ermanno Gherardi, John Edwards and Jonathan Bennett for their helpful discussions and suggestions.