Plasminogen (SEQ ID NO: 1) is the principal serine protease zymogen in the extracellular fluids of vertebrates, and its active form, plasmin, is implicated in pericellular proteolysis associated with a wide range of physiological and pathological processes, including the hydrolysis of fibrin into soluble degradation products and the suppression of tumors by angiogenesis inhibition (Gately, Proceedings of the NationalAcademy of Sciences, USA 1997; 94: 10868-10872). In general, plasminogen expression is fairly stable and the regulation of the activity of the fibrinolytic system occurs mainly via up-and down-regulation of the expression of plasminogen activators and the inhibitors of these activators. Activation of plasminogen is a consequence of cleavage of the Arg56l-Val562 bond, to form a two-chain, disulfide linked plasmin. The two known physiological plasminogen activators are the serine proteases tissue-type plasminogen activator (t-PA) (SEQ ID NO: 2) and urokinase (u-PA or UK) (SEQ ID NO: 3), both of which directly catalyze the hydrolysis of the activation bond. However, plasminogen can also be activated by another completely different mechanism, which requires formation of an activator complex with a molecule such as streptokinase. When streptokinase is complexed with plasminogen, plasminogen spontaneously converts into plasmin. This complexed plasmin is then able to activate free plasminogen. Plasmin on its own cannot activate plasminogen.

Streptokinase (SK) (SEQ ID NO: 4) is a single-peptide secretory protein of 414 amino acid residues produced by various strains of hemolytic Streptococcus (Jackson and Tang, Biochemistry 1982; 21: 6620-6625; Malke et al., Gene 1985; 34: 357-361). SK does not contain cysteine or carbohydrate. Proteolytic digestion, NMR and other biochemical- biophysical studies indicate that SK is a flexible multi-domain protein

(Conejero-Lara et al., Protein Science 1996; 5: 2583-2591; Medved et al., European Journal ofBiochemistty 1996; 239: 333-339; Parrado et al., Protein Science 1996; 5: 693-704; Rodriguez et al., European Journal of Biochemistty 1995; 229: 83-90). SK and human plasminogen can form an equimolar activator complex that catalyzes the conversion of plasminogen from different mammalian species to plasmin. This property renders SK a potent clinical agent for the treatment of blood clotting disorders, such as acute myocardial infarction and stroke (Coleman et al., Hemostasis and Thrombosis, Basic Principles and Clinical Practice J. B. Lippincott Co.: Philadelphia, 1994). The activation of human plasminogen by SK involves the formation of a streptokinase-plasminogen (SK-Plg) complex that alters the conformation of the catalytic domain of the zymogen to complete its enzyme-active center. The SK-Plg complex converts to a streptokinase- plasmin (SK-Plm) complex spontaneously. Both the SK-Plg and the SK-Plm complexes catalyze the hydrolysis of the specific activation bond, Arg56l- Val562, of the substrate plasminogen, resulting in the formation of plasmin.

However, plasmin alone is not a plasminogen activator.

A native plasminogen molecule contains at least seven structural domains, including the N-terminal 77-residue pre-activation peptide, five 'kringles'and a C-terminal trypsinogen-like zymogen domain (Sottrup- Jensen et al., Program of Chemical Fibrinolysis Thrombolysis 1978 ; 3: 191- 209). An isolated catalytic domain of plasmin (ogen) is called micro- plasmin (ogen) (Plm/Plg). Human plasminogen contains 24 disulfide bonds. Human plasminogen is glycosylated at two positions that are located within the third kringle and between the third and fourth kringles, respectively. Many isolated SK and plasminogen fragments, obtained via proteolytic reaction or recombinant methods, have been used to identify the regions involved in the interaction between the two molecules (Rodriguez et al., European Journal of Biochemistry 1995; 229: 83-90; Shi, et al., Journal of Biological Chemistry 1988; 263: 17071-17075; Young et al., Journal of Biological Chemistry 1998; 273: 3110-3116).

Accordingly, it would be useful to have a crystalline structure of the

streptokinase-micro plasmin (ogen) (SK-IlPlg) complex. Such a structure would make it possible to predict the portions of SK that complex with plasminogen and to design modified streptokinases, such as a streptokinase having less antigenicity but which is still able to complex plasminogen and lead to activation of plasminogen.

It is an object of the invention to provide a structure for streptokinase-micro plasminogen complex and identify the plasminogen complexation site and the streptokinase portions that are not essential for plasminogen complexation or activation.

It is an object of the present invention to provide streptokinase derived thrombolytic agents.

It is an object of the present invention to provide a method of making thrombolytic agents derived from streptokinase.

Summary of the Invention Structural information about the streptokinase-micro plasminogen complex has been used to identify the part of the streptokinase structure not involved in plasminogen complexation or activation. These nonessential portions can be modified to reduce antigenicity, for example, by changing the nonessential portions of streptokinase to more human-like polypeptide portions ("humanization of streptokinase"). One way this can be done is to compare the nonessential portion to a structural database of human proteins to find similar structures. Then the streptokinase nonessential structural part is replaced with the human structural part such as by genetic engineering of a mutant encoding the individual streptokinases, which is then expressed in a bacterial host such as E. coli. Alternatively, the nonessential portions can be removed or truncated to simplify streptokinase to a smaller molecule which retains plasminogen activation activity. Such smaller proteins should have reduced antigenicity and be cheaper and easier to produce. The modified streptokinases are useful in treating clotting disorders.

Description of the Drawings Figure 1 is a stereo view of the activation pocket of the plasmin catalytic domain. A 2.9 A resolution 2IFobsI~IFcalcI electron density map, phased with the final refined model and contoured at 1.0, is superimposed on the refined model.

Figure 2 is a stereo view of the crystal structure of the complex of human micro plasmin (uPlm) and streptokinase. The IlPlm molecule is in the middle of the complex. The a-domain of SK is at the top, left side of the complex. The (3-domain of SK is to the right side of the complex. The y- domain of SK is at the bottom of the complex. Only the Ca traces are shown.

Figures 3 (a)- (c) illustrate ribbon diagrams of the three domains of streptokinase. Figure 3 (a) is a ribbon diagram of the a-domain; Figure 3 (b) is a ribbon diagram of the (3-domain; and Figure 3 (c) is a ribbon diagram of the y-domain. The domains are oriented to illustrate the similarity in their overall ß-grasp folding.

Figure 4 (a) is a stereo view of the interaction between human micro plasmin and the a-domain of streptokinase. The micro plasmin molecule is at the bottom of the complex. Figure 4 (b) is a stereo view of the interaction between human micro plasmin and the y-domain of streptokinase. The micro plasmin molecule is at the top of the complex. The side chains which are involved in the interaction are displayed along with the Ca backbones.

Figure 5 is a stereo view of the superposition of the catalytic domains of human plasmin and human two-chain tissue-type plasminogen activator (t- PA) (Lamba et al., Journal ofMolecular Biology 1996; 258: 117-135).

Selected positions are numbered accordingly. Only the Ca traces are shown.

Some commonly used loop names for trypsin-like proteases are also included (Renatus, et al., EMBO Journal 1997; 16: 4797-4805).

Figure 6 (a) is modeled view of the substrate binding site of the micro plasminogen-streptokinase complex showing the molecular surface of the complex. The orientation of the complex is similar to the orientation of the

complex in Figure 1. Figure 6 (b) is a stereo view illustrating docking of a substrate micro-plasminogen into the substrate binding site. The enzyme micro-plasmin is to the left of the complex. The streptokinase molecule is in the center of the complex, from top to bottom. The substrate micro- plasminogen is to the right of the complex.

Figure 7 is a stereo view illustrating a putative active-zymogen form of plasminogen, compared with plasmin. The cleaved activation loop of plasmin, shown in a thicker tube, is at the center of the complex, as is the side chain of Lys698 in plasmin. The corresponding parts of the active- zymogen are towards the right. The rest of plasmin (ogen) is towards the top of the complex and possesses an active conformation around the active site, particularly the peptide segment (shown in a thicker tube) upstream of the nucleophile Set'41. (3-strands in the surrounding region are shown as arrows.

The salt bridge distance between the tips of Lys69R and Asp740 in the active- zymogen form is approximately the same as that between the free amino group of Val562 and Asp740 in plasmin. Upstream of Lys698 is the binding site to the y-domain of streptokinase.

Figure 8 is a stereo view of the a-domain of streptokinase (SK) superimposed on staphylokinase (SAK) (SEQ ID NO: 5) (Rabijns, et al., Nature Structural Biology 1997; 4: 357-360). In addition to the Ca traces, the side chains of SK Val'9 (SAK Met26) and SK Glu39 (SAK Glu46) are plotted.

Detailed Description The flexibility and multidomain nature of both SK and plasminogen have heretofore prevented the crystallization and determination of the crystal structures of plasmin (ogen), SK and the SK-Plg complex. The rapid autolysis of the SK-Plg complex renders the crystallization of the wild-type SK-Plg complex impractical.

It is known that the catalytic domain of human plasminogen can bind and be activated by SK (Shi et al., Nature Structural Biology 1990; 4: 357- 360; Wang and Reich, Protein Science 1995; 4: 1758-1767). A recombinant

human microplasminogen (Plg) was constructed, containing an alanine residue substituted for the active-site serine residue. The streptokinase- microplasminogen (SK-Plg) complex spontaneously, but slowly, converts to a streptokinase-microplasmin (SK-uPlm) complex. However, it does crystallize and the crystal diffracts to atomic resolution. The three- dimensional structure of the SK-, uPlm complex is disclosed herein.

1. Streptokinase Structure i. Streptokinase Portions Complexed With Plasmin (ogen) Streptokinase complexation portions refers to single amino acid residues or polypeptides of streptokinase required for the complexation of plasminogen and plasmin to streptokinase and for the activation of plasminogen to plasmin. ii. Streptokinase Portions Involved In Substrate Specificity Streptokinase substrate specificity portions refers to single amino acid residues or polypeptides required to impart substrate specificity upon plasmin complexed to streptokinase as compared to plasmin alone. iii. Nonessential Portions of Streptokinase Nonessential portions as used herein refers to those portions of streptokinase that can be modified, such as by being removed or replaced, without destroying the ability of the streptokinase to complex with plasminogen and without destroying the ability of the SK-Plm complex to activate plasminogen. Moreover, the nonessential portions include portions that can be modified, such as by being removed or replaced, or by substituting or deleting one or more of the amino acids, without destroying the ability of streptokinase to impart substrate specificity to plasmin when complexed as SK-Plm compared to plasmin alone. The modified streptokinase proteins disclosed and claimed can have one or more of these nonessential portions removed or replaced.

2. Design and Methods of Making the Modified Streptokinase i. Design Native streptokinase can induce formation of anti-streptokinase antibodies following administration of a single dose. Subsequent doses are then attacked by these anti-streptokinase antibodies, making subsequent doses ineffective.

To form a humanized chimeric streptokinase, the nonessential portions are compared against a database of human proteins to identify human proteins or portions thereof which are structurally similar to the nonessential streptokinase portions. A chimeric humanized streptokinase mutant can then be made in which the nonessential portion (s) are replaced with the human protein portions. Alternatively, or in addition, a truncated protein can be made in which one or more nonessential portions, or one or more amino acids therein, have been removed.

Preferably, the human protein or portion thereof has a high degree of structural similarity to the streptokinase portion. However, the human portion does not have to be structurally identical to the streptokinase portion.

Preferably, the human portion does not retain any of the function of the native human protein from which it is derived. Of course, the humanized or truncated protein should retain substantially all or a substantial fraction of its ability to complex and activate plasminogen. ii. Methods of making the modified Streptokinase The humanized or truncated proteins may be made using methods known to those of skill in the art. These include chemical synthesis, modifications of existing proteins, and expression of humanized proteins or truncated proteins using recombinant DNA methodology. The humanized protein can be made as a single polypeptide or the human protein portion can be attached to the base streptokinase polypeptide after separate synthesis of the two component polypeptides.

Where the protein is relatively short (i. e. less than about 50 amino acids) the protein may be synthesized using standard chemical peptide synthesis techniques. Solid phase synthesis in which the C-terminal amino

acid of the sequence is attached to an insoluble support followed by sequential additional of the remaining amino acids in the sequence is the preferred method for the chemical synthesis of the proteins described herein.

Chemical synthesis produces a single stranded oligonucleotide. This may be converted into a double stranded DNA by hybridization with a complementary sequence, or by polymerization with a DNA polymerase using the single strand as a template. One of skill would recognize that while current methods for chemical synthesis of DNA are limited to preparing sequences of about 100 bases, longer sequences may be obtained by the ligation of shorter sequences. Techniques for solid phase synthesis are described by Barany and Merrifield, Solid-Phase Peptide Synthesis; pp. 3- 284 in The Peptides; Analysis, Synthesis, Biology Vol. 2. Special Methods in Peptide Synthesis, Part A, Merrifield, et al., J. Am. Chem. Soc. 1963; 85: 2149-2156, and Stewart et al., Solid Phase Peptide Synthesis, 2nd ed. Pierce Chem. Co.: Rockford, Ill., 1984.

Alternatively, the protein may be made by chemically modifying a native protein. Generally, this requires cleaving the native protein at one or more sites and then annealing desired polypeptides onto the newly formed termini. The desired cleaved peptides can be isolated by any protein purification technique that purifies on the basis of size (e. g. by size exclusion chromatography or electrophoresis). Alternatively, various sites in the protein may be protected from hydrolysis by chemical modification of the amino acid side chains which may interfere with enzyme binding, or by chemical blocking of the vulnerable groups participating in the peptide bond.

In the preferred embodiment, the humanized or truncated proteins will be synthesized using recombinant methodology. Generally, this involves creating a polynucleotide sequence that encodes the protein, placing the polynucleotide in an expression cassette under the control of a suitable expression promoter, expressing the protein in a host, isolating the expressed protein and, if required, renaturing the protein. Additional proteins can be made separately and then ligated to the modified streptokinase, or the

polynucleotide sequence can encode the streptokinase in phase with anther protein.

DNA encoding the protein can be prepared by any suitable method, including, for example, cloning and restriction of appropriate sequences or direct chemical synthesis by methods such as the phosphotriester method of Narang et al., Meth. Enzymol. 1979; 68: 90-99; the phosphodiester method of Brown et al., Meth. Enzymol. 1979; 68: 109-151; the diethylphosphoramidite method of Beaucage et al., Tetra Lett. 1981; 22: 1859-1862; and the solid support method of U. S. Patent No. 4,458,066.

Alternatively, partial length sequences may be cloned and the appropriate partial length sequences cleaved using appropriate restriction enzymes. The fragments may then be ligated to produce the desired DNA sequence.

In a preferred embodiment, DNA encoding the protein will be produced using DNA amplification methods, for example polymerase chain reaction (PCR).

The proteins may be expressed in a variety of host cells, including E. coli or other bacterial hosts, yeast, and various higher eukaryotic cells, such as the COS, CHO and HeLa cells lines, insect cells, and myeloma cell lines.

The recombinant protein gene is operably linked to appropriate expression control sequences for each host. The plasmids encoding the protein can be transferred into the chosen host cell by well-known methods such as calcium chloride transformation for E. coli and calcium phosphate treatment or electroporation for mammalian cells. Cells transformed by the plasmids can be selected by resistance to antibiotics conferred by genes contained on the plasmids, such as the amp, gpt, neo and hyg genes.

Once expressed, the protein can be purified according to standard procedures such as ammonium sulfate precipitation, affinity columns, column chromatography, or gel electrophoresis. Substantially pure compositions of at least about 90 to 95% homogeneity are preferred, and 98 to 99% or more homogeneity are most preferred for pharmaceutical uses.

One of skill in the art would recognize that after chemical synthesis,

biological expression, or purification, the protein may possess a conformation substantially different than the native protein. In this case, it may be necessary to denature and reduce the protein and then to cause the protein to re-fold into the preferred conformation. Methods of reducing and denaturing the protein and inducing re-folding are well known to those of skill in the art. For example, the expressed, purified protein may be denatured in urea or guanidium chloride and renatured by slow dialysis.

To determine which proteins are preferred, the proteins should be assayed for biological activity. Such assays, well known to those of skill in the art, generally fall into two categories; those that measure the binding affinity of the protein to a particular target, and those that measure the biological activity of the protein.

3. Methods Of Using Modified Streptokinase The modified streptokinase molecules can be administered for treatment of blood clot disorders, such as in heart attacks, as known in the art for administration of native streptokinase and tissue-type plasminogen activator (t-PA) and urokinase (u-PA or UK).

The compounds are preferably administered intravenously in appropriate carriers. The appropriate dosages will depend upon the route of administration and the treatment indicated, and can be readily determined by one skilled in the art. Dosages are generally initiated at lower levels and increased until desired effects are achieved.

The present invention is further described by the following nonlimiting examples.

Example 1: Protein preparation and crystallization Recombinant streptokinase and S741A mutant of human IlPlg were constructed. The proteins were expressed in E. coli as inclusion bodies which were washed, dissolved in 8 M urea and combined to be refolded together by rapid dilution. The 1: 1 complex between SK and u. Plg was purified further using S300-chromatography. The protein sample was stored

at 0°C for more than two months before being used for a successful crystallization. The initial crystallization condition was determined by the use of sparse-matrix screens from Hampton Research. Crystals were grown at 20°C from sitting drops with wells containing 1.0 M sodium citrate, 0.2 M HEPES (pH 8.0), 1 mM magnesium chloride. The protein concentration used was 40 mg/ml. Crystals appeared in about two weeks and had typical dimensions of 0. 1 x 0.1 x 0.5 mm. The crystal belongs to space group P2, with cell parameters of a=80.0 A, b=125.1 A, c=86.8 A and ß=105.4°. One crystallographic asymmetric unit contains two SK-Plm complexes with VM=2.9 A3/Da (Matthews, Journal of MolecularBiology 1968 ; 33: 491- 497).

To confirm the crystal content, selected crystals were dissolved in H20 and analyzed by SDS PAGE. The results showed that the u. Plg was converted to IlPlm and SK was digested at two positions to different extents.

The proteolytic cleavage in uPlg renders the complex changed from SK-Plg to SK-, uPlm. The low proteolytic activity in the sample may come from trace-amount protease contamination, and/or the Ser 741 to Ala mutant leaking back to the"wild-type"u. Plg during the E. coli expression.

Example 2: Crystallographic methods and data processing Data collection and heavy-atom derivative screen was conducted at room temperature on a Siemens area detector. The program SAINT was used to process the data. Molecular replacement searches for the IlPlm molecules were carried out with the program MRX (Zhang and Matthews, Acta.

Crystallographica 1994; D50,675-686). The solution clearly showed a local two fold symmetry between the two SK-, uPlm complexes. However the overall quality of the electron density based on this molecular replacement solution alone was strongly biased by the model used and was not useful for obtaining interpretable electron density beyond the region of the search model.

The phases of the crystal structure of the SK-Plm complex were

solved by the use of multiple isomorphous replacement (MIR) techniques, using platinum, mercury and iridium derivatives. The program packages SOLVE (Terwilliger and Berendzen, Acta. Crystallographica 1996; D52: 749-757) and MLPHARE (Otwinowsky, Proceedings of the CCP4 Study Weekend, 25-26 January, (W. Wolf, P. R. Evans and A. G. W. Leslie, eds.), pp 80-86, Daresbury Laboratory: SERC, 1991) were used to identify and refine the heavy atom parameters. The statistics of the X-ray data collection and MIR phasing are summarized in Table 1.

Table 1. Summar of Data Collection, Data Statistics and MIR Phasing<BR> Data set res. complt(%) No. of Rsym(%) Riso #PP#e Rcullis<BR> (A) all/last refl. all/last (%) acen/cen acen/cen<BR> Native 2.9 91.34/80.27 33424 5.3/25.2 - - -<BR> K2PtCl4a 3.6 94.48/84.14 18287 7.6/20.2 20.8 1.22/0.76 0.75/0.78<BR> K2Pt(NO2)4 3.0 91.14/7.76 30246 8.9/29.8 24.5 1.31/0.86 0.77/0.85<BR> C7H5HgO3.Nb 3.6 93.36/86.28 18221 11.9/24.1 18.1 0.92/0.75 0.90/0.84<BR> Pt&Hgc 3.2 82.25/74.26 22873 3.9/24.3 26.4 1.14/0.73 0.77/0.85<BR> K2Pt(CNS)6 3.4 94.97/89.08 21955 10.5/19.4 24.5 1.07/0.67 0.82/0.88<BR> K2PtCl6 3.0 92.40/80.91 31176 7.9/28.3 26.6 1.67/0.93 0.71/0.80<BR> K3lrCl6 3.4 90.51/76.76 20869 7.8/21.2 18.7 0.54/0.43 0.97/0.98<BR> Resolution (A) 19.22 10.85 7.55 5.79 4.70 3.95 3.41 3.00 Total<BR> No. of refl. 119 558 1336 2435 3857 5669 7508 8728 30210<BR> FOMd 0.6418 0.7104 0.7794 0.7755 0.6998 0.6224 0.4924 0.3136 0.5317<BR> a. All heavy atom soaking experiments were performed in the standard mother liquor of 1.2 M sodium citrate, 0.2 M @<BR> 8.0) and 1% glycerol (v/v).<BR> b. p-Hydroxy-mercury-benzoate<BR> c. Soaked in 1.4 mM K2PtCl4 for one day followed by soaking in saturated p-Hydroxy-mercury-benzoate for four day@<BR> d. FOM - figure of merit<BR> e. #PP# - the phasing power for both acentric and centric reflections.

The initial MIR phases at 3.0 A resolution were confirmed by the MR solution of the ßPlm part of the complex. The phases were further improved by electron density averaging over two-fold non-crystallographic symmetry and solvent flatting. The initial mask for the electron density averaging was derived from the MR solution. The electron density improvement was carried out with the programDM (Cowtan and Main, Acta. Crystallographica 1996; D52: 43- 48).

From the improved electron density map, the SK molecule was seen in three distinct domains adjacent to the region of the RPlm molecule. The initial model of, uPlm was built using the crystal structure of chymotrypsinogen (Wang et al., Journal of MolecularBiology 1985 ; 185: 595-624) as a template. The structure of SK was built directly from the electron density map. Model building was performed with the program O (Jones et al., Acta.

Crystallographica 1991; A47: 110-119). Iterative refinement and model building were used to improve the model gradually. SigmaA-weighted maps were calculated with the program SIGMAS (Read, Acta. Crystallographica 1986; A42: 140-149) and used in the initial model building.

Refinement was carried out with the program XPLOR (Brunger et al., Science 1987; 235: 458-460) and TNT (Tronrud et al., Acta. Crystallographica 1987; A43: 489-501). None-crystallographic-symmetry constrains were used throughout the refinement; in the final model, the two copies of the crystallographic independent SK-, uPlm complexes are practically identical. All of the chemically expected 250 residues of the uPlm molecule were included in the final model at 2.9 A resolution; and of the 414 residues of the SK, 322 were modeled. A representing region of the electron density map of RPlm is shown in Figure 1. The final R factor is 20.3% over the 8.0-2.9 A resolution shell (28,600 reflections), and the free R (Brunger, Nature 1992; 355: 472-474) is 30% (3,150 reflections). Bond and angle deviations are 0.01 A and 1.8°, respectively, as determined by XPLOR using Engh and Huber parameters (Engh and Huber, Acta. Crystallographica 1991; A47: 392-400). Structural superposition and solvent accessible surface calculation were carried out with EDPDB (Zhang and Matthews, Journal ofApplied Crystallography 1995; 28: 624-630). Figures were created by using MOLSCRIPT (Kraulis, J. Appl. Cryst. 1991; 24: 96-950), RASTER3D (Merritt and Murphy, Acta. Cryst. 1994; D50: 869-873) and GRASP (Nicholls et al., Proteins 1991; 11: 281-296).

Example 3: Overall Structure of human µPlm The crystal structure of SK-, uPlm complex was determined at 2.9 A resolution using X-ray crystallography. There are two SK-µPlm complexes per crystallographic asymmetric unit, which are practically identical with each other. Figure 2 shows the Ca trace of one SK-Plm complex. The µPlm component of the complex contains the region from residue Ala542 to the C- terminal residue Asn791 of plasmin. The dimensions of the) J. Plm molecule are about 40 x 45 x 50 A. Resembling the architecture of many other trypsin-like serine proteases, µPlm consists of two domains, each of a six-stranded 0-barrel.

The C-terminus of uPlm ends with an α-helix packing against the N-terminal P- barrel. The catalytic residues, His603, Asp646, and Ser74l position confirmed the Ser to Ala substitution.

Consistent with the high sequence homology (i. e. 39% identity), the coordinate root mean square deviation (rmsd) between juPlm and chymotrypsin (Harel et al., Biochemistry 1991; 30: 5217-5225) is 0.7 A for 193 Cα atoms, using a 1.5 A cutoff. Compared to both chymotrypsin and chymotrypsinogen crystal structures (Harel et al., Biochemistry 1991; 30: 5217-5225; Wang et al., Journal of MolecularBiology 1985 ; 185: 595-624), the active site conformation of the catalytic domain of plasmin is indeed in its enzyme form. The activation bond, Arg56'-Val562, has been cleaved as indicated by SDS PAGE and N- terminal sequencing of the crystal contents (data not shown). The new C- terminus of the cleaved loop, containing Pro559, Gly560, and Arg561, is mobile in the crystal and can be seen in the electron density only at a low contour level.

The newly liberated N-terminus (WGG) (amino acids 562-565 of SEQ ID NO: 1) enters the activation pocket that is designed precisely to fit both the main- chain atoms and the divaline side chains. Its stability and proper positioning is reinforced by the solvent inaccessible salt bridge linking the terminal amino group to the carboxylate group of Asp740. This salt bridge ensures that the loop immediately upstream of nucleophile Ser74l changes its conformation to an active form. Consequently, the oxyanion hole is formed by the amide groups of residues 738-740, and the S1 specificity pocket is properly formed.

Six disulfide bonds stabilize the structure ofuPlg. Three of them, Cys58S Cys604, Cys680 Cys747, and Cys737-Cys765, are within six-residue ranges of the catalytic triad residues and function to maintain the platform of the catalytic triad. Another one, Cys558-Cys566, which is absent in many other typsin-like proteases, flanks the activation cleavage site. In the zymogen form, this disulfide bond likely constrains the conformation of the short activation loop such that the Arg561-Val562 bond is confined to be readily cleaved by plasminogen activators. The two new termini liberated by the activation cleavage are also constrained by the disulfide bond.

Immediately beneath the imidazole ring of His603 is residue Ala601 which sits in a pocket perfect for its methyl group side chain. This residue was found to be mutated to a threonine residue in the plasmin of a group of patients with a predisposition of thrombosis (Ichinose et al., Proceedings of the National <BR> <BR> <BR> <BR> Academy of Sciences, U. S. A. 1991 ; 88: 115-119). Such an Ala-to-Thr mutation disrupts the hydrogen bond between His603 and Asp646 and impairs the charge delay network of the catalytic triad.

Example 4: Overall structure of streptokinase In the crystal structure of the SK-pLPlm complex, SK appears as a three domain protein with several segments in the primary sequence disordered in the crystal lattice. The three domains are linked with each other by coiled coil peptides and are likely to fold independently in solution. They are denoted as a- ß-, and y-domains hereafter along the peptide chain from the N-terminus to the C-terminus (see Figure 2).

The a-domain begins at residue Asn and ends at about residue Pro Residues 1-11 and two extra residues (Gly-Ser) adopted from the cloning vector are disordered in the electron density map. Similarly the region of residues 45- 70 has a lack of interpretable electron density. A proteolytic cleavage that occurs at the bond between Lys59 and Ser60 is located in this disordered region.

The (3-domain begins at residue Alaiss and ends at residue Pro283. This domain also contains a cleavage site, Lys257-Ser258, which is cleaved only in a portion of total SK-Plm complexes as shown in our SDS PAGE analysis. No preference of SK molecules, cleaved and non-cleaved, is seen at this site in the two crystallographically independent SK-, uPlm complexes. The y-domain starts at residue Asp285 and becomes invisible beyond residue Arg372, although SDS PAGE suggests that the C-terminal 40 or so residues are attached. The domain boundaries found in the crystal structure are consistent with the results from various protease mediated SK-degradations (Parrado et al., Protein Science 1996; 5: 693-704). It is also consistent with the previous observations that the N-terminal 16 residues and the C-terminal 40 residues of SK are functionally dispensable for plasminogen activation (Kim et al., Biochemical and Biophysical Research Communications 1996; 40: 939-945; Young et al., Journal of Biological Chemistry 1995; 270: 29601-29606). In fact, the N- terminal 16 residues of SK play a role in the secretion of this protein from the host cell (Pratap et al., Biochem. Biophys. Res. Commun. 1996; 227: 303-310).

There is also evidence showing that some fragments in the region of residues 45-70, which is disordered in the complex structure, exist in an inherently flexible state (Nihalani et al., Protein Science 1998; 7: 637-648).

Roughly speaking, every one of the three domains of SK belongs to the ß-grasp folding class (Murzin et al., Journal of Molecular Biology 1995; 247: <BR> <BR> <BR> <BR> 536-540), but with some noticeable differences (see Figure 3). Like a typical ß- grasp protein, the SK a-domain contains a single a-helix packing against a mixed five-stranded ß-sheet, In addition, there is a short two-stranded ß-sheet on the same side of the major ß-sheet as the a-helix. The ß-strands forming the majorαß2,αß1,αß7,αß4,andαß5.Thetopologyofthisß-sh eetare (Richardson et al., Journal of MolecularBiology 1976 ; 102: 221-235) is (+1,- 3x,-1,2x). The hydrogen bond network of the major ß-sheet is disrupted at the middle of the a (32 strand by a bulge at position 36. The a-helix is located between aß3 and aß4 and is thus named aa34. Between the major ß-sheet and the a-helix is the hydrophobic core of the a-domain. Disturbing this hydrophobic core is likely to result in a dysfunctional SK as shown by a Gly24 to His mutation (Lee et al., Biochemical and Biophysical Research Communications 1989; 165: 1085-1090). The SK ß-domain shares the same overall folding with the SK a- domain. The coordinate rmsd between the two domains is 1.7 A for 81 residues, using a 4.0 A cutoff. Some corresponding loops between the two domains, however, have different lengths. The SK y-domain contains a four-stranded major (3-sheet and a short two-stranded (3-sheet. The major (3-sheet has a topology similar to that of the major ß-sheet of the a-domain without a (35.

Between rß2 and Yß3 are some coiled coil loops. The qualities of the electron density in the a-and y-domains of SK are significantly better than that of the (3- domain region. Correspondingly, the average temperature factors of the a-, (3-, and y-domains are 43,80, and 39 A2, respectively. These differences appear correlated with the interactions of each domain of SK with the pPlm molecule in the complex.

Example 5: Interaction between streptokinase and IlPlm The SK molecule has extensive interactions with the IlPlm molecule, mostly through the SK (3-and y-domains. The values of buried molecular surface area are 1,650 A, 950 A2 and 1,500 A between WPlm and the a, (3 and y-domains of SK, respectively.

The SK a-domain is located near the catalytic triad of , Plm. There are three major contact regions between the SK a-domain and µPlm (see Figure 4a).

The first region contains the interaction between the major ß-sheet, particularly the strands of αß1 and αß2, of SK and the loop region of residues 713-721 of RPlm. In this contact region, Arg7l9 of plasmin (SEQ ID NO: 1) forms salt bridges with both Glu39 and Glu134 of SK (SEQ ID NO : 4), and it also has van de Waals interaction with SK Val'9. The uncharged alkyl group side-chain of residue 19 of SK has been shown to be important for plasminogen activation (Lee et al., Biochemistry and Molecular Biology International 1997 ; 41: 199- 207). Arg7l9 of plasminogen has also been identified as an important residue involved in the SK-Plg complex formation (Dawson and Pontin, Biochemistry 1994; 33: 12042-12047). The second contact region contains the interaction between the bulge region in the a32 strand of SK and the 643-645 region of tPIg, which is the upstream region of the catalytic residue Asp646. The positively charged side chain of µPlm Lys644 also protrudes towards the C- terminus of the a-helix, aa34, of SK, presuming a helix dipole-charge interaction. The third contact region is between the loop following the a-helix, αα3,4, of the SK a-domain and the 606-609 region ofjuPlm. The latter is the down stream region of the catalytic residue His603. These close interactions between the SK a-domain and the catalytic triad of ßPlm are likely to contribute to the substrate specificity difference between plasmin and the SK-plasmin complex. The mode of interaction between the SK a-domain and, uPlm is clearly different from that of some other ß-grasp folding proteins. For example, the Ras-binding protein, C-rafl, binds to its target protein, Rap 1a, by forming an extended ß-sheet between the edges of their existing ß-sheets (Nassar et al., Nature Structural Biology 1996; 3: 723-729).

The SK y-domain binds to RPlm near the activation cleavage site of plasmin (see Figure 4b). On the IlPlm part, this interaction mainly involves two loop regions: RPlm (622-628) in the so called calcium-binding loop and u. Plm (692-695) in the so called autolysis loop. The interactions include a salt- bridge between SK Lys332 and Plm Glu623, hydrogen bonds (e. g. SK Glu311 and u. Gln622), and hydrophobic interactions. The amino acid sequence of region 622-628 in human plasmin (ogen) is"QEVNLEP" (amino acids 622-628 of SEQ ID NO: 1), and in bovine plasmin (ogen) is"NEKVREQ" (amino acids 643-649 of SEQ ID NO: 6). Since this region of u, Plm is involved in the SK binding, the sequence difference shown above may provide explanation why the catalytic domain of bovine plasminogen binds with SK significantly weaker than human plasminogen does (Young et al., Journal of Biological Chemistry 1998; 273: 3110-3116). On the other hand, the only close interaction of SK with the activation loop region of u. Plm (i. e. around u. Plm (558-566)) is that between SK Ala342 and pPlm Val567. Therefore human plasminogen activation by SK is unlikely to require direct contact of SK with the activation loop of plasminogen.

Furthermore, the observed C-terminus of the SK y-domain is on the side opposite to these u, Plm binding regions. Hence it is probable that the last 40 or so residues, which are disordered in the crystal, have nothing to do with the SK- Plg complex formation.

Although there is no kringle domain present in the complex crystal, kringle domains have been shown to be involved in plasminogen activation by SK. Since the N-terminus of the catalytic domain of plasmin (ogen) is on the hemisphere opposite to the SK binding sites, the extension of kringle 5 from the catalytic domain is unlikely to disturb the observed interactions between SK and u. Plm. Also, it has been shown previously (Rodriguez et al.. European Journal of Biochemistry 1995; 229: 83-90) that the complex of plasmin with the fragment SK (143-293) (i. e. the ß-domain) or SK (143-386) (i. e. the P-and y- domains) is very rapidly inhibited by a2-antiplasmin, whereas the complex with intact SK is resistant to inhibition. Furthermore in the case of SK (143-386), inhibition by a2-antiplasmin results in dissociation of the SK-Plm complex ; SK (143-293), in contrast, remains associated with the a2-antiplasmin-plasmin complex. The results suggest that the a2-antiplasmin-plasmin interface overlaps with the SK a-domain binding site and partially overlaps with the SK y-domain binding site on the plasmin surface, while the SK ß-domain binding site may have nothing to do with the a2-antiplasmin binding site.

Example 6: Putative substrate binding site of the SK-Plm complex Although it is active against fibrin, plasmin alone can not convert plasminogen to plasmin. This substrate binding specificity can be explained by a comparison of the crystal structure of ßPlm with that of the catalytic domain of human t-PA (Renatus et al., EMBO Journal 1997; 16: 4797-4805). While both are trypsin-like proteases, human t-PA has high specificity towards activation of plasminogen. The overall structures of these two domains are similar (see Figure 5). The coordinate rmsd between the catalytic domain of human two-chain t-PA (Lamba et al., Journal of Molecular Biology 1996; 258: 117-135) and human uPlm is 0.74 A for 177 Ca atoms, using a 1.5 A cutoff.

However there are many structural differences on the enzyme surface around the active site. There are at least three significant backbone differences around the active site. 1) Corresponding to plasmin (ogen) between residues 644 and 645, t- PA has a six-residue insertion with three aspartate residues in a row. This insertion in t-PA is structurally replaced by part of the SK a-domain in the SK- ZPlm complex. 2) The"autolysis loop"region of ZPlm (689-695), which contacts the SK y-domain, is different from the corresponding region of t-PA. 3) The region ofuPlm (711-720) is different from the corresponding region in t-PA, where it is called the"methionine loop". The u, Plm conformation in this region makes it possible to have a complementary contact with the SK a-domain. 4) In the so called"37-loop"region of t-PA which interacts with the natural inhibitor, PAI-1, and is involved in fibrin specificity (Bennett et al., Journal of Biological Chemistry 1991; 266: 5191-5201; Madison, et al., Proceedings of the National Academy of Sciences, USA 1990; 87: 3530-3533), human plasmin (ogen) is four residues shorter around residue 583. Some of these differences, if not all, are likely responsible for the substrate specificity difference between plasmin and t- PA. The SK-plasmin complex may change the substrate specificity of plasmin by compensating for some of these differences. On the other hand, it has been shown that the isolated, synthetically prepared activation loop of plasminogen can not be cleaved by plasminogen activator (Ganu and Shaw, International Journal of Peptide Protein Research 1982; 20: 421-428). This observation suggests that not only the amino acid sequence of the cleavage site but also its conformation and the overall structure of the substrate zymogen contribute to the specificity of the activator.

The crystal structure of the SK-Plm complex shows that the complex has an opened cavity (see Figure 6a) compared with the spherical (convex) shape of the catalytic domain of plasmin (ogen). Therefore the SK-Plg complex should provide more substrate binding surface than the plasmin molecule alone can. A manual molecular model to dock a model micro plasminogen molecule into the substrate binding site of the SK-Plm complex is shown in Figure 6b.

In such a model, the activation bond, Arg56l-Val562, of the substrate (micro) plasminogen is positioned into the active site of the catalytic (micro) plasmin; the N-terminus of substrate pPlg is positioned to be closed to the disordered region of SK (45-70). From the bottom of the substrate binding concave, jPlm contributes approximately 1,050A2 binding area, mostly from the surface of the strand, aß2, and the a-helix, aa34. Since it might be modeled close to the kringle 5 domain of the substrate plasminogen, the flexible SK (45-70) region may provide extra substrate binding surface; that would explain the observed high affinity of the SK a-domain with the kringle domains of plasminogen (Young et al., Journal of Biological Chemistry 1998; 273: 3110-3116) as well as the important role played by residues 45-51 of SK in binding with plasminogen (Nihalani et al., Protein Science 1998; 7: 637-648). Forthe SK ß-domain, the major (3-sheet forms part of the wall of substrate binding concave with its helix side facing outside. The (3-strand on the rim of the R-sheet, ß2, potentially forms hydrogen bonds with the strand of residues 625-629 of substrate plasminogen. The SK (3-domain contributes-550 A2 binding surface in total.

The SK y-domain contributes some coiled coil, around residue 330, to the substrate binding, about 150 A2 in total. Several potential salt bridges can be predicted from this hypothetical model, including Arg561 of the substrate plasminogen (s-Plg) to Asp735 of the catalytic plasmin (c-Plm), s-Plg Lys557 to c- Plm Glu606 and s-Plg Lys750 to SK Asp78. Lys557 of plasminogen was found to be important for plasminogen activation by t-PA (Wang and Reich, Protein Science 1995; 4: 1769-1779) and could be explained if similar binding modes were assumed for binding of t-PA to plasminogen. The Asp735 of c-Plm, interacting with the side chain of Arg561 of s-Plg, defines the substrate specificity of the SK-Plm complex at the S 1 position.

Example 7: A possible activation mechanism of human plasminogen by SK One of the functions of SK is to turn the zymogen plasminogen into an active"enzyme"without cleaving the peptide chain. It appears from the crystal structure of the SK-Plm complex that it is the interaction between the SK y- domain and the catalytic domain of plasminogen that creates the enzymatic activity of human plasminogen-SK complex.

As shown by many crystal structures of trypsin (ogen)-like proteases, one characteristic of the proteases in this family is a buried salt bridge formed in the activation pocket associated with the activation. In the classical case, this salt bridge is formed between the carboxylate group of the aspartate residue that is immediately upstream of the catalytic nucleophile serine residue and the liberated amino terminus after the activation cleavage. The formation of the salt bridge reorients the aspartate residue relative to the zymogen structure and thus restructures the active site, which includes the oxyanion hole, the catalytic triad and the S1 specificity pocket (Freer et al., Biochemistry 1970; 9: 1997-2009).

However, non-cleavage activation has also been found in some proteins in this family, including t-PA and vampire-bat plasminogen activator (v-PA) (SEQ ID NO: 7) (Renatus et al., EMBU Journal 1997; 16: 4797-4805; Renatus et al., Biochemistry 1997 ; 36: 13483-13493). Human t-PA has significantly high plasminogenolytic activity in its single-chain form. v-PA, on the other hand lacks the activation cleavage site. Both of them switch between the active and the inactive stages in response to environmental changes, including the presence of fibrin. In their crystal structures, complexed with high affinity inhibitors, both plasminogen activators were frozen in the"active stage". These structures demonstrate that in such an intact protein, a buried lysine residue replaces the functional role of the released amino terminus of the protease activated by the cleavage. The conformation of this lysine residue, Lys' (numbered as Bode & Renatus (Bode and Renatus, Current Opinion Structural Biology 1997; 7: 865- 872), unlike the newly released amino terminus which usually stays in one form, may switch between the active and the inactive stages depending on environmental conditions such as the local concentration of cofactors (Lamba et al., Journal ofMolecular Biology 1996; 258: 117-135; Nienaber et al., Biochemistry 1992; 31: 3852-3861).

A similar scenario is likely to exist in the SK-Plg complex (see Figure 7).

Based on three-dimensional structural comparison and sequence alignment, it is evident that the same lysine residue is also conserved in plasminogen and is located at position 698. When the activation loop remains intact, and the activation pocket is not occupied by the released amino terminus, the plasminogen molecule is a two-stage proenzyme which predominantly stays in its active form. The binding of SK y-domain to the"autolysis"loop region, which is upstream of the conserved lysine residue is likely to be the trigger for plasminogen to switch from its inactive form to the active form. In such an active form, Lys698 forms the critical salt bridge with Asp740. On the other hand, when the activation pocket is occupied by the released amino terminus, the binding of SK y-domain will have little effect on the amidolytic activity of plasmin (ogen). Some factors, which were proposed to favor the active form of single-chain t-PA and v-PA, seem to have little effect on plasminogen activation by SK. For example, the so called zymogen triad, Asp'94, His40, and Ser32, present in trypsinogen but not in t-PA or v-PA, was assumed to lock Aspe94, and thus the oxyanion stabilizing loop, in its inactive form. These residues are present in plasminogen as Asp740, His586, and Ser578, but they do not prohibit the formation of active-zymogen upon binding with SK.

It is believed that the SK y-domain binds to plasminogen to create an active-zymogen, while the binding of the SK a-and ß-domains changes the substrate specificity of the active-zymogen. Although the detailed mechanism remains to be established, this model can be used to explain the following observations. The combination of SK (220-414) and SK (16-251), but not either peptide alone, effectively activates human plasminogen (Young et al., Journal of Biological Chemistry 1998; 273: 3110-3116). An explanation to this observation might simply be that the a-domain and y-domain have different functions that compensate with each other in plasminogen activation. SK (16- 251) dose-dependently enhanced the activation of plasminogen by SK (16-414) (Young et al., Journal of Biological Chemistry 1998; 273: 3110-3116). SK or SK (16-414) can convert plasminogen to plasmin; however, plasmin alone can not convert other plasminogen to plasmin. The additional SK (16-251) peptides, on the other hand, form complexes with the newly formed plasmin molecules and modulate their substrate specificity such that new plasminogen activators are formed.

Example 8: Structure comparison between the a-domain of SK and staphylokinase Among the ß-grasp folding family (Murzin et al., Journal ofMolecular Biology 1995; 247: 536-540), staphylokinase (SAK) is another bacterial source plasminogen activator, whose crystal structure has been determined recently (Murzin et al., Journal of MolecularBiology 1995 ; 247: 536-540). Like SK, staphylokinase activates human plasminogen by forming a zymogen-activator complex (Lijnen et al., Journal ofBiological Chemistry 1991; 266: 11826- 11832). However, SK and staphylokinase do not share detectable sequence homology. The size of staphylokinase is only one third that of streptokinase, and its binding mode and activation mechanism to human plasminogen are unknown. It is particularly interesting to find that staphylokinase shares a three- dimensional folding with the SK a-domain (see Figure 8). The coordinate rmsd between staphylokinase and the SK a-domain is 1.8 A for 91 Ca atoms, using a 4.0 A cutoff.

Based on the three-dimensional structural similarity between the SK a- domain and staphylokinase, we propose that staphylokinase binds to plasminogen in the same mode as the SK a-domain. Along this line, several observations on the staphylokinase-plasminogen interaction could be explained.

First, SAK Glu46, which corresponds to SK Glu39, was found to be important for the formation of the SAK-Plg complex (Silence et al., Journal of Biological <BR> <BR> <BR> Chemistry 1995; 270: 27192-27198). This residue would be located on the ß2 strand and form a salt-bridge with Arg7l9 of plasmin (ogen). SAK Met26, which corresponds to SK Val'9, was also found to be of crucial importance for the activation of plasminogen by staphylokinase (Schlott et al., Biochemical and Biophysical Acta. 1994; 1204: 235-242). Since the side chain at this position has van de Waals contact with the hydrophobic portion of the Arg719 side chain, disturbing such a contact would result in disturbing the complex contact through both van de Waals and electrostatic interactions. A few more residues of staphylokinase that were found important for the activity of the complex of SAK-Plg, including Lys50, Glu65, and Asp69, all would be located on the substrate binding surface that is similar to that we have proposed for the SK- , complex. In such a scenario, the N-terminal fragment of staphylokinase (i. e. residues 1-20), most of which are disordered in the crystal structure, would be located at a position that potentially could affect the conformation of the active site by allosteric binding. Therefore, a one domain protein like staphylokinase would perform multiple functions that are accomplished by two/three domains in SK.

The teachings of the references cited herein and referenced below are specifically incorporated herein. Modifications and variations of the present invention will be obvious to those skilled in the art from the foregoing detailed description and are intended to be encompassed by the following claims.