Prothrombin Kringle-2 Domain Has a Growth Inhibitory Activity against Basic Fibroblast Growth Factor-stimulated Capillary Endothelial Cells*

Recently, O’Reilly et al.(O’Reilly, M. S., Holmgren, L., Shing, Y., Chen, C., Rosenthal, R. A., Moses, M., Lane, W. S., Cao, Y., Sage, E. H., and Folkman, J. (1994) Cell 79, 315–328; O’Reilly, M. S., Boehm, T., Shing, Y., Fukai, N., Vasios, G., Lane, W. S., Flynn, E., Birkhead, J. R., Olsen, B. R., and Folkman, J. (1997) Cell 88, 277–285) developed a simple in vitro angiogenesis assay system using bovine capillary endothelial cell proliferation and purified potent angiogenic inhibitors, including angiostatin and endostatin. Using a simplein vitro assay for angiogenesis, we purified a protein molecule that showed anti-endothelial cell proliferative activity from the serum of New Zealand White rabbits, which was stimulated by lipopolysaccharide. The purified protein showed only bovine capillary endothelial cell growth inhibition and not any cytotoxicity. This molecule was identified as a prothrombin kringle-2 domain (fragment-2) using Edman degradation and the amino acid sequence deduced from the cloned cDNA. Both the prothrombin kringle-2 domain released from prothrombin by factor Xa cleavage and the angiogenic inhibitor purified from rabbit sera exhibited anti-endothelial cell proliferative activity. The recombinant rabbit prothrombin kringle-2 domain showed potent inhibitory activity with half-maximal concentrations (ED50) of 2 μg/ml media. As in angiostatin, the recombinant rabbit prothrombin kringle-2 domain also inhibited angiogenesis in the chorioallantoic membrane of chick embryos.

Angiogenesis is a process of blood vessel formation in which new vessels sprout from existing blood vessels (1). Blood capillaries are primarily composed of endothelial cells that are normally quiescent in adult mammals under physiological conditions (2). Angiogenesis is required for a variety of physiological processes such as embryonic development, wound healing, tissue regeneration, and organ regeneration. Outgrowth of new blood vessels under pathological conditions can lead to the development and progression of diseases such as tumor growth, diabetic retinopathy, tissue and organ malformation, and cardiovascular disorders (3). The switch of the angiogenesis phe-notype depends upon the net balance between the up-regulation of angiogenic stimulators and the down-regulation of angiogenic suppressors (2).
A variety of growth factors can stimulate angiogenesis in vitro and in vivo (4). Of the known angiogenic factors, fibroblast growth factors (FGFs) 1 and vascular endothelial growth factors are most commonly expressed (5)(6)(7)(8)(9). These factors may function synergistically in promoting tumor growth when angiogenic inhibitors are simultaneously down-regulated (10,11). Multiple factors, including an angiogenic stimulator (FGF) and an inhibitor (transforming growth factor-␤), are called into play during wound healing to bring about regrowth of damaged tissues and a functional vascular bed (12).
The strategy to discover a new angiogenic factor depends upon the generation of neovascularization in the chick chorioallantoic membrane assay and in the corneal assay in vivo and in the endothelial cell tube formation and migration assay in vitro.
Recently, O'Reilly et al. (13,14) developed a simple in vitro angiogenesis assay system using bovine capillary endothelial (BCE) cell proliferation and purified potent angiogenic inhibitors, including angiostatin and endostatin. These proteins suppressed neovascularization in the mouse corneal assay and growth of tumor metastases.
In this report we describe the identification of an angiogenic inhibitor from rabbit serum treated with lipopolysaccharide (LPS) by using a BCE cell proliferation assay. Also, we report that this molecule is identical to the prothrombin kringle-2 domain (PtK2).
The adrenal cortex was extricated from the glands and cut into 1-mm pieces. The sliced tissues were then incubated in 0.5% collagenase at room temperature for 1 h. The detached capillary segment and endothelial cell aggregates were suspended in a culture medium (DMEM containing 10% bovine calf serum, 2 mM L-glutamine, 10 units/ml penicillin G, and streptomycin sulfate) and plated onto gelatinized dishes. After 2-4 days, colonies of endothelial cells were mechanically removed using a pipette equipped with a microtip. Endothelial cell aggregates were seeded onto a gelatinized dish and grown in a culture medium containing 3 ng/ml recombinant human bFGF (R & D Systems, Inc.). The cells over 30 passages were used for the endothelial cell proliferation assay.
Endothelial Cell Proliferation Assay and L929 Cell Cytotoxic Assay-Endothelial cell proliferation assay was performed according to the method of O'Reilly et al. (13). BCE cells were maintained in DMEM containing 10% heat-inactivated bovine calf serum and 3 ng/ml recombinant human bFGF. Cells growing in gelatin-coated six-well plates were dispersed in a 0.05% trypsin solution and resuspended with DMEM containing 10% bovine calf serum. Approximately 12,500 cells in 0.5 ml were added to each well of gelatinized 24-well plates and incubated at 37°C (in 10% CO 2 ) for 24 h. The media were replaced with 0.25 ml of fresh DMEM containing 5% bovine calf serum, and samples were added to each well. After 30 min of incubation, media were added to obtain a final volume of 0.5 ml of DMEM containing 5% bovine calf serum and bFGF at 1 ng/ml. After 72 h of incubation, the cells were trypsinized and counted using a hemocytometer. To ensure that any inhibition observed was not due to detachment of the BCE cells from the plate, all wells of the assay were examined for evidence of cell detachment under an inverted microscope prior to cell counting. Cytotoxic activity was determined by using mouse L929 fibroblast cells as described by Ruff and Gifford (17).
Separation and Isolation of Angiogenic Inhibitor-To purify the rabbit angiogenic inhibitor, we used both TNF cytotoxicity assay and BCE cell proliferation assay, because the endothelial cell growth inhibitory activity co-eluted with fibroblast cell cytotoxic activity on conventional chromatography procedures. Briefly, solid ammonium sulfate was added to LPS-treated rabbit serum to 80% saturation. Pellets were dialyzed against 10 mM NaP i , pH 7.0 containing 50 mM NaCl. The dialyzed proteins were applied to a DEAE-Sephacel column previously equilibrated with a starting buffer (10 mM NaP i , 50 mM NaCl). The column was eluted at 4°C with a linear gradient of 50 -500 mM NaCl in 10 mM NaP i , pH 7.0. The active fractions were pooled, concentrated, and loaded at 4°C onto a Sephacryl S-200 column equilibrated with 50 mM NaP i , pH 7.0. The active fractions were loaded at room temperature onto a Mono Q column equilibrated with 50 mM NaP i , pH 7.0. The column was eluted with a linear gradient of 125-300 mM NaCl in 50 mM NaP i , pH 7.0. The active fractions from the FPLC-Mono Q column were loaded on a preparative native 12.5% polyacrylamide gel (LKB, Inc.). After electrophoresis for 48 h at 50 V and 4°C, the gels were sliced to 5-mm and eluted in phosphate-buffered saline (PBS), and the eluates were assayed for BCE cell proliferation and L929 cell cytotoxicity.
Primary Structure Determination of Rabbit Angiogenic Inhibitor-N-terminal amino acid sequence analysis was performed with 10 g of the protein electroblotted onto a polyvinylidene difluoride membrane (Bio-Rad). For internal peptide sequencing, protein samples were dissolved in 100 l of digestion buffer (20 mM Tris-HCl, 100 mM NaCl, and 5% acetonitrile, pH 7.8). Digestion was performed with trypsin, chymotrypsin, and V8 protease, respectively, at an enzyme to substrate ratio of 1:50 for 12 h at 37°C. The digested proteins were then applied to a C18-reverse phase column for the isolation of digested products. The isolated peptide fragments were sequenced on a model 477A protein sequencer (Applied Biosystems, Inc., Foster City, CA). For C-terminal amino acid sequencing, the purified angiogenic inhibitor was incubated with carboxypeptidase Y at a 1:50 molar ratio for 3, 10, and 30 min, respectively. Each aliquot was then loaded directly onto a Beckman 6300 amino acid analyzer.
For the amino acid composition of the purified angiogenic inhibitor, the carboxymethylated proteins (20 g) were dissolved in 0.2 ml of acid hydrolysis solution (6 N HCl, 0.01% phenol, 0.005% 2-mercaptoethanol) in an acid-washed hydrolysis tube. The sample was then sealed under reduced pressure and hydrolyzed at 110°C for 24 h. The amino acid composition of the sample was determined using a Beckman 6300 amino acid analyzer.
Purification of Rabbit Prothrombin and Kringle-2 Domain-Rabbit prothrombin (Pt) was purified according to the modified method of Weinstein et al. (18). The blood was treated with 0.85% sodium citrate and centrifuged at 1,000 ϫ g for 30 min to remove blood cells and insoluble materials. Benzamidine HCl and P-PACK (D-phenylalanyl-L-prolyl-L-arginine chroromethyl ketone, Calbiochem, Inc.) were added to a final concentration of 5 mM and 1 M, respectively. Ten grams of QMA Accell (Waters, Inc.) were added to 350 ml of plasma, and the mixture was stirred at room temperature for 1 h. After allowing the QMA to settle, the plasma was decanted, and the beads were washed several times with a Tris buffer (20 mM Tris-HCl, 100 mM NaCl, 5 mM benzamidine HCl, 1 M P-PACK, pH 7.5). The QMA-bound proteins were eluted with a Tris buffer containing 0.6 M NaCl. The supernatant was pooled, and solid sodium citrate was added to a final concentration of 0.6%. Barium chloride (8 ml of 1 M BaCl 2 /100 ml) was added dropwise, and the suspension was stirred at 4°C for 1 h. After centrifugation at 4,500 ϫ g for 20 min, the pellet was resuspended in 0.2 M EDTA, 5 mM benzamidine HCl and then dialyzed overnight at 4°C against the Tris buffer. The proteins were applied to a 1 ϫ 16-cm column of QMA Accell and then eluted from the column with a linear gradient of 0.2-0.4 M NaCl using an FPLC apparatus. Prothrombin pools were identified with SDS-PAGE and were dialyzed for 4 h at 4°C against a Tris buffer containing 50 mM NaCl and applied to a Blue-Sephadex column. After being washed with the dialysis buffer, bound prothrombin was eluted with a Tris buffer containing 0.6 M NaCl.
The purified Pt was dialyzed at 4°C overnight against the Tris buffer (100 mM Tris-HCl, 150 mM NaCl, 1 mM CaCl 2 , pH 8.0) and incubated with factor Xa (Sigma) at 23°C overnight on a water bath. The proteolytic fragments were loaded to a preparative native 12.5% polyacrylamide gel (LKB, Inc.). After electrophoresis for 48 h in 50 V at 4°C, the gels were sliced to 5-mm and eluted in PBS. The eluates containing the PtK2 were pooled, dialyzed extensively against doubly distilled water, and lyophilized. The isolated PtK2 was analyzed by N-terminal sequencing and assayed for BCE cell proliferation.
Purification of Plasminogen and Angiostatin-Human plasminogen (Pg) was purified according to the modified method of Deutsch and Mertz (19). Two liters of blood were treated with 0.85% sodium citrate and centrifuged at 1,000 ϫ g for 30 min to remove blood cells and insoluble materials. The sample was diluted 1:2 with PBS and applied to a lysine-Sepharose column equilibrated with PBS. The column was washed by 0.3 M NaP i , pH 7.0, containing 3 mM EDTA and then eluted with 0.2 M aminocaproic acid, pH 7.4. Pg pools were identified with SDS-PAGE. The eluent was diluted with an equal volume of chloroform, and the aqueous phase was removed and dialyzed extensively using 20 mM Tris-HCl, pH 7.4.
Angiostatin was produced from the purified Pg by a limited proteolytic digest according to the method of O ' Reilly et al. (20). Briefly, porcine pancreatic elastase (Sigma) was added (0.8 unit/mg) to 100 mg of Pg in 20 mM Tris-HCl, pH 7.4. The solution was incubated at 37°C for 5 h and then loaded onto a lysine-Sepharose column that had been equilibrated with 50 mM NaP i , pH 7.0. The column was washed and subsequently eluted with 0.2 M aminocaproic acid. The eluent was concentrated and loaded onto a Sephacryl S-200 gel column previously equilibrated with PBS. The angiostatin containing fractions was pooled, dialyzed by doubly distilled water, and lyophilized. Purified angiostatin was analyzed on SDS-PAGE followed by Coomassie staining and by N-terminal sequencing.
Antisera and Immunoblot Analysis-Rats were injected with 80 g of recombinant rabbit PtK2 in Freund's complete adjuvant, followed by 80 g in Freund's incomplete adjuvant at 2-week intervals after the priming injection. After 4 weeks, antibody formation was detected by enzyme-linked immunosorbent assay, and final boosting was followed by 20 g of recombinant rabbit PtK2 in PBS. After 1 week, the rats were bled, and the titer of the antiserum was assessed by enzyme-linked immunosorbent assay.
For immunoblot analysis, the samples were resolved on 15% SDS-PAGE or 12.5% native PAGE and electrophoretically transferred to nitrocellulose membranes. The primary antibody was diluted 1/100 in PBS. The bound primary antibody was detected by a secondary antibody (peroxidase-conjugated goat anti-rat Ig G, Sigma).
cDNA Cloning, Expression, and Purification of Recombinant Rabbit PtK2-Total RNA was isolated from the rabbit liver using the guanidium thiocyanate extraction method described by Chomczynshi and Sacchi (21). Complementary DNA prepared from a New Zealand White rabbit liver RNA was used in each series of polymerase chain reaction experiments. The primers used for polymerase chain reaction were a series of degenerative oligonucleotides based on the amino acid sequences of the angiogenic inhibitor determined by Edman degradation (5Ј-end primers of nnnnnnacyggncaytccgg (TGHSG) and caytcnggngtnaaycarcc (HSGVTQP) and 3Ј primers of nnnnnnccytcdatngcrtaytc (EEAIEG) and gcrtaytcnccnagnacytcytc (EVLGLEEA)). The amplified products were directly ligated to the pGEM-T vector (Promega, Inc.) and sequenced by the dideoxy chain termination method (22).
Recombinant PtK2 was expressed in Escherichia coli strain BL21(DE3) using a periplasmic secretion pET22b vector (Novagen). The crude extract was loaded onto a DEAE-Sephacel column that had been equilibrated with 10 mM NaP i , pH 7.0, containing 50 mM NaCl. Proteins were eluted with an NaCl gradient. Fractions containing PtK2 were pooled, concentrated, and loaded on a Sephacryl S-200 column equilibrated with 50 mM NaP i , pH 7.0. The purified PtK2 was analyzed on SDS-PAGE followed by Coomassie staining and N-terminal sequencing.
Reduction and Alkylation of Recombinant Rabbit PtK2-Reduction and alkylation were performed according to Cao et al. (23). Fifty g of purified protein in 0.3 ml of DMEM in the absence of serum were incubated at room temperature with 15 l of 0.5 M dithiothreitol for 15 min. After incubation, 30 l of 0.5 M iodoacetamide was added to the reaction. The protein solution was dialyzed at 4°C for 6 h against 100 volumes of DMEM. After dialysis, the samples were analyzed on SDS-PAGE under nonreducing conditions and assayed for their inhibitory activities on BCE cell proliferation.
Chick Chorioallantoic Membrane Assay-Three-day-old fertilized eggs were incubated at 37°C in 3% CO 2 , and a window was made after the extraction of ovalbumin (24). After 2 days of incubation, a Thermanox coverslip (Nunc, Inc.) containing rPtK2 (20 g) was applied to the CAM (chorioallantoic membrane) of individual embryos. After 48 h, a 20% fat emulsion was injected into the chorioallantois of the embryos, and CAMs were photographed.

Isolation and Characterization of the Endothelial Cell
Growth Inhibitory Factor (Angiogenic Inhibitor) from LPStreated Rabbit Serum-Previously, we reported a purification of a TNF-like factor from the serum of New Zealand White rabbits, which was stimulated by LPS. Several experimental procedures, including immunoblot analysis, receptor binding assay, and cell line specificity, indicated that the TNF-like factor might be a high molecular mass form of TNF␣ (25).
When a highly purified TNF-like factor was further purified by native PAGE elution, the bioactive fractions were separated into two peaks by BCE cell proliferation assay (Fig. 1B). The upper fractions that showed both BCE cell growth inhibition and L929 cell cytotoxicity were able to bind antibodies against recombinant human TNF␣ (rhTNF␣) on native PAGE under nonreducing conditions (Fig. 1A, right panel). This result was consistent with the previous report that TNF␣ had potent BCE cell growth inhibitory activity (26).
On the other hand, the lower fractions that revealed only BCE cell growth inhibition did not react with antibodies against rhTNF␣ (Fig. 1A, right panel). This result suggests that the lower fractions contain materials that are immunogenically different from TNF␣. When the pooled lower fractions were subjected to SDS-PAGE under reducing conditions, a 22-kDa single band was observed by silver staining (Fig. 2,  inset). Using a native PAGE elution step, we isolated 0.2 mg of angiogenic inhibitor from rabbit serum containing 30,000 mg of total protein. The half-maximal inhibition in BCE cell proliferation was observed when a 2 g/ml concentration of this angiogenic inhibitior was treated.
Primary Structure Determination of Rabbit Angiogenic Inhibitor-The amino acid sequence of the angiogenic inhibitor was determined by automated Edman degradation (Fig. 3A). Approximately 10 g of the purified protein (the pooled lower fractions) was electroblotted to a polyvinylidene difluoride membrane and sequenced on a model 477 protein sequencer. In one extended sequencing run, an essentially single N-terminal amino acid sequence was clearly identified through 27 cycles, suggesting that only one sort of protein molecule was present in the isolated active material. For the internal peptide sequencing, 40 g of the protein was digested with trypsin, chymotrypsin, or V8 protease, respectively. The digested peptides were isolated by high performance liquid chromatography and sequenced by automated Edman degradation on a protein sequencer. The cysteic acid residues were identified from the molar ratio of carboxymethylated cysteine determined by amino acids composition analysis (data not shown) and from mass spectrometer analysis of the proteolytic fragments (data not shown).
Repeated sequence analyses of the overlapping proteolytic digests indicated that the rabbit angiogenic inhibitor consists of approximately 110 amino acids. However, we could not obtain a full sequence corresponding to the molecular size of 22 kDa. Thus, we determined the molecular mass of the angiogenic inhibitor by mass spectrometer analysis. Unexpectedly, this analysis showed two peaks, a minor peak with a molecular mass of 12.3 kDa and a major peak with a molecular mass of 12.6 kDa (Fig. 2). The abnormal behavior of the rabbit angiogenic inhibitor on SDS-PAGE could possibly be due to an unusually large number of proline residues (10% of total amino acids) in the protein (27). However, a high number of glutamate residues on the C-terminal portion of the molecule may also be responsible for the abnormality (28).
Identification of Angiogenic Inhibitor as Kringle-2 Domain (Fragment-2) of Prothrombin-The amino acid sequence of the angiogenic inhibitor showed high homology with the human and mouse PtK2. Since the cDNA encoding rabbit prothrombin was partially cloned and the nucleotide sequence encoding kringle-2 domain (K2) was not available (29), we obtained cDNA encoding rabbit PtK2 from rabbit liver RNA. The nucleotide sequence of the amplified polymerase chain reaction products was determined by the dideoxy chain termination method (Fig. 3B). The amino acid sequence deduced from its cDNA was well matched with the sequence of the angiogenic inhibitor as described above and was very similar to the known human and bovine PtK2 sequence (Fig. 4). The C-terminal amino acid of PtK2 as judged by the factor Xa cleavage site was supposed to be arginine. However, when the angiogenic inhibitor was digested by carboxypeptidase Y, the first released amino acid was glycine (Fig. 3A). This analysis revealed that the C-terminal To confirm the identity between the angiogenic inhibitor and PtK2 at the protein level, we purified rabbit prothrombin and obtained the kringle-2 domain from prothrombin by factor Xa cleavage (Fig. 5, lanes A-C). The molecular mass of PtK2 derived from prothrombin was 22 kDa on SDS-PAGE under reducing conditions (Fig. 5, lane C), and mass spectrometer analysis revealed a 12.4-kDa protein and a 12.7-kDa protein (data not shown). When the N-terminal amino acid sequence of this protein was analyzed by automated Edman degradation, a major sequence was clearly identified through 15 cycles. Two positions were unassigned due to the lack of a signal (Fig. 3C).
A secondary sequence (approximately 1 ⁄10 the primary signal) starting at the fourth amino acid (serine) was also detectable. These N-terminal differences between the two forms can explain the molecular mass heterogeneity detected by mass spectrometer analysis. The PtK2 derived from prothrombin also showed anti-endothelial cell proliferative activity as the angiogenic inhibitor purified from rabbit serum. These results indicate that the angiogenic inhibitor is the same molecule as the PtK2.
Anti-PtK2 Antibody Specificity-We examined the anti-recombinant PtK2 antibody specificity by immunoprecipitation. The antibody reacted not only with prothrombin but also with the kringle-2 domain containing prothrombin derivatives such

FIG. 2. SDS-PAGE and mass spectrometer analysis of an angiogenic inhibitor purified from LPS-induced rabbit serum.
The molecular mass of the purified rabbit angiogenic factor was determined by mass spectrometer analysis. The inset represents the molecular mass of the purified rabbit angiogenic factor by SDS-PAGE under reducing conditions, and the gel was visualized by silver staining. Lane M, molecular mass marker (Bio-Rad); lane B, purified rabbit angiogenic inhibitor.

FIG. 3. Primary structure and cDNA sequence of rabbit angiogenic inhibitor and N-terminal amino acid sequence of kringle-2 domain purified from factor Xa-treated rabbit prothrombin.
A, amino acid sequence of the purified angiogenic inhibitor was determined by Edman degradation. The amino acids identified by the cDNA of the prothrombin kringle-2 domain are indicated by asterisks. The symbol () could not be read by Edman degradation or presented Asn as a weak signal. It was identified as Thr from the cDNA sequence. It was confirmed by C-terminal sequence analysis that the C-terminal amino acid of angiogenic inhibitor was glycine. 7 , V8 protease digestion; Š----‹, trypsin digestion; f---f, N-terminal amino acid sequencing; Š⅐ ⅐ ⅐ ⅐‹, chymotrypsin digestion; O, carboxypeptidase Y digestion. B, the cDNA sequence of rabbit prothrombin kringle-2 domain was determined from the method described under "Experimental Procedures." The lines above the letters indicate bases encoding amino acids not identified by Edman degradation. C: a, N-terminal amino acid sequence of kringle-2 domain purified from factor Xa -treated prothrombin; b, N-terminal amino acid sequence of rabbit TNF reported by Abe et al. (30). as prethrombin. However, the antibody did not react with the kringle 1 domain or thrombin (Fig. 6A). Both the kringle 2 domain purified from factor Xa-treated rabbit prothrombin and the recombinant PtK2 showed similar binding affinity to the anti-PtK2 antibody (Fig. 6B). However, the antibodies did not react with rabbit plasminogen, angiostatin, or the kringle-5 domain by immunoblot analysis (Fig. 6D). Although very weak bands were detected on plasminogen and angiostatin, these bands were thought to be nonspecific, because bands of the same intensity were also found in the blot that incubated with the control sera.

Recombinant Rabbit PtK2 (rPtK2) Inhibits Endothelial Cell Proliferation in Vitro and Angiogenesis in the Chick Embryo in
Vivo-We produced recombinant rabbit PtK2 in an E. coli expression system. Using DEAE-Sephacel chromatography and Sephacryl S-200 chromatography, the recombinant PtK2 was purified to apparent homogeneity from E. coli crude extracts (Fig. 5, lane D). The purified rPtK2 exerted a growth inhibitory effect on bFGF-stimulated BCE cells, and the effect was blocked by an antibody against recombinant PtK2 (Fig.  7A). Also, the activity of rPtK2 for BCE cells was diminished after reduction and alkylation of the protein (Fig. 7C). This result indicates that PtK2 requires appropriate folding by intrachain disulfide bonds to display its inhibitory activity. The anti-endothelial cell proliferative activity of rPtK2 was also compared with that of pPtK2 (proteolytic PtK2 isolated from factor Xa-treated prothrombin). As shown in Fig. 8, the concentrations of half-maximal inhibition (ED 50 ) for rPtK2 and pPtK2 are about 2 g/ml and 4 g/ml, respectively. Although the activity of rPtK2 is two times that of pPtK2, this result sug-gests the folding of recombinant PtK2 might be similar to PtK2 in prothrombin.
Previously, O , Really et al. (13) reported that angiostatin, a proteolytic fragment of plasminogen, including kringle 1-4 domain, was a potent inhibitor of BCE cell growth in vitro and of tumor cell metastases. Thus, to directly compare the inhibitory efficacy, rPtK2 and the proteolytic angiostatin (Fig. 5, lane E) were assayed for suppression of bFGF-stimulated BCE cell proliferation. The concentration of half-maximal inhibition (ED 50 ) for angiostatin is about 2 g/ml. This is similar to what is observed for rPtK2 (Fig. 8). A distinct morphological change of BCE cells by rPtK2 or angiostatin was not observed (Fig. 9). These data indicate that the signal of the BCE cell suppression by PtK2 might be equal to that by angiostatin.
We examined the ability of rPtK2 to inhibit in vivo angiogenesis by using the chick CAM assay (23). At doses of 20 g/coverslip, the rPtK2 showed potent inhibition of angiogenesis (Fig. 10).

FIG. 4. Amino acid sequence alignment of prothrombin kringle-2 and plasminogen kringle-1 domain.
Each kringle domain carries six conserved cysteine residues. The sequences for the kringle domains were aligned according to their conserved cysteine residues. Conserved amino acids are shaded. The amino acid sequence of the rabbit plasminogen kringle-1 domain has yet to be reported.

DISCUSSION
We purified the angiogenic inhibitor, an inhibitor of endothelial cell proliferation, from LPS-treated rabbit serum. The amino acid sequence of the purified angiogenic inhibitor determined by automated Edman degradation showed high homology with the PtK2 from several species. We isolated the cDNA encoding the kringle-2 domain from rabbit liver and found the whole nucleotide sequence. The amino acid sequence deduced from the nucleotide sequence of the cloned cDNA was well matched with that of the angiogenic inhibitor determined by automated Edman degradation. Furthermore, the K2 released by factor Xa cleavage from purified rabbit prothrombin had a similar molecular mass on SDS-PAGE under reducing conditions and mass spectrometer analysis. In addition, the N-terminal amino acid sequence of the two proteins was found to be identical. Interestingly, we found that the C-terminal arginine residue of PtK2 was deleted from the angiogenic inhibitor. We believe that removal of the C-terminal arginine of the angiogenic inhibitor could be processed by a carboxypeptidase-like protease during chromatographic procedures. All the above data indicate that the angiogenic inhibitor is fundamentally the same molecule as the kringle-2 domain derived from prothrombin.
Previously, Abe et al. (30) reported the purification of TNF from LPS-treated rabbit serum. They reported that the protein exerted cytotoxic activity against L929 cells and had a very similar N-terminal amino acid sequence with that of the angiogenic inhibitor (Fig. 3C). However, they could not identify this molecule as a prothrombin kringle-2 domain. In this paper, we demonstrated that this molecule did not have cytotoxic activity against L929 cells and was not antigenically related to TNF␣ (Fig. 1A, left panel). Also, the whole amino acid sequence of the angiogenic inhibitor did not have any meaningful sequence homology to that of TNF␣. However, this molecule was co-eluted at the fractions which contained L929 cell cytotoxic activity in conventional chromatography procedures. The angiogenic inhibitor eluted from gel filtration columns as a 45-kDa oligomer in a manner similar to native TNF␣ reported by Abe et al. (30) (data not shown). The abnormal migration of this molecule on SDS-PAGE under reducing conditions may be due to some unusual inherent property of the protein such as the presence of a high number of proline residues in the molecules.
The TNF-like factor and angiogenic inhibitor (PtK2) appear to exert different effects on BCE cell morphology. In the presence of PtK2, the morphology of BCE cells appeared similar to that of the control cell. In contrast, inhibition by TNF resulted  in an expanded, shallow cytoplasm (26) and extensive actin redistribution (data not shown). These results suggest that signal transduction leading to BCE cell growth inhibition by PtK2 differs from that of TNF receptors.
PtK2 was derived from the proteolytic cleavage of prothrombin by factor Xa and thrombin during blood coagulation. Deguchi et al. (31) reported that proteolytic PtK2 may act as a negative feedback modifier of the clotting cascade by inhibiting prothrombin activation. To confirm appropriate folding of recombinant rabbit PtK2 expressed in E. coli., we examined the inhibition of prothrombin activation by recombinant PtK2. The recombinant PtK2 showed a 1,500-fold inhibition of prothrombin activation. 2 This result is consistent with the report of Deguchi et al. (31). Furthermore, 1 H NMR spectroscopy analysis at 600 MHz revealed that the folding of recombinant PtK2 was comparable with that of the kringle-1 domain of tissue plasminogen activator (32). 2 These data suggest that the recombinant PtK2 might be folded properly in E. coli.
Because LPS administration is followed by intravascular thrombosis (33), we compared the amount of PtK2 in LPStreated serum with that of normal serum by immunoblot analysis. As shown in Fig. 11, the amount of PtK2 derived from LPS-treated serum was obviously more than from normal serum. This result suggests that intravascular thrombosis by LPS is related with PtK2 production.
Kringle is a small protein domain that has three characteristic intradisulfide bonds. The kringle structure was found in other proteins, such as prothrombin (34), plasminogen (35), urokinase (36), hepatocyte growth factor (37), and apolipoprotein(a) (38). These domains appear to be independent folding units, but a general functional role is not yet known. In the case of the plasminogen kringle-1 and -4 domains, they were known to be involved in the binding of plasminogen to fibrin (39). These kringle domains are presumed to interact with fibrin by binding to an exposed lysine side chain. However, it has been reported that prothrombin kringle domains do not adsorb to lysine-Sepharose columns in spite of a very similar amino acid sequence (40).
Recently, it was reported that angiostatin, an angiogenesis inhibitor, suppressed endothelial cell growth in vitro and tumor metastases in vivo (13). Angiostatin consists of the first four kringle domains of plasminogen (1-4 kringle domains). It is generated from cleavage of plasminogen by elastase-like protease (41) or from urokinase and a free sulfhydryl donor (42). When recombinant PtK2 and the proteolytic angiostatin were assayed for suppression of endothelial cell proliferation in vitro and chick CAM neovascularization in vivo, recombinant PtK2 activities were comparable with those of angiostatin. In addition to having a homologous kringle structure, PtK2 is very similar to angiostatin in that both proteins are derived from proteolytic degradation of large abundant circulatory proteins involved in regulation of blood coagulation. It has been reported that single kringle domains within angiostatin and the fifth kringle domain of plasminogen inhibit BCE cell growth as well as intact angiostatin. This inhibition is not related to lysine affinity (23,43).
The first kringle domain of angiostatin suppresses endothelial cell growth more potently than the other kringle domains of angiostatin. It also presents a more abnormal migration than other kringle domains of angiostatin on SDS-PAGE under reducing conditions (23). PtK2 is similar to the first kringle domain of angiostatin in that the molecular masses of the two proteins are overestimated 2-fold by SDS-PAGE under reducing conditions. However, it has not been reported that the single kringle domains of plasminogen and angiostatin exist independently in blood like PtK2. In the presence of PtK2, the morphology of BCE cells appeared similar to that of the control cells or angiostatin-treated cells. From this result, although the molecular mechanism of how angiostatin and its related kringle fragments inhibit endothelial cell growth has yet to be reported, it could be speculated that the signal of endothelial cell suppression by PtK2 might be similar to that of angiostatin and its related molecules. It is possible that a single kringle domain exerts endothelial cell growth inhibitory activity through multimerization. In fact, it has been suggested that the kringle domain could form a kringle-kringle interaction based on its structure (44).
It is known that the FGFs and other angiogenic stimulators can be sequestered in the extracellular matrix of many cell types, including endothelial cells. Presumably these factors are released by proteolytic degradation of the matrix (45). Now, the angiogenic inhibitors also appear to be stored but in a much different fashion. It seems that they may be cryptic parts of larger molecules that are not themselves inhibitors (2). A num-2 T.-H. Lee, T. Rhim, and S. S. Kim, unpublished data. ber of angiogenic inhibitors have been characterized as proteolytic fragments of large molecules. In addition to angiostatin (13) and the prothrombin kringle-2 domain, examples are endostatin (14), which is a C-terminal fragment of collagen XVIII, internal fragments of thrombospondin (46), fragments of laminin (47), and a fragment of fibronectin (48). The capability to release inhibitor fragments from storage as cryptic segments of abundant proteins may contribute to maintaining the normal quiescence of the vasculature and to turn off transitory angiogenic processes such as ovulation and wound healing. This capability may also contribute to the health of the organism by suppressing diseases with angiogenic components including tumorigenesis, rheumatoid arthritis, and diabetic retinopathy.
In this study, we demonstrated that the prothrombin kringle-2 domain has a potent endothelial cell growth inhibitory activity like that of angiostatin. Therapeutic use of this endogeneous inhibitor is less likely to cause side effects such as suppression of hematopoiesis and gastrointestinal symptoms. Future studies will address the usefulness of this approach by analyzing anti-angiogenesis of prothrombin kringle-2 domain in in vivo models.