Cell-associated proteases affect tumour cell migration in vitro

Several studies have shown that tumour cells have a higher level of protease activity associated with them than do normal cells (Unkeless et al. 1973; Ossowski et al. 1973a; Bosmann, Lockwood & Morgan, 1974; Goldberg, 1974a; Chen & Buchanan, 1975; Hatcher et al. 1976; Hatcher et al. 1977). A number of cell properties including morphology (Ossowski, Quigley, Kellerman & Reich, 1973b), growth rate and saturation density (Burger, 1970; Chen & Buchanan, 1976), motility (Ossowski et al. 1973b;Ossowski, Quigley & Reich, 1975; Chen & Buchanan, 1976), and agglutin-ability (Goldberg, 19746) can be altered by addition of exogenous proteases or protease inhibitors to cultures of normal and malignant cells and it has been hypothesized that phenotypic differences between normal and malignant cells may be due, in part, to differences in the level of protease activity associated with the cells (Poste & Weiss, 1976).

In a previous study we compared the migration of malignant mouse fibrosarcoma cells in both foetal calf serum and normal human serum. The migration of the cells was significantly depressed in human serum (Varani, Orr & Ward, 1978 a). In attempting to delineate the mechanism by which human serum suppressed tumour cell migration we observed that the malignant fibrosarcoma cells produced a high level of protease activity and that human serum had low levels of protease inhibitor activity relative to foetal calf serum (Varani, Orr & Ward, 19786). In light of these findings we decided to examine the role of cell-associated proteases in the regulation of cell migration. In this report we present evidence from 2 lines of inquiry which indicates that in vitro migration of the fibrosarcoma cells is affected by cell-associated proteases.

The malignant fibrosarcoma cells used in this study were derived from a tumour induced by 3-methylcholanthrene in a C57 bl/6 mouse. The cells had been cultivated in vitro for approximately 10 months prior to the start of the present studies. The cells were maintained in medium 199 (M199) supplemented with 10 % non-heat-inactivated foetal calf serum, 100 U/ml of penicillin and 100 μg/ml of streptomycin. The cells were grown at 37 °C in 5 % CO₂ in air. The generation time of the cells in log phase was approximately 20-22 h and the saturation density was greater than 1000 cells per mm². A variant cell population with altered levels of protease activity and altered migratory characteristics was established from the parent line as described below. After establishment these cells were grown under the same conditions as the parent cells.

The assessment of cell migration was done using a modification of methods described by Carpenter (1973) and by Harrington & Stastny (1973). Cells were removed from monolayers by trypsinization, washed 3 times and centrifuged into a pellet. A typical cell pellet consisted of approximately 5 × 10⁶ cells in a volume of 50 μI. To this was added 150 μl of M199 containing 10 % serum and 0·2 % (w/v) Seaplaque agarose (Marine Colloids, Inc., Rockland, Maine). One or two microlitre-sized drops of the cell suspension were delivered to the wells of a microtitre dish (Linboro Scientific Co., New Haven, Connecticut). The microtitre dish was then placed in a refrigerator for 10 min to allow the agarose to solidify. After cooling, the agarose droplets were covered with 200 μl of chilled M199 containing 10 % serum. The overlay medium was added gently so as not to disturb the agarose drops. Tests were performed in quadruplicate. The microtitre dishes were incubated in 5 % CO₂ in air at 37 °C. Migration of the cells was measured after 18 h by phase-contrast microscopy using an inverted tissue-culture microscope. The microscope contained a calibrated grid in the eyepiece. The width of grid space represented 100 μm actual distance at a magnification of 100 ×. The distance from the edge of the agarose drop to the leading edge of migrating cells was determined on 4 sides of each droplet. Since tests were done in quadruplicate, 16 readings were obtained for each measurement. The migration of cells out of the agarose drop produced a uniform, expanding corona of cells around the agarose drop. While the majority of the cells migrated out of the agarose drop in the corona a few cells migrated much farther out than the major population of cells. The number of cells that migrated beyond the corona of cells was counted. A detailed description of the migratory characteristics of these cells and of non-malignant fibroblasts has been recently published (Varani et al. 1978 a).

Modified Boyden chambers were used to assess migration based on a method described by Romualdez & Ward (1975) for chemotaxis. Cells were added to the top half of the chamber at a concentration of 5 × 10⁵ cells per chamber in M199 with 10% serum. M199 with serum was added to the bottom half of the chamber. The chambers were incubated for 4 h at 37 °C after which time the membranes were fixed and cell migration assessed in the normal manner. The number of cells that migrated into the filters in 5 high-power fields was counted. Selectron filters (Schleicher and Schuell; Keene, New Hampshire) of i2-/tm porosity were used. Tests were performed in triplicate.

The foetal calf serum used in this study was obtained from 3 different lots of serum purchased from Grand Island Biological Company (Grand Island, New York). There was no detectable variation in serum from the 3 lots. Normal human serum was obtained from 9 healthy donors. Little variation was seen among the sera. The foetal calf serum and normal human serum were not heat-inactivated prior to use in this study.

Four inhibitors of protease activity were used in these studies. They included e-amino-n-caproic acid (EACA), soybean trypsin inhibitor (SBTI) (Type I-S), lima bean trypsin inhibitor (LBTI) (Type II-L) and bovine pancreas trypsin inhibitor (BPTI) (Type I-P). All inhibitors were purchased from Sigma Chemical Company (St. Louis, Missouri).

Cell-associated, plasminogen-mediated protease activity was measured using the procedure described by Goldberg (1974 a) in which casein serves as the substrate. In addition, various enzyme activities were measured in tumour cell extracts. The extracts were prepared with Triton X-100 detergent (Bosmann et al. 1974) and protein concentrations determined by the Lowry method (Lowry, Rosebrough, Farr & Randall, 1951). Protease activity was measured using the ester substrate, N-acetyl-DL-phenylalanine-β-naphthyl ester, a substrate specific for chymotrypsin-like enzyme as previously reported (Becker & Ward, 1969). Other enzymes measured included β-glucuronidase (Musa, Doe & Seal, 1965) and glycosidase (Bosmann et al. 1974).

Serum was depleted of plasminogen by affinity chromatography on lysine-linked Sepharose 4B (Deutsch & Mertz, 1970). The lysine-linked Sepharose 4B was obtained from the Pharmacia Company (Piscataway, New Jersey). After depletion, the amount of residual plasminogen remaining in the serum was determined after conversion to plasmin with urokinase (Wulf & Mertz, 1969).

Fibrosarcoma cells were harvested from culture, washed 3 times and resuspended in agarose-containing medium for migration assays as described in the Methods section. In one group the cells were resuspended in medium containing 10% foetal calf serum; in the other group the cells were resuspended in medium containing 10% normal human serum. The migration of the fibrosarcoma cells was influenced by the type of serum present. The distance from the edge of the agarose drop to the leading edge of the corona of the malignant cells was reduced by about 60% in normal human serum as compared to foetal calf serum. An even greater reduction was seen in human serum when the number of cells beyond the corona was counted (Table 1). To confirm the observations made in the agarose assay, the migration of fibrosarcoma cells in the 2 sera was compared using the Boyden chamber assay. Cells were harvested from culture, washed 3 times and added to the top well of the chamber in M199 with either 10% human serum or 10% foetal calf serum. The bottom well contained the same serum that was in the top well. The results obtained with the Boyden chamber assay were similar to the results in the agarose assay (Table 1). There was, again, a marked reduction in the migration of fibrosarcoma cells in normal human serum.

The difference in migratory activity of the tumour cells in foetal calf serum and normal human serum was reflected in dramatic differences in growth characteristics of these cells in the 2 sera. When fibrosarcoma cells were grown in foetal calf serum (where migration was optimal) the cells did not form distinct colonies; rather, they grew as individual separated cells. When these cells were grown in human serum (where migration was depressed) they grew as distinct colonies. There was little evidence of cells spreading beyond the edges of the colonies and few individual cells were observed. These findings are shown in Figs. 1 and 2.

In a previous study we found that human serum had less inhibitory activity than foetal calf serum when tested against the fibrosarcoma cell-associated protease as well as against trypsin and plasmin (Varani et al. 19786). If differences in antiprotease activity accounted for the difference between foetal calf serum and normal human serum in supporting tumour cell migration, it seemed possible to increase the migration of the cells in human serum by the addition of protease inhibitors. Four inhibitors were added to M199 with 10% normal human serum and migration studies carried out (Table 2). Three of the inhibitors, SBTI, LBTI and BPTI, all enhanced migration of the cells in human serum (Table 2). The enhancement was dose-dependent; the minimum concentration at which enhancement occurred was 100–200 μ g/ml while the optimum concentration was 2·0–2·5 mg/ml of each drug. The distance migrated by the cells in the corona in the presence of optimal concentrations of each inhibitor was approximately twice the distance migrated by cells in control M199 with normal human serum and no inhibitor. This approached the migration rate for cells in medium with 10% foetal calf serum (Table 2). The number of cells that migrated beyond the corona and the number of cells that migrated into the filters in the presence of protease inhibitors was also increased over the baseline number obtained in normal human serum without inhibitors.

In contrast to these results, the fourth inhibitor, EACA, did not enhance migration but rather inhibited migration of the tumour cells in normal human serum (Table 2). Inhibition was dose-dependent and occurred over the range 1–20 mg/ml.

We also looked at the effects of these same inhibitors on the migration of the tumour cells in foetal calf serum (data not shown). The 3 trypsin inhibitors did not further enhance migration in foetal calf serum when added to the medium in the concentrations that enhanced migration in human serum. At higher concentrations (5 ⁿⁱg/^m0 slight inhibition of migration occurred. EACA, on the other hand, inhibited migration in foetal calf serum over the same range of concentrations that inhibited migration in normal human serum.

In addition to the use of protease inhibitors we also looked at the ability of plasminogen-depleted human serum to support migration. Plasminogen was removed from serum using lysine-coupled Sepharose 4B as described in the Methods section. By measuring the amount of plasminogen activatable with urokinase before and after treatment we found that more than 99% of the plasminogen was removed from the serum. The depleted serum was then compared to untreated human serum for its ability to support migration (Table 3). It can be seen that depletion of the plasminogen from human serum did not affect its ability to support fibrosarcoma cell migration.

In contrast to this, when the plasminogen was depleted from foetal calf serum in the same way, there was a slight reduction (about 33%) in the ability of the serum to support migration. The addition of purified plasminogen (obtained from Sigma Chemical Company) to the plasminogen-depleted foetal calf serum reversed the depressed migration.

Fibrosarcoma cells from the parent population were seeded at 5 × 10⁴ cells/flask into several 25-cm² flasks containing M199 with 10% foetal calf serum and 5 mg/ml SBTI. Most of the cells attached in this medium but most did not divide and monolayers were not formed. However, distinct colonies of cells were observed growing in several of the flasks. These flasks were sub-cultured in the normal manner in M199 without SBTI. When the cells in each subculture reached the monolayer stage they were again subcultured in M199 without SBTI. Extracts were prepared from the cells in one of the replicate flasks of each group and tested for various enzyme activities as described in the Methods section. Extracts prepared from the parent fibrosarcoma cells were used as controls in all assays. Extracts prepared from one culture had elevated levels (relative to the parent cells) of activity when measured against the substrate N-acetyl-DL-phenylalanine-β-naphthyl ester at pH 7·4. The other cultures showed no elevation of this activity. Cultures of cells with the elevated esterase activity were subsequently subcultured approximately every 3 to 4 days at a 1:3 split ratio. Over the course of a 3-month period the cells were passaged 25 times. During this time they were assayed for activity at passages 8, 16, and 24 in addition to the original assay which occurred on second-passage cells (Table 4). It can be seen that the levels of activity remained elevated during this period. The activity of the cell extracts measured at each passage relative to the activity of extracts prepared from the parent cells on the same day varied from 129 ± 1% to 239 ± 9%. Cell extracts containing 100 μg of protein had activity equivalent to 1–10 μg of crystallized trypsin (2 × crystallized, Sigma Chemical Company). Extracts from the variant cell population were also compared to the parent population for activity at pH 3·4 and 4·7 as well as for glucuronidase and glycosidase activities. No significant differences were found between these cells and the parent cells with regard to any of these activities. Finally, cells from the variant population were compared with the parent cells for plasminogen-mediated caseinolytic activity. The variant cell population caused more extensive and more rapid clearing of the casein than the parent cells although differences between populations were difficult to quantitate due to the nature of the assay.

In addition to differences measured by the esterase assay and the caseinolytic assay the variant cells were also morphologically different from the parent line. The variant cells were more elongated and tended to grow in whorls more typical of fibroblasts than of the tumour cells. The growth rate of the variant cells was slightly slower and the saturation density was about 20% lower than the parent cells.

The variant cells and the parent cells were compared for migratory activity using both the agarose assay and the Boyden chamber assay in foetal calf serum (Table 5). By any of the 3 criteria used the variant cells migrated less actively than the parent line. A comparison between the 2 lines could not be carried out in normal human serum. When the variant cells were put into agarose drops containing human serum the drops did not adhere uniformly to the plastic surface. Some drops detached completely while only sporadic migration of cells occurred in the drops that did adhere. This type of response is what we observed with the parent cells when trypsin was incorporated into the overlay solution.

In a previous study we showed that human serum did not support the migration of turnout cells as well as foetal calf serum (Varani et al. 1978 a). The tumour cells used in this study were fibrosarcoma cells obtained from a chemically induced tumour in a C57 bl/6 mouse. We subsequently showed that these cells had a high level of protease activity as measured by caseinolysis and that human serum had less antiprotease activity than foetal calf serum. These cells did not attach to plastic flasks when.plated as single cells in medium with human serum but when protease inhibitors with antitrypsin activity were added to the medium the defect was overcome (Varani et al. 19786). Because of these observations we decided to look at the role of cell-associated proteases and the effect of protease inhibitors on the migration of these cells in human serum.

Two assay systems, the agarose drop assay and the Boyden chamber assay, were used to assess migration in the present study. The two were used in conjunction because each has certain advantages and disadvantages and the maximum amount of data can be obtained by using both. As in the previous study it was again found that the tumour cells did not migrate as well in human serum as in foetal calf serum. When the trypsin inhibitors, SBTI, LBTI or BPTI were added to medium with human serum, migration was enhanced to levels comparable to that in foetal calf serum. When EACA was added to human serum, however, migration was not enhanced but was dramatically inhibited. Thus it was possible to influence the migratory activity of the cells either positively or negatively depending on which inhibitor was used. The 3 inhibitors which enhanced migration had been shown previously to have antitrypsin activity and to facilitate the attachment of the fibrosarcoma cells to plastic flasks in medium with human serum. EACA was shown not to have antitrypsin activity and did not facilitate cell attachment (Varani et al. 1978b).

The mechanism by which attachment and migration are inhibited is not clear. The inhibition is probably not due to the activation of plasminogen because removal of the plasminogen from human serum did not enhance migration and the addition of EACA, which blocks the activation of plasminogen (Ambrus, Ambrus, Lassman & Mink, 1968; Okamoto, Oshiba, Mihara & Okamoto, 1968; Varani et al. 1978b), also did not enhance migration. Another possibility is that activation of complement by the tumour cells leads to inhibition of migration. Heat inactivation of human serum did, at least partially, restore migration (Varani et al. 1978 a). However, we have not been able to duplicate the effects of heat inactivation by the addition of goat anti-C3 anti-C5 to human serum (unpublished observation) and more work will have to be done before this point can be clarified.

By whatever mechanism inhibition occurs, it seems likely that the inhibition of attachment and the inhibition of migration are related. Cells must be able to attach to the substratum in order to pull themselves along. It has been postulated, in fact, that the abnormal migratory responses of tumour cells in culture is primarily due to abnormalities in their attachment capabilities (Vasiliev & Gelfand, 1975). In our studies we found that treatments which overcome the defect in attachment also enhanced migration while treatments which did not overcome this defect did not enhance migration. On the other hand the relationship between migration and attachment may be complex. Although the cells did not attach when plated as single cells in human serum they did attach when added to the plates in agarose drops and did attach to the surface of micropore membrane. Perhaps a subtle defect in attachment is all that is necessary to prevent optimal migration.

In addition to the relationship between attachment and migration, other factors must also be involved in the regulation of cell migration. EACA, for example, had no effect on attachment yet dramatically inhibited migration in both normal human serum and foetal calf serum. Furthermore, although we did not further inhibit the ability of human serum to support migration by removing the plasminogen, we were able to reduce the migration of the cells in foetal calf serum by plasminogen depletion. These results are in agreement with previous studies showing a role for plasminogen in the in vitro migration of cells in foetal calf serum (Ossowski et al. 1973b, 1975). The discrepancy between our findings in human serum and our findings and the findings of others in foetal calf serum indicates that the regulation of motility is complex and that multiple factors are involved.

In addition to the studies with the trypsin inhibitors the results we obtained with the variant cell population of fibrosarcoma cells also supports the idea that a cell-associated protease contributes to the reduced migration of the cells in human serum. This population of cells was isolated from the parent cells after an initial passage of the parent cells in medium containing SBTI. Whether or not this selection procedure will prove fruitful in the isolation of additional variant populations is unknown since only one of the many flasks seeded grew cells with altered, stable characteristics. In any event, a population of cells was isolated which had less migratory activity than the parent cells in foetal calf serum and virtually no migration in human serum. In addition, these cells had increased levels of esterase activity when tested against a substrate specific for chymotrypsin-like enzymes (Cohen & Erlanger, 1960). These cells also showed more caseinolytic activity than the parent cells although the difference was impossible to quantitate due to the nature of the assay (Goldberg, 1974a). These cells did not show altered β-glucuronidase or glycosidase activity when compared to the parent cells. It is hoped that several additional variant populations with altered migratory activity can be isolated in order to clarify further the regulation of motility.

The difference in migratory behaviour of the parent cells in normal human serum and foetal calf serum was associated with altered growth characteristics in the 2 sera. In human serum the cells grew as distinct colonies while in foetal calf serum the cells did not form distinct colonies but grew as individual, separated cells. What relation the in vitro growth and migratory characteristics have to the in vivo properties of growth, invasion and metastasis remains to be seen. If the in vitro and in vivo characteristics are related it may be possible to alter the in vivo characteristics by the use of selective inhibitors. We have studies in progress to determine if this is possible. Perhaps more exciting is the fact that now that we have been able to isolate cells with altered migration ability and altered enzyme levels from a single tumour cell population, we can compare the population for differences in their in vivo biological behaviour.

This study was supported in part by grant number AI-09651 from the National Institutes of Health and by a grant from the Connecticut Research Foundation. W. Orr is a fellow of R. Samuel McLaughlin Foundation (Toronto, Canada).