Kringle domains of human angiostatin. Characterization of the anti-proliferative activity on endothelial cells.

Recently we have identified angiostatin, an endogenous angiogenesis inhibitor of 38 kDa which specifically blocks the growth of endothelial cells (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; Folkman, J. (1995) Nat. Med. 1, 27-31). Angiostatin was shown to represent an internal fragment of plasminogen containing the first four kringle structures. We now report on the inhibitory effects of individual or combined kringle structures of angiostatin on capillary endothelial cell proliferation. Recombinant kringle 1 and kringle 3 exhibit potent inhibitory activity with half-maximal concentrations (ED50) of 320 nM and 460 nM, respectively. Also, recombinant kringle 2 displays a significant inhibition, although decreased compared with both kringle 1 and kringle 3. In contrast, kringle 4 is an ineffective inhibitor of basic fibroblast growth factor-stimulated endothelial cell proliferation. Among the tandem kringle arrays, the recombinant kringle 2-3 fragment exerts inhibitory activity similar to kringle 2 alone. However, relative to kringle 2-3, a marked enhancement in inhibition is observed when individual kringle 2 and kringle 3 are added together to endothelial cells. This implies that it is necessary to open the cystine bridge between kringle 2 and kringle 3 to obtain the maximal inhibitory effect of kringle 2-3. An increased (<2-fold) inhibitory activity is observed for the kringle 1-3 fragment (ED50 = 70 nM) compared with kringle 1-4 (ED50 = 135 nM). These data indicate that the anti-proliferative activity of angiostatin on endothelial cells is shared by kringle 1, kringle 2, and kringle 3, but probably not by kringle 4 and that more potent inhibition results when kringle 4 is removed from angiostatin. Thus, in view of the variable lysine affinity of the homologous domains, it would appear that lysine binding capability does not correlate with the relative inhibitory effects of the kringle-containing constructs. However, as we also demonstrate, appropriate folding of kringle structures is essential for angiostatin to maintain its full anti-endothelial activity.

The formation of new blood vessels occurs as a result of the growth of capillaries by vascular sprouting from preexisting vessels, a process called angiogenesis (1,2). Upon growth stimulation, quiescent endothelial cells can enter into the cell cycle, migrate, degrade the underlying basement membrane, and form a lumen. Angiogenesis is required for a variety of physiological processes such as embryonic development, wound healing, and tissue as well as organ regeneration (3,4). These processes depend upon the tightly regulated growth of endothelial cells which can be switched on and off within a short period. Abnormal growth of new blood vessels can lead to the progression of many diseases such as diabetic retinopathy and tumor growth (4). Direct experimental evidence shows that tumor growth and metastases are angiogenesis-dependent (1)(2)(3)(4). The switch to the angiogenic phenotype requires both upregulation of angiogenic stimulators and down-regulation of angiogenesis inhibitors (5,6).
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 factor (VEGF)/vascular permeability factor are most commonly expressed in tumors (7)(8)(9)(10)(11). Tumor cells may overexpress one or more of these angiogenic factors that may function synergistically in promoting tumor growth. To date, VEGF has been reported as an endothelial cell-specific mitogen (12). Although overproduction of angiogenic stimulators is necessary for the angiogenic switch, it is not sufficient for a tumor to become angiogenic. Expression of angiogenesis inhibitors must be simultaneously down-regulated (13,14). An increasing number of negative angiogenesis regulators (nine presently known) have been identified (4). Among these endogenous suppressors, angiostatin (15) is produced in association with a murine Lewis lung carcinoma growth. Angiostatin has been characterized as a potent inhibitor of angiogenesis which specifically targets proliferating endothelial cells (3,15).
Angiostatin is a circulating angiogenesis inhibitor that has been purified from serum and urine of mice bearing a murine Lewis lung carcinoma (15). Amino acid sequence analysis reveals that angiostatin is identical to an internal fragment of mouse plasminogen beginning with amino acid residue Val 79 (Val 98 including the signal peptide). Based on its molecular weight, the carboxyl terminus of angiostatin is predicted to be * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  residue 440 of plasminogen. Thus, angiostatin contains the first four triple loop disulfide-linked structures of plasminogen, known as kringle domains. Proteolytic fragments of human plasminogen comparable to the murine angiostatin, but not intact plasminogen, inhibit the neovascularization and growth of lung metastases in mice (15). In vitro, human angiostatin specifically inhibits endothelial cell proliferation but fails to inhibit proliferation of other cell types, including Lewis lung tumor cells. In vivo, it inhibits neovascularization in the chick chorioallantoic membrane assay and in the mouse corneal assay (15). We report here the characterization of the inhibitory activity of individual and multiple kringle fragments of angiostatin. This study was made possible by the availability of each of the individual angiostatin kringle modules as recombinant proteins in their properly folded and "active" (i.e. endowed with their characteristic ligand binding properties) forms (16 -19). Our data suggest that a functional difference is present among individual kringle structures and that a comparable, if not more potent, anti-endothelial activity can be obtained from a smaller fragment than from intact angiostatin. Integrity of the kringle structures of angiostatin is required to maintain its inhibitor potency.

EXPERIMENTAL PROCEDURES
Proteolytic Fragments of Plasminogen-Fragments containing kringles 1-3 (K1-3), kringles 1-4 (K1-4) and kringle 4 (K4) (Fig. 1) were prepared by digestion of human plasminogen (HPg) with Lys 78 carboxyl terminus (Lys-HPg) (by methods used in Abbott Laboratories) with porcine elastase (Sigma) as published (20,21). Briefly, 1.5 mg of elastase was incubated at room temperature with 200 mg of HPg in 50 mM Tris-HCl pH 8.0 overnight with shaking. The reaction was terminated by the addition of diisopropyl fluorophosphate (Sigma) to a final concentration of 1 mM. The mixture was rocked for an additional 30 min at room temperature and dialyzed overnight against 50 mM Tris-HCl, pH 8.0.
Gene Construction, Expression, and Purification of Recombinant Kringles-A polymerase chain reaction-based method was used to generate the cDNA fragments coding for K1, K2, K3, K4, and K2-3 of HPg ( Fig. 1). Recombinant K1 (rK1) and K4 (rK4) were expressed in Escherichia coli strain DH5␣ using a pSTII plasmid vector as published (16,17). The expressed proteins are obtained periplasmic, i.e. oxidatively refolded. Purification to homogeneity was achieved by chromatography on lysine-Sepharose 4B (Pharmacia Biotech Inc.) and Mono Q (Bio-Rad) resins (16,17). Recombinant K2, K3, and K2-3 were expressed in E. coli bacterial cells strains BL21 (rK2) and M15 (rK3, rK2-3) using the expression vector pQE8 (Qiagen), as reported (18,19). To avoid homodimerization by formation of interkringle disulfide bridges, the cysteine residues 169 in rK2 and 297 in rK3 were mutated to serines (19). The rK2, rK3, and rK2-3 contained an amino-terminal hexahistidine tag, with an inserted factor Xa cleavage site. Initial purification was by chromatography on a Ni 2ϩ -nitrilotriacetic acid-agarose column (18,19). In the case of rK2-3, ␤-mercaptoethanol was added to all extraction buffers to a final concentration of 5 mM. To achieve refolding (22), the purified proteins were adjusted to pH 8.0, and dithiothreitol was added to a final concentration of 5 mM. Following an overnight incubation, the solution was diluted gradually with 5 volumes of 50 mM Tris-HCl, pH 8.0, containing each reduced and oxidized glutathione at a concentration of 1 mM. The renatured proteins were dialyzed against H 2 O and lyophilized. Final purification of rK2 and rK2-3 was achieved by affinity chromatography on a lysine-Bio-Gel column (2 ϫ 13 cm) and eluted with phosphate buffer containing 50 mM 6-aminohexanoic acid. After proteolytic removal of the hexahistidine tag from the rK2, the proteins were desalted by gel filtration on Sephadex G-50f. Recombinant K3 was purified by reverse-phase HPLC on an Aquapore Butyl column (2.1 ϫ 100 mm, widepore 30 nM, 7 m; Applied Biosystems) using a Hewlett Packard liquid chromatograph 1090 and acetonitrile gradients. The fact that rK1, rK2, rK2-3, and rK4 are all retained upon affinity chromatography on lysine-conjugated support gels affords good indication that the recovered proteins are functional. Further evidences of proper folding were provided by their thermal melting behaviors in the presence and absence of specific lysine analog ligands (16,17), NMR spectra by reference to intact modules obtained via proteolytic fragmentation of plasminogen (16 -19), and, in the case of rK1, independent x-ray crystallographic structures determined ligand-free (23) and complexed (24) with specific -aminocarboxylic acid ligands, analogs of lysine.
Reduction and Alkylation-The reduction and alkylation of kringle fragments were performed according to a standard protocol, as published (25). Approximately 20 -80 g of purified protein in 300 -500 l 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 overnight against 20 volumes of DMEM. The solution was dialyzed further at 4°C for additional 4 h against 20 volumes of fresh DMEM. After dialysis, the samples were analyzed on a SDS gel and assayed for their inhibitory activities on endothelial cell proliferation.
Endothelial Proliferation Assay-Bovine capillary endothelial (BCE) cells were isolated as described previously (26) and maintained in DMEM supplemented with 10% heat-inactivated bovine calf serum, antibiotics, and 3 ng/ml recombinant human bFGF (Scios Nova, Moun- tainview, CA). Monolayers of BCE cells growing in six-well plates were dispersed in a 0.05% trypsin solution. Cells were 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 tissue culture plates and incubated at 37°C (in 10% CO 2 ) for 24 h. The medium was replaced with 500 l of fresh DMEM containing 5% bovine calf serum, and samples of individual or tandem kringle fragments in triplicate were added to each well. After 30 min of incubation, bFGF was added to a final concentration of 1 ng/ml. After a 72-h incubation, cells were trypsinized, resuspended in Hematall (Fisher Scientific), and counted with a Coulter counter.

Purification and Characterization of Kringle Fragments of
HPg-Recombinant proteins expressed from E. coli were refolded in vitro and purified to homogeneity using HPLC (Fig.  2). Under reducing conditions, rK2, rK3, and rK4 migrated with molecular masses of 12-13 kDa ( Fig. 2A, lanes 3-5), corresponding to the predicted molecular mass of each kringle fragment. Surprisingly, recombinant K1 migrated with a higher molecular mass of 17 kDa on SDS gel electrophoresis ( Fig. 2A, lane 2). This has been observed previously for a rK1 segment expressed as a different construct (27), which suggests oligomerization of the unfolded K1 unit. The fragments K1-4 (angiostatin) and K1-3, obtained via elastolytic digestion of Lys-HPg, migrated with the predicted molecular masses of 43 and 35 kDa, respectively (Fig. 2B, lanes 2 and 3). Aminoterminal amino acid sequence analysis of the purified fragments yielded an identical sequence, YLSECKTGNGK, for K1-3 and K1-4. The amino-terminal sequence for K4 produced VVQDCYHGDG, with approximately 20% VQDCYHGDG, as predicted from the expected sequence beginning with Val 354 and Val 355 of HPg, as observed previously (28). Recombinant K2-3 (21,564 kDa, determined by mass spectroscopy) also yielded a well defined band on SDS gel electrophoresis, albeit somewhat shifted from the 21.5-kDa standard marker (Fig. 2B,  lane 4).
Anti-endothelial Cell Proliferative Activity of Individual Kringles-Individual recombinant kringle fragments of angiostatin were assayed for their inhibitory activities on BCE cell growth stimulated by bFGF. As shown in Fig. 3A, rK1 inhibited BCE cell proliferation in a dose-dependent fashion. The concentration of rK1 required to reach 50% inhibition (ED 50 ) was about 320 nM (Table I). In contrast, rK4 exhibited a significantly less marked effect on endothelial cell proliferation. The anti-endothelial cell proliferation activity of rK4 was directly compared with that of native K4 derived from proteolytic plasminogen. The lack of inhibitory activity of these two fragments was virtually identical. A dose-dependent inhibition of endothelial cell proliferation was also observed for rK2 and rK3 (Fig. 3B). However, the inhibitory potency of rK2 was consistently lower than that of rK1 or rK3 (ED 50 ϭ 460) ( Fig. 3 and Table I). No cytotoxicity or distinct morphology associated with apoptotic endothelial cells such as rounding, detachment, and fragmentation of cells could be detected, even after incubation with a high concentration of these kringle fragments. These data suggest that the anti-endothelial growth activity of angiostatin is shared by fragments of K1, K2, and K3, but not, or only marginally by K4.
Anti-endothelial Cell Proliferative Activity of K1-3 and K1-4 Fragments-To evaluate the anti-endothelial cell proliferative effect of combined kringle fragments, purified proteolytic fragments of human K1-4, K1-3, and rK2-3 were assayed on BCE cells. In agreement with our previous finding (15), BCE cell proliferation was inhibited significantly by angiostatin (ED 50 ϭ 135 nM) ( Fig. 4 and Table I). An increase of anti-endothelial growth activity is suggested for the K1-3 fragment (ED 50 ϭ 70 nM) (Table I). In both cases, the inhibition of endothelial cell proliferation occurred in a dose-dependent manner. These results indicate that the presence of K4 is not required for anti- endothelial growth activity and that its removal from angiostatin is likely to potentiate its activity.
Additive Inhibition by rK2 and rK3-The rK2-3 fragment displayed only weak inhibitory activity that was similar to that of rK2 alone (Fig. 4). However, both rK2 and rK3 inhibited endothelial cell proliferation (Fig. 3B). This finding suggested to us that the inhibitory effect of K3 might be hidden in the structure of K2-3 as an interkringle disulfide bond links Cys 169 in K2 to Cys 297 in K3 (19,20) (Fig. 5B). To test this hypothesis, we examined the inhibitory effect of rK2 and rK3 in combination. Interestingly, a marked inhibition was seen when individual rK2 and rK3 fragments were added together to BCE cells (Fig. 5A).
Inhibitory Activity of Combined Kringle Structures-To determine whether various fragments have enhancing or antagonistic effects on inhibition of endothelial cell proliferation, various kringle structures of human angiostatin at the concentration of 320 nM were tested on BCE cells. We first analyzed the two lysine-binding kringles rK1 and K4. As shown in Fig.  6A, K4 did not interfere with the inhibitory activity of K1, suggesting that the lysine binding site of rK1 might not be directly involved in the anti-endothelial growth activity. To investigate further if the lysine binding site is involved in the inhibitor activity, we studied the effect of 6-aminohexanoic acid (a ligand specific for the kringle lysine binding site) on kringle 1 and found no difference in its anti-proliferative activity. Similarly, rK2, which also exhibits measurable lysine binding ca-pability, 2 does not interfere with the inhibitory effect of rK1 (Fig. 6A). However, the inhibitory activity of rK1 was decreased significantly by the addition of rK2-3 (Fig. 6B). This finding suggests that the fragment of rK2-3 could compete for the binding site of K1 on endothelial cells. In contrast, the intact K1-3 fragment exerted a potent inhibitory effect on BCE cells.  Thus, integrity of K1 and K2-3 within the same fragment is required to achieve maximal inhibition.
Appropriate Folding of Kringle Structures Is Required for the Anti-endothelial Activity of Angiostatin-To study whether the folding of kringle structures is required for the anti-endothelial proliferation activity, native angiostatin (K1-4) was reduced with dithiothreitol, alkylated with iodoacetamide, and assayed on BCE cells. As shown in Fig. 7A, on SDS gel electrophoresis, the reduced/alkylated protein migrated to a higher position, with a molecular mass of about 42 kDa (lane 2) compared with the native angiostatin with a molecular mass of 33 kDa (lane 1), suggesting that the native angiostatin cystine bridges were significantly disrupted. The anti-proliferative activity of angiostatin was largely abolished after reduction/alkylation (Fig.  7B). Thus, appropriate folding of kringle structures as tandem domains held together by intrachain or interchain (K2-3) disulfide bonds is required for angiostatin to display its potent inhibitory activity.

DISCUSSION
Angiostatin is a proteolytic fragment of plasminogen processed in mice bearing a murine Lewis lung primary tumor. This fragment mediates the suppression of metastatic growth (3,15) and primary tumor growth (55) by blocking tumorinduced angiogenesis. At present, the origin of angiostatin in vivo is not known. It is unlikely that tumor cells produce angiostatin directly because they lack a detectable amount of mRNA for plasminogen. 3 Although elastase cleaves plasminogen into an active form of angiostatin in vitro, it is not yet clear which protease(s) is involved in conversion of plasminogen to angiostatin in vivo. Our data demonstrate that smaller kringle fragments of human angiostatin contain inhibitory activity on endothelial cell proliferation.
Amino acid sequence alignment of the kringle domains of HPg shows that K1, K2, K3, and K4 display considerable sequence similarity (48 -50% identity, Fig. 8). NMR (19,29,30) and x-ray crystallographic (23,24,31,32) studies have demonstrated that the homology translates into a remarkable conformational uniformity. Among these structures, the lysine-binding K1 is the most potent inhibitory segment of endothelial cell proliferation. However, K3, which lacks lysine binding potential, exhibits higher inhibitory potency than either K2 4 or K4 (20,31,33), which show definite lysine binding capability. Indeed, K4, which manifests comparable or even higher affinity for lysine than does K1, exhibits marginal or no inhibitory activity. Hence, the lysine binding site of the kringle structures may not be directly involved in the inhibitory activity.
In view of the functional divergence of the homologous structures, the kringle domains provide a unique system to study the role of mutations caused by DNA replication during evolution (34). Similar divergent activities related to the regulation of angiogenesis exhibited by a group of structurally related proteins are also found in the -CXC-chemokine and prolactingrowth hormone families (35)(36)(37)(38)(39).
K4 is unique among the HPg kringles in that it contains two sets of positively charged lysine pairs, adjacent each to Cys 22 and Cys 80 , respectively (Fig. 8). Inspection of the three-dimensional structure (29,31) shows that these four lysines, together with Lys 59 , configure an exposed, positively charged area in K4, whereas other kringle structures lack such a cationic cluster. Whether this lysine-enriched domain contributes to the loss of inhibitory activity of the HPg K4 remains to be established. K4 was previously reported to stimulate proliferation of other cell types and to increase the release of intracellular calcium (40). The fact that removal of K4 from angiostatin somewhat potentiates its inhibitory activity on endothelial cells suggests that this structure may prevent some of the inhibitory effect of K1-3. Because proteases such as elastase can convert plasminogen to K1-3 in vitro, it would be interesting to investigate if such a conversion also occurs in vivo.
The mechanism underlying how angiostatin and its related kringle fragments specifically inhibit endothelial cell growth is unclear. It is not known whether the inhibition occurs directly at the cell surface or whether angiostatin is internalized by endothelial cells and subsequently inhibits cell proliferation. Both mechanisms could involve a receptor specifically expressed in proliferating endothelial cells. Alternatively, angiostatin may interact with an endothelial cell adhesion receptor such as integrin ␣ v ␤ 3 , thus blocking integrin-mediated angiogenesis (41). Of interest, Friedlander et al. (42) reported recently that in vivo angiogenesis in cornea or chorioallantoic membrane models (induced by bFGF and by tumor necrosis factor) was integrin ␣ v ␤ 3 -dependent. However, angiogenesis stimulated by VEGF, transforming growth factor-␣, or phorbol esters was dependent on integrin ␣ v ␤ 5 . Antibodies to the individual integrins specifically blocked one of these pathways, and a cyclic peptide antagonist of both integrins blocked angiogenesis induced by each cytokine (42). Because bFGF-and VEGFinduced angiogenesis are inhibited by angiostatin, it may block a common pathway for these integrin-mediated angiogenesis.
An increasing number of endogenous angiogenesis inhibitors have been identified in the last few decades (4). Of the characterized endothelial cell suppressors, several inhibitors are proteolytic fragments. For example, the 16-kDa amino-terminal fragment of human prolactin inhibits endothelial cell proliferation and blocks angiogenesis in vivo (43). In a recent paper, D'Angelo and co-workers reported that the anti-angiogenic 16-kDa amino-terminal fragment inhibited the activation of mitogen-activated protein kinase by VEGF and bFGF in capillary endothelial cells (44). Similar to angiostatin, the intact parental molecule of prolactin does not inhibit endothelial cell proliferation, nor is it an angiogenesis inhibitor. Platelet factor 4 inhibits angiogenesis at high concentrations (35,36). However, the amino-terminally truncated proteolytically cleaved platelet factor 4 fragment exhibits a 30 -50-fold increase in its antiproliferative activity over the intact platelet factor 4 molecule (45). Smaller peptide fragments of fibronectin, murine epidermal growth factor, and thrombospondin have also been shown to inhibit endothelial cell growth selectively (46 -48). When combined with our observation that the proteolytically processed fragments of K1-4 and K1-3 inhibit endothelial cell growth, these findings lead us to suggest that proteolytic processing of a large protein may change the conformational structure of the original molecule or expose new epitopes that are anti-angiogenic. Thus, protease(s) may play a critical role in the regulation of angiogenesis. To date, little is known about the regulation of these protease activities in vivo.
In this paper we also show that the disulfide bond-mediated folding of the kringle structures in angiostatin is essential to maintain its inhibitory activity on endothelial cell growth. Kringle structures analogous to those in plasminogen are also found in a variety of other proteins. For example, apolipoprotein(a) has as many as 37 repeats of plasminogen K4 (49). The amino-terminal portion of prothrombin also contains two kringles that are homologous to those of plasminogen (50). The tissue-type and the kidney-type (urokinase) plasminogen activators possess kringle modules that share extensive homology with those of plasminogen (51,52). In addition, TrK-related tyrosine kinases (53) and hepatocyte growth factor also carry kringle structures (54). It remains to be seen whether these kringle proteins or kringle fragments can also inhibit endothelial cell growth.
Endothelial cell-specific inhibitors such as angiostatin or related fragments may be useful for long term therapy in suppression of both primary and metastatic tumor growth. Therapeutic use of endogenous inhibitors is less likely to cause side effects such as suppression of hematopoiesis and gastrointestinal symptoms (4). The combination of various angiogenesis inhibitors may have a synergistic inhibitory effect on tumor growth because they probably block angiogenesis via different mechanisms.