Inhibition of Plasma Kallikrein by a Highly Specific Active Site Blocking Antibody

Background: Unregulated plasma kallikrein proteolytic activity can result from C1-inhibitor deficiency, causing excessive and potentially fatal edema. Results: The antibody DX-2930 potently and specifically inhibits plasma kallikrein and exhibits a long plasma half-life. Conclusion: An antibody protease inhibitor can lead to potent and specific bioactivity. Significance: DX-2930 could be an effective therapeutic for the prophylactic inhibition of plasma kallikrein-mediated diseases.

HAE-C1INH is a rare disease caused by unregulated activation of the KKS that clinically manifests as intermittent attacks of subcutaneous or mucosal edema affecting the face, larynx, gastrointestinal tract, limbs, or genitalia (27). Affected individuals harbor autosomal dominant mutations in the C1-INH gene (SERPING1) that lead to a deficiency in either total (type I HAE-C1INH) or functional (type II HAE-C1INH) C1-INH protein, a protease inhibitor of the serpin superfamily that regulates the contact, coagulation, fibrinolytic, and complement pathways (28,29). During an HAE attack, reduced levels of functional C1-INH are insufficient to inhibit its key targets, both FXIIa and pKal ( Fig. 1), thereby causing the overproduction of bradykinin and ultimately the pain and swelling characteristic of this disease.
Clinically, the role of the plasma KKS in HAE-C1INH pathophysiology has been demonstrated by the approval of therapeutics that inhibit this pathway. For treatment of acute attacks, these include C1-INH replacement, selective inhibition of pKal proteolytic activity with the Kunitz domain inhibitor ecallantide, or bradykinin B2 receptor antagonism with the bradykinin peptidomimetic icatibant (30). Prophylactic treatments include chronic C1-INH replacement, which is limited by the need for frequent intravenous administration that can lead to the placement of an indwelling catheter. C1-INH therapy is also associated with a risk for serious thromboembolic events (31). Oral androgens are an alternative prophylactic option that are more convenient to administer but are associated with serious toxicities and morbidities from unwanted androgenizing effects, particularly in women, as well as hepatic adenomas with malignant potential (32,33). Furthermore, despite chronic treatment with existing prophylactic agents, breakthrough attacks still occur (34). Thus, there remains an unmet medical need for a long lasting, convenient, safe, and effective prophylactic therapeutic to treat HAE-C1INH.
Given the potential for target specificity and long serum halflife with antibody therapeutics, we used phage display to select a potent and highly specific human antibody inhibitor of pKal, DX-2930. Here, we describe the discovery and biochemical characterization of DX-2930, including the x-ray structure of the Fab⅐pKal complex, which provides a rationale for the specificity and effectiveness of DX-2930. Moreover, preclinical studies reveal DX-2930 to exhibit a prolonged half-life in circulation, and we further demonstrate that this translates into a prolonged capacity to inhibit the KKS. Taken together, data presented here support the potential of DX-2930 for the prophylactic inhibition of pKal-mediated edema and provide a highly specific reagent tool to investigate the effects of targeted pKal inhibition in complex biological samples containing highly homologous proteases.

EXPERIMENTAL PROCEDURES
Proteins-Full-length human plasma kallikrein (prekallikrein and activated pKal forms), 1-chain kininogen (intact kininogen), and 2-chain kininogen (treated with pKal to excise bradykinin) were obtained from Enzyme Research Laboratories. C1-INH and pKal orthologues from rat, cynomolgus monkey, and rabbit plasma were purified by Athens Research and Technology. The pKal catalytic domain variant used in crystallographic studies contained three deglycosylation (N396E, N453E, and N494E) and two cysteine (C383S and C503S) mutations and were purified as described previously (35).
M Pro-Phe-Arg-AMC as substrate in Buffer B. Neutrophil elastase (Athens Research and Technology) was assayed using 100 M N-methoxysuccinyl-Ala-Ala-Pro-Val-AMC as substrate in Buffer B. Plasmin (Haematologic Technology Inc.) was assayed using 10 M H-D-CHA-Ala-Arg-AMC as substrate in Buffer A. Thrombin ␣ (Enzyme Research Labs) was assayed using 10 M H-D-CHA-Ala-Arg-AMC as substrate in Buffer A containing 2.5 mM CaCl 2 and 1 mg/ml BSA. Recombinant tissue kallikrein 1 (R & D Systems) was assayed using 10 M Pro-Phe-Arg-AMC as substrate in Buffer B. Recombinant tissue kallikrein 2 (36) was assayed using 10 M Pro-Phe-Arg-AMC as substrate in Buffer B. Recombinant tissue kallikrein 5 (R & D Systems) was assayed using 10 M Z-Val-Val-Arg-AMC as substrate in Buffer A. Recombinant tissue kallikrein 12 (R & D Systems) was assayed using 10 M Val-Pro-Arg-AMC as substrate in Buffer C. Trypsin (Athens Research and Technology) was assayed using 10 M Pro-Phe-Arg-AMC as substrate in Buffer B. Recombinant tissue plasminogen activator (Sigma) was assayed using 10 M Z-Gly-Pro-Arg-AMC as substrate in Buffer A containing 1% BSA. Urokinase (Sigma) was assayed using 10 M benzyl-Ala-Gly-Arg-AMC as substrate in Buffer A.
Phage Display Selection and Primary Screening-Human antibodies against pKal were obtained by panning an antibody phage display library (37) against biotinylated pKal immobilized on streptavidin-coated magnetic beads (Dynal, M280). Selections were performed as described (38), including positive selections with decreasing amounts of biotinylated human or rat pKal and negative selections with untreated streptavidin beads and biotinylated prekallikrein. Phage isolates were screened by ELISA (streptavidin-immobilized pKal with detection by anti-M13 coat protein VIII), positive hits were DNAsequenced, and unique Fabs were batch-processed for Escherichia coli expression as isolated Fab fragments from the pMID21 vector as described (38). An affinity maturation library was constructed around the lead Fab (given its inhibition of both human and rat pKal) by employing mixed nucleotide synthesis of heavy chain variable complementary determining region 3 (HV-CDR3) where the nominal base was present at 85% and each of the others at 5%.
Antibody Expression and Purification-High throughput production of E. coli-secreted Fabs and their subsequent purification by protein A-Sepharose followed previously reported protocols (38). The DX-2930 Fab construct generated for crystallization studies was produced as follows. The DNA sequence for the DX-2930 Fab region was codon-optimized and cloned into the pJexpress 401 expression vector by DNA2.0. SS320 TG1 E. coli cells (Lucigen) were transformed with the Fab expression vector and grown in 10 ml of TBDry media containing 2% glucose and 25 g/ml kanamycin for 18 h at 30°C. 10 liters of TBDry media containing 0.1% glucose and 25 g/ml kanamycin was then inoculated with 10 ml of the starter culture (1:1000 ratio) and grown for 20 h at 30°C. Cells were harvested by centrifugation (15 min at 2000 ϫ g); the pellet resuspended in 10 liters with TBDry media containing 25 g/ml kanamycin, 1 mM isopropyl 1-thio-␤-D-galactopyranoside, and 0.5% Triton X-100, and incubated an additional 20 h at 30°C with shaking. The culture was finally harvested by spinning at 9000 ϫ g for 30 min and filtered through a Corning 0.22-m PES filter. The resulting culture supernatant containing secreted Fab protein was subjected to a three-step chromatography purification. Fab product was first captured by immunoaffinity protein A resin (RepliGen). 80 ml of immobilized protein A resin was added to the 10 liters of filtered supernatant and incubated with rocking for 18 h at 4°C. Resin was captured in a fritted column and extensively washed to remove host cell-related impurities and reduce endotoxin level. Washes successively included 10 column volumes (cv) of PBS, 20 cv of Fab Wash Buffer (50 mM sodium acetate, 0.5 M sodium chloride, pH 5.0), 50 cv of PBS with 0.1% Triton X-100, and then a final wash with 50 cv PBS. Soluble Fab (sFab) was then eluted from protein A with 6 cv of Fab elution buffer (50 mM sodium phosphate, 150 mM sodium chloride, pH 2.5) and instantly neutralized to pH 6.0 with 2 cv equivalents of 1 M HEPES, pH 7.2. Neutralized protein A eluate was loaded onto PBS pre-equilibrated mercaptoethylpyridine HyperCel resin (Pall Lifesciences) and incubated overnight at 4°C with rocking. Mercaptoethylpyridine resin was then washed with 10 cv of PBS, followed by step elutions with 50 mM sodium acetate solutions at varying pH values from pH 5 to pH 3. The major elution peak at pH 5 was collected and directly loaded onto a POROS HS 50 strong CEX resin (Invitrogen) pre-equilibrated with 50 mM sodium acetate, pH 4.5 ("CEX Buffer"), washed first with the same CEX Buffer, and then an additional wash with CEX Buffer ϩ 250 mM NaCl. Fab was subsequently eluted with CEX Buffer ϩ 350 mM NaCl. Purified Fab product was exchanged into a buffer containing 50 mM citrate/phosphate, 100 mM NaCl, pH 6.0, through an ultrafiltration/diafiltration process using a Millipore Amicon Ultra concentration unit (10,000-kDa molecular mass cutoff). Recombinant Fab fragments were reformatted into full-length human IgG1 antibodies (f-allotype) (39), expressed as stable transfections in CHO cells using a fed-batch fermentation strategy, and purified as previously described (40) into formulation buffer (100 mM citrate/phosphate buffer, pH 6.0, 50 mM NaCl, 2% (w/v) trehalose, and 0.01% Tween 80).
Enzyme Inhibition Analysis-Steady state inhibition measurements were performed as described previously (40). Briefly, pKal (nominal concentration of 1 nM as total protein) was incubated with different concentrations of inhibitor for 1 h at 30°C in Buffer B (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% PEG-8000, and 0.1% Triton X-100) in a 96-well plate. The maximum final inhibitor concentration was 10 nM, followed by a 1.5-fold serial dilution (total 11 nonzero inhibitor concentrations) plus a control ([I] 0 ϭ 0). The reactions were initiated by the addition of H-Pro-Phe-Arg-AMC substrate (Sigma, catalogue no. P9273) at a final concentration of 10, 50, or 100 M and monitored in a fluorescence plate reader (Spec-traMax, Molecular Devices) with excitation and emission wavelengths at 360 and 480 nm, respectively. Initial reaction rates were determined by linear fit of raw experimental data traces (fluorescence versus time) as the slopes of the regression lines. The initial reaction rates, v, were fit to Equation 1.
In  (42). In contrast with Morrison's original formulation, the enzyme concentration [E] 0 does not appear in the denominator and instead is subsumed in the empirical constant A. This formulation significantly increases numerical stability in nonlinear regression for the following reason. In our particular case ([E] 0 ϭ 1 nM, K i app Ϸ0.1 nM) the enzyme concentration [E] 0 must be treated as an adjustable parameter, according to the theoretical principles explained elsewhere (43). However, during the iterative regression analysis, [E] 0 could transiently attain values very near zero, which could cause numerical underflow ("division by zero") in the original form of the Morrison equation (42). The Michaelis constant K m for the synthetic substrate was determined to be 315 Ϯ 16 M (data not shown). At a substrate concentration [S] 0 ϭ 10 M, the factor 1 ϩ [S] 0 /K m in Equation 2 (assuming competitive inhibition, see inset in Fig. 2A) is equal to 1.03.
Thus, in our particular case the apparent inhibition constants K i app obtained in the regression analysis are identical to the true inhibition constant within 3%. The activity of 20 different human serine proteases listed above was observed in the absence or presence of 1 M DX-2930 using synthetic chromogenic or fluorogenic substrates.
ELISA Measurement of pKal-mediated Bradykinin Formation-Following the incubation of human pKal (0.15 nM) with varying concentrations of DX-2930 for 1 h at 30°C in Buffer B, 1-chain HMWK was added at a final concentration of 20 nM. After the reaction proceeded for 10 min at 30°C, samples were applied to a 96-well 10-kDa molecular mass cutoff filter plate (Millipore MultiScreen with Ultracel-10 membrane), centrifuged at 400 ϫ g in a table top centrifuge for 45 min to remove unreacted HMWK, and analyzed by an anti-bradykinin ELISA (Peninsula Laboratories) per the manufacturer's instructions.
ELISA Detection of pKal Binding to HUVEC-Human umbilical vein endothelial cells (Lonza) were seeded onto 96-well plates at ϳ2 ϫ 10 4 cells/well, and incubated for 48 h. The wells were blocked with 3% BSA in Buffer D (20 mM HEPES, 150 mM NaCl, 2 mM CaCl 2 , 1 mM MgCl 2 , pH 7.5), washed, and then treated with Buffer E (Buffer D ϩ 50 M ZnSO 4 and 0.5 mg/ml BSA) or HMWK (40 nM in Buffer E) for 1 h at 25°C, followed by a 1-h incubation with 3 nM pKal or pKal pretreated with 1 mM 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride (AEBSF, Thermoscientific). Varying concentrations of biotinylated DX-2930 were then added to the treated cells for 2 h, and bound antibody was detected by streptavidin-alkaline phosphatase and p-nitrophenyl phosphate additions.
Surface Plasmon Resonance-Surface plasmon resonance (SPR) measurements were performed using a BIAcore 3000 (GE Healthcare) with the detection temperature at 25°C and HBS-P running buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, and 0.005% surfactant P20). To measure kallikrein interactions with DX-2930, goat anti-human Fc fragment-specific IgG (Jackson ImmunoResearch) was immobilized by amine coupling on a CM5 sensor chip at ϳ5000 response units (RU), and DX-2930 was subsequently captured by injecting a 5 nM antibody solution for 4 min at 5 l/min. Reference surfaces were activated and blocked in mock amine coupling reactions. 500 nM prekallikrein, pKal, or pKal-AEBSF were injected for 5 min at 20 l/min followed by a 5-min dissociation phase. Surfaces were regenerated with a 30-s pulse of 10 mM glycine, pH 1.5, at 100 l/min. To measure the affinity of pKal to HMWK, 2-chain HMWK was immobilized by amine coupling on a CM5 chip at ϳ2000 RU, and varying concentrations of pKal were flowed over the surface for 5 min at 20 l/min, and dissociation followed for 5 min. Equilibrium RU values for each pKal concentration, [E], were fit to Equation 3 to obtain the observed dissociation equilibrium constant K D . Reference surface and chip regeneration followed the above procedures.

X-ray Crystallographic Structure of the DX-2930⅐pKal
Complex-Purified DX-2930 Fab and the deglycosylated pKal catalytic domain variant (35) were incubated at 1:1 stoichiometry. The resulting complex (150 M) in addition to Fab alone was diluted 1:1 in reservoir precipitant ( Table 2) and crystallized by the hanging drop vapor diffusion method. Diffraction data from resulting crystals were collected with a Rigaku FR-Eϩ Super-Bright x-ray generator with a Saturn 944ϩ CCD detector and VariMax HF optics (complex) and at the Argonne National Laboratory (APS) beam line (Fab alone, refer to Table 2). The complex structure was solved by molecular replacement in Phaser using separate constant and variable domains of a Fab model (derived from PDB 3IDX) and the pKal catalytic domain structure PDB 2ANY (35). The crystallographic model of the 2930 Fab determined from the pKal⅐2930 Fab complex structure was then used to help solve the Fab-only structure. The final models were obtained after iterative building/refinement in REFMAC/COOT (44 -46).
Pharmacokinetic Properties of DX-2930 in Cynomolgus Monkey-Six male cynomolgus monkeys were assigned to two dose groups, each consisting of three animals. All animals (treated at WIL Research) were dosed on day 0, with group 1 receiving a single i.v. injection of 20 mg/kg DX-2930 (2 ml/kg of DX-2930 at 10 mg/ml) and Group 2 receiving a single s.c. injection of 20 mg/kg DX-2930 (also 2 ml/kg of DX-2930 at 10 mg/ml). Blood was collected from all animals in citrated plastic tubes at the following time points: predose, 5 min; 1, 4, 24, 48, 96, 168, 240, 336, 432, 504, 672, 840, and 1008 h post-dose. Plasma samples were analyzed using a qualified sandwich ELISA method. Briefly, 96-well microplates (Costar 3059) were coated with 1 g/ml goat anti-human IgG (Bethyl Labs: preadsorbed with monkey IgG; diluted in 100 mM sodium carbon-ate, pH 9.4) by incubation at 4°C overnight and subsequently blocked with 3% BSA in 1ϫ PBS (incubated 2 h at 37°C). Test sample and antibody standards were then diluted 1:10 with PBST-B (1ϫ PBS, 0.1% Tween 20, and 3% BSA) and captured on the microplate by incubation for 2 h at 37°C. Finally, bound IgG/DX-2930 was detected by the addition of 50 ng/ml HRPlabeled goat anti-human IgG (Bethyl Labs) diluted in PBST-B. Antibody standard dilutions were included to calculate the concentration of DX-2930 in sample plasma. From these data, the pharmacokinetic parameters were calculated using WinNonlin Professional Version 5.3 (Pharsight Inc.). All data were analyzed noncompartmentally.
Ex Vivo Analysis of HMWK Cleavage by pKal-Cynomolgus monkeys were subcutaneously injected with 25 and 50 mg/kg DX-2930, and blood was collected as above at WIL Research. Aliquots of the resulting citrated plasma samples were activated with 150 g/ml kaolin for 10 min at 37°C, diluted 1:20 in 1ϫ PBS, and aliquots analyzed by Western blots blocked in PBST ϩ 2% BSA and probed with 1:2000 HRP-labeled sheep anti-human kininogen polyclonal antibody (Cedarlane) diluted in PBST ϩ 2% BSA.
Carrageenan-induced Paw Edema-Male Sprague-Dawley rats (180 -200 g each, Harlan Laboratories) were divided into different treatment groups (n ϭ 10 each group) and treated at Washington Biotechnology, Inc., with either an i.p. injection of Formulation Buffer, indomethacin-positive control (Sigma, 5 mg/kg prepared in 0.1 M NaHCO 3 at a concentration of 2.5 mg/ml), or an s.c. injection of different doses of DX-2930. The i.p. and s.c. injections were administered 30 min and 24 h, respectively, prior to the injection of 0.1 ml of 1% carrageenan (FLUKA, prepared in distilled water) into the right hind paw. Paw volumes were measured by plethysmography before and 1, 2, 4, 6, and 8 h after carrageenan injection.

Discovery of DX-2930 by Antibody Phage Display and Affinity
Maturation-Phage isolates were screened for interaction with pKal by ELISA, and hits were characterized using purified Fab fragments. Of these, one Fab (M0162-A04) was selected for affinity maturation given that it inhibited the proteolytic activity of both human and rat pKal with equal potency (K i app Ϸ5 nM), while not binding prekallikrein (data not shown). A comparison of apparent inhibition constants (K i app ) and HV-CDR3 amino acid sequences for the most potent inhibitors obtained from the affinity maturation campaign (Table 1) revealed that only a few amino acid substitutions were necessary to significantly improve potency. For instance, the introduction of a valine at position 5 (I103V) in HV-CDR3 alone improved the affinity almost 10-fold (e.g. the M0201-H06 Fab, K i app ϭ 0.56 nM), although the combination of T101I and A108E substitutions along with I103V (isolate M0199-A08) further improved the potency toward pKal (K i app ϭ 0.05 nM). Including rational mutagenesis ( Table 1) and consideration of invariant residues post affinity maturation (e.g. Pro-104 and Arg-105), a common motif for the HV-CDR3 interaction with pKal is evident (shown as a modified SeqLogo in Table 1). Ultimately, the M0199-A08 variant was chosen to be reformatted for production as a germ line-optimized human IgG1 in CHO cells given its favorable potency, specificity, and protein expression/stability properties, and it is hereafter referred to as DX-2930.
DX-2930 Is a Potent Inhibitor of pKal-DX-2930 retains its potent in vitro inhibition of pKal as a full-length human IgG1 (Fig. 2A). The observed linear increase in K i app values with increasing substrate concentrations is consistent with a competitive inhibition mechanism for DX-2930 ( Fig. 2A, inset).  Fig.  2A. b Residue positions were specific for DX-2930. Sequence logo is shown for the affinity maturation output and weighting isolates for their capacity to inhibit pKal (higher weight to lower K i ).
Using Equation 2, the inhibition constant K i was calculated at each substrate concentration and averaged (K i ϭ 0.120 Ϯ 0.005 nM) ( Fig. 2A). DX-2930 also inhibited purified pKal from mouse (K i app ϭ 0.17 nM), rat (K i app ϭ 0.17 nM), rabbit (K i app ϭ 14.2 nM), and cynomolgus monkey (K i app ϭ 0.07 nM) (data not shown). In plasma, a major fraction of circulating pKal is bound to HMWK via sites distal from both the pKal catalytic domain and the bradykinin-containing domain of HMWK (1). Therefore, we also tested whether DX-2930 inhibits pKal complexed with 2-chain HMWK, a form of HMWK that lacks bradykinin ( Fig.  2A). The observed dissociation constant (K D ) for the interaction of pKal with sensor chip-immobilized 2-chain HMWK was determined by surface plasmon resonance experiments (Fig.  2B, K D ϭ 72 nM). Similar, or even higher affinity, dissociation constants (15 nM) have been previously reported for the interaction of pKal with 1-chain HMWK (47). Thus, at the physiologic HMWK concentration used in Fig. 2A (600 nM) (48) the catalytic amount of pKal present is essentially fully bound to HMWK. As shown in Fig. 2A, DX-2930 inhibits both free and HMWK-bound pKal with equivalent K i app values. Consistent with the formation of a saturated pKal⅐HMWK complex, we obtained similar K i app values at higher HMWK concentrations (4.7 M) (data not shown).
DX-2930 also effectively targets the pKal⅐HMWK complex when bound to endothelial cells (Fig. 2C). Binding of pKal to HMWK has been previously shown to lead to the association of the resulting complex with endothelial cells (1). Specifically, ELISA experiments verified that biotinylated DX-2930 only interacted with HUVEC cells pre-loaded with pKal complexed with cell-bound HMWK. In contrast, C1-INH interacted poorly with similarly treated cells.
In addition to the above experiments that utilize a small peptide substrate, we further show that DX-2930 is a potent inhibitor of the pKal-mediated proteolysis of single-chain HMWK (its endogenous protein substrate). Addition of DX-2930 to pKal samples in vitro prevented the proteolysis of HMWK and the consequent release of bradykinin with an IC 50 ϭ 1.3 nM, as measured by bradykinin-specific ELISAs (Fig. 2D). This observed IC 50 value is higher than the K i app value for DX-2930, which may be attributed to technical challenges with using an ELISA for enzyme kinetic measurements.
DX-2930 Only Inhibits Active pKal-Several findings established that DX-2930 specifically targets activated pKal, and in particular the active site of the enzyme. First, DX-2930 did not inhibit any of the 20 different serine proteases listed under "Experimental Procedures" (including the 68% identical FXIa catalytic domain) at a final concentration of 1 M DX-2930. Second, SPR experiments (Fig. 2E) demonstrated that DX-2930 binds active pKal, but binding is markedly diminished to an equivalent amount of prekallikrein or to pKal that has been inactivated with a small molecule, active site-directed inhibitor (AEBSF). This latter finding suggests that the antibody interacts with the pKal active site given that AEBSF covalently modifies the active site serine (49). Importantly, SDS-PAGE experiments that analyze DX-2930 samples preincubated with and without an equimolar concentration (2 M) of pKal demonstrate that the antibody is not cleaved (Fig. 2F).
Structure of the DX-2930⅐pKal Complex-To understand the interaction between DX-2930 and pKal with atomic resolution, we crystallized the Fab portion of DX-2930 both alone and in a complex with a deglycosylated variant of the human pKal catalytic domain (35). Diffraction data for Fab-only crystals (collected to 1.75 Å) and the pKal⅐DX-2930 Fab complex (to 2.1 and 2.4 Å for different crystal forms) was used to solve the structures by molecular replacement methods that employed the coordinates of both a generalized Fab framework without CDR loops and the pKal catalytic domain structure PDB 2ANY as the search models. Subsequent building/refinement resulted in the final complex and apo-Fab structures presented in Fig. 3, with diffraction and refinement statistics presented in Table 2. Overall, the three final complex structures (one in the asymmetric unit of PDB 4OGX and two in the asymmetric unit of PDB 4OGY) are highly similar, with small adjustments in the interface such that alignments centered on either the Fab or pKal yield lower deviations for the respective chains (Fig. 3B) Fig. 3D) that are present in the complex deviate little from their individual apo structures. However, DX-2930 heavy chain CDR loops do not have resolvable density in the apo structure.
Inspection of the complex structure reveals the reason why DX-2930 prevents substrate proteolysis by pKal. Namely, the Fab interface covers ϳ2400 Å 2 of surface and completely obstructs the active site of pKal (Fig. 4A). Direct contributions to this interface are made by Fab CDR and framework (FR) residues from both the heavy chain (HV-FR1, HV-CDR1, and HV-CDR3) and the light chain (LV-CDR2). Hydrophobic interactions of HV-FR1 residues, particularly Phe-27 and Phe-29, dominate the interface of this motif with pKal residues Val-410, Leu-412, Arg-416, Leu-418, Leu-439, Lys-575, and the His-434 of the pKal catalytic triad (Fig. 4B; prekallikrein numbering is used). For reference, the pKal residues His-434, Asp-483, and Ser-578 are equivalent to chymotrypsin catalytic triad residues His-57, Asp-102, and Ser-195, respectively. Moreover, His-31 of CDR1 forms a hydrogen bond with the backbone carbonyl of pKal residue Asp-437, and it consequently flips the Asp-437-Gly-438 peptide bond relative to the apo structure of the pKal catalytic domain. On the opposite side of the complex interface, Lys-50 of LV-CDR2 makes a side chain-mediated interaction with pKal residues Tyr-555 and Glu-600 (Fig. 4C). This interface is also supported by several backbone-mediated hydrogen bonds made by pKal residues Glu-527 (contacting amide nitrogen of Gly-57), Lys-528 (contacting carbonyl oxygen of Ser-56), and Arg-604 (interacting with carbonyl oxygen atoms of Leu-54 and Glu-55).
The most intimate interactions between the DX-2930 Fab and pKal catalytic domain are mediated by the same heavy chain CDR3 residues that led to affinity improvements ( Table  1). The CDR3 loop extends deeply into the active site of pKal (Fig. 4D) and interacts with both the substrate binding region and the catalytic triad. Arg-106 anchors the loop, burying into the canonical S1 binding pocket and forming a bidentate bonding interaction with pKal D572, much like the broad serine protease inhibitor benzamidine (Fig. 4, D and E). This interac- tion is preceded on the N-terminal side of the CDR by contacts of both Pro-104 and Arg-105, which abuts the catalytic triad, and directly bonds with Asp-483, with the S3 and S2 substrate binding pockets of pKal, respectively. Val-103 appears to have enough hydrophobic character to interact with pKal Tyr-555 and Trp-598 but not so large that it disrupts a nearby loop in  pKal (residues 478 -480) that further interacts/stabilizes both HV CDR1 and CDR3. Finally, Glu-108, which is predominantly acidic/polar in the most inhibitory affinity maturation variants, contacts Lys-575 of pKal in an interaction that is likely to be a significant driver of DX-2930 specificity with respect to other serine proteases, most of which have a Gln or Glu at this position (35). Although not directly involved in pKal interactions, the CDR3 residues 99 RRIG, Asp-107, and 109 FDI that were also included in the affinity maturation rounds make Fab intramolecular contacts that stabilize the CDR3 loop conformation and also likely contribute to the overall folding stability of the Fab. Taken together, the composition and extended length of the HV CDR3 allows the intimate contact of this loop with the pKal active site, while being short enough to steer the potentially labile peptide backbone of this loop away from the proteolytic residues of pKal. Of note, the complex structure is consistent with a proposed competitive inhibition mechanism given that binding of this antibody to pKal will exclude even small peptide substrates from the pKal active site.
Pharmacokinetic and Pharmacodynamic Analysis of DX-2930-To develop a long term prophylactic inhibitor of pKal, it was necessary to demonstrate a sustained bioactivity of DX-2930 in vivo. Toward this end, the pharmacokinetic profile of DX-2930 injected both s.c. and i.v. into cynomolgus monkeys at 20 mg/kg was examined, revealing a high bioavailability (66%) and sus-tained half-life of ϳ12.5 and 19.3 days, respectively ( Fig. 5A and Table 3). The injected dose of DX-2930 also retains bioactivity throughout its residence in the circulation. Specifically, plasma samples from monkeys that were dosed with or without s.c. DX-2930 were treated ex vivo with kaolin to activate the contact system, and the proteolytic activity of the resulting pKal was assessed by monitoring the presence or absence of cleaved 2-chain HMWK by Western blot analysis (Fig. 5B). DX-2930 inhibited the pKal-mediated proteolysis of HMWK even at 28 days following s.c. dosing, a finding consistent with the prolonged serum half-life of DX-2930.
DX-2930 Treatment Reduces Carrageenan-induced Paw Edema in Rats-To begin addressing the in vivo efficacy of DX-2930, we investigated the ability of DX-2930 to diminish carrageenan-induced paw edema (CPE) in rats. CPE is not a disease model of HAE but is a commonly used model of general inflammation that involves several mediators including proinflammatory cytokines, histamine, 5-hydroxytryptamine, and prostaglandins, in addition to kinins (50,51). Rats were pretreated with s.c. DX-2930 24 h prior to the injection of 1% carrageenan into the right hind paw to induce edema, and dosedependent reductions in swelling were recorded (Fig. 6). DX-2930 (30 mg/kg) reduced paw edema at 8 h by 55.6% compared to vehicle-treated rats. The COX inhibitor indomethacin was used as a positive control and reduced paw edema at 8 h by

DISCUSSION
Dysregulated contact system activation is central to the pathophysiology of HAE, and HAE attacks can be effectively treated with inhibitors that either prevent pKal-mediated proteolytic production of bradykinin or antagonize the B2 receptor. We used phage display to select and develop a potent, specific, and long-lived antibody inhibitor of pKal, DX-2930. X-ray crystallographic analysis of the DX-2930 Fab region bound to pKal details the molecular basis of our discovery, demonstrating that the antibody binds and occludes the substrate-binding active site region of the protease.
The DX-2930 mechanism of inhibition is similar to that employed by other well described serine protease inhibitors. For instance, amino acid Arg-106 of DX-2930 HV-CDR3 buries deeply into the canonical S1 substrate binding pocket in a manner similar to both benzamidine (Fig. 4, D and E, and Fig. 7A) and the lysine side chain of the broad specificity inhibitor BPTI/ aprotinin (Fig. 7B). With respect to reported antibody inhibitors of serine proteases, Craik and co-workers (52) characterized three antibodies that bind matriptase (MT-SP1) whereby the interacting HV-CDR3 residues align in a C-terminal to N-terminal orientation rather than the substrate-like N-termi-FIGURE 6. DX-2930 inhibits carrageenan-induced paw edema in the rat. DX-2930 was administered via a s.c. injection 24 h prior to carrageenan injection into the rat hind paw, although the vehicle (Formulation Buffer) and indomethacin (5 mg/kg) controls were administered by i.p. injection 30 min prior to carrageenan treatment. Paw volume was measured at the indicated time points following carrageenan injection and averaged for each treatment group (n ϭ 10). Error calculated using the standard deviation. FIGURE 7. Comparison of inhibitor interactions with serine protease active site pockets. Surface rendering of serine proteases are shown docked with stick models of their respective ligands just above the schematic models of these interactions with canonical protease docking sites highlighted. A, small molecule benzamidine docks into the deep S1 binding pocket of KLK4 (PDB 2BDG). B, peptide sequence from the small protein BPTI binds matriptase in a "substrate-like" mode (PDB 1EAW). C, part of the HV-CDR3 of DX-2930 binds the S1-S3 sites of pKal and then abruptly turns away from the catalytic serine. N and C refer to Nand C-terminal ends of docked schematic peptides. Inhibitor stick color scheme by atom: blue, nitrogen; red, oxygen; green, carbon. Red in surface representation and red asterisk in schematic representation denote the catalytic serine residue. nal to C-terminal orientation (e.g. BPTI in Fig. 7B). This reverse orientation allows the antibodies to bind the scission machinery of an active protease without resulting in their proteolysis. In contrast, DX-2930 binds in the substrate-like N-terminal to C-terminal orientation in the S1-S3 subsites, even to the extent that it chemically and structurally mimics the P3 proline and P1 arginine present in the natural pKal substrate, HMWK. However, the orientation of the HV-CDR3 loop in DX-2930 prevents interactions with subsites S1Ј-S3Ј, and thereby it prevents its cleavage by the pKal catalytic serine (Fig. 7C). Importantly from the perspective of an antibody therapeutic, we demonstrate that DX-2930 is indeed not proteolyzed by pKal (Fig. 2F) despite such an intimate complex formation.
The crystal structure of prekallikrein is not available, but the likely structural homology to the related proenzyme form of kallikrein 6, whose crystal structure has been defined (53), suggests that the S1 pocket (and other active site residue conformations) would only be present in the activated pKal form and not in prekallikrein. Thus, given the significant contact of DX-2930 HV-CDR3 residues with the pKal active site, it is not surprising that our SPR binding studies (Fig. 2E) demonstrate that DX-2930 does not bind prekallikrein.
A lack of prekallikrein binding was an intentional primary selection criterion to potentially reduce the therapeutic levels of DX-2930 necessary to inhibit pKal activity. Prekallikrein is present in plasma at a concentration of ϳ500 nM, and we reasoned that binding to prekallikrein would necessitate drug levels at equivalent concentrations. In contrast, a potent antibody inhibitor that bound only the active form of the enzyme could be dosed to levels that match the amount of pKal generated in disease. pKal levels between 30 and 110 nM have been estimated in HAE plasma, and an effective dose of ecallantide for the treatment of acute attacks of HAE has a C max of ϳ80 nM (586 ng/ml) (54,55). Therapeutic pKal inhibition has the potential to attenuate kinin production, as well as to limit the formation of active pKal generated by the cyclical pKal-mediated conversion of FXII to FXIIa. Therefore, it is possible that prophylactic inhibition of pKal activity may require a lower dose of inhibitor than that needed to acutely treat an HAE attack. However, future clinical studies in HAE will help elucidate the DX-2930 dose necessary for successful long term attack prophylaxis.
Certain properties of DX-2930 may support its use as a novel reagent to investigate contact system activation in biological systems or as a potential therapy for the treatment of pKalmediated diseases, such as HAE. First, the specificity of DX-2930 toward binding activated pKal, rather than prekallikrein or any other tested serine protease suggests that it could prove to be a useful reagent to detect biological sites where active pKal is generated. Second, DX-2930 can target free soluble pKal, as well as pKal complexed with HMWK that may be further associated with endothelial cells (Fig. 2, A-C). In contrast, pKal is resistant to inhibition by endogenous inhibitors C1-INH and ␣2-macroglobulin (␣2M) when it is bound to HMWK on endothelial cells, a property it shares with other cell-bound proteases such as plasmin and urokinase-type plasminogen activator (56,57). Such observations support the concept that HAE attack localization may be partially attributed to endothelial cell-bound pKal. Third, our studies in non-human primates demonstrate that DX-2930 has a long pharmacokinetic half-life (12.5 days in non-human primates, Fig. 5A and Table 3). Our recent completion of a phase 1A clinical study further reveals that DX-2930 has an almost 3-week half-life in healthy human individuals (58). Finally, DX-2930 exhibits striking specificity for pKal, which differs from the endogenous C1-INH, which has likely evolved to target numerous serine proteases and has been reported to inhibit 12 of the 20 human serine proteases that were tested in this study (59 -69). Taken together, high therapeutic doses of C1-INH are likely required for the treatment of HAE to offset its limited potency and broad specificity and may account for the occurrence of breakthrough attacks with current prophylactic approaches (70).
Contact system activation is observed in a variety of diseases associated with the hallmarks of inflammation (heat, redness, pain, and swelling) (1), implying that the activation and regulation of the contact system is tightly controlled in normal physiology. Here, we demonstrate using DX-2930-treated rats that plasma kallikrein is a mediator in CPE. Although not a specific disease model of HAE, bradykinin is one of several mediators in CPE (51), and the observed efficacy of DX-2930 in this model supports its on-target in vivo bioactivity. As a highly specific pKal inhibitor with in vivo efficacy, DX-2930 can be used to investigate the therapeutic benefit of pKal inhibition in other diseases reported to have elevated contact system activation. The role of the plasma KKS system in thrombosis has been particularly intriguing given that the targeted disruption of contact system proteins appears to reduce thrombus formation without increasing bleeding in animal models of venous and arterial thrombosis (71)(72)(73)(74)(75)(76)(77).
Interestingly, for such a conserved biological system, a genetic deficiency in prekallikrein or other contact system proteins (FXII and HMWK) is not associated with disease, including risk for bleeding, which suggests that chronic pKal inhibition could be well tolerated (78). Combining its potency with expected long half-life in patient circulation, DX-2930 may prove to be an effective prophylactic treatment of chronic diseases mediated by contact system activation, including HAE.