Angiostatin generation by cathepsin D secreted by human prostate carcinoma cells.

Angiostatin, a potent endogenous inhibitor of angiogenesis, is generated by cancer-mediated proteolysis of plasminogen. The culture medium of human prostate carcinoma cells, when incubated with plasminogen at a variety of pH values, generated angiostatic peptides and miniplasminogen. The enzyme(s) responsible for this reaction was purified and identified as procathepsin D. The purified procathepsin D, as well as cathepsin D, generated two angiostatic peptides having the same NH(2)-terminal amino acid sequences and comprising kringles 1-4 of plasminogen in the pH range of 3.0-6.8, most strongly at pH 4.0 in vitro. This reaction required the concomitant conversion of procathepsin D to catalytically active pseudocathepsin D. The conversion of pseudocathepsin D to the mature cathepsin D was not observed by the prolonged incubation. The affinity-purified angiostatic peptides inhibited angiogenesis both in vitro and in vivo. Importantly, procathepsin D secreted by human breast carcinoma cells showed a significantly lower angiostatin-generating activity than that by human prostate carcinoma cells. Since deglycosylated procathepsin D from both prostate and breast carcinoma cells exhibited a similar low angiostatin-generating activity, this discrepancy appeared to be attributed to the difference in carbohydrate structures of procathepsin D molecules between the two cell types. The seminal vesicle fluid from patients with prostate carcinoma contained the mature cathepsin D and procathepsin D, but not pseudocathepsin D, suggesting that pseudocathepsin D is not a normal intermediate of procathepsin D processing in vivo. The present study provides evidence for the first time that cathepsin D secreted by human prostate carcinoma cells is responsible for angiostatin generation, thereby causing the prevention of tumor growth and angiogenesis-dependent growth of metastases.

Angiogenesis is an important process of new microblood vessel formation under both physiological and pathological conditions. Because angiogenesis is essential for tumor growth and metastases, the discovery of angiogenesis inhibitors is consid-ered to represent a potential approach to treat cancer. It has so far been suggested that the metastatic growth of carcinoma cells is associated with induced angiogenesis and that the rate of tumor growth correlates with the formation of new blood vessels (1). Higher vascular tumors can be stimulated by angiogenic growth factors released by tumors and host tissues, thereby producing metastases at a higher rate than less angiogenic tumors. On the other hand, it has been known that certain types of primary tumors, such as murine Lewis lung carcinoma and human prostate carcinoma cell lines (PC-3), can produce angiogenic inhibitors in order to suppress the tumor growth at metastatic sites, besides angiogenic growth factors (2,3). Therefore, the growth rate of a tumor may be the result of the balance between positive and negative angiogenic effects.
Recently, human prostate carcinoma cells (PC-3) have been shown to express an enzymatic activity in the serum-free conditioned medium that can generate bioactive angiostatin from plasminogen (12), in which plasminogen is suggested to be initially converted to proangiogenic plasmin by cleavage of the Arg 560 -Val 561 bond by plasminogen activators such as urokinase-and tissue-type plasminogen activators and then converted to angiostatin by autoproteolysis of plasmin in the presence of free sulfhydryl donors. The importance of plasmin in cancer-mediated conversion of plasminogen to angiostatin was also described in human fibrosarcoma cells (HT1080) (14).
However, it remains to be determined whether angiostatin production by prostate carcinoma cells is totally dependent on the plasmin action.
Here we report that procathepsin D produced directly by human prostate carcinoma cells is responsible for the generation of angiostatic peptides comprising kringle domains 1-4 of plasminogen and that the isolated angiostatic peptides can inhibit endothelial cell proliferation in vitro and bFGF-induced angiogenesis in vivo. In addition, we address the question about the difference of procathepsin D molecules between human prostate and breast carcinoma cells in the ability of angiostatin generation.

EXPERIMENTAL PROCEDURES
Materials-Human cathepsin D and pepsin were purchased from Sigma. Porcine pancreatic elastase was from Elastin Products Company, Inc. A chromogenic peptide substrate (S-2251) was from Chromogenix AB, Sweden. Lysine-Sepharose 4B, Sephadex G-75, Superdex G-200, Phenyl-Sepharose HP, and hydroxyapatite were from Amersham Pharmacia Biotech. Angiostatin consisting of the first three kringles of plasminogen was from TechnoClone GmbH (Vienna, Austria). Human cathepsin E was prepared from human erythrocyte membranes as described previously (15). Polyclonal antibodies against human cathepsin D and human plasminogen were from Calbiochem and Dako Chemical Co., respectively. The monoclonal antibody against human miniplasminogen was from American Diagnostica Inc. Human prostate carcinoma (PC-3) cells and human breast carcinoma (MCF7) cells were obtained from Dainippon Pharmaceutical Co. (Tokyo). Human umbilical endothelial cells were from Clonetic Co. Seminal vesicle fluid was obtained from two patients with prostate carcinoma in the Department of Urology, Kyushu University Hospital (Fukuoka, Japan), under an Institutional Review Board-approved protocol with subjects providing informed consent.
Cell Culture-PC-3 cells were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum (FCS), penicillin G (100 units/ml), and streptomycin (100 g/ml). MCF7 was cultured in Dulbecco's modified Eagle's medium supplemented with 10% FCS, 1% nonessential amino acids, 1 mM sodium pyruvate, and the antibiotics. Human omental microvascular endothelial cells that were isolated from normal omental tissue removed during surgery as described previously (16) were cultured in M-199 medium containing 10% FCS. Human umbilical endothelial cells were cultured as recommended by the suppliers. All cell cultures were performed at 37°C in 95% air, 5% CO 2 . Culture media were prepared from cultures of various cells. Confluent cell monolayers were washed twice with phosphate-buffered saline (PBS) and cultured in each culture medium for 18 h. The culture media were collected and filtrated through a 0.45-m pore size membrane (Millipore Corp.).
Isolation of Plasminogen and Its Related Fragments-Human plasminogen was purified from the pooled human plasma in the presence of 3 mM benzamidine-HCl, 5 mM EDTA, aprotinin (100 units/ml), and soybean trypsin inhibitor (10 g/ml). The pooled plasma was first subjected to affinity chromatography on lysine-Sepharose 4B in 50 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl and 5 mM EDTA as described (17), followed by gel filtration on Sephacryl S-200. The fragment of the lysine binding site I (LBS I) including kringles 1-3 was purified from human plasminogen according to the method of Sottrup-Jensen et al. (18). The monospecific antibody for LBS I was prepared from the antisera raised in rabbits by affinity chromatography on a protein A-Sepharose column followed by an LBS I-conjugated Sepharose 4B column.
Purification of Procathepsin D-The culture medium from PC-3 cells was subjected to a CM-Sepharose column (2.5 ϫ 30 cm) equilibrated with 10 mM Tris-HCl, pH 7.4, containing 25 mM NaCl (buffer A). Solid ammonium sulfate was added to the nonadsorbed fraction to give a final concentration of 1.0 M. After centrifugation, the supernatant was subjected to a phenyl-Sepharose 4B column (5.0 ϫ 10 cm) equilibrated with buffer A containing 1.0 M ammonium sulfate. After washing with the same buffer, proteins bound to the column were eluted by the linear gradient of ammonium sulfate ranging from 1.0 to 0 M. The angiostatingenerating activity was eluted at a concentration of 5 mM ammonium sulfate. The active fractions were pooled and dialyzed against buffer A. The dialysate was concentrated by ultrafiltration on an Amicon Diaflo YM10 membrane and then subjected to Superdex G-200 (1.0 ϫ 60 cm) chromatography equilibrated with buffer A. The activity was eluted at the position corresponding to an apparent molecular mass of 50 -60 kDa. After concentration and dialysis against buffer A, the active frac-tion was subjected to a Q-Sepharose HP column (0.5 ϫ 6.5 cm) equilibrated with buffer A. The enzyme activity was eluted by the gradient of NaCl ranging from 0 to 1.0 M. The active fractions were pooled and dialyzed against 20 mM sodium phosphate buffer, pH 7.2, and then subjected to a hydroxyapatite column (1.0 ϫ 10 cm) equilibrated with the same buffer. The activity was eluted by the gradient of sodium phosphate ranging from 20 to 200 mM. The enzyme responsible for the generation of angiostatin was eluted at a concentration of about 50 mM sodium phosphate.
Purification of Angiostatic Peptides-The angiostatic peptides were purified from human plasminogen fragments hydrolyzed by the culture medium of PC-3 cells. Human plasminogen (10 mg) was incubated with the culture medium from PC-3 cells (0.8 mg) in 100 ml of citrate/ phosphate buffer, pH 4.0, at 37°C overnight. Then the mixture was subjected to affinity chromatography on a lysine-Sepharose 4B column equilibrated with 50 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl and 5 mM EDTA. Angiostatic peptides consisting of the first four kringle domains of plasminogen were eluted with 0.1 M ⑀-aminocaproic acid in the same buffer. The peptides were further purified by gel filtration on a Sephadex G-75 column. The main peak fractions were pooled, dialyzed against PBS, and stored at Ϫ80°C until use.
Assay Procedures-The angiostatin-generating activity was assayed by enzyme-linked immunosorbent assay with two types of antibodies against human plasminogen: the antibodies to human plasminogen and the monospecific antibodies to LBS I. Angiostatin-generating activity was evaluated by assaying remaining plasminogen molecules after incubation between a known amount of plasminogen and a fraction containing the angiostatin-generating activity for a certain period. Microtiter wells were coated with 150 l of 50 mM carbonate-buffer, pH 9.6, containing the monoclonal antibody to miniplasminogen (5 mg/ml) at 4°C overnight. Unbound antibody was drained off, and the wells were blocked by 300 l of bovine serum albumin (5 mg/ml) in the coating buffer for 1 h at room temperature. The wells were washed four times with PBS containing 0.05% Tween 20. Meanwhile, samples (50 l) were mixed with 50 l of plasminogen (1.0 mg/ml), 100 l of citrate/ phosphate buffer (pH 4.0) and incubated at 37°C for 1 h. The reaction mixture was then diluted 2500-fold with PBS containing aprotinin (10 units/ml), l% bovine serum albumin, and 0.05% Tween 20. The diluted sample (100 l) was added to the coated wells and incubated at 37°C for 1 h. After washing, 50 l of horseradish peroxidase-conjugated antihuman LBS I antibody (20 ng/ml) was added to the wells and incubated at 37°C for 1 h. After washing, 100 l of freshly prepared o-phenylendiamine solution was added and allowed to stand in the dark at room temperature for 20 min. The reaction was stopped by adding 4 N sulfuric acid. The absorbance at 490 nm was determined promptly. 1 unit was defined as the enzyme activity that degrades 20 g of plasminogen for 1 h at 37°C.
Plasmin activity was assayed by the method of Friberger et al. (19). Briefly, samples were mixed with 50 l of urokinase-type plasminogen activator (20 units/ml) in 50 mM Tris-HCl, pH 7.4, containing 10 mM NaCl. After a 15-min incubation at 37°C, 50 l of the chromogenic peptide substrate S-2251 was added to the mixture and incubated for 10 min at 37°C. The reaction was stopped by adding 2% citric acid solution. The liberated p-nitroaniline was determined at 405 nm promptly.
Immunoblotting-Proteins were electroblotted on polyvinylidene difluoride filters. After blocking in 4% skim milk in 10 mM Tris-HCl and 0.15 M NaCl at pH 7.4 containing 0.1% Tween 20 for 2 h at room temperature, the membranes were incubated in 10 mM Tris-HCl-saline, pH 7.4, containing the first antibodies (10 g/ml). After washing, the membranes were incubated for 1 h with the second antibody solution containing horseradish peroxidase-conjugated goat anti-rabbit IgG or rabbit anti-mouse IgG antibody (diluted 1:1000 -1:4000). Enhanced chemiluminescence (PerkinElmer Life Sciences) was used for visualization.
NH 2 -terminal Amino Acid Sequence Analyses-The proteins that had been separated by SDS-PAGE were electrophoretically transferred from the gels onto polyvinylidene difluoride membranes. The protein bands on the membranes were stained with Coomassie Brilliant Blue R-250, excised, and subjected to the NH 2 -terminal amino acid sequence analysis by an automatic protein/peptide sequencer (Applied Biosystems model 477A).
Endothelial Proliferation Assay-The proliferation assay of human omentum microvascular endothelial cells was performed by the method of O'Reilly et al. (2), with a slight modification. Briefly, the cells (5000 cells) were plated onto gelatinized culture plates (24 wells) and incubated in M199 containing 10% FCS, penicillin G (100 units/ml), and streptomycin (100 g/ml) at 37°C for 3 h in 5% CO 2 . The medium was replaced with 0.25 ml of serum-free medium, and further incubated overnight. Then, the samples were added to the wells. After a 20-min incubation, the culture medium containing 10% FCS (0.25 ml) was added to the wells and incubated for 72 h. After treatment with trypsin-EDTA, the dispersed cells were counted using a Coulter counter.
bFGF-induced Mouse Cornea Angiogenesis Assay-Cornea angiogenesis assay was performed using 6-week-old male BALB/c mice, as described previously (2). Briefly, samples (2.0 mg/ml) and bFGF (1.0 mg/ml, Genzyme Corp., Cambridge, MA) were mixed with an equal volume of 12% poly-2-hydroxylethylmethacrylate polymer (Hydron, Interferon Sciences, Inc., New Brunswick, NJ). 1 l of the mixture was pipetted twice onto Teflon sheets. Pellets were air-dried for 1-2 h in a laminar flow hood, cut into nine pellets, and stored overnight at 4°C. A pocket was made within the cornea stroma, and pellets were positioned 1.0 -1.2 mm from the limbs. Corneas were examined for 6 days after implantation, and the angiogenic response was quantitated using the method of Chen et al. (20).

Angiostatin Generation from Human Plasminogen by Secreted Proteinases by Human Prostate Carcinoma Cells-
The culture medium of human prostate carcinoma cells (PC-3) was incubated with human plasminogen at various pH values at 37°C overnight. The reaction products were analyzed by SDS-PAGE under nonreducing conditions and immunoblotting (Fig.  1). The maximal degradation was observed at pH 4.0 -5.0, where three protein bands with apparent molecular masses of 35, 42, and 45 kDa were observed (Fig. 1A). Immunoblot analysis revealed that the 42-and 45-kDa peptides immunoreacted with the antibodies to an elastase-derived fragment (LBS I) comprising kringles 1-3 of plasminogen, whereas the 35-kDa peptide immunoreacted with the monoclonal antibody to miniplasminogen (including the kringle 5 and serine protease domains of plasminogen) (Fig. 1B). Plasminogen was also degraded by the medium at pH values of 6.0 -8.0, but electrophoretic mobilities of the generated peptides were slightly different from those observed at acidic pH. In addition, the extent of the generation of angiostatic peptides at pH 6.0 -8.0 was less than 10% relative to that at pH 4.0 -5.0. The generation of angiostatic peptides at acidic pH was strongly inhibited by pepstatin A (1 M) but not by other protease inhibitors (Table  I). These results indicate that the proteolytic activity responsible for the generation of angiostatic peptides by the culture medium of PC-3 cells at acidic pH is attributable to the aspartic proteinase family and that the 42-and 45-kDa angiostatic peptides and the 35-kDa miniplasminogen are generated by its enzymatic action.

Identification and Purification of Procathepsin D Secreted by Human Prostate Carcinoma Cells as an Angiostatin-generating
Enzyme-To characterize the enzyme activity responsible for the generation of angiostatic peptides, the corresponding enzyme was purified from the culture medium of PC-3 cells (Table  II). Phenyl-Sepharose chromatography provided a rapid and efficient method for removing significant amounts of contaminating proteins. About 850-fold purification over the culture medium was accomplished by this step. The complete elimination of contaminating proteins was accomplished by hydroxyapatite chromatography. The final preparation gave single protein bands with an apparent molecular mass of about 44 kDa when examined by SDS-PAGE under reducing and nonreducing conditions ( Fig. 2A). Immunoblot analysis revealed that these bands were immunoreactive with antibodies to human cathepsin D (Fig. 2B, lanes 1 and 2). In addition, the NH 2terminal amino acid sequence of this protein was found to be LVRIPL, which was consistent with the NH 2 -terminal amino acid sequence of procathepsin D starting with the residue Leu 21 of the predicted amino acid sequence of preprocathepsin D (21). Human mature cathepsin D had an apparent molecular mass of 30 kDa in SDS gels (Fig. 2B, lanes 3 and 4), and its NH 2 -terminal amino acid sequence was found to be GPIPEVL, which corresponded to the amino acid sequence starting at the 65th Thr residue of the predicted amino acid sequence of procathepsin D. These data indicate that the enzyme responsible for the generation of angiostatic peptides released by PC-3 cells is procathepsin D.
To determine whether the purified enzyme can actually generate the angiostatic fragments from plasminogen and also to compare its angiostatin-generating activity with that of mature cathepsin D, both procathepsin D and cathepsin D were incubated with plasminogen (enzyme/substrate ratio of 1:100) at 37°C and pH 4.0 overnight. The reaction products were analyzed by SDS-PAGE under reducing and nonreducing conditions. Not only cathepsin D but also procathepsin D could generate the 45-and 42-kDa angiostatic peptides and the 35-kDa miniplasminogen, which corresponded to the 55-and 52-kDa angiostatic peptides and the 32-kDa miniplasminogen under reducing conditions, respectively (Fig. 2C).  Angiostatin Generation Requires Conversion of Procathepsin D to Pseudocathepsin D-Angiostatic peptides and miniplasminogen were generated time-dependently by incubation of plasminogen with procathepsin D at pH 4.0 and 37°C for up to 1 h (Fig. 3B). The 45-, 42-, and 35-kDa peptides all were resistant to further degradation by prolonged incubation up to 4 h (not shown). These results were consistent with those obtained with the culture medium of PC-3 cells. The generation of these peptides was in fair agreement with the conversion of procathepsin D to the 41-kDa form (Fig. 3A). Within a 15-min incubation, about 50% of procathepsin D was converted to the 41-kDa form. The complete conversion was accomplished within a 30-min incubation. The NH 2 -terminal amino acid sequence of the 41-kDa protein was found to be IAKGP, which corresponded to the NH 2 -terminal amino acid sequence of pseudocathepsin D starting with the 47th Ile in the predicted amino acid sequence of preprocathepsin D. This conversion was accelerated as the procathepsin D concentration was increased and was inhibited by pepstatin A. These results indicated that the 41-kDa protein was generated by a proteolytic cleavage between Leu 46 and Ile 47 in the propart, predominantly through intermolecular reaction. No protein bands corresponding to mature cathepsin D were observed.
Characterization of the Process of Angiostatin Generation by Procathepsin D-The time-dependent profile for the generation of angiostatic peptides from plasminogen was determined by a lower concentration of procathepsin D (enzyme/substrate ratio of 1:250) (Fig. 4). The results indicated that the generation of angiostatic peptides consisted of multiple steps. Intermediates with higher molecular weight masses generated in the earlier stages of incubation (up to 60 min) appeared to be finally converted to the triple peptides of 35, 42, and 45 kDa under nonreducing conditions. The NH 2 -terminal amino acid sequences of 42-and 45-kDa fragments were found to be FEKKVYLSE, which corresponded to the NH 2 -terminal amino acid sequence of angiostatin starting with the 75th Phe residue of the predicted amino acid sequence of plasminogen, suggesting that two forms of angiostatin are produced by the different carbohydrate modification of angiostatin. On the other hand, the NH 2 -terminal amino acid sequence of the 40-kDa peptide revealed under reducing conditions was PDVETPSEF, which corresponded to the sequence starting with the 452th Pro residue of the predicted amino acid sequence of plasminogen. The NH 2 -terminal amino acid sequence of the 40-kDa peptide was consistent with that of the 35-kDa peptide under nonreducing conditions. Furthermore, the relationship between the 32-and 40-kDa peptides was determined. The 40-kDa protein was first generated within 15 min after incubation, and its level was increased up to 1 h. Then it was gradually decreased during the period up to 2 h with increasing amounts of 32-and 10-kDa peptides. The NH 2 -terminal amino acid sequence of the 32-kDa peptide was identical with that of the 40-kDa peptide. The NH 2 -terminal amino acid sequence of the concomitant generated 10-kDa peptide was found to be AQLPYIE, which corresponded to the sequences starting with the 670th Ala residue of the predicted amino acid sequence of plasminogen. The results indicated that the 40-kDa peptide was converted to the 32-kDa The angiostatin-generating activity was determined by enzyme-linked immunosorbent assay as described under "Experimental Procedures." 1 unit was defined as the amount of enzyme required to degrade 20 g of plasminogen at 37°C for 1 h.   fragment by cleavage of the Glu 699 -Ala 700 bond. Based on these data, the sequential cleavage sites of plasminogen by pseudocathepsin D were summarized in Fig. 4C. Plasminogen appears to be cleaved by pseudocathepsin D at the Leu 451 -Pro 452 bond followed by cleavage at the Leu 74 -Phe 75 bond and finally at the Glu 699 -Ala 700 bond.
When plasminogen was incubated with procathepsin D at various enzyme/substrate ratios, the 55-and 52-kDa angiostatic peptides and the 32-kDa peptide were rapidly generated at an enzyme/substrate ratio below 1:20 (Fig. 5A). At ratios between 1:40 and 1:160, the 40-kDa intermediate was produced prior to the 32-kDa peptide formation, besides the angiostatic peptides. The effect of pH on the generation of angiostatic peptides by procathepsin D was examined at the enzyme/substrate ratio of 1:20 (Fig. 5B) and 1:10 (Fig. 5C). The efficient angiostatin generation was observed at pH values between 3.5 and 5.5 (Fig. 5B). Although angiostatin generation was relatively slow at pH values above 6.0, angiostatic peptides, especially the 55-kDa peptide, were clearly produced by this enzyme even at pH 6.8 (Fig. 5C). Therefore, we concluded that pseudocathepsin D produced by procathepsin D autoactivation could generate angiostatic peptides at a wide range of pH values between 3.0 and 6.8.

Generation of Angiostatic Peptides by Cathepsin E and Pepsin-
To determine whether other relevant aspartic proteinases have the ability to generate angiostatic peptides, plasminogen was treated with human cathepsin E and pepsin, as well as cathepsin D, under the same conditions as procathepsin D (Fig.   6). Similarly to pseudocathepsin D and cathepsin D, both cathepsin E and pepsin efficiently generated the angiostatic peptides. However, cathepsin E and pepsin further degraded the angiostatic peptides into smaller fragments by prolonged incubation up to 4 h, whereas cathepsin D, as well as pseudocathepsin D, did not cause the degradation of angiostatic peptides for 4 h of incubation.
Inactivation of Plasmin by Pseudocathepsin D-Plasmin is known to play a key role in producing angiogenic factors such as MMPs and cytokines (22). Cancer invasion is thought to be initiated by the activation of plasminogen and procollagenase cascade (23,24). In addition, plasmin has been shown to be important in cancer-mediated conversion of plasminogen to angiostatin (14,22,25). These results thus suggest that inactivation of plasmin modulates angiogenesis in tumors. If the Glu 699 -Ala 700 bond in the serine protease domain of plasminogen is cleaved by pseudocathepsin D, the plasminogen-plasmin converting activity may be lost. To examine this possibility more directly, the plasmin activity was determined by use of plasminogen activator after incubation of plasminogen with procathepsin D at pH 4.0 for 37°C (Fig. 7). The plasmin activity was lost time-dependently, indicating that angiostatin generation by pseudocathepsin D results in the concomitant loss of plasminogen-plasmin converting activity.

Inhibition of Human Endothelial Cell Proliferation in Vitro and b-FGF-induced Angiogenesis in Mouse Cornea in Vivo by
Purified Angiostatic Peptides-Angiostatic peptides generated by pseudocathepsin D were purified from procathepsin D-treated plasminogen by use of lysine-Sepharose and Sephadex G-75 columns. The NH 2 -terminal amino acid sequence analyses revealed that purified angiostatic peptides had the same NH 2 -terminal amino acid sequences and consisted of kringles 1-4 of plasminogen, suggesting that the difference in their apparent molecular masses in an SDS gel is due to the heterogeneity in the carbohydrate modification. The proliferation of human microvascular endothelial cells was inhibited by these peptides in a dose-dependent manner (Fig. 8A). A similar result was obtained by angiostatin generated by elastase, which was known to consist of kringles 1-3 of plasminogen.
The effect of the purified angiostatic peptides on bFGF-induced angiogenesis in mouse cornea was also investigated (Fig.  8B). Hydron pellets containing 50 ng of bFGF with or without 50 ng of angiostatic peptides were implanted into the cornea of three mice. 6 days after implantation, bFGF alone clearly induced angiogenesis in corneas, whereas bFGF containing 100 ng of angiostatic peptides did not induce the increased angiogenesis. The bFGF-induced angiogenesis was inhibited by angiostatic peptides generated by pseudocathepsin D in a dosedependent manner. The angiogenesis was completely inhibited by 100 ng of angiostatic peptides.
Comparison of Procathepsin D Molecules Secreted by Human Prostate and Breast Carcinoma Cells-It is well known that procathepsin D is abundantly released by human breast carcinoma cells (26 -29). We analyzed whether the culture medium of human breast carcinoma cells (MCF7) could exhibit the angiostatin-generating aspartic proteinase activity found in PC-3 cells. Surprisingly, the angiostatin-generating activity was barely detectable in the culture medium from MCF7 cells, similarly to the control medium from human umbilical vascular endothelial cells (Fig. 9). To determine whether this was due to the inability of procathepsin D secreted by MCF7 cells to generate angiostatic peptides, procathepsin D was also purified from the culture medium of MCF7 cells. The final preparation of procathepsin D gave a single protein band by SDS-PAGE, but its mobility was slightly faster than that from PC-3 cells (Fig. 10A). After N-glycosidase F treatment, both procathepsin D molecules showed the same mobility, suggesting that both enzymes are different in carbohydrate structures.
When MCF7-derived procathepsin D was incubated with plasminogen at pH 4.0 and 37°C, angiostatic peptides were scarcely generated within 12 h incubation (Fig. 10B). Only a small amount of angiostatic peptides was produced by this enzyme after 24 h of incubation. Interestingly, N-glycosidase F-treated procathepsin D from PC-3 cells showed a marked decrease in the angiostatin-generating activity. These data suggest that the structural difference of procathepsin D molecules is associated with their ability to generate angiostatic peptides.
Angiostatin could be generated by procathepsin D secreted by human prostate carcinoma PC-3 cells even at pH values around neutrality (Fig. 5, B and C). However, it remains to be answered whether procathepsin D is secreted extracellularly in vivo in patients with prostate carcinoma. We thus examined if procathepsin D was detected in the seminal fluid from patients with prostate carcinoma. Although PC-3 cells secreted procathepsin D alone in the culture medium in vitro, the seminal vesicle fluid from patients with prostate carcinoma contained not only procathepsin D but also cathepsin D (Fig. 11). When treated at pH 4.0 and 37°C for 1 h, procathepsin D from PC-3 cells and the seminal vesicle fluid completely converted to pseudocathepsin D. The molecular mass of cathepsin D was not changed by the treatment. These results suggest that angiostatic peptides may be generated in vivo by both cathepsin D and procathepsin D in patients with prostate carcinoma.

DISCUSSION
This study provides evidence that cathepsin D and/or pseudocathepsin D are responsible for the angiostatin-generating activity secreted by human prostate carcinoma cells. In vitro studies showed that procathepsin D purified from the culture medium of PC-3 cell could generate angiostatic peptides from plasminogen in a wide range of pH values between 3.0 and 6.8 (Fig. 5). The most efficient generation of angiostatic peptides by procathepsin D was observed at pH 4.0 -5.0, and this process required the concomitant conversion of procathep- sin D to catalytically active pseudocathepsin D. At pH values below 3.0 or above 6.0, however, the rate of plasminogen cleavage by the enzyme was less than 20% of that at pH 4.0 -5.0. Angiostatin generation was strongly inhibited by pepstatin A at each pH value, indicating that the potency of angiostatin generation is dependent on the rate of the conversion of procathepsin D to pseudocathepsin D. The cleavage site at the Leu 46 -Ile 47 bond in the procathepsin D resulted in a loss of 3 kDa. Pseudocathepsin D described here appeared to correspond to the catalytically active intermediate formed when procathepsin D from recombinant (30) and natural sources (31-35) is exposed to acid pH. However, pseudocathepsin D was not con-verted to the mature cathepsin D even by the prolonged incubation. In contrast, the seminal vesicle fluid from patients with prostate carcinoma contained the mature cathepsin D and procathepsin D and did not show pseudocathepsin D (Fig. 11). These results are consistent with the previous findings that the autoproteolysis of procathepsin D alone cannot generate the mature enzyme found in vivo (35) and that pseudocathepsin D is not a normal intermediate of procathepsin D processing in vivo (36). Taken together, the present results suggest that the conversion of procathepsin D to the mature enzyme via pseudocathepsin D in vivo may occur more rapidly as compared with the reaction under in vitro conditions and that the conversion of pseudocathepsin D to cathepsin D may be mediated by other enzyme(s).
It is well known that the extracellular pH of tumors is acidic, although the precise mechanism for the formation and maintenance of the acidic extracellular pH remains unknown. However, recent evidence has indicated that the proton pump of the vacuolar H ϩ -ATPase at the plasma membrane of cancer cells contributes to the formation and maintenance of acidic pH in the extracellular space of tumors (37). This enzyme normally resides in acidic organelles such as endosomes and lysosomes and contributes to the maintenance of acidic environments. In the actively bone-resorbing osteoclasts, this enzyme resides at the plasma membrane of the ruffled border and works to form and maintain an acid environment in the resorption lacuna (38 -41). Therefore, when located at the plasma membrane of the cancer cells, this enzyme may work to extrude acid extracellularly. In addition, anoxia may contribute to maintain an acid environment of tumors (42). Negatively charged macromolecules, such as sialoglycoproteins and mucopolysaccharides, on the plasma membrane of cancer cells are also associated with the formation and maintenance of an acid environment of tumors. Furthermore, the ␤ subunit of mitochondrial ATP synthase (F 1 F 0 -ATPase) has been present on the surface of cancer cells (43) as well as endothelial cells (44). It is thus interesting to speculate that this enzyme at the FIG. 8. Inhibition of human endothelial cell proliferation in vitro and bFGF-induced mouse corneal neovascularization in vivo by purified angiostatic peptides. A, various concentrations of angiostatic peptides generated by pseudocathepsin D (E) and elastase (q) were incubated with human omentum microvascular endothelial cells at 37°C for 72 h. Angiostatic peptides generated by pseudocathepsin D, which consist of kringles 1-4 of plasminogen, inhibited endothelial cell proliferation in a dose-dependent manner. A similar inhibition profile for endothelial cell proliferation was observed with angiostatic peptides generated by elastase, which consist of kringles 1-3. B, concentration-dependent inhibition of the mouse cornea neovascularization by angiostatic peptides. Hydron pellets containing 50 ng of bFGF were implanted in mouse cornea in the presence or absence of various concentrations of angiostatic peptides. The vessel lengths were determined 6 days after bFGF were implanted in mouse cornea. The error bars show the means Ϯ S.D. plasma membrane of cancer cells also may work to produce extracellular ATP and to maintain an acidic extracellular pH. Therefore, it is most likely that cancer cells in tumors provide an acidic pH environment formed by multiple pathways in vivo that is favorable for the activation of procathepsin D, thereby generating angiostatic peptides.
Angiostatic peptides produced by pseudocathepsin D consisted of kringles 1-4 of plasminogen. Pseudocathepsin D initially cleaved at the site between Leu 74 and Phe 75 of plasmin-ogen and subsequently at the site between Leu 451 and Pro 452 . Finally, it cleaved at the site between Glu 699 and Ala 700 . In parallel with angiostatin generation, pseudocathepsin D nicked the serine proteinase domain of plasminogen, resulting in the loss of the potency of plasminogen-plasmin converting activity (Fig. 7). This may cause loss of the angiostatin-generating activity by plasmin. On the other hand, plasmin is known to be a key factor to generate angiogenic factors (22). Therefore, inactivation of plasmin by pseudocathepsin D may also contribute to anti-angiogenesis. Human mature cathepsin D showed the same cleavage profile for plasminogen as the pseudocathepsin D. Interestingly, the relevant human aspartic proteinases cathepsin E and pepsin also degraded plasminogen with a similar susceptibility to generate angiostatic peptides. However, while the generated angiostatic peptides were resistant to proteolysis by pseudocathepsin D, they were further degraded by cathepsin E and pepsin by prolonged incubation. Consistent with these results, neither cathepsin E nor pepsin was detectable in the culture medium of PC-3.
Purified angiostatic peptides with kringles 1-4 of plasminogen inhibited the proliferation of human microvascular endothelial cells and the bFGF-induced angiogenesis in mouse cornea (Fig. 8). Since recombinant angiostatin carrying kringles 1-3 of plasminogen (Leu 93 -Pro 386 ) has also been shown to inhibit the proliferation of bovine capillary endothelial cells and lung metastasis of mouse melanoma B16-BL6 (45), the inhibitory effect on the endothelial proliferation and the growth of several diverse carcinoma appears to be expressed by kringles 1-3 of plasminogen. Importantly, the antiproliferative activity of angiostatin has been reported to be exhibited by individual recombinant kringles 1, 2, and 3 (46), but not by kringle 4. In addition, kringle 5 has been shown to inhibit bFGFinduced proliferation and migration of bovine capillary endothelial cells (46,47). These results suggest that the antiproliferative property of angiostatin is shared by various kringlecontaining fragments of plasminogen. Taken together, the present data suggested that pseudocathepsin D generated angiostatic peptides with kringles 1-4 having the antiproliferative activity on endothelial cells. The antiproliferative effect of angiostatic peptides on endothelial cells appears to be associated with their ability to inhibit tumor angiogenesis.
Interestingly, procathepsin D purified from the culture medium of human breast carcinoma cells (MCF7) showed little or no activity for angiostatin generation (Fig. 10). SDS-PAGE FIG. 10. Difference in the angiostatin-generating activity between procathepsin D molecules derived from PC-3 and MCF7 cells. A, SDS-PAGE of procathepsin D purified from culture media of PC-3 and MCF7 cells before and after treatment with N-glycosidase F. Procathepsin D from PC-3 cells moved more slowly than that from MCF7 cells, but the mobilities of both enzymes were identical after N-glycosidase F treatment. B, procathepsin D molecules from these two cell lines before and after N-glycosidase F treatment were compared with respect to the ability to generate angiostatic peptides. Each procathepsin D was incubated with plasminogen at pH 4.0 and 37°C. Aliquots were withdrawn from the reaction mixtures at the indicated times and then analyzed by SDS-PAGE under nonreducing conditions. Freshly prepared seminal vesicular fluid obtained from two patients with prostate carcinoma (1 and 2) was centrifuged at 3000 ϫ g for 20 min at 4°C. Before and after treatment at pH 4.0 and 37°C for 3 h, the supernatant (30 l), as well as procathepsin D purified from the culture medium of PC-3 cells (3), was analyzed by SDS-PAGE and immunoblotting with antibodies to human cathepsin D.
revealed that procathepsin D derived from PC-3 cells moved more slowly than that from MCF7 cells and that N-glycosidase F treatment resulted in identical mobility. In addition, procathepsin D from PC-3 cells exhibited a plasminogen-hydrolyzing activity much higher than that from MCF7 cells at pH 4.0, although no significant difference in their activities on synthetic peptide substrates was observed. 2 These data suggest the difference in three-dimensional structures between both enzymes, probably due to their different carbohydrate structures. This speculation was further supported by experiments showing that deglycosylation of PC-3-derived procathepsin D was markedly decreased in the potency of angiostatin generation (Fig. 10).
In conclusion, angiostatic peptides could be generated from plasminogen through the action of procathepsin D secreted from prostate carcinoma cells in vitro. This reaction essentially required the concomitant conversion of procathepsin D to pseudocathepsin D. Since both procathepsin D and cathepsin D were detected in the seminal vesicle fluid from patients with prostate carcinoma, these molecules are likely to play a role in angiostatin generation in vivo as well as in vitro.