Cell-surface cytokeratin 8 is the major plasminogen receptor on breast cancer cells and is required for the accelerated activation of cell-associated plasminogen by tissue-type plasminogen activator.

Cytokeratin 8 (CK 8) has been identified on the external surfaces of viable, unpermeabilized epithelial cells (Hembrough, T. A., Vasudevan, J., Allietta, M. M., Glass, W. F., and Gonias, S. L. (1995) J. Cell Sci. 108, 1071-1082). In this study, we demonstrated that CK 8 is the major plasminogen-binding protein in plasma membrane fractions isolated from three breast cancer cell lines, BT20, MCF-7, and MDA-MB-157. To assess the function of CK 8 as a plasminogen receptor, monoclonal antibody 1E8 was raised against the carboxyl-terminal 12 amino acids of CK 8. The 1E8 epitope was present on the external surfaces of breast cancer cells, as determined by immunofluorescence microscopy. 125I-1E8 bound to MCF-7 cells; the maximum binding capacity (1.5 × 106 sites per cell) was comparable with that determined for plasminogen. When MCF-7 cells were incubated with Fab fragments of 1E8, specific 125I-plasminogen binding was decreased up to 82%. Specific plasminogen binding was decreased up to 67%, even when the unbound 1E8 Fab was removed by washing the cells prior to adding 125I-plasminogen. Preincubation with 1E8 Fab decreased plasminogen binding to BT20 and MDA-MB-157 cells, although to a lesser extent than with MCF-7 cells. Plasminogen activation by tissue-type plasminogen activator was greatly accelerated, due to a large decrease in Km, when the plasminogen was bound to MCF-7 cells. Pretreatment with 1E8 Fab decreased the rate of plasminogen activation by up to 83%, implicating CK 8 in the MCF-7 cell-accelerated reaction. These studies identify cell-surface CK 8 as a major plasminogen receptor in breast cancer cells and as a required component for the rapid activation of cell-associated plasminogen by tissue-type plasminogen activator.

Cancer invasion and metastasis are promoted by proteinases that are activated on the tumor cell surface and in the pericellular spaces. These proteinases digest basement membrane and extracellular matrix components, allowing tumor cell penetration through tissue (1). In breast cancer, cellular expression of the serine proteinase, urokinase plasminogen activator (u-PA), 1 has been correlated with an aggressive, malignant phenotype (2,3). This may reflect the ability of cell-associated u-PA to activate plasminogen that, in turn, digests glycoprotein components of the extracellular matrix and may activate other proteinases (4,5). A second plasminogen activator, tissue-type plasminogen activator (t-PA), also binds to cell surfaces and activates plasminogen (6). t-PA may substitute for u-PA in promoting tumor cell invasion (7).
The cellular receptor for u-PA has been cloned and characterized (8,9). uPA receptor is a glycosylphosphatidylinositolanchored membrane protein that binds u-PA with moderately high affinity (K D ϳ1.0 nM). Cellular binding sites for t-PA and plasminogen have also been identified; however, the biochemical nature of these sites remains more vague. Breast cancer cells in vitro bind plasminogen in a specific and saturable manner (10,11). As is typical of most plasminogen/cellular interactions, the binding affinity is modest (K D ϳ1.0 M); however, the binding capacity is high (B max ϳ10 6 sites per cell). Occupancy of breast cancer cell plasminogen-binding sites in vivo probably depends on the concentration of plasminogen in the surrounding tissue, which may be quite high since the plasma concentration of plasminogen is about 1-2 M.
Plasminogen is activated more readily by u-PA and t-PA when both enzyme and substrate are bound to the surfaces of certain cell types (6,(12)(13)(14). Plasmin that is cell-associated reacts very slowly with ␣ 2 -antiplasmin and ␣ 2 -macroglobulin, which are rapid inhibitors of solution-phase plasmin (15,16). Thus, the identification and characterization of major cellular binding sites for plasminogen is important for understanding the role of plasmin in tumor cell invasion and other processes that require cell-surface proteolytic activity.
Plasminogen binding to cellular receptors and other macromolecules, including fibrin and specific extracellular matrix proteins, is mediated by its kringle domains, three of which (K1, K4, and K5) express affinity for lysine (17)(18)(19). The highest affinity interaction usually involves K1 and proteins with carboxyl-terminal lysines (19,20). Several plasminogen receptors have been identified in eukaryotic cells. Annexin II, which was identified as an endothelial cell plasminogen receptor, lacks a carboxyl-terminal lysine in its native form and thus requires proteolytic modification by Lys-specific proteinases in order to acquire plasminogen-binding activity (21,22). ␣-Enolase, which was identified as a plasminogen receptor in U937 monocytoid cells, contains a carboxyl-terminal lysine in its intact form (20). Both ␣-enolase and annexin II have been demonstrated on the external surfaces of cultured cells, using antibody-based technologies (20,21); however, the level of expression of cell-surface ␣-enolase is sufficient to account for only about 10% of the plasminogen binding capacity of U937 cells (23). Cytokeratin 8 (CK 8) is an intermediate filament protein that polymerizes with CK 18 to form a component of the cytoskeleton in simple epithelia and many epithelial cell-derived neoplasms. CK 8 interacts extensively with the internal leaflet of the plasma membrane and is, therefore, one of the major components recovered in the plasma membrane fraction when epithelial cells are subjected to subcellular fractionation procedures (10). CK 8 is unique among cytokeratins in that the intact structure includes a carboxyl-terminal lysine. In a recent study, we demonstrated that CK 8 is the primary plasminogenbinding protein in the plasma membrane fraction isolated from hepatocytes (10). We then demonstrated, by immunoelectron microscopy and immunofluorescence microscopy, that CK 8 is present on the external surfaces of intact, unpermeabilized hepatocytes, hepatocellular carcinoma cells, and breast cancer cells, confirming previous studies (24,25) and identifying CK 8 as a candidate plasminogen receptor.
The purpose of the present study was to determine (i) whether cell-surface CK 8 binds plasminogen; (ii) whether CK 8 contributes substantially to the total plasminogen binding capacity of breast cancer cells; and (iii) whether cell-associated CK 8 promotes plasminogen activation. To accomplish these goals, we prepared a monoclonal antibody (1E8) specific for the carboxyl-terminal 12 amino acids of intact CK 8, which was considered the most likely plasminogen-binding site within the CK 8 structure. We then prepared highly purified Fab fragments of 1E8. This was an important step since the primary structure of all mouse and human IgG isotypes includes carboxyl-terminal Lys residues (26), which, unless exposed to plasma carboxypeptidases, bind plasminogen in solution (27,28). The results presented here demonstrate that CK 8 is the major plasminogen-binding protein in the plasma membrane fraction of breast cancer cells, the primary protein responsible for the majority of the specific binding of plasminogen to intact breast cancer cells in culture, and a required cellular component for the accelerated activation of cell-associated plasminogen by t-PA.
Proteins and Antibodies-[Glu 1 ]Plasminogen was purified from human plasma by the method of Deutsch and Mertz (29). Single-chain tissue-type plasminogen activator (t-PA) was purchased from American Diagnostica. Monoclonal antibody PCK-26, which recognizes CK 5, CK 6, and CK 8, was from Sigma. Of these cytokeratins, only CK 8 is expressed by breast cancer cells (30).
A polypeptide corresponding to the 12 carboxyl-terminal amino acids of human CK 8, preceded by an amino-terminal cysteine, CKLVSESS-DVLPK, was synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry on a Gilson AMS 422 multi-peptide synthesizer and purified by high performance liquid chromatography on a Poros flow-through column (PerSeptive Biosystems, MA). The peptide was cross-linked via its amino-terminal cysteine to maleimide-activated keyhole limpet hemocyanin (Pierce) and used to immunize A/J mice (Jackson Labs, ME). Antibody titers were determined using the same peptide cross-linked to BSA. The mouse with the highest titer was given a final intrasplenic booster. Two days later, spleen cells were harvested and fused with SP2/0 mouse myeloma cells. These cells were selected in HAT-containing medium and cloned.
Eighteen clones generated antibodies that specifically bound CK 8. These antibodies were further screened for ability to detect CK 8 by immunofluorescence microscopy and Western blot analysis. One clone (1E8) was selected for subcloning. Ascites was generated by expanding the subclone of 1E8 (1E8-C6-B4) in CAF1 mice (Jackson Labs, ME), and antibody was purified by Protein A chromatography. Antibody 1E8 isotyped as an IgG 2b .
Generation of 1E8 Fab Fragments-Fab fragments of antibody 1E8 were generated using the Immunopure Fab Preparation Kit (Pierce). Briefly, 5 mg of antibody were incubated with 1 ml of papain-agarose for 2 h at 37°C. Digested antibody was separated from papain-agarose and subjected to protein A-Sepharose chromatography. The flow-through was collected in 1-ml fractions and analyzed by SDS-PAGE. Fractions containing Fab fragments were pooled, aliquoted, and stored at Ϫ20°C until needed. Intact antibody did not contaminate the final Fab preparations, as determined by Coomassie staining of gels that contained up to 20 g of purified Fab per lane. For some experiments, 1E8 Fab preparations were further processed by incubation with CPB-Sepharose for 12 h at 25°C, to remove any residual carboxyl-terminal lysine residues (as might have been present due to trace contamination with Fc domain). The CPB-Sepharose was prepared by coupling 5 mg of CPB/1 g of CNBr-activated Sepharose.
Cell Lines and Tissue Culture-All cell lines were obtained from the ATCC (Rockville, MD). The breast cancer cell lines, MCF-7 and MDA-MB-157, and the fibroblast cell line, AKR-2B, were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine, 100 units/ml penicillin, and 100 mg/ml streptomycin. BT20 breast cancer cells were cultured in RPMI 1640, supplemented as above. Cells were typically passaged by trypsin/EDTA treatment and gentle scraping. In order to minimize the effects of trypsin on cell-surface proteins, cultures were maintained for at least 2 days after passaging before performing experiments.
Subcellular Fractionation-Confluent monolayers of breast cancer cells (3-6 ϫ 10 7 cells) were harvested by gentle scraping into 5 mM EDTA and pelleted at 200 ϫ g. Whole cell extracts (W fraction) were obtained by Dounce homogenization in the presence of 10 M aprotinin, 1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA, 10 M leupeptin, and 0.2 g/ml E-64. The W fraction was centrifuged at 250 ϫ g for 20 min in an Eppendorf microcentrifuge. The resulting pellet consisted of nuclei and unhomogenized cells (N fraction). The supernatant was centrifuged at 1250 ϫ g for 20 min. The resulting pellet was enriched in plasma membrane but partially contaminated with mitochondria, whereas the supernatant consisted of cytoplasm and organelles, including mitochondria (C fraction). The plasma membrane-enriched pellet was resuspended in 1.45 M sucrose (400 l) and overlaid with 0.25 M sucrose in a 500-l airfuge tube. The step gradient was centrifuged at 55,000 ϫ g in an airfuge for 45 min. Enriched plasma membranes (P fraction) were collected from the interface of the two sucrose solutions (31).
SDS-PAGE, Western Blotting, and Plasminogen Overlay Assays-Subcellular fractions from breast cancer cells were subjected to SDS-PAGE and electrotransferred to polyvinylidene difluoride membranes (Millipore) using a Hoefer Transphor apparatus (2 h, 0.5 amp). In some studies, the transferred proteins were stained with 0.2% (w/v) Coomassie Blue R-250 (Bio-Rad). In plasminogen overlay experiments, the membranes were blocked with 5% nonfat dried milk, rinsed twice with 20 mM sodium phosphate, 150 mM NaCl, 0.1% (v/v) Tween 20, pH 7.4 (PBS-T), and then incubated with 125 I-plasminogen (10 nM) and aprotinin (10 M), in the presence and absence of 10 mM ⑀ACA or a 50-fold molar excess of non-radiolabeled plasminogen. To determine radioligand binding, the blots were rinsed three times for 15 min in PBS-T, dried, and analyzed using a PhosphorImager. For Western blot analyses, blocked membranes were incubated sequentially with a 1:500 dilution of CK 8-specific antibody (PCK-26 or 1E8), horseradish peroxidaselabeled goat anti-mouse monoclonal antibody, and diaminobenzidine. 125 I-1E8 Binding to MCF-7 Cells-Antibody 1E8 was radioiodinated using IODO-BEADS (Pierce). The specific activity was 1-2 Ci/g. Confluent MCF-7 cells were subcultured into 48-well plates and grown for 48 h. Increasing concentrations of 125 I-1E8 (0.02-1.0 M), in Earle's balanced salt solution, 10 mM HEPES, 10 mg/ml BSA, pH 7.4 (EHB), were incubated with the cells for 2 h at 4°C. The cells were then rinsed three times with EHB and once with PBS. Cell-associated radioactivity was recovered in 1.0% SDS, 0.1 M NaOH and measured in a ␥-counter. For concentrations of 125 I-1E8 up to 0.2 M, nonspecific binding was determined by competition with a 50-fold molar excess of non-radiola-beled 1E8. Nonspecific binding was plotted as a function of 125 I-1E8 concentration and extrapolated to higher concentrations by linear regression. Specific 125 I-1E8 binding was then determined, at each concentration, by subtracting the nonspecific binding derived from the standard curve. Specific binding isotherms were fit to the equation for a rectangular hyperbola by nonlinear regression and analyzed by Scatchard transformation. 125 I-1E8 Binding to Purified CK 8 -Rat hepatocyte CK 8 was purified by the method of Achtstaetter et al. (32). In the final step of this procedure, CK 8 is eluted from DEAE-Sephacel in a buffer that contains 8.0 M urea. The purified CK 8, at a concentration of 15 g/ml, was diluted 1:3 with PBS and incubated in 48-well tissue culture plates for 1.5 h at 37°C. The plates were rinsed three times and blocked with 10 mg/ml BSA for 1.5 h at 37°C. 125 I-1E8 binding was studied using the method described for the cell cultures in the previous section.

I-Plasminogen Binding to Breast Cancer Cells in the Presence of
1E8 Fab-In order to determine the contribution of cell-surface CK 8 to the total plasminogen binding capacity of breast cancer cells, competition binding experiments were performed. 1E8 Fab (0.5-8.0 M) was preincubated with MCF-7 cells for 2 h at 4°C. In some experiments, the cultures were washed three times with EHB to ensure that all unbound 1E8 Fab was removed. In other studies, the 1E8 Fab was allowed to remain in the cultures. 125 I-Plasminogen (0.2 M) was then added in the presence and absence of 10 mM ⑀ACA or a 50-fold molar excess of unlabeled plasminogen, for 1 h at 4°C. Unlabeled plasminogen and ⑀ACA are equally effective at displacing 125 I-plasminogen from specific cellular binding sites (12,13). At the conclusion of an incubation, the cultures were washed three times. Cell-associated radioactivity was recovered in SDS/NaOH and measured in a ␥-counter. For the protocol in which unbound 1E8 Fab was not removed by washing prior to adding 125 I-plasminogen, the 1E8 Fab was always pretreated with CPB-Sepharose. In control experiments, the effects of monoclonal antibody 1D7 Fab on 125 I-plasminogen binding to MCF-7 cells was studied. 1D7 is specific for human ␣ 2 -macroglobulin. 125 I-Plasminogen binding was also studied in cultures of BT20 cells, MDA-MB-157 cells, and AKR-2B fibroblasts, with and without prior 1E8 Fab treatment. Fibroblasts do not express CK 8 (this was confirmed by Western blotting with PCK-26 and 1E8) and thus cannot bind 1E8 Fab in a specific manner.
Plasminogen Activation by t-PA-Increasing concentrations of plasminogen (0.05-1.0 M) were incubated with MCF-7 cells or in wells without cells for 1 h at 22°C. The MCF-7 cell cultures were washed so that only cell-associated plasminogen remained. t-PA (2 nM) and the plasmin-specific fluorescent substrate, VLK-AMC (0.5 mM), were added simultaneously to the MCF-7 cell cultures and to the wells without cells. Plasminogen activity was detected by continuous monitoring (30-s intervals) of fluorescence emission using a Cytofluor 2350 fluorescent plate reader (Millipore). The excitation wavelength was 380 nm, and the emission wavelength was 480 nm (5-nm slit widths). Curves of fluorescence against time were transformed using a first derivative function so that the resulting plots showed the concentration of active plasmin against time. Amounts of plasminogen, which bound to the MCF-7 cells, were determined separately, by performing equivalent incubations with 125 I-plasminogen. The velocity of VLK-AMC hydrolysis (0.5 mM) by MCF-7 cell-associated plasmin was decreased by 45 Ϯ 3% (n ϭ 4) compared with solution-phase plasmin; the presented graphs were not corrected for this difference in substrate hydrolysis rate. No VLK-AMC hydrolysis was observed when either plasminogen or t-PA were omitted from the reaction.
To determine whether CK 8 affects the kinetics of plasminogen activation by t-PA on the cell surface, MCF-7 cells were preincubated for 1 h at 22°C with increasing concentrations of 1E8 Fab (1.0 -8.0 M) or with vehicle. After washing the cultures, 1.0 M plasminogen was added for 1 h. After washing the cells again, t-PA (2 nM) and VLK-AMC (0.5 mM) were added simultaneously. Plasminogen activation was determined by monitoring fluorescence emission.
Immunofluorescence Microscopy-BT20 and MCF-7 cells were passaged onto 30-mm glass coverslips and grown for at least 48 h. The cells were washed with EHB buffer and incubated with purified 1E8 at 1/200 dilution for 2 h at 4°C. The cells were then washed three times with EHB, incubated with Texas Red-labeled goat anti-mouse IgG (1/1000 dilution) for 2 h at 4°C, rinsed three times again, and fixed in ice-cold paraformaldehyde. Cellular immunofluorescence was imaged using an Olympus 3H2 Microscope. In control experiments, mouse nonimmune IgG was substituted for primary antibody or primary antibody was omitted.

RESULTS
Characterization of Monoclonal Antibody 1E8 -Whole cell extracts from BT20 breast cancer cells were subjected to SDS-PAGE and Western blot analysis using the newly generated monoclonal antibody, 1E8. Only a single protein immunoreacted with 1E8. This protein was CK 8, as determined by its molecular mass (55 kDa) and immunoreactivity with antibody PCK-26 (Fig. 1). In control experiments, identical amounts of protein from whole cell extracts of AKR-2B cells, which lack CK 8, were subjected to Western blot analysis with antibody 1E8. No immunoreactivity was observed (results not shown).
We previously performed indirect immunofluorescence microscopy experiments, using antibodies AB 6.01 and M20, which recognize CK 8, and demonstrated the presence of CK 8 or a CK 8-like protein on the external surfaces of intact, unpermeabilized breast cancer cells (10). To determine whether the carboxyl terminus of CK 8, which we considered the most likely plasminogen-binding site, was exposed on the surfaces of breast cancer cells, immunofluorescence microscopy studies were performed with antibody 1E8. Fig. 2 shows that antibody 1E8 bound to live, nonpermeablized BT20 cells, forming a diffuse and punctate pattern, identical to that seen with other CK8-reactive antibodies (PCK-26, AB 6.01) (10). All of the cells in each preparation showed positive immunofluorescence, irrespective of whether they were in clusters or isolated. When primary antibody was omitted or when mouse nonimmune IgG was substituted for antibody 1E8, no immunostaining was observed. Equivalent results were obtained with MCF-7 cells (results not shown); however, AKR-2B fibroblasts were immunonegative with antibody 1E8, as expected (data not shown).
Plasminogen Overlay Assays-In our previous study (10), we performed plasminogen overlay assays and demonstrated that CK 8 is the major plasminogen-binding protein in the P fraction (plasma membranes) of rat hepatocytes. Fig. 3 shows that similar results were obtained when three separate breast cancer cell lines were analyzed. In the P fraction from each cell line, the major plasminogen-binding protein had an apparent mass of 55 kDa, identical to that of CK 8. The MDA-MB-157 cell P fraction showed a second, plasminogen-binding species, with an apparent mass of 35 kDa. This 35-kDa band was also present as a minor component in the P fraction from the MCF-7 cells. 125 I-Plasminogen binding to the 55-and 35-kDa bands was completely inhibited when ⑀ACA or excess unlabeled plasminogen was added (results not shown).
In the whole cell (W) extract from all three breast cancer cell lines, the 55-kDa band was the major plasminogen-binding FIG. 1. CK 8 specificity of antibody 1E8. BT20 whole cell lysate (100 g) was subjected to Western blot analysis using the well characterized CK 8-specific antibody, PCK-26, and the newly generated antibody, 1E8. Molecular mass standards (in-kDa) are shown. component. The prominent appearance of the 55-kDa band in the W fractions distinguished the breast cancer cells from hepatocytes; in the hepatocyte W fraction, CK 8 was a minor plasminogen-binding component (10). The 55-kDa band was excluded from the breast cancer cell cytoplasmic fractions (C), as would be expected for CK 8. To confirm that the 55-kDa band observed in the plasminogen overlay assays was CK 8, the identical BT20 subcellular fractions were subjected to Western blot analysis with antibody PCK-26. The antibody detected large amounts of CK 8 in both the W and P fractions, while a greatly reduced level of antigen was detected in the C fraction (Fig. 4). The mobility of the major band detected in the Western blot analysis coincided exactly with the mobility of the 55-kDa band in the plasminogen overlay assays.
Specific Binding of Antibody 1E8 to MCF-7 Cells-Radioiodinated antibody 1E8 bound to MCF-7 cells in a specific, saturable manner. A representative binding isotherm is shown in Fig. 5. An unexpected finding in this and two other equivalent studies was that the results transformed into apparently linear Scatchard plots. While not conclusive, this result suggests that each antibody engages only a single CK 8 epitope. The K D for 1E8 binding to MCF-7 cells was 0.4 Ϯ 0.1 M and the B max was 1.5 Ϯ 0.5 ϫ 10 6 sites per cell (n ϭ 3). The B max for 1E8 is comparable with that determined for plasminogen binding to the same cell type (10).
In control experiments, binding of intact 125 I-1E8 to purified, immobilized CK 8 was studied. 125 I-1E8 bound specifically and saturably to the CK 8, whereas no binding was observed in BSA-coated wells in two separate experiments. The data transformed into linear Scatchard plots (r Ͼ 0.95) and a K D of 0.24 M was derived (results not shown).
1E8 Fab Fragments Block Plasminogen Binding to Breast Cancer Cell Lines-The contribution of cell-surface CK 8 to the plasminogen binding capacity of breast cancer cells was studied using 1E8 Fab and MCF-7 cells, as a representative cell line. When 1E8 Fab was coincubated with 125 I-plasminogen in the MCF-7 cell cultures, specific radioligand binding was de-creasedandtheextentofthedecreasewas1E8Fabconcentrationdependent (Fig. 6). With the highest concentration of 1E8 Fab (8 M), specific plasminogen binding was decreased by 82%.
In separate experiments, MCF-7 cells were preincubated with 1E8 Fab and then washed to remove unbound Fab prior to adding 125 I-plasminogen. It was anticipated that this protocol would decrease the level of observed competition since 1E8 Fab dissociation from the cell surface would be favored while the 125 I-plasminogen is present in the cultures. However, as shown in Fig. 6, significant inhibition of specific 125 I-plasminogen binding was still observed. In cells that were pretreated with 8 M 1E8 Fab, 125 I-plasminogen binding was decreased by 67%.
In control experiments, no decrease in 125 I-plasminogen binding was observed when AKR-2B fibroblasts were pretreated with 1E8 Fab. Antibody 1D7 Fab had no effect on 125 I-plasminogen binding to MCF-7 cells, irrespective of whether the Fab was present during the incubation with plasminogen or washed-out prior to adding plasminogen (results not shown). The ability of 1E8 Fab to inhibit plasminogen binding to MCF-7 cells, even after the free Fab was removed by washing, confirms that the observed competition is due to blocking of plasminogen-binding sites on the cell surface and not a solution-phase interaction with 125 I-plasminogen.
Binding of 125 I-plasminogen to BT20 and MDA-MB-157 cells was studied after pretreating the cultures with 4 M 1E8 Fab. The cells were washed prior to adding 125 I-plasminogen to remove unbound 1E8 Fab. As shown in Table I, specific 125 Iplasminogen binding to the BT20 cells was decreased, and the magnitude of the effect was only slightly less than that observed with MCF-7 cells. Decreased plasminogen binding was also observed with the MDA-MB-157 cells; however, the effectiveness of 1E8 Fab was not as great in this cell line. In control experiments, we studied the effects of 1E8 Fab on plasminogen binding to purified, immobilized CK 8. 1E8 Fab (4 M) was incubated with the CK 8. The wells were then washed before adding 125 I-plasminogen. In two separate experiments, specific 125 I-plasminogen binding was decreased by 62 and 68%. Thus, the extent of competition observed in the BT20 and MCF-7 cell cultures was similar to that detected using the identical protocol, in a purified system that includes CK 8 as the exclusive plasminogen-binding protein.
MCF-7 Cells Enhance Plasminogen Activation by t-PA-To study plasminogen activation by t-PA, increasing concentrations of plasminogen were added to MCF-7 cell cultures and to wells without cells. The cell cultures were washed, so that only cell-associated plasminogen remained (less than 3% of the plasminogen originally added to the wells). The wells without cells were not washed. t-PA and VLK-AMC were then added simul-taneously to each well. In the MCF-7 cell cultures, the velocities of plasminogen activation showed an initial lag phase lasting about 2 min, followed by a 5-7-min period when the velocities optimized and the first derivative functions (dRFU/ dt) linearized. RFU indicates relative fluorescence units. The period between 2.5 and 7 min was used to derive the plasminogen activation rates shown in Fig. 7 (panel A).
In wells without cells, plasminogen activation proceeded without an apparent lag phase. Rates of plasminogen activation (determined in the 2.5-7-min time interval) were slightly decreased compared with the rates measured in cell cultures that were initially incubated with the identical concentrations of plasminogen. However, since the wells without cells were not washed before adding t-PA, the amount of available plasminogen was actually about 100-fold greater. Fig. 7, panel B, shows Lineweaver-Burk transformations of the activation data. In the transformations, the actual amount of available substrate (plasminogen) is plotted as opposed to the amount initially added to the wells. Levels of cell-associated plasminogen are converted into units of concentration by dividing the number of moles of bound plasminogen by the total incubation volume (100 l). The most remarkable finding was a 2000-fold reduction in the K m for the activation of cell-associated plasminogen (2 nM) compared with plasminogen in solution (4 M). Although the K m for plasminogen activation in solution is similar to previously reported constants (33,34), this value should be interpreted cautiously since the plasminogen concentration range was selected to match the cell culture studies and is substantially lower than that required for optimal accuracy. The V max for plasminogen activation by t-PA was decreased about 3-fold when the plasminogen was cell-associated.
In order to study the role of cell-surface CK 8 in promoting the activation of plasminogen by t-PA, MCF-7 cells were pretreated with increasing concentrations of 1E8 Fab. The cells were then washed and incubated with plasminogen. After a final wash, t-PA and VLK-AMC were added simultaneously, and fluorescence was monitored. In this assay, some 1E8 Fab probably dissociated from the cell surfaces during the 1-h incubation with plasminogen at 22°C. Nevertheless, as shown in Fig. 8, preincubation with 1E8 Fab caused a concentration-dependent decrease in the ability of the cell cultures to promote plasminogen activation. After incubation with 8 M 1E8 Fab, the velocity of plasminogen activation was decreased by 83 Ϯ 6%.

DISCUSSION
Numerous studies have demonstrated that cell-associated plasmin and plasminogen activators promote tumor cell invasion in in vitro model systems (1,35,36). An increasing number of studies support the function of these same proteinases and their cellular receptors in tumor invasion and metastasis in vivo. For example, Mueller et al. (37) demonstrated that melanoma cells, when transfected to overexpress plasminogen activator inhibitor 2, form subcutaneous tumors that do not metastasize in scid mice, unlike the highly metastatic parent cell line. Evans and Lin (38) blocked implantation of intravenously FIG. 6. 1E8 Fab inhibits 125 I-plasminogen binding to MCF-7 cells. 1E8 Fab was preincubated with MCF-7 or AKR-2B cells for 2 h at 4°C. In the first protocol, 125 I-plasminogen was added to the MCF-7 cell cultures without first washing the cells so that unbound 1E8 Fab remained (q). In the second protocol, the MCF-7 cells were washed to remove unbound 1E8 Fab prior to adding 125 I-plasminogen (E). The AKR-2B cells were washed after incubation with 1E8 Fab and prior to adding the 125 I-plasminogen (Ⅺ). Specific plasminogen binding is expressed as a percentage of that observed without 1E8 Fab treatment. The presented results, for the MCF-7 cells, show the mean Ϯ S.E. for five separate experiments. administered breast cancer cells, in the lungs of rats, by coadministering recombinant plasminogen activator inhibitor 2. Finally, Kobayashi et al. (39) demonstrated that urinary trypsin inhibitor, which neutralizes cell-associated plasmin, also inhibits cancer cell invasion through Matrigel in vitro and blocks formation of lung metastases by Lewis lung carcinoma cells in vivo. By contrast, the potent solution-phase plasmin inhibitors, ␣ 2 -antiplasmin and ␣ 2 -macroglobulin, which are ineffective against cell-associated plasmin, show no activity in the same cancer cell invasion assays (39). These studies emphasize the importance of identifying macromolecules responsible for plasminogen binding in cancer. Our initial studies with rat hepatocytes identified CK 8 as a potential plasminogen receptor in epithelial cells (10). Since prior evidence for the presence of CK 8 on the outer surfaces of cells had been disputed, immunofluorescence and immunoelectron microscopy studies were performed. The fluorescence microscopy studies of hepatocytes and breast cancer cell lines showed that CK 8 is distributed uniformly on the surfaces of viable, adherent cells and not restricted to a fraction of each population, such as injured cells (10). We confirmed that the cell preparations were, in fact, unpermeabilized and that the CK 8-specific antibodies recognized cell-surface CK 8 and not intracytoplasmic CK 8 by electron microscopy. However, since the epitopes recognized by commercially available CK 8-specific antibodies are uncharacterized, it was not possible to conclude that the region of CK 8 responsible for plasminogen binding is exposed on the cell surface. Thus, monoclonal antibody 1E8 was raised utilizing synthetic peptide technology. The similarity in the distribution of 1E8 immunofluorescence, compared with previously studied CK 8-specific antibodies (10), suggests that the carboxyl terminus of CK 8 is intact and available for protein-binding interactions on breast cancer cell surfaces. Furthermore, our binding studies with 125 I-1E8 provide an estimate of the amount of CK 8 on the outer surfaces of MCF-7 cells (1-2 million sites per cell). The density of CK 8 on the MCF-7 cell surface is sufficient to account for the high cellular binding capacity for plasminogen (10).
Competition binding experiments with 125 I-plasminogen and 1E8 were performed using two separate methods. The methods were designed to ensure that activity resulted from antibody binding to cell-surface CK 8 and not interaction with 125 Iplasminogen in solution. In early studies, we routinely screened antibody preparations for plasminogen-binding activity in solution by testing the ability of each antibody to inhibit 125 I-plasminogen binding to Lys-Sepharose. Various CK 8-specific monoclonal antibodies (1E8, PCK-26, AB 6.01, M20) bound plasminogen in solution, as determined by the Lys-Sepharose competition assay. 1E8 Fab preparations demonstrated little or no plasminogen binding activity, and after treatment with CPB-Sepharose, were entirely inactive. The variable amount of residual plasminogen binding activity in Fab preparations, without CPB-Sepharose treatment, may have reflected tracelevel contamination with Fc domain or perhaps proteolytic modification of the Fab itself. When cell cultures were pretreated with 1E8 Fab and then washed to remove free 1E8 Fab, before adding 125 I-plasminogen, any residual plasminogen binding activity that may have been associated with the Fab could only have increased the level of observed plasminogen binding to the cells. Thus, the 1E8 Fab "wash-out" protocol provided the most rigorous demonstration of receptor competition even though this method probably provided a minimum estimate of the contribution of CK 8 to the plasminogen binding capacity, due to 1E8 Fab dissociation from the cell surface during the incubation with plasminogen.
Of the three breast cancer cell lines studied, the greatest effects of 1E8 Fab on plasminogen binding were observed with MCF-7 cells. When CPB-Sepharose-treated 1E8 Fab was retained in the culture medium during the incubation with 125 Iplasminogen, binding was inhibited by up to 82%. Plasminogen binding was inhibited by 60 -70% when the free 1E8 Fab was removed prior to adding 125 I-plasminogen. The difference between these two values suggests that 1E8 Fab may have partially dissociated from the cell surface after Fab washout but that substantial levels of cell-associated 1E8 Fab remained. In our studies comparing cell lines, 10% less competition was observed with BT20 cells and 25% less competition was observed with MDA-MB-157 cells, compared with the MCF-7 cells; however, at least with the BT20 cells, CK 8 still accounted for more than one-half of the plasminogen-binding sites. Since the Fab washout protocol was used in the experiments with MDA-MB-157 cells, it is quite possible that CK 8 accounts for more than 50% of the plasminogen-binding sites in this cell line as well. Thus, we conclude that CK 8 is a major plasminogenbinding protein in various breast cancer cell lines and may be was added for 1 h. After washing again, t-PA and VLK-AMC were added, and fluorescence emission was monitored (mean Ϯ S.E., n ϭ 3). The relative plasminogen activation rate is plotted as a percentage of that observed with cells that were not 1E8 Fab-treated. responsible for the great majority of the plasminogen-binding sites in some.
To demonstrate maximum inhibition of 125 I-plasminogen binding, we used high concentrations of 1E8 Fab; however, we still may not have entirely saturated available CK 8 binding sites for plasminogen. If the affinity of 1E8 Fab and intact antibody for CK 8 are equivalent (ϳ0.4 M), then at equilibrium 1E8 Fab at 4 and 8 M would be expected to occupy 91 and 95% of the available epitopes, respectively. If the affinity of 1E8 Fab for cell-surface CK 8 was reduced compared with intact antibody, then the degree of saturation would be less. In any case, the unoccupied CK 8 may be responsible for a significant fraction of the residual plasminogen binding, especially in experiments with the MCF-7 cells that were the most profoundly affected by 1E8 Fab.
Even though there are data to support an important role for u-PA in determining the metastatic potential of breast cancer cells (2, 3), we chose to assess the functional significance of plasminogen/breast cancer cell interactions by examining the rate of plasminogen activation by t-PA. We and others (6,12,13) have shown that only certain cell types, including hepatocytes, promote plasminogen activation by t-PA. We hypothesized that cell-surface CK 8 might be responsible for the previously demonstrated enhancement of plasminogen activation by t-PA in hepatocytes (12) and that CK 8 might promote plasminogen activation by t-PA in breast cancer cells. Our analysis of MCF-7 cells demonstrated that plasminogen activation by t-PA is greatly accelerated when the plasminogen is cell-associated. The increase in activation rate was attributed to a substantial decrease in K m . From the data presented, it was not possible to conclude whether the enhanced rate of plasminogen activation requires only plasminogen binding to CK 8 or a concomitant interaction of t-PA with CK 8 or another surface binding site. In any respect, the 83% decrease in plasminogen activation rate after 1E8 Fab pretreatment demonstrates that CK 8 plays an essential role in promoting the activation of plasminogen by t-PA on the MCF-7 cell surface. The lag phase observed in the analysis of cell-associated plasminogen activation by t-PA provides suggestive evidence that t-PA may need to bind to a cell-surface site before it acquires optimal catalytic efficiency. If this is true, then one could postulate a model for enhanced plasminogen activation by CK 8 that is similar to that originally proposed for fibrin (40). In this model CK 8 would mediate the formation of a ternary complex in which the K m for plasminogen activation is greatly decreased. In studies with purified CK 8, we recently demonstrated specific t-PA binding (41). CK 8 can be aberrantly expressed in many nonepithelial cancers, including lymphomas, melanomas, gliomas, and sarcomas (42,43). For many of these cancers, expression of CK 8 has been correlated with increased invasiveness in vitro and in vivo. When CK 8 is aberrantly expressed in squamous cell carcinomas, it is localized, by immunohistochemistry, primarily to the invasion front (44,45). In malignant melanoma, in vitro invasiveness has been directly correlated with cellular expression of CK 8 (46). Finally, mouse L cells, a noninvasive CK 8-negative fibroblast cell line, becomes invasive after cotransfection with CK 8 and its co-polymer CK 18 (47). Based upon our work, we hypothesize that the increased invasiveness of CK 8-expressing cells may be partially explained by the small fraction of CK 8 that is available at the cell surface and competent to function as a plasminogen receptor.
The mechanism by which CK 8 is expressed on the cell surface is unknown. The sequence of CK 8 does not include a transmembrane domain; however, our unpublished prelimi-nary data 2 suggest that cell-surface CK 8 is integrally associated with the plasma membrane. The alternate hypothesis is that CK 8, which is released by cells, binds to the external plasma membrane and then functions as an extrinsic receptor for plasminogen. In either case, the functional significance of CK 8, which is its ability to bind plasminogen and promote plasminogen activation on the tumor cell surface, is the same. We propose that cell-surface CK 8 may be important in promoting invasion of breast cancer cells.