Plasmin Converts Factor X from Coagulation Zymogen to Fibrinolysis Cofactor*

Known anticoagulant pathways have been shown to exclusively inhibit blood coagulation cofactors and enzymes. In the current work, we first investigated the possibility of a novel anticoagulant mechanism that functions at the level of zymogen inactivation. Utilizing both clotting and chromogenic assays, the fibrinolysis protease plasmin was found to irreversibly inhibit the pivotal function of factor X (FX) in coagulation. This was due to cleavage at several sites, the location of which were altered by association of FX with procoagulant phospholipid (proPL). The final products were ∼28 and ∼47 kDa for proPL-bound and unbound FX, respectively, which did not have analogues when activated FX (FXa) was cleaved instead. We next investigated whether the FX derivatives could interact with the plasmin precursor plasminogen, and we found that plasmin exposed a binding site only on proPL-bound FX. The highest apparent affinity was for the 28-kDa fragment, which was identified as the light subunit disulfide linked to a small fragment of the heavy subunit (Met-296 to ∼Lys-330). After cleavage by plasmin, proPL-bound FX furthermore was observed to accelerate plasmin generation by tissue plasminogen activator. Thus, a feedback mechanism localized by proPL is suggested in which plasmin simultaneously inhibits FX clotting function and converts proPL-bound FX into a fibrinolysis cofactor. These data also provide the first evidence for an anticoagulant mechanism aimed directly at the zymogen FX.

Blood coagulation and fibrinolysis are opposing hemostatic processes. Whereas the coagulation pathway produces insoluble fibrin clot to seal vascular leaks, the fibrinolysis pathway solubilizes the clot to restore normal blood flow. It is obvious that for either of these pathways to be effective, their amplification must be sequential. Several coordinating mechanisms have been identified that each control the availability of fibrinolysis cofactors. The best described of these is the production of fibrin (1), which, unlike its precursor fibrinogen, is a fibrinolysis cofactor that colocalizes the zymogen plasminogen and an activating protease, tissue plasminogen activator (tPA), 1 for subsequent plasmin generation. Plasmin directly mediates fi-brin solubilization by proteolysis. In order for plasminogen to assemble into the ternary activation complex, the cofactor molecule must contain a COOH-terminal Lys, which is an essential constituent of the plasminogen binding site (2)(3)(4)(5)(6)(7)(8)(9)(10). Carboxypeptidases in plasma have been identified that excise COOH-terminal Lys and thereby limit the generation of plasmin (11)(12)(13) by restricting fibrinolysis cofactor function. In particular, an important regulatory role has been demonstrated for one of these carboxypeptidases, TAFI, which has been shown to first become activated and then later inactivated by the final coagulation enzyme, thrombin (12)(13)(14). Thus, thrombin orchestrates a lag between coagulation and the amplification of fibrinolysis to ensure sufficient clot production.
An additional mechanism of communication between coagulation and fibrinolysis has recently been reported. This involves the enzyme complex responsible for physiological thrombin production, prothrombinase (15,16). Prothrombinase consists of the serine protease factor Xa (FXa), and the non-enzymatic cofactor Va (FVa), which in the presence of Ca 2ϩ associate with each other and a procoagulant phospholipid (proPL)-containing surface. Together the constituents of prothrombinase enhance the FXa-mediated activation of prothrombin to thrombin by 5 orders of magnitude (17). Previous studies have shown that plasmin can inhibit the prothrombinase cofactor function of FVa (18) and the procofactor V (19). We have extended these findings to show that plasmin also proteolytically inhibits the coagulation activity of FXa, and furthermore converts both FXa and FVa into fibrinolysis cofactors (9,20). Interestingly, the acquisition of fibrinolysis function was dependent on the prior association of FXa or FVa with proPL. This may provide a means to accelerate plasmin generation specifically at sites of vascular damage, where the exposure of proPL is selectively triggered.
The physiological generation of FXa requires recruitment of its inactive precursor, factor X (FX), to sites of vascular damage through an association with proPL. Therefore, following sufficient fibrin formation to initiate fibrinolysis, it is conceivable that localized plasmin production may modulate FX function. We now report that FX is a substrate for plasmin, but the final proteolytic products differ from that of FXa. The cleavage of proPL-bound FX by plasmin results in a diametric functional change that inhibits participation in coagulation and simultaneously induces the expression of fibrinolysis cofactor activity.
were prepared and quantified as described (16).
Proteins-Human coagulation factor X (FX) was purified from fresh frozen plasma (21) or from prothrombin complex concentrate diluted to 1 unit/ml factor X clotting activity (obtained as generous gifts from the Canadian Red Cross Society, London Collection Center, and from Bayer, Inc., respectively). For comparison to in-house preparations, FX was also purchased (Hematologic Technologies, Inc.). Human FXa (22,23) and Lys-plasminogen were produced, as described (25). Lys-plasminogen was radioiodinated (50,000 -150,000 dpm/g) using Iodogen (Pierce) and chromatographically desalted (Excellulose 5, Pierce) to remove unincorporated 125 I. Purified human plasmin was purchased (Hematologic Technologies, Inc.).
FX Enzymatic Activity Assays-The effect of plasmin treatment of FX on the generation of FXa coagulation and amidolytic activity was evaluated in plasmin digestion time courses identical to those described above. At each time point, aprotonin (50 KIU/ml) was added to stop further effects of plasmin. Clotting activity was measured by thromboplastin clotting time assays using FX-deficient plasma (Sigma) reconstituted with 5 units/ml purified FX treated with plasmin as a substrate. The amount of FX activation was determined by comparison to a standard curve obtained by titrating purified FXa in the presence of precomplexed plasmin and aprotonin to duplicate the conditions of the time course samples. The generation of FXa amidolytic activity at 22°C in HBS using S2765 (200 M) as a chromogenic substrate was followed after treatment of plasmin-digested FX with RVV as described above. The rate of color development was monitored in a kinetic multiwell plate reader (V max , Molecular Devices).
Ligand Blots-To identify FX-derived species that interact with plasminogen, ligand blotting experiments were performed according to methods published by several laboratories (5,3,2). In our studies, FX fragments were separated by SDS-PAGE and electrotansferred to PVDF (27,28). The PVDF was blocked overnight at 4°C in bovine serum albumin (Sigma, 10 mg/ml) and then incubated with 125 I-Lysplasminogen (0.1 M) for 1 h at 22°C in the presence of the protease inhibitors GEMSA (50 nM) and aprotonin (50 KIU/ml, Calbiochem). Following extensive washing with HBS, the location of bound 125 Iplasminogen was determined by autoradiography and compared with the electrophoretic patterns made visible by staining the PVDF with Coomassie Brilliant Blue R-250.
Effect of Plasmin-treated FX on tPA Activity-The activation of plasminogen (1 M) to plasmin by single-chain tPA (10 nM) was followed as a time course utilizing the chromogenic substrate S2366. To determine whether plasmin-treated FX affects tPA function, FX was treated with plasmin (or left untreated) exactly as described above for 30 min. Just enough aprotonin (9.5 KIU/ml) was added to inhibit the plasmin used in this step prior to the addition of tPA and plasminogen. To control for the possible effects of added aprotonin and plasmin, reactions that did not involve a plasmin treatment step contained aprotonin and plasmin that had been preincubated.
Amino Acid Sequencing-The subunit composition of plasmin-digested FX products was investigated both by nonreducing SDS-PAGE and by two-dimensional SDS-PAGE involving a first dimension under nonreducing conditions and a second dimension under reducing conditions (29,20). To identify proteolytic cleavage sites, electrophoretic patterns were transferred to PVDF, and bands were excised for sequence analysis (27,28).

Inactivation of FX Coagulation
Activity by Plasmin-To determine whether FX coagulation activity is altered upon pretreatment with plasmin, Fig. 1 shows FXa-dependent thromboplastin clotting (panel A) and RVV-initiated amidolytic (panel B) assays. Both methods demonstrated that plasmin inhibits the coagulation enzyme function of FX. This was observed regardless of whether binding of FX to proPL was facilitated by Ca 2ϩ .
Cleavage of FX by Plasmin-The demonstration that FX enzymatic activity is lost due to plasmin pretreatment strongly indicates that FX is cleaved by plasmin. To provide direct evidence, changes in plasmin-treated FX electrophoretic mobility were followed using the same digestion conditions and pretreatment times as in Fig. 1. As demonstrated by the SDS-PAGE shown in Fig. 2, plasmin sequentially cleaves FX at several sites. Interestingly, the proteolytic sites recognized by plasmin depended on whether Ca 2ϩ was present to facilitate FX-proPL interactions. It should be noted that minor amounts of the alternate proteolytic products were detected under both reaction conditions, indicating that one pathway is favored over the other but is not exclusive. The nomenclature adopted previously for Pn-mediated FXa fragments (20) was used for the analogous FX fragments observed here. However, two additional species were observed for FX that had apparent molecular masses of 28 kDa ( Fig. 2A) and 47 kDa (Fig. 2B) and resulted from cleavage by plasmin in the presence of Ca 2ϩ and EDTA, respectively.
To determine whether only binding of FX to Ca 2ϩ was the determinant in altering the cleavage sites recognized by plasmin, the 30-min time points shown in Fig. 2 were compared with identically treated FX in the presence of Ca 2ϩ but no proPL. These data are shown in Fig. 3 and demonstrate that Ca 2ϩ alone is insufficient to generate the fragmentation profile that resulted in the presence of Ca 2ϩ and proPL. Thus, binding to proPL and not just to Ca 2ϩ is suggested to alter the cleavage of FX by plasmin.
To further compare the cleavage of FX and FXa by plasmin, the same fragmentation time courses were conducted as in Fig.  2 except that prior to electrophoresis, the FX-derived fragments generated by plasmin were treated with RVV to remove the activation domain. The resulting SDS-PAGE profiles are shown in Fig. 4, with the mobility of plasmin-mediated species generated directly from purified FXa indicated. As shown in Fig. 4A, when FX␤ Pn, Ca and FX␥ Pn, Ca were treated with RVV under conditions that favored binding to proPL, the resulting species had electrophoretic migrations identical to FXa␤ Pn, Ca and FXa␥ Pn, Ca , respectively. This suggests that plasmin cleaves the same sites in FX and FXa to generate these fragments. However, the 28-kDa product was unique to FX bound to proPL, and RVV had no effect on the mobility of this fragment.
In contrast to FX␥ Pn, Ca , but similar to the further proteolyzed 28-kDa species, the data presented in Fig. 4B show that RVV does not cleave FX␥ Pn, EDTA over the indicated time. This is additional evidence that two pathways of FX proteolysis by plasmin are followed, depending on whether the prior interaction with proPL is facilitated.
Activation of FX Fibrinolysis Function by Plasmin-Because plasmin liberates COOH-terminal basic amino acids and COOH-terminal Lys is a requisite for plasminogen binding to known fibrinolysis cofactors, we investigated whether the proteolysis of FX by plasmin generated species capable of participating in fibrinolysis. As an initial step, we determined whether the plasmin-mediated FX fragments were capable of interacting with plasminogen. To do so, the SDS-PAGE patterns shown in Fig. 2 were transferred to PVDF and probed with 125 I-plasminogen. The resulting autoradiographs are shown in Fig. 5. Binding of 125 I-plasminogen to untreated FX was not detectable. However, under conditions that facilitated proPL binding (Fig. 5A), plasmin induced the expression of a receptor site(s) in FX␤ Pn, Ca , FX␥ Pn, Ca , the 28-kDa fragment, and the ϳ13-kDa fragment. Using this ligand blotting method, the species clearly having the highest apparent affinity for 125 I-plasminogen was the 28-kDa fragment for as yet unknown reasons. These observations suggested that plasmin liberates a COOH-terminal Lys on each of these FX fragments.
Under conditions that did not facilitate binding of FX to proPL during the plasmin pretreatment step (Fig. 5B), we observed 125 I-plasminogen binding only marginally to FX␤ Pn, EDTA and a species that migrated with the dye front. The former appeared to be present as a minor constituent of our starting FX preparation and is likely independent of plasmin pretreatment. No interaction was detectable between 125 I-plasminogen and the other major cleavage products, FX␥ Pn, EDTA and the 47-kDa fragment, suggesting that a COOH-terminal Lys is not made available by proteolysis unless binding to proPL is facilitated .
Based on our ligand blotting experiments, the 28-kDa fragment in particular was suggested to have potential as a fibrinolysis cofactor. Therefore, we followed the effect of plasmincleaved FX on the activation of plasminogen by tPA using a chromogenic assay for plasmin (Fig. 6). As shown, a mixture of FX␥ Pn, Ca and the 28-kDa species produced by pretreatment of proPL-bound FX with plasmin for 30 min (final concentration, 0.5 M FX) resulted in a 27-fold enhancement of tPA activity (Fig. 6A). In contrast, when either native FX or FX pretreated with plasmin to produce FX␥ Pn, EDTA and the 47-kDa fragment (Fig. 6B) were used, insignificant effects on tPA activity were observed. As shown in Fig. 6C, acceleration of tPA by the FX␥ Pn, Ca and the 28-kDa species mixture was titratable to levels below the circulating concentration of FX (0.17 M).
Identity of FX-derived Fibrinolysis Cofactor(s)-Because the cleavage sites recognized by plasmin to produce the 28-kDa fragment are unique to FX-bound to proPL, and because this species was found to be a good tPA accelerator and plasminogen-binder, we conducted amino acid sequence analyses to locate the specific cleavage sites. Table I summarizes these data. When the nonreduced FX␥ Pn, Ca band was analyzed, two equimolar sequences were revealed that corresponded to the NH 2 terminus of the heavy and light subunits of FX. Determination of the ϳ13-kDa species amino acid sequence, which appeared concomitant with FX␥ Pn, Ca , revealed that FX␥ Pn, Ca is generated due to cleavage of the heavy subunit at Gly-331 (based on amino acid numbering adopted previously (30)).
When the nonreduced 28-kDa species was excised from PVDF and sequenced, three equimolar residues per cycle were observed. These corresponded to the NH 2 terminus of the heavy and light subunits, in addition to an internal heavy chain sequence beginning at Met-296. To further investigate this mixture of fragments, two-dimensional SDS-PAGE was conducted in which the 28-kDa species was first electrophoresed nonreduced and then subjected to a second dimension under reducing conditions (not shown). Two species were identified after transfer to PVDF and were sequenced. The mobility of the first was not affected by reduction and yielded a sequence identical to the NH 2 terminus of the FX heavy subunit. The second reduced fragment had an apparent molecular mass of approximately 20 kDa and migrated exactly to the position of the native FX light subunit. NH 2 -terminal amino acid sequence analysis confirmed its identity as the light subunit. These data cumulatively suggested that the nonreduced 28-kDa band was a co-migrating mixture of an NH 2 terminus-derived heavy subunit fragment ending at Arg-295 and a heterodimer consisting of the intact light subunit disulfide-linked to a small heavy chain fragment. Under reducing conditions, the small heavy chain fragment was not detectable after transfer to PVDF, and consequently only the amino acid sequence under nonreducing conditions was obtained.

DISCUSSION
Prothrombinase is the final enzyme complex generated within the coagulation cascade and functions directly to generate thrombin. Consequently, numerous modes of regulation of prothrombinase assembly and activity have evolved to enable fine-tuning of thrombin production (31). These include anticoagulant mechanisms that exclusively target the prothrombinase constituent cofactor FVa (32) and enzyme FXa (33), as well as the upstream tenase complexes, which are responsible for activating FXa from its inactive precursor (32,34). In the current study, we investigated whether the fibrinolysis effector protease plasmin can modulate the properties of the zymogen FX. Utilizing functional assays initiated by tissue factor or RVV FX activator, we show that plasmin is able to function as an anticoagulant through proteolytic inhibition of FX. To our knowledge this is the first evidence for an anticoagulant mechanism that operates directly at the level of FX inactivation.
Because proteolysis by plasmin is capable of producing COOH-terminal Lys, which is an essential feature of known fibrinolysis cofactors (2)(3)(4)(5)(6)(7)(8)(9)(10), the possibility that plasmincleaved FX acquires fibrinolytic activity was investigated. We found that in addition to having an anticoagulant effect on FX, plasmin converted the FX into a fibrinolysis cofactor. This conclusion is based on experiments that showed plasmin induces the expression of a plasminogen binding site on FX and the ability to accelerate the rate of tPA-dependent plasmin generation by approximately 30-fold. The latter effect was observed below the circulating concentration of FX. This supports the hypothesis that at sites of clot formation, where proPL increases the local concentration of coagulation proteins (15), the in vitro effects of Pn-cleaved FX on tPA reported here may have a physiological function in concert with other tPA accelerators, notably fibrin. Because the expression of FX-derived tPA cofactor activity was observed only under conditions that facilitated FX-proPL interactions, the proposed pathway would be advantageously localized to sites of clot formation.
To understand the structural basis by which FX function is altered by plasmin, Fig. 7 shows a fragmentation map of FX bound to proPL. The first plasmin-mediated cleavage produces FX␤ Pn, Ca , which has the same electrophoretic mobility as previously documented FX␤ generated by treatment of FX with FXa (35). The location of the cleavage site that excises the ␤-peptide is inferred and is based partly on the supposition that FX␤ Pn, Ca must have a newly exposed COOH-terminal Lys to account for the observed plasminogen binding. Autolytic cleavages at Lys-435 and Lys-433 have been reported for FXa, with the latter representing a minor product (36). Thus, we have positioned the first plasmin-mediated cleavage of FX to be consistent with the former.
The second cleavage of proPL-bound FX by plasmin produced FX␥ Pn, Ca and a 13-kDa species. Amino acid sequence analysis of the former under nonreducing conditions produced two sequences corresponding to the NH 2 terminus of the FX heavy and light subunits. Sequence analysis of the latter indicated cleavage of the heavy subunit within the "autolysis loop" between Lys-330 and Gly-331, which is the same as that previously reported for FXa cleaved under identical conditions (20). FIG. 4. Comparison of plasmin-mediated FX and FXa cleavage products. Time courses identical to those shown in Fig. 2 were conducted, except that at each time point, aprotonin (50 KIU/ml) was added followed by treatment with RVV (10 g/ml) at 37°C for 1 h in 10 mM CaCl 2 to excise the FX activation fragment. The electrophoretic mobility of plasmin-mediated species generated directly from purified FXa is indicated. The FX was pretreated with plasmin exactly as described in Fig. 2 for 30 min. Just enough aprotonin (9.5 KIU/ml) was added to inhibit the plasmin used in this step prior to the addition of tPA and plasminogen. A, conducted in the presence of Ca 2ϩ (2 mM) and proPL (300 M), as in Fig. 2A. B, conducted in the presence of EDTA (5 mM) and proPL (300 M), as in Fig. 2B. C, FX was plasmin-pretreated as in A for 30 min, and the concentration dependence of tPA activity was monitored. The data were corrected for endogenous tPA activity. (n ϭ 3; mean Ϯ S.D.). Unlike the recently reported proteolysis of FX by FXa (37), cleavage in the autolysis loop by plasmin was not observed by us to be protected by Ca 2ϩ . The exposure of a new COOHterminal Lys at position 330 is consistent with our finding that plasminogen interacts with FX␥ Pn, Ca in ligand blotting experiments. Furthermore, cleavage at this site explains the loss of FX coagulation function because disruption of the protease domain results.
Continued incubation of proPL-bound FX with plasmin resulted in a third sequential cleavage that produced a 28-kDa band under nonreducing conditions. Amino acid sequencing identified a heterodimer consisting of the light subunit disulfide linked to a small piece of the heavy subunit starting at Met-296 and presumably extending to Lys-330. The small heavy subunit fragment encompasses Cys-302, which is known to link the FX subunits (38). The end position at Lys-330 is inferred from the prior cleavage known to expose Gly-331 as the NH 2 terminus of the 13-kDa fragment. Immunoblot analysis utilizing the well studied monoclonal antibody H11 (39) specific for the Gla-containing domains of several proteins, including residues 4 -9 of FX, confirmed that the light chain is a constituent of the 28-kDa heterodimer (not shown). Interestingly, cleavage at Met-296 was not observed after an identical incubation of FXa with plasmin (20), suggesting that the activation domain in FX influences the accessibility of cleavage sites.
Comigrating with the 28-kDa heterodimer was the expected fragment corresponding to the NH 2 -terminal portion of the heavy subunit produced by cleavage at Met-296. Reduction changed the electrophoretic mobility of the 28-kDa heterodimer containing the light subunit, but not the mobility of the fragment corresponding to the NH 2 -terminal heavy subunit fragment. Compared with the other FX fragments generated by plasmin, the apparent affinity observed for plasminogen was clearly strongest for the nonreduced 28-kDa species. The prediction that a COOH-terminal Lys is available in the heterodimer (Lys-330) but not in the single chain heavy subunit fragment (predicted COOH terminus at Arg-295) strongly suggests that the former is responsible for the observed plasminogen binding. Consistent with this conclusion is the lack of plasminogen binding to the 28-kDa monomer under reducing conditions (not shown). The basis for comparatively strong plasminogen binding to the 28-kDa heterodimer is under investigation.
When conditions were used to prevent the interaction of FX with proPL (i.e. in the presence of EDTA), the electrophoretic pattern due to plasmin cleavage was altered. At present, the specific cleavage sites have not been determined by amino acid sequencing. However, immunoblot analysis revealed that upon production of FX␥ Pn, EDTA , antigenic recognition by the Gla domain-specific antibody, H11, is lost (not shown). Therefore, unlike our findings for proPL-bound FX, cleavage of FX in the absence of proPL binding resulted in rapid cleavage of the light chain. This is consistent with our previous report (20) showing that the light subunit of FXa is cleaved by plasmin in the presence of EDTA. Subsequent cleavage of FX␥ Pn, EDTA was observed to result in the appearance of a 47-kDa species that does not have an analogous plasmin-mediated fragment when FXa rather than FX is the substrate (20). Cleavage of the FX light subunit by plasmin was protected only in the presence of both Ca 2ϩ and proPL, indicating that proPL binding and not just divalent cation binding was necessary. The Gla domain has been shown to be involved in activation of FX by tissue factor/FVIIa or RVV FX activator (24), which provides an explanation for inhibition of FX coagulation activity by plasmin under conditions that did not facilitate proPL binding. Further evidence for differences in the sites recognized by plasmin due to binding of FX to proPL comes from plasminogen binding and tPA activity studies. Whereas plasmin cleavage was found to convert proPL-bound FX into a plasminogen receptor and tPA accelerator, neither activity was detectable when FX was cleaved by plasmin in the absence of proPL and/or Ca 2ϩ .
The FX data presented here, together with our earlier reports that showed that plasmin has similar modulatory effects on proPL-bound FXa and FVa (9,20), support a model in which plasmin functions to create a profibrinolytic milieu through direct conversion of specific proteins into tPA cofactors. We have found that neither FX, FXa, nor FVa expresses tPA cofactor function when not bound to proPL. Although the plasmin-mediated cleavage patterns differ, both proPL-bound and unbound FX, FXa, and FVa are inhibited from participating in any further clot formation. In this way, plasmin may participate in a feedback amplified mechanism by producing tPA cofactors localized by the selective accessibility of cellular proPL trapped in the fibrin network. We are currently investigating the relative contribution of fibrin to proPL-bound tPA cofactors within a clot.