Portal de Periódicos da CAPES

Generation of Angiostatin by Reduction and Proteolysis of Plasmin

1997; Elsevier BV; Volume: 272; Issue: 33 Linguagem: Inglês

10.1074/jbc.272.33.20641

1083-351X

Paul Stathakis, Melinda Fitzgerald, Lisa J. Matthias, Colin N. Chesterman, Philip J. Hogg,

Glycosylation and Glycoproteins Research

Extracellular manipulation of protein disulfide bonds has been implied in diverse biological processes, including penetration of viruses and endotoxin into cells and activation of certain cytokine receptors. We now demonstrate reduction of one or more disulfide bonds in the serine proteinase, plasmin, by a reductase secreted by Chinese hamster ovary or HT1080 cells. Reduction of plasmin disulfide bond(s) triggered proteolysis of the enzyme, generating fragments with the domain structure of the angiogenesis inhibitor, angiostatin. Two of the known reductases secreted by cultured cells are protein disulfide isomerase and thioredoxin, and incubation of plasmin with these purified reductases resulted in angiostatin fragments comparable with those generated from plasmin in cell culture. Thioredoxin-derived angiostatin inhibited proliferation of human dermal microvascular endothelial cells with half-maximal effect at approximately 0.2 μg/ml. Angiostatin made by cells and by purified reductases contained free sulfhydryl group(s), andS-carbamidomethylation of these thiol group(s) ablated biological activity. Neither protein disulfide isomerase nor thioredoxin were the reductases used by cultured cells, because immunodepletion of conditioned medium of these proteins did not affect angiostatin generating activity. The plasmin reductase secreted by HT1080 cells required a small cofactor for activity, and physiologically relevant concentrations of reduced glutathione fulfilled this role. These results have consequences for plasmin activity and angiogenesis, particularly in the context of tumor growth and metastasis. Moreover, this is the first demonstration of extracellular reduction of a protein disulfide bond, which has general implications for cell biology. Extracellular manipulation of protein disulfide bonds has been implied in diverse biological processes, including penetration of viruses and endotoxin into cells and activation of certain cytokine receptors. We now demonstrate reduction of one or more disulfide bonds in the serine proteinase, plasmin, by a reductase secreted by Chinese hamster ovary or HT1080 cells. Reduction of plasmin disulfide bond(s) triggered proteolysis of the enzyme, generating fragments with the domain structure of the angiogenesis inhibitor, angiostatin. Two of the known reductases secreted by cultured cells are protein disulfide isomerase and thioredoxin, and incubation of plasmin with these purified reductases resulted in angiostatin fragments comparable with those generated from plasmin in cell culture. Thioredoxin-derived angiostatin inhibited proliferation of human dermal microvascular endothelial cells with half-maximal effect at approximately 0.2 μg/ml. Angiostatin made by cells and by purified reductases contained free sulfhydryl group(s), andS-carbamidomethylation of these thiol group(s) ablated biological activity. Neither protein disulfide isomerase nor thioredoxin were the reductases used by cultured cells, because immunodepletion of conditioned medium of these proteins did not affect angiostatin generating activity. The plasmin reductase secreted by HT1080 cells required a small cofactor for activity, and physiologically relevant concentrations of reduced glutathione fulfilled this role. These results have consequences for plasmin activity and angiogenesis, particularly in the context of tumor growth and metastasis. Moreover, this is the first demonstration of extracellular reduction of a protein disulfide bond, which has general implications for cell biology. Fusion of human immunodeficiency (1Ryser H.J.P. Levy E.M. Mandel R. Di Sciullo G.J. Proc. Natl. Acad. Sci. U. S. A. 1994; 91: 4559-4563Crossref PubMed Scopus (232) Google Scholar) and Sindbis viruses (2Abell B.A. Brown D.T. J. Virol. 1993; 67: 5496-5501Crossref PubMed Google Scholar) with the plasma membrane has been proposed to involve reduction of critical disulfide bonds in the viral envelope glycoproteins. Similarly, cytotoxicity of diptheria toxin appears to be mediated by reduction of the toxins interchain disulfides at the cell surface (3Ryser H.J.-P. Mandel R. Ghani F. J. Biol. Chem. 1991; 266: 18439-18442Abstract Full Text PDF PubMed Google Scholar, 4Mandel R. Ryser H.J.P. Ghani F. Wu M. Peak D. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 4112-4116Crossref PubMed Scopus (197) Google Scholar). In contrast, formation of interchain disulfide bonds in the extracellular domains of the erythropoietin (5Miura O. Ihle J.N. Arch. Biochem. Biophys. 1993; 306: 200-208Crossref PubMed Scopus (47) Google Scholar) and interleukin-3 (6Stomski F.C. Sun Q. Bagley C.J. Woodcock J. Goodall G. Andrews R.K. Berndt M.C. Lopez A.F. Mol. Cell. Biol. 1996; 16: 3035-3046Crossref PubMed Google Scholar) receptors may regulate receptor activation, and nerve growth factor has been found to interact covalently with its receptor through formation of a interchain disulfide bond (7Stach R.W. Wagner B.J. Ennulat D.J. Perez-Polo J.R. Int. J. Neurol. 1985; 26: 87-94Google Scholar). These findings imply that protein disulfide bonds can be manipulated extracellularly. Despite these observations, demonstration of reduction of a native protein disulfide bond in the extracellular milieu has not yet been reported. We describe reduction of disulfide bond(s) in the serine proteinase, plasmin (8Plow E.F. Herren T. Redlitz A. Miles L.A. Hoover-Plow J.L. FASEB J. 1995; 9: 939-945Crossref PubMed Scopus (379) Google Scholar), in the conditioned medium of Chinese hamster ovary (CHO) 1The abbreviations used are: CHO, Chinese hamster ovary K-1; ε-ACA, ε-aminocaproic acid; GSH, reduced glutathione; HDMVEC, human dermal microvascular endothelial cells; MPB, 3-(N-maleimidylpropionyl)biocytin; PDI, protein disulfide isomerase; PVDF, polyvinylidene difluoride; PAGE, polyacrylamide gel electrophoresis; uPA, urokinase plasminogen activator; PAI-1, plasminogen activator inhibitor-1. 1The abbreviations used are: CHO, Chinese hamster ovary K-1; ε-ACA, ε-aminocaproic acid; GSH, reduced glutathione; HDMVEC, human dermal microvascular endothelial cells; MPB, 3-(N-maleimidylpropionyl)biocytin; PDI, protein disulfide isomerase; PVDF, polyvinylidene difluoride; PAGE, polyacrylamide gel electrophoresis; uPA, urokinase plasminogen activator; PAI-1, plasminogen activator inhibitor-1. or HT1080 cells. The reduction of plasmin disulfide bond(s) was followed by proteolysis of the enzyme releasing a fragment containing the first four kringle domains, which is the domain structure of the angiogenesis inhibitor, angiostatin. Tumor growth depends on the development of an adequate blood supply (9Hanahan D. Folkman J. Cell. 1996; 86: 353-364Abstract Full Text Full Text PDF PubMed Scopus (5991) Google Scholar). The regulation of new blood vessel growth by the process of angiogenesis is mediated by several molecules released by both tumor and host cells (9Hanahan D. Folkman J. Cell. 1996; 86: 353-364Abstract Full Text Full Text PDF PubMed Scopus (5991) Google Scholar). This was formally demonstrated for angiostatin, which accumulated in the circulation in the presence of a growing Lewis lung tumor and disappeared when the tumor was removed (10O'Reilly M.S. Holmgren L. Shing Y. Chen C. Rosenthal R.A. Moses M. Lane W.S. Cao Y. Sage E.H. Folkman J. Cell. 1994; 79: 315-328Abstract Full Text PDF PubMed Scopus (3138) Google Scholar). The angiostatin produced by the primary tumor was found to inhibit the neovascularization and growth of its remote metastases. Systemic administration of angiostatin also inhibited growth of three human and three murine primary carcinomas in mice (11O'Reilly M.S. Holmgren L. Chen C. Folkman J. Nat. Med. 1996; 2: 689-692Crossref PubMed Scopus (1146) Google Scholar). 3-(N-Maleimidylpropionyl)biocytin (MPB) was purchased from Molecular Probes (Eugene, OR), whereas all other chemicals were from Sigma-Aldrich (Sydney, NSW, Australia). Plasminogen was purified from fresh frozen human plasma and separated into its two carbohydrate variants according to published procedures (12Castellino F.J. Powell J.R. Methods Enzymol. 1981; 80: 365-378Crossref PubMed Scopus (180) Google Scholar). Both Glu(1) and Glu(2) plasminogen were used in the experiments described herein. Urokinase plasminogen activator (uPA) was a gift from Serono Australia. Plasmin was generated by incubating plasminogen (20 μm) with uPA (20 nm) for 30 min in 20 mm HEPES, 0.14 m NaCl, pH 7.4, buffer at 37 °C. Recombinant human plasminogen activator inhibitor-1 (PAI-1) was made according to Sancho et al. (13Sancho E. Tonge D.W. Hockney R.C. Booth N.A. Eur. J. Biochem. 1994; 224: 125-134Crossref PubMed Scopus (71) Google Scholar). Aprotinin (Trasylol) was obtained from Bayer Australia (Sydney, NSW, Australia). Protein disulfide isomerase (PDI) was purified from human placenta as described previously (14Lambert N. Freedman R.B. Biochem. J. 1983; 213: 225-234Crossref PubMed Scopus (125) Google Scholar), with modifications (15Hotchkiss K.A. Chesterman C.N. Hogg P.J. Biochemistry. 1996; 35: 9761-9767Crossref PubMed Scopus (48) Google Scholar). RecombinantEscherichia coli thioredoxin was from Calbiochem-Novabiochem (San Diego, CA). Rabbit polyclonal antibodies were developed against purified human placenta PDI in New Zealand white rabbits and affinity-purified on a PDI-AffiGel15 matrix (Bio-Rad Laboratories, Hercules, CA). Affinity-purified rabbit polyclonal antibodies against human thioredoxin were a gift from Prof. Arne Holmgren (Karolinska Institute, town, Sweden). Protein concentrations were determined using the Bio-Rad Protein Assay kit. All proteins were aliquoted and stored at −80 °C until use. Amino-terminal sequencing was performed by Auspep (Parkville, Australia), on proteins resolved on SDS-polyacrylamide gel electrophoresis (PAGE) and transferred to polyvinylidene difluoride (PVDF) membrane (16Matsudaira P. J. Biol. Chem. 1987; 262: 10035-10038Abstract Full Text PDF PubMed Google Scholar). CHO and HT1080 cells were purchased from American Type Cell Culture (Rockville, MD). CHO cells were maintained in Dulbecco's modified Eagle's medium/Ham's F-12 medium, whereas HT1080 cells were maintained in Dulbecco's modified Eagle's medium. The media contained 10% fetal calf serum, 2 mml-glutamine, 10 units/ml penicillin G, and 10 μg/ml streptomycin sulfate. All medium components were from Life Technologies, Inc. (Gaithersburg, MD). Experiments with adherent CHO and HT1080 cells at ∼ 80% confluence were performed in 25-cm2 tissue culture plates (Corning, NY) in complete medium or Hanks' balanced salt solution containing 25 mmHEPES, pH 7.4. Conditioned medium was collected by incubating CHO or HT1080 cells at ∼80% confluence with Hanks' balanced salt solution containing 25 mm HEPES, pH 7.4, for 6 h. On some occasions, HT1080 cells at ∼80% confluence were incubated with Dulbecco's modified Eagle's medium containing 3% fetal calf serum, 2 mml-glutamine, 10 units/ml penicillin G, and 10 μg/ml streptomycin sulfate for 3 days to generate conditioned medium. The ratio of number of cells to volume of conditioned medium was ∼3 × 106 cells/ml. All conditioned medium was passed through a 0.22-μm filter prior to storage at −80 °C. Human dermal microvascular endothelial cells (HDMVEC) were a gift from Dr. Chris Jackson (17Jackson C.J. Garbett P.K. Nissen B. Schreiber L.J. J. Cell Sci. 1990; 96: 257-262PubMed Google Scholar), and they were maintained in M199 medium containing 25% human serum, 25 μg/ml heparin, and 50 μg/ml endothelial cell growth supplement. In proliferation assays, HDMVEC at passage 6 were seeded into gelatin-coated 24-well plates and allowed to adhere overnight in M199 medium containing 25% human serum, 25 μg/ml heparin, and 7.5 ng/ml fibroblast growth factor-2 at 37 °C and 5% CO2. The fibroblast growth factor-2 content of the medium was reduced to 2.5 ng/ml, thioredoxin-derived angiostatin was added, and the wells were incubated for 72 h. Cell number was determined using a hemocytometer. Adherent CHO or HT1080 cells (3 × 106cells) were incubated with either plasminogen or plasmin (10 μg) in 1 ml of Hanks' balanced salt solution containing 25 mmHEPES, pH 7.4, at 37 °C. Incubation with plasminogen was for 6 h, whereas incubation with plasmin was for 3 h. ε-Aminocaproic acid (ε-ACA) (25 mm) was added for 20 min to displace plasmin(ogen) and its kringle-containing fragments from the cell surface. After dialysis of the reaction medium overnight against 20 mm HEPES, 0.14 m NaCl, pH 7.4, buffer to remove ε-ACA, the plasmin(ogen) kringle products were collected on 50 μl of packed lysine-Sepharose (Pharmacia Biotech Inc., Uppsala, Sweden) beads by incubation on a rotating wheel for 1 h at room temperature, washed two times with HEPES buffer, and resolved on 10% SDS-PAGE. The proteins were transferred to PVDF membrane and Western blotted using anti-plasminogen polyclonal antibodies as described below. In experiments using CHO- or HT1080-conditioned medium (from 3 × 106 cells/ml), plasmin was added to a final concentration of 10 μg/ml and incubated for 1 h at 37 °C. The plasmin kringle products were collected on lysine-Sepharose and resolved and detected as described above. After incubation of plasminogen or plasmin with adherent cells, the medium was aspirated and incubated with MPB (100 μm) for 30 min at room temperature, followed by quenching of the unreacted MPB with reduced glutathione (GSH) (200 μm) for 10 min at room temperature. Unreacted GSH and other free sulfhydryls in the system were blocked with iodoacetamide (400 μm) for 10 min at room temperature. The kringle containing proteins were collected and resolved as described above, transferred to PVDF membrane, and blotted with streptavidin-peroxidase to detect the MPB-labeled proteins. Experiments using conditioned medium were performed in the same way. Plasmin (5 μm) was incubated with either oxidized or reduced PDI (10 μm), oxidized or reduced thioredoxin (10 μm), or dithiothreitol (10 μm) for 1 h in 20 mm HEPES, 0.14m NaCl, pH 7.4, at 37 °C. The two active-site disulfides of PDI or single disulfide of thioredoxin were reduced by incubation of 10 μm of either protein with equimolar dithiothreitol, with respect to active site disulfide content, for 1 h in the same HEPES buffer at 37 °C. Samples of the reactions (25 μl) were resolved on 10% SDS-PAGE and stained with Coomassie Blue. To detect free sulfhydryl group(s) in angiostatin, plasmin (0.2 μm) was incubated with either reduced PDI (0.4 μm) or reduced thioredoxin (0.4 μm) for 1 h in 20 mm HEPES, 0.14 m NaCl, pH 7.4, buffer at 37 °C. Reactions were labeled with MPB as described above for conditioned medium. Samples of the reactions (25 μl) were resolved on 10% SDS-PAGE, transferred to PVDF membrane, and blotted with either anti-plasminogen polyclonal antibodies or streptavidin-peroxidase to detect MPB-labeled angiostatin. To purify thioredoxin-derived angiostatin, plasmin (10 μm, 5 mg) was incubated with reduced thioredoxin (20 μm) for 3 h in 20 mm HEPES, 0.14m NaCl, pH 7.4, buffer at 37 °C. On some occasions, the reaction was quenched with iodoacetamide (2 mm) for 20 min at room temperature to S-carbamidomethylate the free sulfhydryl groups. The angiostatin was purified by affinity chromatography on a 1 × 20 cm lysine-Sepharose column. The bound proteins were eluted with 20 mm HEPES, 0.14 mNaCl, pH 7.4, buffer containing 25 mm ε-ACA. The angiostatin was separated from a small amount of aggregated protein by gel filtration on a 2 × 50 cm Sephacryl S-100 (Pharmacia) column in 20 mm HEPES, 0.14 m NaCl, pH 7.4, buffer. The aggregated protein represented disulfide-bonded angiostatin multimers, because the multimers were lost when resolved on SDS-PAGE run under reducing conditions. Samples were separated on 10% SDS-PAGE under nonreducing conditions (18Laemmli U.K. Nature. 1970; 227: 680-685Crossref PubMed Scopus (205531) Google Scholar), transferred to PVDF membrane, developed according to the manufacturer's instructions (DuPont, Boston, MA), and visualized using chemiluminescence. Both rabbit anti-human plasminogen polyclonal antibodies and swine anti-rabbit IgG horseradish peroxidase-conjugated antibodies (Dako, Carpinteria, CA) were used at 1:2000 dilution. Streptavidin-horseradish peroxidase (Amersham Australia, Sydney, NSW, Australia) was used at 1:2000 dilution. Plasminogen contains five consecutive kringle domains followed by a serine proteinase module (19Castellino F.J. Urano T. deSerrano V.S. Beals J.M. Lipoprotein a. Academic Press, NY1990: 10-87Google Scholar). Urokinase- or tissue-plasminogen activator convert plasminogen to plasmin by hydrolysis of the Arg561-Val562 peptide bond in the serine proteinase module. Activation results in autoproteolytic release of an amino-terminal peptide by cleavage at the Lys77-Lys78 peptide bond. The five kringle domains (heavy chain) of plasmin are covalently attached to the serine proteinase module (light chain) by two disulfide bonds. Incubation of plasminogen or plasmin with adherent CHO cells resulted in formation of a ∼43-kDa plasmin fragment with an amino terminus of Lys-Val-Tyr-Leu-Ser-Glu, corresponding to that of Lys78plasmin (Fig. 1). A fragment of this size encompasses the first four kringle domains, which is the domain structure of the angiogenesis inhibitor, angiostatin (10O'Reilly M.S. Holmgren L. Shing Y. Chen C. Rosenthal R.A. Moses M. Lane W.S. Cao Y. Sage E.H. Folkman J. Cell. 1994; 79: 315-328Abstract Full Text PDF PubMed Scopus (3138) Google Scholar). The angiostatin fragment derived from plasmin, not plasminogen, because PAI-1 prevented the formation of both plasmin and the angiostatin fragment by CHO cells, whereas PAI-1 did not influence production of angiostatin from exogenous plasmin. In addition, the serine proteinase inhibitor, aprotinin, inhibited formation of angiostatin from both plasmin derived from plasminogen and from plasmin added exogenously, implying that proteolysis (perhaps autoproteolysis) of plasmin is required for production of angiostatin. The angiostatin fragment was also generated by adherent CHO cells incubated with plasminogen and medium containing 10% calf serum, with a rate comparable with that in the absence of serum (not shown). The lysine binding activity of the plasmin(ogen) kringles was not required for angiostatin formation, because the lysine analogue, ε-ACA, did not perturb angiostatin formation (Fig. 1 B). The angiostatin generating activity of adherent CHO cells was comparable with the activity in CHO-conditioned medium, implying that plasmin is converted to angiostatin in the extracellular milieu independent of the cell surface (Fig. 1 B). Generation of angiostatin from plasmin by CHO-conditioned medium was inhibited by aprotinin (not shown), as was observed using CHO cell monolayers (Fig.1 B). This result suggests that angiostatin is derived from plasmin by autoproteolysis, that plasmin activates an enzyme that ultimately cleaves it to generate angiostatin, or that plasmin was cleaved by a unidentified CHO cell-derived enzyme that is inhibited by aprotinin. The results described below demonstrate that plasmin autoproteolysis can account for the angiostatin fragment. These findings are in agreement with those of Gately et al. (20Gately S. Twardowski P. Stack M.S. Patrick M. Boggio L. Cundiff D.L. Schnaper H.W. Madison L. Volpert O. Bouck N. Enghild J. Kwaan H.C. Soff G.A. Cancer Res. 1996; 56: 4887-4890PubMed Google Scholar), who observed that serine proteinase activity was necessary for angiostatin production by human prostate cancer cells. In contrast, Dong et al. (21Dong Z. Kumar R. Yang X. Fidler I.J. Cell. 1997; 88: 801-810Abstract Full Text Full Text PDF PubMed Scopus (447) Google Scholar) recently reported that a macrophage-derived metalloelastase was responsible for the generation of angiostatin in Lewis lung carcinoma. The metalloelastase was inhibited by 5 mm EDTA. The angiostatin generating activity of CHO-conditioned medium was not affected by 5 mm EDTA (not shown), implying that metalloproteinases were not responsible for the activity in this system. The key to the mechanism of formation of the angiostatin fragment was provided by the observation that it contained free sulfhydryl group(s). This was demonstrated by the incorporation of the thiol-specific compound, MPB (Fig. 2). Plasminogen and plasmin did not incorporate MPB, which is consistent with the absence of free sulfhydryl groups in these proteins. A lower molecular mass fragment of ∼38 kDa was also labeled with MPB. This fragment is probably a proteolytic product of the 43-kDa fragment (see below), and may be the human counterpart of the 38-kDa murine angiostatin described by O'Reilly et al. (10O'Reilly M.S. Holmgren L. Shing Y. Chen C. Rosenthal R.A. Moses M. Lane W.S. Cao Y. Sage E.H. Folkman J. Cell. 1994; 79: 315-328Abstract Full Text PDF PubMed Scopus (3138) Google Scholar). No labeling was observed if the maleimide moiety of MPB was blocked with GSH prior to incubation of MPB with angiostatin (not shown), confirming the sulfhydryl-specific nature of the labeling. The demonstration of free sulfhydryl group(s) implies that formation of the angiostatin fragment involved disulfide bond reduction in plasmin. Reduction of protein disulfide bonds usually requires a catalyst. Two proteins that are known to catalyze reduction of protein disulfide bonds and are secreted by cultured cells are PDI and thioredoxin. PDI is a noncovalent homodimer with a subunit molecular mass of 60 kDa that assists the folding of nascent proteins in the endoplasmic reticulum (22Freedman R.B. Hirst T.R. Tuite M.F. Trends Biochem. Sci. 1994; 19: 331-336Abstract Full Text PDF PubMed Scopus (649) Google Scholar). The ability of PDI to catalyze disulfide reduction in proteins resides in two very reactive disulfide bonds that share the common sequence WCGHCK. These disulfide bonds catalyze thiol-disulfide interchanges that can lead to the net formation, the net rearrangement, or the net reduction of protein disulfide bonds depending on the nature of the protein substrate, the redox conditions, and the presence of other thiols and disulfides. Thioredoxin is a 12-kDa protein that functions as a dithiol/disulfide redox protein (23Holmgren A. Annu. Rev. Biochem. 1985; 54: 237-271Crossref PubMed Google Scholar). It has a single reactive disulfide bond in the sequence WCGPCK. Both PDI (24Yoshimori T. Semba T. Takemoto H. Akagi S. Yamamoto A. Tashiro Y. J. Biol. Chem. 1990; 265: 15984-15990Abstract Full Text PDF PubMed Google Scholar, 25Krishna Rao A.S.M. Hausman R.E. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 2950-2954Crossref PubMed Scopus (32) Google Scholar, 26Kröning H. Kähne T. Ittenson A. Franke A. Ansorge S. Scand. J. Immunol. 1994; 39: 346-350Crossref PubMed Scopus (35) Google Scholar, 27Terada K. Manchikalapudi P. Noiva R. Jauregui H.O. Stockert R.J. Schilsky M.L. J. Biol. Chem. 1995; 270: 20410-20416Abstract Full Text Full Text PDF PubMed Scopus (123) Google Scholar) and thioredoxin (28Martin H. Dean M. Biochem. Biophys. Res. Commun. 1991; 175: 123-128Crossref PubMed Scopus (44) Google Scholar, 29Rubartelli A. Bajetto A. Allavena G. Wollman E. Sitia R. J. Biol. Chem. 1992; 267: 24161-24164Abstract Full Text PDF PubMed Google Scholar, 30Dean M.F. Martin H. Sansom P.A. Biochem. J. 1994; 304: 861-867Crossref PubMed Scopus (12) Google Scholar) have been shown to be secreted by cells. Incubation of purified plasmin with either reduced PDI or reduced thioredoxin resulted in formation of 45-, 41-, and 38-kDa plasmin fragments that all contained free sulfhydryl group(s) (Fig.3). The three plasmin fragments had the same Lys-Val-Tyr-Leu-Ser amino terminus, corresponding to the amino terminus of the angiostatin fragment made from plasmin by CHO cells (Fig. 1). In contrast, incubation of plasminogen with either reduced PDI or reduced thioredoxin did not perturb plasminogen structure by SDS-PAGE or generate free sulfhydryl groups in plasminogen that could incorporate MPB (not shown). Therefore, the fragments derive from plasmin by reduction and autoproteolysis of the enzyme. These findings are consistent with the results using CHO cells or conditioned medium (Fig. 1) and imply that autoproteolysis of reduced plasmin is sufficient for angiostatin formation. The reactions shown in Fig. 3were stopped before they had gone to completion to clearly delineate the similarly sized plasmin fragments by SDS-PAGE. No further products were evident when all the plasmin had been converted to the three angiostatin fragments (not shown). Incubation of plasmin with either oxidized PDI or oxidized thioredoxin did not perturb plasmin structure (Fig. 3 A), indicating that the active-site disulfide bond(s) in PDI and thioredoxin must be reduced to catalyze plasmin reduction. The ability of PDI or thioredoxin to generate angiostatin from plasmin is not based on thermodynamic considerations alone, because dithiothreitol, which is a better reductant than either PDI or thioredoxin, did not generate angiostatin (Fig. 3 A). These results support the contention that the mechanism of formation of angiostatin by CHO cells involved reduction of one or more plasmin disulfide bonds, which triggered proteolysis of the enzyme releasing fragment(s) containing the first four kringle domains. To determine whether the angiostatin fragments made by reduction and autoproteolysis of plasmin were active in cell culture, they were tested for their ability to inhibit proliferation of HDMVEC. O'Reillyet al. (10O'Reilly M.S. Holmgren L. Shing Y. Chen C. Rosenthal R.A. Moses M. Lane W.S. Cao Y. Sage E.H. Folkman J. Cell. 1994; 79: 315-328Abstract Full Text PDF PubMed Scopus (3138) Google Scholar) showed that angiostatin inhibited proliferation of cultured bovine capillary endothelial cells, with half-maximal effect at 0.2 μg/ml. Plasmin was incubated with reduced thioredoxin, and the angiostatin fragments were purified by lysine-Sepharose affinity and gel filtration chromatography (Fig.4 A). Thioredoxin was chosen as the plasmin reductase because of its general availability. The thioredoxin-derived fragments were potent inhibitors of HDMVEC proliferation (p < 0.05), with a half-maximal effect at approximately 0.2 μg/ml (Fig. 4 B). Importantly,S-carbamidomethylation of the free sulfhydryl group(s) in thioredoxin-derived angiostatin ablated its inhibitory activity. These findings imply that the free sulfhydryl group(s) in the angiostatin fragments are essential for their anti-proliferative activity. This poses interesting questions about mechanism of action, such as receptor recognition. In contrast to the inhibitory effects of thioredoxin-derived angiostatin on microvascular endothelial cell proliferation, thioredoxin-derived angiostatin did not significantly affect proliferation of macrovascular endothelial cells from bovine aorta or human umbilical vein (not shown). The mechanism for the specificity of thioredoxin-derived angiostatin for microvascular endothelial cells is unknown. Although PDI and thioredoxin can produce angiostatin from plasmin, these enzymes are relatively inefficient (see Fig. 3). It seemed likely that neither PDI nor thioredoxin would be secreted in sufficient quantities to account for the angiostatin generating activity of conditioned medium (compare Figs. 2 and 3). Immunodepletion of CHO-conditioned medium of PDI or thioredoxin did not affect angiostatin generating activity (not shown), confirming that neither of these enzymes are used by CHO cells. This result implies that CHO cells secrete a "plasmin reductase," which probably functions in the same way as PDI or thioredoxin but is much more efficient at catalyzing reduction of plasmin disulfide bond(s). We have partially characterized this plasmin reductase from the conditioned medium of HT1080 cells, a human fibrosarcoma cell line. We chose HT1080 cells because they convert plasminogen to plasmin, and plasmin to angiostatin, severalfold more efficiently than CHO cells (Fig.5). The plasmin reductase secreted by HT1080 cells requires two components for activity, a protein component that can be heat-inactivated and a low molecular mass cofactor that is probably GSH (Fig.6). Low micromolar concentrations of GSH fully activated the plasmin reductase. Cultured cells constituitively secrete GSH and accumulate low micromolar concentrations in their medium (31Bannai S. Tsukeda H. J. Biol. Chem. 1979; 254: 3444-3450Abstract Full Text PDF PubMed Google Scholar, 32Dethmers J.K. Meister A. Proc. Natl. Acad. Sci. U. S. A. 1981; 78: 7492-7496Crossref PubMed Scopus (214) Google Scholar). Therefore, the concentration of GSH required to activate the plasmin reductase is within the expected physiological range. Tumorigenesis is associated with marked changes in the plasminogen activation system at the tissue level. Increases in uPA, uPA receptor, and the inhibitors PAI-1 and PAI-2 are common. Plasmin is thought to play a critical role in facilitation of tumor invasion, metastasis, and angiogenesis (33Mignatti P. Rifkin D.B. Physiol. Rev. 1993; 73: 161-195Crossref PubMed Scopus (1174) Google Scholar). Our results suggest that plasmin can be both a positive and a negative regulator of tumor angiogenesis and metastasis through formation of angiostatin. It is plausible that angiostatin formation may be temporally and spatially controlled by release of plasmin reductase from tumor or host cells. In summary, cultured CHO or HT1080 cells secrete a plasmin reductase that reduces one or more disulfide bonds in plasmin, which triggers proteolysis of the enzyme releasing fragments with the same domain structure and biological activity as angiostatin. These results have consequences for plasmin activity and angiogenesis and are the first demonstration of extracellular reduction of a protein disulfide bond. We thank Dr. Kathryn Quinn for recombinant PAI-1 and Dr. Chris Jackson for the HDMVEC. We also thank Dr. Quinn for comments on the manuscript.