Influence of basement membrane molecules on directional migration of human breast cell lines in vitro

The basement membrane (BM) is an organized, continuous extracellular matrix structure that separates epithelia from their stroma. Its major constituents are laminin, collagen type IV, entactin/nidogen and heparan sulfate proteoglycans (Lee, 1988). The BM may be considered as a barrier that is crossed, for example, during tumor invasion and embryo gastrulation.

Multiple functions have been ascribed to laminin, which is the main BM component (Campbell and Terranova, 1988; Beck et al. 1990). It can serve, amongst other functions, as a substratum for cell attachment and spreading (Fligiel et al. 1985) and as a signal for cell migration (McCarthy et al. 1985). This cruciform glycoprotein consists of one A (∼400×10³M_r) and two B (∼200×10³M_r) subunits (Timpl et al. 1979). Different molecular domains of laminin have binding affinities for collagen type IV, heparin, entactin, other laminin molecules or laminin receptors (Beck et al. 1990).

Various cell surface receptors for laminin have been postulated, e.g. on human breast carcinoma cells (67xlO³Af_r, Terranova et al. 1983), on platelets (VLA-6, Sonnenberg et al. 1988) and on neuronal cells (Kleinman et al. 1988). Some authors stated that different types of laminin receptors might interact with different domains of the laminin molecule (Goodman et al. 1987).

The BM can be altered in pathological situations. During invasion, for example, when malignant tumor cells cross the BM, it often appears fragmented and disorganized (Lee, 1988). Experiments with laminin receptorbinding peptides (Iwamoto et al. 1987), laminin-cell binding (Albini et al. 1989) and cell surface laminin expression (Varani et al. 1983) indicate that laminin is involved in invasion and metastasis (Hunt, 1989). Some authors (Bracke et al. 1986; Zimmerman and Keller, 1987) advanced the hypothesis that the extracellular matrix contains a molecule that can arrest cell migration and therefore acts as a stop signal. Malignant tumor cells would no longer be receptive for or responsive to that signal.

Migration is one of the activities of invasive cells (Sträuli and Haemmerli, 1984; Zimmerman and Keller, 1987; Mareel et al. 1990) that can be analyzed in simple in vitro systems. In order to test the effect of the BM on the migration of human breast cell lines, we examined the directional migration (‘translocative motility’) of these cells onto BM components coated on a glass substratum. We demonstrated that Matrigel™ (reconstituted BM) and its main component laminin arrested migration of one MCF-7 cell line, while two other MCF-7 variants and one HBL-100 cell line were unaffected. Moreover, within the laminin molecule, the Pl fragment particularly seemed to be able to cause this arrest.

The MCF-7 cell line was originally derived from a pleural effusion caused by a human breast adenocarcinoma (Soule et al. 1973). We used three variants of this cell line.

The MCF-7/AZ cell line was kindly provided by P. Briand (The Fibiger Institute, Copenhagen, Denmark). The MCF-7 ras TD5 cell line is a v-Ha-ras-1 MCF-7/AZ transfectant that has been selected for tumor formation in vivo and that expresses the v-Ha-ras oncogene product (Van Roy et al. 1990). The MCF-7/6 cell line was kindly provided by H. Rochefort (Unité d’Endocrinologie Cellulaire et Moléculaire, Montpellier, France).

HBL-100, kindly provided by N. Busso (Centre Médical Universitaire, Genève, Switzerland), is a human cell line derived from the milk of an apparently healthy lactating woman (Gaffney, 1982) and contains a tandemly integrated SV40 genome (Caron de Fromentel et al. 1985).

All cell lines were maintained in 25 cm² tissue culture plastic vessels (Nunc, Roskilde, Denmark). Culture media were: (1) For the MCF-7/AZ and MCF-7 ras TD5 cell lines: Eagle’s Minimal Essential Medium, containing Earle’s salta and non-essential amino acids (EMEM, Flow Laboratories, Irvine, Scotland) supplemented with 5% (v/v) fetal bovine serum, 0.05% (w/v) L-glutamine, 250 i.u. penicillin ml⁻¹ and 6 ng bovine insulin ml⁻¹. (2) For HBL-100 cells: Dulbecco’s modification of Eagle’s Medium (DMEM, Flow Laboratories) supplemented with 10% fetal bovine serum, 0.05% (w/v) L-glutamine, 250 i.u. penicillin ml⁻¹, 100 μg streptomycin ml’⁵ and ImM sodium pyruvate. (3) For MCF-7/6 cells: a 1/1 (v/v) mixture of Haro F12 (Gibco Europe, Gent, Belgium) and DMEM (Flow Laboratories) containing 10% fetal bovine serum, 0.05% (w/v) L-glutamine and 250 i.u. penicillin ml⁻¹.

Since the MCF-7/AZ and the MCF-7/6 cell lines were obtained from different laboratories, we checked their human breast origin and several MCF-7 characteristics using routine procedures. Both MCF-7/AZ and MCF-7/6 cells had estrogen and progesterone receptors, illustrating their breast origin. The respective estrogen and progesterone receptor concentrations m MCF-7/AZ cells were 47.7 and 67 finol mg⁻¹ protein. MCF-7/6 cells contained 26.1 frnol estrogen and 25 finol progesterone receptors per mg of protein. Conditioned medium of both cell lines contained a single human LDH (lactate dehydrogenase) isoenzyme band characteristic of only a few breast tumor cell lines including MCF-7 (Horan Hand et al. 1983). Immunocytochemical detection of intermediate filaments of cells in culture showed that both cell lines were cytokeratin positive and vimentin negative, a combination rarely found in metastatic human epithelial cells from pleural fluids, which are normally positive for both cytokeratin and vimentin (Ramaekers et al. 1983). In addition, both cell lines revealed a positive punctate signal after immunocytochemical staining with a set of four monoclonal antibodies prepared against MCF-7 membrane antigens (Plessers et al. 1986).

Single cell suspensions were prepared by dissociation of monolayer cell cultures with trypsin/EDTA (0.05%/0.02% in modified Puck’s saline A solution; w/w/v) (Gibco Europe). Spheroidal cell aggregates were prepared by shaking cell suspensions (6×10⁵ cells in 6 ml complete culture medium) in 50 ml Erlenmeyer flasks during 3 days on a Gyrotory® shaker (New Brunswick Scientific Company Inc., New Brunswick, NJ) at 70 revs min⁻¹.

AS BM components we used either reconstituted BM (Matrigel™) or the purified BM molecules, laminin, collagen type IV and fibronectin or laminin fragments.

Mouse laminin was purified from Engelbreth-Holm-Swarm mouse sarcoma (EHS) as described earlier (Timpl et al. 1979). E8 and Pl laminin fragments were prepared by limited enzymatic digestion of laminin-nidogen (Fig. 1, lane 1) or laminin (Fig. 1, lane 2) with, respectively, elastase or pepsin as described by Lissitzky et al. (1989). The pepsin Pl fragment contains the intersection of the three laminin subunits, the elastase E8 fragment consists of the terminal part of the long arm of the laminin molecule. 3% to 12% gradient SDS–polyacrylamide electrophoresis, under reducing or non-reducing conditions (Laemmli, 1970), was performed to evaluate the purity of these laminin fragments (Fig. 1). Under non-reducing conditions it revealed a Pl fragment with an apparent molecular weight of about 290×10³ (Fig. 1, lane 3) and an E8 fragment consisting of two major bands of about 80 and 140×10³M_r (Fig. 1, lane 4) comparable to those described earlier (Rohde et al. 1980; Goodman et al. 1987). Human plasma fibronectin and human placental collagen type IV were purchased from Sigma (St Louis, MO). Matrigel™, an EHS preparation containing laminin, collagen type IV, heparan sulfate proteoglycans and entactin/nidogen was obtained from Collaborative Research (Bedford, MA).

Coating of glass coverslips with BM molecules was done in accordance with the Newgreen (1984) procedure, which was slightly modified. This author also determined the optimal coating concentrations for glass. Glass coverslips (35 mm×9 mm×l mm) (Vel, Leuven, Belgium) were washed in 95% ethanol (v/v) and allowed to air dry before use. BM molecules were diluted in PBS at 80 μg ml⁻¹ (laminin and fibronectin), 20 μg ml⁻¹ (E8 and Pl fragments of laminin) or 1 mg ml⁻¹ (collagen type IV).

To make lanes, 7 drops of 1 μl of a BM molecule solution were pipetted along the longitudinal axis of a coverslip, at about 5 mm from one another. These drops were immediately spread out by means of small, rectangular coverslips (9mmx2mmxl mm). The molecules were allowed to adsorb to the glass overnight, at 37 °C in a humid atmosphere containing 5% CO₂ in air. Unadsorbed molecules were removed by washing twice in PBS. Coated coverslips were never allowed to dry out completely during the procedure. PBS coatings were used as a control. The positions of the BM molecule lanes could be revealed only by immunostaining.

Just before use, Matrigel™ was diluted in the appropriate complete culture medium (1/1, v/v) at 4°C to prevent gelation. Matrigel™ lanes were directly pipetted (7×2 μl) onto the coverslip. Gelation was allowed by a 3 h incubation at 37 °C in a humid atmosphere containing 5% CO₂ in air. A practical advantage of Matrigel™ over the above mentioned components is that its lanes can be seen as granular zones on the underlying substratum, so that no immunocytochemical staining is needed for its localization.

This migration assay was performed as described earlier (Storme and Mareel, 1980), with minor modifications. Spheroidal cell aggregates with established cell-cell contacts (De Bruyne et al. 1988) were used because they reflected the situation in vino where cell-cell as cell-substratum interactions also are involved in migration and because their migration front on glass substratum could be easily localized and followed. Aggregates with a diameter of about 0.2 mm were selected under a Macroscope (x50, Wild, Heerbrugg, Switzerland), pipetted onto the coverslip in a minimum volume of complete culture medium (7 aggregates per 120 μl per 315 mm²) and randomly (without knowing the positions of the lanes) distributed with a sterile needle along the longitudinal axis of the coverslip. They were then allowed to attach and spread on the glass overnight, at 37 °C, in a humid atmosphere containing 5 or 10% CO₂ in air. After washing once in culture medium, the coverslips were transferred to Leighton tubes with 1.2 ml of the appropriate complete culture medium. Cells were then allowed to migrate out of the aggregate during the following 7 days. Together with directional migration, cell proliferation also contributed to the increase in diameter of the area covered by the cells. Nevertheless, earlier experiments demonstrated that cells from aggregates treated with proliferation inhibitors (5-fluorouracil or 500 rad ionizing radiation) were still able to migrate and microcinephotography demonstrated that active cell movements were involved (Storme and Mareel, 1980).

At 7 days after seeding, the number of aggregates arrested by the BM component lanes relative to the number of aggregates that met a lane during migration were evaluated after fixation and immunostaining (see below).

Immunocytochemical staining allowed us to localize the BM molecule lanes and the position of the cells with respect to them. Therefore, cultures were washed in 0.1M sodium cacodylate buffer (pH 7.4) and fixed for 10 min in 1% (w/v) glutaraldehyde in the same buffer. Remaining aldehyde groups were inactivated for 30min with 0.5mgml^-1 sodium borohydride. The cultures were then rinsed in 0.1% (w/v) bovine serum albumin (BSA) in phosphate-buffered saline (PBS) without calcium and magnesium (BSA-buffer). Aspecific binding of the antisera was reduced by quenching for 20 min with 5% (v/v) normal goat serum (Amersham, Brussels, Belgium) (for anti-laminin and anticollagen type IV) or 5% (w/v) BSA (for anti-fibronectin) in BSA- buffer. Primary antisera were all raised in rabbits: anti-EHS laminin (1/200; kindly provided by M. Havenith, Rijksuniversit- eit Limburg, Maastricht, The Netherlands), anti-human fibronectin (1/400; Cappel-Wortington, Cooper Biomedical, Malvern, PA) and anti-collagen type IV (1/200; kindly provided by M. Havenith). All antisera were incubated overnight. The polyclonal anti-laminin antiserum also stained the Pl and E8 laminin fragments. We used gold-labeled goat anti-rabbit serum (Auroprobe LM GAR, Amersham) (1/40; 2 h) as a secondary antiserum. Immunogold positivity was visualized with the silver enhancement technique (IntenSE M, Amersham) as described by the manufacturer. Observations were made by a combination of light microscopy (phase-contrast, for visualization of the cells) and epiillumination with polarized light (reflected by the silver particles). For control stainings, the entire procedure was followed, but the primary antiserum was omitted,

For the dynamic observation of cell motility and migration, time-lapse video recordings were made with a Zeiss Inverted Microscope (Oberkochen, FRG) equipped with a camera (National Panasonic, WV-1850C), an Eos animation control unit (Barry, UK), a home-made time-pulse generator and a Sony VO-58508 U-Matic videocassette recorder. The glass coverslips with migrating cells were kept in a tissue culture plastic Petri dish (5.2 mm diameter) (Nunc) with 10 ml of the appropriate complete culture medium. The microscope was set up in a thermostatically controlled room, at 37 °C with 5 or 10% CO? in air. All recordings were made on cells migrating out of an aggregate and localized on the border between a coated and an uncoated zone on the coverslip. Two images per 20 s were taken (250 times acceleration when projected at a rate of 25 images s^-1).

After seeding, cell aggregates attached mostly to an uncoated zone on the substratum. After spreading of an aggregate, cells migrated out in a symmetrical and radial way (Tig. 2A). When meeting a coated lane, cell migration could either be arrested (e.g. by laminin, Fig. 2B) or unarrested. Occasionally, temporal arrest during a timespan of less than one day was observed after which cells migrated imperturbably on the lanes. These situations were also scored as unarrested. Cells migrating from an aggregate met a coated lane mostly within 2-4 days. Fig. 3 shows the data and Fig. 4 the illustrations of how Matrigel™ and different BM molecules affected cell migration.

When meeting Matrigel™ lanes, migration of MCF-7/AZ cells was completely arrested, while migration of all other cell lines was permitted. As Matrigel™ is a crude extract, we wondered whether one of the BM constituents alone could mimic its effect on migration. Fibronectin and collagen type IV did not affect migration of any of the cell lines tested. Laminin, however, mimicked Matrigel™ as it completely arrested the migration of MCF-7/AZ cells while it permitted migration of MCF-7/6, MCF-7 ras TD5 and HBL-100 cells. Also lower laminin coating concentrations (down to 10 μg ml⁻¹) were able to stop MCF-7/AZ cell migration (data not shown). Unlike MCF-7/AZ cells, arrest by laminin was never seen for MCF-7/6 or HBL-100 cells and rarely for MCF-7 ras TD5 cells. To exclude the possibility that the medium composition of MCF-7/AZ and MCF-7/6 cells might be responsible for their differences in arrest by laminin, medium exchange experiments were performed: three out of three MCF-7/AZ cultures were arrested by laminin and three out of three MCF-7/6 cultures were unarrested by laminin. Unlike E8, Pl laminin fragments arrested migration of MCF-7/AZ cells. Nevertheless, its efficiency of arrest was much lower compared to native laminin. None of these fragments arrested MCF-7/6 cells. Most of the MCF-7 ras TD5 aggregates revealed no arrest by Pl or E8 fragments, but as with laminin some background arrest was seen for both fragments. None of the tested cell lines were arrested by PBS lanes (data not shown).

Time-lapse video recordings showed that active migration was involved in the increase in diameter of the area covered by the cells. Although migration of MCF-7/AZ cells was inhibited by laminin or Matrigel™, the arrested cells still proliferated and showed an intensive ruffling activity on these coatings. The ruffles on, e.g., laminin started as broad slabs (Fig. 5A), shaded off into long, narrow ‘tongue-like’ ruffles (Fig. 5B and C), which then retracted into the perikaryon (Fig. 5D). This extension and retraction was continuously repeated. Also in the few situations where MCF-7 ras TD5 cells were arrested by laminin or Pl, such a ruffling pattern was observed. In contrast, all MCF-7 cell lines that were unarrested by laminin or Matrigel™ or that migrated on uncoated glass revealed continuously moving broad ruffles (illustrations not shown).

Our main finding was that Matrigel™ lanes permanently arrested the migration of MCF-7/AZ cells, while they permitted migration of MCF-7/6, MCF-7 ras TD5 and HBL-100 cells. Of several BM components, only laminin but not fibronectin or collagen type IV mimicked this effect. However, we have to take into account the fact that Matrigel™ does not only physically differ from these soluble components, but also contains other molecules such as heparan sulfate proteoglycan and entactin. MCF- 7/AZ cell migration was also arrested by Pl, but not E8, laminin fragments. This occurred, however, with a lower efficiency compared to the native laminin molecule.

Depending on their concentration, extracellular matrix molecules may have opposite effects on cell migration. Perris and Johansson (1987) found that dispersion of amphibian neural crest cells on laminin substrata was optimal at lower concentrations. In our assay, however, laminin concentrations of 10 μg ml⁻¹, eight times lower than standard, were still effective in arresting migration of MCF-7/AZ cells. Arrest of migration was not due to toxic effects or inhibition of cell proliferation by the laminin coating. Arguments for that were provided by time-lapse observations, which revealed a continuous membrane ruffling of these arrested cells onto the border of the laminin substratum and continuing cell division. MCF-7/AZ cells displayed in general a low affinity for laminin. Not only was their migration arrested by laminin lanes, but MCF-7/AZ cell suspensions revealed also a low attachment efficiency to laminin and formed cell-cell clusters instead of spreading on this substratum (our unpublished observations).

Unlike fibronectin and collagen type IV, laminin was also reported to inhibit migration of NIH 3T3 mouse fibroblasts (Goodman and Newgreen, 1985), human epidermal kératinocytes (Woodley et al. 1988) and endothelial cells (Kirkpatrick et al. 1996). In addition, Perris and Johanssen (1987) suggested that neural crest cell migration in the embryo occurs preferentially along fibronectin-rich paths, while laminin-abundant areas correspond to sites of arrest for moving cells. Taken together, these observations support the postulations that laminin is the main BM component affecting migration (McCarthy et al. 1985) and that it can function as a stop signal for cell migration (Bracke et al. 1986; Zimmermann and Keller, 1987).

Albini et al. (1986) reported that transfection with a v-Ha-ras oncogene resulted in an increased interaction (attachment, migration, chemoinvasion) of MCF-7 cells with laminin. Our experiments are in agreement with these data. In contrast with the parental MCF-7/AZ cells, migration of MCF-7 ras TD5 cells was unarrested by Matrigel™ and mostly unarrested by laminin or its fragments. However, not only these v-Ha-ras transfected cells, but also the untransfected MCF-7/6 and HBL-100 cells were unarrested by laminin lanes. MCF-7 ras TD5 cells were not only unarrested by laminin lanes, but their cell suspensions also attached with a significantly higher efficiency to laminin than the parental MCF-7/AZ cells (our unpublished observations). Albini et al. (1986) stated that an increased interaction of v-Ha-ras-transfected cells with laminin was due to a higher number of 67×10³M_r laminin receptors.

A suitable molecular explanation for the diverse migratory response of the different cell lines to laminin and its fragments is still lacking. As suggested already for the MCF-7 ras TD5 cells, an involvement of laminin receptors is one of the possibilities. Goodman et al. (1987) proposed that cells can even have different laminin receptors, each interacting with a distinct laminin domain. Indeed, the VLA-6 (α_Bβ₁) integrin was identified as an E8 receptor (Sonnenberg et al. 1990), whereas the 67×10³M_r receptor was reported to bind laminin as well as its Pl fragment with high affinity (similar K_D values) (Terranova et al. 1983; Barsky et al. 1984). The observation that the Pl fragment, like native laminin, was able to arrest MCF-7/AZ cell migration is an argument in favor of the involvement of a Pl-binding receptor. However, 67×10³M_r receptors were detected on MCF-7/AZ as well as MCF-7/6 cells (unpublished results) and this can therefore not serve as an explanation for their having different migration to laminin and Pl. The involvement of other putative Pl receptors, such as the α_vβ₃ vitronectin receptor as suggested by Sonnenberg et al. (1990), is however not excluded.

A second plausible explanation for migration on laminin is its internalization (Wewer et al. 1987) or degradation (Stephens et al. 1989) by the migrating cells. We observed internalization of gold-labeled laminin, but this occurred by MCF-7/AZ as well as HBL-100 cells (unpublished results). MCF-7 cells were found to produce and secrete the urokinase plasminogen activator (Butler et al. 1983). Plasmin, the cleaved substrate of urokinase, is able to degrade laminin (Liotta et al. 1981). In that context, Corree et al. (1990) reported that MCF-7/6 cells displayed fivefold more plasmin receptors on their surface than MCF-7/AZ cells. This could mean that, by exposing this protease on their cell surface, they focally degrade the laminin substratum (Stephens et al. 1989).

MCF-7/AZ cells, the migration of which is arrested by laminin and Matrigel™, were found to be non-invasive in embryonic chick heart fragments in vitro (De Bruyne et al. 1988). In contrast, the invasive MCF-7/6 (Bracke, M., Van Larebeke, N. A., Vyncke, B. M. and Mareel, M.: Invasion and plasma membrane ruffling of MCF-7 human breast carcinoma cells in vitro: opposite effect of retinoic acid on invasive and non-invasive cell variants; unpublished data) and invasive HBL-100 cells (De Bruyne et al. 1988) were not arrested by laminin and Matrigel™. Taken into account these observations, laminin may play a role in stopping invasion of the basement membrane.

In conclusion, Matrigel™ lanes can arrest MCF-7 cell migration in vitro. This activity was retained by its main component, laminin, and by the Pl laminin fragment. So, within the BM barrier, laminin seems to function as a stop signal for cell migration. A signal to which some cell types have become insensitive.

P.C. is a recipient of a fellowship from the I.W.O.N.L., Brussels, Belgium. F.V.R. is a Senior Research Associate of the NFWO, Brussels, Belgium. This work was supported by a grant from the ‘Sportvereniging tegen de Kanker’, Brussels, Belgium and the ‘A.S.L.K. Kankerfonds’ (36.1131.88), Brussels, Belgium. J. De Boever (Department of Gynecology, University Hospital, Gent, Belgium) is greatly thanked for determining the estrogen and progesterone receptor content of the MCF-7 cells. The authors also thank L. Baeke, B. Buysse, R. Colman, A. Verspeelt and B. Vyncke for the technical assistance, N. Van Larebeke and the “Fondation Philippe et Thérèse Lefevre’ for access to the videoequipment facilities and J. Reels van Kerckvoorde for preparing the illustrations.