Protease Nexin-1 Inhibits Plasminogen Activation-induced Apoptosis of Adherent Cells*

Degradation of adhesive glycoproteins by plasmin is implicated in cell migration. In this study, we further explored the role of plasminogen activation in cell adhesion and survival and show that uncontrolled plasminogen activation at the cell surface may induce cell detachment and apoptosis. We hypothesized that this process could be prevented in adherent cells by expression of protease nexin-1, a potent serpin able to inhibit thrombin, plasmin, and plasminogen activators. Using two- and three-dimensional culture systems, we demonstrate that Chinese hamster ovary fibroblasts constitutively express tissue-type plasminogen activator and ef-ficiently activate exogenously added plasminogen in a specific and saturable manner ( K m (cid:1) 46 n M ). The forma- tion of plasmin results in proteolysis of fibronectin re-adhesion of (cid:1) 60% viable cells for plasmin detachment and 51.4 (cid:6) 2% re-adhesion of (cid:1) 77% viable cells for control Versene detachment. These data suggest that a subpopulation of plas-min-detached cells may resist anoikis.

Apoptosis of adherent cells induced by disruption of integrinmediated cell-matrix interactions, referred to as anoikis, has been described in physiological conditions such as the renewal of intestinal epithelial cells and the involution of the mammary gland after lactation (for a review, see Ref. 1). Recent studies have shown that resistance to anoikis may also contribute to the progression of malignancy (2). Therefore, definition of the intracellular events and extracellular mediators involved in pathological anoikis could be helpful to define therapeutic targets. The intracellular pathways involved in anoikis have been extensively studied and include Bcl-2-related proteins (3), p53 (4), interleukin-1␤ converting enzyme-related proteases (including caspase-3), JNK (c-Jun N-terminal kinase) (5), MEKK (mitogen-activated protein kinase/extracellular signal-regulated kinase kinase kinase) proteolysis by caspases (6), focal adhesion kinase (7,8), and Akt (9). In contrast, the extracellular mediators that can trigger anoikis are less well characterized. It has been shown, however, that growth factors (such as insulin, insulin-like growth factor-1, and epidermal growth factor) and their receptors cooperate with integrins to trigger different intracellular survival signaling pathways (2). Proteases such as matrix metalloproteinases (10,11) and leukocyte elastase (12) could also trigger anoikis by breaking down cell-extracellular matrix interactions. Plasmin, the active enzyme generated by plasminogen activation (13), is able to cleave adhesive glycoproteins such as laminin (14,15) and fibronectin (16) and can interfere with integrin-mediated cell adhesion to adhesive glycoproteins (17)(18)(19). We hypothesized that pericellular plasmin generation could trigger anoikis of adherent cells and sought to demonstrate that inhibition of plasmin formation by transfection of serpins (serine proteinase inhibitors) would prevent this phenomenon. For this purpose, we selected protease nexin-1 (PN-1), 1 an ϳ43-kDa serpin that forms SDS-stable complexes with ␣-thrombin, plasmin, and plasminogen activators (20 -22). PN-1 has been shown to inhibit tumor cell-mediated extracellular matrix destruction (23), to modulate neurite outgrowth (24), and to promote neuronal cell survival (25). The secretion of PN-1 by a variety of anchoragedependent cells, including fibroblasts (26,27), astrocytes (28), neuroblastoma (29), fibrosarcoma (26), and vascular smooth muscle cells (30,31), suggests that it may play a major protective role against thrombin and other serine proteases (30).
Our results show that wild-type Chinese hamster ovary fibroblasts (CHO-K1) (32) constitutively express tissue-type plasminogen activator (t-PA) able to convert exogenously added plasminogen into plasmin. The pericellular generated plasmin induces cleavage of adhesive glycoproteins, followed by CHO-K1 cell detachment and apoptosis. We further show that gene transfer and expression of PN-1 or plasminogen activator inhibitor-1 (PAI-1) inhibited plasminogen activation-induced anoikis of CHO-K1 cells.

Purified Proteins
Human Glu-plasminogen was purified as described (35) with modifications and was considered to be Ͼ99% pure as assessed by SDS-PAGE and N-terminal sequence analysis (36). Human ␣-thrombin was obtained as reported previously (37). Protein concentration was calculated using molecular masses and extinction coefficients (⑀ 280 nm,1 cm 1% ) of 93 kDa and 16.8 for plasminogen and 36 kDa and 1.83 for ␣-thrombin. Elastase-derived plasminogen fragments K 1ϩ2ϩ3ϩ4 (the first four kringles) and K 5 -SP (kringle-5 and the serine protease domain) were prepared according to Sottrup-Jensen et al. (38). Recombinant plasminogen with Ser 741 mutagenized to Ala (rPg-S741A) was obtained as described elsewhere (39) and purified from the cell culture medium by affinity chromatography on lysine-Sepharose.

Cell Culture and Transfections
CHO-K1 cells (American Type Culture Collection CCL-61) were cultured at 37°C in a humidified atmosphere of 5% CO 2 using Ham's F-12 medium supplemented with 2 mM L-glutamine and 10% (v/v) fetal calf serum under two-and three-dimensional conditions.
Three-dimensional collagen matrix gel culture was performed as described (11,40) with minor modifications. Briefly, plastic 24-well culture plates were loaded with 175 l of a solution of 0.5 mg/ml collagen and 40 g/ml fibronectin in Ham's F-12 medium, and gelation was allowed to proceed for 30 min at 37°C. Serum-starved wild-type or sham-or PN-1-transfected CHO-K1 cells were isolated from culture flasks, resuspended in the collagen/fibronectin solution, and overlaid onto pre-gelled wells (250,000 cells in 175 l/well). Medium (200 l/ well) was added after gelation, and the cultures were incubated as described above. The collagen solution and medium were supplemented with 1 M plasminogen (final concentration) when indicated.
The cDNA coding sequence for rat PN-1 preceded by a Kozak consensus translation initiation site (41) was inserted as an EcoRI/XbaI fragment into a pcDNA3 expression vector (Invitrogen). The complete PN-1 coding sequence was built up by ligating, at the SphI restriction site, the 3Ј-end fragment of the rat PN-1 cDNA with a reverse transcription-PCR product coding for the missing 5Ј-end fragment. The 3Ј-end cDNA fragment was excised from the 1468-bp XhoI/XbaI cDNA probe (42). The 5Ј-end fragment was amplified from rat brain cDNA with primers 5Ј-GAATTCGCCACCATGAATTGGCATTTTCCCT-3Ј (sense) and 5Ј-GCATCACCGTTGAGAGCTGCTTCTTCGTC-3Ј (antisense), subcloned in the pGEM-T-Easy vector (Promega), and inserted as an EcoRI/SphI fragment. The pcDNA3 vector containing the complete PN-1 coding sequence or an empty pcDNA3 vector (sham) was transfected into CHO-K1 cells seeded in 75-cm culture flasks or on 96-well plastic plates and cultured as described above. The transfection procedure was performed using FuGENE 6 (Roche Applied Science) according to the manufacturer's protocols.
CHO-K1 cells overexpressing human active PAI-1 were obtained and cultured as described (43). The secreted PAI-1 has very similar or identical physicochemical and functional properties compared with natural PAI-1 (43).

Cellular Plasminogen Activation Assay
Wild-type and sham-and PN-1-transfected CHO-K1 cells cultured as described above were starved of serum for 24 h. Cells were then incubated with varying concentrations of either plasminogen or rPg-S741A (100 l/well) supplemented with inhibitors when indicated. In experiments aimed at neutralizing expressed PN-1, varying concentrations of ␣-thrombin were incubated with cells for the 24-h period of serum deprivation following the transfection procedure. Residual thrombin was inhibited with a 10-fold excess of hirudin (Stago) prior to the addition of plasminogen. The kinetics of plasmin formation at 37°C were monitored by measuring the rate of p-nitroaniline release from the chromogenic substrate CBS0065 (0.75 mM final concentration). The change in absorbance at 405 nm as a function of time (18 h) was monitored using a Dynex MR5000 multiwell plate reader. The plate was then washed with phosphate-buffered saline (PBS), and 0.75 mM CBS0065 (100 l/well) was added to detect cell-bound plasmin. In three-dimensional cultures, plasmin was detected in the conditioned medium after 48 h of incubation with 1 M plasminogen supplemented with 80 mM ⑀-aminocaproic acid or 0.5 M ␣ 2 -antiplasmin when indicated. Rates of p-nitroaniline release were transformed to femtomoles of plasmin as described (44).

PN-1 Expression by CHO-K1 Cells
Inhibition of ␣-Thrombin Activity and Formation of Complexes with 125 I-Labeled ␣-Thrombin-To detect PN-1, cell-free conditioned medium from sham-or PN-1-transfected CHO-K1 cells was incubated with ␣-thrombin at varying concentrations for 15 min at 37°C. Thrombin inhibition was checked by measuring residual thrombin using the chromogenic substrate CBS0065. Alternatively, the conditioned medium was incubated with 5 nM 125 I-labeled ␣-thrombin, and the formation of complexes with PN-1 was detected by SDS-PAGE and autoradiography (31). Free PN-1 was also detected by Western blotting of the culture supernatant using anti-rat PN-1 monoclonal antibody 4B3 as described below.
Immunocapture and Immunoblot Analysis-Following plasminogen activation experiments, cells were washed twice with PBS and lysed with 1% Triton X-100 in PBS for 15 min at 4°C. Samples were centrifuged at 10,000 ϫ g for 10 min at 4°C; the supernatant was removed by aspiration; and similar amounts of protein (Bradford protein assay) were incubated overnight at 4°C with antibodies immobilized on beads coated with either protein A (Amersham Biosciences) or goat antimouse IgG antibody (Dynal, Inc.) according to the manufacturers' instructions. The antibodies used were goat anti-human t-PA polyclonal IgG antibody and monoclonal antibody CPL15 (45) directed against kringle-1 of plasmin(ogen), respectively. The beads were then washed, resuspended in 25 l of SDS sample buffer (62.5 mM Tris base, 2% SDS, 10% glycerol, and 5% 2-mercaptoethanol), and boiled for 5 min. After centrifugation, proteins in a final volume of 20 l were separated by SDS-PAGE using 10% polyacrylamide gels. The separated proteins were transblotted onto a nitrocellulose membrane, which was saturated for 1 h with 5% nonfat milk powder in Tris-buffered saline (pH 7.4) and 0.1% Tween 20, and finally incubated for 2 h with monoclonal antibody 4B3 (specific for rat PN-1) (31). The membrane was then washed with Tris-buffered saline (pH 7.4) and 0.1% Tween 20 and incubated for 1 h with horseradish peroxidase-conjugated goat anti-mouse IgG antibody (1:1500). After five washes, the signal was detected using a chemiluminescence kit (ECL, Amersham Biosciences).

Detection of Cell Survival and Apoptosis
After 24 h (two-dimensional) or 48 h (three-dimensional) of incubation with plasminogen and other reagents, CHO-K1 cells were washed with PBS and submitted to various tests for the detection of cell survival and apoptosis.
Replating of Cells-Control cells (detached by incubation with 0.02% Versene (Invitrogen) for 5 min at 37°C) and plasmin-detached cells (also incubated with Versene) were washed with Ham's F-12 medium, resuspended in medium containing 10% (v/v) fetal calf serum, and replated for 60 min. The proportion of dead cells in the suspension of detached cells was quantitated by phase-contrast microscopy of trypan blue exclusion by living cells in a 1:1 mixture with trypan blue (Invitrogen).
Cell Detachment Assay-CHO-K1 cells in two-dimensional cultures were incubated for 1 h at 37°C with the tetrazolium salt 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; 0.5 mg/ml) in PBS (46). Detached cells were discarded, and the remaining living adherent cells formed formazan crystals, which were dissolved in dimethyl sulfoxide and colorimetrically detected at A 550 nm using a multiwell plate reader. Absorbance readings are proportional to the number of living cells.
Terminal Deoxynucleotidyltransferase-mediated dUTP Nick End Labeling (TUNEL) and 4Ј,6-Diamidino-2-phenylindole (DAPI) Staining-To visualize DNA fragmentation, cells grown on 4-well slide chambers (Labtech) or cells cytospun from collagenase-digested threedimensional cultures were submitted to TUNEL (Roche Applied Science) and DAPI (Sigma) nuclear counterstaining according to the manufacturers' instructions. After washing, the slides were mounted with Fluoprep (Dako), and positivity was counted using an epifluorescence microscope. The apoptotic index was calculated as the percentage of TUNEL-positive nuclei relative to total DAPI-stained nuclei.

Detection of t-PA, Matrix Metalloproteinase (MMP)-9, and PAI-1
Fibrin zymography was performed as described previously (48). Briefly, samples prepared as described below were electrophoresed under nonreducing conditions on SDS-10% polyacrylamide gels. The SDS was then exchanged with 2.5% (w/v) Triton X-100, and the gel was carefully overlaid on a fibrin-agarose gel containing 100 g/ml IgG directed against t-PA when indicated or, for reverse fibrin zymography, 0.5 IU/ml urokinase. Zymograms were incubated at 37°C and regularly photographed over a period of 4 -24 h. MMP-9 activity in the conditioned medium was quantified as described previously (49). PAI-1 and t-PA⅐PAI-1 complexes were determined by enzyme-linked immunosorbent assay (ELISA) as described previously (50).

Statistical Analysis
Statistical analyses were performed using Statview Version 5.0 software. Results are expressed as the means Ϯ S.D. unless otherwise stated. Comparisons used one-way analysis of variance with Scheffe's F test or Wilcoxon signed ranks as appropriate. Statistical significance was set at p Ͻ 0.05.

Wild-type CHO-K1 Cells Are Able to Activate Plasminogen-
Plasminogen incubated with wild-type CHO-K1 cells was activated in a saturable and specific manner (V max ϭ 17.6 nmol/ min, K m ϭ 46 nM) (Fig. 1). rPg-S741A, which is an inactive catalytic site mutant, could be converted to a two-chain molecule, but no degradation of the chromogenic substrate CBS0065 was detected, thus indicating absence of activity. The activator, constitutively expressed by wild-type CHO-K1 cells, was identified as t-PA by (i) the inhibition in cell lysates by specific anti-t-PA polyclonal IgG antibodies (Fig. 2, inset B) of a fibrin zymography lysis band corresponding to the molecular mass of t-PA (Fig. 2, inset A) and (ii) the inhibition of plasminogen activation by 1,5-dansyl-Glu-Gly-Arg chloromethyl ketone (IC 50 ϭ 2 nM), a selective inhibitor for t-PA. No fibrinolytic band was apparent in zymograms of the conditioned medium from 24-h two-dimensional cell cultures, indicating that cell-associated t-PA was converting plasminogen into plasmin at the cell membrane. Plasmin formation was indeed inhibited by the lysine analog ⑀-aminocaproic acid (IC 50 ϭ 0.21 mM), further confirming that plasminogen activation occurred at the cell surface after binding to C-terminal lysine residues of membrane glycoproteins. The amount of plasmin bound to cells increased as a function of the concentration of plasminogen added (Fig. 1, inset) and was progressively released into the supernatant, where it could be specifically inhibited by ␣ 2antiplasmin (IC 50 ϭ 1.4 nM) (Fig. 2). CHO-K1 cells in threedimensional cultures expressed t-PA (detected in the conditioned medium at 48 h), activated plasminogen, and produced an amount of plasmin (40 nM at 48 h) that could be inhibited by ␣ 2 -antiplasmin (500 nM final concentration); plasmin formation was prevented by 80 mM ⑀-aminocaproic acid.
Proteolysis of natural substrates by plasmin was demonstrated. After a 18-h incubation with 500 nM plasminogen, both fibronectin (Fig. 3, left panel) and laminin (data not shown) were degraded, and pro-MMP-9 expressed by CHO-K1 cells was activated in a dose-dependent manner (right panel).
Cellular Plasminogen Activation Induces Cell Detachment-Plasminogen activation by wild-type CHO-K1 cells was followed by cell shrinkage and detachment, which occurred in a dose-and time-dependent manner (between 12 and 24 h) in two-dimensional cultured cells (Fig. 4A). An inverse linear correlation was observed between the amount of cell-generated plasmin and the percentage of remaining viable adherent cells (r ϭ 0.987) as assessed by the MTT test. From the data presented in Fig. 4 and the microscopic observation of cells in experiments parallel to those shown in Fig. 2, it was concluded that cell detachment was directly dependent on both binding of plasminogen to cells and the proteolytic activity of plasmin. Cell detachment was indeed prevented by ⑀-aminocaproic acid and 1,5-dansyl-Glu-Gly-Arg chloromethyl ketone. In contrast, the mini-plasminogen (kringle-5 and the serine protease do- main, K 5 -SP), which lacks the high affinity lysine-binding sites present in kringle-1 and kringle-4; the isolated fragment K 1ϩ2ϩ3ϩ4 , which lacks the proteolytic region; and rPg-S741A were not able to induce any significant cell detachment (Fig.  4A). CHO-K1 cells seeded in three-dimensional collagen matrices appeared as single rounded isolated cells that underwent major morphological events, including sprouting with growth cone-like projections and branching upon 48 -72 h of culture (Fig. 4B, upper panel). The addition of 1 M plasminogen to the three-dimensional gels resulted in regression of the morphogenic changes, i.e. the cells remained rounded and presented vacuolization, membrane blebbing, and no further evidence of morphogenesis (Fig. 4B, lower panel). These plasmin(ogen)induced effects were prevented by ⑀-aminocaproic acid and ␣ 2 -antiplasmin, thus indicating that cleavage of adhesive glycoproteins by plasmin induced the detachment of cells. Activation by plasmin of pro-MMP-9 expressed by CHO-K1 cells (Fig.  3) did not contribute significantly to cell detachment, as the incubation of cells with up to 200 M GM6001, a broad-range synthetic inhibitor of MMPs, did not prevent cell detachment (data not shown). At 60 h of treatment with 500 nM plasminogen, all adherent cells were detached. Using the trypan blue exclusion assay, it was found that 39.5 Ϯ 2.2% of these cells were dead compared with 23.3 Ϯ 3.6% in Versene-detached control cells. When all detached cells were replated, viable cells that were able to adhere were quantified with an MTT test performed after a 1-h incubation at 37°C: 26.3 Ϯ 4.2% readhesion of ϳ60% viable cells for plasmin detachment and 51.4 Ϯ 2% re-adhesion of ϳ77% viable cells for control Versene detachment. These data suggest that a subpopulation of plasmin-detached cells may resist anoikis.
Wild-type CHO-K1 Cell Detachment and Apoptosis-The proteolysis of matrix glycoproteins (Fig. 3) and the cell detachment induced by the cellular activation of plasminogen (Fig. 4) led to apoptosis, as estimated by TUNEL staining and ELISA quantification of histone-associated DNA fragments (Fig. 5). After a 24-h incubation with 500 nM plasminogen, the apoptotic TUNEL index of residual adherent cells and floating cells was 8.2 Ϯ 5.6 and 56.25 Ϯ 4.6%, respectively, compared with 0.35 Ϯ 0.2% for untreated control cells (Fig. 5B). At this plasminogen concentration, caspase activity was significantly high (3.7-fold increase compared with control cells; p Ͻ 0.03 versus controls; n ϭ 6). Similar nuclear apoptotic changes were observed in cytospun cells isolated from three-dimensional collagen gels containing 1 M plasminogen (apoptotic index, 9.6 Ϯ 2.7% versus 0.3% in untreated cells). A dose-dependent increase in DNA fragmentation was documented at varying plasminogen concentrations, but no significant effect was detected in cells incubated with rPg-S741A (Fig. 5C).
Cell Transfection with Protease Nexin-1 Inhibits Plasminogen-induced Anoikis-Plasminogen activator inhibitors were not detected in the conditioned medium from wild-type or sham-or PN-1-transfected CHO-K1 cells, as assessed by reverse fibrin zymography and ELISAs directed against human, rat, or mouse PAI-1 (data not shown). Cellular plasmin formation, cell detachment, and apoptosis were therefore uncontrolled after the addition of plasminogen. We hypothesized that the cellular expression of PN-1 could be used to prevent this phenomenon, as this serpin inhibits both plasminogen activators and plasmin. We first performed a transient transfection of CHO-K1 cells either with a plasmid containing the cDNA of rat PN-1 (PN-1-transfected cells) or with an empty plasmid (sham- transfected cells). Using an immunoblot procedure (Fig. 6A), PN-1 was detected in both the lysates and conditioned medium from transfected cells. In contrast, no PN-1 was detectable in sham-transfected cells. The secreted PN-1 had the expected PN-1 characteristics. (i) SDS-stable complexes were formed with ␣-thrombin (Fig. 6B) and with plasmin (data not shown); and (ii) the cell-conditioned medium blocked the amidolytic activity of plasmin (Fig. 6C) and thrombin (data not shown). Complexes between t-PA and PN-1 were identified after cell lysis and immunocapture with anti-t-PA antibody immobilized on protein A-Sepharose beads, SDS-PAGE separation of re-tained proteins, and immunoblotting using anti-PN-1 monoclonal antibody 4B3 (Fig. 7A). The t-PA⅐PN-1 complexes formed in CHO-K1 cells transfected with PN-1 were also observed by fibrin zymography (Fig. 7B, left panel). Zymograms of the conditioned medium collected from wild-type or sham-transfected CHO-K1 cells in three-dimensional collagen matrices showed a lysis band corresponding to free t-PA that was not observed in PN-1-transfected cells (Fig. 7B, right panel). The decreased amount of free t-PA and plasmin observed in PN-1-transfected cells relative to sham-transfected cells most probably corresponds to consumption by formation of complexes with PN-1.
The rate of plasminogen activation at the surface of PN-1transfected cells in both two-dimensional plates and threedimensional collagen matrix cultures was significantly decreased compared with sham-transfected cells (Fig. 8). Rates were corrected for the number of cells/well (sham-transfected versus PN-1-transfected wells), and the calculated V max values are therefore independent of this parameter and were found to be significantly different (p Ͻ 0.01). Sham-transfected and control cells showed no difference in plasminogen activation. The inhibitory effect of PN-1 on plasminogen activation was related to its serpin activity, as indicated by its reversal in experiments in which PN-1 was neutralized with ␣-thrombin (Fig. 9). For this purpose, PN-1-transfected cells were incubated with varying concentrations of ␣-thrombin for 24 h, followed by neutralization with hirudin prior to plasminogen activation. Under these conditions, the amount of plasmin formed increased proportionally to the amount of thrombin added and reached levels similar to those in sham-transfected cells at a thrombin concentration of 2.5 nM (Fig. 9). Transfection of PN-1 in CHO-K1 cells resulted in protection against cell detachment induced by activation of plasminogen at varying concentrations (Fig. 10A). PN-1 also protected against apoptosis, as indicated by a significant decrease in the DNA fragmentation ratio as estimated by ELISA of histone-associated DNA fragments (Fig.  10B). Furthermore, in cytospun cells isolated from three-dimensional collagen gels containing 1 M plasminogen, the apoptotic TUNEL index of sham-transfected cells decreased from 7.2 Ϯ 1.1 to 3.2 Ϯ 0.4% in PN-1-transfected cells (p ϭ 0.005) (Fig. 10C).
Comparable experiments were performed with PAI-1-transfected CHO-K1 cells, which similarly resisted plasminogen activation and anoikis compared with control cells. Indeed, we observed neither plasminogen activation nor apoptosis in human cells stably transfected with PAI-1 and incubated for 24 h with up to 1 M plasminogen. This complete protection was probably achieved via t-PA inhibition, as indicated by the detection of t-PA⅐PAI-1 complexes, which were undetected in control cells. DISCUSSION The CHO fibroblastic cell line has been widely used to study the mechanisms of cell adhesion and survival (51)(52)(53)(54)(55). Wildtype CHO-K1 cells express adhesive glycoproteins, including fibronectin (51) and laminin (52), and integrins ␣ 5 ␤ 1 and ␣ v ␤ 3 (53), which behave as fibronectin receptors. Attachment of CHO-K1 cells to fibronectin mediated by ␣ 5 ␤ 1 integrin induces cell survival signaling, which protects them from serum starvation-induced cell death (54,55). In contrast, CHO-K1 cells grown in suspension in serum-free medium undergo apoptosis (56). In addition, short-term proteolytic treatment of fibronectin-coated substrata with trypsin results in a decrease in CHO cell adhesion (57). Furthermore, when generated at the cell surface, plasmin can abrogate the cell-binding function of the extracellular matrix proteins vitronectin (17) and laminin (14,15). It has been suggested that the pericellular proteolytic potential of adherent cells may function as an intermediary mechanism in endothelial cell adhesion/migration (58) and detachment (19) and in optic nerve ligation-induced ganglion cell loss (15).
In this study, we have demonstrated that adherent CHO-K1 cells in both two-dimensional cultures and three-dimensional fibronectin-containing collagen matrices express t-PA and may therefore have the potential to develop pericellular proteolytic activity via plasminogen activation. Although CHO-K1 cells may express cell-surface urokinase-type plasminogen activator (u-PA) receptor (59), we did not detect u-PA in cell lysates. The natural expression of t-PA by CHO-K1 cells has not as yet been described. We therefore sought to analyze the consequences of rendering the above-mentioned pericellular proteolytic pathway operative. Because plasmin is able to degrade adhesive glycoproteins such as fibronectin (16) and laminin (14), we hypothesized that pericellular plasmin generation could lead to a disturbance of the survival signaling mediated by integrins via the cleavage of adhesive glycoproteins. We have demonstrated that the generation of pericellular plasmin activity at the surface of CHO-K1 cells is followed by morphogenic events, cell retraction, and apoptosis, i.e. anoikis, which depends on both the cellular activation of plasminogen by t-PA and the proteolytic activity of plasmin. Cleavage of extracellular matrix proteins such as fibronectin and laminin and the survival of non-apoptotic replated cells suggest that plasmin-induced cell detachment was due to digestion of the extracellular matrix rather than loss of adhesive proteins from the cell surface. These results are in agreement with previous observations in vascular smooth muscle cells (18,60). Activation of plasminogen and anoikis were inhibited by competitors of plasminogen binding and a synthetic inhibitor of t-PA, thus indicating that assembly of plasminogen and t-PA activity are necessary to convert plasminogen into plasmin on the cell membrane. The plasmin(ogen) dependence of cell detachment and apoptosis was further confirmed by the absence of effect of rPg-S741A, a plasminogen active-site mutant. The demonstration in threedimensional collagen matrix cultures, a model system that mimics the in vivo cell environment (61), of plasmin-dependent apoptosis and its reversal by PN-1 provides strong support to the possible relevance of this phenomenon in vivo.
The enzymatic activity of serine proteases is tightly regulated by a series of specific endogenous serpins. Plasma proteinase inhibitors have been shown to behave as potential anti-apoptotic factors for vascular smooth muscle cells (62). This finding supports the view that serpins may function as a regulatory mechanism, as uncontrolled proteolysis may lead to cell detachment and apoptosis. A variety of anchorage-dependent cells produce their own regulators of proteolytic activity. For example, PAI-1, a known inhibitor of plasminogen activators may play a significant role in regulating plasmin formation at the cell surface (63) and in the prevention of plasminogeninduced capillary tube regression (11). PN-1, secreted by a variety of cell types, may also regulate plasminogen activation by forming covalent inhibitory complexes with plasmin, u-PA, and t-PA (20, 21, 64 -67). These complexes are internalized and degraded, thus contributing to regulation of serine protease were incubated with anti-t-PA polyclonal IgG antibodies immobilized on protein A beads. Proteins were eluted from the beads with SDS sample buffer and analyzed by Western blotting after SDS-PAGE under reducing conditions. The membrane was probed with mouse anti-rat PN-1 monoclonal antibody 4B3, and bands were detected using peroxidase-conjugated goat anti-mouse IgG antibody and a chemiluminescence procedure as described under ''Experimental Procedures.'' A band of 110 kDa corresponding most probably to t-PA⅐PN-1 complexes and a band corresponding to PN-1 were observed in the lysates from PN-1-transfected cells (arrows). The two nonspecific bands observed in the control sample as well as in the sham-and PN-1-transfected samples may correspond to protein A and IgG dissociated from the Sepharose beads. Protein A has been shown to run anomalously upon SDS-PAGE (78). B, fibrin-agar zymography. Samples were separated by SDS-PAGE; and after renaturation of proteins with Triton X-100, the gel was overlaid onto a fibrin-agarose gel supplemented with plasminogen. Fibrinolytic activity was detected as lysis areas after overnight incubation at 37°C. Left panel, lysates from serum-starved CHO-K1 cells cultured in two-dimensional plates. Both sham-and PN-1-transfected CHO-K1 cells expressed t-PA. One additional fibrinolytic band of slower mobility, corresponding most probably to t-PA⅐PN-1 complexes, was observed in the lysates from CHO-K1 cells expressing PN-1. Complexes of t-PA with serpins may paradoxically exhibit fibrinolytic activity after SDS-PAGE and renaturation with Triton X-100 (41). Right panel, supernatants from three-dimensional cultured sham-and PN-1transfected CHO-K1 cells in the presence of 1 M plasminogen. The upper bands and the lower band observed in the sham-transfected sample correspond to plasmin (Pn) and t-PA, respectively. In the PN-1-transfected sample, t-PA was not observed; and the plasmin band was substantially decreased. activity (22,68) and extracellular matrix destruction (23,69). CHO-K1 cells are remarkable in that they do not express detectable amounts of either PAI-1 or PN-1 and constitute therefore an original model to study the effect of a serpin on cell proteolysis. We took advantage of this particularity to evaluate how plasmin-induced apoptosis could be modulated by transfection of these cells with PN-1. Cell transfection with PAI-1, a well known model of plasminogen activation inhibition, was studied in parallel.
The sham transfection procedure did not alter either the ability of CHO-K1 cells to express t-PA or their attachment properties compared with control cells. Furthermore, no traces of PAI-1 could be detected in these cells. PN-1 was absent in cell lysates and supernatants from sham-transfected CHO-K1 cells, which transformed plasminogen into plasmin at the cell surface and underwent major morphological changes, followed by apoptosis under both two-and three-dimensional culture conditions. In contrast, transfection of rat PN-1 cDNA into CHO-K1 cells resulted in the formation of t-PA⅐PN-1 and plasmin⅐PN-1 complexes, which prevented plasminogen activation and plasmin-induced anoikis. A substantial decrease in the amount of cell-associated t-PA activity or t-PA released in  PN-1 (n ϭ 6). C, shown is the TUNEL apoptotic index. Cytospun control and sham-and PN-1-transfected cells isolated from three-dimensional cultures were submitted to the TUNEL reaction and counterstained with DAPI. Bars represent the apoptotic index of TUNEL-positive cells as a percentage relative to total DAPI cell nuclei as evaluated by epifluorescence microscopy. *, p Ͻ 0.0005 versus controls; **, p Ͻ 0.005 versus sham-transfected cells. The residual apoptotic index of PN-1-transfected cells (3.1%) compared with control cells in the absence of plasminogen (0.3%) was most probably related to the efficiency of transfection. the conditioned medium was observed in PN-1-transfected versus sham-transfected cells by fibrin zymography. An interesting and potentially important observation concerns the effect of ␣-thrombin on the response of PN-1-transfected cells to plasminogen. With the addition of increasing amounts of ␣-thrombin to PN-1-transfected cells, the decrease in plasmin formation and the prevention of apoptosis caused by PN-1 were reversed. These data indicate that PN-1 was responsible for the inhibition of plasmin formation and the maintenance of cell adhesion and survival. Of note, the similar protection against apoptosis observed with PAI-1-transfected cells (this study) or PAI-1 constitutively expressed by endothelial cells (11) further supports the concept that cell detachment and apoptosis are plasminogen activation-dependent.
PN-1 is barely detectable in plasma (70), thus suggesting that its action is localized at the perivascular environment, where its principal targets may be generated (thrombin, plasmin) or expressed (plasminogen activators). PN-1 present in capillaries and vascular smooth muscle cells of arteries in the brain may play a major protective role against thrombin and possibly other serine proteases following cerebrovascular injury (30). For instance, neurite outgrowth inhibited by thrombin is relieved by PN-1 (71). The u-PA-mediated invasion of the amniotic membrane by tumor cells has also been shown to be inhibited by PN-1 (72). Our present data provide experimental evidence indicating that PN-1 may also be protective against the effects of uncontrolled plasmin formation on cell survival. This could be the case in cells and tissues in which PN-1 is the main regulatory serpin (27) and under conditions associated with PN-1 up-regulation. For instance, the recently demonstrated up-regulation of the PN-1 gene by the Prx2 homeoprotein suggests that this serpin might be involved in skeletal and cardiovascular development as well as in adult vascular remodeling (73). Increased expression of PN-1 in tumor cell lines also suggests its involvement in tumorigenesis (74,75). Furthermore, we recently reported up-regulation of PN-1 by vascular smooth muscle cells in hypertensive rats, which display arterial wall hypertrophy (31).
In conclusion, we have shown that the pericellular activation of plasminogen is a powerful proteolytic pathway able to trigger apoptosis. This phenomenon may be relevant in pathological states where plasminogen activation (76) and extracellular matrix proteolysis and cell death by apoptosis (77) coexist, i.e. in the cardiovascular and central nervous systems. The CHO-K1 cell two-and three-dimensional culture proteolytic model may be a useful tool to evaluate therapeutic agents such as the serpins in pathologies involving imbalance between proteases and their inhibitors. In this context, PN-1 appears to be a promising candidate.