A Truncated Plasminogen Activator Inhibitor-1 Protein Induces and Inhibits Angiostatin (Kringles 1–3), a Plasminogen Cleavage Product*

Plasminogen activator inhibitor-1 (PAI-1) is a serpin protease inhibitor that binds plasminogen activators (uPA and tPA) at a reactive center loop located at the carboxyl-terminal amino acid residues 320–351. The loop is stretched across the top of the active PAI-1 protein maintaining the molecule in a rigid conformation. In the latent PAI-1 conformation, the reactive center loop is inserted into one of the beta sheets, thus making the reactive center loop unavailable for interaction with the plasminogen activators. We truncated porcine PAI-1 at the amino and carboxyl termini to eliminate the reactive center loop, part of a heparin binding site, and a vitronectin binding site. The region we maintained corresponds to amino acids 80–265 of mature human PAI-1 containing binding sites for vitronectin, heparin (partial), uPA, tPA, fibrin, thrombin, and the helix F region. The interaction of “inactive” PAI-1, rPAI-123, with plasminogen and uPA induces the formation of a proteolytic protein with angiostatin properties. Increasing amounts of rPAI-123 inhibit the proteolytic angiostatin fragment. Endothelial cells exposed to exogenous rPAI-123 exhibit reduced proliferation, reduced tube formation, and 47% apoptotic cells within 48 h. Transfected endothelial cells secreting rPAI-123 have a 30% reduction in proliferation, vastly reduced tube formation, and a 50% reduction in cell migration in the presence of VEGF. These two studies show that rPAI-123 interactions with uPA and plasminogen can inhibit plasmin by two mechanisms. In one mechanism, rPAI-123cleaves plasmin to form a proteolytic angiostatin-like protein. In a second mechanism, rPAI-123 can bind uPA and/or plasminogen to reduce the number of uPA and plasminogen interactions, hence reducing the amount of plasmin that is produced.

Plasmin, the end product of plasminogen activator-derived proteolysis, is a trypsin-like enzyme that functions in the degradation of fibrin and extracellular matrix (ECM) 1 proteins.
Plasminogen is converted to plasmin by the activity of urokinase plasminogen activator (uPA) or tissue plasminogen activator (tPA) (1,2). Plasminogen, uPA, and plasmin bind to uPA receptors (uPAR) localizing the proteolytic activity of plasmin to the cell surface where it is used by migrating cells to degrade the ECM and basement membrane proteins (1,3,4).
The proteolytic activity of plasmin is regulated by plasminogen activator inhibitors 1 and 2 (PAI-1 and PAI-2). PAI-1 and uPA are coordinately expressed in the normal regulatory pathways of plasminogen activator-derived proteolysis.
PAI-1 is a secreted protein that is rapidly converted to a latent form at physiological temperatures. It becomes stabilized and activated when it binds vitronectin (Vn) (5-7), a cell adhesion glycoprotein (6,8) found primarily in the ECM at acute injury, in tissue repair (9), and in some malignant tumors (10).
PAI-1 is a multifunctional regulatory protein. Vitronectinbound PAI-1 (active) plays a crucial role in regulating the proteolytic degradation of the extracellular matrix, which in turn regulates cell migration and invasion (5,11). Active PAI-1 regulates proteolysis by inhibiting uPA/tPA. Upon binding Vn, activated PAI-1 then binds a uPA⅐uPAR complex, thus forming a PAI-1⅐uPA⅐uPAR complex (12)(13)(14). These interactions result in a cascade of regulatory events: 1) dissociation of PAI-1 and uPAR from their respective vitronectin binding sites (15,16); 2) inactivation of uPA and PAI-1 by internalization and lysosomal degradation of the PAI-1⅐uPA complex (17,18); 3) recycling of uPAR to a new site on the cell surface (19); 4) availability of PAI-1 binding sites on vitronectin for attachment of other molecules, such as ␣ v ␤ 3 integrins (20,21), plasmin, and recycled uPAR. The regulatory mechanisms of PAI-1 are also regulated by other molecules in the proteolytic pathway. PAI-1 and plasmin both have vitronectin binding sites. The serine protease activity of plasmin degrades Vn such that one of the PAI-1 binding sites is mostly lost. This results in destabilization of the Vn⅐PAI-1 complex and a loss of PAI-1 function as an inhibitor of plasmin formation through the plasminogen activator pathway. This same PAI-1 binding site on Vn overlaps a heparin binding site. The protease activity of plasmin completely abolishes the Vn binding site for heparin and exposes a plasminogen binding site (9). Heparin interaction with PAI-1 increases the specificity of PAI-1 for thrombin (22). Variations in Vn conformational states can convert PAI-1 inhibitory specific-ity from the plasminogen activators to thrombin (23,24).
PAI-1 belongs to the serpin family of serine protease inhibitors. The PAI-1 protein contains a strained loop reactive center, amino acids 320 -351, located at the carboxyl terminus of the molecule (25)(26)(27)(28)(29). PAI-1 initially interacts with uPA at Arg-346 (30) and acts as a pseudo substrate "bait." However, uPA is unable to cleave the pseudo substrate and PAI-1/uPA form a stable complex rendering uPA inactive (29,(31)(32)(33)(34). It has been demonstrated that the PAI-1-reactive loop must then insert between strands 3 and 5 of the ␤-sheet to exhibit inhibitory activity (35)(36)(37). Declerck et al. (38,39) showed that the reactive loop can be cleaved by uPA between residues 346 and 347, turning PAI-1 into a substrate. In the inactive configuration of PAI-1 (not bound to vitronectin), the reactive loop is spontaneously inserted into the folds of PAI-1 at physiological temperatures (29,40). This conformation is stable due to the loop insertion into strand 4a of the PAI-1 structure (41). A Lys-323 residue located within the ␤-strand s5A is thought to play an important role in the regulation of the latent/inactive state. This residue forms a hydrogen bond with Asn-150 (or 144 in mature protein) on the s3A/helix F loop of PAI-1 (42).
In an attempt to gain a better understanding of the relationship between PAI-1 structure and its multiple regulatory functions, the porcine PAI-1 gene was systematically dissected to obtain truncated PAI-1 proteins, rPAI-1. The rPAI-1 proteins would be used as tools to identify the regions of the inhibitor that bind (complex) other important regulatory molecules in proteolytic and adhesion mechanisms. In so doing, we hoped to identify other domains that have a biological function.
We have closely examined one deleted PAI-1 gene product (rPAI-1 23 ) that encompasses the porcine PAI-1 gene sequences (46) containing nucleotides 444 -999 corresponding to nucleotides 238 -793 in the human PAI-1 gene (amino acids 80 -265 of mature protein) (47). The recombinant rPAI-1 23 fragment: 1) excluded those sequences coding for the reactive center loop at the carboxyl terminus; 2) eliminated a vitronectin binding site, more than one half of a heparin binding region, and the latency residue at Lys-323 (42); 3) retained vitronectin and a second tPA binding region at amino acids 128 -145, part of a heparin binding site at amino acids 80 -88, a thrombin interactive region at residues 115-148, residues 127-158, which contain the helix F and the loop that connects helix F with helix s3A (48,49).
The data we present show that when a truncated PAI-1 protein, rPAI-1 23 , complexes with plasminogen or plasminogen and uPA, the following results occur: 1) Plasmin is cleaved to produce a proteolytic angiostatin-like cleavage product. UPA enhances the formation of the angiostatin-like protein by increasing the amount of available plasmin. 2) Excess rPAI-1 23 inhibits proteolytic angiostatin formation by binding to uPA and plasminogen such that they are not available to form plasmin that can be cleaved into angiostatin. 3) Cultured endothelial cells exposed to rPAI-1 23 exhibit a decrease in proliferation, a loss in tube formation capability, increased apoptosis, and decreased migration in the presence of VEGF.
The PAI-1 DNA fragments were double-digested with EcoRI and XbaI and ligated into a Pischia pastoris yeast shuttle vector, pGAPZ␣A (Invitrogen, Carlsbad, CA). The TOP 10 strain of Escherichia coli was transformed by electroporation of the ligation reaction in 2-mm cuvettes containing 50 ng of the ligation reaction/100 l of electroporation competent TOP 10 cells using a BTX Electro Cell Manipulator 600 (BTX, San Diego, CA). The electroporation conditions were 2.5-kV resistance, 129 ohm and a charging voltage of 2.45 kV. The pulsed cells were incubated at 37°C for 1 h and then transferred to an agar plate containing low salt Luria broth and 50 g/ml zeocin. Following an overnight incubation at 37°C, colonies were selected and grown in low salt LB broth for 5-7 h at 37°C. The DNA from each colony growth was isolated on a Qiagen miniprep spin column (Qiagen, Inc., Valencia, CA). The purified DNA from each expanded colony was double-digested to verify the presence of the gene insert. Positive isolates from the restriction enzyme digests (Fig. 2) were verified by sequencing.
rPAI-1 23 Protein Expression and Isolation-The DNA from a positively identified ligation product containing the rPAI-1 23 gene fragment was linearized at a specific restriction enzyme site in the pGAPZ␣A glyceraldehyde-3-phosphate dehydrogenase promoter. The linear DNA was transfected into electroporation competent P. pastoris strain of yeast according to the manufacturer's protocol (Invitrogen). The transfected yeast cells were streaked onto YPD/zeocin (100 g/ml) plates. Resulting clones were streaked onto YPD plates a second time to ensure integration of the transfected DNA into the yeast genome. Individual clones were selected after 2-3 days of growth on the YPD plates and grown in YPD medium at 30°C for 1 day. The transfected cells were pelleted, and the supernatant containing the secreted recombinant PAI-1 protein was collected and verified on a silver-stained 4 -20% gradient SDS-polyacrylamide gel (Fig. 3).
PAI-1 Activity Assay to Detect Plasmin Formation-The rPAI-1 23 protein (30 nM) was incubated with 10 nM of uPA (American Diagnostica Inc., Greenwich, CT) in TBS, pH 7.8, containing 100 g/ml BSA and 0.01% Tween 20 at room temperature for 1 h. Following the PAI-1⅐uPA reaction, 33 nM plasminogen and 15 g/ml Chromozym PL (Roche Molecular Biochemicals) were added to the rPAI-1 23 ⅐uPA reaction mix. The reactants continued to incubate at 37°C until color change could be detected in the uPA standards. The colorimetric change was measured at 405 nm on an EL-311 Microplate Autoreader (Bio-Tek Instruments, Burlington, VT). In this assay plasmin cleaves Chromozym PL resulting in a p-nitraniline molecule that absorbs at 405 nm. The colorimetric change is a measurement of plasmin formation.
Zymography Functional Assay-1.3% casein was incorporated into an SDS-polyacrylamide separating gel. The casein serves as a substrate FIG. 1. Truncated rPAI-1 23 gene relative to full-length human and porcine PAI-1. This drawing depicts the rPAI-1 23 gene relative to the human and porcine PAI-1 genes. It also illustrates the PAI-1 gene sequences/regions that code for amino acids known to be important binding sites for various regulatory proteins in the proteolytic and fibrinolytic pathways.
for plasmin proteolytic activity. The rPAI-1 23 protein (30 -120 nM) was bound to 10 nM of uPA at room temperature for 1 h. Following this reaction, 33 nM of plasminogen was added to the rPAI-1 23 ⅐uPA reaction mix and incubated at 37°C for 20 -30 min. The samples were electrophoresed at constant voltage on a polyacrylamide gel containing 1.3% casein. The polyacrylamide gels containing the electrophoresed samples were first washed in 2.5% Triton-X before an overnight incubation at 37°C in a 50 mM Tris-HCl buffer, pH 7.9. The gels were fixed in 40% methanol, 10% acetic acid and then stained in 10% Coomassie Blue for 1 h. The gel was destained in 10% methanol, 10% acetic acid to detect the regions of proteolysis (50,51).
Zymography Analysis of Vitronectin Binding to rPAI-1 23 -Vitronection (Sigma Chemical Co., St. Louis, MO) at a concentration of 25 nM was incubated with 30 nM of rPAI-1 23 for 1-2 h at 37°C before performing the zymography functional assay as described above.
Western Blot Analysis of the 36-kDa Proteolytic Fragment Resulting from rPAI-1 23 , Plasminogen, uPA Reaction-Equal fractions of reaction mixtures for the zymography assay described above were placed on an SDS, nonreducing, 4 -20% gradient polyacrylamide gel. The electrophoresed samples were transferred to a nitrocellulose membrane. The membrane containing the transferred proteins was blocked and washed. The results were visualized on a Western blot probed with 1 g/ml goat anti-human plasminogen kringles 1-3 (angiostatin) (R&D, Minneapolis, MN) for 1 h at room temperature. The membrane was washed and blocked a second time for 1 h at room temperature. A rabbit anti-goat IgG (HϩL) secondary antibody (Pierce, Rockford, IL) at a concentration of 1 g/ml was incubated with the membrane for 1 h at room temperature. A horseradish peroxidase-conjugated tertiary antibody (donkey anti-rabbit IgG, Amersham Pharmacia Biotech, Arlington Heights, IL) diluted 1:1000 amplified the binding reaction, which was ultimately detected by addition of a chemiluminescent substrate (Amersham Pharmacia Biotech) for 45 min at room temperature (52).
Varied Permutations for rPAI-1 23 , uPA, and Plasminogen Interactions-The order in which rPAI-1 23 , uPA, and Plg were added to a reaction mixture was varied: (a) uPA ϩ Plg for 1 h at 37°C, then rPAI-1 23 was added for an additional 1 h at 37°C; (b) rPAI-1 23 ϩ uPA for 1 h at 37°C, then Plg was added for an additional 1 h at 37°C; or (c) rPAI-1 23 ϩ Plg for 1 h at 37°C, then uPA was added for an additional 1 h at 37°C.
In some cases, the amount of uPA and/or rPAI-1 23 was varied. To ensure that either uPA or Plg was exhausted in a reaction mix, an additional amount of either uPA or Plg was added to some permutations The samples were electrophoresed on a zymogram as described.
Endothelial Cell Incubation with Exogenous rPAI-1 23 -Porcine endothelial cells were plated at a density of 0.5 ϫ 10 6 cells/ml in 24-well culture plates containing DMEM supplemented with 10% fetal calf serum. The cells were allowed to adhere to the culture plate before adding exogenous rPAI-1 23 (1 g/ml). The cells were incubated at 37°C for 24 h at which time an additional 1 g/ml rPAI-1 23 was added. This procedure was maintained for 5 days. Controls were endothelial cells alone or endothelial cells to which equivalent amounts of secreted yeast cell protein were added. The cells were observed for proliferation, attachment, and tube formation.
Insertion of rPAI-1 23 Gene into pCMV/Myc/ER Eukaryotic Protein Expression Vector-The 555-bp DNA fragment coding for the truncated rPAI-1 23 was isolated from porcine aortic endothelial cell RNA by reverse transcription as described above. The cDNA was made doublestranded in a PCR reaction containing the 3Ј-porcine PAI-1 primer described and a 5Ј-PAI-1 primer to incorporate a PstI restriction enzyme site. The sequence of the 5Ј-primer is 5Ј-AACTGCAGAAGGAGC-TATGG-3Ј. The PCR conditions were as described. The enzyme-activated PAI-1 DNA fragments were ligated into a pCMV/Myc/ER eukaryotic protein expression vector (Invitrogen) for amplification in E. coli. Isolates from each ligation reaction were double-digested with restriction enzymes XbaI and EcoRI to verify the presence of the gene insert (data not shown). Positive isolates from the restriction enzyme digests were verified by sequencing.

/pCMV/Myc/ER into Endothelial Cells-
The pCMV/Myc/ER eukaryotic protein vector is designed to provide directionality of protein expression to the endoplasmic reticulum. The expressed protein can either be retained or secreted by manipulation of a stop codon in the inserted gene of interest. The rPAI-1 23 /pCMV/ Myc/ER DNA at a concentration of 25-50 g/400 l DMEM supplemented with 10% FCS was placed in a 4-mm gap cuvette and electroporated into bovine endothelial cells using a BTX Electro Cell Manipulator 600. The electroporation conditions were 550 V/capacitance and resistance, 1000 F, 13 ohm, and a charging voltage of 260 V. The pulsed cells were immediately transferred to a culture plate containing DMEM supplemented with 10% FCS. Geneticin (G418) at a concentration of 1 mg/ml was added to the cells 24 h after electroporation.
Proliferation Assay-Bovine endothelial cells stably transfected with rPAI-1 23 were plated into a 24-well culture plate at a density of 2.3 ϫ 10 4 /ml in DMEM supplemented with 10% FCS. The cells were trypsinized and counted on a hemocytometer plate at 24-h time intervals. Further verification of proliferation was performed by use of the CellTiter 96 Aqueous One Solution cell proliferation assay (Promega, Madison, WI) following the manufacturer's protocol. In brief, 1 ϫ 10 4 rPAI-1 23 transfected bovine endothelial cells were plated into a 96-well microtiter plate in DMEM containing 10% FCS. After 24 or 48 h of growth, 20 l of the manufacturer's supplied proliferation reagent was added to each well and incubated at 37°C for 1-4 h. Absorbance was read at 490 nm in an EL-311 Microplate Autoreader (Bio-Tek Instruments, Burlington, VT).
Tube Formation-Bovine aortic endothelial cells stably transfected with rPAI-1 23 were plated in 24-well plates at a concentration of 1 ϫ 10 5 cells/ml in DMEM supplemented with 10% FCS. The cells were fed every 48 h and maintained for a total of 21 days. Tube formation was observed. At day 14, a fluorescent ester probe, BCECF AM (Molecular Probes, Inc., Eugene, OR) was diluted to 10 mM in Me 2 SO. The diluted fluorescent probe was added to the culture medium to obtain a final concentration of 5 M and incubated for 60 min at 37°C. Following the incubation, the cells were washed twice in phosphate-buffered saline. The fluorescing cells were photographed with a 35-mm camera at 200ϫ magnification under a Nikon inverted microscope.
Apoptosis Detection-Bovine aortic endothelial cells and smooth muscle cells were each plated at a density of 1.0 ϫ 10 6 cells/T-75 culture flask containing DMEM supplemented with 10% fetal bovine serum. The cells were incubated for 24 h at 37°C, 5% CO 2 before adding either exogenous rPAI-1 23 (1 g/ml) or yeast supernatant protein (1 g/ml). The cells were incubated an additional 18 h at 37°C. At which time the medium containing the detached cells was collected and placed on ice while performing two HBSS washes. The medium and HBSS washes containing the detached cells were combined, counted, pelleted, and washed twice in cold phosphate-buffered saline. Fluorescein isothiocyanate-conjugated ApopNexin and propidium iodide were added to the cells following the manufacturer's protocol (Integren, Purchase, NY) (53). Adherent cells were trypsinized, resuspended in DMEM containing 10% fetal bovine serum, and incubated at 37°C for 10 min. The pelleted cells were washed twice with cold HBSS. Fluorescein isothiocyanate-conjugated ApopNexin and propidium iodide were added to the pelleted cells. Each cell fraction was analyzed separately on a FACScan (Becton Dickinson, San Jose, CA).
Migration Assay-Endothelial cells stably transfected with pCMV/ Myc/ER/rPAI-1 23 or pCMV/Myc/ER were serum-starved overnight in DMEM. The next day each transfected cell sample was counted and 1 ϫ 10 5 cells were plated onto a 24-well, 8-m pore membrane (Corning, Inc., Corning, NY). Cells were allowed to attach for 30 min. At that time, the top chamber containing the membrane-seeded cells was placed into the bottom chamber of the co-culture well that contained either DMEM supplemented with 10% FCS or DMEM supplemented with 10% fetal bovine serum and VEGF (10 ng/ml). The cells were allowed to migrate through the membrane pores for 4.5 h. The cells were fixed in 4% buffered formalin, stained with toluidine blue, and mounted on glass slides. Five fields from each of three test samples were counted at a magnification of ϫ 200.

Isolation of Porcine PAI-1 Gene Fragments-The porcine
rPAI-1 gene fragments rPAI-1 14 and rPAI-1 23 were ligated and replicated in the P. pastoris protein expression vector. The double-digested DNA isolated from selected colonies corresponded to their respective molecular weight sizes of 1137 and 555 bp as shown in Fig. 2. The sequences of human PAI-1 (huPAI-1) and porcine PAI-1 (poPAI-1) are 91% homologous and have no "obvious functional differences" (46). The rPAI-1 23 region of poPAI-1 contains 16 amino acid variations from the huPAI-1 gene. The sequence we obtained for rPAI-1 23 (data not shown) matches the corresponding sequences in the reported full-length poPAI-1 gene.
Protein Expression and Isolation- Fig. 3 shows the rPAI-1 23 protein that was isolated from the supernatant of P. pastoristransfected cells. The protein corresponds to the appropriate molecular mass of 18.5 kDa.
Chromozym Assay to Detect Plasmin Formation-The graph in Fig. 4 shows the results of the Chromozym assay and compares huPAI-1 inhibition of uPA with uPA inhibition by rPAI-1 proteins rPAI-1 23 and rPAI-1 14 (full-length PAI-1). The rPAI-1 14 protein (lane 6) has activity equivalent to 36.6 IU/g when compared with huPAI-1 activity (rows 2 and 3). Interestingly, rPAI-1 23 (row 5) was able to inhibit ϳ50% of plasmin formation when compared with the activity of the control containing 0 IU of huPAI-1 (row 4). This equates to 8.25 IU when compared with the measured 136 IU/0.44 g of huPAI-1 (row 2). The rPAI 23 protein lacks the reactive center loop shown to be necessary for binding uPA to inhibit the activation of plasmin. The numeric value ascribed to rPAI 23 in this assay is a measurement of plasmin formation, because color formation occurs when the Chromozym PL is cleaved by plasmin.
Zymography Functional Assay-The results of the zymography experiments, which enable visualization of the functionality of rPAI-1 23 (Fig. 5), clearly show that the truncated PAI-1 protein, rPAI-1 23 , inhibits the 80-kDa proteolytic plasmin (the size of the uPA-cleaved plasminogen) when incubated with either plasminogen (Fig. 5, lanes 6 -8) or uPA and plasminogen (Fig. 5, lanes 10 -12). The degree of inhibition is dependent upon the uPA concentration, but is nevertheless dramatically reduced when compared with the plasmin formed with the equivalent concentration of uPA. In all cases, a proteolytic cleavage product slightly greater than the 34-kDa molecular mass marker is visible. When the electrophoretic separation is adjusted, the proteolytic product consists of two distinct bands very close to each other that are ϳ36 -38 kDa. The degree of proteolysis associated with these fragments is substantial. The proteolytic activity associated with the protein(s) is intensified when rPAI-1 23 is first incubated with uPA and then plasminogen (Fig. 5, lanes 10 and 11) as compared with the rPAI-1 23 and plasminogen reaction mixture (Fig. 5, lanes 6 and 7). Increasing amounts (4-fold) of rPAI-1 23 incubated with uPA ultimately inhibited the proteolytic activity (lanes 8 and 12). FIG. 5. Zymography functional assay. The zymography technique allows visualization of plasmin formation. The proteolytic protein near 36 kDa that is formed as a result of rPAI-1 23 interaction with either plasminogen or plasminogen and uPA is clearly visible in lanes 6, 7, 10, and 11. The enhancement of the proteolysis is seen in lanes 10 and 11. Inhibition with increasing amounts of rPAI-1 23 is seen in lanes 8 and 12. Full-length recombinant PAI-1 is named rPAI-1 14 . An rPAI-123 protein with a reactive center loop at the carboxyl terminal is named rPAI-1 24 . In the following description of the lanes, 14 ϭ rPAI-1 14 , 24 ϭ rPAI-1 24 , and 23 ϭ rPAI-1 23  huPAI-1 is deleted in rPAI-1 23 . However, more interior Vn binding sites corresponding to huPAI-1 amino acids 109, 110, 116, and 128 -145 have been maintained. In a folded mature PAI-1 protein, the vitronectin binding regions are thought to be near each other. The loss of one vitronectin binding site could effect the stability of rPAI-1 23 . Therefore, this experiment was performed to analyze the formation of the 36-to 38-kDa proteolytic proteins when rPAI-1 23 is associated with Vn. There is no significant difference observed in the formation of the 36-to 38-kDa proteolytic fragment when rPAI-1 23 is incubated with vitronectin ( Fig. 6, lanes 9 -12). We also addressed the stability of rPAI-1 23 in various buffers (TBS, TBS/Tween 20, TBS/BSA, TBS/Tween 20/BSA) and found no difference that could be attributable to the buffer components with or without vitronectin. If vitronectin is binding rPAI-1 23 , then that binding does not alter the functionality of rPAI-1 23 .
Western Blot Analysis of the 36-kDa Proteolytic Fragment from rPAI-1 23 , Plasminogen, uPA Reaction-Reaction samples that were visualized on the functional zymogram (Fig. 5) were electrophoresed on a 4 -20% SDS, nonreducing polyacrylamide gel. The proteins contained within the polyacrylamide gel were transferred to a nitrocellulose membrane, and the Western blot was probed for plasminogen kringles 1-3 (angiostatin). The results of the angiostatin-probed membrane (Fig. 7) correspond to the results seen on the zymogram (Fig. 5). In lanes 9 and 11 on both gels, the proteolytic proteins that were inhibited by 120 nM of rPAI-1 23 are not visible on the angiostatin probed membrane. However, lanes 8 and 10 clearly show the presence of those bands at lower concentrations (30 nM) of rPAI-1 23 . It should be noted that the angiostatin-probed protein band slightly above the one that was eliminated by rPAI-1 23 remained intact. Furthermore, note the lack of a positive angiostatin band in lane 4 that contains rPAI-1 14 protein (fulllength PAI-1).
Varied Permutations for rPAI-1 23 , uPA, and Plasminogen Interactions-The varied permutations for rPAI-1 23 , uPA, and Plg interactions (Figs. 8 -10) were designed to determine how rPAI-1 23 induces the formation of proteolytic angiostatin. Fig. 8 shows the results of the experiment designed to test the effect of rPAI-1 23 on plasmin formation. In Fig. 8, lanes 3 and 4, uPA and Plg were first reacted to form plasmin before the addition of rPAI-1 23 (30 nM). The concentration of uPA in a reaction mixture was either 5 or 20 nM. In lanes 5 and 6, uPA and 30 nM rPAI-1 23 were first reacted before the addition of 33 nM of Plg. In all four lanes (lanes 3-6), the proteolytic angiostatin is formed at 36 -38 kDa. The plasmin at 80 kDa is nearly gone when compared with the plasmin in lanes 1 and 2. Regardless of the permutation and the concentration of uPA, any differences in the molecular mass or intensity of the visible proteins represented in lanes 3-6 of the zymogram are negligible.
The effects of rPAI-1 23 on uPA formation of plasmin were tested by adding rPAI-1 23 (120 nM) to uPA (5 nM) either before or after an incubation with Plg (30 nM). To ensure that all of the uPA or Plg was exhausted, a second amount of uPA (5 nM) or Plg (30 nM) was added for a third hour at 37°C. The results of those experiments are shown in Fig. 9. Lanes 5, 6, 9, and 10 represent uPA ϩ 120 nM rPAI-1 23  In the samples where uPA ϩ Plg occurred first before the addition of 120 nM rPAI-1 23 (lanes 7, 8, 11, and 12), the amount of plasmin formed remains less than in the uPA ϩ Plg control samples (lane 2). In the angiostatin-forming reaction in which  7. Western blot probed with anti-kringles 1-3 (angiostatin). Samples from the zymogram in Fig. 5 were electrophoresed on a 4 -20% SDS, nonreducing polyacrylamide gel. The analysis is of the proteins formed in the interactions of rPAI-1 23 3 and 4), the amount of plasmin is much greater than either permutation with 120 nM of rPAI-1 23 (lanes 5, 6, 9, and 10).
In experiments that produced the data shown in Fig. 10, the concentration of rPAI-1 23 (30 or 60 nM) and uPA (0.05-5.0 nM) were varied in the reaction mixes. In all samples where rPAI-1 23 was added, it was first reacted with Plg before the addition of uPA. In those samples (Fig. 10, lanes 6 -9 and 11-14) the amount of plasmin that is formed is substantially less than the plasmin produced in the controls, lanes 2, 3, and 10. In lanes 6 and 7, where the uPA is held constant at 5 nM and the rPAI-1 23 concentration is 30 nM, the amount of plasmin is slightly greater than in lanes 8 and 9, where the amount of rPAI-1 23 is 60 nM. In lanes 7 and 9, where excess Plg is added to the reaction, there appears to be a slight increase in the level of plasmin that is not seen in lane 9 that contains twice the amount of rPAI-1 23 . In lanes 11-14, where 30 nM of rPAI-1 23 was first reacted with Plg, then 0.5 nM uPA, the plasmin levels are reduced from that found in lanes 6 and 10. When twice the amount of rPAI-1 23 is added to Plg then 0.5 nM uPA, the plasmin level is near background Plg control levels in lane 1. Attention should be made to the low level of the angiostatin fragment(s) seen in lanes 6 -10. Endothelial Cell Incubation with Exogenous rPAI-1 23 -These experiments were performed to do an initial observation of the effects of rPAI-1 23 on endothelial cell growth, migration, and tube formation. The cell cultures were allowed to grow for 5 days at which time tube formation was already observable in both sets of control cells (Fig. 11A). The untreated control endothelial cells and endothelial cells incubated with protein secreted from yeast cells grew normally and became confluent (Fig. 11B). The endothelial cells grown in the presence of exogenous rPAI-1 23 displayed characteristics very dissimilar to the two controls as follows: 1) The cells easily became detached in the first 1-2 days. Analysis of the rPAI-1 23 -treated endothelial and smooth muscle cells showed 47% of the combined detached and adherent endothelial cells were apoptotic compared with 12% of apoptotic smooth muscle cells treated with the same concentration of rPAI-1 23 (Fig. 11C). 2) The rate of proliferation was reduced 30% at the end of 96 h (Fig. 11D), and the cells never reached confluency. 3) Tube formation did not occur in those endothelial cells treated with exogenous rPAI-1 23 (Fig. 11B).
Effects of rPAI-1 23 Expression from Endothelial Cells-Expression and secretion of rPAI-1 23 from endothelial cells was verified in a Chromozym Pl assay compared with endothelial cells transfected with the empty vector and endothelial cells overexpressing PAI-1 (54). The medium from each cell culture was collected and assayed. The medium containing rPAI-1 23 contained 0.04 IU/ml plasmin as compared with undetectable plasmin in the endothelial cells overexpressing PAI-1 and 0.5 IU/ml in medium from endothelial cells transfected with the empty vector (data not shown).
Tube Formation- Fig. 12 depicts tube formation in normal endothelial cells (Fig. 12A) and endothelial cells expressing rPAI-1 23 (Fig. 12B) after 14 days of growth. The tube formation is extensive in the control cells and marginal in the transfected cells. The rate at which the formation occurs is greatly reduced in the rPAI-1 23 -expressing cells. The BCECF AM probe is an esterase substrate for cell viability and is used to determine the ability of surviving cells to proliferate. Fig. 12 (C and D) verifies the viability and proliferative capability of both cells. However, it also shows the difference in the density of the rPAI-1 23 (Fig.  12D) compared with the control (Fig. 12C).
Proliferation-Endothelial cells stably transfected with rPAI-1 23 were tested for cell proliferation. The differences in proliferation of endothelial cells transfected with empty vector are compared with proliferation of rPAI-1 23 -transfected endothelial cells in Fig. 12E. During the initial 24-h growth period, the cells secreting rPAI-1 23 grew at a slightly faster rate. However, that difference was overtaken by the control endothelial cells in the second 24-h period. The control cells continued to proliferate faster than the rPAI-1 23 -transfected cells with a 170% greater rate of proliferation by the 96-h time point. All cells were maintained in culture for an additional 10 days to determine whether the control cells had reached plateau growth phase while the rPAI-1 23 transfectants continued to proliferate. At 336 h, the control cells had increased in number by 288%, while the rPAI-1 23 -transfected cells had increased their number by 201%. The net result was greater than twice the number of control cells when compared with the rPAI-1 23 cells at 336 h.
Migration-The numbers of migrated endothelial cells transfected with rPAI-1 23 is vastly reduced when compared with the control cells that contain the empty vector (Fig. 12F). However, the numbers of migrating control cells treated with VEGF increase 300%, and the rPAI-1 23 -transfected cell migration rate is reduced by one half. The effect of VEGF on migration of endothelial cells transfected with rPAI-1 23 is a 20-fold decrease when compared with the empty vector control. DISCUSSION Angiostatin, a 38-kDa cleavage product of plasminogen, functions as a regulator of angiogenesis by inhibiting endothelial cell proliferation, migration, and tube formation (55,56). It is known that angiostatin is a cleavage product of plasminogencontaining kringle domains 1-3 (57). However, the angiostatin- FIG. 9. Binding of rPAI-1 23 to uPA. The effects of rPAI-1 23 on uPA formation of plasmin were tested by adding either 30 or 120 nM rPAI-1 23 to 5 nM uPA either before or after an incubation with 33 nM Plg. To ensure that all Plg or uPA was exhausted, a second amount of either one was added for an additional hour at 37°C. In the following description of the lanes, 23 ϭ rPAI-1 23  related mechanism(s) for inhibition of endothelial cell proliferation, migration, and tube formation remains unknown. Nor is it known which proteases convert plasminogen to angiostatin (57).
The data that we have presented show that a truncated PAI-1 protein, rPAI-1 23 , mediates the formation of proteolytic angiostatin-like protein fragments of an approximate 36-to 38-kDa molecular mass and inhibits 80-kDa plasmin. The mechanism by which this occurs is 2-fold. The first and seem- FIG. 11. Endothelial cell incubation with exogenous rPAI-1 23 . The function of angiostatin is to inhibit endothelial cell proliferation, migration, and tube formation. The series of experiments in A-E were performed to do an initial observation of exogenous rPAI-1 23 effects on endothelial cell growth, attachment/detachment, and tube formation. After 5 days in culture, the cells treated with exogenous rPAI-1 23 displayed obvious differences in proliferation, morphology, tube formation, and detachment. A and B, tube formation. A represents the untreated control cells displaying confluency and tube formation; B shows rPAI-1 23 -treated endothelial cells. Their rate of proliferation is reduced such that they did not reach confluency by day 5 and, therefore, did not display tube formation. C, apoptosis was evaluated after observing detachment of endothelial cells treated with rPAI-1 23 . The results of the annexin V binding assay showed that 65% of the combined rPAI-1 23 detached and adherent cells were undergoing apoptosis, whereas only 15% of the detached and adherent control cells were apoptotic. This effect was not detectable in smooth muscle cells. D, proliferation was determined by counting the cells in each test sample at various time points. The results were verified by performing a CellTiter 96 Aqueous One Solution cell proliferation assay. The rate of proliferation of rPAI-1 23 -treated cells was reduced ϳ30% in 96 h as compared with control endothelial cells that were untreated or to which yeast protein was added to culture medium (D). ingly preferred mechanism shown in Figs. 8 and 13 is due to rPAI-1 23 cleavage of plasmin into 36-and 38-kDa fragments. This takes place regardless of the permutation for combining rPAI-1 23 , uPA, and plasminogen. Increasing amounts of uPA enhance the formation of angiostatin simply by producing more plasmin for rPAI-1 23 to cleave. However, increasing amounts of rPAI-1 23 inhibit the formation of proteolytic angiostatin by reducing the amount of 80-kDa plasmin that is formed. This occurs in a second mechanism where rPAI-1 23 binds either uPA (Fig. 9) or plasminogen (Fig. 10), thus reducing the number of uPA and plasminogen interactions, which in turn reduces the amount of plasmin that is produced (Figs. 9, 10, and 14). Even though the number of rPAI-1 23 molecules exceeds that of uPA by 6-to 24-fold, plasmin is still formed at approximately onethird the amount normally formed in the absence of rPAI-1 23 . This suggests that plasminogen is able to dissociate uPA from rPAI-1 23 , making uPA available for binding plasminogen. Given sufficient quantities of rPAI-1 23 , increasing numbers of plasminogen become bound to rPAI-1 23 and are unavailable for binding uPA. Likewise, increasing amounts of rPAI-1 23 have an enhanced opportunity to bind both uPA and plasminogen. The rPAI-1 23 protein interaction with uPA and plasminogen inhibits the formation of the normal 80-kDa plasmin and the formation of the 36-to 38-kDa proteolytic angiostatin-like fragments. These data strongly suggest that rPAI-1 23 exposes binding sites for uPA, plasminogen, and plasmin that provide the truncated inhibitor with two pathways for regulating plasmin.
When rPAI-1 23 protein was added exogenously to endothelial cells, proliferation was decreased by ϳ30% in the initial 48 h and tube formation was inhibited. Endothelial cells transfected with rPAI-1 23 contained within pCMV/Myc/ER vector displayed the same profile of reduced proliferation and tube formation. Moreover, the rPAI-1 23 -transfected cells show a 20-fold reduction in migration in the presence of VEGF when compared with the control cells. Additionally, we demonstrate endothelial cell specificity in the apoptosis assay where we determined that 47% of the rPAI-1 23 -treated endothelial cells underwent apoptosis within the first 24 -48 h, while only 12% of the rPAI-1 23 -treated smooth muscle cells were apoptotic. The 36-kDa proteolytic fragment that we have identified exhibits those properties inherent to angiostatin.
The significant apoptotic index in rPAI-1 23 -treated cells provides a potential explanation for the reduction in proliferation of rPAI-1 23 -treated and -expressing endothelial cells. The reduction in proliferation in turn reduces the tube formation, because the tubes do not begin to form until endothelial cells reach quiescence/confluence. The apoptotic and proliferation results that we have documented are in concurrence with results reported by Claesson-Welsh et al. (58).
The reactive loop containing the uPA binding site needed for inactivation of uPA was deleted from rPAI-1 23 , thus releasing the strained loop conformation as well as the 59 amino acids on the amino terminus containing important binding sites. By all definitions of "active" PAI-1, rPAI-1 23 would not be expected to have activity. However, we have shown that this truncated protein is capable of inhibiting the 80-kDa plasmin by two mechanisms. First, it cleaves plasmin to produce angiostatinlike proteolytic proteins. The proteolytic activity of the 34-kDa angiostatin is ultimately inhibited by increasing molecules of rPAI-1 23 that are available for binding uPA and/or plasminogen. In this second mechanism, rPAI-1 23 reduces the numbers of uPA⅐Plg interactions, thus reducing the amount of plasmin production. This truncated PAI-1 conformation appears to be exposing sites that participate in a functional role for PAI-1 in generating angiostatin fragments from plasmin. It has been suggested that, in the transition between latent and active states, such sites may become available for interaction with other regulatory proteins involved with proteolysis, fibrinolysis, adhesion, or angiogenesis (48, 58 -60).
There are reports that metalloproteinases are responsible (61-65) for cleavage of plasminogen into angiostatin. In some reports, it has been demonstrated that neoplastic cells produce an enzyme that converts plasminogen to angiostatin (66,67). In other findings, metalloproteinases expressed by tumor-associated macrophages produce angiostatin fragments in tumor cells (62). Falcone et al. (68) recently reported that angiostatin fragments (ϳ36 kDa and 48 kDa) are formed as "by-products" FIG. 14. Inhibition of angiostatin by rPAI-1 23 . The truncated PAI-1 protein, rPAI-1 23 , has a binding affinity for both uPA and plasminogen. Plasminogen can dissociate uPA from rPAI-1 23 . The affinity of uPA for plasminogen enables plasmin to be formed at a reduced level. Increasing molecules of rPAI-1 23 reduce the potential of uPA and plasminogen interactions.
FIG. 13. The mechanism by which rPAI-1 23 inhibits plasmin and induces angiostatin. Varied permutations for rPAI-1 23 , uPA, and Plg interactions were tested to determine whether the order in which 30 -60 nM rPAI-1 23 , uPA, and Plg were added to a reaction mixture altered the appearance of the proteolytic angiostatin-like product. Regardless of the order, 30 -60 nM of rPAI-1 23 is able to cleave plasmin. The cleavage product is proteolytic angiostatin. Increasing the amount of uPA in the reaction enhances the formation of angiostatin by providing more plasmin for rPAI-1 23 to cleave. The affinity of uPA for plasminogen is greater than for rPAI-1 23 , because either permutation results in comparable amounts of angiostatin. of membrane-bound plasmin regulation in normal cells in the absence of metalloproteinases. They also showed that the Ͻ36-kDa fragment has enzymatic activity but does not bind to the cell surface. The larger fragment also has enzymatic activity and is able to inhibit endothelial cell proliferation (68). There are reports that urokinase and sulfhydryl donors are responsible for the conversion of plasminogen into angiostatin (69). A reductase secreted by fibrosarcoma and Chinese hamster ovary cells reduced plasmin disulfide bonds to generate angiostatin (67,70). O'Mahony et al. (71) have shown that transforming growth factor-␤1 inhibits angiostatin from forming in human pancreatic cancer cells. The addition of exogenous PAI-1 inhibited the cleavage of plasminogen into angiostatin (71). Collectively, these data suggest that cleavage of angiostatin from plasminogen occurs by more than one mechanism. We show a role for PAI-1 in the induction of proteolytic angiostatin from plasmin. We are pursuing the in vivo events that could result in a PAI-1 conformation having the function we have demonstrated, because this would be an important finding relative to tumorigenesis and vascular disease.