WO1997002340A2 - Methods for inhibiting factor xiii activity - Google Patents

Abstract

Methods for inhibiting Factor XIII activity feature a ligand that forms at least one contact, at a distance of about 5 Å or less, with at least one amino acid residue of a Factor XIII monomer. Factor XIII inhibitors are selected or designed using the three-dimensional structure of Factor XIII as a guide. In one approach, inhibitory ligands are selected or designed based on the Factor XIII b-sandwich:core interface. In a second approach, inhibitors are selected or designed based on the catalytic site of Factor XIII. In a third approach, inhibitory ligands are selected or designed based on the Factor XIII dimer interface. In a fourth approach, inhibitory molecules were designed to occupy the Factor XIII binding site, and include an electrophilic moiety susceptible to nucleophilic displacement by the reactive Cys-314.

Description

Description METHODS FOR INHIBITING FACTOR XIII ACTIVITY

Technical Field

The invention relates to methods for inhibiting Factor XIII activity. These methods advantageously use the three-dimensional crystal structure of recombinant human Factor XIII to identify molecules that reduce or eliminate Factor XHI-catalyzed cross-linking of biological materials.

Background of the Invention

Factor XIII is a transglutaminase, a class of thiol enzymes that catalyze the formation of amide cross¬ links between giutamine and lysine residues in a variety of proteins. Such γ-glutamyl-ε-lysyl amide cross-links provide structural and mechanical stability to biological materials.

Factor XIII is the last enzyme to be activated in the blood coagulation cascade. Activated Factor XIII ("FXIIIa") catalyzes a number of covalent cross-linking reactions of fibrin in blood clots. These covalent fibrin cross-linking reactions render blood clots mechanically stable and greatly increase clot resistance to plasma degradation (fibrinolysis) . Specifically, FXIIIa catalyzes linear cross-linking between the gamma ("γ") chains of fibrin to form γ dimers. This reaction is very quick (in plasma it occurs within 30 minutes or less) , and gives mechanical strength and proteolytic resistance to the clot. Simultaneous with catalyzing γ chain cross¬ linking, FXIIIa catalyzes cross-linking of fibrin alpha ( "a" ) chains to a plasminogen inhibitor, ₂-PI. This α ₂-PI incorporation into a cross-linked clot gives a clot immediate protection against fibrinolysis. Subsequently, and at much slower rates (up to 6 days) in a clot, FXIIIa catalyzes lateral cross-linking of α chains to form clusters of five to seven α chains. In addition, FXIIIa catalyzes (i) α chain cross-linking to γ dimers; and (ii) γ dimer cross-linking with other γ chains to form γ trimers and tetramers. In general, the older the clot, the greater the amount of the fibrin cross-linking and the greater the resistance to clot lysis.

After an initial clot has been proteolytically degraded or lysed, reformation of a second or third clot at the same site becomes more life-threatening than the formation of the first clot. More specifically, when the first occluding clot is lysed, commonly the clot is not completely dissolved. As blood flow is restored over the undissolved portion of the clot, some blood constituents cause platelet deposition and promote further thrombosis. When this happens, the natural hemostatic balance (the balance between clot formation and clot degradation) is shifted from fibrinolysis toward thrombosis. The surface of any undissolved portion of the first clot is intrinsically thrombogenic, and more fibrin is often deposited before resolution of the first clot is complete. This newly deposited fibrin gives the older, undissolved portion of the clot an enhanced opportunity to become more cross-linked. As a clot persists over time, more fibrin cross-linking occurs, and the denser and more resistant to fibrinolysis the clot becomes, until a hard clot is formed, i.e., a very mature, fibrin- dense clot.

To prevent formation of fibrinolysis-resistant clots, a commonly administered oral anticoagulant is dicoumarin. Dicoumarin inhibits the proper formation of clotting proteins via mechanisms in the liver involving γ carboxylation of clotting proteins. In general, dicoumarin impairs the body's ability to make coagulation factors. However, the beneficial effects of dicoumarin are not immediately achieved -- generally a three to eight day period of dicoumarin administration is needed for effective clot inhibition. Additionally, administration of dicoumarin is associated with toxicity. Dicoumarin affects homeostatic balance, such that thrombosis is reduced, i.e., clot formation is diminished or inhibited. Consequently, the patient is more prone to internal bleeding.

Factor XIII catalyzes other physiologically- relevant cross-linking reactions. For instance, Factor Xllla cross-links fibrin to the adhesive glycoproteins fibronectin and thrombospondin to anchor the blood clot to the site of injury. Factor Xllla also cross-links fibrin to other coagulation substrates, such as Factor V, platelet actin, and von Willebrand factor.

Outside of the blood coagulation system, Factor Xllla cross-links fibronectin,- vitronectin, collagen and lipoprotein A in the extracellular matrix, and is important in the stimulation of connective tissue cells and the steering of their processes involved in inflammation and wound healing.

Deficiencies in Factor XIII lead to rebleeding after initial clot formation; umbilical and intracranial bleeding in children; and a 100% spontaneous abortion rate in women. Abnormal levels of Factor XIII or of Factor XIII activity have been linked to a variety of diseases, including ulcerative colitis and other gastrointestinal disorders; liver cirrhosis; acute leukemia; rheumatoid arthritis; retinal disease; sclerosis; atherosclerosis; and heart disease.

Factor XIII, in combination with fibrinogen, thrombin and other clotting proteins, may be used as a component of "fibrin glue," a tissue adhesive that is used as a replacement or adjuvant to conventional 97/02340 PC17US96/11182

sutures. In addition, macrophages associated with some cancerous tumors have been found to contain Factor XIII, and have been shown to be involved in the deposition of cross-linked fibrin or fibrinogen in breast cancer and Hodgkin's disease.

Two agents that are often used to prevent clot formation are tissue plasminogen activator and streptokinase. These agents activate normal plasmin, but have no effect on resolution of clots that have formed (e.g., no effect on clot cross-linking and stabilization) . In contrast to these presently available agents, FXIIIa inhibitors or inactivators do not prevent clot formation; instead, FXIIIa inhibitors/inactivators aid in clot resolution if one or more clots form. Since FXIII inhibitors do not affect homeostatic balance, such inhibitors do not have the undesirable side-effect of overt bleeding or toxicity.

Accordingly, there is a need for compounds, compositions and methods that reduce or eliminate FXIIIa- catalyzed cross-linking of biological materials. The present invention fulfills this need, and further overcomes the disadvantages associated with presently available agents and methods.

Brief Description of the Drawings

Figures 1 and 2 provide the molecular coordinates of Factor XIII obtained by three-dimensional X-ray crystallography to 2.65 A resolution. Figure 1 represents Factor XIII monomer 1; Figure 2 represents Factor XIII monomer 2. The monomers 1 and 2 are related by 2-fold symmetry (i.e., the first can be generated by rotating the second by 180 degrees) .

Figure 3 is a stereo diagram showing the overall and secondary structure of the Factor XIII A- subunit. Figure 4 provides two orthogonal views of the Factor XIII A-subunit dimer, represented as coil structures.

Figure 5 is a close-up stereoview of the active-site cavity of one A-subunit in the Factor XIII dimer, wherein the backbone is depicted as a coil structure, and secondary structure features are indicated by thick helices and arrows. Side chain atoms of denoted residues are illustrated. Figure 6 is an expanded stereoview of the active site of one A-subunit in the Factor XIII dimer. Side chain atoms of Cys-314, His-373, Asp-396, Tyr-560 and Arg-11' are depicted.

Figure 7 is a first-generation series of Factor XIII inhibitory molecules.

Brief Summary of the Invention

It is an object of the present invention to provide methods for inhibiting Factor XIII activity. Within one aspect of the invention, the methods comprise the steps of: providing a ligand that forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue of a Factor XIII monomer, wherein the amino acid residue is selected from the group consisting of Cys-314, His-373 and Asp-396; and combining Factor XIII and the ligand, thereby inhibiting the activity of Factor XIII. The step of combining Factor XIII and the ligand can be performed in vitro. In a preferred embodiment, the ligand additionally forms a contact, at a distance of about 5 A or less, with at least one amino acid residue selected from the group consisting of His-342, Glu-434, Asp-343 and Trp-279. In another preferred embodiment, the ligand is a peptide that is selected to correspond to a portion of an inhibitory Greenberg peptide or to a portion of a Factor XIII core and/or sandwich domain that contributes to the 97/02340 PCΪ7US96/11182

Factor XIII β-sandwich:core interface. In yet another preferred embodiment, the ligand is a peptidomimetic or non-peptide molecule having optimal complementarity with a binding site of Factor XIII. More preferably, the ligand forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue selected from the group consisting of Trp-279, Tyr-372 and Gln-313. A preferred ligand also has an electrophilic moiety susceptible to nucleophilic displacement by Cys-314.

A second aspect of the invention provides methods for inhibiting Factor XIII activity, comprising: providing a ligand that forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue of a Factor XIII monomer, wherein the amino acid residue is selected from the group consisting of Arg-77, Arg-78 and Asp-191; and combining Factor XIII and the ligand, thereby inhibiting the activity of Factor XIII. Preferably, the ligand additionally forms a contact, at a distance of about 5 A or less, with at least one amino acid residue selected from the group consisting of Tyr-69, Cys-327, Gln-80, Asn-185, Try-181 and Asp-236.

Within a third aspect of the invention, methods for inhibiting Factor XIII activity, comprising: providing a ligand that forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue of a Factor XIII monomer, wherein the amino acid residue is selected from the group consisting of Lys-257, Glu-401, Lys-113, Asp-367, Arg-260 and Asp-404; and combining Factor XIII and the ligand, thereby inhibiting the activity of Factor XIII, are disclosed.

These and other aspects of the invention will become evident upon reference to the following detailed description. Detailed Description of the Invention

Before describing the present invention in detail, it may be helpful to define certain terms that are used herein: The term "modulator" is used herein to denote a molecule that either up-regulates or down-regulates a biological activity, and particularly Factor XIII enzymatic activity. The terms "ligand" and "modulator" are used interchangeably herein. The term "inhibitor" is used herein to denote a molecule that down-regulates a biological activity. The term "inhibitor" encompasses inactivators, antagonists, repressors, and the like. An inhibitor herein may exhibit about 5% to 100% inhibition of a selected biological activity; preferably, the amount of inhibition is experimentally or clinically significant.

The term "FXIII" is used herein to denote natural or recombinant Factor XIII (or portions thereof) , including zymogen and activated forms of Factor XIII. The term "FXIIIa" refers to activated Factor XIII.

In platelets, placenta, and other tissues, Factor XIII is a dimer of two A-chains, each containing 730 amino acid residues (A. Ichinose et al. , Biochemistry 2ϋ:6900-06, 1986), while in the circulating plasma Factor XIII is a heterodimer of two A- and two B-subunits. Both forms are activated by thrombin in the presence of Ca²⁺, with release of a 37-residue amino-terminal peptide; in the case of the plasma form, dissociation of the B- subunits occurs as well. The three-dimensional structure of human recombinant Factor XIII (EC 2.3.2.13 zymogen; protein- glutamine:amine γ-glutamyltransferase A-chain; a transglutaminase zymogen) was originally solved at 2.8 A resolution by X-ray crystallography. Further work resulted in a structure refined against data to a higher resolution of 2.65 A (see Figures 1 and 2). These atomic coordinates have been deposited in the Protein Data Bank, Chemistry Department, Brookhaven National Laboratory, Upton, NY 11973 (reference 1 GGT) . This structure shows that each chain of the homodimeric protein is folded into four sequential domains. A catalytic triad reminiscent of that observed in cysteine proteases has been identified in the core domain. The amino-terminal activation peptide of each subunit crosses the dimer interface and partially occludes the opening of the catalytic cavity in the second subunit, preventing substrate binding to the zymogen.

Two additional coordinate sets have been obtained by continuing X-ray diffraction studies of Factor XIII, and these are also informative starting points for the structure-based design of Factor XIII ligands. One is a 2.2 A resolution structure of the Factor XIII zymogen, and the other is a 2.5 A resolution structure of thrombin-cleaved Factor XIII. There are two A-subunits in each crystal structure.

IDENTIFICATION OF INHIBITORS OF FACTOR XIII ACTIVITY

Achyuthan et al. have reported that certain Factor Xllla-derived peptides inhibit transglutaminase activity (J. Biol. Chem. 2£8_:21284-92, 1993) . In particular, two peptides, Asn-72 - Asp-97 (peptide-4) and Asp-190 - Phe-230 (peptide-7) , inhibited Factor Xllla cross-linking of fibrin, N, N' -dimethylcasein or fibronectin. These peptides are sometimes referred to as "Greenberg peptides." Achyuthan et al. stated that peptides-4 and -7 had no effect on binding of Factor Xllla to fibrin, but were identified as potential Factor Xllla substrate recognition sites. Thus, the fibrin binding sites and the substrate recognition sites of Factor Xllla appeared to be distinct. More specifically, Achyuthan et al. speculated that the inhibitory peptides reacted with giutamine-containing substrate, thereby decreasing the effective concentration of substrate in the reaction.

As described further herein, an alternative explanation for the inhibitory effect of the Greenberg peptides is that they bind to the sandwich or core domain of Factor XIII and thereby act as a "wedge" between these two parts of the enzyme. Insertion of this "peptide wedge", rather than a fibrin recognition exosite, results in inhibition of FXIII enzymatic activity. This enzyme inactivation may result from destabilization of the sandwich:core interface.

Upon removal of the activation peptide from Factor XIII, the accessible surface area buried between the sandwich domain and the core domain was determined to be 768 A². Some of this area is artifactual, due to cleavage of the F184-N185 peptide bond (which makes a contribution to the accessible surface area determination) . This relatively small amount of accessible surface implies that there is limited flexibility in this region of the molecule. However, the amount of accessible surface buried by collapsing the barrels onto the core domain is substantial (i.e., 2,184 A²-¹. This area is comparable to the amount buried when the two Factor XIII monomers make the dimer (i.e., 2,900 A²) . Therefore, based on these calculations, the barrel domains should not dissociate. Considering that the Greenberg peptides bind to fibrin, one hypothesis features movement of the barrel domains to increase accessible surface area, leading to Factor XIII activity. While the exact biological function of the barrel domains is unknown, barrels are present in all members of the transglutaminase family. The barrel domains may function to isolate the transglutaminase reaction from water (a necessary condition for the transglutaminase reaction, preventing the equilibrium from shifting to a giutamine deamidase reaction) . However, in many instances, the activity of the transglutaminase enzymes increases when the barrel domains are removed or clipped at the linker peptide (residues 501-515 of Factor XIII) . Within the current theory, the presence of the barrel domains depresses Factor XIII activity. For instance, it has been reported that barrels 1 and 2 (a 19 kDa fragment) have no affinity for fibrin binding (C. Greenberg et al. , Biochem. J. 2_5_£:1013-19, 1988) . If substrate is not present, the barrel domains collapse onto the surface of the molecule and clog the active site, preventing the formation of random amide cross-links between giutamine and lysine residues in a variety of proteins. This inactive form of Factor XIII is the form that was crystallized and analyzed herein.

Peptide-4 corresponds to residues at the interface of the sandwich and core domains. Peptide-7 contacts the sandwich domain and the two barrel domains. While effects on the sandwich domain may explain the inhibitory action of peptide-4, such explanation for inhibition by peptide-7 was strained. However, with the finding that peptide-7 also contacts the R-78 residue intimately, and that F-230 makes strong contacts with Y- 69 and E-70, including main-chain H-bonds, a single interaction suffices to rationalize all of the results found by Achyuthan et al.

First, Achyuthan et al. found that peptides-4 and -7 inhibited γ-γ fibrin crosslinking, as well as 5'- (biotinamido)pentylamine (BP) incorporation into N, N' - dimethyl casein and fibronectin, all with IC50 values of 5-25 micromolar. The "wedge" theory proposed herein explains this result as a Ki for binding the peptides and destabilizing Factor XIII. More specifically, the inhibition of both the dimethyl casein and fibrin cross- linking by peptides-4 and -7 could be overcome by adding large amounts of fibrin to the assay. If fibrin stabilizes Factor XIII by binding at some "fibrin-binding site," this result is plausible. Since fibrin has profound effects on the kinetics of calcium-dependent thrombin activation, the existence of such a site is reasonable.

Second, for peptide-7, Achyuthan et al. found that both the N-terminal and, independently, the C- terminal portions were inhibitory. However, these portions of peptide-7 were much less inhibitory than the complete, intact peptide-7. It is noteworthy that the major contacts of peptide-7's N- and C-terminal regions with the sandwich:core interface involve D-191 and F-230.

Finally, Achyuthan et al. demonstrated that peptides-4 and -7 were inhibitory for guinea pig liver tissue transglutaminase. With the extensive conservation of peptide-7 residues among the transglutaminase family and Factor XIII (see below) , this result is quite reasonable in view of the "peptide wedge" theory.

97/02340 PC17US96/11182

12

TABLE 1 Peptide-7 Sequence Analysis

Peptide-7; Residues 19Q-232;

DDAVYLDNEKEREEYVLNDIGVIFYGEVNDIKTRSWSYGQFED FXIII eDaVYLDsEeERqEYVLNqqGflYmGSakqlkgrPWNFGQFEd Consensus T'gases

* *#@@ % #* **** * *$ *! * *#$*** Conserved TtTTtcchHHHHHHHhTteEEEEEEEETTEEEEEEEEetTTtT 2° structure mBmBmBammabBmBBBbmbBBBBBbBBmmbmbmBBBBBBBBBm Access'blty; intact FXIII o o Barrel 1 & 2 contact

««« Sand . dom. contact

"Conserved" denotes conserved amino acid sequence among the transglutaminase family. The symbols mean that the factor XIII amino acid is: conserved in 15 out of 15 sequences; conserved in 11 out of 15 sequences;

# conserved in 9 out of 15 sequences; @ conserved in 7 out of 15 sequences; $ conserved in 3 out of 15 sequences; and

I conserved in 1 out of 15 sequences. "2"structure" indicates secondary structure of the protein. The symbols represent: T, turn; t, end of turn; c, coil; H, helix; h, end of helix; E, beta strand; and e, end of beta strand.

"Access'blty" is the static accessibility scale, which indicates percent accessibility for the native Factor XIII zymogen having all domains in place: B, 0-10; b, 10-30; m, 30-70; a, 70-90; A, 90-100.

"%" denotes residues that change exposure upon removal of the barrel domains or the sandwich domain. For the peptide-7 residues that change accessibility when the barrel domains are collapsed onto the core domain, the following results have been obtained:

TABLE 2 Peptide-7 Contacts With Barrel Domains

Residue Contacts in Barrels D-208 Good H-bonds to guanidino group of R-703 by both 0D1 and 0D2

1-209 3.5-4 A contacts to K-704, P-670 and R-703

Y-214 3.3-3.8 A contacts to H-716 and Y-718

G-215 > 4 A contacts to V-626 E-216 3.3-3.6 A contacts to S-624 (possible H- bond of E-2160E2 ... S-6240 {3.6 A})

N-218 > 4 A contacts to K-623

D-219 Weak contacts to V-626

K-221 > 4 A contacts to V-626 and T-628 R-223 3.6-4.0 A contacts to Y-718 ring atoms

S-224 3.1-3.9 A contacts to K-704 and E-720

S-226 3.3-3.8 A contacts to 1-706, P-670 and D-

668

G-228 > 4 A contacts to P-670 [END OF PEPTIDE-7]

E-231 3.2-3.8 A contacts with W-698 (no H-bonds)

D-232 Good H-bonds to R-674 guanidino group

Distances were measured from the June 3,

1994 orthorhombic coordinate set, Molecule 2. The resolution was 2.65 A; R-factor was 0.21; no waters had been placed in the coordinate set. Peptide-7 also contacts the sandwich domain, and these contacts involve the RRGQ sequence (residues 77-80) . An analysis of the peptide-7 sandwich domain contacts produced the following results .

TABLE 3 Peptide-7 Contacts With the Sandwich Domain

Residue Contacts in Sandwich Domain D-190 3.9 to > 4 A contacts with P-150

D-191 ODl and 0D2 make good, short H-bonds to

R-78 NH₂ and NE A-192 4.0 A contacts to R-78

V-193 3.8-4 A contacts to F-184 E-198 4.5 A contacts to NE2 of H-65

V-205 Long 4.6 A contacts to R-77

L-206 3.6-3.9 A contacts to H-65, 0..NZ H-bond to K-68 (3.1 A) D-208 > 5 A contacts to K-68 Q-229 3.9 A weak contacts to Y-69

F-230 Strong hydrophobic contacts to Y-69 and

K-68; O..NH H-bond to E-70 (2.9 A) [END OF PEPTIDE-7]

E-231 Weak contacts to E-70

D-232 Hydrophobic contacts to E-70 (3.8-4.1 A)

Peptide-4 (residues 72-97 of Factor XIII), another inhibitory Greenberg peptide, contains the critical RRGQ sequence. In fact, these are the only residues in peptide-4 that change accessibility when the sandwich domain is joined to the rest of the protein. TABLE 4 Peptide-4 Sequence Analysis

Peptide-4r Residues 72-97;

NKLIVRRGQSFYVQIDFSRPYDPRRD FXIII ekLWRRGQPFwltLhFeRgYEasvD Consensus T'gase family * **★* ★ * Conserved

SSEEEEeTtcEEEEEEEeScctTTTt 2° structure abBBBBBbmmBmBbBmBmmmbmmAAB Static accessibility ⁱb 'ξ 'S Sandwich contact to core and barrel domains

For symbol definitions, see legend of Table 1.

Analysis of the contacts to the RRGQ sequence provides the following results.

TABLE 5 RRGQ Contacts With the Core Domain

Residue Contacts in Core Domain

R-77 Intimate contacts with N-185 and W-187; weak contacts with G-239; no charge neutralization.

R-78 4 H-bonds to D-191, G-253 and N-185; D-191 neutralizes the guanidino charge G-79 3.5-4.0 A contacts with C-188 SG atom

Q-80 Contacts and H-bond to N-185

This "peptide wedge" explanation of all of the results of Achyuthan et al. is more satisfactory than the interpretation that peptides-4 and -7 correspond to the fibrin-binding domain of Factor XIII. There are numerous arguments against the latter interpretation. First, a fibrin-binding peptide should not influence the activity of Factor XIII with respect to casein or fibronectin, because, in contrast to fibrin, these substrates would be unlikely to bind peptides-4 and -7. Second, peptides-4 and -7 are located at the back of the Factor XIII molecule, far from the catalytic site. Further, most of the sequences of peptides-4 and -7 are buried beneath the sandwich and barrel domains. Third, the postulate that peptide-7 serves as the fibrin-binding site requires that barrel 2, and much of barrel 1, dissociates from the core for Factor XIII activity. The extensive interface of these two barrels with the core makes this proposal unlikely. While barrel 1 (with its Y-560 finger) may move to allow substrate access to the catalytic residue C-314, full dissociation of barrel 1 from the molecule upon Factor XIII activation probably does not occur. The accessible surface area buried when the barrel domains associate with the surface of the core domain is too large to make full dissociation of the barrels a tenable hypothesis. Moreover, if full dissociation did occur, the activity of the barrel-less protein would be expected to be greater than the activity of native Factor XIII, and this is not true. In fact, there is little difference in the Km's of the barrel-less and native proteins, implying that the affinity of Factor XIII for substrate is not affected by the barrel domains.

Accordingly, the "peptide wedge" theory of inhibition by the "Greenberg" peptides is consistent with all available data. That is, peptides-4 and -7 bind to the sandwich domain (peptide-7) or the core domain (peptide-4) of Factor XIII, and act as a wedge between the two parts of the molecule, resulting in inhibition of enzymatic activity.

Based on this theory, two approaches to Factor XIII inhibition are described herein. One approach features distortion of the β-sandwich/core interface, while the second features interference with the catalytic site.

The inhibitory activity of Greenberg peptides-4 and -7 is proposed to be related to disruption of the interface between the β-sandwich and core domains of Factor XIII. Accordingly, sequences within or associated with these "Greenberg peptides" were identified that may exhibit the same net inhibitory effect. Several peptides from the core and sandwich domains that appear to make the most contacts with their counterparts were selected for further study.

The second approach to enzyme inhibition relies heavily on the crystal structure of Factor XIII. The three amino acids thought to be involved in the catalytic mechanism are Cys-314, Asp-396, and His-373 (L. Pedersen et al., Protein Sci_^ .1:1131, 1994). Given this information, along with a data set consisting of Factor XIII inhibitors, a negative image mold suitable to accommodate the 3-D shape of potential ligands and to interact with their functional groups was obtained. This negative image mold is in essence a representation of the ligand binding site of Factor XIII. Thus, molecules that would form strong bonding interactions with those amino acids identified as crucial for ligand binding were designed.

CHARACTERIZATION QF INHIBITORS QF FACTOR XIII ACTIVITY

Several Factor XIII ligands have been identified. However, the binding site of these ligands is not known. With knowledge of the three-dimensional crystal structure of Factor XIII, the binding pocket and/or mechanism of action of these ligands can be implicated. With this increased understanding of FXIII activity at the molecular level, novel and/or optimized inhibitors can be designed. As cumulative knowledge of the function of FXIII and of the structure-activity profile of FXIII ligands increases, Factor XIII-based research and clinical agents and methods will improve.

FACTOR XIII LIGANDS Down-regulators of Factor XIII activity that are identified or characterized by the methods of the present invention are contemplated to be advantageous for use in therapeutic applications in which clot resolution is desired. More specifically, such compounds are useful when inhibition of FXIIIa-mediated cross-linking of fibrin in clots in a recipient is needed. Such applications for FXIIIa inhibitors of the present invention include diseases, disorders, and surgical interventions where enhanced fibrinolysis of new and/or old clots is useful. More specifically, the FXIIIa inhibitors of the invention may be advantageously administered to patients that have suffered a myocardial infarction, or that have recently undergone angioplasty or other cardiovascular procedures that can lead to arterial stenosis or restenosis. In addition, these FXIIIa inhibitors may be used as general profibrolytic agents, and may be beneficially administered to recipients suffering from stroke, phlebitis, deep vein thrombosis, disseminated intravascular coagulation, arterial occlusion or pulmonary embolism.

Up-regulators of Factor XIII activity that are identified or characterized by the methods of the present invention are contemplated to be advantageous for use in therapeutic applications wherein enhancement of clot formation is desired. For instance, such up-regulators may be advantageously administered, either alone or in combination with exogenous Factor XIII, to recipients who are defective in clot formation in general, and in Factor XIII-mediated clot formation in particular. FXIIIa ligands identified or characterized by the methods of the present invention can be formulated with a pharmaceutically acceptable carrier for parenteral, oral, nasal, rectal, or transdermal administration or the like, according to conventional methods. Formulations may further include one or more diluents, fillers, emulsifiers, preservatives, buffers, excipients, and the like, and may be provided in such forms as liquids, powders, emulsions, suppositories, liposomes, transdermal patches and tablets, for example. Because of their ease of administration, tablets and capsules represent the most advantageous oral dosage unit form. Accordingly, suitable carriers for tablets and capsules include starches, sugars, diluents, granulating agents, lubricants, binders, disintegrating agents and the like. One skilled in the art may formulate FXIII ligands in an appropriate manner, and in accordance with accepted practices, such as those disclosed in Remington's Pharmaceutical Sciences, Gennaro (ed.), Mack Publishing Co., Easton, PA 1990 (which is incorporated herein by reference in its entirety) . FXIII ligands may also be administered in the form of liposomes or sustained release formulations, or by means of implantable pumps or implantable slow release devices, to achieve a consistent dosage over an extended period of time. Exemplary therapeutic dosage levels of FXIII inhibitors are those that produce an acceleration of clot fibrinolysis. The exact dose for an individual patient will be determined by the clinician according to accepted standards, taking into account the nature and severity of the condition to be treated, age, weight, and general health of the patient, and other factors evident to those skilled in the art. Determination of dose is within the level of ordinary skill in the art. For instance, data from accepted animal models are generally predictive of doses in humans to within one order of magnitude. The dog coronary artery thrombosis model is an exemplary animal model in this regard. Therapeutic doses in humans can range from about 1 μg to about 100 mg FXIIIa inhibitor per kg body weight of recipient per day. In some instances, after initial resolution of one or more formed clots, a lower dosage level may be administered to maintain the desired fibrinolytic effect.

The oral absorption of FXIIIa inhibitors can be further enhanced by use of drug permeation enhancers, such as salicylates; surfactants such as bile acids and their salts, polyoxyethylene fatty acids or fatty acyl ethers; chelating agents such as ethylenediamine tetraacetic acid; and solvents such as dimethylsulfoxide and ethanol (Verhoef et al., Eur. J. Drug Metb. Pharmat.okin.af.. 15:83-93, 1990) . Pharmaceutical compositions containing FXIII inhibitors are administered in unit dosage form at daily to weekly intervals, in single or multiple doses, or by continuous infusion. An "effective amount" of such a pharmaceutical composition is an amount that provides clinically significant clot resolution and/or fibrinolysis.

The invention is further illustrated by the following non-limiting examples.

EXAMPLES

Example 1. Materials and Methods

Recombinant human Factor XIII was obtained from Saccharomyces cerevisiae in its zymogen form, which is a dimer containing two A-subunits with a total molecular weight of 166 kDa. Crystals of the protein grown from ammonium sulfate at pH 5.8 are orthorhombic (space group P2₁2₁2; cell dimensions a = 101.2 A, Jb = 182.7 A, and c = 93.4 A) There are two monomers in the asymmetric unit, giving a calculated value of 53% for the solvent content. Using multiple isomorphous replacement (MIR) techniques, starting phases obtained from five heavy atom derivatives were used to calculate 2.9 A resolution electron density maps (Tables 6 and 7) .

TABLE 6 Data Collection

Resolut'n easurems Unique % R

Data set limit no. reflec 'n complete (merge)* A

Native

Siemens, 6 2.4 490,649 44,344 63.3 0.123 crystals

SSRL 2.4 217,433 65,937 94.1 0.086

Derivative

Uranyl nitrate

Crystal 1 2.9 145,075 62,858 85.6 0.095

Crystal 2 3.4 65,429 34,868 85.0 0.077

PIP 3.4 63,245 32,578 76.3 0.086

K₂PtCl₄ 3.1 86,405 45,383 76.1 0.094

EUCI3 2.8 39,950 36,410 47.1 0.081

YbCl3 3.0 49,872 41,309 74.3 0.053 Early native data and all heavy-atom derivative data used in phasing were collected on a Siemens area detector equipped with a Rigaku RU200 rotating-anode generator and a graphite monochromator. Data processing was carried out using the XENGEN package (A. Howard et al. , J. Appl. Crystalloσr. 2£:383-87, 1987) and the program MACRO (S. Rao, J. McAlister, and E. Merritt, personal communications) . A native data set was also collected on Beamline VII-1 at the Stanford Synchrotron Radiation Laboratory (SSRL) .

*R ⁽merge⁾ = ∑h^∑i \ ^τh, i " <^∑h>l /^∑h^∑iI ^τh, iI where h measurements are unique reflection indices (I) , and i represents symmetry equivalent indices. PIP, di-μ -iodo-bis(ethylenediamine)diplatinum(II) nitrate. TABLE 7 Heavy-Atom-Phasing

Useful Average

Sites, resolution phasing

Derivative no. limit, A R(centric) * power

Uranyl crystal 1

Isomorphous 15 2.9 0.581 2.35

Anomalous 15 2.9 - 1.64

Uranyl crystal 2

Isomorphous 13 3.6 0.587 1.84

Anomalous 13 5.0 - 1.24

PIP 12 4.6 0.524 1.70

K₂PtCl₄ 12 3.1 0.621 1.47

EuCl₃ 3 4.5 0.628 1.46

YbCl₃ 4 3.2 0.628 1.70

Refinement of heavy-atom parameters was carried out by using PHASES (W. Furey and S. Swaminathan, Am. Crystallogr. Asso . Mtσ Abstr. (Am. Crystallogr. Assoc, New York), Series 2, Vol. 18, p.73 (abstr.), 1990). The combined figure of merit before solvent flattening was 0.67; that after solvent flatttening was 0.79 (B. Wang, Meth. Enzymol . H5_:90-112, 1985). Only data with F > 2σ (F) were used. Density averaging techniques exploiting the noncrystallographic 2-fold symmetry were also employed; map calculations were carried out with the CCP4 package of crystallographic software (SERC Daresbury Laboratory, CCP4 V2.2

Manual JL. Suite of Programs fo Protein

Crystallography (Daresbury Lab., Warrington, U.K.), 1993) . The program 0 was used for model building

(T. Jones et al. , Acta Crystallogr. Δ l:iιo-i9,

1991) . Isomorphous and anomalous refer to isomorphous and anomalous differences between native and heavy-atom derivative data; unless indicated, phasing information is based on isomorphous differences.

*FJcentric) = ∑ (|F_PH - -Fp| - |/_H|⁾/∑ |Fp_H " ^FPI • Phasing power = /_rms/^Erms' where /_{rms =} [ (Σf² ) n] ¹ / ² and E_rms = [∑(F_PH - |Fp + f_K \ ) ²/n] ^l/^'< Phases derived from an initial polyalanine model were combined with MIR phases to yield an improved 2.8 A resolution map from which the final model was built. An R factor of 0.434 (10.0 to 2.8 A resolution) for native data collected at a synchrotron source was calculated for this initial model. Preliminary refinement by conjugate gradient minimization was carried out with the program X-PLOR (A. Brunger et al. , Science 23_5_:458-60, 1987) . At the current stage of refinement [R = 0.217 for data from 10.0 to 2.65 A resolution, F > 2 σ(F) , with 48,797 reflections and 11,400 atoms] , the model has good geometry [rms deviation from ideality = 0.012 A (bonds) and 1.847° (angles)] , and the average value of the individually refined B factors is 26.7 A². The atomic coordinates corresponding to the structural model of Factor XIII at 2.65 A resolution are provided in Figures 1 and 2.

A second crystal form of Factor XIII zymogen was grown from 1,2-propanediol (space group P2]_; cell dimensions a = 101.1 A; b = 72.65 A; c = 135.08 A; beta = 106.1°) . Similar crystals of thrombin-cleaved Factor XIII were grown from ethanol (space group P2]_; cell dimensions a = 101.1 A; b = 73.13 A; c = 134.72 A; beta = 106.4°) . The P2^ Factor XIII zymogen structure has been refined against 10.0 to 2.2 A resolution diffraction data to give R = 0.200, for 11,434 protein atoms and 762 water molecules, and 82,329 reflections with F > 2σ(F) . The rms deviations from ideality for this model are 0.011 A

(bonds) and 1.8° (angles) ; the average value of the refined B factors is 32.5 A². The thrombin-cleaved Factor XIII structure has been refined against 49,681 reflections from 10.0 to 2.5 A resolution with F > 2σ(F) to give R = 0.182. The model contains 11,372 protein atoms and 557 water molecules and has good geometry [rms deviation from ideality = 0.011 A (bonds) and 1.9 (angles)]; the average value of the refined B factors is 29.9 A².

Exa ple 2. Subunit Structure and Activation Pen ide The current X-ray models of the Factor XIII A- subunit consist of residues 8-727 with one gap between residues 30 and 43. In the 2.65 A zymogen structure, the C-terminus of the model has been extended to residue 729; in the 2.2 A zymogen structure, weak electron density could be seen for residues 31-42 in one subunit, and the model includes this portion of the structure. In each crystal form, the two crystallographically independent subunits are very similar: the rms differences between their α-carbon atoms are about 0.5 to 0.6 A. Four distinct and sequential domains make up the Factor XIII monomer 1: the β-sandwich, core, barrel 1, and barrel 2. Figure 3 is a stereodiagram showing the overall and secondary structure in the Factor XIII monomer 1 (P. Kraulis, J. Appl. CryataHoar. 4.:946-50, 1991) . From the amino terminus, the four domains in order are the β-sandwich (top left) , the core domain (dark grey shading, right center) , barrel 1 (bottom) , and barrel 2 (left center) . The one gap in the model occurs between the amino terminus activation peptide (dark grey shading, top right) and the β-sandwich. In three of these domains, the secondary structure is predominantly β-sheet, while in the remaining domain (core) , it is mixed α-helix and β-sheet. Thus, the Factor XIII A- subunit can be thought of as a flattened trianglular wedge about 100 A long, 60 A wide, and 50 A in the third dimension, with a β-sheet domain placed at each vertex and the mixed helix and β-sheet domain at the core.

Situated at the amino terminus of the Factor XIII A-subunit is a short activation peptide. Activation by thrombin cleavage occurs at Arg-37, which is in the gap of 12 residues that could not be located in most of the various electron-density maps (since weak density was observed for this loop in 1 case out of 6) . The 23- residue section of the activation peptide in the model does not possess any regular secondary structure. Half a dozen short oxygen-to-nitrogen hydrogen bonds, with distances less than or equal to 2.9 A, hold residues Asn- 17 through Thr-28 of the activation peptide to the first domain of the A-subunit, the β-sandwich.

Example 3. Domain Structure

The β-sandwich domain consists of 140 residues from Glu-43 through Phe-184, and is characterized by two four-stranded antiparallel sheets twisted about 50° with respect to each other. There are only two turns of helix in this domain; they make up part of a 28-residue loop between the first and second β-strands of the sandwich. Overall, this domain is about 40 A in each dimension.

Two other domains that are of predominantly β structure have a seven-stranded β-barrel fold, and are located at the carboxyl terminus of the monomer. There are about 100 residues in each of these domains: barrel 1 contains residues Ser-516 through Thr-628 and is a little larger than barrel 2, which runs from Ile-629 through Arg-727. Each barrel is just under 50 A long and about 25 A wide. These domains have the "3+4" sheet structure that distinguishes the fibronectin III fold from the "4+3" β-sheet pattern seen in the immunoglobulin constant domains (D. Leahy et al. , Science 2JLS.:987-91, 1992). The two barrels in the Factor XIII A-subunit are quite similar (the rms difference between 81 equivalent α- carbon atoms within 3.0 A is 1.57 A); the larger number of residues in barrel 1 is due to two connecting segments that loop away from the body of the barrel more than any observed in barrel 2. One of these loops is between strands 3 and 4 and contains a tyrosine at position 560. While barrel 1 and barrel 2 are structurally very similar, there is no significant sequence similarity between the two domains, and neither is similar in sequence to fibronectin type III domains (P. Bork and R. Doolittle., Proc. Natl. Acad. Sci . USA 89:8990-94. 1992) . The largest and central domain in the factor

XIII A-subunit spans about 330 amino acid residues, from Asn-185 near the β-sandwich domain to Arg-515, after which barrel 1 begins. In this core domain, there are 10 short segments of α-helix; a two-turn helix is found in a loop near the β-sandwich, while the other helices are arranged in two groups, which are separated by a central β-sheet. This domain thus has several distinct structural features. At its amino-terminal end, close to the β-sandwich, are five helices. In this region two antiparallel β-strands that flank barrel 2 approach the second feature of the core, a twisted β-sheet consisting of six antiparallel strands located at the center of the domain. In turn, the β-sheet is curled along the side of the third region in the core, which contains the remaining four helical segments. The other prominent secondary structure in this region is a three-stranded mixed β-sheet near the carboxyl terminus of the core that flanks one end of barrel 1. It is in this domain that the active site of the enzyme has been located, which is described in Example 6, below.

Example 4■ Conserved Residues

Amino acid sequence alignments of 15 members of the transglutaminase family, including the non-enzymatic human erythrocyte membrane protein band 4.2, reveal that the proteins share 25-45% identity with Factor XIII. The distribution of conserved residues in members of this family among the four structural domains has been examined by sequence alignments. More specifically, amino acid sequences for the following members of the transglutaminase family were used: Factor XIII, human and mouse tissue transglutaminase, guinea pig tissue transglutaminase, bovine endothelial transglutaminase, chicken erythrocyte transglutaminase, human and rat keratinocyte transglutaminase, human and rat prostate transglutaminase, human and mouse epidermal transglutaminase (transglutaminase 3) , Limulus hemocyte transglutaminase, annulin, and human p4.2 protein.

A small number of the conserved residues are located in the β domains, but a dramatic majority (136 of 180) are found in the central-core domain. This result indicates that the fold of the core domain and the catalytic mechanism remain consistent throughout the transglutaminase family, and implies that the differing substrate specificity among the transglutaminases is localized in the bordering domains.

Example 5. Dimer Structure

Each of the two Factor XIII subunits comes together with a crystallographically related mate to form the physiological dimer. Two orthogonal views of the Factor XIII dimer, represented as peptide coils, are shown in Figure 4. The activation peptide and domains in the left subunit are colored (shaded) and oriented as in Figure 3. The second (right) subunit in the dimer is in mirror image orientation to the left subunit. The location of the active site in each subunit is marked by an asterisk.

The dimer is formed when the long edges of two subunits align, and thus has a hexagonal shape in one view (see Figure 4, left panel) . There is a long interface between the monomers that involves both sandwich and core domains. Accessible surface area calculations (M. Connolly, J. Appl. Crystallogr. l£:548- 58, 1983) give values of 1710 and 1880 A² for the area buried upon formation of the two crystallographically independent dimers. These numbers are comparable to those reported for other protein-protein complexes (W. Tulip et al., £_ Mol. Biol. 222:122-48, 1992). An interesting feature of the dimer is that the activation peptide of each subunit crosses the interface to associate with the core in the other half of the dimer; in so doing, the peptide passes over the entrance to the catalytic cavity in the second monomer and precludes substrate binding. Five short hydrogen bonds hold each activation peptide to the other subunit in the dimer. In another view of the dimer (see Figure 4, right panel), the hexagon is not flat but puckered. Both activation peptides cross the large (90 A wide) concave top face; it is from this face that the active site cavities of each subunit are accessible.

Example 6. Active Site

A small number of residues that are conserved in the active transglutaminases come together to form the active site. A stereoview of the active site cavity of one A-subunit in the Factor XIII dimer is shown in Figure 5. The backbone of the protein is shown as a coil structure, with secondary structure features drawn as thick helices and arrows. Domains are colored (shaded) as in Figure 3 (core domain in light grey shading throughout the Figure, barrel 1 is lined and shaded, and the activation peptide of the second subunit in the dimer is represented by dark grey shading at the lower portion of the Figure) . Side-chain atoms of catalytic triad residues (Cys-314, His-373, and Asp-396) , several other active site residues taking part in interesting hydrogen- bond interactions (Tyr-560, His-342, Asp-343, and Glu- 434), and Arg-11¹ in the activation peptide of the second subunit in the dimer are drawn as ball-and-stick structures (nitrogen = large black, oxygen = small black, sulfur = grey at Cys-314 and carbon = remaining grey) . The apostrophe in the designation "Arg-11'" indicates that the denoted amino acid residue is within the second A-subunit of the FXIII dimer.

At the amino terminus of the longest helix in the structure sits Cys-314, which is known to be essential for the enzyme's catalytic activity (J. Folk,

Adv. Enzymol . Relat. Areas Mol. Biol. 51:1-56, 1983) .

This helix is bordered on one side by the central twisted β-sheet in the core domain. The sulfhydryl group of Cys- 314 forms a hydrogen bond to His-373 (the SΪ-N^δl distances in the two crystallographically independent monomers are 3.4 and 3.4 A in the 2.2 A structure) , which is in a strand in the middle of the β-sheet. In turn, the other nitrogen atom in the histidine ring forms a hydrogen bond with a terminal oxygen atom of Asp-396 Ctjε 2-0^ = 2.9 and 3.0 A in the 2.2 A structure), a residue that sits on a neighboring β-strand in the same sheet. Side-chain atoms of this Cys-His-Asp triad can be easily superimposed onto those for the Cys-His-Asn triad in papain (rms fit of 0.61 and 0.49 A) . Thus, the catalytic triad in FXIII, a transglutaminase, is very similar to the arrangement in papain, a cysteine protease.

In Factor XIII, these three catalytic residues are at the base of a cavity bounded by the core and barrel 1 domains. Also participating in the hydrogen- bonding pattern in the active site cavity is a tyrosine residue in one of the long loops in the barrel 1 domain: the distance from the hydroxyl oxygen atom of Tyr-560 to the sulfur atom of Cys-314 is 3.4 and 2.9 A in the 2.2 A structure for the two subunits. Such OH...S hydrogen bonds are unusual and generally found only if the sulfur atom is negatively charged (L. Gregoret et al. , Proteins 1:99-107, 1991; and S. Mangani and K. Hakansson, Eur. J. Biochem. 210 :867-71. 1992) . This result suggests that the hydrogen-bonding interaction between Cys-314 and His- 373 is that of a thiolate-imidazolium ion pair as in the cysteine proteases (E. Baker and J. Drenth in Biological Macromolecules and Assemblies. F. Jurnak and A. McPherson (eds.), Wiley, New York, pp. 314-68, 1987) . Figure 5 also shows a group of conserved residues surrounding His-

342, which forms a hydrogen bond to Glu-434 (N^δl-O^ε = 2.5 and 2.5 A in the 2.2 A structure) . The N^ε2 atom of His-

342 has no hydrogen bond to residues in the immediate area, and its distance to Cys-314 (N^ε2-SY = 8.0 and 8.0 A in the 2.2 A structure) is too far for participation in a second catalytic triad. Asp-343 is hydrogen-bonded to Arg-11¹ in the activation peptide of the second subunit in the dimer. This pair of residues, His-342 and Asp-

343, may guide the lysyl substrate residue into the active site.

The transamidation reaction catalyzed by Factor XIII is similar to the reverse of proteolytic cleavage. However, water must be excluded from the reaction to avoid deamidation side reactions. While the catalytic triad of Factor XIII resembles that of the cysteine proteases, the active site of Factor XIII is less accessible than that of the proteases. The transamidation reaction of Factor XIII involves two very large protein molecules as substrates. After removal of the activation peptide, the active site cavity can be accessed from two directions. For instance, the catalytic Cys-314 may be approached by the giutamine substrate macromolecuie from the direction of the two β- barrel domains (from the left in Figures 5 and 6) , most likely displacing Tyr-560 by movement of barrel 1, and possibly barrel 2 as well. A second large substrate (the lysyl substrate) can access the active site cavity from the opposite direction, either from a long groove that runs along the subunit-subunit interface or by bending along the β-sandwich domain.

Figure 6 is a stereoview of the active site of one A-subunit in the Factor XIII dimer. Side-chain atoms of the catalytic triad residues (Cys-314, His-373, and Asp-396), Tyr-560 from barrel 1, and Arg-11¹ from the activation peptide of the second subunit in the dimer are shown. The catalytic triad residues become accessible to substrate upon removal of (i) the second subunit's activation peptide (center) from the opening to the active site cavity, and (ii) Tyr-560 from the nucleophilic Cys-314 of the triad. Removal of the activation peptide occurs upon proteolysis by thrombin

(at Arg-37) at a site in the gap between the activation peptide and the rest of the second subunit. Residue Tyr-

560 is removed from Cys-314 by shifting barrel 1 (light grey, bottom of Figure) away from the active site.

Movement of barrel 1, with or without barrel 2 (left margin, lined and shaded) as well, may occur upon a conformational change of the peptide linking the core domain to barrel 1. For instance, movement of one or both of the barrel domains may occur upon activation of the Factor XIII zymogen in the presence of calcium. In the Factor XIII zymogen, the catalytic triad residues are not accessible to substrate (see Figure 6) .

Three observations are consistent with additional structural rearrangements being connected with enzymatic activity. First, transglutaminase 3 is activated 15-fold by proteolysis of the linker peptide joining the core and barrel 1 domains (I. Kim et al. , J_._

Biol. Chem. 2£≤.:12682-90, 1993) . In transglutaminase 3, the linker is longer than in the other transglutaminases, indicating that the length of this segment is critical to enzyme activity. Second, a 52 kDa thrombin- or trypsin- cleaved Factor XIII fragment (corresponding to the core and sandwich domains) possesses catalytic activity (C.

Greenberg et al. , Biochem. J. 25L:1013-19, 1988), demonstrating that the core domain retains its structure independently of the presence of the barrel domains. Third, prolonged incubation of Factor XIII with thrombin leads to a Ca²⁺-dependent cleavage at Lys-513 (C. Greenberg et al. , supra; and N. Takahashi et al., Proc. Natl. Acad. Sci. USA £1:8019-23, 1986) . In the zymogen structure, this residue is located in the peptide linking the core domain to barrel 1. This residue points into the protein, participating in a salt bridge with Glu-453, while its nonpolar neighbors Val-511 and Met-512 are directed out toward solvent. Upon binding of Ca²⁺ and substrate, the peptide linking the core and barrel 1 domains may undergo a conformational change, causing the barrel 1 domain or possibly both barrel 1 and 2 domains to rotate away from the core. This movement may then make the Lys-513 residue accessible for proteolysis by thrombin.

Thus, movement of one or both barrel domains permits removal of Tyr-560 from the giutamine binding pocket, allowing the glutamyl substrate access to Cys- 314. Thereafter, the acyl enzyme can be formed and NH₃ can diffuse away from the enzyme. From computer superpositions with papain-inhibitor complex crystal structures, another Factor XIII active site residue, Trp- 279, has been identified as being likely involved in the transglutaminase mechanism (L. Pedersen et al. , Protein Sci. 2:1131, 1994). In the papain family of cysteine proteases, a conserved giutamine residue stabilizes the oxyanion intermediate in the proteolytic reaction by forming a hydrogen bond to the negatively charged intermediate. There is no giutamine residue in the Factor XIII active site that can carry out this function; instead, a conserved tryptophan residue at position 279 is most likely to perform the function of stabilizing the proposed oxyanion intermediates which form during the transglutaminase reaction. The lysyl substrate then approaches along either the β-sandwich domain or the long subunit interface groove, where it is steered to the active site by Asp-343 and the His-342/Glu-434 diad. Finally, reaction of the lysyl substrate occurs, and the cross-linked product then diffuses away.

Example 7. Designed Peptides for Interface Destab lization

A. Sandwich:Core Interface Destabilization

The "Greenberg peptides" are not fibrin recognition exosites, but rather are peptide "wedges" that inactivate Factor XIII via destabilization of the sandwich and core interface. Greenberg peptide-4

(residues 72-97) corresponds to some critical, conserved residues at the interface of the sandwich and core domains. In particular, residues 77-80 (RRGQ) are conserved in all transglutaminases. Similarly, Greenberg peptide-7 (residues 190-230) is in the vicinity of both the sandwich and the two barrel domains. There is an important interaction of residue D-191 with R-78, but there are other hydrophobic interactions as well. Analyses of the contacts that peptides-4 and -7 make with the core and sandwich domains, respectively, are presented in Tables 8 and 9, below.

TABLE 8 Greenberg Peptide-7

Residue Contacts in sandwich Distance (A)

D190 P150 3.9 D191 R78 A192 R78 4 0 V193 F184 3 8-4, E198 H65 4 5 V205 R77 4.6 L206 H65 3 6-3 D208 K68 > 5.0 Q229 Y69 3.9 F230 Y69,K68 hydrophobic TABLE 9 Greenberg Peptide-4

Residue Contacts in sandwich Distance (A)

R77 N185,W187,G329

R78 D191,G253,N185

G79 C188 3.5-4.0

Q80 N185

The above data show that the "Greenberg peptides" bind to the sandwich domain (peptide-7) or the core domain (peptide-4) of Factor XIII, thus distorting the interface between these domains by competition. Furthermore, Factor XIII is unstable when activated and quite insoluble. Therefore, peptides designed from certain sequences of the "Greenberg peptides" will disrupt this interface and destabilize Factor XIII at 37° C, the thermal denaturation temperature of the core domain. With that rationale, the peptides shown in Table 10 were synthesized and tested in a Factor XIII assay. For these peptides, C-152 was replaced by nor-valine. The sequences 150-156 and 229-236 were selected, based upon accessibility changes when one domain is unplugged. The sequence 68-71, although not part of a "Greenberg peptide," was selected based upon its hydrogen bond with C-327. The same characteristics apply to the sequence 150-156, where an accessibility change (in particular at residue 153) was noted that may affect the stability of Factor XIII. TABLE 10 Designed Peptides

Lys-Tyr-Glu-Asn (residues 68-71)

Acetyl-Lys-Tyr-Glu-Asn-NH₂

Arg-Arg-Gly-Gln (residues 77-80)

Acetyl-Arg-Arg-Gly-Gln-NH2

Gln-Phe-Glu-Asp-Gly-Ile-Leu-Asp (residues 229-236)

Acetyl-Gin-Phe-Glu-Asp-Gly-Ile-Leu-Asp-NH2

Pro-Lys-norVal-Ile-Val-Gly-Lys (residues 150-156)

Acetyl-Pro-Lys-norVal-Ile-Val-Gly-Lys-NH₂

Asp-Asp-Ala-Val (residues 190-193)

Acetyl-Asp-Asp-Ala-Val-NH2 Asn-Gly-Arg-Arg

Phe-Gly-Gly-Arg-Arg-Ala

Cyclo- (Phe-Gly-Gly-Arg-Arg-Ala)

Ser-Gly-Arg-Arg

Ser-Gly-Arg-Lys

B. Dimer Destabilization

The Factor XIII dimer interface can also be a target for FXIII inhibition via disruption. Polar residues at the dimer interface that are involved in hydrogen bonds can provide good starting points for designing inhibitors that disrupt the Factor XIII dimer interface. For instance, the following residues are useful for inhibitor design: Lys-257, Glu-401 Lys-113, Asp-367

Arg-260, Asp-404

Additionally, residue Lys-257 is the site of a Factor

XIII deficiency mutation.

Example 9. Designed Inhibitors Using Crystal Structure

The second approach to designing molecules that inhibit Factor XIII uses molecular modeling techniques. A data set representing known Factor XIII inhibitors was minimized in the TRIPOS force field using Casteiger- Marsilli charges. A set of 50 conformers of each inhibitor was generated using MULTISEARCH in Sybyl v6.1, and the resultant conformers were submitted to DISCO (Y. Martin et al. , J. Comp. -Aided Molec. Desiσn 2:83, 1993). This program identifies both the bioactive conformations and the common features responsible for bioactivity for a set of molecules. Sufficient evidence for the catalytic pocket has been implicated from the crystal coordinates of the FXIII zymogen (V. Yee et al. , Proc. Natl. Acad. Sci. USA H:7296-7300, 1994) . The process described herein determines a common alignment of the known ligands, which in turn provides a good predictive model of activity via the CoMFA method for new agents (R.D. Cramer et al. , J. Am. Chem. Soc. 110:5959-67. 1988). Establishment of the CoMFA model allowed identification of the amino acids involved in Factor XIII-ligand binding, as discussed below. Once these residues were determined, a de novo design approach was used. The proposed structures were designed with the added criterion of optimal complementarity with the receptor binding site.

Optimization of several variables resulted in the CoMFA summary shown in Table 12. Misalignment of molecules resulted in a lower cross-validated R2, indicating the importance of the alignment rule in generating reliable predictions. The structure of penicillin (inactive) that was not included in the DISCO data set was generated to match the CoMFA superposition template, and was added to the CoMFA predictive data set. This analysis used two components for optimal fitting, with a standard error in the cross-validated analysis of 0.322. No outliers were removed from the analysis. For this data set, steric and electrostatic runs were performed separately, with the steric component playing a more important role.

TABLE 12

CoMFA Summary

S = 2

9 cross-validation groups

R2 Comps F.

Full Analysis 0.367 2 23, .391 Steric 0.444 1 55.683 Electrostatic 0.021 4 79.053

The CoMFA model makes "chemical sense" for the following reasons. An examination of the CoMFA fields predicted: (1) a steric wall toward Trp-279, beyond which added bulk will decrease potency; (2) that placing hydrogen-accepting groups in close proximity to Tyr-372 will increase potency; and (3) that electron-rich moieties will interact favorably with Gln-313.

Additionally, the field representations indicated that filling the Factor XIII active site and pointing toward barrel-2 with steric bulk will have a favorable effect on activity. This finding is in agreement with experimental evidence that the two barrels occlude the entrance of the giutamine substrate to the binding pocket, and that deletion of both barrels still results in an active enzyme. Consequently, the CoMFA results pointed toward designing inhibitors that would occupy a bigger volume within the binding site. A first-generation series of inhibitory molecules was designed (see Figure 7) . Each of these molecules included an electrophilic moiety susceptible to nucleophilic displacement by the reactive Cys-SH, in accord with the proposed mechanism.

Exa l io. Assays of Factor XIII Activity

PRIMARY FXIIIa SCREENING ASSAY

Fibronectin (Enzyme Research Labs, Inc., South Bend, IN) is purified according to F. Grinell et al., Cell IS.:517-25, 1980. Purified fibronectin is diluted to 1 μg/ml in 0.1 M Na2C>₃, pH 9.7, and added to control and test assay wells of 96-well plates at 100 μl/well. All wells are incubated overnight at 4°C.

After incubation, the plate wells are washed twice with 200 μl/well of wash buffer per wash to remove all fibronectin not bound to the wells. The wash buffer contains 0.135 M NaCl, 2.7 mM KCl, 25 mM Tris at about pH 7.4, and 0.1% Tween 20® (Sigma Chemical Co., St. Louis, MO; polyoxyethylenesorbitan monolaurate) . The plate wells are then blocked with 200 μl/well blocking buffer and incubated for 1-2 h at 37°C. The blocking buffer contains 0.135 M NaCl, 2.7 mM KCl, 25 mM Tris at about pH 7.4, 0.1% Tween 20®, and 0.5% BSA (bovine serum albumin, Sigma fraction V; Sigma) . After incubation, the plate wells are again washed with 200 μl/well of wash buffer to remove excess blocking buffer. Biotin is conjugated to fibrinogen (using D- biotin-N-hydroxysuccinimide ester (BNHS, Boehringer Mannheim, Indianapolis, IN) ) according to the method of

K. Hoffman et. al. , J m_ Chem. SOC. 1 :3585, 1978.

Briefly, fibrinogen is dialyzed against 0.1 M NaHCC>₃, pH 8.5 (coupling buffer) at a concentration of 1 mg/ml. BNHS is dissolved in dimethylformamide at a concentration of 1.7 mg/ml, and added to fibrinogen in coupling buffer at a ratio of 30 μl BNHS to 1 ml fibrinogen. The mixture is incubated with rocking for 4 h at room temperature, dialyzed against PBS, and stored in aliquots at -80°C. Biotinylated fibrinogen (40 μl of 20 μg/ml solution) is added to each plate well. Except in control wells (which received 10 μl of 10% DMSO alone) , 10 μl of test sample are added to duplicate wells after the addition of biotinylated fibrinogen. FXIIIa (recombinant human FXIIIa2

(ZymoGenetics) , as described in European Patent No. 0 268 772, granted April 26, 1995, and incorporated herein by reference) is diluted to 100 μg/ml in a solution of 50 mM Tris, 5 mM CaCl2 at pH 7.4 containing l unit/ml thrombin (bovine; Enzyme Research Labs) and incubated for 20 min at 37°C. FXIIIa then is diluted to 1 μg/ml in a cross¬ linking buffer consisting of the blocking buffer described above plus 20 mM CaCl2 and 75 ng/ml D- phenylalanyl-L-prolyl-L-arginine chloromethyl ketone (PPACK; Calbiochem, La Jolla, CA) . Fifty microliters of FXIIIa in cross-linking buffer is added to each test and control well . All wells then are incubated for about 10 min at 37°C.

After incubation, the wells are washed twice with 200 μl/well of wash buffer to remove all biotinylated fibrinogen that is not cross-linked to fibronectin. Then 100 μl streptavidin-alkaline phosphatase (Amersham, Arlington Heights, IL) diluted to 1:2000 in wash buffer plus 0.5% crystalline BSA (American Research Products, Solon, OH) is added to each control and test well. All wells are incubated for about 45 min at 37°C.

The plate wells are washed three times with 200 μl/well of wash buffer to remove streptavidin-alkaline phosphatase not bound to cross-linked biotinylated fibrinogen. Substrate is prepared by adding 6 mg phosphatase substrate (Sigma) per ml of 96 ml diethanolamine (Sigma), 56 mg gCl2 and H2O to one liter at about pH 9.8. About 100 μl of phosphatase substrate is added to each plate well. If streptavidin-alkaline phosphatase is bound to biotinylated fibrinogen, the substrate is cleaved, producing a colorimetrically detectable product. Absorbance is read at 405 nm on a ThermoMax microplate reader (Molecular Devices, Menlo Park, CA) . Inhibition is measured as a percent reduction of cross-linking by a test sample as compared to control wells containing only FXIIIa and test sample diluent.

The advantage of this primary FXIIIa screening assay over fluorometric or other signal assays is that unbound reagents and test samples are washed from the wells before signal development and detection. Potential FXIIIa inhibitors may be retested in the primary FXIIIa screening assay using 100- to 1000-fold serial dilutions of test sample.

tTG SCREENING ASSAY

Because FXIII is a member of the tissue transglutaminase (tTG) family, all potential FXIIIa inhibitors identified in the primary FXIIIa screening assay are further screened for tTG (non-FXIII-specific) inhibitory activity. The tTG screening assay is identical to the primary FXIIIa screening assay, except that tTG (Sigma) is used instead of FXIIIa and no thrombin or PPACK is used.

Determination Of IC_5Q Values

IC5₀ refers to the molar concentration of a test compound that gives 50% inhibition of FXIIIa. The determination of IC5₀ values is done the same as in the primary FXIIIa screening assay. The IC5₀ values allow one to make comparisons of one compound to another. Suitable controls for this determination may include iodoacetamide (Sigma) , iodoacetic acid (Sigma) and cerulenin (2,3-epoxy-4-oxo-7,10-dodecadienamide, Sigma).

Secondary Screening Procedures A panel of secondary screening assays is used to confirm FXIIIa inhibitory activity. Specifically, these secondary screening assays include tests for inhibition of fibrinogen clotting; inhibition of other blood clotting factors (Factor Vila ("FVIIa"), Factor IX ("FIX"), Factqr Xa ("FXa"), and thrombin); inhibition of papain (which utilizes a cysteine thiol active site that is similar to one present in FXIII and tTG) ; and plasma clot lysis.

Iodoacetamide and iodoacetic acid are alkylating agents, and are used as comparative (FXIIIa- inhibiting) controls in some of the secondary screening assays. Cerulenin is a transglutaminase inhibitor and also may be used as a comparative control.

Fibrinogen clotting assay

Clots are formed in 1.5 ml Eppendorf tubes by adding reagents to give the following final assay concentrations in a total volume of 100 μl: 2 mg/ml

FXIII-free fibrinogen purified on a rabbit anti-FXIII column (Enzyme Research Labs); 20 mM CaCl2; 1 unit/ml thrombin (bovine, Enzyme Research Labs) ; and either 1 μg/ ml FXIIIa (recombinant human FXIIIa₂ (ZymoGenetics) ) or 20 μg/ml tTG (Sigma) in 120 mM NaCl plus 25 mM Tris at about pH 7.2. In addition, 10 μl of one of the following is added to each tube in varying concentrations to determine a dose response: test sample, iodoacetamide, or cerulenin. The final reaction volume in each Eppendorf tube is adjusted to 100 μl with 120 mM NaCl plus 25 mM Tris at about pH 7.2. A positive (complete cross-linking) control contains fibrinogen, enzyme (FXIIIa or tTG) , calcium, and thrombin at final assay concentrations listed above, and is tested in the absence of any inhibitor. A negative (no cross-linking) control contains a solution of FXIII- free fibrinogen, thrombin and 20 μM EDTA [ (ethylenediamine)tetraacetic acid, disodium salt, dihydrate (J.T. Baker, Phillipsburg, NJ) ] in the absence of calcium and enzyme. Both positive and negative controls contain a final DMSO concentration of 1%. The FXIIIa-treated clots and their controls are incubated for about 10 min at about 37°C, while the tTG-treated clots and their controls are incubated for about 4 h at about 37°C. After incubation, all of the clots and controls are frozen to quench reactions.

Frozen clots and controls are boiled for about 3 min, centrifuged at 14,000 RPM for 5 min, and then the supernatant is removed. The clots are resuspended in 100 μl lysis buffer and incubated overnight at 37°C. The lysis buffer contains 10 M urea, 10 mM EDTA, 1% sodium dodecyl sulfate ("SDS"), and 1% 2-mercaptoethanol. After incubation, approximately 5 μl of each sample is brought to a final volume of 20 μl with a reducing sample buffer (containing 125 mM Tris, pH 6.8, 20% glycerol, 4% SDS, 10% 2-mercaptoethanol, and 0.025% bromophenol blue), and applied to 4-15% gradient gels (Integrated Separation Systems, Natick, MA) for polyacrylamide electrophoresis (PAGE) at 200 volts for approximately 1 h or until the tracking dye reached the bottom of the gel.

After staining with Coomassie blue, the alpha, beta, and gamma chains of fibrin are visualized as bands on the gel. The presence of FXIIIa-catalyzed fibrin cross-linking is observed by the appearance of a gamma- gamma dimer, and the disappearance of the monomeric gamma band from the gels. The tTG cross-linking occurs by formation of alpha chain polymers and alpha2-gamma oligomers. Since there are often many bands, they are difficult to visualize. A better indication of tTG cross-linking may be the disappearance of the monomeric alpha band from the gels.

The gels are also scanned on a densitometer (Bio Image Whole Band Analyzer, Millipore Corp., Bedford, MA) . The relative intensity of gamma-gamma bands are scanned and compared for FXIIIa inhibition, while alpha bands are scanned and compared for tTG inhibition. No cross-linking is expected in the negative controls containing only FXIII-free fibrinogen, thrombin, and DMSO. Complete cross-linking is expected in positive controls containing fibrinogen, enzyme (FXIIIa or tTG) , calcium, and thrombin in the absence of inhibitor.

FVIIa Inhibition Assay . For the FVIIa, FXa and thrombin inhibition assays described below, an inhibition assay buffer containing 50 mM Tris, 5 mM CaCl2 150 mM NaCl, 0.1% Tween 20, and 0.1% BSA (albumin, bovine, Sigma fraction V (Sigma)) at about pH 8.0 is usecT. For each of the inhibition assays, the total reaction volume is 100 μl and the assays are performed in 96-well plates.

For the FVIIa inhibition assay, about 30 μl of inhibition assay buffer, 25 μl of 20 nM recombinant FVIIa (Walt Kiesel, University of New Mexico), and 10 μl of one of a series of dilutions of test sample are added to plate wells. A negative (no inhibition) control contains 30 μl inhibition assay buffer and 25 μl of 20 nM FVIIa.

The test and control wells are incubated for about 10 min at room temperature. After incubation, 25 μl of 20 nM tissue factor (Walt Kiesel, University of New Mexico) is added to the wells and incubated for a second 10 min period at room temperature. After the second incubation period, 10 μl of 10 mM H-D-isoleucyl-L- arginine-p-nitroanilide-dihydrochloride (S2288 sub- strate, Kabi, Stockholm, Sweden) is added to each well. Inhibition of FVIIa activity is determined by reading absorbance at 405 nm on a ThermoMax microplate reader.

Factor TX Inhibition Assay 60 μl buffer, 10 μl of 40 nM FIX, and 10 μl of one of a series of dilutions of test sample are added to plate wells. The buffer contains 150 mM NaCl, 5 mM CaCl2 25 mM Tris, pH 8.0, and 0.1% polyethylene glycol (PEG) . A negative (no inhibition) control contains 60 μl buffer and 10 μl of 40 nM FIX.

The test and control wells are incubated approximately 10 min at room temperature. After incubation, approximately 10 μl of a mixture containing 200 nM FX (Walt Kiesel, University of New Mexico), 500 pM FVIII (Novo Nordisk) and 10 μM PCPS (phosphatidylcholine : phosphatidylserine (Sigma) in a 1:1 ratio) is added to each well. Then 10 μl of 1 mM S2765 substrate (Kabi) is added to each well.

In this assay, FIX converts FX to FXa. Inhibition of FIX activity is determined by reading absorbance at 405 nM on a ThermoMax microplate reader.

Factor Xa Inhibition Assay

About 30 μl inhibition assay buffer, 50 μl of 10 nM FXa and 10 μl of one of a series of dilutions of test sample are added to plate wells. A negative (no inhibition) control contains 30 μl inhibition assay buffer and 50 μl of 10 nM FXa.

The test and control wells are incubated for about 10 min at room temperature. After the incubation period, 10 mM N-a-benzyloxycarbonyl-D-arginine-L-glycyl- L-arginyl-p-nitroanilide-dihydrochloride (S2765 substrate, Kabi) is added to each well. Inhibition of FXa activity is determined by reading absorbance at 405 nM on a ThermoMax microplate reader. Thrombin Inhibition Assay

About 30 μl inhibition assay buffer, 50 μl of 2 μg/ml thrombin (bovine, Enzyme Research Labs) , and 10 μl of one of a series of dilutions of test sample are added to plate wells. A negative (no inhibition) control con¬ tains 30 μl inhibition assay buffer and 50 μl of 2 μg/ml thrombin.

The test and control wells are incubated for about 10 min at room temperature. After the incubation period, 20 μl of 1 mM SPECTROZYME TH (H-D-hexahydro- tyrosyl-L-alanyl-L-arginine-p-nitroanilide-diacetate salt, American Diagnostica, Inc., New York, NY) is added to each well. Inhibition of thrombin activity is determined by reading absorbance at 405 nM on a ThermoMax microplate reader.

Plasma Clot Lysis Assay

For this assay, 50 μl of one of a series of concentrations of iodoacetic acid, cerulenin, or test sample is added to 0.5 ml human citrated plasma (prepared from one blood donor) in 5 ml polypropylene tubes. Approximately IO⁶ cpm ¹²⁵I-fibrinogen (Amersham) , 50 mM CaCl2 (20 mM above citrate concentration) and 5 units thrombin (bovine, Enzyme Research Labs) are added to each test and control tube. The tubes are incubated while rocking for 20 h at 37°C. Plasma clots formed in the tubes are washed three times with about 1 ml phosphate buffered saline (PBS) . The plasma clots in each tube are counted after the PBS washes. Approximately 2 ml of 50 mM Tris, pH 7.4, containing 50 mM NaCl, 28 μg plasminogen (human, Calbiochem) , and 280 ng urokinase (human, Calbiochem) are added to the clots in test and control tubes. At 1, 2, and 4 h intervals, 25 μl aliquots of the samples are removed from each tube and counted. Papain Inhibition Assay

For the papain inhibition assay, a buffer containing 0.1 M NaP0₄ and 2.5 mM EDTA at about pH 7.0 is used. An Azocoll™ solution is prepared by adding 5 mg Azocoll™ (Calbiochem; an insoluble, powdered cowhide to which a bright red dye is attached (azo-dye-bound collagen) ) to the buffer while stirring. The Azocoll is prewarmed to about 37°C. 10 μl of 0.1 M L-cysteine, 10 μ 1 of 50 μg/ml papain (Boehringer Mannheim), 0.5 ml of Azocoll solution, and 50 μl of one of a series of concentrations of test sample, iodoacetamide, iodoacetic acid, or cerulenin are added to duplicate 13 x 75 mm glass tubes. The test and control tubes are incubated for 30 min at 37°C in an agitated water bath. After incubation, all of the tubes are centrifuged at 2000 RPM for 5 min. The supernatant is removed and its absorbance is read at 520 nm on a Beckman DU640 spectrophotometer to determine the amount of papain inhibition.

From the foregoing, it will be appreciated that, although specific embodiments of the invention have been described herein for purposes of illustration, various modifications may be made without deviating from the spirit and scope of the invention. Accordingly, the invention is not limited except as by the appended claims.

Claims

Claims

1. A method for inhibiting Factor XIII activity, comprising: providing a ligand that forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue of a Factor XIII monomer, wherein the amino acid residue is selected from the group consisting of Cys-314, His-373 and Asp-396; and combining Factor XIII and the ligand, thereby inhibiting the activity of Factor XIII.

2. The method of Claim 1, wherein the step of combining Factor XIII and the ligand is performed in vi tro .

3. The method of Claim 1, wherein the ligand additionally forms a contact, at a distance of about 5 A or less, with at least one amino acid residue selected from the group consisting of His-342, Glu-434, Asp-343 and Trp-279.

4. The method of Claim 1, wherein the ligand is a peptide that is selected to correspond to a portion of an inhibitory Greenberg peptide or to a portion of a Factor XIII core and/or sandwich domain that contributes to the Factor XIII β-sandwich:core interface.

5. The method of Claim 1, wherein the ligand is a peptidomimetic or non-peptide molecule having optimal complementarity with a binding site of Factor XIII.

6. The method of Claim 5, wherein the ligand forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue selected from the group consisting of Trp-279, Tyr-372 and Gln-313.

7. The method of Claim 5, wherein the ligand has an electrophilic moiety susceptible to nucleophilic displacement by Cys-314.

8. A method for inhibiting Factor XIII activity, comprising: providing a ligand that forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue of a Factor XIII monomer, wherein the amino acid residue is selected from the group consisting of Arg-77, Arg-78 and Asp-191; and combining Factor XIII and the ligand, thereby inhibiting the activity of Factor XIII.

9. The method of Claim 8, wherein the ligand additionally forms a contact, at a distance of about 5 A or less, with at least one amino acid residue selected from the group consisting of Tyr-69, Cys-327, Gln-80, Asn-185, Try- 181 and Asp-236.

10. A method for inhibiting Factor XIII activity, comprising: providing a ligand that forms at least one contact, at a distance of about 5 A or less, with at least one amino acid residue of a Factor XIII monomer, wherein the amino acid residue is selected from the group consisting of Lys-257, Glu-401, Lys-113, Asp-367, Arg-260 and Asp-404; and combining Factor XIII and the ligand, thereby inhibiting the activity of Factor XIII.