Interaction of the Factor XIII Activation Peptide with α-Thrombin

The serine protease thrombin proteolytically activates blood coagulation factor XIII by cleavage at residue Arg37; factor XIII in turn cross-links fibrin molecules and gives mechanical stability to the blood clot. The 2.0-Å resolution x-ray crystal structure of human α-thrombin bound to the factor XIII-(28–37) decapeptide has been determined. This structure reveals the detailed atomic level interactions between the factor XIII activation peptide and thrombin and provides the first high resolution view of this functionally important part of the factor XIII molecule. A comparison of this structure with the crystal structure of fibrinopeptide A complexed with thrombin highlights several important determinants of thrombin substrate interaction. First, the P1 and P2 residues must be compatible with the geometry and chemistry of the S1 and S2 specificity sites in thrombin. Second, a glycine in the P5 position is necessary for the conserved substrate conformation seen in both factor XIII-(28–37) and fibrinopeptide A. Finally, the hydrophobic residues, which occupy the aryl binding site of thrombin determine the substrate conformation further away from the catalytic residues. In the case of factor XIII-(28–37), the aryl binding site is shared by hydrophobic residues P4 (Val34) and P9 (Val29). A bulkier residue in either of these sites may alter the substrate peptide conformation.

The serine protease thrombin plays a central role in the blood coagulation process (1,2). It proteolytically activates blood coagulation and plasma factors such as factor V, factor VIII, factor XIII, and protein C (3). Its proteolytic activity is also responsible for catalyzing the conversion of fibrinogen to fibrin (4) by the cleavage of fibrinopeptides A and B from the N termini of the fibrinogen ␣ and ␤ chains, respectively. In addition, thrombin is the most potent stimulator of platelet aggregation (5). Thrombin achieves this diverse yet specific recognition of substrates with a deep active site cleft and by exploiting an apolar binding site near the catalytic residues as well as an anion binding exosite distant from the active site cleft (6).
Crystal structures of ␣-thrombin complexed to a number of natural and synthetic inhibitors have been studied in detail and characterized to near atomic resolution (for example, see Refs. [7][8][9][10][11][12][13][14]. These structures have provided a wealth of information about the interactions between inhibitors and enzyme. However, more limited knowledge about the atomic level interactions between substrate and thrombin comes mainly from the crystal structures of thrombin bound to fibrinopeptide A. Structures of human and bovine ␣-thrombin bound to the Nterminal peptide of the fibrinogen ␣-chain (15,16) show that the C-terminal region of the peptide runs anti-parallel to the Ser 214 -Glu 217 segment of thrombin and that the arginine at the cleavage site occupies the S1 specificity pocket. The N-terminal region of the peptide adopts a compact ␣-helical conformation, folding back toward the active site cleft to bring the hydrophobic residues Leu 9 and Phe 8 to the apolar binding site (15,16). In the present study a 2.0-Å resolution crystal structure of the factor XIII activation peptide, encompassing residues 28 -37, complexed to human ␣-thrombin, sheds light on further details of thrombin-substrate interactions. This work describes the second unique structure of a thrombin-substrate complex, reveals conserved structural features as well as conformational differences in substrate and enzyme, and extends understanding of thrombin-substrate interactions.
Factor XIII is a transglutaminase and the last enzyme to be activated in the blood coagulation cascade. The zymogen circulates in human plasma as a heterotetramer of two catalytic A-subunits and two carrier B-subunits and is found intracellularly in platelets, megakaryocytes, and macrophages as an A 2 dimer (17)(18)(19)(20). Factor XIII is activated by thrombin, which cleaves the A-subunit between Arg 37 and Gly 38 , releasing the N-terminal activation peptide in the presence of Ca 2ϩ ions (21,22). The activated enzyme catalyzes the formation of ␥-glutamyl-⑀-lysyl amide cross-links between polypeptide chains in adjacent fibrin molecules, rendering the blood clot mechanically stable and resistant to fibrinolysis.
The residues surrounding the thrombin cleavage site in factor XIII are of particular interest. Val 34 has drawn much recent attention, because a common Leu 34 polymorphism has been associated with a lower incidence of myocardial infarction in patients with coronary artery disease (23)(24)(25) and of deep venous thrombosis (26,27). In addition, the Val 34 Leu polymorphism was found to be more common in patients with primary intracerebral hemorrhage in one study (28) and to be less common in patients with brain infarction in another (29). Recombinant and purified factor XIII Leu 34 variant has a significantly higher specific activity than the Val form (30,31) and is activated more quickly (32,33). The proximity of Val 34 to the thrombin cleavage site at Arg 37 suggests that the Leu 34 substitution may affect thrombin cleavage and be responsible thus for the difference in activation and specific activity (33,34).
Crystal structures of factor XIII A 2 in zymogen, thrombincleaved, and Ca 2ϩ -bound forms have been determined (35)(36)(37). In all structures, the 37-residue N-terminal activation peptide remains in the same position and conformation with respect to the rest of the molecule, even after thrombin cleavage or cation binding. As well, in all structures the activation peptide near the Arg 37 cleavage site is highly disordered and poorly defined in the electron density maps. The present crystal structure of the factor XIII-(28 -37) decapeptide bound to human ␣-thrombin reveals the detailed atomic level interactions between the activation peptide and thrombin and provides the first high resolution view of this functionally important part of the factor XIII molecule.

EXPERIMENTAL PROCEDURES
Crystallization-Human ␣-thrombin (Enzyme Research Laboratories, Inc., South Bend, IN) and the factor XIII (fXIII 1 ) peptide as a chloromethylketone derivative (Thr 28 -Val 29 -Glu 30 -Leu 31 -Gln 32 -Gly 33 -Val 34 -Val 35 -Pro 36 -Arg 37 -CMK, AnaSpec, Inc., San Jose, CA) were obtained from commercial sources. The enzyme was concentrated to 8 mg/ml and mixed with a 10-fold molar excess of peptide dissolved in water. Crystals of the thrombin-peptide complex were grown by hanging drop vapor diffusion at room temperature, from 24% polyethylene glycol 8000 and 200 mM sodium chloride in 50 mM sodium citrate buffer at pH 5.5. Typically crystals grow to full size in about 2 weeks, to maximum dimensions of 0.1 ϫ 0.2 ϫ 0.4 mm.
Data Collection and Processing-A crystal stabilized in a cryoprotectant containing 15% glycerol was cooled in a stream of cold nitrogen (Ϫ180°C) for data collection. Diffraction data were collected on an in-house Rigaku R-AXIS IV imaging plate system using CuK␣ radiation ( ϭ 1.5418 Å) from a Rigaku H3R rotating anode x-ray generator operated at 50 kV, 100 mA and equipped with Yale focusing mirrors. A total of 127 images was collected with 60-min exposure, 1.5°oscillation step, and a crystal-detector distance of 100 mm. Data to 2.0-Å resolution were processed using the HKL program suite (38); data processing statistics are listed in Table I. The unit cell parameters are a ϭ 54.07, b ϭ 81.26, c ϭ 85.46 Å, and ␤ ϭ 101.62°with space group P2 1 . The calculated Matthews number of 2.7 (39) corresponds to a solvent content of 55% and Z ϭ 4, indicating that there are two independent molecules in the asymmetric unit.
Structure Solution and Refinement-The structure was solved by molecular replacement (40) using the program AMoRe (41), with human ␣-thrombin coordinates from the 1.6-Å resolution crystal structure of its complex with hirugen and p-amidinophenylpyruvate, and Protein Data Bank code 1AHT (10) as the search model. The cross rotation calculation showed two peaks with correlation coefficients 34.5% and 28.3%, respectively, which were significantly higher than the rest. Translation function calculations for these peaks gave two clear solutions and the overall correlation coefficient increased to 61.3% with an R factor of 37.0%. Rigid body refinement gave a correlation coefficient and R factor of 69.9% and 33.4%, respectively, for 10.0-to 3.5-Å resolution data and indicated the solution was correct. A rigid body refinement with the two molecules in the asymmetric unit treated as independent rigid groups was then carried out using X-PLOR (42). The model was refined further by the gradual extension of data to 2.0-Å resolution, alternating with manual model fitting to the 2͉F o ͉ Ϫ ͉F c ͉ and ͉F o ͉ Ϫ ͉F c ͉ electron density maps using the program O (43). A molecular dynamics refinement using the simulated annealing technique was also carried out. The fXIII-(28 -37) residues initially built into 2͉F o ͉ Ϫ ͉F c ͉ and ͉F o ͉ Ϫ ͉F c ͉ omit electron density contoured at 1.0 and 2.5 , respectively ( Fig. 1), before being included in the refinement. Water molecules were selected from ͉F o ͉ Ϫ ͉F c ͉ peaks higher than 4; the sodium ions were selected from positive ͉F o ͉ Ϫ ͉F c ͉ peaks, which corresponded to the binding site characterized elsewhere (44,45). These positions were checked after each cycle of refinement and retained only if the electron density remained clear in the 2͉F o ͉ Ϫ ͉F c ͉ map. The final R and free R values are 19.3% and 25.6%, respectively, for 10-to 2.0-Å resolution data. The model quality was checked using PROCHECK (46) and DDQ (47). All the main-chain torsion angles for non-glycine and non-proline residues are within the allowed regions of the Ramachandran plot (48). The refinement and geometric parameters are given in Table I.

RESULTS
The asymmetric unit consists of two thrombin molecules (MOL1 and MOL2), two peptide molecules (PEP1 and PEP2), 479 waters, and 2 sodium ions. The residue numbering in thrombin is based on that of chymotrypsin (7). Peptide residues are numbered 28 to 37 to conform to the intact factor XIII sequence. Terminal residues 1H-1B and 14L-15 in the thrombin light chain and residues 245-247 and 148 -149E in the heavy chain lack continuous and well defined electron density, and were not included in the model. Well defined electron density was observed for PEP1 ( Fig. 1) and the C-terminal portion of PEP2. However the electron density for N-terminal residues 28 -30 of PEP2 was found be relatively weak, as reflected in high B factors (Table I).
Thrombin Structure-The structures of the two thrombin molecules observed in this complex crystal are quite similar to other thrombin crystal structures reported in the literature. The r.m.s. deviation for the C␣ atoms of the first and second molecules from those of the initial search model are 0.59 and 0.46 Å, respectively, whereas the r.m.s. deviation between the two molecules in the asymmetric unit is 0.51 Å. The two protein molecules in the asymmetric unit of this crystal structure participate in substantially different crystal packing interactions. with its 6 surrounding protein molecules, whereas only 175 such contacts with 4 surrounding protein molecules are observed for MOL2. The tight packing of MOL1 compared with MOL2 in the crystal structure is also reflected in the different average B factors for the two molecules of 22.6 and 35.5 Å 2 , respectively (Table I).
Peptide Structure-The fXIII-(28 -37) peptides PEP1 and PEP2 are bound to thrombin molecules MOL1 and MOL2, respectively. The two peptides have roughly the same backbone conformation: the r.m.s. deviation of the ten C␣ atoms after superposition of the two peptides is 0.75 Å. Side-chain atoms of N-terminal residues show considerable conformational differences between the two peptides, as evidenced by a high r.m.s. deviation for all atoms between the two peptides of 1.80 Å. The C-terminal segment Val 34 -Arg 37 of the peptides adopts an extended ␤-sheet arrangement with main-chain hydrogen bonds with thrombin and runs anti-parallel to the Ser 214 -Glu 217 seg-ment of the enzyme (Fig. 2). The peptide N-terminal residues Thr 28 to Gln 32 form a short stretch of irregular ␣-helix with main-chain hydrogen bonds between residues Thr 28 -Leu 31 , (Thr 28 -Gln 32 in the case of PEP2), Val 29 -Gln 32 , and Glu 30 -Gly 33 . The N-terminal segment folds back toward the active site cleft and Gly 33 facilitates this folding, bringing Val 29 and Val 34 , rather distant along the peptide chain, close enough to form main-chain hydrogen bonds. The hydrophobic side chains of both these valine residues are in close proximity and point in the same direction. The side chains of residues Thr 28 , Glu 30 , and Leu 31 are clustered together and point in the opposite direction. This arrangement gives a compact structure for the segment Thr 28 -Val 34 (Fig. 1).
Thrombin-Peptide Interactions- Table II lists the hydrogen bond interactions between the peptide and thrombin in each complex; MOL1-PEP1 interactions are illustrated in Fig. 2. Arg 37 makes a number of hydrogen bonds with the thrombin residues constituting the active site cleft. It forms main-chain hydrogen bonds with thrombin residues Ser 214 , Gly 193 , and Ser 195 ; its side chain participates in hydrogen bonds with the carbonyl oxygen of Gly 219 and side-chain oxygen atoms of Asp 189 . Glu 192 of MOL2 swings in to interact with both Arg 37 and Pro 36 in PEP2. The peptide structure is stabilized also by two water hydrogen bonds in the specificity pocket. These hydrogen bonds are made by N-⑀ and N-H1 of Arg 37 . Other interactions include Val 35 in the peptide forming main-chain hydrogen bonds with Gly 216 of thrombin, and the terminal nitrogen of Thr 28 hydrogen bonding to thrombin's carbonyl oxygen of Arg 97 . A large number of van der Waals interactions (Յ4.0 Å) also are observed between the peptide and the enzyme molecule, as schematically represented in Fig. 3.
Significant differences in crystal contacts involving the Nterminal regions of the two fXIII-(28 -37) peptides are observed. In PEP1, the peptide nitrogen of Thr 28 and terminal nitrogen of Gln 32 hydrogen bond to O-⑀1 of Gln 38 in MOL2; Gln 32 also hydrogen bonds to the carbonyl oxygen of Lys 36 in MOL2. In addition, a number of van der Waals interactions are observed between PEP1 and MOL2. Thr 28 , Glu 30 , Leu 31 , and Gln 32 in PEP1 make contacts with Lys 36 , Gln 38 , Leu 65 , Arg 67 , Tyr 76 , and Ile 82 in MOL2. In contrast, no intermolecular interactions are observed between PEP2 and MOL1. The lack of these stabilizing crystal contacts partially explains the high B factors for PEP2. It should be noted that MOL2 also has a higher average B factor than MOL1.
The carbonyl carbon atom of fXIII-(28 -37)'s P1 residue Arg 37 makes a hemiketal bond with the active site residue Ser 195 O-␥, and is also linked to His 57 N-⑀2 through a methylene, because the original peptide was derivatized with a reactive chloromethylketone group. The Arg 37 side chain occupies the S1 2 spec-   29 . Solvent Structure-The final structure contains 489 water molecules and two sodium ions. Each sodium ion coordinates with carbonyl oxygen atoms of Arg 221A and Lys 224 in a thrombin molecule. The waters which surround each sodium ion are part of a solvent cluster that occupies an internal cavity enclosed by residues Tyr 184A -Gly 188 , Asp 221 -Tyr 224 , and Asp 189 , as reported earlier (44,45). This cluster and its linkage to the bulk water has previously been suggested as a route to expel water molecules through the back of the specificity pocket upon insertion of a P1 side chain during substrate/inhibitor binding. Among the many other water clusters also observed in the current structure is a large clustering of solvent molecules between the two thrombin molecules in the asymmetric unit. Many of these water molecules are hydrogen bound to exosite II residues Arg 126 , Asp 178 , and Arg 233 of MOL1 and Lys 135 , Glu 164 , Asp 186A , Glu 186B , and Lys 186D of MOL2. DISCUSSION The structure of this thrombin-fXIII-(28 -37) complex can be compared with that of the only other substrate complexes studied, those of fibrinopeptide A-(7-16) (FPA) bound to ␣-thrombin (15) (Fig. 3). The fXIII- (28 -37) and FPA follow roughly the same backbone conformation (Fig. 4) despite the fact that the sequence identity between the two peptides is only 20%. The r.m.s deviations of the C␣ atoms between FPA and PEP1, and between FPA and PEP2, are 1.00 and 0.57 Å, respectively. Although PEP1 and PEP2 show similar r.m.s deviations from FPA for P1-P5 residues after superposition of this segment (0.47 and 0.52 Å), the PEP1 shows a larger deviation (0.69 Å) than PEP2 (0.33 Å) in the case of N-terminal residues. The difference in similarity between the two fXIII-(28 -37) peptides  and FPA may not be significant given the higher B-factors and likely greater positional uncertainty for the atoms in PEP2.
A comparison of the fXIII-(28 -37) with PPACK bound to thrombin (7) shows that residues P1-P3 superimpose well. This indicates that the C-terminal residues, which are bound to the active site cleft and also involved in the hydrogen bonding with the enzyme, have essentially the identical conformation in fXIII- (28 -37), FPA, and PPACK. The arginine side chain of PPACK fills the specificity pocket, and the proline is encapsulated in a hydrophobic cage. The same mode of interaction is observed between the P1 Arg 37 residue and thrombin in the FPA and fXIII-(28 -37) complexes. The P2 residues, Pro 36 in fXIII- (28 -37) and Val 15 in FPA, occupy the same position as proline in PPACK. Thus the conformations of the C-terminal residues in fXIII- (28 -37) and in FPA are determined mainly by the active site geometry of thrombin, because this region has direct, specific interactions with the enzyme. The N-terminal residues, however, have only weak van der Waals interactions with thrombin and are not engulfed by thrombin molecules to give a tight packing. One probable reason for the observed folding of the substrate peptides is the stability offered by the clustering and burial of hydrophobic residues, rather distant along the peptide chain, in the apolar binding site. In PPACK, the benzyl side chain of D-Phe fills the residual part of the hydrophobic cavity called the aryl binding site (6). The side chain of a substrate's L-amino acid at the P3 position would interact with a different thrombin site as compared with that of PPACK. In the fXIII- (28 -37), the side chain of P3 residue Val 35 points toward the bulk solvent, and the side chain of P4 residue Val 34 occupies the aryl binding site instead. Based on modeling studies, it has been suggested that the side chain of an Lstereoisomer at P3 would extend into bulk solvent due to steric hindrance by the P2 side chain and the Trp 60D indole group of thrombin (7), and probably the side chain of the P4 residue in a polypeptide substrate would point toward the apolar binding site (6). Because the P3 and P4 residues in FPA are glycines, this fXIII-(28 -37) structure provides the first experimental confirmation for this prediction of substrate conformation when bound to thrombin. The aryl binding site in the fXIII- (28 -37) complex is occupied by the aliphatic side chains of P9 residue Val 29 and P4 residue Val 34 and is bordered by the side chains of thrombin residues Tyr 60A , Ile 174 , Trp 215 , and Glu 217 , and main-chain atoms of residues Arg 97 -Asn 98 and Gly 216 . In the case of FPA, the P9 residue Phe 8 occupies almost the same relative position as Val 29 , and the P8 residue Leu 9 is displaced from both the Val 29 and Val 34 positions in fXIII-(28 -37) (Fig.  5). The crystal structure of the thrombin-hirudin complex (8) shows that the phenolic side chain of the inhibitor's Tyr 3 is located in the aryl binding site (Fig. 5). Finally, the naphthyl and tosyl moieties of benzamidine-derived inhibitors, and other aromatic groups in synthetic inhibitors interact in a similar favorable manner with this apolar site (6, 49 -52).
In the present crystal structure, Ile 174 shows side-chain conformational differences and a shift of about 1 Å compared with thrombin in complexes with FPA or PPACK (Fig. 6). This may be due to the steric bulk of the Val 34 side chain in the aryl binding site. In addition, Trp 215 shows a slight shift in its position. Therefore, a bulkier side chain at P4 position may not be a favorable condition for the observed conformation of the substrate peptides. We have modeled a leucine residue in place of Val 34 and have calculated distances to surrounding atoms. This calculation showed that the Leu 34 side chain in various orientations makes unfavorable short contacts with either thrombin or N-terminal segments of the peptide or both. Thrombin residues that are involved in these short contacts are Glu 217 , Ile 174 , Trp 215 , and Gly 216 , and peptide residues involved are Val 29 and Gln 32 . The short contacts between the modeled Leu 34 and Val 29 of the peptide indicate that the Val 34 Leu mutation likely leads to a different peptide conformation. Eventually this may affect properties of factor XIII as a substrate. For example, in the crystal structure of thrombin bound to fibrinopeptide A with phenylalanine replaced by the larger tyrosine at P9, a disordered peptide conformation results (52). The observed peptide conformation is incompatible with nucleophilic attack by the catalytic Ser 195 and explains how the bulkier tyrosine mutation renders the peptide a less susceptible substrate.
Finally, the side-chain arrangement of Glu 30 in fXIII-(28 -37) is different from that of the corresponding Leu 9 in FPA. Although the hydrophobic side chain of leucine in FPA folds toward the apolar binding site, the charged Glu 30 side chain in fXIII-(28 -37) points toward the bulk solvent (Fig. 4). In FPA Glu 11 makes salt bridges with Arg 173 of thrombin. This interaction has been thought of as an important factor for the binding of fibrinogen to thrombin (15,16). In fXIII- (28 -37), the corresponding residue Gln 32 adopts a different side-chain conformation and makes only a few van der Waals interactions with thrombin. The loss of binding energy due to the absence of this peptide-thrombin electrostatic interaction may be at least partially compensated by the dual hydrophobic insertions into the aryl binding site in the fXIII-(28 -37) complex. It has been observed that glycine at P5 position is conserved in fibrinogen, and a non-glycine residue would be unfavorable to the observed peptide folding (15,16). In fXIII-(28 -37) also, the P5 position is occupied by a glycine and adopts a similar main-chain conformation. This confirms an important role for glycine at P5 for substrate conformation and thus for substrate-thrombin interaction.

CONCLUSION
The fXIII-(28 -37)-and FPA-bound thrombin crystal structures, along with those for inhibitor complexes, highlight the several important determinants of thrombin substrate structure. First, the P1 and P2 residues must be compatible with the geometry and chemistry of the S1 and S2 specificity sites in thrombin. Second, a glycine in the P5 position is necessary for the conserved substrate conformation seen in both fXIII- (28 -37) and FPA. Finally, the hydrophobic residues, which occupy the aryl binding site determine the substrate conformation further away from the catalytic residues. In FPA, the P8 residue Leu 9 occupies the aryl binding site and interacts with thrombin's apolar binding site, but a similar interaction was not observed for the corresponding Glu 30 in fXIII- (28 -37). In the case of fibrinopeptide A, a hydrophobic P9 residue is crucial for thrombin cleavage; a bulkier residue in this position disrupts the geometry necessary for nucleophilic attack by Ser 195 and makes the peptide a less favorable substrate (52). In the case of fXIII- (28 -37), the aryl binding site is shared by hydrophobic residues P4 (Val 34 ) and P9 (Val 29 ).
Interestingly, the Leu 34 variant of factor XIII involves a substitution of one of the aryl binding site residues, Val 34 . The Leu 34 variant has a higher specific activity and is more quickly activated by thrombin than the more common Val 34 polymorphism (30 -33). Kinetic studies show that thrombin cleaves the Leu 34 factor XIII variant twice as efficiently as the Val 34 form (33), and with peptide substrates, reveals a lower K m and higher k cat for a factor XIII peptide containing Leu 34 than for a similar Val 34 peptide (34). These observations suggest that the Leu 34 substitution increases thrombin's affinity for its substrate, the catalytic cleavage may be facilitated, and/or the product may be removed from the active site more quickly. Comparison of the crystal structure of thrombin bound to the Leu 34 variant of fXIII-(28 -37) with this reported Val 34 form may bring some valuable insight into which explanation is more likely. Our modeling indicates that a Leu 34 substitution on the fXIII-(28 -37) likely leads to a different peptide conformation. However, modeling cannot accurately predict the details of the altered conformation of the Leu 34 peptide. Preliminary observations with crystals of the factor XIII Leu 34 peptide bound to thrombin show that they have a different morphology, space group symmetry, and diffraction behavior than for the Val 34 peptide complex, 3 and thus support the expectation of a different peptide conformation.