Human Procarboxypeptidase U, or Thrombin-activable Fibrinolysis Inhibitor, Is a Substrate for Transglutaminases

Procarboxypeptidase U (EC 3.4.17.20) (pro-CpU), also known as plasma procarboxypeptidase B and thrombin-activable fibrinolysis inhibitor, is a human plasma protein that has been implicated in the regulation of fibrinolysis. In this study, we show that pro-CpU serves as a substrate for transglutaminases. Both factor XIIIa and tissue transglutaminase catalyzed the polymerization of pro-CpU and the cross-linking to fibrin as well as the incorporation of 5-dimethylaminonaphthalene-1-sulfonyl cadaverine (dansylcadaverine), [14C]putrescine, and dansyl-PGGQQIV. These findings show that pro-CpU contains both amine acceptor (Gln) and amine donor (Lys) residues. The amine acceptor residues were identified as Gln2, Gln5, and Gln292, suggesting that both the activation peptide and the mature enzyme participate in the cross-linking reaction. These observations imply that transglutaminases may mediate covalent binding of pro-CpU to other proteins and cell surfaces in vivo. In particular, factor XIIIa may cross-link pro-CpU to fibrin during the latter part of the coagulation cascade, thereby helping protect the newly formed fibrin clot from premature plasmin degradation. Moreover, the cross-linking may facilitate the activation of pro-CpU, stabilize the enzymatic activity, and protect the active enzyme from further degradation.

Pro-CpU, 1 also known as plasma procarboxypeptidase B (1) and thrombin-activable fibrinolysis inhibitor (2), is a single chain 60-kDa glycoprotein secreted by the liver. The active enzyme is specific for Lys or Arg residues and is therefore referred to as a "basic" carboxypeptidase. Pro-CpU displays high protein sequence identity to other known members of the metallocarboxypeptidase family, including carboxypeptidase B from pancreas (1). The protein was discovered as a plasminogen-binding protein, and this property has led investigators to explore a role in fibrinolysis (1)(2)(3)(4)(5)(6)(7)(8).
Previous studies have shown that basic carboxypeptidases in general affect the stability of fibrin clots in vitro (9 -14). The mechanism most likely involves the removal of exposed COOHterminal Lys residues on the surface of a "mature" fibrin clot, thus limiting the number of Lys binding sites. COOH-terminal Lys residues are not found on a newly formed clot. However, initial partial plasmin degradation exposes these residues, which then act as binding sites for tissue plasminogen activator, plasminogen, and plasmin (15,16). Accordingly, the interactions between COOH-terminal Lys residues and Lys binding sites are recognized as potential targets for the regulation of the fibrinolytic system. Early studies examined the ability of pancreatic carboxypeptidase B and carboxypeptidase N to reduce the binding capacity of the fibrin clot (9,10,12,13). More recently, it has been shown that CpU similarly affects the stability of a fibrin clot in vitro (2)(3)(4)(5)(6)(7).
Pro-CpU binds to plasminogen (1), and active CpU binds to ␣ 2 -macroglobulin and pregnancy zone protein (22). Consequently, it is uncertain how CpU is maintained in the clotting milieu. In this study, we present evidence that pro-CpU is a substrate for transglutaminases and is cross-linked to fibrin in vitro. These observations may have important biological significance, and they suggest that pro-CpU is incorporated into the fibrin clot during the latter part of the coagulation cascade. The cross-linked pro-CpU may assist other proteins, including ␣ 2antiplasmin in the protection of the newly formed clot from premature plasmin degradation. Protein Purification-Pro-CpU was purified from human plasma by plasminogen-Sepharose affinity chromatography as described previously (1,24). Pro-CpU was separated from ␣ 2 -antiplasmin by anion exchange chromatography on a Mono-Q HR 5/5 column (Amersham Pharmacia Biotech) connected to a Amersham Pharmacia Biotech fast protein liquid chromatography system. The column was equilibrated in 20 mM Tris-Cl, pH 7.4 (Buffer A) and developed using a 0.3% Buffer B/min linear gradient from Buffer A to 20 mM Tris-Cl, pH 7.4, containing 1 M NaCl (buffer B) at a flow rate of 1 ml/min. Plasminogen, ␣ 2 -antiplasmin, and extracellular superoxide dismutase were purified as described previously (24,25,26).
Incorporation of Dansylcadaverine or Dansyl-PGGQQIV-Tissue transglutaminase or factor XIIIa was incubated with pro-CpU at 37°C in 50 mM Tris-Cl, 100 mM NaCl, pH 7.4, containing 10 mM Ca 2ϩ , 0.5 mM DTT, and 0.5 mM dansylcadaverine (28) or dansyl-PGGQQIV (23) at increasing enzyme:substrate concentrations (see Fig. 1). The reactions were continued for 3 h at 37°C before 10 mM EDTA was added. The reaction products were analyzed by reducing SDS-PAGE and examined under UV light.
Transglutaminase-induced Polymerization-Tissue transglutaminases or factor XIIIa was incubated with 125 I-labeled pro-CpU for 4 h at 37°C in 50 mM Tris-Cl, 100 mM NaCl, pH 7.4, containing 10 mM Ca 2ϩ , and 0.5 mM DTT at increasing enzyme:substrate concentrations in the presence of a protease inhibitor mixture composed of 2 mM phenylmethanesulfonyl fluoride and 0.02 mM E-64 (see Fig. 2). The reaction was terminated by the addition of 10 mM EDTA, and the reaction products were analyzed by reducing SDS-PAGE. The gels were stained for protein using Coomassie Blue, and radioactive bands were visualized in a PhosphorImager (Molecular Dynamics 410A).
Transglutaminase-catalyzed Incorporation of 1.4-[ 14 C]Putrescine-One mg of pro-CpU was incubated with 100 g of factor XIIIa or guinea pig tissue transglutaminase in 50 mM Tris-Cl, 100 mM NaCl, pH 7.4, containing 50 mCi 1.4-[ 14 C]putrescine, 10 mM Ca 2ϩ , and 0.5 mM DTT. The reaction was continued for 5 h at 37°C and chased for an additional 5 h with 10 mM putrescine. The reaction was stopped by the addition of 10 mM EDTA.
Identification of Acceptor Residue(s) within the Primary Structure-The 1.4-[ 14 C]putrescine-labeled pro-CpU (1 mg) was reduced by adding 5 mM DTT in the presence of 6 M guanidinium chloride in 100 mM Tris-Cl, pH 8. The reaction was quenched after 30 min by the addition of 10 mM iodoacetic acid. The reduced and alkylated sample was then dialyzed into 20 mM ammonium bicarbonate and digested with trypsin (1:50, w/w) for 4 h at 37°C. The generated peptides were lyophilized and separated by high performance size exclusion chromatography using a Superdex, Peptide HC 10/30 (Amersham Pharmacia Biotech) connected to a SMART system (Amersham Pharmacia Biotech) equilibrated in 20 mM phosphate buffer, 250 mM NaCl, pH 7.2. Aliquots of the collected fractions were counted in a scintillation counter (Beckman LS 5000 TD) and fractions containing 14 C radioactivity were pooled. These pools were further separated by reverse phase HPLC using a combination of columns including Vydac C18/5 m, 300 Å (250 ϫ 2.1 mm), Nucleosil C18/5 m, 300 Å (250 ϫ 2.0 mm), and Jupiter 300 C18/5 m, 300Å (250 ϫ 2.0 mm) (all columns were from Phenomenex, Torrence, CA). The columns were connected to an Applied Biosystems 130A HPLC system and eluted using linear gradients in 0.1% trifluoroacetic acid and acetonitrile. Peptides were monitored at 220 nm and collected manually. Radioactive peptides were further analyzed by Edman degradation and plasma desorption time of flight mass spectrometry.
Transglutaminase-catalyzed Cross-linking of Pro-CpU to Fibrin-MaxiSorp™ polystyrene tubes (Nunc) were coated with fibrin by incubating 200 l of human fibrinogen at 500 g/ml and 0.1 National Institutes of Health units of thrombin for 20 min. The tubes were emptied, washed three times with 50 mM Tris-Cl, 100 mM NaCl, pH 7.5 (Buffer A) and blocked overnight with 1% bovine serum albumin in Buffer A at 4°C. The next day, the tubes were washed with Buffer A, and 0.2 g of iodinated pro-CpU or 0.225 g of iodinated ␣ 2 -antiplasmin were added in triplicates. The cross-linking reactions were initiated by adding tissue transglutaminase or FXIIIa ranging from 0.05 to 10 g in Buffer A containing 10 mM CalCl 2 and 0.5 mM DTT. The volumes were adjusted to 100 l, and the reactions were continued for 4 h at 37°C. The tubes were washed three times in 1) Buffer A containing 10 mM EDTA, 2) 100 mM glycine-HCl, pH 3, and 3) Buffer A containing 1% SDS. After the last washing step, the tubes were counted for ␥-radioactivity.
Amino Acid Sequence Analysis-Automated Edman degradation was carried out in an Applied Biosystems 477A sequencer with on-line analysis of the phenylthiohydantoin (PTH) amino acids using an Applied Biosystems 120A HPLC. The remaining PTH-derivatives were collected and counted for 14 C radioactivity. In selected runs, the eluate from the 120A HPLC system was collected at timed intervals and counted for 14 C radioactivity to determine the retention of PTH-Gln-[ 14 C]putrescine. Some samples were transferred to polyvinylidene difluoride membranes before Edman degradation (27).
Amino Acid Analysis and Determination of Protein Concentrations-The concentrations of proteins were determined by amino acid analysis (29). The samples were desalted and the absorbances at 280 nm was determined. Aliquots were dried and hydrolyzed for 24 h in 6 M HCl, and the amino acid composition was determined in a Beckman 6300 amino acid analyzer. The determined absorbance coefficient for pro-CpU was A 1%, 1 cm ϭ 20.6. The concentration of other proteins were calculated according to the following absorbance coefficients: ␣ 2 -antiplasmin, A 1%, 1 cm ϭ 13.0 (30) and a fibrinogen A 1%, 1 cm ϭ 13.9 (31).
Plasma Desorption Mass Spectrometry-Peptides were analyzed in a BioIon 20K plasma desorption instrument (Applied Biosystems AB, Uppsala, Sweden). The peptides were dissolved in 0.1% trifluoroacetic acid and applied to nitrocellulose-covered targets. The spectrums were recorded at 15 kV for 1 ϫ 10 6 primary fission events. Further details of the instrumentation and spectral analysis have been described elsewhere (32).

Pro-CpU Is a Substrate for Tissue Transglutaminase and
Factor XIIIa-To test whether pro-CpU serves as a substrate for transglutaminases, we monitored the incorporation of fluorescent donor (dansylcadaverine) or acceptor (dansyl-PGGQ-QIV) substrates by tissue transglutaminase or factor XIIIa. Pro-CpU and the fluorescent probes were incubated with increasing amount of factor XIIIa or tissue transglutaminase. The reactions were continued for 3 h, and the products were analyzed by reduced SDS-PAGE and examined under UV light (Fig. 1). It is apparent that transglutaminases catalyzed the incorporation of both probes into pro-CpU in a Ca 2ϩ -dependent manner. Other proteins, not expected to be transglutaminase substrates, including human extracellular superoxide dismutase, bovine serum albumin, bovine carbonic anhydrase, ␤-amylase, hog thyroglobulin, horse ferretin, and bovine catalase, failed to incorporate the probes (not shown). However, the known acceptor (␣ 2 -antiplasmin) (33) and donor (plasminogen) (34,35) substrates acted as expected (Fig. 1). Taken together, these experiments suggest that pro-CpU incorporated dansylcadaverine and dansyl-PGGQQIV in a specific manner.
Because pro-CpU contains both reactive Gln (amine acceptor) and Lys (amine donor) residues, we tested whether the transglutaminases were able to form inter-pro-CpU N ⑀ (␥-glutamyl)lysine cross-links and produce pro-CpU homopolymers. Iodinated pro-CpU was incubated with increasing amounts of tissue transglutaminase or factor XIIIa (not shown), and the reaction products were analyzed by reduced SDS-PAGE (Fig.  2). The result indicated that transglutaminases catalyzed the formation of high molecular weight polypeptides and thus confirmed that pro-CpU is a substrate for transglutaminases. Significantly, the result also suggests that pro-CpU is capable of forming inter N ⑀ (␥-glutamyl)lysine cross-links with other proteins.
Identification of the Reactive Gln Residue(s) in Pro-CpU-Pro-CpU (1 mg) and tissue transglutaminase or factor XIIIa (100 g) were incubated in the presence of [ 14 C]putrescine, as described under "Experimental Procedures." The [ 14 C]putrescine labeling was performed under nondenaturing conditions using a low concentration of reducing agent (0.5 mM DTT) to sustain the activity of the transglutaminases. The disulfide integrity was not affected, even by an extended incubation in 0.5 mM DTT, as judged by the inability of pro-CpU to incorporate iodo[ 14 C]acetic acid (data not shown). Following the [ 14 C]putrescine labeling reaction, pro-CpU was reduced and alkylated and digested with trypsin. The peptides were separated by high performance size exclusion chromatography, and pools containing radioactive peptides were further purified by reverse phase HPLC, as described under "Experimental Procedures." The purified radioactive peptides were characterized by Edman degradation and mass spectrometry (Table I). Two radioactive tryptic peptides were recovered from both the tissue transglutaminase and the factor XIIIa labeling of pro-CpU: Phe 1 -(Gln 2 )-Ser-Gly-(Gln 5 )-Val-Leu-Ala-Ala-Leu-Pro-Arg 12 and Ala 283 -Tyr-Ile-Ser-Met-His-Ser-Tyr-Ser-(Gln 292 )-His-Ile-Val-Phe-Pro-Tyr-Ser-Tyr-Thr-Arg 302 . The yield of the two peptides (ϳ4 nmol) suggested that these represent the major amine acceptors. During Edman degradation of the two peptides PTH-Gln 2 , PTH-Gln 5 , and PTH-Gln 292 were not detected, suggesting that these residues were modified. This conclusion was substantiated by three observations: (i) the PTH-derivatives released during Edman degradation in the positions corresponding to Gln 2 , Gln 5 , and Gln 292 contained 14 C radioactivity; (ii) the retention time of the PTH-derivative eluting in these positions was consistent with the retention time of PTH-Gln-[ 14 C]putrescine (Table I); and (iii) the masses of peptide 1 and peptide 2 are consistent with the [ 14 C]putrescine modification described above (Table I).
Because some of the predicted tryptic peptides were rather large, we performed a second digest using pepsin to generate smaller, more manageable peptides. Pro-CpU and tissue transglutaminase or factor XIIIa were incubated in the presence of [ 14 C]putrescine, as described under "Experimental Procedures," and digested with pepsin. The analysis of these digests resulted in the purification of two major radioactive peptides containing Gln 2 and Gln 292 , as identified in the tryptic digests.
In addition, small amounts of [ 14 C]putrescine modified peptides, including Gln 16 and Gln 81 -Gln 82 -Gln 83 , were purified from the tissue transglutaminase-catalyzed reaction. The Gln 81 -Gln 82 -Gln 83 -containing peptide was obtained in several cleavage variants due to the low specificity of pepsin, and some of these fragments contained modified Gln 81 , whereas in others Gln 82 or Gln 83 was modified.
The analysis of several separate [ 14 C]putrescine labeling experiments suggested that Gln 2 , Gln 5 , and Gln 292 are the  a The numbers in parentheses represent the yields of each PTHderivative in pmol. The residues in brackets were detected as PTH-Gln-[ 14 C]putrescine but were not quantified (nq). During the standard protocol for separating PTH-derivatives, PTH-Gln eluted between PTH-Ser and PTH-Thr. In this system PTH-Gln-[ 14 C]putrescine coelutes with methylphenylthiocarbamate. The calculated mass does not include the mass of incorporated putrescine (ϩ71.1 Da).

FIG. 2. Evidence for the transglutaminase-induced polymerization of pro-CpU.
Iodinated pro-CpU was incubated with increasing amounts of tissue transglutaminase (TG) as indicated in the presence of protease inhibitors. After 4 h, the reaction products were analyzed by reduced SDS-PAGE and stained for protein using Coomassie Blue (A). Radioactive bands were visualized in a Phospho-rImager (B). It is apparent that pro-CpU was converted to high molecular weight polymers that did not enter the gel. EDTA inhibits the enzymatic activity of transglutaminases and was included in one of the reactions mixtures to show that the observed polymerization was transglutaminase-dependent. Similar results were obtained using factor XIIIa (data not shown). major amine acceptor sites in transglutaminase-catalyzed cross-linking reactions involving pro-CpU. These residues are positioned both in the activation peptide and in the mature enzyme, implying that N ⑀ (␥-glutamyl)lysine cross-links prevent the release of CpU upon activation (see Fig. 4).
Transglutaminase-catalyzed Cross-linking of Pro-CpU to Fibrin-The biological relevance of the observations described above was investigated by comparing the tissue transglutaminase-or FXIIIa-catalyzed incorporation of pro-CpU and ␣ 2antiplasmin in immobilized fibrin. The cross-linking was examined by incubating iodinated pro-CpU or ␣ 2 -antiplasmin and increasing amount of transglutaminase in fibrin coated polystyrene tubes. The tubes were washed extensively and the amounts of bound pro-CpU or ␣ 2 -antiplasmin were estimated by counting the ␥-radioactivity associated with the tubes (Fig.  3). The results showed that binding of the two proteins depended on the concentration of the transglutaminases and the availability of Ca 2ϩ . Significantly, the amount of cross-linked pro-CpU was comparable to that of ␣ 2 -antiplasmin, suggesting that the incorporation of pro-CPU is biologically relevant.

DISCUSSION
In this study, we present evidence that pro-CpU is a substrate for transglutaminases and is cross-linked to fibrin during the latter part of the coagulation cascade. Incorporation of dansylcadaverine [ 14 C]putrescine and dansyl-PGGQQIV demonstrates the presence of both amine acceptor and amine donor sites. This observation was substantiated by the formation of transglutaminase induced pro-CpU polymers. We determined the location of the amine acceptor sites to ascertain whether both the zymogen and the mature enzyme participated in the cross-linking reaction.
Three major amine acceptor sites, Gln 2 , Gln 5 , and Gln 292 , were preferred by both tissue transglutaminase and factor XIIIa. Some additional minor [ 14 C]putrescine incorporation was observed in Gln 16 and Gln 81 -Gln 82 -Gln 83 after a detailed analysis of the tissue transglutaminase reaction products (see Fig. 4). This incorporation accounted for only a minor fraction of the total incorporated radioactivity and might reflect differences in the substrate requirement of tissue transglutaminase or factor XIIIa. However, the major reactive Gln residues preferred by both transglutaminases were Gln 2 , Gln 5 , and Gln 292 . Two of the major reactive Gln residues are found near the NH 2 -terminal of pro-CpU. The presence of closely spaced reactive Gln residues is quite common and has been shown in other proteins, including plasminogen activator inhibitor-2 (36), ␤ A3-crystallin (37), the ␥-chain of fibrinogen (38), and vitronectin (39). It is not possible to derive a consensus sequence for the amine acceptor sites in proteins, but it is common that reactive Gln residues are positioned near the NH 2 or COOH terminus (40).
The transglutaminase-catalyzed cross-linking may influence the biology of pro-CpU in several important ways. Specifically, the reactive Gln residues are located both in the activation peptide and in the mature enzyme, suggesting that CpU remains associated with the matrix after activation. The enzymatic activity of CpU has been reported to be transient in vitro (4,41) because of a change in the thermodynamic stability after activation (42). It is likely that the N ⑀ (␥-glutamyl)lysine crosslinks and other noncovalent interactions promoted by the cross-linking increase the conformational stability and stabilize the enzymatic activity of CpU. Proteolytic inactivation of CpU is similarly less likely as a result of both increased thermodynamic stability and steric hindrance. These factors may also influence the activation kinetics of pro-CpU.
Fibrin functions as a matrix during fibrinolysis and helps confine the activities of the involved proteins to prevent potential damaging effects elsewhere. This is exemplified by the factor XIIIa-catalyzed cross-linking of ␣ 2 -antiplasmin to fibrin during the final stages of the blood coagulation (33,43,44). Similarly, pro-CpU activated outside the immediate vicinity of the clotting/fibrinolysis environment is likely to be bound to ␣ 2 -macroglobulin or pregnancy zone protein and removed (22). The fibrin incorporation of ␣ 2 -antiplasmin is presumably mediated by plasminogen, which has affinity for both fibrin and ␣ 2 -antiplasmin (45), thus bringing ␣ 2 -antiplasmin in close proximity to the fibrin. Significantly, plasminogen also has affinity for pro-CpU (1), and it is likely that this interaction similarly promotes a factor XIIIa mediated incorporation of pro-CpU in the newly formed fibrin clot.
The observations described in this study suggest that transglutaminases may mediate covalent binding of pro-CpU to

FIG. 3. Pro-CpU is cross-linked to fibrin by transglutaminases.
Polystyrene tubes were coated with fibrin and incubated with iodinated pro-CpU or iodinated ␣ 2 -antiplasmin in the presence of increasing amount of tissue transglutaminase (TG) as indicated. The tubes were washed and counted for ␥-radioactivity, and the results of three identical experiments were averaged. Both ␣ 2 -antiplasmin (q) and pro-CpU (ϫ) were cross-linked to fibrin by tissue transglutaminase in a concentration-dependent manner. Similar results were obtained using FXIIIa. In addition, the incorporation of ␣ 2 -antiplasmin (E) and pro-CpU (f) were abolished by the addition of 10 mM EDTA. These experiments show that pro-CpU is cross-linked to fibrin by transglutaminases. Moreover, the binding efficiency of the two proteins appeared to be similar, suggesting that the incorporation of pro-CpU is physiologically relevant.
FIG. 4. Schematic showing the location of amine acceptor sites. The major sites, including Gln 2 , Gln 5 , and Gln 292 , are utilized by factor XIIIa, and tissue transglutaminase are positioned near the NH 2 terminus and toward the COOH terminus. In addition, tissue transglutaminase incorporated [ 14 C]putrescine in Gln 16 and in a cluster of Gln residues, including Gln 81 , Gln 82 , and Gln 83 . Accordingly, because amine acceptor sites are found in both the activation peptide and in the mature enzyme, it is likely that pro-CpU remains associated with the fibrin clot following removal of the activation peptide.
other proteins and cell surfaces in vivo. In particular, the data suggest that pro-CpU, CpU, or both are incorporated into fibrin during a coagulation event and function in concert with other proteins (e.g. ␣ 2 -antiplasmin) to prevent premature dissolution of the fibrin clot. This interaction may also have important implications for the pro-CpU activation kinetics and may stabilize CpU after activation.