Structural and Inhibitory Effects of Hinge Loop Mutagenesis in Serpin-2 from the Malaria Vector Anopheles gambiae*

Background: Serpin-2 (SRPN2) is a key regulator of mosquito immunity and contains an inserted hinge region linked to activation in other serpins. Results: Structure/function analyses of hinge mutations refute a hypothesized activation mechanism in SRPN2. Conclusion: SRPN2 hinge insertion provides a thermodynamically stable conformation without restricting inhibitory capability. Significance: To effectively utilize SRPN2 for vector control, its mode of action must be understood. Serpin-2 (SRPN2) is a key negative regulator of the melanization response in the malaria vector Anopheles gambiae. SRPN2 irreversibly inhibits clip domain serine proteinase 9 (CLIPB9), which functions in a serine proteinase cascade culminating in the activation of prophenoloxidase and melanization. Silencing of SRPN2 in A. gambiae results in spontaneous melanization and decreased life span and is therefore a promising target for vector control. The previously determined structure of SRPN2 revealed a partial insertion of the hinge region of the reactive center loop (RCL) into β sheet A. This partial hinge insertion participates in heparin-linked activation in other serpins, notably antithrombin III. SRPN2 does not contain a heparin binding site, and any possible mechanistic function of the hinge insertion was previously unknown. To investigate the function of the SRPN2 hinge insertion, we developed three SRPN2 variants in which the hinge regions are either constitutively expelled or inserted and analyzed their structure, thermostability, and inhibitory activity. We determined that constitutive hinge expulsion resulted in a 2.7-fold increase in the rate of CLIPB9Xa inhibition, which is significantly lower than previous observations of allosteric serpin activation. Furthermore, we determined that stable insertion of the hinge region did not appreciably decrease the accessibility of the RCL to CLIPB9. Together, these results indicate that the partial hinge insertion in SRPN2 does not participate in the allosteric activation observed in other serpins and instead represents a molecular trade-off between RCL accessibility and efficient formation of an inhibitory complex with the cognate proteinase.


Serpin-2 (SRPN2) is a key negative regulator of the melanization response in the malaria vector Anopheles gambiae. SRPN2
irreversibly inhibits clip domain serine proteinase 9 (CLIPB9), which functions in a serine proteinase cascade culminating in the activation of prophenoloxidase and melanization. Silencing of SRPN2 in A. gambiae results in spontaneous melanization and decreased life span and is therefore a promising target for vector control. The previously determined structure of SRPN2 revealed a partial insertion of the hinge region of the reactive center loop (RCL) into ␤ sheet A. This partial hinge insertion participates in heparin-linked activation in other serpins, notably antithrombin III. SRPN2 does not contain a heparin binding site, and any possible mechanistic function of the hinge insertion was previously unknown. To investigate the function of the SRPN2 hinge insertion, we developed three SRPN2 variants in which the hinge regions are either constitutively expelled or inserted and analyzed their structure, thermostability, and inhibitory activity. We determined that constitutive hinge expulsion resulted in a 2.7-fold increase in the rate of CLIPB9 Xa inhibition, which is significantly lower than previous observations of allosteric serpin activation. Furthermore, we deter-mined that stable insertion of the hinge region did not appreciably decrease the accessibility of the RCL to CLIPB9. Together, these results indicate that the partial hinge insertion in SRPN2 does not participate in the allosteric activation observed in other serpins and instead represents a molecular trade-off between RCL accessibility and efficient formation of an inhibitory complex with the cognate proteinase.
Anopheles gambiae mosquitoes are dominant insect vectors for the most virulent species of human malaria parasites, Plasmodium falciparum, in Africa (1)(2)(3). Malaria continues to be a devastating disease, responsible for over 800,000 deaths in 2013, mostly among children in sub-Saharan Africa (4). The lack of vaccines and substandard medical resources in infected areas, coupled with drug resistance, complicate successful treatment of infected patients (5,6). Vector control remains the foremost method for controlling the spread of malaria, but resistance to all four classes of insecticides has been reported in malaria vectors (4,7,8). This highlights the need for new insecticides that are less susceptible to the selective pressures that drive resistance. Theoretical studies have suggested that late life-acting insecticides that target females would be the most effective measure to impact vector populations without facilitating resistance (9,10). The serine proteinase inhibitor SRPN2 3 is a particularly promising potential late life-acting insecticide target in A. gambiae (11,12). SRPN2 is a negative regulator of the mosquito melanization response, and depletion of SRPN2 in A. gambiae causes spontaneous melanization and significantly shortens the life span of adult female mosquitoes (11). Chemically targeting SRPN2 in field mosquito popula-tions therefore shows promise in limiting the spread of malaria in endemic areas.
SRPN2 is part of a complex regulatory pathway that modulates the insect immune response (13)(14)(15). Insects lack an adaptive immune system and must rely solely upon innate immune reactions, including melanization, to combat infectious organisms (14). Melanization is employed to encapsulate and kill invading pathogens and is initiated upon pathogen detection (16). Recognition proteins in the insect hemolymph recognize non-self biomolecules and activate a clip domain serine proteinase cascade (17)(18)(19)(20). This cascade culminates in the activation of prophenoloxidase-activating proteinases, which convert prophenoloxidase to phenoloxidase (21)(22)(23). Phenoloxidase hydroxylates monophenols to catechols and oxidizes catechols to quinones, which polymerize to form eumelanin (24,25). Although melanization is an efficient mechanism for targeting foreign pathogens, it adversely affects insect longevity (11). Uncontrolled melanization is likely to be physiologically detrimental due to the production of reactive and toxic byproducts (16) and, thus, the probable cause of the decreased life span of SRPN2-depleted mosquitoes.
The specific SRPNs and cognate functional clip domain serine proteinases (CLIPBs) that interact to regulate the melanotic response in A. gambiae are still largely uncharacterized (26). However, SRPN2 inhibits CLIPB9 both in vitro and in vivo and is the most well characterized regulatory interaction in the mosquito melanization pathway (12). CLIPB9 contains a single N-terminal clip domain and a C-terminal serine proteinase catalytic domain and is synthesized as a zymogen, becoming activated upon proteolytic cleavage at the beginning of the catalytic domain (12). Serpins inhibit proteinases via a suicide inhibitory mechanism that results in permanent inactivation of both the serpin and its cognate proteinase (27,28). Serpins generally contain 350 -400 amino acids and adopt a conserved native fold consisting of three ␤-sheets (A, B, and C) surrounded by up to nine ␣-helices (A-I) with a reactive center loop (RCL) that acts as bait for target proteinases. This native serpin fold exists in a stressed, metastable form. Upon cleavage of the RCL scissile bond (P1-P1Ј) by a target proteinase, the acyl-intermediate undergoes a remarkable 70-Å translocation whereby the RCL is inserted into ␤-sheet A as an additional ␤-strand (29). This conformational change is achieved because the relaxed, cleaved form is more thermodynamically stable than the native fold. The translocation disrupts the integrity of the proteinase active site, rendering it inactive and covalently linked to the serpin in an SDS-stable complex (30). SRPN2 uses this mechanism to permanently inactivate CLIPB9, thereby limiting phenoloxidase activation and the melanotic immune response (12).
The crystal structure of A. gambiae SRPN2 was previously determined to a resolution of 1.75 Å in its native, active form (31). As expected, SRPN2 adopts the conserved serpin fold. The most notable difference between SRPN2 and most other native serpins is the conformation of the N-terminal region of the RCL, the hinge region, composed of residues Leu 356 -Ala 360 . The SRPN2 hinge region is partially inserted between strands 3 and 5 of ␤-sheet A (␤A3 and ␤A5). A similar partial hinge insertion has only been found in a small number of inhibitory serpins: non-heparin-bound antithrombin III (ATIII) (32, 33), heparin cofactor II (34), murine antichymotrypsin (35), and Spn48 from Tenebrio molitor (36). In ATIII, the partial hinge insertion is linked to heparin-mediated activation. This insertion in ATIII restricts the flexibility of the RCL, which limits accessibility of the P1-P1Ј bond to the target proteinase (37)(38)(39). Binding of heparin pentasaccharide (H5) to helix D induces significant conformational changes, including C-terminal helix D extension and expulsion of the hinge region, resulting in extension of the RCL. As a result of these conformational changes, ATIII activity dramatically increases against factors IXa and Xa (34,37,40,41). Further evidence for H5-mediated activation of a hinge-inserted serpin is reported for Spn48, in which heparin binding significantly increases inhibition against its target proteinase (36). However, SRPN2 does not contain a heparin binding site, and H5 does not increase the activity of SRPN2 against CLIPB9 Xa (31). Therefore, although SRPN2 contains a partial hinge insertion linked to allosteric inactivation of some other serpins, its mechanistic role in SRPN2, if any, is entirely unknown.
In previous studies, mutagenesis was employed to either constitutively expel or stabilize the hinge insertion of ATIII to determine the concomitant effect on inhibitory activity (40,(42)(43)(44). Mutation-induced expulsion of the hinge region mimicked heparin activation in ATIII, and mutation-induced restriction of the hinge region significantly diminished its inhibitory capability (42,44). In the current study, we used equivalent mutations in SRPN2 to elucidate the influence of the partial hinge insertion on SRPN2 inhibition of CLIPB9 Xa . We created SRPN2 mutations that constitutively expelled (S358E) or restricted (S358W, E359C/K198C) the hinge region, investigated the structural and thermodynamic effects of these mutations, and determined the mutants' ability to inhibit CLIPB9 Xa . The results indicate that the partial hinge insertion in SRPN2 is not a structural regulatory mechanism as observed in ATIII. Instead, the data suggest that the partial hinge insertion in SRPN2 is thermodynamically stable, permits P1-P1Ј accessibility to CLIPB9 Xa , and maintains the ability to efficiently form inhibitory complexes.
All SRPN2 variant proteins were expressed and purified as described previously with the following modifications (12,45). The plasmid constructs were used to transform Escherichia coli strain BL21(DE3) competent cells (Invitrogen). Bacteria were grown in 1 liter of LB medium at 37°C until an A 600 of 0.6 -0.8 was reached. 0.1 mM isopropyl ␤-D-1-thiogalactopyranoside was then added, and the cells were shaken at 20°C and 150 rpm for 16 h. The cells were then lysed by sonication and purified by nitrilotriacetic acid-agarose chromatography (Qiagen), followed by ion exchange chromatography on a Q-Sepharose column (GE Healthcare). The fractions were examined by 10% SDS-PAGE and Coomassie Blue staining to confirm homogeneity. Fractions containing sufficiently pure recombinant protein were pooled for later use.
The cloning, expression, and purification of recombinant A. gambiae proCLIPB9 Xa were performed as described previously with the following modifications. Full-length A. gambiae proCLIPB9 was cloned into a pFastBac1 (Invitrogen) vector, and the activation cleavage site (IGMR) was mutated (IEGR) via a QuikChange multisite-directed mutagenesis kit (Stratagene) to generate A. gambiae proCLIPB9Xa and permit controlled activation by factor Xa. Recombinant baculoviruses were generated by using the resulting plasmid according to the manufacturer's instructions (Invitrogen). Recombinant protein was expressed in Sf9 cells following the manufacturer's protocol and purified by nitrilotriacetic acid-agarose chromatography (Qiagen) followed by ion exchange chromatography on a Q-Sepharose column (GE Healthcare). The fractions were examined by 10% SDS-PAGE and Coomassie Blue staining to confirm homogeneity. Fractions containing sufficiently pure recombinant protein were pooled for later use.
Intensities were integrated using XDS (46), and the Laue class check and data scaling were performed with Aimless (47). Structure solution was conducted by molecular replacement with Phaser (48) via the Phenix (49) or Molrep (50) interface using the SRPN2-WT structure (PDB entry 3PZF) (31) as the search model. Refinement and manual model building were carried out with Phenix and Coot (51), respectively. TLS (transition/liberation/screw) refinement (52) was incorporated in the later stages to model anisotropic atomic displacement parameters. Structure validation was conducted with Molprobity (53). Disordered side chains were truncated to the point where electron density could be observed. Figures were prepared using PyMOL (54) and the CCP4MG package (55). Relevant crystallographic data are provided in Table 1. The structures were deposited in the Protein Data Bank with the following accession numbers: SRPN2-S358E (4RO9), SRPN2-S358W (4ROA), and SRPN2-K198C/E359C (4RSQ).
The diffraction data for SRPN2-K198C/E359C were initially indexed in an orthorhombic P lattice (a ϭ 97, b ϭ 164, c ϭ 186.23) and displayed a large pseudotranslation peak of ϳ30% of the origin at 0.263, 0.5, 0.456, and the self-rotation function indicated 3-fold non-crystallographic symmetry parallel to the crystallographic b axis. The two non-crystallographic symmetry trimers were positioned by molecular replacement with Phaser and Molrep using a single chain from PDB entry 3PZF as the search model. The same solutions were obtained with both programs with the top score in the space group P222. However, following refinement, the electron density was clear for one trimer (chains A-C) but poor for the second trimer (chains D-F), and the R-factors converged at R/R free ϭ 29%/36%. The molecules were arranged in rows that stacked along the crystallographic b axis and consisted of alternating sets of two A-C trimer rows and two D-F trimer rows. Inspection of the crystal packing revealed that chain D of the D-F trimer overlapped a symmetry mate related by a crystallographic 2-fold axis along b. We considered that this might be a case of lattice translocation disorder given that the diffraction spots were streaked along the b* reciprocal lattice direction. However, this condition is typified by alternating strong and diffuse spots in the diffraction pattern (56), which was not observed for the SRPN2-K198C/ E359C crystals. The data were reprocessed in a P1 unit cell, and a molecular replacement solution consisting of 24 molecules (8 trimers) was obtained with Molrep. This allowed us to determine the packing arrangement of the trimers and subsequently apply the results of this solution to a higher symmetry monoclinic P lattice (P2 1 , a ϭ 97.93, b ϭ 164.39, c ϭ 186.18, ␤ ϭ 90.04°) containing 12 molecules (4 trimers) in the asymmetric unit. Further confirmation was obtained with the program Zanuda (57), which yielded the same top solution for the mono-clinic P2 1 lattice. Twin refinement with Refmac (58) using the pseudomerohedral twin law (h, Ϫk, Ϫl), determined by Xtriage within the Phenix package, yielded a final model with R/R free ϭ 19.4/25.5% and a twin fraction of 45%. Thus, it appeared that the crystal structure was best modeled as a monoclinic P lattice with a ␤-angle near 90°, which "mimicked" a higher symmetry orthorhombic P lattice.
Differential Scanning Calorimetry-To determine the thermal stability of the SRPN2 mutants, purified protein (0.5 mg/ml) was dialyzed in 10 mM sodium phosphate, pH 7.6, and 150 mM NaCl overnight, degassed, and loaded into a VP-DSC microcalorimeter (MicroCal Inc.). The temperature was scanned at 1°C/min from 20 to 80°C. Heat capacity (Cp) (kcal/mol/K) was plotted after subtraction of a blank experiment without protein. A second scan for each protein was performed but excluded from the final data due to aggregation. To compare T m of each protein, the baseline Cp was normalized to SRPN2-WT.
Activation of proCLIPB9 Xa -The activation of proCLIPB9 Xa was performed as described previously (12) with the following modifications. 5 g of purified proCLIPB9 Xa was incubated with 2 g of commercial bovine factor Xa (New England Biolabs) in a total volume of 100 l (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2 mM CaCl 2 ) at 37°C for 16 h. Cleavage of the zymogen was examined by Western blot using anti-His antibody. To examine the amidase activity, 2 l of the above reaction were transferred to a 96-well flat bottomed microplate (Corning, Inc.), followed by the addition of 200 l of 1,000 M acetyl-IEAR-p-nitroanilide (BioWorld) in buffer (0.1 M Tris-HCl, pH 8.0, 0.1 M NaCl, 5 mM CaCl 2 ). Absorbance changes at 405 nm were monitored immediately in a microplate reader (Bio-Tek Instruments, Inc.) every 30 s for 20 min. One unit of amidase activity was defined as ⌬A 405 /min ϭ 0.001. Amidase activity of the serine proteinase was defined as the activity of enzyme minus the activity of factor Xa alone.
Serpin Inhibition Assays-To examine the inhibitory effect of serpins on proteinase activity, 100 ng of activated CLIPB9 Xa in a volume of 2 l was incubated with 4 l of recombinant SRPN2 proteins at different molar ratios with the addition of 1 l of BSA (2 g/l) at 23°C for 20 min. The reaction was then subjected to an amidase activity assay as described above. Substitution of recombinant serpin with 4 l of buffer (20 mM Tris-HCl, pH 8.0,100 mM NaCl) was used to determine 100% enzyme activity. Residual amidase activity was plotted against the ratio of SRPN2 to activated CLIPB9 Xa , and the stoichiometry of inhibition (SI) was determined as the x intercept of the linear regression fit. All experiments were carried out with at least three independent replicates.
The second-order rate constant (k a ) of interactions between SRPN2-WT or SRPN2-S358E and CLIPB9 Xa was determined under pseudo-first order conditions as described previously (59). A fixed amount of CLIPB9Xa (2.4 pmol) was mixed with different concentrations of recombinant SRPN2 in 1,000 M IEAR-p-nitroanilide. The progress of product formation (P) at each concentration of SRPN2 was measured immediately as described above. For each combination of enzyme and inhibitor, a k obs value was calculated by nonlinear regression using the following equation, where P represents the amount of product formation, V is initial velocity, t is time, and k obs is reaction rate. The k a was determined by plotting a series of k obs against the respective SRPN2 concentration and measuring the slope of the linear regression fit.
The association rate constant for the interaction between CLIPB9 Xa and A. gambiae SRPN2-S358W or A. gambiae SRPN2-K198C/E359C was determined by a discontinuous second order rate constant inhibition assay as described previously with the following modifications (60). Briefly, A. gambiae SRPN2-S358W or A. gambiae SRPN2-K198C/E359C was added to activated CLIPB9 Xa at a molar ratio of 100:1 and incubated at room temperature for varying periods of time (t), including 0, 10, 20, 20, 40, 60, 80, 100, and 120 min. The residual amidase activity was measured at each time point (V t ) as described above. Initial enzyme activity (V 0 ) was measured by replacing SRPN2 protein in the reaction with the same volume of buffer. The slope was calculated by plotting a series of ln(V t /V 0 ) values against the respective incubation time. The k a was calculated by dividing this negative slope by the concentration of the SRPN2 variants.
Detection of inhibitory complexes between SRPN2-CLIPB9 Xa was performed as described previously (12). Activated CLIPB9Xa was mixed with purified SRPN2 at a molar ratio of 1:6 and incubated at room temperature for 10 min. The reaction mixtures were separated by 10% SDS-PAGE and stained with Coomassie Blue.

RESULTS
Structure of SRPN2-S358E-The previously determined crystal structure of A. gambiae SRPN2-WT revealed a characteristic serpin fold consisting of three ␤-sheets (A, B, and C) flanked by nine ␣-helices (A-I) (31). The SRPN2-WT RCL hinge region (residues Leu 356 -Ala 360 ) was inserted into ␤-sheet A between strands ␤A3 and ␤A5. P14 residue Ser 358 is located at the apex of the loop involved in the hinge insertion. The corresponding P14 residue in ATIII (Ser 380 ) was also located in this position, and its mutation to a glutamate (ATIII-S380E) resulted in an expulsion of the hinge region. Importantly, ATIII-S380E inhibitory activity against factor Xa is increased nearly 200-fold, to a level comparable with wild-type ATIII upon H5 activation (32,44). To investigate the consequences of the hinge region insertion for SRPN2 structure and activity, we mutated Ser 358 to a glutamate (SRPN2-S358E) in an attempt to constitutively expel the residue from ␤-sheet A in the native serpin conformation.
The conformation of the hinge region in SRPN2-S358E was characterized from the crystal structure, determined to a resolution of 2.0 Å (Table 1 and Fig. 1A) and contained three molecules in the asymmetric unit (Fig. 1B). The three subunits are nearly identical except for the number of residues that could be traced in the RCL hinge region (Fig. 1C). Although most of the RCL was disordered in each subunit, the RCL in chain A could be traced to residue Ala 360 (Leu 356 -Ala 360 were inserted in SRPN2-WT) (Fig. 1D). Therefore, the A-chain was used for all further analysis of the SRPN2-S358E structure.
The overall structure of SRPN2-S358E is very similar to the SRPN2-WT structure. Superposition of the two structures using secondary structure matching (61) yielded a root mean square deviation of 0.76 Å between C␣ atoms of the 353 residues aligned ( Fig. 2A). As predicted, the most notable difference between the wild type and SRPN2-S358E structures is the RCL hinge region (Fig. 2B). The SRPN2-S358E hinge region is indeed expelled from ␤-sheet A, translocated ϳ12 Å from the inserted position in SRPN2-WT. Glu 358 forms hydrogen bonds with Lys 47 in helix A of the A-chain of a symmetry-related molecule, Arg 252 located between ␤B2 and ␤B3, and Lys 114 located between helix D and ␤A1 (Fig. 2C). Overall, these structural data confirm that SRPN2-S358E assumes the native serpin fold and contains the predicted expulsion of the hinge region.
Structure of SRPN2-S358W-The fact that the SRPN2-S358E structure revealed an expelled RCL hinge region prompted us to develop a mutant in which the hinge region insertion could be further stabilized compared with the wild-type protein. Therefore, we generated a SRPN2-S358W mutant, in which Trp 358 could potentially interact with the hydrophobic residues between strands ␤A3 and ␤A5 and thus be buried within the hydrophobic interior of the serpin.
The crystal structure of SRPN2-S358W was determined to a resolution of 1.9 Å (Table 1) and is isomorphous to SRPN2-WT with a single molecule in the asymmetric unit. The root mean square deviation between C␣ atoms in SRPN2-S358W and SRPN2-WT was 0.33 Å for 358 residues aligned (Fig. 3A). As predicted, the RCL hinge region in SRPN2-S358W adopts a similar conformation compared with SRPN2-WT, buried within the interior of ␤-sheet A (Fig. 3B). The SRPN2-S358W RCL could be modeled to residue Glu 374 , and Trp 358 side chain density was clearly defined (Fig. 3C). Trp 358 is located between ␤A3 and ␤A5 and is embedded within a hydrophobic pocket composed of Ile 351 and Ile 353 on ␤A5, Phe 197 on ␤A3, Phe 400 on ␤B5, Ile 390 on ␤B4, and Tyr 251 located between ␤B2 and ␤B3 (Fig. 3D). Residues Phe 197 , Phe 400 , and Tyr 251 are highly invariant and conserved in ϳ75% of serpins (62). Subtle conformational changes are observed in multiple residues within this hydrophobic pocket to accommodate the tryptophan side chain. The structure of SRPN2-S358W suggests that Trp 358 may stabilize the hinge region in an inserted conformation.
Structure of SRPN2-K198C/E359C-Despite the structural data from SRPN2-S358W, it is possible that the region could become intermittently expelled in a dynamic equilibrium as observed previously in ATIII-S380W (40). Therefore, we also is the intensity measured for the ith reflection, and ͗I(hkl)͘ is the average intensity of all reflections with indices hkl. c R meas ϭ redundancy-independent (multiplicity-weighted) R merge (47,70). R pim ϭ precision-indicating (multiplicity-weighted) R merge (71,72). d CC1 ⁄ 2 is the correlation coefficient of the mean intensities between two random half-sets of data (73,74). e R factor ϭ ⌺ hkl ʈF o (hkl) ͉ Ϫ ͉F c (hkl) ʈ/⌺ hkl ͉ F o (hkl)͉; R free is calculated in an identical manner using 5% of randomly selected reflections that were not included in the refinement.
created a SRPN2-K198C/E359C mutant in which a disulfide bond is introduced between the hinge region residue Glu 359 and the adjacent K198C residue located on ␤A3. Hypothetically, a disulfide bond at this position would strengthen the interaction and maintain the inserted hinge loop region with limited possibility for its expulsion. The introduction of a disulfide bond at this position was used previously for similar purposes in a study of ATIII (42). The crystal structure of SRPN2-K198C/E359C was determined to a resolution of 2.9 Å (Table 1), containing 12 molecules in the asymmetric unit. The overall conformation of SRPN2-K198C/E359C was very similar to SRPN2-WT with a root mean square deviation of 0.63 Å between C␣ atoms for 351 residues aligned (Fig. 4A). As predicted, the RCL hinge region in the SRPN2-K198C/E359C structure is inserted into ␤-sheet A (Fig. 4B). The hinge region of the RCL in the SRPN2-K198C/ E359C structure could be modeled up to residue Ala 360 and shows clear density corresponding to the disulfide bond introduced between Cys 359 and Cys 198 (Fig. 4C). We further confirmed the presence of the disulfide bond by analyzing SRPN2-WT, SRPN2-K198C, SRPN2-E539C, and SRPN2-K198C/ E359C on SDS-PAGE under reducing and nonreducing conditions. All proteins ran at the same size under both conditions except for SRPN2-K198C/E359C, which ran faster under nonreducing compared with reducing conditions, indicating the presence of a disulfide bond (Fig. 4D). Therefore, SRPN2-K198C/E359C exists in a native conformation and contains a disulfide bond that stabilizes the partial RCL hinge insertion.
Differences in Thermostability and Structure among the SRPN2 Variants-The SRPN2 variant structures clearly reveal either expulsion or stabilization of the RCL hinge region. To investigate the thermodynamic effects of these mutations, we used differential scanning calorimetry to determine the midpoint transition temperature (T m ) of each variant (Fig. 5A). The thermal unfolding transition was absent during repeated scans (data not shown) due to irreversible aggregation of the samples at high temperature. The T m of SRPN2-WT was 55.9°C, and a similar T m of 54.6°C was determined for SRPN2-S358W. The   expulsion of the hinge region in SRPN2-S358E resulted in a decrease in T m to 52.9°C. Whereas Glu 358 formed new hydrogen bonds with Lys 114 and Arg 252 , the network of stabilizing bonds found in the hinge of SRPN2-WT was lost, reducing overall thermodynamic stability. As a consequence, hinge expulsion led to 16 residues in the RCL being left unresolved (residues 361-376), twice as many as were unresolved in the SRPN2-WT structure (residues 367-374). Last, we determined that SRPN2-K198C/E359C had a significant increase in thermostability, with T m of 60.7°C. This T m increase reflects the enhanced protein stability provided by the disulfide bond. Together, these results indicated that differences in the RCL conformation and stabilizing intramolecular interactions between the SRPN2 variants are reflected in their respective thermostabilities.
The structural differences found in the hinge region of the SRPN2 variants also affect the respective conformations of ␤-sheet A. During the formation of the inhibitory complex, the RCL must insert between ␤A3 and ␤A5 as an additional ␤-strand. Therefore, the distance between the ␤-strands must increase in order to accommodate insertion of the RCL. In SRPN2-WT, ␤A3 and ␤5 are separated by a distance of ϳ8.6 Å at the location of the partial hinge insertion (Fig. 5B). Similar distances are found for SRPN2-S358W (9.0 Å) (Fig. 5C) and SRPN2-K198C/E359C (8.7 Å) (Fig. 5D). However, with the hinge expelled in SRPN2-S358E, this distance was reduced to 6.3 Å (Fig. 5E).
Stoichiometry of Inhibition of SRPN2 Variants-Having established the structure and thermostability of the SRPN2 variants, we determined their ability to form an inhibitory complex with the activated cognate proteinase CLIPB9 Xa and the SI of this interaction (Table 2 and Fig. 6). SRPN2-WT forms an SDS-stable complex with CLIPB9 Xa with an SI of 1.7 Ϯ 0.1. SRPN2-S358E was also able to inhibit CLIPB9 Xa and form an SDS-stable complex with the proteinase with a slightly increased SI of 2.5 Ϯ 0.7, indicating that the expelled hinge mutant is capable of forming an inhibitory complex at a level comparable with SRPN2-WT. Conversely, SRPN2-S358W did  not form an inhibitory complex with CLIPB9, reflected in its significantly increased SI of 50 Ϯ 12 (Table 2 and Fig. 6). However, CLIPB9 Xa efficiently degraded SRPN2-S358W, leaving a cleavage product at a size consistent with cleavage at the RCL P1-P1Ј. Therefore, although SRPN2-S358W cannot inhibit CLIPB9 Xa , it is efficiently cleaved at the RCL. Surprisingly, SRPN2-K198C/E359C was able to form an inhibitory complex with CLIPB9 Xa , although the SI was greatly increased to 20 Ϯ 5. SRPN2-K198C/E359C also contained a band in SDS-PAGE consistent with cleavage at the RCL.
Rate of CLIPB9 Xa Inhibition by SRPN2 Variants-Having established that the SRPN2 variants, with the exception of SRPN2-S358W, each form an inhibitory complex with CLIPB9 Xa , we determined the second order rate constants (k a ) of the SPRN2 variants' inhibition against CLIPB9 Xa (Table 2). SRPN2-WT was able to inhibit CLIPB9 Xa with a k a of (1.5 Ϯ 0.2) ϫ 10 3 M Ϫ1 s Ϫ1 . Combining this reaction rate with the SI (SI ϫ k a ) provides a measure of total flux down both the substrate and inhibitory pathways of SRPN2-WT with CLIPB9 Xa of (2.6 Ϯ 0.4) ϫ 10 3 M Ϫ1 s Ϫ1 . This level of activity is relatively low compared with ATIII, which has inhibitory activity against factor Xa in the 1.0 ϫ 10 7 M Ϫ1 s Ϫ1 range when fully activated (40). To determine whether the expelled hinge region increases the inhibitory activity of SRPN2, we measured the k a of SRPN2-S358E against CLIPB9 Xa . The k a of SRPN2-S358E was (2.8 Ϯ 0.3) ϫ 10 3 M Ϫ1 s Ϫ1 , a 1.9-fold increase compared with the k a of SRPN2-WT. Combining the k a and SI of SRPN2-S358E, (7.1 Ϯ 0.3) ϫ 10 3 M Ϫ1 s Ϫ1 , indicates that the overall reaction kinetics of SRPN2-S358E are increased nearly 3-fold compared with SRPN2-WT.
We also investigated the second order rate constants of the hinge-inserted SRPN2 variants SRPN2-S358W and SRPN2-K198C/E359C (Table 2). Based on its high SI and inability to form an SDS-stable inhibitory complex, it is not surprising that SRPN2-S358W had a k a of (6.1 Ϯ 0.2) ϫ 10 1 M Ϫ1 s Ϫ1 . These results confirm that SRPN2-S358W has effectively no inhibitory activity against CLIPB9 Xa . We determined that the k a of SRPN2-K198C/E359C was (2.3 Ϯ 0.4) ϫ 10 2 M Ϫ1 s Ϫ1 , which is a 6.7-fold decrease compared with SRPN2-WT. However, the SI of SRPN2-K198C/E359C was significantly increased, indicating that the disulfide bond may hinder this variant's ability to form an inhibitory complex independent of the effect of the inserted hinge region that we are investigating. Combining the SI and k a for SRPN2-K198C/E359C, (4.6 Ϯ 0.4) ϫ 10 3 M Ϫ1 s Ϫ1 , indicates that this hinge-inserted variant has a similar flux down inhibitor and substrate pathways when interacting with CLIPB9 Xa , comparable with SRPN2-WT and SRPN2-S358E.

DISCUSSION
SRPN2 is a critical negative regulator of the melanization pathway in A. gambiae. The phenotype of SRPN2 depletion from adult female mosquitoes shows accelerated mortality rates and decreased feeding propensity at the time when malaria-infected mosquitoes are able to transmit the parasite to the next human host (11,12). 4 SRPN2 may therefore be a promising target for small molecule inhibitors to be utilized as late life-acting insecticides for vector control.
The structure of SRPN2-WT revealed a partial insertion of the RCL hinge region into ␤-sheet A. A very similar partial RCL insertion was linked previously to allosteric activation of ATIII. The ATIII data revealed a model whereby heparin-induced conformational changes resulted in expulsion of the hinge region increasing accessibility of the RCL P1-P1Ј and exosites, thereby boosting inhibition of target proteinases. Due to the relative rarity of the inserted hinge region among serpins, we hypothesized that it likewise functions as a regulatory mechanism in SRPN2. To test this hypothesis, we developed SRPN2 mutants with constitutively expelled (SRPN2-S358E) or stabilized (SRPN2-S358W, SRPN2-K198C/E359C) hinge regions, analyzed the structure and thermostability of the variants, and determined the effects of the mutations on inhibition of CLIPB9 Xa . Our data reveal that a hingeexpelled SRPN2 variant had a limited increase in inhibitory activity against CLIPB9 Xa and that stabilization of the inserted hinge region did not affect RCL accessibility. Together, these results strongly suggest that SRPN2 does not follow the ATIII model of allosteric activation by H5.
The hinge-expelled SRPN2-S358E mutant did indeed gain an ϳ3-fold increase in CLIPB9 Xa inhibition, indicating that serpin activity does generally benefit from increased RCL extension and flexibility. However, this increase is much less than the 190-fold increase in the equivalent ATIII-S380E mutant against factor Xa (44). An explanation for the limited advantage in SRPN2 is the possibility that hinge expulsion is achieved by exosite interactions between SRPN2 and CLIPB9 prior to cleavage of the scissile bond. SRPN2-S358E would therefore circumvent this step, gaining 3-fold efficiency. Such exosite interaction would increase the specificity of SRPN2 for CLIPB9, which may be crucial in the physiological context of an open circulatory system. Further structural and mutagenesis studies and inves-4 K. Michel, unpublished observations. tigations into the SRPN2-CLIPB9 interaction will be necessary to test this hypothesis. An equally parsimonious explanation for the minimal activation of SRPN2-S358E is that accessibility of the P1-P1Ј bond is not strongly hindered by the potential hinge insertion. Therefore, CLIPB9 inhibition simply does not require extension of the RCL. We found that CLIPB9 Xa cleaves the RCL of both SRPN2-S358W and SRPN2-K198C/E359C, despite their stabilization of the hinge insertion. SRPN2-S358W is unable to form an inhibitory complex with CLIPB9 Xa , probably due to the bulky P14 residue inhibiting the complete insertion into ␤-sheet A, which has been reported for other such mutants (63). However, SRPN2-K198C/E359C can form an inhibitory complex with CLIPB9 Xa , and its increased SI is probably due to effects of the disulfide bond on complete insertion. Furthermore, when SI is taken into consideration, SRPN2-K198C/ E359C has an efficiency of interaction with CLIPB9 Xa comparable with SRPN2-WT. Therefore, these variants indicate that the partial hinge insertion does not hinder the ability for CLIPB9 to target and cleave the SRPN2 RCL. These data can be interpreted in the context of a recently proposed expansion of the allosteric ATIII activation model (64). This model envisions that lowered activity of hinge-inserted ATIII in the absence of H5 is largely due to repulsive exosite interactions with factors IXa and Xa. These proposed negative exosite interactions are diminished physiologically by conformational changes induced by H5 and presumably are overcome by mutational expulsion of the hinge region as seen in ATIII-S380E. Following this model, efficient cleavage of SRPN2-S358W and SRPN2-K198C/E359C can be explained by the lack of such repulsive exosites in the interaction of SRPN2 with CLIPB9.
Consistent with the idea that RCL accessibility is not a limiting factor of SRPN2 activity, the structural data suggest that the hinge-inserted conformation of SRPN2 provides increased efficiency for complete hinge insertion after P1-P1Ј cleavage, thus facilitating inhibitory complex formation. The SRPN2-WT hinge insertion naturally results in a local increase in the distance between ␤A3 and ␤A5 that decreases once the hinge is expelled in SRPN2-S358E. Previous studies on ATIII indicated the requirement for coordination of ␤A strand separation and RCL insertion (44). However, the drastically increased level of RCL cleavage upon hinge expulsion in ATIII would overcome negative effects due to increased ␤A strand interaction. However, because the SRPN2 RCL is cleaved at similar rates in inserted and expelled forms, the local separation of ␤A strands may provide an energetic advantage toward rapid insertion. The SRPN2-S358E inhibitory data supports this, with a 1.9-fold increase in SRPN2-S358E k a (perhaps due to increased RCL accessibility) and a 1.5-fold decrease in SI (due to decreased insertion). The partially inserted hinge region in SRPN2-WT therefore gains an energetic head start toward complete insertion that is not negated by diminished RCL accessibility.
Although the data did not reveal that the inserted hinge in SRPN2 functions as a regulatory mechanism against CLIPB9 inhibition, it is nevertheless conceivable that hinge expulsion is required for inhibition of a yet to be identified proteinase target. Although this possibility is actively pursued, the low level of CLIPB9 inhibition in vitro remains intriguing and suggests that a different unidentified mode of SRPN2 regulation may exist in vivo. Activation of ATIII against thrombin requires a bridging mechanism mediated by full-length heparin that binds both proteins, in contrast with the allosteric activation mediated by H5 that governs inhibition of factors IXa and Xa (reviewed in Ref. 41). This divergence has been attributed to the poorer suitability of ATIII as a thrombin substrate (44) and the inability of thrombin to utilize exosites critical for contact between ATIII and factors IXa and Xa (41). Molecules that facilitate the initial interaction between thrombin and ATIII are therefore far more likely to increase ATIII activity against thrombin than those that extend the RCL (44). The ATIII-S380E mutation only increases the k a against thrombin 1.9-fold (44), which correlates intriguingly well with our SRPN2-CLIPB9 Xa data. It is tempting to speculate that a bridging mechanism rather than allosteric regulation modulates SRPN2 activity. There are a number of inactive proteinases in the CLIPA family that coordinate protein-protein interactions in different pathways within insect immune systems (65)(66)(67)(68)(69). The specific molecular roles of CLIPAs within melanization in mosquitoes await elucidation, and some may indeed regulate SRPN2 activity and thus phenoloxidase activation in A. gambiae.