Identification of the domains of tissue-type plasminogen activator involved in the augmented binding to fibrin after limited digestion with plasmin.

The binding of recombinant tissue-type plasminogen activator (rt-PA) to fibrin increases upon digestion of fibrin with plasmin. Optimal binding is observed following a limited plasmin digestion of fibrin, coinciding with the generation of fibrin fragment X polymers. We studied the involvement of the separate domains of the amino-terminal "heavy" (H) chain of rt-PA in this augmentation of fibrin binding. The fibrin-binding characteristics of a set of rt-PA deletion mutants, lacking either one or more of the structural domains of the H chain, were determined on intact fibrin matrices and on fibrin matrices that were subjected to limited digestion with plasmin. The augmented fibrin binding of rt-PA is partially abolished when the plasmin-degraded fibrin matrices are subsequently treated with carboxypeptidase B, demonstrating that this increased binding is dependent on the generation of carboxyl-terminal lysine residues in the fibrin matrix. Evidence is provided that this increase of fibrin binding is mediated by the kringle 2 (K2) domain that contains a lysine-binding site. Further increase of the fibrin binding of rt-PA is independent of the presence of carboxyl-terminal lysines. It is shown that the latter increase is not mediated by the K2 domain. Based on our data, we propose that the increase in fibrin binding, unrelated to the presence of carboxyl-terminal lysine residues, is mediated by the finger (F) domain, provided that this domain is correctly exposed in the remainder of the protein.

The binding of recombinant tissue-type plasminogen activator (rt-PA) to fibrin increases upon digestion of fibrin with plasmin. Optimal binding is observed following a limited plasmin digestion of fibrin, coinciding with the generation of fibrin fragment X polymers. We studied the involvement of the separate domains of the amino-terminal "heavy" (H) chain of rt-PA in this augmentation of fibrin binding. The fibrin-binding characteristics of a set of rt-PA deletion mutants, lacking either one or more of the structural domains of the H chain, were determined on intact fibrin matrices and on fibrin matrices that were subjected to limited digestion with plasmin. The augmented fibrin binding of rt-PA is partially abolished when the plasmin-degraded fibrin matrices are subsequently treated with carboxypeptidase B, demonstrating that this increased binding is dependent on the generation of carboxyl-terminal lysine residues in the fibrin matrix. Evidence is provided that this increase of fibrin binding is mediated by the kringle 2 (K2) domain that contains a lysine-binding site. Further increase of the fibrin binding of rt-PA is independent of the presence of carboxylterminal lysines. It is shown that the latter increase is not mediated by the K2 domain. Based on our data, we propose that the increase in fibrin binding, unrelated to the presence of carboxyl-terminal lysine residues, is mediated by the finger (F) domain, provided that this domain is correctly exposed in the remainder of the protein.
The fibrinolytic process ultimately leads to complete solubilization of the fibrin network of a thrombus (1). During this process, the serine protease plasmin gradually cleaves fibrin into distinct degradation products. Besides being a substrate for plasmin, fibrin also acts as an important cofactor in fibrinolysis. Both plasminogen, the inactive form of plasmin, and tissue-type plasminogen activator (t-PA)' can bind to fibrin. This binding is postulated to induce the formation of a so-called ternary complex that facilitates plasmin formation, thereby accelerating the degradation of fibrin (2, 3). The fibrinolytic process may be considered a dynamic process in which the cofactor function of fibrin could alter during ongoing degradation. This concept is supported by several observations. It has been shown for plasminogen that its binding to fibrin increases after limited degradation of fibrin by plasmin (4)(5)(6)(7). Binding to intact fibrin is mediated by a so-called aminohexyl-binding site that exhibits affinity for internal lysine residues in fibrin (8,9). Tran-Thang et al. (7) have shown that the increased binding is due to the exposure of a new type of high affinity binding sites in fibrin. The lysinebinding sites in plasminogen are required for this interaction. Previously, we have demonstrated that the amino-terminal "heavy" (H) chain of t-PA consists of autonomous, structural, and functional domains encoded by an exon or two adjacent exons (10). Such domains can be deleted or transposed to other molecules without grossly affecting the functional properties of the remainder of such proteins (11,12). This approach allowed us to conclude that the fibrin-binding property of t-PA and the acceleration of its activity is mediated by the "finger" (F) and the kringle 2 (K2) domains. The nature of the interaction of the F domain with fibrin is unknown, whereas the K2 domain contains a lysine-binding site (13). Recently, a lysine-binding site has also been located in the K1 domain of t-PA that may be involved in fibrin binding (14).
Based on our data, we suggested a model for the role of different t-PA domains during the process of fibrinolysis (13). Initially, binding of t-PA to intact fibrin would be mediated by the F domain. Subsequently, upon cleavage of fibrin by plasmin, carboxyl-terminal lysine residues are generated. Increased binding of t-PA to fibrin would then be achieved by the lysine-binding site present in the K2 domain. In addition, it has been shown that the K2 domain also exhibits affinity for internal lysine residues, conceivably mediated by an aminohexyl-binding site (15). Consequently, the K2 domain may play a role in the initial binding to intact fibrin as well. In contrast, Higgins and Vehar (16) have shown, explicitly using plasmin-degraded fibrinogen, that increased binding of t-PA to fibrin does not involve a lysine-binding site.
To examine the fibrin binding of t-PA during ongoing fibrin degradation in further detail, we employed a fibrin binding system that has been previously described for plasminogen (7). In this assay, a matrix of intact fibrin is pre-treated with plasmin to generate discrete intermediate stages of fibrin degradation. The involvement of different t-PA domains in fibrin binding has been studied using t-PA deletion mutants lacking one or more of these domains. Furthermore, we ana-

Fibrin
Binding of t-PA 12605 lyzed the effect of the exoprotease carboxypeptidase B in this system. Carboxypeptidase B specifically removes carboxylterminal lysine and arginine residues from polypeptides. The application of carboxypeptidase B allows us to discriminate between a potential involvement of aminohexyl-and lysinebinding sites in fibrin binding. Consequently, this approach is an extension of our previous studies, employing the lysine analogue e-aminocaproic acid as a competitor for fibrin binding of t-PA (13).
Here, we report that the increased binding of t-PA, formed on limited digestion of fibrin by plasmin, can be attributed in part to the lysine-binding site within K2. Furthermore, we propose that the remainder of the increased binding, not dependent on carboxyl-terminal lysines, is mediated by the F domain, provided this domain is correctly presented to digested fibrin.

EXPERIMENTAL PROCEDURES
Reagents-Iscove's modified minimal medium, minimal essential medium without leucine, Ultroser G, fungizone, and fetal calf serum were purchased from Gibco (Paisley, Scotland). Trasylol (10,000 kallikrein inhibiting units (KIU)/ml) was obtained from Bayer (Leverkusen, Federal Republic of Germany). Synthetic oligonucleotides were made on an automated DNA synthesizer (Applied Biosystems, model 381A). Plasmid pAGO, employed for co-transfection, contains the sequence of the active thymidine kinase gene of herpes simplex virus type 1 as described by Colbere-Garapin et al. (17). Hypoxanthine, aminopterin, and thymidine were supplied by Sigma, and radioactive materials were obtained from the Radiochemical Centre (Amersham, United Kingdom).
Radiolabeling of Proteins-Purified monoclonal antibody was iodinated using the chloramine-?' method (19). After labeling, the protein was dialyzed against PBS, containing 0.01% (v/v) Tween 80 and 0.01% (w/v) KI. Diisopropylfluorophosphate-treated fibrinogen (10 mg/ml) was diluted with a n equal volume of 0.5 M sodium phosphate (pH 7.2). To this mixture, 1 mCi of carrier-free ['251]NaI and 10 pg of chloramine T (1 mg/ml in PBS) were added for 1 min at room temperature. The reaction was terminated by the addition of 10 pg of sodium bisulfite. Then, 50 p1 of 1% (w/v) KI and 100 p1 of 0.5 M sodium phosphate (pH 7.2) were added. Free ['251]KI was removed by extensive dialysis at room temperature against 0.3 M NaC1, 10 mM Tris-HC1 (pH 7.4). The lZ5I-labeled proteins were stored at -20 "C.
Construction of rt-PA Deletion Mutants-Plasmid pSV2/t-PA, containing full length human t-PA cDNA (10) was used for the construction of the t-PA deletion mutant cDNAs. To construct the mutant cDNAs, a 1357-base pair (bp) fragment of the pSV2/t-PA full length cDNA, extending from the Hind111 site (position 61; see Table I) to the Sac1 site (position 1417) was inserted into the HindIIIl SacI-digested polylinker of the double-stranded replicative form of M13mp18am4 bacteriophage DNA (20). Subsequently, the '"13 gapped duplex out-looping" mutagenesis procedure (21) was employed using both recombinant M13mp18am4 (single-stranded) and M13mp18 (double-stranded, linearized) and different synthetic oligonucleotides (36-mers) to delete the desired parts of the t-PA cDNA (Table I). DNA sequence determinations were performed to verify correct fusion sequences (22). Finally, different mutated fragments (Table I) were exchanged in the pSV2/t-PA full length t-PA cDNA to obtain the set of t-PA deletion cDNAs under the control of the SV40 "early" promoter. The plasmid, encoding the mutant del.Kl, has been constructed and sequenced similarly to the plasmids mentioned above and was a kind gift from Dr. Sheila Little (Eli Lilly, Indianapolis, IN). In this construct, the cDNA sequence bp 448-716 has been removed. Consequently, it encodes the rt-PA variant with Gly instead of Asp a t position 87 and amino acids Thr88-Gly176 deleted.
Construction of Stable Cell Lines and Tissue Culture-Mouse Ltkcells were maintained in Iscove's modified medium, containing penicillin, streptomycin, and 10% (v/v) fetal calf serum. Semi-confluent monolayers of these cells were exposed for 4 h to a calcium phosphate co-precipitate of pSV2/t-PA DNA (or variants thereof) and plasmid pAGO DNA (20 and 1 pg/80-cmz flask, respectively), essentially according to the procedure of Graham and van der Eb (231, except that no carrier DNA was used. The cells were "shocked" for 2 min with 15% (v/v) glycerol (in PBS). After a recovery period of 24-48 h in medium containing serum, the cells were trypsin-treated and replated in selective medium containing hypoxanthine (12 pg/ml), aminopterin (1 pg/ml), and thymidine (8 pg/ml) (HAT). After 7-14 days, colonies resistant to this selection were taken. Separate colonies were grown and incubated for 2-3 days with medium without serum, before samples were analyzed for rt-PA production. The clones that secreted the highest amount of rt-PA activity were selected and grown in Iscove's modified medium, supplemented with penicillin, streptomycin, HAT, and 2% (w/v) Ultroser G. For metabolical labeling, the cells were grown to confluency, washed with PBS, and starved for 16 h in minimal essential medium lacking leucine, but additionally containing 6 mg/ml HEPES (pH 7.5), penicillin, fungizone, streptomycin, HAT, and 60 KIU/ml Trasylol. The medium was removed and replaced by fresh medium (lacking leucine) and [3H]leucine was added to a final concentration of 14 pCi/ml. After 3 days, the medium was harvested and centrifuged to remove cell debris and 0.01% (v/v) Tween 80 and 0.02% (w/v) sodium azide were added.
Purification of Metabolically Labeled Recombinant Proteins-The %labeled rt-PA (variant) proteins were purified by immunoaffinity chromatography, using monoclonal antibody ESP2 IgG (Bioscot, Edinburgh, Scotland) coupled to CNBr-activated Sepharose, yielding 1.3 mg of ESP2/ml of Sepharose. This monoclonal antibody is directed against an antigenic determinant on the t-PA L chain present on each of the deletion mutants (24). The conditioned media (70-140 ml) of different rt-PA preparations were loaded on separate columns of 0.1 ml of ESP2-Sepharose. Subsequently, the columns were washed with 75 ml of PBS containing 0.01% (v/v) Tween 80, 50 ml of PBS with a final NaCl concentration of 1 M, 0.01% (v/v) Tween 80, and 10 ml of PBS containing Tween 80 and 0.4 M KSCN. The proteins were eluted with a buffer containing PBS, 0.01% (v/v) Tween 80, and 3 M KSCN, and fractions of 500 p1 were collected. The fractions with rt-PA activity were pooled and extensively dialyzed against 50 mM sodium phosphate (pH 7.5), 0.01% (v/v) Tween 80 and subsequently stored at -70 "C until use. The reduced samples of rt-PA (variant) proteins were examined by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (25) followed by fluororadiography. The rt-PA (mutant) preparations were shown to be homogeneous and in the singlechain form (Fig. 2).
Determination of the Concentration of the rt-PA (Deletion) Proteins-For t-PA antigen determinations, we applied two monoclonal antibodies directed against different epitopes on the t-PA L chain, as determined according to the method described by Van Zonneveld et al. (24). These two antibodies were ESP2 (IgG) and MPWVPAl in 24-well tissue culture plates was done essentially according to Tran-Thang et al. (7). To prevent nonspecific binding, the wells were coated overnight a t 4 "C with 20 mg/ml BSA in a 50 mM sodium carbonate buffer (pH 9.4). The "BSA coating" buffer was removed, and the wells were washed two times with PBS, 0.01% (v/v) Tween 8 0 , l mg/ml BSA (buffer A). In each well, 300 p1 of fibrinogen (1 mg/ ml) in buffer A was added to 25 pl of human thrombin (12 units/ml).
A fibrin matrix was formed that was air-dried for 16 h a t 37 "C. It should be noted that fibrinogen preparations may contain some heterogeneity in the a-chains due to limited plasmin degradation either in plasma and/or during purification (26,27). T o eliminate potential pre-existing carboxyl-terminal lysines, in all experiments the fibrin matrices were pre-treated with 300 pl of 1 unit/ml carboxypeptidase B in buffer A for 1 h a t 37 "C. Plasmin-treated fibrin matrices in wells were prepared by adding 200 pl of buffer A, containing 6.8 nM plasmin, for 30 min a t 37 "C (see next paragraph). The plasmin digestion was arrested by removing the supernatant and, subsequently, the matrices were incubated for 30 min a t 37 "C with 300 pl of buffer A containing 0.2 M e-aminocaproic acid and 25 KIU/ ml Trasylol to elute and inactivate remaining plasmin from fibrin. The plates were then washed five times with 1.5 ml of buffer A. Fibrin matrices that were not degraded with plasmin were treated in an identical manner. Carboxypeptidase B treatments were done for 1 h a t 37 "C in 300 pl of buffer A containing carboxypeptidase B (1 unit/ ml) and subsequently the matrices were washed three times with 1.5 ml of buffer A. The release of lysine residues by carboxypeptidase B treatment of plasmin-treated fibrin was shown by amino acid analysis. To that end, the fibrin matrices were treated for 1 h with carboxypeptidase B in the absence of BSA, and the supernatant was removed for amino acid analysis. As controls, fibrin matrices were incubated either with buffer alone or treated with carboxypeptidase B but not with plasmin.
Determination of the Extent of Fibrin Degradation-Fibrin plates were prepared as described above except that '251-fibrinogen (approximately 150,000 counts/min/ml) was added to the unlabeled fibrinogen solution before clotting with thrombin. After the carboxypeptidase B pre-treatment (no release of counts due to the presence of carboxypeptidase B), the fibrin matrices were treated with increasing amounts of 0.2-35 nM plasmin in 200 pl of buffer A. The degradation of fibrin was expressed as the percentage of radioactivity that was released into the supernatant. The radioactivity incorporated into the fibrin matrices was taken as the 100% input value. The release of radioactivity from the fibrin matrices after plasmin incubation corresponds with the degradation of the nonlabeled fibrin, as shown by analysis of samples of the released fibrin fragments and the remaining fibrin matrix by SDS-PAGE (Fig. 1).
Binding of 3H-Labeled rt-PA Deletion Proteins to Fibrin-Approximately 0.15 pmol of each of the metabolically labeled rt-PA (mutant) proteins (corresponding to 1500-2000 counts/min) in 200 pl of buffer A, was added to a fibrin-coated well and incubated for 1 h at 37 "C. The supernatant was then quantitatively collected, diluted with 200 pl of buffer A containing 0.5% (w/v) SDS, and the radioactivity was determined by scintillation counting. The wells were rapidly washed three times with 1.5 ml of buffer A. The bound rt-PA (variants) were eluted together with the fibrin matrix from the respective well by gently shaking for 1 h at room temperature with 200 pl of buffer A, containing 0.5% SDS (w/v). The eluates were diluted with 200 pl of buffer A and the radioactivity was determined. Less than 10% of the radioactivity was lost during the washing steps. Nonspecific binding was determined using wells that were prepared identically to the fibrin-coated wells; however, the fibrinogen was omitted. The percentage of bound radioactivity was calculated and the nonspecific binding percentage was subtracted.

RESULTS
Fibrin Binding of rt-PA as a Function of Plasmin Degradation of Fibrin-To study the fibrin binding of rt-PA during ongoing fibrin degradation we devised an assay that employs immobilized noncross-linked fibrin pre-treated with increasing amounts of plasmin prior to rt-PA binding. Fibrin degradation was determined quantitatively by incorporating radiolabeled lZ5I-fibrinogen in the matrix. We show that an increased release of radioactivity correlates with the extent of fibrin degradation. We analyzed both the fibrin degradation products remaining in the well and the fragments released upon plasmin treatment by SDS-PAGE under reducing conditions (Fig. 1). Upon limited plasmin incubation, the achains of fibrin are cleaved predominantly, probably at Lys-208, Lys-221, and/or Lys-232 (28). This results in the generation of amino-terminal fragments ( M , 25,000-27,000) of the a-chains of fibrin that remain linked by disulfide bonds to the apparently intact pand y-fibrin chains (Fig. 1A). It can be observed that the carboxyl-terminal fragments (Mr 40,000-43,000), derived from the a-chains of fibrin, and several other fragments originating from the pand y-chains, are released from the matrix (Fig. 1 B ) . This indicates that predominantly X-fragment polymers are present in the wells. Moreover, the fragments released from the matrix contain fragments Y, D, and E as well, as was shown by SDS-PAGE under nonreducing conditions (data not shown).
To conveniently monitor the binding to fibrin of rt-PA, and subsequently of rt-PA deletion mutants, we used purified recombinant proteins that had been metabolically labeled with [3H]leucine. For that purpose, the conditioned media of mouse Ltk-cells, stably transfected either with rt-PA cDNA or variants thereof (depicted in Fig. 2A), were employed and subjected to a one-step immunoaffinity chromatography procedure. This protocol yielded apparently homogeneous singlechain preparations of rt-PA and of rt-PA deletion mutants (Fig. 2B). In accord with other reports (4, 16), we observe an increase in binding of rt-PA to fibrin upon pre-incubation of the fibrin matrix with plasmin (Fig. 3). Optimal rt-PA binding is detected in a broad range between approximately 10 and 70% fibrin degradation, coinciding with the predominant appearance of the fibrin fragment X. A more extensive digestion of fibrin, causing the release of over 70% of the fibrin radiolabel, dramatically decreases the binding of rt-PA. In our subsequent experiments we employed a plasmin treatment that induces a release of 10-16% of the incorporated radioactivity, corresponding with an initial degradation of the achains and an optimal rt-PA binding. We determined the binding of rt-PA to an initially plasmin-degraded fibrin matrix of a fixed amount of metabolically labeled t-PA, supplemented with an increasing concentration of unlabeled t-PA (Fig. 4). The binding increases linearly to a t-PA concentration of approximately 0.5 PM and subsequently reaches saturation. Consequently, all binding experiments using 0.75 nM rt-PA were performed under nonsaturable conditions. The same holds for each rt-PA variant since an identical binding percentage was found when either 0.375 or 1.5 nM of labeled rt-PA variant was used in the fibrin-binding experiments. (4-7). Furthermore, evidence has been   Fig. 1, in which in another experiment the fibrin matrices and the released fibrin degradation products after incubation with increasing amounts of plasmin were characterized by SDS-PAGE. Experimental details are outlined under "Experimental Procedures," and the actual amount of t-PA bound to fibrin was calculated.

The Importance of Carboxyl-terminal Lysine Residues in Fibrin for the Binding of rt-PA-It has been shown before that plasminogen exhibits increased fibrin binding upon plasmin treatment of fibrin
provided showing that carboxyl-terminal lysine residues are involved in the increased binding of plasminogen to fibrin (9). To study the importance of carboxyl-terminal lysine residues for t-PA binding, we analyzed the effect of the exopeptidase carboxypeptidase B on fibrin binding by rt-PA (Fig. 5, first  four bars). The binding of rt-PA to intact fibrin in this system amounts to 33.7 & 1.1%, probably mediated by the F domain and the aminohexyl-binding site within the K2 domain, as shown previously (10,15). The extent of binding to intact fibrin is not altered when the fibrin matrix is incubated extensively with carboxypeptidase B (34.2 f 1.2%). Provided that carboxyl-terminal lysine residues affect the binding of rt-PA to fibrin, this observation indicates that intact fibrin is indeed devoid of carboxyl-terminal lysines. Significantly,

Fibrin
Binding of t-PA FIBRIN BINDING distribution of the gene (29). 'Due to this strategy and based on the position of introns within the translational reading frame, it is anticipated that at the junctions of the out-looped regions novel codons are created, The resulting constructs and their subsequent mutant proteins are outlined in Table   I. All mutants were expressed and secreted by stably transfected mouse Ltk-cells and metabolically labeled using [3H] leucine. As mentioned before, immunoaffinity chromatography of these rt-PA deletion mutant proteins yields apparently homogeneous, single-chain preparations as shown by the analysis of reduced samples on SDS-PAGE (Fig. 2B).
Several groups have shown that the single-chain form o f t -PA displays a higher affinity for fibrin than the two-chain form (16,30). Hence, to ensure that the changes in binding after pre-treatment of the fibrin matrix with plasmin are not due to conversion of the rt-PA (variants) from the single- . Therefore, all our data concern fibrin binding by with plasmin (third bar); fibrin after limited digestion with plasmin fibrin matrices were prepare& an untreated fibrin matrix and and subsequently treated with carboxypeptidase B (fourth bur). For with the calculated standard error are given. Background, i.e. aspecific fibrin both either Or not by an binding to coated bovine serum albumin, has been substracted. +, extensive incubation with carboxypeptidase B. The quantieither a plasmin and/or a carboxypeptidase B treatment; -, without tative fibrin-binding data are summarized in Fig. 5. As shown enzymatic treatment. PLZ, incubation with plasmin at a concentrabefore, the deletion mutant lacking the complete H chain, carboxypeptidase B incubation at 1 unit/ml for 1 h at 37 "C.
tion that induces 10-16% fibrin degradation in 30 min at 37 "C. CPB, del.FEK1K2, exhibits no affinity for fibrin (10). If only the K1 domain is coupled to the L chain, del.FEK2, then hardly any binding to fibrin is detected. However, all rt-PA derivawhen the fibrin matrices were mildly digested with plasmin, tives that contain either the F or the K2 domain or both an increased binding was observed up to 51.2 k 1.4% of the display affinity for intact fibrin in this assay system. It should applied metabolically labeled rt-PA. AS expected, a subsebe noted that for all these rt-P.4 variants the extent of binding quent incubation of the plasmin-treated fibrin matrices with to intact fibrin is not affected by treatment with carboxypepcarboxypeptidase B causes the release of free lysines (data tidase B. The small rt-PA variants del.KlK2 and del.EKlK2 not shown). The binding of rt-PA to fibrin matrices that were that contain either only the F or the F and the E domain of mildly digested with plasmin and, in addition, treated with the H chain bind to the same extent to intact and degraded carboxypeptidase B decreases to 41.6 f 1.6% relative to fibrin. In contrast, del.FEK1 that contains K2 exhibits inmatrices that were not incubated with carboxypeptidase B. creased binding to degraded fibrin. Interestingly, this increase Clearly, incubation of these plasmin-treated matrices with is completely dependent on the presence of carboxyl-terminal carboxypeptidase B does not fully reduce the binding of rtlysine residues in fibrin, since a carboxypeptidase B treatment P A to the level of the binding to intact fibrin (i.e. 33.7 & completely abolishes this augmentation of binding. These 1.1%). Hence, we conclude that the increased binding of rt-results demonstrate the involvement of the lysine-binding site PA to fibrin which has undergone limited plasmin proteolysis in the K2 domain in increased binding of rt-PA to degraded is at least partially due to the occurrence of carboxyl-terminal fibrin. Similar observations were made for fibrin binding of lysine residues in the fibrin matrix. This observation is indic-the other K2 domain-containing mutants de1.F and del.Kl. ative for a role of the lysine-binding site in the K2 domain, On the other hand, the fibrin binding of deLK2, which also as we have postulated before (13). Moreover, our data dem-increases after limited degradation of the fibrin matrix, is not onstrate that part of the increased binding of rt-PA to plas-affected by the carboxypeptidase B treatment of plasminmin-digested fibrin is independent of the occurrence of car-digested fibrin. Such augmented fibrin binding, independent boxyl-terminal lysines. It is of particular interest to establish of carboxyl-terminal residues, had been encountered in rt-PA whether distinct autonomous domains of the rt-PA protein as well. From these data, we assume that the K2 domain is mediate either the increased binding dependent on the pres-not involved in the increased fibrin binding that is independence of carboxyl-terminal lysines or the increased binding ent of carboxyl-terminal lysines. This particular increased that is independent of such residues. T o address this issue we binding is observed provided that the F, E, and the K1 domain have employed a series of rt-PA deletion mutant proteins, are present in the same variant. In view of the basal fibrin precisely lacking structural and functional domains of the H binding of del.EKlK2 and del.KlK2 and the lack of fibrin chain, and have assayed their ability to bind to fibrin that binding of the del.FEK2 deletion mutant protein, it is conhas been digested with plasmin and/or with carboxypeptidase ceivable that the increased carboxyl-terminal lysine-inde-  pendent binding is mediated by the F domain, provided that our previous studies (IO), we constructed yet another set of digested fibrin. rt-PA deletion mutant proteins that lack one or more structural and functional domains of the H chain. Here, employing full length rt-PA cDNA, the corresponding cDNA sequences In the fibrinolytic process fibrin acts both as a substrate were "looped out" precisely according to the exon-intron for plasmin and as an assembling surface for plasminogen and DISCUSSION TABLE I Details on the construction and composition of the rt-PA deletion mutants The nomenclature for the mutant rt-PA proteins complies with the recommendations of the Subcommittee on Fibrinolysis of the International Committee on Thrombosis and Haernostasis (Washington, November 1988). All mutants were constructed from the full length rt-PA cDNA by the "out-looping'' procedure (21). The sequence of the primers for mutagenesis used for this technique are given in the 5' to the 3' direction. An asterisk indicates the position of the fusion, created by the deletion of a cDNA segment. The exact position where the cDNA sequences were looped out are depicted as the fusion site. Fusions were made exactly at the positions of the introns within the t-PA gene (29). Consequently, novel codons were generated, creating a single alteration in the amino acid sequence of the expressed proteins, listed in the table. After the mutagenesis procedure, different mutated fragments were inserted into the expression vector pSV2/tPA.

Mutant
The plasmid pSV2/t-PA contains the t-PA cDNA sequence from bp 78 to 2165 preceded by 17 bp of the polylinker of pUC9 resulting in a Hind111 site arbitrarily positioned at 61 (10). t-PA. In the latter function, fibrin strongly potentiates the activity of t-PA and is thus considered an obligatory cofactor. It is generally believed that binding of plasminogen and t-PA to fibrin, conceivably at specific sites, results in the formation of a ternary "cyclic" complex that effectively converts plasminogen into plasmin (2,3). In this paper we have investigated the effect of ongoing degradation of fibrin by plasmin on the characteristics of fibrin binding by t-PA. The degradation of fibrin proceeds according to well defined, sequential proteolytic steps, resulting in a number of discrete intermediate degradation products (31). The initial degradation product, denoted fragment X, consists of apparently intact fibrin pand y-chains and cleaved or-chains. Clearly, fragment X polymers perform a key role in the assembly of fibrinolytic components and in the acceleration of the fibrinolytic process. Suenson and Petersen (32) have shown that, among naturally occurring intermediate degradation products, fragment X polymers are most effective in the acceleration of t-PA-mediated plasminogen activation. Furthermore, these investigators have demonstrated that the generation of polymerized fragment X coincides with maximal conversion of Glu-plasminogen into Lys-plasminogen (33). Employing the same fibrin-binding assay as reported in this paper, Tran-Thang et al. (7) show that Glu-plasminogen hardly binds to intact fibrin, whereas binding is stimulated 15-30-fold upon limited (less than 10%) plasmin digestion of fibrin. Although, these authors do not correlate the extent of fibrin digestion with the appearance of particular fibrin fragments, in our hands the indicated extent of degradation again coincides with the generation of fragment X polymers. Our data on binding of t-PA to fibrin are also in accord with the view that polymerized fragment X is a crucial determinant in assembly of fibrinolytic components. Optimal binding of t-PA to plasmin-digested fibrin again coincides with a limited digestion and the generation of polymerized fragment X. Consequently, it is conceiv-able that a ternary cyclic complex, required for accelerated activation of plasminogen, functions optimally if fibrin has been plasmin-digested to a limited extent. Subsequently, we have addressed the issue of the involvement of different domains of t-PA in augmentation of binding to plasmin-digested fibrin, as well as to the nature of these interactions. A major conclusion from our results is that about half of the increased binding is due to the generation of carboxyl-terminal lysine residues, whereas the other half is independent of the presence of these chain termini. By employing a set of t-PA deletion mutants, lacking one or more of the domains of the amino-terminal H chain, we could clearly demonstrate that the increase dependent on carboxylterminal lysines is mediated by the lysine-binding site within the K2 domain. An apparent contradiction between this conclusion and that of Higgins and Vehar (16) can be envisaged based on the results of seemingly similar experiments. Those authors concluded that a lysine-binding site on t-PA is only involved in binding to intact fibrin and not in increased binding to fibrin upon plasmin digestion. However, it should be noted that those studies were performed with fibrinogen pre-treated with plasmin before clotting. As indicated before (7), such fibrin matrices may not reflect the physiological fibrinolytic process occurring in uiuo. Moreover, it has been demonstrated that the physico-chemical properties of such matrices are quite different from those of pre-formed fibrin matrices (34). The rigidity (elasticity) of the matrices prepared from fibrinogen pre-treated with plasmin before clotting is only about 1% of those prepared from plasmin-digested fibrin. These distinct physico-chemical properties of the respective fibrin matrices may explain the observed biological discrepancies.
In agreement with our previous findings (13), the results presented in this paper indicate the presence of a single lysinebinding site within the t-PA protein, located on the K2 Fibrin Binding of t-PA domain. Gething et al. (14) have claimed an additional site within the K1 domain that exhibits affinity for lysine-Sepharose. The fibrin binding data of the t-PA deletion mutants, still containing the K1 domain (del.K2 and del.FEKB), do not support the view that the K1 domain contains a lysinebinding site. In contrast to mutant proteins containing the K2 domain, these mutant proteins do not reveal increased fibrin binding upon limited plasmin digestion that is dependent on the presence of carboxyl-terminal lysines. As yet, a satisfactory explanation for the observed differences in the properties of the K1 domain is not at hand. Both rt-PA and the mutant protein del.K2 display increased binding to plasmin-digested fibrin that is independent of carboxyl-terminal lysine residues. At present, an unambiguous assignment of a particular domain responsible for this increase cannot be advanced, since several options can be considered. Actually, this effect is only observed for proteins (rt-PA and del.K2) that contain the amino-terminal domains F, E, and K1, while it is not detected with the other deletion mutant proteins. We favor the view that this increase is mediated by the F domain. For that option, we assume that the K1 domain (and possibly E) acts as a obligatory "spacer" for correct positioning of the F domain toward other parts of the molecule and, as a consequence, for its proper functioning. This assumption is in agreement with data reported by a number of groups that have constructed chimeric molecules aimed at creating proteins that have combined the fibrinbinding features of t-PA with the catalytic activity of urokinase-type plasminogen activator (u-PA), located on its carboxyl-terminal ("B") chain. To that end, e.g. a construct was made in which the F domain has been fused to the u-PA B chain (35). Moreover, t-PA variants have been constructed in which the position of the K1 and K2 domains has been "switched" (36). In all instances, the fibrin-binding capacity of these chimeric proteins was severely hampered, indicating that the "spacing" of the F domain toward other parts of the molecule is an important condition for its functioning. However, at this point we cannot exclude other possibilities for the role of the K1 and the E domain, in particular the increased binding to plasmin-digested fibrin, independent of carboxyl-terminal lysine residues. Currently, we are investigating the potential spacer function of the indicated domains by the construction and expression of t-PA substitution mutant proteins.