Interactions of serine proteases with cultured fibroblasts

This review summarizes the mechanisms by which several serine proteases, particularly urokinase, thrombin, and elastase, interact with cultured fibroblasts. Many of these studies were prompted by findings that interactions of these proteases with cells and the extracellular matrix are important in a number of physiologic and pathologic processes. Two main pathways have been identified for specific interactions of these proteases with fibroblasts. One involves surface binding sites for the free protease that appear to bind only one particular protease. An unusual feature collectively shared by the binding sites for urokinase, thrombin, and elastase is that the bound protease is not detectably internalized by the fibroblasts. The other pathway by which serine proteases interact with fibroblasts involves proteins named protease nexins (PNs). Three PNs have been identified. They are secreted by fibroblasts and inhibit certain serine proteases by forming a covalent complex with the protease catalytic site serine. The complexes then bind back to the fibroblasts via the PN portion of the complex and are internalized and degraded. Recent studies showing that the fibroblast surface and extracellular matrix accelerate the inactivation of thrombin by PN-1 support the hypothesis that the PNs control protease activity at and near the cell surface. The PNs differ from plasma protease inhibitors in their molecular properties, absence in plasma, site of synthesis, and site of clearance of the inhibitor:protease complexes.     

Inhibitors of Urokinase and Thrombin in Cultured Neural Cells

Recent studies have suggested important roles for certain proteases and protease inhibitors in the growth and development of the CNS. In the present studies, inhibitors of urokinase or thrombin in cultured neural cells and serum-free medium from the cells were identified by screening for components that formed sodium dodecyl sulfate-stable complexes with 125I-urokinase or 125I-thrombin. Rinsed glioblastoma possessed two components that complexed 125I-urokinase. One was type 1 plasminogen activator inhibitor (PAI-1), because the 125I-urokinase-containing complexes were immunoprecipitated with anti-PAI-1 antibodies. The other component formed complexes with 125I-urokinase that were not recognized by antibodies to PAI-1 or protease nexin-1 (PN-1). Its identity is unknown. In addition to these cell-bound components, the glioblastoma cells also secreted two inhibitors that formed complexes with 125I-urokinase; one was PAI-1, and the other was PN-1. The secreted PN-1 also formed complexes with 125I-thrombin. It was the only thrombin inhibitor detected in these studies. Human neuroblastoma cells did not contain components that formed detectable complexes with either 125I-urokinase or 125I-thrombin. However, human neuroblastoma cells did contain very low levels of PN-1 mRNA and PN-1 protein. Added PN-1 bound to the surface of both glioblastoma and neuroblastoma cells. This interaction accelerated the inhibition of thrombin by PN-1 and blocked the ability of PN-1 to form complexes with 125I-urokinase. Thus, cell-bound PN-1 was a specific thrombin inhibitor.(ABSTRACT TRUNCATED AT 250 WORDS)     

Effects of fibroblasts and endothelial cells on inactivation of target proteases by protease nexin-1, heparin cofactor II, and C1-inhibitor

Previous studies have shown that glycosaminoglycans in the extracellular matrix accelerate the inactivation of target proteases by certain protease inhibitors. It has been suggested that the ability of the matrix of certain cells to accelerate some inhibitors but not others might reflect the site of action of the inhibitors. Previous studies showed that fibroblasts accelerate the inactivation of thrombin by protease nexin-1, an inhibitor that appears to function at the surface of cells in extravascular tissues. The present experiments showed that endothelial cells also accelerate this reaction. The accelerative activity was accounted for by the extracellular matrix and was mostly due to heparan sulfate. Fibroblasts but not endothelial cells accelerated the inactivation of thrombin by heparin cofactor II, an abundant inhibitor in plasma. This is consistent with previous suggestions that heparin cofactor II inactivates thrombin when plasma is exposed to fibroblasts and smooth muscle cells. Neither fibroblasts nor endothelial cells accelerated the inactivation of C1s by plasma C1-inhibitor.     

Protease nexin-1 inhibits plasminogen activation-induced apoptosis of adherent cells

Degradation of adhesive glycoproteins by plasmin is implicated in cell migration. In this study, we further explored the role of plasminogen activation in cell adhesion and survival and show that uncontrolled plasminogen activation at the cell surface may induce cell detachment and apoptosis. We hypothesized that this process could be prevented in adherent cells by expression of protease nexin-1, a potent serpin able to inhibit thrombin, plasmin, and plasminogen activators. Using two- and three-dimensional culture systems, we demonstrate that Chinese hamster ovary fibroblasts constitutively express tissue-type plasminogen activator and efficiently activate exogenously added plasminogen in a specific and saturable manner (K(m) = 46 nm). The formation of plasmin results in proteolysis of fibronectin and laminin, which is followed by cell detachment and apoptosis. Protease nexin-1 expressed by transfected cells significantly inhibited the activity of plasmin and tissue-type plasminogen activator via the formation of inhibitory complexes and prevented cell detachment and apoptosis. In conclusion, protease nexin-1 may be an important anti-apoptotic factor for adherent cells. This cell model could be a useful tool to evaluate therapeutic agents such as serpins in vascular pathologies involving pericellular protease-protease inhibitor imbalance.     

Binding of type 1 plasminogen activator inhibitor to the extracellular matrix of cultured bovine endothelial cells

The binding of type 1 plasminogen activator inhibitor (PAI-1) to the extracellular matrix (ECM) of cultured bovine aortic endothelial cells was investigated using purified 125I-labeled or L-[35S]methionine-labeled PAI-1 as probes. Little specific binding of latent PAI-1 to ECM previously depleted of endogenous PAI-1 could be demonstrated. In contrast, the guanidine-activated form of PAI-1 bound to ECM in a dose- and time-dependent manner, and binding was saturable. The dissociation constant (Kd) for this interaction was estimated to be 60 nM by Scatchard analysis, and approximately 6 pmol of activated PAI-1 was bound per cm2 of ECM. Binding was relatively specific since unlabeled, activated PAI-1 competed with 35S-labeled PAI-1 for binding to ECM, but latent PAI-1 did not. Moreover, PAI-2, protein C inhibitor (i.e. PAI-3), protease nexin-1, and alpha 2-antiplasmin were not able to compete. Tissue-type plasminogen activator (tPA) also inhibited binding, but diisopropyl fluorophosphate-inactivated tPA did not. Pretreatment of ECM with tPA, urokinase-type PA, or thrombin had no effect on its ability to subsequently bind PAI-1, whereas trypsin, plasmin, and elastase pretreatment greatly reduced its ability to bind PAI-1. Guanidine-activated, radiolabeled PAI-1 resembled active endogenous PAI-1 since it was unstable in solution but stable when bound to ECM. In addition, it formed complexes with tPA that had a relatively low affinity for ECM. These data suggest that ECM of bovine aortic endothelial cells contains a protease-sensitive structure that binds active PAI-1 tightly and relatively selectively and that this association stabilizes PAI-1 against the spontaneous loss of activity that occurs in solution.     

The phosphatidylethanolamine-binding protein is the prototype of a novel family of serine protease inhibitors

Serine proteases are involved in many processes in the nervous system and specific inhibitors tightly control their proteolytic activity. Thrombin is thought to play a role in tissue development and homeostasis. To date, protease nexin-1 is the only known endogenous protease inhibitor that specifically interferes with thrombotic activity and is expressed in the brain. In this study, we report the detection of a novel thrombin inhibitory activity in the brain of protease nexin-1(-/-) mice. Purification and subsequent analysis by tandem mass spectrometry identified this protein as the phosphatidylethanolamine-binding protein (PEBP). We demonstrate that PEBP exerts inhibitory activity against several serine proteases including thrombin, neuropsin, and chymotrypsin, whereas trypsin, tissue type plasminogen activator, and elastase are not affected. Since PEBP does not share significant homology with other serine protease inhibitors, our results define it as the prototype of a novel class of serine protease inhibitors. PEBP immunoreactivity is found on the surface of Rat-1 fibroblast cells and although its sequence contains no secretion signal, PEBP-H(6) can be purified from the conditioned medium upon recombinant expression.     

Detection of serine proteinase inhibitors in human cornified cells

A screening test for serine proteinase inhibitors revealed trypsin and urokinase inhibitors in the extract of human cornified cells. No inhibition for α-chymotrypsin, thrombin or plasmin was detected. Characterization of the inhibitors separated with a Sephacryl S-200 gel column demonstrated that: 1) trypsin inhibitor with a molecular weight of 45,000 was labile to heat, acid and alkali and showed temporary inhibition, and 2) urokinase inhibitor with a molecular weight of 35,000 was found relatively stable and exhibited time dependent inhibition. Both were distinct from a known thiol proteinase inhibitor which showed high stability and immediate inhibition. Regulatory roles of serine proteinase inhibitors are postulated.     

Characterization of the receptor for protease nexin-I:protease complexes on human fibroblasts

Fibroblasts as well as several other cell types, secrete a number of protease inhibitors into their culture media. Among these inhibitors are the protease nexins, a class of proteins which covalently bind serine proteases, thereby inactivating their specific targets. Protease nexin-I, first discovered in human foreskin fibroblasts, binds thrombin, plasmin, and urokinase with high affinity, forming covalently linked complexes. Human fibroblasts bind complexes of protease nexin-I and its target protease via a cell-surface, high-affinity receptor. We have analyzed a number of characteristics of this receptor, and found them to be typical of class II receptors in general. At 4 degrees C binding of PN-I:protease complexes was competed by heparin. In addition, binding was independent of the particular protease bound to the PN-I; purified complexes of PN-I with thrombin or urokinase competed equipotently for [125]I-thrombin:PN-I binding. As the pH of the binding buffer was lowered, binding to cells increased. A twofold increase in binding was attained by lowering the pH from 7.5 to 4.5. This phenomenon was not due to irreversible, pH-induced changes to either the cell surface or the labeled complexes. At 37 degrees C, the removal of labeled complexes from culture medium was rapid; approximately 80% was removed by 4 hours under given conditions. The internalization of complexes was also very rapid, with an estimated ke (endocytic rate constant) of 1.0 min-1. At neutral pH, fibroblasts bind complexes in a saturable manner. Scatchard analysis yields a receptor number of 250,000 per cell and a Kd of 1 nM.     

Thrombospondin forms complexes with single-chain and two-chain forms of urokinase

Thrombospondin (TSP), an adhesive glycoprotein found in platelets and extracellular matrix, has been shown previously to interact with plasminogen and tissue plasminogen activator, resulting in efficient plasmin generation. We now demonstrate specific complex formation of TSP with both the single-chain and two-chain forms of urokinase (scuPA and uPA). Binding of uPA and scuPA to immobilized TSP was detected and quantified using colorimetric immunoassays and a functional amidolytic assay. Binding was time and concentration dependent with apparent affinity constants of 40-50 nM. Binding was not affected by serine protease inhibitors, EDTA, or epsilon-aminocaproic acid. scUPA and uPA bound to TSP retained functional activity. Using a sensitive amidolytic assay we found that TSP. scuPA complexes were efficiently converted to TSP. uPA by catalytic plasmin concentrations. Additionally, TSP.uPA complexes were found to have plasminogen-activating activity equivalent to fluid-phase uPA and to be protected from inhibition by plasminogen activator inhibitor type 1, the major plasma and matrix plasminogen activator inhibitor. Using immunohistochemical techniques, we also demonstrated co-distribution of TSP and uPA in normal and malignant breast tissue. Complex formation of TSP with uPA may serve to localize, concentrate, and protect these enzymes on cell surfaces and within the extracellular matrix, thereby providing a reservoir of plasminogen activator activity.     

Cell surface complex of cathepsin B/annexin II tetramer in malignant progression

The cysteine protease cathepsin B is upregulated in a variety of tumors, particularly at the invasive edges. Cathepsin B can degrade extracellular matrix proteins, such as collagen IV and laminin, and can activate the precursor form of urokinase plasminogen activator (uPA), perhaps thereby initiating an extracellular proteolytic cascade. Recently, we demonstrated that procathepsin B interacts with the annexin II heterotetramer (AIIt) on the surface of tumor cells. AIIt had previously been shown to interact with the serine proteases: plasminogen/plasmin and tissue-type plasminogen activator (tPA). The AIIt binding site for cathepsin B differs from that for either plasminogen/plasmin or tPA. AIIt also interacts with extracellular matrix proteins, e.g., collagen I and tenascin-C, forming a structural link between the tumor cell surface and the extracellular matrix. Interestingly, cathepsin B, plasminogen/plasmin, t-PA and tenascin-C have all been linked to tumor development. We speculate that colocalization through AIIt of proteases and their substrates on the tumor cell surface may facilitate: (1) activation of precursor forms of proteases and initiation of proteolytic cascades; and (2) selective degradation of extracellular matrix proteins. The recruitment of proteases to specific regions on the cell surface, regions where potential substrates are also bound, could well function as a 'proteolytic center' to enhance tumor cell detachment, invasion and motility.     

Characterization of the dexamethasone-induced inhibitor of plasminogen activator in HTC hepatoma cells

Incubation of HTC rat hepatoma cells with the synthetic glucocorticoid dexamethasone rapidly inhibits plasminogen activator (PA) activity secondary to the induction of a specific acid-stable inhibitor of plasminogen activation (Cwikel, B. J., Barouski-Miller, P.A., Coleman, P.L., and Gelehrter, T.D. (1984) J. Biol. Chem. 259, 6847-6851). We have further characterized this inhibitor with respect to its interaction with both urokinase and tissue plasminogen activator, and its protease specificity. The HTC PA inhibitor rapidly inhibits urokinase and tissue plasminogen activator with an apparent second-order rate constant of 3-5 x 10(7) M-1 X s-1. The inhibitor forms stable covalent complexes with both urokinase and tissue plasminogen activator, with which plasmin, trypsin, and factor Xa apparently do not compete. Complex formation is saturable and requires the active site of the PA. The mass of the inhibitor-PA complex is 50,000 daltons greater than that of PA alone, consistent with an Mr for the PA inhibitor of 50,000 as demonstrated directly by reverse fibrin autography. The HTC PA inhibitor does not inhibit thrombin and differs in its kinetic and biochemical properties from protease nexin.     

Reciprocal Modulation of Astrocyte Stellation by Thrombin and Protease Nexin-1

When cultured astroglia are treated with agents that elevate intracellular cyclic AMP, they become process-bearing stellate cells and resemble differentiated astrocytes in vivo. Thrombin rapidly reversed the stellation induced by dibutyryl cyclic AMP, forskolin, or isoproterenol in cultured rat astrocytes; half-maximal and maximal effects occurred at 0.5 and 8 pM, respectively. The proteolytic activity of thrombin was required for stellation reversal, as thrombin derivatized at its catalytic site serine with a diisopropylphospho group was inactive. Two thrombin inhibitors, protease nexin-1 and hirudin, blocked and reversed the effect of thrombin. The stellation reversal effect of thrombin was specific, as 300-1,000-fold higher concentrations of other serine proteinases, including plasmin, urokinase, trypsin, and T cell serine proteinase-1, were ineffective. Thrombin is a mitogen for astrocytes at concentrations in excess of 30 pM. Thrombin increased both cell number and ornithine decarboxylase activity, an early marker for mitogenic stimulation, in astrocyte cultures. The lowest thrombin concentrations that completely reversed astrocyte stellation, however, did not increase ornithine decarboxylase activity. Moreover, several other mitogens for astrocytes did not reverse dibutyryl cyclic AMP-induced stellation. Thus, the stellation reversal effect of thrombin is distinct from the mitogenic response.     

Neurite outgrowth activity of protease nexin-1 on neuroblastoma cells requires thrombin inhibition

Protease nexin-1 (PN-1) is a protein proteinase inhibitor recently shown to be identical with the glial-derived neurite-promoting factor or glial-derived nexin. It has been shown to promote neurite outgrowth in neuroblastoma cells and in sympathetic neurons. The present experiments were designed to further test the hypothesis that this activity on neuroblastoma cells is due to its ability to complex and inhibit thrombin. It has been suggested that PN-1:thrombin complexes might mediate the neurite outgrowth activity of PN-1. However, the present studies showed that such complexes, unlike free PN-1, did not promote neurite outgrowth. The neurite outgrowth activity of PN-1 was only detected in the presence of thrombin or serum (which contains thrombin). PN-1 did not affect the rate or extent of neurite outgrowth that occurred when neuroblastoma cells were placed in serum-free medium. Retraction of neurites by thrombin was indistinguishable in cells whose neurites had been extended in the presence or absence of PN-1. The neurite-promoting activity of PN-1 was inhibited by an anti-PN-1 monoclonal antibody, which blocks its capacity to complex serine proteinases. The plasma thrombin inhibitor, antithrombin III, stimulated neurite outgrowth but only when its thrombin inhibitory activity was accelerated by heparin. The neurite outgrowth activity of both antithrombin III and PN-1 corresponded to their inhibition of thrombin. Together, these observations show that PN-1 promotes neurite outgrowth from neuroblastoma cells by inhibiting thrombin and suggest that this depends on the ability of thrombin to retract neurites.     

The glioma cell-derived neurite promoting activity protein is functionally and immunologically related to human protease nexin-I

Protease nexin-I (PN-I, Mr approximately 43,000) is representative of a newly described class of cell-secreted protease inhibitors. PN-I has been purified to apparent homogeneity, partially sequenced, and monospecific antibodies have been raised against it. PN-I is a potent inhibitor of urokinase, thrombin, plasmin, and trypsin. In addition, cells have specific receptors that mediate the uptake of covalently linked complexes formed between PN-I and its protease substrates. In the present studies, we have investigated the relationship between human PN-I and a protease inhibitor derived from C6 glioma cells in culture that has neurite-promoting activity. On the basis of co-purification on heparin-Sepharose, identical molecular weight, antibody cross-reactivity, and receptor cross-reactivity, we conclude that PN-I and the glioma-cell-derived inhibitor are equivalent molecules.     

Extracellular proteolytic cleavage by urokinase is required for activation of hepatocyte growth factor/scatter factor.

The extracellular protease urokinase is known to be crucially involved in morphogenesis, tissue repair and tumor invasion by mediating matrix degradation and cell migration. Hepatocyte growth factor/scatter factor (HGF/SF) is a secretory product of stromal fibroblasts, sharing structural motifs with enzymes of the blood clotting cascade, including a zymogen cleavage site. HGF/SF promotes motility, invasion and growth of epithelial and endothelial cells. Here we show that HGF/SF is secreted as a single-chain biologically inactive precursor (pro-HGF/SF), mostly found in a matrix-associated form. Maturation of the precursor into the active alpha beta heterodimer takes place in the extracellular environment and results from a serum-dependent proteolytic cleavage. In vitro, pro-HGF/SF was cleaved at a single site by nanomolar concentrations of pure urokinase, generating the active mature HGF/SF heterodimer. This cleavage was prevented by specific urokinase inhibitors, such as plasminogen activator inhibitor type-1 and protease nexin-1, and by antibodies directed against the urokinase catalytic domain. Addition of these inhibitors to HGF/SF responsive cells prevented activation of the HGF/SF precursor. These data show that urokinase acts as a pro-HGF/SF convertase, and suggest that some of the growth and invasive cellular responses mediated by this enzyme may involve activation of HGF/SF.     

Glycosaminoglycans on fibroblasts accelerate thrombin inhibition by protease nexin-1.

Protease nexin-1 (PN-1) is a proteinase inhibitor that is secreted by human fibroblasts in culture. PN-1 inhibits certain regulatory serine proteinases by forming a covalent complex with the catalytic-site serine residue; the complex then binds to the cell surface and is internalized and degraded. The fibroblast surface was recently shown to accelerate the rate of complex-formation between PN-1 and thrombin. The present paper demonstrates that the accelerative activity is primarily due to cell-surface heparan sulphate, with a much smaller contribution from chondroitin sulphate. This conclusion is supported by the effects of purified glycosaminoglycans on the second-order rate constant for the inhibition of thrombin by PN-1. Also, treatment of 35SO4(2-)-labelled cells with heparitin sulphate lyase or chondroitin sulphate ABC lyase demonstrated two discrete pools of 35S-labelled glycosaminoglycans; subsequent treatment of plasma membranes with these glycosidases showed that heparitin sulphate lyase treatment abolished about 80% of the accelerative activity and chondroitin sulphate ABC lyase removed the remaining 20%. These results show that two components are responsible for the acceleration of PN-1-thrombin complex-formation by human fibroblasts. Although dermatan sulphate is also present on fibroblasts, it did not accelerate the inhibition of thrombin by PN-1.     

Vitronectin regulates the synthesis and localization of urokinase-type plasminogen activator in HT-1080 cells.

The effect of extracellular matrix composition on the location, amount, and activity of cell-associated urokinase-type plasminogen activator was tested using HT-1080 cells adherent to either fibronectin or vitronectin. Specific immunoprecipitation of newly synthesized urokinase indicated that cells adherent to fibronectin synthesized 2-3-fold more urokinase than cells adherent to vitronectin. Complexes of urokinase and plasminogen activator inhibitor type 1 (PAI-1) were detected in cell layers of vitronectin-adherent but not fibronectin-adherent cells. Inhibition of PAI-1 using a neutralizing monoclonal antibody resulted in a 3-fold increase in urokinase enzymatic activity on vitronectin adherent cells. Urokinase activity on fibronectin adherent cells was only slightly increased following PAI-1 neutralization. Examination of both HT-1080 and normal human fibroblast cells by immunofluorescent microscopy localized urokinase-type plasminogen activator to discrete, focal areas underneath cells adherent to vitronectin. Urokinase was not detectable by immunofluorescence on cells adherent to fibronectin. The addition of exogenous prourokinase to locate urokinase receptors on adherent HT-1080 cells indicated that the focal localization of cell-surface urokinase resulted from the clustering of urokinase receptors following adhesion to vitronectin but not fibronectin-coated substrates. These results suggest that vitronectin can contribute to the control of cell-surface plasmin activity by regulating the synthesis of urokinase and directing the localization of urokinase receptors.     

Binding uptake and degradation of antithrombin III X protease complexes by cultured corneal endothelial cells

Interaction of 125I-labeled human antithrombin III (125I-AT III) X protease complexes with bovine corneal endothelial cells has been studied in tissue culture. 125I-AT III does not bind to endothelial cells, but its complexes with either thrombin or trypsin bind specifically to the cultures. The binding of 125I-AT III X protease complexes is not via the moiety of the free antithrombin III (AT III) or the free protease, since neither AT III nor thrombin compete on the binding of 125I-AT III X thrombin complexes. Only unlabeled AT III X thrombin complexes compete on the binding of the iodinated ligand. 125I-AT III X trypsin complexes bind with a KD of 1.4 X 10(-7) M to high affinity-binding sites present on the cell surface of corneal endothelial cells. Saturation of binding to the cell surface is observed at a concentration of 2.5 X 10(-7) M 125I-AT III X trypsin complexes and the number of binding sites per cell is about 4 X 10(4). The cell surface binding reaches a maximum by 15 min and then decreases with time. The cells, when incubated at 37 degrees C, appear to internalize the bound complexes by adsorptive endocytosis which proceeds at a rate of 0.5-0.8 pmole/1 X 10(6) cells/h. The internalization process of 125I-AT III X protease complexes is saturated at a concentration of 2.5 X 10(-7) M. Since the cells release 125I-labeled material into the extracellular media which cannot be precipitated by trichloroacetic acid (TCA), it probably represents degradation of 125I-AT III X protease complexes into small fragments at a linear rate of about 0.5 pmole/1 X 10(6) cells/h. The described process of AT III X protease complexes binding, internalization and subsequent degradation by corneal endothelial cells may represent a clearing mechanism for extracellular AT III X protease complexes formed under pathological conditions.     

An inhibitor of plasminogen activation from human placenta. Purification and characterization

Placental extracts contain inhibitors of human urinary urokinase. These extracts form a heterogeneous population of complexes with 125I-urokinase that are recognizable by changes in gel filtration profile and mobility during sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Treatment with reducing agents eliminated the size heterogeneity without loss of activity, thereby allowing the placental inhibitor to be purified. Active inhibitor has been isolated in apparently homogeneous form after an eight-step procedure that included salt extraction, ammonium sulfate fractionation, column chromatography on CM-cellulose, DEAE-Sepharose, and hydroxylapatite, chromatofocusing, preparative gel electrophoresis, and hydrophobic chromatography. The purified inhibitor has Mr = 47,000. The inhibitor is relatively specific for plasminogen activators since it does not inhibit the action of plasmin, factor XIIa, plasma kallikrein, or thrombin. The inhibitor forms complexes with 1:1 stoichiometry that block the active sites of urokinase (but not prourokinase) and both one- and two-chain forms of tissue plasminogen activator. The stability of these complexes in sodium dodecyl sulfate-polyacrylamide gel electrophoresis suggest that they are based on covalently bonded structures. Although both types of plasminogen activator are inhibited, the rate of interaction is significantly faster with urokinase, tissue plasminogen activator being inhibited less efficiently. The complexes formed can be dissociated by mild alkali or hydroxylamine, thereby regenerating both enzymes and inhibitor at their original molecular weights. The results suggest that the complexes are stabilized by ester-like bonds; these might involve the hydroxyl of serine at the active site of the proteases and a carboxyl group in the inhibitor.