Engineering exosite peptides for complete inhibition of factor VIIa using a protease switch with substrate phage.

Limitations of current anticoagulant therapies have led us to develop two distinct classes of exosite peptide inhibitors for the initiator of the clotting process, the tissue factor-factor VIIa (TF.FVIIa) complex (Roberge, M., Santell, L., Dennis, M. S., Eigenbrot, C., Dwyer, M. A., and Lazarus, R. A. (2001) Biochemistry 40, 9522-9531). Although both peptide classes are potent and selective inhibitors of TF.FVIIa, neither showed 100% inhibition at saturating concentrations. Crystal structures of these peptides in complex with the FVII/FVIIa protease domain revealed their distinct binding sites and close proximity to the active site. The favorable orientation of the 15-mer A-site peptide A-183 (EEWEVLCWTWETCER) suggested that a C-terminal extension into the FVIIa active site could yield a chimeric inhibitor that was not only potent and selective but complete as well. A novel two-step "protease switch" approach using substrate phage display was developed by first binding all phage containing A-183 and C-terminal extension libraries to immobilized and inactive FVIIa. Upon altering pH and adding TF to switch on FVIIa enzymatic activity, only those phage released by proteolytic cleavage within the extension were propagated. This process selected for both preferred sequence and length in the extension, leading to a 27-mer peptide A-183X (EEWEVLCWTWETCERGEGVEEELWEWR) with a C-terminal 12-mer extension containing an Arg in the P1 position. A-183X was a more potent and complete inhibitor of FX activation, having a maximal extent of inhibition of approximately 99% with an IC50 of 230 pm versus A-183 which maximally inhibited to 74% with an IC50 of 1.5 nm. A-183X also had a maximal prolongation of the prothrombin time of 7.6- versus 1.9-fold for A-183, making it a more effective anticoagulant.

The limitations of current anticoagulant therapies have stimulated the search for new alternatives based on selective inhibition of serine proteases in the coagulation pathway (1)(2)(3)(4). Commonly used anticoagulants, such as coumarin and heparin, lack specificity and have a narrow therapeutic window, leading to undesirable side effects such as bleeding. The tightly regulated coagulation cascade consists of several highly homologous serine proteases and their cofactors and inhibitors (5,6). The development of a specific inhibitor for a single coagulation factor could reduce side effects and improve the therapeutic profile. The architecture of the active site in all these proteases is very similar, which makes the development of a specific small molecule active site inhibitor challenging. Although relatively nonselective with respect to small chromogenic substrates, these proteases are highly specific for their natural macromolecular substrates. In order to achieve this, exosites on these enzymes play an important role in substrate recognition and catalysis (7)(8)(9)(10)(11)(12). Blocking such important interactions could result in the specific inhibition of a single protease in this pathway.
Factor VIIa (FVIIa) 1 is the initiator of the extrinsic blood coagulation pathway. Upon vascular damage, zymogen FVII binds to tissue factor (TF), a membrane-bound protein cofactor that is expressed in the surrounding tissue (13). This results in the formation of the active complex TF⅐FVIIa, which activates FX to FXa, FIX to FIXa, and FVII to FVIIa through specific proteolytic cleavages. Ultimately, this cascade leads to the generation of thrombin, resulting in the formation of fibrin and eventually a blood clot. Because FVII/FVIIa is present only in relatively low concentrations in blood (10 nM) and it initiates the coagulation cascade (13), this enzyme is an attractive drug target for the development of anticoagulants.
Recent focus on the search for specific inhibitors has lead to the development of two new classes of exosite peptide inhibitors of human FVIIa (14 -16). These cyclic peptide classes, termed E-series and A-series, are typified by the 18-mer E-76 and the 15-mer A-183 (EEWEVLCWTWETCER), respectively. They bind tightly to two distinct exosites on the protease domain of FVIIa and potently inhibit the activation of FX to FXa by TF⅐FVIIa with IC 50 values in the single digit nanomolar range. Similar potency was observed for the inhibition of the amidolytic activity of TF⅐FVIIa. Although these peptides were potent inhibitors of TF⅐FVIIa, their inhibition was incomplete. At saturating concentrations, E-76 and A-183 showed a maximal extent of inhibition of FX activation of 90 and 78%, respectively, whereas the maximal extent of inhibition of the amidolytic activity was 50 and 32%, respectively. The partial inhibition by these exosite peptides has been explained by their mechanisms of inhibition involving both allosteric effects on the active site of FVIIa as well as steric hindrance with the substrate FX (14 -16).
The extent of inhibition of the A-series peptides correlated with the degree of anticoagulant activity determined by prolongation of the TF-dependent prothrombin time (PT) (15). * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: Dept. of Protein Engineering, Genentech, Inc., 1  Whereas partial inhibition and moderate PT prolongation has some potential advantages, more complete inhibition may be required to achieve effective anticoagulation in vivo. We recently pursued one strategy to obtain more complete inhibition by linking the A-and E-series peptides together to make fusion peptides (17). However, these bifunctional fusion peptides were relatively large and required 2 disulfide bonds. Here we have pursued a new strategy to obtain less complex peptide exosite inhibitors that are both specific and complete. The crystal structure of the protease domain of FVII in complex with A-183 revealed that its C terminus was in close proximity to the active site (16,18). This suggested that extending its C terminus into the active site region should result in a chimeric peptide having a high degree of specificity and potency due to its exosite interactions and more complete inhibition due to greater steric hindrance in the substrate binding cleft of the active site. Furthermore, such a chimeric peptide inhibitor could also have a higher affinity due to a more extensive binding surface.
In order to determine both the preferred length and sequence for such an extension, we developed a novel strategy using substrate phage display. Substrate phage display has been successfully employed to determine preferred cleavage sequences for a variety of proteases (19 -29). The substrate phage library is displayed between the p3 coat protein of the phage and a protein or peptide affinity tag. Following addition of protease to the phage library, only phage containing preferred sequences are cleaved from the tag. Libraries immobilized prior to proteolysis release phage directly into the supernatant, whereas libraries in solution require capture of uncleaved phage after proteolysis. In either case, phage in the supernatant are propagated, and preferred substrate sequences are determined by sequencing of selected phage.
In this report, we describe a new substrate phage approach in the context of the peptide exosite inhibitor A-183 and FVIIa, which serves as both the binding target and the protease. A-183 was used as the anchor to bring the substrate library in close proximity to the active site of FVIIa under conditions where the enzyme is not active (Fig. 1). Upon activation of FVIIa by addition of its obligate cofactor TF and modification of the pH, only phage with suitable sequences for enzymatic cleavage by TF⅐FVIIa are released and propagated. Employing this strategy, we panned 4 initial libraries and one consensus library to ultimately obtain a 27-residue peptide, which completely (ϳ99%) inhibited FX activation with enhanced affinity and improved the TF-dependent anticoagulant profile.

EXPERIMENTAL PROCEDURES
Molecular Modeling-A plausible and short route from the C terminus of A-183 to an Arg in the S 1 subsite 2 of FVIIa was sketched out in Xfit (30). This was carried out by first combining two halves of the protease domain, one from the A-183⅐FVII zymogen structure (16,18) and the other from the TF⅐FVIIa structure (31). The rationale for this derives from the protease architecture (two distorted ␤-barrels) and the largely independent effects of A-183 in one barrel and the presentation of the S 1 subsite in the other barrel. Despite the conceptual simplicity of this approach, the resulting hybrid structure is very likely to faithfully represent the salient features of the A-183⅐FVIIa structure.
Substrate Phage Library Design-A His 6 -A183-(protease-resistant spacer)-p3 fusion was constructed for monovalent display of an A-183 extension peptide library on filamentous phage. The previously described clone AD (15) was used as a template for Kunkel mutagenesis (32) to create a template phagemid containing DNA encoding His 6 , peptide A-183, 3 stop codons, and the protease-resistant sequence TPT-DPPTTPPT (33), which was fused to the N terminus of g3. The following primers were used: 5Ј-TGC TGG ACG TGG GAG ACC TGC GAA CGT GGT GAA GGT CAG TAA TAA TAA ACC CCG ACC GAT CCG CCG ACC ACC CCG CCG ACC GAT TTT GAT TAT-3Ј and 5Ј-ACA AAT GCC TAT GCA CAT CAC CAT CAC CAT CAC TCC GAA GAG TGG GAG-3Ј.

FIG. 1. Phage selection strategy.
Substrate phage libraries in binding buffer, pH 6.0, were incubated with immobilized FVIIa for 1 h (step 1). Unbound phage were removed by repetitive washing with binding buffer (step 2). Bound phage were immediately incubated for 5 min with cleavage buffer (sTF, pH 8.5) (step 3). Supernatants were incubated with immobilized anti-His 4 mAb to pull out any uncleaved phage (step 4). Supernatants were then subjected to propagation in E. coli and further rounds of selection (step 5).
were propagated in XL1-Blue cells with VCSM13 helper phage (Stratagene, La Jolla, CA). After several rounds of panning, individual clones from each library were picked and sequenced. Enrichment was monitored by titering the number of phage released by FVIIa in cleavage buffer compared with phage released by FVIIa in binding buffer.
Statistical Analysis of Sequence Data-The percent occurrence of each amino acid at each position was calculated after normalization for codon bias. The significance of the standard deviation was calculated essentially as described (35).
Construction and Panning of a Partially Randomized Consensus Sequence-Based on the preferred residues found by panning libraries A and D, a consensus library was designed for monovalent display of His 6 -A183-Gly 1 -Glu 2 -X 3 -X 4 -X 5 -X 6 -Glu 7 -X 8 -X 9 -Glu 10 -Trp 11 -Arg 12 -Gly 13 -X 14 -X 15 -Gly 16 -Gly 17 -(protease-resistant spacer)-p3 random peptides. Codons GDG, GDC, GAS, GRG were used for positions X 3 , X 4 , X 5 , and X 6 , respectively, and codon NNS was used for positions X 8 , X 9 , X 14 , X 15 , respectively, where D stands for nucleotides A, G or T; N, S, and R nucleotides are defined above. This consensus library was sorted as described above, and individual clones were sequenced.
Binding and Amidolytic Activity of Immobilized FVIIa-MaxiSorp plates were coated with 100 l of FVIIa at different concentrations (1, 5, 10, and 20 g/ml) in 50 mM carbonate buffer, pH 9.6, at 4°C overnight. Amidolytic activity was monitored at 405 nm over 5 min with 0.5 mM Chromozym t-PA in 50 mM MES, pH 6.0, or 50 mM Tris, pH 8.5, containing 5 mM CaCl 2 , 100 mM NaCl, and 0.1% BSA in the absence and presence of a 10-fold molar excess of sTF over FVIIa in a total volume of 100 l. Binding of N-terminal biotinylated A-183 to immobilized FVIIa was carried out at various pH values, ranging from pH 4 to 11 as described previously (16).
A second substrate (A-183X-IVGGSGG-Z) was made as above using the 12-mer extension consensus sequence derived from substrate phage followed by the 7-residue sequence IVGGSGG. A-183X-IVGGSGG-Z (EEWEVLCWTWETCERGEGVEEELWEWRIVGGSGG-Z) was expressed and purified as described below for the Z-A-183 extension peptides. A-183X-IVGGSGG-Z (25 g) was incubated in 50 mM MES, pH 6.0, or 50 mM Tris, pH 8.5, and 5 mM CaCl 2 , each containing 150 mM NaCl for 20 h at room temperature at a 20:1:10 molar ratio of A-183X-IVGGSGG-Z/FVIIa/sTF. The final reaction volume was 50 l. The reaction mixture was analyzed by MALDI-TOF mass spectrometry and non-reducing SDS-PAGE using a 10 -20% NOVEX gel.
Z-A-183 Extension Peptides-DNA sequence encoding a linker (GGS-GGDDDDK), A-183, and a stop codon were inserted into the vector pZCT (37) after the C terminus of the Z-domain using Kunkel mutagenesis (32) to make an expression plasmid for Z-A-183 (pZCT-A-183), which was confirmed by DNA sequencing. The extension plasmids pZCT-A-183X, pZCT-A-183-[GGS] 3 GGR, pZCT-A-183-[GGS] 4 , pZCT-A-183X-R12A, and pZCT-A-183X-⌬R12 were constructed by Kunkel mutagenesis using pZCT-A183 as template by inserting DNA sequence encoding GEGVEEELWEWR, GGSGGSGGSGGR, GGSGGSGGSGGS, GEGVEEELWEWA, and GEGVEEELWEW, respectively, between the C terminus of A-183 and the stop codon. All constructs were transformed into E. coli strain 27C7 and grown in low phosphate minimal media (15). All Z-fusion peptides were secreted into the media and purified using IgG-Sepharose (Amersham Biosciences) as described previously (38) followed by size exclusion on Superdex 75 (Amersham Biosciences) at 0.5 ml/min flow rate using Tris buffer, pH 7.5, and 200 mM NaCl. Fractions from gel filtration containing monomeric Z-fusion peptides were loaded on a 1-ml Resource Q column (Amersham Biosciences) for concentration and final purification by ion exchange chromatography, using a salt gradient from 0.2 to 0.8 M NaCl in 20 mM Tris, pH 7.5, over 12 column volumes at a flow rate of 3 ml/min.
To obtain A-183X, we replaced the 10-residue linker above preceding the A-183X sequence with one having a FXa cleavage site (GGSG-GIEGR) in pZCT-A-183X. A-183X was cleaved from Z-A-183X by overnight incubation at room temperature with bovine FXa in a 50:1 molar ratio (Z-A-183X/FXa). A-183X was then purified by size exclusion, using PBS containing 300 mM NaCl at 0.5 ml/min on a peptide Superdex column (Amersham Biosciences). Peaks were collected and analyzed by SDS-PAGE, MALDI-TOF mass spectrometry, and reversed phase HPLC, which resulted in a single peak.
Inhibition of TF⅐FVIIa Activity-Inhibition of FX activation by TF⅐VIIa was determined in triplicate with 800 pM relipidated TF-(1-243), 8 pM FVIIa, and 330 nM FX at 25°C as a function of peptide concentration essentially as described (14,39). TF was incorporated into phospholipid vesicles and quantified as described previously (39). The linear rates of FXa generation (mA 405 ⅐min Ϫ1 ⅐min Ϫ1 ) were converted to percent activities (100% ϫ v i /v 0 ). Control experiments showed that the peptides tested did not inhibit the FXa amidolytic activity.
Inhibition of amidolytic activity to measure the IC 50 value and extent of inhibition was carried out in triplicate at pH 7.8 essentially as described previously except that 30 nM sTF and 1 mM Chromozym t-PA were used (14). The linear rates of the increase in absorbance at 405 nm are expressed as percent activities (100% ϫ v i /v 0 ).
Clotting Assays-The PT assay was performed in citrated and pooled normal human plasma as described previously (14). Clotting times were determined in triplicate using an ACL 6000 Coagulation Analyzer (Coulter Corp., Miami, FL) using Innovin (human relipidated TF and Ca 2ϩ ) from Dade Behring Inc. (Newark, DE) to initiate the assay.

Structure-based Design and Substrate Phage Strategy-
The crystal structure of A-183 in complex with the protease domain of FVII revealed the location of its binding site as well as the orientation of its N and C termini (16,18). The C-terminal Arg of A-183 was in close proximity (ϳ17 Å direct line) to the protease-active site. We modeled a hypothetical extension onto the C terminus of A-183, attempting to find the shortest reasonable path to the active site of FVIIa (Fig. 2). Considering the shape of the surface and the constraints in flexibility due to peptide bonds, we estimated that a 40-Å extension of ϳ11 residues should be sufficient to place an Arg into the S 1 subsite.
In order to select for the preferred peptide length and sequence that would extend into the active site, we utilized a substrate phage display approach. The challenge of this approach is to bind A-183 phage libraries to FVIIa under conditions when it is inactive and then activate the enzyme in the presence of the bound phage libraries to select for only those phage that are released due to enzymatic cleavage. The latter condition becomes even more challenging since A-183 inhibits  (16) complexed with peptide A-183 (green) having a C-terminal extension of 11 residues (yellow); C␣ atoms are depicted as red spheres. The extension is modeled into the active site where its C-terminal Arg (red) is at the S 1 subsite.
TF⅐FVIIa activity, albeit incompletely (15). The amidolytic activity of FVIIa has been studied under a variety of different conditions such as changes in pH, different salts at various concentrations, and the effect of TF (40). In the absence of TF, FVIIa exhibits very low enzymatic activity. Addition of TF enhances the activity for the cleavage of FX by over 10 4 -fold, whereas the increase in amidolytic activity ranges from ϳ20to 100-fold, depending on the pH and the substrate (13,40). Importantly, in the absence of TF, FVIIa showed no detectable amidolytic activity below pH 6.0 or above pH 10.5 (40).
First, we tested the binding of biotinylated A-183 to immobilized FVIIa at various pH values, ranging from pH 4 to 11. No substantial differences in binding were observed between pH 6.0 and 8.5. We then determined whether immobilized FVIIa is still active when coated on a plate and whether its activity can be specifically regulated (Fig. 3). MaxiSorp plates were coated with 4 different concentrations of FVIIa and tested in the absence and presence of sTF at pH 6.0 and pH 8.5. At 10 g/ml FVIIa, the presence of sTF showed a 15-fold increase in activity at pH 8.5 versus no sTF, whereas amidolytic activity was essentially undetectable at pH 6.0 in the absence of sTF (Fig. 3).
As described previously, A-183 shows potent but incomplete inhibition of amidolytic activity (15). To test whether the residual activity is still sufficient for cleavage in the presence of A-183, we designed a TF⅐FVIIa control substrate (A-183-CS-Z) comprising A-183, an 18-residue linker with the 8-residue FVIIa cleavage sequence found in FX (NLTRIVGG), and the Z-domain of protein A, which was used as an IgG binding affinity tag to facilitate purification (36). The putative P 1 arginine in A-183-CS-Z was 11 residues from the C-terminal Arg of A-183. A-183-CS-Z was incubated with FVIIa in the presence and absence of sTF at pH 6.0 and 8.5. The reaction mix was incubated at room temperature for 2 h and separated by HPLC (data not shown). Collected peaks were analyzed by SDS-PAGE, N-terminal sequencing, and mass spectrometry, which confirmed that the fusion peptide was cleaved after the putative P 1 arginine only under enzymatically favorable conditions (sTF, pH 8.5); no cleavage was observed at pH 6.0 and the absence of sTF. Thus, suitable conditions for controlling both binding (pH 6.0) and catalysis (sTF, pH 8.5) were determined.
Based on these results, substrate phage libraries (see below) were captured on immobilized FVIIa at pH 6.0 (Fig. 1). In principle, the change to enzymatically favorable conditions (sTF, pH 8.5) should result in the release of only those phage displaying preferred substrate sequences. However, the addition of cleavage buffer resulted in the reestablishment of a new binding equilibrium, as evidenced by the release of uncleaved phage. In order to propagate only cleaved phage, we introduced an N-terminal His 6 tag in the A-183 phage libraries to remove any uncleaved phage, i.e. those containing both the His tag and A-183, from solution using an immobilized anti-His mAb. The cleaved phage, which remained in solution, were then propagated and sorted further for enrichment. Control phage displaying His 6 -A-183 were 1000-fold more selectively captured by the anti-His mAb versus phage displaying only A-183. About 2 ϫ 10 6 colony-forming units were captured by 500 ng of immobilized antibody, which was determined by propagating a serial dilution of both test phage.
Library Design and Sorting Results-Four different libraries were designed to determine the preferred length and sequence of the extension (Fig. 4). Libraries A-C contained a putative P 1 arginine at position 12, 11, or 10, respectively; the first 7 residues were fully randomized. Position 8 was randomized with residue subsets S/N/K/R, N/K, and L/Q in libraries A-C, respectively, based upon sequences taken from human FVII, FIX, and FX substrates (40). Positions prior to the P 1 arginine were fixed according to the cleavage sequence from FX, whereas those following the P 1 arginine had the IVGG sequence found from P 1 Ј to P 4 Ј at the FVII and FX cleavage sites. The linker for library D contained a flexible (GGS) 2 G spacer followed by fully randomized residues from positions 8 to 14, addressing the question of which residues are preferred in the cleavage site. In natural and synthetic FVIIa substrates proteolysis occurs after Arg, defining it as the residue at the P 1 position. Thus, the position in which Arg was observed most frequently could also be used to determine the preferred length of the extension.
The results from libraries A and D were combined to create a consensus library for residues 1-15 of Gly 1 -Glu 2 -(Gly/Val/ Glu) 3 -(Gly/Asp/Val) 4 -(Glu/Asp) 5 -(Glu/Gly) 6 -Glu 7 -X 8 -X 9 -Glu 10 -Trp 11 -Arg 12 -Gly 13 -X 14 -X 15 (Fig. 4), where the Arg at position 12 was considered to be the P 1 residue. In this library, positions 1, 2, 7, and 10 -13 were fixed; positions 3-6 were partially randomized, and positions 8,9,14, and 15 were completely randomized, leading to a theoretical diversity of 3.8 ϫ 10 7 variants. After panning this library for 6 rounds, 200 clones were randomly sequenced and analyzed (Table II). A consensus sequence was derived from this library from analysis at positions 8 and 9 first where Leu and Trp were most highly preferred both individually and in combination with each other; positions on the PЈ side were not taken into account for this analysis. Further analysis in a sequence context-dependent manner (i.e. Leu 8 -Trp 9 -Glu 10 -Trp 11 -Arg 12 ) led to the most preferred residues at the remaining positions. The consensus sequence for positions 1-15 determined by these criteria was GEG-VEEELWEWRGFA.
Inhibition of TF⅐FVIIa Activity-In order to characterize the activity of the phage-selected peptides, we isolated the preferred A-183 extension peptide A-183X (EEWEVLCWTWET-CERGEGVEEELWEWR). A-183X was fused to the C terminus of the 58-residue Z-domain of staphylococcal protein A (36) via a linker containing an FXa cleavage site to facilitate expression and purification. Following purification on IgG-Sepharose, the fusion was treated with FXa and subjected to size exclusion chromatography to isolate the 27-mer peptide A-183X. Compared with the A-183 control, there was a marked improvement in the inhibition of TF⅐FVIIa-catalyzed activation of FX in both the maximal extent of inhibition and the affinity (Fig. 5A). A-183X showed complete (ϳ99%) inhibition of FX activation with an IC 50 of 230 pM, whereas A-183 inhibited to a maximal extent of 74% with an IC 50 of 1.5 nM.
To assess whether the extension has to be specific to obtain complete inhibition of FX activation, we tested arbitrary 12residue extensions of Z-A-183 consisting of a GGS motif with and without an Arg at the putative P 1 position (Z-A-183-[GGS] 3 GGR and Z-A-183-[GGS] 4 , respectively). These two controls were deemed necessary since prior data showed that the Z-domain C-terminally fused to A-183 increased the extent of inhibition in FX activation to 89% at saturating peptide concentrations (15). Consistent with these findings, both random control extensions also improved upon the 78% extent of inhibition of FX activation observed for Z-A-183 (Fig. 5B), but only up to 87%, essentially the same as that observed for the Zdomain fused to the C terminus of A-183. These effects were independent of the Arg at the putative P 1 position. In addition, the control extensions provided no substantial improvement in affinity, having IC 50 values of 1.5 and 1.1 nM for Z-A-183-[GGS] 3 GGR and Z-A-183-[GGS] 4 , respectively, versus 1.7 nM for Z-A-183. Z-A-183X had an IC 50 of 240 pM, exhibiting the same 7-fold improvement in affinity as A-183X and completely (ϳ99%) inhibited FX activation. Thus, we conclude that the 12-mer peptide extension is responsible for the specific improvements (complete inhibition and improved affinity for TF⅐FVIIa) sought by our substrate phage selection strategy. A-183 extension peptide sequences were obtained by selection from the random linker libraries. From 100 selected clones in each library, 55 (library A) and 85 (library D) readable sequences were aligned and the occurrence of each amino acid was tabulated. The percent occurrence of each amino acid at each position was calculated after normalization for codon bias. Residues where the fraction found exceeded the fraction expected by at least 2 n (boldface) and 3 n (boldface and shaded) are highlighted.
a Dashes represent amino acids that were not included in the library.

Engineering Peptide Exosite Inhibitors of Factor VIIa
We tested the ability of a number of extension peptides to inhibit amidolytic activity using the small chromogenic substrate Chromozym t-PA as well since this may give us further insight into assessing effects at the active site. This assay has a larger dynamic range since the maximal inhibition by A-183 at saturating concentrations is only 32% instead of 78% found in the FX activation assay (15). The data for inhibition of amidolytic activity is shown in Fig. 6. In agreement with the FX activation data, we see a significant improvement of the maximal extent of inhibition by Z-A-183X to 54%. The arbitrary 12-residue extension control peptides Z-A-183-[GGS] 4 and Z-A-183-[GGS] 3 GGR maximally inhibited to 37 and 40% inhibition, respectively. Most importantly, either removal (Z-A-183X-⌬R12) or alanine substitution (Z-A-183X-R12A) of the C-terminal Arg of Z-A-183X resulted in peptides with maximal extents of inhibition (38 and 39%, respectively) comparable with the control peptides and significantly less than the 54% for Z-A-183X (Fig. 6). For completeness, the maximal inhibition of amidolytic activity at saturating concentrations by A-183 and A-183X was determined to be 31 and 55% inhibition, respectively (data not shown), indicating that the N-terminal Z-domain had no effect. In the amidolytic assay, the IC 50 value for Z-A-183 was 6.0 nM, whereas all of the extension peptides had slightly better IC 50 values of ϳ3.2 nM (Fig. 6).
Mechanistic Implications-The fact that the C-terminal Arg in the extension is important for maximal activity is consistent with the proposal that the extension lies in the substrate binding cleft with the Arg in the P 1 position. To address this point, we expressed and purified A-183X-IVGGSGG-Z, which con-tains the 12-mer extension consensus sequence derived from substrate phage followed by the 7-residue sequence IVGGSGG as the linker between A-183X and the Z-domain of protein A. Following incubation of A-183X-IVGGSGG-Z with FVIIa under different conditions, reaction mixtures were analyzed by non- TF⅐FVIIa using a consensus library A-183 extension peptide sequences were obtained by selection from the consensus linker library. From 200 selected clones, 154 readable sequences were aligned, and the occurrence of each amino acid was tabulated. The percent occurrence of each amino acid at each position was calculated after normalization for codon bias. Residues where the fraction found exceeded the fraction expected by at least 2 n (boldface) and 3 n (boldface and shaded) are highlighted.
a Dashes represent amino acids that were not included in the library at these positions.

Engineering Peptide Exosite Inhibitors of Factor VIIa
reducing SDS-PAGE and MALDI-TOF mass spectrometry. At pH 6.0 in the absence of sTF, we found a single band at ϳ12 kDa and a molecular mass of 11,745.0 daltons (calculated 11,744.8 daltons). At pH 8.5 in the presence of sTF, we observed 2 new bands of ϳ8 and ϳ3 kDa and a molecular mass of 8,263.9 daltons, which corresponds to the calculated mass of 8,265.0 daltons of the product IVGGSGG-Z. Therefore, the Arg at position 12 is indeed the P 1 residue, which binds at the S 1 subsite of FVIIa in the TF⅐FVIIa complex.
Although our results in completely inhibiting FX activation with A-183X were satisfying, it was initially somewhat surprising to find that the extent of inhibition of amidolytic activity was not 100%, as might be expected if the extension blocks the active site. We offer the following explanations for this result. Whereas the extension peptide was selected based on sequence preference as a substrate, the A-183X peptide itself has a carboxylate at its C terminus instead of an amide and is thus more akin to a product, which likely has a relatively fast off-rate. Furthermore the 12-mer extension peptide itself had no inhibitory activity (data not shown). Therefore, any binding energy due to the extension itself is relatively weak and certainly much weaker than A-183. Consistent with this, only moderate increases in potency of 7-and 2-fold were found for A-183X in the FX activation and amidolytic assays, respectively. At saturating concentrations of A-183X, where A-183 provides essentially all of the binding energy, the C-terminal tail is mostly "free" in solution. The increase in the extent of inhibition for A-183X can be ascribed to the fraction of time that the tail is bound and blocks the active site. The fact that the C-terminal Arg is important for this interaction is consistent with its importance for binding at the S 1 subsite, as noted for many serine proteases. The fact that we see complete inhibition for FX activation is likely due to the fact that this macromolecular substrate has a much larger binding site on TF⅐FVIIa compared with Chromozym t-PA. This would be especially important in the substrate binding cleft where the extension lies.
Prolongation of TF-dependent Clotting Time in Normal Human Plasma-The effect of more completely and potently inhibiting TF⅐FVIIa activity in vitro by the extension peptides was further elaborated in TF-dependent clotting assays in normal human plasma. The peptides showed a concentration-dependent prolongation of the PT, consistent with their ability to inhibit TF⅐FVIIa activity. Furthermore, the improvement in inhibiting FX activation was correlated with a significant improvement in the maximal extent of prolongation of the PT and in potency (Fig. 7). Peptide A-183X showed a maximal prolongation in the PT of 7.6-fold versus 1.9-fold for A-183. There was also a notable increase in potency of A-183X versus A-183 (as well as for Z-A183X versus Z-A-183) as observed by a more prolonged PT at lower concentrations (Fig. 7). Clotting assays were also carried out in the context of Z-domain fusions to include a control extension peptide. In this case the extension peptide Z-A-183X had a maximal PT fold prolongation of more than 8.6-versus 2.7-fold for Z-A-183. Z-A-183X also showed an increase in potency versus Z-A-183. The maximal PT fold prolongation for the control extension peptide Z-A-183-[GGS] 4 was 3.3-fold, slightly above that for Z-A-183, but significantly below that for Z-A-183X. The slight differences in the PT fold prolongation observed for identical peptides with and without the Z-domain may result from the more complex reactions taking place in plasma. The observation that clotting does eventually occur at saturating concentrations of A-183X or Z-A-183X may be due to any residual catalytic activity of approximately Յ1% and/or TF-independent clotting (17).
Sequence Rationale-The preferred extension sequence con-tained the predicted P 1 arginine located 12 residues from the C terminus of A-183, in agreement with the design criteria. Whereas other selected residues in the extension were considerably less predictable, one might think that residues at P 2 , P 3 , P 4 , etc. would resemble sequences from the natural substrates FVII, FIX, and FX (40). In fact, there is no resemblance to these sequences for several possible reasons. The selected consensus extension reflects several influences, among which are alignment of a suitable P 1 residue with the S 1 subsite and sufficient flexibility to traverse the FVIIa surface. Library designs included a relatively short connection between the C terminus of A-183 and the P 1 residue in order to minimize entropy loss upon binding. It is likely that the orientation of A-183X as it enters the active site is somewhat different from that of the natural substrates. Furthermore, binding of A-183 induces a significant conformational change in the 60s loop (16,18), which likely results in an allosteric effect at the active site that alters substrate binding and leads to inhibition (15). Thus, in the presence of A-183, extension residues that are preferred in the substrate binding cleft may be quite different from those in its absence. Residues in the extension are composed of hydrophobic and negatively charged residues. We speculate that the Trp and Leu residues may provide moderate binding energy in the substrate binding cleft, whereas Glu residues could provide favorable interactions with polar surface residues and more likely benefit from solvent exposure. Implications for Other Proteases-Proteases bind their macromolecular substrates at both their active sites and exosites (7)(8)(9)(10)(11)(12). Other approaches to develop chimeric inhibitors of TF⅐FVIIa include linking a TF mutant that cannot bind FX with a Kunitz domain optimized for inhibition of TF⅐FVIIa activity (41) as well as linking A and E exosite peptide inhibitors together (17). In both of these cases, more effective anticoagulation was observed. Chimeric protease inhibitors are not limited to FVIIa. Bivalirudin, also known as hirulog, a 20residue peptide derived from fusing the C terminus of hirudin with an active site peptide inhibitor (42), specifically and reversibly inhibits thrombin by binding both to its anion-binding exosite and its catalytic site through a complex set of kinetic and conformational steps (43). Bivalirudin has recently been approved as an intravenous anticoagulant given in combination with aspirin to prevent clot formation in patients with unstable angina who are undergoing coronary angioplasty.
The protease switch strategy we have taken to develop potent, selective, and complete inhibitors of TF⅐FVIIa may have utility for other cofactor-dependent proteases. In particular, other members of the coagulation cascade have protein cofactors required for optimal catalytic activity (5, 6). The clinical need for improved anticoagulants that are potent and selective may benefit from our chimeric protease inhibitor approach.